Innovative Developments

Bartolo FINAL 08-09-11 13:47 Pagina 1
Brtolo
et al.
Paulo Brtolo is Professor of Advanced Manufacturing Processes at the

Polytechnic Institute of Leiria (IPL), Director of the Centre for Rapid and
Sustainable Product Development of IPL and President of the Scientific Council
of the Institute for Post-graduation Studies and Research of IPL.
Paulo Brtolo is Editor-in-Chief of the Journal Virtual & Physical Prototyping,
published by Taylor & Francis (ISSN 1745-2759)
an informa business
Innovative Developments
in Virtual and Physical
Prototyping
Innovative Developments in Virtual and Physical Prototyping presents essential

research in the area of Virtual and Rapid Prototyping. The volume contains
reviewed papers presented at the 5th International Conference on Advanced
Research in Virtual and Rapid Prototyping, hosted by the Centre for Rapid and
Sustainable Product Development of the Polytechnic Institute of Leiria, Portugal,
from September 28 to October 1, 2011. A wide range of topics is covered, such
as CAD and 3D Data Acquisition Technologies, Additive and Nano Manufacturing
Technologies, Rapid Tooling & Manufacturing, Biomanufacturing, Materials for
Advanced Manufacturing Processes, Virtual Environments and Simulation,
Applications of Virtual and Physical Prototyping Technologies.
Innovative Developments in Virtual and Physical Prototyping is intended for
engineers, designers and manufacturers who are active in the areas of
mechanical, industrial and biomedical engineering.
Prototyping
Edited by
Paulo Jorge Brtolo et al.
INNOVATIVE DEVELOPMENTS IN VIRTUAL AND PHYSICAL PROTOTYPING
This page intentionally left blank
PROCEEDINGS OF THE 5TH INTERNATIONAL CONFERENCE ON ADVANCED RESEARCH

AND RAPID PROTOTYPING, LEIRIA, PORTUGAL, 28 SEPTEMBER 1 OCTOBER, 2011
Innovative Developments in Virtual

and Physical Prototyping
Editors
Paulo Jorge Brtolo
Ana Cristina Soares de Lemos
Ana Patrcia Oliveira Tojeira
Antnio Mrio Henriques Pereira
Artur Jorge Mateus
Ausenda Lus Avelar Mendes
Cyril dos Santos
Dino Miguel Fernandes Freitas
Helena Maria Brtolo
Henrique de Amorim Almeida

Igor Marques dos Reis
Juliana Rosa Dias
Marco Andr Neves Domingos
Nuno Manuel Fernandes Alves
Ruben Filipe Brs Pereira
Tatiana Marisa Fernandes Patrcio
Telma Margarida Dias Ferreira
Centre for Rapid and Sustainable Product Development

Polytechnic Institute of Leiria, Portugal
CRC Press/Balkema is an imprint of the Taylor & Francis Group, an informa business
2012 Taylor & Francis Group, London, UK
Typeset by MPS Limited, a Macmillan Company, Chennai, India
Printed and bound by CPI Group (UK) Ltd, Croydon, CRO 4YY
All rights reserved. No part of this publication or the information contained herein may be
reproduced, stored in a retrieval system, or transmitted in any form or by any means,
electronic, mechanical, by photocopying, recording or otherwise, without written prior
permission from the publisher.
Although all care is taken to ensure integrity and the quality of this publication and the
information herein, no responsibility is assumed by the publishers nor the author for any
damage to the property or persons as a result of operation or use of this publication
and/or the information contained herein.
Published by:
CRC Press/Balkema
P.O. Box 447, 2300 AK Leiden, The Netherlands
e-mail: Pub.NL@taylorandfrancis.com
www.crcpress.com www.taylorandfrancis.co.uk www.balkema.nl
ISBN: 978-0-415-68418-7 (Hbk)

ISBN: 978-0-203-18141-6 (eBook)
Innovative Developments in Virtual and Physical Prototyping Brtolo et al. (eds)

2012 Taylor & Francis Group, London, ISBN 978-0-415-68418-7
Table of Contents
Preface
Committee Members
XIII
XV
Keynotes
Product
Lab A low cost alternative to introduce AM in South Africa
D.J. de Beer
Idea 2
Additive Manufacturing-assisted scaffold-based Tissue Engineering

C.K. Chua, M.J.J. Liu & S.M. Chou
3
13
Biomanufacturing
Ultrastructural analysis of the hDSC interactions with biodegradable 3D scaffolds
S.E. Duailibi, M.T. Duailibi, L.M. Ferreira, F.A.O. Tanaka, J.P. Vacanti & P.C. Yelick
25
Individual contour adapted functional implant structures in Titanium

C. Schoene, R. Stelzer, P. Sembdner, L. Betrol, J. Markwardt, B. Reitemeier & G. Engel
29
New approaches to prototype 3D vascular-like structures by additive layer manufacturing

E. Bassoli, L. Denti, A. Gatto, A. Paderno, G. Spaletta, N. Zini, V. Strusi, D. Dallatana & R. Toni
35
A novel protein-based scaffold with macro- and micro-structural features for tissue
engineering applications
M.J.J. Liu, S.M. Chou & C.K. Chua
Surgical training and post-surgery evaluation using rapid prototyped biomodels
L. Queijo, Joo Rocha, Paulo Miguel Pereira & Manuel San Juan
The calibration of continuous Digital Light Processing (cDLP) for the highly accurate
additive manufacturing of tissue engineered bone scaffolds
D. Dean, J. Wallace, A. Siblani, M.O. Wang, K. Kim, A.G. Mikos & J.P. Fisher
43
51
57
Fabrication and characterization of biodegradable composite scaffolds for Tissue Engineering

T. Serra, M. Navarro & J.A. Planell
67
Spatially varying porosity with continuous path plan for hollowed tissue scaffolds
A.K.M.B. Khoda, I.T. Ozbolat & B. Koc
73
Heterogeneous tissue scaffolds for spatiotemporally controlled release kinetics

I.T. Ozbolat, A.K.M.B. Khoda, M. Marchany, J.A. Gardella & B. Koc
79
Medical application of rapid prototyping in orthopedics surgical planning

C.B.L. Ulbrich, C.A.C. Zavaglia, T.P. Leivas & F. Teixeira
85
Polycaprolactone-based scaffold plus BMP-2 in a sheep thoracic spine fusion model

M. Yong, F. Melchels, C. Vaquette, D. Hutmacher, C. Adam, M. Domingos & P. Bartolo
89
Development of functional graded device of PCL/PG by Selective Laser Sintering for drug
delivery applications
G.V. Salmoria, P. Klauss, K. Zepon, L.A. Kanis & C.R.M. Roesler
Flexible PCL tube scaffolds by winding of micro-extruded filaments
K. Ragaert, L. Cardon & J. Degrieck
93
99
Automation design and simulation of a stent

E.L. Melgoza, L. Seren, J. Ciurana & A. Rosell
105
Production and in-vitro characterization of micro-structured implant surfaces

M. de Wild, Th. Mller, S. Tschumi, R. Schumacher & H. Albrecht
111
A study of mechanical and biological behavior of porous Ti6Al4V fabricated on EBM

V. Petrovic, J.R. Blasco, L. Portols, I. Morales, V. Primo, C. Atienza, J.F. Moreno & V. Belloch
115
Enabling technologies for robotic organ printing

R.A. Rezende, F.D.A.S. Pereira, B.D.T. Kemmoku, J.V.L. Silva, V. Mironov, V. Kasyanov & T. Vilbrandt
121
The semi-automated design & manufacture of patient-specific intervertebral disc implants

N. de Beer
131
Additive manufacturing of soft tissue geometries for reconstruction purposes

M. Truscott, G.J. Booysen & D.J. de Beer
139
Development of patient-specific implants using Direct Metal Laser Sintering in Titanium

G.J. Booysen, M. Truscott, J. Els & D.J. de Beer
145
CAD and 3D Data Acquisition Technologies

A prototype of a spherical tippe top
M.C. Ciocci, B. Malengier & B. Grimonprez
157
An innovative methodology for laser scanner integration in a robot cell for small batch
production of sculpture artworks
C. Cenati, G. Borroni, L. Cevasco, D. Parazzoli & M. Danesi
163
Application of time compression techniques to dental restoration: A procedure for quality evaluation
A. Salmi, E. Atzeni & L. Iuliano
169
Reverse engineering of casting equipment for process simulation

F. Calignano, P. Minetola, A. Salmi, E. Atzeni & L. Iuliano
177
Patient specific parametric geometric modelling of cementless hip prosthesis

G. Saravana Kumar & M. Gupta
187
Cross-section morphological study

B. Bauer, A. Tibi & U. Shavit
195
Cellular structure design for lightweight components

J. Nguyen, S.-I. Park & D.W. Rosen
203
No models, no moulds!
A.T. Estvez
211
Touchless gesture user interface for 3D visualization using the Kinect platform
and open-source frameworks
G.C.S. Ruppert, P.H.J. Amorim, T.F. Moraes & J.V.L. Silva
215
Multi-material blending for complex-shaped heterogeneous objects

I.T. Ozbolat & B. Koc
221
Gaussian Vault geometry: Integrated approach in design and fabrication of physical prototypes
M.P. Sharman & S. Datta
227
Application of contour tracing algorithm for assisting non-contact data acquisition

S. Rianmora, P. Koomsap & P. Kuagoolkijgarn
233
Image-based direct slicing of a single line drawing for rapid prototyping

N. Chansri & P. Koomsap
241
Design and additive manufacturing of cellular lattice structures

L. Hao, D. Raymont, C. Yan, A. Hussein & P. Young
249
The Corinthian capital according to Alberti: Generative modelling and digital prototyping
of classical architectural elements
E. Castro e Costa, F. Coutinho, J.P. Duarte & M. Krger
VI
255
Evaluation of different fitting algorithms using CMM and white fringe projection systems
F. Domingues, C. Silva, N.M. Alves, H.A. Almeida & P.J. Brtolo
263
Robust STL processing for extrusion-based manufacturing

M.B. Gaspar & N. Martins-Ferreira
273
Materials
A review of different techniques to characterise the mechanical properties of SLS parts focus
on resistivity measurements
E. Boillat, D. Fivat, Jamasp Jhabvala, M. Matthey & R. Glardon
281
Influence of hygrothermal aging on the mechanical properties of Nylon 12 composites processed

by selective laser sintering
R. Seltzer, J.S. Escudero & F.M. de la Escalera Cutillas
287
Structural and functional properties of NiTi shape memory alloys produced by Selective Laser Melting
H. Meier, C. Haberland & J. Frenzel
291
Investigation on the inclusions in maraging steel produced by Selective Laser Melting

L. Thijs, J. Van Humbeeck, K. Kempen, E. Yasa, J.P. Kruth & M. Rombouts
297
SMARTerials for high performance buildings

N. Lazarovich, G. Capeluto & M.S. Silverstein
305
Porous ceramic filters through 3D printing

A. Withell, O. Diegel, I. Grupp, S. Reay, D. de Beer & J. Potgieter
313
Production of alumina parts through Selective Laser Sintering of alumina-polyamide

composite powder
J. Deckers, S. Khuram, J. Vleugels, J.-P. Kruth & S. Boury
319
Material characterisation of Additive Manufacturing components made from a polyetherketone

(PEK) high temperature thermoplastic polymer
M.A. Beard, O.R. Ghita, J. Bradbury, S. Flint & K.E. Evans
329
Material study of laser cladded Inconel 625

M. Rombouts, G. Maes & R. Persoons
333
A preliminary investigation on Selective Laser Melting of M2 high speed steel

Z.H. Liu, C.K. Chua, K.F. Leong, K. Kempen, L. Thijs, E. Yasa, J. Van-Humbeeck & J.P. Kruth
339
Fatigue behavior of additive manufacturing parts. A preliminary analysis

J. Mungua, K.W. Dalgarno & R. Reid
347
Characterisation of carbon fibre reinforced nylon-12 composites for selective laser

sintering process
C. Yan, L. Xu, L. Hao & Y. Shi
355
New material development for laser additive manufacturing

L. Hao
359
Mechanical behavior of Epoxy-Aluminum composite for rapid tools applications

G.V. Salmoria, F.A. Yaez-Villamizar, A. Sabino-Netto & G.M.O. Barra
365
Thermal characterization of laser sintering of nylon-12

T.T. Diller, M.M. Yuan, D.L. Bourell & J.J. Beaman
369
Metallurgical, accuracy and cost analysis of Ti6Al4V dental coping fabricated

by electron beam melting process
W.P. Syam, A.M. Al-Ahmari, M.A. Mannan, H.A. Al-Shehri & K.A. Al-Wazzan
375
Thermal and dynamic-mechanical behavior of Fullcure 3D Printing Resin post-cured by

different methods
L.F. Vieira, R.A. Paggi & G.V. Salmoria
385
Rheological study of poly vinyl alcohol with two thermo-initiator for use in rapid prototyping
L.L. Lima, C.B.L. Ulbrich, C.A.C. Zavaglia, V.P. Bavaresco, J.G.M. Coelho & C.G.B.T. Dias
VII
389
Rapid Tooling & Manufacturing

A study on the mouldability of technical parts using hybrid moulds and structural foams
A.A. Nogueira, P.G. Martinho, A.M. Brito & A.S. Pouzada
399
Rapid tooling in metal forming processes using 3D-printed tools

S. Junk, R. Wagner, M. Trnkle & S. Ct
405
Conformal cooling in moulds with special geometry

M.A. Garca, C. Garca-Pando & C. Marto
409
Thermoplastic Resin Transfer Moulding in a rapid manufactured mould

E. Atzeni, F. Calignano, L. Iuliano, P. Minetola, A. Salmi, E. Bassoli, L. Denti & A. Gatto
413
A creative solution for rapid and inexpensive model-making in product design processes
F. Veloso, .M. Sampaio, F. Chaves & R. Simoes
423
Fabrication of a Biopsy Micro-Forceps prototype with Incremental Sheet Forming

R. Perez-Santiago, M. Garca-Romeu & I. Bagudanch
429
Using additive manufactured tooling in the fabrication of poly (L-Lactide-co-Glycolide) implants

S.F. Khan, K.W. Dalgarno & M.J. German
437
Developing Ti jewelry through additive manufacturing and conversion technologies

B. Paiva, R.J. Neto & J. Lino
443
Automation for building manufacturing

F. Craveiro, J.M. Matos, H. Brtolo & P.J. Brtolo
451
Additive and Nano Manufacturing Technologies

Optimization of Selective Laser Melting technology using design of experiments method
M. Averyanova, E. Cicala, Ph. Bertrand & D. Grevey
459
From functional specifications to optimized CAD model: Proposition of a new DFAM methodology
J.Y. Hascoet, R. Ponche, O. Kerbrat & P. Mognol
467
A novel approach to produce Functionally Graded Materials for additive manufacturing

P. Mognol, P. Muller & J.Y. Hascoet
473
Electrochemical micromilling with ultra short pulses

R. Zemann, F. Bleicher, C. Habersohn & R. Zisser-Pfeifer
479
Functionally Graded Rapid Prototyping

N. Oxman, S. Keating & E. Tsai
483
Continuous contour printing versus layer by layer printing in an Additive Manufacturing Technology
G. Vallicrosa, J. Delgado, L. Seren & J. Ciurana
491
Methodology for analyzing the depth of sintering in the building platform

J. Delgado, L. Sereno, J. Ciurana & L. Hernandez
495
Influence of process parameters in the first melting layer of a building platform in a SLM machine
J. Delgado, L. Sereno, J. Ciurana & L. Hernandez
499
Surface finishing techniques for SLM manufactured stainless steel 316L components
W. Morton, S. Green, A.E.W. Rennie & T.N. Abram
503
Variable Fused Deposition Modelling analysis of benefits, concept design and tool path generation.
H.L. Brooks, A.E.W. Rennie, T.N. Abram, J. McGovern & F. Caron
511
Innovative features in implants through Beam Melting a new approach for Additive
Manufacturing of endoprostheses
B. Mueller, T. Toeppel, M. Gebauer & R. Neugebauer
Investigating the influence of build parameters on the mechanical properties of FDM parts
J. Giannatsis, K. Sofos, V. Canellidis, D. Karalekas & V. Dedoussis
VIII
519
525
Porosity as a key to increase material properties of laser sintered parts

S. Rsenberg, L. Schmidt, H. Hosse & H.-J. Schmid
531
Software tools for rapid Prototype as Design

A. Withell, O. Diegel & S. Reay
539
Direct Manufacturing Design Rules

D. Zimmer & G. Adam
545
A priori process parameter adjustment for SLM process optimization

S. Clijsters, T. Craeghs & J.-P. Kruth
553
Surface roughness analysis in Selective Laser Melting

G. Strano, L. Hao, R.M. Everson & K.E. Evans
561
AsTeRICS a rapid prototyping platform for Assistive Technologies

C. Wei & C. Veigl
567
Comparison of five rapid prototype techniques (SLS/FDM/DLP/3DP/Polyjet)

C.B.L. Ulbrich, C.A.C. Zavaglia, P.I. Neto, M.F. Oliveira & J.V.L. Silva
573
Droplet impact dynamics in ink-jet manufacturing

W. Zhou, D. Loney, A.G. Fedorov, F.L. Degertekin & D.W. Rosen
581
Rapid Prototyping techniques for individualized medical prosthesis manufacturing

A. Fiorentino, G.P. Marenda, R. Marzi, E. Ceretti, D.T. Kemmoku & J.V.L. Silva
589
Towards surface topography: Quantification of Selective Laser Melting (SLM) built parts
A. Diatlov, D. Buchbinder, W. Meiners, K. Wissenbach & J. Bltmann
595
An Additive Manufacturing method based on xerography

Y.E. Tan & C.K. Chua
603
Preliminary investigation on cellular support structures using SLM process

A. Hussein, C. Yan, R. Everson & L. Hao
609
Microchannels fabrication in Direct Metal Laser Sintering (DMLS)

A.R.R. Bineli, A.L. Jardini, A.P.G. Peres, L.F. Bernardes & R.M. Filho
613
Knowledge based process planning and design for Additive Manufacturing (KARMA)
B. Singh & N. Sewell
619
A comparison of laser additive manufacturing using gas and plasma-atomized Ti-6Al-4V powders
M.N. Ahsan, A.J. Pinkerton & L. Ali
625
Thermoplastic filament extruder head for desktop Additive Manufacturing machines

P.I. Neto, A.L.L. Filho, F.D.A.S. Pereira, J.V.L. Silva & Z.C. Silveira
635
Manufacturing of defined porous metal structures using the beam melting technology
J.T. Sehrt & G. Witt
639
Virtual Environments and Simulation

An ontology-based automotive troubleshooting configuration system development
J.S. Liang
647
A tactile display for texture perception in virtual environments

M. Mengoni, M. Germani, B. Colaiocco & P. Morichetti
653
Virtual Reality technology as a support to the maintenance activity of buildings

A. Zita Sampaio & A.R. Gomes
661
Some studies on dislocation density based finite element modeling of Ultrasonic Consolidation
D. Pal & B.E. Stucker
667
Virtual prototyping of force-feedback robotic instruments for surgery

J.M. Gmez-de-Gabriel, V.F. Muoz, W.S. Harwin & A. Barrow
677
IX
A comparison between BioCAD and some known methods for finite element model generation
P.Y. Noritomi, T.A. Xavier & J.V.L. Silva
685
Towards simulation of a bioreactor environment for biofabricated tissue maturation

R.A. Rezende, C.A.R. Laureti, J.V.L. Silva, V. Mironov, V. Kasyanov & R.M. Filho
691
Prototyping process of a virtual-reality treadmill system for exploration of real world

panoramic environments
J. Hu, G.M. Fadel, I. Wood, P. Napieralski & S. Babu
699
Numerical simulations applied in a protocol for virtual prototype of a femoral prosthesis

stem fatigue life test
C.A.R. Laureti, D.T. Kemmoku, P.Y. Noritomi & J.V.L. Silva
709
BioCAD techniques: Example on maxilla for rapid expansion simulation

D.T. Kemmoku, C.A.R. Laureti, P.Y. Noritomi & J.V.L. Silva
715
Finite Element Synthesis

N. Oxman
719
Comparison of bone remodeling algorithms for hip implants

J. Frazo, H. Almeida, P. Brtolo & N. Alves
725
Computer modelling and simulation of reaction injection moulding: Filling and curing stages
R.T. Dias, A. Mateus, G.R. Mitchell & P.J. Brtolo
731
Warped hexahedral meshing of an ellipsoidal inclusion for design of composite material

L. Podshivalov, A. Fischer & P.Z. Bar-Yoseph
737
Applications
Rapid manufacturing of removable complete denture components
S. Wahab, N. Kassim & Z.A. Rajion
745
Investigation of wear behaviour of FDM fixtures

P. Minetola & L. Iuliano
749
The application of Laser Sintering for archaeological model-making

S.P. Soe, D.R. Eyers, A.T. Potter, T. Jones & N. Nayling
757
Novel concept for manufacturing lightweight centrifuge rotors for laboratories

U. Klaeger & V. Galazky
763
Use of rapid prototype techniques for large prosthetic cranioplasty

C.B.L. Ulbrich, C.A.C. Zavaglia, G.H.L. Paschoal, J.V.L. Silva & J.F.D. Zullo
767
A-footprint: Ankle and foot orthotic personalization via rapid manufacturing

J. Mungua, K.W. Dalgarno, J. Pallari & S. Cook
771
Physical prototypes in cross-functional team collaboration: A study of the Model-T2

concept car project
S. Datta, S. Hanafin, B. Rolfe & T. de Souza
779
Production of functional parts using SLM Opportunities and limitations

A.B. Spierings, G. Levy, L. Labhart & K. Wegener
785
The integration of solutions that evolved by nature, into innovative products of the future
R. Becker
791
Comparative analysis between a CAD model design and physical models obtained by
manufacturing additive technologies using optical scan
F. de Alencar & P.J. Brtolo
Rapid prototyping for original design
N.G. Harris & T.J. Coole
797
801
Integrated strategy for sustainable product development

A.L. Santos, H. Almeida, H. Brtolo & P. Brtolo
807
3D scanning and digital manufacturing technologies applied in research projects in archaeology

J.R.L. Santos, S.A.K. Azevedo, S. Belmonte, A.B. Junior & R.C. Fontes
813
Fabrication of a soft morphing structure using a Shape Memory Alloy (SMA) wire/polymer
skeleton composite
J.S. Kim & S.H. Ahn
819
Design and development of large rigid plastic packaging for Olives

M.F. Castro, C.I. Martins & A.J. Pontes
825
The potential for additive manufacturing in jewellery design

T. Ferreira, P. Brtolo & R.I. Campbell
829
Customization tool for people with special needs

B. Providncia, J. Ciurana & J. Cunha
837
Rapid prototyping in collaborative architectural design process

R. Pupo, R. Ruschel & M. Andrade
845
Author index
849
XI

Preface
Innovative Developments in Virtual and Physical Prototyping contains papers presented at the 5th International
Conference on Advanced Research in Virtual and Physical Prototyping (VR@P 2011), held by the Centre for
Rapid and Sustainable Product Development, School of Technology and Management of the Polytechnic Institute
of Leiria, Portugal. The Centre for Rapid and Sustainable Product Developed is a Centre of Excellence of the
Polytechnic Institute of Leiria. Its mission is to contribute to the advancement of science and technology leading
to more suitable, effective and efficient products, materials and processes, helping to generate added-value for
Industry, and to promote the awareness of the role and importance of rapid and sustainable product development
in society.
The International Conference on Advanced Research in Virtual and Physical Prototyping was designed to
be a major forum for the scientific exchange of multi-disciplinary and inter-organisational aspects of virtual
and rapid prototyping and related areas, making a significant contribution for further development of these
fields. It joined participants from more than 20 countries. Such diversity was parallel to the various multidisciplinary contributions to the conference, whose subjects enclose a wide range of topics like CAD and 3D Data
Acquisition Technologies, Additive and Nano Manufacturing Technologies, Rapid Tooling & Manufacturing,
Biomanufacturing, Materials for Advanced Manufacturing Processes, Virtual Environments and Simulation,
Applications of Virtual and Physical Prototyping Technologies, all of them making a significant contribution
for future development. This research community has been strongly engaged in the development of innovative
solutions to solve Industrys problems, contributing to a more pleasant and healthy way of living.
I am deeply grateful to authors, participants, reviewers, the International Scientific Committee, Session chairs,
student helpers and Administrative assistants, for contributing to the success of this conference. The conference
was endorsed by:
The Polytechnic Institute of Leiria (IPL)

The Centre for Rapid and Sustainable Product Development (CDRsp)
Portuguese Foundation for Science and Technology
The Global Alliance of Rapid Prototyping Associations (GARPA)
XIII

Committee Members
CONFERENCE CHAIR
Paulo Brtolo,
Centre for Rapid and Sustainable Product Development
Polytechnic Institute of Leiria
HONORARY CHAIRS
Alain Bernard
cole Centrale de Nantes, France
Gideon Levy
University of Applied Sciences St. Gallen, Switzerland
Jean-Pierre Kruth
KU Leuven, Belgium
Terry Wohlers
Wohlers Associates, USA
Yongnian Yan
Tshingua University, China
INTERNATIONAL SCIENTIFIC COMMITTEE

Anath Fischer
Technion, Israel
Antnio Pontes
University of Minho, Portugal
Antnio Pouzada
Alain Bernard
cole Centrale de Nantes, France
Bernard Hon
University of Liverpool, UK
Bopaya Bidanda
University of Pittsburgh, USA
Brent Stucker
Utah State University, USA
Chua Chee Kai
Nanyang Technological University, Singapore
Ciro Rodriguez
Tecnolgico de Monterrey, Mxico
David Bourell
Texas University, USA
David Rosen
Georgia Institute of Technology, USA
Dietmar Hutmacher
Queensland University of Technology, Australia
Dimitri Dimitrov
University of Stellenbosch, South Africa
Duc Pham
Cardiff University, UK
Elisabetta Ceretti
University of Brescia, Italy
Gideon Levy
University of Applied Sciences St. Gallen, Switzerland
Goran Putnik
Ian Campbell
Loughborough University, UK
Ian Gibson
National University of Singapore,
Singapore Center for Rapid and Sustainable
Product Development, Portugal
Imre Horvth
Delft University of Technology, Netherlands
Jean-Pierre Kruth
KU Leuven, Belgium
Joaquim de Ciurana
University of Girona, Spain
Luca Luliano
Politecnico di Torino, Italy
XV
Ricardo Gonalves
New University of Lisbon, Portugal
Richard Hague
Loughborough University, UK
Rubens Maciel Filho
University of Campinas, Brazil
Suman Das
Georgia Institute of Technology, USA
Steinar Killi
Oslo School of Architecture and Design, Norway
Tahar Laoui
King Fahad University of Petroleum & Minerals,
Saudi Arabia
Terry Wohlers
Wohlers Associates, USA
Tugrul Ozel
Rutgers University, USA
Wei Sun
Tsinghua University, China
Xavier Fischer
ESTIA, France
Yuan-Shin Lee
North Carolina State University, USA
Jorge Vicente Lopes da Silva
CENPRA, Brazil
Jukka Tuomi
Helsinki University of Technology, Finland
Jung-Hoon Chun
Massachusetts Institute of Technology, USA
Kenneth Dalgarno
Newcastle University, UK
Mamoru Mitsuishi
University of Tokyo, Japan
Ming Leu
Missouri University of Science and
Technology, USA
Olaf Diegel
Auckland University of Technology, USA
Olivier Jay
Danish Technological Institute, Denmark
Rajurkar Kamlakar
University of Nebraska-Licoln, USA
XVI
Keynotes

Product
Idea 2
Lab A low cost alternative to introduce AM in South Africa
D.J. de Beer
Technology Transfer and Innovation: Vaal University of Technology, Vanderbijlpark, South Africa
ABSTRACT: South Africa in general, but more specifically poverty-stricken areas such as the Southern
Gauteng, is challenged with low levels of skills development, underpinned by insufficient job opportunities.
As part of an innovation and job creation strategy-driven solution, the VUT introduced an Idea 2Product Lab
where, as a strategic intervention, individuals from the region can be provided with appropriate skills development; infrastructure for entrepreneurs to develop new products that can be tested and modified in the market
place according to customer needs; increased adoption and transfer of technology and research and development
into new tools for entrepreneurs. The project (through its dedicated and virtually linked Lab facilities) also
provides infrastructure to produce small batches of niche products.
1
1.1
INTRODUCTION
Historical development and positioning of the
Product
Lab initiative at VUT
Idea 2
In a recent article reflecting on Additive Manufacturing (AM) development in South Africa, Campbell
et al [Campbell et al, 2011] points out that the RSA
has embraced Rapid Prototyping (RP) as a method
of revolutionizing its industry. They concluded that
good progress has been made against a previously
reported road map and that current development plans
should remain in place, with the addition of a greater
educational dimension.
In this paper, the author reports on the development
of a unique Idea 2Product Lab at the Vaal University of
Technology, and how it bridges the gap left between
the existing SA AM industry/landscape, and the current SA Government-funded FabLab. The paper goes
on to identify future lessons that can be learnt and
proposes further steps to be included in the AM development road map, and more specifically how to use
the Idea 2Product Lab to stimulate educational involvement, as well as how to attract small and micro-scale
companies towards AM usage.
Figure 1. Caption of the SA AM landscape from 1991

2011.
level AM machines (driven by machine cost, ease of

operation and the very low cost of consumables), are
driving a 3rd generation boom in the SA AM industry, accounting for approximately 25%30% of the
available technologies in SA.
1.3 SA FabLab initiatives history
The South African Department of Science and Technology (DST), under its Advanced Manufacturing
Technology Strategy (AMTS), started FabLabs in
different regions of the country with a two-prompt
approach, namely to encourage high-level industrial
Research and Development (R&D) on the one hand,
with product invention, skills transfer, job creation and
solutions to crossing the digital divide, on the other
hand (Le Roux, 2006).
The SA initiatives were part of an international rollout by the Massachusetts Institute of Technologys
(MIT) Centre for Bits and Atoms (CBA) as part of their
educational outreach programme, and include involvement (and support from MIT) in India, Norway, the
US, Ghana, Japan, Spain and Costa Rica at that time.
Whilst the FabLab programme as implemented in SA
1.2 Relevance of the current SA AM landscape

The rapid growth in SA RP machine installations
observed in 2004, continued in subsequent years as
shown in Figure 1, with an increase in the dominance
of 3D printers. As an example, 92% of the RP systems
sold during 2007 were 3D printers. This trend continued in subsequent years, and by mid 2011 (Fig. 1),
94% of the estimated 270 machines available in the
RSA can be reported as 3D printers. The predominant
reason for this would appear to be cost, but improved
part quality and easier materials handling have also
played a role (Wohlers, 2008). Currently, the entry
The obvious chasm is that although research driven

by high end systems resulted in enabling and innovative applications, it did not really promote new
system/materials development thus far (again due to
obvious reasons), and did not really allow enough education (as well as exploration and investigation) of
junior or senior students on inexpensive/lower cost 3D
printers.
stimulates the development of (functional) 3D prototypes, it uses off-the-shelf industrial grade fabrication
and electronic tools to create layers of e.g. laser cut
material that can be added manually, opposed to AM
platforms.
2
COST AS A DRIVER FOR FUTURE GROWTH
2.1 AM system cost as driver of ownership and

application
2.2 Emerging of Entry Level Rapid Prototyping

(ELRP) systems
Towards the end of 2008, the author reported (De Beer,

2008) that of the approximately 140 AM systems available in SA, 88% were 3D printers. For the total number
of machines however, it is estimated that the replacement value (calculated on the same years machine
costs) was approximately US$9.25 million.
Replacement costs of high end machines constitute
approximately 54% of this cost, whilst the replacement
costs of 3D printers constitute the remaining 46%.
Bearing in mind that high end machines only constitute
12.5% of the total installations, the remaining 87,5%
constitute 3D printers. A further important part of the
analysis shows that by far, university/research institutions own the high-end systems, whilst again by a
significant majority, industry owns the 3D printers.
Whilst being an excellent indicator that universities/
research institutions drive the RSAs research agenda,
it may simultaneously signal a further weakness.
Due to the very high installation/acquisition and running costs, the normal trend is that only students in
advanced stages of an undergraduate degree or even
postgraduate degree, have access to these systems.
As for industry ownership, the opposite is obviously true. Due to a much higher investment in 3D
printing systems, owners/users of these technologies
may end up having a much more hands-on approach,
which may even result in industry making a more significant contribution to training (especially hands-on
training) of the lower end users. It does however also
implicate that industry is still using AM technologies
in a concept/form-fit-function/master pattern mode,
opposed to direct manufacturing.
Various critical signals can be derived from the
above-mentioned analysis:
Up until recently, 3DP has been seen as the entry

level RP process (Pei et al 2011). As reported by
Dimitrov et al (2006) and Wohlers (2010), 3DP has
been recognised as a very competitive process in
terms of cost and speed, and the number of 3D printing machines sold during 2009 had risen by 18% over
2008. 3DP systems from Z-Corporation have a fast
fabrication time, a low material cost, the option of
colour, and build parts that can be used directly for
casting (Azari and Nikzad, 2009). It was also reported
that if companies were to invest in RP, the most popular choice (28%) would be 3DPs (Strategic Direction,
2009).
Despite significant improvement across the AM
technology industry, Wohlers (2003) indicated that for
AM to better penetrate new markets, systems must
become more affordable, easier to use and simpler
to maintain. More importantly, the cost of ownership must drop further and quality of materials must
improve. As a timeous response, Entry Level Rapid
Prototyping (ELRP) systems came forward to provide
a significantly lower-cost alternative for entry level
users. ELRP machines are now gaining popularity, and
whilst during the early/initial stage were not taken seriously by the established OEMs, have made significant
progress in terms of ease of use, with as a significant
improvement in part quality.
Over and above the very reasonable acquisition costs (ranging from approximately $520$4000),
ELRPs also offer material costs that resemble typical
raw polymer material costs, instead of highly inflated
(and electronically chipped) cartridges of material that
allows testing, experimentation and exploration, and
which offer a low risk entry to education and innovation. A further development within the ELRP systems
group is the so-called DIY AM systems. Figure 2
shows the RAPMAN 3D system as sold by Bits From
Bytes in the UK, and which is despatched in kit form,
to be assembled upon arrival. The RAPMAN 3D system currently makes a significant contribution to the
development of the SA AM landscape, with reported
successes in academia and industry.
Comparisons of six systems under review are shown
in Table 1 (used in the VUTs Idea 2Product Lab ) and
Table 2 (part of a literature review for comparison
purposes).
The common denominator here is that the systems
are fairly open w.r.t. architecture, and programming/
control is open for user adjustment and experimentation. A further commonality is that the material
on the one hand side, academia is leading the

way and introducing the high end systems and
applications;
due to implementation costs, although the benefits
are significant, industry do not want to risk the necessary investment for high end systems, but rather
choose to invest in lower risk 3D printing systems;
apart from participating in government supported
technology transfer, pilot or contract research
projects, industry remains on the lower of the value
chain in terms of AM applications;
although academia is driving the demonstration of
high end systems, not enough take up realises;
in a similar mode than the above-mentioend, not
enough students are being trained (especially due to
the risk involved in damage of high end systems);
Table 1. Comparison of cost and technical data for the

systems used in the VUTs Idea 2Product Lab .
System
RapMan 3D
(Model 1)
BFB 3000
UP! (PP3DP)
US$1495
Thermoplastic
Filament
Extrusion
Positioning Cartesian
(X, Y, Z)
3-axis
system
US$3895
Thermoplastic
Filament
Extrusion
Cartesian
(X, Y, Z)
3-axis
system
US$2690
Thermoplastic
Filament
Extrusion
Cartesian
(X, Y, Z)
3-axis system
Input
Type
Maximum
print
dimensions
Print
resolution
SD Card
SD Card
275 mm
205 mm
210 mm
0.1 mm
275 mm
275 mm
200 mm
0.125 mm
140 mm
140 mm
135 mm
0.2 mm
Printing
Speeds
7 mm3
17.0 mm3
per second
7 mm3
15 mm3
per second
27.778 mm3
per second
Materials
ABS,
HDPE,
LDPE, PP,
uPVC
(in 3 mm
filament)
ABS,
HDPE,
LDPE, PP,
uPVC
(in 3 mm
filament)
ABS (in
1.7 mm
filament)
Cost
Build
Process
Figure 2. Caption of the RAPMAN DIY 3D printing

system.
was implemented. Off-the-shelf equipment generally

found in FabLabs are laser cutters for the cutting of
two dimensional parts that can be linked or joined
using clever slot and clip-on mechanisms which can be
designed as part of developing an intelligent system,
or on the very basic level, glued. Further additions
are high resolution desk-top milling machines for the
manufacturing of circuit boards, precision parts or
moulds (typically for resin gravity casting/Reaction
Injection Moulding (RIM) or Vacuum Casting of parts,
as well as a sign cutter that plots in copper for the fabrication of flexible circuit boards, antennas, etc., (Le
Roux, 2006).
Le Roux continue to point out that at the current moment, the SA FabLab facilities do not use
fabbers, but instead employ machines that make
two-dimensional parts that can be assembled as
into three dimensional parts. Le Roux continues to
debate that besides assisting the CBA to explore the
potential application in the developing world, the
programme is also assisting governments and communities to empower communities with technology. As
the FabLab provides a rapid prototyping platform,
local entrepreneurs are encouraged take their own
ideas from the drawing board to prototypes and start
their own businesses
Whilst all of the envisioned benefits, such as to:
Figure 3. Caption of the UP! PP3DP printing system.
used is extruded polymer filament, available under

US$ 50/kg a significantly reduced price compared to
the existing commercial 3D Printer category of materials, and as such, very applicable to the educational
market. Local manufacturing of consumables remains
an option. Figure 3 shows a caption of the UP! Personal
Portable 3D Printer (PP3DP).
2.3
Relevance of ELRP systems with regards to the

FabLab philosophy
The FabLab programme brought prototyping requirements to under-serviced communities in SA, as well
as in other places world-wide where the programme
create a platform from which a communitys technical challenges can be shared with an international
the creation of capacity to support local FabLabs can

be structured to involve and benefit stakeholders. The
table was thus set with an unspoken challenge for
local academia, science councils and/or government
programmes to come up with local developments for
creation of a sustainable model.
In taking the idea of manufacturing education and
exposure of school learners, adult basic education and
training programmes or students a step closer to the
reality of creation of new inventions to solve reallife problems and also turning these into small new
businesses, and important link can be made between
the existing FabLab model, and ELRP technologies.
Apart from the micro-electronic design and manufacturing facilities found in FabLabs, the author is of the
opinion that the new generation ELRP technologies as
described in section 2.2 yield the potential to successfully complement or even replace more conventional
technologies found in a traditional FabLab. The higher
the numbers of these new generation (low cost) AM
machines, the closer one gets to the ideal solution
wanted for training groups of individuals from the designated groups as identified above.Also, incorporating
the proven potential of the FabLab micro-electronics
facilities, the initiative under investigation and discussion (the VUTs Idea 2Product Lab ) offers a fresh,
locally developed approach that can support the complete value chain starting from education of AM,
to producing final grade parts or products in small
batches. Linking more industrial level technologies
(both higher-end AM technologies, as well as conventional manufacturing technologies) in the very same
FabLab philosophy, but used in a virtual linked centre through existing and planned infratsructure, the
possibilities becomes endless to produce final grade
functional parts and products.
Based on the above-mentioned, as well as the
aspect of uncertainty in terms of further government
roll-out of existing FabLab models in South Africa,
together with the uncertainty of whether an individual/institution may roll out a FabLab on own initiative,
the VUT took a bold step to learn from the existing models, add the experience gained in developing
a Design and Advanced Additive and Digital Manufacturing Centre, to produce a state-of-the-art unique
Product
Lab , that can continue along the initial
Idea 2
goals set by the FabLabs, but to reach even further.
Table 2. Comparison of cost and technical data for

similar systems as used in the VUTs Idea 2Product Lab .
System
Fab@Home RepRap II:

(Model 1)
Mendel
Makerbot
CupCake CNC
Cost
US$2400
US$520
US$750
Build
Process
Syringebased
extrusion
Thermoplastic Thermoplastic
Filament
Filament
Extrusion
Extrusion
Positioning
Cartesian
(X, Y, Z)
3-axis
system
Cartesian
(X, Y, Z)
3-axis
system
Cartesian
(X, Y, Z)
3-axis system
Input Type
USB
Connection
USB
Connection
SD Card
and USB
Connection
Maximum
print
dimensions
203 mm
203 mm
102 mm
200 mm
200 mm
140 mm
100 mm 100
mm
130 mm
Print
resolution
0.1 mm
0.1 mm
0.08mm
Printing
Speeds
Depending
on fluidity
of material
15.0 cm3
per hour
X/Y feed rate

5000 mm per
minute
Z feed rate 200
mm per minute
Materials
Any kind
of liquid
or paste
that can be
dispensed
from a
syringe
ABS, PLA,
HDPE, etc
(in 3 mm
filament)
ABS, PLA,
HDPE, CAPA
(PCL) (in
3 mm filament)
contingent of engineers or other users that can help

them to assist problems using own experience;
enable communities to tackle specific problems or
challenges and create a local network in which
the individuals and SMMEs can share ideas and
brainstorm solutions;
support manufacturing in rural areas;
expose children, school learners and students to the
use of computers and machinery to make (manufacture) goods; etc.
became a reality, the SA programme roll-out has
stagnated with the eventual FabLabs, together with a
mobile version.
Further ideals were to create a Fab certificate
an informal advanced technical qualification, together
with the possibility that FabLab users should be able to
gain recognition of skills sets developed and demonstrated while using the FabLab facilities. On a local
front, formal recognition of such competencies should
assist in opening up employability options or even
position the candidates to progress up the skills ladder (Le Roux, 2006). All along, the idea was that SA
should be pioneering best practise for the implementation of FabLabs worldwide by demonstrating how
2.4 Current Idea 2Product Lab infrastructure and

operational model
Based on initial research into and experience gained
in using a RAPMAN DIY kit from Bits From Bytes in
the UK, together with research conducted to find applicable (industrial grade) CNC equipment and a Laser
Cutter, internal (institutional) funding was sourced
from the VUT to establish an Idea 2Product Lab (including registering the Trade mark name). Through exposure gained of the UP! PP3DP during the 1st NZ RPD
Conference held in Auckland, 78 February 2011, the
initial planning of the equipment list was completed.
Figure 4. Caption of VUTs Idea 2Product Lab AM facilities.
Figure 6. Caption of the Table-top industrial grade CNC

lathe.
Figure 5. Caption of the table-top CNC lathe.
2.5
Existing Idea 2Product Lab Infrastructure
Currently, the Idea 2Product Lab has the following

equipment and infrastructure:
Four RapMan 3.1 3D Printer kits;

Six BFB 3000 Plus 3D Printers;
Ten UP! PP3DP machines;
Twenty CAD workstations;
LG laser cutting and engraving machine;
SP 2120 PC-based CNC Lathe with MACH3 system;
SP 2213 PC-based CNC Milling Machine with
MACH3 system;
Applicable hand tools and Micro-electronics workstation.
Figure 4 shows a caption of the AM unit within
the VUTs Idea 2Product Lab , whilst Figures 5, 6 and 7
shows captions of the table-top industrial grade CNC
lathe, table-top industrial grade CNC milling machine
and Laser Cutter.
2.6
Figure 7. Caption of the industrial grade Laser Cutter.
Sketchup, Sculptress and Autodesk 123D last

mentioned offers full solid modelling capbilities. For
export to .STL files from Google Sketchup, MeshLab
v1.3 is used (also open source). Autodesk 123D has
the advantage that it can export .STL files (as well as
a host of other useful files formats, such as STEP or
DWG formats).
Figure 8 shows a caption of a chair created in
Autodesk 123D, Figure 9 shows a caption of
the .STL file exported with Autodesk 123D and
CAD software
Due to the fact that the Idea 2Product Lab will predominantly serve students, school learners and individuals
from the VUTs surrounding communities (and still
within the spirit of the FabLab development) open
source (free ware) CAD will be used. Currently,
the Idea 2Product Lab is experimenting with Google
in MiniMagics2 . Over and above viewing the .STL

files, MiniMagics2 offers the following benefits:
Figure 10 shows a caption of the model printed with

an UP! PP3DP machine.
Materialise also offers an .STL viewer as freeware (MiniMagics2 2.0), which can be used to varify
or fix .STL files. In various design files exported,
Autodesk 123D showed no bad edges when viewed
Import and save .STL, .magics or .mgx files

Compress STL files up to factor 20
View parts and rotate, pan or zoom
Detect bad edges and flipped triangles
Make a section
Request part information, such as XYZ measurements, volume, surface area or number of bad
edges
Measure parts
Add text annotations
Create a print-out of parts
See all information included in the magics file (like
pictures, annotations, etc.)
IDEA 2
PRODUCT
LAB PLANNING
3.1 Positioning of the Idea 2Product Lab to impact on

national imperatives/strategies
Figure 8. Caption of a chair created in Autodesk 123D.
South Africa in general, but more specifically povertystricken areas such as the Southern Gauteng, is
challenged with low levels of skills development,
underpinned by insufficient job opportunities.
As part of the solution, the VUT proposes the
introduction of an Idea 2Product Lab where, as a strategic intervention, individuals from the region can be
provided with:
appropriate skills development;

infrastructure for entrepreneurs to develop new
products that can be tested and modified in the
market place according to customer needs;
increased adoption and transfer of technology and
research and development into new tools for
entrepreneurs
3.2
Broad aims and objectives
The project (through its dedicated and virtually linked

Lab facilities) also provides infrastructure to produce
small batches of niche products. The following benefits are foreseen as part of the projected project
outcomes:
Figure 9. Caption of the .STL file created in Autodesk

123D, and viewed in MiniMagics2 .
Figure 10. Caption of the chair as printed on the UP! PP3DP

printing system.
To create jobs and help the regions very poor

communities improve their poverty situation;
To concentrate on high value-added manufacturing
initiatives in the Southern-Gauteng region;
To support people doing buisness in the 2nd economy to move to the main stream in the SouthernGauteng region;
To develop human resources and build multi
inter trans-disciplinary or cross-cutting skills and
research base;
To facilitate the development of new industries
and product development initiatives by encouraging
joint research and community interaction;
To develop and support competitive small enterprises; and
To provide an enabling infrastructure for innovators.
The VUTs Idea 2Product Lab is available to:
digital revolution will go beyond the current wave of

personal computation to personal fabrication. Ideas
on how South Africa could move in the same direction to facilitate support and infrastructure for learners
and local communities to keep up with development
were based on a much wider acceptance and rollout of the FabLab programme. The VUTs Idea 2Product
Lab now provides a new home-grown (scaleable and
as such more economic) model that will empower
people through having access to personal fabrication
through the availability of entry level AM platforms.
The Idea 2Product Lab should be seen as a catalyst to
create an innovation culture within the host university
(also future hosts when the project is duplicated) and
in the relevant region. The existing ideas that innovation can be born by blending indigenous knowledge
with technological inputs from the developed world,
are getting more substance with the VUTs Idea 2Product
Lab initiative. Key to this idea will be to facilitate
proper understanding (and protection) of the indigenous knowledge through applicable infrastructure and
processes, and to assist South Africa to decentralize
manufacturing, and to excel to the level of personalised
manufacturing, led by local innovation and innovative
product development.
Innovators;
Students;
School learners;
SMEs;
Industry;
Entrepreneurs;
In short, the VUTs Idea 2Product Lab is available to any

member from the local communities in need of support
for the development of a specific idea.
3.3 Training objectives

The VUTs Idea 2Product Lab aims to follow a holistic
product development approach, as the focus will be
on manufacturing of the complete product including
design, fabrication, testing and debugging, monitoring and analysis and documentation of the process.
As such, it will provide/facilitate the availability of
market-ready products. It will also facilitate peerto-peer learning, which will contribute to personal
development of the participating students. This is also
a significant attribute to the projects sustainability.
The lab and its infrastructure are not just about material manipulation, but also allow for rapid development
of innovative concepts. In addition, rapid learning is
possible and innovations can be prototyped in hours
by literally anybody.
The VUTs Idea 2Product Lab therefore also serves as
training infrastructure that will teach students, innovators, entrepreneurs and business people how to use
the tools and to come up with local innovation models, as real innovation should come from within. As
such, a home-grown model of the MITs how to make
almost anything will be developed within the VUTs
Product
Lab . The VUTs staff and students will
Idea 2
be doing ongoing research to introduce new tools and
facilities to the lab that can assist users to make almost
anything and can be turned into a very powerful marketing brand for the Idea 2Product Lab . It will
also serve as a research platform to stimulate innovation and development of new AM technologies or
platforms.
The VUT is also using the Idea 2Product Lab to
engage with other universities (both nationally and
internationally) to participate in basic research for the
Product
Lab , and then to introduce new innovaIdea 2
tions to benefit the local community or solve local
industry problems. It furthermore will support the
overarching goal to instil innovation as core curriculum in all learning programmes. (Various international
institutions have already reacted on the initial presentation of the concept, and would like to engage in
collaborative development work).
3.4
PROJECT RESULTS
Various positive results can be reported since the

inception of the Idea 2Product Lab project, and will be
discussed below.
4.1 Using the Idea 2Product Lab as extension of

undergraduate laboratories
The Technology Transfer and Innovation directorate
of the VUT was approached to make ELRP equipment available for undergraduate teaching of advanced
manufacturing processes in the Industrial Engineering course. Instead, students were invited to use the
Product
Lab to execute design projects, and in such
Idea 2
a way to understand the impact and role of Design
and AM in advanced manufacturing process development. Figure 11 shows a caption of the first students
attending the Idea 2Product Lab Saturday session, whilst
Figure 12 shows some first products built by students,
using the facilities in the Idea 2Product Lab . A further
point of interest is that students with no prior exposure
to AM took on significant challenges. More importantly, these students devoted a Saturday morning to
attend the Idea 2Product Lab , which indicates the importance of the facility in orientating students towards
advanced processes. Furthermore, the the students
were amazed by the potential of the Idea 2Product Lab ,
through experiencing designed products, action figures and other creatures or devices appearing on
the ELRP machines.
Figure 13 shows a hinged boat rudder developed
by a student from the Electrical Engineering Masters
group.
Planned local economic impact
With the advent of the FabLab development, it was

anticipated that internationally, the next phase of the
Figure 14. Serviete holder developed as part of a student

design competition.
Figure 11. Caption of the first students using VUTs

Product
Lab AM facilities.
Idea 2
FUTURE ACTIVITIES
5.1 Competitions to stimulate ideas and innovation

The VUTs Innovation Centre is establishing an IdeaLAB, as part of the Idea 2Product Lab to simulate
innovative ideas as part of the innovation value chain.
A key enabler will be student- and local/national
competitions to stimulate ideas and innovation.
5.2 Evaluation the applicability of the Fab
degree
Figure 12. Caption of some student projects.
Following the successful development of Fab labs,

the Fab Academy was launched to provide access
to advanced instruction for students in these labs
exceeding the educational resources locally available to them. It uses innovative instructional methods
and media such as online video collaboration, which
makes global lecturing (and access to global expertise)
possible.
According to information available from the Amsterdam Fab lab, the Fab Academy offers Certificates
on relevant technical topics, and a Diploma aimed
at vocational and technical training for employment
and investment (along with assistance to its graduates in those areas). These are accredited by the Fab
Academy; it is anticipated that they will be followed
by a Bachelors degree to be offered under applicable regional accreditation, along with post-graduate
study.
A very interesting facet of the Fab Academy is
that its Certificate programmes provide familiarity with technical options and capabilities, handson experience, and direction for further study. Each
requires, and is evaluated by, developing and documenting projects, and are periodically renewed to
reflect best practices.The Certificates typically require
12 weeks, and are combined for the awarding of a
Diploma. According to the Fab Academy, Diploma
equates to MITs rapid-prototyping course How
To Make (almost) Anything. Furthermore, learning
Figure 13. Caption of a hinged boat rudder developed in

the Idea 2Product Lab AM facilities.
4.2
Facilitating a student product design

competitition entry
As part of a student product design competition, one of

the Design Interns developed a product similar to and
for evaluation by the Carol Boyes production range.
The serviet holder as shown in Figure 14 was designed
in Sculptress, and built in the Idea 2Product Lab , using
the UP! PP3DP.
10
progress is evaluated by skills and projects rather than

time or credits.
Linking to the Short Learning Programme (SLP)
activities within the Enterprise Development Unit
(EDU) within the VUTs Technology Transfer and
Innovation (TTI) Directorate, and following the
existing South African Recognition of Prior Learning (RPL) principles, the applicability of a similar
Product
Lab certificate and skills sets will be
Idea 2
evaluated, for possible future inclusion (Amsterdam
FabLab).
5.3
In addition, the Idea 2Product Lab has already moved

beyond just being a group or combination of exciting
technologies, as intelligent control systems is being
developed by participating post-graduate systems, and
which will aid in the management and operation of
the facility.
Significant progerss has been made since its establishement, and the initial goals have been achieved in
a reasonably short time. As such, the project has fulfilled its initial development goals, and is set to support
the next step, being local AM platform and materials
development.
Funding and association/commercialisation
Currently, the Idea 2Product Lab is a registered entity

of the VUTs TTI. First initiatives are being taken to
link it to regional development programmes, through
a combination of funding possibilities from regional
and provincial government, supported by corporate
foundation funding opportunities.
As such, the Idea 2Product Lab initiative is becoming
a commercial knowldege product, and the VUT is
ready to deliver this to its associates, but also to partner
universities. The Idea 2Product Lab also currently creates possibilities to be associated with renowned CAD
and other software companies, which will support its
standing in the local and internatioanl communities.
6
Expose children, school learners and students to the

use of computers and machinery to make (manufacture) goods.
REFERENCES
Amsterdam FabLab. http://fablab.waag.org/content/fabacademy. Accessed 29 July 2011.
Azari, A and Nikzad, S. 2009. The evolution of rapid prototyping in dentistry: a review. Rapid Prototying Journal,
15(3): 216225.
Campbell, R.I., De Beer, D.J, and Pei, E. 2011. Additive manufacturing in South Africa: building on the foundations.
Rapid Prototyping Journal 17(2): 156162.
De Beer, D.J. 2008. SA RP 2008. In Wohlers, T. 2008. Wohlers
Report 2008: State of the Industry Report Annual Worldwide Progress Report, Wohlers Associates, Fort Collins,
CO.
Dimitrov, D., Scherve, N. and De Beer, N. 2006. Advances
in three dimensional printing state of the art and future
perspectives. Rapid Prototyping Journal, 12(3): 136147.
Le Roux, H. 2006. Fabulous Fabrication. Engineering News,
26(27): 1617, 79.
Pei, E., Campbell, R.I. and De Beer, D.J. 2011. Entry-level
RP machines: how well can they cope with geometric
complexity? Assembly and Automation, 31(2): 153160.
Strategic Direction. 2009. The maturing of rapid prototyping:
market for additive fabrication pases $1 billion, Strategic
Direction, 25(8): 3840.
Wohlers, T. 2003. Wohlers Report 2003: State of the Industry Report Annual Worldwide Progress Report, Wohlers
Associates, Fort Collins, CO.
Wohlers, T. 2010. Wohlers Report 2010: State of the Industry Report Annual Worldwide Progress Report, Wohlers
Associates, Fort Collins, CO.
CONCLUSIONS
In evaluation the development approach taken, it is

clear that the initial objectives have been achieved.
The Idea 2Product Lab , although still new, already has
proven that it can:
Offer a platform from which the universitys expertise be shared with individual users to the extent
that it can help them to assist problems using own
experience;
Offer such individuals the opportunity to develop
much-needed skills-sets;
Offer a platform for a local interest group to become
active in peer-to-peer learning, user support and
networking to stimulate innovative thinking and to
cerate innnovative solutions for local problems;
11

Additive Manufacturing-assisted scaffold-based Tissue Engineering

C.K. Chua, M.J.J. Liu & S.M. Chou
School of Mechanical and Aerospace Engineering, Nanyang Technological University, Singapore
ABSTRACT: Additive manufacturing (AM) or rapid prototyping (RP) has been widely used for the fabrication
of tissue substitute in orthopedic and craniofacial surgeries. The AM techniques are used so as to eliminate the
medical complications arising from the use of allografts and autologous tissues. To date, AM methods can be
classified into several classes such as melt-dissolution deposition, particle-bonding, indirect and photosensitive
techniques. These approaches are commercially available or custom-built AM systems and they have been shown
to display micro-scaled features which are advantageous for cell migration, attachment as well as mass transportation of essential nutrients within the tissue engineered construct. In contrast to traditional tissue engineering
(TE) approaches, AM-built constructs have significantly improved the cell proliferation rate in the interior of
the tissue scaffolds. Modern AM approaches have since emerged as potential TE solutions to construct complex
tissues and organs by the inclusion of living cells suspension or growth factor(s) as raw materials into the AM
technology. The newly developed AM methods, which are also grouped as bioprinting and organ printing, are
still at its infancy as reflected by existing literature. Nevertheless, more innovative AM-related approaches are
anticipated in the future and, as suggested by the author, functional CAD-based modeling may possibly be the
next state-of-the-art to AM technology.
1
INTRODUCTION
TE methods, AM techniques remain a preference to

the manufacture of tissue constructs due to its high
reproducibility and production efficiency.
In a general perspective, medical professionals, scientists and engineers have been making significant
research effort to provide the best solution to tissue
repair and regeneration in orthopedic and craniofacial
surgeries. In orthopedic applications, musculoskeletal
tissues such as cartilage, bone and osteochondral sites
have unique extra cellular matrix (ECM) arrangements
which allow them to withstand physiological loadings
and subsequently execute their specific functionalities.
On the other hand, craniofacial tissue defects associated with the human ear, nose, cranial and dental
tissues can affect the quality of lives in both physical and psychological aspects (Zuk 2008, Warren et
al. 2003). These clinical scenarios thus motivated the
relevant TE professionals to seek options to resolve
issues in the biological and engineering areas. The
typical requirements of a TE scaffold, regardless of
the field of surgery, are: (1) the ability to mimic
the anatomical geometry of the native tissue for the
essential mass transport mechanisms, (2) providing
sufficient mechanical integrity so as to allow the regain
of functionality rapidly, (3) appropriate degradation
rate that matches the tissue regrowth rate and, (4)
non-toxic by-products so as to eliminate medical complications during the tissue healing process (Yang et al.
2001, Sachlos et al. 2003a). These requirements, in
short, can be addressed individually or in combinations depending on the area of focus. AM has been
explored extensively in order to meet these requirements. Existing AM techniques can be distinctively
In recent decades, tissue engineering (TE) has emerged

as a multi-disciplinary field for both biomedical and
biotechnological purposes such as drug delivery systems for pharmaceutical administrations, prosthetics and scaffolds for tissue repair and regeneration.
Particularly, the development and application of tissue scaffolds mainly for orthopedic and craniofacial
surgeries via additive manufacturing (AM) or rapid
prototyping (RP) has been identified as one of the
classes of bio-fabrication methods within the TE field
(Yang et al. 2002, Bartolo et al. 2009). AM for TE
is the manufacturing process of biological substitutes
in a layer-by-layer manner in conjunction with the
use of computer-aided design (CAD). The innovative additive approach allows substantial control to
the dimensional precision of the resultant scaffold
constructs. In addition, the fabricated biological constructs from the AM technology serve as viable tissue
replacements which eliminate issues such as donor
site morbidity and chronic rejection of allografts or
autologous implants (Leong et al. 2003, Chua et al.
2005, Mikos et al. 2006). Images from existing modalities such as computer tomography (CT) scan and
magnetic resonance imaging (MRI) can be manipulated in commercialized CAD software packages for
the customization of the TE constructs. The ability
to incorporate CAD provides the opportunity to yield
tissue substitutes of geometrical and mechanical properties similar to that of the native tissues, hence leading
to the production of patient-specific tissue constructs.
Despite the high cost in comparison to conventional
13
categorized into two classes (Yeong et al. 2004):

the melt-dissolution deposition and particle-bonding
approaches. AM methods are known to be applicable
in direct or indirect manner dependent on processing conditions and choice of materials. Although AM
has established itself with a wide range of possible
solutions to TE applications today, further enhancements and improvisation of these techniques are still
on-going. Improvements in AM technologies can vary
from in-house developed AM systems to bioprinting
and organ printing techniques (Mironov et al. 2003,
Mironov et al. 2009). As there are limited reviews
on the AM techniques for TE to date, this article
aims to provide an updated portfolio of AM systems
which have been implemented or explored to provide
medical professionals, scientists and tissue engineers
with a reference source for the progression of AM
technology.
2
2.1
Figure 1. Scanning electron microscopic (SEM) image of

3D interconnected pore microstructure created by the FDM
technique.
composite constructs consisting of PCL and tricalcium

phosphate (TCP) were introduced (Shao et al. 2006).
The TCP, in the study, serves as a reinforcement to
increase the mechanical integrity of PCL for the bone
segment while PCL itself is only mechanically adequate for the cartilage ECM. The auricular frameworks
constructed using autologous tissues, which are recognized as the state-of-the-art in reconstructive surgery,
were often lacking of delicate details to the external
profile of a human ear. Nevertheless, the auricular
scaffolds manufactured via the FDM technology were
able to replicate the CT scan data of the patients contralateral ear satisfactorily (Zeng et al. 2008). This
observation indicated that the melt-dissolution deposition AM method is a potential substitute to traditional
human ear reconstruction techniques. Since the applications of AM technology in the area of craniofacial
surgery is considerably new, further work should be
conducted to identify suitable internal morphologies
for regeneration of the targeted tissues.
Despite the ability to form TE scaffolds with precise
dimensions, the FDM technique does have its limitation. One of the disadvantages includes the need to
reshape materials into filaments of specific diameters
before the start of the manufacturing process. Next,
the rapid solidification of the extruded fibers which
form the resultant constructs provides insignificant
topological features for crucial tissue regrowth processes such as cell attachment and neovascularization
(Figure 2). Moreover, the use of naturally-derived and
hydrophilic materials is prohibited due to the systems
high operating temperature.
The drawbacks of the melt-dissolution deposition techniques have motivated the innovation of
in-house AM systems specifically used for TE applications. For instance, the 3D fiber deposition (3DF)
was developed to produce porous articular cartilage
scaffolds using biodegradable poly (ethylene glycol)terephthalate/poly (butylenes-terephthalate) (PEGT/
PBT) copolymers (Malda et al. 2005).
In contrast to FDM, the 3DF technique requires
material feed in granules form and melting of the
PRIMITIVE AM APPROACHES
Melt-dissolution deposition methods
Fused deposition modeling (FDM), till today,

remained a viable melt-dissolution deposition method
to TE scaffold engineering. This AM technique creates tissue constructs by the extrusions of semi-liquid
state or molten material through a nozzle moving along
the x- and y- axes, while the building platform lowers along the z-axes as each layer is fabricated and
adjoined to the previous layer as it cools and solidifies. The FDM technique has been investigated for
its feasibility in TE scaffold building by the authors
group. The initial FDM investigations were first carried out using acrylonitrile-butadiene-styrene (ABS),
a non-biocompatible material which is proprietary to
the FDM system (Too et al. 2002) (Figure 1). The
effects of process parameters were found to influence
the porosity, pore size and mechanical strength of the
built parts. An empirical mathematical model relating
the process parameters (raster gap) and structural properties (porosity) was subsequently established. The
mechanical characteristics relating to the structural
properties were then integrated into the mathematical model in a later study for a better understanding of
the resulting FDM builds (Ang et al. 2006). Following this investigation, a FEA model was performed on
the FDM process using polycaprolactone (PCL), an
approved biomaterial by the Food and Drug Administration (FDA), to explore and affirm the possibility of
manufacturing biocompatible TE scaffolds (Ramanath
et al. 2007, Ramanath et al. 2008). In orthopedic
surgery, tissue replicates of the osteochondral defective sites have been built via the FDM. PCL scaffolds
of coral-like structures produced by the FDM presented interconnected networks (of pore sizes between
300580 m) that supported the co-culturing of
osteoblasts and chondrocytes as well as integrated
the rich ECM of both cell types (Cao et al. 2003).
To match the mechanical properties of tissue scaffolds with that of the native medial femoral condyle,
14
2.2
Particle-bonding methods
Particle-bonding techniques such as selective laser sintering (SLS) and 3D printing (3DP) have evolved as
viable advanced TE alternatives to obtain scaffolds
with complex geometries and desirable surface morphologies. Both AM methods exhibited the ability to
use powdered-based materials which can be available commercially or produced via mixing, milling
or spray drying. Another similarity of the two techniques includes the need to have fully filled and
leveled powder beds before beginning the fabrication processes. Nevertheless, the working principles of
these additive techniques differ distinctively. The SLS
operates by selectively sintering the powder materials
using a carbon dioxide laser beam to form individual 2D sliced layers on the part-built chamber or
part-bed. The part-bed is encompassed with a constant heat source which serves to bond the adjacent
layer together. The resultant 3D SLS physical model
is then progressively produced as the building platform lowers along the z-axis after each sintered layer
is constructed. The SLS processing conditions such
as the laser powder and scan speed of the laser beam
are important controlling factors to the resultant build.
The 3DP, on the other hand, joins the neighboring particles and layers by the injection of a liquid adhesive
or binder. For both processes, the powder particles
which are not bonded are suggested to serve as temporary supports to the model and are subsequently
removed via compressed air or post-treatment procedures after the builds are completed (Leong et al.
2003). Since the SLS and 3DP are not specifically
designed for TE applications, the proprietary materials used in the commercialized systems are often
non-biocompatible. This hence has led to substantial efforts for the identification of suitable material
choices and their processing conditions for the respective particle-bonding procedures. For instance, the
material choices for SLS have been widely researched
by the authors group. The material selections include
poly(etheretherketone) (PEEK), Poly(vinyl alcohol)
(PVA), poly(L-lactic acid) (PLLA), poly(L-lactideco-glycolide) (PLG) and PCL. These biomaterials
were first evaluated for their intrinsic properties such
as glass transition and melting temperature before
defining the suitable SLS processing parameters. A
relevant work has been conducted to identify the optimal processing parameters for the polymeric materials
mentioned (Tan et al. 2005). As PEEK is known for
its bio-inertness, the bioactivity of the PEEK scaffolds
can be increased by the incorporation of hydroxyapatite (HA) particles via physical blending in a roller
mixer (Tan et al. 2003) (Figure 3a). The formation
of composite blends such as PVA/HA by physical
blending was justified to be more feasible in contrast to spray-dried composite powders (Chua et al.
2004) (Figure 3b). The physically blended materials
were used to build scaffolds addressing to craniofacial
and joint defects via the SLS and its TE potential of
the SLS constructs was verified (Wiria et al. 2008).
Figure 2. A microscopic image illustrating the dense

deposited fibers produced by the FDM.
materials is conducted in the presence of nitrogen

gas so as to preserve the intrinsic properties of the
hydrophilic components. Furthermore, on top of the
AM technology, the co-polymeric system was implemented so that tailor-made properties such as swelling
ratio, biodegradability and mechanical strength can be
met (Moroni et al. 2007). To provide a wider range of
composite blends for TE constructs, a desktop-based
RP (DBRP) technique was introduced (Hoque et al.
2009). Using the DBRP approach, PCL-poly(ethylene
glycol) (PEG) cartilage constructs were created and
resulted in a significant increase in hydrophilicity
which reflected higher yield of DNA content during
in vitro culture as compared to the PCL homopolymers. It is also important for deposition-based AM
systems to accommodate non-thermosensitive materials so that the biocompatibility of AM-built scaffolds
can be enhanced with more choices of hydrophilic
constituents, which are often characterized with lower
melting point. An example of such AM method is
the liquid-frozen deposition manufacturing (LFDM)
system, which is a fiber-based deposition technique
and involved no heating process (Yen et al. 2009).
In the LFDM method, poly(glycolic acid) (PLGA)
polymeric solution was used as the material feed and
each layer built was frozen via the low temperature
platform. The resultant frozen built was subsequently
lyophilized to vaporize the unwanted solvent. The
LFDM PLGA constructs were observed to preserve
the intrinsic properties of PLGA as compared to the
FDM PLGA builds.
Despite the wide applications of melt-dissolution
and fiber-based deposition methods, these techniques
will face issues when intricate surface topographical features are necessary. Moreover, the deposition
methods create scaffold constructs based on the additive stacking of 2D lay-down patterns which will thus
result in difficulties faced when CAD automation
involving the complex assembly of polyhedral unit
cells is required to be implemented (Leong et al. 2008,
Sudarmadji et al. 2011).
15
PCL SLS fabricated constructs were first explored for

drug delivery devices (Leong et al. 2001, Low et al.
2001, Cheah et al. 2002, Ang et al. 2007, Leong et
al. 2006). These findings were subsequently applied
to create laser sintered PCL/HA TE scaffolds which
are examined to be feasible for bone TE (Wiria et
al. 2007) (Figure 3c). The research group further
challenged the SLS system with PLG/HA/TCP blend
for more options for bone repair and regeneration
(Simpson et al. 2008) (Figure 3d). More recently,
PCL homopolymeric constructs built via the SLS,
together with the incorporation of the author groups
in-house Computer-Aided System forTissue Scaffolds
(CASTS) (Chua et al. 2003b, Cheah et al. 2004, Chua
et al. 2003a, Naing et al. 2005), were demonstrated
to provide the necessary architectural features for the
treatment to myocardial infarction (MI) (Yeong et al.
2010).
One of the TE-associated disadvantages of the SLS
technique includes harsh processing conditions such
as high temperature and laser power. These factors
are often reasons for the limited range of material
choices for SLS fabrication. Secondly, the topological characteristics of the resultant SLS-built scaffolds are restricted due to the particle size of the
powder-based feed since micro-sized granules can be
obtained commercially or via grinding, spray drying and roller mixing. The existing issues faced are
recently circumvented using biocomposites with nanofeatures. This strategy, similarly to the mature SLS
method, requires a pre-processing step to yield microspheres as raw materials for the SLS process (Duan
et al. 2010). The microspheres can then combined to
form a nanocomposite blend of poly(hydroxybutyrateco-hydroxyvalerate) (PHBV) and calcium phosphate
(CaP) which demonstrated improved bioactivity as
well as mechanical properties for bone TE. Further
comparison done between PHBV/CAP and HA/PLLA
nanocomposite scaffolds revealed the former composite as a preferred biomimetic system as bone
substitutes (Duan et al. 2011). The spot size of the
laser beam of commercialized SLS system may possibly contribute to the inability to achieve nano-scaled
features. Thus, customized SLS system with minimum
spot size of 50 m has since been built to manufacture
nano-HA models as biomimetic substrates for bone
TE (Shuai et al. 2011). In addition, the SLS working
principles were also improvised to include ceramics
such as bioactive glass for achieving fully dense crania
implants (Comesaa et al. 2011). The laser scanning
concept was also attempted to fabricate human ear
model cartilages for auricular reconstruction (Liu et
al. 2009).
3DP systems, in general, are innovated by different commercial sources or research institutes so as to
provide more material or composite blends forTE scaffold fabrication. TheriformTM processing is one of
the 3DP methods used to produce poly(L-lactic acid)
(PLA) scaffolds of different pore size and void fraction
for the culturing of dermal fibroblasts, smooth muscle cells and endothelial cells. The porosity of each
Figure 3. SEM images of (a) PEEK/HA, (b) PVA/HA, (c)

PCL/HA and (d) PCL/HA/TCP SLS composite scaffolds.
16
individual construct was increased by the inclusion

of particulate-leaching technique; sodium chloride
(NaCL) salt particles were blended within the PLA
matrix and were subsequently dissolved to form additional pores (Zeltinger et al. 2001). To eliminate the use
of organic solvents, starch-based polymers were built
using a 3DP by ZCorp. with water as a binder to the
powder particles (Lam et al. 2002). 3DP PLGA constructs manufactured from the same organization were
evaluated to provide structural characteristics for the
facilitation of bone tissue formation and growth (Ge
et al. 2009). An attempt to build HA interconnected
scaffold networks for bone using custom-made 3DP
demonstrated to meet both structural and mechanical
requirements (Seitz et al. 2005). Although the 3DP
technique does not subject the physical build to harsh
processing conditions like the SLS method, the postprocessing step which is used to strengthen the TE
construct or remove the binder may result in shrinkage
as high as 27% (Fierz et al. 2008). The shrinkage was
found to be isotropic but the density of the supposed
build was significantly increased. The compromise in
accuracy was avoided by using an alternate 3DP system, known as the Thermojet (Maher et al. 2010),
which is capable of producing micro-assay channels
for effective cell guidance.
INDIRECT AND PHOTOSENSITIVE AM

APPROACHES
Indirect and photosensitive AM methods are often

adopted as alternatives to direct AM technology as
they offers wider range of material choices and, in
majority cases, eliminate the need for toxic solvent
during the fabrication procedure. These techniques, in
general, require the use of a negative mould which
will be removed after the scaffold material has been
cast and solidified. Moreover, the indirect approaches
integrate both conventional and advanced TE fabrication methods, therefore resulting in the formation of
macro- as well as micro-structural features (Taboas
et al. 2003). The negative moulds, made of thermoplastic, have been explored and built commonly via a
3D inkjet printing or drop-on-demand printing (DDP)
technology. Collagen type I scaffold using the indirect method had been examined for the presence of
contamination from the sacrificial mould (Sachlos
et al. 2003b). The mould removal was performed using
ethanol, an organic solvent which is used for sterilization of biomedical devices.The authorsgroup had also
developed collagen scaffolds using a similar technique
and a mathematical model to illustrate the theoretical
aspects of the inkjet printing technology (Yeong 2006,
Yeong et al. 2007) (Figure 4).
More recently, silk fibroin (SF) scaffolds with
pre-defined interconnected networks were shown to
provide an additional option for AM-based naturally
derived TE constructs (Liu et al. 2011) (Figure 5). The
indirect AM fabrication concept has also been used
to produce other biomaterial types such as ceramics,
Figure 4. Images showing (a) a sacrificial mould fabricated

via the inkjet printing technique, (b) the resultant collagen
construct with interconnected channels and (c) a SEM image
illustrating the formation of internal macro-channel and the
preservation of porous structure arising from conventional
TE methods.
synthetic polymers or a combination of the materials

(Taboas et al. 2003, Jeong et al. 2010, Mondrinos et
al. 2006, Wilson et al. 2011, Lee et al. 2010, Schumacher et al. 2010). The indirect method was carried
out using a 3DP system to yield PLGA scaffolds
with villus features (Lee et al. 2005). The manufactured mould was composed of plaster while increased
porosity was achieved by inclusion of the particulateleaching technique during the casting process. The
17
the indirect techniques, which involve multiple procedural steps, may result in contamination issues. The
photosensitive processes, on the other hand, can possibly lead to degradation of the biomaterials intrinsic
properties. In short, these techniques should be closely
monitored during processing and further characterization of the post-processed constructs will determine if
the resultant TE scaffold is feasible for their relevant
applications.
Figure 5. (a) Printed thermoplastic mould (after removal
of support mater) and (b) SF scaffold with interconnected
macro-channels obtained using the sacrificial mould.
MODERN AM APPROACHES
BIOPRINTING AND ORGAN PRINTING
The two methodological terms, bioprinting and organ

printing, have been used interchangeably as they are
considerably new to the TE field. Nevertheless, this
class of AM techniques can be categorized into two
distinctive groups: (1) a technique aimed to fabricate
scaffolds equipped with biological cues to enhance
cell attachment, followed by neo-tissue formation,
and/or (2) a method to produce tissue constructs
evenly encapsulated with living cells. Nevertheless,
the two categories aim to hasten the recovery process by promoting functional healing. The former class
of bioprinting and organ printing is commonly used
to distribute growth factors spatially within scaffold
networks for the guidance of cell differentiation. For
instance, an inkjet-based AM technique has been conducted to deliver growth factors such as the bone
morphogenetic protein and heparin-binding epidermal
growth factor for bone formation and stem cell differentiation respectively (Cooper et al. 2009, Miller et al.
2011). Otherwise, the growth factors can be delivered
into the microchannels of the scaffold after the bioprinting process, which was demonstrated in dental
TE (Kim et al. 2010).
The direct writing of cell encapsulated materials has
created gel-based scaffold constructs via stereolithography, 3DP, inkjet printing, laser-assisted systems and
commonly with custom-built AM systems. This perspective of AM advancement is anticipated to resolve
limitations faced with TE of organs with complex
structures. The ability to pattern cells evenly within
scaffold constructs has been shown to yield higher
cell viability and eliminate the need for in vitro cell
culturing. A commercially available stereolithography
and 3DP has been applied to PEG-based hydrogels
(Dhariwala et al. 2004, Censi et al. 2011) while
custom-built bioprinters with or without laser assistance used gel-like materials such as pluronic F-127,
and alginate (Smith et al. 2004, Cohen et al. 2006,
Guillotin et al. 2010). The suspensions used to encapsulate cells for inkjet printers include collagen type I
and culture mediums (Xu et al. 2005, Saunders et al.
2008).
The current status of bioprinting and organ printing, despite their existence for several years, can be
observed to be in the exploratory stage. This is because
efforts to fine tune the processing parameters for
optimal cell viability is still in progress based on
Figure 6. 3D and top view of the (a) photopolymeric RP

mould, (b) paraffin wax mould and (c) gelatin TE scaffold.
mould and porogens were removed by immersion of

the TE constructs in deionized (DI) water. The effects
of scaffold architecture and pore size on smooth cell
growth were further investigated (Lee et al. 2008).
Photosensitive AM techniques such as polyjet printing system and stereolithography have provided as
viable options for the applications of indirect and direct
approaches. The authors research has also included
using the polyjet printing technology for the fabrication of gelatin (Figure 6) and SF scaffolds (Tan et al.
2009, Liu et al. 2010a, Liu et al. 2010b). The photosensitive moulds are usually removed by a photopolymer
stripper used widely in the semiconductor industry and the resultant scaffold builds were shown to
induce negligible cytotoxicity of the scaffold construct
(Melchels et al. 2010).
The indirect application of stereolithgraphy was
established to produce structurally-controlled TCP
ceramic scaffolds for orthopedic applications (Li et
al. 2008). Stereolithography is a well-known photosensitive technique for the manufacturing of TE
scaffolds for both direct and indirect TE scaffold fabrication. Poly(D,L-lactide) (PLLA) resin substrates
created using the direct photo-crosslinked method
exhibited the ability for osteoblast attachments and
proliferation (Melchels et al. 2009). Novel scaffold
designs were then incorporated to improve mass transport of essential nutrients, for effective cell processed
and prevention of cell growth only at the peripheries
The indirect and photosensitive methods have
demonstrated their viability in TE scaffold. However,
18
Chua, C. K., Leong, K. F., Cheah, C. M. & Chua,

S. W. (2003a) Development of a tissue engineering
scaffold structure library for rapid prototyping. Part
1: Investigation and classification. International Journal of Advanced Manufacturing Technology, 21(4),
291301.
Chua, C. K., Leong, K. F., Cheah, C. M. & Chua, S. W.
(2003b) Development of a tissue engineering scaffold
structure library for rapid prototyping. Part 2: Parametric library and assembly program. International Journal
of Advanced Manufacturing Technology, 21(4), 302312.
Chua, C. K., Leong, K. F., Tan, K. H., Wiria, F. E. &
Cheah, C. M. (2004) Development of tissue scaffolds using selective laser sintering of polyvinyl alcohol/hydroxyapatite biocomposite for craniofacial and
joint defects. Journal of Materials Science-Materials in
Medicine, 15(10), 11131121.
Chua, C. K., Yeong, W. Y. & Leong, K. F. 2005. Rapid prototyping in tissue engineering: A state-of-the-art report.
In Proceedings of the 2nd International Conference on
Advanced Research and Rapid Prototyping. Leiria, Portugal: Taylor & Francis Group plc.
Cohen, D. L., Malone, E., Lipson, H. & Bonassar, L. J.
(2006) Direct Freeform Fabrication of Seeded Hydrogels in Arbitrary Geometries. Tissue Engineering, 12(5),
13251335.
Comesaa, R., Lusquios, F., del Val, J., Lpez-lvarez, M.,
Quintero, F., Riveiro, A., Boutinguiza, M., de Carlos, A.,
Jones, J. R., Hill, R. G. & Pou, J. (2011) Three-dimensional
bioactive glass implants fabricated by rapid prototyping
based on CO2 laser cladding. Acta Biomaterialia, 7(9),
34763487.
Cooper, G. M., Miller, E. D., DeCesare, G. E., Usas, A.,
Lensie, E. L., Bykowski, M. R., Huard, J., Weiss, L. E.,
Losee, J. E. & Campbell, P. G. (2009) Inkjet-Based Biopatterning of Bone Morphogenetic Protein-2 to Spatially
Control Calvarial Bone Formation. Tissue Engineering
Part A, 16(5), 17491759.
Dhariwala, B., Hunt, E. & Boland, T. (2004) Rapid Prototyping of Tissue-Engineering Constructs, Using Photopolymerizable Hydrogels and Stereolithography. Tissue
Engineering, 10(910), 13161322.
Duan, B. & et al. (2011) Optimized fabrication of
CaP/PHBV nanocomposite scaffolds via selective laser
sintering for bone tissue engineering. Biofabrication, 3(1),
113.
Duan, B., Wang, M., Zhou, W. Y., Cheung, W. L., Li, Z. Y.
& Lu, W. W. (2010) Three-dimensional nanocomposite
scaffolds fabricated via selective laser sintering for bone
tissue engineering. Acta Biomaterialia, 6(12), 44954505.
Fierz, F. C., Beckmann, F., Huser, M., Irsen, S. H.,
Leukers, B., Witte, F., Degistirici, ., Andronache, A.,
Thie, M. & Mller, B. (2008) The morphology of
anisotropic 3D-printed hydroxyapatite scaffolds. Biomaterials, 29(28), 37993806.
Ge, Z., Tian, X., Heng, B. C., Fan, V., Yeo, J. F. &
Cao, T. (2009) Histological evaluation of osteogenesis of
3D-printed poly-lactic-co-glycolic acid (PLGA) scaffolds
in a rabbit model. Biomedical Materials, 4(2), 17.
Guillotin, B., Souquet, A., Catros, S., Duocastella, M.,
Pippenger, B., Bellance, S., Bareille, R., Rmy, M.,
Bordenave, L., Amde, J. & Guillemot, F. (2010)
Laser assisted bioprinting of engineered tissue with high
cell density and microscale organization. Biomaterials,
31(28), 72507256.
Hoque, M. E., San, W. Y., Wei, F., Li, S., Huang, M.-H.,
Vert, M. & Hutmacher, D. W. (2009) Processing of Polycaprolactone and Polycaprolactone-Based
existing literature. Moreover, there is a lack of research

findings associated with bioprinted scaffolds in vivo.
Nevertheless, the advanced AM technology has since
shown its potential for TE of artificial human tissues
or organs.
5
CONCLUSION
In this article, an overview of AM technology has been

presented to provide medical professionals, scientists
and engineers with a reference framework to the TE
biofabrication approaches. Based on existing literature, it can be inferred that AM has indeed contributed
significantly to enhance the healing of defective orthopedic and craniofacial tissues. More recently, improvisations made to the existing AM techniques have
demonstrated viable potential to the engineering of
complex tissues and organs. It is foreseen that AM
can be brought to a much advanced level where the TE
builds are integrated with functionally graded morphologies and mechanical properties that match the
native tissue. This aspect shall be explored in the automated CAD perspective and together with a suitable
AM technique, a functional CAD-based TE scaffold
can be yielded. The novel CAD-related research is
presently proposed and in midst of investigation by
the authors group.
REFERENCES
Ang, K. C., Leong, K. F., Chua, C. K. & Chandrasekaran, M.
(2006) Investigation of the mechanical properties and
porosity relationships in fused deposition modelingfabricated porous structures. Rapid Prototyping Journal,
12(2), 100105.
Ang, K.C., Leong, K.F., Chua, C.K. & Chandrasekaran,
M. (2007) Compressive properties and degradability
of poly(-caprolatone)/ hydroxyapatite composites under
accelerated hydrolytic degradation. Journal of Biomedical
Materials ResearchPart A, 80(3), 655660.
Bartolo, P. J., Chua, C. K., Almeida, H. A., Chou, S. M. &
Lim, A. S. C. (2009) Biomanufacturing for tissue engineering: Present and future trends. Virtual and Physical
Prototyping, 4(4), 203216.
Cao, T., Ho, K. H. & Teoh, S. H. (2003) Scaffold Design and
in Vitro Study of Osteochondral Coculture in a ThreeDimensional Porous Polycaprolactone Scaffold Fabricated by Fused Deposition Modeling. Tissue Engineering,
9 (supplement 1), 103112.
Censi, R., Schuurman, W., Malda, J., di Dato, G.,
Burgisser, P.E., Dhert, W. J. A., van Nostrum, C. F.,
di Martino, P., Vermonden, T. & Hennink, W. E.
(2011) A Printable Photopolymerizable Thermosensitive
p(HPMAm-lactate)-PEG Hydrogel for Tissue Engineering. Advanced Functional Materials, 21(10), 18331842.
Cheah, C.M., Leong, K.F., Chua, C.K., Low, K.H. &
Quek, H.S. (2002) Characterization of microfeatures in
selective laser sintered drug delivery devices. Proc Institution Mechanical Engineers, Part H: Journal of Engineering
in Medicine, 216(6), 369383.
Cheah, C. M., Chua, C. K., Leong, K. F., Cheong, C. H. &
Naing, M. W. (2004) Automatic algorithm for generating complex polyhedral scaffold structures for tissue
engineering. Tissue Engineering, 10(34), 595610.
19
Copolymers into 3D Scaffolds, and Their Cellular

Responses. Tissue Engineering Part A, 15(10), 3013
3024.
Jeong, C. G. & Hollister, S. J. (2010) A comparison of
the influence of material on in vitro cartilage tissue
engineering with PCL, PGS, and POC 3D scaffold architecture seeded with chondrocytes. Biomaterials, 31(15),
43044312.
Kim, K., Lee, C. H., Kim, B. K. & Mao, J. J. (2010) Anatomically Shaped Tooth and Periodontal Regeneration by Cell
Homing. Journal of Dental Research, 89(8), 842847.
Lam, C.X.F., Mo, X.M., Teoh, S.H. & Hutmacher, D.W.
(2002) Scaffold development using 3D printing with a
starch-based polymer. Materials Science & Engineering C-Biomimetic and Supramolecular Systems, 20(12),
4956.
Lee, K.-W., Wang, S., Dadsetan, M.,Yaszemski, M. J. & Lu, L.
(2010) Enhanced Cell Ingrowth and Proliferation through
Three-Dimensional Nanocomposite Scaffolds with Controlled Pore Structures. Biomacromolecules, 11(3),
682689.
Lee, M., Dunn, J. C. Y. & Wu, B. M. (2005) Scaffold fabrication by indirect three-dimensional printing. Biomaterials,
26(20), 42814289.
Lee, M., Wu, B. M. & Dunn, J. C. Y. (2008) Effect of scaffold
architecture and pore size on smooth muscle cell growth.
Journal of Biomedical Materials Research Part A, 87A(4),
10101016.
Leong, K.F., Phua, K.K.S., Chua, C.K., Du, Z.H. & Teo,
K.O.M.(2001), Fabrication of porous polymeric matrix
drug delivery devices using the selective laser sintering
technique. Proceedings of the Institution of Mechanical
Engineers, Part H: Journal of Engineering in Medicine
215(2), 191201.
Leong, K.F., Cheah, C.M. & Chua, C.K. (2003) Solid
freeform fabrication of three-dimensional scaffolds for
engineering replacement tissues and organs. Biomaterials,
24(13), 23632378.
Leong, K.F., Chua, C.K., Gui, W.S. & Verani (2006) Building
porous biopolymeric microstructures for controlled drug
delivery devices using selective laser sintering. International Journal of Advanced Manufacturing Technology,
31(56), 483489.
Leong, K.F., Chua, C.K., Sudarmadji, N. & Yeong, W.Y.
(2008) Engineering functionally graded tissue engineering scaffolds. Journal of the Mechanical Behavior of
Biomedical Materials, 1, 140152.
Li, X., Li, D., Lu, B. & Wang, C. (2008) Fabrication of bioceramic scaffolds with pre-designed internal architecture
by gel casting and indirect stereolithography techniques.
Journal of Porous Materials, 15(6), 667671.
Liu, M. J. J., Chou, S. M. & Chua, C. K. 2010a. Development
of silk fibroin scaffolds using an indirect rapid prototyping technology. In International Conference on Cellular &
Molecular Bioengineering. Singapore.
Liu, M. J. J., Chou, S. M. & Chua, C. K. 2010b. Proteinbased scaffold fabrication using Polyjet technology. In 4th
International Conference PMI. Belgium.
Liu, M. J. J., Chou, S. M. & Chua, C. K. 2011. The study
of micro- and macro-structural features of silk fibroin
scaffolds. In First International Symposium on Bioengineering, eds. C. K. Chua, M. Y. Kee, Y. L. Guan, J.P. Chen,
K. Q. Luo, C. S. Lai, C. K. Kwoh, K. S. Chuan & S. L.
Wu. Singapore: Research Publishing Services.
Liu, Y., Zhang, L., Zhou, G., Li, Q., Liu, W., Yu, Z., Luo, X.,
Jiang, T., Zhang, W. & Cao, Y. (2009) In vitro engineering
of human ear-shaped cartilage assisted with CAD/CAM
technology. Biomaterials, 31(8), 21762183.
Low, K.H., Leong, K.F., Chua, C.K., Du, Z.H. & Cheah, C.M.
(2001) Characterization of SLS parts for drug delivery
devices. Rapid Prototyping Journal, 7(5), 262267.
Maher, P. S., Keatch, R. P. & Donnelly, K. (2010)
Characterisation of rapid prototyping techniques for studies in cell behaviour. Rapid Prototyping Journal, 16(2),
116123.
Malda, J., Woodfield, T. B. F., van der Vloodt, F., Wilson, C.,
Martens, D. E., Tramper, J., van Blitterswijk, C. A. &
Riesle, J. (2005) The effect of PEGT/PBT scaffold architecture on the composition of tissue engineered cartilage.
Biomaterials, 26(1), 6372.
Melchels, F. P. W., Barradas, A. M. C., van Blitterswijk, C. A.,
de Boer, J., Feijen, J. & Grijpma, D. W. (2010) Effects of the
architecture of tissue engineering scaffolds on cell seeding
and culturing. Acta Biomaterialia, 6(11), 42084217.
Melchels, F. P. W., Feijen, J. & Grijpma, D. W. (2009)
A poly(d,l-lactide) resin for the preparation of tissue
engineering scaffolds by stereolithography. Biomaterials,
30(2324), 38013809.
Mikos,A. G., Herring, S. W., Ochareon, P., Elisseeff, J., Lu, H.
H., Kandel, R., Schoen, F. J., Toner, M., Mooney, D., Atala,
A., Dyke, M. E. V., Kaplan, D. & Vunjak-Novakovic, G.
(2006) Engineering ComplexTissues.Tissue Engineering,
12(12), 33073339.
Miller, E. D., Li, K., Kanade, T., Weiss, L. E.,
Walker, L. M. & Campbell, P. G. (2011) Spatially directed
guidance of stem cell population migration by immobilized patterns of growth factors. Biomaterials, 32(11),
27752785.
Mironov, V., Boland, T., Trusk, T., Forgacs, G. & Markwald,
R. R. (2003) Organ printing: computer-aided jet-based
3D tissue engineering. Trends in Biotechnology, 21(4),
157161.
Mironov, V. & et al. (2009) Biofabrication: a 21st century
manufacturing paradigm. Biofabrication, 1(2), 116.
Mondrinos, M. J., Dembzynski, R., Lu, L., Byrapogu,V. K. C.,
Wootton, D. M., Lelkes, P. I. & Zhou, J. (2006) Porogenbased solid freeform fabrication of polycaprolactonecalcium phosphate scaffolds for tissue engineering. Biomaterials, 27(25), 43994408.
Moroni, L., Hendriks, J. A. A., Schotel, R., de Wijn, J.
R. & van Blitterswijk, C. A. (2007) Design of Biphasic Polymeric 3-Dimensional Fiber Deposited Scaffolds
for Cartilage Tissue Engineering Applications. Tissue
Engineering, 13(2), 361371.
Naing, M. W., Chua, C. K., Leong, K. F. & Wang, Y. (2005)
Fabrication of customized scaffolds using computer-aided
design and rapid prototyping techniques. Rapid Prototyping Journal, 11(4), 249259.
Ramanath, H. S., Chandrasekaran, M., Chua, C. K., Leong, K.
F. & Shah, K. D. (2007) Modeling of extrusion behavior of
biopolymer and composites in fused deposition modeling.
Key Engineering Materials, 334335, 12411244.
Ramanath, H. S., Chandrasekaran, M., Chua, C. K., Leong,
K. F. & Shah, K. D. (2008) Melt flow behaviour of
poly--caprolactone in fused deposition modeling. Journal of Materials Science Materials in Medicine, 19(7),
25412550.
Sachlos, E. & Czernuszka, J. T. (2003a) Making Tissue Engineering Scaffolds Work. Review: The application of solid
freeform fabrication technology to the production of tissue engineering scaffolds European Cells and Materials,
5, 2939.
Sachlos, E., Reis, N., Ainsley, C., Derby, B. & Czernuszka,
J. T. (2003b) Novel collagen scaffolds with predefined
internal morphology made by solid freeform fabrication.
20
Saunders, R. E., Gough, J. E. & Derby, B. (2008) Delivery of

human fibroblast cells by piezoelectric drop-on-demand
inkjet printing. Biomaterials, 29(2), 193203.
Schumacher, M., Deisinger, U., Detsch, R. & Ziegler, G.
(2010) Indirect rapid prototyping of biphasic calcium
phosphate scaffolds as bone substitutes: influence of
phase composition, macroporosity and pore geometry
on mechanical properties. Journal of Materials Science:
Materials in Medicine, 21(12), 31193127.
Seitz, H., Rieder, W., Irsen, S., Leukers, B. & Tille, C. (2005)
Three-dimensional printing of porous ceramic scaffolds
for bone tissue engineering. Journal of Biomedical Materials Research Part B: Applied Biomaterials, 74B(2),
782788.
Shao, X., Goh, J. C. H., Hutmacher, D. W., Lee, E. H. &
Zigang, G. (2006) Repair of Large Articular Osteochondral Defects Using Hybrid Scaffolds and Bone MarrowDerived Mesenchymal Stem Cells in a Rabbit Model.
Tissue Engineering, 12(6), 15391551.
Shuai, C. & et al. (2011) Structure and properties of nanohydroxypatite scaffolds for bone tissue engineering with a
selective laser sintering system. Nanotechnology, 22(28),
285703.
Simpson, R. L., Wiria, F. E., Amis, A. A., Chua, C. K.,
Leong, K. F., Hansen, U. N., Chandrasekaran, M. & Lee,
M. W. (2008) Development of a 95/5 poly(L-lactideco-glycolide)/hydroxylapatite and -tricalcium phosphate
scaffold as bone replacement material via selective laser
sintering. Journal of Biomedical Materials Research Part
B: Applied Biomaterials, 84B(1), 1725.
Smith, C. M., Stone, A. L., Parkhill, R. L., Stewart, R. L.,
Simpkins, M. W., Kachurin, A. M., Warren, W. L. &
Williams, S. K. (2004) Three-Dimensional BioAssembly
Tool for Generating Viable Tissue-Engineered Constructs.
Sudarmadji, N., Tan, J.Y., Leong, K. F., Chua, C. K. & Loh,Y.
T. (2011) Investigation of the mechanical properties and
porosity relationships in selective laser-sintered polyhedral for functionally graded scaffolds. Acta Biomaterialia,
7(2), 530537.
Taboas, J. M., Maddox, R. D., Krebsbach, P. H. & Hollister, S.
J. (2003) Indirect solid free form fabrication of local and
global porous, biomimetic and composite 3D polymerceramic scaffolds. Biomaterials, 24(1), 181194.
Tan, J. Y., Chua, C. K. & Leong, K. F. 2009. Indirect
fabrication of tissue engineering scaffolds using rapid
prototyping and a foaming process. In 4th International
Conference on Advanced Research in Virtual and Rapid
Prototyping, 5157. Leiria, Portugal.
Tan, K. H., Chua, C. K., Leong, K. F., Cheah, C.
M., Cheang, P., Abu Bakar, M. S. & Cha, S. W.
(2003) Scaffold development using selective laser sintering of polyetheretherketone-hydroxyapatite biocomposite
blends. Biomaterials, 24(18), 31153123.
Tan, K. H., Chua, C. K., Leong, K. F., Cheah, C. M.,
Gui, W. S., Tan, W. S. & Wiria, F. E. (2005) Selective
laser sintering of biocompatible polymers for applications in tissue engineering. Bio-Medical Materials and
Engineering, 15(12), 113124.
Too, M. H., Leong, K. F., Chua, C. K., Du, Z. H., Yang,
S. F., Cheah, C. M. & Ho, S. L. (2002) Investigation
of 3D Non-Random Porous Structures by Fused Deposition Modeling. The International Journal of Advanced
Manufacturing Technology, 19(3), 217223.
Warren, S. M., Fong, K. D., Chen, C. M., Loboa, E. G., Cowan,

C. M., Lorenz, H. P. & Longaker, M. T. (2003) Tools and
Techniques for Craniofacial Tissue Engineering. Tissue
Wilson, C., van Blitterswijk, C., Verbout, A., Dhert, W. & de
Bruijn, J. (2011) Scaffolds with a standardized macroarchitecture fabricated from several calcium phosphate
ceramics using an indirect rapid prototyping technique.
Journal of Materials Science: Materials in Medicine,
22(1), 97105.
Wiria, F. E., Chua, C. K., Leong, K. F., Quah, Z. Y., Chandrasekaran, M. & Lee, M. W. (2008) Improved biocomposite development of poly(vinyl alcohol) and hydroxyapatite
for tissue engineering scaffold fabrication using selective
laser sintering. Journal of Materials Science-Materials in
Medicine, 19(3), 989996.
Wiria, F. E., Leong, K. F., Chua, C. K. & Liu, Y. (2007)
Poly-e-caprolactone/hydroxyapatite for tissue engineering scaffold fabrication via selective laser sintering. Acta
Biomaterialia, 3(1), 112.
Xu, T., Jin, J., Gregory, C., Hickman, J. J. & Boland, T. (2005)
Inkjet printing of viable mammalian cells. Biomaterials,
26(1), 9399.
Yang, S. F., Leong, K. F., Du, Z. H. & Chua, C. K. (2001) The
design of scaffolds for use in tissue engineering. Part 1.
Traditional factors. Tissue Engineering, 7(6), 679689.
Yang, S. F., Leong, K. F., Du, Z. H. & Chua, C. K. (2002) The
design of scaffolds for use in tissue engineering. Part II.
Rapid prototyping techniques. Tissue Engineering, 8(1),
111.
Yen, H.-J., Hsu, S.-H., Tseng, C.-S., Huang, J.-P. & Tsai, C.-L.
(2009) Fabrication of Precision Scaffolds Using LiquidFrozen Deposition Manufacturing for Cartilage Tissue
Engineering. Tissue Engineering Part A, 15(5), 965975.
Yeong, W. Y., Chua, C. K., Leong, K. F., Chandrasekaran,
M. & Lee, M. W. (2006) Indirect fabrication of collagen scaffold based on inkjet printing technique. Rapid
Prototyping Journal, 12(4), 229237.
Yeong, W. Y., Chua, C. K., Leong, K. F. & Chandrasekaran,
M. (2004) Rapid prototyping in tissue engineering: challenges and potential. Trends in Biotechnology, 22(12),
643652.
M. & Lee, M. W. (2007) Comparison of drying methods
in the fabrication of collagen scaffold via indirect rapid
prototyping. Journal of Biomedical Materials Research
Part B: Applied Biomaterials, 82B(1), 260266.
Yeong, W. Y., Sudarmadji, N., Yu, H. Y., Chua, C. K., Leong,
K. F., Venkatraman, S. S., Boey, Y. C. F. & Tan, L. P.
(2010) Porous polycaprolactone scaffold for cardiac tissue
engineering fabricated by selective laser sintering. Acta
Zeltinger, J., Sherwood, J. K., Graham, D. A., Meller, R. &
Griffith, L. G. (2001) Effect of pore size and void fraction
on cellular adhesion, proliferation, and matrix deposition.
Zeng, W., Lin, F., Shi, T., Zhang, R., Nian, Y., Ruan, J. &
Zhou, T. (2008) Fused deposition modeling of an auricle framework for microtia reconstruction based on CT
images. Rapid Prototyping Journal, 14(5), 280284.
Zuk, P. A. (2008) Tissue Engineering Craniofacial Defects
With Adult Stem Cells? Are We Ready Yet? Pediatric
Research, 63(5), 478486.
21
Biomanufacturing

Ultrastructural analysis of the hDSC interactions with biodegradable

3D scaffolds
S.E. Duailibi, M.T. Duailibi & L.M. Ferreira
Federal University of So Paulo,UNIFESP-CTCMol, So Paulo, Brazil
Nacional Institute of Science and Technology Biofabrication Institute, INCT-Biofabris, So Paulo, Brazil
F.A.O. Tanaka
Escola Superior de Agricultura Luiz de Queiroz, Universidade de So Paulo, So Paulo, Brazil
J.P. Vacanti
Massachusetts General Hospital, Harvard Medical School, Boston, MA, US
P.C. Yelick
Tufts University, MA, US
ABSTRACT: This study observed the behavior of hDSCs on 3 biodegradable 3D scaffolds surfaces coated
with collagen: PGA; PHB; Silk and PCL. Based on the demonstrated ability for scaffold surfaces to dramatically
influence cell viability and adhesion, in this study we characterize hDSC viability, affinity, and differentiation
on 3D scaffolds. Isolated human tooth cells were obtained from patients scheduled extraction (CONEP-13083)
and plated and expanded in standard tissue culture conditions. After 20 and 30 days, hDSCs were observed under
inverted microscopy, followed by TEM and SEM analyses, which revealed the activity of hDSCs on each scaffold
type, including cells adhesion on the biomaterial surface. The high cells activity was observed where we found
more mitochondria and rugous endoplasmatic reticulum representing high energy and mRNA production for
cells proliferation in similarity with the control group. The PGA and PHB exhibited the better hDSCs adhesion
with 3D scaffold interface.
INTRODUCTION
Cell transplantation utilizing synthetic biodegradable polymer substrates has successfully been used
as a means of tissue reconstruction and repair.
Using this approach, autologous donor tissues, when
available, are harvested, dissociated into single cell
suspensions, seeded onto appropriate scaffold materials, and implanted at the functional tissue site.
Traditionally, scaffold materials are chosen whose
physical properties closely match those of the replacement organ, in order to provide structural support
until engineered materials can form. After implantation, the biodegradable scaffold gradually dissolves
and is eventually replaced with regenerated tissue,
whose functions ideally contribute to those of the
original organ. This approach has been successfully
used to regenerate a variety of tissues including liver,
esophagus, small intestine, heart valve, heart muscle,
and cartilage, demonstrating that tissue engineering
approaches may surpass those of existing conventional
therapies (Mikos et al., 1993; Sittinger et al.,1996;
Kim et al., 1998; Kim et al., 1999).
In the field of dentistry, tissue engineering strategies
have successfully been used for a variety of applications, including regenerating periodontal defects
in dogs, and treating periodontal defects caused by
The tooth is a complex organ, consisting of both highly

mineralized and soft tissue components. Humans
cannot regenerate permanent teeth, and human dental tissues do not exhibit significant regenerative
capabilities. Tissue engineering, a relatively new science for tissue and organ regeneration, integrates
knowledge learned from the overlapping fields
of developmental biology, molecular/cellular biology, molecular genetics, and chemical engineering
(Langer & Vacanti,1993; Sittinger et al.,1996; Bohl
et al., 1998; Kim & Vacanti 1999; Mooney & Mikos,
1999; Stock & Vacanti, 2001 and Vacanti et al., 2001).
Although surgical techniques for successful organ
transplantation have improved significantly, facilitated
in part by the availability of immunosuppressive drugs,
organ donor shortage remains a worldwide problem
(Langer & Vacanti,1993). To address this shortage,
strategies for tissue regeneration, as first described by
Langer and Vacanti (1993), and Kaigler and Mooney
(2001), can be classified into three typesconductive,
inductive and cellular transplantationthe latter of
which has guided the approach taken by this group
for tooth tissue engineering.
25
Glutamax, 50 units/ml penicillin, 50 mg/ml streptomycin, 2.5 mg/ml ascorbic acid, and 50% F12 medium
(Sigma-Aldrich Corp, St. Louis, MO, USA). Singlecell tooth bud suspensions were generated by filtration
through a Falcon 40-micron cell strainer.
Cells were resuspended in the same media, plated
into 75 cm2 (T75) culture flasks (Costar, Cambridge,
MA, USA) at 2.5 105 cells/ml, and grown in 5% CO2
at 37 C until the cells reached confluence. The culture
medium was changed 2 times per week. We obtained
averaged cell yields of 1.0 106 cells/impacted third
molar tooth.
oral tumors, tooth decay, and periodontal disease

(Lynch et al., 1999; Murphy & Mooney, 1999). The
existing challenges for successful dental tissue engineering include identifying the ideal combination of
biodegradable scaffold materials, cells, and soluble
growth factors, to address particular clinical problems. For long term repair and regeneration of dental
tissues, it is necessary to find the combination of
cell types, scaffold materials, and scaffold design, to
bioengineer complex dental tissues exhibiting similar
physical, mechanical, and functional characteristics
to naturally formed dental tissues. Suitable scaffold
materials must exhibit proper porosity, chemical composition, mechanical properties, and biodegradative
properties to support the growth and differentiation of
dental tissues, as well as serve as gene and/or growth
factor delivery vehicles.
We have previously reported that dissociated
porcine dental progenitor cells, seeded onto biodegradable PGA/PLGA scaffolds and grown for 30 weeks
in the omentum of adult rat hosts, formed organized
tooth crowns (Young et al., 2002, 2005a, 2005b). More
recently, we reported the formation of organized bioengineered tooth crowns from cultured postnatal rat
tooth bud cells (Duailibi et al., 2004, 2006, 2008).
Based on the need to identify alternative scaffold
materials that can guide the size and shape of bioengineered human dental tissues, and to better understand
the manner through which scaffold materials instruct
dental progenitor cell differentiation, the objective of
the present study was to characterize the cell attachment and differentiation, with ultrastructural analysis
to study cells interactions with Poly-glycolic acid
(PGA), Poly Hydroxi butirate acid(PHB), silk and
Poly caprolactone (PCL) as biomaterial surfaces and
hDSC viability, affinity, and differentiation on 3D
scaffolds with collagen pre-treatment.
2
2.1
2.2 hDSC seeding onto PGA scaffolds and

characterization
At confluence, hDSCs were harvested by trypsinization (0.25% trypsin/EDTA, Gibco-Invitrogen Corp.,
Tulsa, OK, USA) for 10 min at 37 C, washed twice
with the same medium, and counted using a hemacytometer and Leica DM IRB microscope. Harvested
hDSCs were split into equal portions, and statically
seeded at concentration of 106 cells/cm3 scaffold, onto
electrospun nanofiber PGA, PHB, silk and PCL scaffolds. All scaffolds were previously pre-coating for 12
hours with Collagen (Invitrogen, Carlsbad, CA, USA).
After allowing hDSCs to adhere for one hour at 37 C,
the scaffolds were analyzed.
2.3 Scaffold fabrication
PGA scaffolds were generated by electrospinning, a
relatively simple and cost-effective method for creating nanofiber scaffolds with inter-connected pore
structure and fiber diameters in the sub-micron range
(Pham et al., 2006). SEMs of the original Electrospun PGA microfibers fabricated with and without
collagen as surface treatment and used in these studies are shown (Figure 1 and 2). Electrospun PGA
scaffolds are ideal for tissue engineering applications
due to the fact that the large surface area facilitates
hDSC attachment and allows for multiple focal adhesions to form on multiple, micron-size diameter fibers.
Additionally, fibers of micron and sub-micron size
more closely mimic the size scale of naturally formed
extracellular matrix (ECM) components. Functional
modifications of electrospun microfibers can also be
used to improve polymer biocompatibility and cytocompatibility, and to promote cell proliferation and
METHODS AND MATERIAL

Human dental stem cell (hDSC) isolation
and culture
Human tooth tissues were collected from patients

scheduled for diagnostic third molar extraction, as
determined by professional dentists from public and/or
private clinics, according the ethic committee (CEP
0595/01). Isolated human enamel and pulp organ tissues were placed in a petri dish containing pre-warmed
(37 C) Hanks balanced salt solution (HBSS, Gibco
BRL, Gaithersburg, MD, USA). Human tooth tissues
were digested for 30 to 40 minutes with 0.4 mg/ml type
II collagenase (Sigma-Aldrich, St. Louis, MO, USA),
and 0.2 mg/ml Dispase I (Boehringer Mannheim,
Indianapolis, IN, USA). Following enzyme digestion, the tissues were dissociated by trituration, and
washed 5 times in 50% Dulbeccos modified Eagle
medium (DMEM, Gibco BRL, Gaithersburg, MD,
USA) containing 10% fetal bovine serum (FBS), 5 ml
Figure 1. and 2. Electrospun PGA microfibers (SEM). Figure 1, on left, 500, and Figure 2, on right, PGA with collagen
as surface treatment, 100.
26
differentiation (Pham et al., 2006). The scaffolds were

sterilized with ethylene oxide.
The PHB (Figure 3), Silk (Figure 4) and PCL (Figure 5) scaffolds were developed by Estadual University
of Campinas (UNICAMP), Engineering of Materials
and were gently provided for this study.
The cells were harvested and they were seeded on
the PGA, PHB, Silk and PCL scaffold as a co-culture
during approximately 2 weeks, as we can show on the
figures of optical microscopy (Figure 6 and 7).
After 2 weeks, the constructs, cells attached on the
scaffolds three-dimensional were fixed on Karnovsky
solution at 4 C overnight. To capture the SEM images,
the samples were dehydrated in acetone, dried at critical point in CO2 . The samples were glued in stubs and
cover with metallic gold. The analyses were done in
emvLE435VP.
The ultrastructural studies, TEM, Tramission Electronic Microscopy, samples were washed in buffer
cacodilate 0.05%, then, ps fixed in osmio tetroxide and steined in uranila acetate 0,5% overnight. The
dehidratation was done with crescent concentration of
acetone and polymerized in resin epoxy. The sections
ultra thin were done in ultramicrotome Leica Ultracut (UCT with 70 nm of thickness and stain in uranila
acetate 3% and lead citrate in pH 12. The analysis were
done in TEM Zeiss EM900.
Figure 3. PHB scaffold with collagen as surface treatment

(TEM) on right 500 nm and left 200 nm.
Figure 7. Silk (left) and PCL (right) scaffold with cells in

co-culture (optical microscopy)1.
Figure 4. Silk scaffold with collagen as surface treatment

(TEM) on right 2000 nm and left 1000.
Figure 8. Control group, only cells on the plaque surface, N,

nucleus; M, mitochondria and RE, endoplasmic reticulum.
Figure 5. PCL scaffold with collagen as surface treatment

(TEM) on right 500 nm and left 200 nm.
Figure 9. PGA and PHB constructs.
Figure 6. PGA (left) and PHB (right) scaffold with cells in

co-culture (optical microscopy).
Figure 10. Silk and PCL constructs.
27
RESULTS AND DISCUSSION
function of hepatocytes on a novel three dimensional synthetic biodegradable polymer scaffold with an intrinsic
network of channels. Ann Surg. 28:813.
Kim SS, Vacanti JP. 1999. The current status of tissue
engineering as potential therapy. Semin Pediatr Surg.
8(3):119123.
Langer R, Vacanti JP.1993. Tissue engineering. Science
260:920926.
Lynch SE, Genco RJ, Marx RE. 1999. Tissue engineering.
In:Applications in maxillofacial surgery and periodontics.
1st ed.Chicago, IL: Quintessence Publishing. Mooney DJ,
Mikos AG. 1999. Growing new organs. Scientif Amer
6065.
Mikos AG, Sarakinos G, Leite SM, Vacanti JP, Langer
R. 1993. Laminated Three-dimensional biodegradable
foams for use in Tissue Engineering. Biomaterials. 14(5):
323330.
Murphy WL, Mooney DJ. 1999. Controlled delivery of
inductive proteins, plasmid DNA and cells from tissue
engineering matrices. J Periodontal Res. 34: 413419.
Pham QP, Sharma U, Mikos AG. 2006. Eletrospinning of
Polymeric Nanofibers for Tissue Engineering Applications: A Review. Tissue Engineering. 12(5):11971211.
Sittinger M, Bujia J, Rotter N, Reitzel D, Minuth WW,
Burmester GR. 1996. Tissue engineering and autologous
transplant formation: practical approaches with resorbable
biomaterials and new cell culture techniques. Biomaterials 17:237242.
Stock UA, Vacanti JP. 2001.Tissue engineering: current state
and prospects. Ann Rev Med. 52:143151.
Vacanti CA, Bonassar LJ, Vacanti M, Shufflebarger J. 2001.
Replacement of an avulsed phalanx with tissue engineered
bone. N Engl J Med. 344:15111514.
Young CS, Terada S, Vacanti JP, Honda M, Bartlett JD, Yelick
PC. 2002. Tissue engineering of complex tooth structure
on biodegradable polymer scaffolds. J Dent Res. 81:695
700.
Young, C.S., Kim, S.-W., Taylor, R., Vacanti, J.P., Bartlett,
J.D., and Yelick, P.C. 2005a. Developmental Analysis and
Three-Dimensional Computer Modeling of Tooth Crowns
Grown on Biodegradable Polymer Scaffolds. Archives in
Oral Biology. 50(2):259265.
Young, C. S., Abukawa, H., Asrican, R., Ravens, M.S.,
Troulis, M.J., Kaban, L.B., Vacanti, J.P., and Yelick, P.C.
2005b. Tissue-engineered hybrid tooth and bone. Tissue
Eng. 11(910):1599610.
We observed cells with dense cytoplasm, with mitochondria and rough endoplasmic reticulum in high
quantity, demonstrating high activity promoting perfect adaptation in interface with the scaffold. The
high cells activity was observed on PGA and PHB
(Figure 9), than others 3D scaffolds of Silk and
PCL (Figure 10), where we found more mitochondria
and rough endoplasmic reticulum representing high
energy and mRNA production for cells proliferation
in similarity with the control group (Figure 8).
4
CONCLUSION
The PGA and PHB exhibited the better hDSCs adhesion with 3D scaffold interface.
These studies were supported by the UNIFESP,
So Paulo, Brazil, CNPq, FAPESP 07/512274 and
07/588567,Tufts University, Boston, MA,The Center
for the Integration of Medicine and Innovative Technology (CIMIT), NIH/NIDCR grants TW007665 and
DE016132 (PCY), USP ESALQ, So Paulo Brazil.
REFERENCES
Bohl KS, Shon J, Rutherford B, Mooney DJ. 1998. Role
of synthetic extracellular matrix in development of engineered dental pulp. Biomater Sci Polymer Edn. 9:749
764.
Duailibi MT, Duailibi SE, Young CS, Bartlett JD, Vacanti JP,
Yelick PC. 2004. Bioengineered Teeth from Cultured Rat
Tooth Bud Cells. J. Dental Research USA. 83(7):523528.
Duailibi SE, Duailibi MT, Vacanti JP, Yelick PC. 2006.
Prospects for tooth regeneration. Periodontology 2000.
41:177187.
Duailibi SE, Duailibi MT, Zhang W, Asrican R, Vacanti JP,
Yelick PC. 2008. Bioengineered Dental Tissues Grown in
the Rat Jaw. J. Dent. Res. 87(8);745750.
Kim SS, Utsunomiya H, Koski JA, Wu BM, Cima MJ, Sohn
J, Mukai K, Griffith LG, Vacanti JP. 1998. Survival and
28

Individual contour adapted functional implant structures in Titanium

C. Schoene, R. Stelzer, P. Sembdner & L. Betrol
Department of Mechanical Engineering, Institute of Machine Elements and Machine Design,
Dresden University of Technology
J. Markwardt
Department of Oral and Maxillofacial Surgery, Dresden University of Technology
B. Reitemeier
Department of Prosthetic Dentistry, Dresden University of Technology
G. Engel
Product Development Company Hofmann and Engel GmbH in Boxdorf near Dresden
ABSTRACT: Apart from applications in mechanical engineering, now even the domain of medicine may
benefit from the option of using metallic materials for Direct Manufacturing. In the medical domain, the use
of biocompatible materials, such as titanium or titanium alloys is essential to produce individual implants. As a
result of this development, it is now possible to generate new patient-specific geometries fitted to the contour.
This paper elucidates the process chain to derive individual design variants and to produce patient-specific bone
replacement implants for the lower jaw-bone regions by using innovative reverse engineering and manufacturing
methods. For this interdisciplinary project, technical scientists, medical scientists at the university hospital and
engineers from a product development firm work together.
1
INTODUCTION
Scientific studies of CAD/CAM applications in

medicine and dental prosthetics, which have been ongoing for approximately 15 years, focus on models, tooth
crowns and bridges (Schoene & Stelzer 2005, Luthardt
et al. 2002; Bernard & Fischer 2002). The use of
CAD/CAM technology with CNC-milling and rapid
manufacturing in dental industry is now very common.The majority of scientific approaches to use rapid
manufacturing since 2006 have been aimed at endoprostheses in CoCr and titanium alloys (SLM 2006;
Wirtz 2005; Warnke 2009; Rehme 2006, Bertol 2010).
In cases when jaw implants are required due to
disruption of continuity in the lower jaw bone, the relevant anatomical regions are commonly represented
by means of imaging techniques, such as Computer
Tomography (CT). Based on these data, which are specific to each patient, we analyse the range of available
standard care programs, choose the appropriate reconstruction plates for the jaw region and implant them.
As a rule, the reconstruction plates currently in use are
characterised by great variations in stiffness between
implant and bone (Figure 1).
In Figure 2 a patient is illustrated with aesthetic and
clinical deficits as a result of the damage of a standard
reconstruction plate after surgery and tumour resection
on the left facial side. Additionally, since the geometry
is not significantly individualised, we also see obvious
Figure 1. Standard reconstruction plate for jaw region.
functional and structural-mechanical deficits, as well

as aesthetic disadvantages. Consequently, an application that protects the tissue and is also highly stable,
which is a necessity for optimal treatment, cannot be
provided by any of the methods currently available
(Markwardt et al. 2007, Schieferstein 2003, Maurer
2004).
As a result of ongoing globalisation, the greatly
expanding market for medical implants made of
biocompatible high-performance materials is under
29
Individual design and modifications of the implant

are performed based on the 3D model. The fundamental steps necessary to generate individual implants are
listed below:
Figure 2. Patient after damage of standard reconstruction

plate.
ever-increasing pressure from competitors. In this context, the reconstruction of bone defects, in particular in
the oral, jaw and facial region, by means of osteosynthetic plates is regarded as a great challenge. Here,
special advantages may accrue to a new implant design
whose contour and stiffness are tailored to specific
geometric and elastic conditions, since in this way it
is possible to reduce complications during ingrowth.
The LaserCUSING method (Gebhardt 2008;
Wohlers 2009) provides the first technological
approach to manufacturing new filigree implants that
are perfectly aligned with the contour and gradually
modified in stiffness. LaserCUSING is an innovative technique, following a generative approach, which
is able to realise structures according to the direction
of force action.
One objective of the planned research project is
aimed at the development of a process chain that
extends all the way from CT layer images of a diseased patient up to the manufacturing of individual
bone substitute implants for the patient while taking
into consideration a Rapid Manufacturing technique.
Thus, the rapid manufacturing of individual implants
that repair defects is primarily emphasized in order
to keep the waiting periods for patients as short as
possible.
Imaging of the diseased area and surrounding

regions by means of CT
Creation of a discrete surface model from the CT
image stacks
Alignment of the lower jaw model for design in a
defined co-ordinate system
Definition of cutting planes to isolate the defective
regions
Generation of a mathematical surface representation of the lower jaw contour in the affected area
and the surrounding regions
Definition of the positions of the holes that will later
secure the implant in the residual bone
Implant design with CAD system (Computer Aided
Design)
Design of the cutting templates, which are applied
to the jaw before resecting the diseased bone
Production planning for the LaserCUSING system incl. placement in the working space
Creation of the support structure taking into account
the building layer configuration
Manufacturing of the implant and the cutting templates with LaserCUSING
Removal of the implant from the building plate by
means of erosion
Manual removal of the support structure
Corundum blast finishing of the implant surfaces.
2.1
From the CT image stack to the CAD

solid model
When generating the individual implant, the first step

consists of mapping the defective bone regions and
the surrounding soft part tissues by means of CT
techniques. The result of this data acquisition procedure is made available in single images in the DICOM
format. The next step is to read these images by means
of VoXim (VoXim 2009). The following step is 3D
soft tissue segmentation, wherein materials of different density, such as soft tissue and bone, are separated
from each other. Afterwards, a faceted 3D model of the
segmented bone regions is output in the STL format.
This model is important for ongoing design.
Thus, discrete data is first made available. Now it
is necessary to reverse these data into a solid model of
mathematically correct representation for CAD modelling. To do this, the faceted data are processed with
the Geomagic Studio (Geomagic 2010). The polygons
are subject to various repair and filtering measures.
Here, the following steps are to be run in order:
2 APPROACH
The project consortium began with mandibular
implants in 2006.
The work we are doing in Dresden is unique in
that it features close interdisciplinary co-operation
among radiologists and oral and maxillofacial surgeons, dentists and engineers. The study includes an
ethics proposal for animal experiments and a patent.
The CT data required for diagnosis are also used
to generate the virtual 3D model of the jaw bone.
30
Diversify (deletes stand-alone object regions)

Remove peaks
Repair cuts
Smooth
Manually rework
Figure 4. left: Registration of 5 different jaws, right curves

after virtual jaws cutting in position between tooth 31 and 41.
Figure 3. Sequence from physical object to CAD solid
model [15].
At this point, we also define a uniform co-ordinate

system and the cutting planes in co-operation with
the surgeon. These cutting layers define the region
that is to be removed in surgery and replaced with an
implant. In this cutting region, the contour of the inner
bone structure, the cancellous bone, is reversed as a
Spline curve and simultaneously stored separate from
the lower jaw model. This information is significant,
among other things, for the definition of the positions
of the holes that will later hold the implant in place.
These positions should be in the region of maximal
layer thickness of the corticalis (outer bone region).
In the strictest sense, Reverse Engineering describes
the procedure of 3D digitising of workpieces with
sculptured surfaces, conditioning the 3D point data
and converting them to CAD models (Schoene, 2009).
Positive results will only be obtained by 3D data
recording in conjunction with qualified and problemoriented data conditioning and application in the
follow-up computer-aided strategy. Another equally
important fact is that the CAD representation successfully withstands production planning, manufacturing
and quality inspection (Fichtner et al. 2005).
Reverse engineering is performed then to generate
a parametric solid model. The solid model is stored
in the STEP format and is now available for design.
The path from physical object to CAD solid model is
represented in Figure 3.
2.2
Figure 5. Different implant variant, left outer attachment to

the residual bone, right inner attachment to spongious tissue.
It is also possible to intentionally include discontinuities in the enveloping geometry.

The shell design makes it possible to implement an
inner filigree structure as well. These structures inside
the implant are expected to offer greater reliability in
the bone regeneration. In this approach, these structures may have a different geometric shape, as well as
stochastic discontinuities and different dimensioning.
Design is carried out considering the positive contact with the residual jaw on both sides. In general,
we are investigating two different ways to attach the
implant to the residual jaw (Figure 5). The first connecting type (Figure 5, left) is based on a variant in
which the implant is secured to the outer lower jaw
contour. In the second variant (Figure 5, right), the
implant is shifted into the bone and cemented with a
suitable bone substitute material. The choice of variant
must be based on the individual case.
The principal design procedure for an individual
implant is described below using the variant in which
the implant is attached to the residual bone:
Design of the individual implant
Current designs of implants that are identical in contour orient themselves to the organic bone structure
configuration. They consist of an outer mounting
shell design and an inner filigree tissue structure. The
designed outer shell of the implant correspondingly
follows the contour of the removed jaw region. Manufacturing technology allows this to be very thin in
shape in order to reproduce the stiffness and strength
of the bone. Thin-walled envelope geometry of about
0.3 mm thickness is realized; thickness in the areas
attached to the residual bone is 0.4 mm.
In cases that the tumour destroyed the bone, there
is no useful geometry. Therefor a database, containing
characteristic curves, is planned (Figure 4).
Insert section layers in lower jaw model

Section the jaw model
Prepare lower jaw stubs of residual bone (remove
milling region)
Transfer surface information of the milled over
lapping regions
Offset surfaces by wall thickness value
Define length of the overlapping regions for the
implant
Fill in enveloping geometry according to the cut
section of the jaw regions
Insert inner filigree structure.
Before repairing the defect with an implant, it is

necessary to remove the corresponding jaw region in
a surgical operation. The section layers required have
31
amorphous material, such as powder. In this way,

LaserCUSING is able to produce functional models.
The material characteristics obtained are commensurable with those of the series material and make it
possible to use the parts thus produced even under the
conditions of production. LaserCUSING is a technology that works using a layer-by-layer technique,
wherein layer thickness values vary from 30 m to
50 m.
Depending on the technology and the material, it is
very difficult or simply impossible to produce surfaces
lying under an angle of 45 to the building plate. For
these surfaces, we need special supporting structures
which have to be generated in the CAD system and
later on by means of Magics, which is a type of Rapid
Prototyping software.
First, the implant is placed in the CAD system just
as it is to be built in the LaserCUSING system. Then
it is shifted in Z direction by 0.5 mm so that is can
be removed from the building platform by means of
wire erosion later on. Thus it is possible to add the
supporting structure to the overhung surfaces. After
the building procedure, this supporting structure has
to be removed again. For this reason, this additional
structure should be kept as small as possible in order
to reduce necessary rework. Afterwards, the implant
model is exported as an STL file. Magics is used
to generate the remaining supporting geometry. The
parameters for this supporting structure have to be
dimensioned and modified as a function of the shape
type and position. As a function of the focus diameter,
the supporting structure is only fused in the building
process as a line structure. Consequently, it may be easily removed afterwards. After this step, these generated
data are virtually cut into layers. The LaserCUSING
system is filled with pure titanium powder and fitted
with a titanium plate intended to be use as a building plate. Manufacturing of the mandibular implant
using pure titanium, is a technological challenge since
it requires inert gas. In contrast to other body regions,
however, for the oral and maxillofacial zones, pure
titanium is preferred due to allergic reactions.
Next, the layer data are entered into the software of
the machine, and the implant is positioned on a virtual
building plate. The implant and the support structure
are assigned the corresponding manufacturing parameters. Thereby, laser power and rate are defined, among
other parameters. The entire process, from setup to
removal, is performed in an inert gas atmosphere to
guarantee manufacturing free of oxidation.
In the first step, the building plate is lowered down
by one layer element, and new powder is introduced. In
the next step, the powder is surfaced with a lamination
plate (coating). In the last step, the deposited powder
coating, which has a constant thickness, is selectively
fused by laser (exposing). This procedure is repeated
until the component is complete.
After completion of the building procedure, the
building plate is removed, and the implant erodes from
the plate. After that, the support geometry is removed
and the implant is cleaned.
Figure 6. Cutting patterns, made by direct manufacturing,

here in the model test carried out on a model of the lower jaw.
Figure 7. Implant was attached to the jaw of pig cadaver,

connecting type is based on a variant in which the implant is
secured to the outer lower contour.
already been defined virtually in an earlier step. Since

the operating team does not have access to these virtual
layers during implantation, the position of the section
layers is predefined in the form of cutting templates
(Figure 6). These templates guarantee an unambiguously positive contact at the jaw. Figure 7 demonstrates
one variant of the attached implant to the pig cadaver
jaw.
The cutting templates are also designed with the
CAD model of the lower jaws. Beginning with the
parting planes where the resection will be performed,
the adjacent surfaces are derived. From this, we create a
two-piece body. One lateral surface of the body is used
as the cutting surface along which the medical doctor
moves the saw. Two-piece performance is necessary in
order to secure the cutting templates to the jaw and to
prevent undercuts.
Afterwards, the design results are also subject to
finite element analysis to evaluate the implants stability. We calculate using various extreme values due
to the different strength values for the bone that are
necessary for the computations. Finally, mechanical
strength trends can be abstracted.
2.3
Production planning and manufacturing

of the implant
The implants are produced by means of LaserCUSING, which is a generative technique based on
32
Figure 9. Titanium implant for outer attachment and cutting

patterns to the residual bone of living animal.
Figure 8. Direct Manufacturing jaw implant created with

LaserCUSING, made of titanium and drilling patterns
(Stainless steel) for attachment to the spongious tissue.
Building of the cutting patterns is performed analogously, with the difference that stainless steel, processed in a nitrogen atmosphere, is used as material.
Figure 8 elucidates examples for an implant made of
titanium and the corresponding drilling patterns made
of stainless steel to be secured to the bone.
Figure 10. Titanium implant is attached to the residual bone.
SUMMARY AND OUTLOOK
manufacturing. The coating system is also subject to

continuous improvement.
Other recent research topics focus on the computeraided modelling of inner filigree support structures
that stimulate growth. In the future it is expected
that such structures will be efficiently created in an
automated manner using a CAD system. A stable
attachment to the residual bone is essential to the
function of the individual implant. The presentation
outlines possible design variants. In design, strength,
biocompatibility and operating conditions are to be
considered. Future tests will determine the stability of
these connections. The authors are currently developing a test bench for investigations of the jaw model.
The test bed is also used to validate the results obtained
in the FE analyses.
The process chain introduced here shows the path

from the CT image of a diseased patient via design
of individual implants to the production of titanium
implants by means of generative manufacturing techniques. This approach has been tested in eight lower
jaws of pig cadavers, one human model lower jaw and
two macerated human lower jaws up to now. Further
operations were performed on the jaws of 10 living test
animals (miniature pigs). So it was carried out, how the
titanium implant is attached to the residual bone, the
positions of the number of screws and the usability of
the tools (Figures 910).
The results of these experiments show that we have
succeeded in achieving a general fitting accuracy. At
present, we are testing the process chain in animal
experiments and are verifying the suitability of the
implants in living beings.
It takes about 32 hours to carry out the entire process to produce individual implants. This span includes
7 working hours to prepare the CT data for the solid
model of the jaw region. 13 hours are required to design
the implant and the cutting patterns, while 12 hours are
allotted for production planning and manufacturing.
In the future, the process may be optimised in the
field of CT layered image processing. Process time
should be positively influenced by segmentation and
creation of the 3D model of the lower jaw. Adequate
interpolation and filtering methods should contribute
to higher data quality. Design of implants and cutting patterns should be improved by the generation of
new software tools. To keep lead times to a minimum,
we are currently optimising the support structures and
their process parameters for production planning and
ACKNOWLEDGMENT
We acknowledge Dipl.-Ing. Gerd Engel, managing director of the product development company
Hofmann & Engel Produktentwicklung GmbH in
Boxdorf near Dresden, for his innovative ideas and his
forward-looking decision to invest in an advanced laser
CUSING system at an early date. We also acknowledge Dipl.-Ing. Thomas Jahn, project manager in this
scientific project in the company. Gerd Engel also contributed to the distribution of this modern technology
in an industrial environment and now for applications
in the medical domain as well.
The research topic is sponsored by the Saxon Bank
for Reconstruction and Development SAB Dresden
(promotional ident:13363/2273).
33
REFERENCES
Rehme, O. & Emmelmann, C. 2006, Rapid manufacturing

of lattice structures with selective laser melting, Laserbased Micropackaging, edited by Friedrich G. Bachmann,
Willem Hoving, Yongfeng Lu, Kunihiko Washio, Proc. of
SPIE Vol. 6107, 2006.
Schieferstein, H. 2003, Experimental analyse of the human
masticatory system, Report of the Institute of machine
tools and science of management, TU Mnchen, Band
180, Mnchen.
Schoene, C. 2009, Reverse Engineering for freeform surfaces in process chains of production engineering, Dr. Hut,
Mnchen, ISBN 9783868531039.
Schoene, C. & Stelzer, R. 2005, 3D-Digitising and
Reverse Engineering, Aspects of Production Engineering in dental industry. Darmstadt: PACE-Forum Digital
Manufacturing.
SLM-Medical Applications, 2006, Patients in Australia
receive industrial rapid Manufacturing hip implants, MCP
Newsletter 07/2006, Lbeck, Germany.
VoXim, Version 5.6.1., 2009, IVS Solutions GmbH, Anwenderhandbuch, Chemnitz.
Warnke, P. & Douglas, T. 2009, Rapid Prototyping: Porous
Titanium Alloy scaffolds produced by selective laser
melting for bone tissue engineering, Tissue engineering,
Part C, Volume 15, Number 2, 2009.
Wirtz, P. 2005, Manufacturing of bone implants from titan
material with laser forming technology, Dissertation,
RWTH Aachen.
Wohlers, T. 2009, Rapid Prototyping and Manufacturing,
State of the Industry, Wohlers Associates, Inc.
Bernard, A. & Fischer, A. 2002, New Trends in Rapid Product

Developement, CIRP AnnalsManufacturing Technology,
Volume 51, Issue 2.
Bertol, L., Junior, W. da Silva, F. & Aumund-Kopp, C. 2010,
Medical design: direct metal laser sintering of Ti-6Al-4V,
Material and Design, 31 (2010) 39823988, Elsevier.
Fichtner, D., Schne, C., Schreiber, S. 2005, 3D-Digitizing
and Reverse engineering,Aspects of Production Engineering, Proceedings of the VR@P 2005, Leiria, Portugal.
Gebhardt, A. 2008, Generative Fertigungsverfahren, Rapid
Prototyping, Rapid Tooling, Rapid Manufacturing, ISBN10 10:3446-226664.
Geomagic, 2010, Benutzerhandbuch Geomagic Inc. Version
12.
Luthardt, R., Weber, A. & Schne, C. 2002, Design and production of Dental Prosthetic Restaurations: Basic research
on dental CAD/CAM technology, International Journal of
Computerised Dentistry 2002; 5.
Markwardt, J. & Pfeifer, G. & Eckelt, U. & Reitemeier,
B. 2007, Analysis of complications after reconstruction
of bone defects involving complete mandibular resection using finite element modeling, Onkologie 2007, 30,
212126.
Maurer, P. G. W. 2004, Clinical and computeraided study
for design optimisation of osteosynthetic plates for
mandible, Universittsklinik und Poliklinik fr Mund-,
Kiefer-und Plastische Gesichtschirurgie an der MartinLuther-Universitt Halle-Wittenberg, Habilitation, Halle.
34

New approaches to prototype 3D vascular-like structures

by additive layer manufacturing
E. Bassoli, L. Denti & A. Gatto
Department of Mechanical and Civil Engineering, University of Modena and Reggio Emilia, Modena, Italy
A. Paderno & G. Spaletta

Department of Mathematics, University of Bologna, Bologna, Italy
N. Zini
Institute of Molecular Genetics-Section of Bologna, National Research Council, IOR-Bologna, Italy
V. Strusi, D. Dallatana & R. Toni

Laboratory of Regenerative Morphology and Bioartificial Structures, Department of Human Anatomy,
Pharmacology and Forensic Medicine, University of Parma School of Medicine, Parma, Italy
ABSTRACT: In this paper preliminary results are summarized on the use of a combined Additive Layer
Manufacturing (ALM) and indirect replication methodology to reconstruct reticular-like, three-dimensional (3D)
structures mimicking the 3D vascular network of the adult human thyroid gland. In a first step, we developed
a fractal-like algorithm capable of modeling the native arterial distribution of the adult thyroid lobe, allowing
for vascular growth within its geometrical domain. Although some arbitrary simplifications were adopted, yet
the vascular density of the computational simulation showed good consistency with that of a native thyroid
lobe. In a second step, single vascular branches were prototyped based on the STL output of the algorithm and
ALM techniques, up to the achievement of a model having satisfactory geometrical/morphological accuracy. In
a third step, the problem of reproducing the vascular geometry with a biocompatible polymer was addressed,
and different protocols of replication technology were evaluated. Limits and possible methodological solutions
are discussed.
1
INTRODUCTION
allowing for best cell survival and growth (Sachlos

et al., 2003; Yeong et al., 2010). Besides many others, polycaprolactone (PCL) and the natural fibrous
protein collagen are biomaterials able to promote cell
growth, are reabsorbed at an appropriate rate without
toxic residues in vitro and in vivo, and depict sufficient
mechanical strength to be handled and implanted in
animal models. Therefore, they have been considered
reasonable options to generate biocompatible scaffolds (Eshraghi & Das, 2010; Kleinman et al., 1981;
Yeong et al., 2006;).
Additive Layer Manufacturing (ALM) techniques
inherently comply with the high levels of geometrical complexity needed to produce biocompatible
scaffolds, and have found successful applications in
the biomedical field (Giannatsis & Dedoussis, 2009).
In particular, Fused Deposition Modeling (FDM)
has been adopted to produce PCL honeycomb scaffolds, eventually colonized by human fibroblasts (Hutmacher et al., 2001), and similar results have been
obtained by selective laser sintering (SLS) in osseous,
hepatic and cardiac tissues engineering (Duan et al.,
2010; Huang et al., 2007; Yeong et al., 2010). In all
these instances, bioabsorbable scaffolds were ALM
Regenerative medicine of highly vascularized, soft

tissue and endocrine organs is one of the most challenging and ground-breaking areas in biomedical
research. New exciting frontiers can be opened by
ex-situ (i.e on the laboratory bench) bioengineering of
entire bioartificial organs, eventually usable for clinical transplantation. Using either stem cells, progenitor
cells or differentiated cells grown in a 3D culture system under the guidance of a 3D biocompatible scaffold,
tissue self-assembly and differentiation may occur up
to the formation of an entire viable organ (Toni et al.,
2011). In accordance with previous evidence on the
role played by the 3D geometry of the environment
where the cells grow (Ingber, 2005; Vogel & Sheetz,
2006), we have suggested that the ex situ assembly of
the thyroid parenchyma could be highly regulated by
the geometrical properties of a biocompatible physical support mimicking its natural stromal/vascular
scaffold or SVS (Toni et al., 2007).
Currently, a number of studies addressed to the production of biocompatible scaffolds have been focussed
on the type of biomaterial, molecular size and porosity
35
modeling of its SVS was obtained. Starting from

planar input data, represented by a real scintigraphic
or ultrasonographic in-vivo image of a single thyroid
lobe, the external profile of the lobe is found by edge
detection techniques. Points on the lobe contour are
sampled and approximated via splinecurves, consisting of a sequence of polynomials that smoothly join
into each other, so that the overall function presents
no jumps or ridges. The approximation spline curves
are opportunely roto-translated, to form an analytically known surface, whose internal volume can be
computed and whose spatial rendering represents the
observed lobe.
Then, a fractal-like approach has been adopted to
reconstruct the intralobar arterial system, whose 3D
geometry corresponds to that of the thyroid SVS; in
particular, the rule of diffusion limited aggregation
(DLA) has been employed. In such a method, the
fractal system is modeled as a Brownian tree starting from a seed (initial attractor) that represents the
SVS source-vessel, progressively joined via a random
walk by single elements of the fractal system, i.e. the
various branches of the SVS.
The algorithm is constrained by morphometric variables including the number of intraglandular arterial
branches, ramification angles and vasal calibers. The
simulated SVS grows inside the geometrical limits
(volume and shape) of the reconstructed thyroid lobe;
as a result, it may yield a model of the normal arterial
vascularisation inside the volume of an adult thyroid
lobe. Growth of vascular arborizations is restricted to
2nd and 3rd order intra-glandular arteries and the origin of 2nd order branches is fixed to a given number
onto the surface of the lobe.
The simulated spatial model of the thyroid lobe and
SVS was made available in the STL format.
A simplified model was also produced, where a portion of the arterial structure is obtained by the growth
of a branch from a single seed.
engineered with the intent to reconstruct simple thin

tissues (Giannatsis & Dedoussis, 2009; Sachlos et al.,
2003; Yeong et al., 2006;). In contrast, for highly
branched 3D structures, like those of the vasculature
in a thick 3D thyroid lobe, the required ALM technology appears so sophisticated to go nearby, if not
beyond, the current technological limits (Park et al.,
1998). Therefore, it results that the state-of-the art of
direct ALM processes for production of biocompatible scaffolds with 3D vascular architecture is still far
from being satisfactory.
To overcome these limitations, an indirect replication methodology has been recently introduced
(Sachlos et al., 2006), that enables to engineer a collagen scaffold with predefined internal channels, acting
as an artificial vascular system. Initially, a sacrificial
mould with the negative shape of the searched scaffold
was 3D printed, cavities of the mould were emptied
from the support material using an appropriate solvent, liquid collagen was cast into the mould, and then
frozen. Later, the mould was dissolved with ethanol
and the collagen scaffold dried at critical point with
liquid carbon dioxide. In this manner, a vascular morphology and vessel as small as 200 m in diameter
were obtained. With a similar approach, it has also
been shown that the geometry of these scaffolds may
be manipulated up to a size scale corresponding to the
cellular level (Yeong et al., 2006).
Based on these premises, we decided to investigate
the feasibility of a combined ALM and indirect replication procedure for bioengineering ex situ the SVS
of an adult human thyroid lobe. By using a recently
developed algorithm capable of modeling the natural
distribution of the intralobar arteries of the human thyroid, we exploited the STL output of this software to
prototype by ALM and synthetic resins simulations of
single vascular branches. Then, scaffold reconstruction was pursued through two different approaches
including: a) production of the negative of the scaffold
and its use as a mould (single replication) for biomaterial application; b) production of scaffold replicas,
to be used to generate a mould and then obtain a
biocompatible scaffold by double replication. Both
approaches have critical aspects due to the need to
set-up and control multiple process steps, and evaluate the error sources during the technological chain.
In the case of a thyroid SVS an additional difficulty
is represented by its extremely complex geometry and
microscopic dimensions. Finally, differential removal
of the master/mould with sufficient detail and accuracy further increase the challenges of this procedure
(Yeong et al., 2006).
2
2.2 Construction of prototypes by ALM

The virtual lobe reconstruction and the simplified
model in the STL format were used as a direct input
for ALM techniques. The files were inspected and
repaired through software tools dedicated to ALM
processes, considering the following errors: missing,
degenerate or overlapping facets, incorrect normal
vectors and non-manifold geometrical entities. The
fixed digital models were then analyzed to check their
viability for production, in terms of minimum detail
dimension and edge radii.
Layer processes and materials with the potential
capability of building a physical prototype of the two
branched structures were identified. The aim was to
evaluate the geometrical/dimensional capabilities of
ALM processes if applied to the considered models. The following techniques and machinery were
selected:
MATERIALS AND METHODS
2.1 Thyroid arterial vascularisation modeling

Based on the authors previous working experience
in the numerical-computational reconstruction of the
human thyroid gland morphology (Spaletta, 2004;
Della Casa et al., 2006), an hybrid fractal/deterministic
Stereolitography (SL):Viper si2, 3D Systems Inc.;

Drop-On-Demand (DOD): R66 plus, SolidScape
Inc.;
36
Table 1. Technical specifications of ALM equipment evaluated for prototype construction.

min. layer
thickness
[mm]
nominal
accuracy
[mm]
build
matter.
support
mater.
Viper
0.050
0.100
R66 plus
0.013
0.025
epoxy
resin
wax
ProJet
HD 3000
Eden350
0.016
0.050
honeycomb
epoxy
wax +
fatty esters
wax
0.016
0.042
acrylic
resin
acrylic
resin
was attempted. A solid block of acrylic resin containing the hollow branched SVS was obtained. The
SVS cavity remained full of support material, to be
eventually dissolved before pouring a biocompatible
material.
2.3.2 Double replication

The procedure for double replication was applied
through the following steps:
1. Production of the master in acrylic resin with MJM,
via ProJet HD 3000;
2. casting of a mould that englobes the master. The
main requirements are the capability of copying the
master geometry with a high detail resolution and
the lack of cytotoxicity of possible residues;
3. master removal to obtain a cavity in the mould.
Differential removal (either thermal or chemical)
has to occur without damaging the mould material;
4. casting of a biocompatible material in the cavity to
obtain the scaffold;
5. differential removal of the mould without damaging the scaffold. Due to the variable mechanical
properties of biocompatible materials this step is
particularly critical.
wax
Multi Jet Modelling (MJM): ProJet HD 3000,

3D Systems Inc., and Eden350 , Objet Geometries
Ltd.
Table 1 lists the main technical specifications and
nominal accuracy for each system, as well as the material used to build the prototype and the supports for
cavities and overhangs. In this step the research was
focused on geometric issues and not on materials, so
industrial polymeric resins were adopted. Nevertheless, the possible developments towards single/ double
replication procedures require to identify which combinations of build/support material can guarantee the
easiest differential removal.
The models were then measured and compared to
the STL file for the assessment of dimensional accuracy. The optical measuring microscope Kestrel 200
by Vision Engineering, equipped with Quadra-check
metrology software, was used for the measurements.
The system ensures 0.5 m stage repeatability in X and
Y axis, and uncertainty for the nominal values of the
considered specimen results in 10 m. Attention was
paid to the identification of a reference frame integral
with the branch structure, to ensure a common orientation between the physical prototype and the STL file
during the measurements. The following entities were
measured:
As to the mould construction (step 2), the following

materials have been tested:
The first two materials have been deposited in multiple steps by dipping the master in aqueous dispersions
of the ceramic material, up to a thickness of few mm.
Calcium sulphate has also been used in acetone dispersions, to increase the drying rate. In addition, moulds
in the form of solid blocks have been tested pouring the ceramic dispersions on the master. CaCO3 and
NaCl have been deposited from supersaturated aqueous solutions, as well as poured in the form of fine
powders onto the master.
These ceramic moulds are expected to ensure easy
thermal removal of the master. Differential removal
of the mould around the scaffold might, then, be
achieved either by chemical dissolution or by applying
a mechanical load (e.g. pulsed or continous ultrasound
waves) up to the fragile rupture of the mould.
Finally, also a metallic shell mould built by electroforming up to a thickness of few mm (Ippolito et al.,
1996) has been tested. The master has been covered
with a conductive silver layer and used as a cathode
in a bath containing copper sulphate (125220 g/l)
and sulphuric acid (4075 g/l). Current densities of
1 A/dm2 were adopted. This approach could ensure
thermal removal of the master and electrochemical
inverse reaction for the dissolution of the mould from
the final scaffold.
box dimensions of the model perpendicular to the

axes of the reference frame (height, width, length);
diameters of the main thick branch and of an end
one;
length of the main thick branch and of a middle one
Measures of these geometric features were repeated
on three identical specimens built by ALM.
2.3
calcium sulphate
blends of clay and silica
calcium carbonate
sodium chloride
Replication procedures for scaffold production
At this stage all the replication procedures used to

obtain a test-scaffold were limited to the simplified
SVS model.
2.3.1 Single replication
Starting from the negative of the scaffold produced
with ProJet HD 3000, a single replication procedure
37
RESULTS
In a CAD environment the STL model analysis

showed missing facets (holes in the model surface)
and incorrect normal vectors inhibiting importation
in the ALM machine software. In addition, the model
was characterized by sharp edges in branch crossings
(Figure 3A), giving rise to stress concentration areas
in the physical model and high risk of failure. Surface
repair and edge rounding were achieved though the
CAD operation of sphere sweeping along the branch
axis, and a tubular model was obtained (Figure 3B).
Viper , R66 plus and Eden350 systems were
unable to produce the simplified SVS structure,
whereas only Viper provided a model at double
scale. In contrast, the MJM system ProJet HD 3000
allowed for the construction of a simplified SVS
model (Figure 4). The material used was a UV-curable
resin composed of ethoxylated bisphenol A diacrylate
(1535%), urethane acrylate oligomers (2040%),
and tripropyleneglycol diacrylate (1.53%).
The results of size measurements are listed in
Table 2, together with nominal values obtained from
the STL file and dimensional deviation, which is calculated in the last column. Figure 4 clearly shows that
the diameter of end branches is comparable to the
machine accuracy, leading to substantial variations in
prototyped diameters. Maximum and minimum diameters of a single thin branch were measured and are
listed in Table 2.
Generally, the measurements showed low standard deviation, that means good repeatability, and
dimensional deviation below 0.6 mm. A singularity
can be noticed as to width, since all the models are
around 4 mm smaller than the STL file. This can
3.1 Thyroid SVS modeling

An example of the numerical-computational model of
a human thyroid SVS is shown in Figure 1. Despite
the limitations introduced, the similarity between the
density of the vascular network obtained with this
simulation and that of a real thyroid lobe (analyzed
by either in vivo digital multilayer angiotomography
or injection/corrosion casts from isolated anatomical
specimens) resulted very consistent.
The simplified model of a single SVS branch is
shown in Figure 2.
3.2
Construction of prototypes by ALM
The model shown in Figure 1 consists of a highly tangled structure with branch diameters from 1mm down
to 40 m. This geometry could not be produced with
any of the considered ALM techniques. For this reason
we decided to focus on the simplified SVS model of
Figure 2. It represents a portion of the arterial structure
with branches down to the 3rd order (around 100 m in
maximal diameter) and box dimensions of 16 19
28 mm3 .
Figure 1. Computer modelling of the intraglandular arteries

in the human thyroid lobe (Toni et al., 2011).
Figure 3. A) STL model obtained by numerical modeling; B) Detail of the model after surface closure and edge
rounding.
Figure 2. Simplified model of single SVS branches (Toni

et al., 2011).
38
Figure 5. Mould (negative of the simplified model) built on

ProJet HD 3000 with acrylic resin.
Figure 4. Prototype of the simplified model built on

ProJet HD 3000 with acrylic resin.
Figure 6. Shell mould in calcium sulphate.
Table 2. Dimensional measurements of the prototypes,

compared to nominal values.
width
height
length
main branch dia.
end branch dia.
nominal
measured
from STL mean (SD)
[mm]
[mm]
dimensional
deviation
[mm]
28.11
19.35
16.01
0.38
0.10
4.18
0.58
0.54
+0.03
min +0.01
max +0.06
0.03
0.13
main branch lgth 0.88

mid branch lgth 2.72
23.93 (.91)
18.77 (.69)
15.47 (.11)
0.41 (.02)
min 0.11 (.01)
max 0.16 (.01)
0.85 (.01)
2.59 (.01)
3.3.2 Double replication

Considering materials deposited by dipping, the best
results were obtained for moulds made with calcium
sulphate. The experiments with calcium sulphate in
water provided shells with good consistency, but some
difficulties were encountered: the long time for drying and solidification prevented the achievement of a
constant thickness. In addition, these moulds broke
during the baking cycle for the master removal. The
use of acetone for dispersing calcium sulphate allowed
for a higher evaporation rate and easier construction of
the shell. A high mould porosity was obtained, favoring removal of combustion gases during burning of the
master. Figure 6 shows a middle step of the shell construction, and the final mould containing the master
with average thickness of 3 mm. The mould was dried
for 24 hours at room temperature and then baked to
remove the master. A low heating rate of 1.7 C/min
was adopted up to 400 C to avoid thermal shock and
failure, then temperature was maintained for 2 hrs.
Residues of the master were still present, and a second
cycle including an isothermal of 5 hrs at 400 C was
applied. This treatment was effective in emptying the
channels of the SVS simplified model.
Other ceramic materials were discarded due to an
insufficient consistency of the shell.
The electroforming process of a Cu shell on the silver coated master yielded promising results. The main
thick branch and an end branch were left uncoated to
not be ascribed to dimensional accuracy; instead, it

can easily be explained with the failure of an end
branch, probably during handling and transport of the
prototypes.
3.3
Replication procedures for scaffold production
3.3.1 Single replication

Although we attempted a dissolution in tetrahydrofuran up to 24 hours at 50 C with the aid of sonication,
the mould built by MJM (Figure 5) could not be
emptied from the support material.
39
approximately 0.5 mm. As a consequence, the dimensional constraints of the thyroid SVS are considerably
less critical if the negative geometry is chosen, like in a
single replication approach. In our hands only the MJM
system ProJet HD 3000 allowed for the construction of a physical prototype of the simplified structure.
Dimensional measurements proved that these parts are
built with close tolerance and a degree of geometrical/morphological accuracy consistent with that of the
computational simulation.
Replication is another critical aspect of the innovative procedures for prototyping that we analyzed.
Differential removal, in fact, needs to be achieved on
complex thin geometries. In the case of the single replication approach, a mould was successfully built by
MJM but the cavity inside it could not be emptied
from the support material, and dissolution within the
branches was not achieved.
Two promising solutions for double replication were
identified. The first is based on a calcium sulphate
shell covering the master. In this very case strong attention has to be paid to the thermal removal of the master,
and the shell needs to provide sufficient permeability
to the combustion gases. In addition, small thickness
and sufficient porosity are required to avoid the shell
failure.
The second process is electroforming of a metallic
shell on the master. The tests proved the feasibility
of this procedure that enables to produce a shell with
an inlet and several outlet holes. Thus, injection of
a biocompatible polymer could be performed under
slight vacuum. Further experiments are required to set
up deposition parameters for obtaining the best detail
reproduction and finest shell structure. Previous experiences in this field (Ippolito et al., 1996) proved that
a thorough control of current density, and the adoption of reverse polarity at a certain frequency provide
the best quality of the deposited layer. The set up
of the process for this specific application will be the
objective of future developments of our research, considerably different from previous industrial uses. The
relevance of this procedure relies on its reversibility,
i.e. in the opportunity of achieving mould creation and
removal with the same technology simply reversing
the current flow.
To conclude, based on current ALM technology
we have explored innovative replication approaches to
prototype with biocompatible materials 3D, vascularlike structures. These physical supports are expected
to act as organomorphic scaffolds for ex situ bioengineering of soft tissue organs. In particular, we have
focussed our attempts to the reconstruction of the SVS
of the human thyroid gland. We have been able to collect encouraging data on the possibility to prototype
a simplified SVS model (single branches). However,
our procedures have been designed to be compatible
with more complex model structures, like an entire
SVS. Limitations intrinsic to the accuracy of current
ALM procedures urges further technological development for their efficient application to soft tissue and
organ engineering.
Figure 7. Inlet and outlet holes in the Cu shell.
provide inlet and outlet holes useful for subsequent

filling with scaffold materials (Figure 7).
DISCUSSION
The numerical and computational model, developed

for the thyroid SVS, adopts the idea that physical and
biological structures can be represented via fractal
geometry.At the same time, a deterministic use is made
of all the currently known anatomical information; further deterministic data (should they become available)
could be included and the modelling of other glands
could also be implemented. The overall algorithm currently yields a probabilistic 3D simulation of the SVS,
within the bounds of the 3D reconstructed lobe, built
on a minimal set of planar image data. A more deterministic validation of the simulated SVS model is still
under consideration, possibly employing 3D microCT
information.
From an engineering point of view, reconstruction
of a complete SVS model is beyond the current technological capabilities of ALM processes. Therefore, a
simplified SVS model, as that we used, represents an
ideal benchmark for the development of new prototyping procedures, allowing for a robust estimate of
the feasibility and accuracy of the process chain.
The investigated geometry depicts many critical
aspects: small overall dimensions, thin branches, and
high entanglement. All these variables are challenging
factors for layer construction, requiring a minimum
size entity of 40 m together with a complex support
structure for overhangs and tangled branches, to be
easily removed without damaging the model. In addition, the model material needs to be strong enough
to sustain its own weight. For this reason the STL
file obtained by the numerical modeling requires the
introduction of edge radii. Finally, due to numerical
scattering in the fractal algorithm the model leads to
missing triangles and inverted normals, all to be fixed
for manufacturability.
Although all the investigated ALM techniques guarantee that the minimum building detail is around
0.01 mm, this limit has to be referred to a thin detail
laying on a bulky substrate, i.e. a hole or a step of a
web. If a self-standing branch is considered, the difficulties in thickness regularity, part consistency and
supports removal rise the minimum dimension up to
40
ACKNOWLEDGEMENTS
Liu, C.Z., Han, Z.W., Hourd, P., Czernuszka, J.T. 2008. On

the process capability of the solid free-form fabrication:
a case study of scaffold moulds for tissue engineering,
J.E.I.M. 222 (3): 377391.
Park, A., Wu, B., Griffith, L.G., 1998. Integration of surface
modification and 3D fabrication techniques to prepare
patterned poly(l-lactide) substrates allowing regionally
selective cell adhesion. J Biomater Sci Polym E 9:89110.
Sachlos, E., Reis, N., Ainsley, C., Derby, B., Czernuszka,
J.T. 2003. Novel collagen scaffolds with predefined internal morphology made by solid freeform fabrication.
Biomaterials 24 (8): 14871497.
Spaletta G. 2004. Reconstruction in space and visualization of a planar image: a mathematical and computational
introduction. Acta Biomed 78 (1): 2631.
Taylor, P.M., Sachlos, E., Dreger, S.A., Chester, A.H.,
Czernuszka, J.T., Yacoub, M.H. 2006. Interaction of
human valve interstitial cells with collagene matrices
manufactured using rapid prototyping. Biomaterials 27:
27332737.
Toni, R., Della Casa, C., Spaletta, G., Marchetti, G., Mazzoni,
P., Bodria, M., Ravera, S., Dallatana, D., Castorina, S.,
Riccioli, V., Castorina, E.G., Antoci, S., Campanile, E.,
Raise, G., Rossi, R., Ugolotti, G., Martorella, A., Roti, E.,
Sgallari, F., Pinchera, A., 2007. The bioartificial thyroid:
a biotechnological perspective in endocrine organ engineering for transplantation replacement. Acta Biomed. 78
(1): 129155.
Toni, R., Della Casa, C., Bodria, M., Spaletta, G., Vella, R.,
Castorina, S., Gatto, A., Teti, G., Falconi, M., Rago, T.,
Vitti, P., Sgallari, F. 2008. A study on the relationship
between intraglandular arterial distribution and thyroid
lobe shape: Implications for biotechnology of a bioartificial thyroid. Annals of Anatomy 190 (5): 432441.
Toni, R., Tampieri, A., Zini, N., Strusi, V., Sandri, M., Martin,
I., and the Italian Study Group on Ex Situ Regenerative
Biology of Glandular/ Parenchymal Organs (PRIN2008)
MorphoSynthe/Organomorpho Research Groups on Scaffolds for Regeneration of Skeletal Tissues (FRIB2010),
2011. Ex situ bioengineering of bioartificial endocrine
glands: a new frontier in regenerative medicine of soft
tissue organs. Annals of Anatomy (in press).
Vogel, V., Sheetz, M., 2006. Local force and geometry sensing
regulate cell function. Nat. Rev. Mol. Cell Biol. 7: 265
275.
Yeong, W.Y., Chua, C.K., Leong, K.F. 2006. Indirect fabrication of collagen scaffold based on inkjet printing
technique. Rapid Prototyping Journal 12(4): 229237
Yeong, W.Y., Sudarmadji, N., Yu, H.Y., Chua, C.K., Leong,
K.F., Venkatraman, S.S., Boey, Y.C.F., Tan, L.P., 2010.
Porous polycaprolactone scaffold for cardiac tissue engineering fabricated by selective laser sintering. Acta
Biomaterialia 6(6): 20282034.
The authors like to thank Dr Lavrijsen (TNO Science

and Technology, Rapid Manufacturing Demo Centre)
for technical support during the preparation of the
ALM prototypes. This study has been possible by
Grant MIUR PRIN 2008ZCCJX4 entitled: Ex situ
regenerative biology of glandular/parenchymal organs:
the model of the organomorphic scaffold
REFERENCES
Della Casa, C., Spaletta, G., Bodria, M., Ravera, S.,
Dallatana, D., Castorina, S., Martorella, A., Toni, R., 2006.
A fractal model for bioengineering of the stromal/vascular
scaffold of a bionic human thyroid gland. Ital. J. Anat.
Embryol 11 (23): 75.
Duan, B., Wang, M., Zhou, W., Cheung, W., Li, Z., Lu,
W., 2010. Three-dimensional nanocomposite scaffolds
fabricated via selective laser sintering for bone tissue
engineering. Acta Biomaterialia 6(12): 44954505.
Eshraghi, S., Das, S., 2010. Mechanical and microstructural properties of polycaprolactone scaffolds with onedimensional, two-dimensional, and three-dimensional
orthogonally oriented porous architectures produced by
selective laser sintering. Acta Biomaterialia 6(7): 2467
2476.
Giannatsis, J., Dedoussis, V., 2009. Additive fabrication technologies applied to medicine and health care: a review. Int
J Adv Manuf Technol. 40: 116127
Huang, H., Oizumi, S., Kojima, N., Niino, T., Sakai, Y., 2007.
Avidin-biotin binding-based cell seeding and perfusion
culture of liver-derived cells in a porous scaffold with a
three-dimensional interconnected flow-channel network.
Biomaterials 28(26): 38153823
Hutmacher, D.W., Schantz, T., Zein, I., Ng, K.W., Teoh, S.H.,
Tan, K.C., 2001. Mechanical properties and cell cultural
response of polycaprolactone scaffolds designed and fabricated via fused deposition modeling. J Biomed Mater
Res 55:20316.
Ingber, D.E., 2005. Mechanical properties of tissue growth:
function follows form. Proc. Natl. Acad. Sci. USA 102:
1157111572.
Ippolito, R., Iuliano, L., Gatto, A., 1996. Edm tooling by solid
freeform fabrication and electroplating techniques. Proc.
7th Solid Freeform Fabrication Symposium, Austin, 1214
August 1996: 199206.
Kleinman, H.K., Klebe, R.J., Martin G.R., 1981. Role of collagenous matrices in the adhesion and growth of cells.
J Cell Biol 88:47385.
41

A novel protein-based scaffold with macro- and micro-structural

features for tissue engineering applications
M.J.J. Liu, S.M. Chou & C.K. Chua
School of Mechanical and Aerospace Engineering, Nanyang Technological University, Singapore
ABSTRACT: Silk Fibroin (SF) protein, which is derived from the cocoons of the Bombyx Mori silkworms, has
been used widely for biomedical applications, such as surgical sutures, owning to its high mechanical strength and
biocompatibility. Recent advances in the Tissue Engineering (TE) field have acknowledged and recognized SF
as a feasible candidate for tissue repair and regeneration; the regenerated form of SF protein in sponge networks
has demonstrated potential for both soft and hard tissue replacements. However, adequate cell growth within
the SF foams remained as a challenge to TE. Thus, in the present study, novel SF scaffolds with macro-channels
were fabricated using the Additive Manufacturing (AM) or Rapid Prototyping (RP) technology. This work is, to
our knowledge, the first generation of SF tissue constructs with both micro- and macro-scaled structures. The
SF scaffolds were formed by casting aqueous SF solution to negative moulds made from a 3D inkjet printer. In
vitro examinations of the scaffold specimens revealed that the indirect approach does not induce toxicity and
were able to maintain the intrinsic properties of the naturally-derived biomaterial.
INTRODUCTION
Additive Manufacturing (AM) or Rapid Prototyping (RP) technology has been recognized as a viable
alternative to the conventional TE approaches (Yang
et al. 2002, Leong et al. 2003, Yeong et al. 2004, Liu
et al. 2007). Existing RP approaches combined with
Computer Aided Designs (CAD), which consists of
pre-designed features, can fabricate TE constructs in a
layer-by-layer manner. This allows precise control over
the dimensional and structural aspects of the resultant
TE scaffolds. The CAD design process can also be
automated to produce scaffold assemblies of different
unit cell shapes and porosity gradients (Chua et al.
2003a, Chua et al. 2003b, Cheah et al. 2004, Naing
et al. 2005, Leong et al. 2008, Sudarmadji et al. 2011).
These advantages have led to the feasibility in yielding
patient-specific scaffold constructs for tissue repair
and regeneration. To date, RP techniques are classified
into the two categories: particle-bonding and meltdissolution deposition (Yeong et al. 2004). Briefly,
particle bonding techniques include selective laser sintering (SLS) and 3-dimensional printing (3DP), where
powder particles are selectively sintered and bonded
respectively (Chua et al. 2004, Tan et al. 2005, Wiria
et al. 2007, Lam et al. 2008, Yeong et al. 2010). The
unprocessed powders within the resultant tissue constructs will appear as voids or macro-channels after
a simple removal step. The melt-dissolution deposition methods, such as the Fused Deposition Modeling
(FDM), require the biomaterial in semi-liquid state in
order to produce fiber-assembled scaffold networks
(Cao et al. 2003, Shao et al. 2006, Ramanath et al.
2007, Zeng et al. 2008). Overall, the RP technology has brought about substantial improvements to
Over the recent years, Silk Fibroin (SF) protein has

emerged as a potential biomaterial for biomedical
applications. Particularly, in Tissue Engineering (TE),
the naturally-derived material has been explored as tissue constructs for ligament, cartilage and bone repairs,
owing to its versatility in processing and superior biocompatibility (Altman et al. 2003, Chen et al. 2008,
Fan et al. 2009). The SF protein can be used as tissue
scaffolds either in its native or regenerated form. The
mechanical properties of regenerated SF porous foams
have since showed feasibility in bone and cartilage tissue regeneration (Wang et al. 2010, Leal-Egaa et al.
2010). These foams are often processed using conventional approaches where user-defined control over
the internal architecture of the SF sponges is limited.
The absence of desirable macro- and micro-structural
features within the SF sponges, may lead to compromises in cellular attachment, migration and finally
neo-tissue regeneration (Yang et al. 2001, Yeong et al.
2007, Tan et al. 2010, Liu et al. 2010b). The limitations
are generally attributed to the lack of mass transfer
of nutrients and removal of metabolic wastes within
the foam constructs. As a result, cell growth and proliferation are commonly seen only at the peripheries
of the SF sponges. Moreover, it can be foreseen that
the foam networks may encounter more challenges
when the engineering of organs and tissues (which
are large and structurally complex) are considered.
Therefore, it will be useful to incorporate customized
design options within the SF sponges for wider range
of TE applications.
43
3D inkjet printer (Model: T612, Benchtop, Solidscape Inc.). Using the droplet-based approach, each
thin layer formed consists of two distinct materials,
InduraCast (thermoplastic) and InduraFill (wax),
to produce the mould and temporary supporting features respectively. The support materials were removed
by immersing the printed part into mineral oil at 60 C,
with constant stirring. The unwanted solutes and oil
residues within the mould structures were blown off
using an air gun. Subsequently, the moulds were left
to dry in the fume hood for 3 days.
the design and development of TE scaffolds; they

include: (1) high reproducibility, (2) macro-sized features or channels to facilitate cell growth within
scaffolds and (3) customizable designs to achieve suitable mechanical strength and better biocompatibility
of the tissue constructs (Peltola et al. 2008, Bartolo
et al. 2009, Liu et al. 2010a). Nevertheless, the RP
approaches involved harsh processing conditions or
the presence of binders. The drawbacks therefore limit
the material choices mainly to synthetic polymers as
naturally-derived materials tend to denature or degrade
when subjected to unsuitable conditions such as high
temperature and exposure to toxic solvents (Yeong
et al. 2006). Furthermore, the surface topography of
each scaffold construct, which is critical to cell attachment, is highly dependent on the RP technique used
(Taboas et al. 2003). For instance, the rapid solidification of the extruded fibers via the FDM can lead to
formation of struts with in significant surface roughness. Although the morphologies of the SLS and 3DP
scaffold models appeared to be more distinctive as
compared to the FDM, the topographical features are
limited by the processing conditions as well as the
average particle size of the raw powder material (Lee
et al. 2005). To gain sufficient surface roughness for
adequate cell attachment, conventional TE techniques
perhaps will provide as favorable approaches since
the concentration effects of biomaterials (for both
synthetic and naturally-occurring) has demonstrated
to influence the micro-porosities of foam or sponge
structures (Mandal et al. 2009, Liu et al. 2011). The
combination of traditional and advanced TE methods,
also known as the indirect RP approach, is thus suggested as a feasible option to fabricate TE scaffold
builds with both macro- and micro-architectures.
The proposed study is motivated to introduce interconnected macro-channels into naturally-derived SF
without altering the intrinsic properties of the protein peptides. Even though indirect RP methods have
been widely applied for TE, SF protein has yet to be
explored with this advanced manufacturing technique.
Thus, this present work encompassed an original innovation by producing pre-designed SF tissue constructs
using the RP technology. The interconnected channels
within the SF tissue constructs are created by casting
aqueous SF solution into sacrificial moulds built using
a 3D inkjet printer. It was observed that the 3D SF scaffolds maintained the intrinsic properties of its native
form despite the multiple fabrication steps involved.
In addition, a cytotoxicity assessment revealed that the
novel protein-based tissue scaffolds is non-toxic and
therefore is suitable for tissue repair and regeneration.
2
2.1
2.2 Manufacturability evaluation of the

3D inkjet printer
To assess the dimensional deviation between the
designed and measured feature size, a slab of 1 mm
in thickness with circular through-hole profiles of 0.5,
0.75, 1 and 2 mm diameters were printed. The slab
consists of five sets of through-hole profiles. The
built features were imaged using the scanning electron microscope (SEM) (Model: JCF1600, JEOL)
and subsequently measured using a public domain
image-processing program (ImageJ, National Institute
of Health (NIH), USA). The roundness profiles of the
measured feature were also evaluated using the outof-roundness (OOR) measurement according to the
ANSI Y14.5M1994 and ANSI B89.3.11972. The
establishment of the dimensional measurements
served to provide a basis to understand the geometric tolerance of the RP builds as well as to identify the
effects of control parameters on the resulting builds.
2.3 Fabrication of 3-dimensional SF scaffolds
with internal channels
Aqueous SF, derived from the cocoons of Bombyx
Mori silkworms (Treenway Silks, Canada) with concentration of 10% (wt/wt) was cast into the sacrificial
thermoplastic moulds. The cast moulds were allowed
to stand for 15 minutes under room temperature to
ensure full penetration of the regenerated SF within
the scaffold template. The moulds containing the SF
solution were subsequently frozen at 80 C for at
least 12 hours and freeze dried at 85 C for 24 hours
to obtain porous silk sponges. After the lyophilisation process, the specimens were immersed in ethanol
(95% in methanol) for 20 minutes to induce -sheet
structure of silk and insolubility in aqueous solutions.
This procedure was followed by another 2 hours of
freeze drying to sublimate the solvent before removing
the moulds.
Upon achieving the cross-linked SF constructs,
the specimens were immersed into boiling deionized
water to remove the thermoplastic mould materials.
Once the mould materials were completely removed,
the SF scaffolds were freeze dried to yield tissue
constructs with defined macro-channels and morphological micro-features. For the assessment of the fine
structural characteristics of the 3D SF tissue scaffold,
crystallized and non-crystallized SF foams were used

Sacrificial mould fabrication
The negative sacrificial moulds were designed using

commercial computer aided design (CAD) software (ProEngineer, PTC) and manufactured using a
44
to form colored formazan product that is soluble in the

tissue culture medium. The quantity of the formazan
product, which is directly proportional to the number
of living cells, can be measured by the absorbance at
490 nm. Finally, the cell viability of each culture well,
in terms of percentage, is evaluated using the following
formula:
for comparison. It should be understood that the mould

removal step was not applied to the SF sponges and
thus the proportions of secondary structure formed
may differ slightly.
2.4 Thermal analysis of SF tissue scaffolds
and foams using a differential scanning
calorimeter (DSC)
Viability (%) = (100 OD490e ) OD490b
The thermal properties of the SF specimens were

analyzed using a differential scanning calorimeter
(Model: Diamond DSC, Perkin Elmer). Trimmed SF
portions of approximately 2.5 mg each were sealed in
aluminum cell and the course of investigation was conducted under a nitrogen atmosphere, at a heating rate
of 10 C/min from 30 to 400 C.
where OD490e is the mean value of the measured optical density of the 100% extracts of the test sample and
OD490b is the mean value of the measure optical density of the negative control. If the viability of the extract
is reduced to less than 70% of the negative control, it
is concluded to have a cytotoxic potential.
2.5
2.7
Structural analysis of SF tissue constructs and

foams using a fourier transform infrared (FTIR)
spectrometer
The macro- and micro-structural features of the SF

scaffolds were examined using a scanning electron
microscope. The specimens were air dried, placed on
an aluminum holder and sputtered with gold before
mounting onto the SEM and viewed using an acceleration voltage of 10.0 kV.
The analysis of the secondary structures of the SF specimens was conducted using FTIR spectrometer. For
each sample set, approximately 1 mg of regenerated
SF was pressed into a pellet and each spectrum was
recorded in transmittance mode with an accumulation
of 10 scans with a resolution of 4 cm1 and a spectral
range of 4000 to 400 cm1 .
2.6
Imaging of SF tissue scaffolds using the

scanning electron microscope (SEM)
In the present study, a 3D inkjet printer which operates based on RP technology was used to fabricate
3D SF protein tissue scaffolds. The initial phase
of the proposed work included the assessment of
the manufacturability of the inkjet printer system.
Table 1 displayed the deviation between the dimensions between the designed and as-built features. It
was shown that the percentage diameter error increases
as the designed featured size decreased. Nevertheless,
the measured roundness demonstrated better circular profile with decrease in diameter of through-hole
samples; the observation suggests that the shape configuration is maintained by the printing system. The
current work attributed the dimensional discrepancies
to the change in properties of the build and support
material, which had been evaluated previously by Liu
et al. 2007.
Using the thermoplastic builds fabricated via the
3D inkjet printer (Figure 1a), 3D SF tissue constructs
with well-defined internal macro-channels were successfully created (Figure 1b) and replicated the input
CAD model. It can be seen that the presence of blue dye
pigments (from the sacrificial mould) did not adhere
to the SF structure. Hence, suggesting that the thermoplastic mould dissolves completely upon exposure to
boiling deionized water.The observation can be further
justified by assessing the SEM images of the SF constructs. As presented in Figure 2a, the macro-channels
were formed upon the dissolution of the sacrificial
mould. Concurrently, the micro-scaled porosities of
the SF protein scaffolds were similar to that of conventional SF foams; SF sponges are understood to exhibit
Cytotoxicity assessment
The RP-built SF tissue scaffolds were tested for cytotoxicity towards the growth of fibroblasts with reference to the ISO 10993 (Part 5 and Part 12) guidelines.
Briefly, the SF 3D constructs were immersed in culture
medium (DMEM supplemented by 10% (v/v) fetal
bovine serum and 1% (v/v) penicillin/streptomycin
mixture, Invitrogen) and incubated at 37 1 C for
72 2 hours. The extract of Para rubber was used as
the positive control while the negative control was the
standard culture medium.
Using a 96-well tissue culture plate, 1 104 3T3
cells per 100 l of standard culture medium (mouse
embryo fibroblasts) were seeded into each well to
achieve a sub-confluent monolayer after 24 2 hours.
The cells were examined under the microscope to
ensure even cell growth and morphologies before proceeding to the following steps. Once the consistencies
among the wells were verified, the standard culture
medium was aspirated and the cells were challenged
with the extracts of the scaffold, positive control and
negative control. To ensure reproducibility, eight sets
of cells (n = 8) were used for each type of extract. The
culture plate was incubated for another 24 2 hours.
The quantitative results of the cytotoxic test were
obtained by the use of the Celltiter 96 AQueous
One Solution cell proliferation assay (MTS, Promega),
which contains a tetrazolium compound and an electron coupling reagent. The MTS compound, upon
added to each cultured well plates, is bio-reduced by
the hydrogenase enzymes in metabolically active cells
45
-sheet configuration. The induction of -sheet, which

are identified in the amide I (16101645 cm1 ) and III
(12201620 cm1 ) regions, aimed to increase the dissolution resistance of SF in aqueous medium and also
to enhance the mechanical property and biodegradability of the protein structure. The shoulders at
1265 cm1 of Figure 3a and 3b were also identified as
clear markers to the presence of crystalline characteristic bands (Nam and Park 2001). The non-crystallized
SF sponges, on the other hand, showed distinctive peak
of -helical structure at 1650 cm1 and no distinct
shoulder at the amide III band.
leave-like morphologies which facilitates cell anchorage and subsequently tissue growth (Nam et al. 2001).
In addition, it was noted that the topographies of the
defined channels appeared to be denser or more compact as compared to SF foams (Figure 2b and 2c).
The increase in solidity of the macro-sized features
was contributed by the constant bombardment of water
molecules to the channel walls. The effects of temperature have been suggested to influence the chemical
or molecular configuration of SF, therefore resulting
in the densification of the channel walls (Putthanarat
et al. 2002). However, it should be re-emphasized that
the purpose of the channels is to encourage cell migration and mass transport of essential nutrients within the
SF scaffolds. Thus, minimal cell proliferation is anticipated on the channel walls while the cell processes
will be more prominent in regions with the leave-like
SF morphologies.
The fine structures of the 3D SF scaffolds were
also analyzed to ensure that (1) there were no traces
of the sacrificial mould and (2) the intrinsic properties of the protein were maintained. The crystallinity
of the 3D SF constructs was verified by using the
FTIR. The FTIR spectra recorded serve to identify
the molecular configurations within the amide I, II
and II and IV modes (Singh 2000). As the protein peptides were aggregated due to the formation of -sheets,
the spectral data were de-convolved in order for the
examination of the secondary structures. Figure 3ac
displayed the characteristic bands of the cross-linked
3D SF scaffolds, cross-linked SF foams and noncross-linked SF foams respectively. The de-convolved
spectra indicated that the 3D constructs exhibited similar molecular configuration to that of the crystallized
10% wt/wt SF foams, thus indicating the formation of
Table 1. Dimensional measurements of the slab template
manufactured by the 3-D inkjet printer.
Designed Measured
Diameter Diameter
(m)
(m)
2000
1000
750
500
Diameter
error (%)
Measured
out-ofroundness
(m)
1887.34 19.27 5.63 0.96 118 21.61

889.17 14.30 11.08 1.43
78 16.14
610.12 8.88 18.65 1.18 68.5 12.07
361.34 7.98 27.73 1.60 51.5 8.22
Figure 2. Scanning electron microscopic images of the

(a) top view (30), (b) close-up view (75) and
(c) cross-sectional profile of the SF TE scaffolds. The arrows
indicate the width of the macro-channels.
Figure 1. (a) Printed thermoplastic mould (after removal

of support mater) and (b) SF scaffold with interconnected
macro-channels obtained using the sacrificial mould.
46
Figure 4. The DSC thermogram of the (a) SF tissue scaffold is represented by the lined plot. The (b) dashed and (c)
dashed-dot lined plots are obtained from the 10% (wt/wt)
cross-linked and non-cross-linked SF foams respectively.
Figure 3. (a) Typical FTIR spectral data of SF tissue constructs. (b) and (c) are spectra representing the 10% (wt/wt)
cross-linked and non-cross-linked SF sponges respectively.
The indicated arrows highlight the shoulder which resembled the -sheet conformation.
The thermal properties of SF specimens were measured by dynamic heating at a constant rate. The DSC
procedure is useful in identification of any phase
transitions such as vaporization, crystallization and
decomposition during the heating process (Agarwal
et al. 1997). As presented in Figure 4a and 4b, which
resembled the thermal plots of the 3D SF scaffold
and crystallized SF foams respectively, two endothermal peaks were observed. In particular, there was no
exothermal peak seen for the 3D SF construct hence
indicating minimal or negligible traces of the sacrificial mould (which melting temperature is around
100 C). The first peak which occurred at approximately 50 C is attributed to the vaporization of water
molecules while the second endothermal peak signified the thermal decomposition of the SF samples. It
was interestingly noted that the 3D SF constructs and
cross-linked SF foams decomposed at 298 and 288 C
respectively. The current work proposed that the difference in decomposition temperatures is affected by the
process treatments used to produce both crystallized
structures. The slight variation in process conditions
may lead to the change in the proportion of secondary structures (such as -turn, -helical and -sheet
structures) formed. In contrast, the non-crystallized
SF foams displayed an additional exothermal peak
at 224 C (Figure 4c) which had been recognized as
the release of heat energy upon the formation of
stable -sheets. This observation further illustrates
how the molecular arrangements of the SF protein
can be affected by the presence of heat or elevated
temperature.
The cytotoxicity examination of the 3D SF scaffolds
was conducted using the 3T3 fibroblast cells. From
the quantitative results displayed in Figure 5, it can be
inferred that the 3D scaffolds portrayed cell viability
effectiveness similar to that of the negative control.
Since the cell survival rates of the scaffold extracts
Figure 5. The relative cell viability percentage of cells

cultured in extracts derived from standard culture medium
(negative control), scaffold extracts (100, 50, 25 and 10%
v/v respectively) and Para rubber extracts (positive control).
A total of 8 samples were tested (n = 8) were tested in each
group. *Indicates a significant difference is observed when
compared to the negative control group (p < 0.05).
exceed 70%, it can be postulated that the indirect RP

method imposed on the SF protein induce insignificant toxicity. In contrast, cells exposed to the positive
control experienced obviously lower cell viability, thus
showing indications of toxic potential.
The thermoplastic builds from the 3D inkjet printer
have been demonstrated to produce TE scaffolds
with interconnected macro-channels. The varieties of
biomaterials used for the tissue constructs include
synthetic polymers, ceramics as well as naturallyoccurring proteins and polysaccharides (Lee et al.
2010, Jeong et al. 2010, He et al. 2008, Xu et al. 2007).
In particular, collagen type I appeared to be the most
commonly used natural biomaterial for the indirect RP
approach (Sachlos et al. 2003, Yeong et al. 2006, Liu
et al. 2008). However, it is important to understand
that collagen type I possessed weaker mechanical
properties as well as lower temperature resistance as
compared to the SF protein (Altman et al. 2003). Under
higher temperature ambient such as 100 C, collagen
type I may denature and lose its intrinsic characteristics
47
Fabricated by Fused Deposition Modeling. Tissue Engineering, 9(supplement 1), 103112.

Cheah, C.-M., Chua, C.-K., Leong, K.-F., Cheong, C.-H. &
Naing, M.-W. (2004) Automatic Algorithm for Generating Complex Polyhedral Scaffold Structures for Tissue
Engineering. Tissue Engineering, 10(34), 595610.
Chen, X., Qi, Y.-Y., Wang, L.-L., Yin, Z., Yin, G.-L., Zou, X.H. & Ouyang, H.-W. (2008) Ligament regeneration using
a knitted silk scaffold combined with collagen matrix.
Biomaterials, 29(27), 36833692.
Chua, C. K., Leong, K. F., Cheah, C. M. & Chua, S. W.
(2003a) Development of a tissue engineering scaffold
structure library for rapid prototyping. Part 1: Investigation and lassification International Journal of Advanced
Manufacturing Technology, 21(4), 291301.
Chua, C. K., Leong, K. F., Cheah, C. M. & Chua, S.
W. (2003b) Development of a tissue engineering scaffold structure library for rapid prototyping. Part 2:
Parametric library and assembly program International
Journal of Advanced Manufacturing Technology, 21(4),
302312.
Chua, C. K., Leong, K. F., Tan, K. H., Wiria, F. E. & Cheah,
C. M. (2004) Development of tissue scaffolds using selective laser sintering of polyvinyl alcohol/hydroxyapatite
biocomposite for craniofacial and joint defects. Journal of Materials Science: Materials in Medicine, 15(10),
11131121.
Fan, H., Liu, H., Toh, S. L. & Goh, J. C. H. (2009) Anterior
cruciate ligament regeneration using mesenchymal stem
cells and silk scaffold in large animal model. Biomaterials,
30(28), 49674977.
He, J., Li, D., Liu, Y., Gong, H. & Lu, B. (2008) Indirect
fabrication of microstructured chitosan-gelatin scaffolds
using rapid prototyping Virtual and Physical Prototyping
3(3), 159166.
Jeong, C. G. & Hollister, S. J. (2010) A comparison of
the influence of material on in vitro cartilage tissue
engineering with PCL, PGS, and POC 3D scaffold architecture seeded with chondrocytes. Biomaterials, 31(15),
43044312.
Lam, C. X. F., Olkowski, R., Swieszkowski, W., Tan, K. C.,
Gibson, I. & Hutmacher, D. W. (2008) Mechanical and in
vitro evaluations of composite PLDLLA/TCP scaffolds
for bone engineering. Virtual and Physical Prototyping,
3(4), 193197.
Leal-Egaa, A. & Scheibel, T. (2010) Silk-based materials
for biomedical applications. Biotechnology and Applied
Biochemistry, 55(3), 155167.
Lee, K.-W., Wang, S., Dadsetan, M.,Yaszemski, M. J. & Lu, L.
(2010) Enhanced Cell Ingrowth and Proliferation through
Three-Dimensional Nanocomposite Scaffolds with Controlled Pore Structures. Biomacromolecules, 11(3), 682
689.
Lee, M., Dunn, J. C. Y. & Wu, B. M. (2005) Scaffold fabrication by indirect three-dimensional printing. Biomaterials,
26(20), 42814289.
Leong, K. F., Cheah, C. M. & Chua, C. K. (2003) Solid
freeform fabrication of three-dimensional scaffolds for
engineering replacement tissues and organs. Biomaterials,
24(13), 23632378.
Leong, K. F., Chua, C. K., Sudarmadji, N. & Yeong, W.
Y. (2008) Engineering functionally graded tissue engineering scaffolds. Journal of the Mechanical Behavior of
Biomedical Materials, 1(2), 140152.
Liu, C. Z., Sachlos, E., Wahl, D. A., Han, Z. W. & Czernuszka, J. T. (2007) On the manufacturability of scaffold
mould using a 3D printing technology. Rapid Prototyping
Journal, 13(3), 163.
and functions while the SF protein, in the present study,

was shown to decompose at 298 C. In addition, the
SF protein, which is derived from silkworm cocoons
served as a more economical choice since the raw
materials are readily available and SF sponges have
also been identified potentially for bone TE applications as well (MacIntosh et al. 2008). The TE potential
of collagen type I, conversely, is dependent upon its
antigenicity due to the animal origin (Pulkkinen et al.
2010). Thus, the present study recommends the SF
protein as a biomaterial suitable for the integration
of RP-based internal architectural features. The RP
technology has provided an effective pathway to cell
migration and mass transport of essential nutrients
within the SF tissue scaffolds. Nevertheless, future
work should address the design optimization of the 3D
scaffold for its intended TE purposes. The enhancement procedures shall include identifying effective
cross-linking densities of SF for cell attachment and
assessing the appropriate macro-channel sizes for cellular transport mechanisms. The concept of functional
gradient may also be introduced. In addition, in vitro
cell-based investigations and in vivo transplantations
should also be conducted to observe the cell response
to the SF constructs and any presence of foreign body
reaction respectively.
4
CONCLUSION
This study successfully manufactured SF tissue scaffolds with 3D interconnected macro-features for the
facilitation of cellular-based mass transport movements. The novel pre-designed SF scaffolds were
yielded using sacrificial moulds that were fabricated
using a 3D inkjet printer. The microscopic examinations conducted in the current work revealed the
formation of both macro- and micro-structures within
the SF tissue scaffolds while the fine structure analysis
of the constructs revealed that the intrinsic properties
of the protein were preserved despite the multi-step
fabrication procedures. Moreover, extracts of the SF
scaffolds showed no toxicity potential, hence deeming
the SF protein as a suitable material for the integration
of AM or RP technology.
REFERENCES
Agarwal, N., Hoagland, D. A. & Farris, R. J. (1997) Effect
of moisture absorption on the thermal properties of Bombyx mori silk fibroin films. Journal of Applied Polymer
Science, 63(3), 401410.
Altman, G. H., Diaz, F., Jakuba, C., Calabro, T., Horan, R.
L., Chen, J., Lu, H., Richmond, J. & Kaplan, D. L. (2003)
Silk-based biomaterials. Biomaterials, 24(3), 401416.
Bartolo, P. J., Chua, C. K., Almeida, H. A., Chou, S. M. &
Lim, A. S. C. (2009) Biomanufacturing for tissue engineering: Present and future trends. Virtual and Physical
Prototyping, 4(4), 203216.
Cao, T., Ho, K.-H. & Teoh, S.-H. (2003) Scaffold Design
and in Vitro Study of Osteochondral Coculture in a
Three-Dimensional Porous Polycaprolactone Scaffold
48
Liu, C. Z., Xia, Z. D., Han, Z. W., Hulley, P. A., Triffitt, J. T. & Czernuszka, J. T. (2008) Novel 3D collagen scaffolds fabricated by indirect printing technique
for tissue engineering. Journal of Biomedical Materials Research Part B: Applied Biomaterials, 85B(2), 519
528.
Liu, M. J. J., Chou, S. M. & Chua, C. K. 2010a. Development
of silk fibroin scaffolds using an indirect rapid prototyping
technology. In International Conference on Cellular &
Molecullar Bioengineering. Singapore.
Liu, M. J. J., Chou, S. M. & Chua, C. K. 2010b. Proteinbased scaffold fabrication using Polyjet technology. In 4th
International Conference PMI. Belgium.
Liu, M. J. J., Chou, S. M. & Chua, C. K. 2011. The study
of micro- and macro-structural features of silk fibroin
scaffolds. In First International Symposium on Bioengineering, eds. C. K. Chua, M. Y. Kee, Y. L. Guan, J.P.Chen,
K. Q. Luo, C. S. Lai, C. K. Kwoh, K. S. Chuan & S. L.
Wu. Singapore: Research Publishing Services.
MacIntosh, A. C., Kearns, V. R., Crawford, A. & Hatton, P.,
V. (2008) Skeletal tissue engineering using silk biomaterials. Journal of Tissue Engineering and Regenerative
Medicine, 2(23), 7180.
Mandal, B. B. & Kundu, S. C. (2009) Cell proliferation
and migration in silk fibroin 3D scaffolds. Biomaterials,
30(15), 29562965.
Naing, M. W., Chua, C. K., Leong, K. F. & Wang, Y. (2005)
Fabrication of customised scaffolds using computer-aided
design and rapid prototyping techniques. Rapid Prototyping Journal, 11(4), 249259.
Nam, J. & Park, Y. H. (2001) Morphology of regenerated
silk fibroin: Effects of freezing temperature, alcohol addition, and molecular weight. Journal of Applied Polymer
Science, 81(12), 30083021.
Peltola, S. M., Melchels, F. P. W., Grijpma, D. W. &
Kellomki, M. (2008) A review of rapid prototyping
techniques for tissue engineering purposes. Annals of
Medicine, 40(4), 268280.
Pulkkinen, H. J., Tiitu, V., Valonen, P., Jurvelin, J. S., Lammi,
M. J. & Kiviranta, I. (2010) Engineering of cartilage in
recombinant human type II collagen gel in nude mouse
model in vivo. Osteoarthritis and Cartilage, 18.
Putthanarat, S., Zarkoob, S., Magoshi, J., Chen, J. A., Eby, R.
K., Stone, M. & Adams, W. W. (2002) Effect of processing temperature on the morphology of silk membranes.
Polymer, 43(12), 34053413.
Ramanath, H. S., Chandrasekaran, M., Chua, C. K., Leong,
K. F. & Shah, K. D. (2007) Modeling of extrusion behavior
of biopolymer and composites in fused deposition modeling. Key Engineering Materials, 334335, 12411244.
Sachlos, E., Reis, N., Ainsley, C., Derby, B. & Czernuszka,
J. T. (2003) Novel collagen scaffolds with predefined
internal morphology made by solid freeform fabrication.
Shao, X., Goh, J. C. H., Hutmacher, D. W., Lee, E. H. &
Zigang, G. (2006) Repair of Large Articular Osteochondral Defects Using Hybrid Scaffolds and Bone MarrowDerived Mesenchymal Stem Cells in a Rabbit Model.
Singh, B. R. 2000. Basic aspects of the technique and applications of infrared spectroscopy of peptides and proteins. In
Infrared analysis of peptides and proteins. Principles and
applications, ed. B. R. Singh, 237. Washington, D.C.:

Oxford University Press.
Sudarmadji, N., Tan, J.Y., Leong, K. F., Chua, C. K. & Loh,Y.
T. (2011) Investigation of the mechanical properties and
porosity relationships in selective laser-sintered polyhedral for functionally graded scaffolds. Acta Biomaterialia,
7(2), 530537.
Taboas, J. M., Maddox, R. D., Krebsbach, P. H. & Hollister, S.
J. (2003) Indirect solid free form fabrication of local and
global porous, biomimetic and composite 3D polymerceramic scaffolds. Biomaterials, 24(1), 181194.
Tan, J. Y., Chua, C. K. & Leong, K. F. (2010) Indirect
fabrication of gelatin scaffolds using rapid prototyping
technology Virtual and Physical Prototyping, 5(1), 4553.
Tan, K. H., Chua, C. K., Leong, K. F., Cheah, C. M., Gui, W.
S., Tan, W. S. & Wiria, F. E. (2005) Selective laser sintering of biocompatible polymers for applications in tissue
engineering. Bio-Medical Materials & Engineering, 15,
113124.
Wang,Y., Bella, E., Lee, C. S. D., Migliaresi, C., Pelcastre, L.,
Schwartz, Z., Boyan, B. D. & Motta, A. (2010) The synergistic effects of 3-D porous silk fibroin matrix scaffold
properties and hydrodynamic environment in cartilage
tissue regeneration. Biomaterials, 31(17), 46724681.
Wiria, F. E., Leong, K. F., Chua, C. K. & Liu, Y. (2007)
Polycaprolactone/hydroxyapatite for tissue engineering
scaffold fabrication via selective laser sintering. Acta
Xu, S., Li, D., Lu, B., Tang, Y., Wang, C. & Wang, Z.
(2007) Fabrication of a calcium phosphate scaffold with
a three dimensional channel network and its application to perfusion culture of stem cells. Rapid Prototyping
Journal.
Yang, S., Leong, K.-F., Du, Z. & Chua, C.-K. (2001) The
Design of Scaffolds for Use in Tissue Engineering. Part I.
Traditional Factors. Tissue Engineering, 7(6), 679689.
Yang, S., Leong, K.-F., Du, Z. & Chua, C.-K. (2002) The
Design of Scaffolds for Use in Tissue Engineering. Part II.
Rapid Prototyping Techniques. Tissue Engineering, 8(1),
111.
Yeong, W.-Y., Chua, C.-K., Leong, K.-F. &Chandrasekaran, M.
(2004) Rapid prototyping in tissue engineering: challenges and potential. Trends in Biotechnology, 22(12),
643652.
Yeong, W. Y., Chua, C. K., Leong, K. F., Chandrasekaran, M.
& Lee, M. W. (2006) Indirect fabrication of collagen scaffold based on inkjet printing technique. Rapid Prototyping
Journal, 12(4), 229237.
M. & Lee, M. W. (2007) Comparison of drying methods
in the fabrication of collagen scaffold via indirect rapid
prototyping. Journal of Biomedical Materials Research
Part B: Applied Biomaterials, 82B(1), 260266.
Yeong, W. Y., Sudarmadji, N., Yu, H. Y., Chua, C. K., Leong,
K. F., Venkatraman, S. S., Boey, Y. C. F. & Tan, L. P.
(2010) Porous polycaprolactone scaffold for cardiac tissue
engineering fabricated by selective laser sintering. Acta
Zeng, W., Lin, F., Shi, T., Zhang, R., Nian, Y., Ruan, J. &
Zhou, T. (2008) Fused deposition modelling of an auricle framework for microtia reconstruction based on CT
images. Rapid Prototyping Journal, 14(5), 280284
49

Surgical training and post-surgery evaluation using rapid

prototyped biomodels
Lus Queijo & Joo Rocha
Instituto Politcnico de Bragana, Bragana, Portugal
Paulo Miguel Pereira

Servio de Neurocirurgia do Hospital de S. Joo, Porto, Portugal
Manuel San Juan

ESTII Universidad de Valladolid; CIBER Centro de Investigacin Biomecnica y Ergonoma, Valladolid, Spain
ABSTRACT: The biomedical use of Rapid Prototyping Technologies (RP) had great developments in the last
years, especially as supportive tools for tissue growth, direct or supportive technology for implant fabrication
or as tool for personalized biomodels production applied to studies, this research will focus on this last type of
usage in continuation of previous work developed with RP as an aid of surgery procedures. Biomodels can play
an important role as a complementary diagnostic method to medical staff (Queijo et al. 2010). The usage of RP
technologies for biomodels production, in Lytic Spondylolisthesis surgical training and as a tool for post-surgery
evaluation, is presented in this paper.
1
INTRODUCTION
thinner part of the posterior vertebral arch and hence,

the least resistant to fatigue.
Accordingly, stress fracture of the isthmus is a common occurrence among young active adults and particularly, among participants in some sports like diving,
swimming, weightlifting, gymnastics and running.
With this fracture the vertebral body, pedicle and
superior articular processes become separated from
the inferior articular processes and hence from the vertebra below. This condition creates the possibility of
slippage between the vertebrae.
The term spondylolisthesis is used to identify the
anterior translation in the sagittal plane of a vertebra
(and the spine above it) relative to the vertebra below.
There are several aetiologies but the lytic or isthmic
type, as described above, is the most common. In a
lytic spondylolisthesis there is a bilateral defect of the
isthmus (pars interarticularis).
The slippage between the vertebrae can cause the
exiting nerve roots (the nerves exiting the spinal canal
at this level, through the intervertebral foramina) to be
squeezed causing leg pain and difficulty in walking.
When this occurs, a surgical treatment may be necessary to decompress the nerves and stabilize the spinal
segment (to avoid further slippage) with or without
reduction of the deformity.
3D replicas of vertebral spine sections are useful in

diagnosing, planning and surgery simulation, visualization and manipulation. To patients it is important,
allowing them to understand their pathologies nature,
surgical proceedings performed by surgeon (Madrazo
et al. 2008) as well to reduce anxiety facing surgery
need (Queijo et al. 2010). In this study it will be
presented a procedure to manufacture the needed
biomodels to fulfill these requirements. These allowed
a better support to surgery practice by a previous
analysis to the patient condition in the form of a
segmented biomodel and the surgery planning by a
biomodel representing the corrected spine section and
the medical devices applied.
All physical biomodels were constructed based
in the 3D digital models reconstructed from patient
CT (Computerized Tomography) scans, where image
segmentation techniques were performed.
The manufacturing technique used was Three
Dimensional Printing (TDP) once this technique can
provide the needed resolution and surface finishing
needed to proper visualization and manipulations,
associated to a manufacturing low cost.
1.1 Lytic spondylolisthesis
1.2 From CT images to 3D digital models
With bipedal posture most of the loads transmitted to

the lumbar spine pass through the posterior elements of
the vertebrae. The isthmus (pars interarticularis) is the
The fabrication ability of 3D physical models depends

exclusively of the existence, in proper format, of a
3D digital model. This digital model, used as .stl type
51
file can have several proveniences, most commonly

segmentation image or CAD software. As this last
one allows the manipulation of 3D digital models by
adding, subtracting or changing some of his features,
the first ones are essential to perform image reconstruction and obtain the digital models from 2D image
data files.
Most of these images come from CT or MRI scans
in the form of cross sectional images from the study
areas, according human axis (axial, coronal and sagittal) and obeying to the international standards in
Digital Imaging and Communications in Medicine
DICOM.
Models quality is directly connected with the interval between images that is also a compromising solution between object structure size and the amount of
radiation to which the patient is subjected.
3D digital models are obtained recurring to image
segmentation techniques where several operations are
performed in order to distinguish the main object
bone structures, in this case, from the main tissues by
the application of masks according with the gray gradient in Hounsfield scale (HU). These masks allow
a posterior rendering process that reconstructs the
structure in a tridimensional image.
1.3
Figure 1. ZCorps TDP printer in action.
This process allows biomodels fabrication with good

visualization characteristics, manipulation possibility
and with a cost that is substantially lower than other
alternate methods.
ZCorps TDP technique uses a composite powder
as building material bonded by an aqueous media that
is jetted by a printing head, similar to any 2D printer,
as can be seen in figure 1. Once the models are built
some stability must be provided to the surfaces (that
are pulverous) in order to give rigidity and manipulation ability. This stabilization is, usually fulfilled by
the impregnation with another bonding media such as
cyanoacrylate.
Rapid prototyping three dimensional printing
The logical following step after the 3D digital model

rendering is the fabrication of the physical biomodel
that will allow a better visualization allied to his manipulation. The technology applied to this fabrication
is rapid prototyping, often called additive manufacturing or fabrication that represent a new group of
non-conventional techniques introduced in the medical field and provide high reproducibility and elevated
capacity to quickly produce very complex 3D shapes
(Gibson 2006) (Brtolo et al. 2009).
Once 3D digital model is obtained, his form is
exported in the format of a .stl file where all the surfaces are converted into a triangle mesh allowing to be
imported by any 3D print manager software. Next step
is a hidden process where the print manager divides the
model contained in the .stl file in several slices that are
re-encoded and will constitute another file type that is
sent to the printera .sli file.
Manufacturing processes available today, as
Fused Deposition Modeling (FDM), Stereolithography (SLA), Selective Laser Sintering (SLS), Tridimensional Printing (TDP or 3DP) and Laminated
Object Manufacturing (LOM) among other specific
processes, bring a wide range of choice over building
materials and outcome biomodel mechanical characteristics, as well as costs. A more detailed description
over each one of this processes and over biomodeling
process can be found in previous work (Queijo et al.
2010).
Despite several rapid prototyping techniques available, each one with his own characteristics, the chosen
to produce these biomodels have been three dimensional printing (TDP) in the variation from ZCorp.
1.4 Medical case description

The images belong to a 30 year-old male complaining
of bilateral pain along the inferior limbs progressing
over the past 5 years.
The imaging studies yielded a bilateral L5 isthmic
lysis and an anterior slip of L5 over S1 of about 50%
of the superior endplate of S1 (lytic spondylolisthesis). Moreover, there was a small posterior slip of L4
over L5. Together, these two slips result in an anterior
translation of the vertebral body of L5 compared to L4
and S1.
In face of a progressive clinical course, a surgical treatment was proposed to the patient in order to
decompress the nerve roots and to fix the vertebrae,
avoiding further slips among them.
During the surgery the facet joints L4L5 and L5
S1 were removed unilaterally and the discs L4L5 and
L5S1 were excised as well. Multiaxial pedicle screws
in titanium alloy (MAST Legacy , Medtronic Inc,
Minneapolis, MN, USA) were inserted bilaterally in
the pedicles of L4, L5 and S1 and connected to
two longitudinal 5.5 mm diameter rods using a percutaneous system (Sextant Reduction , Medtronic
Inc.). PEEK cages (Capstone , Medtronic Inc.) were
inserted in L4L5 and L5S1 disc spaces and were
filled with local bone. Extra bone chips were inserted
in the disc spaces around the cages to enhance interbody fusion rate. Translational torque was applied to
52
Figure 3. Image segmentation Mask processing and 3D

previewing.
Figure 2. Image segmentation Rendering without mask
processing.
the L5 screws through the Sextant Reduction system in order to try to achieve a better sagittal alignment
of the vertebrae.
In the postoperative CT scan we found an appropriate placement of the implants, but a very limited
correction of the deformity.
METHODOLOGY
Along methodology description, each biomodel production phase is intercalated with the facts related in
medical description. For better contextualization it was
decided to divide themes.
Ethically, patient should provide authorization for
TC images to be used and these should be, as soon
as the process allow, made anonymous (once TC systems register patients information). In this process, 2D
images are imported to segmentation software where
will be treated to rebuild the desired spinal area.
2.1
Figure 4. Pedicle cut-sections 3D digital model.
is carried out, as seen in figure 3. Unwanted structures

as CT table and iliac bone showed in figure 2 have
been removed once those are irrelevant to the study. In
this process and to allow a better and individualized
visualization we have decide to isolate each one of the
structures with an individual mask.
After obtained the reconstructed 3D digital model
it is already visible the constriction in L5S1 conjugation holes as well as the front slippage of L5 vertebra.
This way, it was decided to produce a biomodel with
pedicle cut sections where the constriction and the
slippage could be better observed. Digital model have
been manipulated by orthogonally cutting all masks in
L5 pedicle regions. The result can be seen in figure 4
where critical areas are marked.
Once 3D digital model is fully defined it has been
exported as a .stl file and manipulated by 3D print
manager.
Image segmentation
First of phases in the procedure consists in defining gray values interval corresponding to HU units in
Hounsfield scale to isolate the maximum of our object
of work the section of spine, including L3, L4 and L5
vertebrae and most of sacrum. As first iteration, have
been adopted a range of values defined as standard to
bone tissue and set between 226 and 1196 HU. This
process created a mask that has covered each image
pixel which value is included in the chosen range and
painted it in a selected color. Rendering process based
on the created mask showed us that despite the chosen
range is close to the desired, once has rejected most
of the unwanted surrounding tissues there still having some adjustments to be made before following to
the next step figure 2. Range HU values where then
settled between 200 and 1196.
The following step consisted in rectifying each
structure, frame by frame until a coherent digital model
2.2 Evaluation biomodel

In 3D print manager, as long as get an estimated building time and materials consumption, it is possible to
define model orientations over building area.
53
Figure 7. Surgical planning 3D digital model.
Figure 5. Evaluation biomodel. A conjugation hole constriction; B L4L5 and L5S1 slippage.
Figure 8. Medical device positioning validation biomodel.
Figure 6. Multiaxial pedicle screw, PEEK cages and longitudinal rod: A real devices; B 3D modeled devices.
Figure 9. Validation biomodel details: A critical nerve

root areas; B PEEK cages positioning; C Possible facet
removal area; D Pedicle screw orientation and positioning.
Once settled building parameters, biomodel fabrication is done, layer by layer, followed by surfaces stabilization as final procedure, as described in previous
work (Queijo et al. 2010).
Through physical biomodel, shown in figure 5, has
been visualized patient condition in the constricted
conjunction holes and vertebra slippage and evaluated
possible pedicle vertebral screw-bars insertion points
and angulations that would bring L5 vertebra to a most
favorable position.
Also, this model has been shown to the patient to
explain him the nature of his pathology and the need
for a surgery where the lower vertebrae would be fixed
with medical devices, represented in figure 6.
2.3
3D CAD software (figure 6) and exported to .stl files,

all bone structures are repositioned.
As defined previously, vertebrae repositioning is
done to allow a bigger clearance in the conjugation
holes and a reduced slippage between L4L5 and L5
S1 and, then, matched with medical devices resulting
in a complete 3D digital model as sown in figure 7 that
have been built for a new evaluation and positioning
validation (figure 8).
In surgical planning biomodel it is possible to
see, after repositioning, the interference between bone
structures that will conduct to possible bone removal,
the detailed pedicle screw positioning and orientation
and the need in accessibility to position intervertebral
cages, as can be seen in details from figure 9.
Surgical planning biomodel
After those definitions, with previously CAD modeled

pedicle screws, PEEK cages and longitudinal rods, in
54
Figure 12. Biomodel comparison front view: pre-surgery

(left) and post-surgery (right). A PEEK Cage placement
detail.
Figure 10. Post-surgery biomodel.
Figure 13. Biomodel comparison left view: pre-surgery

(left) and post-surgery (right). A pedicle screw positioning;
B conjugation hole clearance.
allowing not only a complementary diagnose method

for 3D visualization of complex areas but also valuable
tools for surgical applications.
In this case, it has been possible to show evaluation biomodel to the patient and explain, supported by
a simplified form of visualization, the cause of pain
and the reason of surgery need. Patient becomes aware
that the only way to relief long time pain was fixing
vertebrae to avoid a continuous degeneration of his
condition.
Another advantage of biomodels is their uses as
a complimentary diagnose allowing medical staff to
observe medical images in a 3D way.
Patient condition was properly defined through
evaluation biomodel that enabled to identify the
amount of slippage between L4-L5 and L5-S1 vertebrae as well the compression in the nerve roots caused
by this slippage that constricted conjugation holes. It
then becomes clear that it would be difficult to bring
L5 vertebra close to a satisfactory position.
Through positioning biomodel were defined adequate insertion points and the needed orientation to
pedicle screws. Also, in this case, were defined which
bone structures needed to be removed to allow proper
cage insertion as well the needed vertebra repositioning. It becomes more or less clear that nerve root
decompression would be achieved not all by vertebrae repositioning but also by bone structures removal
(right L5 pedicle and facet) as can be seen in figure 9
detail C.
Post-surgery biomodel allowed a comparison
between what was planned and the achieved result
(figures 12 and 13). It was confirmed that possible vertebrae repositioning was limited, allowing a
Figure 11. Post-surgery biomodel details: A PEEK cage

fixing L4L5 vertebrae; B Facet and partial pedicle removal
to allow cage positioning; C Pedicle screw implants with
longitudinal rods.
2.4
Post-surgery biomodel
Post-surgery biomodel (figure 10) is obtained following the same procedure as for evaluation biomodel.
Based in patient post-surgery TC images, 3D digital
model has been reconstructed with the particularity of
being needed two distinct range in Hounsfield scale
one for the bone, as settled previously and another
to metallic medical devices settled in 14002976 HU.
Reconstruction, in this case, become harder once there
is a considerable presence of noise in TC images, due
to the ray dispersion in presence of metallic devices.
Once combined the two masks generated for bone
and for metal, the result is a 3D digital model that
is exported to a .stl file and then fabricated. This
biomodel allow a comparison between what has been
planned and what was achieved in the surgery due
to all conditionings. Figure 11 shows the details in
post-surgery biomodel.
CONCLUSIONS AND FURTHER WORK
As previously demonstrated, through rapid prototyping

technologies and a multidisciplinary collaboration it is
possible, in a short period of time, to build biomodels
55
REFERENCES
medium 3 mm increase in axial vertebrae distance

but that was granted the needed conjugation holes
clearance to fully decompress nerve roots.
Further work will focus in the evaluation and
surgery planning biomodels in order to make them as
functional as possible. If succeeded, this will allow the
inclusion of real medical devices directly in biomodels as a simplified form of simulating surgery phase
of medical devices placement.
This type of surgery becomes frequently dependent
on surgeons decisions and unknown factors occurring
during surgery, so there is a need to investigate further
cases with personalized manufacturing tools to allow
proper pedicle screw placement and orientation, also
known as surgical guide.
Brtolo, P. J. S., Almeida, H. & Laoui, T. 2009. Rapid prototyping and manufacturing for tissue engineering scaffolds.
Int. J. Comput. Appl. Technol., 36, 19.
Gibson, I. 2006. Rapid prototyping: from product development to medicine and beyond. Virtual and Physical
Prototyping, 1, 3142.
Madrazo, I., et al. 2008. Stereolithography in spine pathology:
a 2-case report. Surgical Neurology.
Queijo, L. et al. 2009. A prototipagem rpida na modelao
de patogenias. 3. Congresso Nacional de Biomecnica.
Bragana. Portugal.
Queijo, L. et al. 2010. A surgical training model manufacture
using rapid prototyping technology. Innovative Developments in Design and Manufacturing-Advanced Research
in Virtual and Rapid Prototyping, 175179.
56

The calibration of continuous Digital Light Processing (cDLP)

for the highly accurate additive manufacturing of
tissue engineered bone scaffolds
D. Dean
Departments (Dept.) of Neurological Surgery, Case Western Reserve University, Cleveland, OH, US
J. Wallace
Biomedical Engineering, Case Western Reserve University, Cleveland, OH, US
A. Siblani
envisionTEC Inc. Ferndale, Michigan, US
M.O. Wang
Fischell Dept. of Bioengineering, Jeong H. Kim Engineering Bldg (JHKEB), University of Maryland,
College Park, MD, US
K. Kim
Department of Chemical and Biomolecular Engineering, JHKEB, University of Maryland,
A.G. Mikos
Department of Bioengineering, Rice University, Houston, TX, US
J.P. Fisher
Fischell Dept. of Bioengineering, Jeong H. Kim Engineering Bldg (JHKEB), University of Maryland,
ABSTRACT: It is helpful if the rendering of both the external and internal geometry of bone tissue engineering scaffolds is highly accurate. The external geometry must accurately fit the defect site if the scaffold is
to be incorporated by the host tissue. It may also be useful to load internal pore spaces with cells and growth
factors prior to implantation. Optimal pore space size has been judged to be between 200 and 1600 microns.
Continuous Digital Light processing (cDLP) is one of the most promising methods for the highly accurate rendering of tissue engineering scaffolds using biocompatible dye-initiator packages and resorbable polymers. The
high accuracy of cDLP scaffold rendering results in part from two technical developments. The first technical
achievement reported here is the integration of DLP (Texas Instruments, Dallas, TX) technology using a Digital Micromirror Device (DMD) chip into an additive manufacturing device, such as the Perfactory SXGA+
Standard UV device (envisionTEC, Ferndale, MI). The Perfactory UV device is capable of continuously polymerizing 35.5 35.5 50 m voxels. The second technical achievement reported here is the development of
a biocompatible dye-initiator package for the rendering of resorbable polymer tissue engineering scaffolds. A
dye is used to block light, thereby limiting the depth of polymerization. In this study we demonstrate the fabrication of scaffolds from the well-studied resorbable polymer, poly(propylene fumarate) (PPF). We have used a
Perfactory UV device to render porous cylindrical PPF scaffolds with a diameter of 6 mm and a length of either
1.2 mm (N = 10) or 12.4 mm (N = 8) with either 2 or 4 minute exposure using a plate and post geometry. Our
Computer Aided Design for this scaffold is rendered on the Perfactory UV via 50 m thick layers. We used a
120 m curing depth to ensure sufficient overcuring (inter-layer binding). A yellow azo chromium or titanium
dioxide (TiO2 ) dye, Irgacure 819 (BASF [Ciba], Florham Park, NJ) initiator, and diethyl fumarate solvent
were added to the primary material, PPF, and used for scaffold production. A 500-195-20 Mitutoyo (Aurora,
IL) caliper was used to measure scaffold features. The 12.4 mm long azo chromium scaffolds were micro-CT
(CT) scanned. The 1.2 mm long scaffolds were imaged via scanning electron microscope (SEM). We found that
qualitative analysis of these CT images presented anisotropic but predictable shrinkage. Qualitative analysis
of SEM images presented thinning at layer margins. The 1.2 mm azo chromium scaffolds presented an average
observed post diameter (expected 0.4 mm) of 0.43 mm (0.02 std dev) and an average observed plate diameter
(expected 0.6 mm) of 0.63 mm (0.01 std dev). The 12.4 mm azo chromium, 4 minute exposure, scaffold group
presented an average diameter (expected 6 mm) of 6.03 mm (0.03 std dev). The 12.4 mm TiO2 average diameter
was 5.92 mm (0.07 std dev). Accurate calibration of overcuring ensures interlayer binding and full formation of
the smallest, 400 m in this study, scaffold features.
57
1
1.1
COMPUTER AIDED DESIGN OF BONE

TISSUE ENGINEERING SCAFFOLDS
to be extremely accurate. All of these issues may

reduce resolution seen in the z direction or in directions
oblique to the x, y, and z directions.
Shape design for different scale spaces

1.2 Functional design for different scale spaces
There is a great deal of interest in the use of additive

manufacturing [1] technology to render fully or partially resorptive implants to fit patient-specific defect
sites. The focus of this paper is bone tissue engineering, with the particular application being cranial
repair. However, the issues that are discussed have
implications for a wide variety of tissues.
The current standard-of-care is to use nonresorbable materials for intra- or pre-operatively prepared cranial implants. Often the external shape
of pre-operatively manufactured cranial implants is
determined via 3D CT-scan [2]. High accuracy (e.g.,
200500 m) is useful when rendering these implants
as they must fit tightly against the remaining skull to
protect against trauma without compressing the adjacent brain and scalp. While this level of computer
aided design (CAD) and computer aided manufacturing (CAM) accuracy is impressive, even higher levels
of accuracy have been sought for tissue engineering
applications.
Current studies have shown that internal porous
space in resorbable tissue engineering implants is
desirable, as it provides more surface area for seeded
cells, growth factors, and host tissue and vasculature. Given that the effective diameter for pore spaces
and/or vascular channels is likely to range between
2001600 m [3], it is useful if scaffolds can be rendered with higher accuracy (e.g., 25200 m) than
is needed for current non-resorbable implants. The
computer aided design of these pore spaces can be
optimized for cell and growth factor loading as well as
nutrient perfusion during pre-implantation culturing in
a bioreactor [4].
Scaffold accuracy may be challenged by the need
for supports when using an additive manufacturing
device to render tissue engineering scaffolds. It is not
uncommon for a scaffold design to include an isolated
(i.e., unconnected to the overall implant) feature in a
layer that is expected to join the overall object when
a subsequent layer is rendered. Snap-off supports are
often used to hold these isolated pieces in place until
they join the overall implant. However, the need for
supports within an internal pore or vascular channel
might disrupt the overall geometry.
Another challenge to manufacturing accuracy may
be anisotropy. It is common for additive manufacturing device resolution to differ between the x and
y directions versus the z direction (i.e., anisotropy).
This commonly occurs when controlling layer thickness, the z direction, is more difficult than controlling
lateral feature dimensions, the x and y directions.
Z-resolution may be affected by minimum depth of
polymerization energy (e.g., light-based systems) or
material melting (e.g., devices based on sintering technology), both of which may affect inter-layer binding,
sometimes referred to as stitching. Additionally, as
layers are built, the build-plate stepper motor needs
Most methods used to add functional features to scaffolds require an even higher level of resolution than is
required for external surface, pore, or vascular channel
geometry. These functional features may be designed
to affect cell attachment, proliferation, or maturation.
This includes surface features that interact directly
with cells such as scaffold stiffness, roughness, or
hydrophilicity. These features may relate to directly
rendered surface morphologies or the inclusion of
particles or biological materials that range in size from
nanometers to 10s of microns. It is also possible to
determine some functional features by adjusting the
chemistry (e.g., see Section 4) of the liquid resin from
which scaffolds are polymerized.
Other functional scaffold features are created
through the incorporation of particulates such as
tricalcium phosphate crystals, cell-specific ligands,
antibiotics that are suspended in the resin during polymerization, or a coating or other treatment of the scaffold surface applied after polymerization [5, 6]. It may
be possible to modify hydrophobic scaffold surfaces to
promote/encourage cell attachment either by radio frequency glow discharge (Harrick Plasma, Ithaca, NY)
or protein adsorption [7]. Protein adsorption can be
brought about by pre-soaking scaffolds in serum.
The mechanical function of scaffolds can be controlled by adjusting scaffold geometry or material
properties at different scale spaces. The design of vascular and cell, and/or growth factor loading/coating,
channels can also be optimized to allow for shear
stimulation of cells via flow or the application of
compressive, tensile, or shear forces [8]. Similarly the
thickness of walls, supports, or other scaffold features
can be used to guide tissue formation and/or resist
bioreactor or in vivo loading requirements. Depending
on the mechanism of resorption, geometric, mechanical, and/or chemical modifications to the scaffold can
be used to influence the rate and timing of scaffold
degradation. Higher molecular weight polymers will
offer more strength but may also prevent the degradation necessary for resorption, which in turn may be
necessary for tissue remodeling and/or repair. Since
the solvent, DEF monomer, is incorporated into the
scaffold in our work, it changes the cross-linking mesh
structure and reduces rigidity.
2
2.1
POLYMERS FOR ADDITIVE

MANUFACTURING OF SCAFFOLDS
Solid polymers for additive manufacturing
technologies
Additive manufacturing strategies that use heat to

produce scaffolds from solid polymer filaments or
powders have been reviewed by Peltola et al [9]. They
58
and N-vinyl-2-pyrrolidone [11]. Resorption kinetics and toxicity will likely limit the list of lightpolymerizable polymers that can be used for rendering
tissue engineering scaffolds.
The most obvious difference between SLA and
cDLP is the hardware (Figure 1). For SLA a part
descends into the resin whereas it is pulled out the
resin in a cDLP device. The SLA process is gentler on
the forming implant than the cDLP process because
the part must attach much more firmly to the build
platform that is lifted from the basement plate after
every layer is formed. However SLA requires that the
resin surface be uniformly flat before the exposure of
each new layer, often requiring the use of a wiper or
recoating blade to mechanically smooth the surface. In
comparison, in cDLP a new layer of resin must form
underneath the scaffold on the light transmitting basement plate after each layer is rendered. The current
device allows resin to spread over the basement plate
passively.
Another difference between SLA and cDLP is in the
users ability to set the level of curing during rendering. The overall strategy with the SLA is to draw parts
in relatively high molecular weight resins as quickly
as possible so that they can be moved to a post-curing
device (e.g., a UV bath). Because we have used a dye
to limit the depth of polymerization in the cDLP, we
have had the option of using higher levels of irradiance without losing resolution in the z direction. The
use of higher levels of energy deposition may increase
scaffold green strength (i.e., strength immediately
after rendering but before post-curing). In general,
increased green strength will make it easier to handle
and, more specifically, to mechanically clean unpolymerized resin from the scaffold pore space prior to
post-curing [10].
Figure 1. Stereolithography and cDLP based systems both

rely on photocrosslinking for freeform fabrication. However,
the methods differ as illustrated here. Stereolithography typically requires a deep vat of resin. As parts are built, they attach
to an elevator which moves downward through the polymer
resin as each layer is rendered at the surface by a moving
laser. In contrast, cDLP systems render parts by projecting
an image through a clear basement containing the resin, curing at the bottom surface rather than the top surface. The parts
attach to a build platform which moves upward, away from
the basement, after each projection.
discuss the advantages and disadvantages of the available methods and note their use and importance in
the creation of surface roughness. However, they also
note that scaffolds with smooth surfaces can be modified or coated after rendering. They note that the
highest resolution modalities are SLA (stereolithography) (Figure 1) and Two Photon Polymerization (TPP).
However, apparently, in regards to SLA they combine
the technology provided by 3D Systems (Rock Hill,
SC) with continuous Digital Light Processing (cDLP)
technology (envisionTEC, Gladbeck, Germany) under
a single SLA rubric.
2.2
Liquid polymers for SLA and cDLP

2.3
Both SLA and cDLP require light polymerizable polymers. As additive manufacturing technologies, both
are very different than the use of translucent molds
to cure parts in a UV bath, in that the polymerization reaction must be initiated in a small, temporarily
irradiated region, and by the energy emitted from the
available light source. However, the photochemistry of
this time- and space-delimited reaction can be similar
or the same in both devices.
The UV laser in 3D Systems SLA devices generates roughly 6 orders of magnitude higher power than
the lamp-based system in envisionTEC cDLP devices.
SLA devices are most commonly used to render parts
quickly and then post-cure them in a UV bath. On a
layer to layer basis, a much longer exposure time is
used in cDLP devices. Layer rendering time in our
application is seconds in the former and minutes in
the latter. Total energy deposited with either device is
roughly similar.
The production of tissue engineering scaffolds
via SLA or cDLP has focused on resorbable polymers. The polymers studied to date include PPF
[10] and fumaric acid monoethyl ester (FAME) endfunctionalized poly(D,L-lactide) (PDLLA) oligomers
System for use of poly(propylene fumarate)
The resin for SLA or cDLP that we have used to render

scaffolds contains between a 1:1 and a 2:1 weight ratio
of PPF to DEF and between 0.52.0 wt% photoinitiator. In all cases the photoinitiator used in our studies
has been bis(2,4,6-trimethylbenzoyl)phenylphosphine
oxide (BAPO) (Ciba Specialty Chemicals, Tarrytown,
NY). Additionally, we have used a dye for cDLP
rendering (see Section 2.5).
2.4
Role of solvent
There are scaffold rendering issues that SLA and cDLP

share and those where they differ. Both processes
are susceptible to issues when working with a highly
viscous polymer. In the SLA device, highly viscous
polymer may cause mounding if the wiper blade is not
used, or is unsuccessful in creating a flat resin surface
of the required thickness.
In cDLP a missed layer will result if insufficient
polymer recoats the basement plate or there are bubbles in that layer. We have observed that when the resin
used for cDLP is too viscous, a vacuum-like seal forms
between the scaffold and the basement plate. Viscous
59
layer of calibrated thickness. In most cases this will be

accomplished by light absorption. Thus the dye will
compete with the initiator for photons. With one dye
that we have studied, titanium dioxide (TiO2 ), light
is primarily scattered rather than absorbed. Since the
light is not only reflected directly back towards the
light source, there is a possibility that there will be
polymerization near the scaffold in areas which are
not intentionally or directly exposed. Indeed, we have
observed this, and refer to it as lateral overcuring
(see Section 5).
We are in the process of determining how much dye
is incorporated into a scaffold and lost with each build.
From our current observations it appears that it is less
than 1:1 with polymer, solvent, or initiator. This was
evidenced by the finding that when using a chromium
azo dye (see Section 3.2), the depth of polymerization
slowly decreased for a fixed exposure level each time
a scaffold build process was completed, necessitating
frequent recalibration. This would suggest that the dye
concentration was gradually increasing as a result of
unbalanced incorporation of resin components into the
cured scaffolds. This phenomenon has not yet been
observed for resin containing TiO2 as the dye. Since
the dye is incorporated it will determine the color of
the implants. A critical property of the dye is its ability
to stay suspended throughout the rendering process. It
may be necessary to stop the process and re-stir the
resin if the dye is settling out.
Figure 2. The basic steps necessary in the calibration of

a cDLP system are shown here. The blue arrows indicate a
logical calibration order, while the dashed arrow indicates
that Steps 5 and 6 may feed back into subsequent iterations
of the calibration loop.
resins may require a longer pause between layers, as

more time is required for the flow into void spaces left
in the areas where the previous layer was cured.
Use of a solvent can alleviate these issues by reducing the resins viscosity. We have used the monomer
precursor, diethyl fumarate (DEF), to poly(propylene
fumarate) (PPF) in our work. This monomer crosslinks into the scaffold, but once cross-linked poses little
to no toxicity risk.
3.2 Secondary properties of dye

2.5
Role of dye-initiator package
Our initial work was with a yellow azo chromium

dye, which may be toxic and/or carcinogenic [16]. As
expected, it provided excellent light-blocking properties. Although we did not measure the amounts that
were incorporated, we did observe that it colored the
resulting scaffolds a dark orange to red color. Similarly, we did not conduct rheological studies; however,
we found that this dye started to come out of suspension within one day, suggesting that it might be
beneficial to stir the resin during long scaffold rendering sessions. In our experiments we found that dye
concentrations between 15% were needed to reduce
single PPF layer thickness to the desired value of
120 m. This thickness would provide 70 mm of overcuring which would be useful for binding 50 m thick
layers.
Since the dye used in a dye-initiator package is
likely to be incorporated into the scaffold, it may be
helpful to utilize dyes that can also positively influence scaffold surface roughness, act as antibiotics,
or positively affect the scaffold degradation environment (e.g., buffer the pH if it would otherwise be
too acidic or basic). We have experimented with three
dyes that would be both biocompatible and, we expect,
would have beneficial properties during cell attachment, tissue incorporation, and/or scaffold resorption
(Figure 3). Two of these dyes, doxycycline hyclate and
amphotericin B, have antibiotic properties. We were
unable to reach our goal of a 120 m thick layer of
The relationship between depth of polymerization

and exposure in SLA is determined for each resin
via a calibration procedure referred to as rendering
WINDOWPANESTM (3D Systems) [10]. Once an
acceptable Z resolution is obtained, the calibration
is set for all cases. As noted in Section 2.2, the user
can set the level of irradiance in cDLP. By varying
irradiance versus initiator and dye concentration, one
can determine curing depth. Rather than a single setting, curing depth can be experimented with until an
optimal overcure (i.e., stitching between layers)
is found relative to layer thickness (Figure 2). The
use of dye-initiator packages has been a constant in
industry [12], where toxicity issues for internal use are
less prevalent. However, we expect that this topic will
receive renewed interest in light of the need for biocompatible dyes and initiators for use in cDLP-based,
and possibly with SLA-based, applications of additive manufacturing technologies for tissue engineering
[1315].
3
3.1
DYE-INITIATOR PACKAGES FOR ADDITIVE

MANUFACTURING TECHNOLOGIES
Primary properties of dye
The primary function of the dye is to block light so

that as much energy as is useful can be deposited in a
60
Figure 4. The plate-and-post test scaffold design shown

from A) isometric, B) front, and C) top viewpoints. The following dimensions characterize the geometry of this design:
plate thickness = 400 m; distance between plates (or post
height) = 800 m; vertical circular pore diameter = 800 m;
post diameter = 600 m; overall scaffold diameter = 6 mm;
overall scaffold height = 12.4 mm. The height-to-diameter
ratio of the scaffold approximately 2:1, which is useful for
mechanical compression testing applications. The small size
of the test scaffold also lends itself to small animal model
testing.
Figure 3. The relationship between cure depth (m) and

concentration (wt%) for three biocompatible dyes (amphotericin B, doxycycline hyclate, and rutile titanium dioxide).
The data for an effective, but toxic, yellow azo chromium dye
is also provided for comparison. Doxycycline hyclate proved
ineffective for this application, while amphotericin B and titanium dioxide proved to be promising candidates for further
calibration studies (see Figure 6 for a deeper investigation of
titanium dioxide). The following parameters were held constant throughout these tests: BAPO concentration = 0.5wt%;
irradiance = 200 mW/dm2 , exposure time = 300 s. Data represents mean standard deviation (n = 3).
in general, and potentially reducing the need for postrendering photo-curing.

3.4 Role of solvent
PPF even when using very high concentrations

of doxycycline hyclate and initiator. Amphotericin
proved more effective, but the relatively high concentration needed gave us concern about the possibility of
toxicity, so other dyes were evaluated.
The third dye evaluated, titanium dioxide, is widely
available in food- or pharmacologically-safe, highly
pure, formulations. Small particles of TiO2 are thought
to be easily metabolized. Since it is used as a food
or drug coloring agent, its optical properties are well
understood. It blocks light more by scattering rather
than by absorbing photons. There are different configurations of TiO2 crystals. The rutile form with
250305 nm sized crystals has very useful optical
properties for UV blocking.
3.3
The solvent DEF is used primarily to reduce viscosity. As with the dye, the solvent is incorporated into
the resulting scaffold. It also has an effect on PPF
crosslinking, tending to reduce scaffold rigidity as
more is used. Thus, without a dye present we have
determined that the mechanical properties of PPF were
improved by increasing the DEF concentration due to
an increase in crosslinking density. However, beyond
a 75:25 PPF:DEF ratio those material properties start
to degrade due to the increased distance between PPF
chains during the crosslinking reaction [18].
4
CALIBRATION OF cDLP-BASED ADDITIVE

MANUFACTURING
Our goal in the calibration study was to render scaffolds with the plate and post geometry (Figure 4) [4].
These cylindrical test scaffolds are 6.0 mm in diameter
12.4 mm in length. The diameter of the large vertical channels seen in Figure 4 is 800 m. The plates
are 400 m thick and 800 m apart from each other.
The posts running between the plates, are 600 m
in diameter.
Properties of initiator
We have a long experience with BAPO for UV-based

photo-crosslinking of PPF for use in tissue engineering
applications [17]. Most of our work using translucent
molds or in stereolithography [10] has utilized 0.5
1.0% BAPO with no dye present. Upon introducing
a dye we found that these levels of BAPO were too
low to achieve sufficient green strength and that it was
useful to increase the level to between 1.02.0%. With
the azo chromium and amphotericin dyes, one could
determine the maximum amount of dye and initiator
needed so that higher light exposures did not increase
layer thickness. Being able to infuse more light allowed
us to increase green strength, increasing the success
rate of layer stitching and the manufacturing process
4.1
Calibration of a cDLP
Calibration of a cDLP process consists of at least six

steps (Figure 2). The first step in the calibration procedure is to polymerize single layers of the cDLP resin,
i.e., PPF, DEF, BAPO, and the dye (Figure 2, Step 1).
There are at least three variables to study, i.e., dye
61
rendering. If this susceptibility to mechanical force

during scaffold rendering becomes an issue, it may be
possible to counteract those forces with an increased
number and density of support structures.
The fifth step is to render a multi-layer scaffold.
Enough of the scaffold should be rendered to determine if a larger, more useful scaffold, can be prepared.
However, it is likely that a useful test part can be rendered in 4-6 hours. Not only can one determine layer
binding with a test part, but also the resolution (accuracy) of scaffold geometry relative to build parameters
can be tested. For example, how much overcuring is
useful to ensure that there is no failed lamination, or
subsequent delamination, may be determined. Another
question is how accurately are the varying geometries
of desired features rendered, especially in directions
oblique to the build plane (x and y) and inter-build
plane (z) directions. Finally, one can look at how software tools, such as anti-aliasing, affect the accuracy of
the shape and dimensions of designed features.
The sixth and final step is to test the scaffolds both
in vitro and in vivo. As mentioned, in vitro testing
could include mechanical tests, biological environments without cells or tissues, and biological environments with cells, growth factors, and/or tissues. It
is possible, indeed likely, that the results of this step,
as well as the fifth step, will feed back into further
optimizing of the first four steps.
concentration, initiator concentration, and irradiance

duration. Other factors that could be varied would
be polymer molecular weight and polydispersity as
well as irradiance level (i.e., the amount and rate at
which light is applied). It is not yet clear whether these
variables will always interact in a linear fashion. For
example, it has been reported that TiO2 can act as a
catalyst [19]. The goal in using TiO2 is to have a layer
thickness that ensures adequate overcuring between
layers, yet is thin enough to allow for a desired z
step size and the generation of accurate geometries.
Resolution in x, y, and z will determine the accuracy
of desired external scaffold, and internal pore, surface
morphology.
The second step is to ensure that the material properties of the chosen resin configuration will provide
useful scaffolds. In some cases scaffolds will be loaded
with cells and/or growth factors and immediately
implanted. In other cases scaffolds will be pre-cultured
(e.g., in a bioreactor) prior to implantation. It may
be useful to undertake mechanical tests on both bulk
material samples as well as fully rendered scaffolds.
It may also be useful to study the interaction between
mechanical loading of the scaffold and its degradation.
Finally, it may be useful to study scaffold degradation
byproducts (e.g., does scaffold degradation release
byproducts or affect pH in a way detrimental to seeded
cells, neotissue, or the host).
The third step is to study the use of the resin to form
a burn-in patch on the basement plate on the upper
elevator of the cDLP device. We have not been able to
directly cure a burn-in patch on the build plate. We,
therefore, do this by overcuring resin on the basement
plate. The overcured resin membrane or patch is then
transferred to the build platform and cured onto that
platform using a UV bath (ProcureTM 350, 3D Systems) followed by warming with a heat gun. Heat is
used to ensure that the patch center cures to the underlying build platform, as the dye content of the resin
may prevent UV penetration at the patch edges. Care
must be taken to allow the heated layer and platform
to cool to prevent accelerated curing kinetics when
the patch is reintroduced to the device. This procedure
allows scaffolds to bind to cured PPF resin directly,
rather than the metal build platform itself.
The fourth step is to transfer the scaffold CAD file
to the cDLP device for rendering. The CAD file will
likely contain support structures spanning the space
between the scaffold and the burn-in patch. The support structures must rise sufficiently above the burn-in
patch to allow resin to circulate between the burn-in
patch and the scaffold during rendering of the scaffold
and to allow washing out of unpolymerized resin following that procedure. It is common to use pin-point
supports in SLA work. We expect that, in the cDLP
device, because the build plate is pulled up out of the
resin, peeled off the basement plate, and subsequently
pressed back down into the resin against the basement
plate, there may be more force applied to the scaffold, and especially to a region like a pin-point support
structure, during cDLP rendering than during SLA
CALIBRATION STUDY
The PPF was synthesized and purified as per previously described methods [20]. Briefly, DEF (Acros,
Pittsburgh, PA) and propylene glycol (Acros) were
reacted in a 1:3 molar ratio with hydroquinone and
zinc chloride as a crosslinking inhibitor and a catalyst,
respectively. This reaction created the intermediate,
bis(hydroxypropyl) and ethanol as a byproduct. The
intermediate was then transesterified under a vacuum
to produce poly(propylene fumarate) and propylene
glycol as a byproduct. The PPF was then purified
and gel permeation chromatography was used to calculate the number average molecular weight (Mn =
1200 Da).
We used Sachtleben (White Plains, NY) R320 TiO2
which is a 320 nm crystal. Our study found that we
obtained a 133 m layer of PPF 4.8% TiO2 (range
tested: 04.8%), 2% BAPO (range tested 0.52%),
33% DEF (range tested: 33 and 50%), and an irradiance level of 200 mW/dm2 for 300 seconds (60 s
and 300 s were tested) (Figures 5 and 6). In our single layer tests we observed something we had not seen
with any other dye-initiator package, a lateral spreading (i.e., in x and y) of polymerization (i.e., lateral
overcuring) beyond the intended layer boundaries.This
area increased most quickly at higher concentrations of
TiO2 , especially with increased light input at those high
dye concentrations (Table 1, Figure 7). The area of lateral overcuring was not as thickly or as strongly cured
as the expected area of exposure. As with overcuring
62
Table 1.
Figure 5. Depth of polymerization (m) was characterized

as a function of titanium dioxide concentration (wt%) for five
different combinations of BAPO concentration (wt%) and
exposure time (s). From these tests, it was determined that a
2wt% titanium dioxide concentration with 2wt% BAPO and
a 60s exposure time would yield an average depth of polymerization equal to 133.3 m. These settings could therefore
be used to build in 50 m layers with 83.3m of overcuring. A 200 mW/dm2 irradiance was used for these tests. Data
represents mean standard deviation (n = 3).
Percent of Lateral Overcuring*
[TiO2 ]
wt%
[BAPO]
wt%
Sample Number
1
Avg.
Std.
Dev.
0
0.1
0.2
0.3
0.4
0.8
1.2
1.6
2
2.4
2.8
3.2
3.6
4
4.4
4.8
0
0.1
0.2
0.3
0.4
0.8
1.2
1.6
2
2.4
2.8
3.2
3.6
4
4.4
4.8
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
0
0
36.36
36.36
54.55
63.64
59.09
77.27
72.72
81.82
72.73
72.73
72.72
72.73
68.18
63.64
0
16.67
25.00
36.36
45.46
54.55
54.55
63.64
72.73
68.18
81.82
81.82
81.82
81.82
81.82
81.82
0
0
36.36
45.45
54.55
63.64
72.73
72.72
77.27
72.72
72.73
72.73
63.64
63.64
68.18
72.73
0
16.67
25.00
40.91
45.46
54.55
63.64
63.64
63.64
72.73
77.27
77.27
81.82
81.82
72.73
72.73
0
0
36.36
45.45
54.55
63.64
72.73
72.72
72.72
72.72
72.73
63.64
68.18
63.64
63.64
72.73
0
16.67
30.68
36.36
45.46
54.55
63.64
63.64
72.73
68.18
77.27
72.72
81.82
77.27
81.82
81.82
0
0
36.36
42.42
54.55
63.64
68.18
74.24
74.24
75.76
72.73
69.70
68.18
66.67
66.67
69.70
0
16.67
26.89
37.88
45.46
54.55
60.66
63.64
69.70
69.70
78.79
77.27
81.82
80.30
78.79
78.79
0
0
6.8E-15
5.25
0
0
7.87
2.62
2.62
5.25
1.4E-14
5.25
4.55
5.25
2.62
5.25
0
0
3.28
2.62
0
0
5.25
0
5.25
2.62
2.62
4.55
0
2.62
5.25
5.25
*Lateral overcuring is characterized as a function of TiO2

concentration for two different BAPO levels, 1 and 2%. Three
(n = 3) curing tests were performed for each TiO2 /BAPO
combination.
Figure 6. Increasing titanium dioxide concentration led to
an increased amount of lateral overcuring. Testing was performed using a 200 mW/dm2 irradiance and a 300 s exposure
time. Two levels of BAPO, shown above, were tested for each
titanium dioxide concentration. Data shown here represents
mean standard deviation (n = 3).
in the z direction, we expected that at an appropriate

setting this would not diminish the accuracy of the
rendered scaffold.
In order to quantify this phenomenon, an extra
step was added to the normal curing test calibration procedure. In addition to measuring cured layer
thickness, i.e. the Z dimension, XY dimensions were
also measured. The curing test procedure uses a small
square-shaped test pattern of UV exposure. At each
TiO2 concentration increment, the length and width
of the cured square-shaped thin layer were recorded.
Additionally, the length and width of the total cured
area, including those areas affected by lateral overcuring, were also measured. With these data, it was
possible to calculate the percent overcure. The length
Figure 7. A curing test sample is shown. The superimposed,

red, dashed line encloses the square test pattern which is projected during the test. The material outside of this boundary
was not directly exposed, but rather was polymerized due the
lateral overcuring caused by scattering.
and width, or X and Y, measurements were averaged

for each part, and this process was repeated three
times (n=3) for each TiO2 and BAPO concentration
(Table 1).
Next we went to building 2 plate test parts like those
seen in Figure 4, which are 11 plates. The two layer
test parts also included support structures between the
first plate and the burn-in patch. These test parts were
successful, although we found that it was useful to
63
decrease the thickness of the burn-in patch. We chose

to do that instead of increasing the size of the supports
as it was both possible and reduced PPF usage.
After rendering the 2 plate scaffolds we moved on
directly to the full 11 plate scaffolds. On our first
experiment we found an incomplete build and a membrane of polymerized material had formed on the
basement plate. This was corrected by: (1) regularly
straining out polymerized resin, (2) cleaning the basement plate at set intervals, and (3) monitoring the
basement plate throughout the 16 hour build cycle. We
found cleaning unpolymerized polymer from both the
2 and 11 plate scaffolds from the internal pore space
to be a simple procedure using an ultrasonic alcohol
bath. The scaffolds rendered for this study were shown
Figure 8. A photograph of a full plate-and-post scaffold

rendered using a 1:1 PPF/DEF ratio with 1.5wt% BAPO and
0.75wt% TiO2 . A 200 mW/dm2 irradiance and 150s exposure
time were used.
Table 2. Azo-Chromium Dye PPF Scaffold Accuracy

Measurements of Plates and Posts for Single Plate Scaffolds
Sample
Number
Plate
Thickness
(mm)
1
0.45
2
0.41
3
0.42
4
0.41
5
0.41
6
0.40
7
0.43
8
0.46
9
0.41
10
0.45
Avg.
0.43
Std. Dev. 0.02
Post Diameter
(mm)
Post Number*
Table 4. TiO2 Dye PPF Scaffold Accuracy 2-Plate Test

Scaffold Diameter Measurements*
Average
Post
Diameter
(mm)
0.59
0.65
0.62
0.60
0.59
0.61
0.64
0.63
0.63
0.67
0.67
0.66
0.61
0.63
0.63
0.61
0.62
0.64
0.59
0.66
0.61
0.64
0.63
0.59
0.63
0.63
0.64
0.61
0.61
0.61
0.67
0.67
0.64
0.66
0.65
0.60
0.63
0.64
0.58
0.63
Sample
Number
0.64
0.66
0.63
0.62
0.63
0.61
0.63
0.63
0.60
0.64
0.63
0.01
1
2
3
4
Avg.
Std. Dev.
Plate 1
Diameter
(mm)
Plate 2
Diameter
(mm)
5.90
6.02
5.93
5.97
N/A**
5.83
5.92
5.84
5.92
0.07
*Four 2-plate test scaffolds were rendered using a 150 s exposure time and a 200 mW/dm2 irradiance. The resin used for
these samples contained a 2:1 PPF/DEF ratio with 2wt%
BAPO and 1 wt% TiO2 . For each test scaffold, the diameter of
each plate was measured using digital calipers. The expected
diameter was 6.0 mm. The observed error is approximately 2
times the Perfactory devices tolerance of 35 microns.
**This plate rendered well but was damaged during postprocessing.
*For each test scaffold, four post diameter measurements

were collected.
Table 3. Azo-Chromium Dye PPF Scaffold Accuracy Full Scaffold Diameter

Measurements
Sample
Number
Exposure
Time
(min)
Diameter Measurements*
(mm)
6.01
6.01
6.03
6.01
5.84
5.87
5.85
5.80
5.89
5.85
5.80
5.83
5.85
5.79
5.89
5.82
6.08
6.02
6.02
6.02
5.79
5.80
5.87
5.81
5.87
5.77
5.86
5.82
5.83
5.84
5.82
5.87
6.03
6.03
6.01
6.00
5.81
5.81
5.80
5.85
5.79
5.84
5.82
5.80
5.80
5.85
5.84
5.90
6.08
6.00
6.05
6.04
5.85
5.82
5.80
5.85
5.87
5.88
5.83
5.83
5.83
5.83
5.86
5.83
6.04
6.02
6.02
6.00
5.86
5.84
5.79
5.86
5.85
5.86
5.81
5.82
5.85
5.82
5.84
5.87
Std.
Avg.
(mm)
Dev.
(mm)
6.03
0.03
6.02
0.02
5.83
0.03
5.83
0.03
5.85
0.04
5.82
0.02
5.83
0.02
5.85
0.03
*The diameter of each sample was measured in ten random locations along the
12.4 mm length of the scaffold.
64
in the near UV (400300 nm) and/or middle UV (300

200 nm). Since the dye is incorporated into the scaffold
it may be desirable if that dye: has antibiotic properties, causes surface roughness, has a cell-signaling
function, or has a chemical function (e.g., pH buffering). In choosing TiO2, we were confident in its
biocompatibility and light-blocking capabilities.
We were initially concerned by what appeared to
be an increased level of unintended polymerization
(i.e., lateral overcuring). However, we hypothesize that
polymerization is not only carried out beyond the build
envelope in x and y but that it is also occurring in
oblique directions in the layer of polymer that is present
in the basement tray of the cDLP device. We expect
that the high refractory index of the TiO2 is causing
light scattering. However, while this scattering is in all
directions, rather than only in the z direction, if the
amount of solid layer curing continues to occur only
in the z direction, we see no reason to be more concerned than if one were trying to calibrate the amount
of interlayer overcuring alone. There will be no interlayer overcuring in other directions as their will be no
additional layers to the sides and the layers above the
current location are not yet in existence.
Of greater concern was the production of a second
membrane on the base plate. While we could manage
this, we found that we were also straining the resin
more frequently than has been our experience with
an SLA device. This may be because the resin chamber is not cooled, has a lot of surface area exposed
to warm air, and possibly there was light not filtered
by the devices housing during our use of the device.
Polymerization is always occurring, but given these
factors and the relatively high initiator concentration,
if it is possible, it might be best to have a good supply
of fresh polymer during calibration procedures and to
run the machine as continuously as possible at other
times to reduce light and heat exposure. In addition to
light blocking it may be interesting to investigate dyes
that absorb light. Light absorbing dyes, initiators, or
solvents might better preserve the shelf life of the resin.
Current cDLP devices can provide native accuracies
of up to 13 m in z and 71 m in x and y, and up to
35.5 m in x and y when using anti-aliasing software
[21]. Resolution has increased as newer generations of
these devices have been released. This is clearly sufficient resolution to prepare patient-specific implants
[2]. This resolution may be sufficient to render features
(e.g., surface roughness) that cells might respond to.
Figure 9. Scaffolds rendered using an azo-chromium based

dye are illustrated here in several formats. A) Scanning electron microscope (SEM) image of full scaffold (Note fish eye
lens artifact); B) SEM zoomed view of Full Scaffold features. Note pixelation of surfaces (in plane); C) Top view of
plate; D) Oblique view of scaffold features. E) Side view
of scaffold features as reconstructed from CT. F) Oblique
view of the CT data set.
to be accurate to within 80 m (Table 4) as measured

by micrometer. One of the resulting implants is shown
in Figure 8.
As a control the same calibration study was performed using a yellow azo chromium dye, Meco Fast
Yellow Y-57 (Ming Zu, Taipei, Taiwan). The plate and
post diameters results of this study have been previously reported (Table 2) [21] and we report the
plate thicknesses here (Table 3). These scaffolds are
depicted in Figure 9.
6.1
Conclusions
Our experiments with SLA rendering of PPF scaffolds

did not use a dye-initiator package. In those experiments using 800 Da PPF [10], we were unable to render
layers of less than 400 m. Using 1200 Da PPF, we
are able to use a cDLP device to render layers as thin
as 60 m. The resulting highly accurate scaffolds are
likely to allow improvements in the modeling, prediction, and eventual design of scaffold-specific cell
attachment, proliferation, maturation, and resorption
parameters.
DISCUSSION
The creation of dye-initiator packages for cDLP rendering of tissue engineering scaffolds requires, at
minimum, a dye that is biocompatible and blocks light
65
The increase in green strength over SLA-rendered,

low molecular weight, PPF scaffolds was also a significant improvement [10]. We hypothesize that there
are at least two reasons for the increased resolution
and improved green strength over SLA-rendered scaffolds. First, because of the cDLP devices software and
the fact that we did not use a dye-initiator package.
Second, we expect that the SLA controller software
results in the deposition of less energy in the x, y,
and z directions prior to post-curing. It is optimized
to quickly drawing the part so that it can be put in a
post-cure UV bath. This saves rendering time and use
of the UV laser. Green strength is less of an issue with
the relatively higher molecular weight polymers commonly used in many commercial rapid prototyping and
additive manufacturing activities.
We expect that scaffold rendering resolution will
continue to increase with further iterations of cDLP
devices. Part of the increase will be from refinements to cDLP technology and part of the increase
may be from the use of dye-initiator packages which
allow the production of highly accurate features with
sufficient green strength to allow aggressive postrendering handling, especially the removal of supports
and unpolymerized resin. Finally, we expect that cDLP
dye-initiator packages can be further optimized to
bring about post-implantation benefits to scaffold
resorption and neotissue incorporation.
[8]
[9]
[10]
[11]
[12]
[13]
ACKNOWLEDGMENTS
This research was partially supported by the Research
Foundation of the Department of Neurological
Surgery, Case Western Reserve University, Cleveland,
OH and by NIH grant R01-DE013740.
[14]
[15]
REFERENCES
[16]
[1] ASTM F279210. Standard Terminology for Additive Manufacturing Technologies. American Society
for Testing and Materials. West Conshohocken, PA.
[2] Dean, D., Min, K.-J. & Bond, A. 2003. Computer Aided
Design of Pre-fabricated Cranial Plates. J. Craniofacial
Surgery 14: 819832.
[3] Hollister SJ. 2005. Porous scaffold design for tissue
engineering. Nat. Mater. 4: 518524.
[4] Anderson, E.J. & Knothe Tate, M.L. 2007. Design
of tissue engineering scaffolds as delivery devices
for mechanical and mechanically modulated signals.
Tissue Eng. 13(10): 25252538.
[5] Schek RM, Wilke EN, Hollister SJ, Krebsbach PH.
2006. Combined use of designed scaffolds and adenoviral gene therapy for skeletal tissue engineering.
Biomaterials 27(7):11601166.
[6] Vehof JW, Fisher JP, Dean D, van der Waerden JP,
Spauwen PH, Mikos AG, Jansen JA. 2002. Bone formation in transforming growth factor beta-1-coated
porous poly(propylene fumarate) scaffolds. J Biomed
Mater Res 60(2):241251.
[7] Alves CM, Yang Y, Marton D, Carnes DL, Ong JL,
Sylvia VL, Dean DD, Reis RL, Agrawal CM. 2008.
[17]
[18]
[19]
[20]
[21]
66
Plasma surface modification of poly(D,L-lactic acid)

as a tool to enhance protein adsorption and the attachment of different cell types. J Biomed Mater Res B
Appl Biomater 87(1):5966.
Sun, M., Lv, D., Zhang, C., Zhu, L. 2010. Culturing functional cartilage tissue under a novel bionic
mechanical condition. Medical Hypotheses 75: 657
659.
Peltola, S.M., Melchels, F.P., Grijpma, D.W.,
Kellomki, M. 2008. A review of rapid prototyping
techniques for tissue engineering purposes. Ann Med.
40(4): 268280.
Dean, D., Wallace, J., Kim, K., Mikos, A.G., Fisher, J.P.
2010. Stereolithographic Rendering of Low Molecular
Weight Polymer Scaffolds for Bone Tissue Engineering. In P.J. da Silva Brtolo, A.C.S. de Lemos, A.M.H.
Pereira, A.J. dos Santos Mateus, A.L.A. Mendes;
C.S.M. de Moura, C.A.B. Capela, C.S.G. da Silva,
F.A.C. Domingues, H.M.C.daR.T.G. Brtolo, H. deA.
Almeida, I.S.C. Ferreira, J.M. Matias, N.M.F. Alves
& S.C.S.F. Rodrigues, Innovative Developments in
Design and Manufacturing: Advanced research in
virtual and rapid prototyping, CRC Press (Taylor &
Francis), Boca Raton, FL, pp. 3743.
Jansen, J., Melchels, F.P., Grijpma, D.W., Feijen, J.
2009. Fumaric acid monoethyl ester-functionalized
poly(D,L-lactide)/N-vinyl-2-pyrrolidone resins for the
preparation of tissue engineering scaffolds by stereolithography. Biomacromolecules 10(2):214220.
Professional Experience: A. Siblani.
Heller, C., Schwentenwein, M., Russmueller, G.,
Varga, F., Stampfl, J., Liska, R. 2009. Vinyl esters:
low cytotoxicity monomers for the fabrication of
biocompatible 3D scaffolds by lithography based additive manufacturing. J Polym Sci A Polym Chem
47(24):69416954
Melchels, F.P., Feijen, J., Grijpma, D.W. 2010. A review
on stereolithography and its applications in biomedical
engineering. Biomaterials 31(24):61216130.
Seck, T., Melchels, F., Feijen, J., Grijpma, D.
2010. Designed biodegradable hydrogel structures
prepared by stereolithography using poly(ethylene
glycol)/poly(d,l-lactide)-based resins. Journal of
Controlled Release, 148(1), 3441.
Paschoal, F.M., Anderson, M.A., Zanoni, M.V. 2009.
Simultaneous removal of chromium and leather dye
from simulated tannery effluent by photoelectrochemistry. J Hazard Mater. 166(1):531537.
Fisher, J.P., Holland, T.A., Dean, D., Engel, P.S., Mikos,
A.G. 2001. Synthesis and properties of photocrosslinked poly(propylene fumarate) scaffolds. J Biomater
Sci Polym Ed 12:673687.
Fisher, J.P., Dean, D., Mikos, A.G. 2002. Photocrosslinking Characteristics and Mechanical Properties of Diethyl Fumarate/Poly(Propylene Fumarate)
Biomaterials. Biomaterials 23: 43334343.
Fujishima, A., Honda. K. 1972. Electrochemical
Photolysis of Water at a Semiconductor Electrode.
Nature 238(5358): 37.
Kasper, F.K., Tanahashi, K., Fisher, J.P., Mikos, A.G.
2009. Synthesis of poly(propylene fumarate). Nat
Protoc 4(4), 518525.
Wallace, J.E., Siblani, A., Wang, M., Kim, K., Fisher,
J.P., Mikos, A.G., Dean, D. 2010. Highly Accurate Rendering of Tissue Engineered Scaffolds via Continuous
DLP Polymerization. 2010 International Conference
on Biofabrication, Philadelphia, PA 10/05/10.

Fabrication and characterization of biodegradable composite

scaffolds for Tissue Engineering
T. Serra & M. Navarro
Institute for Bioengineering of Catalonia (IBEC), Barcelona, Spain
J.A. Planell
Institute for Bioengineering of Catalonia (IBEC), Technical University of Catalonia, CIBER-BBN, Barcelona, Spain
ABSTRACT: In this study, polylactic acid (PLA) and polyethylene glycol (PEG) were combined with soluble
CaP glass particles and processed by rapid prototyping to obtain fully biodegradable structures for Tissue Engineering applications. The obtained 3D biodegradable structures were characterized in terms of their architecture
and mechanical properties. The scaffold morphology, internal micro-architecture and mechanical properties were
evaluated using Scanning Electron Microscopy (SEM), micro-computed tomography (micro-CT) and mechanical testing, respectively. Well defined structures with pore size of 350400 m (in the axial view), struts width
of approximately 7080 m, and a porosity ranging between 6065% were obtained. The combination RP and
PLA/PEG/CaP glass turned into promising fully degradable, mechanically stable, bioactive and biocompatible
composite scaffolds for TE.
INTRODUCTION
tool is an interesting and versatile technique to build

three-dimensional heterogeneous and complex tissue
models (Butscher 2011).
By reviewing the literature on RP fabricated
scaffolds, numerous degradable polymers such as
polycaprolactone (PCL), polylactic acid (PLA), polyglycolic acid (PGA), chitosan and their copolymers
have been used to fabricate scaffolds (Hutmacher
2007, Yeong 2004, Taboas 2003, Geng 2005, Moroni
2006, Seal 2001).
In particular, PLA is a currently used biodegradable
polymer that has been approved by the FDA for different biomedical applications. Though this polymer
has been extensively studied, its use in the elaboration
of rapid prototyping scaffolds and particularly those
elaborated through nozzle-based systems has been
limited and scarcely reported (Melchel 2009, Xiong
2002).
One of the strategies to improve bioactivity and
mechanical integrity of polymer scaffolds is by incorporating an inorganic phase such as calcium phosphate
particles (Shor 20072009). In this sense, calcium
phosphate glasses are a good candidate given their
controlled biodegradability and bioactive potential.
In special, calcium phosphate glasses in the system P2 O5 -CaO-Na2 O-TiO2 have shown excellent biocompatibility both in vitro and in vivo (Navarro 2003a,
Sanzana 2008).
This work aims at fabricating a fully biodegradable 3D scaffold based on PLA as major polymer
The development of new temporary 3D porous scaffolds with precise and reproducible geometries that
support cell in-growth, differentiation and ultimately
tissue regeneration is of paramount importance in
Tissue Engineering in order to study the effect of scaffolds architecture in cell response and also to have
well defined architectures that allow studying and
predicting scaffolds behavior by computer modeling
(Hollister 2005).
Most scaffolds developed over the last 30 years
mainly fabricated by conventional techniques such as
solvent casting and particulate leaching, thermal
induced phase separation and gas foaming among others have shown irregular geometries and poor mechanical properties (Hutmacher 20002007). Recently,
rapid prototyping (RP) has emerged in the biomaterials field as a new tool that allows the fabrication of scaffolds with well defined architectures in a
reproducible layer-wise fashion. Moreover, RP opens
the possibility to build custom-made scaffolds based
on patient-specific tissue defect. Several rapid prototyping techniques have been developed and the
elaboration of different polymer and ceramic scaffolds with different geometries have been reported
(Hollister 2005, Hutmacher 20002007, Yeong 2004,
Zein 2002, Taboas 2003). In particular, a nozzledeposition approach consisting in a dispensing system
integrated with pumping technology and a CAD/CAM
67
matrix component and polyethylene glycol (PEG)

together with calcium phosphate soluble glass particles by a rapid prototyping method to obtain a
3D composite structure with superior mechanical
properties and bioactivity. In addition, the obtained
structures are characterized by evaluating their processing, their final architecture, and mechanical
behaviour.
2
2.1
2.2.1 Optimization of deposition parameters

Prior to fabrication, standard calibration curves for
the polymer and composite solutions were generated
by depositing material lines at various speeds and pressures to establish the range of line widths obtainable
for a given tip diameter. Different PEG concentrations
(51020% w/w) were mixed with PLA and G5 in
order to evaluate the effect of PEG in the material
processing. Two sets of two-dimensional structures
composed of parallel lines were created: one by fixing
the deposition rate at 3 mm/s and varying the extrusion pressure between 10 and 100 psi, the second by
keeping the pressure constant at 50 psi and varying
the deposition rate between 1 and 10 m/s. For each
value of pressure and deposition speed, three lines
were deposited and their widths were measured using
a profilometer (Dektak 6M, Veeco). The calibration
lines of the system were obtained for all polymer solutions and were used to establish the optimal printing
parameters.
A 0/90 lay-down pattern (Fig. 2) was implemented
in order to produce a pattern of fully interconnected
square pores. Regarding the scaffold design, a nominal
orthogonal layers configuration (ORTH) with distance
between roads (D) of 500 m and diameter of the
struts () around 200 m was adopted.

Materials
Poly(95L/5DL)lactic acid from PURAC and PEG

(Mw = 400) from Sigma Aldrich were dissolved in
chloroform (5% w/v) in order to obtain a homogeneous polymer solution. PEG was used as a plasticizer in order to facilitate the scaffold processing.
A titania-stabilised, completely degradable, calcium
phosphate glass with the following molar composition: 44.5P2 O5 -44.5 Ca2 O-6Na2 O-5TiO2 coded G5
was used in the form of particles (<40 m) and
added to the solution (Navarro 2003b). Materials were
combined according to the compositions shown in
Table 1.
2.2
Scaffolds design and fabrication
3D scaffolds were fabricated by a nozzle-deposition

system (direct-print tool) (Tissue Engineering 3Dn300 by Sciperio/nScrypt. Inc., Orlando, Florida)
shown in Figure 1. It is a dispensing system
integrated with a pumping technology to conformably deposit materials on various substrates.
It utilizes a computer-aided-design/computer-aidedmanufacturing (CAD/CAM) approach to build threedimensional structures. The dispensing process is
controlled by the motion control software and the
CAD program allowing flexible alteration of parameters such as 3D geometry of the deposition pathways,
speed of deposition, and air pressure in the pneumatically actuated pump, and dispensing height. It features
accuracy and repeatability of the XYZ positioning of
the dispensing nozzle with a resolution within a few
microns (Li 2007).
The fabrication of 3D scaffolds via polymer deposition involved the adjustment of several parameters
such as printing speed, polymer flow, height of the
single line deposed, polymer concentration, nozzle
size and temperature of the heating jacket connected
to the syringe.
Table 1.
Figure 1. 3Dn-300 direct-print tool (designed by Sciperio/

nScript Inc., Orlando, USA).
Composition of the studied materials.
Material
Polymer matrix
(w/w %)
G5 particles
(w/w %)
PLA/PEG
PLA/PEG/G5
95 PLA/5PEG
95 PLA/5PEG
50
Figure 2. (a) Lay-down pattern of 0/90 forming square

honeycomb pores view in the axial direction of the printing
process. (b) Cross-section view of the printing process.
68
2.3
Scaffold characterization
where , L, w, and h refer to the strut diameter,

strut length, scaffold width, and scaffold height in
millimeters, respectively. Furthermore Nc represents
the number of cylinders (struts) per layer, while Nl
represents the number of layers per scaffold.
Porosity was also measured using mass/volume
techniques according to the following relationship:
2.3.1 Differential scanning calorimetry

Differential scanning calorimetry (DSC-2910, TA
Instruments) was used to determine the thermal
properties of the materials pre-and post-processing.
Samples (510 mg) of PLA/PEG 95/5 films produced
by solvent casting method and 3D-deposited scaffolds, were firstly heated from 10 C to 200 C then
cooled to 25 C and heated to 200 C, at a heating
rate of 10 C/min in aluminium pans with nitrogen as a
purge gas. The resulting DSC curves were analyzed to
determine glass transition (Tg ) and crystallization (Tc )
temperatures, and the crystallinity (Xc ) of the scaffolds. All values were taken from the thermograms
corresponding to the second heating cycle.
Being
scaffold = apparent scaffold density (g/cm3 ) =
m/(Lwh).
Therefore,
2.3.2 Morphological SEM study

Morphological analysis of the 3D structures was carried out by scanning electron microscopy (SEM, JEOL
JSM 6400, Tokyo, Japan) to visualize and evaluate
the physical integrity of the material struts and layers, as well as to evaluate the changes between the
theoretically defined pore geometry and size and the
one obtained after processing. SEM observation also
allowed verifying the distribution of glass particles.
where m represents the mass of the scaffold (g) and

bulk represents the material density (g/cm3 ), in the
case of polymeric scaffolds PLA = 1,231 g/cm3 while
in the case of composite scaffolds 1/bulk = Wt %PLA /
PLA + Wt %G5 /G5 . being G5 = 2,903 g/cm3 .
2.4
Mechanical properties of scaffolds
A Universal Testing Machine (MTS-Bionix 858, MTS

Systems Corporation, Eden Prairie, USA) with a 10N
load cell was used to evaluate the mechanical properties of scaffolds. The samples were tested at a speed of
1 mm/min without preloading. Stress-strain data were
computed from load-displacement measurements. The
compressive modulus was determined based on the
slope of the stress-strain curve in the elastic region.
For each material composition, three cubic scaffolds
(5 5 5 mm3 ) were tested. Cubic samples were
cored from larger 3D printed blocks initially designed
in the CAD software. The real accurate dimensions of
the specimens were measured before the test.
2.3.3 Porosity
The theoretical volume porosity percentage
(%Voltheoretical ) was calculated for each scaffold using
the initially designed geometries based on a unit
cube (Fig. 3), whereby the strut diameter and spacing
between layers were equal (i.e., no overlapping due to
the fusion between struts from one layer to the adjacent
was assumed).
where
Vt = true volume (mm3 ) =Vc Nc Nl = (2 /4) L Nc Nl
Vc = cylinder volume (mm3 )
and
Va = apparent volume (mm3 ) = Lwh
L = Nc + D (Nc 1)
D = distance between roads (struts)
h = Nl
Therefore,
2.4.1 Microstructure analysis and 3D

reconstruction by micro-CT
Two scaffolds (one for each material composition)
were scanned using a microCT X-Tek HMX225
(Digisens) with a voxel resolution of 8 8 8 m3 .
Computer 3D reconstruction of the scaffolds was made
using Mimics 14.0 software (Materialise, Leuven,
Belgium) to determine the porosity as well as the percentage and distribution of the glass particles within
the 3D composite structures.
3
3.1
RESULTS
Optimization of deposition parameters
(calibration curves)
Figure 3 shows the measured line widths of solutions of

PLA/PEG blend at different concentrations (20%, 10%
and 5% w/w of PEG) and PLA/PEG/G5, respectively,
as a function of the applied pressure. It is observed
that in general, the lines width increases linearly as
the applied pressure increases. Furthermore, for the
same value of pressure, the width increases slightly by
Figure 3. Cross section view of the dispensing process.
69
Table 3. Thermal characteristics of PLA/PEG 95/5 during

the second heating scan.
PLA/PEG 95/5
Tg ( C)
Tc ( C)
%Xc
Nonprocessed
Processed
40
44.61
120.1
108.4
2.1
2.4
(Table 2). The syringe temperature was set at 4045 C

while ambient temperature was 25 2 C.
Figure 3. Line widths of PLA/PEG blend at different concentrations and PLA/PEG G5 as a function of applied pressure. Deposition speed: 3 mm/s; needle diameter: 200 m.
3.2
Thermal characterization
DSC analysis confirmed that the crystalline fraction

of PLA/PEG 95/5 did not changed significantly when
the polymer was processed. There was no significant
difference between the glass transition temperature
value before and after the process. The values of crystallinity for non-processed and processed PLA/PEG
95/5 ranged between 2.1% and 2.4%, respectively. The
thermal characteristics of PLA/PEG 95/5 samples both
before and after processing are indicated in Table 3.
3.3
According to the SEM analysis, with both polymeric and composite materials, well defined structures
with pores size of 350400 m in the axial view
and struts width of approximately 7080 m were
obtained (Fig. 5). The total spacing between the struts
axes was therefore about 500 m, according to the
nominal design.
In the case of the material with G5 glass, a fairly
homogeneous distribution of the particles within the
matrix was observed with the glass incorporated and
embedded by the polymer (Fig. 5). Furthermore, these
3D structures showed coexistence of porosities ranging from the macroscale (due to the designed pore
size) to the micro and nanoscale due to the presence
of glass particles and to the pores left by solvent
evaporation (Fig. 6).
All the scaffolds showed full interconnected porous
structures. Theoretical volume porosity percentage
based on the deposition paths ranged between 87.21%
and 85,72% considering distance between struts
400 m and 350 m respectively.
Experimentally measured porosities where 65.8
4.8% for the PLA/PEG 95/5 material and 60.0 3.7%
for the PLA/PEG/G5 according to equation (4).
Figure 4. Line width of PLA/PEG blend at different concentrations as a function of the deposition speed. The points
refer to experimental data Driving pressures: 50 psi; needle
diameters: 200 m.
Table 2. Detailed processing parameters for the fabrication
of scaffolds.
Dispensing pressure
Temperature
Printing rate
Nozzle size
Pore size
Morphological evaluation by SEM and porosity
4080 psi
4045 C
2 mm/s
200 m
350400 m
decreasing PEG concentration (and increasing PLA

concentration) while G5 in troduction had a more
significant effect.
Figure 4 shows line widths as a function of the deposition rate for the same set of solutions. It is observed
that the lines width decreases remarkably as deposition speed increases. The blends with 20 and 10%
of PEG (PLA/PEG 80/20 and 90/10) showed similar
behaviours. The blend with 5% PEG (PLA/PEG 95/5)
displayed a remarkable increase in line width in comparison to the other compositions. A similar behaviour
was observed when G5 glass was added to the polymer
blend.
An optimal set of parameters for the pumping equipment was selected from the obtained results, i.e. a
motor speed in a range between 40 and 80 psi and
a deposition speed of 2 mm/s. Such combination of
parameters was used with a G27 (200 m) nozzle
3.4
Mechanical properties of scaffolds
The average data in terms of compressive modulus

are plotted in Figure 7 for both the material with and
without glass particles. The compressive modulus of
PLA/PEG 95/5 scaffolds was found to be 92.32 2.18
MPa whereas for PLA/PEG/G5 the value increased to
99.81 3.55 MPa.
70
Figure 8. 3D reconstructed image of a PLA/PEG 95/5

(a) and PLA/PEG/G5 (b) scaffold.
confirmed from the observation of the 3D reconstructions.

Figure 5. SEM micrograph of 3D deposited scaffolds. a, b)
PLA/PEG 95/5 material; c, d) PLA/PEG/G5. (a,c) top view;
(b,d) cross-section view.
In this study a method to produce 3D polymeric

(PLA/PEG) and composite (PLA/PEG/G5) scaffolds
by a rapid prototyping technique was carried out and
optimized. Scaffolds were characterized using SEM,
micro-CT and DSC to characterize the morphology,
internal geometry, mechanical and thermal properties.
An optimized printing process involves complex
interactions among the hardware, software and material properties (Zein 2002, Comb 1994). PEG was used
as plasticizer to decrease the Tg of the blend in order
to improve processing conditions at low temperature
(T = 40 C). After trying different PLA/PEG compositions, it was found that 5% w/w was the best. The
addition of 5% w/w of PEG was enough to improve
the blend processing without modifying excessively
the PLA/G5 material. DSC results showed that the
processing parameters used did not induce significant
changes in terms of Tg and fraction of crystallinity
due to degradation at elevated temperatures or shear
forces during material deposition, which in turn could
compromise scaffold in vitro and in vivo behaviour.
Figure 3 and 4 showed the line width for each material
as a function of pressure and velocity, respectively. As
expected and predicted by the model by Mironov et al.
(2003), the line width increases linearly with increasing pressure and decreases with increasing velocity
(Mariani 2006).
Regarding the morphological analysis by SEM, well
defined structures with pores around 350400 m and
struts of approximately 70 m were obtained in agreement with the initial design. In general, it was observed
that the thickness of the struts obtained by this RP
technique was thinner than the ones reported in the
literature. This fact could be due to the evaporation of
the solvent that takes place once the material has been
deposited. The solvent found in the polymer solution
evaporates immediately after the solution flows out
from the syringe leading to an important shrinkage
of the struts, and leaving behind pores at the scaffold
surface. Thus, differences between the nominal diameter of the struts and the final one could be explained
by this solvent evaporation fact. Indeed, in the case
of fused deposition systems (FDM) where polymers
Figure 6. SEM micrograph of (a) PLA/PEG 95/5 and (b)

PLA/PEG/G5 surface.
Figure 7. Compressive modulus of PLA/PEG 95/5 and

PLA/PEG/G5 scaffolds.
3.5
DISCUSSIONS
Microstructure characterization by micro-CT
Micro-CT enabled 3D characterization of RPfabricated scaffolds. A 3D reconstructed model is

displayed in Figure 8.
The evaluation of porosity from the 3D reconstruction revealed a value of 65% for the PLA/PEG 95/5
blend scaffold and 60% for the PLA/PEG/G5 one.
Moreover, an amount of glass particles corresponding
to a percentage of 43% was obtained. The distribution of the particles was homogeneous as observedby
microCT analysis, and full pore interconnectivity was
71
Geng, L. 2005. Direct writing of chitosan scaffolds using a

robotic system. Rapid Prototyping Journal, 11(2): 9097.
Hollister, S.J. 2005. Porous scaffold design for tissue engineering. Nature Materials 4 518524 .
Hutmacher, D.W. 2000.Rapid prototyping in tissue engineering: challenges and potential. Biomaterials 21, 2529
2543.
Hutmacher, D.W. 2007. State of the art and future directions
of scaffold-based bone engineering from a biomaterials
perspective. J. Tissue Eng. Regen. Med.1: 245260.
Koh, Y. H. 2006. Fabrication of poly(-caprolactone)hydroxyapatite scaffold using rapid direct deposition
Mater. Lett. 60 11847.
Landers, R. 2002. Fabrication of soft tissue engineering scaffolds by means of rapid prototyping techniques. J Mater
Sci. 37(15): 310716.
Li, B. 2007. Robust direct-write dispensing tool and solutions
for micro/meso-scale manufacturing and packaging. Proceedings of the ASME International Manufacturing Science and Engineering Conference, October 2007. Atlanta,
Georgia.
Mariani, M. 2006. Characterization of tissue-engineered
scaffolds microfabricated with PAM. Tissue Eng.
Mar;12(3): 54757.
Melchels, F,P. 2009. A poly(D,L-lactide) resin for the preparation of tissue engineering scaffolds by stereolithography.
Biomaterials 30(2324): 38019.
Mironov, V. 2003. Organ printing: computer-aided jet-based
3D tissue engineering. The layer-by-layer assembly of
biological tissues and organs is the future of tissue
engineering. Trends Biotechnol. 21, 157161.
Moroni, L. 2006. 3D fiber-deposited scaffolds for tissue
engineering: Influence of pores geometry and architecture on dynamic mechanical properties/Biomaterials 27,
974985.
Navarro, M. 2003a. Cellular response to calcium phosphate
glasses with controlled solubility. J. Biomed. Mater. Res.
67A: 10091015.
Navarro, M. 2003b. Physico-chemical degradation of
resorbable phosphate glasses stabilized with TiO2.
J. Amer. Ceram. Soc. 86(8): 134552.
Sanzana, E. 2008. Of the in vivo behaviour of calcium
phosphate cements and glasses as bone substitutes. Acta
Biomaterialia. 4: 19241933.
Seal, B.L. 2001. Polymeric biomaterials for tissue and organ
regeneration. Mater Sci Eng R: Rep;34: 147230.
Shor, L. 2007. Fabrication of three-dimensional polycaprolactone/hydroxyapatite tissue scaffolds and osteoblastscaffold interactions in vitro. Biomaterials 28(35):
52915297.
Shor, L., 2009. Precision extruding deposition (PED) fabrication of polycaprolactone (PCL) scaffolds for bone tissue
engineering. Biofabrication,1(1): 10.
Taboas, J.M. 2003. Indirect solid free form fabrication of
local and global porous, biomimetic and composite 3D
polymer-ceramic scaffolds. Biomaterials; 24: 18194.
Xiong, Z. 2002. Fabrication of porous scaffolds for bone tissue engineering via low-temperature deposition, Scripta
Materialia 46 (11): 771776.
Woodfield, T.B.F. 2004. Design of porous scaffolds for
cartilage tissue engineering using a three-dimensional
fiber-deposition technique. Biomaterials 25, 41494161.
Yeong, W.Y. 2004. Rapid prototyping in tissue engineering: challenges and potential. Trends Biotechnology. Dec;
22(12): 64352. Review.
Zein, I. 2002. Fused deposition modeling of novel scaffold architectures for tissue engineering applications.
Biomaterials 23: 11691185.
are melted and subsequently extruded without using

solvent, this shrinkage phenomenon is not as significant (Zein 2002, Moroni 2006, Shor 20072009,
Woodfield 2004).
As shown in Figure 6, the presence of G5 particles
added an interesting topography to the scaffold surface and, these 3D structures showed a combination
of porosities ranging from the macroscale due to the
pores initially designed to the micro and nanoscale due
to the pores left by solvent evaporation.
Theoretical porosity was higher than the one experimentally measured using mass/volume techniques.
Due to the fusion between struts (of deposited material) and the underlying layers, measured porosity
values were somewhat lower than the theoretical ones.
Similar results have been reported when dispensing
3D structures with hydrogels (Landers 2002). Differences between porosity values can be further explained
by the fact that the theoretical calculations assume a
unit cube as opposed to the actual deposition of struts,
which generated edge effects (Woodfield 2004). In
fact, a decrease of strut diameter, strut spacing or layer
thickness results in an increase of deposited polymer
and a consequent decrease of pore volume within the
same scaffold volume (Moroni 2006).
Nevertheless, the evaluation of porosity by microCT reconstruction confirmed the values obtained by
the experimental approach.An estimation of the homogeneous dispersion of G5 particles in the polymer
matrix was carried out and the obtained values were in
agreement with the theoretical ones. The incorporation
of G5 glass particles into the polymer matrix improved
the mechanical behaviour of the scaffolds significantly. Indeed, the compressive modulus increased
8.11% when the glass was added. In general, the
structures obtained in this study showed higher values of compressive modulus than other RP structures
reported in literature where a similar pattern deposition
was used and similar porosity percentage was obtained
(Zein 2002, Shor 2009).
CONCLUSION
Rapid prototyping, in particular the nozzle-based

deposition system used in this work is suitable to
process 3D composite scaffolds based on PLA/PEG
and glass (G5). This technique/materials combination
lead to the attainment of promising fully degradable, mechanically stable, bioactive and biocompatible
composite scaffolds with well defined architectures
for TE.
REFERENCES
Butscher,A. 2011. Structural and material approaches to bone
tissue engineering in powder-based three-dimensional
printing. Acta Biomaterialia 7(3): 907920.
Comb, J.W. 1994. Layered manufacturing control parameters
and material selection criteria. Manufacturing Science and
Engineering Vol. 2, PED-vol. 682, ASME.
72

Spatially varying porosity with continuous path plan for hollowed

tissue scaffolds
A.K.M.B. Khoda
University at Buffalo, Buffalo, NY, US
I.T. Ozbolat
The University of Iowa, Iowa City, IA, US
B. Koc
Sabanci University, Istanbul, Turkey
ABSTRACT: In this paper, a bio-inspired scaffold design has been proposed by incorporating its functional
(biological requirement) and fabrication constraint into the design processes. The proposed new methodology
generates functionally gradient porosity along the hollowed scaffold architecture with desired level of control
by combining two geometrically oriented consecutive layers. The resulting set of layers address the repeatable,
interconnected and controllable pores. Modeling of the first layer starts with discretizing internal region by
connecting ruling lines between internal and external features using a dynamic programming algorithm. These
feature connecting ruling lines act as resolution for the proposed method. Then by accumulating those ruling
lines, the region has been divided into equal area sub-regions. In the second layer, the desired pore size and hence
the porosity has been achieved by geometrically partitioning those sub-regions. And thus the combined set of
layers achieves the desired controlled variational porosity along the scaffold architecture. To ensure a continuous
and interconnected tool path, optimized zigzag pattern followed by a concentric spiral like optimal layers are
generated based on the required biological and mechanical properties. Several examples will be presented to
demonstrate the proposed methodology. The designed examples will also be fabricated layer-by-layer using a
micro-nozzle biomaterial deposition system.
INTRODUCTION
necessitates the well-connectivity of internal network

channels. Also any possible unmatched connection
would create blockade in flow resulting congestion and
should be avoided. Most importantly well-designed
porous scaffolds can eliminate or reduce the foreign
body reaction/oppenheimer effect which could avoid
scarring tissue around it (Madden, Mortisen et al.
2010).
Determining the optimum porosity of the designed
scaffold is a challenging problem in directing cell
attachment and growth (Lin, Kikuchi et al. 2003)
of tissue. And consensus has yet to be met by the
scientific community as several researcher advocate
diverse acceptable range mentioned in Khoda et al.
(Khoda, Ozbolat et al. 2011). Obviously, higher pore
size or porosity supports better cell growth, generation and better flow transportation of nutrients and
metabolic wastes but limits the structural integrity and
the load bearing capacity of scaffolds (Davis, Hsieh
et al. 2005). Thus identifying the porosity level over
the scaffold structure to satisfy both the biological
and mechanical requirements (Gomez 2007) simultaneously is a multi objective criterion requires much
sophistication. And by adding the bio-fabrication
requirements into the design constraint introduce
Successful design of tissue scaffolds determines its

functionality and requires comprehensive consideration of macro and micro-structural properties and
their interaction with natural tissue. Far from being
a passive component, the scaffold material and the
porous architecture play a significant role in tissue
regeneration by preserving tissue volume, providing temporary mechanical function, and delivering
bioactive agents (Hollister 2005). Designing a directly
fabricated scaffold with proper interconnectivity and
continuity between pore along the scaffold architecture is one of the major research themes within Tissue
Engineering.
Cell seeded on the scaffold structure needs nutrients, proteins, growth factors and waste disposal that
make mass and fluid transport vital to cell survival.
The size, geometry, orientation, interconnectivity and
surface chemistry of pores and channels determine the
nature of nutrient flow (Taboas, Maddox et al. 2003).
Moreover, interconnectivity of pores has implications
as far as cell migration and tissue ingrowth are considered (Karande 2007). Thus, the mass and the fluid
need to have a pathway to and from the cell, which
73
geometry is then sliced with required resolution suitable for Solid Freeform Fabrication (SFF) system.
The consecutive layer has been discretized with subregions and pore-cell based on the designed porosity,
using the methodology described in our earlier work
(Khoda, Ozbolat et al. 2011). And corresponding
zigzag pattern and concentric spiral like pattern toolpath have been generated for fabricating the design
structure. Modeling of zigzag pattern layer ensures the
biological and mechanical requirement of the scaffold
and the consecutive concentric spiral like pattern layer
mainly enforces the designed porosity of the scaffold
supporting the structural integrity.
further challenges.Attaining those conflicting target in

scaffold architectures results a complex design, which
is difficult to fabricate with traditional polymeric scaffold fabrication processes (Khoda, Ozbolat et al. 2011;
Sobral, Caridade et al. 2011). Solid free-form fabrication (SFF) techniques reduce this gap between the
complex and critical design with reality and viewed as
an efficient way of reproducibly generating scaffolds
of desired properties with highly intricate architecture and compositional variation (Hutmacher, Sittinger
et al. 2004).
A number of researchers have successfully used
various SFF techniques along with a wide range of
bio-material to fabricate scaffold and a comprehensive review can be achieved in literature (Hutmacher,
Sittinger et al. 2004; Yeong, Chua et al. 2004; Gomez
2007). Most of these proposed scaffold structures
do not conform to the shape of the damaged tissue
as they are mostly uniform rectangular patterns and
are designed with regular porosity. During fabrication
of these designed scaffolds, jumps or motion without deposition is also highly substantial due to the
nature of the tool-path, which is independent of the
geometry. And such non-continuous and un-natural
movements during fabrication degenerate uniformity
of the deposited material shape and need to be avoided
(Khoda, Ozbolat et al. 2011). On the other hand, few
literatures (Sobral, Caridade et al. 2011) achieved
improved cell seeding and distribution efficiency
through the scaffold by implementing gradient porosity. A novel variational porosity with interconnected
and continuous channel structure capable of fabrication by solid free form technique has been proposed in
Khoda et al (Khoda, Ozbolat et al. 2011). Even though
gradient in both porosity and pore sizes are highly recommended, but few research have been done in this
area due to considerable limitations in design and fabrication processes (Khoda, Ozbolat et al. 2011; Sobral,
Caridade et al. 2011).
In this paper, a bio-inspired scaffold design has
been proposed by incorporating its functional (biological requirement) and fabrication constraint into
the design processes. The methodology has been proposed for hollowed organs and tissues such as aorta,
diaphysis of femur bone, heart, or human ear. Porosity architecture in this paper is built in stacks of two
consecutive layers: (i) ruling line based zigzag pattern
and (ii) concentric spiral like pattern. And combination of these two layers result controllable pore size
with proper interconnectivity and continuity along the
scaffold structure, which is suitable for layer-based
bio-fabrication processes.
2.1 Accumulation of sub-region section

The scaffolding area between the internal and external feature has been divided with finite number of
non-intersecting ruling lines by the method proposed
in (Khoda, Ozbolat et al. 2011). The generated ruling lines act as resolution for discretizing the scaffold
architecture based on area. The idea here is to accumulate the area generated between two ruling lines
and formed equal area sub-region meeting the upper
and lower width of each sub-region within the biological requirement range. While accumulating the
equal area sub-regions, attention must be given to the
resulting pore size which has a significant role to play
for effective scaffold functions. Since the presented
methodology achieve the designed porosity with two
consecutive layer stacks, final pore size of biological
preference can only be achieved by the contribution
from both layer in the scaffold. If the sub-regions area
increases so as the pore size and vice versa which eventually represent the conflicting requirement of scaffold
parameter.
Due to the nature of the geometry, the area generated
by two consecutive ruling lines are generally converge
towards inner direction and thus the sub-region section
has a similar converging pattern i.e. area in unit length
decreases form outer to inner directions. To ensure that
such convergence do not restrict the permeability (i.e.
nutrient and waste flow), especially towards the inner
feature of the scaffold, the convergence factor for each
sub-regions need to be within an acceptable range.
2.2 Pore-cell generation
To ensure the designed porosity in the model, the only
controllable parameter is the pore size and this section
deals with controlling the pore size of the scaffold.
Pore size in the scaffold can be defined as the maximum distance from any points inside the scaffold to the
nearest edge of the substrate. To generate the porosity
in the scaffold each sub-region section area has been
considered and divided across the ruling line in equal
number. These segments of each sub-section can be
considered as a pore cell of the scaffold and area sum
of all pore cells in a sub-region must be the same as the
area of the sub-region. Due to the convergence and the
accumulation pattern of the sub-region the pore cell
COMPUTER-AIDED BIO MODELING
Firstly, medical image obtained from Computed

Tomography (CT) or Magnetic Resonance Imaging
(MRI) is used to obtain the geometric and topology
information of the replaced tissue/organ. The resulted
74
would be anisotropic in nature. And to deal with such

anisotropy, emphasize has been given to each individual pore cell during modeling process and has been
designed with geometrical constraints as discussed in
(Khoda, Ozbolat et al. 2011).
OPTIMUM TOOL PATH DESIGN

Figure 1. (a) Zigzag tool path for first layer (b) circular
tool path (c) combined continuous and interconnected deposition path for the stack together with controlled variational
porosity.
The proposed two layer stack porosity design represents the controlled variational porosity along the scaffold architecture. In earlier works, variational porosity
design lacks the continuity and connectivity between
the features and thus fabrication procedure with existing fabrication techniques remain a challenge (Sobral,
Caridade et al. 2011). In our earlier work (Khoda,
Ozbolat et al. 2011), we try to address this issue by
designing region based variational porosity with continuous and interconnected deposition path planning
but material over deposition in transitional region was
a concern. In this work, a novel continuous deposition path planning method has been developed for the
fabrication of the designed scaffold with a computercontrolled system ensuring a well-connected internal
channel network (Figure 1).
3.1
The proposed techniques have been implemented on

2.3 GHz PCs using Rhino Script and Visual Basic
programming languages. For illustration, two example geometries diaphysis section of femur bone and
aorta were generated using ITK-Snap 1.6 software
(www.itksnap.org 2008) from the MRI data obtained
from Mimics Software (http://www.materialise.com/
mimics 2008). Each of the solid models were then
converted to STL file format using a CADs system
converter and were sliced by a set of intersection plane
to find the slicing contours for layer-based manufacturing processes. The methodology has been used in
these generated slices and of which two consecutive
layers represent the desired porosity.
During the modeling two consecutive slices with a
uniform distance of 300 m has been considered for
modeling and the substrate diameter has been considered as 150 m for both examples. For both cases,
two gradient porosity distribution architectures have
been generated with (i) increasing gradient, and (ii)
decreasing gradient porosity from outer curve to inner
curve.
Figure 2 shows the first example of femur slice
from diaphysis section which has marrow cavity. Total
2000 number of ruling line has been generated as initial resolution. A total 140 number of sub-region has
been generated with the average area of 10.27 mm2 .
The maximum and minimum allowable area has been
determined as 1.32 mm2 and 0.18 mm2 respectively
and the corresponding porosity range 72%90% can
be applicable with this design area.
For aorta, total 2000 number of ruling line has been
generated as initial resolution. A total 100 number of
sub-region has been generated with the average area
of 2.16 mm2 . The maximum and minimum allowable
area has been determined as 0.3 mm2 and 0.12 mm2
respectively considering aorta as soft tissue. And the
corresponding porosity range 66%78% has been
determined that can be applicable with this design area.
Bio-fabrication of the proposed methodology has
been performed by using a 3D micro-nozzle biomaterial deposition system. Fabrication process is biologically compatible. Sodium alginate (Sigma-Aldrich,
USA), a type of hydrogel widely used in cell immobilization, cell transplantation, and tissue engineering,
Zigzag tool path
Generating the tool line for the first layer i.e. subregion section layer is generated using a continuous zig-zag pattern. A zigzag pattern has been used
connecting all the sub-region section gradually. Subregions are adjacent to each other sharing a common
edge with its neighbor and the whole area is covered by these sub-region sections. By connecting the
single shared edge from each sub-region in an alternative order will result the zigzag path resembling the
designed sub-region section shown in figure 1(a).
3.2
IMPLEMENTATION AND BIO-FABRICATION
Concentric spiral tool path
Due to the free form shape of the geometry, the pore

cell generating lines are not connected between subregions and are discrete. By connecting those lines
consecutively, without any transformation, would generate stepped dominated discrete circular path which
might not be favorable for fabrication. Besides these
stepping will generate material over-deposition by limiting the tool motion and will restrict the permeability
in the internal channel network. Thus, a continuous
tool-path need to be generated fitting through these
pore-cell generating lines to make the spiral tool
path suitable for smoother fabrication with minimum
shifting of cell boundaries shown in figure 1(b). An
area weight based point insertion algorithm (Khoda,
Ozbolat et al. 2011) has been used for smoothing this
stepped dominated circular path and generate final
concentric spiral tool-path with the combination of
first degree polynomials which is very suitable for
SFF-based control systems.
75
Table 1. Comparison between proposed method and conventional Cartesian coordinate system porous structure.
Overall Avg % Error
Porosity modeling technique
Pore size
Porosity
Femur
0.8
1.4
5.5
0.3
3.8
2.5
6.0
4.2
Aorta
Figure 2. Sub-regions divided with gradient of 89.7, 89.4,
89, 88.7, 88.2, 87.5, 87.5, 87.5, 86.8, 85.9, 84.7, and 81%
(a) increasing porosity from outer to inner (b) decreasing
porosity from outer to inner with continuous tool path.
Proposed method
Conventional Cartesian
coordinate
Proposed method
Conventional Cartesian
coordinate
Figure 3. Sub-regions divided with gradient of 78, 77.7,

76.9, 76, 75, 73.7, 72.2, 69.5, 68.5 and 66% (a) increasing
porosity from outer to inner (b) decreasing porosity from
outer to inner with continuous tool path.
is preferred as biomaterial due its biocompatibility

and formability. 3.3% (w/v) alginate solutions were
prepared by suspending 0.33mg sodium alginate (SA)
into 10 ml de-ionized (DI) water and stirred in room
temperature for 20min. And then the solution has been
filled in a reservoir and a pneumatic system has been
deployed to flow the solution via the micro-nozzles
(150200 m Nozzle tip purchased from EFD ). The
system runs in room temperature under low pressure
(08 psi). The reservoir is mounted on the dispensing
system that is driven by a 3D motion control. A PC is
connected to the system to control the motion in 3D.
And by controlling the motion of the dispensing system, the deposition of the material can be controlled.
Calcium chloride solution (0.6% (w/v)) is then dispensed onto printed alginate structure through another
nozzle to provide cross-linking between the alginate
anion and the calcium cation to form the hydrogel.
A pink color pigment has been used just for proper
visualization purpose.
Furthermore, bio-fabrication of sample examples
has been performed using a micro-nozzle biomaterial deposition system shown in figure (46). For
demonstration purpose, two consecutive slices with a
uniform distance of 400 and 300 m has been considered for modeling femur and aorta respectively. For
proper visualization three gradient type (i) increasing, (ii) decreasing and (iii) constant porosity for both
femur and aorta have been fabricated. For femur slice,
200 m substrate diameter and for aorta slice 150 m
Figure 4. Bio-fabrication of two consecutive slices generated with constant porosity; Femur slice with 85% porosity
(a) by proposed methodology (b) by conventional Cartesian
coordinate system; Aorta slice with 64% porosity (c) by proposed methodology (b) by conventional Cartesian coordinate
system.
substrate diameter has been selected as design parameter to generate the sample tool-path. A total of 110
and 120 number of equal area sub-regions with the
average area of 12.88 mm2 and 1.8 mm2 have been
accumulated for femur and aorta slice respectively.
The proposed method has also been compared with
the conventional uniform-unit-based porous structure
with constant porosity shown in figure 4.Table 1 shows
the porosity comparison of the fabricated model and
the designed model. The characterization has been
performed by using IMAGEJ software. Five random
sub-region samples are taken from each example over
the printed structure and each pore cell inside those
sub-regions are measured. Their average has been
used to calculate the fabricated model porosity. The
result shows an encouraging outcome with minimal %
error while comparing the designed porosity with the
fabricated porosity.
From these fabrication samples, it is clear that the
porosity architecture in Cartesian coordinate do not
follow the shape of the geometry. But the proposed
76
CONCLUSION
In this paper, novel design and path planning methods

are developed to design porous scaffolds conforming to the targeted damaged tissue. The proposed
methods design scaffolds with continuous and interconnected variational porosity. The generated conforming scaffolds provide better structural integrity,
improved oxygen diffusion during cell regeneration,
cell differentiation and guided tissue regeneration.
Most importantly, the proposed methodology generates interconnected and controlled pore size with
desired accuracy and a continuous deposition path
planning appropriate for 3D biofabrication processes.
The developed novel techniques could be used to generate the porous structures with gradient (linear or
non-linear), variational or constant porosity that can
provide localized control of material concentration
along the scaffold architecture.
Figure 5. Bio-fabrication of two consecutive femur slices

divided with gradient of 78, 80, 80, 82, 85, 85, 86, 86, 86, 87,
87 and 87% (a) increasing porosity (b) decreasing porosity
from outer to inner feature.
REFERENCES
Davis, M. E., P. C. Hsieh, et al. (2005). Custom design of
the cardiac microenvironment with biomaterials. Circulation Research 97: 815.
Gomez, C. (2007). A Unit Cell Based Multi-scale Modeling
and Design Approach for Tissue Engineered Scaffolds.
Mechanical Engineering Department, Drexel University
Hollister, S. (2005). Porous scaffold design for tissue
engineering. Nature Material 4: 518526.
http://www.materialise.com/mimics (2008). Mimics.
Hutmacher, D. W., M. Sittinger, et al. (2004). Scaffoldbased tissue engineering: rationale for computer-aided
design and solid free-form fabrication systems. Trends
in Biotechnology 22(7): 354362.
Karande, T. S. (2007). Effect of Scaffold Architecture on
Diffusion of Oxygen in Tissue Engineering Constructs.
Biomedical Engineering, The University of Texas at
Austin.
Khoda, A., I. T. Ozbolat, et al. (2011). A new Functionally
Gradient Variational Porosity Architecture for Hollowed
Scaffold Design. Biofabrication (Accepted).
Khoda, A. K. M. B., I. T. Ozbolat, et al. (2011). Engineered
Tissue Scaffolds With Variational Porous Architecture.
Journal of Biomechanical Engineering 133(1): 011001.
Lin, C. Y., N. Kikuchi, et al. (2003). A Novel Method
for Biomaterial Internal Architecture design to match
bone plastic properties with desired porosity. Journal of
Biomechanics 37: 623636.
Madden, L. R., D. J. Mortisen, et al. (2010). Proangiogenic scaffolds as functional templates for cardiac tissue
engineering. R. Langer. 107: 1521115216.
Sobral, J. M., S. G. Caridade, et al. (2011). Threedimensional plotted scaffolds with controlled pore size
gradients: Effect of scaffold geometry on mechanical performance and cell seeding efficiency. Acta Biomaterialia
7(3): 10091018.
Taboas, J. M., R. D. Maddox, et al. (2003). Indirect solid free
form fabrication of local and global porous, biomimetic
and composite 3D polymerceramic scaffolds. Biomaterials 24(1): 181194.
www.itksnap.org (2008). ITK-SNAP. 1.6.
Yeong, W.-Y., C.-K. Chua, et al. (2004). Rapid prototyping
in tissue engineering: challenges and potential. Trends in
Biotechnology 22(12): 643652.
Figure 6. Bio-fabrication of two consecutive aorta slices

divided with gradient of 53, 55, 57, 60, 61, 64, 65, 66, 68
and 70% (a) increasing porosity (b) decreasing porosity from
outer to inner feature.
method in this paper generates the porosity along the

scaffold architecture considering the feature geometry
and thus avoids such error. Furthermore, the proposed
methodology generates a continuous and interconnected tool path while achieving the designed porosity
in spite of the internal feature. But for conventional
structure, the presence of such internal feature forced
the fabrication tool path to jump or create motion
without deposition resulting degeneracy in deposited
material uniformity.
There has been some material over deposition along
the outer and inner edge, which is visible in these above
figures (Figure (46)). This is due to the sharp and
sudden changes in the tool-path direction along the
edge. This is due to the sharp and sudden changes
in the tool-path direction along the edge. Increasing
the feed rate particularly at the edges could alleviate this over deposition. Moreover due to the systems
backlash, few non-uniform and narrow sub-regions
have been generated. By using more precise system
a perfect fabrication is easily achievable. The biofabrication system uses a CNC-based motion system to
deposit bio-compatible material directly from the generated paths, and thus the fabricated structure closely
conforms to the designed scaffolds.
The proposed design algorithm generates the internal points of the designed scaffold sequentially. Thus
most motion control system can handle those points
with or without some simple modification and eventually follow the tool-path to achieve the desired
porosity.
77

Heterogeneous tissue scaffolds for spatiotemporally controlled

release kinetics
I.T. Ozbolat
A.K.M.B. Khoda, M. Marchany & J.A. Gardella

B. Koc
ABSTRACT: In this research, hybrid tissue scaffolds are modeled, designed and fabricated to control release
kinetics spatially and temporally for improved tissue regeneration. Heterogeneous porous tissue scaffolds with
varying characteristics to mediate the release of base material and enclosed biological modifiers are proposed
based on tissue engineering requirements. Firstly, a computer-aided biodegradation model is developed to simulate degradation process of micro-patterned polymeric membranes. Next, scaffolds are bio-mimetically designed
to control internal porous architecture with varying porosity. A new optimized internal architecture scheme is
developed to enhance fluid transport with continuous base material deposition plan. A sample scaffold encapsulated with microspheres is fabricated to explore the natural distribution of microspheres and a stochastic
distribution model of biological modifiers is adapted from the image based microsphere distribution model. This
study is extended for the development of hybrid scaffolds for spatial control of microspheres and base material
to synchronize the release kinetics with tissue regeneration. Finally, a pressure-assisted multi-chamber single
nozzle solid freeform fabrication (SFF) technique is utilized to fabricate hybrid scaffolds.
INTRODUCTION
spatial gradient to guide specific cues to the cellular

microenvironment.
In this paper, proteins, growth factors and cells etc.
are named with a general term biological modifiers.
Development of engineered tissue scaffolds with superior control over cell-biological modifier interactions
is still very much infancy. Advancing through heterogeneous multifold scaffolds with controlled release
fashion enables synchronization of regenerating tissue
with the release kinetics of loaded biological modifiers. This might be an engineering challenge and
promising approach for improved and efficient tissue
regeneration. The most critical limitations: the selection of proper biological modifier(s) incorporation,
precise control over concentration gradient and timing should be overcome. Hence, tissue scaffolds need
to be designed and fabricated in a way that biological modifiers should be incorporated and released in a
specific spatial and temporal orientation to mimic the
natural tissue regeneration process. Spatial and temporal control over heterogeneous tissue scaffolds can be
achieved by controlling three important parameters:
(i) degradation of scaffold base material, (ii) internal architecture with controlled fluid transport, and
(iii) distribution of scaffold base material and loaded
modifiers.
Current medical procedures aim to restore tissue function to patients with diseased or damaged tissues
through tissue transplantation and implants. Tissue
Engineering, an interdisciplinary field of biology,
biomaterials and engineering, is seeking to restore
tissue functions by developing engineered scaffolds
providing optimum environment for cell attachment
and growth, tissue regeneration, fluid movement and
structural integrity (Ozbolat, Marchany et al. 2009).
Engineered scaffold attempts to mimic the complexity
of both external and internal architecture of replaced
tissue in a way that optimal microenvironment is
designed for the cells to culture and develop into
tissues. Optimal microenvironment can be achieved
by signaling cellular activity through delivering vital
biological modifiers such as proteins, growth factors
and drugs. Release of these modifiers with spatial
and temporal gradient concentration mediates tissue
regeneration process. It develops a mental biology, in
which cells are guided by a mechanism with respect
to obtained spatial and directional cues (Sun, Silva
et al. 2010). As a result, scaffolds in tissue engineering
should be developed in way that they enable controlled
release scheme of biological modifiers with distinct
79
CONTROLLED DEGRADATION
Ficks second law as follows (Siepmann, Herrmann

et al. 2008):
Immediately after implantation into the host, properties of a scaffold prone to changes during the degradation process due to physical, chemical, mechanical and biological interactions between biomaterial
and the surrounding environment. This change may
induce different biological responses to cellular matrix
(Azevedo and Reis 2004). Controlling the degradation by varying the scaffold geometry in micro
scale could have a big impact on cell growth and
proliferation such as biological response, functional
and mechanical behavior, and the biological modifier
release rate. The functional behavior of the scaffold
and the amount of mechanical loading during regeneration of tissue structure change over time, and are
affected by the micro-scale geometry (Gomez 2007).
In addition, scaffolds can be designed in a way to
release biological modifier choreographed over a time
period to improve healing while viable tissue adheres
and grows into degrading construct (Anderson,
Rosenholm et al. 2008). Therefore, it is essential to
release encapsulated biological modifier in a controlled fashion since the incorporated modifiers need
to maintain their integrity to stimulate desirable cells to
achieve biological response and thereafter the degradation of the release system (Tessmar and Gopferich
2007). In this research, micro-patterned membranes
are used to control the cellular microenvironment and
allow cells to align themselves to the shape of geometrically modeled patterns. While cells proliferate
through the patterns, the degradation process proceeds
simultaneously leading changes in the microenvironment. To maintain appropriate microenvironment, the
degradation process needs to be controlled. In the following section, modeling of the degradation process is
presented.
The degradation process of polymeric biomaterials in vivo can be characterized as the penetration
of surrounding media through the polymer matrix.
A hydraulic degradation phase follows as chain scission in the polymer backbone to form small chains
(oligomers). Oligomers are media soluble fragments
which diffuse from polymer matrix into the media
(Azevedo and Reis 2004). During hydraulic degradation phase, diffusion of media through the polymer
takes crucial place as the accelerated media absorption
results in higher hydraulic degradation rate (Correlo,
Pinho et al. 2007). The enzymes in blood media break
small fragments into natural monomeric acids, which
are then excreted from human body following some
cyclic processes (Wang, Pan et al. 2008). The latter
phase is directly related with biological environment
and associated media properties. Based on our earlier
work (Marchany, Gardella et al. 2009), this research
assumes the degradation process mainly follows surface erosion mechanism due to the micro-scale feature
size of the polymeric membrane. Furthermore, modifier release is also assumed to be directly associated
with the diffusion process and the diffusion of media
through biomaterial network can be modeled by using
M is the media concentration at time t for a diffusion

distance d. Here D denotes the diffusion coefficient
which depends on several factors such as pressure,
temperature and the composition of the system (Sun
1996). In this research, a modified Finite Element
Method (FEM) is used to obtain solutions. The reader
is referred to our earlier work for details (Huebner,
Dewhirst et al. 2001).
3
INTERNAL ARCHITECTURE WITH

CONTROLLED FLUID FLOW
Natural tissues and organs are not homogenous rather

they have layers of tissue with one or more specific
function that generate natural functional gradients over
their structure (Leong, Chua et al. 2008). Thus, scaffolds with controlled variation in porosity over their
architectures could have significant impact on functional, mechanical and biological behavior of scaffolds
(Ozbolat, Khoda et al. 2009). In this section, a new
method has been presented to provide functional gradients by controlling the porosity distribution over
their complex architectures with an optimum interconnected network channel that provides improved
connectivity between each functional gradient region
to achieve enhanced permeability and improves transport of fluid through the structures. This enables
flow of the media through the internal regions and
mediates the release of loaded modifier in internal
regions.To mimic the functional gradient in porosity of
the replaced tissue, a porosity distribution (PD) function can be approximated and the scaffold area can
be discretized into n number of uniform regions with
respect to a maximum error constraint over a parametric domain u of the scaffold architecture. A porosity
level based on PD is determined for each of these n
regions, where the number of discretized region represents the resolution for gradient porosity (Fig. 1(a)
(b)). To achieve these uniform porosity regions Ri ,
i = 1, . . . , n a non-uniform-offsetting operation (Koc
and Lee 2001) is proposed to avoid error in distance
based offsetting for free-form surfaces and complex
shapes.
Then the original and newly generated surface is
sliced with a uniform slicing distance by a set of
intersection plane perpendicular to the longitudinal
direction and contours were generated by connecting
the intersection points between the planes. Such crosssectional contours might have self-intersecting loops
and overlaps based on the surface property and offsetting distance, and can be eliminated by outward
normal loop direction method proposed in Khoda et
al. (Khoda, Ozbolat et al. 2011) followed by a cleaning operation. The cleaned regions are then ready
80
Figure 1. (a) Discretization of the porosity distribution

function, (b) transverse view of a slice.
for porosity assignment based on the discritized PD

function.
For material deposition, cylindrical filament
assumption has been used as design and modeling
parameter common in SFF system (Khoda, Ozbolat
et al. 2011). Porosity in each region is then achieved
by determining the number of filaments and the distance between filaments. Since porosity in each region
is different so as the number of filament, and thus a
new toolpath design is proposed to ensure continuity and connectivity by a network of interconnected
channels and continuous filament deposition. The proposed continuous deposition-path plan would enhance
the nutrient flow and waste disposal through the internal channels network directly affecting permeability
of a scaffold (Agrawal, McKinney et al. 2000) and
at the same time would minimize the start and stop
interruption during solid free-form fabrication.
Figure 2. (a) A schematic of multi-chamber single nozzle assembly, (b) a dark field image showing a sampled
region (c) normality analysis and (d) stochastic distribution
of microspheres in a single filament.
3D motion control. A multi-nozzle assembly is developed to dispense a mixture of two different biomaterial
concentrations as shown in Fig. 2(a). Material flow and
concentration through mixture chamber is controlled
by regulating positive nozzle pressures. Sodium alginate, a hydrogel based biomaterial extensively used
in tissue engineering applications (Ozbolat and Koc
2010), is used to fabricate hybrid scaffolds.
While exact location of each microsphere cannot be controlled in dispensing systems, a stochastic approach needs to be developed to locate the
microspheres. Thus, statistical analysis is performed
over selected representative lengths of deposited filament through the nozzle system to explore the
distribution of microspheres over the cross-section
of filaments with respect to the filament boundary
(See Fig. 2(b)). Microsphere distribution over a randomly selected region (See Fig. 2(b)) is analyzed and
results are obtained as in Fig. 2(c). Normal distribution
is obtained with p-value < 0.05 showing statistical significance, in which microspheres are located mostly in
center regions. Then, a stochastic modeling approach
developed in our earlier work (Ozbolat and Koc 2010)
is used distribute biological modifiers in filaments
designed in CAD (See Fig. 2(d)).
CONTROLLED DISTRIBUTION OF
BIOLOGICAL MODIFIERS AND
BIOMATERIALS
Target delivery of biological modifiers in a spatiotemporal fashion to induce functional tissue repair can
be enabled by using the appropriate means such as
loading in a device or a carrier (Guldberg 2009).
Modeling of 3D freeform porous devices with localized control of encapsulated biological modifiers has
thus a potential to affect release kinetics of biological modifiers spatially in tissue engineering and drug
delivery platforms. Release systems in regenerative
medicine and tissue engineering should be designed
in way that they provide various profiles of modifier
release with distinct spatial gradient to guide specific
cues to the cellular microenvironment. Precise control of concentration and spatiotemporal gradient of
their presence improves the quality and quantity of
regenerating tissue.
In this research, biofabrication of tissue engineered
porous scaffolds is performed by utilizing a multinozzle dispensing system (Ozbolat and Koc 2010).
Fabrication process is biologically compatible. The
system runs in room temperature under low pressure
(05 psi) to reduce fluidic shear forces that can damage
incorporated biological modifiers or diminish their
active properties. The dispensing system is driven by a
RESULTS
5.1 Controlled degradation

In this research, Poly(lactic Acid) (PLA) is used to
fabricate the polymeric membrane.The proposed algorithm for degradation analysis is implemented using
Visual C++ and OpenGL library functions. As shown
in Fig. 3, degradation starts with etching process at the
corners and proceeds with a circular profile shown in
Fig. 3(a), and eventually completes in 2 days. A 3D
81
Figure 3. (a)(b) Top and side view of a degradation process simulation based on 6 hour time increments for a single
PLA block (c) a DIC image of the fabricated micro-patterned
membrane (d) daily DIC images for degradation of the fabricated micro-patterned membrane with uniform block size of
200 200 5 m.
Figure 4. Flowchart of the applied methodology from

reverse engineering through biofabrication.
simulation of the degraded block is also presented in

Fig. 3(b). Daily images of degraded micro-patterned
films are obtained using a differential interference
contrast microscope (DIC) with 5X and 10X dry objective lenses (Marchany, Gardella et al. 2009). Fig. 3(d)
depicts degradation results at 0h, 24h, 48h and 72h.
Degradation process starts with etching process immediately after immersing in the buffer solution at the
corners and proceeds through inner regions showing
a surface erosion mechanism. Images are taken from
different blocks on the film and blocks show variability in terms of degradation rate and intermediate
geometry. Some blocks indicate higher degradation
rates compared to the others and intermediate geometries possess differences. Variability in the degradation
results thus reveals a random nature of the process.
The results from the experimental analysis of degradation closely match the results from the developed
computer-aided biodegradation model.
One of the most challenging differences between
the model and the experimental result is the randomness part of the degradation process which is
assumed to be negligible in the developed computeraided degradation model. The results reveal that if
the feature size is reduced to nano or micro scale,
then degradation follows surface erosion as soon as
blocks are exposed to the solution and lasts in few
days, although same polymer shows relatively longer
degradation periods if the device size is relatively big
(VonBurkersroda, Schedl et al. 2002).
5.2
Figure 5. (a) Two consecutive fabricated vertebra slices

with five different porosity region of 84.6%, 83.6%, 79.3%,
78.1% and 76.2% (b) zoomed view for continuous and
inter-connected tool path (c) 3D fabricated vertebra with 18
slices.
laser scanner is used to generate scan data and converted into CAD model through Pixform software.
Then, the model is inputted into NURBS modeling software Rhinoceros 4.0. The desired number of
regions is generated and sliced. Next, the deposition
path planning method presented in Section 3, is used
to achieve the optimum interconnected and continuous toolpath. The optimization problem is solved using
ILOG Cplex 12 software and exported to CAD environment again to generate NC code for the toolpath.
The generated toolpath plan is then verified using
Flashcut software and exported into solid freeform
fabrication system to build the designed scaffold.
Overall applied methodology is also presented as a
flowchart in Fig. 4.
Finally, a single nozzle sodium alginate deposition
system is used to fabricate the model with nozzle
tips diameter of 150 m with controlled variational
porosity along the scaffold architecture shown in
Fig. 5.
The presented tool path optimization method
enhances the permeability of the designed vertebrae by
21.1% compare to that without toolpath optimization
(Khoda, Ozbolat et al. 2011).
Controlled internal architecture with enhanced

fluid flow
For implementation purpose, a lumbar vertebra solid

model is obtained by extracting the 3D geometry
through reserve engineering. Roland Picza LPX-60
82
All concentration levels follow similar release fashion starting with a burst release and continuing with
decreasing rate of release kinetics. Release of insulin
through 1% alginate concentration begins with 57%
burst release in 2h and finalizes around 18h. Release
of insulin through 1.5% alginate concentration however starts with 55% burst release and is completed in
24h. Increasing alginate concentration to 2% reduces
the burst release and reduces the burst release to 53%
at time 2h, in which release of insulin stops at 28h.
The slowest release kinetics is attained at 3% alginate
concentration in which 50% burst release is observed
in 2h and release of insulin finalizes at 32h, which is
almost double of total release time of insulin in 1%
alginate concentration. As a result, change in concentration affects the release kinetics. Increasing alginate
concentration slows down the release fashion. Based
on equation presented in (Ozbolat 2011), increase in
filament diameter also brings deceleration in release
kinetics that can be explained by the increased diffusion distance of biological modifiers through the
filament.
Figure 6. (a) 3D model in CAD based on (b) alginate

and loaded microsphere concentration profile (c) fabricated
porous scaffold with highlighted differentiation in microsphere concentration between 0.30.6% (w/v) (d) and dark
field image highlighting loaded microsphere.
In this paper, release kinetics through engineered tissue scaffolds is controlled spatiotemporally to enhance
tissue regeneration process. Spatial and temporal control enables synchronization of tissue regeneration
with the release kinetics to advance cell proliferation
and tissue healing. Spatiotemporal control of release
kinetics in this paper is achieved by controlling (i) the
degradation of scaffold base material, (ii) the structural geometry and the internal architecture, and (iii)
distribution of base material and loaded biological
modifiers.
Transition from incorporation of a single modifier to multiple modifiers for scaffolding applications
might be an engineering challenge and promising
approach for improved and efficient tissue regeneration. As a future direction, tissue scaffolds needs to be
designed and fabricated in a way that multiple proteins
or growth factors should be impregnated and released
in a specific spatial and temporal orientation to mimic
the natural tissue regeneration process.
Figure 7. (a) Diffusion constant change with change in alginate concentration and (b) release kinetics profile of insulin
under corresponding concentration change.
5.3
CONCLUSION
Controlled distribution of biological modifiers

and biomaterials
In this research, variation in material concentration

is characterized by incorporating color inks to differentiate the colors. Fig. 6(a) illustrates a sample
model developed in CAD based on alginate and loaded
microsphere concentration profile given in Fig. 6(b).
In this paper, user input concentration profile is
assumed to follow tissue engineering needs. Based on
the concentration profile given in Fig. 6(b), certain
number of microspheres are distributed stochastically
inside filaments in each region according to our earlier quantification study (Ozbolat and Koc 2010).
Fig. 6(c) illustrates a printed porous hybrid scaffold
and Fig. 6(d) highlights loaded microspheres in printed
alginate micro-filaments.
In this research, effect of alginate concentration
change over the release kinetics is analytically determined by applying the developed methodology in
Section 4. For sake of demonstration, release kinetics of insulin is analyzed while its treatment is critical
in diabetic wounds (Ozbolat 2011). Release kinetics of
insulin is analyzed analytically through 250 m alginate filament under various concentration values. As
the concentration of alginate increases, release kinetics slows down (See Fig. 7(b)). Release of loaded
insulin initializes with burst release and more than
half of the drug releases within the first 2 hours.
ACKNOWLEDGEMENT
The research was partially funded by U.S. Army Medical Research Grant #: W81XWH-0510401. The
authors would like to thank Dr. Robert Hard in the
Department of Pathology and Anatomical Sciences at
University at Buffalo for micro-scale dark field image.
REFERENCES
Agrawal, C. M., J. S. McKinney, et al. (2000). Effects of fluid
flow on the in vitro degradation kinetics of biodegradable
83
scaffolds for tissue engineering. Biomaterials 21(23):

24432452.
Anderson, J., J. Rosenholm, et al. (2008). Mesoporous silica:
An alternative diffusion controlled drug delivery system.
Topics in Multifunctional Biomaterials & Devices. N.
Ashammakhi. Oulu, Finland, University of Oulu: 119.
Azevedo, H. S. and R. L. Reis (2004). Understanding the
enzymatic degradation of biodegradable polymer and
strategies to control the degradation rate Biodegradable
Systems in Tissue Engineeing and Regenerative Medicine
R. L. Reis and J. S. Roman. Boca Raton FL, CRC:
177201.
Correlo, V. M., E. D. Pinho, et al. (2007). Water absorption
and degradation characteristics of chitosan-based polyster
and hhydroxyapatite composites. Macromol. Bioscience
7(3): 354363.
Gomez, C. (2007). A Unit Cell Based Multi-scale Modeling
and Design Approach for Tissue Engineered Scaffolds.
Mechanical Engineering Department, Drexel University
Guldberg, R. E. (2009). Spatiotemporal Delivery Strategies for Promoting Musculoskeletal Tissue Regeneration. Journal of Bone and Mineral Research 24(9):
15071511.
Huebner, K. H. H., D. L. Dewhirst, et al. (2001). Finite
Element Method for Engineers. NewYork, J. Wiley&Sons.
Khoda, A. K. M. B., I. T. Ozbolat, et al. (2011). Engineered
Tissue Scaffolds With Variational Porous Architecture.
Journal of Biomechanical Engineering 133(1): 011001.
Koc, B. and Y. S. Lee (2001). Non-uniform offsetting
and hollowing objects by using biarcs fitting for rapid
prototyping processes. Computers in Industry 47: 123.
Leong, K. F., C. K. Chua, et al. (2008). Engineering functionally graded tissue engineering scaffolds. Journal of
the Mechanical Behavior of Biomedical Materials 1(2):
140152.
Marchany, M., J. A. Gardella, et al. (2009). Determination
of Spatially Resolved Hydrolytic Degradation Kinetics
of a Micropatterned Bioresorbable Polyester Membrane.

7th International Conference on Secondary Ion Mass
Spectrometry, SIMS XVII. Toronto, Canada.
Ozbolat, I., A. Khoda, et al. (2009). Geometric Modeling of
Complex Tissue Engineering Scaffolds with Controlled
Porosity Distribution. Industrial Engineering Research
Conference, Miami, Fl.
Ozbolat, I. T. (2011). Tissue Scaffolds for Spatiotemporally
Controlled Release Kinetics. Department of Industrial
Engineering. Buffalo, University at Buffalo PhD.
Ozbolat, I. T. and B. Koc (2010). Modeling of Spatially
Controlled Biomolecules in Three-Dimensional Porous
Alginate Structures. Journal of Medical Devices 4(4):
041003.
Ozbolat, I. T., M. Marchany, et al. (2009). Feature-based
Design of Bio-degradable Micro-patterned Structures.
Journal of Computer-aided Design & Applications 6(5):
661671.
Siepmann, F., S. Herrmann, et al. (2008). A novel mathematical model quantifying drug release from lipid implants.
Journal of Controlled Release 128(3): 233240.
Sun, Q., E. Silva, et al. (2010). Sustained Release of Multiple Growth Factors from Injectable Polymeric System as
a Novel Therapeutic Approach Towards Angiogenesis.
Pharmaceutical Research 27(2): 267271.
Sun, Y. M. (1996). Sorption/desorption properties of water
vapor in poly(2-hydroxyethyl methacrylate): 2. Two-stage
sorption models. Polymer 37: 39213928.
Tessmar, K. J. and A. M. Gopferich (2007). Matrices
and scaffold for protein delivery in tissue engineering
Advanced Drug Delivery Reviews 59(45): 274291.
VonBurkersroda, F., L. Schedl, et al. (2002). Why degradable polymers undergo surface erosion or bulk erosion.
Biomaterials 23(21): 42214231.
Wang,Y., J. Pan, et al. (2008). A phenomenological model for
the degradation of biodegradable polymers. Biomaterials
29(23): 33933401.
84

Medical application of rapid prototyping in orthopedics surgical planning

C.B.L. Ulbrich & C.A.C. Zavaglia
Department of Materials Engineering, State University of Campinas, DEMA/FEM/UNICAMP, Brazil
T.P. Leivas & F. Teixeira

Clinical Hospital of the Medical School of the State University of So Paulo, HCFMUSP, Brazil
ABSTRACT: The use of rapid prototyping has become a useful tool to help and support medical activities. It
can be used as an auxiliary tool for the diagnosis of certain diseases and the development of complex surgical
procedures. The aim of this paper is to show the use of a Rapid Prototype model to help visualize the anatomical
structures in an orthopedics case. A virtual 3D model was created from Computer Tomography images using
the Invesalius software. After that, a physical model was created using a 3DP machine. The patient and the
medical team would discuss this surgical procedure using the images and the biomodel. The main purpose of the
prototype model was to introduce the idea of a new educational environment using RP to clarify the complexity
of this procedure for both doctors and patient.
1
INTRODUCTION
Orthopedic joint replacement is the surgical procedure of replacing a diseased and/or dysfunctional
human orthopedic joint with artificial components, so
as to restore function and mobility to that joint. The
most common joints targeted by this procedure are the
hips and knees.
As shown in Figure 1, implants can either be fixed
to the cavity of femoral by an interfacing layer of flexible polymeric cement like PMMA (Frost & Sullivan,
2005). Another option is to be fixed directly forced
(press-fitted) into the cavity of femoral (Kowalczyk,
2001).
However, in some cases, the region affected by the
disease must be simply removed because there is no
need for an implant.
Rapid prototype application in medicine is an
increasingly growing demand (Wohlers, 2008). With
this technology it is possible since the construction
of models for surgical planning to custom prostheses.
The aim of this paper is to show the use of a Rapid
Prototype model to help visualize the anatomical
structures in an orthopedic case.
Figure 1. Schematic of a total hip replacement.
significant enhancement by intravenous iodinated

contrast. A tomography image is shown in Figure 2.
2.2 Creating a virtual model using CAD systems

2
The patient was submitted to exam of computerized tomography multislice (Figure 3). For the exam,
an acquisition protocol was used using 1 mm for
increment between slices and 1 mm of thickness.
Data obtained in the exams (DICOM format) were
converted in three-dimensional using InVesalius software (CTI ProMED, Brazil). The software made
possible to isolate woven of the bone structure through
the segmentation for threshold to export them in a STL
file as showed in Figure 4.
2.1 Patient
Our patient is a 29-years-old male. Tomography report
indicates the presence of a heterogeneous mass with
soft tissue components, with an epicenter in the
region of the superior-posterior iliac right, extending to the adjacent musculature. The lesion measures
approximately 10.5 10.4 7.2 cm and showed no
85
Figure 2. Image of the CT examination report.
Figure 5. Virtual model.
Figure 6. Physical model (in construction) and after

finished.
Figure 3. DICOM images and multislice CT scanner.
Figure 7. Physical model with marking area to be removed.
Figure 4. InVesalius segmentation.
complexity of the case and the lack of available solutions in the market capable to assist the patients
specific needs.
2.3
2.4 Explaining the planned surgery to the patient
Creating a physical model
The STL file were edited using Magics 15.0 software

(Materialise, Belgium), in order to select just the study
area as showed in Figure 5.
A physical model was rapid manufactured using
rapid prototype as showed in Figure 6. This model was
created on Z Printer 510 (manufactured by Z Corporation). Materials used were ZP 130 (powder), ZB58
(binder) and Z Bond (resin).
The option for the use of three-dimensional technologies for the project and production was due to
Surgical planning can be used in several steps: decision

making of surgical procedures, selection of implants
(type, quantity, dimensions, position and attachment
points), positioning and fixation, etc.
In this case, the model was used to explain to the
patient the execution of the surgical procedure and
which areas would be removed (Figure 7).
The surgical planning model can also be used to
assist in teaching and training of doctors and specialists in graduate and extension courses, among others.
86
2.6
Patient three months after surgery
Three months after surgery, the patient was again subjected to further examination. Figure 9 illustrates the
results obtained with surgery. The region removed, is
quite consistent with the model plan.
3
CONCLUSIONS
The main purpose of the prototype model was to introduce the idea of a new educational environment using
RP to clarify the complexity of this procedure for both
doctors and patient.
The use of a surgical planning model for surgical
planning can provide benefits such as: reduce duration of surgery, improve its accuracy, decrease patient
morbidity and hence improve quality of surgery.
The model created virtually proved to be very consistent when compared to the final outcome after
surgery.
Various technologies (such as rapid prototyping and
virtual simulation) can also be used to disseminate
knowledge through courses to training doctors and
specialists in graduate and extension courses, among
others.
Figure 8. Virtual simulation.
ACKNOWLEDGMENTS
Authors would like to acknowledge (INCT-BIOFABRIS)
and CNPq for financial support.
REFERENCES
Frost & Sullivan. 2005. U.S. & Asian Markets for Orthopedic
Joint Replacement 2005. Reference A75154.
Gelalis L.D, Xenakis T.A, Hantes M., Vartziotis K., Soucacos
P.N. 2001. Three-dimensional computerized selection of
hip prostheses in patients with congenital dislocated hips.
Kowalczyk, P. 2001. Design optimization of cementless
femoral hip prostheses using finite element analysis.
J Biomech Eng 123(5):396402.
Wohlers, T. 2008. Wohlers Report 2008. State of the Industry,
Annual Worldwide Progress Report, ISBN 09754429
45.
Figure 9. Patient image three months after surgery.
2.5 Virtual simulation for the surgical planning

To better understand the procedure, a virtual simulation was also performed. This simulation is shown
in Figure 8. It has been removed the region marked
previously for the virtual model.
87

Polycaprolactone-based scaffold plus BMP-2 in a sheep thoracic spine

fusion model
M. Yong, F. Melchels, C. Vaquette, D. Hutmacher & C. Adam
Queensland University of Technology, Brisbane, Queensland, Australia
M. Domingos & P. Bartolo

Polytechnic University of Leiria, Leiria, Portugal
ABSTRACT: We report the application of a novel scaffold design in a sheep thoracic spine model for spine
deformity correction. The combination of the calcium-phosphate coated polycaprolactone scaffolds with recombinant human bone morphogenic protein-2 are intended as a future bone graft substitute in ensuring the stability
of bony intervertebral fusion. A solid free-form fabrication process based on melt extrusion has been utilized in
the manufacturing of these scaffolds.
1
1.1
INTRODUCTION
Background
2.1 Anatomical considerations

Sheep (ovis aries) have been widely used in a large
number of interbody spinal fusion models (Cunningham 1998a, b; Sandhu 2000, c; Sawyer 2009, d; Abbah
et al. 2009). The physical size of the sheep spine is
deemed sufficient to allow spinal surgery to be readily
carried out and to allow for assessment of the success
of the study using biomechanical testing, radiological and histological analysis. In addition, sheep are a
large animal model which simulates an environment
representative of an a dolescent child, while securing
optimal application and fixation techniques closely
corresponding to that in the human spine (Easley
2008). Sheep have also been established for use in both
open chest spinal surgery as well as minimally invasive
endoscopic assisted surgery (Cunningham 1998).
Anterior spinal surgery is a well-recognized and effective approach for adolescent idiopathic scoliosis correction (Dubousset 2001a, b; Lowenstein et al. 2007).
The success of anterior scoliosis surgery depends
on achieving a solid bony fusion between adjacent
vertebrae after the intervertebral discs have been
surgically cleared and the disc spaces filled with
graft material. The golden standard for bone grafting in spinal fusion surgery is autograft (host graft
material), however limited availability and donor site
morbidity make synthetic alternatives to autograft
desirable.
Current research focuses on the development of
synthetic scaffolds in combination with growth factors such as recombinant human bone morphogenic
protein-2 (rhBMP-2) to achieve solid bony fusion following scoliosis surgery (Sandhu 2000a, b; Sawyer
2009, c; Abbah et al. 2009). To date there are no studies examining the use of such biodegradable implants
in a sheep thoracic spine model.
2.2 Scaffold design

The scaffold design has been conceptualized through
cadaveric dissections. Solid Acrylonitrile Butadiene
Styrene (ABS) prototypes of the scaffold were first
designed and manufactured in different thicknesses
by fused deposition modeling (Dimension SST 768),
and after trial implantations of the spinal thoracic
levels of T5/6, T7/8 and T9/10, the optimal scaffold
geometry was confirmed and a thickness of 2.5 mm
was decided to be most appropriate for this animal
model. The final scaffold design was based on a
0 90 lay-down pattern plus scaffold contour to
confer additional strength for surgical handling and
implantation of the prepared disc space. The scaffolds (Figure 1) were fabricated using PCL and a
Dual_BioExtruder, a computer-controlled extrusionbased additive manufacturing device developed
1.2 Aim
The aim of the current project is to design and fabricate biodegradable polycaprolactone (PCL) scaffolds
by using a solid free-form fabrication process based
on melt extrusion. These scaffolds are coated with
a biomimetic calcium phosphate (CaP) layer which
actively promotes bone ingrowth and regeneration.
Following functionalization of these scaffolds with
rhBMP-2, surgical implantation is undertaken at one
of the predefined thoracic spinal levels of either T5/6,
T7/8 or T9/10.
89
Figure 1. CT 40 scan of a fabricated scaffold.
in-house at the Polytechnic University of Leiria, Portugal (Domingos et al. 2010). The semicircular shape
conforms to the cleared anterior intervertebral disc
space ensuring a low-profiled construct under compression.
2.3
Figure 2. Representative force-strain diagram of compression test on uncoated scaffold.
Surgical procedure
In this in vivo sheep study, three thoracic intervertebral spaces (T5/6, T7/8, T9/10) in each animal
receive either (i) PCL + CaP scaffold + rhBMP-2
(ii) PCL + CaP scaffold alone, or (iii) rib head autograft. The treated intervertebral disc spaces are stabilized with a 5.5 mm titanium rod secured with two
vertebral screws.
3
RESULTS
3.1 MicroCT analysis

The porosity of uncoated scaffolds measured
55.6 0.9% with an average pore size of 406 6 m
and average strut diameter of 328 6 m. The calcium phosphate coating of the scaffolds conferred
an increase in scaffold strut size (340 4 m) which
resulted in a decrease of pore size (386 5 m)
and porosity (49.9 2.0%). The CT was also
used to determine the cross-sectional surface area
(103.3 mm2 ) that was used for the compressive modulus and yield stress calculations. A representative 3D
rendered image of a scanned scaffold is presented in
Figure 1.
Figure 3. SEM of a coated scaffold (sectioned) demonstrating calcium phosphate coating of individual strut filaments.
and consistency were largely preserved. This observation is important as the preservation of the scaffolds
overall shape once implanted surgically within the
sheeps intervertebral disc space and placed under
compression is a condition sine quanon.
3.3 Scanning electron microscopy (SEM)
Biomimetic properties of the calcium phosphate
coating of the scaffolds actively promote bone regeneration. The coating of these scaffolds involve immersion in concentrated simulated body fluid (10) and
scaffold surface activation (Yang et al. 2008). Confirmation of the coating is seen with SEM presented
in Figure 3. Preliminary results show a heterogeneous
coating whereby there is a clear difference between
lightly coated outer layers of the scaffold contrasting with well-rounded coatings of central (inner) strut
filaments.
3.2 Compressive testing of scaffolds

From the compression tests, no influence of the coating procedure on the mechanical properties could be
observed. Figure 2. depicts a representative forcestrain diagram, showing an initial alignment phase
(up to 10 %) followed by elastic deformation with a
stiffness value of 9.41.3 N/% or an elastic modulus
of 9.1 1.25 MPa. The scaffolds yielded at 25 1%
strain and 117 14 N (1.14 0.13 MPa), after which
permanent deformation progressively increases as the
load is increased.
Macroscopic observation of compressed scaffolds
revealed thinning of the construct by shearing between
the deposited layers of material, while both the internal pore architecture and the overall scaffold shape
3.4 Fusion assessments

To date, two pilot surgeries with implantation of
earlier scaffold prototypes have been successfully
90
critical to a stable biomechanical environment for

vascularization and bone remodeling. This has been
successfully demonstrated by Abbah et al. in an analogous porcine lumbar interbody fusion model whereby
complete bony fusion was seen as early as 3 months
with advanced bone remodeling at 6 months (Abbah
et al. 2009).
A scaffold pore size of 406 +/6 microns with
100% pore interconnectivity fulfills the required minimum of 300 micron pore size recommended for
sufficient vascularization of tissue-engineered graft
(Karageorgiou et al. 2005). The scaffold architecture consists of a honeycomb configuration which is
a regular two-dimensional array of polygonal pores,
each defined by a wall shared between adjacent pores
(Gibson et al. 1997). Structurally this consists of
a 90 degree lay-down pattern which offers a hexagonal pore architectural pattern conferring desirable
physiological and mechanical properties by way of
combining high stiffness and strength with open vertical pore channels that facilitate bone ingrowth. Yeo
et al. in an in vivo application of scaffolds fabricated
to similar dimensions matched closely the exhibited
compressive strength and modulus of cancellous bone
confirming its potential as a bone graft substitute
(Yeo et al. 2008).
Calcium phosphate coating of these PCL scaffolds
is intended to enhance its bioactivity hence promoting
bone formation. The external scaffold geometry has
been customized to allow for a low-profiled construct
confirming to the anatomical landmarks of the spine.
This aims at a greater integration and guided spinal
fusion. (Hutmacher et al. 2000)
Figure 4. PCL-based scaffold in situ within a predefined

intervertebral disc space. Internal fixation with a 5.5 mm titanium rod and two vertebral screws stabilize the functional
spinal unit. The 5.5 mm titanium rod has been removed for
illustrative purposes.
undertaken. Triple bone fusion assessments

(i.e. biomechanical testing, radiological imaging and
histological analysis) of individual functional spinal
units at the 6-month mark is currently underway. A
functional spinal unit (FSU) consists of an instrumented intervertebral disc level as well as the adjoining top and bottom vertebrae. (See Figure 4).
DISCUSSION
Bony fusion is essential for long-term stability of

instrumented spinal segments in the setting of scoliosis
deformity correction.
Increasingly being studied are biologically active
substances intended to extend, enhance, or even
replace autologous graft. The use of rhBMP-2 as an
osteoinductive implant component has been on the
increase because of its proven potency in vivo (Burkus
2004).
The biomedical application of these PCL-based
scaffolds aims at providing a structural support that
promotes the repair and regeneration of tissues in
combination of living cells and biologically active
molecules. Bone formation is actively guided with the
ensuing cell colonization, migration, growth and differentiation. This forms the basis of a viable tissue
engineered construct (TEC) (Hutmacher 2007).
Scaffolds are required to provide sufficient initial mechanical strength and stiffness to substitute for
the mechanical function of bone. This permits cell
migration and population of the scaffold in vivo. In
addition, the scaffold material should be adequately
robust to resist deformation upon cell infiltration as
well as wound contraction forces in vivo. An internal fixation construct stabilises the instrumented disc
space thus reduces the mechanical role of the scaffold
in situ. This maintains sufficient structural integrity
CONCLUSION
The combination of biologics and scaffold engineering represent a novel approach to promoting bony
fusion in the setting of thoracic spine deformity correction.Application of computer-controlled extrusionbased additive manufacturing devices pave the way in
customization of future spinal bone graft substitutes.
REFERENCES
Abbah, S.A. 2009. Biological performance of a polycaprolactone-based scaffold used as fusion cage device in a large
animal model of spinal reconstructive surgery. Biomaterials 30(28): 50865093.
Burkus, J. K. Bone morphogenetic proteins in anterior lumbar
interbody fusion: old techniques and new technologies.
Invited submission from the Joint section meeting on disorders of the spine and peripheral nerves. J Neurosurg
Spine 1(3): 254260.
Cunningham, B. W. 1998. Video-assisted thoracoscopic
surgery versus open thoracotomy for anterior thoracic
spinal fusion. A comparative radiographic, biomechanical
and histological analysis in a sheep model. Spine (Phila
Pa 1976) 23(12): 13331340.
Domingos, M. et al. 2010. Innovative developments in design
and manufacturing Advanced research in virtual and
91
rapid prototyping, ed. Boca Raton: Crc Press-Taylor &

Francis Group: 6773.
Dubousset, J. 2001. Scoliosis and its pathophysiology: do we
understand it? Spine (Phila Pa 1976) 26(9): 1001
Easley, N. E. et al. 2008. Biomechanical and radiographic
evaluation of an ovine model for the human lumbar spine.
Proc Inst Mech Eng H 222(6): 915922.
Gibson, L. J. et al. 1997. Cellular Solids Structure and
Properties. Cambridge University Press: New York.
Hutmacher, D. W. 2000. Scaffolds in tissue engineering bone
and cartilage. Biomaterials 21(24): 25292543.
Hutmacher, D. W. 2007. State of the art and future directions
of scaffold-based bone engineering from a biomaterials
perspective. J Tissue Eng Regen Med 1: 245260.
Karageorgiou, V. et al. 2005. Porosity of 3D biomaterial scaffolds and osteogenesis. Biomaterials 26:
547491.
Lowenstein, J. E. et al. 2007. Coronal and sagittal plane

correction in adolescent idiopathic scoliosis: a comparison between all pedicle screw versus hybrid thoracic hook
lumbar screw constructs. Spine (Phila Pa 1976) 32(4):
448452.
Sandhu, H. S. 2000. Anterior lumbar interbody fusion with
osteoinductive growth factors. Clin Orthop Relat Res 317:
254260.
Sawyer, A. A. 2009. The stimulation of healing within a rat
calvarial defect by mPCL-TCP/collagen scaffolds loaded
with rhBMP-2. Biomaterials 30(13): 24792488.
Yang, F. et al. 2008. Biomimetic calcium phosphate coating
on electrospun poly (e - caprolactone) scaffolds for bone
tissue engineering. Chem Eng J 137: 154161.
Yeo, A. et al. 2008. The degradation profile of novel, bioresorbable PCL-TCP scaffolds: an in vitro and in vivo study.
J Biomed Mater Res A 84: 208218.
92

Development of functional graded device of PCL/PG by Selective Laser

Sintering for drug delivery applications
G.V. Salmoria
Laboratrio de Engenharia Biomecnica, Hospital Universitrio, Universidade Federal de Santa Catarina,
Florianpolis-SC-Brasil
Laboratrio CIMJECT, Dep. Engenharia Mecnica, Universidade Federal de Santa Catarina,
P. Klauss & K. Zepon

Laboratrio de Engenharia Biomecnica, Hospital Universitrio, Universidade Federal de Santa Catarina,
L.A. Kanis
Grupo de Desenvolvimento em Tecnologia Farmaceutica, Universidade do Sul de Santa Catarina UNISUL,
Tubaro SC Brazil
C.R.M. Roesler
ABSTRACT: Drug delivery device is a technology employed to conduct the drug to a specific site of the body,
which should be released and absorbed. Reservoir type systems are devices in which the drug is in a nucleus
isolated from the outside by a polymeric thin layer, which will regulate a process of functional gradient diffusion.
In this work two drug delivery devices reservoir type were developed using Polycaprolactone for the release of
progesterone, using the technique of Selective Laser Sintering to investigate the influence of the functional
gradient concentration promote during manufacturing in the drug delivery profile. Two types of gradient devices
were fabricated, the first containing only the polycaprolactone polymer on the walls, called R, and the second
containing polycaprolactone and 15% of progesterone on the walls. Both reservoirs were filled with 40 mg of
progesterone at this core. The study of degradation was conducted to assess their influence on the release process.
The results of drug delivery showed that the amount of drug released for both reservoirs was linear with time,
featuring a zero order release kinetic. The highest mass loss and the addition of drug in the wall of the reservoir
R* quickly provide the drug to the middle. The feasibility of building three-dimensional parts using the SLS
technique, allowed the construction of reservoir type devices for the functional controlled release of the drug.
1
INTRODUCTION
years, the implant has 4 cm long and 2 mm diameter

(Adams and Beal, 2009).
Rapid Prototyping is a generic term for a group of
techniques that can generate a physical model directly
from computer aided design data. It is an additive process in which each part is constructed in a
layer by layer until the final geometry is obtained
(Jacobs, 1992; Jacobs, 1999). This type of construction
allows to obtain three-dimensional parts with complex
geometries with high accuracy and with varied sizes
and shapes (King e Tansey, 2003). Selective Laser Sintering is a Rapid Prototype process that creates 3D
objects, using a infrared laser beams to process powder materials (Jacobs, 1992; Yeong et al. 2004; Hur
et al. 2001; Gibson and Shi, 1997; Salmoria et al,.
2007). The structure and properties of the parts to be
manufactured by selective laser sintering should be
Drug delivery device is a technology employed to conduct the drug to a specific site of the body, which
should be released and absorbed (Ansel et al., 2000).
Reservoir type systems are devices in which the drug
(solid or liquid) is in a nucleus isolated from the
outside by a polymeric thin layer (Massod, 2007;
Florence e Atwood, 2003), that will regulate a process of functional gradient diffusion. In this type of
system, high concentration of drug inside the device is
essential to maintain a constant concentration gradient
across the membrane. (Bajpai et al., 2008). The reservoir type systems are commonly used in subcutaneous
contraceptive delivery systems. As an example, subcutaneous rod device of EVA copolymer containing
etonogestrel is designed for hormone release in three
93
were shaken horizontally (60 Hz) and kept at a temperature of 37 C. Every 48 hours withdrew from
all the liquid from bottles and filled again with
20 mL of the buffer solution. The absorbance of each
solution was determined by UV spectrophotometry
(246 nm) (HITACHI, model 2010, Japan), using the
equation (1):
considered according to their application. The level of

control over the microstructure and porosity of the SLS
parts depends on the process parameters (laser power,
scan speed and spot diameter of the laser beam and bed
temperature), particularly the powder properties, since
these can influence the parameters. Particle shape and
size distribution influences the powder packing density, while the melting flow behavior and the thermal
stability define the laser power and scan speed (Jacobs,
1992; Salmoria et al,. 2007)
In recent years the SLS process has shown
great prominence in the medical field and several
researchers have conducted studies showing a wide
diversity of materials and applications, such as the
manufacture of drug delivery devices (Low et al.,
2001; Cheach et al., 2002; Williams et al., 2005; Leong
et al., 2007). In this work two drug delivery devices
reservoir type were developed using Polycaprolactone
for the release of progesterone, using the technique
of Selective Laser Sintering to investigate the influence of the functional gradient concentration promote
during manufacturing in the drug delivery profile.
where Q is the amount of drug released in mg, fd is the

dilution factor (20 mL) and the Abs is a absorbance.
To determine the mass loss, used the same samples
contained in buffer pH 7.4 for drug release. After the
conclusion of the tests, the samples were dried on filter
paper in the vacuum oven and temperature 36 C until
constant weight. The mass loss was determined by the
difference between the masses of the parts before and
after testing release
In this work two different devices (type reservoir)
were constructed: 1) preparation of a rectangular piece
with pure PCL and pure progesterone in the nucleus
(center), 2) elaboration of a rectangular piece composed on the walls, base and top by PCL with 15%
of progesterone and pure progesterone in the nucleus
(center). In both reservoir, primarily was fabricated the
base and walls, with 0.144 J/mm2 energy density of the
laser and the after the top with 0.112 J/mm2 energy
density of the laser, to avoid dimensional deformation. The speed of the laser beam was used 50 mm/s.
The progesterone (40 mg) was added to the center of
the device after fabrication by SLS. The ends of the
devices were sealed by welding plates of polystyrene
(PS). min.
EXPERIMENTAL
The drug selected is Progesterone; a hormone

lipophilic used as contraceptive, post-menopausal
therapies and treatment of endometrial cancer. It has
a melting point 128132 C and molecular weight
314.46 g/mol. The progesterone was manufactured
by IFFECT CHEMPHAR (PRC) and provided by
Henrifarma Products Chemicals and Pharmaceuticals.
The powder polymer used in this study was Polycaprolactone, supplied by the Sigma-Aldrich, has a
molar mass around 70,00090,000 g/mol, the melting
temperature of 60 C and melt flow rate 1.00 g/10 min.
The particles size was 150212 m.
The mixing process of powders, for manufacture of
devices containing progesterone on the walls, was held
proportionally up to a total of 85% polymer and 15%
drug. For a better homogenization of the mixture, the
recipient, containing the powder, was inserted into a
cylinder connected to a mechanical mixer responsible
for to rotate at a constant speed of about 105 rpm, a
total of 2 hours and 30 minutes.
The Selective Laser Sintering equipment used in
this work was a prototype developed in CIMJECT
laboratory belonging to the Department of Mechanical Engineering of the Federal University of Santa
Catarina-UFSC.
The morphology of the device specimens and
the cryogenic fracture surface were examined with
a scanning electron microscope (SEM) using two
equipment XL 30 Phillips e Jeol JSM-6390LV. The
specimens were coated with gold in a Bal-Tec Sputter
Coater SCD005.
For measures of drug release, the reservoir devices
were placed in bottles with 20 mL of phosphate buffer
pH 7.4, properly sealed and identified. The bottles
3.1 Manufacturing of reservoir type device

Two models of reservoir type device were built. The
first, called R, was fabricated using the polycaprolactone polymer on the walls, base and top. The second,
called R*, was fabricated contained 85% of polycaprolactone polymer and 15% of progesterone on the walls.
Both contained 40 mg of progesterone into its core.
The proposed schemes are shown in Figure 1.
Figure 1. Schematic of reservoir type devices (R and R*).
Figure 2. Design for manufacturing the devices.
94
The specimens were built in the SLS equipment

from the design shown in Figure 2. The SLS program
sends the information that the laser must draw the
regions that are in black. First the base was designed,
after the walls and finally the top (same base design).
The Figure 3 shows the images of the device R
where (a) is top and (b) the view of the core with the
measurements of each wall.
The dimensions measured of the specimens R
showed higher values than project, which may be associated with the sintering of adjacent particles due to
heat transfer during sintering. The thicker the top is
due to heat transfer to the powder located below.
The Figure 4 shows the micrographs of the surfaces
(left) and cryogenic fracture (right) of device R. The
specimen sintered maintained a structural homogeneity, coalescence of particles and interconnected pores
distributed in the sintered structures.
After the devices fabricated, 40 mg of progesterone
were placed into the device and after the side were
sealed with PS.
3.2
The highest percentage of mass loss presented by

the reservoir R*, for a total of days below than reservoir R (49 and 78 days, respectively) is attributed to
the presence of progesterone in the walls that promotes the amorphous form, favoring the diffusion of
water into the matrix, which further eases the process
of hydrolysis of ester linkages, promoting the erosion
of the device. In the literature is that progesterone has
high solubility in PCL, about 0.02 g/g (Chang et al.,
2005) and can act as a plasticizer, lodging between the
polymer chains, and thus, reduce crystallinity.
3.3
Functional drug delivery for gradient reservoirs

of the polycaprolactone with progesterone
The Figure 7 shows the progesterone release profile

from the reservoirs type devices R and R*. The test
Determination of mass loss of the reservoirs

of polycaprolactone (R e R*)
The Figures 5 and 6 show the micrographs of the base

and top of the devices R (with 40 mg of progesterone in
the core) after 78 days in phosphate buffer (pH 7.4) and
R* (with 15% of progesterone on the walls and 40 mg
in the core) after 49 days. The average percentage of
mass loss for R and R* was 1.82% 0.09 (SD) and
2.70% 0.35 (SD), respectively.
Figure 4. Micrographs of the surfaces (left) and cryogenic

fracture (right) showing the (a) base, (b) wall and (c) top of
the reservoir type device R.
Figure 5. Micrographs of base and top surface of the reservoir type device (R), with increases of 30 and 100 times, after
78 days, in phosphate buffer (pH 7,4), at 37 C.
Figure 3. Images of the reservoir type device showed (a)

top (b) view of the core.
95
Table 1. Values of correlation coefficients and release

constants of the reservoirs R and R*.
Reservoir
R
R*
Correlation
Coeficient (r)
Release Constant
(g/cm/dia)
0.999
0.997
21.28
29.52
this phase occurs initially the diffusion of water until

the core of the device, a process that can occur through
the pores, but mainly by diffusion through the polymer
matrix, followed by dissolution of the drug and the
formation of a saturated solution. From this moment
the drug begins to diffuse from the center of the device
to the outer region initiating drug release, increasing
the release rate up to saturation of the all device.
In the second phase, the release rate constant
observed is maintained by saturated solution of drug
in the core of the reservoir, which ensures a constant concentration on the surface of the reservoir. The
maintenance of drug concentration on the surface of
the reservoir is an indispensable factor to ensure zero
order release kinetics (Ma et al., 2006; Siepman and
Siepman, 2008).
The Table 1 shows the values of the release constants, in g/cm/day, of the reservoirs R and R* and
the correlation coefficients to evaluation the linearity
of the release process. For R the values were calculated
from the stabilization of the rate of release (29 days).
The amount of drug released for both reservoirs
was linear with time, featuring a zero order release
kinetic (Costa, 2002). The release rate of progesterone increased by approximately 39% in R* device,
when compared with the device R, fact associated
with the addition of progesterone in the wall of the
device. The values of constant release for R was
21.28 g/cm/day and R* was 29.52 g/cm/day. Based
on these results and compared with several commercial devices such as Norplant that has length 3 cm
(Ma et al., 2006), we can increase the amount of drug
released by changing the length of the device up to
this size, reaching a maximum of 63 and 90g/day
of progesterone released. These results are similar
to that achieved by the commercial Progestasert ,
which release 65g per day. This intrauterine implant,
contains 38 mg of progesterone and provides controlled release by non-biodegradable membranes poly
(ethylene-co-vinyl acetate) (Villanova et al., 2010).
Figure 6. Micrographs of base and top surface of the reservoir type device (R*), with increases of 30 and 100 times,
after 49 days, in phosphate buffer (pH 7,4), at 37 C.
Figure 7. Progesterone release profile of reservoirs type

device (R e R*).
for R was conducted for 78 days and for R* for 49

days.
The Figure 7 shows that for both systems studied,
the cumulative amount of drug released to the environment increases with time. The R* system released 13%
of the total amount of drug incorporated in 49 days,
while the system R released 7% in 78 days. The largest
amount of drug released from R* can be explained
by the addition of drug in the wall of the reservoir,
which provides fast the drug to the medium. As shown
above, the reservoirs R and R* had a total mass loss of
2.70% 0.09 and 1.82% 0.35, respectively, which
may also have influenced the higher amount of drug
released by R*.
In the progesterone release profile from the device
R are observed two distinct phases of release. In the
first phase, non-linear, the results show a progressive
increase in the rate of drug release up to 290 day.
After this period, the device reaches a steady state flow
where the release becomes constant.
The behavior observed in the first phase of release
from R can be explained by the need for diffusion of
progesterone from the device core to the surface. In
CONCLUSIONS
The feasibility of building three-dimensional parts

using the SLS technique, allowed the construction of
reservoir type devices for the functional controlled
release of the drug. In tests of the progesterone release
from gradient reservoir type devices, which were
called R (40 mg of the Progesterone in the core) and R*
(containing 15% progesterone in the walls and 40 mg
in the core) showed that the amount of drug released
96
Costa, P.J.C. (2002). Avaliao in vitro da lioequivalncia de formulaes farmacuticas. Revista Brasileira de
Cincias Farmacuticas, 38(2).
Gibson, I., Shi, D. (1997). Rapid Prototyping Journal, 3,
129136.
Hur, S. M., Choi, K. H., Lee, S. H., Chang, P. K.
(2001). Journal of Materials Processing Technology, 112,
236243.
King, D., e Tansey, T. (2003) Rapid tooling: selective laser sintering injection tooling. Journal of Materials Processing
Technology, 132, 4248.
Jacobs, P.F. (1999). From Rapid Prototyping to Rapid
Tooling, ASME, New York.
Leong, K.F., Wiria, F.E., Chua, C.K., Li, S.H. (2007). Characterization of a poly-epsilon-caprolactone polymeric drug
delivery device built by selective laser sintering. Biomed
Mater Eng., 17(3), 14757.
Low, K.H., Leong, K.F., Chua, C.K., Du, Z.H., Cheach, C.M.
(2001) Characterization of SLS parts for drug delivery
devices. Rapid Prototyping Journal, 7(5), 262267.
Ma, G., Song, C., Sun, H., Yang, J., Leng, X. (2006) A
biodegradable levonorgestrel-releasing implant made of
PCL/F68 compound as tested in rats and dogs. Contraception, 74, 141147.
Massod, S.H. (2007) Application of fused deposition modelling in controlled drug delivery devices. Assembly
Automation, 27(3), 32522.
Salmoria, G.V., Leite, J.L., Ahrens, C.H., Lago, A., Pires,
A.T.N. (2007). Polymer Testing, 26, 361368.
Salmoria, G. V., Ahrens, C. H., Klauss, P., Paggi, R. A.,
Lago, A. (2007). Rapid Manufacturing of Polyethylene
Parts With Controlled Pore Size Gradients Using Selective
Laser Sintering, Materials Research, 10(2), 211214.
Siepman, J., Siepman, F. (2008). Mathematicl modeling of
grug delivery. International Journal of Pharmaceutical,
364, 328343.
Yeong, W. Y., Chua, C.K., Leong, K. F., Chandrasekaran,
M. (2004) Rapid Prototyping in Tissue Engineering:
Chalenges and Potentional. TRENDS in Biotechnology,
22, 643652.
Williams J.M., Adewunmi, A., Schek, R.M., Flanagan, C.L.,
Krebsbach, P.H., Feinberg, S.E., Hollister, S.J., Das, S.
(2005). Bone tissue engineering using polycaprolactone
scaffolds fabricated via selective laser sintering. Biomaterials, 26, 48174827.
for both reservoirs was linear with time, featuring a

zero order release kinetic. The highest mass loss and
the addition of drug in the wall of the reservoir R*
quickly provide the drug to the middle, justifying the
higher amount of progesterone released, compared to
the reservoir R. The design and manufacturing of a
gradient reservoir with 3 different regions concentration of progesterone (i.e.: PCL, PCL/15%PG, and PG)
are in route to investigate the influence in the gradient
concentration and delivery profile to improve the modeling of controlled functional gradient drug delivery
devices.
ACKNOWLEDGEMENTS
The authors would like to thank FAPESC, CNPq and
AEB (Brazilian Aerospace Agency) for the financial
support.
REFERENCES
Ansel, H.C., Popovich, N.G., Allen Jr., L.V. Formas farmacuticas & Sistemas de liberao de frmacos. 6 ed. So
Paulo: Editorial Premier, 2000.
Adams, K. e Beal, M.W. (2009). Implanon: A Review
of the Literature With Recommendations for Clinical
Management. Journal of Midwifery & Womens Health,
54(2).
Bajpai, A.K., Shukla, S.K., Bhanu, S., Kankane, S. (2008).
Responsive polymers in controlled drug delivery. Progress
in Polymer Science, 33, 10881118.
Cheah, C.M., Leong, K.F, Chua, C.K., Low, K.H., Quek,
H.S. (2002). Characterization of microfeatures in selective laser sintered drug delivery devices. Proc Inst Mech
Eng H. 216(6), 369383.
Chang, H.I., Williamson, M.R., Perrie, Y., Coombes, A.G.
(2005). A Precipitation casting of droag-loaded microporous PCL matrices: Incorporation of progesterone
by co-dissolution. Journal of Controlled Release, 106,
263272.
97

Flexible PCL tube scaffolds by winding of micro-extruded filaments

K. Ragaert & L. Cardon
CPMT group, Faculty of Applied Engineering Sciences, University College Ghent, Ghent, Belgium and Department
of Materials Science & Engineering, Ghent University, Ghent, Belgium
J. Degrieck
Department of Materials Science & Engineering, Ghent University, Ghent, Belgium
ABSTRACT: An important requirement for tissue engineering scaffolds is matching of the functional mechanical properties to their natural tissue counterpart. Specifically for arteries this comprises the elastic response of
the vessel wall to blood pressure. Human aorta has a low elastic modulus when compared to some FDA-approved
synthetic polymer materials frequently used in tissue engineering. The current research endeavours to expand
the existing production technology of 3D plotting to winding of micro-extruded filaments in order to obtain
flexible polymer tubes with continuous fibre. Tube scaffolds are manufactured by conventional 3D plotting and
by winding. Their structure and quasi-static mechanical properties are evaluated and compared to human aorta.
Winded tubes are found to be far more suitable for application as a blood vessel scaffold than their 3D plotted
counterparts.
INTRODUCTION
are compared to those found in literature for human

aorta.
Some older reports describe measurements of
mechanical properties for the human aorta, most of
which are based on the principles of pressure-diameter
experiments [911]. For the modulus in the circumferential direction, values of 0.75 MPa to 1.7 MPa
are noted for thoracic and abdominal aorta segments
[1013].
Whilst several groups have reported on the creation
of PCL tube scaffolds with randomly oriented fibre
through electrospinning [1416], to our knowledge
this manuscript presents the first results on adapting
the 3D plotting technique for winding of the extruded
filament into PCL tubes with continuous fibre.
The creation of a functional Tissue Engineered Blood

Vessel (TEBV) for arteries with lumen diameters of
6 mm and smaller is widely researched, yet has met
with little clinical success up to date [13]. Next to
biocompatibility and biodegradability, an important
aspect of TEBV properties is elastic-mechanical compliance with its natural tissue counterpart. The TEBV
must be strong enough to withstand rupture at elevated
blood pressures (functionality of collagen) and at the
same time be able to follow the elastic movement of
a natural artery (functionality of elastin). Where synthetic scaffolds are concerned, it has proven difficult
to obtain both qualities in the same part. Often, ultimate mechanical properties are more than sufficient
but flexible behaviour under smaller loads cannot be
emulated.
In the current research, we investigate the mechanical suitability of polymeric tubes for use as TEBV.
Poly--caprolactone (PCL) is a suitable material for
applications that require more pliant polymer materials. Both at room and body temperature, PCL is above
its glass transition and quite flexible when compared
to FDA-approved aliphatic polyesters like poly-(lactid
acid) PLA of poly-(glycolic acid) PGA [4]. PCL has
equally shown promise during in vivo experiments of
an electrospun TEBV [5].
PCL tubes are created by both conventional 3D
plotting and filament winding. Their static mechanical properties are assayed by tensile testing of ring
segments, an oft-used for mechanical assay of TEBV
[68]. The resulting circumferential modulus values
METHODS & MATERIALS
2.1 Materials
Thermoplastic polymer material used is granulated
PCL CAPA 6500 from Perstorp with number averaged
molecular mass of 50000 Da and a melt temperature
of 60 C [17].
2.2 BioScaffolder apparatus
All tube scaffolds were produced on a BioScaffolder
apparatus (SysEng). The manufacturing technique
of the BioScaffolder process has been previously
described in literature [18, 19]. In brief, polymer granulate is extruded by a mobile dispense head into fine
filaments (order 100800 m). These filaments are
99
Figure 1. Winding of tube scaffolds on BioScaffolder

apparatus.
deposited on a plotting table layer by layer, building

up a 3D part in an additive fashion.
PCL was processed at 110 C with a thermoplastics dispense head. A 28 g needle was used, producing
extruded filaments 180 m in diameter. Spindle speed
of the extrusion screw was 22 rpm. Parameters specific
to each tube series are described below.
2.3
3D plotted tube scaffolds
Three series of tube scaffolds were created by conventional 3D plotting with the following dimensions:
inner diameter din = 6 mm, wall thickness t = 0.5 mm
and tube height h = 5 mm. For the first series the orientation shift between two consequent layers was 45 , for
the second it was 30 and for the third 15 . These series
are respectively labelled 3Dplotted-45, 3Dplotted30 and 3Dplotted-15. Most important processing
parameters were set as follows: dispense head XY feed
80 mm/min, dispense pressure 5 bar, layer thickness
160 m and strand distance 324 m.
Figure 2. Tensile test setup with pins.
2.5 Scaffold inspection

Visual evaluation of scaffolds was performed with a
Keyence VHX 500 microscope. Scaffold height and
wall thickness was assessed with a digital calliper.
2.6 Tensile testing
Evaluation of mechanical properties was done on a
Instron 5565 tensile testing apparatus, with a load cell
of 10 N. Tubes were mounted over a double pin setup,
which is shown in Figure 2.
Start length was calculated as the circumference of
the tube and strain was determined in relation to this
length. Tubes were loaded at rate of 10 mm/min to a
maximum load of 10 N. E-modulus was determined as
the slope of the linear ascending part of the stress-strain
curve, meaning 1 MPa for 3D plotted tubes and
0.4 MPa for the winded tubes. Contrary to natural
tissue, it was not necessary to precondition the tubes
prior to measurement, as the elastic behaviour does not
vary with increasing load cycle. Modulus was determined for three tubes per production series and results
averaged.
2.4 Winded tube scaffolds

Tubes with continuous fibre were realized by the
implementation of a rotational axis on the BioScaffolder apparatus, onto which the extruded filaments
are winded (shaft speed = 3 rpm). This is illustrated
in Figure 1.
Tube scaffolds were designed with din = 6 mm,
t = 0.5 mm and h = 910 mm. Different filament
winding angles were accomplished by adapting the
dispense head feed F. Three series of winded tube scaffolds were created on a rotating axis of 6 mm with
the following winding angles: 10 , 45 and a combination of both. These series are respectively labelled
winded-10, winded-45 and winded-C.
3
3.1
RESULTS
Manufactured tube scaffolds
An overview of the dimensions of the fabricated tubes

is given in Table 1. All dimensions remain within
an acceptable standard deviation of less than 1% for
height and less than 5% for wall thickness, which
speaks well for the reproducibility of the manufacturing technique.
Figure 3 shows the resulting geometry of a 3D plotted tube with the 45 stacking sequence. To obtain a
flexible tube scaffold, there are several disadvantages
to structures created by conventional 3D plotting.
100
Table 1.
dimensions of fabricated tubes.
tube type
3Dplotted-45
3Dplotted-30
3Dplotted-15
Winded-10
Winded-45
Winded-C
dimension tube1 tube2 tube3 mean stdev

h [mm]
t [mm]
h [mm]
t [mm]
h [mm]
t [mm]
h [mm]
t [mm]
h [mm]
t [mm]
h [mm]
t [mm]
5.01
0.52
5.02
0.52
5.01
0.53
9.53
0.52
9.08
0.54
9.65
0.52
5.00
0.51
5.04
0.51
5.01
0.54
9.63
0.48
8.97
0.53
9.70
0.48
5.03
0.54
5.01
0.54
5.00
0.52
9.54
0.51
9.04
0.52
9.71
0.51
5.01
0.52
5.02
0.52
5.01
0.53
9.57
0.50
9.03
0.53
9.69
0.50
0.02
0.02
0.02
0.02
0.01
0.01
0.06
0.02
0.06
0.01
0.03
0.02
Figure 5. Static elastic modulus values for all scaffold types.
Figure 3. 3D plotted tube of 45 type. (a) isometric view,

(b) side view and (c) top view. All scale bars 1 mm.
Figure 4. From left to right: tubes from series winded-10,

winded-45 and winded-C.
For one, the layer-wise production method effectively limits polymer filament distribution to the individual layers; there is no continuous filament in the
longitudinal/axial direction of the tube. This may cause
delamination of the scaffold layers when axially loading the tube and could be the case for suturing the
TEBV into position or mounting it on a bioreactor for
conditioning. Moreover, it leads to expect a high stiffness in the radial & circumferential direction which is
not desirable for arterial replacements.
Secondly, the dispense head is forced to manoeuvre
back and forth in the thin section of the tube wall,
which leads to a very dense structure with little to no
remaining porosity in the scaffold wall.
Finally, in every layer there are two sections of the

tube wall in which the deposited filament does not
meander back and forth within the wall thickness,
but draws a straight line. The phenomenon is inherent to the postprocessor, which attempts to deposit as
much continuous filament as possible in the direction
aligned to the current layer orientation. This straight
line is repositioned in each subsequent layer according to the stacking angle. Hence, a 45 -stacked tube
wall tube will display this anomaly every 4th layer and
within 4 longitudinal sections of the tube, whilst a 15 stacked part will only exhibit it every 12th layer for a
single wall section, but also in 12 different locations.
We summarize this phenomenon as follows:
with the scaffold stacking angle and nw both the

number of wall sections in the circumference where
and every how many layers the weakness occurs within
these sections.
Figure 4 shows the three types of winded scaffolds.
The tube winded under low angle (winded-10), not
unlike the 3D plotted tubes, turns out to be a very
dense and closed structure. The one winded under
45 (winded-45) however, displays increased porosity in the tube wall. Finally, porosity in the tube with
combined winding angles (winded-C) is governed
by the density of the outer layerwinded under low
anglewhich will affect pore size and geometry.
All of the winded tubes present continuous polymer
fibres over the entire length of the scaffold.
These winded tubes feel much more pliant to the
touch than their 3D plotted counterparts, hinting at
lower elastic modulus values.
3.2 Tensile properties
The results of the quasi-static tensile tests are summarized in Figure 5.
The 3D plotted tubes exhibit stiffness values
between 16 MPa and 26 MPa, which is about 200 times
101
the earlier reported moduli for arterial tissue [1013].

Stiffness tends to rise with diminishing stacking angle.
As shown in Figure 6, visual inspection of these scaffolds after testing reveals a peculiar mode of straining.
A plastic deformation is observed within a single section of the tube wall. One may note from Figure 6-c that
exactly every nth
w layer is broken, causing the tensile
force to be redistributed over the remaining scaffold
layers and further weakening the structure.This section
corresponds to the locations of the straight-deposited
filament discussed above, confirming them out as
influential flaws in the wall structure.
For the winded scaffolds, remarkably lower modulus values are achieved. The open structure of the
winded-45 series attains tensile modulus values of
4.53 0.26 MPa, less than a quarter of the stiffness of the 3D plotted tubes. Even the winded-10
series which is closest in structure to the dense plotted tubes, displays a distinctly reduced modulus of
12.60 0.68 MPa. As could be expected, the windedC series with combined winding angles shows values
between those of its separate composing structures,
with a modulus of 7.73 0.24 MPa.
There is less variation in modulus values between
samples within the different winded series than within
the 3D plotted series and the tubes did not display
plastic deformation after tensile testing.
Figure 6. Evaluation of 3D plotted tube with 45 stacking sequence. (a) cut-open tube before tensile loading. (b)
cut-open tube after loading and (c) enlargement of strained
region in tube. All scale bars are 1mm.
head feed and speed of the rotational axis, working towards either larger or smaller pores. Both this
porosity and the directional orientation of the polymer
filaments leads to lower values for the circumferential modulus. The most pliant result was obtained for
the winded-45 series, with an average modulus of
4.53 MPa. While this does not yet emulate the low
stiffness values of the arterial tissue, it most certainly
indicates that this adapted production technique is a
step in the good direction for creating flexible tube
scaffolds for TEBV with synthetic polymer materials.
When abstracting the tube scaffolds to a thin-walled
pressure vessel, the following relation between circumferential tensile stress and equivalent internal fluid
pressure may be made :
DISCUSSION
Compared to natural arteries, 3D plotted tube scaffolds for TEBV display a radial stiffness that is much
too high, due to a very dense wall structure and the
layer-wise assembly. They do not sport a continuous
fibre throughout the tube height and the structural
weakness induced by the periodical deviation from
the meandering pathway within a layer causes them to
partially fracture and then deform plastically during
loading. It is possible to minimize this flaw by using
smaller stacking angles, distributing the flaw along the
tube circumference but this leads to overall stronger
and stiffer tubes, exactly the property we are trying
to downplay for use as functional TEBV. The lack of
wall porosity will equally have a negative effect on
nutrient transport and cell migration throughout the
scaffold. Quite clearly, the 3D plotted tubes fabricated
here are unfit for use as flexible TEBV scaffolds.
On the other hand, winding of extruded PCL filament offers auspicious results in term of both wall
porosity and mechanical properties. A useful advantage of the winding technique which is evident from
these simple experiments is the presence of a continuous fibre over the entire tube length, resulting
in a more homogeneous straining behaviour instead
of local failure. Lack of aforementioned localized
structural weakness also ensures smaller variance in
stiffness results.
It is clear from visual inspection that the tubes
are more porous than their plotted counterparts. Such
porosity may be fine-tuned by playing with dispense
with the circumferential stress component induced

by a fluid pressure p on a thin-walled tube with outer
diameter dout and wall thickness t and t, the stress
induced on two circumferential tube wall sections by
the tensile testing with a load force F on a tube with
height h and wall thickness t.
As such, the onset values for the linear elastic
behaviour and the determination of elastic modulus
correspond to roughly 1100 mmHg blood pressure for
3D plotted tubes and 430 mmHg for the winded tubes.
This high value sustains the earlier conclusion that the
plotted tubes are not at all suited to mimicking the
behaviour of natural arteries and that winded tubes
show greater promise, even whilst they are still some
way off from reaching the very flexible nature of the
arterial tissue itself.
It would seem that two of the main different
approaches for constructing aTEBV may be advancing
towards one another. Synthetic polymers are generally too stiff, but adequate processing through filament
winding or electrospinning on a rotating mandrel [5,
1416] is able to produce more and more flexible
structures. On the other hand, biological polymerbased endeavours like collagen-based scaffolds for
TEBV which display a much greater at tractiveness
102
to cells , have proven to be too weak in terms of

tensile properties [2, 20, 21] but researchers are able
to gradually improve these properties, for example
by mechanical preconditioning [21]. Perhaps hybrid
scaffolds that combine a synthetic polymer backbone
for strength with natural material for flexibility and
cell interaction will turn out to be a feasible solution
for this intricate challenge of creating a functional
TEBV, both towards cell interactions and mechanical behaviour. Efforts have already been made in this
direction, reporting tensile moduli of 3.8 MPa for an
electrospun PCL/collagen combined scaffold [22].
5
CONCLUSIONS
In this research, it was demonstrated that the winding

method for tube scaffolds is not only viable but also
renders distinctly more flexible products than with traditional 3D plotting. This makes them more suitable
as a replacement for the very elastic natural arterial
tissue, even though close approximation of the circumferential modulus of arteries was not yet achieved.
Further research should delve into attaining these low
elastic moduli by adapting filament sizes, processing
parameters and winding angles.
ACKNOWLEDGEMENTS
The authors would like to thank Mustafa Erko for his
help with the development of the tensile test setup,
as well as Adrian Badia & Jose Luis Moreno for
their assistance with the scaffold production & tensile
testing.
This research was funded by Faculty of Applied
Engineering Sciences at University College Ghent.
REFERENCES
[1] Isenberg, B.C., C. Williams, and R.T. Tranquillo,
Small-diameter artificial arteries engineered in vitro.
Circulation Research, 2006. 98(1): p. 2535.
[2] Gong, Z.D. and L.E. Niklason, Blood vessels engineered from human cells. Trends in Cardiovascular
Medicine, 2006. 16(5): p. 153156.
[3] Yow, K.H., et al., Tissue engineering of vascular conduits. British Journal of Surgery, 2006. 93(6): p.
652661.
[4] Van de Velde, K. and P. Kiekens, Biopolymers:
overview of several properties and consequences on
their applications. Polymer Testing, 2001. 21: p. 433
442.
[5] Pektok, E., et al., Degradation and Healing Characteristics of Small-Diameter Poly(epsilon-Caprolactone)
Vascular Grafts in the Rat Systemic Arterial Circulation. Circulation, 2008. 118(24): p. 25632570.
103
[6] Assoul, N., et al., Mechanical properties of rat thoracic

and abdominal aortas. Journal of Biomechanics, 2008.
41(10): p. 22272236.
[7] Buttafoco, L., et al., Physical characterization of vascular grafts cultured in a bioreactor. Biomaterials,
2006. 27(11): p. 23802389.
[8] Wille, J.J., E.L. Elson, and R.J. Okamoto, Cellular and
matrix mechanics of bioartificial tissues during continuous cyclic stretch. Annals of Biomedical Engineering,
2006. 34(11): p. 16781690.
[9] Bergel, H., The properties of Blood Vessels, in Biomechanics, Its Foundations and Objectives, Y.C. Fung, N.
Ferrone, and M. Anliker, Editors. 1972, Prentice Hall:
New Jersey.
[10] Langewouters, G.J., K.H. Wesseling, and W.J.A. Goedhard, The Static Elastic Properties of 45 Human
Thoracic and 20 Abdominal Aortas Invitro and the
Parameters of a New Model. Journal of Biomechanics,
1984. 17(6): p. 425435.
[11] Silver, F.H., D.L. Christiansen, and C.M. Buntin,
Mechanical-Properties of the Aorta-a Review. Critical Reviews in Biomedical Engineering, 1989. 17(4):
p. 323358.
[12] Learoyd, B.M. and M.G. Taylor, Alterations with age
in viscoelastic properties of human arterial walls.
Circulation Research, 1966. 18(3): p. 278-&.
[13] Drangova, M., et al., Elasticity and geometry measurements of vascular specimens using a high-resolution
laboratory CT scanner. Physiological Measurement,
1993. 14(3): p. 277290.
[14] Murugan, R. and S. Ramakrishna, Design strategies
of tissue engineering scaffolds with controlled fiber
orientation. Tissue Engineering, 2007. 13(8): p. 1845
1866.
[15] Stitzel, J., et al., Controlled fabrication of a biological vascular substitute. Biomaterials, 2006. 27(7): p.
10881094.
[16] Vaz, C.M., et al., Design of scaffolds for blood vessel
tissue engineering using a multi-layering electrospinning technique. Acta Biomaterialia, 2005. 1(5): p.
575582.
[17] Perstorp, Capa 6000 series data sheet, P.U. Ltd, Editor
2003.
[18] Ragaert, K., et al., Machine design & processing
considerations for the 3D plotting of thermoplastic
scaffolds. Biofabrication, 2010. 2 (1).
[19] Landers, R., et al., Fabrication of soft tissue engineering scaffolds by means of rapid prototyping techniques. Journal of Materials Science, 2002. 37(15): p.
31073116.
[20] Kakisis, J.D., et al., Artificial blood vessel: The Holy
Grail of peripheral vascular surgery. Journal of Vascular Surgery, 2005. 41(2): p. 349354.
[21] Seliktar, D., et al., Dynamic mechanical conditioning of
collagen-gel blood vessel constructs induces remodeling in vitro. Annals of Biomedical Engineering, 2000.
28(4): p. 351362.
[22] Lee, S.J., et al., Development of a composite vascular scaffolding system that withstands physiological
vascular conditions. Biomaterials, 2008. 29(19): p.
28912898.

Automation design and simulation of a stent

E.L. Melgoza
Instituto Tecnolgico y de Estudios Superiores de Monterrey, Monterrey, Mexico
L. Seren & J. Ciurana

Grup dEnginyeria de Producte, Procs i Producci, Universitat de Girona, Spain
A. Rosell
Servei de Pneumologia, Hospital Universitari de Bellvitge;
Institut dInvestigaci Biomdica de Bellvitge (IDIBELL), Barcelona, Spain
ABSTRACT: Stents have been used so far as a palliative solution to maintain the human tubes unblocked and
improve patients quality of life. Nevertheless, there are still several complications associated with available
airway stents: stent migration, stent obstruction, and granulation tissue formation. In this work, a combination
of methodologies based on Attribute Listing, QFD and TRIZ methods was applied in order to develop an
innovative tracheal stent. The final design, which met the most important requirements in order to avoid the usual
complications, was proposed and custom-manufactured using an additive technology known as Fab@Home. The
development of the stent, described herein, could be considered a time and cost-effective approach to produce
customized airway stents.
1
INTRODUCTION
Partial or total obstruction of the main airway can be

caused by different agents including either benign or
malignant diseases, foreign bodies or trauma. These
lesions occlude the trachea and result in a reduction of
ventilation and oxygen intake. Airway stenting might
provide prompt tracheal stabilization by pressing gently against the tracheal wall and thereby maintaining
the trachea open during the respiratory movements.
Stents may be used in both benign and malignant airway processes, although the most frequent indication is
malignant tracheobronchial obstruction, specifically
bronchogenic carcinoma (1). Malignant obstruction
often presents dyspnea, suffocation, chest pain and
sepsis. These symptoms dictate an extremely limited survival expectancy making airway stenting an
adequate palliation treatment for these cases while
treating the primary cause with chemotherapy or radiation (2). Airway stent placement for benign conditions
is performed only to selected patients unsuitable for
curative surgical treatment since they have longer life
expectancies and many anatomical variations.
At present, there are basically two types of stents
clinically available: tubular stents made of plastic (generally silicone) and expandable metallic stents (uncovered or covered by plastic material). Their properties
and consequently their advantages and disadvantages
depend on the material from which they are built. Since
1990 the Dumon stent (Novatech, Abayone, France),
a cylindrical silicone tube with studs in its external
surface, has probably been the most frequently used
tracheal silicone stent worldwide (3). Nevertheless,

several other silicone stents are commercially available (4, 5) and new designs have arisen trying to end
with the drawbacks associated with the silicone stents,
such as the dynamic C shaped stent named Natural
stent, more flexible, easier to insert and with higher
contact between the tracheal wall and the stent than the
Dumon (6). The Natural stent geometry resembles, in
a more accurate way, the tracheal natural rings considering that the human trachea does not have a perfectly
circular transversal profile, but an elliptical, D-shaped,
triangular or U-shaped contour depending on sex and
age (7, 8). In addition, the trachea is composed by
different wall morphologies and functions. The anterior wall is curved and thick due to the cartilaginous
rings, which prevent the trachea from collapsing and
obstructing the airway, and the posterior wall is soft
due to the smooth muscle, which allows the expansion
during the normal movements of the trachea (Figure 1).
Considering the wall functions, the Natural stent does
not differentiate between the anterior and posterior
wall.
Interestingly, despite all the possibilities, several
complications such as stent migration, formation of
granulation tissue around the stent and retained secretions inside the stent are still the major problems of
stenting. Since there is no ideal stent available in the
market, a new improved design is required. The new
design should avoid the aforementioned complications
and should be customized to fit snugly, due to the high
rate of patients with unusual stenosis or anatomical
variations.
105
The aim of this work is to present an integration of design methodologies, based on several tools
such as Attribute Listing, QFD, and TRIZ, in order to
develop functional and innovative customized tracheal
stent designs. Additionally, the additive Fab@Home
machine is presented as a suitable technology to
produce customized airway stents.
2
EXPERIMENTAL WORK
The experimental work consisted of two parts: the

design of a customized stent and the manufacturing
process to fabricate it.
Figure 1. Anterior and posterior human tracheal wall
morphology.
2.1 Customized stent design
The complexity of making customized designs for

biomedical applications goes far beyond from covering only the patients necessities. The conventional
design process is heavily dependent on experiences
and skills of the designer. In order to decrease the
dependence on design experience and to enhance the
participation of pulmonologists in the design process,
a customization methodology should be applied (9).
There are several methods available for a design team
to understand the customer needs, such as questionnaires, group discussions and interviews. The following step is to use this information to set targets for a
new product development. The attribute listing method
and the Quality Function Deployment (QFD) can be
used as tools to reinforce the concept design stage.
QFD is a systematic process that helps planning the
development of new products, by fully understanding
the necessities and requirements of the final user and
showing any physical contradiction. Finally, a problem solving tool, known as TRIZ, aids in solving those
physical contradictions found and closes the conceptual design process (10). In summary, this combination
of design methodologies is adequate to find the best
customized stent shape and to select the most suitable
material and manufacturing process.
Recently, the use of Additive Manufacturing (AM)
technologies in medical applications has become more
conspicuous (11). AM facilitates the fabrication of
individual complex geometrical models with economical viability and almost no constraints. Within this
context, AM technologies seem to be the most suitable manufacturing methodology to customize airway
stents for patients facing both malign and benign
diseases in order to improve their quality of life.
Specifically, the open-source Fab@Home additive
machine, is a versatile and low-cost AM machine
capable of producing three dimensional (3D) objects
from computer-aided design (CAD) data by successively piling material up in a relatively short time
without the need of expensive moulds (12). Within this
context, this machine represents an optimal solution
for manufacturing stents fairly adapted to the patients
anatomic characteristics at a relatively low cost and
for producing stents both time and cost-effective for
patients with expected short-term survival (13).
Commercial airway stents are available in different

shapes and sizes. The Natural stent was considered
the starting model for a new design conceptualization
due to its C shape adapted to the tracheal wall (6).
In the following sections the procedure for designing
patient-customized stents is described.
2.1.1 Attribute Listing methodology
There are several techniques that control the organization and generation of innovative ideas such as Morphological Analysis, Matrix Analysis, and Attribute
Listing. In this study the Attribute Listing, which
focuses on the attributes of an object, was selected
as the starting method to develop the customized stent
design. A state of the art, a benchmarking study, and
various interviews with doctors were performed in
order to gather enough information to generate an
attributes list with the main necessities (Whats) and
its pertinent functional requirements (Hows).
2.1.2 Quality Function Deployment methodology
The QFD methodology was used to structure and
prioritize the information generated in the Attribute
List, translating it into quantified specifications, establishing targets for meeting these specifications and
finally recording expected technical contradictions
(Figure 2). QFD was performed to ensure that the variety of specifications given by doctors and specialists
converge to a successful stent.
2.1.3 Theory of Inventive Problem Solving
methodology
The Theory of Inventive Problem Solving, known as
TRIZ and based on the principles of invention derived
from the study of more than 1.5 million patents worldwide, was used to find out design patterns and solve
the technological contradictions raised with the QFD
methodology. This methodology pairs the contradictions found in the QFD and gives tips based on design
principles to solve the conflict (10).
2.2 Stent additive manufacturing procedure
Following the aforementioned methodology, several
stents were designed using CAD software (SolidWorks
2010, Concord, MA, USA). In order to customize
106
Table 1.
Stent attributes list.
Avoid migration
When coughing, the stent verical

displacement has to be less than
a tracheal ring.
Displacement 10 mm
Avoid obstruction The inner diameter should be bigger
than 8 mm to allow the mucus flow
Avoid granulation Do not allow physically in growth of
tissue
tumor or reactive tissue:
Hydrophobicity = 90130 contact angle
Easy to insert and The material must be flexible
remove
Young modulus, E = 115 MPa
Figure 2. Quality Function Deployment for the stent development process.
Dynamic
The wall near the muscular tracheal tissue

must have a displacement of;
During breathing = 1 mm
During coughing = 4 mm
Biocompatible
As stated by international regulations, the

matcrial must comply with:
ISO 10993-1
FDA approved for long term implants
Biodurable
The material must have a thermal stability

between 3545 C
The material must have chemical stability:
do not react with the environment (air,
mucus, body cells), ISO 10993-18:2005
Radiopaque
The material must able reachable: inability

of X ray to pass through the material
Manufacturability Lead time 10 hours

Low cost
Cost $1,100
Average silicone stents cost is $250$600,
plus using rigid bronchoscope could
increase the net reimbursement by $1,000
Figure 3. The Fab@Home Model 1 machine.
3
the stents, four different parameters were considered
as important values which can determine a tracheal
geometry. The parameters were obtained from Computed Tomography (CT) data of a 40-year-old man.
Briefly, the optimal stent length was approx. 60 mm,
the inner diameter approx. 16 mm, the outer diameter
approx. 20 mm and the distance between rings approx.
4 mm.
The models geometries were saved in Stereolithography format (STL) and sent to the Fab@Home Model
1 machine (Koba Industries,Alburquerque, NM, USA)
(Figure 3). This AM technology machine consists in a
three-axis Cartesian gantry positioning system driven
by stepper motors attached to lead screws. The lead
screws move the piston which extruders the material
through a syringe deposition tool with a tip diameter of
0.25 mm. The motion and positioning control is provided by a computer with an application that displays
the real-time state of the machine and allows to control
and to position the axes, to import the STL file, and
to execute tool paths in order to fabricate the stents.
The material used to fabricate the stents was common
household silicone (Fischer Iberica, Cambrils, Spain).
RESULTS
3.1 Stent designs

Information from a state of the art about tracheobronchial stents, a benchmarking study, and from
several meetings with doctors was collected as a first
step to generate the attributes list, which includes
eleven requirements concerning customer needs
(flexible, easy to insert, dynamic, with inner low
roughness, biodurable, hydrophobic, avoiding migration, radiopaque, and inexpensive), legal issues
(biocompatible), and manufacturing specifications
(flexible, dynamic, and with reduced manufacturing
time). Each attribute was quantified in order to translate it into technical requirements. All of them are
shown in Table 1.
The overall eleven product specifications were
considered for generating a QFD matrix. The QFD
approach was used to determine the priorities to
achieve the design goals. From this QFD several issues
may be highlighted:
107
Biocompatibility and low migration were determined as the two product specifications of greater
importance, with scores higher than nine out of ten.
Figure 4. Final 3D CAD model with its dimensions.

Figure 5. Final stent model manufactured with a wavy
geometry with its dimensions.
Fourteen engineering requirements were detailed

and 23 strong, 8 moderate, and 6 weak relationships
with product specifications were found.
Other 6 commercial stents were used to perform
a competitive benchmarking. The majority accomplished badly in migration and stent dynamism. Two
factors to improve in the new design.
Low cost and final position when coughing were
ranked as the two technical parameters more difficult to achieve.
A correlation between engineering requirements
was made in order to determine interrelationships of
design requirements. Five extremely negative correlations were found. These negative correlations or
contradictions are dependent of other variables that
embody a physical effect in the design task.
The final result of the QFD approach is a selection of several areas for improvement such as reducing
wall thickness, lowering manufacturing hours, and
reducing the cost of the overall process. Engineering
requirements, technical difficulty, correlation matrix,
and benchmark may be used to guide the development
and improvement for the new design.
Finally, someTRIZ tips were applied to the selection
of the manufacturing process, the material selection,
and the shape of the stent. At the end, twelve new
stent sketches were generated, although only one final
design was chosen by three experts in manufacturing,
four designers and four surgeons (Figure 4).
Concerning the shape, the final model was designed
forming a D geometry. The posterior straight wall
was described as a 1.0 mm flexible membrane able to
reproduce the behavior of the muscular membrane in
order to deform and compress the airway when coughing, to avoid secretions retention, and to facilitate the
insertion and the removal of the stent. The anterior
wall, 2.0 mm thick, was designed as a wavy curved
geometry intended to distribute the pressure of the
stent through the length of the trachea while increasing the contact surface. The resultant 16 mm inner
diameter was considered enough to allow the air and
mucus flow.
3.2 Stent manufacturing

Several manufacturing requirements such as wall sections thickness between 1 and 2 mm, lead time below
10 hours, and cost under $1000 were addressed to
select the Fab@Home machine as the most convenient
additive technology for this case study. The results
using the Fab@Home machine Model 1 are shown in
Figure 5. This additive technology was able to produce
a posterior wall section thickness of about 1.00 mm,
an anterior curved wall of about 2.00 mm and a length
of 60 mm with an inner diameter of 16 mm.
The total manufacturing time spend to fabricate the
stent model was approximately 45 minutes. From a
visual point of view, and because of the deposition
process associated to this technology, the surface of the
manufactured stent was layered giving a high roughness to the model. Concerning the total costs, this
open-source machine, which costs is roughly $2,300,
provides an economic solution to fabricate, at low cost,
airway stents requiring only basic skills and tools.
DISCUSSION
As mentioned before, the ideal stent does not yet exist.

All available commercial stents are associated with
certain complications such as stent migration, stent
obstruction due to retention of secretions and formation of granulation tissue around the stent (14).
The goal of this work was to present a combination
of design methodologies which considered and faced
almost all requirements in order to understand and
technically interpret the most important challenges
when using airway stents and to propose a fast and
inexpensive technology to manufacture them. A final
stent model was obtained aiming to minimize the typical stent-associated complications, and to ensure an
effective normal tracheal behavior.
Firstly, customer needs and technical specifications
were determined, studied, and interpreted in order to
come up, using the TRIZ methodology, with a final
108
design which tries to satisfy all the requirements.

Regarding the drawbacks of the silicone stents, both
the Dumon stent and the Natural stent present a similar migration rate confirming the difficulty of silicone
stents to remain at the occluded zone once inserted.
The manufacturing procedure used in this work to fabricate the D-wavy-shaped stent allows a customization
of the new stent and consequently the adaptation of
this to the precise dimensions of the patients trachea.
Moreover, the surface roughness of a stent produced by
the Fab@Home machine is visually appreciable, and
probably it can increase the friction with the tracheal
wall. Both tactics may work together to reduce the
migration rate. A second complication is the obstruction of the stent due to retention of mucus inside, also
occurring in the two stent, the Dumon and the Natural.
The poor flexibility of the stents and the low ratio inner
to outer diameter might be the reasons for the high
incidence of obstruction. The design described in this
work, with a 1 mm flexible posterior wall, was intended
to allow the deformation of this section into the stent
while reducing the pressure into the stent area during cough. The deformation stimulates the movement
of the mucus through the stent and therefore avoids
the obstruction of its interior. The inner diameter was
maintained around 16 mm to allow the mucus and the
air flow normally. Finally, the growth of granulation
tissue around the upper and lower part of the stent is
associated with impaired mucosal blood supply (15).
In order to minimize this drawback it is necessary a
reduction of the pressure on the mucosa. The Dumon
stent with its studs as an external fixation or the Natural stent with its rectangular rings may damage the
mucosa and stimulate the production of granulation
tissue (5). The wavy geometry without sharp edges
designed herein may support the distribution of the
pressure homogenously throughout the length of the
stent while decreasing the overall load and probably
the airway inflammation. Nevertheless, this work had
some obvious limitations. The material used for the
manufacture was ordinary silicone. Further studies are
needed with medical grade silicone or polymeric materials in order to optimize the manufacturing process.
Moreover, a sterilization post-process should be considered. Mechanical tests should be performed to test
the stent behavior and in vivo studies should be carried
out to determine the biocompatibility and efficiency
of this new stent.
CONCLUSIONS
In conclusion, the combination of design methodologies presented herein has been used to theoretically
guide the development of a customized silicone stent
suitable for the palliation of airway obstructions as it
fulfills two important requirements: a wavy D-shape
intended for avoiding migration and a posterior wall
which may potentiate the dynamism of the trachea and
prevent of its obstruction. Nevertheless, future work is
needed to improve the manufacturing technique, the
biocompatibility of the material and the efficiency of

the entire process.
ACKNOWLEDGEMENTS
The authors would like to express their gratitude
to the Product, Process and Production Engineering
Research Group from University of Girona, the Institut dInvestigaci Biomdica de Bellvitge and the
Research Chair in Intelligent Machines from Tecnolgico de Monterrey for the facilities provided during
the experiments and all their valuable support. This
work was carried out with the grant supports from
IREBID project (FP7-PEOPLE-2009-IRSES-247476).
REFERENCES
[1] Shin JH, Song H, Shim TS. Management of tracheobronchial strictures. Cardiovasc Intervent Radiol.
2004;27(4): 31424.
[2] Kim H. Stenting therapy for stenosing airway disease.
Respirology. 1998;3(4):2218.
[3] Dumon J. A dedicated tracheobronchial stent. Chest.
1990;97(2):32832.
[4] Cooper JD, Pearson FG, Patterson GA, Todd TRJ,
Ginsberg RJ, Goldberg M, et al. Use of silicone stents
in the management of airway problems. Ann Thorac
Surg. 1989;47(3):3718.
[5] Ryu YJ, Kim H, Yu C, Choi JC, Kwon YS, Kim J, et al.
Comparison of natural and dumon airway stents for
the management of benign tracheobronchial stenoses.
Respirology. 2006;11(6):74854.
[6] Park HY, Kim H, Koh W, Suh GY, Chung MP, Kwon OJ.
Natural stent in the management of post-intubation
tracheal stenosis. Respirology. 2009;14(4):5838.
[7] Mackenzie CF, McAslan TC, Shin B. The shape of
the human adult trachea. Anesthesiology. 1978;49(1):
4850.
[8] Mehta S, Myat HM. The cross-sectional shape and circumference of the human trachea. Ann R Coll Surg
Engl. 1984;66(5):3568.
[9] He Y, Ye M, Wang C. A method in the design and
fabrication of exact-fit customized implant based on
sectional medical images and rapid prototyping technology. Int JAdv ManufTechnol. 2006;28(56):5048.
[10] Webb A. TRIZ: An inventive approach to invention.
Eng Manage J. 2002;12(3):11724.
[11] Rengier F, MehndirattaA,VonTengg-Kobligk H, Zechmann CM, Unterhinninghofen R, Kauczor H-, et al.
3D printing based on imaging data: Review of medical applications. International Journal of Computer
Assisted Radiology and Surgery. 2010;5(4):33541.
[12] Malone E, Lipson H. FabHome: The personal
desktop fabricator kit. Rapid Prototyping Journal.
2007;13(4):24555.
[13] Chiang WM, Lim CS, Chua CK, Lo PL, Eng CT.
Application of rapid prototyping and tooling in customised airway management. Rapid Prototyping Journal. 2005;11(2):10612.
[14] Zakaluzny SA, Lane JD, Mair EA. Complications of
tracheobronchial airway stents. Otolaryngology-Head
and Neck Surgery. 2003;128(4):47888.
[15] Freitag L, Eicker R, Linz B, Greschuchna D. Theoretical and experimental basis for the development of a
dynamic airway stent. European Respiratory Journal.
1994;7(11):203845.
109

Production and in-vitro characterization of micro-structured

implant surfaces
M. de Wild, Th. Mller, S. Tschumi, R. Schumacher & H. Albrecht
School of Life Sciences, University of Applied Sciences Northwestern Switzerland, Muttenz, Switzerland
ABSTRACT: The following study reports on the cell-biological profile of cells cultured on titanium structures
whose surfaces topographies were CAD micro-structured and directly manufactured with the Selective Laser
Melting (SLM) process. Human osteoblast-like MG-63 and SaOS-2 cells which are known to be sensitive
to structural features were assessed for cell viability, proliferation and differentiation when cultured on three
designed surface microstructures. Based on cell viability, as well as the production of bone specific biomarkers
like transforming growth factor-1 (TGF-1), osteocalcin, prostaglandin E2 (PGE2 ) and alkaline phosphatase
(ALP), it was found that the biological profiles are cell-line specific, and depend on the titanium surface
topography. SLM shows potential for the manufacture of customized titanium implants with surface topographies
that are compatible with specific cell types.
1
INTRODUCTION
The success of bone implants depends in part on the

stable osseointegration of bone onto the implanted biomaterial (Albrektsson 1981). This anchorage between
the artificial material, e.g. titanium, and the biological
tissue should be fast and enduring. As cells are sensitive to structural features, surface topographies of
implants at the nano, micro and macro level must satisfy cell requirements to ensure bone growth (Gittens
et al. 2011, Fadeeva et al. 2010, Mendonca 2010,
Schwartz 2008, Zhao et al. 2007, Keller et al. 2003).
Conventional implants are produced by milling and
turning followed by structuring, e.g. via roughening
with titanium plasma spraying.
In the following study the use of additive manufacturing to simultaneously create metallic macro structures with a designed surface micro topography was
investigated. Specifically, clinically established submicron topography was superimposed by hot acid etching in an attempt to further improve the osteoblastic
response. The influence of the resultant surface topography on cells was verified with human osteoblast-like
MG-63 (originally from an osteosarcoma) and SaOS-2
cells, which are known to be sensitive to structural
features at the growth surface (Zhao et al. 2006).
2
EXPERIMENTAL WORK
2.1 Design
Samples were designed with the Software NX
(Siemens, Version 7.5). The CAD data included information for both the outer shape and surface topology. Four types of surfaces were CAD modeled and
represented as stl files: A: unstructured planar CAD
shell (reference SLM native surface), B: dots of

100 m, separated by 200 m in a hexagonal pattern, C: pits of width 50 m, separation of 200 m and
length of 1000 m, D: lines with a width of 50 m and
a spacing of 200 m, see figure 1. These scales were
chosen to reflect the dimensions of human bone cells.
2.2
Sample production
The starting material was clinically pure titanium

grade 2 according to (ISO 5832-2 1999). Titanium
disks with a diameter of 14 mm and a thickness of
1 mm were produced in a flat, horizontal position
with a circular support structure on a SLM 250HL
machine (SLM-Solutions, Germany) equipped with
a 200 W Ytterbium fibre laser with a wavelength of
= 1075 nm under an inert argon atmosphere. This
allowed for handling of the titanium powder in a closed
material recycling loop under a permanent protective
inert gas atmosphere.
The laser parameters (focus, trajectories, speed,
power and energy density) and geometry (line width,
distances and depth) were optimized to reproducibly
create small and periodic microstructures. The settings
are given in table 1.
2.3
Post treatment
The SLM samples were ultrasonically cleaned for

five minutes in ultrapure water (18.2 M cm,
PURELAB Option-Q7). Thereafter hot acid etching
was performed by immersing the Ti disks into a mixture of HCl and H2 SO4 at elevated temperature for
several minutes (Li et al. 2002, Kieswetter et al. 1996)
before rinsing with ultrapure water two minutes and
ultrasonic cleaning for ten minutes, both repeated three
times. This led to a superimposed fractal structuring
111
Figure 2. BSE micrograph showing the starting titanium

powder.
Figure 1. CAD representation of the virtual surface designs,

top overview, bottom detail illustration. B: dots of 100 m,
separated by 200 m in a hexagonal pattern, C: pits of width
50 m, separation of 200 m and length of 1000 m, D: lines
with a width of 50 m and a distance of 200 m.
Table 1.
AD.
Laser parameters for the surface modifications
parameter
setting
maximum power
inert gas
layer thickness
hatching strategy
orientation
200 W
argon 4.6
30 m
checker board
flat laying horizontal
An MTT (Thiazolyl Blue Tetrazolium, Sigma

#M5655) cell viability assay was used to study cell
fitness. Furthermore, the production of transforming growth factor-1 (TGF-1, Immunoassay, #G7590
from Promega), osteocalcin (ELISA, #BMS2020INST
from Bender Med-System), prostaglandin E2 (PGE2 )
(Kit htrf, #62P2APEB, from Cisbio) and alkaline
phosphatase (ALP) was assessed using commercially
available kits as indicated. To determine cell density the cells were either counted manually (with a
Neubauer counting cell chamber) or via total protein
assay (Bradford test, #500-0006, Bio Rad). Cell densities were normalized to 1 Mio cells/well or mg/ml
protein/well for better between experiment comparability. All tests were measured with SpectraMax
FlexStation3 spectrophotometer (Molecular Devices).
3
of the surface in the m range. The samples were then
rinsed with ethanol (J.T. Baker absolute), packed into
peel bags (Milian SUD) and autoclaved (Advantage
Lab, Al02, 134 C, 20 minutes).
2.4
Physico-chemical material characterization
The purity of the starting material was verified by

X-ray diffraction (XRD, Bruker Phaser D2). The particle size and morphology as well as the surface
topography and chemistry of the solid SLM samples were analyzed by Scanning Electron Microscopy
(SEM, Hitachi TM-1000) equipped with an energy
dispersive X-ray spectrometer (EDX, Na U). The
3D-Image Viewing software (Version 2.0.1 for TM,
Denshi Kougaku Kenkyusyo Co, Ltd) allowed stereographic reconstruction of the four subimages recorded
by a split four-segment BSE detector. The roughness
was further quantified with a profilometer (Rugosurf
G90, cut-off = 2.5 mm).
2.5
Biological material characterization
Human osteoblast-like MG-63 and SaOS-2 cells were

used to analyze the influence of surface topography on
cell viability, proliferation and differentiation.
RESULTS
3.1 Physico-chemical material characterization

The purity of the starting material was verified by
X-ray diffraction and showed no phases other than
titanium in the hexagonal phase. In SEM images
the titanium particles exhibited spherical shape with a
mean diameter of 30 m, see figure 2.
Elementary EDX measurements on solid samples revealed a clean titanium surface free of foreign elements. The surface topography of the disks
was investigated on both length scales: the designed
CAD structure in the upper m range (figure 3) and
the chemically induced etching structure in the sub
m range (figures 4a and 4b). The resultant structures deviated from the virtually planned geometries
because of the laser width and melting region, and
due to the track width compensation, as can be seen
in figure 5. However, the periodicity was consistent
with the planned geometry. The laser parameters were
chosen to optimize the microstructure, i.e. to create
non-intermittent lines and prevent welding of nearby
structures. The depth of the structure was quantitatively determined in the D type morphology: The
profilometric Rz value was 157.3 9.4 m, whereas
the height measured in metallographic cross sections
112
Figure 3. BSE micrograph showing a SLM sample with pit

structure (type C), after shot peening.
Figure 5. Three dimensional stereoscopic representation of

the created structures type B, C and D. B: dots of 130 m,
separated by 170 m in a hexagonal pattern, C: pits of width
120 m, separation of 120 m and length of 1000 m, D:
lines with a width of 120 m and a distance of 120 m.
Figure 6. Profile of the line structure type D by means of

imaging a cross-section.
3.2 Cell-biological investigations

Figure 4a. BSE micrograph showing a SLM sample type B,
with superimposed etching topography.
Measured cell-biological profiles differed between

cells cultured on plastic, smooth titanium, and structured titanium. Although the profiles of MG-63 and
SaOS-2 cells were very different, for both cell types
the three structured surfaces exhibited similar profiles which strongly differed to that observed with the
smooth surface (see figures 7 and 8).
All four types of SLM surfaces showed no sign of
cytotoxicity, and surface structure did not affect cell
differentiation or TGF- levels in either cell line.
4
DISCUSSION
In the current investigation the following was found:
Figure 4b. BSE micrograph showing the fractal etching

topography.
was approximately 175 m, see figure 6. The profilometric method cannot determine the high aspect ration
configuration as the diamond tip cannot resolve the
microstructure.
1. The cell-biological profile of the smooth surface

type A differs strongly from that of plastic.
2. The structured surface types B, C and D show
relatively similar cell-biological profiles, however
these differ strongly from the profiles corresponding to the smooth surface type A and plastic.
3. The two cell lines show different cell-biological
profiles on the structured surfaces B, C and D.
In contrast to the current investigation, in a previous
study (de Wild et al., 2010) it was reported that surface
113
of the School of Life Sciences, University of Applied

Sciences Northwestern Switzerland.
REFERENCES
Figure 7. Comparison of markers from MG-63 cells, all

values were normalized to surface type A (mean values SD,
n = 4, partly a strongly deviating value was removed).
Figure 8. Comparison of markers from SaOS-2 cells, all

values were normalized to surface type A (mean values SD,
n = 4, partly a strongly deviating value was removed).
modification of SLM titanium disks by sandblasting,

hot acid-etching, vibratory grinding or polishing, had
a strong influence on cell differentiation and TGF-
levels of MG-63 cells. A likely explanation for the
less pronounced influence found in the current study
is that the m scale of structures B-D investigated here
was at the upper limit of cell recognition.
The generative SLM method demonstrates potential for direct manufacturing of customized titanium
implants with complex architectures and specially
designed surface topographies to achieve optimal
compatibility with cells. However, in future studies
the length scales need to be further minimized by optimizing the SLM process, and the topology should be
applicable to all surfaces in all spatial directions.
ACKNOWLEDGEMENT
The multi-disciplinary team gratefully acknowledges
the financial support of the strategic funding program
Albrektsson T., Brnemark P., Hansson H.-A., Lindstrm

J. (1981). Osseointegrated Titanium Implants: Requirements for Ensuring a Long-Lasting, Direct Bone-toImplant Anchorage in Man. Acta orthop. scand: 52(2),
155170.
de Wild M., Maier K., Schneider M., Tschumi S., Schumacher R., Albrecht H. (2010). Surface Modification and
In-vitro Investigation of Generatively Produced Implants.
Biomaterialien: 11, 157.
Fadeeva E., Schlie S., Koch J., Chichkov B. (2010). Selective Cell Control by Surface Structuring for Orthopedic
Applications. J Adhesion Science and Technology: 24,
22572270.
Gittens R.A., McLachlan T., Olivares-Navarrete R., Cai Y.,
Berner S., Tannenbaum R., Schwartz Z., Sandhage K.H. &
Boyan B.D. (2011). The effects of combined micron/submicron-scale surface roughness and nanoscale
features on cell proliferation and differentiation. Biomaterials: 32(13), 3395403.
ISO 5832-2:1999(E). 1999. Implants for surgery Metallic materials Part 2: Unalloyed titanium. International
standard: ISO 5832-2:1999(E).
Keller, J.C., Schneider, G.B., Stanford, C.M., Kellogg,
B. (2003). Effects of implant microtopography on
osteoblast cell attachment. Implant Dentistry: 12,
175181.
Kieswetter K, et al. (1996). Surface roughness modulates
the local production of growth factors and cytokines
by osteoblast-like MG-63 cells. Journal of Biomedical
Materials Research: 32, 5563.
Kim H., Choi S.H., Ryu J.J., Koh S.Y., Park J.H., Lee I.S.
(2008). The biocompatibility of SLA-treated titanium
implants. Biomed Mater: 3, 25011.
Li D., Ferguson S.J., Beutler T., Cochran D.L., Sittig C.,
Hirt H.P., Buser D. (2002). Biomechanical comparison
of the sandblasted and acid-etched and the machined and
acid-etched titanium surface for dental implants. J Biomed
Mater Res. 60, 32532.
Mendonca G., Mendonca D.B.S., Aragao F.J.L., Cooper L.F.
(2010). The combination of micron and nanotopography
by h2so4/h2o2 treatment and its effects on osteoblastspecific gene expression of hmscs. J Biomed Mater Res:
94A(1), 16979.
Schwartz Z., Raz P., Zhao G., Barak Y., Tauber M., Yao H.,
et al. (2008). Effect of micrometer-scale roughness of the
surface of ti6al4v pedicle screws in vitro and in vivo.
J Bone Joint Surg Am. 90A(11), 248598.
Zhao G., Zinger O., Schwartz Z., Wieland M., Landolt D.,
Boyan B. (2006). Osteoblast-like cells are sensitive to
submicron-scale surface structure. Clin Oral Implants
Res: 17, 25864.
Zhao G., Raines A.L., Wieland M., Schwartz Z., Boyan B.
(2007). Requirement for both micron- and submicron
scale structure for synergistic responses of osteoblasts to
substrate surface energy and topography. Biomaterials:
28, 28219.
114

A study of mechanical and biological behavior of porous Ti6Al4V

fabricated on EBM
V. Petrovic, J.R. Blasco & L. Portols
Metal Processing Research Institute AIMME, Valencia, Spain
I. Morales, V. Primo & C. Atienza

Instituto de Biomecnica de Valencia, Valencia, Spain
J.F. Moreno
Bio-Vac & Instituto Tecnolgico de Materiales-UPV, Valencia, Spain
V. Belloch
ERESA, Valencia, Spain
ABSTRACT: EBM is a ALM technology capable of processing ferrous and non ferrous powders to fullydense material using layer-by-layer principle. It is highly suitable for manufacturing of medical implants since
it can fabricate designed and controlled porosity and tailored surface quality for the purpose of better bone
in-growth. Due to great advantages in controlled geometry it has become important rival to other porous titanium
materials available on the market, processed with different technologies such as space holder method [1], sintered
microsphere porous coating, etc. This paper brings one of the most thorough analysis of mechanical testing of
porous material made by EBM for the time being. So as to obtain a full picture of porous EBM Ti64, the results
of testing are then compared to the properties of commercially available materials as well as to the human bone
properties. In addition, some basic results of in vivo testing on bone in-growth of EBM specimens are also
mentioned in this paper.
INTRODUCTION
Additive Manufacturing (AM) technologies have been

available on the market for many years. Initially, these
technologies were considered only for prototyping
since the first technologies that appeared on the market were capable of fabricating only polymer parts of
low quality and low resistance. However, in the last
decade, the sector of AM has experienced an important evolution with constant growth in sales of machine
systems and rapid products. Numerous advantages
of freeform fabrication have driven new developments in processing principles and materials. New
value-added materials have been adapted for layer-bylayer processing. On the other hand, new technologies
have been developed to process demanding materials
for different sectors. New energy sources have been
introduced in order to process high melting point metals such as Titanium, Cobalt Chromium, etc. One of
the most powerful active principles is Electron Beam
Melting that has been commercialized by the Swedish
company Arcam .
Electron Beam melting (EBM) is a free-form fabrication technology capable of processing ferrous and
non ferrous powders to fully-dense material, using
layer-by-layer principle.A 3D model is sliced electronically and the slices are printed out one upon each
other to form a final part. This 3D printing is done by
selective melting of powder. Only the part of powder
layer that corresponds to a part slice is melted; the rest
of the powder remains un-melted. In the case of EBM,
the energy source consists of an electric circuit that
is formed between a tungsten filament placed inside
of the electron gun and the building plate (Figure 1).
A high voltage unit supplies 60 kV to the filament
which emits a beam of electric current that may vary
between 0 and 50 mA.
The electron beam is conducted by a set of different
coils that guide, focus and orient it until it impacts the
powder surface. During the impact, electric energy is
transformed to heat which fully melts the powder. In
order to prevent dispersion and deflection of the beam,
the working chamber is kept under deep vacuum (order
of magnitude 104 mbar). Hence, as in other layer-bylayer processes, powder is released from containers
and distributed by a powder dispatcher over the build
platform in fine 70100 m layers. The beam melts
powder to a solid slice, merging it with previous slices.
The build platforms descends for the value of layer
thickness and a new powder layer is dispatched. The
process repeats until the part is completed.
115
Figure 1. Fusing on EBM machine (a); scheme of an

additive machine (b).
EBM has three major advantages in comparison

with other AM processes:
Due to high power (up to 3000 W), the nominal speed of processing is 4060 cm3 /h which is
substantially higher than in laser machines.
Due to processing under vacuum (absence of oxygen), the processed material has very high purity
which results in higher properties and better biocompatibility.
The processing temperature in the build chamber is
very high (in the case of Titanium alloys, around
650 C). Hence, there is less difference between
melting temperature and powder temperature and
less thermal stresses, causing almost no warpage in
processed material.
STATE OF THE ART
In numerous previous works, it was proven that

freeform fabrication offers a step forward in providing additional value to the design of medical devices:
prosthesis, orthosis and implants ([6], [7], [8]). However, for the time being no exhaustive study was found
that demonstrates that advanced design is accompanied by good mechanical properties of porous material.
In addition, there are other porous titanium materials
available on the market. These materials are processed with different technologies such as the space
holder method [1], sintered microsphere porous coating, etc. The present work was conceived to demonstrate that the porous material fabricated with EBM is
equal or better in properties to commercially available
alternatives.
In addition, for medical applications, a thorough
analysis of any material is essential. There is an abundance of previous work regarding characterization of
titanium foam. One of the most complete works has
been published by Winkelried [1]. In porous tantalum foam, which is also commonly used in human
implants, one of the most complete studies has been
offered by Zardiackas et al. [3]. Hence, these two works
were used as a point of reference in this study. Regarding other efforts, Heinl et al. [9] characterized porosity
of Ti64 samples made on selective EBM with acid
etched surface and performed bioactivity testing in
Figure 2. Modelling of porous material in Netfabb: a) model

divided in cubes; b) model divided in different porous zones;
c) types of cells; and, d) selected cell frontal and isometric
views.
SBF. Ponader et al. [10] evaluate in vivo Ti64 samples

made by sEBM by implanting in craneal zone of pigs.
By micro X rays and hystomorphometrics study, the
authors prove good interconnectivity of Ti64 scaffolds.
Finally, Haslauer [12] tested biocompatibility of titanium alloy discs made using direct metal fabrication
in vitro.
Regarding mechanical properties of porous titanium fabricated with EBM, only recently some
relevant work has been encountered [4], [5], [11], [13].
However, the present study shows to be the most complete mechanical study of porous EBM titanium for
the time being. Performed within the framework of
the MEDIFUTUR project and initiated at the beginning of 2009, it was aimed to contribute with the
full biomechanical characterization of EBM porous
titanium.
For that purpose, the authors present and analyze the
results of complete mechanical testing of porous material made by EBM. These results are then compared
to the properties of commercially available materials
as well as to the properties of human bone. In addition, some basic results of in vivo testing on bone
in-growth of EBM specimens are also mentioned in
this paper.
3
METHODOLOGY
3.1 Selection of the pore type

Porous material engineering is enabled by the use of
specific software solutions for lattice structure generation. For the purpose of this study, Netfabb by
FIT was employed. This software allows definition of
zones of different porosities on a model imported in
STL (Figure 2b). The model is divided into cube units
and a type of cell is assigned to each cube (Figure 2a).
There is a huge number of cells that can be designed in
Netfabb (Figure 2c). For the purpose of this project a
square cell was selected (Figure 2d). This cell is determined by cube width (L), bar diameter (d) and pore
diameter (D).
3.2 Selection of the pore size
The power of electron beam is huge. Previous experience with EBM porous material indicates that,
116
Table 1.
Preliminary pore size measurement.
Table 3.
Dimension
D
m
L
m
d
m
Dreal
m
1
2
3
4
5
1067
1067
814
814
614
2500
2500
2000
2000
2000
700
700
600
600
800
985
987
645
658
521
Table 2.
List of test specimens and cell dimensions.
Test type
Test specimen description
N of t.s.
Tensile
74 13 6.5 mm
load zone: 8 4 20 cells
Compression
Bending
Fatigue
Preselected pore size measurement.
d
m
D
m
dreal
m
Dreal
m
%p
m
450
450
450
600
700
800
666
647
577
376
504
681
49.7
57.5
66.5
although designed to certain value, pore sizes are

somewhat smaller when fabricated. It is supposed
that this fact is due to excessive heat that cannot be
evacuated instantly. More material is melted, bar crosssection is increased and consequently the pore size
is smaller. In order to predict the deviation between
designed pore and real pore, previous test fabrications
have been performed.Table 1 shows some of the results
of preliminary testing.
A thorough revision of existing porous materials
and living tissues brought us to the optimal range
of pore size for bone in-growth is from 300 to
700 m. This was the target range for this study. After
the analysis of preliminary results, 3 different pore
size samples were chosen in the above mentioned
range. Accordingly, specimens with three nominal
pore sizes: 600, 700 and 800 m (P-600, P-700 and
P-800) were preselected and tested again. As shown
in Table 2, the measured values of pore size (Column 4) were inside the recommended range and the
obtained porosity (Column 5) was relatively high
(5065%).
3.3 Testing
The purpose of this study was the complete testing
of porous titanium fabricated on EBM for biomedical
use. On one side, this entails testing of porous material to all types of loads that implants are commonly
exposed in the human body: tensile test, bending test,
compression test and fatigue test. However, the biological behaviour of porous material is crucial for
medical use. As mentioned before, this aspect was very
important for choosing the testing samples porosity.
Hence, the samples were used for in vivo testing of
bone in-growth and compared to commercial materials. All details on this are shown in the Results
section.
5
5
1
Cell type
L [m]
d [m]
D [m]
P-600
P-700
P-800
1050
1150
1250
450
450
450
600
700
800
Figure 3. Test specimens for tensile and compression test.
RESULTS & DISCUSSION
4.1 Test specimens

Due to the absence of standards related to lattice structure testing, the decision was made to emulate the
shape and size of the specimens used in previous
work. Among relevant works used for the definition of
mechanical tests are those of Imwinkelried et al. [1],
Hong et al. [2] and Zardiackas et al. [3]. However, in
contrast to other commercial materials (foams, porous
coating, etc.), the porous material made by EBM is a
composition of predefined cells and, as such, it has a
regular and controlled porosity.
For each of the preselected porosities (P-600,
P-700 and P-800), the cell size is different (1.05, 1.15
and 1.25 mm, respectively). Therefore, it is not possible to define unique dimensions of specimen which
would contain a round number of cells in all cases.
Consequently, the specimens were designed as a function of the number of cells. Figure 3 shows different
test specimens.
All test specimens were fabricated on the Arcam A2
machine. The material used for fabrication is Ti6Al4V
in powder form with 52108 m granulometry. The
chemical composition of this material corresponds to
the Ti6Al4V defined in Ti6Al4V-ISO 58323 Implants
for surgery Metallic materials Part 3: Wrought
titanium 6-aluminium 4-vanadium alloy.
4.2 Test results
Tensile test
Table 4 represents the summary of tensile test results
for porous titanium while the Figure 4 shows the
117
Table 4.
foam.
Results of tensile tests and comparison to titanium
Cell type
Force [N]
Section [mm2 ]
Rm [MPa]
P-600
P-700
P-800
7483
4082
5949
79.38
95.22
114.40
94.27
42.87
61.75
Ti64 (EBM)
P-600
P-700
P-800
Ti64 (foam)[1]
Rm [MPa]
%p
Rm [MPa]
%p
94.27
42.87
61.75
49.7
57.5
66.5
70
62.5+0.5
Figure 5. Comparative view of compressive strength.
Table 5. Results of compression tests and comparison to

commonly used materials.
Cell type
P-600
P-700
P-800
Section [m]
Rm [m]
E [MPa]
166.0
139.4
114.9
159.2
183.5
230.2
2615
2927
3288
Table 6.
foam.
Results of bending tests and comparison to titanium
Cell type
Force [N]
Section [mm2 ]
Rm [MPa]
P-600
P-700
P-800
651
611
401
33.6
36.8
40
149.67
101.98
64.23
Ti64 (EBM)
P-600
P-700
P-800
Ti64 (foam) [1]
Rm [MPa]
%p
Rm [MPa]
%p
149.67
101.98
64.23
49.7
57.5
66.5
105
62.5 + 0.5
Bending test
Figure 4. Comparative view of tensile test results.
comparison in achieved porosity and tensile strength

to Ti foam.
The values shown in the table represent the average
value of four test samples per each porosity. Unlike
other tests where mechanical properties decrease with
the increase of porosity, the lowest average tensile
strength was detected in P-700. Nevertheless, the order
of magnitude of tensile strength of EBM samples is
similar to that of titanium foam.
Figure 6. Comparative view of flexural strength.
As presented in Table 6 and Figure 6, porous EBM

titanium shows very good flexural properties which
decrease with the increase of porosity. Compared to Ti
foam, it shows a similar order of magnitude.
Compression test
Fatigue test
The following table shows the summary of the compression test results. In addition, a graph is shown to
compare Ti64 made by EBM with other commonly
used materials as well as trabecular and cortical bone.
It may be perceived that the elastic modulus of EBM
porous titanium is similar to that of porous tantalum,
while the compressive strength is superior to that of
tantalum, and to cortical and trabecular bone.
Compression fatigue test was performed according to

the Haversine sine cycle with the maximum force of
Fm = 3820 N and a stress ratio of R = 0.1. The equipment is shown on Figure 7. The working frequency was
10 Hz. A limit of 5.000.000 cycles was established as
the condition to end the test. All specimens passed the
limit without any problem, showing reasonably good
fatigue properties.
118
Figure 7. Compression fatigue test equipment.
Figure 9. a) Part of a rabbit femur after extraction; b) TAC

of in vivo sample; micro TAC reconstruction of EBM (c) and
Biovac sample.
Figure 8. Test specimens for in vivo testing.
4.3
Table 7.
Bone ingrowth testing
The experimental model used in this study was condilar femoral medial defect in rabbits. Due to the
elevated costs of in vivo testing and after the analysis of the mechanical results, the P-700 sample was
selected for implantation. Five samples (shown at
Figure 8) were fabricated to be implanted during 8
weeks, as this period was sufficient to show correct osseointegration. The samples were designed for
pull-out testing: one part of the sample is implanted
and the other has a hole to introduce the wire to
perform the test. Porous titanium samples made on
EBM were implanted together with two commercially
available materials: samples with chemically etched
surface and samples with sintered microsphere surface
coating.
For evaluation of bone in-growth, EBM control
samples were compared to samples provided by two
medical device manufacturers, Biovac and Eckermann. Five samples of each type were implanted in
the femur of rabbits (Figure 9a and 9b). The control period was 8 weeks. To compare the results of
osseointegration between samples, a Region of Interest
(ROI) was selected as a part of the sample completely
introduced to the bone (delimited by fuchsia color
on figure 9c and 9d). In the Table 7, VROI corresponds to a total volume of ROI, Vimplant is the volume
Specimen
Ecker 1
Ecker 2
Biovac 1
Biovac 2
P-700-1
P-700-2
Preliminary pore size measurement.

VROI
Vimpl
Vbone
V
% Vbone
void
Fpullout
mm3
mm3
mm3
101.81 96.20
110.04 108.71
107.66 35.35
103.53 41.15
4.93
1.14
45.96
53.86
87.88
85.71
63.56
86.34
21265
21265
967323
967323
687231
687231
taken by the implant and Vbone is the volume taken

by bone cells. Finally, the last two columns show the
percentage of remaining (void) volume taken by bone
cells and the pull-out force necessary to extract the
implant.
The results show that the bone occupied a very high
percentage of void space. Bearing in mind that the
void volume in the case of EBM samples is much
bigger than in the case of Bio-Vac, it has reached a
surprisingly high level of in-growth (up to 86% of
the void space was filled). Finally, no adverse biological effects, such as tissue inflammation or implant
rejection, were noticed in these 8 weeks.
119
REFERENCES
CONCLUSIONS
In this paper, a full biological and mechanical characterization of porous titanium made with EBM is
offered. The results of this study show that the porous
titanium made with EBM has mechanical properties
similar to those of commercially available materials
and can compete with them on an equal basis. Also, the
biological behaviour of the studied material showed it
to be very good with outstanding osseointegration and
admirable pull-out results and, of utmost importance,
without any adverse biological effects. These results
back up the previous work already published in the
literature [5].
On the other hand, engineered porous material built
with EBM offers the possibility to have a designed
and completely controlled, well-interconnected porosity with constant pore size. Also, freeform fabrication
offers the possibility to build implants and prosthesis fully adapted to a patients data with minimum
post-processing.
In general terms, EBM represents a very good
alternative to conventional processes used in medical device manufacturing. The results of this study
show that, apart from being superior to alternative
porous material regarding control and design of interconnected porosity, the porous titanium made on
EBM is on the same level of mechanical properties as its commercial alternatives. In addition, the
porous titanium made on EBM has shown to be bone
ingrowth friendly. Hence, it is expected that in the
forthcoming years EBM is converted into one of
the references in the medical device manufacturing
sector.
ACKNOWLEDGMENTS
The development of the MEDIFUTUR project, which
results are presented in this paper, has been possible
thanks to the financial support of the Government of
the Valencian Community and its Council of Health
through the framework of Industrial R&D Projects
of Special Relevance for the Valencian Autonomous
Region.
Imwinkelried T. et al. 2007. Mechanical properties of openpore titanium foam. Journal of Biomedical Materials
Research Part A Vol. 81A, Iss. 4, pages 964970.
Hong T.W. et al. 2008. Fabrication of porous titanium scaffold
materials by a fugitive filler method, J Mater Sci: Mater
Med 19: 34893495.
Zardiackas L. et al. 2001. Structure, Metallurgy, and Mechanical Properties of a Porous Tantalum Foam, Journal
of Biomedical Materials Research Vol. 58, Iss. 2, pg
180187.
Parthasarathy, J. et al. 2010. Mechanical evaluation of porous
titanium (Ti6Al4V) structures with electron beam melting
(EBM), Journal of the mechanical behavior of biomedical
materials 3 (2010) 249259.
Thomsen P. et al. 2009. Electron beam-melted, freeform-fabricated titanium alloy implants: Material surface
characterization and early bone response in rabbits, Journal of Biomedical Materials Research Part B: Applied
Biomaterials Vol 90B, Iss 1, pg 3544.
J. Ferris et al. 2010. Development of an innovative generation of customized medical devices to be produced by
rapid manufacturing technologies. 9th Int. Symposium
of Computer Methods in Biomechanics and Biomedical
Engineering, 24-27 February 2010 Valencia.
J. Delgado et al. 2010. FABIO project: Development of
innovative customized medical devices through new biomaterials and additive manufacturing technologies. 3rd
Int Conference on Additive Technologies, iCAT 10, 2224
September 2010 Nova Gorica, Slovenia.
Petrovic V. et al. 2011. Additive Layered Manufacturing:
Industrial applications through case studies. Int. Journal
of Production Research 49(4).
Heinl P. et al 2008. Cellular Ti6Al4V structures with interconnected macro porosity for bone implants fabricated
by selective electron beam melting. Acta Biomater 4(5):
153644.
Ponader S. et al. 2010. In vivo performance of selective electron beam-melted Ti-6Al-4V structures. J Biomed Mater
Res A. 92(1): 5662.
Li X. et al. 2010. Fabrication and compressive properties
of Ti6Al4V implant with honeycomb-like structure for
biomedical applications. Rapid Prototyping Journal. Vol.
16 Iss. 1.
Haslauer C.M. et al. 2010. In vitro biocompatibility of titanium alloy discs made using direct metal fabrication. Med
Eng Phys. 32(6): 64552.
Marin, E. 2010. Characterization of cellular solids in Ti6Al4V
for orthopaedic implant applications: Trabecular titanium. Journal of the Mechanical Behavior of Biomedical
Materials. Vol. 3, Iss. 5, 373381.
120

Enabling technologies for robotic organ printing

R.A. Rezende, F.D.A.S. Pereira, B.D.T. Kemmoku & J.V.L. da Silva
Renato Archer Information Technology Center CTI, Campinas, Brazil
V. Mironov
Medical University of South Carolina MUSC, Charleston, US
V. Kasyanov
Riga Stradins University, Riga, Latvia
Turlif Vilbrandt
Uformia, Norway
ABSTRACT: Organ printing is defined as a computer-aided layer-by-layer additive robotic biofabrication of

functional human 3D tissue and organ constructs using self-assembling tissue spheroids as building blocks.
During last decade organ printing has been rapidly emerged as a potentially superior alternative of classic solid
scaffold-based approach in tissue engineering. It is becoming increasingly obvious that organ printing could not
be reduced anymore to just simple one step bioprinting process and it rather represents an integrated complex of
enabling technologies which could be arranged into organ biofabrication line. The information technologies such
as computer-aided design, related software and mathematical models and computer simulations are critically
important integral technological components of organ printing technology which are enabling precision placement of tissue spheroids in 3D space according specially designed instructive blueprint. Another technologies
enabling organ printing include clinical cell sorters, scalable tissue spheroids biofabricators, microfluidics-based
tissue spheroid encapsulators, robotic bioprinters, irrigation dripping perfusion bioreactors integrated with noninvasive and non-destructive biomonitoring systems and biosensors. This review outlines most important recent
advances and potential future directions in the development of enabling technologies for organ printing.
1
INTRODUCTION
The classic solid scaffold-based approach in tissue

engineering is in essence a top-down approach. Precision placement of cells with resulted high density
as well as unsolved problem of effective vascularization is the main intrinsic limitations of classic solid
scaffold-based approach. Organ printing or robotic
bioprinting is a bottom-up approach and it represents
very promising and potentially superior alternative to
classic approach in tissue engineering. Organ printing
is usually defined as a computer-aided robotic
layer-by-layer additive biofabrication of 3D functional human constructs using self-assembling tissue
spheroids as building blocks (Mironov et al. 2003;
Mironov et al. 2008; Mironov et al. 2009). Organ
printing is a biomedical variant of rapid prototyping
technology currently known as additive manufacturing
which includes several steps: pre-processing, processing and post-processing. Initially it was assumed that in
order to bioprint functional 3D human tissue or organ
construct using organ printing technology it will be
sufficient to have the most essential elements: bioink,
both authors made equal contribution to this paper.

correspondence author.
Figure 1. Organ Biofabrication cycle (CTI, Campinas,

Brazil).
biopaper and bioprinter (Mironov et al. 2009). However, the pathway from computer-aided design or
blueprint to implanted inhuman body bioprinted construct is more complex and includes several essential
phases: design, production, maturation and clinical
phase (Fig. 1).
It is becoming increasingly obvious that every step
on this pathway requires development or employing of
special technologies which can be defined as enabling
technologies for organ printing. Thus, the successful
121
clinical translation of organ printing technology will

depend not only on the development of robotic bioprinter but rather on all complex of related enabling
technologies and their seamless integration. In this
context, organ printing technology represents not just
single tool or device but rather group of seamlessly
integrated enabling technologies or organ biofabrication line (Mironov et al. 2011). In this review we
outline recent advances in development of enabling
technologies for organ printing.
2
BIOCAD OR BLUEPRINTS FOR ORGAN

PRINTING
Organ printing as any rapid prototyping technology in

essence is an information technology because it transforms virtual reality of computer-aided design (CAD)
or organ blueprint into physical reality or bioprinted
organ (Figs 2a,b). Although CAD is well established
technology in additive manufacturing with standard
software, there is no detailed blueprint for printing complex 3D human organs. Human anatomy is
also well established discipline with detailed knowledge microscopic organization of all organs and tissues. Clinical bioimaging technologies are constantly
improving and resolution of Magnetic Resonance
(MRI), Computed Tomography (CT) and Micro-CT is
constantly improving or already approaching closely
the desirable level of resolution which is sufficient
enough to get necessary anatomical and histological
information for organ bioprinting (Fig. 2c). Computeraided reconstructions of serial histological sections
of human organs can provide information about histological organization of human organ which is not
yet possible to get using in vivo bioimaging. Finally,
mathematical modeling of anatomical and histological
structures is increasingly used as additional powerful
tool to design blueprint. Thus, the real challenges in
designing CAD of human organs is to not absence
of required anatomical information, but rather the
ways how to transform this accumulated knowledge of
human anatomy and histology into a viable blueprint
with sufficient and necessary instruction for robotic
printer how to print human organ.
Organ printing using tissue spheroid as building
block is based not on continuous or analog bioprinting
but rather on digital or quantal (particular, dropletbased) bioprinting. According our best knowledge the
most of existing CAD software are not designed for
digital bioprinting. There are several intrinsic limitations of existing CAD software based on traditional
approach especially in case of creating STL file for
digital or droplet-based bioprinting. Recent advances
in development of software based on novel approach
so-called function representation allow overcoming
these limitations (Fig. 2d). The main advantages of
using novel function representation software are that
it enables digital bioprinting without STL file with
better resolution and right data.
Thus, blueprint for digital bioprinting human organ
using tissue spheroids as a building blocks or digital
Figure 2. Software for organ printing (CTI, Campinas,

Brazil): (a) Computer-aided design of vascular tree; (b)
CAD of Y-shape segment of vascular tree; (c) 3D model
by medical imaging obtained with InVesalius; (d) Computer simulation of post-printing evolution of bioprinted
tubular construct bioprinted from self-assembling tissue
spheroids using function representation software (Uformia,
Norway).
materials does not exist yet, but it is technologically

feasible to develop such blueprint using novel function
representation software.
CLINICAL CELL SORTERS
Finding clinically relevant cell source is the main problem of any tissue engineering technology. Ideal cell
122
Figure 3. Commercially available clinical cells sorters:

(a) Commercial clinical cell sorter Celution developed
by Cytori Therapeutics, San Diego, USA in collaboration
with Olympus, Japan and commercialized by General Electric Health Care, USA; (b) Commercial clinical cell sorter
developed by TissueGenesis Inc, Honolulu, Hawaii, USA.
source for tissue engineering must provide enough

amounts of cells; it must have or it could be possible to turn these cell source by directed differentiation
into desirable variety of histo- and organo-specific cell
phenotypes; finally, ideal cell source must be autologous and not induce undesirable and unwanted side
effect such as immune response and rejection (Visconti et al. 2010). In case of organ bioprinting all
these requirements are even more important because
for printing human organ we need millions of and
billions of human cells with extremely diverse phenotype. For bioprinting of certain human organs we
need large amount of several dozens of different cell
phenotypes. The good news that there are already commercially available clinical cell sorters developed by
Cytori Therapeutics, San Diego, USA and TissueGenesis, Honolulu, Hawaii, USA (Fig. 3). These clinical
sorters are focusing on exploring adipose (fat) tissue
as clinically relevant source of human cells and enable
rapid (during just 12 hours) isolation of large amounts
of adipose tissue derived adult stem cells. Clinical cell
sorters are subject of FDA approval and approval of
such devices by regulatory agencies can dramatically
enable collection of necessary amount of human cell
types and, thus, make organ printing technology even
more feasible.
BIOREACTOR FOR STEM CELLS

PROPAGATIONS
Bioreactors for stem cells propagation are necessary

if amount of human cells which is possible to get from
human cell source are not sufficient in numbers in
order to bioprint human organs for example human
dental pump stem cells or bone marrow derived stem
cells. In essence it is an issue of scalability. Commercially available bioreactors for stem cells propagation
are already reality. For example, Aastrom Inc, Ann
Arbor, Michigan, USA (Fig. 4). Development of new
more efficient type of bioreactor for human stem cell
propagation is important challenge and it is essential for enabling organ printing. It remains to be seen
how functionality of these bioreactor will evolve. Will
Figure 4. Bioreactor for stem cells propagation (Aastrom

Inc, Ann Arbor, MI, USA).
bioreactor be able always designed only to propagate

cells and increase their number or they will also be
able to direct their differentiation in desirable cell
phenotype.
5
SCALABLE TISSUE SPHEROIDS

BIOFABRICATORS
Simple calculations based on known size and volume of human organ such as kidney, liver and lung
and diameter and volume of tissue spheroid strongly
indicate that in order to bioprint human organ of desirable size it will be necessary to develop technology
for scalable biofabrication of million tissue spheroids.
Thus, scalable tissue spheroid biofabrication must be
considered as one on critically important technology
enabling organ printing. There are many different technologies for tissue spheroid biofabrications (Linz &
Chang 2008) but not all of them could be considered
as scalable methods of tissue spheroids biofabrication.
We are focusing our attention here on presenting three
technologies which we believe are most promising
emerging methods for scalable tissue spheroids biofabrication.
5.1 Modified classic hanging drop method
Classic hanging drop method is one of most simple
and most popular methods of tissue spheroids biofabrication based on gravity induced cell aggregation
(Kelm et al. 2003). However, it has certain limitations which include heterogeneity in tissue spheroid
size and potential of formation of undesirable satellite tissue microspheroids as well as tissue spheroids
with irregular shape (Mehesz et al. 2011). Recently
this technology was improved and modified by making
these spheroids droppable. This modification makes
hanging drop technology suitable for robotic automatization (robotic dispensing of cell suspension) and
very user friendly especially for collection of fabricated tissue spheroids. This technology is now commercialized by Swiss start-up company InSphero
(http://www.insphero.com). Tissue spheroids can be
used not only as building blocks in organ printing technologies but also as 3D human tissue assay
for drug discovery and toxicity studies, modeling
human disease such as cancer and testing novel
123
nanotechnology based drug delivery vesicles. It is

safe to predict that development of enabling technologies for organ printing such as scalable robotic
tissue spheroid biofabrication can lead to many spinning off applications in regenerative medicine and
tissue engineering beyond organ printing (for example such as extracorporeal liver device). Finally, tissue spheroids fabrication technologies can be also
used for development of industrial scale production in vitro meat (Mironov et al. 2009). The most
interesting fact that there are already several companies (InSphero, Switzerland; Microtissue, USA;
Nano3D Biosystems, USA; Stem Cell Technologies,
USA; Co.don, Germany; OsteoSphere, USA; Shrink
Nanotechnologies, USA) which are already actively
commercializing tissue spheroids biofabrication technologies. Moreover, some of these companies such
as Co.don (Germany) have commercially available
products (chondrospheres) already in clinical trial.
5.2
Biofabrication tissue spheroids using

non-adhesive hydrogel
In order to overcome limitation of hanging drop

method many group are developing tissue biofabrication technologies using microfabricated devices or
molds. Jeff Morgan group from Brown University
has developed molds using rapid prototyping technology which allow creating recession in non-adhesive
agarose hydrogel and thus enables biofabrication of
standard size tissue spheroids (Figs 5a,b) (Napolitano
et al. 2007). This technology has been commercialized by start-up company Microtissues. We recently
slightly modified original Jeff Morgan group technology and make it more suitable for robotic dispensing
by employing 96 microwell and Eppendorff (eMotion
5075) robotic pipetting device (Fig. 5c) (Mehesz et al.
2011). Canadian stem cells group of Peter Zandstra
has successfully developed and tested molding device
which looks like inverted pyramid and which also
has been commercialized by Stem Cell Technologies
(Dang et al. 2004).
USA company Shrink Nanotechnologies recently
developed and commercialized honeycomb-like pattern or so-called StemDisk technological platform
(see for details Shrink Nanotechnologies website
http://www.shrinknano.com/products/product-tools/
stemdisc). Our team is working on improving this
design keeping in mind that rounded bottom is essential for rapid biofabrication of tissue spheroid of
ideal ball like shape and the fact that some cells
can be spuriously deposited on empty space between
recessions in non-adhesive agarose hydrogel and can
generate undesirable satellite tissue microspheroids
of irregular size and shape. CTI is using as a mold
a rapid manufacturing technology (Objet Geometries Inc.) that can have very good resolution and
small details of more complex geometries. Comparative analysis of different methods of scalable tissue
spheroid biofabrication using non-adhesive surfaces
and hydrogels is a necessary step in further improving
Figure 5. Scalable robotic biofabrication of tissue

spheroids. (CTI, Campinas, Brazil): (a) tissue spheroids
biofabrication technology; (b) prototyped mol 96 microwell mold; (c) robotic pippeting device eMotion-5075
(Eppendorf, Germany).
and advancing this technology. Robotization of tissue

biofabrication technology is another important task.
Tissue spheroid biofabricated from stem cells can be
use to generate chondrospheres, osteospheres, cardiospheres, myospheres, neurospheroids , hepatospheres
using directed differentiation by differentiation cocktails and other tissue- and organo-specific building
blocks for organ bioprinting.
Recently developed at Centro de Tecnologia
da Informao Renato Archer, Campinas, Brazil
novel concept of lockable micro-scaffold or simply Lockyballs (Fig. 6) represents bottom-up
solid scaffold technology in combination with tissue
spheroid biofabrication technology. This integrated
approach can lead to development of lockable tissue spheroids or novel simple and elegant method of
rapid in vivo 3D tissue biofabrication. This approach
could be especially very promising for in vivo cartilage
and bone tissue engineering.
5.3 Microfluidics based tissue fabrication
technologies
Microfluidics especially digital or droplet-based
microfludicis opens unique opportunities for large
scale biofabrication of tissue spheroids. Microfluidcs
device allow to biofabricate 10,000 droplets per second. Although microfluidics have been already used
for cell encapsulation and fabrication colloid droplet
of complex internal structure using elegant device
124
Figure 6. Computer-aid design of lockable micro-scaffold

spheres or Lockyballs (CTI, Campinas, Brazil).
digital droplets but rather the way to prevent preliminary and undesirable tissue spheroid fusion in robotic
bioprinter cartridge before actual printing. Encapsulation of coating of tissue spheroids prevents tissue
spheroid fusion and turning them into rod-like or linear fiber structure in bioprinter cartridge. For example,
encapsulation into hyaluronan will provide lubrication effect and prevent preliminary tissue spheroids
fusion in bioprinter and the same time will permit tissue spheroids fusion after printing. Thus, biomaterials
or hydrogels for tissue spheroids encapsulation must
from one side prevent tissue fusion during processing or bioprinting and at the same time be permissive
for tissue fusion during post-processing after finishing bioprinting process. Thus, it is obvious that such
encapsulation hydrogel must be biodegradable and
sacrificial.
Figure 7. Microfluidic device for scalable biofabrication of

uni-lumenal and janus-like tissue spheroids (CTI, Campinas,
Brazil).
developed by David Weitz group a Harvard University,

application of digital microfluidics for scalable biofabrication of tissue spheroid is only started (Rhutesh
et al. 2008). We are developing design of microfluidics
device which will enable scalable biofabrication of
solid, uni-limenal and even janus-like tissue spheroids
(Fig. 7).
Manufacturing and experimental testing of such
microfluidics device together with funding proper
photo-sensitive and sacrificial hydrogel can made
important contribution to development of novel
method of scalable tissue spheroid biofabrication.
Combination of microfluidics with nanotechnology by
using magnetic nanoparticle offer even more exciting perspective to develop tissue spheroids capable for
self-directed self-assembly as it was proposed in one
of our reviews on application of nanotechnology in
tissue engineering (Mironov et al. 2008). Finally, it
could eventually lead to development of high speed
magnetic bioprinter and thus further improve and
enable organ printing.
6 TISSUE SPHEROIDS ENCAPSULATORS

Cell encapsulation is well established technology. In
context of organ printing there are several possible
ways to explore encapsulation technology. It is possible to create tissue spheroids by encapsulation of cells.
However, the cell density is usually low. More interesting is the idea of encapsulation of tissue spheroids
which already have high level of cell density. In this situation encapsulation is not the way to create printable
ROBOTICS BIOPRINTERS
Robotic bioprinter is a most essential element of

emerging bioprinting technology.
Bioprinter can be defined as a computer-aided
robotic device for precision layer by layer placement
of cells and minitissue such as tissue spheroids into
3D space. In this context bioprinter is some a sort
of robotic tissue bioassembler. The most essential
component of any bioprinter is X-Y-Z axis robot.
Another two important essential elements are nozzles
or dispensing system and system of operational control. Thus, organ printing belongs to group of robotic
manufacturing technologies. One of authors predicted
several years ago that bioprinter will be soon as common in biomedical lab as microscopes. However, the
absence of commercially available and affordable bioprinter was the main limitation and impediment for
rapid development of organ printing technology. Now
situation is improved. There are already commercially
available professional bioprinters (Fig. 8) as well as
variety of relatively cheap and affordable bioprinters
(Fig. 9).
Thus, production of commercial bioprinters is
already viable real business and direct manifestation of commercial potential of still evolving organ
printing and bioprinting technology. Advanced Tissue Biofabrication Center at MUSC together with
CTI, Campinas is developing and testing novel high
resolution robotic bioprinter which will allow additive biofabrication of 3D tissue and organ constructs
by sequential spraying hydrogel layers and precision
computer controlled punching tissue spheroids in it
(Fig. 10). CTI is also modified cartridge system of very
popular Fab@home 3D printer and created Fab@CTI
3D bioprinter capable for dispensing hydrogel with
living cells and fabricating hydrogel based scaffolds
(Figs 9a,b).
It is safe to predict that design and manufacturing of novel type of 3D bioprinters as well as testing
appropriate biomaterials and hydrogels for such bioprinters is a booming area in biofabrication field.
125
Figure 10. Medical University of South Carolina (MUSC)

bioprinter on development and testing at CTI, Campinas,
Brazil: (a) General view of MUSC robotic bioprinter fabricated by Izumi International, Greenville, SC, USA; (b)
X-Y-Z axis robot (Janome, Japan); (c) Three nozzles (two
Nordson sprayers and one Fishman robotic dispenser in the
center) MUSC Robotic Bioprinter will be commercialized by
start-up company CUSPIS LLC, Charleston, SC, USA.
Figure 8. Commercial bioprinters: (a) BioFactory (Delta

Robotics, Switzerland); (b) NovoGen MMX (Organovo,
San Diego, CA, USA); (c) The BioiAssembly Tool (BAT)
(Sciperio/nScript, Orlando, FL, USA); (d) 3D-Bioplotter
(Envisiontech, Germany).
Figure 11. Conceptual design of irrigation dripping tripled

perfusion bioreactor for bioprinted organs (Mironov et al.,
2009).
Figure 9. Affordable commercial 3D printers: (a,b)

Fab@CTI 3D printer; (c) RepRap 3D printer; (d) World
smallest 3D printer (Technical University of Vienna, Austria);
(e) DesktopFab 3D printer.
Comparative analysis of existing bioprinting systems

is beyond original scope of this article and deserves a
special comprehensive review.
IRRIGATION DRIPPING TRIPLED

PERFUSION BIOREACTOR
Bioreactor is an essential component of classic tissue engineering. There are several reviews (Martin
et al. 2004; Mironov et al. 2006) and at least two
books specially devoted to bioreactor technologies
(Chaudhuri & Al-Rubeai 2005).
Initially bioreactors have been used as tool for
enhancing cell seeding on solid scaffolds. Perfusion
bioreactors additionally have been used as a tool
for providing mechanical conditioning of tubular tissue engineered constructs. Finally, bioreactor is also
some sort of container which allows keeping tissue
engineered constructs in wet environment and thus
maintains their viability and also serves as packing and
transportation device for matured tissue engineered
constructs. In case of organ printing the function of
perfusion bioreactor is to buy time necessary for
post-printing tissue fusion, remodeling and maturation of bioprinted constructs. The bioprinted tissue
construct even with build in vascular system is not
ready for immediate intravascular perfusion because it
takes certain time for vascular tissue spheroids to fuse
and resulted vascular tree to be sufficiently mature
and ready for intravascular perfusion. We introduce
novel concept of irrigation dripping tripled perfusion
bioreactor in order to allow bioprinted tissue construct
including its vascular tree to maturate before initiating of biomimetic intravascular perfusion (Fig. 11).
Three perfusion circuits in this novel type of perfusion bioreactor serve three purposes: one perfusion
system provides wet environment around the printed
constructs; second perfusion system is designed for
intravascular perfusion of maturated build in vascular
tree; and, finally, third perfusion circuits is designed
for enabling the temporal interstitial flow through
removable temporal porous minitubes. These removable porous tubes also provide temporal support and
serve as some sort of non-biodegradable but removable
supporting structure or serve as an analog of scaffold
in classic tissue engineering. The distance between
these tubes as well as their porosity must be designed
126
Figure 12. Computer-aided design of porous tube for irrigation dripping tripled perfusion bioreactor (CTI, Campinas,
Brazil).
based on mathematical modeling and computer simulation. The design of porous removable minitubes for
irrigation bioreactor is shown on Figure 12.
The removable minitubes must be as thin as possible and novel strong material (composite with carbon
nanotubes) must be probably used for manufacturing of such minitubes. Finally, these minitubes must
be coated with non-adhesive porous inert Teflon-like
biomaterial which will avoid cell adhesion and consequently their non-destructive removal after bioprinted
tissue maturation and switching from interstitial to
intravascular perfusion. Rationale design of irrigation
dripping tripled perfusion bioreactors is first step on
the way to build such novel type of bioreactor. Selection and testing proper materials for removable porous
minitube is a second step. Finally, fabrication and testing bioreactor is third and last step. It is interesting
that removable porous tube can have additional functions: they can as biosensors of functional maturation
of bioprinted tissue, they can be used for delivery soluble extracellular matrix molecular and even cells for
accelerating tissue maturation, and finally they can be
use for providing electric stimulation and other physical methods enhancing functional tissue maturation.
The design, manufacturing, testing and further optimization of such novel type of perfusion bioreactor
are not trivial but feasible engineering task. At least the
conceptual design of such bioreactor has been already
accomplished. Economic irrigation dripping approach
which was already successfully used in agricultural
irrigation will allow dramatically reduce cost of perfusion by more rational using of expensive perfusion
cell culture media. Thus, it is obvious that development
of perfusion bioreactor is critically important and it is
essential enabling technology for organ printing.
CONCEPT OF ACCELERATED TISSUE

MATURATION
It is important to realize that outcome of bioprinting is

not a functional tissue construct immediately suitable
for implantation. In order to transform bioprinted 3D
tissue and organ construct into functional tissue the
bioprinted construct must undergo accelerated issue
maturation. This process is called accelerated tissue
maturation. It is based on fundamental assumption that

tissue spheroid can fuse and relatively fast can be turn
into functional issue using specially designed accelerated tissue maturation technologies. Those chemical
and physical factors which enable accelerated tissue maturation are logically to define as maturogenic
factors or simply maturogens. It has been shown
that relatively simple in vitro tissue fusion and tissue
enveloping assays can be employed for systematic discovery and screening maturogens (Hajdu et al. 2010).
Moreover, it has been already demonstrated sign these
assays that TGF betta, serotonin, periostin and transglutaminase are candidate maturogenic molecules.
There are strong evidence that 3D confined minitissues
such as tissue spheroids have propensity for accelerated tissue maturation due to hypoxia induced collagen
synthesis and/or so-called excluding volume effect of
molecular crowding effect which can be enhanced
by perfusion soluble extracellular matrix molecules
such as fibronectin (Lareu et al. 2007a,b; Robinson
et al. 2004). It has been also demonstrated that cell
to cell signaling and communication (for example,
between endothelial and smooth muscle cells in case
of vascular tissue) is essential in dramatic enhancing synthesis, deposition and bioassembly of main
structural components of extracellular matrix such as
collagen and elastin which determines material properties of bioprinted tissues. Finally, it has been also
shown that maturogens such as serotonin, periostin and
transglutaminase are enhancing extracellular matrix
molecules. These molecules are possible ingredients
of future effective maturogenic cocktail. Further discovery, experimental validation and high throughput
screening of potential candidate maturogenic factors
is an important task in development accelerated tissue
maturation technology. Thus, methods of accelerated
tissue maturation are important enabling technologies
for organ printing.
10
STRATEGIES FOR NON-INVASIVE AND

NON-DESTRUCTIVE BIOMONITORING
OF BIOPRINTED TISSUE MATURATION
One of principal and not properly addressed question in emerging organ printing technology is how
to estimate that bioprinted 3D tissue and organ construct is became functional and ready to implantation.
There are several possible explorable strategies in this
context. Using parallel development of several bioprinted 3D tissue and organ constructs with sequential
scarifying some of them for invasive and destructive
testing of tissue maturation and level of functionality is initially possible but it will be too expensive.
Thus, the ideal evaluation methods and technologies
must be of course non-invasive and non-destructive.
One possible non-invasive approach is a biochemical evaluation of perfusion and search for biomarker
of tissue maturation. Using of small amount of randomly redistributed sentinel-cell with self-reporting
gene which start to express fluorescent markers such as
127
GFP when certain level of maturation and functionality have been achieved is another interesting approach.
Finally, removable perfusion tube in proposed bioreactor (see above) can be used as a biosensor for estimation of tissue maturation. For example, biosensor
can estimate impedance or tissue electroconductivity and thus judge tissue maturation. Non-invasive
and non-destructive bioimaging technology such as
ultrasound and MRI can be also employed. Perfusion technique with inert tracer in combination of
bioimaging (two photon microscopy) can be also valuable tool for estimation tissue integrity and maturation
and functionality. Finally, non-invasive estimation of
bioprinted tissue and organ constructs material properties is another important marker of maturation. For
example, non-invasive clinical modalities for estimation liver cirrhosis and fibrosis can be adapted for
estimation level of maturation of bioprinted tissue. It
is safe to predict that with the progress and advances
in development of organ printing technology the role
of non-destructive and non-invasive methods for estimation tissue maturation and functionalities will only
increased. Without development of effective methods of pre-implantation and post-implantation quality
control of bioprinted organs functionality, maturation and integrity their approval for clinical use and
implantation by regulatory agencies will be practically
impossible.
11
CONCLUSIONS AND PERSPECTIVES
Organ printing is a biomedical variant of 3D rapid prototyping technology or computer-aided layer-by-layer

additive biofabrication using tissue spheroid as building blocks. It is becoming obvious that organ printing
could not be reduced just to robotic bioprinting. It
is either over-simplification (reduction all biological aspects to viability issue) or misunderstanding of
intrinsic biological complexity (dynamic and evolving
nature of directly self-assembled tissues and organs)
associated with organ bioprinting technology.
Rapidly emerging organ printing is enfolding rather
as an integrated complex of multiple enabling technologies. These technologies have, however, several
common characteristics: it is automated computeraided robotic technologies; they are based on special
information technologies and specially designed software; they are processing and operating with living
biological materials which imposes severe technological restraints; finally, the effective employing of these
enabling technologies will depend on their seamless
integration and high degree of compatibilities. Ideally,
these technologies must be flexible enough to be seamlessly integrated and be used for various cells, tissue
and organs. It is not possible to exclude development of
tissue- and organo-specific bioprinting technologies.
It is also important to underline that bioprinting and
biofabrication technologies as all rapid prototyping
technologies in essence are information technologies,
because they are transforming virtual information or
organ blueprint into physical biological reality bioprinted organ. It means that automation of bioprinting
processes and development of integrated organ biofabrication line will require sophisticated operational control including development of novel type of software.
Moreover, designing of blueprint for organ printing as well as high resolution and reproducible digital
(or droplet-based) bioprinting methods will require
optimization of existing CAD software and development of new generation of BioCAD software based on
novel function representation approach. Biofabrication technologies as any other modern manufacturing
technologies are heavily depended on computer-aided
design, computer simulation, mathematical and computational modeling and in silico or virtual testing.
Thus, it is safe to predict that the first complex human
organ such as kidney will be bioprinted at first
in silico.
Development of virtual organ biofabrication line is
another interesting approach and logical and advanced
way for industrial and clinical translation of emerging
organ printing technology. Virtual manufacturing (for
example, virtual refinery) is already well established
approach for designing industrial process in oil industry. Virtual organ biofabrication line must combine all
possible visual information about machines, devices
and processes in for of interactive computer generated virtual animated presentation including capacity of virtual visit to organ biofabrication plant as
an avatar to visually observe and virtually interact
with all components of organ biofabrication line. Virtual organ biofabrication line could be an important
computational tool for screening optimal design and
biofabrication process engineering of future organ
biofabrication line and associated biofabrication and
bioprinting processes, excellent educational tool and it
will allow to integrated national and even international
research team. The virtual organ biofabrication line
will be a necessary step toward development of
real organ biofabrication industrial plant. Thus,
computer-aided design, computer simulation, mathematical modeling, virtual reality methods and informational technologies in general are essential tools
for development of organ printing technologies and
industrial scale biofabrication process engineering.
ACKNOWLEDGEMENTS
This work was partially funded by NSF R-II grant
South Carolina Project for Organ Biofabrication,
The So Paulo Research Foundation (FAPESP), The
Brazilian Institute of Biofabrication (INCT-Biofabris),
and The National Council for Scientific and
Technological Development (CNPq) through CTI/PCI
program.
REFERENCES
Chaudhuri, J. Al-Rubeai, M. (editors). 2005. Bioreactors for
Tissue Engineering: Principles, Design and Operation.
Springer; 383 pages.
128
Dang, S.M., Gerecht-Nir, S., Chen, J., Itskovitz-Eldor, J.,

Zandstra, P.W. 2004. Controlled, scalable embryonic stem
cell differentiation culture. Stem Cells. 22(3): 27582.
Hajdu, Z., Mironov, V., Mehesz, A.N., Norris, R.A.,
Markwald, R.R., Visconti, R.P. 2010. Tissue spheroid
fusion-based in vitro screening assays for analysis of tissue
maturation. J Tissue Eng Regen Med. Dec; 4(8): 65964.
Kasper, C., Van Griensven, M. Prtner, R. (editors). 2009.
Bioreactor Systems for Tissue Engineering (Advances in
Biochemical Engineering/Biotechnology), Springer; 271
pages.
Kelm, J.M., Timmins, N.E., Brown, C.J., Fussenegger, M.,
Nielsen, L.K.. 2003. Method for generation of homogeneous multicellular tumor spheroids applicable to a wide
variety of cell types. Biotechnol Bioeng. Jul 20; 83(2):
17380.
Lareu, R.R., Subramhanya, K.H., Peng, Y., Benny, P., Chen,
C., Wang, Z., Rajagopalan, R., Raghunath, M. (2007).
Collagen matrix deposition is dramatically enhanced in
vitro when crowded with charged macromolecules: the
biological relevance of the excluded volume effect. FEBS
Lett. Jun 12; 581(14): 270914.
Lareu, R.R., Arsianti, I., Subramhanya, H.K., Yanxian, P.,
Raghunath, M. 2007. In vitro enhancement of collagen
matrix formation and crosslinking for applications in
tissue engineering: a preliminary study. Tissue Eng. Feb;
13(2): 38591.
Lin, R.Z., Chang, H.Y. 2008. Recent advances in threedimensional multicellular spheroid culture for biomedical
research. Biotechnol J. Oct; 3(910): 117284.
Martin, I., Wendt, D., Heberer, M. 2004. The role of bioreactors in tissue engineering. Trends Biotechnol. Feb;
22(2): 806.
Mehesz, A.N., Brown, J., Hajdu, Z., Beaver, W., da Silva,
J.V.L., Visconti, R.P., Markwald, R.R., Mironov, V. 2011.
Scalable robotic biofabrication of tissue spheroids. Biofabrication. Jun; 3(2): 025002. Epub 2011 May 12.
Mironov, V., Boland, T., Trusk, T., Forgacs, G, Markwald, R.R.
2003. Organ printing: computer-aided jet-based 3D tissue
engineering. Trends Biotechnol. Apr; 21(4): 15761.
Mironov, V., Kasyanov, V.A.,Yost, M.J., Visconti, R., Twal, W.,

Trusk, T., Wen, X., Ozolanta, I., Kadishs, A., Prestwich,
G.D., Terracio, L., Markwald, R.R. 2006. Cardiovascular tissue engineering I. Perfusion bioreactors: a review.
J Long Term Eff Med Implants. 16(2): 11130.
Mironov, V., Kasyanov, V., Drake, C., Markwald, R.R. 2008.
Organ printing: promises and challenges. Regen Med. Jan;
3(1): 93103.
Mironov, V., Kasyanov, V., Markwald, R.R. 2008. Nanotechnology in vascular tissue engineering: from nanoscaffolding towards rapid vessel biofabrication. Trends
Biotechnol. Jun; 26(6): 33844.
Mironov, V., Visconti, R.P., Kasyanov, V., Forgacs, G,
Drake, C.J., Markwald, R.R. 2009. Organ printing: tissue
spheroids as building blocks. Biomaterials. Apr; 30(12):
216474.
Mironov, V., Trusk, T., Kasyanov, V., Little, S., Swaja,
R., Markwald, R 2009. Biofabrication: a 21st century manufacturing paradigm. Biofabrication. Jun; 1(2):
022001.
Mironov, V., Kasyanov, V., Markwald, R.R. 2011. Organ printing: from bioprinter to organ biofabrication line. Curr
Opin Biotechnol. Mar 16. [Epub ahead of print].
Napolitano, A.P., Dean, D.M., Man, A.J., Youssef, J., Ho,
D.N., Rago, A.P., Lech, M.P., Morgan, J.R. 2007. Scaffoldfree three-dimensional cell culture utilizing micromolded
nonadhesive hydrogels. Biotechniques. Oct; 43(4): 494,
496500.
Robinson, E.E., Foty, R.A., Corbett, S.A. 2004. Fibronectin
matrix assembly regulates alpha5beta1-mediated cell
cohesion. Mol Biol Cell. Mar; 15(3): 97381
Shah R. K., Shum, H.C., Rowat, A.C., Lee, D., Agresti,
J.J., Utada, A.S., Chu, L-Y., Kim, J-W, FernandezNieves, A., Martinez, C.J., Weitz, D.A. 2008. Designer
emulsions using microfluidics. Materials Today 11
1827.
Visconti, R.P., Kasyanov, V., Gentile, C., Zhang, J., Markwald,
R.R., Mironov, V. 2010. Towards organ printing: engineering an intra-organ branched vascular tree. Expert Opin
Biol Ther. Mar; 10(3): 40920.
129

The semi-automated design & manufacture of patient-specific

intervertebral disc implants
N. de Beer
Laboratory for Rapid Product Development, Department of Industrial Engineering, Stellenbosch University,
Stellenbosch, South Africa
ABSTRACT: Rapid Manufacturing has been successfully applied to produce patient-specific implants for
several orthopedic cases including cranio-maxillofacial, dental, hip, knee and foot requirements, but very little
has been published on its application for spinal implants. This may in part be due to the complex nature of spinal
bio-dynamics, but probably due to the rigorous clinical approval process required for spinal implant devices.
This paper reports on the development of a process chain for the semi-automated design and manufacture of
patient-specific intervertebral disc implants, and includes the results of in-vitro vertebral compression tests done
to evaluate the reduced risk of implant subsidence for patient-specific devices. The paper concludes with an
economic assessment of the South African artificial disc implant market and the potential advantages as well as
challenges of using customized implants.
1
INTRODUCTION
A growing number of people across the world continue to suffer from debilitating back pain, where in
many cases, the pain can become unbearable resulting
in major lifestyle adjustments. Pain may occur as a
worn disc becomes thin, narrowing the space between
the vertebrae.
If conservative or minimally invasive treatment is
unsuccessful, a total disc replacement (TDR) procedure may be prescribed in preference to a disc fusion
procedure due to a growing concern that fusion may
affect degeneration in the adjacent discs, (Cheh, et al.,
2007, Harrop, et al., 2008, Matsumoto, et al., 2009,
Higashino, et al., 2010).
These intervertebral disc implants are however not
without their own set of concerns. Typical complications that are still observed in some cases include:
subsidence and anterior migration of the disc (Van
Ooij, et al., 2003). Bertagnoli suggests that one of the
most prevalent reasons for such disc failures is incorrect positioning of the implant (Bertagnoli, 2005).
This is made more difficult by the fact that every
patients anatomy and condition requiring surgery is
unique. Manufacturers of disc implants compensate
for these dissimilarities by creating different standard
size implants. Surgeons then try to select the most suitable match during surgery, by pushing various trial
sizes into the vertebral space before placing the final
implant. This trial-and-error technique relies heavily
on the level of experience of the surgeon and could
lead to TDR device under sizing and inaccurate positioning of the implant, which could lead to implant
subsidence (Leary, et al., 2007, Shim, et al., 2005, Van
Ooij, et al., 2003, Van Ooij, et al., 2007, Cinotti, et al.,

1996, Zeegers, et al., 1999).
Most existing disc implants consist of endplates
that are designed relatively flat in comparison to the
concave boney endplate geometry. In order to accommodate the implant, the bone endplates are often
surgically reduced to a flat plane and a slot is cut to
receive the implant keel (fin-like protrusion that facilitates implant fixation). This action compromises the
strength of the vertebras cortical shell and reduces its
ability to withstand pressure and can lead to implant
subsidence or vertebral fracture (Auerbach, et al.,
2010, Lowe, et al., 2004). A more elegant solution will
be to leave the endplates as intact as possible and rather
adapt the shape of the implant to match the geometry
of the vertebrae.
One approach may be to design the implant endplates with some measure of generic concavity to
match that of the bone, based on morphometric studies of different population groups. However Van der
Houwen contends that data on the prevalent shapes
of the vertebral surfaces are scarce, citing 10 studies
that have investigated the morphometry of vertebral
bodies and their endplates, using a variety of methods
(cadaver, CT, MRI, and X-Ray). He finally concludes
that the future may lie in custom-made implants, with
for every person a perfect fit based on a pre implant
measurement using CT data (Van der Houwen, et al.,
2010).
The design and manufacture of patient-specific
implants is not a new phenomenon, with examples
of the successful application of Rapid Manufacturing
(RM) in areas such as cranioplasty implants (Poukens,
et al., 2003, Hieu, et al., 2003), oral and maxillofacial implants, finger joint implants (Gibson, et al.,
131
Figure 1. Clinical process chain for design & manufacture of patient-specific IVD endplates.
2006), customized hip and knee implants (Sercombe,

et al., 2008, Harrysson, et al., 2007), (Harrysson,
et al., 2008), dental implants (Vandenbroucke, et al.,
2007), foot implants (Schindel, et al., 2005), and
recent research within the CustomIMD Framework 6
EU project on customized intervertebral disc nucleus
replacements. Despite these examples however, to the
authors knowledge, very little work has been done to
develop patient-specific intervertebral disc implants.
This research was therefore conducted to contribute
and provide much needed progress in this area of
application.
2.2 Imaging and scanning
The process chain for customizing the design & manufacture of an intervertebral disc implant was developed
and is shown in Figure 1 below. Each step will be
discussed shortly.
The third step after acquiring CT scans is to convert

the two-dimensional images into a 3D (STL file)
model by means of a segmentation process. A variety
of software solutions are available to perform this operation. In the case of our study, Mimics (Materialise,
Belgium) was used.
2.1
2.4 Design and customization
Since information about the patients bone geometry is

required for the implant design, it is necessary to have a
CT scan taken. CT scans should be acquired at a high
spatial resolution with thin, contiguous image slices
(0.751.25 mm is ideal) and as small a field of view
(FOV) as possible while still including the patients
external contour. No gantry tilt should be applied, and
the patient must remain completely still through the
entire scan.
2.3 Data transformation
PROCESS CHAIN DEVELOPMENT
Diagnosis
The first step in the process is to ascertain the patients

condition and eligibility for a TDR procedure. After
diagnostic tests and three to six months of conservative
treatment with no significant improvement, the patient
is offered surgery based on a shared decision-making
process. Based on a set of indications or contraindications, TDR may be prescribed as a surgical solution.
The design and customization phase incorporates userfriendly Surgical Planning Tool software which was
developed in Matlab to allow surgeon involvement
during the pre-surgical planning stage. Following the
digital realignment of vertebrae, the surgeon identifies
seven anatomical landmarks on the surface of each
vertebral endplate (Figure 2). Six of these points are
132
Figure 3. Simulation process flow in relation to regulatory

approval.
Figure 2. Anatomical landmarks for design of patientspecific IVD endplates.
then used to define a spline curve which forms the

footprint profile for the endplate of the intervertebral
disc prosthesis. The seventh landmark on each vertebra defines the centre line on which the centre point
of the spherical ball-and-socket joint connection is
defined.
Once the fourteen landmarks have been identified, their coordinates are exported to a semiautomated parametric 3D CAD model that has been
designed using Autodesk Inventor Professional 2009
(Autodesk, California). The final step in the design
process is to modify the implant endplates to match
the geometry of the bone endplate surfaces. This is
done by performing a simple Boolean subtraction
between the implant and the vertebrae. STL files
of the implant along with the bones were exported
to 3-Matic software (Materialise, Belgium) where
the subtraction was performed. The subtraction step
was then followed by an undercut removal function,
to ensure that the implant can be inserted without
obstructions caused by undercuts.
2.5
Biomechanical simulation
An understanding of the vertebral movements is useful

in the assessment of typical spinal disorders (such as
instability) and the prediction of treatment outcomes.
Several simulation models of the spine have
already been reported in literature (Sun, et al., 2004),
(De Jongh, et al., 2008), and as such, this study did
not develop a new simulation model, but discusses
its relevance and application in the process chain of
implant customization as a whole.
Figure 3 depicts an expansion of step 5 of Figure 1,
showing the basic process flow during design improvements to the customized intervertebral disc implant
using simulation. Anthropometric patient information
is collected, and by making use of movement capturing technologies, patient kinematics can be collected
by allowing the patient to perform a set of predefined
movements. Given the patients information, a simulation model of the patients condition is derived through
the use of inverse kinematics. A second generic simulation model can be scaled to match the patients
basic anatomy, age and weight. This generic healthy
simulation model is then compared with the model
of the patient in terms of ranges of motion for the
same set of exercises. Based on an initial comparison, an assessment is then made in terms of what
corrective action is needed to rectify spinal alignment
and vertebral positions and in doing so, recommendations and boundary conditions are derived for the
design and placement of an intervertebral disc implant.
The implant is designed using the Surgical Planning Tool and 3D parametric CAD model described
previously.
Once designed, the patients simulation model is
updated to include the implant. A new comparison
between the implanted patient model and the generic
healthy model is then made as before. This iterative
process is repeated until a satisfactory resemblance
between the models is achieved.
2.6 Regulatory approval

An inherent difficulty for customized products lie in
the fact that each design is by definition, essentially
unique but still needs to adhere to stringent performance and safety standards. FDA and CE certification
procedures require long and rigorous sets of tests
which, at present, is not feasible for a customized
implant design. Currently however, allowance is made
for customized implants in general by having both the
patient and surgical team sign consent before implantation proceeds. Examples of such implants occur
commonly for operations in the cranio-maxillo/facial
regions, where patient anatomy already require unique
geometry solutions. This situation is however not
ideal as a long term solution. As simulation models increase in their ability to more closely resemble
the natural kinematic and dynamic behavior of the
human body, it seems apparent that such techniques
should become standard practice during the testing and
approval of a spinal implant. A specific set of simulation outcome demands must be established which
an implant design needs to adhere to before approval
can be granted. Once established, these passing criteria can be built into the simulation model process flow
as shown in Figure 3.
133
investigate improvements to the achievable contact area and stiffness between the implant-vertebral
construct.
3.1 Materials and methods
Figure 4. Rapid Manufactured patient-specific Ti6Al4V

intervertebral disc implant.
2.7
Manufacturing, post processing, & surgery
The ability of Rapid Manufacturing to produce end-use

products directly in biocompatible materials such as
titanium and cobalt-chrome, makes it an ideal choice
for manufacturing patient-specific spinal implants.
The essential material issues in the medical field
relate mostly to biocompatibility and in many cases
poor surface finish and porosity is a desirable feature
for implants to allow osteointegration of implant and
bone. In other cases where articulating surfaces (such
as knee or hip joints) are involved, surface roughness
must be very low, and may require polishing or grinding. After surface treatment, the implant components
are assembled, sterilized and packaged ready for
surgery.
As a demonstrator of the process chain described
above, a generic, non-clinical ball-and-socket type
implant was designed with an emphasis on endplate customisation. Figure 4 shows an example
of the implant that was manufactured using Direct
Metal Laser Sintering (DMLS) technology. The parts
were produced on an EOSINT M270 machine, using
Ti6Al4V powder material at the Centre for Rapid
Prototyping and Manufacturing (CRPM) of the Central University of Technology (Bloemfontein, South
Africa).
REDUCED RISK OF SUBSIDENCE
One of the key benefits expected from using patient

specific intervertebral disc implants, is the potential it
may have to reduce the risk of implant subsidence post
surgery. Subsidence depends, in part, on the stiffness
and strength of the implant-end plate interface, and
factors that influence this interface include bone mineral density, amount of cartilaginous end plate removal
during surgery, anteroposterior position of the implant
on the vertebral end plate (i.e., variable regional bone
strength), implant shape, and implant size. Therefore, in vitro laboratory compression tests were performed (both non-destructive and failure tests) to
Four spines from male cadavers (ages 4565 years,

average 52 years), which had a total of n = 88 vertebrae
(C3 to L5) were acquired under approved institutional
protocol from the Division of Anatomy and Histology,
Department of Biomedical Sciences, at the University of Stellenbosch. Bone Mineral Density (BMD)
scans revealed that three out of the four cadaver specimens did not qualify to be included in the study as one
was osteoporotic, one was osteopenic, and the third
showed the presence of kyphoplasty. Therefore one
cadaver, with a BMD of 1.081 g/cm2 remained for the
study (n = 22). From the remaining spine, two vertebrae (L4 & L5) were used in a separate sub-study
to assess the accuracy of the manufactured implants
using Rapid Manufacturing, and therefore 20 vertebral bones remained for mechanical pressure testing.
Detailed geometry information of the vertebrae for
customizing implant designs was acquired using a
calibrated Siemens Somatom Emotion 16-slice CT
scanner.
After imaging, the vertebra specimens were dissected from one another and all surrounding soft
tissue was removed. Posterior elements, which contribute toward vertebral compression mechanics, were
kept intact. A blunt scalpel was used to remove the
remaining cartilaginous end plates while taking care
not to damage or remove any bone. This preparation
technique was performed according to the method prescribed by a spine surgeon with extensive experience
in disc replacement surgery.
Each vertebral specimen was then potted by placing
the superior body centrally within a plastic container
and filling it with a liquid epoxy resin (Prime 20 LV
Epoxy Resin, with Prime 20 ULV slow hardener, AMT
Composites, Cape Town). In order to ensure that each
vertebral endplate was orientated horizontally, perpendicular with respect to the vertically applied pressure,
a support structure was 3D printed using the Z-Printer
310 system (Z-Corporation, Burlington, MA), using
ZP150 powder material (not infiltrated). The support
structure was designed using 3-Matic software (Materialise, Belgium) in such a way that one end matched
the geometry of the superior vertebral endplate and
would therefore accommodate the bone and orientate
it correctly while the resin hardened around it to secure
properly. Figure 5 shows in (a) the footprint profile
(blue spline) on a sketch plane with centroid position and (b) the final implant for L3 and orientation
support.
The implants were designed (10 flat and 10 conformal) using 3-Matic software (Materialise, Belgium)
as simple prismatic extrusions of the footprint profiles, each having a spherical cut at the centroid
position of its top surface to transmit the applied
force.
134
Figure 5. L3 Implant design showing (a) footprint profile &

centroid, and (b) final implant & support.
3.1.1 Mechanical testing

First, non-destructive tests were performed at low
loads in order to investigate the distribution of force
based on the percentage of surface contact between
the implant and the vertebral endplate. Secondly, failure testing was performed to determine failure loads,
displacement and calculated stiffness.
Non-destructive tests were performed using a
Schulz hydraulic hand press with a pump stroke of
0.4 mm. A calibrated 20 kN load cell was used to measure the applied force and an I-Scan sensor (Tekscan
Inc., Massachusetts) was used to record the contact
load distribution between the implant and the bone
endplates. An initial preload of 50 N was applied, after
which the force was increased slowly at an even rate
until a predetermined non-destructive load of 400 N
was reached. As a precaution, this load was decreased
by 50 N in groups of four vertebrae to accommodate the decreasing surface area from lumbar to the
cervical region. The selection of these loads were
based on previous literature with similar experimental procedures (Auerbach, et al., 2010, Tan, et al.,
2005, Hasegawa et al., 2001). After completing nondestructive testing, a failure test was performed once
on each vertebral specimen. Testing was performed
using an MTS hydraulic pressure tester (MTS, Minnesota, USA). A calibrated 20 kN load cell was used to
measure the applied load. Displacement sensors were
used to measure displacement as vertebral fracture
occurred. After a preload of 100 N was applied, a slow
ramp was applied at 0.1 mm/s using the MTS Model
407 Controller. The test was stopped when fracture
had occurred or the load-displacement curve dropped.
The stiffness was calculated using linear regression of
the elastic portion of the load-displacement graph.
3.2
Results
3.2.1 Non-destructive tests

A typical example of load distributions that were
observed (in this case T3 and T2) is shown in
Figure 6.
The scan result images were scaled uniformly and
superimposed over images of the vertebrae. The profiles of the implants are represented by a blue line. A
clear distinction is visible between the contact percentages of the different implants, with the conformal ones
Figure 6. Typical load distributions from I-Scan sensors.
showing a marked improvement over their flat counterparts. What is also evident in the case of the flat
implants, is that all the loads were transmitted onto the
peripheral sides of the endplates where the high rims
of the concave bone geometry made first contact and
created stress concentrations. This was especially true
in the case of the upper thoracic and cervical vertebrae
where the concavity is quite pronounced.
The differences in the percent contact area achieved
by the two implant designs were analyzed statistically. The average contact for conformal implants
was 45.27%, while flat implants only made an average of 10.49% contact. By hypothesis testing, it was
confirmed that this difference is indeed statistically
significant (p < 0.001). Figure 7 shows the graphical
depiction of the measured results.
The solid bars indicate the percentage contact for
conformal implants, and it is significant to note that
this level of contact increases from the lumbar to the
cervical bones. This is due to the fact that the endplates
of the vertebrae become increasingly concave from
lumbar to cervical, allowing for a more distinct and
comprehensive fit between bone and implant. Despite
the increase in concavity of the endplates, the percentage contact that the flat implants made (light blue bars),
remained relatively constant at above or below 10%.
Thus the vertebral level did not play a significant role
in the performance of the flat implants with regards to
percent contact.
3.2.2 Destructive tests
As expected, the lumbar and low-thoracic vertebrae
withstood the highest loads. Figure 8 shows the stiffness values that were derived from the linear regression calculations on each vertebras load-displacement
curve. Stiffness is higher for conformal implants in
each case and the difference becomes more pronounced from the mid-thoracic to the cervical vertebrae. The growing difference is due mostly to a
decrease in stiffness of samples using flat implant
designs.
Hypothesis testing was also done on the destructive
test results and the difference observed in the stiffness when using conformal implants compared to flat
implants was statistically significant with a p-value <
0.0001.
135
Figure 7. Percent contact made per vertebra at maximum non-destructive loads applied.
Figure 8. Stiffness values per vertebra during failure testing.

Figure 9. Process chain for the diagnostic phase.
ECONOMIC EVALUATION
Along with benefits already discussed, it is important

to investigate the cost implications of patient-specific
implants for TDR treatments. Similar cost model studies have been done (Guyer, et al., 2007), (Soegaard,
et al., 2007), however specific studies for the spine are
limited for the South African context.
Soegaard proposed a framework using activity
based costing with four different phases of perioperative care, namely diagnosis, admission, surgery and
follow-up. Guyers model assesses direct costs from
both the hospital and the payer perspective (insurer or
patient). This study combined Soegaards framework
with the South African National Health Reference
Pricelist (NRP) to determine a cost model. The NRP
is a baseline against which medical insurers can determine benefit levels and health service providers can
individually determine fees charged to patients. This
is a very structured document whereby medical procedures are listed and carefully documented by means
of a coding system.
Literature sources as well as extensive interviews
with surgeons were used to derive the individual steps
involved during the four perioperative stages mentioned. An overview of these are shown below in
Figures 912. The calculation of costs during each
stage is a complicated process. Each diagnosis is
assigned an ICD code (the International Statistical
Class of Disease and Related Health Problems, 10th
Figure 10. Process chain for the admission phase.
Revision), while procedures are assigned a CPT (Current Procedural Terminology) code. All clinical procedures are linked to the NRP by means of a tariff coding
system. In certain cases these tariffs may be modified
when a patients condition will affect such procedures.
Finally, any pharmaceutical, surgical and healthcare
consumables used in the procedure are assigned a
NAPPI (National Pharmaceutical Product Interface)
code. The detail cost breakdown for each phase is
not shown in full, but is summarized in Figure 13 by a
cumulative graph.All costs refer to 2009 SouthAfrican
Rand values.
During the diagnostic phase, a general practitioner
will refer the patient to a specialist who performs a
series of tests and prescribes a number of conservative treatment options. Once the patient is eligible
for surgery, the admission phase is entered. Just prior
136
Figure 11. Process chain for the surgical phase.

Figure 13. Cumulative costs for lumbar TDR surgery.
5
Figure 12. Process chain for the postoperative follow-up
phase.
Table 1.
Implant manufacturing cost breakdown.
Quantity
Machine setup
Material cost
Running cost
Postprocessing
TOTAL COST
Cost per implant
R500
R231
R1257
R900
R2888
R2888
R500
R462
R2095
R900
R3957
R1979
R500
R1154
R2933
R900
R5488
R1098
R500
R2078
R3771
R900
R7249
R805
to hospitalization, the design and manufacture of the

patient-specific disc implant is initiated, as described
earlier in Figure 1.
The calculation of the manufacturing cost of the
disc implant is based on machine setup, material
cost, running cost, and postprocessing. Material cost
is estimated from the part volume while running
cost is factored against the total time to manufacture the part(s). Table 1 shows a comparative
breakdown of manufacturing costs, indicating the
inherent savings when producing multiple implants
simultaneously.
The surgical phase of the TDR process is filled
with variability that influences the cost of the phase.
Factors that cause this variability include:
The location of the degenerated disc (cervical,
thoracic or lumbar)
The operating time
The type of decompression surgery i.e. laminectomy, osteophyte removal, or discectomy
The Body Mass Index (BMI) of the patient
The age of the patient
Whether the patient is a smoker or not
CONCLUSIONS
The results of this study show that there are indeed

significant potential benefits that can be achieved
through the use of customization during the design
and manufacture of intervertebral disc implants. With
the design and manufacturing process that has been
proposed, these and other potential benefits can and
should be developed for the improvement of existing disc implant designs. Further research is required
to develop the necessary insertion tools and fixation
designs, of which there are many alternatives that
Rapid Manufacturing can offer with its ability to create
complex geometry.
Mechanical test results revealed that conformal
implants achieved significantly higher surface contact than flat implant designs. Furthermore, conformal
implants showed on average, a 137% increase in
stiffness over flat implants during destructive tests.
A brief overview of a cost model for TDR within a
South African context was also presented. A cumulative breakdown of costs for TDR indicated that 69%
of costs are incurred during the surgery and implant
manufacturing phase. Up to 72% reduction in manufacturing costs can be achieved when producing
multiple implants simultaneously. Furthermore, the
potential time that can be saved during surgery by
using a patient-specific implant is still a significant
topic for further study.
Although there are several significant challenges
ahead (both technical and regulatory) the prospect
for customisation in the spine is indeed an exciting
one with far reaching potential benefits to the patient
specifically and to health care in general.
REFERENCES
Finally, the postoperative and follow-up phase includes

intensive care and general ward recovery. Follow-up
visits and subsequent tests assist the patient during
long term recovery.
Auerbach, J. D., Ballester, C. M., Hammond, F., Carine, E.

T., Balderston, R. A., Elliott, D. M. (2010). The effect of
implant size and device keel on vertebral compression
properties in lumbar total disc replacement. The Spine
Journal, 10, 33340.
Bertagnoli, R. (2005). Complications and Rescue Strategies in TDR Procedures. Proceedings of the 20th Annual
Meeting of the North American Spine Society (NASS),
Philadelphia, USA.
137
Cheh, G., Bridwell, K. H., Lenke, L. G., Buchowski, J. M.,

Daubs, M. D., Kim, Y., Baldus, C. (2007). Adjacent Segment Disease Following Lumbar/Thoracolumbar Fusion
With Pedicle Screw Instrumentation A Minimum 5-Year
Follow-up. Spine, 32(20), 225357.
Cinotti, G., David, T., Postacchini, F. (1996). Results of Disc
Prosthesis After a Minimum Follow-Up Period of 2 Years.
Spine, 21(8), 9951000.
De Jongh, C. U., Basson, A. H., Scheffer, C. (2008). Predictive Modelling of Cervical Disc Implant Wear. Journal of
Biomechanics, 41, 317783.
Gibson, I., Cheung, L. K., Chow, S. P., Cheung, W. L.,
Beh, S. L., Savalani, M., Lee, S. H. (2006). The Use of
Rapid Prototyping to Assist Medical Applications. Rapid
Guyer, R. D., Tromanhauser, S. G., Regan, J. J. (2007). An
economic model of one-level lumbar arthroplasty versus
fusion. The Spine Journal, 7, 55862.
Harrop, J. S., Youssef, J. A., Maltenfort, M., Vorwald, P.,
Jabbour, P., Bono, C. M., Goldfarb, N., Vaccaro, A.
R., Hilibrand, A. S. (2008). Lumbar Adjacent Segment
Degeneration and Disease After Arthrodesis and Total
Disc Arthroplasty. Spine, 33(15), 17017.
Harrysson, O. L. A., Hosni, Y. A., Nayfeh, J. F. (2007).
Custom-Designed Orthopedic Implants Evaluated Using
Finite Element Analysis of PatientSpecific Computed
Tomography Data: Femoral-Component Case Study.
BMC Musculoskeletal Disorders, 8, 91.
Harrysson, O. L. A., Cansizoglu, O., Marcellin-Little, D. J.,
Cormier, D. R., West, H. A. (2008). Direct Metal Fabrication of Titanium Implants with Tailored Materials
and Mechanical Properties Using Electron Beam Melting
Technology. Materials Science & Engineering C, 28(3),
36673.
Hasegawa, K., Abe, M., Washio, T., Hara, T. (2001). An
Experimental Study on the Interface Strength Between
Titanium Mesh Cage and Vertebra in Reference to Vertebral Bone Mineral Density. Spine, 26, 95763.
Hieu, L. C., Bohez, E., Van der Sloten, J., Phien, H. N.,
Vatcharaporn, E., Binh, P. H., An, P. V., Oris, P. (2003).
Design for Medical Rapid Prototyping of Cranioplasty
Implants. Rapid Prototyping Journal, 9(3), 17586.
Higashino, K., Hamasaki, T., Kim, J. H., Okada, M., Yoon, S.
T., Boden, S. D., Hutton, W. C. (2010). Do the Adjacent
Level Intervertebral Discs Degenerate After a Lumbar
Spinal Fusion? An Experimental Study Using a Rabbit
Model. Spine, 35(22), E114452.
Leary, S. P., Regan, J. J., Lanman, T. H., Wagner, W.
H. (2007). Revision and Explantation Strategies Involving the CHARITE Lumbar Artificial Disc Replacement.
Spine, 32(9), 100111.
Lowe, T. G., Hashim, S., Wilson, L.A., OBrien, M. F.,
Smith, D. A. B., Diekmann, M. J., Trommeter, J. (2004). A
Biomechanical Study of Regional Endplate Strength and
Cage Morphology as It Relates to Structural Interbody
Support. Spine, 29(21), 238994.
Matsumoto, M., Okada, E., Ichihara, D., Watanabe, K.,
Chiba, K., Toyama, Y., Fujiwara, H., Momoshima, S.,
Nishiwaki, Y., Iwanami, A., Ikegami, T., Takahata, T.,

Hashimoto, T. (2009). Anterior Cervical Decompression
and Fusion Accelerates Adjacent Segment Degeneration
Comparison With Asymptomatic Volunteers in a Ten-Year
Magnetic Resonance Imaging Follow-up Study. Spine,
35(1), 3643.
Poukens, J., Haex, J., Riediger, D. (2003). The Use of
Rapid Prototyping in the Preoperative Planning of Distraction Osteogenesis of the Cranio-Maxillofacial Skeleton.
Computer Aided Surgery, 8(3), 14653.
Schindel, R., Lampert, C. H., Gross, W. (2005). Case Report
TALUS: Application of Selective Laser Sintering Technology in Orthopaedics. European Cells and Materials,
10(Suppl. 5), TM4.
Sercombe, T., Jones, N., Day, R., Kop, A. (2008). Heat
treatment of Ti-6Al-7Nb components produced by selective laser melting. Rapid Prototyping Journal, 14(5),
300305.
Shim, C. S., Lee, S., Maeng, D.H., Lee, S.-H. (2005). Vertical Split Fracture of the Vertebral Body Following Total
Disc Replacement Using ProDisc: Report of Two Cases.
Journal of Spinal Disorders & Techniques, 18(5), 4659.
Soegaard, R., Christensen, F. B., Christiansen, T., Bunger,
C. (2007). Costs and effects in lumbar spinal fusion. A
follow-up study in 136 consecutive patients with chronic
low back pain. European Spine Journal, 16, 65768.
Sun, L. W., Lee, R. Y. W., Lu, W., Luk, K. D. K. (2004). Modelling and Simulation of the Intervertebral Movements
of the Lumbar Spine using an Inverse Kinematic Algorithm. Medical & Biological Engineering & Computing,
42, 7406
Tan, J.-S., Bailey, C., Dvorak, M., Fisher, C., Oxland, T.
(2005). Interbody Device Shape and Size Are Important
to Strengthen the VertebraImplant Interface. Spine, 30,
63844.
Van der Houwen, E. B., Baron, P., Veldhuizen, A. G.,
Burgerhof, J. G. M., Van Ooijen, P. M. A., Verkerke, G. J.
(2010). Geometry of the Intervertebral Volume and Vertebral Endplates of the Human Spine. Annals of Biomedical
Van Ooij, A., Oner, F. C., Verbout, A. J. (2003). Complications
of Artificial Disc Replacement - A Report of 27 Patients
with the SB Charit Disc. Journal of Spinal Disorders &
Techniques, 16(4), 36983.
Van Ooij, A., Schurink, G. W., Oner, F. C., Verbout, A. J.
(2007). Findings in 67 patients with recurrent or persistent symptoms after implantation of a disc prosthesis for
low back pain. Nederlands Tijdschrift voor Geneeskunde,
151(28), 157784.
Vandenbroucke, B., Kruth, J.-P. (2007). Selective Laser
Melting of Biocompatible Metals for Rapid Manufacturing of Medical Parts. Rapid Prototyping Journal, 13(4),
196203.
Zeegers, W. S., Bohnen, L. M. L. J., Laaper, M.,
Verhaegen, M. J. A. (1999). Artificial disc replacement
with the modular type SB Charit III: 2-year results in 50
prospectively studied patients. European Spine Journal,
8(3), 21017.
138

Additive manufacturing of soft tissue geometries for reconstruction

purposes
M. Truscott & G.J. Booysen
Central University of Technology, Free State, Bloemfontein, South Africa
D.J. de Beer
Vaal University of Technology, Vanderbijlpark, South Africa
ABSTRACT: The paper describes the use of specialised imaging software (MAGICS and MIMICS ) in parallel with Computer Tomography (CT) or Magnetic Resonance Imaging (MRI) scanning, to produce representative
geometries of patient specific soft tissue or muscle abnormalities. Many patients suffer from abnormalities caused
by diseases or traumatic accidents, resulting in deformed soft tissue or representative muscle volumes, which
can either have an aesthetical result, or even cause chronic pain or discomfort, due to abnormal body posture or
related causes. It furthermore evaluates various case studies performed on patients with unique problems, where
an appropriate process chain have been developed, using CT or MRI scans, imaging through MIMICS and
volume manipulation using mirroring techniques to identify a-symmetrical (volumetric) tissue/muscle volumes.
Through using Boolean operations, supportive volumetric structures can be developed, using AM techniques to
develop master patterns for indirect moulding processes, to produce casts for implantations.Several successful
interventions are discussed to illustrate the successful interventions and supportive process, as developed by the
Central University of Technology (CUT) Integrated Product Development (IPD) research team.
INTRODUCTION
1.1 Historical development of the CUTs CRPM

and IPD Research Focus Areathat led to
medical product development success
in South Africa
Additive Manufacturing (AM) has been introduced
to the South African industry in 1991, through a
private consortium, making the services from a 3D
Systems SLA 250 available to support product development. Formal research programmes started through
a combined research project between the CSIR (owning 2 Stratasys FDM 1500 machines) and the CUT
(then Technikon Free State). Initial exposure led to
the CUTs 1st investment in 3D Inkjet printing (Solidscape). Industrys involvement and funding helped to
attract matching government funds, whilst access to
RP equipment at CUT created a better understanding of the then (1996) new technology. This improved
knowledge and understanding created the opportunity
to purchase an SLA 250 (3D Systems). As research
activities evolved, limitations (and based on opportunities identified through a typical SWOT analysis
approach) led to the acquisition of a DTM 2000 Sinterstation, using government research infrastructure
funding initiatives. The combined CUT infrastructure
at that stage, together with early versions of Materialise
MIMICS and MAGIGS RP software, created a
perfect opportunity to develop a Medical Product
Development activity, within the CUTs broader

Integrated Product Development (IPD) Research
group, as well as an expanded AM research base within
the university. Through active R&D and Technology
Transfer, last mentioned created a chain reaction in
industry, leading to purchasing of an EOS P380, followed by the purchasing of an EOS S700, EOS M250,
EOS P385 and EOS M270X(Ti) respectively all justified by industry oriented research projects or industry
needs for technology development and support. Concurrently, extensive investments in CAD/CAM, specialised software, reverse engineering equipment and
support software were made, to effect medical product
development, reverse engineering, quality inspections
and modeling and simulation capabilities.
1.2
Early medical product development success
Early experimentation started through collaboration

with Prof Butow from University of Pretoria, where
different technologies (FDM and SLS) were used
to prepare surgical planning models as part of the
surgery planning for a 3 year old boy that needed
cranio-facial surgery. Specialists requested a prototype of the skull for pre-surgical planning. The first
attempt in handling medical CT data was through using
3 mm slices from a Toshiba Spiral X-Vision/GX CAT
scanner, but had to be repeated at the finest allowable slice-thickness to supply more accurate data. The
139
specialists were extremely pleased with the 3D computer image created, and approval was given to proceed
to the prototype stage. Due to limited availability of
AM platforms in SA (only SLA and FDM at that stage),
and due to the amount of supports needed to grow the
skull, a first attempt was made using Stratays ICW06
wax material. The skull was initially also grown in two
halves, to ease the removal of support material. The
material was however, very brittle, which restricted
the handling capabilities and concerns were expressed
about the applicability for surgery usage. As such, it
was only used for visual inspection. This early work
supported the CUTs expansion planning, and a decision was taken to purchase a DTM Sinterstation 2000,
due to its diverse material handling capability. Delivery to SA took longer than planned, and as the model
was required urgently, DTM arranged that the model
be grown by a 3rd party. The SLS prototype gave very
good results especially since there were no supports
to remove, and the surgeons could work with ease
on the Nylon prototype. The fact that planning drawings and measurements could be made directly on the
Nylon skull, together with the fitment of metal brackets, made it a winning combination. Both the maxilla
craniofacial and neuro-surgeons frequently referred
to the skull during the eight-hour operation, concluded
in approximately 60% of the conventional operation
time.
This early success, achieved in 1998, has created
a benchmark study that set the table for many more
successful developments using Selective Laser Sintering and Laser Sintering, together with proving the
technology for further development and implementation by the CUT team [Schenker et al. 1999, De Beer
2011]. Further development focused on the ability to
create a quick response approach, as normally the CUT
team had to attend to a diverse range of support applications where custom developments had to be done
for extreme cases, leading to a unique development
approach where the CUT team interacted in a team
approach with surgeons and bio-engineers, to set the
trend for other teams in SA to follow [Truscott et al.
2007].
Whilst proven methods and process chains have
been established to develop surgical planning models, models to be used as prototypes and prototype
castings for implants, including direct implantation of
bone geometries created through Direct Metal Laser
Sintering in Ti-alloys, an evident gap in the development was the use of AM technologies to create patterns
or even moulds for the development of soft tissue
implant-geometries for reconstruction purposes.
2
2.1
SOFT TISSUE DEVELOPMENT USING AM

Confirmation of the need for soft tissue
development
Sun et al. (2011) confirm that patients suffer from

emotional pressure and abnormal psychology caused
by severe facial defects involving the partial ortotal

loss of some facial organs. To assist them in removing thepsychological load and restoring their selfconfidence, effectiverehabilitation therapy is adopted
to rebuild their features.
Although modern surgery can repair defective features by implanting some other tissues to satisfactory
levels, a number of patients may not select such therapy because of advanced age,medical conditions or the
expenses involved. The authors conclude that the use
of facial prostheses is a better alternative to surgeries;
facial prostheses can also supplement the surgical
reconstruction of facial defects (Sun et al. 2011)
2.2 International progress in soft tissue

development using AM technologies
In a paper written by Gibson et al. (2006), the
authorsconclude that RP has been helpful in a number of ways to solve medical problems. However, the
technology has numerous limitations that have been
analysed in order to establish how the technology
should develop in the future. They furthermore state
that systems that can model elastomeric as well as
plastic components would be useful for creating soft
tissuemodels, which can extend applications beyond
orthopaedics.
On similar development, Bibb et al. (2010) report
that there was a wide array of early research and
development to create anatomical forms using AM
processes. The authors mention that production of
anatomical forms have been used as patterns for
replication into more appropriate materials via secondary processes such as silicone moulds or vacuum
casting. Later research attempted to use AM methods toproduce moulds from which prosthesis forms
could bemoulded. The authors furthermore state that
such developments required extensive time and CAD
facilities. They also hint that there was a need for
critical evaluation of the physical properties of the
prostheses produced or the technical capabilities and
appropriateness of the RPtechnologies used for developing an integrated and efficient digital prosthetic
process.
METHODOLOGY
3.1 Data processing

The authors have developed various methods based on
the use of MIMICS and MAGIGS RP to firstly,
create representative data to be used for identifying
and isolation of soft tissue volumes, and secondly, to
manipulate these and to create changed or manipulated replacements. The process chain involves CT or
MRI data (preferable MRI), using MIMICS . Once
a walk-through of the data has been done, on a layered approach (following the CT or MRI layered data),
affected data can be mirror-copied onto data representing the un-affected or healthy data. Due to the fact
140
that the human body and its representative physiological data is not symmetrical, the selection of mirroring
planes is important, and requires careful consideration and significant experience in using the related
software. The issue of a-symmetry has been addressed
by Ferraro et al. (2001), in previous studies focusing
on human faces, and the authors concluded that the
degree of symmetry differ individually. Benz et al.
(2002) and Koch et al. (1993), concluded that the deviation from perfect symmetry is not very large, and that
a mid plane could be used to mirror the healthy side
to the deformities, which concur with and support the
authors work.
3.2
Data manipulation
Once an overlay of the two data sets have been created, Boolean operations can be used to determine the
difference between the healthy and deformed side or
tissue, from which a new geometry can be created to
represent the geometric volume that will be used to
manufacture the representative prosthesis.
3.3
Figure 1. Image of male patient with Polands Syndrome.
Prototype and prosthesis development
Geometries developed as describe above, are used to

produce .STL data, from which LS or SLA models can
be created, to be used for the manufacturing of silicone
moulds, for casting of the appropriate materials.
In an alternative approach, the data can also be
used to develop moulds, for casting of the appropriate
materials.
Figure 2. Image of female patient with Polands Syndrome.
4
4.1
CASE STUDIES
Congenital defect Polands Syndrome
pectoralis muscle development male
and female
In the two case studies conducted, the patients presented with Poland Syndrome. Poland Syndrome is
described as the underdevelopment or absence of the
chest muscle (pectoralis major) on one side of the
body (Mersch 2011). Surgery was required for aesthetic purposes, and to attempt to resolve posture
issues. As such, the surgeons required a prototype of
the missing 3D volume in order to manufacture an
insert. Although pectoralis silicon inserts are available commercially, these case studies required a more
customised approach1 .
Analyses of the pectoralis muscle data were performed and surgical planning for one of the studies
performed to see the visual effect of such an operation.
CT data of a female patient were used for the isolation
of the soft tissue. The isolation of the data proved to
be very difficult because only CT data and not MRI
data were available. Figure 1 shows an image created
1
As in the case of the pectoralis muscle development case

studies, the CUTs IPD team normally has to deal with
extreme cases that need custom developments.
using CT data from the male patient case study, whilst

Figure 2 shows the image created from the female
patients CT data.
The representative volume of the pectoralis major
muscle was obtained and mirrored using MIMICS ,
and a prototype built of the volume. Surgical simulation was performed on the case study involving the
female pectoralis muscle using the simulation module
of MIMICS . The prototype of the female patient was
presented to the medical practitioner. A silicon mould
was manufactured in Brazil and has consequently been
implanted successfully into the patient.
The process chain followed included:
CT Scan of patient;
Isolation of pectoralis data
Mirroring/Surgical simulation;
Conversion to .STL file;
LS in Nylon on EOS P380;
Mould preparation/Silicon Mould
Figure 3 shows a layer from the CT scans.

Figure 4 shows the top view rendering in MAGICS
RP . Figure 5 shows further image manipulation and
mirroring of data done in MAGICS RP . Figure 6
shows the .STL data and placement in MAGICS RP
for growing on the EOS P380 LS machine.
141
Figure 3. CT Image of one layer of the patients pectoralis

major muscle.
Figure 6. Image showing the scheduling of the part geometries in MAGICS RP for growing on the EOS P380 LS
machine.
Figure 7. Image of the Nylon prototype development.

Figure 4. Image showing the development in MAGICS
RP .
Figure 8. Post-surgery image.
4.2 Working with Children of Fire to develop ear

prosthesis
Figure 5. Image showing the external mirroring of data.
Figure 7 shows the prototype developed in Nylon

Polyamide 2200 using the EOS P380 LS machine,
whilst Figure 8 shows a post surgery image of the
patient.
Through the support of Children of Fire, and through

the use of 3D imaging, an ear prosthesis was developed for a fire-victim with severe facial burn-injuries.
Figure 9 shows an image of the ear prosthesis/mould
as developed for silicone casting, as well as shows an
image of the patient.
142
Figure 9. Images of the ear prosthesis development and the

patient/burn victim.
Figure 11. Image showing the development of the fitment.
Figure 12. Image showing final fitment and comparison.

Figure 10. Image showing the calf development in Nylon
Polyamide 2200.
4.3
Calf muscle development
The CUTs IPD research team was approached to

develop calf-muscle geometry for a patient. The
patient presented with a-symmetric calf muscles, and
surgery was recommended for aesthetic considerations. 3D Photography was used to develop the geometries, from which the solid models and .STL data were
generated to produce the prototypes that represented
the difference in geometry of the two calf muscles.
Figure 10 shows an image of the prototype as produced in Nylon Polyamide 2200, whilst figures 11 and
12 show the fitment after the surgeons intervention.
CONCLUSIONS
Following the authors research work, a firm strategy

and process chain is in place to support similar developments. Whilst the process is still done in isolation
from orthopaedic practice, the process has been developed working with teams of specialists, and taking
their input and expertise into account.
Through the case studies presented, backed by similar conceptual and modelling exercises, the feasibility
of using AM in parallel with CT, MRI and 3D photography as input, together with CAD, MIMICS and
MAGICS RP has proven to be capable to create a support system for patients suffering from severe defects
caused by abnormal symmetry.
RECOMMENDATIONS
The authors recommend that further development be

done, including the development of local biocompatible casting techniques, together with more futuristic direct plotting/printing of similar geometries as
presented. AM in biocompatible materials will significantly shorten the product development time, and
will have as a significant impact on the cost and
effectiveness of the process.
ACKNOWLEDGEMENTS
The authors would like to acknowledge the contribution of fellow staff and students from CRPM, the
143
support and collaboration of Drs T van Rooyen, E van

der Walt and P Coetzer.
Furthermore, the authors would like to acknowledge the National Research Foundation (NRF), Medical Research Council (MRC), Technology and Human
Resources for Industry Programme (THRIP) and DST
through the Tshumisano and Technology and Innovation Agency (TIA) support given to make the
infrastructure and research possible.
Disclaimer: Any opinion, findings and conclusions
or recommendations expressed in this material are
those of the authors and thereby the NRF does not
accept any liability in regard thereto.
REFERENCES
Benz, M., Laboureux, X., Maier, T., Nkenke, E., Seeger, S.,
Neukam, F., Hausler, G. 2002. The symmetry of faces.
In Gnther Greiner (ed.), Proc. of Vision, Modeling and
Visualization Conf., Erlangen, Germany, 2022 November
2002.
Bibb,R., Eggbeer, D., Evans, P. 2010. Rapid prototyping technologies in soft tissue facial prosthetics: current state of
the art. Rapid Prototyping Journal 16(2), 130137.
De Beer, D. J. 2011. Establishment of rapid prototyping/additive manufacturing in SouthAfrica:The Journal of
the Southern African Institute for Mining and Metallurgy

111, 211215.
Ferrario, V. F., Sforza, C., Colombo,A., Schmitz, J. H., Serrao,
G. 2001. Morphometry of the orbital region:a soft-tissue
study from adolescence to mid-adulthood, Plastic and
Reconstructive Surgery 108, 285292.
Gibson, I., Cheung, L. K., Chow, S. P., Cheung, W. L., Beh,
S. L., Savalani, M., Lee, S. H. 2006. The use of rapid prototyping to assist medical applications. Rapid Prototyping
Journal 12(1), 5358.
Koch, A., Wiese, K.G., Rasse, M. 1993. Zur symmetrie
des menschlichen mittelgesichts mit 1 abbildung und 3
tabellen im text, Zeitschrift fur Morphologie und Anthropologie. 79, 313320.
Mersch, J. (2011). Poland syndrome. Available at:.
http://www.medicinenet.com/poland_syndrome/article.
htm. (Retrieved 26 July 2011).
Schenker, R., De Beer, D. J., Du Preez, W. B., Thomas, M. E.,
Richter, P. W. 1999. Novel combination of reverse engineering and rapid prototyping in medicine. SA Journal of
Science 95(8), 327328.
Sun, J., Xi, J., Chen, X., Xiong, Y. 2011. A CAD/CAM system for fabrication of facial prostheses. Rapid Prototyping
Journal 17(4), 253261.
Truscott, M., De Beer, D. J., Vicatos, G., Hosking, K.,
Barnard, L., Booysen, G., Campbell, R. I. 2007. Using RP
to promote collaborative design of customised medical
implants. Rapid Prototyping Journal 13(2), 107114.
144

Development of patient-specific implants using Direct Metal Laser Sintering

in Titanium
G.J. Booysen, M. Truscott & J. Els
Central University of Technology, Free State, Bloemfontein, South Africa
D.J. de Beer
Vaal University of Technology, Vanderbijlpark, South Africa
ABSTRACT: The CUTs IPD research team has been involved in the development of surgery models to support
intervention surgery in a significant number of traumatic patient-specific cases. Although very successful, it
remained an incomplete support process/option, as up to the point of using AM technologies to produce the
surgery model, all operations are executed in a digital domain. After providing the surgery model, developments
mainly resolve on mechanical operations almost contradicting the digital development that has taken place.
The paper will discuss successful developments that led to Ti machining of patient-specific implants, and which
paved the way to develop a direct digital manufacturing approach to remain in a digital environment with
both time and cost savings as a further result. Afore-mentioned will be illustrated through a series of successful
developments of direct-produced Ti implants, and which were used in successful surgery applications. The EOS
M270 Xtended (Ti) Direct Laser Sintering system was instrumental in achieving these results, and the case study
discussions will be supplemented by related process development and improvements, together with a discussion
of growing parameters (like support structure placement and sacrificial parts build around the parts).
1
INTRODUCTION OF CUTS MEDICAL

DEVELOPMENT APPROACH
1.1 Historical development of the CUTs CRPM

and IPD Research Focus Areathat led to
specialisation in direct laser sintering
technologies
Following a position of being part of a very late adoption approach, the CUTs IPD research team became
only the third group to have access to in-house Additive Manufacturing (AM) technologies, late in 1996.
This has been almost five years after the technologys
introduction to the South African industry in 1991.
The CUTs first acquisition was during 1995,
through a co-investment from industry to acquire a
3D Inkjet wax ink-jet printing (Solidscape) technology. It also created the start of a successful Private Public Partnership (PPP) or Triple-Helix model
demonstrated by the IPD team, on how academia can
work with industry. Following industrys needs for
technology and process development, brought their
involvement and subsequently funding support, which
in turn brought matching government funds. Slowly
but surely, necessary building blocks (read AM platforms) came in place, to result in one of the best
equipped centres in the Southern Hemisphere.
The breakthrough came in 1996, when the CUT
seized the opportunity to purchase an SLA 250 (3D
Systems) through an international reduced offer to
academic institutions for the then very low price of
US$ 99 000. Access to this platform gave the team

a solid understanding of RP/AM, and as research
activities evolved, an increased footprint was created in industry, through a series of novel (action)
research-based application case-studies.
Due to the limited material choices at that stage, and
following increased international activities, together
with results used from a strategic SWOT-analyses on
the future of the IPD research team, a DTM 2000
Sinterstation was acquired once again through a partnership with industry and government. Having access
to more material options brought even more industry involvement, and very importantly, exposure to
polymer, sand and metal sintering. With this came
more industry challenges (and opportunities), and also
industry partnerships.
1.2 ADEPT Airmotive partnership and CRPMs
consequent specialisation in Direct Laser
Sintering technologies and applications
A further breakthrough in the technology development
curve occurred when ADEPT Airmotive approached
the CUT team with an offer to co-invest in dedicated
polymer and sand sintering technologies, which lead to
a long-standing partnership with EOS, and the initial
purchasing of an EOS P380, followed by the purchasing of an EOS S700. It furthermore led to a number
of breakthroughs in applications-oriented research
projects attempted especially due to the increased
145
Figure 1. Collage showing images of the various stages of

the ADEPT Airmotive Engine Development.
Figure 2. Various sand casting patterns, cores and the

polystyrene engine block prepared for investment casting.
production volume gained from the DTM Sinterstation 2000. The ADEPT Airmotive project helped to
give significant credibility and exposure to the team,
and helped to have the first major PPP-funding, which
at that stage was approximately 750 000 and quite a
high project value for a rather small university. Figure 1
shows a collage that portrays some of the exciting development stages and products created through
the ADEPT Airmotive project, and include the CAD
development, geometrical prototypes created as part
of the final stages of the engine design, which aided
significantly in reviewing the design for fitment purposes, the grown polystyrene engine block and a 1/4
scale fully prototyped engine model to ensure full
fitment and correctness of the design.
Figure 2 shows various sand casting patterns and
cores, together with the polystyrene engine block
prepared for investment casting.
1.3
CRPMs early medical contribution
From the early successes described above, further

investment led to the acquisition of an EOS M250
and EOS P385 respectively both again justified by
Figure 3. Pelvic, skull and femur pre-operative models in

nylon polyamide.
industry oriented research projects or industry needs

for technology development and support.
Concurrently, extensive investments in CAD/CAM,
specialised software, reverse engineering equipment and support software were made, to effect
medical product development, reverse engineering,
quality inspections and modeling and simulation
capabilities.
Having access to larger growing/sintering volumes
and the specialised software, led to an increased level
of involvement with the medical fraternity, which
in turn led to another field of specialisation for the
research team. The increased growing volume facilitated larger medical models and consequently led to
extensive medical product development involvement,
through the provision of surgical models to support
a diverse range of successful surgical procedures.
Figure 3 shows some of the unique applications where
the CUT team has been involved in supporting extreme
cases.
1.4 Medical development and evolvement into

direct metal laser of Ti alloys
A number of case studies conducted also required the
development of geometries for implantation, where
complex reverse engineering and CNC tool-path
creation for machining of a customised implant in
Titanium became part of the CUT teams development
process.
1.4.1 Pelvis and femur development
Various case studies involving the pelvis and femur
have been performed. In most instances these are
extreme cases involving cancerous growths, polio as
well as rectifying mistakes made during previous
operations. The 3D model in nylon polyamide was
used for surgical planning and for implant design as
can be seen in the following sequence of photographs
in Figure 4 and 5. The implant was manufactured
in Titanium using a DMU60T 5-axis CNC milling
machine and was done in collaboration with ISIQU.
146
Figure 4. Collage of images showing the CT data of the

deteriorated pelvis, as well as the grown part.
Figure 5. Collage of images showing the pelvis data manipulation, mechanical design, machined and grown parts and
final implant (X-ray).
1.4.2 Elbow development

The CUT IPD team was approached to assist with
a model where the patient has lost a major part of
his elbow through a motorbike accident, where only
remnants of his right elbow were left in the arm. The
CUT IPD team assisted by developing surgical models
in PA 2200 using the CT scan of the injured elbow,
as well as a mirror image of the left healthy elbow.
Through interaction with the surgeon, a final model
that represented the implant geometry required.
The CUT IPD team reverse engineered the manually altered models for development of the CNC
toolpaths, which in turn was given to ISIQU to
machine the implant from a Titanium billet. Machining
took approximately two weeks, and required 8 hours
of machining per day. The Ti machined elbow was
implanted with significant success, and the patient
regained 90% of the use of his arm, where previously, at best, it would have resulted in a fused limb.
Figure 6 show images of the development, which
amongst others resulted in a bench-mark exercise
with respect to collaboration between engineers and
Figure 6. Collage of images showing the elbow development stages from CT to implant.
surgeons in a team approach, when attempting medical

product development.
Furthermore, the case study emphasised the need
for direct metal laser sintering in Titanium, as it
became clear that a significant amount of time, energy
and material were wasted. The team did the original
development in a digital environment. This advantage
was lost when the development went back to work on
a physical prototype, which was then scanned, to produce CNC toolpaths, to once again machine a part
through a material removal process, where approximately 80% of the material is wasted. Figure 6 shows
the left and right elbow knuckles developed using the
scanned and mirrored CT data, the carved knuckle representing the residual bones, the machined Titanium
implant using the reverse engineered data, as well as
the X-ray showing the successful implant.
1.4.3 Scapula and pelvis development
For this intervention, the CUT IPD team was
approached to support the development of three
implants after a wide resection removing the patients
scapula, shoulder joint and proximal humerus. The CT
data of the right hand side scapula, shoulder joint and
proximal humerus was mirrored and the .STL file and
3D model presented to the surgeons. The .STL file
was then used to create surfaces using COPYCAD
and design modelling using Power SHAPE was performed.The industrial partner informed the CUT IPD
team that the pelvis implant was a perfect match to
the resected bone. The collaboration and help they
received from the research team has proven not only
beneficial to the speedy manufacturing of the implant,
but also to the precise anatomical fit of the implant
into the patient.
According to the surgeons, such an implant has
never been made before and such an operation could
only have been successful with the knowledge of the
3D skeletal model of the pelvis.
The industrial partner further commented that
Reconstructing the pelvis did not only give me the precise location of the resection margin, the acetabulum
147
and the pubis, but also gave me the ability to design the
implant in three major pieces to assist access into the
patient. Also due to the size of the defect, it would have
been impossible to manage the implant otherwise.Your
assistance has been proven invaluable once more.
Although all of these cases played a significant
role to establish trust in the process and in the team
members ability to accurately and quickly execute the
translations from CT/MRI data to .STL and beyond for
either prototyping or machining purposes, it also highlighted the absence of a single manufacturing process,
where direct implants could be grown in Ti-alloys,
or other biocompatible materials. These developments
and the obvious missing link were used to develop
both the case to become involved in a Beta-programme
to test and implement and EOS M270 Xtended (Ti)
Direct Metal Laser Sintering platform, as well as yet
another successful PPP-benchmark research project
where more than 750 000 was contributed by government and industrial partners through the NRFs
THRIP programme.
2 APPLICABILITY OF DIRECT AM FOR

BIOMEDICAL APPLICATIONS
LITERTAURE REVIEW
a long timeowing to significant material removal. The

authors furthermore maintain that in many individual cases, where patient-fittedcomponents are required
or if new constructional solutionsare under development, low effectiveness and relatively highcost of
conventional (removal) machining is disadvantageous.
2.3 Advancement from RP to AM in the

biomedical field
Vandenbroucke & Kruth (2007) continues to state that
because of technical improvements of layer manufacturing(LM) processes and the possibility to process
different metals (and compounds), Rapid Prototyping
(RP) has moved beyond its initial applications into
rapid manufacturing (RM). They also point out that
the progress made could benefit medical and dentalapplications beyond polymer applications for visual
(anatomical) models or single-use surgical guides, to
also support the manufacturing of functionalimplants
or prostheses.
2.4 Proven advanced manufacturing and

biomedical applications
2.1 Applicability of sintering technologies

for metal powders
In an article by Tolochko et al. (2003), the authors
confirm that sintering technology can be used to
create functional metal parts through direct laser sintering methods. They point out that the further study
andimprovement of this technology are of specialinterest. Furthermore, the authors identify the significant
potential of using both single- and double-component
metal powders without complex process difficulties
when double-component metal powders are used.
2.2 AM vs Conventional processes

Uklejewski et al. (2011) report that mass-produced
orthopaedic end-osseous implants have mainly been
produced from wrought or cast bar stock by five- or
six-axis computerised numerical control (CNC), Computer aided design (CAD)-drivenmachining, or powder metallurgy production methodologies; including
hot isostatic pressing and powder injection moulding
of near-net-shape components. The authors maintain
that the traditional process chains followed in using
conventional operations (mainly materials removal
processes), are failing due to its long lead time
and material wastage especially where complex
geometries need to be created. In an article by
Murr et al. (2009), the authors confirm that through
material removal processes, often 20% or less of
the feedstock is utilised. However, the set raditional
technologies often fail if the manufacture of animplant component involves complex shape, including
thin-walled sections where cutting operations can take
In a number of articles reviewed, the authors all

agree (Gibson et al. 2006, Vandenbroucke & Kruth,
2007, Murr et al. 2009) that extensive development of the direct digital manufacturing has recently
provided new (promising) perspectivesfor advanced
manufacturing technology including biomedical
applications. The authors are in agreement that recent
material/compounds and process advances creates
both new challenges and significant opportunities
for orthopaedic implant designs, (Hollander et al.
2003), cranioplasty implants (Hieu et al. 2003, 2005),
oral andmaxillofacial implants, finger joint implants
(Gibson et al. 2006; Singare et al. 2006), bone substitute implants (Hoeges et al.2008), customised hip
implants and knee implants (Sercombe et al. 2008;
Murr et al. 2009); dental implants (Santos et al. 2006;
Vandenbroucke & Kruth, 2007).
2.5 Comparison with own results

The sentiments expressed by some of authors quoted
above are well in line with the findings of the IPD
Research Team, and as expressed in an article by
Truscott et al. (2007). It confirms the hypothesis
of future use of AM processes to directly produce
implants in biomaterials.
3 TITANIUM SINTERING CASE STUDIES

3.1 Mechanical test pieces
Figure 7 shows a mechanical tensile test-piece sintered
on the EOS M270 Xtended in Ti.
148
Figure 7. Image of a Ti-sintered mechanical tensile test

piece.

pieces with cylindrical protective rings.
The main concerns about these parts are that the

middle section has to be straight because the test pieces
were required to be built upright and thread cutting was
required afterwards to clamp the test pieces. Considering building this component in any other direction than
upright, there is a possibility of damaging the middle
section when trying to remove the inevitable support
material in a lathe.
On the first platform one single component was
produced without any problems. A second platform
was prepared with 4 components, of which some were
bent backwards by the recoater, due to jamming that
occurred on the beginning layer of the top grip section. It was then decided to put a protective cylinder
around the stem section, spanning between the two
grips sections for the third platform. A further problem encountered with this was that approximately at
the same height than the second platform, the stem
sections were bent backward and the recoater stuck
on the parts. It was then decidedto add a brace part
(indicated by the arrows A in figure 8) to keep the
shaft section connected to the outside cylinder, which
solved the problem completely. Figure 8 shows the
development approach followed.
3.2
Further process development: Ti slender beam

case study for vibration tests
As part of process evaluation, a slender bar with dimensions 203 mm 20 mm 0.5 mm was chosen for a
benchmark study (vibration tests had to be done on
sample). A 3 mm support base was added along the
entire length of the sample, with 4 mm added to the
height of the sample in order to cut it off to size from
the platform via the wire cutter, as can be seen in
figure 9. On the first platform only one part was produced. During the growing process it was noted that
the part started warping at a height of approximately
15 mm from the base. It is assumed that the warping was caused by internal/residual stresses induced
in the part as a result of the rapid heating and cooling
cycle.
Figure 9. Images of Ti-sintered thin-wall test pieces.
As the part started warping, the parts orientation

in relation to the laser changed and caused the laser
to build on powder no longer supported by the bed of
powder underneath, causing the delaminating as seen
clearly in figure 9(A).
It was decided to grow a second platform with 4
parts close to each other to investigate whether the
parts alongside each other would support the recoater
to move across the exposed layers surface without
damaging the parts. At this stage the assumption was
that the size of the part caused the recoater to damage it
and was not fully aware that the problem/damage were
as a result of internal stresses. Realizing this, a third
platform was attempted. Figure 10 shows the altered
designs for the third platform with added design features to parts to keep it upright throughout the building
process. The features were small enough so that it can
be removed by hand after the parts were cut off with
the wire cutter. This did in fact solve the problem.
Protectors were also added on each side of the parts
to prevent the recoater from damaging the ends. This
proved to be unnecessary and rather resulted in a hurdle, as growing had to be stopped before completion,
due to the recoater getting stuck on it.
149
Figure 10. Image of Ti-sintered thin-wall test pieces with

protective shields.
Figure 12. Collage of images showing the scapula CT image

data manipulationAM in Nylon Polyamide (A), Conventional
Machining (D), AM in Ti (B) final AM Ti part with supports
removed, final product, exhibition model and final implant
(X-ray).
Table 1. Growing parameters used for
support generation.
Hatching
Rotation Angle
Hatching Teeth
Height:
Top length
Base Length
Base Interval
Fragmentation
X-interval
Y-interval
Separation Width
Border Teeth
Height:
Top length
Base Length
Base Interval
Figure 11. Collage of images showing the scapula CT image

data manipulationAM in Nylon Polyamide (A), Conventional
Machining (D), AM in Ti (B) final AM Ti part with supports
removed, final product, exhibition model and final implant
(X-ray).
3.3
Scapula case study
The CUT IPD teams involvement in the development

of a scapula as described in section 1.4.3 (Scapula and
Pelvis development) led to the development of two
process chains as shown in Figure 11 below. Image
A represents the Nylon prototype, from where two
options are available, namely A-D, as conventional
route, or A-B-C as AM route. Figure 11 shows the
process chain going from Mimics generated data to the
pre-operation model, when a decision must be made
on the process to follow. For the scapula implant developed, the conventional CNC machining option would
cost approximately 8000, whilst theAM development
route would limit the cost to approximately 1500.
The conventional approach has a material wastage of
approximately 75% due to the use of a solid titanium billet. Machining the billet to the actual part
will take approximately 5 days, including roughing
and finishing of the part.
X = 0.7 mm
Y = 0.7 mm
45
1.5 mm
0.15 mm
0.5 mm
0.1 mm
4 mm
4 mm
0.7 mm
1.5 mm
0.2 mm
0.5 mm
0.1 mm
3.4 Scapula support manipulation study

Due to the cumbersome process to remove supports
(both physical as well as in terms of safety), further developments have been done to ease the support
generation so as to deliver the model in the shortest
possible time, together with minimization of stresses
that may occur in the part. Figure 12 shows a collage of images that explains the support generation
development.
Table 1 shows some of the adjusted settings and
parameters to enable the growing of the part.
3.5 Titanium fork case study
The forks, as shown in figure 13 below, are used as part
of a mechanical design assembly to replace resected
bone. It requires very fine structures to encourage
osteo-integration. These structures were supported at
150

piece.

piece.
3.7 Orbital floor plate case study

piece.
the bottom by a ring (marked C on the image). To

protect these structures from the recoater a supporting ring was also grown around each fork as can
be seen in images A and B. The bottom surface
was extruded onto the substrate to prevent separation
from it.
3.6
The CUT IPD team was approached by the University

of Wales to assist with a model for a patient from the
Moriston hospital (UK) that required an eye orbital
floor implant. The geometries were developed by the
University of Wales, and the CUT IPD team sintered
the orbital floor plate in Titanium. Again, using the
previous experience gained, the case did not present
any challenges, and the sintered implant was presented
to the UK team.
During implantation, the orbital periosteum was
raised through 360 around the right orbit via a coronal
incision. A previous facial scar was opened from the
lower eyelid up to the upper lip. The implant was fixed
in situ to the orbital rims with two 1.5 mm diameter
titanium screws. Cathal ligaments were identified and
repositioned superiorly with a screw on the right and a
manipulator on the left. The lip scar was revised and
the skin and other soft-tissues advanced over the titanium implant tissues were then closed in layers and
the scar revised. The surgery was completed without
any complications.
A remarkable aspect is that the Ti orbital floor plate
as sintered at CRPM (SA), was a perfect fir to the SLA
skull shown in figure 16, which also proves interoperability of AM systems, and significant contribution to
future AM-driven batch/flexible manufacturing.
Mandibular reconstruction case study
The CUT IPD team was approached to assist with

a model for a patient that required a manidibular reconstruction. Previous experience gained in
using asymmetry studies through manipulations and
Boolean operations using Materialise MIMICS and
MAGICS , assisted the team in collaborating with
the University of Maribor, Slovenia. Due to extensive
experience gained through a high number of studies
completed, the case presented very few problems, and
an implant could be created in sintered Ti to support
a successful surgery. Figures 14 and 15 shows images
representing the process followed from using scanned
data to final implantation.
3.8 Nose bone orbital floor plate case study

Figure 17 shows a nose bone implant created for a
Moriston patient.
Following the previous successful intervention, the
CUT IPD team was again approached by the University of Wales to assist with a model for a patient
from the Moriston hospital (UK), who required a
nose bone implant. Once again, the geometries were
developed by the University of Wales, and the CUT
IPD team sintered the nose bone implants shown
in figure 17. Again, using the previous experience
gained, the case did not present any challenges, and
the sintered implant was presented to the UK team.
151

piece.
Figure 18. Flow chart showing the proposed work flow for
the CUTs IPD team in CRPM to adhere to certification
requirements.
Figure 17. Image of a Ti-sintered nose bone implant.
CONCLUSIONS AND RECOMMENDATIONS
The research case studies reviewed forms part of an

action research methodology used by the CUTs IPD
team, and cover a fairly wide field ranging from traditional industry applications (but in a new paradigm
using AM), to support a very diverse and wide array
of medical applications.
It also shows that AM and conventional manufacturing technologies can be used in parallel, to achieve
results. In some of the results produced, it also shows
that AM does in fact become competitive when considering costs effectiveness. Results achieved from the
case studies above, compare favourably with those
reviewed, as well as the recommendations and forecasts made by internationally recognised scientists in
the AM and materials field, as cited in the literature
review.
Also, it once more prove the movement from RP
to AM, and more specifically for Biomedical applications.
In reviewing the results, it again shows that the
CUTs IPD team remains one of a handful of institutions world-wide (and still the only one in SA) that can
Figure 19. Flow chart showing the proposed process development from RP to AM of certified DMLS production
parts.
directly sinter Ti or any other bio-compatible material,

to affect Medical Product Development.
The IPD team played a key role in a number of difficult product development and surgeries for local and
152
international patients. Many of these surgeries would

have been almost impossible without the research
teams intervention.
The case studies however, also revealed that there
are a number of process and organisational changes
that can be made to improve the outputs generated,
and to support DMLS processes (and AM in general)
towards certification.
The flow charts shown in figures 18 and 19 are part
of the IPD teams recommendation towards improvement and optimisation of the DMLS process for
Biomedical applications.
ACKNOWLEDGEMENTS
The authors would like to acknowledge the contribution of fellow staff and students from CRPM, the
support and collaboration of Drs G Vicatos and K
Hosking, as well as Drs R Bibb and D Eggbeer, as
well as Prof I Drstevensek.
Furthermore, the authors would like to acknowledge the National Research Foundation (NRF), Medical Research Council (MRC), Technology and Human
Resources for Industry Programme (THRIP) and DST
through the Tshumisano and Technology and Innovation Agency (TIA) support given to make the
infrastructure and research possible.
Disclaimer: Any opinion, findings and conclusions
or recommendations expressed in this material are
those of the authors and thereby the NRF does not
accept any liability in regard thereto.
REFERENCES
Gibson, I., Cheung, L. K., Chow, S. P.,Cheung,W. L., Beh,
S. L., Savalani, M. & Lee, S. H. 2006. The use of rapid prototyping to assist medical applications, Rapid Prototyping
Journal 12: 5358.
Hieu, L. C., Bohez, E., Vander Sloten, J., Phien, H. N.,
Vatcharaporn, E., Binh, P. H., An, P. V. & Oris, P. 2003.
Design for medical rapid prototyping of cranioplasty
implants. Rapid Prototyping Journal 9: 175186.
Hieu, L. C., Zlatov, N., Vander Sloten, J., Bohez, E., Khanh,
L., Binh, P. H., Oris, P. & Toshev, Y. 2005. Medical
rapid prototyping applications and methods. Assembly
Automation, 25: 284292.
Hoeges, S., Lindner, M., Fischer, H., Meiners, W. & Wissenbach, K. 2008. Manufacturing of bone substitute
implants using selective laser melting, in Vander Sloten, J.
Hollander, D. A., Wirtz, T., Von Walter, M., Linker, R.,
Schultheis, A. & Paar, O. 2003. Development of individual
three-dimensional bone substitutes using SLM.European
Journal of Trauma 4: 228234.
Murr, L. E., Quinones, S. A., Gaytan, S. M., Lopez, M. I.,
Rodela, A., Martinez, E. Y., Hernandez, D. H., Martinez,
E., Medina, F. & Wicker, R. B. 2009, Microstructure and
mechanical behavior of Ti6Al4V produced by rapid-layer
manufacturing, for bio-medical applications. Journal of
Mechanical Behavior Biomedical Materials 2: 2032.
Santos, E. C., Shiomi, M., Osakada, K. & Laoui, T. 2006.
Rapid prototyping of metal components by laser forming,
International Journal of Machine and Tools Manufacture
46: 145968.
Sercombe, T., Jones, N., Day, R. & Kop, A. 2008. Heat
treatment of Ti-6Al-7Nb components produced by selective laser melting. Rapid Prototyping Journal 14:
300304.
Singare, S., Yaxiong, L., Dichen, L., Bingheng, L., Sanhu,
H. & Gang, L. 2006. Fabrication of customized maxillofacial prosthesis using computer-aided design and rapid
prototyping techniques. Rapid Prototyping Journal 12:
206213.
Tolochko, N. K, Arshinov, M. K, Gusarov, A. G. Victor I.
Titov, V. I. Laoui, T & Froyen, L. 2003. Mechanisms of
selective laser sintering and heat transfer in Ti powder.
Truscott, M., de Beer, D. J. Vicatos, G.,Hosking, K., Barnard,
L., Booysen, G. & Campbell, R. I. 2007. Using RP
to promote collaborative design of customized medical
implants. Rapid Prototyping Journal 13: 107114.
Uklejewski, R., Winiecki, M., Rogala, P., Mielniczuk, J.
2011. Selective laser melted prototype of original minimally invasive resurfacing hip endoprosthesis, Rapid
Prototyping Journal 17(1): 7685.
Vandenbroucke, B. & Kruth, J.-P. 2007. Selective laser
melting of biocompatible metals for rapid manufacturing of medical parts. Rapid Prototyping Journal 13(4):
196203.
153
CAD and 3D Data Acquisition Technologies

A prototype of a spherical tippe top

M.C. Ciocci
University College West Flanders, Kortrijk, Belgium
B. Malengier
University of Ghent (UGent), Ghent, Belgium
B. Grimonprez
University College West Flanders, Industrial Design Center, Kortrijk, Belgium
ABSTRACT: Among spinning objects, the tippe top exhibits one of the most bizarre and counterintuitive
behaviours. The commercially available tippe tops basically consist of a section of a sphere with a rod. After
spinning on its rounded body, the top flips over and continues spinning on the stem. It is the friction with the
bottom surface and the position of the center of mass below the centre of curvature that cause the tippe top to rise
its centre of mass while continuing rotating around its symmetry axis (through the stem). The commonly used
simplified mathematical model for the tippe top is a sphere whose mass distribution is axially but not spherically
symmetric, spinning on a flat surface subject to a small friction force that is due to sliding. According to the
eccentricity of the sphere and the Jellet invariant (which includes information on the initial angular velocity)
three main different dynamical behaviours are distinguished: tipping, non-tipping, hanging (i.e. the top rises but
converges to an intermediate state instead of rising all the way to the vertical state.). Subclasses according to
the stability of relative equilibria can further be distinguished. Our concern is the quantitative verification of
the mathematical model. We applied 3D-printing to manufacture a 3-in-1 toy that could catch the three main
characteristics defining the three main groups in the classification of spherical tippe tops as mentioned above.
This toy allows verification by experimentation.
INTRODUCTION
Spinning toys are among the most ancient toys, and

there is a great variety of them. It is quite simple to
start spinning objects like a top or a gyroscope, and
though it is simple to explain their motion in general,
it is challenging to write down the detailed equations of
motion. Among spinning objects, the tippe top exhibits
one of the most bizarre and counterintuitive behaviour.
This top, similar to a mushroom (see Figure 1), flips
over on its stem after spinning for a while. Although a
detailedexplanation of the flip-over is not simple, we
note that rotation is only stable around the symmetry
axis. It is the friction with the bottom surface and the
position of the centre of mass below the centre of curvature which causes the top to rise. Remarkably, at the
inverted state, the center of mass lies higher than at the
initial condition, defying gravity. Experimentally, it is
known that such a transition occurs only if the initial
spin exceeds a certain critical threshold.
The commercially available tippe tops were
patented in Denmark in the 50 s, they always basically consist of a section of a sphere with a rod and
they always tip, see e.g. Figure 2. The commonly
used simplified mathematical model for the tippe top
is a sphere whose mass distribution is axially but
Figure 1. The tippe top, showing inversion.
not spherically symmetric, spinning on a flat surface subject to a (small) friction force that is due to
sliding. Several theoretical articles have been published on the topic, see e.g. (Cohen 1977; Gray and
Nickel 2000; Or 1994; Bou-Rabee et al. 2004; Routh
1905; Ciocci and Langerock 2007; Ciocci et al.; Branicki et al. 2006a; Branicki et al. 2006b; Ebenfeld
and Scheck 1995; Rauch-Wojciechowski et al. 2005;
Moffatt et al. 2004) and references therein for a survey
of the literature. The bifurcation diagrams and phase
diagram given in (Ciocci and Langerock 2007; Ciocci
et al.) provide a global geometric understandig of the
dynamical system modelling the tippe top as an eccentric sphere. Summarizing the results, what is found is
157
Figure 2. From the left: wooden and plastic commercially

avaliable tippe top, spherical inox/rubber salt pot, plastic
spherical shell with a round weight glued on the bottom.
the following. The type of asymptotic dynamics is a

function of the Jellett invariant (which includes information on the initial angular velocity) and eccentricity
of the sphere. Three main different regimes are distinguished: no tippe top phenomenon occurs no matter
what the initial spin is, tippe top dynamics may occur
if the Jellett invariant (which is proportional to the initial spin) is sufficiently large, or incomplete tippe top
behaviour occurs, where the top rises but converges
to an intermediate state instead of rising all the way
to the vertical state. So according to the mathematical
model there are different types of tippe tops and they
do not always tip. Since it seems very easy to make
a tipping spherical tippe top starting from a spherical shell, for example by gluing an candy (M&M)
in a small christmas ball, or spinning a nice spherical salt pot, see Figure 2. We also wondered if it
was possible to produce a 3D object which could capture the three main characteristics defining the three
main groups in the classification of spherical tippe
tops as mentioned above. We wondered if the mathematical model was not only qualitatively correct but
also quantitatively correct. In (Ciocci et al.) computer
simulations were implemented based on the system of
differential equations describing the system. Though
very accurate, these do not guarantee the correctness
of the model. The question about the predictive capabilities of computer simulations was yet unanswered.
Rapid prototyping gave us the possibility to validate
the theoretical model by experimentation.
In our case we used printing with the commercial
available Dimension SST1200es, the printing technology is based on the FDM principle (Fused Deposition
Modeling) in ABSplus. The rapid conversion of a digital design into solid reality gave us great flexibility
in testing different design and their effects. Everyone from architects to product designers can use the
technique to make models, we challenged ourselves
in producing a 3-into-1 toy that brings a mathematical
model into life verifying the theoretical classifcation
of tippe tops as presented in (Ciocci and Langerock
2007; Ciocci et al.).
A second but not less important goal was to make
the toy also visually impressive so that it could be
used as visual aid in physics or mechanics lessons.
Figure 3. Eccentric sphere version of the tippe top. R is the

radius of the sphere, the center of mass O is off center by
.
The top spins on a horizontal table with point of contact Q.
The axis of symmetry is Oz and the vertical axis is OZ, they
define a plane I (containing OQ) which precesses about OZ
with angular velocity . The height of O above the table is
h().
The students can visualize physical phenomena and

carry out tests within the classroom environment.
We believe that the use of this type of prototyping improves communication between teacher and
students and makes abstract concepts more accessible.
2 THE TIPPE TOP MODEL

In Figure 3, we sketch an eccentric sphere as considered in the tippe top model presented in (Ciocci and
Langerock 2007). According to the literature, the three
main groups of tippe tops are distinguished by conditions involving the so-called inertia ratio A/C = Ix /Iz
and eccentricity
of the sphere. Here, Ix = Iy = A,
Iz = C are the moment of inertia of the body referring toa principal axis system Oxyz, whose z-axis is
the symmetry axis of the sphere. The eccentricity
is
the distance between the center of mass and the geometric center of the sphere, with 0 <
< R, where R
is the radius of the sphere. The three groups are as
follows.
1. (A/C 1) <
/R: the tippe tops belong to the
so-called group I. For these tops the position = 0
is always stable, which implies that if these tops are
spun with the symmetry axis in vertical position
and with the heaviest part down, they wont flip
over.
2.
/R < (A/C 1) <
/R: the tippe tops belong to
the so-called group II. These tops show complete
inversion if spun fast enough. All the commercially
available tippe tops belong to this group.
3. (A/C 1) >
/R: the tippe tops belong to group
III, these tops may show incomplete inversion. That
is, they might flip but up to a certain angle < .
158
The classification can be refined as in the following

result, see (Ciocci and Langerock 2007; Ciocci et al.)
for the proof. Denote
PROTOTYPE DESIGN
3.1 Requirements
We aim to design a spherical tippe top that by a control/
change is part of group I, II or III. It is clear that the
problem will be coupled: changing a moving part in
the sphere will change A/C and
/R at the same time.
Furthermore, the toy should be easy to handle. The
target group are young people, so the toy should grasp
their imagination and curiosity. As pointed out, to realize the toy we need an axially symmetric sphere with
control over A/C and
/R. Two possible designs are
natural to consider:
Define also
a solid sphere with a cilindrical hole through the

center wherein a weight can move;
a hollow sphere with a cilindrical rod on which a
weight can move.
and b := arccos(b), c := arccos(c). Also, denote the

Jellet invariant J by
where n is the spin of the top around its axle.

Result (Ciocci and Langerock 2007) The bifurcation
diagrams of the eccentric sphere spinning on a flat
surface with small friction fall in one of the following
six categories:
Group I: A/C 1 <
/R
The vertical state = 0 is stable for any value of J.
The vertical state = is stable if |n | > n+ , and
unstable otherwise.
Intermediate states exist for all values of satisfying > c . Group Ia: b < 1; The entire branch of
intermediate states is unstable. Group Ib: b > 1;
The branch of intermediate states has a fold point at
= b . The branch with > b is stable, while the
branch with < b is unstable.
From the bifurcation diagram presented in (Ciocci

et al.) it is clear that it is possible to hit the three
groups by fixing
/R and changing A/C. Therefore, it
is important to understand for both designs how A/C
and
vary with respect to each other when the weight
is pushed up or down. We set up a Maple worksheet based on the mathematical description as given
in (Ciocci and Langerock 2007; Ciocci et al.) and calculated for both designs A/C and
in function of the
position of the midpoint of the moving weight with
respect to the center of the sphere; this will be further
on denoted by Z2 . We took into account the physical
parameters: dimensions of the different parts (radii,
heights, thickness) and the density of the materials.
From this we realized that for the solid sphere the goal
of three types is within reach, whereas for the hollow
sphere the design has to be significantly changed.
We now discuss our findings for the prototypes
realized according to the two suggested designs. Our
realizations were all printed with the commercial
available Dimension SST1200es with printing technology based on the FDM principle (Fused Deposition
Modeling) in ABSplus.
Group II:
/R < (A/C 1) <
/R.
The vertical state = 0 is stable if |n0 | < n and
unstable otherwise.
The vertical state = is stable if |n| > n+ and
unstable otherwise.
Intermediate states exist for all . We distinguish the following three subgroups. Group
IIa: (A/C 1) < (
/R)2 and |b| < 1. A fold
bifurcation of intermediate states occurs. Group
IIb: (A/C 1) > (
/R)2 or b > 1. The entire
branch of intermediate states is stable. Group
IIc: (A/C 1) < (
/R)2 and b < 1. The entire
branch of intermediate states is unstable.
Group III: (A/C 1) >

/R.
The vertical state = 0 is stable if |n0 | < n .

The vertical state = is unstable for all J.
Intermediate states exist for < c , and are all stable.
3.2 Sphere with cylindrical hole and setscrew

For the first design we realized three different tops,
varying the geometrical dimensions. This was done
because the calculation showed that for the given
materials some zones are hard to achieve or are very
narrow, see Figure 6.
The prototype consists of a sphere with a cylindrical hole through the center, together with a piece of
adjustable cylindrical iron wire (setscrew), see Figure 4. With a caliper, it can be checked how deep the
setscrew is set in the hole. The position of the midpoint of the setscrew with respect to the center of
the sphere is Z2 . The hole is suitable for a setscrew
M12. The dimension of the toy were chosen based on
the mathematical calculations derived from the model.
The diameter of the sphere was chosen so that one can
comfortably spin the toy by hand. With a sphere of
159
Figure 4. CAD design of the first prototype.
Figure 5. The first prototype.
diameter 50 mm, good values for the chosen design

are a hole of radius 5.5 mm, filled with the setscrew of
height 15 mm; or a hole of radius 1.5 mm, filled with
the setscrew of height 3 mm, see Figure 6. The densities are 1.08 g/cm3 for ABSplus, and 7.87 g/cm3 for
the setscrew.
The prototype is axially symmetric, therefore only
the eccentricity
and the moment of inertia A are
function of Z2 , they are easily calculated; C remains
constant when moving the setscrew up and down. In
Figure 6 the quantities A
1, R
and (
(Z2 )/R)2
C
are plotted as functions of Z2 , top for the prototype
with a M12 set screw and bottom for the prototype
with the M3 set screw. The printed prototype, Figure 5,
is of the first type and exhibits the predicted behaviour
as follows: for Z2 between 0 mm and 7.95 mm the
toy does not show tippe top dynamic no matter what
the initial spin is (type I), for Z2 between 7.95 mm
and 17.76 mm complete tippe top dynamic is observed
(type IIc). For Z2 above 17.76 mm the top is of type IIb
(incomplete tipping is observed if initial spin is not sufficiently high). The prototype behaves as predicted by
the model. For a prototype fit for a M3 screw, the intervals for Z2 are as follows: for Z2 between 0 mm and
3.33 mm the toy does not show tippe top dynamic no
matter what the initial spin is (type I), for Z2 between
Figure 6. Different tippe top regimes in function of the position of the midpoint of the setscrew. The straigth lines are
2)
(Z2 )/R, and the positive parabola is A(Z

1, the negaC
tive parabola is (
(Z2 )/R)2 . Top the result for the printed
prototype with a M12 setscrew, see Figure 5. Bottom, the
result for a prototype of the same shape but a setscrew M3.
3.33 mm and 18.20 mm complete tippe top dynamicis

observed (type II). For Z2 above 18.20 mm the top is
of type III (incomplete tipping is observed).
We conclude that this prototype can give a working
3-into-1 toy, but has some disadvantages.
160
The setscrew can come loose after intensive use.

When there are three zones present, the stable zone
is very small, while type III is best when the setscrew
comes out of the sphere.
Using a caliper to know if the setscrew is in the
center is not practical.
Figure 7. CAD rendering of the second prototype.

Figure 9. Different tippe top regimes in function of the
position of the midpoint of the setscrew. Curves as in Figure 6.The physical parameters for the construction are here:
radius of the spherical shell in ABS 25 mm, thickness of the
shell 2.5 mm, radius of the iron cylindrical rod 3 mm, radius
of the (iron) weight 10 mm, height of the weight 10 mm.
Figure 8. CAD design of the second prototype.
3.3
Sphere with cylindrical rod
The second design consists of a spherical shell with a

cilindrical solid rod through the center along which a
symmetric bead is spun, this bead can be put at different heights along the rod. In this design, the user must
open the sphere, change the position of the weight by
screwing it up or down, after which the sphere can be
closed and spun. See Figure 7 and Figure 8.
The advantage of this design is that different ABS
colors can easily be used for both sides making the
tipping very visual, and that the rod can be marked
at the critical positions. Where in the first design
setscrews of different length can be used, in this design
weights of different lengths and different widths can
be considered within one toy.
However, computations show that a 3-into-1 toy was
difficult to obtain with the chosen materials. Different promising improvements of the design are now
being tested. We briefly summurize our findings that
form the basis of further improvements. We tried
three different possibilities for the rod: iron, nylon
and ABS. Nylon and ABS seems towork best.With an
iron rod and physical pararmeters that allow easy playing with the toy we didnt succeed in catching all the
three zones. See below for the specific values of the
parameters. As Figure9 shows, only zone three is hit,
Figure 10. Different tippe top regimes in function of the

position of the midpoint of the setscrew. Curves as in Figure 6.The physical parameters for the construction are here:
radius of the spherical shell in ABS 25 mm, thickness of the
shell 2.5 mm, radius of the nylon cylindrical rod 2.5 mm,
radius of the (iron) weight 10 mm, height of the weight
10 mm.
which means that this tippe top never shows complete

inversion but may tip up to a certain angle.
The prototype is axially symmetric, therefore only
the eccentricity
and the moment of inertia A are function of Z2 ; C remains constant when moving the weight
up and down. The physical parameters for the construction are here: radius of the spherical shell 25 mm,
thickness of the shell 2.5 mm, radius of the cilindrical
161
rod 3 mm, radius of the weight 10 mm, height of the

weight 10 mm.
In the case of a nylon or ABS rod, it was possible to
obtain a tipping top. As illustrated in Figure 10, both
zones II and zone III are hit, which means that complete tipping and incomplete tipping may be observed
according to the position of the weight. Note that the
section of the curve in zone III is very small, which
make it very difficult to observe the hanging phenomenon. A similar obesrvation holds for the section
of the curve in zone I; this is so small that non-tipping
behaviour cannot in practice be observed.
We conclude by remarking that different promising
improvements of this design are now being tested.
REFERENCES
Bou-Rabee, N., J. Marsden, and L. Romero (2004). Tippe
top inversion as a dissipation induced instability. SIAM J.
Applied Dynamical Systems 3, 352377.
Branicki, B., H. Moffat, and Y. Shimomura (2006a). Dynamics of an axisymmetric body spinning on a horizontal
surface. iii. geometry of steady state structures for convex
bodies. Proc. R. Soc. A 462, 371390.
Branicki, B., H. Moffat, and Y. Shimomura (2006b). Dynamics of an axisymmetric body spinning on a horizontal
surface. iv. stability of steady spin states and the rising

egg phenomenon for convex axisymmetric bodies. Proc.
R. Soc. A 462, 32533275.
Ciocci, M. and B. Langerock (2007). Dynamics of the
tippe top via routhian reduction. International Journal of
Bifurcation and Chaos 12(6), 602614.
Ciocci, M., B. Malengier, and B. Grimonprez. Towards a
prototype of a spherical tippe top. submitted to JAM.
Cohen, C. (1977). The tippe top revisited. Am. J. Phys. 45,
1217.
Ebenfeld, S. and F. Scheck (1995). A new analysis of the
tippe top: Asymptotic states and lyapunov stability. Annals
of Physics 243, 195217.
Gray, C. and B. Nickel (2000). Constants of the motion for
nonslipping tippe tops and other tops with round pegs.
Am. J. Phys. 68, 821828.
Moffatt, H., Y. Shimomura, and M. Branicki (2004).
Dynamics of axisymmetric body spinning on a horizontal surface. i. stability and the gyroscopic approximation.
Proc. R. Soc. Lond. A 460, 36433672.
Or, A. (1994). The dynamics of a tippe top. SIAM Journal on
Applied Mathemathics 54, 597609.
Rauch-Wojciechowski, S., M. Skldstam and T. Glad (2005).
Mathematical analysis of the tippe top. Regul. Chaotic
Dyn. 10, 333362.
Routh, E. (1905). Dynamics of a system of rigid bodies. New
York: MacMillan.
162

An innovative methodology for laser scanner integration in a robot

cell for small batch production of sculpture artworks
C. Cenati, G. Borroni, L. Cevasco, D. Parazzoli & M. Danesi
National Research Council (CNR) Institute of Industrial Technologies and Automation (ITIA), Milano, Italy
ABSTRACT: Nowadays SMEs have a high quality production but relatively small batches respect to current
automation technologies that are tailored on big production batches.
There is the request to develop innovative methods and tools to increase flexibility and reconfigurability of
systems for SMEs; the goal is short delivery time for high quality customized products.
In this paper we discuss a case study of a robotic system for complex surfaces of statues extraction, model
digitalization and generation of part program for rapid prototyping or small batch series production.
A complete robotic cell was developed based on a milling manipulator equipped with a rotary table and a
laser scanner.
In the first section the challenging application, on the antique stone carving, is presented; the scanning system
completely self-made at ITIA-CNR laboratories is able to scan, in a reliable way, in an uncontrolled environment
and characterized by an open and modular architecture.
In the second section the integration of the industrial robot with the laser scanner is analyzed within the
calibration algorithm. The robot has to manipulate the scanner in order to cover all parts of the target object;
normally each sensor position is described through an individual and local coordinate system, an innovative
calibration routine was developed to allow transformation of all the data collected in a global and unique
coordinate system.
The system becomes fully autonomous and able of dealing with small production series with minimum human
intervention.
In the third section the prototype is presented with particularly attention on the ITIA-CNR laser scanner
solution: a low cost instrument with high performances.
With an innovative calibration methodology it has been possible to shift the focus from the components and
the technological processes to the power of the calculus because it requires the management of a huge quantity
of data.
In the last section experimental results showed the solution flexibility.
For example, the 3D data can be used by artists in their work and as a tool to realize their ideas. The artist can
create sculptures one tenth of the size of the final sculptures and employ the milling system for realize a small
batch production of their works of art.
INTRODUCTION
1.1 The automation on small bathes and

customized production
In the last decade, 3D CAD-CAM software solutions
have radically changed the design procedures and,
as a consequence, production methodologies. Among
other innovations, CAD systems have introduced the
use of free-form or sculptured surfaces in the product design, and they can be considered a key-feature
in the shift from high-volumes production systems
to the new small-batch-size paradigm. Nevertheless,
to exploit the new CAD Features, CAM full integration in the production system, and, indirectly an high
plant flexibility are needed. In fact, to face the increasing production variability (in time and volumes) high
degree of automation, and extremely fast time in plant

reconfiguration is mandatory.
However, by now industrial manipulators have still
to work, manipulate and check parts with complex
shapes, and it is an hard limit both in the programming phase of extremely complex task and in the
execution phase where accuracy is loss.Therefore, new
design paradigm is needed to address higher demands
in robotic including safety and robustness, as well
as flexibility and dexterity. In addition, such features
should also allow the robot to be used in unstructured task, that is, in operations where all requirements
are imposed both in planning and in the execution
phase.
In this scenario, the vision technology is probably
the most important research field adjacent to industrial
automation.
163
1.2
3D laser scanning integration in robotic cells

for antique stone carving
As example, a challenging application field for standard industrial robots is the antique stone carving,
where the complexity is both in the acquisition of the
master statue and in the robot programming. The application requires, usually, that the measure system has
to be moved in different positions in order to cover the
entire carving perimeter and different measure angles
have to be imposed to avoid undercuts. In this scenario
various issues are still open, due to the fact that the
correct trade-off between accuracy and costs is challenging, and since extremely critical time requirements
are imposed to face small/medium production batches.
Although the apparently simplicity, antique statue
carving points out the limits of the integration of
current scanning technologies with industrial manipulators, and by now, in the knowledge of the authors,
robust solutions based on off-the-shelves devices are
far away to be available. In fact, a still hard lack consists on the availability of low cost measuring system
in order to allow the use of them in a wide field of
application.
The paper presents both an innovative laser scanner
prototype completely self-made at ITIA-CNR laboratories and its fully integration with standard industrial
robot.
The laser scanner can be easily mounted at the endeffector of an anthropomorphic industrial manipulator
and interfaced with the robot controller. An innovative
calibration procedure of the robot-laser scanner integrated system is described, and experimental results
for antique stone carving are reported. Results demonstrate the robustness and the performance reached
by the developed prototype. The prototype design
has shifted the measure accuracy requirements from
high quality components to an extrinsic-model of the
measures-space, that is, an innovative non-parametric
calibration method, based on complete mapping of
the working scanner volume has been developed. The
drawback of this approach consists on the management
of a considerable volume of data and identification of
efficient algorithms for processing procedures. To face
these problems, the calibration calculation has been
implemented by the means of a self-developed embedded system based on a scalable multi-CPU cluster, with
GNU/Linux O.S.
2
2.1
CALIBRATION ALGORITHM
Scenario of measuring system in a robot cell
integration
To clarify the demonstration scenario, consider a laser

scanner device (or in general a generic measure system) connected to the end-effector of the robot. The
main problem consists on the fact that robot has to
be moved in order to allow a whole acquisition of
the objects avoiding undercuts effects. It is trivial that
the measure accuracy is given by (i) the accuracy of
the measuring system, by (ii) the robot position precision, and by (iii) the estimation of the relative position
of the measuring system with respect to the robot
end-effector.
To overcome the problem of the calibration of the
acquisition system with respect to the robot, usually
the reconstruction of the 3D object model on the basis
of the data coming from the various acquisition is
hand-made and only partially automated.
Use of special software (e.g. Geomagic, MeshLab
etc.) for replacement of individual point clouds is the
straightforward solution.
Unfortunately, this approach imposes that scans
overlap each other for a sufficient portion. Three main
disadvantages:
Qualified operator, able to use programs for handling points clouds correctly, on a computer outside
the system, are required
Large number of scans to get enough overlap are
required
Low reliability of point clouds registration using
such software systems, minimization might not
converge.
To overcome these limitations, passive markers can be
integrated for referencing different points clouds but
it does not allow a complete process automation. In
fact, programs for points clouds manipulation, cannot
be left and the markers must be placed in the working
volume in order to have a number of markers overlapping between two adjacent scans greater or equal to
three.
2.2 The new calibration procedure
With the aim to develop an high autonomous roboticintegrated laser scanner device, these approaches do
not fulfill and a new calibration of the measuring system with respect to the robot is needed. The proposed
calibration algorithm of the integrated system consists
on an innovative non-parametric algorithm that provides the analytical expression of the device in the
whole work volume. The function is given by a two
step calibration procedure:
self-calibration of the laser scanner device (see the

paragraph below)
calibration of the laser scanner measures in the robot
reference frames.
Concerning the latter, a self-calibration procedure, for
points back projection in a single reference system
that allow a native self-alignment of different point
clouds acquired was implemented (patent pending,
application number TO2010A000638 date 23/07/2010
entitled System and Method for three-dimensional
objects reconstruction).
For acquisition procedure an artifact with wellknow geometry that identifies a set of significant
points is used.
The procedure allows to compute the 3-D rigid
body transformation that aligns two sets of points
164
for which correspondence is known. The remarkable

points describing the artifact, in the base coordinate
system, are fixed and well known for geometric construction. For each scan pose it is necessary to process
the artifact point cloud, acquired from the scanner,
through an extraction and a fitting process of checking
surfaces.
Two corresponded point set can be defined:
{psi } remarkable points in the scanner coordinate

frame, with i = 1, . . . , N
{pbi } remarkable points in the base coordinate
frame, with i = 1, . . . , N
It is possible to write the relationship between the two
point set as follow:
where R is a 3 3 standard orthonormal rotation

matrix, T is a 3 1 translation vector and V i is a
vector that allows to consider any noise; in fact, the
extraction and fitting surface could be affected by a
minimum error. The optimal solution for [R, T] transformation allow the mapping of set points {psi } on to
{pbi } and the scanning point back projection on to the
base coordinate system.The solution typically requires
minimizing a least squares error criterion given by:
Defines the centroids ps, and pb of the sets points {psi },

and {pbi } respectively it is possible to write:
Like the single points, eq (2) can be expressed as
From algebra theory, whether we define the correlation

matrix K as
it is possible to assert that maximizing the trace:
is equivalent to the equation (2) minimization.

The unknown rotation matrix R that maximizes
the trace can be calculated by the singular value
decomposition of correlation matrix K .
and, as a consequence, the translation vector is estimated by
by aligning the centroids of the two reference systems.

It should be emphasized that the calibration can
ensure automatic points clouds assembling on the
ronde-bosse target guaranteeing high accuracy (less
than 0.1 mm). It is worth to underline that the robot
positioning accuracy is significantly lower than the
result but the calibration use only the manipulator
repeatability.
3
PROTOTYPE DESCRIPTION AND

APPLICATION SCENARIO
3.1 Heritage of 3D scanner system

Although structured light scanners are faster in the
acquisition procedure, they present various drawbacks
and, among others, a clean and lighting controlled
environment is needed. Hence, in the prototype design
choice, laser scanner technology has been preferred for
the acquisition system since it guarantees fast, accurate and reliable target data capture, and it ensures
robustness and flexibility also in actual industrial
environments. However, off-the-shelves laser scanners
display various limits for challenging application as
antique stone carving. In fact, they are usually sophisticated measurement tools, assembled by using high
quality components and manufactured by the means of
high precision machining (worldwide leader companies are located in Germany, Usa and Japan). Another
drawback consists on the fact that the most of available
laser systems are stand-alone measuring devices based
on proprietary transmission protocols, that is, they
do not allow an easy integration into robotic systems
(or full automated set-ups). Hence, the uses of these
typologies of 3D laser scanning devices are limited in
niche sectors where high costs are sustainable.
3.2 The ITIA-CNR laser scanner solution
Despite these various drawbacks, laser scanning technology is a mature and roust solution, and the prototype
developed is aimed to demonstrated that a new a
family of laser scanners that displaying low cost, easiness of integration into a robotic cell and adequate
performances respect to the task purposes are possible.
To reach these targets, we faced the accuracy problem by the development of a non parametric calibration
algorithm. The key-idea, is that by the means of a calibration station, analytical functions (3 /4 polynomial
series) map the 2D points acquired by the camera to
the 3D reference frame of the device.
Therefore, it is possible to design and build a
laser scanner by using off-the-shelves components
165
Figure 2. The cell work flow.
Figure 1. ITIA-CNR scanner on Kuka robot.
(e.g. low cost) because the accuracy and high quality

of components is not needed.
Otherwise this method presents the drawback that
various Gbytes of data have to be processed, but
this software problem has been solved by design
of properly HW/SW multi-cluster solution (Linux
Based).
The set-up developed is briefly described below.
The laser scanner is composed by a unit box containing all the sensors, devices and the PC embedded
that control all the system.
The prototype is provided by camera (1600 1200
pixels) and a laser (Class 3R) emitters; each laser emitter is moved by the means of a stepper-motor. The
module is provided with a set of lithium battery modules that allow a discrete power autonomy, and with
a Wi-Fi communication modules. This features allow
the prototype to be completely wireless and this guarantees an easy interfacing with industrial robots since
it can be used like a standard tool (e.g. as a drill, mill,
etc). In addition, the prototype design has been already
made taking into account industrial requirements as:
low device footprint

protection level (IP68)
safety issues (avoided clamping of body parts like
hands and fingers)
electric shock.
Interfaces for three major manufacturers of robots
Abb, Kuka, and Comau, for directly command the
laser scanner from the robot controller have be
developed.
These interface consist on a socket server daemon

running on the robot controller and developed in the
proprietary robot language.
This server allows an exchange of commands and
data from the robot and the scanner.
Despite the general purpose aim of the laser scanner
integrated system developed, the application tested for
the concept demonstration displays some peculiarity.
In fact, stone carving reconstruction and production
is characterized by the fact that the raw object the
robot has to mill presents unstructured default geometry. This means that the CAM, i.e. the identification of
the milling trajectories, has to be integrated with the
measure of the actual raw geometry. The laser scanner integrated with the robot allows the digitalization
of both the template model (finished product geometric model) and the raw product leading two important
advantages:
accurately know the morphology of the raw

avoid zero piece tool
Finally, the offline programming, using a commercial CAM (Alphacam in particular) make the system
highly flexible; it facilitate the starting set-up and
make easier and faster any changes and any process
optimization of robot machining parameters.
4
EXPERIMENTAL RESULTS
4.1 The prototype cell set-up

The manufacturing platform developed consists of a
6-dof robot manipulator equipped with an appropriate
flange to accommodate the laser probe. The laser scanner can capture subjects placed at 500 and 1500 mm
distance with an accuracy average of 0.1 mm using a
Class 2 laser.
The robotic arm is free to move through its working volume to allow the acquisition of subject for
360 deg. The calibration procedure has been specifically designed to take advantage from the excellent
repeatability of the manipulator. To make the system
more flexible, the cell has also been equipped with a
rotary table with a position control closed-loop.
166
Figure 5. Venus de Milo digitalized.
Figure 3. Worse head digitalized.
Figure 4. Worse head realized with the system.
The subject/model was mounted on a rotary

table for digitalization; a 6 d.o.f. robot 16 kg payload was used for laser scanner manipulation; a 6
d.o.f. robot 180 kg payload was used as machining
center.
Figure 6. Venus de Milo realized with the system.
4.2
Some results
A sequence of tests on statues with heterogeneous

dimension and particular were carried out; below a
list of few examples is presented.
Using 3D laser scanning and digital modelling data
sets have been well produced, the organic nature of the
statues and artwork is such that it would not have been
possible to achieve the desired results using traditional
CAD modelling and standard geometries.
Worse head: Abele Vadacca art work (contemporary
author) 2008 post Christ.
Venus de Milo: Alessandro di Antiochia art work

130 before Christ.
The results was presented in BIMU 2010 exhibition of machine tools, robots, automation and ancillary
products (NC units, tools, components and accessories), hosted by the new fieramilano fairground.
The laser scanning system has been incorporated
into a cell composed of two manipulator robots and a
rotary table.
167
Figure 8. Stand ITIA-CNR in BIMU 2010.
Figure 9. Stand ITIA-CNR in BIMU 2010.
REFERENCES
scale public art in tudor square sheffield. International

Archives of Photogrammetry, remote sensing and spatial
information sciences vol1 XXXVIII, Part 5 Commission V
Symposium New-castle upon Tyne, UK 2010Conference
on 3D Digital Imaging and Modeling, 2001.
Park J. & Guilherme N. & Avinash C.K. 2001. Dual-Beam
Structured-Light Scanning for 3-D Object Modelling.
International Conference on 3D Digital Imaging and
Modeling, 2001.
Peng T. & Gupta S.K. 2007. Model and algorithms for point
cloud construction using digital projection patterns. Journal of Computing and Information Science in Engineering,
7(4): 372381, 2007.
Umeyama S 1991. Least-squares estimation of transformationn parameters between two points patterns, IEEE Trans
Pattern Anal Machine Intell 13: 376380.
Chan S.F. & Kwan R. 2003. Post-processing methodologyies for off-line robot programming within computer
integrated manufacture. Journal of Material Processing
Technology 139 (2003) 814.
Ciocca L & Scotti R. 2004. CAD-CAM generated ear cast by
means of a laser scanner and rapid prototyping machine. J
Prosthet Dent 2004; 92: 5915 December. Bologna, Italy.
Goryn D & Hein S. 1995. CAD-CAM On the estimation of
rigid body rotation from noisy data. IEEE Trans Pattern
Anal Machine Intell 17: 12191220.
LA Pense A. & Parsons J. & Cooper M. & Broadbent S.
& Bingham R. and Watts R. & Meyers J. and Berry J.
2001. The use of 3D Laser scanning and 3D modeling in
the realization of an artistic vision; production of large
168

Application of time compression techniques to dental restoration:

A procedure for quality evaluation
A. Salmi, E. Atzeni & L. Iuliano
Politecnico di Torino, Department of Production Systems and Business Economics (DISPEA), Torino, Italy
ABSTRACT: In this work a procedure for evaluating the accuracy of novel manufacturing methods of dental
crowns is proposed. The procedure is based on a benchmark specifically designed for dental application. The
evaluation procedure starts with the achievement of the Reference Model (ReM), consisting of actual dimensions
of the reference artefact. Then the benchmark is digitized. The point-cloud (Virtual Benchmark VB) is inspected
by using a specific RE software to evaluate the accuracy of the digitization step. Thereafter caps are designed and
manufactured. After fabrication, the coupling surface of each manufactured cap is inspected and compared with
the VB and the ReM to provide information about the accuracy of the manufacturing process and of the whole
crown restoration production process, respectively. Two recent techniques are analysed: investment casting of
cobalt-chromium (CoCr) alloy from drop-on-demand (DoD) wax patterns and milling of zirconia.
1
INTRODUCTION
Recently, in restorative dentistry a great attention has

been focused on Reverse Engineering (RE) and Additive Manufacturing (AM) techniques and the possibility of integrating them into traditional procedures. An
emerging application can be noticed in the crown manufacturing process. The geometry of a stone replica
is digitized by a three-dimensional scanner. From the
point cloud a virtual model is generated and a tooth cap
could be designed on it and manufactured by Computer
Aided Manufacturing (CAM) or using AM techniques.
Since the quality of the product is critical for the
success of the patient treatment (Fransson et al. 1985,
Bindl & Mormann 2005, Tsitrou et al. 2007, Laurent
et al. 2008), it is important to evaluate deviations introduced by different steps of the production path. In fact,
only if deviations and related causes are known corrective actions can be performed at the design stage to
optimize the process. Especially, dimensional errors
are induced by surfaces digitization, by data elaboration, and by the fabrication sequence. However, the
information given by systems suppliers or present in
literature is lacking and does not allow comparison
among different manufacturing systems. Therefore, it
becomes essential to investigate new technologies to
quantify their accuracy and precision.
Currently, there is not a standard methodology for
evaluating the error introduced by novel manufacturing cycles, from the scanning of the replica to
the final dental device. It should be observed that,
because crown restorations are customized products,
not reproducible, it is not possible to define a standard procedure of inspection using real parts. Classical
features (such as spheres, cylinders, cones, or planes)
are fundamentals because geometric and dimensional
errors are defined on them. To have comparable data,

a standard benchmark must be employed (Miyazaki
et al. 2009).
In this work a novel procedure for evaluating the
accuracy of the manufacturing sequence for dental
caps is proposed. Deviations introduced by digitizing,
as well as cap design and fabrication, are studied. This
analysis is based on an innovative benchmark defined
in a previous study by the authors (Atzeni et al. 2009).
2 A NOVEL APPROACH FOR DENTAL

RESTORATION
The recent introduction of RE and AM techniques in
the dental field has marked a turning point in the prosthesis production methodology, from the traditional
hand made to the innovative automated approach. As
regards the fabrication of single dental restorations, the
novel RE/AM based crown manufacturing sequence
(Fig. 1) begins with the achievement of the virtual
model of a stone replica of the prepared tooth. As a
matter of fact, intra-oral scanners are now available
on the market, allowing for the direct digitization of
the oral cavity, but results quality is not yet satisfying, because the digitizing process is highly affected
by the real clinical condition, such as saliva, blood
or movement of the patient (Luthardt et al. 2005,
Kuhmstedt et al. 2007). On the other hand, the digitization of the oral impression is arduous because of
the concave geometries and could result in a poor virtual representation. Consequently, at the moment the
indirect procedure with impression taking and replica
production is still more accurate and thus the stone
replica is commonly digitized (Brosky et al. 2002,
Thongthammachat et al. 2002, Chen et al. 2004).
169
Figure 1. RE/AM based procedure for dental restoration.
Several companies have been introduced on the

market optical scanners specifically developed for
dental application (Dentalwings, Solutionixs Rexcan
DS, Maestro 3D dental scanner, ShapeGrabber 3D
dental scanner, 3shape R700 scanner, Delcams iMetric scanners, etc.). Within a few minutes, threedimensional data of the dental model is obtained,
which can be used for the cap design process (DeLong
et al. 2003, Quaas et al. 2007, Persson et al. 2008).
From the virtual model of the prepared tooth, a technician can design the cap using a specific CAD
tool, typically provided by the same supplier of the
three-dimensional scanner. In fact the common manufacturers strategy is to offer integrated solutions. In the
cap design, special attention is paid on the manufacturing process by which the dental part will be fabricated.
In fact each technique has specific requirements that
must be met. For instance, if the cap fabrication process involves a heat treatment of the cap, the related
shrinkage must be considered at the design stage.
The availability of the mathematical model of the
cap allows to produce the dental device by Computer Aided Manufacturing (CAM) tools or Additive
Manufacturing (AM) processes.
2.1
Recent techniques for caps production
Dental caps can be made of ceramic or metal. For a

long time metal caps have been preferred because of
the higher toughness and longer life, while ceramic
is brittle and cracks easily. Recently, the advent of
Zirconia ceramic and improvements in CAD/CAM
technologies have overcame the ceramic limits (Filser
et al. 2001, Luthardt et al. 2004, Christensen 2009,
Koutayas et al. 2009, Vagkopoulou et al. 2009, Zarone
et al. 2011), making zirconia caps a valid alternative
to metal ones.
2.1.1 Cobalt-chromium (CoCr) caps
The traditional technique to produce metal caps is
investment casting from wax patterns modelled by
hand. Recently, the introduction of RE and CAD/CAM
technologies in the dental sector has made possible to
produce caps by machining (Rekow 2006), but it is
important to notice that this is a costly process that
produces a large amount of waste. Currently, the attention is focused on the use of additive manufacturing
processes: AM techniques could be used directly to
produce the final cobalt-chromium caps or indirectly
to fabricate wax patterns for investment casting.
Cobalt-chromium caps can be produced by Direct
Metal Laser Sintering (DMLS) (Ucar et al. 2009). Due
to the need of supports in DMLS, marks will remain
on the supported surface of the cap. Thus the cap is

oriented upside down so that the inner surface is free
of support structure, in order to ensure the required fit.
Post processing includes the support removal and other
finishing operations such as the manual shot-peening
of the outer surface to improve the coating adhesion
and eventually the heat treatment to relieve residual
stresses.
The indirect procedure involves the fabrication
of wax pattern of the cap by additive techniques.
Wax pattern is then processed by investment casting,
cleaned and polished according to the traditional practice. At the present time, the two mainly used AM
techniques are Stereolithography (SL) and Drop-onDemand (DoD), because of the good balance between
cost and accuracy.
The direct manufacturing of the final cap is
extremely promising, but it should be observed that
the fabricated cap is a completely new product, primarily from the chemical and metallurgical points of
view. The consequence is that the process must be optimized and also the ceramic/porcelain coating process
must be redefined. Moreover long-time results will not
be available for years. On the other hand, the use of AM
to produce only wax patterns does not alter the characteristics of the final part, while the production time
is shortened and the restoration fit is improved. Taking into consideration the general reluctance shown by
dental technicians to the introduction of completely
new processes, at the present the indirect procedure
appears to be the more widespread.
2.1.2 Zirconia caps
The unique way to produce zirconia caps is by milling
from a blank of Zirconia. Several compact 5 axis
CNC milling machine tools for dental application are
available on the market. Typically a green block is
machined, that is a compact of zirconia powders subjected to a pre-sintering treatment to ensure handling
properties as well as good machinability. The milled
cap is then subjected to heat treatment at high temperature to complete the sinterization. The final part has
very high hardness and strength resistance. Zirconia
shrinkage after sintering reaches around 25% and must
be compensated at the design phase. To complete the
restoration fabrication, the cap is veneered with porcelain. This last operation is critical: in literature it is
reported that the main cause of zirconia restoration
failure is the veneer cracking (Denry & Kelly 2008).
From the 3D model of the cap to the physical part
the manufacturing process involves several steps
(toolpath programming, machining, sintering, and
finishing) and commonly requires a few days.
170
EVALUATION PROCEDURE
The specifically designed benchmark used in the evaluation procedure is illustrated in Figure 2. Classical
features are used to resemble real prepared teeth, both
anterior and posterior. The benchmark includes tooth
orientation and provides oblique surfaces similar to
those of a real prepared tooth (Atzeni et al. 2009).
The first step of the manufacturing sequence of
crown restorations consists in the achievement of the
virtual model from the stone replica. Errors in this
phase will highly affect next cap design and fabrication. Thus, the evaluation procedure (Fig. 3) begins
with the analysis of the digitization phase of the benchmark model previously produced. The inspection of
the physical benchmark with a coordinate measuring
machine (CMM) is required to evaluate the deviation contribute of the scanning operation. In fact, it
is not possible to compare the point-cloud directly
to the original 3D CAD model, because results will
include errors from the benchmark fabrication. In this
way, a Reference Model (ReM), consisting of actual
dimensions of the benchmark, is yielded for next comparisons. The point-cloud from digitization process
(Virtual Benchmark-VB) is inspected as well by using
a specific software to evaluate the same dimensions.
The comparison between inspections results and Reference Model allows to identify the accuracy of the
digitization step.
From the VB, caps are designed disregarding the
installation requirements (i.e. marginal and internal
gaps are set to zero, while actually gap is variable as
a function of the fabrication technique). As a consequence, caps may not fit the original studs, but the
use of a unique geometry makes possible to compare
different processes. Cap design is a merely CAD modelling operation and induced deviations are associated
to the software accuracy, typically on the order of
micrometres. These very small deviations are negligible in dental applications, thus this step is not analysed
in the evaluation procedure. Thus it is assumed that the
inner surface of the cap is the exact copy of the outer
surface of the tooth.
Consequently next comparison is made between
inspection results of the manufactured caps and measurements on the virtual benchmark. The inner surface
of each physical cap is inspected by CMM. Nowadays
contactless scanner could be used as an alternative to
CMM for inspection (Iuliano et al. 2010), but in this
application the concave small geometries of the caps
make difficult the three-dimensional optical measuring. Moreover, the accuracy of the CMM inspection
tool is higher. The comparison between inspection
results of the physical cap and the corresponding 3D
data provides information about the deviation induced
by the manufacturing process.
Finally, to evaluate the deviation of the whole
cap design and manufacturing process, a comparison is made between the measurements results of the
fabricated cap and the Reference Model.
4
Figure 2. Benchmark for dental application.
ZIRCONIA AND COBALT-CHROMIUM

CAPS MANUFACTURING
The evaluation procedure proposed in the previous

paragraph is applied to two different methods to
Figure 3. Evaluation procedure.
171
Figure 4. Physical benchmark produced by SLS (a) and Virtual benchmark scanned with DentalWings (b).
fabricate dental caps, which are establishing themselves. Especially, milled Zirconia caps and CoCr caps
produced by investment casting from DoD wax pattern
are analysed.
The physical copy of the benchmark (Fig. 4a) is
obtained from its CAD model by Selective Laser
Sintering (SLS) using the EOS FORMIGA P100, a
system suitable for dental application, recently introduced on the market. Dental models can be produced
from PA2200, a fine powder on the basis of Polyamide
12, which is biocompatible with ISO 10993-1. (International Organization for Standardization 2009) and
features chemical resistance along with good mechanical properties. Moreover, its white colour allows to
achieve excellent results in the reverse engineering
process and has high long-term constant behaviour.
The layer thickness of 0.1 mm guarantees good surface
quality.
The physical benchmark is then measured by CMM
to evaluate the actual shapes and dimensions of the
real part. A DEA CMM (model GLOBAL Image
07.07.07) is used for this operation, equipped with
an indexable swinging head and a touch trigger
probe (diameter 0.5 mm) whose resolution is 1 m.
The machine declared volumetric length measuring
uncertainty MPEE according to ISO-10360/2. (International Organization for Standardization 2009) is
1.5 + L/333 m, where MPE is the acronym for Maximum Permissible Error and L is the measured length.
Basically diameters and height of each tooth, conicity
and flatness are inspected (Fig. 5). Each measurement is repeated three times to have information about
dispersion of results. As result, the Reference Model
(ReM) is built.
4.1
Digitizing of the benchmark
The digitization of the physical benchmark, that is first

step of the cap fabrication procedure, is carried out
using Dental Wings 5 Series system. Dental Wings
is a laser scanner specifically developed for dental
applications. The contactless scanning device exploits
laser triangulation in a fix scanning working volume
of 140 140 95 mm3 . Up to 100,000 points are
measured per single dies scan (up to 1,000,000 points
Figure 5. Geometries of idealized prepared teeth (grayed

out areas refer to surfaces coupling the cap).
per full dental arch) and the declared accuracy on the

device data sheet is 20 m. Thanks to the synchronized 5-axes rotary table, multiple views are merged
into one point cloud automatically. After a preliminary
scanning to calibrate the laser intensity according to
sample surface characteristics, the acquisition carries
on with a second scan of the whole dental arch with
high precision and accuracy. Finally individual teeth
are acquired to have finer point clouds of each tooth.
At the end of the scanning process, all the point clouds
are joined into a single one containing the entire model
(Fig. 4b). It takes a bit more than five minutes to scan
the benchmark.
The STL model of the benchmark is imported into
Rapidform for inspecting digitized parts.This software
tool has in-depth inspection functionalities combined
with highly accurate algorithms certified and tested by
most important metrology laboratories (INUS Technology Inc. 2011, Minetola et al. in prep.). Especially,
the inspection module of RapidForm is used to measure the same features previously measured by CMM
for comparison. For diameter, a cylinder is created by
best fitting the corresponding regions points and its
diameter value is computed. Regarding conicity and
flatness, a high number of points on a conical or planar surface are selected so the software computes the
best fit cone or plane and then the geometrical tolerance by summing the maximum positive and negative
distances of points to the feature. Up to this step, the
procedure is independent of the cap manufacturing
technique.
172
Table 1.
Cap design parameters.
Angle
Cement gap
Collar position
Extra horizontal gap
Extra vertical gap
Margin thickness
Minimum thickness
4.2
Table 2. Reference Model and dimensional deviations of

the virtual benchmark.
65 deg
0 mm
1.5 mm
0 mm
0 mm
0.2 mm
0.6 mm
Entity
Caps design and manufacturing
The measured entities refer to the labels shown in

Figure 5. The results of the CMM inspection on
the physical benchmark, which define the Reference
Model, are listed in Tables 23, where the average
value and standard deviation (SD) for each dimension,
conicity and flatness are detailed. The small standard
deviations indicate that data are consistent. Geometrical tolerances are on the order of few hundredths of
a millimetre, attesting that measured features deviate
only very slightly from a perfect tapered cone or an
ideal plane despite of the step-like appearance of a
three dimensional additive part.
5.1
VB vs. ReM
Average (mm) SD (mm) Average (mm) SD (mm)
Afterward, caps are designed using a DentalWings

proprietary application that full integrates the scanning environment and the Crown & Bridge design
module, by setting parameters as listed in Table 1. The
designed caps are then exported into STL.
Subsequently, the STL files of the caps are used to
produce two series of wax (dentaCast) models by
DoD using the D76+ of Solidscape Inc. Wax patterns
are then sent to a dental laboratory to produce CoCr
caps by investment casting. Similarly, STL files are
sent to another prosthodontic laboratory for fabricating two series of zirconia (3Y-TZP) caps.The geometry
differs from the previous case for the presence of a
groove in the bottom of the cavity cap, required to
ensure the correct fit for mating parts (prepared tooth
and cap). Moreover the model is scaled (26.4%) to
compensate the shrinkage due to the heat-treatment.
The inner surfaces of the final caps, both of CoCr
and Zirconia, are inspected by CMM to measure
diameters, heights and conicity once more. Each measurement is repeated three times to evaluate dispersion
of results.
ReM
Digitization
Results of Rapidform inspection of the Virtual Benchmark are listed in Table 2, in terms of average error and
standard deviation (SD) with respect to the Reference
Model. The comparison shows that the point-cloud is
an accurate description of the real benchmark, with
errors ranging from minus 20 to around 50 m. The
Incisor
R1I
R2I
R3I
HI
Premolar
D1P
D2P
D3P
HP
Molar
D1M
D2M
Table 3.
0.96
1.41
2.40
8.83
0.03
0.01
0.02
0.05
0.05
0.02
0.01
0.01
<0.01
0.01
<0.01
<0.01
3.30
3.85
5.83
6.37
0.01
0.02
0.02
0.03
0.01
0.01
0.01
0.02
0.02
< 0.01
0.01
< 0.01
5.38
5.77
0.07
0.13
0.02
<0.01
0.06
0.08
Geometric tolerances of ReM and VB.
Entity
ReM
VB

Incisor
PL1I
PL2I
PL3I
CI
Premolar
CP
Molar
CM
0.02
0.03
0.04
0.01
0.01
0.02
0.02
0.01
0.06
0.08
0.07
0.08
<0.01
0.02
0.01
<0.01
0.02
0.01
0.07
<0.01
0.06
0.02
0.09
0.02
deviations range is consistent with the declared accuracy of DentalWings device (15 m). Considering the
deviation with sign (negative sign indicates that the virtual feature is smaller than the physical one), it could
be observed that the incisor and the molar teeth in
the VB appear bigger than in the ReM, while premolar is smaller. Bigger absolute deviations are found
on the dimensions of the incisor. In fact, compared to
other teeth, in this case diameters are extracted only
from a partial feature (half truncated-cone). Moreover,
because the density of the point cloud describing teeth
is uniform, less data are available for smaller surfaces.
The rough aspect of the benchmark surfaces, amplified
by the digitization (noise), leads on the virtual model
to standard deviations (SD) of the same order of magnitude of the error. For the same reason geometrical
tolerances get worse as well. The flatness of the lateral
faces of the incisor is 70 m on average, probably due
to the combined effects of the surface roughness and
laser reflection. It could be stated that the accuracy
of the digitization accomplished on the benchmark is
good and adequate for dental purpose.
5.2 Manufacturing
The measure of deviations induced by the cap manufacturing process gives information on the ability of
173
Table 4. Dimensional deviations of CoCr caps with respect

to the VB and the ReM.
Table 5.
Entity
Entity
CoCr cap vs. VB
CoCr cap vs. ReM
Geometric tolerances of CoCr and Zirconia caps.

CoCr caps
Zirconia caps

Incisor
R1I
R2I
hI
Premolar
D1P
D2P
hP
Molar
D1M
D2M
hM
n/a
0.01
0.14
n/a
<0.01
0.10
n/a
0.04
n/a
n/a
<0.01
n/a
0.04
0.06
0.16
<0.01
0.01
0.01
0.05
0.05
n/a
<0.01
0.01
n/a
0.11
0.15
0.17
0.06
0.01
0.01
0.13
0.14
n/a
0.06
0.01
n/a
Incisor
PL1I
PL2I
PL3I
CI
Premolar
CP
Molar
CM
0.05
0.14
0.04
0.14
0.02
0.10
0.01
0.01
0.08
0.04
0.01
0.20
0.06
0.02
<0.01
0.06
0.26
0.12
0.18
0.08
0.09
0.01
0.16
0.06
Table 6. Dimensional deviations of Zirconia caps with

respect to the VB and the ReM.
Entity
the process to reproduce the designed geometry. On

the other hand, deviations with respect to the Reference
Model are an indication of interference or clearance fit
between the prepared tooth and the corresponding cap,
as a result of several production steps. These experimental evidences are of paramount importance for the
definition of the best design parameters as a function
of the specific process (such as the gaps values).
5.2.1 Cobalt-chromium caps
Results of the inspection of the inner surfaces of the
CoCr caps are listed in Table 4 in terms of deviations
with respect to the Virtual Benchmark and the Reference Model. The entities in Table 4 differ slightly
from the set presented in previous tables, since only
the coupling features are analysed (grayed out surfaces
in Figure 5). Especially h is the depth of the cap cavity, measured along its axis. This value is compared
with the height h of the corresponding prepared tooth
(see Figure 5), measured on the virtual benchmark.
Because the same dimension cannot be inspected by
CMM, the analogous comparison with the ReM is not
possible.
The absolute deviation of CrCo caps with respect
to the Virtual Benchmark ranges from 0.01 mm to
0.17 mm. Higher errors are associated to bigger dimension (heights and molar diameters), maintaining the
percentage error at values of 23%. The bottom of the
cavity of the incisor cap showed an irregular surface
and the measure of the radius R1I was not possible.
This anomaly also affected the measurement of the cap
cavity depth, resulting in a larger standard deviation.
From the outcomes, it can be assessed that the accuracy
of metal caps produced by investment casting from
DoD wax patterns is adequate for crown restoration.
It could be observed that all errors are positive
in sign, meaning that produced caps are larger than
the corresponding three-dimensional data. However,
deviations introduced by the manufacturing process in
some cases are comparable with the deviations induced
by the digitizing step. Because the virtual benchmark
could be smaller than the ReM, it is not guaranteed that
CoCr cap vs. VB
CoCr cap vs. ReM

Incisor
R1I
R2I
hI
Premolar
D1P
D2P
hP
Molar
D1M
D2M
hM
0.17
0.03
0.33
0.05
0.03
0.06
0.12
0.05
n/a
0.05
0.03
n/a
0.10
0.05
0.21
0.01
0.01
0.05
0.10
0.04
n/a
0.01
0.01
n/a
0.25
0.18
0.14
0.02
<0.01
0.04
0.27
0.17
n/a
0.02
<0.01
n/a
the final cap will be larger than the prepared teeth. Nevertheless, by comparing CoCr dimensions and ReM, in
this study a clearance fit is observed. This is an interesting result that could allow the installation of the cap
even if the cap design did not take into consideration
the gap required for cement. As regards geometrical
tolerances (Tab. 5), the flatness is good and comparable with the one of the VB, while the conicity get
worse, meaning that the conical part differs from a
perfect cone.
5.2.2 Zirconia caps
Results obtained for zirconia milled caps show bigger
deviations then CoCr caps. The absolute deviation of
Zirconia caps with respect to the Virtual Benchmark
ranges from 0.03 mm to 0.33 resulting in a percentage
error of around 45%. High deviations are observed
for entities belonging to the bottom of the caps cavity, as well as for heights. In fact heights are measured
from the top to the bottom of the cap cavity and errors
in the measurement of the bottom plane highly affect
the evaluation of the cavity depth. Analogously, the
conicity errors (Tab. 5) appear amplified by the manufacturing process.These behaviours may be ascribed to
the milling process and especially to the tool geometry.
Really, a bottom groove was designed to counterbalance the ball end mill radius (grove depth is equal to
174
the tool radius), in order to obtain the cylindrical feature. As a consequence, the area for the inspection of
the bottom plane is narrowed and small irregularities
could highly affect the measurement results. Another
source of error could be attributed to the heat treatment
that causes a large shrinkage.
The deviations related to the marginal entities (R2I ,
D2P , D2M ) are smaller and comparable with the deviations on same entities induced by the investment
casting process. Because the percentage error is almost
constant, the bigger the dimension, the higher the deviation. It should be observed that the marginal fit is
the most important factor to measure the restoration
fit quality and the clinically acceptable range is about
100 50 m (Fransson et al. 1985, Karlsson 1993).
Internal gap is wider and has a larger tolerance.
CONCLUSIONS
An evaluation procedure was defined and applied to

assess errors introduced by each step of two dental restoration production methods based on time
compression technologies.
Results of this investigation show that the pointcloud (Virtual Benchmark) is an accurate description
of the real benchmark, with errors ranging from minus
20 to 50 m. Thus the RE process is an effective way
to obtain the stone replica representation for crown
restoration design. It is also expected that improvements in intraoral scanners will lead to the direct digital
impression, giving up the traditional impression taking
and replica production procedure (Syrek et al. 2010,
Ender & Mehl 2011). As regards manufacturing processes, the investment casting from DoD patterns is
proven to be able to reproduce the designed geometry
with deviations on average 100 m, while the milling
of Zirconia has similar characteristics but the bottom
of the cap cavity presents higher errors. Both processes
produce caps with clinically acceptable values of the
marginal fit (on average around 80 m).
Further investigations are needed to distinguish the
effects of each manufacturing step of Zirconia caps,
in order to assess if improvement in the milling path
could have positive repercussion on the final result or
to optimize the scale factor to properly compensate the
shrinkage due to the heat treatment.
ACKNOWLEDGEMENTS
The authors are grateful to Mr Leonardo Gomez Orea
for his collaboration during preliminary inspections,
and to Dr. Paolo Minetola, whose advice and help
on RE inspection was invaluable. Thanks also to Dr.
Vito Chinellato (EOS S.r.l., Milano, Italy), Dr. Alberto
Lazzaro (WisilDent S.r.l., Torino, Italy), Dr. Sergio
Pieri (CMF Marelli S.r.l., Cinisello Balsamo MI,
Italy), and Mr. Giuseppe Chiauzzi (Dentalabor S.r.l.,
Torino, Italy) for supporting the experimental part of
the research.
REFERENCES
Atzeni, E., Gatto, A., Iuliano, L., Minetola, P. & Salmi,
A. 2009. A benchmark for accuracy evaluation of dental crowns up-to-date manufacturing. In Bartolo, P.J.,
et al. (eds.), Innovative Developments in Design and
Manufacturing: 425430. London: CRC Press.
Bindl, A. & Mormann, W. H. 2005. Marginal and internal fit of all-ceramic CAD/CAM crown-copings on
chamfer preparations. Journal of Oral Rehabilitation 32:
441447.
Brosky, M. E., Pesun, I. J., Lowder, P. D., DeLong, R. &
Hodges, J.S. 2002. Laser digitization of casts to determine
the effect of tray selection and cast formation technique on
accuracy. Journal of Prosthetic Dentistry 87(2): 204209.
Chen, S. Y., Liang, W. M. & Chen, F.N. 2004. Factors affecting the accuracy of elastometric impression materials.
Journal of Dentistry 32(8): 6039.
Christensen, G. J. 2009. Porcelain-Fused-to-Metal Versus
Zirconia-Based Ceramic Restorations, 2009. The Journal
of the American Dental Association 140(8): 10361039.
DeLong, R., Heinzen, M., Hodges, J. S., Ko, C. C. & Douglas,
W.H. 2003. Accuracy of a system for creating 3D computer models of dental arches. Journal of Dental Research
82(6): 438442.
Denry, I. & Kelly, J. R. 2008. State of the art of zirconia for
dental applications. Dental materials: official publication
of the Academy of Dental Materials 24(3): 299307.
Ender, A. & Mehl, A. 2011. Full arch scans: Conventional
versus digital impressions-An in-vitro study. Ganzkieferaufnahmen: Konventionelle versus digitaleAbformtechnikEine In-vitro-Untersuchung 14(1): 1121.
Filser, F., Kocher, P., Weibel, F., Luthy, H., Scharer, P. &
Gauckler, L. J. 2001. Reliability and strength of allceramic dental restorations fabricated by direct ceramic
machining (DCM). Int J Comput Dent 4(2): 89106.
Fransson, B., Oilo, G. & Gjeitanger, R. 1985. The fit of metalceramic crowns, a clinical study. Dent Mater 1(5): 1979.
International Organization for Standardization. 2009.
Biological evaluation of medical devices In Part 1:
Evaluation and testing within a risk management process.
International Organization for Standardization. 2009. Geometrical product specifications (GPS) Acceptance and
reverification tests for coordinate measuring machines
(CMM) In Part 2: CMMs used for measuring linear
dimensions.
INUS Technology Inc. 2011. Rapidform XOV/Verifier:
Overview. INUS Technology Inc. 2011 [cited 23
May 2011]. Available from http://www.rapidform.
com/portal/default/Products/index?Category=Products_
RapidformXOVVerifier_XOVSub1.
Iuliano, L., Minetola, P. & Salmi, A. 2010. Proposal of
an innovative benchmark for comparison of the performance of contactless digitizers. Measurement Science and
Technology 21: 105102105114.
Karlsson, S. 1993. The fit of Procera titanium crowns: An in
vitro and clinical study. Acta Odontologica Scandinavica
51(3): 129134.
Koutayas, S. O., Vagkopoulou, T., Pelekanos, S., Koidis, P. &
Strub, J. R. 2009. Zirconia in dentistry: part 2. Evidencebased clinical breakthrough. Eur J Esthet Dent 4: 34880.
Kuhmstedt, P., Brauer-Burchardt, C., Munkelt, C., Heinze,
M., Palme, M., Schmidt, I., Hintersehr, J. & Notni,
G. 2007. Intraoral 3D scanner. In Huang, P.S. (ed.),
Two-and Three-Dimensional Methods for Inspection and
Metrology: 67620E-9. Boston, MA, USA: SPIE.
Laurent, M., Scheer, P., Dejou, J. & Laborde, G. 2008. Clinical
evaluation of the marginal fit of cast crowns validation of
175
the silicone replica method. Journal of Oral Rehabilitation

35(2): 11622.
Luthardt, R. G., Holzhter, M. S., Rudolph, H., Herold, V.
& Walter, M.H. 2004. CAD/CAM-machining effects on
Y-TZP zirconia. Dental Materials 20(7): 655662.
Luthardt, R. G., Loos, R. & Quaas, S. 2005. Accuracy of
intraoral data acquisition in comparison to the conventional impression. International journal of computerized
dentistry 8(4): 28394.
Minetola, P., Calignano, F. & Iuliano, L. in prep. A customer
oriented methodology for Reverse Engineering software
tools selection in the computer aided inspection scenario.
Miyazaki, T., Hotta, Y., Kunii, J., Kuriyama, S. & Tamaki, Y.
2009. A review of dental CAD/CAM: current status and
future perspectives from 20 years of experience. Dental
Materials Journal 28(1): 4456.
Persson, A. S. K., Andersson, M., Odn, A. & SandborghEnglund, G. 2008. Computer aided analysis of digitized
dental stone replicas by dental CAD/CAM technology.
Dental Materials 24(8): 11231130.
Quaas, S., Rudolph, H. & Luthardt, R. G. 2007. Direct
mechanical data acquisition of dental impressions for
the manufacturing of CAD/CAM restorations. Journal of
Dentistry 35(12): 903908.
Rekow, E. D. 2006. Dental CAD/CAM systems: a 20-year
success story. Journal of the American Dental Association
137 Suppl: 5S6S.
Syrek, A., Reich, G., Ranftl, D., Klein, C., Cerny, B. &
Brodesser, J. 2010. Clinical evaluation of all-ceramic
crowns fabricated from intraoral digital impressions based
on the principle of active wavefront sampling. Journal of
Dentistry 38(7): 553559.
Thongthammachat, S., Moore, B. K., Barco, M. T., Hovijitra,
S., Brown, D. T. & Andres, C. J. 2002. Dimensional accuracy of dental casts: Influence of tray material, impression material, and time. Journal of Prosthodontics 11(2):
98108.
Tsitrou, E. A., Northeast, S. E. & van Noort, R. 2007. Evaluation of the marginal fit of three margin designs of resin
composite crowns using CAD/CAM. Journal of Dentistry
35(1): 6873.
Ucar, Y., Akova, T., Akyil, M. S. & Brantley, W.A. 2009.
Internal fit evaluation of crowns prepared using a
new dental crown fabrication technique: Laser-sintered
Co-Cr crowns. The Journal of Prosthetic Dentistry 102:
253259.
Vagkopoulou, T., Koutayas, S. O., Koidis, P. & Strub, J. R.
2009. Zirconia in dentistry: part 1. Discovering the nature
of an upcoming bioceramic. The European Journal of
Esthetic Dentistry 4 (2): 13051.
Zarone, F., Russo, S. & Sorrentino, R. 2011. From
porcelain-fused-to-metal to zirconia: Clinical and
experimental considerations. Dental Materials 27(1):
8396.
176

Reverse engineering of casting equipment for process simulation

F. Calignano, P. Minetola, A. Salmi, E. Atzeni & L. Iuliano
ABSTRACT: In this work a procedure to obtain the three-dimensional CAD model of a casting from the
physical equipment is proposed. In fact, the actual geometry of the cast part cannot be used for process simulation
because of the shrinkage and deformation occurring during the casting process. The procedure is based on
reverse engineering tools and involves four steps: pre-digitizing, digitizing, surfaces reconstruction, and threedimensional CAD modelling. Critical issues are identified and discussed in this paper. To verify this procedure,
it is applied to an aeronautical part fabricated by sand casting. Results prove that this RE approach is adequate
to virtually reconstruct the single geometry of each equipment part and of the whole assembled mould. The
three-dimensional CAD model is demonstrated to be suitable for CAE analysis, leading to a finite element mesh
that meets requirements of casting simulation. Furthermore it is easily editable to adapt the design according to
numerical outputs.
INTRODUCTION
In last decades, casting process simulation through

Computer Aided Engineering (CAE) codes has been
adopted by few foundry, initially for the analysis of
the solidification phase, in order to optimize risers
location (Shehata & Abd-Elhamid 2003), and then
extending its use to predict defects, deformation, residual stresses, and microstructure (Kermanpur et al.
2008, Sun et al. 2008, Bichler & Ravindran 2010).
On the market, several simulation software packages
are available for foundry applications, based on finite
elements or finite volume/difference methods. In both
cases the three-dimensional CAD models of the casting parts are required, because meshes are generated
on them. However, in foundry a critical issue is often
the unavailability of the mathematical model of the
whole cast assembly.This is the case of old casts, where
the original bi-dimensional drawings define only partially the part geometry (i.e. coupling surfaces). In
other cases, a pre-existing product is redesigned due
to a revision of the process (mould geometry should
be modified or the gating system redesigned) and
CAD models are not updated. In these circumstances
three dimensional CAD models can be reconstructed
starting from available equipments (cores, core boxes,
patterns, moulds, etc.) using Reverse Engineering
(RE) techniques (Ferreira & Alves 2003, Sokovic &
Kopac 2006).
Reverse Engineering is a widespread technology
that enables us to generate a virtual model of a physical part (Durupt et al. 2008, Iuliano & Minetola
2009). There are several application areas of RE,
including metrology, quality control, design, virtual
reality, medical and surgical applications, and preservation of Cultural Heritage. RE process begins with the
digitization of an object, which is the acquisition of

points (point cloud) belonging to the part surfaces.
Information obtained by the scanning device must first
be optimized through appropriate filters, and then the
point cloud can be converted into standard formats
or imported directly into the software used for surfaces reconstruction. Then a set of surfaces, fitting
the point clouds automatically or semi-automatically,
creates the CAD model of the part (Vrady et al. 1997,
Chant et al. 1998, Tai & Huang 2000).
Up to now, RE techniques have been applied in
foundry within a limited range of applications. For
example RE is used for inspection purpose to retrieve
the actual geometries of tools or cast parts. In fact,
through periodical scans, tool wear can be monitored
and worn tools can be replaced by copies of the original ones. Furthermore, as-cast parts are digitized to
inspect the actual geometry for evaluating if quality requirements are met in terms of dimensional and
geometrical tolerances. Another specific application,
reported in literature, is the generation by digitizing
process of three-dimensional CAD surfaces for rapid
tooling (Ferreira & Alves 2003). As regards numerical simulation, the application of RE techniques to
create the geometrical model of the casting was studied by several authors (Wu et al. 2001, Ferreira &
Alves 2003, Sokovic & Kopac 2006, Suresh Babu &
Thumbanga 2011), but the approach is typically the
generation of the Finite Element (FE) model directly
from scan data, bypassing the surface generation step:
by this way the geometry is quickly obtained, but it
cannot be easily edited. In addition to this, the availability of an editable three-dimensional CAD model
is a mandatory requirement of the redesign process.
In fact, because simulation is intended to help engineers during the design phase, the three-dimensional
177
CAD model should be easily updated according to

simulation outputs. This last topic has not been studied in depth yet in scientific papers, but it is of
paramount importance in a concurrent engineering
scenario.
Focusing on these last aspects, a procedure to obtain
the mathematical CAD model and the FE tetrahedral
mesh of a casting from physical equipments is proposed in this paper. The procedure involves the use of
a three-dimensional scanner based on structured light
(ATOS Standard) to digitize mould parts. Point clouds
are aligned into a global reference system and then processed to obtain an editable three-dimensional model
of the assembled mould. If needed, the reconstructed
model is integrated by importing standard geometries
in a 3D CAD environment. This novel procedure is
then applied to an aerospace component, selected as
case study, to let benefits and limitations coming out.
2
REVERSE ENGINEERING PROCEDURE
The RE procedure to obtain the CAD model from the

actual casting is detailed in the following four sections:
pre-digitizing, digitizing, surfaces reconstruction, and
three-dimensional CAD modeling.
2.1
Pre-digitizing
A sand casting mould is an assembly of different parts:

cope and drag mould halves and cores (defining the
cavity geometry including gates and risers), filters,
chillers, and insulation sleeves. Some parts are standard components, such as filters and sleeves, and their
three-dimensional CAD models are generally supplied
by manufacturers. On the contrary, other parts are
developed for the specific cast product. If the threedimensional CAD models of these expressly designed
parts are not available, RE is the best practice to be
adopted for digitizing them, because of the ability of
this technique to accurately copy complex geometries,
such as free-form surfaces that are common in foundry.
However the most challenging issue of this approach
is the correct alignment of each digitized part according to its actual location in the assembled mould. This
problem arises because the scan data of each part has
its own reference system. Therefore to overcome this
problem the procedure illustrated in Figure 1 is proposed. As a matter of fact the aim of a preliminary
digitization is to acquire a limited number of views of
each part of the mould when all parts are in their location in the assembly. Pre-digitization data is then used
to refer all the point clouds according to a unique reference system (hereafter referred to global reference
system). This procedure, named pre-digitizing, starts
with the mould fully assembled. At this initial stage
the counter i is set equal to the number N of parts
that make up the assembled mould. After defining initial conditions, the operator evaluates if the condition
i = 1 is true. If it is not true, a pre-digitization of visible parts of the assembled mould is performed. During
Figure 1. Pre-digitizing procedure.
pre-digitization, the number of views and the points

density is much lower than the ones of a complete
digitization: scan data, in fact, is only used for parts
alignment and not for surfaces generation. Afterwards
the operator identifies the number j of parts with sufficient scan data to be aligned and removes all these parts
except one: such part will be the reference element in
the next pre-digitization step. The counter i is then
updated according to the equation i = i (j 1) that
is the number of remaining parts to be pre-digitized.
The execution of pre-digitization steps continues until
i = 1. When this condition is true all the parts of the
assembled mould are pre-digitized and referenced in
the global reference system. Now we are ready to go
to next step: the digitization phase.
178
2.2
Digitizing
In the digitization phase, each part of the equipment

whose CAD model is unavailable (i.e. pattern plates,
Figure 2. Digitizing procedure for each part of equipment.
core boxes and chillers) is singularly digitized. As

regards equipment parts for sand forming, the scanning is limited to the active surfaces of the equipment,
that is the ones defining the shape of the sand parts.The
procedure followed in this phase is shown in Figure 2.
The surfaces of equipment parts can consist of convex or concave shapes. It is well known that convex
shapes can be easily digitized, while concave ones
(especially small deep holes) could complicate the
scanning process or even prevent it. To overcome these
problems, where possible, the missing information on
the scan data can be obtained by the digitization of
reverse shapes.Thus, positive parts, such as sand cores,
cope and drag moulds, are digitized to complete the
geometries of core boxes and pattern plates.Afterward,
the complete digital model of each part can be generated by aligning all the partial scans (from equipment
and corresponding sand part) with respect to the same
reference system and merging them together. If the
geometry of the part cannot be completed by the digitization of the reverse shapes, a partial point cloud
is obtained. Therefore the digitization output for each
part is a complete or partial point cloud. In this phase
M parts are digitized. M is greater or at least equal
to N L (where L is the numbers of parts whose 3D
CAD model is available) because some elements (e.g.
chillers and insulation sleeves) are partially included
in the pre-digitized sand parts of the assembled mould,
while they are singularly digitized in this phase. If a
partial point cloud is available for K parts, the geometry of the remaining (M K) parts is completely
defined by digitization data.
2.3 Surface reconstruction
To construct editable CAD surfaces from point data
(the output of the previous phase), the (M K) complete point clouds and K partial point clouds should be
first aligned in the global reference system by means
of the pre-digitization data (Fig. 3). Once aligned, single point clouds can be pre-processed for denoising,
smoothing and density reduction. These software operations are needed to ease the surfaces reconstruction
process since they reduce the complexity of the point
cloud (Yu et al. 2011). However, particular attention
should be paid in applying such operations not to alter
excessively the accuracy of the scan data. The point
clouds are then converted into polygonal models by
generating an unstructured triangular mesh.
Considering the polygonal models of the parts in
their assembled state, small gaps could be detected
among adjacent parts. Some gaps are due to the clearance provided in the design stage to ease the assembly
operation and to avoid mould distortion due to thermal
stresses during filling and solidification. These kind
of gap can be identified because the distance between
adjacent parts is bigger than half a millimeter. Other
smaller gaps are the result of the deviation with respect
179
Figure 3. Surface reconstruction and three-dimensional CAD modelling procedure.
to nominal dimension that is induced by the manufacturing process: in fact the scanning process accurately
copies actual geometries (Iuliano et al. 2010). However, all these gaps, both provided in the design stage
and due to the manufacturing operation, must be
removed in the three-dimensional model for numerical
simulation (simplification of the model), because they
complicate the model without significantly affecting
simulation outcomes. Moreover, in foundry practice
visible cavity gaps are removed during the manual
assembling operation by using a filler. Thus, in the
virtual environment, the triangulated model of two or
more adjacent parts is combined to create a single
entity. The small gap (Fig. 4a) is reduced by building one or more narrow bridges. In such a way, the gap
is divided into separately fillable holes (Fig. 4b).
The gap removal operation causes the loss of the
interface. However the identification of the interface information (geometry, distance between adjacent
parts and position) is important to allow, in the next
CAD modelling phase, the splitting of the mould
geometry again into single parts. For instance the
presence of core prints is lost when the mating surfaces of the core and the sand mould are merged by
removing the gap between them. Nevertheless, core
prints modelling is required for process simulation.
In order to re-introduce the core prints, it is necessary to identify and extract some cutting features for
splitting the mould model. Cutting feature recognition
is performed by computing the best fit feature from
gap interfaces (Fig. 4) that are usually classic geometries (planes, cylinders, cones, etc.). In the case of a
Figure 4. Gap removal and cutting feature identification.
cylindrical core print the cutting feature will be represented by the best fit cylinder computed from the
corresponding cylindrical surfaces of the core (inner
cylinder) and the mould (outer cylinder). While gap
interfaces between sand parts are normally planar, so
cutting planes will be identified.
Another requirement is the absence of overlapping
surfaces. However there are M (N L) elements
that are partially included in the sand parts and thus
overlapping polygonal faces are detected on them. For
instance, in the forming stage a chiller is placed on
the surface of the pattern plate; the digitization, alignment and assembly phases will reproduce this mating.
Thus the polygonal faces of the chiller are overlapping
the ones of the pattern plate. To solve this ambiguity,
the more regular faces are kept and the others are discarded. If a gap results from this operation, it is closed
by the gap removal procedure previously described.
At this moment, the cavity of the mould is completely described by an unstructured triangular mesh
and surfaces can be defined on it. In the step of surface generation, primitive and/or free-form surfaces
180
are defined on the polygonal mesh by using a RE tool.

Generally casting parts can be described either by classical shapes connected by highly curved features (it is
the case of the gating system geometry) or by free-form
surfaces (usually defining the mould cavity). Algorithms developed in recent years for the reconstruction
of the CAD model from a polygon mesh are based
on a segmentation process (Edelsbrunner et al. 2000,
Edelsbrunner et al. 2001, Vrady et al. 2007). It is the
process of partitioning the polygonal mesh into several
planar, concave or convex regions, according to curvature changes, thus simplifying the surface generation.
There are several types of elaboration modes (region
detection, curvature detection, and automatic), as a
function of the desired reconstruction accuracy and of
the computational time. After segmentation, two different methods are selected and applied to reconstruct
the parts surfaces: autosurfacing for free-form generation and feature recognition for classical shapes. As
a matter of fact, autosurfacing procedure accurately
models the complex actual geometry of the mould cavity, whereas by feature recognition an approximated
representation is created for accessory elements (such
as runners, risers, etc.) that are generally defined by
elementary surfaces.
2.4 Three-dimensional CAD modelling
All the reconstructed surfaces and the cutting features
(identified during the previously detailed interface recognition phase) are imported into a threedimensional CAD environment. After this task, a
validity check on all the imported entities has to be performed in order to detect any corruption (e.g. geometry
faults, trim loop faults, unfixed blend faults, selfintersection faults, generic geometry-topology faults,
etc.). As a matter of fact, the presence of a corrupted body may result in unexpected outcomes when
a modelling operation is performed on it.
After the validation, the RE reconstructed geometry
is integrated by available 3D CAD models of standard
components (such as those of filters and insulation
sleeves) to complete the mould assembly. In order to
mate these 3D CAD models to the RE reconstructed
geometry, an edge-to-edge or face-to-face alignment,
or a combination of both the methods is performed. It
is important to point out that the alignment is possible
since mating surfaces of the RE reconstructed model
are classical geometries. This is a consequence of
the segmentation strategy adopted during the surface
generation phase.
After the alignment operation, special attention
should be paid again to the overlapping (mating) surfaces and intersections. Mainly intersections could
be the consequence of deviations between analogous
CAD entities of the 3D CAD model and the RE geometry. Especially, the greater the digitization error, the
greater the average deviation between the two geometries. Overlapping surfaces must be merged just into
one and, if needed, new edges are defined. Intersections are solved by maintaining one geometry and
extending/trimming the other one.
Once the simplification is completed, it is necessary to verify for each part if surfaces which describe
its geometry could become a solid entity. If this operation fails, it means that the solid is not closed: there is
at least one opening in the set of surfaces. The opening
often arises from the RE surface generation on partial
point clouds. For closing solid purpose, new surfaces
from neighbouring entities respecting the local curvature have to be modelled. This operation is a potential
source of errors because the generation of new surfaces
is performed without fitting any scan data. Anyway,
the closure of the opening is compulsory to obtain a
closed solid.
After that operation, the resulting geometry is an
editable model of the assembled mould (including
risers and gating system), with chillers, filter and insulation sleeves. The procedure continues with the CAD
modelling of core prints based on cutting features data.
Firstly the core print is modelled by editing the surfaces extracted during the interface recognition phase.
The resulting set of surfaces is used to cut the solid
mould.
From the point of view of numerical simulation, the
CAD of the assembled mould is fulfilled and it can
be exported to FEM environment. On the other hand,
in order to achieve the CAD model of the assembled
mould for cast redesign, the splitting of the cope and
drag mould halves and of the cores should be done.
Again, this task is simple to perform: the CAD model
of the assembled mould for numerical simulation has
just to be cut/split with the cutting features identified at the beginning of the surface reconstruction
phase. By this way each sand part that makes up the
assembled mould will be recognized as a single solid
entity.
CASE STUDY
To verify its feasibility, the procedure is applied to

an existing casting equipment. The assembled mould
consists of cope and drag parts, seven sand cores,
thirteen chillers, five insulation sleeves and one filter. The insulation sleeves and the filter are standard
elements and the associated 3D CAD models are provided by manufacturers, whereas the other parts have
to be reconstructed. In this work, an ATOS Standard
optical scanner is used for digitization. The scanning
accuracy declared by the device data sheet is 0.06 mm.
Each scan takes about 8 seconds and retrieves as many
as 400,000 points on the object surface. This allows
accurate measurements and capture of the shape and
size of the visible surfaces of almost any 3D object.
Following the previously described RE procedure, the
assembled mould is first subjected to pre-digitizing to
locate all parts of the assembly into a unique global reference system (Figure 5). To obtain sufficient data for
parts alignment four loops of the flow chart of Figure 1
are carried out. The time required for this first phase
is around one hour and a half, as shown in Table 1.
Then, pattern plates, core boxes, and chillers are
singularly subjected to scanning (Figure 6). Because
181
Figure 5. Pre-digitizing of the assembled mould.
Figure 6. Digitizing of equipment parts.
Table 1.
of the cavity (Figure 7). As a matter of fact, flash can be

observed in actual cast parts as shown in Figure 7a, but
the CAD model required for cast re-design and process simulation must be filled to obtain a closed and
continuous volume. With the aim of having a reference to reconstruct sand interfaces, the corresponding
cutting features are extracted and temporarily stored.
After interface removal, the remaining polygonal faces
describe the cavity of the casting, the volume enclosing
the mould, and chillers. Overlapping faces are detected
on the cavity surface next to chillers and are solved
by maintaining the polygons belonging to the pattern
plates or core boxes.
Then, surfaces are generated on the polygon mesh.
Chillers, risers and gating systems, that can be reduced
to sets of primitives, are modelled as classical CAD
features (feature recognition). On the contrary, complex surfaces defining the final cast shape can be
only modelled by NURBS (autosurfacing). In this
operation, special attention is given to the definition
of a regular patch structure, because high curvature
changes could produce auto-intersecting surfaces and
will prevent mesh generation in the next simulation
phases.Thus, this operation, that requires a high skilled
operator, is the most time-consuming step in the RE
process and requires about two weeks for the case
study.
Finally, all the surfaces are imported into VISI 18
(Vero Software Limited) CAD environment for the
assembling with standard geometries, such as filter
and insulation sleeves. Again, filter and sleeves are
overlapping cavity surfaces. A simplification is made
by deleting reconstructed surfaces while maintaining
the CAD ones. Whereas, each sleeve, that is included
in or exceeding the volume enclosing the mould is
extended/trimmed by the mould surface.
Procedure phases tools and duration.
Phase
Hardware/Software
Duration
Pre-digitizing
Digitizing
Surface reconstruction
3D CAD modelling
Simulation
ATOS Standard
ATOS Standard
Geomagic Studio 11
VISI 18
ProCAST 2010.0
1.5 hours
6 days
2 weeks
6 hours
2 days
of the complexity of their geometries, some areas of

pattern plates and core boxes cannot be scanned and
thus, in order to complete the partial point cloud, corresponding sand parts are digitized as well. The scanning
of reverse shapes is limited to the lacking areas on the
scan data of plates and boxes. To evaluate the induced
error, a comparison between corresponding portions
of scan data belonging to the pattern plate and the
sand mould is performed, resulting in an average deviation of 0.05 mm. The error is lower than the accuracy
required for sand casting, thus point clouds of original and reverse shapes can be used without distinction
in this application. About six days were required to
complete this second phase of the RE procedure. Complete point clouds were obtained for each part of the
equipment subjected to digitizing.
The next phase is the point cloud elaboration. In
this study, Geomagic Studio 11 software has been
used for the CAD model construction. Firstly, predigitizing data are used to align each point cloud to the
global reference system through a best-fit algorithm
limited to overlapping portions of scan data. Then,
point clouds are pre-processed and polygonized. At
this stage, the attention is shifted to the cavity shape.
Gaps are detected on the polygonized representation
182
Figure 7. Physical casting (a), mould partial cavity without applying gap removal (b), and gap removal example (c).
Figure 8. Three-dimensional CAD model of the assembled mould (a) and a detail of internal chillers (b).
After such operation, the geometry of the assembled

mould is completely defined (Fig. 8). In order to have a
model adequate for casting simulation, core prints are
modelled via CAD on the basis of the interface features
previously extracted. In fact, the corresponding interface must be modelled to obtain accurate simulation
results. The splitting of the cope and drag mould halves
is also accomplished, whereas subdivisions inside each
mould half are less important from a simulation perspective, and thus they can be made subsequently to
have the complete CAD model of the assembled mould
for cast redesign. Three-dimensional CAD modelling
phase takes about six hours.
The 3D CAD model of the assembled mould is
imported into ProCAST, a Finite Element software
suite for casting simulation by ESI Group. The link
between the CAD model and the mesh is the 3D tetrahedral mesh generator, which includes a CAD analysis
module. A multiple density mesh (Fig. 9 and Tab. 2) is
created to improve the simulation in terms of running
time and accuracy of results. The analysis of the mesh
(Fig. 10) shows an aspect ratio below 32 (about 90%
below 4), a dihedral angle ranging from 71 to 172 deg,
and radii ratio of 70% of elements greater than 0.6
proving that the obtained tetrahedral mesh has an high

quality.
Simulation is then performed by setting A357 material and a filling condition obtained from real pouring
data (variable metal flow rate). Common values of heat
transfer coefficients for sand casting process of aluminium alloys are used, taken from ProCAST library.
Filling, solidification and stress analyses are finally
executed and completed without any convergence
error.
CONCLUSIONS
It is well known that in foundry the availability of the

3D CAD model of a product and of the equipment to
fabricate it is often lacking or not constantly updated
to reflect design changes. However the mathematical
CAD model is the starting point to simulate the casting process for pre-existing parts for optimization or
redesign purposes. To overcome this problem, in this
work an RE-based procedure to obtain the geometry of a casting from the related physical equipment
183
Figure 9. Mesh with multiple density of the assembled mould geometry.
is proposed. As a matter of fact, the cast part cannot be used directly for the creation of the virtual
model for CAE analysis because of the shrinkage and
deformation occurring during solidification, cooling,
demoulding, and gating system removal. The proposed
procedure involves four phases: pre-digitizing, digitizing, surface reconstruction, and three-dimensional
CAD modelling. Each phase is detailed in this paper
and critical aspects are discussed. To verify the proposed procedure, it is applied to an aeronautical part
fabricated by sand casting. Results prove that this RE
approach is adequate to virtually reconstruct the single
geometry of each equipment part and of the assembled
Table 2.
Superficial mesh parameters.
Entity
# Edges
Length
(mm)
Cast (main component)

Gating system
Filter
Chillers
Cores
Insulator sleeves
Parting plane
Mould
10121
2748
16
1511
19
85
9
30
2.5
4.0
4.0
4.5
4.5
6.0
8.0
40.0
184
Bonavia during the digitizing operations and modelling in this study. Moreover the authors extend their
thanks to Dr. Alessandro Ghio and Dr. Giuseppe
Argentieri (Avio SpA, Torino, Italy) and to Dr. Lorenzo
Valente and Dr. Cristian Viscardi (ECOTRE sas,
Brescia, Italy) for their help and support in the
research. Financial support to this work provided by
the Piemonte Regional project GReen Engine for
Air Transportation 2020 (GREAT 2020) is gratefully
acknowledged.
REFERENCES
Figure 10. Mesh quality plot.
mould. The 3D CAD model is demonstrated to be suitable for CAE analysis and it is easily editable to adapt
the design according to numerical outputs.
ACKNOWLEDGEMENTS
The authors would like to acknowledge the assistance of Mr Giovanni Marchiandi and Mr Fabrizio
Bichler, L. & Ravindran, C. 2010. New developments in

assessing hot tearing in magnesium alloy castings. Materials & Design 31 (Supplement 1): S17S23.
Chant, A., Wilcock, D. & Costello, D. 1998. The determination of IC engine inlet port geometries by reverse
engineering. The International Journal of Advanced
Manufacturing Technology 14 (1): 6569.
Durupt, A., Remy, S. & Derigent, W. 2008. From a 3D Point
Cloud to a Real CAD Model of Mechanical Parts, a Product Knowledge BasedApproach. InYan, X.-T., et al. (eds.),
Global Design to Gain a Competitive Edge: 805813:
Springer London.
Edelsbrunner, H., Harer, J. & Zomorodian, A. 2001. Hierarchical morse complexes for piecewise linear 2-manifolds.
7079.
Edelsbrunner, H., Letscher, D. & Zomorodian,A. 2000. Topological persistence and simplification. In Proceedings of
the 41st Annual Symposium on Foundations of Computer
Science: IEEE Computer Society.
Ferreira, J. C. & Alves, N.F. 2003. Integration of reverse
engineering and rapid tooling in foundry technology.
Journal of Materials Processing Technology 142 (2):
374382.
Iuliano, L. & Minetola, P. 2009. Enhancing moulds manufacturing by means of reverse engineering. The International
Journal of Advanced Manufacturing Technology 43(5):
551562.
Iuliano, L., Minetola, P. & Salmi, A. 2010. Proposal of
an innovative benchmark for comparison of the performance of contactless digitizers. Measurement Science and
Technology 21: 105102105114.
Kermanpur, A., Mahmoudi, S. & Hajipour, A. 2008. Numerical simulation of metal flow and solidification in the
multi-cavity casting moulds of automotive components.
6268.
Shehata, F. & Abd-Elhamid, M. 2003. Computer aided
foundry die-design. Materials & Design 24 (8): 577583.
Sokovic, M. & Kopac, J. 2006. RE (reverse engineering) as necessary phase by rapid product development.
398403.
Sun, Z., Hu, H. & Chen, X. 2008. Numerical optimization of gating system parameters for a magnesium
alloy casting with multiple performance characteristics.
Journal of Materials Processing Technology 199(13):
256264.
Suresh Babu, T. & Thumbanga, R. D. 2011. Reverse Engineering, CAD\CAM & Pattern Less Process Applications in Casting-A Case Study. International Journal of
Mechanics 5(1): 4047.
185
Tai, C. -C. & Huang, M. -C. 2000. The processing of data

points basing on design intent in reverse engineering.
International Journal of Machine Tools and Manufacture
40(13): 19131927.
Vrady, T., Facello, M. A. & Terk, Z. 2007. Automatic extraction of surface structures in digital shape reconstruction.
Computer-Aided Design 39 (5): 379388.
Vrady, T., Martin, R. R. & Cox, J. 1997. Reverse engineering
of geometric modelsan introduction. Computer-Aided
Design 29 (4): 255268.
Wu, M., Tinschert, J., Augthun, M., Wagner, I., SchdlichStubenrauch, J., Sahm, P. R. & Spiekermann, H. 2001.
Application of laser measuring, numerical simulation
and rapid prototyping to titanium dental castings. Dental
Materials 17 (2): 102108.
Yu, D. M., Li, X. J., Xiong, Y., Gao, Z. H. & Wang, D. 2011.
Reconstructing of Prototype Surface with Reverse Engineering and Data Process Technology. Key Engineering
Materials 458: 368373.
186

Patient specific parametric geometric modelling of cementless hip prosthesis

G. Saravana Kumar & M. Gupta
Indian Institute of Technology Madras, Chennai, India
ABSTRACT: Total hip replacement is a surgical procedure to replace a diseased or injured hip by a prosthetic
implant. Though the geometry of commercially available implants has been standardized, due to changes in
biological features, the available choice of implant may not be the best-fit for a particular patients hip joint. The
present work aims at creating a framework for subject specific femoral implant design. Initially a set of prespecified femoral features that can be used for custom implant design have been identified for this study based
on interaction and discussion with orthopaedic surgeons. The identified features are extracted from the femur
geometry reconstructed from computed tomography images. A parametric implant design is proposed based
on the extracted femur features. The framework has been implemented in SolidWorks using its application
programming interface. An example case study is presented to demonstrate and validate the framework and its
implementation.
1
INTRODUCTION
Surgical treatment for human hip joint to cure severe

arthritis and trauma injured patients is an important orthopaedic procedure. This surgical procedure
known as total hip replacement replaces a diseased
or injured hip by a prosthetic implant. The surgery
consists of removal of femoral head and acetabular
surface and placement of an artificial joint consisting of acetabular cup and femoral stem with head, in
cavities of appropriately drilled bones. Implants can
either be directly forced press-fitted into the cavity
or fixed to its walls by an interfacing layer of flexible polymeric cement. Various researches over the
period of last seven decades have focused on materials and their biocompatibility, tribology, geometry
of the design and the means of avoiding the implant
failures. Recently orthopaedic device manufacturers
are increasingly investing in computer aided design
(CAD), finite element analysis (FEA) and biomechanics simulation technology for design validation and
reliability of the implants and regulatory approvals.
The geometry of the commercially available implants
has been standardized and commercial templates available for implants to help the surgeon make the choice
of implant. However, due to changes in biological
features, the templates may not be the best-fit for
a particular patients hip joint. For example, quite a
large number of designs created in accordance with
the anatomical data of patients in the United States
or Europe are not directly applicable to patients in
South or South-east Asia. Recent advances in CAD,
biomedical image processing and FEA will pave way
to perform effective computations for custom designs
so that post operative recovery is fast and long term
bone loss, prosthetic failure and revision surgery can
be avoided (Ma et al. 2005).
The present work aims at creating a framework
for subject specific femoral implant design based on
the computed tomography (CT) images for specific

patients. Initially a set of pre-specified femoral features that can be used for custom implant design have
been identified for this study based on interaction and
discussion with orthopaedic surgeons. Some of the
features identified are femoral anatomical axis, headneck shaft angle, size of the cross section of medullary
canal, location of lower Trochanter, and femoral head
offset from femoral anatomical axis. These identified
features are extracted from the femur geometry reconstructed from CT images. 3D reconstruction of the
subject specific femur model is accomplished using
Amira, a medical image processing software. A parametric implant design is proposed and based on the
extracted femur features a set of parameters are computed for implant design. Some parameters associated
with the model are stem length, head-neck shaft angle,
radii of cross section of stem at specific locations along
the length, medial curvature, head offset length etc.
This proposed parametric model is made available as
a library part in a solid modelling package. The process
of extraction of the femoral features is semi-automatic
with manual intervention using 3D modelling software SolidWorks. The process of implant model
instantiation is automated through a macro encoded
in C++ language using SolidWorks application programming interface (API). An example case study is
presented to demonstrate and validate the framework
and its implementation.
CUSTUM IMPLANTS FOR TOTAL HIP

REPLACEMENT
2.1 Total hip replacement

The components of implant used in total hip replacement (THR) are shown in Figure 1. In late 1950s,
187
Figure 1. Typical components of a THR.
Sir John Charnley, developed a implant made of stainless steel with Polymethylmethaacrylate (PMMA) as
bone cement for an improved bond between bone and
implant. This model is still considered as one of the
gold standards for the reasons of its very high success
rate (85%) after 1520 years (Gard 2000). Conventionally, stainless steel and cobalt-chrome alloy
were used to manufacture implant. In the past two
decades, various forms of ceramics, high molecular
weight polythene, and titanium alloys have been used
to manufacture implants. The main considerations for
choice of implant material apart from strength, corrosion resistance are wear resistance, bio-compatibility
and ability for osseo-integration. In the recent decades,
cementless implants have gained popularity. Compared to cemented implants, cementless type take
a longer post-operative recovery time but they are
believed to provide more strength and longer implant
life and have been preferred in the recent times.
The success of a joint replacement is directly related
to the various factors such as, septic loosening due to
infection in early stages of recovery, degradation of
implant-bone interface due to micro-motions caused
by poor implant fit, mechanical failure, aseptic loosening caused by wear and corrosion and bone loss due
to stress shielding. Stress shielding is a phenomenon
referring to reduction in bone density owing to reduced
normal stresses. This occurs when there is a large shift
in the stress distribution pattern in the bone before and
after the operation. Hence, implant design should consider these various factors so as increase the success
rate of THR.
2.2
Custom implant design
Implant design in the past 50 years have followed

more or less similar basic trends. The geometries were
designed keeping in view, insertion of implant into
the femur, lesser volume (thus, making it lighter),

attempting an improved press fit to prevent loosening. Conventionally, manufacturers provide design
templates for the available variants and a surgeon
makes a choice based on the fit and fill analysis using
X-ray images of the patients hip joint. In the recent
years, considering a wide diversity in the anthropometric measurements of femur population, there has been
attempts to customize the design in order to improve
the implant fixation and prevent its loosening (Kenzi
et al. 2009, Jun & Choi 2010). The study by Kenzi
et al. reports that customized implant design provide
appreciable results for patients whose femurs have significant anatomical deformity. In an early study Chang
et al. (1999) demonstrated a possibility of getting better results by increasing distal stem length of standard
implants as compared to customizing implants.
The earliest attempt to produce customized implant
employing a CAD system taking CT data as input was
made by Anderson (1991). A software environment
HIDE was created by Viceconti et al. (2001) to create custom designs. However, there is no analysis and
report of success of these implants. Though, there has
been a recent work in the direction of femoral feature based implant design (Jun & Choi 2010), there
has not been an attempt on parametric CAD modelling of implant design, design automation which will
reduce development time. The present work envisages
accomplishing the same.
METHODOLOGY
3.1 Framework
As shown in Figure 2, the overall framework adopts an
approach which starts from patient specific anatomy
data. CT medical scan data is used for reconstruction of
subject specific femur model. This 3D reconstruction
of the subject specific femur model is accomplished
using Amira medical image processing software.
This model which is a surface model (mesh model) is
operated on is SolidWorks CAD software environment for various feature extraction procedure as well as
implant modelling. The implant design process which
is the main focus of this work (portion inside the box
as shown in Figure 2) starts from this input.
The process of feature extraction involves extracting relevant features from the given femur. The process
of extraction is user interactive and semi automatic.
Various computational geometry algorithm routines
along with user interventions and reconstructions are
used for this purpose. A basic template for parameterization of implant geometry is proposed. The
parameterized model takes numerical inputs based on
various extracted features to generate a solid model of
the desired implant geometry. An automated platform
is created to use these parametric inputs in order to
generate such an instantiation in SolidWorks using
application programming interface (API). Entire process of feature extraction and implant instantiation
188
Anatomical femoral axis and neck axis- Various

centroids obtained from cross sections facilitate creation of femoral and neck axes, which in turn help in
creating a reference plane and coordinate system, and
aid in extraction of more features from femur.
Femoral head centre location and head sphere
radius- Anatomical femoral head is not an ideal
sphere. However, it can be approximated to one. This
sphere fitting yields head centre location and head
radius. Head centre serves as an important reference
location in creating reference coordinate system for
the femur. Head radius may not be directly used
for designing femoral part of the implant but is used for
designing acetabular part of THR.
Lower Trochanter reference- Lack of specific geometrical reference in CT scan reconstructed model,
and wide variations in anatomy of femur population
pose a challenge in referencing. Conventionally, tip
of the lower Trochanter has been used as reference in
radiographs for placing an implant into the femur (Dor
2006).
Medial curvature- An improved fixation requires
close interrelation with anatomy of femur. The medial
side is important for the load transfer. The curvature
of the femoral geometry towards the medial side is
extracted in order to guide implant geometry to take
shape closer to femoral geometry.
Figure 2. Schematic of the methodology.
3.3 Parametric implant model
Figure 3. Relevant features of femur for custom implant

design.
allows user intervention at different stages in order

to accommodate the changes in the design. Surgeons
inputs, for example, would be very useful for such a
platform in order to fulfil better the expectations from
an implant. Virtual assembly of the instantiated stem
with femur model is done to check for the form fit.
3.2
Femoral feature extraction
Figure 3 shows relevant feature parameters which have

been identified and extracted from the reconstructed
femur. They are:
Femoral & neck cross sections and radii- cross sections, centroids and mean radii at selected locations
for femoral shaft and neck shaft are needed in order to
define the cross sections of implant geometry so as to
have good fit and fill.
In general, a direct thought to design the implant could

be, to design it so that it exactly matches all the internal
cross sections and dimensions of femur so as have a
good fill and fit upon insertion. However, this design
would be very bulky. It will be over design considering the fact that the implant will be manufactured with
high strength materials. More importantly, it would
be very difficult to insert such an implant into the
femur. Hence in this work a design template is proposed which will have a good fit on the medial side
on the so as to have a better load transfer characteristics and will be thinner along the sagital plane.
The design template proposed is shown in Figure 4
upon which parameters are defined. These parameters
are then customized for specific patients based on the
parameters extracted from the femur model. Though
some conventional design do have modular designs
and templates to attempt a better fit for femur but they
are not customized and this work differs in this context. Also the complete design is instantiated with little
user intervention in a CAD environment.
The design template is instantiated with O as the reference co-ordinate frame. The frontal plane being zx
and the sagital plane being yz and z along the femoral
axis. The origin O is located at the intersection of
the femoral and neck axis. The femur stem shape is
defined with cross sections at three sections and length
of separation between them. The lowermost section
Section 1 is the end of trabaculae bone with circular
cross section of diameter r1 . The second cross section Section 2 (Figure 4) separated by a length L1
189
and with radius rh (is much smaller than the femoral

head radius). The neck length and radius (rn ) is determined based on the distance of the head from this cut
plane and load carrying capability required respectively. The above described parametric model can be
instantiated based on the parameter values computed
from the extracted femur feature parameters as well as
surgeons input.
3.4 Implementation
Figure 4. Parametric implant design.
is located at the z level of lower Trochater refernce

point P. An elliptical cross section with diameter r1
along the anteriorposterio axis (y axis) and r2 along
the mediolateral (x axis) is chosen for this section. This
section is located at a z level of L2 below the origin
O. The third section Section 3, which is the top most
section is chosen rectangular with width r1 and length
r3 . This section is located at a distance of L3 above
the origin O. The z height (L3 ) location for this section is chosen such that the neck cut perpendicular to
neck axis (shown by chain lines) at an offset distance
of D from lower Trochanter reference point P yields
a section to support the implant neck. The neck cut
is itself decided in consultation with surgeon. These
cross sections have uniform thickness r1 along the
anteriorposterio axis. These cross section dimensions,
lengths and locations are directly computed from the
extracted femur parameters in the co-ordinate frame
O. The three cross sections are lofted to obtain the
solid model of the stem. The loft between Section 1
and Section 2 is linear and that between Section 2
and Section 3 follows the arc of radius R to follow
the medial curvature of the femur. The lateral side of
this loft section is linear to reduce the volume of the
implant and brings the top cross section edge to the
femoral line in the frontal plane (Figure 4). This solid
is cut at an offset distance of D from P, perpendicular to the neck axis. This cutting provides a reference
point Q in the implant for the proper surgical placement by surgeon and or pre-operative. post-operative
study in radiograph. The spherical head is located at C
3D model of femur reconstructed in medical image

processing and 3D reconstruction software Amira
is initially taken to CAD modelling package SolidWorks. SolidWorks interface is used to operate
on the femur surface mesh model, reconstruct a few
geometric aspects with the help of below mentioned
algorithms and extract the relevant features. This operation is semi automatic with user intervention and uses
data extraction in SolidWorks and computation using
algorithms coded in C++ and compiled and run outside SolidWorks. The extraction of cross sections
(femoral and neck) is accomplished using prominent
cross sections (PCS) (Sellamani et al. 2010). The
approach obtains a locally swept cross section starting
from a selected point. An unspecified plane is initially
chosen at this point and normal vectors associated with
the intersection of this plane and all other mesh triangles are obtained. Then a sectional Gaussian map
which comprises of all these normal vectors is created
and least square method is used to fit a plane. The
plane, thus obtained, is again used to intersect the surface mesh and this process continues till convergence
is reached within the tolerance limits. The plane, thus,
finally obtained is called as PCS. This method is used
to extract various cross sections in neck and femoral
shaft. The computational geometry algorithms library
(CGAL), an open source project (http://www.cgal.org)
is extensively used in the present implementation.
Once the cross sections are extracted they are used
to obtain centroids and least square fitting (LSQ) of
the same is done to obtain the femoral axis (F) and
neck axis (N) (Figure 4). The coordinate frame O is
created using the plane obtained from these axes. The
femoral axis is taken as negative z and the lateral to
medial axis as x. Points are sampled from the femur
head surface mesh and a LSQ sphere is fitted to give
its centre C and radius rh . The lower Troachanter reference point P is obtained by visual inspection and the
neck cut offset distance D is obtained from surgeons
suggestions. All points are obtained in the reference
plane O. Proximal regions of femur have irregular
cross sections. It becomes practically very difficult to
define these cross sections and their dimension. To
parameterize the implant geometry in these regions,
dimension of the cross section from medial side to
lateral side is considered (r13 ) (refer Figure 4). The
sectional dimensions for Section 13 are obtained
by measuring distances on the cross sections obtained.
To define the medial curvature an intersection curve is
created by the intersection of Reference plane (zx) with
190
the femoral cancellous surface. A three point arc is

defined to obtain the radius of curvature R.The various
lengths (L13 ) are computed from the obtained information. Several parts of the codes were written in C++
and wherever apt open source codes such as Meshlab
(http://meshlab.sourceforge.net/) were adapted.
To create the parametric implant model, SolidWorks API has been used to write macros which
generate a CAD model in the SolidWorks interface. Creating a macro follows the approach of first
recording it while first time manual modelling in
SolidWorks interface and then editing it into a Windows form based application that can run along with
SolidWorks. These macros have been encoded in
Visual Basic.NET or C#.NET.
CASE STUDY
A case study is illustrated to demonstrate the framework and its implementation for custom femoral
implant design. The input femur model taken for
this demonstration is a standardized femur model
(Viceconti et al. 1996) procured from biomedical community resources. Sequentially, all the steps described
in order to create a femur specific implant are demonstrated below.
Initially relevant cross sections in the femur shaft
and neck were obtained using PCS algorithm from
the femur mesh model (Figure 5a). The centroids of
these sections were computed and the femoral and
neck axis F and N were determined to create the reference coordinate frame with origin at O. The femur
head was sampled and a sphere fit obtained to locate
its centre C and compute its radius (Figure 5b). The
lower Trochanter reference P was located visually
(Figure 5c) and the neck cut offset D was fixed with
surgeons help.
Some computational results of the feature extraction process are tabulated in Table 1. The values shown
are as computed and will be transformed to the coordinate frame O. A line is represented as vector passing
through a point and a plane is represented by its normal
vector and a point through which it passes. Standard
vector notations are used and all units are in m.
Once the parameters are available for modelling
the implant geometry, the parametric implant model
is instantiated with O as the reference coordinate. The
parameter values are inputted to the forms as the Windows form based client is run in Visual Basic.NET. The
macros written in SolidWorks API take these input
parameter values and create the geometry. Initially the
reference plane and the location of the various cross
sections Section 13 are created at distances (L13 )
as shown in Figure 6a. The cross section curves namely
circle, ellipse and rectangle with parameters r1 , r1 and
r2 , and r1 and r3 respectively. These sections are used
to loft and create the initial solid model of the stem
(Figure 6b). The neck cut and the neck geometry are
added later with the lower Trochanter reference point
P and offset distance of D as shown in Figure 6c. The
Figure 5. Feature extraction from femur mesh model.
head at location C and radius rh and neck with radius

rn are added later and finally all sharp edges removed
by adding fillets to create the final geometry as shown
in Figure 6d. The implant geometry parameter values
computed from the femur feature parameters fro this
case study are listed in Table 2.
The implant thus designed was analyzed for its
form and fitness with respect to the femur from which
the parameters were extracted. Since the main intention for subject specific implant design is to achieve
better results for implant fit and fill, this is an significant analysis and validates the design. This analysis
can be done in two ways. Firstly, by virtual assembling, placement and interaction of implant and femur
model one can evaluate the goodness of fit of the
THR assembly. Secondly, a physical prototype of the
implant model can be used along with the conventional
method of using radiographs to compare the fit and
fill of implant inside the femur. The former method is
adopted here and a physical prototype was fabricated
191
Table 1.
Some results of feature extraction.
Features
Data points
Computational result
Femoral shaft axis
(0.0387731,0.00804201,0.185)
(0.03903718,0.00671027,0.165) ,
(0.00794232,0.0027007,0.01628417)
(0.00971076,0.00217994,0.01534767)
(0.01138615,0.00165519,0.01422039),
(0.008029,0.021254,0.041424)
(0.002772,0.01895,0.040588)
(0.007568,0.02136,0.030176)
(0.003159,0.024253,0.017729),
(0.00794232,0.0027007,0.01628417)
(0.00971076,0.00217994,0.01534767)
(0.01138615,0.00165519,0.01422039),
F = (0.0223,0.0712,0.9972)
through P = (0.0397,0.0048,0.1390)
N = (0.8293,0.2506,0.4994)
through P = (0.0114,0.0017,0.0142)
Neck axis
Femoral head centre and
radius
Reference plane
C = (0.00961172,0.00254071,0.0246312)
rh = 0.0266
Y = (0.1039,0.9935,0.0457)
through P = (0.0268,0.0023,0.0082)
Table 2. Parameters and their values for the case study

implant design.
Parameters
Values in the reference

coordinates O Units in mm
Section locations, L13

Section parameters, r13
Medial curvature, R
Lower Trochanter location, P
Neck cut line offset, D
Femur head center, C
Head radius, rh
Neck radius, rn
Neck shaft angle
70, 35, 27.5

17, 20, 54
61
(33, 0, 20)
38
(54, 0, 35)
16
5
59
Figure 6. Step by step instantiation of the parametric

implant model in SolidWorks.
to do form analysis. The results of virtual assembly

of the implant model with the femur model after neck
cut is shown in Figure 7. The vertical section and two
cross sections of the assembly (one at Section 1 and
the other at Section 2) are shown. A qualitative visual
inspection shows that the implant geometry follows the
femur along the medial side and also the implant stem
is well placed in the trabacular region and the medulary
canal. A quantitative analysis using the SolidWorks
interference evaluation tool was done and the results
predicted proximal bone loss of 32000 mm3 (excluding the neck cut) and 600 mm3 of cortical bone loss.
Figure 7. Femur and implant model assembly cross section

views in SolidWorks.
This is the interference which reamed during the operational procedure. Further analysis is needed to evaluate
the bone loss with surgeons inputs. A rapid prototyping model was also fabricated to evaluate the form and
fit (Figure 8). Further analysis of this model with radiographs will have to be performed to further ascertain
the quality of fit.
192
geometry routines have been adapted from open source

codes as well as some routines written in SolidWorks
API to accomplish the task of femur feature extraction and instantiation of parametric THR stem model.
A case study has been presented to demonstrate the
methodology as well as to validate the design outcome.
Improvements in the present implementation can be
made in term of further automation and development of graphical user interface (GUI). An alternative
approach without using commercial 3D modelling
package can also be explored.
REFERENCES
Figure 8. Physical prototype model of implant geometry

fabricated using a rapid prototyping machine from Objet.
CONCLUSIONS
A methodology for subject for custom THR implant

design specific to patients has been proposed and
implemented. Specifically, a parametric geometric
model of the THR implant that can be instantiated
by the implant designer by supplying specific values to the parameters has been accomplished. In
the present framework the various geometric parameters are automatically computed based on the feature
parameters extracted from the femur geometry. User
also has the choice of inputting some parameters
based on surgeons suggestions. Various computational
Anderson, J., 1991. Custom designed hip implants through

knowledge based engineering, Materials and Design, 12,
103104.
Chang, P.B., et al., 1999. Preclinical cost analysis of
orthopaedic implant: a custom versus standard cementless femoral component for revision total hip arthroplasty,
Journal of Biomechanics, 32, 13091318.
Dorr, L. D., 2006. Hip Arthroplasty-Minimally Invasive
Techniques and Computer Navigation, Elsevier Saunders.
Gard, P.J., 2000. Hip replacement: choosing an implant.
Operative Techniques in Orthopaedics, 10, 94101.
Jun, Y. & Choi, K., 2010. Design of patient-specific hip
implants based on the 3D geometry of the human femur.
Advances in Engineering Software, 41, 537547.
Kenzi, K., et al., 2009. Computed tomography based custom
made stem for dysplastic hips in Japanese patients. The
Journal of Arthroplasty, 24, 6570.
Ma, R., et al., 2005. Design and manufacture of custom hip
prostheses based on standard X-ray films. International
Journal of Advanced Manufacturing Technology, 27,
7074.
Sellamani, S., et al., 2010. PCSprominent cross sections for
mesh models, Computer Aided Design & Applications, 7,
601620.
Viceconti, M, et al., 2001. HIDE: a new hybrid environment
for the design of custom-made hip prosthesis, Computer
Methods and Programs in Biomedicine, 64, 137144.
193

Cross-section morphological study

B. Bauer, A. Tibi & U. Shavit
Technion, Israel Institute of Technology, Haifa, Israel
ABSTRACT: We all experience steamy, soot shrouded metropolises. Although we enjoy their advantages, due
to heavy air pollution, none of us will want to spend their life in them. The rise in temperature and air pollution
in large cities is caused by different reasons. One of them is the air flow regime within the city. Control over the
air flow can contribute significantly to the improvement of bio-climate comfort. In contrast to the city, whose
soaring building climbing toward the sky from obligatory grid, branched corals channel the surrounding water
flow into harmonic and asphyxiation free environment. This article presents a preliminary comparison between
those morphological structures, under the assumption that this can lead to new urban structures, and in attempt
to break the smog dome that engulfs our cities.
INTRODUCTION
1.2 The logic of mimicking corals in urban design
Contemporary urban design faces tremendous challenge. The urbanization of humanity forces us to plan
dense cities. Unfortunately this kind of design is often
not considering the city environment (Makoto et al.
2000). An example of that process can be seen in
the heat island phenomena. The causes for this phenomenon are varied and include the built area, building
materials, and even the number of inhabitants in the
city contributes to the gap between the city and its surrounding temperatures (Oke, 1982). One of the factors
that enable the lingering effect of the heat island is the
urban air flow regime.
1.1
Urban heat islands
The city geometric-morphologic characters have an

influence over the city air flow, and over the air
exchange between the city and its surroundings. The
air exchange is important for a couple of reasons,
among them the reduction of the urban air pollution
(Sini et al. 1996) and the reduction of the heat island
effect (Brunstein & Johnson, 1977).
The influence of the heat islands is effecting the
power consumption and economy (Hassid et al. 1999)
but resent research suggest that even public health is
affected by the temperature rise (Conti et al. 2004).
Facing the importance of the urban air flow to the
city and its inhabitants, it seems logical that greate
investment will find their way into the field, but things
are never that simple. Air flow analysis in complex
morphologic systems, such as the city, is very complex (Sini et al. 1996). This complexity is turning
the urban design process under such constraints to
virtually impossible.
The notion of mimicking nature is not new. Even

the idea of mimicking nature in urban design is not
new (Brown & Kellenberg, 2009). The difficulty with
mimicking nature in urban design is the elusiveness
of the urban system, as demonstrated above, in the
urban air flow regime. On top of that, the complex
urban morphology makes it difficult to conduct physical experiments to examine the effects that different
morphology have on urban air flow. One of the known
procedures for that kind of experiments is abstraction
of the urban system into a schematic model. These
models can be compared to natural morphological systems and such a study may produce new solutions for
changing the air flow in the city and overcoming the
heat island effects.
Coral colonies are dependent on the water flow
around them and within (Sebens et al. 1996). For this
reason they evolved into structures that enable optimal
water flows. This mechanism can be study in context
of the urban morphology in attempt to break the smog
dome that engulfs our cities.
1.3 Water flow in corals

Corals are colonies of polyps that live symbiotically
with algae inside their body (Sorokin, 1993). There
are a large variety of corals. This paper is focusing
on branched coral colonies and their ability to control water flow all over the surface of the coral in
fixed speed. Those polyps sink a lime skeleton which
they live on. Due to their fixation to the skeleton the
colony is immobile and is depend completely on the
water flow for nutrient, minerals and oxygen supply
and removal, and disposal of waste. Indeed the symbiotic algae provide the polyps with oxygen that it creates
in the photosynthetic process, but even the algae needs
195
the minerals that the coral extract from the water flow.
On top of that during the night both the algae and the
polyps respire and need oxygen supply and waste disposal (Monismith, 2007). The ability of the coral to do
all that is linked directly to the water flow speed: fast
currents wont let the coral catch food; slow currents
wont dispose the waste in satisfying speed.
Since everywhere on the sea floor the speed and
direction of the water flow is rapidly changing, the
corals manipulate the current to suit their needs.
Those mechanisms are expressed in variety of parameters in the colony morphology such as: size, branch
diameter, distance between branches etc. (Sebens
et al. 1996, Kaandorp & Kubler, 2001). Additionally, the currents affect the survival probability of
the colony, metabolism, energy, and prey capturing
success (Sebens & Johnson, 1991).
2
BACKGROUND
The research objective is to compare two different

morphological systems, the city and the branched
coral. Apparently two, completely different, systems:
from the morphology level to the inhabitant and their
characters.
These two systems, despite the differences, rely on
public goods for the basic functions of their inhabitants. Surprisingly, in both systems, it is done by
flow. City residents enjoy fresh air supply and pollution disposal by the breeze-city air flow. The polyps
enjoy oxygen and nutrients supply and carbon dioxide
disposal by the sea water flow.
Therefore the two systems are comparable, at least
based on some parameters, and it is possible to learn
and imply from one system to the other.
But the question remains: how can it be done? What
are the parameters that are relevant to comparison
between system that is based on polyps that lives on
external skeleton to a system that is based on humans
who build external construction and live inside them?
2.1
Parameters comparison dilemmas
Two major problems arise during the comparison

between the systems: one is the sampling of complex
systems; the second is how to create a platform to
compare two different sampling methods. The sampling problem is not unique to these systems. It exists
in complex morphological systems or extreme scales.
The sheer size of the city prevents us from sampling
the city directly. Even if we lower the accuracy of the
sample method we wont be able to build scanner
that can contain entire cities. In a similar manner there
is a problem to sample tiny systems; where the problem
is not the size of the scanner but its precision and
adequate resolution to complex systems.
In corals we encounter another problem (that exist
in morphological analysis of cities too, but in lower
intensity) self shading. Since coral branches creates
complex formation, it is impossible to see the whole
formation by external scan. This is the reason that light
base scanner are not adequate for the sampling process.

Other problems such as the difficulty of measuring
the live tissue of the corral are wider than the scope of
this research.
Even if we solve the topological analysis of each
system separately another question remain: how can
we compare a coral with diameter that reaches few
dozens of centimeters with a city that can reach
tens of kilometers? How can we compare the coral
branch perimeter that reaches several centimeters with
buildings that may reach tens of meters?
2.2 The base for the analysis system
Because of the dilemmas mentioned above it was
decided to develop a comparison system that is based
on relative parameters. By using this non-dimensional
method the different systems are not compared to fixed
ruler, instead from each system parameters which
describe relations between the characters of the system are extracted. These parameters are compared to
similar ones that extracted from other systems.
This method creates pure comparison system that
is not related to the morphological system characters
specifically. By this way it is possible to characterize
every morphological system with relative parameters
that are comparable.
3
METHODOLOGY
The base for the data collection was finding a way to

sample these complex morphological systems, understanding it, and creating a process base on it that
enables an analysis of different systems in different
formats.
The biggest challenge was the coral skeleton sampling because of the self-shading quality of the corals
that prevents light base scanning (such as laser). In
order to sample the coral skeletons MRI scan and
CT scan were examined. These methods use magnetic
resonance and X-ray resonance, respectively, and produce graphical cross sections of the scanned object
(Fig. 1).
Since the output quality of the CT scan was higher
(less background noise), and since the MRI scan
requires to submerge the coral in a water tank, the
CT scan was chosen for the sampling process in this
research.
The CT scanner provides graphical information in
a two-dimension pixels matrix. The Matlab program,
that has internal functions for graphic analysis, was
chosen for the analysis process. In a similar manner all
the morphologic systems where sliced into graphical
sections, uploaded into Matlab, and morphologically
analyzed.
3.1 Coral species in the research
The most important reason for choosing coral species
for this research was their morphological structure.
Branched corals with visual similarity to the structure
196
3.3
City sampling
The urban models were taken from existing

urban models of the three-dimension database
of Google Sketchup (at http://sketchup.google.com/
3dwarehouse), and were chosen from the biggest
metropolises in the world.
In order to maintain the same sampling method as
used in the coral scan, each urban model was sliced into
horizontal sections, one meter apart. In every section
each pixel represents one square. Since the sections
were positioned one meter apart each pixel is actually
representing one cubic meter, and can be referring to
as one voxel.
Figure 1. Section from the CT scan of Pocillopora

damicornis.
of building rising from the ground where chosen. Two

additional criteria for choosing specific species are:
the location of the coral in the reef and its availability
to the research.
The availability of the coral to the research relies
upon coral skeletons that confiscated, from smugglers,
by the Israel Nature and Parks Authority.
The third criterion was the origin of the coral from
the reef outskirts. This species do not rely on nearby
corals to swirl the water; they cope with the water flow
as is.
3.2
Corals sampling
As mentioned above (section 2), CT scans have been

made in order to sample the coral skeleton. CT (computed tomography), is a non-invasive photographic
method that is common in medicine. The computed
tomography was invented by Godfrey Hounsfield in
the early 1970s at EMI labs in Britain. The invention
made possible by the wide use of microprocessors in
computers and medicine.
In this method the scanned object is projected with
X-rays from a source that rotates around the object,
sensors surrounding the object receive the radiation
and convert it into a picture of the object. As the
object projected from more angles the picture becomes
sharper. In medicine the risk of causing new health
problems by over projecting the patient limits the
exposure to the X-rays in each scan to a minimum.
The current research doesnt involve living organisms, so the scan limitations where negligible, and
the scan resolution was very high (less than a 0.1
millimeter).
Coral skeletons have been scanned in the past
(Kaandorp & Kubler 2001). From the scans, threedimensional models were extracted in order to understand the morphological development of the colonies.
In our research the coral skeletons were scanned in
the Imaging Unit of Rambam hospital (Haifa, Israel),
and after preliminary processing, basic database was
established.
3.4
Matlab analysis tool
A variety of tools were used for the data analysis. Some

of the tools are Matlabs toolboxs; others were developed by us for this research. In each section all the
patches were recognized, measured, and similarly
the relations between the patches were measured.
Each picture that was uploaded into Matlab was
cleaned from background noise, afterward a general
data of the picture produced (table 1, lines 15), then
data for every patch has been produced (Table 1, lines
815), and finally the relations between the patches
were obtained (Table 1, lines 67).
3.4.1 The what and how of the analysis criteria
The extraction of the criteria data in each section is
done by separating the patches from the background.
Each patch symbolized by a sequential number, and
a new matrix were each cell is associated with the
matching branch number.Afterward the rest of the data
were extracted based on a pixels count.
3.4.2 The criteria meaning
The purpose of the chosen criteria, beside the obvious role as analysis parameters, is to bridge between
the worlds, between coral morphology and urban morphology. To achieve this goal, each criterion that was
chosen is both morphologically important and the connection to one of the systems contributes a new layer
of meaning to it. If we consider the criterion mean
object area it describes the mean area of every branch
or building. But this criterion contain more information: in the coral the branch area connected directly to
the water flow speed, as the water flow speed rises the
branch area is becoming larger (Sebens et al. 1996). In
the city, however, the building area represents the space
were we, the humans, use the building. The functional
area in a building (floor area) is one of the most important parameters in an era that the world population
increase rapidly and the urban density is rising.
3.5 The standard output

The morphological analysis output composed from
two parts: overall analysis, and section analysis.
197
Table 1. Parameters that have been gathered during the

morphological analysis.
The data
Coral meaning
Height
The distance
between the cross
sections in meters.
The number of
buildings in a
given cross section.
The ratio between
the built area and
the vacant space
in a given cross
section.
The average distance, in pixels, that a particle
travels in order to cross the section vertically
without traveling through solid parts.
The average distance, in pixels, that a particle
travels in order to cross the section horizontally
without traveling through solid parts.
The average
The average number
number of
of branches that
buildings
can be reached in a
that can be reached
straight line from a
from a specific
specific branch.
building in a
straight line.
The shortest
The shortest disdistance
tance between two
between two
neighboring
neighboring
branches, on average.
buildings, on
average.
The mean area of a
The mean area of
building in a given
a branch in a given
cross section.
cross section.
The mean
The mean perimeter
perimeter
of a branch in a given
of a building in a
cross section.
The mean length
The mean length of
of the maximal
the maximal axis that
axis that crosses a
crosses a branch in a
building in a given
cross section.
The mean length of the perpendicular axis to
the major axis in it center.
The mean ratio between the perimeter and area
of a coral/ building as described in the formula:
Cir = (4*pi*area)/(perimeter2)
The mean ratio between the perimeter and area
of a coral/ building as described in the formula:
Com = 1/Cir
The mean ratio
The mean ratio
between the major
between the major
axis and perimeter
axis and perimeter of
of a building in a
a branch in a given
cross section.
The mean ratio
The mean ratio
between the major
between the major
and minor axes
and minor axes of
of a building in a
a branch in a given
cross section.
Number of
branches
Solid/void
ratio
Tortuosity X
Tortuosity Y
Neighbors
Distance
Branch area
Branch
perimeter
Branch
major axis
Branch
minor axis
Circularity
Compactness
Tortuosity
Axis ratio
City meaning
The distance between

the cross sections in
millimeters.
The number of
branches in a given
cross section.
The ratio between the
coral and the water
volume in a given
cross section.
Figure 2. Overall output, the relation between the branch

area to the rest of the parameters.
Figure 3. Section analysis of Acropora humilis.
relation between one parameter to the rest. In each

graph the main parameter displayed on the X axis,
the rest of the parameters are displayed on the Y
axis. This output enables seeing all the parameters in
one glance, identify the important ones, and focus on
them in future analysis. Figure 2 presents the relations
between branch area to the rest of the parameters. It
is easy to detest that there is a correlation between the
branch area and the height, diameter and the number
of neighbors, and less to the axes ratio.
3.5.2 Section analysis
The analysis of each section is divided into four
parts (Fig. 3): (a) Overall datasequential numbering, section height, solid/void ratio, etc. (b) Visual
datasection picture, branches center, section grid,
etc. (c) Patch concentrated datadiameter, perimeter, neighbors number, etc. (these data presented in
a table containing minimal, average, and maximal values). (d) Graphsthe graphs present the correlation
between the patch area and perimeter to the number
of neighbors and the mean distance between them.
4
3.5.1 Overall analysis

The overall analysis presents a set of graphs that
compares between all the criteria (Fig. 3). The output is comprised from 15 tabs; each tab presents the
PRELIMINARY RESULTS
The following results refer to one Acropora humilis

coral analysis and to the analysis of southern Manhattan. In Table 2 some of the results are presented.
As we mentioned above (section 2.2), the end results
are in pure numbers, therefore its very important to
pay close attention to the tendencies and relation of
the parameters.
198
Table 2. Preliminary comparison betweenAcropora humilis

and south Manhatan.
Units
Millimeters/
Numeric
Numeric/
Numeric
NumericNumeric
Pixels as
distanceMillimeters
MillimetersMillimeters
Acropora humilis
New-York
Graph 1. HeightNumber of branches
Graph 2. HeightNumber of
buildings
Graph 3. Number of
branches- Neighbors
Graph 4. Number
of buildingsNeighbors
Graph 5. Solid/
vacant ratioNeighbors
Graph 6. Solid/
vacant ratioNeighbors
Graph 7. Tortuosity in
X axis- Circularity
Graph 8. Tortuosity
in X axisCircularity
Graph 9. Distance
between branchesHeight
Graph 10. Distance

between buildingsHeight
to the number of patches is simple in both systems.

While in the city this parameter is linear in the coral its
parabolic. The structure of a coral starts at one point,
expends and branches out till half its height, than its
shrink back while creating typical perimeter.
Similar relationships, having different standard
deviation (Graphs 3, 4) The relation between the
number of patches to the number of neighbors is
parabolic both, in the coral and in the city. But while
in the city the standard deviation is very small, in the
coral its significant. In both graphs the significant
part, where the correlation changes from horizontal to
vertical, is located in the center of the graph. Despite
the similarity, small changes create difference in the
overall equation that describes each system.
Partial resemblance (Graphs 5, 6) The relation
between the solid ratio to the number of neighbors
is similar in the coral and in the city, but only in the
lower values. In the high values the coral is changing
the direction of the parabola while the city remains in
the same direction. Due to the spherical shape of the
coral some parameters appear twice and receive two
different matching values. Because of that the average between those two values lowering the parabola
position in the coral relatively to the city.
Absolute different (Graphs 7, 8) The relation
between the tortuosity and circularity in the coral is
zero; while in the city there is strong linear connection. Interestingly the parameter that describes the air
floe in the city and the water flow in the coral doesnt
have correlations between the systems.
Simple but different (Graphs 9, 10) the relation
between the distance and the height is simple in both
systems. But in contrast to Graphs 1 and 2, here the
connection in the city is parabolic and in the coral its
almost linear. The different correlation in the two systems can be explained by their construction method.
In the city as we climb up the number of buildings in a
section will reduce, but its not necessary that only the
closest building will remain, and as a result the distance between the buildings expands. In the coral its
spherical shape limits the branches in each section to
specific area and therefore the distance between them
remain.
4.2 The results meaning
4.1
Comparison results
The most prominent property of the parameters is the

large standard deviation in the coral. In almost all
the parameter there wasnt a line or a curve that can
describe the parameters of the coral. Nevertheless, it
is interesting to see that there is obvious trend in the
graphs (Table 2). In the city, on the other hand, the
correlations are much simpler, and can be described,
most of the times, as linear connection.
When looking at the results of the coral and the city
analysis, one may see that there are several ways that
they are differ (the data is taken from Table 2).
Simple yet different relationships between parameters (Graphs 1, 2) The relation between the heights
Flow is ruled by tortuosity, in the city and in the

coral. But the tortuosity itself is affected by a wide
variety of parameters that correlates or differs with the
basic grid of the system. When we correlate the city
to the coral there is one character that creates white
noise and differs the systems: the coral foundation.
The coral system is based on one foundation point that
connects the coral to the seabed, from this point the system branching out and shrinks back again. This fact,
for example, creates a big different between graph 1
and graph 2, but if we take only the top half of the coral,
we will see that the two systems are almost identical.
Graphs 3, 4 and 5, 6 are very similar too. The differences between graphs 9 and 10 are derived from the
199
coral foundation too. The coral is based on one point,

so when the system becomes smaller its shrinks to
one small area and the distance between the branches
remain. In the city each building have separate foundations, in order to build high-rises building one should
find appropriate soil to build the foundations, so the
distance between the tall buildings is determined by
the underground substance and not by one base point.
And again we encounter the tortuosity parameter
that can be seen in graphs 7, 8 in this paper. Those
graphs present absolute different between the parameters in the coal and the city. But here exactly lies the
meaning of the air and water flow. How the tortuosity
effect and affect from the different parameters of the
coral system, and is it possible to satisfy those parameters in urban system? If we were able to detect some
unexplainable parameters the differences in the flow
may be attributes to them, but in light of those preliminary findings its seems that the tortuosity is the main
parameters that needs further examination.
CONCLUSIONS
Its fascinating to see so different systems stand side by

side on the same scales, and the differences between
them start to get clear.
5.1
Differences and similarities in the coral and

the city typologies
The difference between the coral and the city is obvious

and clear. But the dipper and more profound differences are a result of the different in the relation between
the parameters of the two systems. The modern city is
organized and precise, streets cross each other orthogonally, the buildings are shrinking while soaring up
to the sky, and the urban grid is determined only on
the ground level. This uniformity probably has a great
effect on the city air flow; it creates canyons where
fierce winds sweep the streets and streets with no
wind at all. In the coral, however, the water flows are
not continuative since the morphology is much more
complex. It seems that this complexity gives the coral
some control over the water flow. This morphology
evolved during millions of years and the coral uses it
for its different needs.
Surprisingly, there are some similarities between
the systems. One example is that when a connection
between two parameters is strong in one system, usually, its also strong in the other (see Table 2 lines
13, 5). It doesnt mean that the connation will have
the same trend, force, or standard deviation, but something in the urban morphology and its connections is
similar to the branched coral one.
5.2
creation tangible appearance (Burt, 1966). As Architects we occupied daily by two aspects of the building:
functionality and esthetics.
Every architect dream is to design a project that
includes most of the relevant ingredients to the
designed subject. Designing a building that each detail
will compliment its neighbor and all of them will create functional and esthetical harmony. This paper seeks
the base of that harmony, by learning one morphologic
system and implementing the knowledge in another.
The coral morphology analysis and the understanding of how its integrated harmonically with the water
flow is the base for the breathe-taking beauty of the
coral reef. Developing analysis tools that this paper
presents, may lead to the implementation of these principals in the city that we inhabits. If we really be lucky
we may create healthier, harmonic city, and maybe
breathe-taking too.
Developing these tools open to us new knowledge,
that if we understand it we can use to solve our problems. Many critical urban problems are waiting at our
door for their solutions. As the cities grow denser by
the day, the need for high quality solutions is critical. The assumption that the evolution provide us with
a set of solutions that were tested and implemented
successfully, leads to the search for the secrets of life
in many fields.
If we succeed to use this knowledge, we might,
through nature, know ourselves better.
Future research
The preliminary meaning of beauty is the perfect

match between the true nature of phenomena and the
ACKNOWLEDGEMENT
Bnaya Bauer is grateful to the Azrieli Foundation for
the award of an Azrieli Fellowship.
REFERENCES
Bornstein, R.D. & Johnson D. S. 1977. Urban-Rural Wind
Velocity Differences. Atmospheric Environment 11: 597
604.
Brown, I. & Kellenberg S. 2009. Ecologically Engineering
Cities Through Integrated Sustainable Systems Planning.
Journal of Green Building 4(1): 5875.
Burt, M. 1966. Spatial Arrangement and Polyhedra with
Curved Surfaces and their Architectural Applications.
Haifa: Technion, Israeli Institute of Technology.
Conti, S. Meli, P. Minelli, G. Solimini, R. Toccaceli, V. Vichi,
M. Beltrano, C. & Perini L. 2005. Epidemiologic Study
of Mortality During the Summer 2003 Heat Wave in Italy.
Enviromental Research 98: 390399.
Hassid, S. Santamouris, M. Papanikolaou, N. Linardi, A.
Klitsikas, N. Georgakis, C. & Assimakopoulos D. N. 2000.
The Effect of the Athens Heat Island on Air Conditioning
Load. Energy and Buildings 32: 131141.
Kaandorp, J. A. & Kubler J. E. 2001. The Algorithmic Beauty
of Seaweeds, Sponges, and Corals. New-York: Springer.
Makoto, Y. Kazuhiko, T. Takashi, W. & Shigehiro, Y. 2000.
Beyond greenbelts and zoning: A new planning concept
for the environment of Asian mega-cities. Landscape and
Urban Planning 47: 159171.
Monismith, S. G. 2007. Hydrodynamics of Coral. The Annual
Review of Fluid Mechanics 39: 3755.
200
Oke, T. R. 1982. The energetic basis of urban heat island.

Quarterly Journal of the Royal Meteorological Society
108(455): 124.
Sebens, K. P. & Johnson A. S. 1991. Effects of Water Movement on Prey Capture and Distribution of Reef Corals.
Hydrobiologia 226: 91101.
Sebens, K. P. Grace, S. P. Helmuth, B. Maney Jr, E. J. &
Miles J.S. 1996. Water Flow and Pray Capture by Three
Scleractinian Corals, Madracis Mirabilis, Montastrea

Cavernosa and Porites Porites, in a Field Enclosure.
Marine Biology 131(2): 347360.
Sini, J. F. Anquentin, S. & Mestyer P. G. 1996. Pollutant
Dispersion and Thermal Effects in Urban Street Canyons.
Atmospheric Environment 30(15): 26592677.
Sorokin, Y. I. 1993. Coral Reef Ecology. Berlin: Springer.
201

Cellular structure design for lightweight components

J. Nguyen, S.-I. Park & D.W. Rosen
Georgia Institute of Technology, Atlanta, US
ABSTRACT: Cellular material structures, such as honeycombs and lattice structures, enable unprecedented
stiffness and strength characteristics, for a given weight. Such materials have very complex geometries and can
be fabricated only by additive manufacturing processes to produce them. New design and CAD technologies to
construct cellular materials are presented in this paper. A method is presented for the lay-out of cells along a given
part model surface or surfaces. Methods are presented also for constructing lattice or honeycomb structures from
the cellular lay-out. The software that embodies cellular structure design has been integrated into a commercial
CAD system and interfaces have been developed for finite element analysis codes. Such integration enables
stress, deflection, modal, and thermal analyses, as well as size optimization to be performed. Several test and
aerospace parts have been designed, analyzed, using the software and fabricated using Selective Laser Sintering.
Design results demonstrate that a wide variety of part shapes can be reinforced with cellular structure. Analysis,
fabrication, and testing results indicate that commercial FEA codes are good at identifying locations of high
stress, but have limitations in describing cellular structure failure mechanisms.
1
1.1
INTRODUCTION
Cellular materials
The concept of designed cellular materials is motivated by the desire to put material only where it is
needed for a specific application. From a mechanical engineering viewpoint, a key advantage offered by
cellular materials is high strength accompanied by a
relatively low mass. These materials can provide good
energy absorption characteristics and good thermal
and acoustic insulation properties as well (Gibson &
Ashby 1997). Cellular materials include foams, honeycombs, lattices, and similar constructions. When the
characteristic lengths of the cells are in the range of
0.1 to 10 mm, we refer to these materials as mesostructured materials. Mesostructured materials that are not
produced using stochastic processes (e.g. foaming) are
called designed cellular materials. In this paper, we
focus on designed lattice materials.
Cellular materials have a biomimetic or bioinspired origin. Many naturally occurring materials
have a porous construction that enables them to be
light, stiff or compliant, and multifunctional, for example to conduct heat well or to convey fluids and
nutrients.Ashby and co-workers (2000) provide a good
motivation for the study of cellular materials: When
modern man builds large load-bearing structures, he
uses dense solids; steel, concrete, glass. When nature
does the same, she generally uses cellular materials;
wood, bone, coral. There must be a reason for it. In
this context, the term cellular does not just mean
that the material is composed of living cells; rather
it refers to the construction using struts, webs, pores,
and/or channels.
Figure 1. Octet-truss unit cell and example parts with

octet-truss mesostructures.
In the past 10 years, the area of lattice materials has

received considerable research attention due to their
inherent advantages over foams in providing light,
stiff, and strong materials (Ashby et al. 2000). Lattice
structures tend to have geometry variations in three
dimensions; some of our designs are shown in Figure 1. As pointed out in (Deshpande et al. 2001), the
strength of foams scales as 1.5 , whereas lattice structure strength scales as , where is the volumetric
density of the material. As a result, lattices with a =
0.1 are about 3 times stronger than a typical foam. The
strength differences lie in the nature of material deformation: the foam is governed by cell wall bending,
while lattice elements stretch and compress.The examples in Figure 1 utilize the octet-truss (shown on the
left), but many other lattice structures have been developed and studied (e.g., kagome, Kelvin foam). We have
developed methods for designing lattice mesostructure for parts (Rosen 2007, Wang & Rosen 2002)
and have developed design-for-manufacturing rules
for their fabrication in SL.
203
1.2
Design for additive manufacturing
Design for manufacturing (DFM) has typically meant

that designers should tailor their designs to eliminate manufacturing difficulties and minimize costs.
However, the improvement of rapid prototyping, or
Additive Manufacturing (AM), technologies provides
an opportunity to re-think DFM to take advantage of
the unique capabilities of these technologies (Gibson
et al. 2010). Several companies are now using AM
technologies for production manufacturing. For example, Siemens, Phonak, Widex, and the other hearing aid
manufacturers use selective laser sintering (SLS) and
stereolithography (SL) machines to produce hearing
aid shells, Align Technology uses stereolithography
to fabricate molds for producing clear braces (aligners), and Boeing and its suppliers use SLS to produce
ducts and similar parts for F-18 fighter jets. In the
first three cases, AM machines enable one-off, custom
manufacturing of 10s to 100s of thousands of parts.
In the last case, AM technology enables low volume
manufacturing and, at least as importantly, piece part
reductions to greatly simplify product assembly. More
generally, the unique capabilities of AM technologies
enable new opportunities for customization, improvements in product performance, multi-functionality,
and lower overall manufacturing costs. These unique
capabilities include:
Shape complexity: very complex shapes, lot sizes

of one, customized geometries, and shape optimization are enabled.
Material complexity: material can be processed one
point, or one layer, at a time, enabling the manufacture of parts with complex material compositions
and designed property gradients.
Hierarchical complexity: hierarchical multi-scale
structures can be designed and fabricated from
the microstructure through geometric mesostructure (sizes in the millimeter range) to the part-scale
macrostructure.
In this paper, we cover two main topics. First, we
present geometric construction methods that enable
designers to take advantage of the shape complexity
capabilities of AM processes. Specifically, we develop
a method for constructing cellular materials that conform to the shapes of part surfaces; when restricted to
lattice structures we call such constructs Conformal
Lattice StructuresTM (CLS). Second, we present the
computer-aided design (CAD) tool that embodies the
geometric construction method and illustrate its application on two examples. The designed parts are meant
to be fabricated by Selective Laser Sintering (SLS).
CLS DESIGN METHOD
The basic idea of how cellular materials are created

is presented here. Four example primitive cell types
are shown in Figure 2, three of which are lattice
structures and the fourth is a foam. These cell types
Figure 2. Cellular primitives: three lattice structures and

one web structure.
are 2-dimensional for simplicity of presentation. The

octet lattice in Figure 1a is an example 3-D cell type.
Lattice structures consist of a set of struts (beams) that
connect the nodes of the lattice.
To generate the cellular designs in Figure 1, the
primitive cell types must be mapped into a mesh. In
2-D, the mesh consists of a set of connected quadrilaterals. In 3-D the mesh consists of hexahedra (6-sided
volume elements with planar sides). The uniqueness of
our work is our use of conformal cellular structures,
rather than uniform lattice block materials, that can
be used to stiffen or strengthen a complex, curved
surface. To see the difference between conformal and
uniform structures, Figure 3a is an example uniform
lattice structure, while Figure 3b shows a conformal
lattice. Meshes for uniform structures consist of cube
elements in 3-D (squares in 2-D), while for conformal
structures, the mesh elements are general hexahedra.
We have developed a new algorithm for generating
conformal meshes that are used to create conformal
lattice and cellular structures. An older algorithm for
generating conformal lattice structures based on a
mapped meshing approach (Wang & Rosen 2002) has
been updated significantly. We prefer that mesh elements are as cubic as possible; i.e., are of uniform
thickness and uniform size. Such meshes are typically
not generated by the free meshing methods in finiteelement analysis codes, while typical part geometries
are too complex for mapped meshing methods.
2.1 CLS construction method

The overall method for generating conformal cellular
structures is shown in Figure 4. It consists of two main
steps, indicated by the shaded rectangles: computing
3D conformal mesh, and populating the mesh with
cells. Inputs and outputs of the steps are shown as
ovals. Each step is detailed below.
The objective of the meshing algorithm is to generate a conformal hexahedral mesh into which cells
from the cell library can be placed. One or more layers
of cellular structure can be placed to support the parts
skin. The input to the algorithm may be a CAD solid
204
Figure 5. Algorithm for constructing a conformal mesh.
Figure 3. Uniform and conformal lattice structures.
Figure 4. CLS construction method.
model of the part, a surface model of the part, or a triangulated surface model of the part. The triangulated
surface model is very common in the rapid prototyping
industry; it is known as an STL model and consists of
a collection of triangles that approximate the curved
surfaces of the part CAD model. A method of constructing solid and STL models of lattice structures
was presented in (Wang et al. 2005); the method utilized the conformal lattice generation algorithm from
(Wang & Rosen 2002).
2.2
Construct 3D conformal mesh
The algorithm to generate a 3-D conformal mesh is

shown in Figure 5. The first step is to divide the part
boundary into regions, each of which is relatively flat.
It is easier to control the mesh generation method if
regions are flat. We implement an absolute angular
deviation measure between surface or triangle normal
vectors to determine if that surface or triangle should
be added to the region being generated. For the purposes of presentation, we will assume that a STL file
was given. Then, one triangle is chosen as the first
triangle of a region. The normal vector of each connected triangles is compared to the normal of the first
triangle; if they differ by less than a given tolerance,
the triangle is added to the region. As an example, the
tolerance for the simple part in Figure 4 was chosen so
that the model consists of two regions: the cylindrical
surface and the planar surface.
For each region, three main steps are performed as

indicated by lines 3, 4, 5 of Figure 5. The first of these
steps is to compute the offset of the object boundary.
An offset is a collection of points that are at a specified distance away from the starting surface (distance
is called the offset distance). As an example, an offset
of a circle is a circle that is concentric with the first circle. If you want the offset of a circle of radius 10 inches
centered at the origin, and the offset distance is 1 inch,
you would get a circle of radius 11 inches also centered
at the origin. The offset of the curve from Figure 3 is
shown in Figure 6. The positive offset curve (shown
with a long-dashed curve) is offset in the direction of
the original curves normal vector. Generally speaking, a positive offset results in a larger object, while a
negative offset results in a smaller object. Offsets are
standard geometric modeling operations that are typically available in mechanical CAD systems. We use
an offset method developed for tessellated part surfaces (Chen 2007), but any offsetting method could be
applied.
The second step is to construct a tri-parameter
volume between the original surface and its offset. Conceptually, this is simpler than it may sound.
Typically, curves and surfaces in CAD systems are
defined using parametric equations. For example, the
curves in Figure 3 would be parameterized by one
parameter, u, that varies from 0 to 1, as in the well
known Bezier, B-spline, or NURBS curves. Surfaces
are parameterized using two parameters. Volumes are
parameterized using three parameters and, hence, are
called tri-parametric solids. This step of the algorithm
assumes that the original boundary surface of the
object is parameterized. The same parameterization
is transferred to the offset surface. Finally, a lofting
is constructed from the original surface to the offset
surface. Parameterized equations and the operations
of lofting and offsetting are all standard, well known
geometric operations in the field.
However, a complication arises since the original
boundary surface of the object consists of triangles.
That is, the original boundary surface, which was
assumed to be parameterized in the paragraph above
actually is not parameterized. Hence, a parameterization must be developed, the details of which are
presented in (Engelbrecht 2009).
205
The third step is to generate the conformal mesh by

dividing the parameterized volume into individual elements (hexahedra). This is a very straightforward step.
After selecting either the number of elements or their
typical size, increments in each of the three parameters
are computed, which are used to generate elements of
the mesh. For example, if the increment in u is chosen to be 0.1, then 10 elements in the u direction will
be generated, since (1 0)/0.1 = 10. Note that the
numerator of the expression (1 0) is the difference
between the maximum and minimum u values. By
successively incrementing each of the three parameters
that define the tri-parametric volume, mesh elements
are created.
The final step in Algorithm Construct Conformal
Mesh (line 7) is to ensure that region boundaries
match by adjusting node positions and by adding elements, if necessary. Since the regions are parametrized
separately, the hexahedral elements may not match
well. Nodes from neighboring regions may be moved
and merged to achieve matched boundaries. Also,
a series of hexahedral or tetrahedral elements may
be added in gaps between meshes in neighboring
regions.
2.3 Construct CLS
The algorithm for the second step (populate mesh
with cells) of the overall conformal cellular structure
design method from Figure 4 is shown in Figure 6.
One input to the algorithm is the conformal hexahedral mesh that was generated in the first step. The
other input is the cell types contained within a library.
The first step is to partition the mesh elements into
regions such that within each region the loading conditions are similar on each element. These need not
be the same regions that were used for mesh construction. For each region, a cell type from the cell type
library is selected to populate the mesh elements in
that region. The idea is to match the regions loading
conditions to the cell type, such that the cell type is
effective at supporting the loading conditions. In this
manner, the resulting cellular structure is more likely
to be lighter for a given level of stress or deflection.
The dashed gray box surrounding the Select Mesh
Elements and Select Cell Types operation boxes
indicate that these operations may be performed concurrently or maybe performed sequentially, depending
upon the designers preference. Note that each of the
Select operations may be performed automatically by
an algorithm or may be performed by the designer
directly.
The final step in Figure 6, Apply Selected Cell
Types to Selected Mesh Elements, is where the actual
cellular geometric model is constructed. This operation has been called population of mesh elements
earlier. This operation is straightforward. One simply
maps a cell type into a mesh element. Since both the
cell type and the mesh element are defined parametrically, a simple parametric mapping algorithm can be
applied to directly construct cell geometry.
Figure 6. Algorithm to compute the conformal cellular

geometric model.
To construct a STL or solid model of the CLS,

additional geometric construction operations must be
performed. We utilize the approach described in (Wang
et al. 2005), where solid models of the half-struts incident at each node of the mesh are constructed using
Boolean operations is a solid modeling system. Then,
each solid node is tessellated and the triangles are
written to a STL file.
The resulting conformal cellular geometric model
can be subjected to optimization methods in order to
reduce weight, increase strength, increase stiffness,
achieve some other objective, or achieve some combination of objectives. Such methods have been applied
by a number of research groups (Chu et al. 2010,
Patel & Choi 2009).
EXAMPLE OF CLS CONSTRUCTION

METHODS
As an example, a simple cylindrical surface will be

stiffened by one layer of conformal lattice structure
using the method presented in Section 3. Figure 7a
shows the original cylindrical surface defined by a STL
file. The surface is 2 units in diameter and 2.5 units
long. Constructed lattice structure is intended to be
0.25 units on a side. With an angular deviation measure of 50 degrees, the surface is partitioned into three
regions, as shown in Figure 7b. A quadrilateral mesh
is created in each region, then the mesh boundaries
are adjusted and merged. The result is shown in Figure 7c, where the different colors indicate the different
regions. After offsetting the surfaces and mesh, the
tri-parameter volumes and hexahedral mesh can be
constructed (Figure 7d). The final step is inserting lattice geometry into each mesh element, resulting in the
CLS model in Figure 7e.
206
Figure 7. Example of CLS construction method.
SLS FABRICATION
The concept of reinforcing components with CLS has

an inherent fabrication challenge due to the complexity of the geometries and features that are impossible
to manufacture via conventional fabrication process.
Indirect casting techniques can generate some complex CLS from a master pattern, but the approach
is very expensive, time consuming, and requires a
lot of trials before an accurate part can be achieved.
SLS technology has become one of the most acceptable additive manufacturing technologies for the
fabrication of end-use parts and functional components. In addition, the increasing number of composite
materials available for SLS made this technology
and ideal candidate for the fabrication of engineered
components optimized with CLS.
Paramount Industries Inc. identified a part on an
unmanned air vehicle (UAV) with properties and a
geometry suitable for reinforcement and optimization
with CLS. The part has a characteristic saddle surface
and serves as a hatch-cover on the UAV. The CAD files
of this part were optimized using the design algorithms
from Georgia Tech. Figure 8 illustrates the hatch cover
207
Figure 8. Paramount Industries Inc. Hatch Cover Component Reinforced with CLS.
Figure 10. Interface Flowchart.
Figure 9. Testing SLS hatch cover part.
files ready for SLS fabrication after optimization with

CLS.
Paramount Industries Inc. successfully fabricated
the hatch cover via SLS and developed internal processes to overcome the many significant challenges
when considering SLS for the fabrication of CLS.
The SLS process has 4 variables that are critical when
the requirement is repeatable mechanical performance
from a truss-like structure made out of struts that are
1 mm thick. The most critical variable is the spatial
orientation of the CLS during fabrication. Any CLS
has an optimal orientation in the x-y-z plane which
yields the best mechanical response when compared
to FEA models performed prior to fabrication. The
next three most important variables of our SLS process are laser power, scan-speed or hatch speed, and
scan-spacing or hatch distance. Other critical process
variables include the material quality, heat distribution,
laser beam profile, the control of scales-and-offsets,
and outline settings.
Several sets of hatch-covers were fabricated. Tests
were performed on the hatch covers by compressing
the center of the plate in an Instron machine, as shown
in Figure 9. Early sets exhibited significant inconsistencies which were determined to originate from
multiple process variables. Using infrared imaging

equipment and thermocouple sensors it was determined that the build envelop and the machine had hot
and cold spots that had to be improved before uniform
properties could be achieved. It was also determined
that the parts on top created thermal gradients that
affected the mechanical properties of the parts at the
bottom of the cake.
The next sets of hatch covers were fabricated after
all the process parameters and variables were dialed-in,
optimized, and measured.These hatch covers exhibited
very consistent mechanical properties. Their flexural
load vs. deflection characteristics were within 5 percent of one another. The flexural load also increased
from about 120 psi to 270 psi.
INTEGRATION INTO CAD
The overall procedure for using TrussCreator in UG

NX is shown in Figure 10. For each step, dialog boxes
are designed and created for user input. The user can
access these dialog boxes through the TrussCreator
menu (Figure 11). In the general parameter setting
menu, there is a dialog box for number of parts, tolerances and faceting parameters for selected surfaces
(Figure 12a). In the lattice parameter setting menu, the
user can input information about diameter of struts,
the size of unit cells, strut tweak parameters (for
adjusting node positions relative to the surface), and
partition angle (step 1 in Figure 5) for given surfaces.
Figure 12b shows a dialog box for lattice settings.
After entering the parameter settings, the user can
select the lattice structure type from a set of eight
types (more are being added). In surface selection
stage, the class selection dialog is loaded that makes
selection for only surfaces in the current working
208
Figure 11. TrussCreator Menu.
Figure 12. Dialog box for parameter setting.
Figure 13. Surface Selection Dialog Boxes.
part (Figure 13a). The user can select and deselect

the surfaces in NX CAD system. The API function
UF_UI_select_with_class_dialog() is used for the dialog box.After selecting the surface, the user can choose
what direction to add the lattice structure. The arrow
indicates the normal direction in which the lattice
structure will be added; the direction can be flipped by
selecting the surface that user wants to change (Figure 13b). Selecting surfaces and their directions can be
repeated until the user has included all of the surfaces
s/he wants. TrussCreator generates lattice structures
based on input information mentioned above. After
generating the lattice structure, it is displayed with its
parent surfaces in the NX environment.
209
mechanical properties of CLS are highly dependent

on appropriate part orientation, as well as process
variables of laser power, scan speed, scan spacing,
and several others; by fine-tuning these variables it
is possible to repeatedly build arrays of CLS parts
with consistent, good mechanical properties.
Future work includes improving the CAD system user interface and developing additional lattice
structure types. More investigation into the SLS
fabrication of CLS is needed so that little, if any,
experimentation is required to build a new CLS
design.
REFERENCES
Figure 14. Example of lattice structure; top: line graphics,

bottom: STL file display.
The user can look at the detail of the lattice structures using NX functionality and save the structure as
an STL file. Lattice structure and its STL version are
shown in Figure 14.
6
CONCLUSIONS
Methods to construct cellular material structures,

specifically conformal lattice structures, were presented in this paper. Their fabrication using SLS
was described and it was described how process
conditions and settings have a significant impact
on mechanical properties of fabricated structures.
Based on this work, the following conclusions can be
made:
the presented CLS construction methods produced

lattice structures on a wide variety of surface shapes,
demonstrating the generality of the methods; these
methods produced CLS that was of high quality,
since the unit cells were of nearly uniform size;
the geometric construction methods resulted in
standard STL files that have been built in several AM machines, including stereolithography and
SLS machines;
Ashby, M. F., Evans, A., Fleck, N. A., Gibson, L. J.,

Hutchinson, J. W. & Wadley, H. N. G. 2000. Metal Foams:
A Design Guide, Woburn, MA: Butterworth-Heinemann.
Cantley, R. W. Molded Plastic Truss Work, US Patent
6,993,879.
Chen, Y. 2007. An accurate sampling-based method
for approximating geometry. Computer-Aided Design
39(11): 975986.
Chu, C., Engelbrecht, S., Graf, G. C., Rosen, D. W. 2010.
A Comparison of Synthesis Methods for Cellular Structures with Application to Additive Manufacturing. Rapid
Deshpande V. S., Fleck N. A. & Ashby M. F. 2001. Effective
Properties of the Octet-Truss Lattice Material. J. Mechanics and Physics of Solids 49(8): 17471769.
Engelbrecht, S. S. 2009. Design of Meso-Scale Cellular
Structure for Rapid Manufacturing, MS thesis, Georgia
Institute of Technology.
Gibson, L. J. & Ashby, M. F. 1997. Cellular Solids: Structure
and Properties, Cambridge: Cambridge University Press.
Gibson, I, Rosen, D. W., Stucker, B. 2010. Additive Manufacturing Technologies: Rapid Prototyping to Direct Digital
Manufacturing, Springer.
Patel, J. & Choi, S. 2009. Optimal Synthesis of Mesostructured Materials under Uncertainty, 50thAIAA/ASME/
ASCE/AHS/ASC Structures, Structural Dynamics, and
Material Conference and 5th AIAA Multidisciplinary
Design Optimization Specialist Conference, Palm Springs,
May 47.
Rosen, D. W. 2007. Computer-Aided Design for Additive
Manufacturing of Cellular Structures. Computer-Aided
Design & Applications 4(5): 585594.
Wang, H., Chen, Y. & Rosen, D. W. 2005. A Hybrid Geometric Modeling Method for Large Scale Conformal
Cellular Structures. In ASME Computers and Information in Engineering Conference, paper DETC2005-85366,
Long Beach, CA, Sept 2428.
Wang, H. & Rosen, D. W. 2002. Parametric Modeling Method
for Truss Structures. In ASME Computers and Information in Engineering Conference, paper DETC2002/CIE34495, Montreal, Sept. 29Oct 2.
210

No models, no moulds!
Alberto T. Estvez
ESARQ (UIC), Barcelona, Spain
ABSTRACT: Working with artificial DNA (software), with computing elements, with application of real
cybernetic processes to architecture, for automation of the robotized production of architecture digitally designed:
Digital design and production seen as a genetic process. Knowing that what can be drawn can be built, because
what can be drawn using digital tools has a digital DNA, that allows automated emergence, robotized selfconstruction and artificial growth. Using digital technologies for not producing more models or moulds as is
habitual in todays production systems [NO MODELS, NO MOULDS!], but to produce real architecture at the
natural scale of 1:1, illustrated with the Barcelona Biodigital Pavilion, Barcelona Consulting rooms, Biodigital
Barcelona Chair, Biodigital Barcelona Furniture, Biodigital Barcelona Lamps, etc. This is a move beyond the
mass production of uniform elements, since digital design and production can equally produce 100 identical or
100 different parts.
1
1.1
NO MODELS, NO MOULDS!
Introduction
No models, no moulds! is a Manifesto, is a cry, is a

desire of advanced research in virtual and rapid prototyping that our entire world needs, for sustain-ability
and efficiency using new biodigital technologies, and
is what our Genetic Architectures Research Group,
Biodigital Architecture Masters Degree and Ph.D.
Program wants: Related with topics like Biomimetics,
Genetics and Bionic, Design for Bio Manufacturing & Design for Sustainability, Architecture and Art
in application of biological and digital techniques,
Model-making drawn from academia and professional
practice. This is what this article contains: the presentation of new and innovative work that we do,
from School and Office, from different aspects and
some examples, Architecture, Design and Art digital approaches from Nature to digital fabrication
(Estvez, 2002, 2003, 2010a, b, c).
1.2
New bio & digital techniques
Thanks to new bio & digital techniques, human beings

can transcend a historical barrier, from only millenary action over the order of the surface, to a new
action over the molecular order (Estvez, 2005a, b).
Now humans can work on an intramolecular level,
towards the knowledge of the genetic orders of order:
when the general visible order of bio & digital beings
is controlled by bio & digital informations chains
(Estvez 2009a, b). Working on this level, one of
the main advantages is that the control of this informations chains allows the structure, form and skin
emergence, with biological or digital processes, to
become architecture. Done with materials that emerge,
that grow alone thanks to self-organization systems, biological or digital, towards more precision,
more efficiency, more sustainability. Also with the
possibility of fusion & link of the mentioned informations chains in biodigital architecture. Is a vast
potential in a biological world if we work with DNA
as if it was biological software, and in a digital realm
if we work with software as if it was digital DNA:
This is the most peculiar application of genetics in
architecture. Understanding that DNA and software
are the same, informations chains, natural or artificial, that produces orders (order) for self-organization,
for autonomous growth, for emergence processes, for
structure, form and skin emergence. Knowing that
what can be drawn can be built, because what can be
drawn using digital tools has a digital DNA, that allows
automated emergence, robotized self-construction and
artificial growth. And more than this today: what can
be captured (digitally) can be built. You dont need
to draw anymore, if you can capture digitally something, even if it is impossible to draw, it is enough to
make it digitally. We have seen it already done: you can
build an instant, a movement, a twister, because you
can capture it. Using digital technologies for not producing more models or moulds as is habitual in todays
production systems [No models, no moulds!], but to
produce real architecture at the natural scale of 1:1.
1.3 For the first time, architects are waiting

engineers
As the machines we have in our Digital Fabrication
Workshop are of the known type of CNC machine,
3D Rapid Prototyping and laser cutter, we need to do
a supplementary research for building in one to one
natural scale without machines specially designed for
211
Figure 1. After biolearning process, CAD-CAM technologies for producing directly real 1:1 scale architecture.
Left, digital drawing; centre/right, real CNC digitally fabricated and real installed in big architectural scale. Images:
Alberto T. Estvez, Biodigital Barcelona Pavilion,
Barcelona, 200809.
Figure 2. After biolearning process, CAD-CAM technologies for producing directly real 1:1 scale architecture,
from genetic architectures points of views. Left, digital
drawing; centre, real CNC digitally fabricated (with human
scale behind) and real installed with Biolamps application;
right,interior view. Images: Alberto T. Estvez, Consulting
rooms, Barcelona, 200809.
Figure 4. Left, digital drawing; above right, microscopic

biostructures research done with scanning electron microscope; below right, CNC digitally fabricated and installed in
big architectural scale, conformed by real grass. Drawings
and photos: Alberto T. Estvez Alejandro Muio Diego
Navarro, Biodigital Barcelona Chair, Barcelona, 2010.
Figure 5. Left, digital drawing; right, digitally fabricated

lamp at our Digital Architecture Workshop with 3D Printer.
Drawings and photos: Alberto T. Estvez & Diego Navarro,
Biodigital Barcelona Lamps, Barcelona, 2010.
1.4 Examples
Figure 3. Pollen details, left, 6000x, and centre left, 20000x,

made with scanning electron microscope. Centre right and
right, CNC digitally fabricated natural scale 1:1 ceiling with
Biolamps application. Photos: Alberto T. Estvez.
(The mentioned Biolamps of the figures above are from our
genetic architecture research, from our work with geneticists
at our Genetic Architectures Laboratory, searching the real
application of genetics to architecture: in this case research
about real application of bioluminescence.)
architecture. How many time architects need to wait

that engineers prepare the machines that architecture
needs for it? This time architects go ahead of engineers
in the visualisation of new constructions, not as the
last centuries, because we know today exactly what
we want, but engineers havent prepared tools really
applied for the scale that architecture need.
When I have seen the big 3D printer that has
the entire building scale, Contour Crafting, I have
asked quickly to the engineer that had created it,
Behrokh Khoshnevis, located in USA: I want to
build the first entire 3D printed building of Spain
(of Europe), how much costs your machine? Half
million Euros Ok, nothing to do with our South
European research standards.
Illustrated on this pages, for Architecture with the

Biodigital Barcelona Pavilion (Fig. 1), for Interior
Design with the Barcelona Consulting rooms (Fig. 2
and 3), for Furniture Design with the Biodigital
Barcelona Chair (Fig. 4), for Objects Design, with the
Biodigital Barcelona Lamps (Fig. 5) and for teaching
(Fig. 6). This is a move beyond the mass production
of uniform elements, since digital design and production can equally produce 100 identical or 100 different
parts. And also illustrated for Art, wanted to integrate
3D pieces digitally produced from the images obtained
after the research with the scanning electron microscope in the foreground of a photos series called still
alive (Fig. 7).
RESEARCH AND PROFESSION
As it was discussed, we need to research first how to

draw digitally and how to produce digitally, with the
machines that we have, the architecture that we want at
the natural scale of 1:1. Escaping from the temptation
to do it manually, because sometimes it seams easier to do it in a manual way. This becomes a question
of rigor and research discipline, to maintain the challenge of the digital production: the effort of arriving at
a successful digital DNA. Then, the digital emergence
advantages will give the advanced difference. This is
what we research and do, at our Genetic Architectures
212
4 AS CONCLUSION, AS EPILOGUE . . .
Figure 6. Academic models digitally fabricated at our Digital Architecture Workshop with 3D Printer. Photo: Alberto
T. Estvez, Biodigital Barcelona Islands, Barcelona, 2011,
designed by Homer Jos Garca Santana and Serra Avsever,
Biodigital Architecture Masters students at Prof. Alberto T.
Estvez Studio.
Figure 7. Different images from an integration (in the foreground of a photos series) of 3D pieces digitally produced
from the images obtained after the research with the scanning electron microscope. Photo: Alberto T. Estvez, Still
alive series, 2010.
Research Group, Ph.D. Program and Office (founded

in ESARQ, UIC, Barcelona, year 2000). Beginning
also with a microscopic research, for example about
radiolarian and pollen structures, like natural systems
that have been perfected during millions years of evolution, until their high relation of efficiency/economy:
heir application will take also these advantages. In this
way started the studies for extracting the genetic rules
and the structural parameters for application with digital tools. Having its digital DNA that let also the
architectural design emerge alone, we have produced it and fabricated digitally, with the effort to
fulfil with the manifesto No models, no moulds!
Only directly digital produced elements in real scale
1:1 are allowed: after biolearning process, CADCAM technologies, CNC machines, 3D Printers, for
producing directly real 1:1 scale architecture, from
genetic architectures points of views.
3 TEACHING
When we talk about an academic context, at our
Biodigital Architecture Masters Degree and Ph.D.
Programme (founded in year 2000, like the first systematic official post-graduate programme on these
subjects, at ESARQ, Universitat Internacional de
Catalunya, Barcelona) we allow models (Fig. 6), only
waiting for engineers. We allow the advanced research
in virtual and rapid prototyping of models: knowing
and explaining the nowadays availability limitations of
todays technology. Yes, the architecture is waiting for
the engineering, far away for an easy real application of
digital fabrication today in the conventional development of our profession as architects. At the same time
that, of course, we teach also research objectives for
producing real architecture at the natural scale of 1:1.
But its not enough to work with the last digital techniques. The change to the fusion of the biological
techniques with the digital techniques must be initiated, and seams that artists go ahead of architects.
Humanity has the responsibility of having a future.
And this will only happen across biodigital architecture, which will use the advantages that are given by the
new biological and digital techniques. In fact, genetic
engines are the ones that move both, similar genetic
principles that are in the basis of biological and digital
(Estvez 2007a, b).
As the expressionists of the beginning of XX Century saw on the Christ of Grnewald a precedent, the
geneticists of the beginning of XXI Century must see
in the Garden of El Bosco the same precedent, both
with half a millennium of antiquity. It is normal that
starting with this change I am addressing to a qualified
forum that knows already about digital techniques.
The forum that must begin to be worried for crossing it with biology, genetics, real life, not only virtual,
which was a necessary first step. An effort is needed
for reaching maturity also with the use of biological
techniques applied to architecture, in the way that also
The Bioplasticity Manifesto justified. For not losing
a precious time, I hope that this wont be a question
for only new generations
Now that extended reality and augmented reality
are fashions ideas, for an understanding of ourselves
in front of nature, when we can take profit of genetics,
we can think that human beings becomes extended
nature and augmented nature.
From one side, every form, structure and skin in
nature follows a function, yes, but across infinite variations (biodiversity) for the same physicalphysiological function, for the same necessity, like
digital techniques already allow. And we, like human
beings, after the physical-physiological necessities,
have psychological-emotional necessities (often existentially most important) that, on the other hand, art &
architecture need also to solve: artistic and architectural idea & form follows function, necessities that
only human beings have, the most human functions.
At the end this is an efficient justification against
the superficial criticism that are normal about digital architecture, about the new possibilities that digital
techniques allow.
These are also the functions that art & architecture
need to follow: not more form only follows function, in
the physical-physiological dogmatic way that rationalfunctionalism have understand. We can understand
that, as human beings, we have other functions, other
extended and augmented necessities that animals
and plants dont have. But we are also nature,
and in our functions of solving necessities ourselves
and what we can do with genetic techniques become
extended nature and augmented nature
Alberto T. Estvez (Barcelona, 1960), Architect
(UPC, 1983), Architecture Ph.D. of Sciences (UPC,
1990), Art Historian (UB, 1994), Art History Ph.D.
213
first time-geneticists working with architectural objectives, in a real application of genetics to architecture.
He is now Director of the Biodigital Architecture
Masters Degree and of the Genetic Architectures
Research Group & Ph.D. Program, ESARQ(UIC),
Barcelona (www.albertoestevez.com).
REFERENCES
Figure 8. Left above and right below, microscopic biostructures research done with scanning electron microscope; right
above and left below digital drawing. Images: Alberto T.
Estvez & Diego Navarro, Kindergarten, Vilob del Peneds,
2009 (for CNC digital fabrication).
Figure 9. People using the soft and furry Biodigital

Barcelona Chair, here installed in CCCB (Barcelona), 2010:
for a soft and furry architecture!, exactly what human beings
like extended nature and augmented nature really need.
Photo: Alberto T. Estvez.
of Arts (UB, 2008): With professional office of

architecture and design (Barcelona, 1983-today: see
examples Figures 8, 9). Teaching in different universities for more than 25 years, in the knowledges
areas of architectural design, architectural theory
and art history, until founding the ESARQ (UIC,
1996), where he teaches like Professor in Architecture, after being its first Director. He founded also
two research lines with two masters degrees and
Ph.D. programmes: Genetic Architectures/Biodigital
Architecture (UIC, 2000-today) and History, Architecture and Design (UIC, 1998-today). He has written
more than one hundred publications and has participated in a large number of exhibitions, congresses
and conferences around Europe, America and Asia.
He has created the worlds first Genetic Architectures Laboratory (UIC, Barcelona, 2000), with-for
Estvez, Alberto T. 2002. Genetic Architectures. In VV.AA.

Memorias: 406409. Orlando (USA): CISCI.
Estvez,AlbertoT. 2003. GeneticArchitectures/Arquitecturas
genticas. In VV.AA. Genetic Architectures/Arquitecturas
genticas: 417. Santa Fe (USA)/Barcelona: SITES
Books/ESARQ (UIC).
Estvez, Alberto T. 2005a. Biomorphic Architecture/
Arquitectura biomrfica. In VV.AA., Genetic Architectures II: digital tools and organic forms/Arquitecturas
genticas II: medios digitales y formas orgnicas: 1880.
Santa Fe (USA)/Barcelona: SITES Books/ESARQ (UIC).
Estvez, Alberto T. 2005b. Genetic Barcelona Project/
Proyecto Barcelona Gentica. Metalocus, 17: 162165.
Madrid.
Estvez, Alberto T. 2007a. Arquitecturas Genticas: la casa
perfecta, o una casa no es una caja In VV.AA.,
Arte, Arquitectura y Sociedad Digital: 1 and 117
122. Barcelona: Publicacions i Edicions Universitat de
Barcelona.
Estvez, Alberto T. 2007b. The genetic creation of
bioluminescent plants for urban and domestic use.
Leonardo, vol. 40, num. 1: 18 and 46. San FranciscoCalifornia/Cambridge-Massachusetts (USA): The MIT
Press.
Estvez, Alberto T. 2009a. Al margen: Escritos de Arquitectura. Madrid: Abada.
Estvez, Alberto T. 2009b. Genetic Architectures III:
new bio & digital techniques/Arquitecturas genticas III: nuevas tcnicas biolgicas y digitales. In
VV.AA., Genetic Architectures III: new bio & digital
techniques/Arquitecturas genticas III: nuevas tcnicas
biolgicas y digitales: 1433. Santa Fe (USA)/Barcelona:
SITES Books/ESARQ-UIC.
Estvez, Alberto T. 2010a. Genetic Architectures Research
Group, ESARQ (UIC). In VV.AA., eme3_2010: International Architecture Festival: 7879. Barcelona: eme3/
CCCB/MACBA.
Estvez, Alberto T. 2010b. Application of Life information
in Architecture: Biodigital Architecture and Genetics. In
VV.AA., LIFE information/On Responsive Information
and Variations in Architecture ACADIA2010: 168173.
New York (USA): ACADIA.
Estvez, Alberto T. 2010c. Still alive, landscapes and
other fleshinesses/Naturalezas vivas, paisajes, y otras
carnosidades. Barcelona: ESARQ (UIC).
214

Touchless gesture user interface for 3D visualization using the Kinect

platform and open-source frameworks
G.C.S. Ruppert, P.H.J. Amorim, T.F. Moraes & J.V.L. Silva
Renato Archer Center for Information Technology, Campinas, Brazil
ABSTRACT: In this work we present a natural user interface system using hand gestures to provide touchless
control of 3D visualization software using the Microsoft Kinect device. Kinect is a relatively low-cost device
developed by Microsoft for the gaming industry to provide controller-free gaming, but it has potential applications
in many other areas, such as: data visualization, augmented reality, accessibility and robotics. Although our
solution can be used with any 3D software, the present work is primarily focused in one particular application
which is the visualization of 3D medical images during a surgical procedure. The operating room is a cleansed
and sterilized environment and the contact of the surgeon with traditional computer interfaces (like mouse and
keyboard) could lead to contamination increasing the risk of patients infection. A touchless interface is an ideal
solution since it does not demand any physical contact and still can provide the necessary control features.
For the visualization of 3D medical images, we used the InVesalius software that provides 3D reconstruction
of medical images with features that enable the rapid prototyping of medical models and provides high-quality
visualization of 3D volumes. This software is open-source, cross-platform, multi-language and is freely available.
Using open-source software libraries and image processing techniques, we implemented the hands tracking and
gestures recognition from the images provided by the Kinect device and enabled the surgeon to successfully
navigate through the image.
1
INTRODUCTION
Computer interfaces based on gestures have been

intensively researched (Kadobayashi 1998, Keskin
2005) but limitations like high cost, bad accuracy
and setup complexity have contributed to keep this
technology impractical for real world applications.
This situation changed drastically recently with the
release of a low-cost, small and easy-to-setup device
by Microsoft, called Kinect.
Gesture user interfaces have applications in many
areas, but the present work is primarily focused in
one particular application which is the visualization of
3D medical images during a surgical procedure. This
scenario is particularly important because the operating room is a cleansed and sterilized environment and
the contact of the surgeon with traditional computer
interfaces (like mouse and keyboard) could lead to
contamination increasing the risk of patient infection.
Before initiating the operation, surgeon and nurses
are required to follow a strict hand scrubbing procedure which takes some minutes, and have to be
repeated whenever they get in touch with any external
material that can possibly be contaminated.
Also, during the surgical procedure, the physician
often needs to examine the medical images. This is
traditionally done using previously printed images on
the negatoscope. However, medical image software
can provide a far richer and more suitable visualization of medical images, not only providing a better
view but also providing important information such

as distance and angle measurements, tissue densities
and others. Nevertheless, these software are usually
controlled using mouse and keyboard, requiring an
undesired physical contact during the surgery.
In this work, we propose a solution to this problem, by presenting a touchless gesture user interface
that allows the surgeon to control the medical imaging
software InVesalius just by performing hand gestures
in midair. The system is also open-source, low-cost
and simple to deploy which allows this solution to be
widely adopted by surgeons and hospitals allowing the
use of medical image software in the operating room
and yet reducing the risks of contamination. Although
we just focused on the medical scenario, the solution
can also be applied as a replacement of the mouse to
any other software like: CAD, 3D modeling, games,
media centers and web browsers.
The paper is organized as follows. Section 2
presents the Kinect device in details. The software
InVesalius is presented in Section 3. The description
of the proposed solution is presented in Section 4.
And the last Section presents the conclusion and
futures work.
2
KINECT
Kinect is a relatively low-cost device developed

by Microsoft for the gaming industry to provide
215
Figure 2. The RGB camera image (a), the image captured

by the IR camera (b), and the depth image in colormap (c).
Figure 1. Kinect main components.
joystick-free control of the Xbox 360 console just

using body movements. However, given the high number of embedded sensors, this device has a great
potential of application in other areas, such as data
visualization, augmented reality, accessibility and
robotics.
The device is connected to the computer through the
USB port and provides an RGB camera (color camera)
and a depth sensor, both with a resolution of 640 480
and a frame-rate of 30 fps. It also provides a multiarray microphone that enables the Kinect to perform
acoustic source localization and ambient noise suppression. In addition, it features a motor on the base
that can control the tilt of device in order to try to keep
the players inside the camera angle of view. Kinect
has also a 3-axis accelerometer that is used mainly for
detecting if the sensor was moved, that may require
new calibration of players.
All the higher level features attributed to Kinect, like
objects, players and gesture recognition, are not really
performed by the Kinect device. The device itself
just provides the mentioned sensors, leaving all the
higher level features to be executed in the videogame
by software that implement computer vision methods.
Despite of having all the mentioned sensors, the feature that makes this device really novel and interesting,
especially in the context of this paper, is the depth sensor. Since the device provides the depth information
combined with the RGB camera, the whole system
becomes a powerful, accurate and fast computer vision
device.
2.1
Depth sensor
The popularly called depth sensor of Kinect, can be

seen as a depth camera and provides a 11-bit singleband image (i.e., from 0 to 2048). In this image,
the intensity of each pixel is inversely proportional
to the depth, which in this case is the distance from the
device to the object on the scene in the corresponding pixel position. The depth range that this sensor can
detect ranges from 60 cm up to more than 8 meters,
but the relation between the pixel intensities and the
actual depth is exponential, which means that the depth
differences are more precisely detected closer to the
device, therefore giving an effective working range
between 1.0 and 3.5 meters.
The depth detection implemented in Kinect is provided by an infrared (IR) projector and an infra-red
camera. The projector projects a fixed structured
light, which in this case consists of a pattern with
millions of small dots. The reflections of this pattern

is captured by the infrared camera. Figure 2 shows
the projected pattern in the scene. The microprocessor inside Kinect then performs stereo triangulation
(Zhang 2008) between the image captured by the IR
camera and the original pattern, obtaining the depth
information for each point on the field of view. The
output of this triangulation is interpolated and sampled
to a discrete space, which is the depth image. Kinect
does not use the RGB camera or any other source of
information to generate the depth image.
This ability to obtain the depth information of the
scene is extremely powerful for a computer vision
system. The automatic recognition and delineation
of different objects in the scene using regular RGB
images is very hard but using the depth images this
task becomes much more trivial and robust.
3
INVESALIUS
InVesalius is an open-source software for visualization

of 3D medical images developed by the Renato Archer
Information Technology Center. This software enables
the accomplishment of diagnostics, surgery planning, elaborate prosthesis, scientific studies, forensic
analysis, among other applications. The name of the
project is a tribute to the belgium physician Andreas
Vesalius (15141564), considered the founder of the
modern human anatomy.
InVesalius is free software. It is free of charge and
its source-code is open for download in the official
website1 . It is multi-platform (Windows, Linux and
MacOS) and also multi-language (english, portuguese,
spanish, french and many others). The software features an online community of more than 5000 users
and it has been downloaded by more than 70 countries.
Because it is an open-source software, this software
also has a great potential to serve as basis for the
development of other scientific projects.
Developed in Python language, the software uses
other open-source libraries such as VTK, ITK,
Wx-Widgets, GDCM and Numpy. It allows importing
images in DICOM (DICOM 2009) and also the Analyze formats, which are the most common standards,
and enables the visualization in 2D slices (multi-planar
reconstruction) and also the 3D volume using a high
quality volume rendering. The user can also segment
tissues of interest and generate a surface rendering
1
216
http://www.portaldosoftwarepublico.gov.br
Figure 3. Screenshots of InVesalius showing different 2D and 3D visualization of a cranial image.
visualization, as well as exporting these surfaces in

STL file format which can be materialized using
the technology of rapid prototyping (3D printing). The
software also offers measurements tools, enabling the
user to perform linear and angular measurements on
the 2D and 3D views.
4
GESTURE USER-INTERFACE
In this work, we present two prototypes of gesture

interface we developed using two different technologies to allow the Kinect to control InVesalius. The
first one uses a simpler but completely open-source
solution based on depth threshold to detect the hand.
The second prototype uses a non-opensource skeleton recognition software to detect the position of the
hands.
In both versions, the hand of the user is tracked
and the corresponding position is used to move the
mouse cursor. So, although we used InVesalius in our
scenario, the proposed gesture interface can be used
with any visualization software that can be controlled
by the mouse movements. The next sections describe
each of these solutions.
4.1
Libfreenect
Kinect was released just for use with the Xbox game
console and no official support (drivers, software,
library, etc) was provided by Microsoft to connect this
device to a computer. But just a few days after the
release, some developers were able to hack the system
and release drivers to have access to all sensors from a
personal computer. This project is named Libfreenect
and was written in C, however it has wrappers to all
most popular languages.
Basically this library offers basic functions to get
the RGB and depth images, as well as to get and/or
set the tilt motor angle, the color of the front led, and
also read the accelerometer values. The images are
provided through callback functions that are called
for every frame and are assigned using the functions freenect_set_depth_callback() and freenect_set_
video_callback(). These functions then receive a
pointer to the buffer where the image is stored. The
function has to operate over this image in the timeframe of 33 ms to handle the 30 fps frame-rate of
Kinect, otherwise frames are dropped.
Figure 4. The colour image (left) and the result for

the threshold in the depth image (right), followed by
post-processing operations (in red) and center of gravity
computation.
Libfreenect does not offer any higher-level functionally like user recognition and hand tracking. All
this features have to be implemented by the software
using libfreenect.
The first prototype we implemented uses libfreenect
and captures only the depth image and applied a threshold according to a predefined distance to the person in
a way to get just a 50 cm range in front of the persons
body. So when the user moves the hand in front of
the body, just the hand pixels are selected. However, a
post-processing step is required since the result of this
threshold is still noisy, unstable and present outlier
pixels. We applied a minimum component filter to
remove outlier regions smaller than 10 pixels and a
mean filter (Gonzalez 2007) over time was applied
between frames to get better stability on the hand
detection.
After the post-processing steps, the center of gravity of the remaining image is calculated, giving the
coordinate of the center of the user hand. This position
was used to generate a virtual mouse event moving the
mouse cursor to that relative position. Figure 4 shows
the processed image.
The mouse press event also is virtually generated
when the user holds still the hand in the same position
for 1 second.As it is almost impossible to hold the hand
perfectly still, a displacement of up to 20 pixels was
considered as no movement. The mouse press event is
generated when the user removes the hand back outside
the threshold area.
This solution showed to be effective in terms of
successfully enabling the user to control the visualization features in the software. Tasks like controlling
3D rotation, zoom, slice changes, and contrast adjustments could be successfully performed. The drawback
of this solution is that the user has to be located in an
217
Figure 5. The OpenNI architecture. The tracking algorithms

and the hardware devices are abstracted for the application.
specific spot so the threshold can separate the hands

movements from the rest of the scene. Also, the solution is dependent on the arms length of the user,
requiring different calibrations.
4.2
OpenNI/NITE
OpenNI and NITE are software components provided

by PrimeSense, which is the company that initially
developed the Kinect technology which later was
transferred to Microsoft. Both are multi-platform and
free of charge, but although OpenNI is distributed as
open-source, NITE is closed and distributed only as a
binary executable.
OpenNI (Open Natutal Interaction) is an opensource framework that provides a middleware for
programming natural interfaces. It is capable of
abstracting the tracking algorithms and the hardware,
as shown in Figure 5. It provides common programming interface for developing skeleton and hand tracking algorithms (plugin components) with a predefined
list of body joints that the algorithms can provide.
However, OpenNI does not come with any algorithm
already embedded, so it is necessary to install these
components so OpenNI can actually be useful.
PrimeSense also distributes freely the package
NITE, which are OpenNI-compliant modules that
implements real-time skeleton and hand tracking
methods. Although not officially revealed, these
algorithms are likely to be similar to the ones that
PrimeSense and Microsoft included in the Xbox
Kinect middleware. Unfortunately, this package is not
open-source and the internals of the algorithms are not
revealed. These algorithms use probabilistic templates
of the human body shape (based on thousands of modeled bodies samples) to track the users in the image. It
is capable of detecting and segmenting different users
on the scene and can also track 15 body joints (hands,
elbows, shoulders, neck, head, torso, hips, knees and
feet) in real-time with a impressive robustness and
accuracy. So for each frame, a 3D coordinate of each
joint can be obtained. Figure 6 shows one example of
the skeleton being tracked.
We implemented a gesture interface that uses
OpenNI and NITE to track the skeleton of the user
Figure 6. User detection and skeleton tracking using

OpenNI/NITE. The first user (in blue) is being tracked and
the skeleton joints are shown connected with a white line.
and generate mouse event (move and clicks) based on

the position of the joints. The right hand of the user
controls the position of the mouse cursor. As we have
all the joints coordinates, we can calculate the length
of the arms and calibrate the range of the hand movement to fit each user. Furthemore, the cursor position
is calculated as the relative position of the users hand
to his torso joint position, so the user can be anywhere in the camera angle of view and he can even
move around the room, and the calibration remains the
same.
The user can also use his left hand to generate virtual mouse clicks. Raising the left hand up to the
level of his torso, a right button press event is generated. When the user returns the left hand down to
relaxed position, a left-button release is performed.
This feature allows the user to perform 3D rotations,
2D slices change and other toolbar selections. When
the user raises the left hand up to the head level, a
right-button event is generated, enabling the user to
control the zoom. These combinations of the left hand
position are easy to detect just by comparing the joint
coordinates.
Using this approach, the user is able to successfully
control all visualization features of InVesalius. The
user experience in this setup is better than the solution
we presented on section 4.1, because the responsiveness is more stable, does not require the user to be
fixed in the same spot and is much more robust. Also,
the ability to track both hands and generate right and
left-button clicks with the left hand showed to be
efficient in the usability of the interface. The main
drawback of this solution is that the package NITE is
not open-source which makes this solution subject to
the availability provided by PrimeSense.
CONCLUSION
In this work, we presented a touchless user interface

solution to enable a surgeon to control the software
218
InVesalius in an operating room just using hand gestures. The Kinect device showed to be very efficient
and enabled a very low-cost and easy solution. Two
prototypes were developed using different software
approaches and, in our experiments, both succeeded
in providing the functionality. However, the solution
using OpenNI/NITE to track the skeleton and the
hands was far more robust and easier to use.
Although we were focused on the surgery application, this solution can potentially be used with any
software that can be controlled by the mouse, or it
could be adapted to generate other kind of events to
control other software.
As future works, we plan to integrate the OpenNI/
NITE protype into InVesalius and explore different
combinations of gestures to provide easier experience.
Furthemore, other technologies like fingers tracking, face detections and speech recognition can be
considered for incorporation.
REFERENCES
DICOM, Digital Imaging and Communications in Medicine
Standard, National Electrical Manufacturers Association,
2009.
Gonzalez, R. and Woods, R. Digital Image Processing,
Prentice-hall (3rd Edition), New York, 2007.
Keskin, C., Aran, O. and Akarun, L., Real time gestural interface for generic applications, European Signal Processing
Conference, 2005.
Kadobayashi, R., Nishimoto, K., and Mase, K., Design and
evaluation of gesture interface of an immersive walkthrough application for exploring cyberspace, 3rd IEEE
International Conference on Digital Object Identifier,
1998.
Zhang, W., Pan, Zhen-Kuan, and Wang, Yu, Recovery of
camera poses and 3D reconstruction based on image
sequence, International Conference on Machine Learning
and Cybernetics, pp. 30113016, 2008.
219

Multi-material blending for complex-shaped heterogeneous objects

I.T. Ozbolat
B. Koc
ABSTRACT: This paper presents a novel multi-directional blending method for heterogeneous object modeling. Contrary to earlier studies, this research introduces material blending through multiple features with different
heterogeneous material composition. This research could have the potential to enrich the theoretical system of
complex-shaped heterogeneous objects and advance the systematic approaches in heterogeneous object design
with irregular heterogeneities. Feature-based method is used to represent and design heterogeneous objects with
multi-directional material composition. Multi-directional material composition is mapped onto ruled surface
based geometric domain based on a set of relations such as neighborhood relations and material presence relations in addition to one to one correspondence of geometric variation driven material composition variation. The
proposed methodologies are implemented and illustrative examples are given in this paper.
1
INTRODUCTION
Heterogeneous objects have been great deal of interest

in CAD, CAE and CAM communities. Heterogeneous
objects often prevail over their traditional homogenous counterparts in certain performances such as
mechanical, thermal, electrical or any combination of
them (Samanta & Koc 2005). Heterogeneous objects
synthesize different materials into one monolithic
component and fulfill critical functional requirements,
in which a single material usually fails. For example,
material incompatibilities, stress concentrations and
non-uniform thermal expansion can be avoided with
continuous gradual composition of multiple materials
and structural variations. Due to these unique features
and advantages, heterogeneous objects have gained
great popularities in numerous applications such as
design of turbine blade (Qian & Dutta 2002), injection mold cooling system (Huang & Fadel 2001), and
tissue scaffolds (Ozbolat et al. 2009).
Many of current heterogeneous object modeling
represents synthetic and hypothetical heterogeneous
objects with regular and simple material distribution.
Constructing complex and irregular heterogeneities
is still a non-trivial task and has attracted enormous
attentions in recent years. Most of existing methods
proposed that complex heterogeneous objects could
be constructed by introducing additional geometric
primitives. Material composition in these methods,
however, is still no more complicated than the primitives and follows a simple variation while each
primitive inherits other primitives material compositions. Assigning different material composition on
each primitive on the other hand deserves particular
attention. In this case, material composition should
vary in several directions to satisfy overall material

composition. Irregular heterogeneity is thus defined
as the variation of material composition in a number of
different directions simultaneously and interactively.
Increasing irregularity and complexity in material
composition for superior multifold performances is of
particular interest in heterogeneous object modeling.
In cases where material composition is governed by
several features simultaneously, single or double material inclusion in the material composition does not
satisfy the composition need. This raises a certain need
for a blend of several materials to achieve desired composition spatially for improved material performance.
In this paper, a new multi-directional blending method
is used to control material composition inside the
object. Variation in material composition is dictated
by existence of multiple features, in which material
composition alters in several blending directions. In
addition to continuous spatial orientation of material
variation, imposed relations and constraints between
distinct features increase heterogeneity in material
composition. This brings further irregularity to the
overall material composition due to effect of material
imposition between several features interactively.
GEOMETRIC MODELING
2.1 Feature-based design

A feature can be defined as a generic shape or a
characteristic of a product with which engineers can
associate certain attributes and knowledge useful for
reasoning about that product (Shah & Mantyla 1995).
Features encapsulate the engineering significance of
221
portion of the product geometry and they are applicable in product design such as heterogeneous object
design. There are several methods proposed for heterogeneous object development such as voxel model,
volume mesh model, feature-based model, function
based model, and composite models. The main advantage of feature-based design over other methods is that
there is no need to develop a new design from scratch
if a design of a similar shape product is available. A
new design can be obtained by changing the features of
the existing design. For detailed comparison of existing methods, the reader is referred to (Kou & Tan
2007). The design of heterogeneous objects involves
developing the object model and obtaining changes
in the material composition. It is possible to represent the continuous material transition with relation
to the objects geometric features. In such a scheme,
the geometric features can be changed to obtain different designs. In order to exploit the advantages of
feature-based design in the context of heterogeneous
object modeling, material attributes were developed
with relation to the geometric features in our earlier
work (Samanta & Koc 2005). This brought further
diversity in design process and different designs were
obtained by varying the geometry of features and their
material attributes.
In feature based representation (Samanta & Koc
2005), a heterogeneous object {O} can be expressed
as a collection of four components: Features {F},
Materials {M}, Relations {R} and Constraints {C}.
Features are geometric entities that constitute the entire
object in geometric domain G E3 , where E3 is
3-dimensional Euclidean space. The main driving
forces in feature-based design are the object features.
The object under design consideration is required to
possess some properties so that it functions properly in
its respective applications. These properties are specified explicitly or implicitly at some specific features
of the object. During the design process, these features
dictate the ultimate shape and the material composition of the object geometry. Features can be subdivided
to three categories: Form features {FFa }, generator
features {GFb } and target features {TFc }. Form features represent the geometry as geometric entities such
as points, curves, surfaces etc. Generator features and
target features can be considered as form features in
geometric domain; however they dominate the material composition change in G . Generator features and
target features are the source and destination shapes
with distinct material properties, respectively. Material composition changes between generator and target
features and a smooth change is required for improved
material properties.
In addition to the object geometry, material composition needs to be defined over the geometric domain.
The design space can be defined as E3 ED , where the
geometry space E3 is the base space, ED is the material
space and D 1 is the number of primary materials. Thus, a set of materials {Md } in material domain
M (where M ED ) constitutes the heterogeneous
object. The heterogeneous object also encompasses a
Figure 1. Increasing irregularity in material composition

of a heterogeneous object and its corresponding entities for
feature-based representation.
set of relations R of its existing features. These relations can be material presence relation, one to one
correspondence of material composition change and
neighborhood relation. If a material presents on a feature in the geometric domain, then the overall material
composition is affected and is called material presence relation. One to one correspondence of material
composition change is, however, defined as the material composition change between a generator feature
and a target feature. This change can also be affected
due to the proximity of other target features and this
scheme is called as neighborhood relation. In other
words, a feature can affect the material composition
change around its neighborhood.These relations determine the overall material composition of the object.
These relations however can be limited by a set of
constraints {Ce } in form of features or materials. For
instance, a material can be restricted to be added into
the composition due to its detrimental effect. Thus,
the object needs to be constructed without inclusion
of this material. The following schematic highlights
the evolution of heterogeneous object modeling with
its components in feature-based design.
Figure 1 shows the evolution of material composition irregularity in heterogeneous object development
in the last few decades (highlighted by arrows). For
the rest of the paper, materials in this study are represented by color codes in RGB space. Figure 1(a)
represents a heterogeneous object with simple material
composition change between a generator (GFb ) and a
target feature (TFc ). Stated by a recent review study
(Kou & Tan 2007), plethora of work has been done in
this concept; however, current state of art necessitates
more complex and irregular composites with enhanced
effectiveness. Researchers have advanced the concept
in Figure 1(a) and introduced geometric complexities in terms of constraints with additional primitives
(Ce ) as shown in Figure 1(b) (Kou & Tan 2007).
Irregular material distribution however deserves particular attention and effort for improved multifold
performance spatially in addition to continuous spatial
orientation of material composition change. In cases
where material composition is governed by more than
one generator-target feature couple (where TFc+1 is
222
Figure 2. Sample set of relations among (a) features F and

(b) branching points B
introduced as illustrated in Figure 1(c)), such that each

couple dominates different material blending, there is
a conflict of material selection. In this case, single
or double material inclusion into the material composition does not satisfy the aforementioned conflicts
simultaneously. Thus, a blend of several materials is
needed to achieve desired composition at each location to enhance material performance by satisfying
imposed relation constraints to material selections and
geometrical complexities. Materials are selected from
a set of materials M based on relations and constraints.
Material presence relation dictates the presence of a
material inside geometric domain such as presence of
M1 (Fig. 1(c)). Material constraint, on the other hand,
restricts the presence of a material i.e. M2 . One-toone correspondence of material composition change
relation dictates the transition of material composition from the generator feature (GFb ) to target features
(TFc and TFc+1 ) separately. Thus, material blending
is progressed in dual-directions shown by one-to-one
correspondence of material composition change relation symbols. Neighborhood relation, however, brings
an interactive relation to these relations and an additional change in material composition is introduced
between target features (TFc and TFc+1 ).
2.2
Ruled surface construction
In our recent work (Ozbolat & Koc 2011), we proposed a method to create ruled surface for hollowed
objects with multiple internal features (Fig. 2(a)). In
(Ozbolat & Koc 2011), the Voronoi diagram was generated and a number of Voronoi cells were obtained
for each individual internal feature as shown in Figure 2(b). Next, curve matching was developed to insert
optimum ruling lines for each unit cell. Then, a ruling line alignment technique was developed to align
ruling lines inserted between the external curve and
the Voronoi while material properties change from
external curve to the internal hollowed features.
Material composition modeling presented in the
next section is then used to map heterogeneous material properties into the constructed ruled surface.
3
MATERIAL COMPOSITION MODELING
Heterogeneous material distribution of an object can

be defined by mapping from the geometric domain
to the material domain by function g : G M .
Figure 3. Sample set of relations among (a) features F and

(b) branching points B
Material composition mapping of a heterogeneous

object is initialized with assigning boundary (feature)
conditions. Material composition of a feature can be
expressed using the following equation:
where ra,d (s), rb,d (u) and rc,d (v) stand for material
composition fraction of material Md at a parametric distance s, u and v on a form feature FFa , on a
generator feature GFb and on a target feature TFc
respectively. Material composition of set of features
F in geometric domain G does not necessarily show
homogeneity on their respective domain as proposed
by (Kou & Tan 2005). Heterogeneity can be thus
imposed by introducing composition fraction as a
function of parametric distance on the feature domain.
More importantly, material composition between
features shows variability depending on set of relations such as neighborhood relation, material presence
relation and one-to-one correspondence of material
variation.
Figure 3 demonstrates a sample relations diagram
on the features and branching points. The Voronoi diagram possesses the medial property of material composition throughout the object geometry and shows
variability depending on the feature location and their
interactions. We continue with mapping material properties of branching points B = {Bj }j=1,..,J in which
each Bj has equal and medial material composition of
its corresponding neighborhood n(Bj ). Corresponding material composition of a neighborhood feature is
evaluated at a parametric distance u where the disc centered at Bj is tangent to that feature at u. In this paper,
neighborhood set of a branching point is assumed to
have at most 3 elements while each branching point
has only one disc that is assumed to be tangent to 3
features simultaneously. Figure 3(a) simply highlights
homogenous features of an arbitrary boundary with
many internal curves. These features are used to map
material composition of branching points at junctions
of the Voronoi Diagram as shown in Table 1.
223
Table 1. Material composition mapping on branching points

based on neighborhood relations.
Material Composition
of Features
Neighborhood Set of
Branching Points
g(GF 1 ) = {M1 }
g(TF 1 ) = {M2 }
g(TF 2 ) = {M3 }
g(TF 3 ) = {M4 }
g(TF 4 ) = {M5 }
n(B1 ) = {GF 1 , TF 1 , TF 2 }
n(B2 ) = {GF 1 , TF 2 , TF 3 }
n(B3 ) = {TF 1 , TF 2 , TF 4 }
n(B4 ) = {TF 2 , TF 3 , TF 4 }
n(B5 ) = {GF 1 , TF 1 , TF 4 }
n(B6 ) = {GF 1 , TF 3 , TF 4 }
Figure 5. Material Composition Mapping for a Unit Cell.
where h = h(u) is a user inputted distance based function to control material composition. The same procedure can be applied between any branching points
over the Voronoi diagram. Thus, material property of
discretized vertices over G can be obtained.
While the material composition information for
each vertex on each directices (a Voronoi cell and a
target feature) is known, variational material composition of a ruling line RLt can be interpolated at each
Iso-condition curve Ik (Fig. 5(a)). Material composition of Ik at the intersection point wijk with RLt is
calculated by Equation (17):
Material Composition of Branching Points

g(B1 ) = {1/3M1 , 1/3M2 , 1/3M3 }
g(B2 ) = {1/3M1 , 1/3M3 , 1/3M4 }
g(B3 ) = {1/3M2 , 1/3M3 , 1/3M5 }
g(B4 ) = {1/3M3 , 1/3M4 , 1/3M5 }
g(B5 ) = {1/3M1 , 1/3M2 , 1/3M5 }
g(B6 ) = {1/3M1 , 1/3M4 , 1/3M5 }
Figure 4. Material Composition Mapping for a Voronoi

Cell.
For instance, B1 has a neighbor set n(B1 ) = {GF1 ,

TF1 , TF2 } and its composition is directly affected by its
neighborhood set. Thus, we map the material composition of B1 as g(B1 ) = {1/3M1 , 1/3M2 , 1/3M3 } with
medial properties of the neighbor set. This means that
contribution of a feature FFa into the material composition of an arbitrary point diminishes when X moves
away from FFa and converges to 0 when X reaches
another feature FFs:s=a . After mapping material composition at the branching points, heterogeneous material composition over the Voronoi can be computed.
In Figure 4, let B3 and B5 are located on the
Voronoi cell at parametric distances pa = VC(ua ) and
pb = VC(ub ) respectively. Note that pa and pb are not
necessarily be a discretized point during parameterization process. They are the point where there is no
C1 continuity. Thus for any arbitrary point in the parametric domain ua u ub , the material property is
calculated using function based interpolation:
where wij0 and wijK are the initial boundary conditions where g(wij0 ) = g(ui ) and g(wijK ) = g(vj ). Once
the material composition is obtained for all vertices
on Iso-condition curves, a quadrilateral unit cell cijk
is constructed as illustrated in Figure 5(b). Overall
material composition of a unit cell g(cijk ) is calculated
using interpolation with respect to the location of its
centroid. Centroid of a quadrilateral is determined by
calculating the centroid of each triangle as the intersection point of medians followed by the weighted average
of these centroids.
In this section, the developed method for mapping
the material composition is introduced. The material
composition of the generated ruled surface between
the generator feature and the Voronoi diagram is
mapped similarly. Next section presents the implementation of overall methodology with illustrative
examples.
4
IMPLEMENTATION AND RESULTS
The proposed algorithms and methodologies are coded

in VB based Rhinoscript and examples are generated in Rhinoceros 4.0 software. Several examples of
multi-directional blending are generated by assigning
different material properties represented by different
values in RGB color space as shown in Figure 6.
Examples are generated on a computer with Intel Pentium 2 GHz processor and 1 GB of RAM. Computation
time of generated example in Figure 6(a) is 28 sec.
Figure 6(b) is developed in 48 sec.
Although change in material properties is dominated by the closest two-features, effects of other
224
composition in this case varies in several ways due

to metamorphosis from GF1 to TF1 and GF1 and
TF2 , effect of neighborhood relations and imposed heterogeneity on TF1 and TF2 . Similarly, heterogeneity
can also be imposed on GF1 to enhance irregularity
further. Besides, the proposed methodology enables
imposing additional form features. Figure 8(b) illustrates an arbitrary geometry with imposing 3 form
features as a material point source to control the spatial
orientation of material composition independent from
blending between GF1 to TF1 and GF1 and TF2 .
Figure 6. Examples of multi-directional blending between

a generator and multiple targets.
Figure 7. Highlighted irregular heterogeneity of material

composition.
Figure 8. Examples of multi-directional blending for

enhanced irregularity in material composition variation:
imposing (a) heterogeneous target feature and (b) material
as a point source.
neighbor properties are also desirable as illustrated in

the examples. Figure 7 illustrates a blending example
with 5 features, each has different material composition. Although change in material composition is
directed from GF1 to each target features gradually,
neighborhood relation is also critical in material composition. For instance, material composition of region
between GF1 and TF4 is altering from {M1 } to {M5 }.
Presence of neighbor features TF1 and TF3 affects
the material composition and introduced {M2 } and
{M4 } highlighted with arrows in red and blue
respectively.
Thus, material composition variation is also guided
by presence of neighborhood relations independent
from geometric blending between the generator feature and target features. Degree of neighborhood
effect can be controlled by controlling the user inputted
function. In this paper, a cubic polynomial function is
used in Equation (2) to represent material composition
variation along the Voronoi diagram.
In Figure 8(a), a heterogeneous object with heterogeneous target features is illustrated. Material
CONCLUSION
In this paper, we propose a novel multi-material blending approach with multiple target features, each with
different homogenous or heterogeneous material composition in its respective domain. Irregular variation
in material composition is then mapped to generate heterogeneous objects based on a set of relations
such as neighborhood relations and material presence in addition to one to one correspondence of
geometric variation driven material variation. Thus,
material composition varies in a number of directions smoothly and elegantly in a complex geometric
domain. Irregularity in the material composition is also
increased by imposing above-mentioned relations. The
proposed methodologies will enable fabrication of
heterogeneous parts while the continuous material
variation over the part is known. Complex heterogeneous part will have a great potential in many different
applications.
REFERENCES
Huang, J. & Fadel, G.M. 2001. Bi-objective optimization
design of heterogeneous injection mold cooling systems. ASME Transactions, Journal of Mechanical Design
123(2): 226239.
Khoda et al. 2011. Engineered Tissue Scaffolds with variational porosity architecture. Journal of Biomechanical
Engineering 133(1):011001(1)(12).
Kou, X. & Tan, S. 2007. Heterogeneous object modeling: A
review. Computer-Aided Design 39(4): 284301.
Kou, X.Y. & Tan, S.T. 2005. A hierarchical representation for
heterogeneous object modeling. Computer-Aided Design
37(3): 307319.
Ozbolat, I. T. & Koc, B. 2011. Multi-target Blending for Heterogenous Object Modeling. Computer-Aided Design (In
Press).
Ozbolat et al. 2009. Feature Based Bio-Modeling of Micropatterned Structures for Tissue Engineering. ComputerAided Design and Applications 6(5): 661671.
Qian, X. & Dutta, D. 2002. Design of heterogeneous turbine
blade. Computer-Aided Design 35(3): 319329.
Samanta, K. & Koc., B. 2005. Feature-based design and
material blending for free-form heterogeneous object
modeling. Computer-Aided Design 37(3): 287305.
Shah, J. J. & Mantyla, M. 1995. Parametric and feature based
CAD/CAM, Wiley: New York.
225

Gaussian Vault geometry: Integrated approach in design and

fabrication of physical prototypes
M.P. Sharman
School of Architecture and Building, Deakin University, Geelong, Australia
S. Datta
School of Built Environment, Curtin University, Perth, Australia
ABSTRACT: This paper reports the second part of a study on the digital design and fabrication of scaled
architectural prototypes. The first paper reported techniques in the realization of a double curved vault surface, the
Gaussian Vault. The aims of the research here further extend this body of knowledge to a better understanding of
constructible components. It addresses the problem of fabricating complex curved forms through the integration
of the basic building elements, skin and structure, to achieve a scaled physical prototype. The focus of the
experimentation is to investigate the process from which a digital surface form is conceived, to its preparation
for fabrication and eventual construction in the fashion of a scaled model or workable prototype.
INTRODUCTION
Recent advances in parametric modeling, architectural geometry and digital fabrication methods have
seen the prototyping of structure envelopes of complex
curvature. Key to the investigation is the ability of processing the model from a three dimensional entity and
into a relatable two dimensional set of assembly elements. The best method of envisaging the progression
of this research is through the construction of physical prototype models that represent the eventual full
scale construction of the desired surface form. Using
computer-aided modeling functions, the design, representation, and fabrication methods of scaled models
can be significantly enhanced (Aish & woodbury,
2005). However, within the digital representation of
prototype form, the translation of the geometry to
formats suitable for digital fabrication poses considerable challenges for architects and designers (Kieran &
Timberlake, 2004).
The primary focus of this research is to explore the
design geometry, investigate methods for preparing the
geometry for fabrication, and constructing the physical prototypes. The focus of the experimentation is
to investigate the process from which a digital surface
form is conceived, to its preparation for fabrication and
eventual construction in the fashion of a scaled model
or workable prototype (Kolarevic & Klinger, 2008).
An integrated design approach, utilizing computeraided modeling and suitable fabrication processes
such as reported by Bechthold (2004, 2008) are investigated. The forms used for the experimentation in
design and fabrication of scaled prototypes were based
on Eladio Dieste and his work in masonry vaults and
shells (Pedreschi, 2000). Of Diestes developed forms,
this paper focuses on the application of prototyping

methods to realize the geometry of the Gaussian Vault.
The intention of the paper is to understand the different processes for translating digital surface geometry
of Gaussian Vaults into physical prototypes.
GAUSSIAN VAULT GEOMETRY
This paper reports the second part of a study on the

digital design and fabrication of scaled prototypes.
The first paper (Sharman & Datta, 2009) reported
techniques in the realization of a double curved vault
surface, the Gaussian Vault (Anderson, 2004). The
aims of the research here further extend this body of
knowledge to a better understanding of constructible
components, but in relation to curved surfaces. This
paper expands on the problem of fabricating forms
such as the Gaussian Vault to the development
of physical mockups. The choice of the Gaussian Vault
as a prototype was motivated by three qualities of the
form: the strict mathematical/structural basis of the
form, the presence of double curvature on the surface
and the possibility to explore the vault prototype using
ruled surface geometry.
The surface geometry of a Gaussian Vault is based
on two catenary curves of differing heights that sweep
across the span of the form. Profiling the progression
(perpendicular to the span), the vault generally starts
and ends with straight line profiles (Figure 1). However, as the profile reaches the apex of the span curves,
the shape transforms into an S-Curve, which creates
the degree of double curvature that characterizes this
surface form (Anderson, 2004).
227
Figure 1. Surface geometry of gaussian vault with catenaries and S-Curve profile.
Figure 3. Curved surface rationalization with planar

polygons.
Figure 2. Rapid prototype of Gaussian Vault.
Figure 4. Paper fabrication of vault geometry with planar

strips in both directions and their assembly.
PROTOTYPING METHODS
This research has a strong focus that aims to engage

with manufacturing and rapid prototyping devices. As
reported in the first paper, three methods of digital
fabrication were explored; (solid) Rapid-Prototyping,
Surface (paper) Fabrication and Structural (component) Fabrication. Parametric tooling was crucial in
these methods due to its capabilities in automated
adaptability to the framework of given surface form.
3.1
Rapid-Prototyping
The use of Rapid-Prototyping within Architectural

fields is not necessarily new or unique. The process
that is presently involved requires very little work to
transfer a design surface into a solid physical form. In
terms of digital process once the Gaussian Vault surface is conceived, the only tooling that is required is
a thickening to add depth to the geometry. With this
established, the digital form can be exported via the
manufacturing requirements of the rapid-prototyping
machine, in this case a Z-Corp Z450 3D Printer
(Figure 2).
3.2
Surface fabrication
One of the more interesting methods discovered from

the investigations into digital fabrication involves an
attempt at generating the surface form into planar polygon elements. From within most Parametric software,
there is the ability to extract geometric information
associated to a grids of points that is applied to a curving surface form.The points enable an establishment of
meshing the surface into a polygon grid which in turn
can be extrapolated and generated into a series of planar strips. This, again, is not necessarily a new process
and has gone a long way in recent times to rationalize
and realize curved form in buildings (Figure 3).
With this the evaluation elements enabled, the fabrication planning could take place. In this case the
method of the fabrication planning involved the software to intellectually unfold each band of the polygon
mesh into a planar strip. To further facilitate an accurate representation of the scaled prototype, planar
strips were developed in both the lateral and longitudinal direction of the surface geometry. This was
permitted by switching the skin direction of the surface within the software parameters (i.e. paths became
profiles and profiles became paths).
The formulation of the planar strips has two distinct
attributes, the first was a set of vector lines that needed
to be severed, and the second was another set of vector
lines that needed to be engraved to show where the
material was to be folded. With the use of a device such
as a laser cutter, each set of vector lines (cut or etched)
could be distinguished and allow the production of
individual planar strips of material.
Where the interests lies is that by making a simple
change in the direction that the surface skin generates,
the output planar strips self generate in an opposing direction (perpendicular). The resulting prototype
model is constructed by fabricating each set of polygon strips using a malleable material such as paper on
a laser cutter. The assembly of these flat elements is
weaved together and reformed back into the intended
3D design form (Figure 4).
GRID SHELL PROTOTYPE
The next stage in the prototyping process was to

build a 1:10 scale model of the prototype using a
structure-skin strategy rather than a purely surface
strategy as outlined in the previous section (Schittich,
2006). The production of a reticulated prototype
228
involves the design of structural members as developable components. The aim of the prototype is to
develop a reticulated grid shell that can be fabricated
using planar technology. The Gaussian surface geometry is formulated into a discrete polygon mesh. The
mesh is similar to the thin surface method, with the
additional constraint that each polygon has to be oriented in strict lateral and longitudinal directions. In
other words, the point grid needs to maintain straight
directions and not be able to curve away in a third
direction (Figure 6).
4.1
Dual carrier component
The use of a dual carrier component to formulate and

produce a scaled prototype form is not a method that
is as straight forward as the previously mentioned processes. The theory behind this method involves the
design and development of a digital model component
that then could relate in terms of formulating geometry to the desired surface. The definition and control of
the component geometry is integral to creating correct
tessellation in the final array. As previously mentioned,
parametric modelling has the ability to calculate a
components deformation when applied to a controlling surface. This means that to successfully control
the performance of a component, it has to be built
within controlling surface planes that will correlate
to the planes in the final application. These controlling planes are referred to as dual carrier surfaces.
These surfaces create a bounding box that constrains
the components geometry in order to act as an interface between component and governing surface mesh
(Pitts & Datta, 2009).
Since the mesh is derived from surface geometry,
no depth has yet been included to represent the body
of the prototype form (i.e. thickness). As the intention
of this experiment is to fabricate structural members
that would represent the form, depth is of vital importance to the calculation of the construction component.
In being able to do this, the depth is formulated by
including an inner and an outer surface geometry to
what would represent the prototype model.
After the process of including point grid and polygon mesh evaluation, these surfaces now present themselves similar to parallel meshes (Pottman, Asperl &
Kilian, 2007). In terms of the digital construction component, this also meant that the primary input would be
developed between two single polygon shapes (quads).
The depth of components would then be determined by
the distance between the meshes; an input that could
be controlled within the digital surface form model
(Figure 5).
The aim of the component was to simulate structural members that will make up the prototype form,
yet still be able to be fabricated using planar technology. Further evaluation into the surface geometry was
required for the component to be accurately applied.
This meant that the surface geometry still had to be
formulated into a polygon mesh, similar to the previous method, yet the polygon had to remain in strict
Figure 5. Dual carrier component used to represent structural thickness.
lateral and longitudinal directions. If a point grid was

directly applied to the surface geometry, the resultant
bands of points would still be able to curve away from
the straight directions required.
To be able to constrain the point grid in terms of
its lateral and longitudinal bearings, a rectangular projection mesh was created in parallel to curved surface
geometry. This mesh produced the required point grid
that, once projected in a controlled single direction
towards the surface, would apply the points where they
were required (Figure 6).
4.2 Component prototype
The development of the construction component had
to address a way to apply solid members between the
four vertices points of each of the polygon shapes. To
make the process more difficult, the directions used
within component also had to adhere to individual
229
Figure 7. Assembly of the grid shell parts.
Figure 6. Difference in point grid application between direct

application or use of a projection grid.
or combined vertex projections rather than environmental directions. This meant that the use of a base
coordinate system of X, Y or Z could not be used.
During the conception of the digital component,
further objectives were explored as to the control of
member sizes and the addition of joinery, between
crossing members. The parameters for these were
setup by variable change within the component model,
which then presented themselves as inputs once
applied in the overall form model.
With the component in place over the point grids
of the surface geometry, the end result was a simple three dimensional model that could be handled
within any standard computer-aided modeling package. With this being a Grid Shell structure, each of the
structural member that formulates a grid shell portion
could be extrapolated for arrangement on a fabrication
sheet layout. The information then can be transferred
to a planar fabrication device such as a laser cutter
or CNC router, whereby each member is cut from the
decided material. Once the manufacturing is complete,
each part of the Grid Shell is reformulated by joining
together at the corners (with the aid of the lap joint).
As each Grid Shell is reconstructed, they are in turn
screwed together with the pre-positioned holes, and
the prototype model slowly takes shape (Figure 7).
Figure 8. Gaussian shell prototype model.
4.3 Scaled model of Gaussian Vault geometry

Overall, the primary goal of the construction component was allowing it to apply to the evaluated surface
geometry autonomously, without need for manual correction on any individual part or portion (Figure 8).
The ability to include surface planar fabrication into a
component, such used in this research, has been tried
and tested with the expected results suitable in a scaled
model format. Further inclusions to the component in
relation to possible full scale manufacturing may take
into account for different joinery methods, the standardization of fittings, and an exploration in detail to
the inside and outside surfaces.
4.4 Conclusions
The analysis of the results of each of the fabrication processes allowed for a process of thinking that
was directed towards a fabrication process that could
relate to the expectation in actual manufacturing the
desired building. Whilst there still is benefit, from a
230
prototyping standpoint, in the use of the of rapid prototyping devices, the end product is simply a form
model that only gives representation to size and shape.
The two processes that develop the surface geometry
with expectation of planar fabrication gives a better realization to the prototype model because they
are a formulation of either the skin or structure in
which a building is comprised of. In saying that, the
combination of both techniques could allow for a selfgenerating component that considers the important
aspects to a building envelope (Schittich, 2006).
The direction of further research would be to
develop the construction component in a way to
account for better integration between the structure
and skin relationship (Stephan, 2005). The ability
for the component to autonomously formulate over
given surface geometry would ease the process of
thinking by the architect or designer in realizing
their building. The ability for architects and designers
to integrate conceptual form making with structural
and material characteristics and its consequent production is enhanced through a better understanding
of the processes of digital to physical translation
(Shelden, 2002). The scaled prototypes investigated
in this paper develop planar fabrication models of
complex geometry. For achieving a better integration
between the structure and skin, their implications for
full-scale modeling and exploration of architectural
and structural details, further research into reticulated
components (Stephan et al, 2005) and parallel meshes
(Pottman, Asperl & Kilian, 2007) is necessary.
REFERENCES
Aish, R. and Woodbury, R.: 2005, Multi-level Interaction in
Parametric Design in Smart Geometry, Springer-Verlag,
Berlin.
Anderson, S. (ed.): 2004, Eladio Dieste: Innovation in

Structural Art, Princeton Architectural Press,
New York.
Bechthold, M.: 2004, Surface Structures: Digital Design and
Fabrication, ACADIA04, Ontario, pp. 8899.
Bechthold M, Innovative Surface Structures: Technologies
and Applications, Taylor & Francis, Oxon, 2008.
Kieran S,Timberlake J, RefabricatingArchitecture, McGrawHill, New York, 2004.
Kolarevic B, Klinger K, Manufacturing Material Effects:
Rethinking Design and Making in Architecture,
Routledge, New York, 2008.
Kolarevic, B (ed): 2003, Architecture in the Digital Age:
Design and Manufacturing, Spon Press, New York.
Pedreschi, R.: 2000, The Engineers Contribution to Contemporary Architecture: Eladio Dieste, Thomas Telford,
London.
Pottman, H., Asperl, A. and Kilian, A. (ed.): 2007, Architectural Geometry, Bentley Institute Press, Exton.
Pitts, Greg and Datta, S. 2009. Parametric modelling of
architectural surfaces, in Chang, Teng-Wen; Champion,
Eric; Chien, Sheng-Fen and Chiou, Shang-Chia (eds),
CAADRIA 2009 : Between man and machine-integration,
intuition, intelligence : Proceedings of the 14th International Conference on Computer Aided Architectural
Design Research in Asia, 635644, National Yunlin
University of Science and Technology, Yunlin, Taiwan
Schittich, C. (ed.): 2006, in Detail: Building Skins,
Birkhuser, Berlin.
Shelden, D.R.: 2002, Digital Surface Representation and the
Constructability of Gehrys Architecture, Ph.D. thesis in
Department of Architecture, Massachusetts Institute of
Technology, Cambridge.
Sharman M, Datta S. 2009. Gaussian Vault Geometry:
Digital Design and Fabrication of Scaled Prototypes,
in Proceedings of the 14th International Conference of
Computer Aided Architectural Design Research in Asia,
267276.
Stephan, S., Snchez-Alvarez, J., and Knebel, K.: 2005,
Reticulated Structures on Free Form Surfaces, Mero
GmbH & Co., Wrzburg.
231

Application of contour tracing algorithm for assisting non-contact

data acquisition
S. Rianmora
Sirindhorn International Institute of Technology, Thammasat University, Pathumthani, Thailand
P. Koomsap & P. Kuagoolkijgarn

Asian Institute of Technology, Pathumthani, Thailand
ABSTRACT: For acquiring surface data, non-contact laser scanning and 3D optical digitizer are preferable
as they offer acquisition speed with less human labour required. However, they cannot properly capture surface
data around sharp edges and corners of an object. Consequently, extra effort is required to fill up the missing
data and to correct the constructed surface. This paper presents an application of image processing technique for
obtaining data points around sharp edges and corners of an object. In the process, views that are perpendicular
to these features are captured and converted to be edge images. Contour tracing algorithm is then applied to
track these features that appear as closed contours on the edge images. These contours are then transformed to
be simple formats, with the known height, they can be extruded to be 3D CAD parts. Boolean operations are
applied to register all 3D parts to form a 3D model.
INTRODUCTION
Reverse engineering (RE) has been introduced as

a process that allows 3D CAD model to be constructed quickly from its physical object. The result
can be used to analyze the structure, function, and
operation (Rho et al. 2001). In general, RE process starts with capturing the geometric shape of an
existing object by acquisition devices which can be
categorized into two groups; contact and non-contact
methods (Bardell et al. 2003). The coordinate measuring machine (CMM), a multi-jointed mechanical
arm with a touch probe, is an example of the contact
method. Using CMM is very time-consuming (Yau &
Menq 1995), but this is mandatory when high accuracy is required. In terms of speed and less human
labour required, non-contact methods such as computed tomography (CT) scans, magnetic resonance
imaging (MRI), optical digitizer and ultrasound are
more efficient (Chan et al. 2001, and Son et al. 2003).
The detected distance is converted into a point cloud
map representing the 3D surface (Lee et al. 2000).
Point cloud data, acquired from these techniques,
require preprocessing operations, such as filtering
outliers, smoothing and blending of existing points
before being registered into one coordinate system
(Schoene & Hoffmann 1997).
Using non-contact scanning devices, however, cannot properly acquire surface data around some surface
features such as small holes, sharp edges, or corners
of an object, as illustrated in figure 1. In figure 2,
the scanning results show some missing information
around the triangular features. These problems have
Figure 1. Mechanical part that contains small triangular

features on the top surface. (a) Original part. (b) Coated part.
direct effects on maintaining the original shape of the

object after reconstructing surface. The accuracy of
laser scanning technique can be lost when the laser
hits the edge of an object because the information
that is sent back to the scanner is from two different
locations for one laser pulse. The coordinate relative
to the scanners position for a point that has hit the
edge of an object will be determined based on an average which will set the point in the wrong place. To
increase the chances of the beam hitting an edge, a
high resolution scan is recommended, however, it contains some noises behind the edges of the object (Blais
et al. 2004). To solve the problems, therefore, an extra
effort is required to fill up the missing data and to correct the constructed surface as illustrated in figure 3
and 4.
Besides manual correction, there was attempt to use
scanners with a smaller beam width but it might be
233
Figure 2. Point cloud data obtained from laser scanner.
Figure 3. Point cloud data are added around sharp edges.
Figure 4. Polygon phases generated after adding points.

(a) Top view, and (b) Side view.
limited by range as the beam width will increase over

distance (Silfvast 2004). In order to make data acquisition process more robust, presented in this research
is an application of image processing to induce geometric shape of sharp edges or corners of an object to
appear explicitly through edge images. These images
can be used to generate 3D part while maintaining surface accuracy. The implementation of this proposed
approach is also presented and discussed.
FEATURE DETECTION
A feature appearing on the object surface can be

easily detected by visual inspection but when the
information of the feature needs to be extracted and
retrieved for isolating regions, mapping or identifying defects, image processing methods have been
introduced (Langer & Kuhnert 2008, and Tuytelaars
2010).
The various features currently employed as general

features (e.g. color, texture, and shape), and domainspecific features (e.g. human faces, fingerprints, and
conceptual features) (Gauthier & Poulin 2009). Shape
is an important visual feature and it is one of the fundamental features for image content description. Shape
can be divided into edges, corners (interest points),
blobs (regions of interest or interest points), and ridges
(Ando 2000, and Kovesi 1999).
Edges are the points which have a strong gradient magnitude and represent as a boundary between
two image regions. That boundary (one dimensional
structure) may include junctions, and present as shape.
Some common algorithms will then chain high gradient points together to form a more complete description of an edge (Arora 2009, and Darshni & Kaur
2010).
Corners and interest points are analyzed as the edges
whose directions are changed rapidly, and referred to
point-like features in an image which have a local two
dimensional structure (Ruzon & Tomasi 2001). Blobs
provide a description of image structures in terms of
regions which contain a preferred point or a center of
gravity, as opposed to corners that are more point-like.
A ridge can be thought of as 1D curve (a grey-level
image) that represents an axis of symmetry or a medial
axis for elongated object. It is difficult to extract ridge
features from general classes of grey-level images
than edge, corner or blob features (Lindeberg 1998).
Ridge descriptors are frequently used for road extraction in aerial images and for extracting blood vessels
in medical images (Chen et al. 2010).
Data obtained by imaging techniques are calibrated
against established, classical measuring procedures
where reliable interpretation and calibration process
are required to enable appropriate decision-making.
Applying the concept of image binarization to induce
object features might be required less time than using
complicated methods. This has led to the proposed
approach that applies image processing algorithm to
obtain geometric shape of an object by converting 2D
images to be formed in 3D information while maintaining the surface accuracy. The concept of imaging
technique of this proposed approach will be presented
and discussed in the next section.
CONTOUR TRACING ALGORITHM
Presented in this section is an algorithm proposed

by Kuagoolkijgarn et al. (2010) for contouring a set
of 1-pixel-wide closed contours that are connected
and nested. The key component of this algorithm is
to apply basic morphological operations to achieve
the locations on an image of all contours instead of
using component labeling, a common approach for
searching for additional contour.
This algorithm is composed of four main steps;
identify area of interest, trace contour(s), identify the
presence of additional contour, and determine the
coordinates of traced contours on the original image.
234
Figure 5. Contour tracing algorithm.
Figure 6. Three phases of the proposed approach.
The input to the algorithm is a binary image of 1-pixelwide closed contours. In the case of color or grayscale
images, they are required to be preprocessed with
edge detection operation (e.g. canny edge detection).
Morphological boundary extraction can be applied on
binary region image to obtain the required input. It
is possible that edge detection operation may deliver
several open contours instead of closed contours. They
must be corrected before feeding into the algorithm.
The outputs are ordered sequences of boundary pixels
representing all closed contours.
The process starts with identifying the working
area. A smallest rectangle that contains all contours is
drawn, and the coordinate of its top left corner is
recorded as a local reference. The contour search
is then activated to search for a starting point and to
continue tracing on a contour. The tracing is continued
on its adjacent contours if they connect. The contour or
a cluster of contours is then subtracted from its original
image. If there still remains a contour in the result, its
image will be fed back to repeat the previous processes
until the result is empty.The algorithm then proceeds to
translate the coordinate of all closed contours back
to the positions on the original image. Illustrated in
figure 5 is contour tracing algorithm.
Recently, contour tracing algorithm has been
applied in rapid customization of multi-color nested
pattern products (Kuagoolkijgarn & Koomsap 2010),
and in 3D modeling from a paper-based overtraced
freehand sketch (Chansri & Koomsap 2011).
RESEARCH CONCEPT
The concept presented in this proposed approach is

about obtaining small feature or shape of the object
from its orthographic views by applying image processing algorithm and CAD model reconstruction. The
feature employed in this concept should contain uniform cross-section where the depth and height can be
measured. However, if the feature contains slope or
taper shape where non-uniform cross-section is found,
the concept of homotopy has been applied to generate the intermediate contours inside each section
(Soonanon & Koomsap 2009).
In the process, views that are perpendicular to the
feature are captured and converted to be edge images.
Contour tracing algorithm is then applied to track this
feature that appears as closed contours on the edge
image. The obtained contours on 2D-pixel coordinates
are then transformed to the formats that are properly
applied to input, store, and display model geometry.
With known heights or depths, the contours are then
extruded to be 3D CAD parts. Boolean operations (e.g.
union, subtract, or intersect) are applied to register
all 3D parts to form a 3D model. Figure 6 shows the
three main processes of the proposed approach; contour identification, extrusion process, and CAD model
reconstruction.
In contour identification process, three main activities are involved; capturing image of the object,
extracting edge image, and tracing contour. On each
activity, the concept of image processing is used for
accomplishing the process. The process starts to capture the images of the views that are perpendicular to
these features. In order to generate clean and clear
images which are free of blurs and low intensities,
the light intensity of the image should be increased
by using larger aperture with fast exposure time
(Rianmora & Koomsap 2008).
To extract edge image of the object features, among
the edge detector algorithms, Canny edge detector
has been selected for this study since its results outperformed other well-known edge detectors (Health
et al. 1998, Rahim et al. 2004, Maini & Aggarmal
2009, and Shriram et al. 2010). To execute the Canny
235
Table 1. The overall concepts of the proposed approach.
Figure 7. The position of the proposed approach in RE

process.
edge-detection algorithm, a user is asked to assign

the values for two parameters; sensitivity threshold
(0.010.99), and sigma (a positive value representing
standard deviation of the Gaussian filter) (Rianmora
et al. 2009).
Then, in activity 3, contour tracing algorithm is
used to convert the extracted edge profiles on 2D-pixel
coordinates to the world coordinates. The results are
the set of points which represent the geometric shape
of the object.
In extrusion process, a traced contour (representing
a feature profile) is extruded with the known height to
generate 3D part, as illustrated in activity 4.
The last process is CAD model reconstruction
which requires one activity; applying Boolean operations, to register all 3D parts and modify the obtained
surfaces to form a complete 3D model.
Illustrated in figure 6 is a position of the proposed
approach in the main path of RE process where the
extruded parts of the features of interest (i.e. subcomponents) will be launched to the path that joins to
the last phase. These sub-components can be joined
with the other 3D parts obtained from the conventional RE process. Moreover, in case of a simple model
that contains flat surface of the entire part, using this
proposed approach can show the benefits for easily
extracting geometric shape which can be constructed
3D CAD model directly without requiring any other
acquisition methods.
Illustrated in table 1 are the three main processes
employed in the proposed approach where five activities are involved. The images of a sample model are
captured from different views, as shown in figure 8.
5 APPLICATION OF THE PROPOSED
APPROACH
The aforementioned part that cannot be acquired data
around triangular features and small holes by laser
scanner will be used to demonstrate the proposed
approach. Using image processing technique can
capture and induce geometric shapes of those features.

In first process top of the part were captured and stored
into bitmap file. The captured image was then fed
into Canny edge detection algorithm which has been
236
Figure 8. Different views of a sample.
Figure 10. The contour points traced in sequence.
Figure 9. Application of contour tracing process on binarized image of the object features.
implemented on LabVIEW platform for extracting

edge image of object features.
After extracting all edges which represent as binarized images (one-pixel-wide closed contour) of the
object features, the contour tracing algorithm was performed to convert edge image information into simple
format (i.e. 3D space) which is ready for creating
model surface.
Illustrated in figure 9 is the contour tracing process.
The result (C0) obtained from the first iteration of the
program is shown at the bottom-left part of the contour
tracing process scene. The external contour is removed
from the image, leaving in eighteen contours. After the
second iteration, all contours are found as shown in
figure 10.
Since eight contours representing small hole features (i.e. C1, C5, C7, C9, C11, C13, C15, C17) are
identical, to minimize the processing time, selecting
only one contour to perform extrusion activity is recommended. For triangular contours which are all the
same shape, contour C2 was used for generating 3D
sub-part as shown in figure 11. With different orientations, all 3D sub-parts were then registered at the
common reference to the existing 3D model obtained
from traditional RE process (figure 12(a)).
For this particular example, moreover, 3D model
can be directly generated where height and depth of
each feature are known as illustrated in figure 12(b).
Applying the proposed approach, after performing
surface reconstruction process, all triangular features
and other small features can be revealed explicitly
compared to the traditional RE process that applies
non-scanning technique to retrieve surface details.
With the same process, a tape dispenser is used to
demonstrate the proposed approach where five images
of different views were taken. Three main phases; contour identification, extrusion process, and CAD model
reconstruction, presented in table 2, were applied.
237
Table 2. The application of this proposed approach on

captured images.
Figure 11. 3D Sub-parts.
Figure 12. 3D CAD model. (a) 3D sub-parts are joined into

the model obtained by traditional RE process. (b) Extruded
model by the proposed approach.
Figure 13. Tape dispenser used to demonstrate this proposed approach.
more complex shapes, some other algorithms have to

be developed in the future works.
CONCLUSIONS
Image processing technique is applied for assisting non-contact acquisition techniques to obtain data
points around the sharp edges and the corners of the
object from 2d images. Contour tracing algorithm is
applied to transform edge contours on pixel coordinates to be on the world coordinates which are
available for constructing 3d part through extrusion
process. The study shows the promising results that
encourage further research where the generated 3d part
can preserve the original shape. In this research, the
geometric and symmetric shapes of an object have
been focused since these shapes are available for extrusion technique. To acquire the curvature surfaces or
REFERENCES
Arora, M. 2009. Detection of Abnormalities in MRI Images
using Texture Analysis, Master Thesis, Electrical and
instrumentation engineering Department. Thapar university, Patiala.
Ando, S. 2000. Image field categorization and edge/corner
detection from gradient covariance. IEEE Transactions
on Pattern Analysis and Machine Intelligence, vol. 22,
pp. 179190.
Bardell, R., Balendran, V. & Sivayoganathan, K. 2003. Accuracy analysis of 3D data collection and freeform modelling
methods. Journal of Materials Processing Technology,
vol. 133, pp. 2633.
238
Blais, F., Picard, M. & Godin, G. 2004. Accurate 3D

acquisition of freely moving objects. 2nd International
Symposium on 3D Data Processing, Visualization, and
Transmission, 3DPVT 2004. Thessaloniki, Greece. Los
Alamitos, CA: IEEE Computer Society.
Chan, V., Bradley, C. & Vickers, G. 2001. A multi-sensor
approach to automating co-ordinate measuring machinebased reverse engineering. Computers in Industry, vol. 44,
pp.105115.
Chansri, N., Koomsap, P., & Kuagoolkijgarn, P. 2011. Application of Contour TracingAlgorithm in 3D Modeling from
a paper-based overtraced freehand sketch. 21st International Conference on Flexible Automation and Intelligent
Manufacturing (FAIM 2011). Taichung, Taiwan.
Chen, H., Yin J., Shu X., Hu C. & Li Y. 2010. An orientationbased ridge descriptor for fingerprint image matching, International Conference On Computer Design And
Appliations (ICCDA 2010).
Darshni, P. & Kaur, G. 2010. Implementation Of Aco Algorithm For Edge Detection And Sorting Salesman Problem. International Journal of Engineering Science and
Technology, vol. 2, pp. 23042315.
Gauthier, M. & Poulin, P. 2009. Preserving sharp edges in
geometry images. Proceedings of Graphics Interface.
Heath, M., Sarkar, S., Sanocki, T. & Bowyer, K. 1998. Comparison of edge detectors: A methodology and initial
study. ComputerVision and Image Understanding, vol. 69,
pp. 3854.
Kuagoolkijgarn, P., Koomsap, P., & Chansri, N.2010. A new
algorithm for tracing nests of interconnected contours,
International Journal of Advanced Manufacturing Technology, vol. 50, pp. 717727.
Kuagoolkijgarn, P., & Koomsap, P. 2010. Applying
image processing for rapid customization of multi-color
nested products. The 17th ISPE International Conference on Concurrent Engineering, Cracow, Poland,
September 2010.
Kovesi, P. 1999. Phase congruency detects corners and edges,
Journal of Computer Vision Research, vol. 1.
Langer, M. & Kuhnert, K.D. 2008. New Hierarchical
Approaches in Real-Time Robust Image Feature Detection
and Matching. 19th International Conference on Pattern
Recognition (ICPR), Tampa, Finland.
Lee, K. & Park, H. 2000. Automated inspection planning of free-form shape parts by laser scanning.
PERGAMON Robotics and Computer Integrated Manufacturing, vol. 16, pp. 201210.
Lindeberg, T. 1996. Edge detection and ridge detection
with automatic scale selection. International Journal of
Computer Vision, vol. 30, pp. 117154.
Lindner, C., Arigita, J. & Len, O.P. 2005. Illuminationbased segmentation of structured surfaces in automated
visual inspection. Proc. SPIEOptical Measurement

Syst. Industrial Inspection IV, W. Osten, C. Gorecki and
E. Novak, Eds., vol. 5856, pp. 99108.
Maini, R. & Aggarmal, H. 2009. Study and comparison of
various image edge detection techniques. International
Journal of Image Processing (IJIP), vol. 3, pp. 112.
Rahim, M., Ismail, N., & Idris, M. 2004. The use of hybrid
technique: Thresholding and edge detection for identifying river from aerial photo. Journal Teknoloki, vol. 41,
pp. 120.
Rho, H., Jun, Y., Park, S. & Choi, H. 2001. A rapid reverse
engineering system for reproducing 3D human busts.
CAD/CAM Research Centre, Korea Institute of Science
and Technology, Seoul, Korea.
Rianmora, S., & Koomsap, P. 2008. Investigating the
influence of color light in data acquisition. The 9th Asia
Pacific Industrial Engineering and Management Systems
Conference (APIEMS 2008). Bali, Indonesia.
Rianmora, S., Koomsap, P. & Van Hai, D. 2009. Selective data
acquisition for direct integration of reverse engineering
and rapid prototyping. Virtual and Physical Prototyping,
vol. 4, pp. 227239.
Ruzon, M.A., &Tomasi, C. 2001. Edge, Junction,And Corner
Detection Using Color Distributions. IEEE Transactions
on Pattern Analysis and Machine Intelligence, vol. 23.
Schoene, C. & Hoffmann, J. 1997. Reverse engineering
based on multiaxis digitizer data. International Conference on Manufacturing Automation (ICMA 97), vol. 1,
pp. 909914.
Shriram, K.V., Dharmendra, T., Sivaraman R. & Karthick
S. 2010. Automotive image processing technique using
Cannys edge method. International Journal of Engineering Science and Technology, vol. 2, pp. 26322644.
Silfvast, W. T. 2004. Laser Fundamentals, 2nd Ed. Cambridge
University, Cambridge.
Son, S., Kim, S., & Lee, K. 2003. Path planning of multipatched freeform surfaces for laser scanning, The International Journal of Advanced Manufacturing Technology,
vol. 22, pp. 424435.
Soonanon, P., & Koomsap, P. 2009. Towards direct transformation of orthographic-view drawing into a prototype.
Virtual and Physical Prototyping. vol. 4, pp. 7590.
Tuytelaars, T. 2010. Dense Interest Points. IEEE Conference on Computer Vision and Pattern Recognition 2010,
pp. 22812288.
Yau, H. & Menq, C. 1995. Automated CMM path planning for dimensional inspection of dies and molds having
complex surfaces. International Journal Machine Tools
Manufacturing, vol. 35, pp. 861876.
239

Image-based direct slicing of a single line drawing for rapid prototyping

N. Chansri
Kasetsart University, Chonburi, Thailand
P. Koomsap
Asian Institute of Technology, Pathumthani, Thailand
ABSTRACT: Sketch-based modeling has been researched for transforming a sketch to be a 3D CAD model.
When it is applied alongside rapid prototyping (RP), a physical prototype can be fabricated quickly and easily.
This paper presents a research on direct interfacing sketch-based modeling with rapid prototyping. It focuses
on generating a stack of contours after going through a sketch based modeling process when 3D coordinates
of all vertices, edges and faces are achieved from the 2D drawing. Rather than reconstructing a 3D model and
feeding it into a regular RP process, obtained faces are projected onto vertical principal planes. Slicing is then
performed directly on the images of these faces. The pixel coordinates where a horizontal plane cut the faces
are transformed into objects world coordinates and connected to form a closed contour representing each layer.
The approach has been successfully implemented and some examples are illustrated in this paper.
1
INTRODUCTION
Technology advancement makes it happen for a 3D

CAD model to be created from its physical object,
engineering drawing and sketch as alternatives to manual creation on commercial CAD software. Direct
transformation of a physical object to be a 3D CAD
model is known as reverse engineering (RE) where
the surface data of an object are captured by using
scanning or measuring devices, and after enhancing
their conditions, these point cloud data are then used
to generate surface of the 3D model (Sokovic & Kopac
2005). A technology for reconstructing a 3D model
from a 2D drawing is geometrical reconstruction
(GR) that was initially developed to collect knowhow from blue prints in drafting rooms (Company
et al. 2005). Boundary representation (B-rep) and constructive solid geometry (CSG) are two reconstruction
schemes. For B-rep, 2D points on an engineering drawing are transformed into vertices in a 3D space before
being connected to form edges, faces and finally a
model (Chen & Feng 2003). For CSG, a model is
formed by applying Boolean logics on a set of primitive objects built from closed regions on a drawing
(Tam & Atkinson 2003).
For a sketch that is widely used for expressing and
communicating ideas, a technology for transforming it
to be a 3D model is called sketch-based modeling. Two
major groups of this modeling are evocative-base modeling and constructive-based modeling. Based on prior
knowledge, evocative-based modeling reconstructs a
3D model from a set of primitive objects or templates that is pre-registered in a database. A pattern of
intended strokes is compared with registered patterns,
and the information of the matched pattern is retrieved.

Two main approaches for evocative-based 3D modeling are iconic (Shesh & Chen 2004, Bae et al. 2009),
and template retrieval (Kara et al. 2006, Cheon & Han
2008).
In case of constructive-based modeling, a 3D
model is reconstructed systematically by applying
visual rules on obtained sketch information. Two main
approaches are optimization and progressive. For optimization approach, the depth information of an object
is determined from an objective function that is constrained by geometrical regularity rules (Lipson &
Shpitalni 1995 & 1996, Company et al. 2004, Kang
et al. 2004, Masry et al. 2005, Yuan et al. 2008,
Tian et al. 2009). Instead of identifying 3D coordinates of all vertices simultaneously, the progressive
approach determines the coordinates of the vertices in
order by propagating from a reference vertex (Lamb &
Bandopadhay 1990, Ku et al. 2006, Chansri et al.
2011).
When these technologies are applied alongside rapid prototyping (RP), one of five core
technology areas supporting rapid product development (Bernard & Fischer 2002), a physical prototype can be fabricated quickly. Several researches have been conducted to make
RE-RP direct integration work more effective (Lee &
Woo 2000, Liu et al. 2003). Similar concept has been
applied to form direct GR-RP interface (Soonanon &
Koomsap 2009).
Presented in this paper is an initial attempt to make
direct integration between sketch-based modeling and
rapid prototyping. A stack of contours is created
directly from the information of a single line drawing
241
without going through a CAD model reconstruction

and STL conversion steps. In the next section, related
works on rapid prototyping process and its interfacing with RE and GR are presented. The third section
presents constructive-based modeling, followed by
image-based direct slicing and implementation. The
conclusion is addressed in the last section.
2
RAPID PROTOTYPING AND ITS

INTERFACING WITH REVERSE
ENGINEERING AND GEOMETRICAL
RECONSTRUCTION
Rapid prototyping is an additive manufacturing process that constructs a physical prototype directly from
a CAD model layer by layer without using molds and
dies. As illustrated in Figure 1, RP process starts from
creating a 3D CAD model which can be done on any
commercially available CAD software. The created
model is then tessellated to STL format that is the
input to the slicing step. The sliced contours are used to
generate machine commands for fabricating a physical
prototype. Post processes may be required depended
upon the selected RP technique. Besides traditional RP
process, the concept of direct slicing, in which slicing
operation is performed directly on a CAD model, has
been introduced to alleviate errors resulted from the
approximation in STL conversion step (Jamieson &
Hacker 1995, Sabourin et al. 1996). Direct slicing utilizes the exact boundary of an original CAD model
instead of the approximated boundary of its tessellated
model.
Rapid prototyping has been used closely with
reverse engineering, and research progress in RP has
influenced its integration with RE. As aforementioned,
several researches have been conducted to make RERP direct integration work more effective. It can be
classified into four paths as illustrated in Figure 2.
The first path is commonly found in practice. Point
cloud data are acquired from an object and gone
through regular RE process to reconstruct a 3D CAD
model. The obtained model is then fed into RP process. Long reconstruction time, utilizing 9095% of
the RE processing time (Schoene & Hoffmann 1997),
has motivated the second path where model reconstruction step is skipped and an STL file is generated
from entire point clouds directly (Liu et al. 2006).
Direct slicing concept has been applied in the third path
to form direct interface between RE and RP. Rather
than going through STL conversion step, sliced file
is created directly from point cloud data (Lee & Woo
2000, Liu et al. 2003). Selective data acquisition has
been introduced as an alternative to the first three
paths. Instead of acquiring entire point cloud data as
for the techniques in the first three paths, it scans an
object selectively from the bottom up according to the
complexity of the object (Rianmora et al. 2009). The
scanning result appears as a stack of contours which
can be directly used to generate machine commands.
Figure 3 illustrates two existing paths for GRRP interface. Initially, GR has been researched as a
Figure 1. Basic steps for RP process.
Figure 2. RE-RP interface (Rianmora et al. 2009).
Figure 3. GR-RP interface (Soonanon & Koomsap 2009).
stand-alone process focusing on a 3D model reconstruction. Its interface with RP is through the CAD
model. Recently, layer-based geometrical reconstruction (LBGR) has been introduced as an attempt to
link GR with RP (Soonanon & Koomsap 2009).
LBGR generates a stack of contours directly from a
2D orthographic views drawing. Information on five
orthographic views (i.e., front, back, left, right and top
views) are acquired locally in a top-down direction
242
to create a stack of slabs before homotopy concept is

applied to generate contours from each slab.
CONSTRUCTIVE-BASED MODELING
Freehand sketch is a natural and intuitive way for

portraying ideas and is an effective tool for communicating ideas during the conversion. This simple quick
draft, according to stroke type and medium, can be
classified in to four groups: a non-overtraced online
sketch, an overtraced online sketch, a non-overtraced
offline sketch and an overtraced offline sketch. The
sketch is normally converted to be a single line drawing when constructive-based modeling is applied for
sketch-based modeling as shown in Figure 4.
Among the four types of sketch, obtaining a single
line drawing from a non-overtraced online sketch is
considered to be the simplest one because each stroke
represents a segment on the intended sketch and stroke
information (i.e., starting point and endpoint) is available for all strokes (Shpitalni and Lipson 1995). Usually, these strokes are beautified before their endpoints
are connected to form a single line drawing. In case of
an overtraced online sketch that strokes are randomly
drawn, additional steps are required to group overtraced strokes and to determine their representatives
(Matsuda et al. 1997, Ku et al. 2006, Purch et al. 2007).
On the other hand, a batch of points is only
available information for a paper-based sketch. Few
researchers have developed methods for identifying
a single line drawing from a non-overtraced freehand sketch (Marti et al. 1995, Haron et al. 2003,
Farrugia et al. 2004). With prior knowledge that these
points represent closed contours on a sketch image,
they all have attempted to detect corners and then
link them to form a single line drawing. The number of research is very limited for identifying a single
line drawing from a paper-based overtraced freehand sketch. Recently, Chansri and Koomsap (2009)
have introduced an algorithm for identifying a single
line drawing from a paper-based overtraced freehand
sketch. The algorithm combines a batch of points,
representing overtraced lines, to form a thick line
sketch before boundary extraction is applied to obtain
dual line sketch. Segments representing edges can
be obtained by expanding internal boundary contours
and/or shrinking external boundary contour simultaneously. The obtained discrete segments will then be
connected to form the line drawing of the sketch.
After a single line drawing is obtained, the 3D model
is reconstructed by using one of the two approaches,
mentioned earlier. For optimization-based approach,
the depth information of an object is determined from
an objective function that is constrained by geometrical regularity rules (Lipson & Shpitalni 1995 & 1996,
Company et al. 2004, Kang et al. 2004, Masry et al.
2005, Yuan et al. 2008, Tian et al. 2009). Geometrical regularities are geometrical relationships between
a single line drawing on a 2D plane and a 3D model
on a 3D space. They are extracted from relationships
Figure 4. Constructive-based modeling.
among lines on the drawing, and their values are

obtained from coordinates of junctions before weighting coefficients to an objective function. By optimizing the objective function, 3D coordinates of all
vertices are determined. Several regularities have been
applied, including face planarity, line parallelism,
line verticality, isometry, corner orthogonality, skewed
facial orthogonality, skewed facial symmetry, line
orthogonality, minimum standard deviation of angles,
face perpendicularity, prismatic face, line colinearity
and planarity of skewed chains.
Instead of identifying 3D coordinates of all vertices simultaneously and iteratively, the progressive
approach determines the coordinates of the vertices in
order by propagating from a reference vertex (Lamb &
Bandopadhay 1990, Ku et al. 2006, Chansri et al.
2011). The principal axes are pre-defined before the
selection of a reference vertex is required. A junction with the lowest position from the largest principal
plane (Lamb & Bandopadhay 1990), a junction with
angular distribution graph (Lipson & Shpitalni 1995,
1996, Kang et al. 2004, Masry et al. 2005, Chansri et al.
2011), and a junction with the shortest length from the
longest lines (Ku et al. 2006) have been applied to
represent a the reference vertex. The 3D coordinates
of all vertices then are propagated from the selected
reference vertex by applying plane equation (Ku et al.
2006).
243
Figure 6. A single line drawing and its projected face

images on XY - and YZ-planes.
Figure 5. The steps of image-based direct slicing concept.
IMAGE-BASED DIRECT SLICING
Normally, after the coordinates of all vertices are

obtained from a single line drawing, a 3D CAD model
is reconstructed and when a physical prototype is
needed quickly, the created model is fed into regular
RP process. Presented in this section is an image-based
direct slicing concept for direct interfacing sketchbased modeling with RP. This concept is intended to
generate a stack of contours directly from a single line
drawing by skipping the model reconstruction and STL
conversion steps.
For this image-based direct slicing concept, a stack
of contours are identified from faces obtained during
identifying the 3D coordinates of vertices (Chansri
et al. 2011). These faces are projected onto the two
vertical principal planes (XY - and YZ-planes) where
slicing operation is conducted. The contours are constructed by linking segments obtained from slicing the
faces.
This slicing approach as illustrated in Figure 5 composes of three main steps: preparing a slice pattern,
slicing face images, and constructing a stack of contours. The input to this approach is an isometric single
line drawing and its projected face images on the vertical principal planes as shown in Figure 6. The process
starts from identifying a slice pattern which is a set
of horizontal lines whose space between two lines
represents a layer thickness. Slicing process is done
by intersecting the slice pattern on the face images.
The intersection points are used to form contours. The
details of all steps are as follows.
4.1
Preparing a slice pattern
The main purpose of this step is to create a slice

pattern. The process starts from identifying and then
creating the image of the smallest rectangular boundary that can contain the largest face. 1-pixel-wide
Figure 7. Slicing process.
horizontal lines are drawn at equidistance on the image

from the bottom up. This distance is a layer thickness
defined by a user.
4.2 Slicing face images
For this step, the faces are sliced one by one. For each
face, its two projected faces are intersected with the
slice pattern. The results are then recorded layer by
layer from the bottom-up. Unless a face is a simple
convex shape, intersecting a slice pattern with the face
boundary will create some difficulty. They may appear
several intersection points and line segments of which
addition process is required to identify pairs of two
intersection points forming segment that belong to the
face. Therefore, interiors inside the boundary are filled
for all projected faces prior to performing slicing operation as illustrated in Figures 7 and 8. The results from
244
Figure 10. The result from applying contours creation

process on the sample.
Figure 8. Slicing a complicated face.
Figure 11. The user interface screen of image-based direct

slicing program.
4.3 Creating contours
Figure 9. Applying the slicing process on the sample

drawing.
the intersection will be only points and segments that

belong to the faces.
Figure 9 illustrates the execution of this step on
the single line drawing in Figure 6. For this particular example, the six faces, excluded the topmost face
(F0) and the bottommost face (F7), are projected on
XY - and YZ-planes, and filled their interiors before
intersecting with the slice pattern.
By using intersection points obtained from the previous step, contours are constructed one by one. Starting
from the bottommost contour, intersection points are
retrieved from associated faces. The horizontal coordinate of each of these intersection points are determined
from the two projected faces that it appears. All coordinates are rearranged; transformed into objects world
coordinates and connected to form a closed contour.
This process repeats until reaching the topmost contour. All contours then are combined to form a stack of
contours. Figure 10 shows the result of applying this
step to the intersection points obtained from Figure 9.
IMPLEMENTATION
The three steps of this image-based direct slicing have

been implemented on LabVIEW program. Figure 11
245
Figure 12. Examples of image-based direct slicing

program.
shows the user interface screen of this image-based

direct slicing program. The input to the program, a
single line drawing, is displayed on the top-left side
of the screen. The five sections at the bottom part of
the screen show work in process. A stack of contours,
the output, is displayed on the top-right corner. The
program has been initially tested with several examples
and a couple of them are shown in Figure 12.
6
CONCLUSION
Image processing technique has been applied in this

research for generating a stack of contours from a
single line drawing without going through a CAD
model reconstruction and STL conversion steps. For
its implementation, a program has been developed and
tested. Extended works are expected for handling a
sketch that contains features.
REFERENCES
Bae, S.H., Balakrishnan, R. & Singh, K. 2008. ILoveSketch:
as-natural-as-possible sketching system for creating 3D
curve models. Proceedings of the 21st Annual ACM
Symposium on User Interface Software and Technology
Monterey, California, USA. 151160.
Bernard, A. & Fischer, A. 2002. New trends in rapid product development. CIRPAnnals ManufacturingTechnology
51(2): 635652.
Chen, K.Z. & Feng, X.A. 2003. Solid model reconstruction
from engineering paper drawing using Genetic Algorithm.
Computer-Aided Design 35: 12351248.
Cheon, S.U. & Han, S.H. 2008. A template-based reconstruction of plane-symmetric 3D models from freehand
sketches. Computer-Aided Design 40: 975986.
Company, P., Contero, M., Conesa, J. & Piquer, A. 2004. An
optimization-based reconstruction engine for 3D modeling by sketching. Computers & Graphics 28: 955979.
Company, P., Piquer, A., Contero, M. & Naya, F. 2005. A
survey on geometrical reconstruction as a core technology
to sketch-based modeling. Computer & Graphics 29(6):
892904.
246
Chansri, N. & Koomsap, P. 2009. Dual line extraction

for identifying single line drawing from paper-based
overtraced freehand sketch. Proceedings of the 16th ISPE
International Conference on Concurrent Engineering,
Taipei, Taiwan 455463.
Chansri, N., Koomsap, P. & Kuagoolkijgarn, P. 2011. Application of contour tracing algorithm in 3D modeling from
a paper-based overtraced freehand sketch. Proceedings of
the 21th International Conference on Flexible Automation
and Intelligent Manufacturing, Taichung, Taiwan.
Farrugia, P.J., Borg, J.C., Camilleri, K.P., Spiteri, C. &
Bartolo, A. 2004. A cameraphone-based approach for
the generation of 3D models from paper sketches. Proceeding of EUROGRAPHICS Workshop on Sketch-Based
Interfaces and Modeling 3342.
Haron, H., Mohamed, D. & Shamsuddin, S.M. 2003. Extraction of junctions, lines and regions of irregular line
drawing: the chain code processing algorithm. Jurnal
Teknologi 38(D):128.
Jamieson, R. & Hacker, H. 1995. Direct slicing of CAD models for rapid prototyping. Rapid Prototyping Journal 1(2):
412.
Kang, D., Masry, M. & Lipson, H. 2004. Reconstruction of
a 3D object from a main axis system. Proc. of the American Association for Artificial Intelligence Fall Symposium
on Making Pen-based Interaction Intelligent and Nature,
6369.
Kara, L.B., DEramo, C.M., Shimada, K. 2006. Pen-based
styling design of 3D geometry using concept sketches
and template models. Proceedings of the 2006 ACM
Symposium on Solid and Physical Modeling Conference,
Cardiff, Wales, United Kingdom.
Ku, D.C., Qin, S.F. & Wright, D.K. 2006. Interpretation of overtracing freehand sketching for geometric
shapes. Proc. WSCG 2006 Conf. in Central Europe on
Computer Graphics, Visualization and Computer Vision,
Plzen, 263270. Czech Republic.
Lamb, D. & Bandopadhay, A. 1990. Interpreting a 3D object
from a rough 2D line drawing. Proc. the 1st IEEE Conf.
on Visualization, San Francisco, California, USA. 5966.
Lee, K.H. & Woo, H. 2000. Direct integration of reverse
engineering and rapid prototyping. Computer & Industrial
Engineering 138: 2138.
Lipson, H. & Shpitalni, M. 1995. A new interfaced for conceptual design based on object reconstruction form a single freehand sketch. Annals of the CIRP-Manufacturing
Technology 44: 133136.
Lipson, H. & Shpitalni, M. 1996. Optimization based reconstruction of a 3D object from a single freehand line
drawing. Computer Aided Design 28: 651663.
Liu, G.H., Wong, Y.S., Zhang, Y.F. & Loh, H.T. 2003.
Modeling cloud data for prototype manufacturing. Journal of Materials Processing Technology 38: 5357.
Liu, Z., Wang, L. & Lu, B. 2006. Integrating cross-sectional
imaging based reverse engineering with rapid prototyping.
Computers in Industry 57: 131140.
Masry, M., Kang, D. & Lipson, H. 2005. A freehand sketching interface for progressive construction of 3D objects.
Computers & Graphics 29: 563575.
Marti, E., Regions, J., Lopez-Krahe, J. & Villanueva, J.J.
1995. Line drawing interpretation as polyhedral objects
to man-machine interaction in CAD systems. Pattern
Recognition and Image Analysis 112.
Matsuda, K., Sugishita, S., Xu, Z., Kondo, K., Sato, H. &
Shimada, S. 1997. Freehand sketch system for 3D geometric modeling. Proceedings of International Conference
on Shape Modeling and Applications, Aizu-Wakamatsu,
Japan 5562.
Pusch, R., Samavati, F., Nasri, A. & Wyvill, B. 2007. Improving the sketch-based interface: forming curves from
many small strokes. The Visual Computer: International
Journal of Computer Graphics 23: 955962.
Rianmora, S., Koomsap, P. & Hai, D.P. 2009. Selective data
acquisition for direct integration of reverse engineering
and rapid prototyping. Virtual and Physical Prototyping
4: 227239.
Sabourin, E., Houser, S.A. & Bhn, J. H. 1996. Adaptive slicing using stepwise uniform refinement. Rapid Prototyping
journal 2: 2026.
Schoene, C. & Hoffmann, J. 1997. Reverse engineering
based on multiaxis digitizer data. Proc. Int. Conf. on
Manufacturing Automation (ICMA 97).
Shpitalni, M. & Lipson, H. 1995. Classification of sketch
strokes and corner detection using conic sections and
adaptive clustering. Journal of Mechanical Design 119:
131135.
Shesh, A., & Chen, B. 2004. SMARTPAPER: an interactive
and user friendly sketching system. Computer Graphics
Forum 23: 301310.
Sokovic, M. & Kopac, J. 2005. RE (Reverse Engineering)

as necessary phase by rapid product development.
Journal of Material Processing Technology 175:
398403.
Soonanon, P. & Koomsap, P. 2009. Towards direct transformation of orthographic-view drawings into a prototype.
Virtual and Physical Prototyping 4: 7590.
Tam, K.S. & Atkinson, J. 2003. An approach for creating
solid models from orthogonal views by identification
of Boolean operations. Journal of Materials Processing
Tian, C., Masry, M. & Lipson, H. 2009. Physical sketching:
reconstruction and analysis of 3D objects from freehand
sketches. Computer Aided Design 41: 147158.
Yuan, S., Tsui, L.Y. & Jie, S. 2008. Regularity selection
for effective 3D object reconstruction from a single line
drawing. Pattern Recognition Letters 29: 14861495.
247

Design and additive manufacturing of cellular lattice structures

Liang Hao, David Raymont, Chunze Yan, Ahmed Hussein & Philippe Young
College of Engineering, Mathematics and Physical Sciences, University of Exeter, Exeter EX4 4QF, Devon,
United Kingdom
ABSTRACT: The concept of designed cellular lattice materials is motivated by the desire to put material only
where it is needed for a specific application. From a mechanical engineering viewpoint, a key advantage offered
by cellular materials is high strength accompanied by a relatively low mass. These materials can provide good
energy absorption characteristics and good thermal and acoustic insulation properties as well. Designed cellular
structures typically exhibit strong structure strength per unit weight than typical foam structures. However, due
to their complexity, these structures are often difficult to generate using existing CAD packages. Furthermore,
metallic additive manufacturing techniques, such as selective laser melting process which shows the great
capability to fabricate strong and lightweight metallic lattice structures, are still facing certain process limitations
in terms of the geometrical capability and support structure requirement for the fabrication of cellular lattice
structures. This paper presents an efficient approach to generate and design periodic lattice structures and
investigates the manufacturability of some selected structures using selective laser melting (SLM) process.
The design of cellular structures is based on image-based algorithms to efficiently generate implicitly defined
periodic lattice structure and rapidly construct volume and surface meshes. The experimental investigation on
the SLM fabrication has studied the effects of unit cell type and cell sizes on the manufacturability of some
typical cellular structures.
1
INTRODUCTION
The design and manufacturing of cellular lattice structures is motivated by the desire to save the expensive
functional materials, build time, energy consumption, and provide high performance such as high
strengths accompanied by a relatively low mass, good
energy absorption characteristics and good thermal
and acoustic insulation properties to aerospace, medical and engineering products. However, due to their
complexity, these structures are often difficult to
generate using existing CAD packages. Currently,
there is great interest in manufacturing cellular lattice
structures with tailored properties. A number of manufacturing methods including investment casting, deformation forming, brazing etc. are proposed to make
cellular structures. Limitations of such techniques
include the fitness of the structures, and the actual cell
geometry (Santorinaios, et al. 2006). Additive Manufacturing (AM) technologies have been developed
over the years to produce three dimensional objects
directly from a digital model by the successive addition
of materials without the use of a specialized tooling. Selective laser melting (SLM) is one of the most
widely used metal AM processes and has the potential to manufacture cellular lattice structures with fine
features at a resolution of 50 microns.
Therefore, this paper presents an efficient approach
to the design and generation of periodic lattice structures and investigates the manufacturability of some
selected structures using the SLM process.
BACKGROUND
2.1 Design of cellular structures

Recent advances in AM have allowed for the creation
of complex geometries to a relatively high level of precision. These manufacturing processes are particularly
well-suited to fabricating computationally generated
cellular lattice structures and have allowed for the
recent development of methods for doing so.
As with the design of the vast majority of components that are to be manufactured, cellular lattice
structures have previously been created using traditional commercial CAD packages. However these
packages have proven to be unsuitable for potentially
large complex micro-architectures due to the vast number of Boolean operations required, as shown in Wang,
et al. 2005. To overcome the difficulties in using
generic CAD packages Wang, et al. 2005 developed
a hybrid geometric modelling method for conformal
truss structures. Their method was demonstrated to be
able to create large triangulated surfaces of repeating
unit cells, such as the tetrahedron and Kelvin Foam
structures. By using open-ended cylinders for trusses
and sealing joins with spheres, repeatable unit cells
were generated. This, however, relied on the use of a
solid modeller (ACIS) to perform the Boolean operations on the trusses and spheres. These methods were
further developed in Chen 2006 to remove the need for
a solid modeller by developing methods for handling
the tessellation of truss-sphere intersections.
249
Voxel modelling is an alternative approach to the

generation of cellular lattice structures and is often
used in the creation of scaffold architectures. While
high resolution images or volumes are normally
required to sufficiently represent geometries using
voxels, they have the advantage of being particularly
straightforward to modify, particularly when using
Boolean operations.
A relatively simple image-based approach to the
generation of conforming scaffold architectures is presented in Starly 2006. In this work Starly slices the
bounding geometry, as defined using a CAD model,
into a number of equally spaced binary images. By
using Boolean operations on each slice a number of
simple unit cells are then introduced into the geometry.
The unit cells themselves are typically solid cubes with
spherical or cuboidal voids, thus ensuring they remain
stackable. This slice-based approach avoids the need
to handle triangulated surfaces for the creation of an
STL file. However, this is likely limited to 3D printing
where image-based slices may be used. As with any
purely voxel-based method, it also results in a poorly
defined geometry at the boundaries.
Another less frequently used approach to the generation of micro-architectures is through the use of
implicit functions. This is the approach taken in the
works by Gabbrielli (Gabbrielli 2008; Gabbrielli, et al.
2009) and more recently by Pasko (Pasko, et al. 2010;
Pasko, et al. 2011). Gabbrielli uses a set of periodic
implicit functions, such as the Schoen Gyroid (Schoen
1970), to create porous micro-architectures to be used
as bone substitutes. By introducing functional variations to the equations Gabbrielli was also able to create functionally graded micro-architectures. However,
there were no methods given for precisely controlling the grading, such as the minimum and maximum
volume fractions. In comparison to other techniques
these methods provide a compact representation of the
complex structures and, through the use of an appropriate iso-surfacing algorithm, a straightforward way
of producing triangulated surfaces.
2.2
SLM fabrication of cellular structures
Recently, some attempts have been made to create cellular lattice structures using SLM process.
Santorinaios, et al. 2006 studied the manufacturability
of open cellular lattice structures with a simple geometry with vertical struts and cross bracing. Three cell
sizes of 1.25, 2.5 and 5 mm were considered, but the
cell size of 5 mm proved to be problematic to fabricate.
The struts tended to sag during the SLM process.
Brooks, et al. 2005 designed and manufactured regular metallic lattice structures with unit cell sizes in
the range 0.8 mm to 5 mm and truss elements of 100
500 m in diameter through the SLM of 316L stainless
steel. McKown, et al. 2008 made a range of metallic
lattice structures based on [45 ] and [0 , 45 ] unitcell topologies by SLM process. The unit cell sizes
of their lattice structures were 1.5 mm and 2.5 mm.
However, the cellular structures manufactured in the
current studies might not exhibit good manufacturability in SLM, and therefore current cellular structures
with large unit cell sizes (greater than 5 mm) could
not be built using the SLM process because of the
occurrence of serious deformation.
In this paper, two unit cell types are chosen to design
periodic cellular lattice structures with a volume fraction of 15% and the manufacturability of cell sizes in
the range of 2 mm to 8 mm are investigated for SLM.
3
DEVELOPING DESIGN TOOLS FOR

CELLULAR STRUCTURE GENERATION
3.1 Surface representations

When dealing with solid 3D objects it is often convenient to only model the objects boundaries using
a mathematical representation of the surfaces. The
choice of surface representation is particularly important for the computational modelling of 3D objects as
each has its own advantages and disadvantages. These
include the availability and complexity of operations
that can be used to manipulate the surface (e.g. smoothing, Boolean operations) as well as the efficiency of the
representation. Each representation also has an impact
on how models are visualised and ultimately realised
(e.g. via rapid manufacturing). The most common representations can be classified as one of the following
forms: explicit, parametric or implicit.
In explicit surface representations points which lie
exactly on the surface are explicitly stored. The most
common type of explicit surface is a mesh of polygons;
typically these are triangles or quadrilaterals although
others can be used. These polygons are often stored as
an order list of vertex indices the order being used
to define the direction the polygon is facing (i.e. the
surface normal).
Unlike explicit surfaces, parametric surfaces do
not store points on a surface. Instead, points on a
parametric surface are expressed as a function of
the parametric variables (u, v), which can be generalised to lie on the unit square [0,1] [0,1] (Zheng
2008). Non-Uniform Rational B-Spline (NURBS) surfaces are a form of parametric surface commonly used
in CAD packages due to their compact representation, smooth surfaces and easy of manipulation. Other
forms of parametric surfaces exist, such as Rational Gaussian (RaG) surfaces (Goshtasby 1993) and
Fourier Shape Descriptions.
Implicit surfaces are defined as an iso-surface of
some function f . In 3D the surface is defined by a set
of points p R3 satisfying the equality:
where f : R3 R
As with parametric forms, implicit surfaces provide
a compact representation for potentially complex surfaces. They also offer a number of advantages, notably
their flexibility (as will be demonstrated later in this
work) and well-defined Boolean operations. However,
250
Table 1.
Implicit surface in/out convention.
Condition
Interpretation
f(x, y, z) = 0
f(x, y, z) < 0
f(x, y, z) > 0
On surface
Inside
Outside
unlike parametric forms they offer little local shape

control and manipulating them can be unintuitive.
The implicit formulation of a unit sphere is given
below:
In this instance the implicit form is not only more

compact, but potentially more useful as the sign of the
function can be used to designate points as either inside
or outside the surface. For this purpose the following
convention is adopted:
The implicit functions of interest to this work are the
set of infinitely periodic surfaces. The most notable
of which are those discovered by Schoen 1970 and
Schwarz 1890. In addition to being infinitely periodic these surfaces are also approximations of minimal
surfaces, that is, the surfaces have a mean curvature
of zero. Using a combination of trigonometric functions in the form given in Eq. (3) a number of periodic
surfaces can be generated (Gabbrielli, et al. 2008).
The simplest triply periodic (or dual periodic in 2D)

function in this form is the Schwarz Primitive; cos(x)+
cos(y) + cos(z) + 1 = 0.
3.2
Mesh generation
Accurate and robust mesh generation is an important

step towards the fabrication of lattice structures. The
work presented in this section looks at methods for
generating image volumes representing the implicit
functions so that both volume and surface meshes
can be constructed. An entirely image-based approach
is taken to exploit the advantages of image-based
meshing.
Methods previously used in (Gabbrielli, et al. 2008)
relied upon the generation of a floating-point volume
which was then iso-surfaced. While this method is
straight-forward, simply requiring that the function be
sampled at regular intervals, it becomes difficult to
generate a volume mesh and integrate with other image
data.
To overcome these difficulties we require that the
generated volumes data-type matches that used by
+
ScanFE from Simpleware Ltd. In the C programming language this is unsigned char, an 8 bit integer.
By using this data-type the generated volumes can easily be combined with data from other sources, such as
medical imaging devices and meshed with + ScanFE.
The most straight-forward translation to imagespace that can be made from an implicit function is
Figure 1. Binary slice representing the Schwarz Primitive

using 20 20 px.
the generation of a binary volume. By evaluating the

function, f , over a range of values voxels can be determined to be either inside or outside and their value set
accordingly.
Despite being efficient to generate, the binary representation yields a poor reconstruction, as can be seen
in Figure 1. The reconstructed surface can be improved
by introducing greyscale values into the volume. This
can be achieved using a smoothing algorithm such
as Gaussian smoothing, however these algorithms can
have adverse effects such as shrinking the volume and
removing small features. A more appropriate solution is to generate the volume with greyscale values
such that they result in the reconstructed surface being
placed as close as possible to the ideal surface. The
marching cubes algorithm will be used to generate a
triangulated surface, as such, the volume can be generated so as to best utilise the greyscale values. As the
position of the reconstructed surface is only dependent
on the two voxels either side of it, a small region of
greyscale values should be placed either side of the
ideal surface. These greyscale values should reflect
their distance to the surface, mimicking the partial volume effect. Voxels further from the surface may simply
be marked as inside or outside. However, unlike many
implicit functions used in computer graphics, the functions of interest are not distance functions.That is, their
value does not reflect a linear measure of distance from
the surface. To overcome this a point is chosen, from
the discretising volume, that is close to the ideal surface (i.e. such that f(x, y, z) 0). The gradient at this
point is then computed, allowing the greyscale values
to be set such that their values reflect their distance
from the surface. Values inside of the surface may be
calculated as follows:
where Gp is the gradient near the surface. Figure 2

shows an example from a volume generated using
this method. The reconstructed surface is shown in
Figure 3.
251
Figure 2. Greyscale slice representing the Schoen gyroid.
Figure 3. Reconstructed surface of the Schoen Gyroid.
4
4.1
Figure 4. Two types of unit cells of Schoen Gyroid and

Schwartz Diamond used in cellular lattice structures.
EXPERIMENTAL STUDY ON SLM OF

METALLIC CELLULAR STRUCTURES
Materials
Cellular lattice structures were made from 316L

stainless steel powders with average particle size of
45 m 10 microns, which was gas atomized and
produced by Sandvir Osprey Ltd., UK.
4.2 The selective laser melting process
The manufacturing process was carried out on a
Realizer SLM Workstation made by MTT Technologies Group, UK. The SLM machine uses a 100 W
CW Ytterbium fibre laser. All processing occurs in
an Argon atmosphere with less than 1.0% O2 . The
processing parameters used in this study were as follows: the laser power was 95 W; the scan speed was
500 mm/s; the scan spacing was 75 m; the layer
thickness was 75 m.
This paper investigates the manufacturability of two
unit cell types for SLM. The STL models of the two
unit cells, the Schoen Gyroid and Schwartz Diamond,

are shown in Figures 4 (a) and (b), respectively. The
volume fraction was set as a constant of 15%. Cellular
lattice structures with the unit cell sizes of 2 mm, 5 mm
and 8mm were manufactured via SLM. The samples
dimensions were 25 25 15 mm3 .
4.3 Effect of unit cell types and cell sizes on the
manufacturabilitys of cellular structures
by SLM process
The Schoen Gyroid cellular structures with a volume
fraction of 15% and unit cell sizes of 2 mm, 5 mm and
8 mm were successfully manufactured using SLM, as
shown in Figures 5 (a)(c), respectively. When the volume fraction is set as a constant of 15%, the amount of
unit cells in the cellular lattice structures will increase
with the decrease in unit cell sizes, and the struts in the
cellular structures become thinner and thinner. When
the unit cell size is very small, it may result in the
loss of connectivity between adjacent cells. During
252
Figure 5. Schoen Gyroid cellular structures with a volume

fraction of 15% and unit cell sizes in the range of 2 mm to
8 mm were successfully manufactured through SLM.
the SLM process, the thin struts may be damaged by

the wiper, resulting in it becoming difficult to remove
loose powders following the completion of the build.
From Figure 5, it can be seen that the Gyroid cellular structure with a cell size of 2 mm has the greatest
number of unit cells and the thinnest struts. However, the struts were not damaged by the wiper of
SLM machine and the loose powder was completely
removed. When the unit cell size is increased, the number of unit cells in the cellular structures decreases and
the struts become stronger and longer. However, as the
struts become longer, the amount of overhang in the
structures increases and deformation may occur during the build process. This is why Santorinaios, et al.
2006 could not build their cellular structures with a
cell size of 5 mm. From Figure 6(c), it can be seen
that the cellular structure with a cell size of 8 mm has
the smallest amount of unit cells and strongest struts,
and no obvious deformation was observed during the
build process. This can be attributed to the fact that
the Schoen Gyroid unit cell has smooth struts rather
than straight beam-like struts, and therefore enable
SLM process to manufacture larger overhangs. Therefore the Gyroid cellular structures with bigger unit cell
sizes can be manufactured by SLM.
The Schwartz Diamond cellular structures with the
same volume fraction of 15% and different unit cell
sizes of 2 mm, 5 mm and 8 mm were also successfully manufactured through SLM process, as shown
in Figures 6 (a)(c), respectively. As with the Schoen
Gyroid cellular structures, the Schwartz Diamond
cellular structure with a cell size of 2 mm was manufactured with no struts damaged and the loose powders
were completely removed, and the Diamond structure
with a cell size of 8 mm was also built without any
deformation.
Figure 6. Schwartz Diamond cellular structures with a volume fraction of 15% and unit cell sizes in the range of 2 mm
to 8 mm were successfully manufactured through SLM.
CONCLUSIONS
The use of implicit modelling, particularly with triply

periodic functions, has been shown to be highly flexible for the generation of open cellular structures.
Several image-based algorithms were developed and
evaluated and it was clearly shown that, to achieve
accurate and smooth models, the generation of the
volume with greyscale values would reconstruct more
ideal surface. By taking an image-based approach it
has been possible to ensure that the generation and
manipulation of the cellular lattice structures remains
robust, despite any geometric complexities.
The effect of unit cell types and cell sizes on the
manufacturability of cellular structures by SLM was
investigated. Two unit cell types, the Schoen Gyroid
and Schwartz Diamond, were considered. The results
reveal that the Schoen Gyroid and Schwartz Diamond
cellular structures with the same volume fraction of
15% and different unit cell sizes of 2 mm, 5 mm and
8 mm are manufacturable through the use of SLM. The
cellular lattice structures with a cell size of 2 mm were
manufactured with no struts damaged and the loose
powders were completely removed. The cellular structures with a cell size of 8 mm were built without any
obvious deformation.
ACKNOWLEDGEMENTS
This work has been supported by EPSRC Industrial CaseAward (CASE/CAN/07/86), Simpleware Ltd
Studentship Sponsorship and UK Technology Strategy Board (TSB) Research Project (BA036D). The
TSB funded project is entitled SAVING Sustainable
253
product development via design optimisation and

additive manufacturingand is a collaboration between
the Simpleware Ltd, Delcam PLC, University of
Exeter, 3T RPD, Crucible Industrial Design Ltd, EOS
Electro Optical Systems Ltd and Plunkett Associates
Ltd.
REFERENCES
Brooks, W.K., Tsopanos, S., Stamp, R., Sutcliffe, C.J.,
Cantwell, W.J., Fox. P., Todd, J. 2005. The production
of open cellular lattice structures using selective laser
melting. 6th National Conference on Rapid Design, Prototyping, and Manufacturing, Buckinghamshire Chilterns
University College, June 2005.
Chen, Y. 2006. A mesh-based geometric modeling method
for general structures. ASME Conference Proceedings:
9281.
Gabbrielli, R. 2009. Foam geometry and structural design of
porous material. Ph.D. thesis, Department of Mechanical
Engineering, University of Bath.
Gabbrielli, R., Turner, I.G. & Bowen, C.R. 2008. Development of modelling methods for materials to be used as
bone substitutes. Key Engineering Materials 361363:
903906
Goshtasby, A. 1993. Design and recovery of 2-d and
3-d shapes using rational gaussian curves and surfaces.
International. Journal of Computer Vision (IJCV) 10:
233256.
McKown, S., Shen,Y., Brookes W.K., Sutcliffe, C.J., Cantwell

W.J., Langdon, G.S., Nurick, G.N., Theobald, M.D. 2008.
The quasi-static and blast loading response of lattice structures. International Journal of Impact Engineering 35(8):
795810.
Pasko, A., Vilbrandt, T., Fryazinov, O., Adzhiev, V. 2010.
Procedural function-based spatial microstructures. 2010
Shape Modeling International Conference : 4756.
Pasko, A., Fryazinov, O., Vilbrandt, T., Fayolle, P. & Adzhiev,
V. 2011. Procedural Function-based Modelling of Volumetric Microstructures. Graphical Models.
Santorinaios, M., Brooks, W, Sutcliffe, C.J., Mines, R.A.
2006. Crush behaviour of open cellular lattice structures manufactured using selective laser melting. High
Performance Structures and Materials Iii 85: 481490.
Schoen, A.H. 1970. Infinite periodic minimal surfaces without selfintersection. Tech. rep., National Aeronautics
and Space Administration, Electronics Research Center,
Cambridge, Mass. 02139.
Starly, B. 2006. Biomimetic Design And Fabrication Of Tissue Engineered Scaffolds Using Computer Aided Tissue
Engineering. Ph.D. thesis, Department of Mechanical
Enginerring and Mechanics, Drexel University.
Schwarz, H. 1890. Gesammelte mathematische abhandlungen. Springer-Verlag, Berlin.
Wang, H., Chen, Y. & Rosen, D.W. 2005. A hybrid geometric modeling method for large scale conformal cellular
structures. ASME Conference Proceedings: 421427.
Zheng, B. 2008. 2D Curve And 3D Surface Representation
Using Implicit Polynomial And Its Applications. Ph.D.
thesis, The Graduate School Of The University Of Tokyo.
254

The Corinthian capital according to Alberti: Generative modelling and

digital prototyping of classical architectural elements
Eduardo Castro e Costa
TU Lisbon, Portugal
Filipe Coutinho
University of Coimbra, Portugal
Jos Pinto Duarte

TU Lisbon, Portugal
Mrio Krger
University of Coimbra, Portugal
ABSTRACT: The main objective of the ongoing investigation described in this paper is to produce a Corinthian
capital through generative modelling and digital prototyping, and according to Leon Battista Albertis treatise on
architecture, De re aedificatoria. This investigation is part of the Digital Alberti research project, which aims to
decodeAlbertis treatise through the use of new technologies.This treatise can be interpreted as a set of instructions
regarding the art of building. Some of these instructions were translated into computational models, capable
of generating digital instances of column elements according to the classical canons prescribed in the treatise.
These instances were then digitally produced, materializing Albertis theory. Despite Albertis thoroughness, some
detailing information is missing, namely for the Corinthian capital; observation and computational geometry
were important for filling in the gaps. Such investigation allows determining the suitability of different modelling
strategies, as well as the potential of different prototyping technologies.
Keywords: Alberti, De re aedificatoria, generative modelling, digital prototyping, computational geometry
CONTEXT
The activities described further are but a small part

of a research project named Digital Alberti. The
projects main objective is to determine the influence of Albertis theory on Portuguese architecture,
by decoding his treatise on architecture, De re aedificatoria, through the use of digital computational
technologies. Namely,it aims to determine such an
influence by converting the treatise into a shape grammar and then determining the changes to the shepa
grammar required to account for the generation of
Portuguese classical buildings.
Currently, two shape grammars are being developed, one for the column system and another for
churches, describing the different levels of detail found
in Albertis treatise. Eventually the two grammars
will converge, serving as a base for a compositional
analysis tool.
Among the many tasks necessary to achieve the
main objective is the production of models that materialize, both virtually and physically, the treatises
content. The work describe further is part of an effort
to model and produce all the column system elements,
so to support the development of the corresponding

shape grammar.
Albertis treatise holds his considerations on the
art of building. This document was first published
in 1485, without illustrations, although in many later
editions illustrations were added by other authors.
Therefore, besides aiding to the construction of the
shape grammar, the tasks outputs will serve as the
treatises missing illustrations, in the form of drawings, digital three-dimensional models and interactive
computational and virtual models. These elements will
be featured in an exhibition to be held at the end of the
Digital Alberti project, as well as in a published book
and in an educational software about the subject.
2
METHODOLOGY
Prototyping the Corinthian capital implied two main

processes: a) translating the part of the treatise describing the capital into a computational model, and b)
generating the physical models. The translation of
the treatise comprises three phases: i) understanding
the treatise, decoding the instructions prescribed by
255
Table 1.
Methodology.
Treatise Computational model Physical model

Translating the treatise
Understanding Building the
the treatise
computational model
Filling in
the gaps
Figure 2. Spreadsheet and visual programming.
three, and at least one of the elements is a leaf, that

is, it has no hierarchically inferior elements, or children. Also a pattern was identified, in which rules
follow a general form
Figure 1. Reading and annotation.
Alberti; ii) building the computational model by implementing such instructions into a computer program;
and iii) filling in the gaps by completing the algorithms
where instructions are missing (Table 1).
3
UNDERSTANDING THE TREATISE
The first step in understanding Albertis treatise was its

analytical reading, focusing on Books Six and Seven,
which encompass chapters describing the column
systems. In these chapters, one can find several definitions regarding these elements, which will henceforth
be called rules. In order to better understand these
rules, a thorough analysis was undertaken of one of
the elements, the Doric base, described in chapter 7 of
Book Seven. The same process was then applied to the
Corinthian capital, a more complex element.
3.1 Reading and annotation
Consider the following passage from the treatise: The
height of the base was then divided into three parts,
one of which was taken up by the thickness of the
die (page 207, Book Seven, Chapter 8 Alberti,
1988). This rule defining the height of the die, can
be mathematically described as
in which the height of the die corresponds to hdie ,

and the rest of the phrase corresponds to the latter part
of the equation. Therefore, while reading the text, the
rules were highlighted, and translated into equations.
Essential to the treatises analysis was the annotation effort, composed of sketching the column elements, so as to clarify the proportional relations prescribed by the rules, and scribbling their mathematical
translations, allowing them to be put in perspective,
and seen as a whole (Figure 1).
This led to the identification of a structure, in which
rules are organized in a hierarchical tree. Each hierarchical level comprises few elements, typically two or
where A represents a parameter of the column element (typically a dimension), parent(A) represents
As hierarchical superior (its reference), and k represents a constant value, prescribed by Alberti, usually
in the form of a fraction (multiplier). In the previous
example, A would be the height of the die, parent(A)
would be the height of the base, and k would be one
third.
3.2 Structuring the rules
As the complexity of the system grew, writing down
the rules on paper or through word processing proved
insufficient. A relational database, however, provided
the structure needed for a more efficient record of the
rules, surpassing the two dimensions of the notebook.
This upgrading was crucial for eliminating redundant
variables, and therefore to a better understanding of
the links among rules.After debugging, the rule system
was tested in a spreadsheet, namely for the Doric base,
calculating its values and identifying the trunks and
leaves of the tree structure.
The next step would be the implementation of the
rules describing the Doric base through a computational algorithm. The selected tool was Grasshopper,
a visual programming plugin for the geometric modelling software Rhinoceros (Figure 2).
Due to the previous structuring effort, the move
to Grasshopper proved quite simple. The rules were
implemented so that the elements parameters could
be adjusted, hence analyzing if Albertis prescriptions
are in fact ideal. The result was a preliminary version
of the computational model, from which subsequent
models were then developed.
4
BUILDING THE COMPUTATIONAL MODEL
In the development of the computational model for

the Corinthian capital, the same methodology was
followed. Albertis rules for this element, found in
256
Table 2.
Table 2.
Rules for modelling the Corinthian capital.
Continued.
English edition of the treatise (Alberti, 1988). Accompanying these tables is a scheme that summons the
hierarchical structure of the capitals components and
the corresponding rules (Table 3).
Since the capital is itself part of a hierarchical structure the column system , the system that generates
it needs some kind of input. Two initial variables were
selected, the diameters of the imoscape and of the
sumoscape, respectively the lower and upper end of
the columns shaft, and represented by Dimoscape
and Dsumoscape. Actually,Dsumoscape depends on
Dimoscape, according to the column height, as stated
in Book Seven, Chapters 6 and 7, pages 201-202. However, for the sake of simplicity, they are considered as
independent variables, disregarding the dimensional
proportion between them.
Figure 3 details the steps in the modelling of the
Corinthian capital. Three main types of operations
can be identified: proportion, subdivision, and details
the steps in the modelling of the Corinthian capital.Three main types of operations can be identified:
proportion, subdivision, and detailing. These operations were implemented through a visual programming
interface, Grasshopper, into a highly parameterized
system, allowing variations in the parameters connected to proportion and subdivision operations, thus
obtaining plausible variations of the prescribed capital.
The detailing operations, however, are based on
something other than the treatise alone. This is because
Alberti prescribes only a few features for some elements, hardly enough to determine their exact shape.
This is the case for the sprouting stalks, the acanthus
leaves and the sprouting flower. In order to achieve
a satisfactory level of detail, some gaps had to be
filled in.
5
Chapter 8 of Book Seven of the treatise, were systematized, generating the correspondent hierarchical
structure. The tables shown next (Table 2) describe the
systematized rules, internally numbered, along with
their mathematical translation, and their location in the
FILLING IN THE GAPS
As the treatise was insufficient for generating the missing pieces, it was necessary to look into other sources,
such as illustrations of later editions of De re aedificatoria, and observation of built examples. Illustrations
of other treatises were also consulted, but only aspects
257
Figure 3. Modelling process for the Corinthian capital.
not conflicting with Albertis rules were taken into

account.Then experimentation began by modelling the
missing pieces as close to such sources as possible,
trying to infer the rules that could generate them
according to a series of self-established principles that
guided the modelling process and decision-making,
and explained below.
Table 3.
5.1
Curve modelling principles
Modelling and implementation of the missing pieces

took under consideration the following principles:
Continuity geometrical shapes should keep continuity on highest possible order;
Simplicity geometrical shapes should be as simple
as possible;
Recycling missing rules should be adapted from
existing rules whenever possible, so as to minimize
the use of sources other than Albertis treatise;
Versatility rules should generate transformations that
can be applied to any surface. Versatility will, in turn,
allow extending the Recycling principle.
Although inexistent at the time of Alberti, computational geometry provided important tools for modelling some of the missing pieces, particularly the ones
to which the principles of simplicity and continuity
were relevant. The mathematical model NURBS (nonuniform rational basis spline) proved useful in keeping
these two principles (Pottman,Asperl, Hofer, & Kilian,
2007).
This was thus the framework for detailing the
complex elements of the Corinthian capital.
5.2
Sprouting stalks
The Corinthian capital features two types of stalks

(#18, #19).According to the principle of recycling, and
Rules for modelling the sprouting stalks.
#15
The remaining space is taken up by the stalks sprouting out from the leaves to the full height of the
vase.
#17
four of them unfold on each face of the capital, two

from the same knot on the right, and two from the
same knot on the left;
#18
the two end ones hang below the corners of the

abacus in a form of spiral,
#19
while the middle ones also curl, so that their ends

meet in the center.
for the sake of coherence, both types were modelled

in the same way, the differences between them being
solely in terms of proportions and the inscribing solid
(hull). This implied the development of an algorithm
according to the principle of versatility. For better
understanding, the larger stalk modelling process will
be described. The generic stalk was modelled as an
extrusion of a profile, or section, along a generating
curve, or path as explained in following.
Path curve: Regarding the path curve, it is possible to
extract two features from the rules above: its endpoints
(starting point on leaves: #15, ending point on abacus:
#18) and its shape at the end (a spiral: #18). Everything
else had to be filled in. The path was defined as a threedimensional (3D) curve, resulting from a combination
of two 2D-curves contained on orthogonal planes.
Horizontal projection: In the few rules prescribed
for the stalks, Alberti points to a relation between
the larger ones and the abacus (#18). Therefore, in
the model, the horizontal projection of the path curve
258
Table 4.
Figure 4. Stalks path curve horizontal projection.
Rules for modelling the acanthus leaves.
#13
Which cover it with two interlapping rows of leaves

standing out in relief; each row contains eight leaves.
#14
The first row is two modules high, as is the second.
#23
Each leaf should be articulated into five or, possibly,

seven lobes.
#24
The tip of the leaves hang forward half a module.
#25
As with all carving, deeply incised lineaments will

add great charm to the leaves of the capital.
Figure 5. Stalks path curve vertical projection.
Figure 7. Application of acanthus leaf algorithm on vase

and sprouting flower.
benefit from the material qualities of the fillet. In fact,

the use of the fillet, in this case as in many others,
prevents the occurrence of acute angles, (Figure 6)
which are prone to breaking.
Both the profile and spirals algorithms can be used
later in the detailing of the Ionic capital, again honoring
the principle of versatility.
Figure 6. Lineaments; imoscape illustration in Morolli &

Guzzon, 1994.
was designed to conform with the abacus contour

(Figure 4), according to the principle of Recycling.
5.3 Acanthus leaves
Vertical projection: The stalks are said to be sprouting out from the leaves (#15), hinting a vertical
direction for the path curves tangent at its starting
point.They should hang in the form of a spiral or
curl. Following the principle of Recycling, this spiral was designed according to Albertis definition for
the Ionic capital (Book Seven, Chapter 8, page 207),
which generates a two-point involute, to be exact.
In order to complete the vertical projection, connecting the spirals to the leaves, a curve was needed
that had the following features: vertical tangent at start
point; inward tangent at endpoint, so to maintain the
continuity with the spiral curve (principle of continuity); and least possible number of control points
(principle of simplicity). A curve that meets these
requirements is a NURBS curve of the 2nd degree
(see curve modelling principles), with 3 control points
(Figure 5).
Profile: No explicit rules can be found inAlbertis treatise regarding the stalks so-called profile. Therefore,
its shape was adapted from the lineaments prescribed
in the treatise for the bottom of the column shaft (Book
Six, Chapter 13, pages 186188; (Figure 4), on the one
hand to honor the principle of recycling, and also to
Like the Corinthian capital, the algorithm that generates the acanthus leaf by itself is a work in progress,
as it is constantly being upgraded, aiming at two main
objectives: similarity and versatility.
The generated acanthus leaf should be indistinguishable from the numerous sculpted examples found
in classical architecture. This implies an ongoing
two-fold investigation, by extracting the rules both
from a thorough observation of sculpted examples,
preferably in Albertis buildings,and from trying to
understand some of natures geometrical rules. This
second approach has not yet been taken, and it will be
addressed later in the investigation.
The principle of versatility is present in the algorithm on two levels. On the one hand, it is developed
so that its transformations can be applied on any surface, thus allowing it to be used later in the detailing
of the sprouting flower and, in the future, in detailing of the scroll in Ionic capital, or of elements in
entablature (Figures 7 and 8). On the other hand, the
algorithm outputs a surface whose complexity is kept
to a minimum, allowing for a wider range of subsequent geometric operations, such as the subdivision of
the leaf into its lobes (#23).
259
Figure 9. Draft 2 digital and physical models.
Figure 8. Detail of ionic capital; geometry of stylized leaf.

Table 5.
Rules for modelling the sprouting flower.
#20
Between these middle two a flower sprouts prominently from a vase,
#21
as far as the top of the abacus.

Figure 10. Broken leaf tips.
As previously stated, the algorithm carries out a

series of geometric operations on a given base surface, transforming it into a leaf (actually the base
surface is not itself transformed; it merely acts as a
container of geometric parameters used by the algorithm to generate a new shape). The leaves on the
Corinthian capital derive from the sub-surfaces resulting from previous successive subdivision operations
to its vase, culminating in rule #13.
The acanthus leaf was modelled as a NURBS surface, allowing it to keep a reduced number of control
points, and thus conforming to the principle of simplicity. It is the result of a lofting operation between
two curves, the lower one being derived directly from
the vases horizontal sections, and the upper one being
a transformation of the base surface (Figure 8). The
loft was selected for its coherent and thus predictable
behavior concerning UV coordinates. UV control is
crucial for guaranteeing further geometric operations
on the output geometry, the leaf, such as its subdivision
into lobes, and all the subsequent detailing necessary
to shape it according to the sculpted references.
5.4 Sprouting flower
As previously stated, the sprouting flower would benefit from the principle of versatility. In fact, its petals
were modelled using the same algorithm that generates
the stylized acanthus leaves. The base geometry for the
flower was a simple sphere conforming to rule #21.
6
PRODUCING PHYSICAL MODELS
The modelling process described previously featured

three milestones, called drafts. These milestones
represent the states when an output was generated
by the computation model for a specific purpose.The
first drafts output consisted of a video capture of
the computational models manipulation, that was presented at the colloquium Gneses das Racionalidades
Modernas em torno de Alberti in Brazil, in April
2011. The purpose of the second and third drafts was
the digital prototyping of physical models. As a work

in progress, these were also draft physical models,
both to test the development of the digital model and its
conversion into a physical model using two different
prototyping techniques, Fused Deposition Modelling
(FDM) and 3D Printing (3DP). While both are considered additive prototyping techniques, they differ on the
state of material used (Pupo, Celani, & Duarte, 2009).
The testing of different prototyping techniques had
two main goals: a) to determine the physical qualities of the geometrical and modelling decisions made
during the modelling phase; and b) to assess the suitability of each technique for fabricating the different
elements of the column system, eventually extrapolating the conclusions for similar shapes, other than
classical elements.
Since the scope of this paper is not the techniques
themselves, we will not go into detail about the functioning or the procedures needed by each technology,
except on what is relevant for the goals.
6.1 Second draft
The second milestone of the modelling process was
set by the production of a first physical model of the
Corinthian capital using the FDM.
The available equipment allowed for the production of models that could fit inside a 20 20 30 cm
envelope. In order to decrease the amount of building material, the model was produced at half scale
(10 10 cm), and made hollow. Production took about
12 hours.
FDM was shown to be a suitable technology for
prototyping the complex geometry of the Corinthian
capital, producing a model apparently resistant to
shock.However, FDM prototypescan be considered
anisotropic, in the sense that results differ according
to the direction along which they are produced.The
capital was produced along its axis direction, hinted
by the layers perpendicular to it. This property is not
necessarily a disadvantage. In the capitals case, however, it caused for the tips of the acanthus leaves to
break (Figure 10), because its upper layers were too
260
Table 6.
Comparative analysis of physical models.
Figure 11. Draft 3 digital and physical models.
small and, thus, very fragile. This suggested two solutions, one for the modelling phase and one for the
prototyping phase.
The modelling approach comprised preventing the
occurrence of acute angles on the leaf edges, by extruding the outer surface along an inward vector. This
solution was actually implemented on the third draft
of the capital.The prototyping approach required positioning the model for production in a way that allowed
for the shape to be decomposed in larger layers. In the
capitals case, however, due to its multiple predominant
directions, the only option would be to prototype each
leaf separately, which could be an option for larger
models but redundant for this one.
6.2 Third draft
The third draft represented an upgraded model of the
Corinthian capital, from which a physical model was
produced. Also as an upgrade, a different prototyping technique was used, 3D Printing (3DP). Compared
with the second draft model, this one was larger and
geometrically more detailed; because of the higher resolution of 3DP compared to FDM, eventual mistakes
were expected to become more obvious.
The size of the physical model was limited to a
20 20 25 cm envelope, allowing for a 13 cm high
and approximately 19 cm wide capital, which took 8
hours to produce.
In terms of material characteristics, the model produced in 3DP has some similarities with a ceramic
object, both in the producing process, and in its
fragility. In fact, one of the smaller stalks was broken
right after the capitals production.
Despite the fragility of the material, the stalks
demise could have been prevented through geometrical options in the modelling phase. Figure 11 (on the
left) shows how the stalks are sprouting out of the vase.
While the larger stalks are supported by the abacus,
the smaller ones have no support whatsoever, rendering them vulnerable.A possible solution would be to
increase the thickness as they sprout out of the vase.
A more elegant solution, to be tested later is to model
the smaller stalks adjacent to the sprouting flowers.
6.3 Analysis of results
3D Printing is not much different from Fused Deposition Modelling. Both decompose the model in horizontal sections, or slices, and process each slice in
sequence. However, results differ in at least three characteristics: resolution, rigidity and isotropy (Table 6).
In terms of resolution, 3DP is a more precise technology. However, a model produced through FDM is
more resistant to shock than the fragile 3DP models. In
terms of isotropy, a model produced by 3D Printing can
be considered isotropic, whereas the FDM-produced
model, due to the lower resolution and to the technology itself, can be very different depending on the
orientation of production layers in reference to the
geometry.
Regarding the production process, 3DP revealed
to be much faster than FDM, generating a model in
8 hours, compared with the 12 hours that took FDM
to generate a model with half the size.
RESULTS AND FUTURE DEVELOPMENTS
The computational model developed was shown to

be operational, and actually quite versatile, setting a
standard for the development of algorithms for other
complex classic elements.Although the computational
model could still be improved, all the canonical elements are present, which means that, from now on,
further refinements of the model can occur only on the
level of detail.The main objective will be the perfect
detailing of the acanthus leaves.
As detailing progresses, gaps will need to be filled
in, for which Albertis built work, as well as his theory on musical consonances, innate correspondences,
and the perfect numbers 6 and 10, will be important
references.
The physical modelling process helped to gain a
better understanding of how the geometry influences
production quality and vice-versa, which is important
when moving from prototyping on to fabrication. In
fact, as mentioned above, the last physical model produced, although of satisfactory quality, is still not the
final goal. This is to produce a detailed model of the
capital in stone to emulate the original handcrafted
techniques.As such, future research will address the
use subtracting fabrication techniques, like milling
and lathing (Pupo, Celani, & Duarte, 2009), instead
of additive prototyping ones. While the geometrical
complexity of the Corinthian capital suggests the use
of also complex machining techniques like 5-axis
milling, lathing is expected to be suitable for simpler elements like the Doric base. Since subtractive
technologies allow working with nobler materials like
261
stone, they can become a very interesting alternative

to additive prototyping technologies. However, it will
be interesting to see to which extent they can match
the geometrical versatility and sophistication of FDM
or 3DP.
In summary, the data gathered throughout the
modelling and prototyping of the Corinthian capital
described in this paper will be of great value for the
continuation of the project, namely for the development of the computational models and the production
of physical models of all the other elements from
Albertis column system.
ACKNOWLEDGEMENTS
This work is funded by FEDER Grants through
COMPETE Programa Operacional Factores de Competitividade and by National Grants through FCT
Fundaopara a Cincia e a Tecnologia, as part of the
Digital Alberti project(PTDC/AUR-AQI/108274/
2008 FCOMP-01-0124-FEDER-008842).The project
is hosted by CES at the University of Coimbra and
coordinated by Mrio Krger. Filipe Coutinho
isfunded by FCT with PhD grant SFRH/BD/66029/
2009.
REFERENCES
Alberti, L. B. (2011). Da arte edificatria. (A. E. Santo, &
M. J. Krger, Trans.) Lisbon: Calouste Gulbenkian Foundation.
Alberti, L. B. (1988). On the Art of Building in Ten Books.
Trans. Engl. of De re aedificatoria. (J. Rykwert, N.
Leach, & R. Tavernor, Trans.) Cambridge, Massachusetts:
MIT Press.
Duarte, J. P., Celani, G., & Pupo, R. (forthcoming 2011).
Inserting computational technologies in architectural curricula. In N. Gu, & X. Wang, Computational Design
Methods and Technologies: Applications in CAD, CAM
and CAE Education. Hershey, PA, USA: IGI Global
Publishing.
Morolli, G., & Guzzon, M. (1994). Leon Battista Alberti: I
nome e le figure: ordini, templi, fabbriche civili: immagini
e architteture dai libri VII e VIII del De reaedificatoria.Firenze: Alinea Editrice.
Pottman, H., Asperl, A., Hofer, M., & Kilian, A. (2007).
Architectural Geometry. Exto, PA, USA: Bentley Institute
Press.
Pupo, R., Celani, G., & Duarte, J. P. (2009). Digital materialization for architecture: definitions and techniques. In E.
S. Nardelli, & C. C. Vincent (Ed.), Proceedings of the 13th
Congress of the Iberoamerican Society of Digital Graphics, on From modern to digital: the (pp. 439-442). So
Paulo, Brazil: Mackenzie Presbyterian University, School
of Architecture and Urban Planning.
262

Evaluation of different fitting algorithms using CMM and white fringe

projection systems
F. Domingues, C. Silva, N.M. Alves, H.A. Almeida & P.J. Brtolo
Centre for Rapid and Sustainable Product Development, Polytechnic Institute of Leiria, Portugal
ABSTRACT: Purpose A common problem of automated inspection of parts with circular features is the
measurement of its roundness deviation, which corresponds to the distance between two concentric circles
touching and enclosing the extracted circumferential line at the minimum radial distance of each other. This
paper aims at evaluating different fitting algorithms to measure cylindrical features (male block) and holes
(female blocks) in physical components, which are part of a same assembly using different data acquisition
systems.
Design/methodology/approach Physical blocks were measured using a coordinate measuring machine, which
is a conventional technique for automated inspection and a white fringe projection system, a non-conventional
technique. Four fitting algorithms were considered: nonlinear least squares methods (Levenberg-Marquardt
method and Gauss-Newton method), the minimum circumscribed method and the maximum inscribed method.
Findings Findings suggest that the fringe projection technique is a suitable method for automated inspection,
generating less data dispersion and consequently a reduced fitting error. The Levenberg-Marquardt and GaussNewton fitting methods define the best approximate circle regarding the measured points, obtained through a
coordinate measuring machine or a fringe projection system.
Originality/value The output of this work provides a major insight into the effect of different fitting algorithms, used to measure and demonstrate that the fringe projection technique is a viable process for automated
inspection.
INTRODUCTION
Reverse engineering technology is a rapidly evolving

field, which makes it possible to recreate an existing object by reconstructing its surface geometry in
a virtual environment (3D CAD model) (Yang et al.,
2010; Bagci, 2009; Park and Chang, 2009; Alves and
Brtolo, 2008, 2006; Lin et al., 2005). This technology enables to create a computer digital model from
an object either with no design data or which went
through many design changes (Kim et al., 2002). While
conventional engineering (Forward Engineering), usually described as the application of scientific and
mathematical principles to practical ends such as the
design, manufacture, and operation of efficient and
economical structures, machines, processes, and systems (Zhang, 1994), transforms engineering concepts
and models into real parts, in reverse engineering the
real parts are transformed into engineering models
and concepts (Figure 1). The existence of a computer
model provides enormous gains improving the quality
and efficiency of design, manufacturing and analysis.
In general, reverse engineering technology comprises different phases, as shown in Figure 2. The
flowchart indicated in this figure represents a simplified idealisation of the entire process. In fact, the
process is not truly sequential, several iterations are
required and the overlapping of phases usually occurs

(Varady et al., 1997).
There are several application areas where reverse
engineering is often necessary. Feng and Xiao (2000)
point out nine cases where reverse engineering is
necessary:
i) Old products with no prints, i.e. products
designed before CAD/CAM;
ii) Products modified during production without
anyone recording the changes;
iii) Design reuse and redesign;
iv) Benchmarking;
v) Automated inspection, where a reverse engineering system automatically inspect a manufactured
part after a successful prototype is produced;
vi) Dimensional control;
vii) Ergonomic and aesthetic design of engineered
parts;
viii) Video and movie production;
ix) Overseas products. Products with no documentation of information suitable for local production.
Ideally, a reverse engineered model should exhibit
exactly the same geometric properties present in the
original model. However, due to both the inaccuracies
in the measured object data and the approximation and
numerical errors during the reconstruction process,
263
Figure 2. Phases of reverse engineering.
Figure 1. Engineering/design approaches. a) Forward engineering or design. b) Reverse engineering or design.
this model is approximate (Figure 3). The main error

sources can be described as follows:
measurement equipment;
data acquisition or sampling strategy;
measurement environment;
data processing.
This paper focuses on data processing errors in

reverse engineering used for dimensional control
applications, evaluating several geometric fitting algorithms through the use of a coordinate measuring
machine (CMM) and a white light fringe projection
system.
FITTING ALGORITHMS
To digitally represent an existing physical object, data

acquisition systems can capture the coordinates of several points located on the objects surface. This data is
then processed to determine the objects dimensions
and geometry. The scanned data is then processed
Figure 3. Reverse engineering and potential errors.
through appropriate software, which usually performs

a geometric fit to the gathered data. According to Hopp
and Levenson (1995), geometric fitting is the process
of computing the representation parameters of a geometric element that, in some sense can best represent
a set of point coordinate data. This representative geometric element is called the substitute geometry for
the data points (Barari et al., 2007). A manufactured
hole, for instance, is usually not perfectly cylindrical
because the process to produce it can never be totally
perfect. Inspection of the hole might involve measuring the coordinates of selected points on the surface
of the hole, fitting a cylinder to the measured points
to minimize the sum of squares of the orthogonal distances, from the cylinder points, and comparing the
position, orientation, and size of the fitted cylinder to
the dimensions and tolerances of the part specification.
Different fitting algorithms have being proposed by
several authors (Xiuming and Zhaoyao, 2009; Goch
264
and Lbke, 2008; Kelley, 1999). In this research

work, the following algorithms were used: nonlinear
least squares methods (Levenberg-Marquardt method
and Gauss-Newton method), minimum circumscribed
method and maximum inscribed method.
The least squares method assumes that the best-fit
curve of a given type is the curve that has the minimal sum of the deviations squared from a given set of
data. The objective consists of adjusting the parameters of a curve (model function) to best fit a set of data
points (xi,yi), i = 1,,n, where xi is the independent
variable and yi is the dependent variable. The model
function, y = f (x,), depends on the independent variable x and on m parameters = (1,, m) with n
m. According to the nonlinear least squares method,
the best fitting curve has the following property:
where ri2 are the squared residuals, the residual

being the difference between an observed value and the
fitted value. For nonlinear problems, to solve equation
(1) it requires that:
i) Initial estimation of each parameters;
ii) Produce the fitted curve for each set of parameters;
iii) Improving the fitting trough the adjustment of the
model function parameters. In this case, this step
is performed using both the Lavenberg-Marquardt
and the Gauss Newton Methods.
In each iteration, the parameter vector is replaced
by a new estimate + . The functions f(xi, + )
are given by:
where
Figure 4. a) Minimum circumscribed circle; b) Maximum

inscribed circle.
circumferential line. The maximum inscribed method

(Figure 4b) defines the reference circle as the maximum circle tangent to the extracted circumferential
line, where the roundness deviation is the radial difference between the extreme points on the extracted
circumferential line (Xiuming and Zhaoyao, 2009;
Goch and Lbke, 2008).
3
is the gradient (Jacobian matrix) of f with respect to
. The increment can be expressed in terms of the
Jacobian function as follows:
In the case of the Lvenberg-Marquardt method,

equation (4) is replaced by:
where is a damping factor adjusted at each iteration.

The minimum circumscribed method (Figure 4a)
defines the reference circle as the minimum circumscribed circle tangent to the extracted circumferential
line (Xiuming and Zhaoyao, 2009; Goch and Lbke,
2008). In this case, roudness deviation is the radial difference between the extreme points on the extracted
EXPERIMENTAL WORK
3.1 Measuring blocks

Two aluminum blocks were machined as illustrated in
Figure 5. These blocks are part of an assembly object,
which enables to evaluate different features like holes
and cylindrical surfaces as well fit characteristics of
the assembly. To reverse engineering a product, it is
necessary and fundamental to look at the type of fit
and function of the larger assembly. For example, if the
product loosely fits into an assembly, then the accuracy of the reproduction can be low. However, if it
is an accurate fit part, the precision must be much
higher.
In this research work, both the cylindrical element
on the male block and the hole element on the female
block were measured using the same environmental
conditions with a different set of points (3 to 12 with
increments of one). Each set of points were measured ten times. In the case of the fringe projection
265
Figure 5. Measuring blocks. a) Female block; b) Male

block.
system, these points were defined using the following

increment:
3.2
Measuring systems
Two acquisition systems were considered: a coordinate

measuring machine, DEA Mistral Swift, and a white
light fringe projection system, COMET 5 1.4M (Figure 6 andTables 1 and 2).The first system uses a trigger
probe to carry out a point-by-point data collection.
The second system involves projecting patterns of light
upon a surface of interest, capturing the image of the
resulting pattern as reflected by the surface. The image
is then analyzed to determine the three-dimensional
coordinates of the acquired data points.
3.3
Computer implementation
The least squares methods (Levenberg-Marquardt

method and Gauss-Newton method), the minimum
circumscribed method and the maximum inscribed
method were all implemented using MATLAB. The
information flow is illustrated in Figure 7.
4
RESULTS
Figure 8 compares the nominal (real) values of the

radius of the cylindrical element in the male block
and the hole in the female block, and the radius of
the approximated circle, using a CMM system and
different fitting methods. In the case of the female
block (Figure 8a), the initial approximated values are
lower than the nominal value, increasing by rising
the number of measured data points. For more than 7
scanned points, the approximate radius remains constant and higher than the nominal value, whenever the
Levenberg-Marquardt, Gauss-Newton and minimum
circumscribed methods are used. The best approximation is obtained with the maximum inscribed
method, which is independent on the number of measured points. in this case. The difference between
Figure 6. Data acquisition systems. a) Coordinate measuring machine, DEA Mistral Swift; b) white fringe projection
system, COMET 5 1.4M.
the nominal radius and the approximated value is

20 m, when the maximum inscribed method is used,
and 40 m when both the Levenberg-Marquardt and
Gauss-Newton methods are used. For the male block
(Figure 8b), the initial approximated values are higher
than the nominal value, decreasing by rising the number of measured data points. Similarly to the female
266
Table 1.
Specifications of the DEA Mistral system.
Measuring range
Accuracy
Overall dimensions
Table 2.
660 1000 460 mm

0, 014m
1125 1785 2350
Specifications of the COMET 5 1.4M.
Camera Resolution
Measuring Volume in mm3
50
100
200
400
800
3D Point Distance in m
50/100/200/400/800
Shortest Measuring Time
in Seconds
Sensor Positioning
Automatic Object Positioning
1360 1024
65 50 50
100 75 60
210 160 140
480 360 250
900 660 500
40/75/150/350/650
0.6
Tripod or sensor stand
with manual turn and
tilt axis, robot
Rotation table, robot
block, for more than 7 scanned points the approximated radius remains constant, but lower than the
nominal value whenever the Levenberg-Marquardt,
the Gauss-Newton and the maximum inscribed methods are used. The best approximation is obtained with
the minimum circumscribed method, which is independent on the number of measuring points in this
case. The difference between the nominal radius and
the approximated value is 20 m when the minimum
circumscribed method is used, and 40 m when both
the Levenberg-Marquardt and Gauss-Newton methods
are used.
In the case of CMM data, the best fitting of
the acquired data points is obtained using both the
Levenberg-Marquardt and the Gauss-Newton methods, even if the generated circle is not the closest
one to the nominal radius (Figure 9). The error indicated in Figure 9 corresponds to the difference between
the approximated circle and all the acquired digital
data. The results are due to the fact that the goal of
the minimum circumscribed method and the maximum inscribed methods is to find, respectively, an
external circle (the smallest circle that contains the
measured profile) and an internal circle (the largest
circle inscribed in the measured profile) regarding
the set of captured points, while the goal of both the
Levenberg-Marquardt and the Gauss-Newton methods
is to find the circle that produces less dispersion and
less errors regarding the measured points.
Figure 10 compares the nominal values of the radius
and the approximated values for both female and male
blocks using a fringe projection system. In the case of
the female block (Figure 10a), thee best approximation
is obtained with the minimum circumscribed method.
The difference between the nominal radius and the
Figure 7. The information flow.
Figure 8. Measured radius using the CMM system. a)

Female block; b) Male block.
approximated value is 40 m when the minimum circumscribed method is used and 50 m when both
the Levenberg-Marquardt and Gauss-Newton methods are used. For the male block (Figure 10b) the best
approximation is again obtained with the minimum
circumscribed method, which converges towards the
nominal value. The difference between the nominal
radius and the approximated value is 40 m when both
267
Figure 9. Normalized errors for CMM measures. a) Female

block; b) Male block.
Figure 10. Measured radius using the fringe projection

system. a) Female block; b) Male block.
the Levenberg-Marquardt and Gauss-Newton methods are used. Similarly to the case of CMM data,
the best fitting of the acquired data points using
the fringe projection technique is obtained through
both the Levenberg-Marquardt and the Gauss-Newton
methods, as indicated in Figure 11.
A comparison between the nominal radius and the
approximate values for the female block, obtained
Figure 11. Normalized errors for fringe projection measures. a) Female block; b) Male block.
using the CMM, the fringe projection system and

different fitting methods are indicated in Figure 12.
Looking at Figure 12, it is possible to observe that
the approximate values, obtained from the data produced using the fringe projection technique, are always
lower than the nominal value. For the CMM data,
the approximate values obtained using the minimum
circumscribed and maximum inscribed methods are
lower than the nominal value, while the values obtained
using the Levenberg-Marquardt and Gauss-Newton
are higher. For the male block (Figure 13), the approximate values produced using the Levenberg-Marquardt
and Gauss-Newton methods for both CMM and fringe
projection measures converge towards similar values
for a set of points higher than seven. In the case of
the male block, the minimum circumscribed produces
approximate values higher than the nominal, while
the maximum inscribed method tend to produce lower
values as we increase the number of measured points.
Figure 14 compares the normalized errors for the
female block using the different fitting methods.
Results show that the fringe projection technique generates less data dispersion and consequently a reduced
fitting error. A similar behaviour can be observed in
Figure 15, which compares the normalized errors for
the male block.
CONCLUSIONS
This paper evaluates different fitting algorithms to

measure a male block and female blocks, which are
part of a same assembly, using a coordinate measuring machine and a white light fringe projection
system. According to the experimental work, a good
268
Figure 12. Comparison between the nominal radius and the

approximated values for the female block obtained using the
CMM and the fringe projection system and different fitting methods. a) Gauss-Newton; b) Levenberg-Marquardt;
c) Maximum inscribed; d) Minimum circumscribed.
Figure 13. Comparison between the nominal radius and the

approximated values for the male block obtained using the
CMM and the fringe projection system and different fitting methods. a) Gauss-Newton; b) Levenberg-Marquardt;
c) Maximum inscribed; d) Minimum circumscribed.
approximation can be obtained by measuring only

seven points uniformly spaced. Results also show that:
the minimum difference between the nominal values of the radius and the approximate values for data
obtained through a coordinated measuring machine
system depends on the type of measure (internal
or external measure). For the female block (internal measure) best results were obtained using the
269
maximum inscribed method, while for male blocks

(external measure), the best results were obtained
using the minimum circumscribed method;
the minimum difference between the nominal values of the radius and the approximated values
for data obtained through a fringe projection system is obtained using the minimum circumscribed
method;
Figure 14. Comparison of normalized errors for the female

block using different fitting methods. a) Gauss-Newton
Method; b) Levenberg-Marquardt Method; c) Maximum
Inscribed Method; d) Minimum Circumscribed Method.
the Levenberg-Marquardt and Gauss-Newton fitting methods define the best approximated circle
to the set of measured data points using a CMM
system and a fringe projection system;
for both CMM and fringe projection the difference
between the nominal values and the approximate
values using the Levenberg-Marquardt and GaussNewton fitting methods are alike. This is particularly relevant as the substitute geometry produced
by most of the inspection software use those fitting
algorithms;
the fringe projection technique generates less data
dispersion and consequently a reduced fitting error.
Figure 15. Comparison of normalized errors for the male

block using different fitting methods. a) Gauss-Newton
Method; b) Levenberg-Marquardt Method; c) Maximum
Inscribed Method; d) Minimum Circumscribed Method.
ACKNOWLEDGEMENTS
This work was supported by the Portuguese Foundation of Science and Technology through the project
Sample Sizes for Interchangeable Production of
Reverse Engineered parts (PTDC/EME-PME/73357/
2006).
REFERENCES
Alves, N.F. and Brtolo, P.J. (2006), Integrated computational tools for virtual and physical automatic construction, Automation in Construction, Vol. 15, pp. 257271.
270
Alves, N.F. and Brtolo, P.J. (2008), Automatic shape recover

for rapid prototyping, Virtual and Physical Prototyping,
Vol. 3, pp. 123137
Bagci, E. (2009), Reverse engineering applications for
recovery of broken or worn parts and re-manufacturing:
three case studies, Advances in Engineering Software,
Vol. 40, pp. 407418.
Barari, A., ElMaraghy, H. and Knopf, G. (2007), Evaluation
of geometric deviations in sculptured surfaces using probability density estimation, Proceedings of the 9th CIRP
International Seminar on Computer-Aided Tolerancing,
Arizona, USA.
Feng, C.-X. and Xiao, S. (2000), Computer-aided reverse
engineering with CMM for digitization and LOM for
duplication, Proceedings of the 4th International Conference on Frontiers of Design and Manufacturing,
Hangzhou, China.
Goch, G. and Lbke, K. (2008), Tschebyscheff approximation for the calculation of maximum inscribed/minimum
circumscribed geometry elements and form deviations,
CIRP Annals Manufacturing Technology, Vol. 57,
517520
Hopp, T.H. and Levenson, M.S. (1995), Performance Measures for Geometric Fitting in the NISTAlgorithm. Testing
and Evaluation Program for Coordinate Measurement
Systems, Journal of Research of the National Institute
of Standards and Technology, Vol. 100, 563574.
Kelley, C.T. (1999), Iterative Methods for Optimization,
SIAM Press, Philadelphia.
Kim, H.C., Hur, S.M. and Lee, S.H. (2002), Segmentation

of the measured point data in reverse engineering, International Journal of Advanced Manufacturing Technology,
Vol. 20, pp. 571580.
Lin, Y.-P., Wang, C.-T. and Dai, K.-R. (2005), Reverse
engineering in CAD model reconstruction of customized
artificial joint, Medical Engineering & Physics, Vol. 27,
189193.
Park, S.C. and Chang, M. (2009), Reverse engineering
with a structured light system, Computers & Industrial
Engineering, Vol. 57, pp. 13771384.
Varady, T., Martin, R.R. and Cox, J. (1997), Reverse
engineering of geometric models an introduction,
Computer-Aided Design, Vol. 29, pp. 255268
Xiuming, L. and Zhaoyao, S. (2009), The relationship
between the minimum zone circle and the maximum
inscribed circle and the minimum circumscribed circle,
Precision Engineering, Vol. 33, pp. 284290
Yang, Z., Seo, Y.-H. and Kim, T.-W. (2010), Adptative
triangular-mesh reconstruction by mean-curvature-based
refinement from point clouds using a moving parabolic
approximation, Computer-Aided Design, Vol. 42,
pp. 217.
Zhang, Z. (1994), Iterative point matching for registration
of free-form curves and surfaces, International Journal
of Computer Vision., Vol. 13, pp. 119152.
271

Robust STL processing for extrusion-based manufacturing

M.B. Gaspar & N. Martins-Ferreira
CDRSP Polytechnic Institute of Leiria, Portugal
ABSTRACT: The slicing and tool-path generation are two fundamental processing steps in rapid prototyping/
rapid manufacturing (RP/RM), as well as in CAM. Nevertheless, the specific requirements of the diverse RP/RM
techniques are not always addressed by existing tools for CAM. We present a set of efficient and robust algorithms
for processing STL files in RP/RM.
INTRODUCTION
Several applications are available for processing STL

files some targeting specifically one RP technique,
others not specific to RP with very wide level of
maturity. Some examples for RP are Magics, NetFabb,
FabStudio, RepRap, or RP3. Some of these solutions
are very powerful, but closed and very expensive. In
our research we needed something with more flexibility, leading us to the development of our own solution.
We chose Matlab for its flexibility and suitability for
rapid prototyping (in the software engineering sense).
The main application we target is the fabrication
of porous scaffolds for biomedical purposes, using
BioExtruder an extrusion-based layered manufacturing machine (Almeida, 2008; Domingos, 2009;
Mateus, 2008).
This process has specific constraints. In other processes the layer thickness can be easily varied, the tool
paths can be interrupted at will (by switching off a
light source), the thickness may be varied, the orientation of the scan is arbitrary, the outer surface is closed,
and eventually the interior is compact. On the other
hand, in this process, the uninterruptible deposition,
the absence of a closed outer surface, and the constrained orientations of scanning in successive layers
result in a limited set of usable strategies, and dictate
the need for specific optimizations.
The algorithms developed for this application have
proved useful in other additive and subtractive RP/RM
applications, namely in producing tool paths for
incremental sheet forming or for research in rapid
fabrication of buildings (Craveiro, 2011).
2
OVERVIEW OF STL PROCESSING

FOR RP/RM
The process starts with a description of a solid volume

or its boundary, either in high-level primitives or as
a discretized polygonal representation. The most used
format for interchange of 3D models for RP/RM is the
STL format. STL consists of a triangular tessellation

of a surface, that is, the boundary of a solid is represented as a number of connected triangular facets.
Each facet is described by a sequence of nine values,
the x, y and z coordinates of each of its three vertices,
in arbitrary vertex order. The STL format also mandates that these be followed by the three components
of the facet normal, in a similar structure, although
some STL exporters do not include consistent data for
the normals, as that can be encoded in the order of
the vertices. Also, the STL format does not require a
surface to be closed, which is an obvious requirement
for RP/RM.
The STL format specification includes some additional restrictions to the tessellation: no vertex lies on
the interior of an edge or facet, and the interior of any
two facets do not intersect.
In our work we deal with source data in STL format;
other formats that can encapsulate triangular tessellations (e.g. collada) are also suitable for our algorithms,
although they have only limited support
Most RP/RM techniques consist of layered, sliceby-slice additive fabrication. Thus, a very common
first processing step is slicing. Several different methods for slicing can be used. In the most general case,
slices are the result of the intersection of a set of parallel (non-intersecting) surfaces with the part. In most
cases, the distance between slices should be constant
(either along a fixed axis or along the normal to the
slicing surfaces). The distance between different adjacent slices may be constant or not. For uniform slice
thickness, the slices should be planar, otherwise a preform will be necessary in order to support the first
slice.
In the BioExtruder, since gravity plays a role in
the flow of the filament, the slicing surfaces have to
be horizontal planes; also, since the slice thickness is
determined by the filament diameter, all slices must
be of equal thickness (see Figure 1).
After a set of slices is generated, it can be used
directly, for instance in stereolitography, defining a
mask for projecting light, or by cutting its contour from
273
sheets of material; in most processes, however, fabrication is one-dimensional, which means that these
regions have to be covered (scanned) by a path, as
detailed in section 4.
In each of the following two sections we describe
the procedures to compute the slices of an object and
to obtain the scanning path for those slices.
3 THE ALGORITHM FOR SLICING

The slicing process is easy to describe.A formal definition of the process is to find an optimal approximating
decomposition of a volume into prismatic regions,
i.e. regions limited by a slice contour and the plane
above it (see figures below). Still, some pitfalls must
be avoided in a computational implementation, as we
will see in what follows.
The way to assess the quality of that approximation
is not clear: we could for instance minimize the volumetric deviation, or minimize the roughness of the
surface. The optimal prism contour would probably be
a combination of the contours of adjacent slices, taking into account the surface normals of the intervening
facets.
It should be noted that to the best of our knowledge this issue has not been addressed in the literature
before partly because in subtractive manufacturing
(CAM) the tolerances are always negative. Indeed, the
Figure 1. A set of contours, obtained by slicing.
layer-by-layer roughing phase of machining always

leaves some material that is to be removed in a finishing phase, usually using contour-parallel trajectories
(Pandey, 2003; Chen, 2010).
In that case the fidelity of the contour to the original surface is not that relevant. The same occurs in
additive manufacturing of solid and closed objects,
where the outer surface is built from contour-parallel
paths. On the other hand, in our case, using a contourparallel strategy is not compatible with the requirement
of porous scaffolds. We use a zig-zag strategy (Held,
1991).
The slices will be represented by their boundaries,
a set of polygonal closed contours for each slice. The
contours are composed by edges that arise from the
intersections of facets with a plane; the contours may
also be regarded as a sequence of vertices, arising from
the intersection of facet edges with the plane.
The vertices of each contour are determined by
interpolation in the edges of the facets. A few issues
should be considered:
1. Not all edges intersect a plane in only one point.
If an edge is parallel and contained in the slicing
plane, it must be handled differently (see section
3.1).
2. The resulting contour vertices need to be chained
in a proper sequence, in order to define the edges
of the contour.
3. The edges of a facet contained in the slicing plane
do not necessarily belong to the boundary of the
slice. If such a facet is adjacent to other facets also
contained in the plane, only some, if any, of its edges
need to be considered (see section 3.1).
4. Processing all facet edges for each plane is computationally intensive.
In order to deal with these issues, we formally
define what our slices should be. One possible definition for the slice is: the closure (with respect to the
topology of the slicing plane) of the interior of the 3-D
object intersected with the plane. By this definition,
facets lying on the plane will always be discarded.
Each slicing plane divides our solid in two regions
the above one and the one below. For the fabrication purpose, we consider that the below region has
already been fabricated, and the above one is yet to
be fabricated. One region of the slicing plane is to
be considered as part of the slice only if it has some
material above it. This means that not all facets lying
on the slicing plane are to be considered for the slice
generation (Section 3.1).
Our slicing algorithm was developed to meet the
following criteria: obtain the chained contours bounding the region with material above; minimize the
occurrence of zero-length edges (duplicate vertices)
or zero-area contours (symmetrical edges).
We start by indexing all the vertices in the STL file,
since each one necessarily occurs in more than one
facet. This significantly reduces the amount of data to
process, and also provides some degree of validation.
The result is a list of unique vertices defined by triplets
274
Table 1. The eleven equivalence classes for the facet vertex

positions relative to a slicing plane. 1, 0 and 1 meaning that
the vertex is below, above, or lies on the plane.
i)
ii)
iii)
iv)
v)
vi)
[(0,0,0)]
[(1,0,0)]
[(1,1,1)]
[(1,1,0)]
[(1,1,0)]
[(1,1,1)]
vii)
viii)
ix)
x)
xi)
[(1,0,0)]
[(1,1,1)]
[(1,1,0)]
[(1,1,0)]
[(1,1,1)]
Figure 2. Some possible configurations of vertices relative

to a slicing plane.
of coordinates, and a list of facets expressed as triplets

of vertex numbers (indices to the list of vertices).
Then, given a set of slice levels Z = {z_1, z_2, . . .,
z_k}, we compute the contour for each slice level z_i
as follows:
1. We determine the relative position of each unique
vertex relative to z_i, assigning to it a 0, 1 or 1 as
detailed below.
2. Apply the mapping of facets to vertices, in order
to obtain triplets of vertex configurations for each
facet (see table with the 27 possible configurations
at the end).
3. Use the vertex configuration triples from 2. and
compute a configuration number for each facet (see
table 2 with the 27 possible cases).
4. Find the facets that are trivially not intersected, and
exclude them from further analysis.
5. Take the remaining set of facets and, by looking up a
table based on the configuration number obtained
in 3., get the ordered pairs of the edges that are
intersected, for each facet.
6. Chain the pairs of edges obtained (section 3.2).
7. Obtain the contour by interpolation of those edge
vertices.
We shall further explain steps 5 and 6.
3.1
Selecting intersected facets
Each vertex is above, below or lying on the slicing

plane; let us denote it accordingly by 1, 1 or 0. Each
facet has three vertices and falls into one of the 27
different configurations {1,0,1}3 (see the figure at
the end). These may be reduced to eleven equivalence
classes by noting that the first vertex is arbitrarily chosen, provided that their cyclic order be preserved. The
eleven distinct configurations (equivalence classes)
are presented in the table below:
The configurations ii) to vi) and vii) to xi) are
symmetric about the slicing plane.
Some of those configurations are illustrated in
Figure 2.
Configurations (1,1,1) and (1,1,1) are ignored,
since they do not intersect the plane. The configuration (0,0,0), that is, a facet contained in the
plane, is also ignored; if it has edges that are
in the boundary of the slice, those edges can be
accounted for by adjacent facets not lying on the
plane. Other configurations are obviously considered,
Figure 3. Pairs of chainable edges. In the case depicted aL is

a zero-length edge defined only for the purpose of chaining.
as they are trivially intersected by the plane: [(1,1,1)],

[(1,1,1)], [(1,1,0)] and [(1,1,0)]. The configurations [(0,1,1)] and [(0,1,1)] produce zero-length
intersections, so they are ignored. The configurations
[(1,0,0)] and [(1,0,0)] are the only pair of symmetric
configurations that we treat differently, thus breaking
the up-down symmetry: we must include the one that
represents having material above the plane [(1,0,0)],
and ignore the other [(1,0,0)].
This set of rules allows us to select a minimal subset
of facets, still generating the intersection of the desired
volume with the slicing plane. In order to construct a
polygonal contour we have to chain these facets. Or,
given that each facet in that subset is intersected along
at least two different edges, we may select a subset of
those facet edges forming a closed chain, with each
edge connected to one edge of each adjacent facet. In
order to be able to construct such chain of edges, we
shall associate with each facet vertex configuration a
pair of values, representing a pair of the facet edges.
See Figure 3, in relation with Table 2.
By the inspection of each one of the above configurations, it is easy to find such a pair of edges that
have an unambiguous intersection in all cases, except
for the configurations [(1,1,0)] and [(1,1,0)], in
which there is an adjacent facet sharing a vertex. For
reasons that will become clear when we explain the
chaining process, we will handle this by considering
a fake edge of zero-length, starting and ending in that
vertex. The result of this can be represented as a table
that associates with each relevant facet configuration
a pair of edge numbers (Table 2).
3.2 Chaining edges

The pairs of edges obtained in step 5 must now be
chained to produce closed contours. Although it may
275
Table 2. The pairs of edges (vp1 ,vp2 ) and (vn1 ,vn2 ) selected for each facet vertex configuration. Letters a,b and c denote the
vertices of a facet. Empty columns are configurations not contributing to the minimal set of facets.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27
vp1
vp2
vn1
vn2
b
c
c
a
b
b
c
c
b
b
c
a
a
b
b
c
a
b
c
c
a
b
c
a
c
c
a
a
b
c
a
a
a
a
b
b
a
a
b
c
Figure 4. Some of the possible configurations of edge

positions relative to a scan line.
seem not necessary for the construction of the scan

trajectories, we will see that having chained contours
is indeed very convenient.
In order to chain the contour segments (or the pairs
of edges), we must start with one pair, take its second
edge aR = (vR,1 , vR,2 ), and look for another pair whose
first edge aL = (vL,1 , vL,2 ) matches that. Doing this
search repeatedly for all pairs was one of the most time
consuming tasks in our first implementation. So, we
developed an optimized algorithm that takes advantage
of the fact that the pairs are consistently oriented, and
are guaranteed to have exactly one matching pair. The
algorithm goes as follows:
The list of all first edges and of all second edges
are guaranteed to have exactly the same elements,
only in different order. Thus, if they can be sorted in
some unambiguous way, we may associate an ordinal
to each; using then a direct lookup to find the next pair
of the chain. See also [Gaspar 2011].
4 THE ALGORITHM FOR SCANNING

Given a set of slices/contours, we want to construct
trajectories in order to cover the interior of the given
contours. The trajectories for producing porous scaffolds with uniform pore geometry and size, for our
purpose, consist of parallel segments with different
orientations in successive layers. These are generated
by intersecting a set of parallel straight lines. In theory,
any set of parallel curves could be used, but that would
not produce uniformly sized and shaped pores.
c
a
a
b
c
c
a
b
b
c
a
b
c
a
b
b
c
a
b
c
Figure 5. A tesselation of the surface of an ear, obtained

from medical data (top) and the tool paths generated for two
slices.
One of the goals of our algorithm is to obtain a minimal number of discontinuities in the scan trajectory,
as well as to avoid self-intersections. Several publications address a similar requirement in the context of
CAM, namely the minimization of tool retractions or
air-time [Tang 1998, Held 1991]. In order to achieve
that, we will have to connect the parallel segments,
along the contour (see Figure 5).
This algorithm is divided in two parts, as in the
slicing algorithm: first, select the proper subset of
intersected edges of the contour, and then construct
a set of trajectories by chaining the segments obtained
from the intersection. We briefly describe those algorithms below; for a more thorough reference see
[Gaspar 2010].
4.1 Selecting the intersected edges
The first part is similar to what we have already presented in the selection of intersected facets. The set
of intersected edges is defined in such a way that it
ensures an even number of intersections for each scanline, so that we can easily form segments by connecting
consecutive intersections along a scanline. We start by
considering all the possible configurations of the positions of the edge vertices relative to a scan line (see
Figure 4). In some of the configurations, we have to
take into account the relative positions of an edge and
also the previous one. This results in 27 cases, since
two edges are defined by three vertices (see Table 3).
276
Figure 6. The twenty-seven possible dispositions of facet vertices relative to a slicing plane.
277
Table 3. Selection of intersected edges from a contour. * means the edge (vi,vi + 1) should be intersected, < or > means that
edge is intersected only if the vertices lying on the scanline have the corresponding orientations.
1
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27
vi1 0 + 0
vi
0 0
vi+1
<
+
0
0
+
+
+
0
*
0
0
<
As in the case for slicing, there are ambiguities in the

definition of the intersections; in order to deal with it,
we must choose either to break the symmetry between
inside and outside, or to break the symmetry between
the left and the right side of the scanlines. Both solutions can be implemented with only minor changes to
the algorithm, and by changing the table 3. Since this
step is repeated for each slice, we use the fact that the
edges of the slices are chained to compute the intersections with all scanlines at once, taking advantage
of Matlabs efficiency for matrix computations. For a
description of this algorithm see (Gaspar, 2010b).
+
0
0
>
0
0
0
CONCLUSIONS
The algorithms developed have been used successfully

for processing a large number of STL files. They form
the basis of a toolbox that has empowered us to explore
new techniques for rapid fabrication in general, and in
particular for the fabrication of porous scaffolds.
REFERENCES
Almeida, H.A., Mota, C., Mateus, A., Brtolo, P.J., Ferreira,
N., & Domingos, M. 2008. Portuguese Patent 104247.
Brtolo, Paulo Jorge, Marco Domingos, Tatiana Patrcio,
Stefania Cometa, and Vladimir Mironov. Biofabrication
Strategies for Tissue Engineering. Tissue Engineering
(n.d.): 140.
+
0
0
+
0
+
+
0
+
*
+
*
+
*
0
+
*
0
0
+
>
+
0
+
+
+
0
+
+
+
+
+
Brtolo, P. J., C. K. Chua, H. a. Almeida, S. M. Chou,

and a. S. C. Lim. Biomanufacturing for tissue engineering: Present and future trends. Virtual and Physical
Prototyping 4, no. 4 (December 2009): 203216.
Chen, Jack Szu-Shen, and Hsi-Yung Feng. Contour generation for layered manufacturing with reduced part
distortion. The International Journal of Advanced Manufacturing Technology (2010).
Craveiro, F., Matos, J.M., Brtolo, H., Brtolo, P.J., Automatisation de la construction de btiments, Proceedings of
AEPR 2011, 16th European Forum on Rapid Prototyping
and Manufacturing, Paris 2011.
Domingos, M., Dinucci, D., Cometa, S., Alderighi, M.,
Brtolo, P.J. & Chiellini, F. 2009. Polycaprolactone scaffolds fabricated via bioextrusion for tissue engineering
applications. Internationl Journal of Biomaterials.
Gaspar, M.B., Ferreira, N.M., Sobre estratgias de varrimento
contnuo para preenchimento de regies no plano, Actas
do Encontro Nacional da SPM 2010, pp. 7886.
Gaspar, M.B., Ferreira, N.M., Sobre estratgias de varrimento
contnuo II (in preparation).
Gaspar, M.B., Biofab Toolbox: Construo de trajectrias
contnuas de varrimento, Master thesis, Polythecnic Institute of Leiria, 2010.
Held, Martin. A geometry-based investigation of the tool
path generation for zigzag pocket machining. The Visual
Computer 7, no. 5 (1991): 296308.
Mateus,A., Almeida, H.A., Ferreira, N., Brtolo, P.J., Mota,
C. & Sousa 2008. Bioextrusion for tissue engineering applications. Virtual and Rapid Manufacturing, pp.
171176.
Pandey, Pulak Mohan, N. Venkata Reddy, and Sanjay G.
Dhande. Slicing procedures in layered manufacturing:
a review. Rapid Prototyping Journal 9, no. 5 (2003):
274288
278
Materials

A review of different techniques to characterise the mechanical properties of

SLS parts focus on resistivity measurements
Eric Boillat, Dylan Fivat, Jamasp Jhabvala, Marc Matthey & Rmy Glardon
Laboratoire de Gestion et Procds de Production, IPR-STI-EPFL, Lausanne, Switzerland
ABSTRACT: As SLS has now been established as a rapid freeform fabrication technique in a growing application field (aerospace, tooling, bone and dental implants etc.), one feels more severely the lack of non-destructive,
reliable and recognised diagnostic methods for the mechanical characterisation of SLS parts (quality control).
This paper presents an attempt to establish the theoretical foundations and to develop a method based on
resistivity measurements to assess the properties of SLS parts and to detect their possible defects.
The principle of resistivity measurements will be presented and the correlation between resistivity and
mechanical properties (density, shear modulus etc.) will be analysed from the theoretical point of view and
illustrated.
The possibility of using resistivity measurements for on-line diagnostics during fabrication will be discussed.
Other diagnostic possibilities based for instance on ultrasounds will also be referred to.
INTRODUCTION
Several methods can be considered to characterise SLS

parts. They all seem to be based on a simple idea.
A certain form of energy is introduced into the part
to be analysed. That energy is absorbed, transmitted,
transferred or even transformed and the needed informations are deduced by detecting and measuring these
mechanisms. From this point of view, the diagnostic
methods can be classified into four categories, according to the nature of the energy introduced into the
system:
1. X-ray tomography, scanning electron-beam
microscopy or optical microscopy form a first class
of methods based on the detection of the energy
transported by electromagnetic waves.
2. Ultrasonic methods consider the way in which the
energy associated to acoustic waves is propagated
into the part.
3. Fundamental elasticity tests (tensile test, . . .) as
well as the identification of natural vibration frequencies are techniques close to ultrasonic methods. The main difference is that the energy comes
either from a progressive loading like a tap (elasticity tests) or from an impulse excitation (resonance
frequency analysis).
4. The so-called electrical methods form the last category. They will be the main topic of this paper. The
idea is to observe the behaviour of the part inside an
electrical LCR-circuit. Their common characteristic is that the energy they use comes from a power
supply.
In the next two sections, X-ray tomography, electronbeam and optical microscopy, ultrasounds, elasticity
tests and resonance frequency analysis are reviewed
and discussed. Section 4 is devoted to the electrical methods (eddy current testing and impedance
computer tomography). A simplified version of computer impedance tomography called 4 wires method
is proposed. An experimental demonstration of the
ability of the 4 wires method to detect defects in
sintered parts is given in section 5 and a theoretical approach is also discussed. In section 6, other
types of experimental results are presented in order
to illustrate the correlation existing between electrical conductivity and mechanical properties like shear
modulus.
OPTICAL METHODS
Scanning electron-beam or optical microscopy (SEM,

OM) as well as X-ray tomography (XRCT) are direct
methods with obvious and safe interpretations. Unfortunately, they often require metallographic sections
(SEM or OM) (J. Jhabvala, E. Boillat, and R. Glardon
2011), (Niu and Chang 2000) and are therefore
destructive. If they are non-destructive like XRCT,
then they are usually very slow (C. Andr 2007) and
their application makes a sense only in case of very
high added value parts (e.g. assessment of conformal
cooling channels in injection moulds). Finally, those
methods show very little interest when thinking of
developing on-line process controls.
281
Real time process monitoring with an infra-red

camera is however a promising technique (Craeghs,
Yasa, Clijsters, Bechmann, Berumen, and Kruth 2011),
(Chivel and Smurov 2010) but it falls outside the class
of methods we consider: it relies on melt pool temperature measurements and is therefore an assessment of
the process itself and not a mechanical characterisation
of the part.
In the same way, the X-Ray diffraction method
(XRD) which is used for chemical and not for
mechanical testings (Zhu, Lu, and Fuh 2003) is not
considered here.
3 ACOUSTICS, ELASTICITY AND VIBRATION

Ultrasonic measurements are efficient to detect
localised defects in homogeneous media (Page and
McCulloch 1986), (Mukhopadhyay, Ray Chaudhuri,
Seal, Dalui, Banerjee, and Phani 2001). Unfortunately,
SLS (or SLM) parts seem not appropriate for such
investigations. Instead of searching localised cracks,
we have to analyse a randomly distributed porosity
(especially in SLS parts). Such structures interact in a
chaotic way with the applied ultrasounds and absorb
them. Therefore, the interpretation of the backscattered signal is difficult. If we want to characterise
SLS(M) parts by ultrasonic techniques, significant
developments are still necessary.
Fundamental elasticity tests (tensile, torsion, hardness) are basic methods to get mechanical information
about a material, but they are destructive and require
standardised samples. Therefore, they cannot be taken
into consideration for individual testings of arbitrary
shaped parts. However, they are commonly used as
references for the validation process of other methods.
In some way, resonance frequencies measurements
can be considered as a non-destructive version of the
fundamental elasticity tests (Spinner and Tefft 1961).
Commercial devices such as the GrindoSonic , tester
are able to perform this kind of operations almost
immediately but require standardised shapes. Customised methods can also be developed by combining
piezoelectric transducers and accelerometers. In that
case, the measured natural frequency is a function of
the elastic modulus, the Poisson ratio, the density and
the geometry of the part (Rao and Rao 1990). Numerical simulations are used to take the geometry into
account and to identify the average (or equivalent)
mechanical properties of the part once the density is
known. The problem is that the obtained informations
are global and fail to identify properly possible defects.
Local characterisations can of course be obtained, but
in that case a deeper reflection about proper clampings
(boundary conditions) is required and other quantities
like free boundary displacements have to be measured. The integration of those elements would make
the set-up very complicated. Electrical measurements
are obviously better alternatives to analyse metallic parts locally. These methods are discussed in the
next section.
ELECTRICAL METHODS
Electrical methods are based on the assumption that the

electrical conductivity1 of a part is strongly correlated
to its mechanical properties (density, shear or Youngs
modulus, strength, etc.). This point has been extensively discussed in literature (Jernot, Chermant, and
Coster 1982), (Chermant, Coster, and Jernot 1984)
(Jernot and Chermant 1982), (Ke, Cheng-Feng, and
Zhen-Gang 2007) and we will illustrate it in this paper
(see paragraph 6).
If it is true that a correlation exists between electrical conductivity and mechanical properties, we still
must define a way to measure without destroying
the part.
There are basically two techniques to measure the
electrical conductivity.
The first one uses the property that a coil undergoes major changes in electrical impedance when it
approaches a conductive medium. The idea is to identify the material conductivity by monitoring the coil
impedance (Moulder, Uzal, and Rose 1992), (Dodd
and Deeds 1968). This technique is often referred to
as eddy current testing. The reason is that eddy current
generation inside the conductive material is the physical mechanism responsible for the modifications of
the coil impedance.
The other approach which will be emphasised in
this paper is called impedance computed tomography (ICT). Its theoretical basis is that the electrical
conductivity of a medium occupying a 3d-domain
can be reconstructed by measuring the voltage resulting from the application of certain electric currents
(York 2001). If the medium is isotropic, a Theorem
says that current injections as well as voltage measurements can be restricted to the boundary of
the domain (Kohn and Vogelius 1984), (Borcea 2002)
and the references therein. This Theorem facilitates
the application of the technique (internal measurements/excitation would pose accessibility problems),
but it is strongly related to the considered dimension
d = 3 of the medium. In particular, it fails in dimension d = 1 unless the medium is not only isotropic but
also homogeneous.
In theory, the identification of the conductivity in
a SLS(M) part through ICT requires a lot of simultaneous current and voltage measurements. Moreover,
the numerical algorithms used to reconstruct the conductivity by processing the measured data (Wexler,
Fry, and Neuman 1985) may result in very high computer complexity. The reason is that the ratio between
the dimension of the sintered parts to evaluate (some
tenths to some hundredths of mm) and the typical
variation scale of the conductivity (some tenths of
m because of residual porosity) is large (typically
>1000). In this paper, our aim is to propose a simplified version of ICT called four wires method. It only
gives statistical information like average (or equivalent) values of over specific areas. But we believe
1
282
The multiplicative inverse of the electrical resistivity.
Table 1.
Data characterising the seven cylinders.
# Part
# Holes
mm
Vsint /Vtot
MS/m
01
02
03
04
05
06
07
0
200
200
300
400
500
250
1.0
1.5
1.5
1.5
1.5
2.0
1.000
0.986
0.955
0.932
0.909
0.886
0.865
0.890
0.243
0.189
0.130
0.108
0.109
0.114
Figure 1. The 4W method applied on a cylindrical part.
that this information is sufficient to yield a valuable

characterisation of the mechanical properties of the
part (see paragraph 6).
5 THE FOUR WIRES METHOD FOR DEFECT
IDENTIFICATION
The four wires (4W) method is a simplification of ICT
with only one current and voltage measurement: we
inject a current I0 at point A1 , we collect it at point A2
and we measure the voltage difference U between
B1 and B2 (see Fig. 1).
The direct output of the 4W method is a resistance or, better, a conductance C = I0 / U expressed
in Siemens (S). It can be transformed into equivalent
conductivity after division by a factor L:
Since the conductivity is measured in S/m, L has the

dimension of a distance and depends on the geometry
of the sample and of the location of the four wires. It
can be evaluated for arbitrary shaped parts by a finite
element technique (FEM) (Matti 2008): we assume
that the sample has a reference conductivity ref . We
compute the conductance Cref by solving numerically
a Poisson problem. Because of linearity arguments, we
have = ref C/Cref which means:
Observe that L does not correspond to a physical

distance in the part and that it is a signed number.
It is of course possible to move the wires and repeat
the measurement. In that case, more precise or even
local information can be obtained from the 4W method.
5.1 Application of the method
In this paragraph, we demonstrate the capabilities of
the 4W method to detect defects. We have built seven
cylinders:
out of the EOS-Direct Metal 50 powder with the

EOSSINT M250 machine (see Fig. 1). Defects have
Figure 2. Cross section of the defective samples #02#03.
Figure 3. Electrical conductivity vs sintered density.
been introduced in cylinders #02 to #07 in the form

of randomly distributed unsintered spherical holes of
diameter = 1, 1.5 or 2 mm (see Tab. 1).
The holes were already designed at the CAD level
and considered as external features when transferring
the CAD to STL file (see Fig. 2).
The conductivity of each piece has been measured
by the 4W method (with the Agilent Multimeter N
34420A). The injection and collection points A1 , A2
(see Fig. 1) were located at the centre of the bottom and
the top sections of the cylinder. The points of voltage
measurement B1 , B2 were chosen on the same generator line at a symmetric position: 2 mm from the
bottom and the top section respectively. The results
are summarised in Tab. 1. In Fig. 3, the conductivity of each part is reported as a function of the ratio
between the sintered volume Vsint and the total volume
Vtot = d 2 h/4.
A significant drop in conductivity can be observed
between the sound (#01) and the bad samples (#02
to #07). It proves that the 4W method presented in
283
Figure 4. Organisation of unsintered powder leading to

different conductivities.
Fig. 1 is sensitive enough to detect macroscopic (or

even mesoscopic) defects in SLS(M) parts.
However, and because of measurement errors
(about 10% for both variables), we cannot reasonably
deduce a correlation between the conductivity and
the ratio Vsint /Vtot from the results in Fig. 3. There
are actually physical evidences that does not only
depend on Vsint and Vtot (E. Tsotsas and H. Martin
1987). Let sound denote the electrical conductivity of
the sound material (sample #01). Then we can arrange
the unsintered powder to vary the conductivity measured by the 4W method between zero and the finite
value max sound Vsint /Vtot :
Since loose powder is almost insulating (Batchelor
and OBrien 1977), we can achieve zero conductivity by arranging the unsintered material in a
(small) disk perpendicular to the current direction
(see Fig. 4(a)).
Fig. 4(b) explains how to place the powder to reach
maximum conductivity.
Yet, the probability of meeting the regular arrangements in Fig. 4 is low. If the unsintered powder
distribution is random, it can be expected (by the law
of large numbers) that the measured conductivity will
not deviate too much from an average value which
could be predicted if the ratio Vsint /Vtot is known. If
we model the unsintered material as a perfect insulator, the part can be considered as a dilute dispersion
of spherical pores in a solid phase (see (C. Argento
and D. Bouvard 1996) and the references therein). In
that case, a rule derived by Bauer (Bauer 1993) may
be applied:
This formula can be considered as a theoretical evidence that conductivity measurements can identify
sintering defects. Unfortunately, rule (2) systematically over-estimates the results of Tab. 1: it predicts
a conductivity ratio /sound of at least 80% for any
of the defective samples #02 to #07 while the measured values are less than 30%. Taking the conductivity
of the unsintered material into account would deteriorate the situation further and produce even higher
estimates of /sound . A possible explanation for this
deviation is that relation (2) has been established under

the assumption that the interactions between pores are
negligible. In our situation, this hypothesis might not
be fulfilled. In particular, it cannot be excluded that the
pores organise themselves so as to create an insulating
barrier perpendicular to the current direction as in Fig.
4(a), causing a significant decrease in the measured
conductivity. Another point which might be of some
importance is that the solid phase where the spherical
pores are diluted is not a continuous medium, like it
is assumed by Bauer. In our case, the solid phase is
actually a sintered material which is itself porous but
at a smaller scale.
6
ELECTRICAL CONDUCTIVITY AND

MECHANICAL PROPERTIES
The cylindrical samples of paragraph 5.1 (see Fig. 2)

are not adapted to tensile tests because their dimensions (1) are not sufficient. Therefore, it was decided to
machine the samples and to give them a shape adapted
to torsion tests (see Fig. 5).
The applied torsion test was performed on a MTS
809 machine and was piloted by the angular displacement varying between max . The limit value max
has been set to 1 . It is sufficient to ensure a good resolution of the measure and it is small enough to keep
the sample far away from the elastic limit.
The torque MT is plotted as a function of the angle
. According to the rule (Del Pedro, Gmr, and Botsis
2004):
with l the length of the sample and Ip the polar moment

of its (constant) section, the plot MT = MT () should
be a line through the origin with a slope proportional to
the shear modulus G. The slope is identified from the
plot (least squares), the quantity l and Ip are computed
from the geometrical data (Fig. 5) and the shear modulus G is deduced from (6). The results are presented
in Tab. 22 .
In Fig. 6 we plot the shear modulus (see Tab. 2) as
a function of the electrical conductivity (see Tab. 1).
In the considered range, the obtained relationship is
almost linear.
Based on this observation, we can deduce the shear
modulus of a sintered part from the conductivity measurement , provided the corresponding properties
(Gm and m ) are known for the massive material:
Formula (4) can be explained with arguments exposed

in (C. Argento and D. Bouvard 1996). In this paper, it
2
The last three samples #05#07 were actually too fragile to

be machined
284
Figure 5. Adapted geometry for torsion tests.

Figure 7. Shear modulus vs electrical conductivity.
by the geometry of the necks. A basic numerical simulation of a single array of connected grains has shown
that the shear modulus (as well as the Youngs modulus) are directly proportional to the ratio x/a. This
remark and (6) finally justify formula (4).
7
Figure 6. Various illustrations.
Table 2. The shear modulus of the defective samples.
# Part
01
02
03
04
shear mod. G, GPa
6.61
0.9
0.46
0.18
is mentioned that the electrical resistance of a sintered

part is mainly due to the contact resistances through
the junctions (necks) connecting the grains. Under this
hypothesis, the sintered part may be seen as a resistance network (C. Argento and D. Bouvard 1996). This
model can be used to compute the electrical resistivity. It is found to be proportional to the average contact
resistance Rneck through the necks:
The constant of proportionality depends on the average number Nc of neighbours of a given particle
(coordination number).
The next step is to relate the contact resistance Rneck
to the geometrical features of the necks. This problem
has been addressed by many authors (see (C. Argento
and D. Bouvard 1996) and the references therein). The
results is that Rneck is inversely proportional to the
relative neck size x/a (see Fig. 6(b)). Combining this
information with (5), we conclude that the electrical
conductivity = 1 is proportional to x/a:
There is now an intuitive evidence that the mechanical strength of a sintered part is essentially influenced
CONCLUSION
Among all the non-destructive diagnostic techniques,

the 4W method seems appropriate to assess sintered
parts. It is fast and the required manipulations are so
simple that it can be easily integrated in an on-line
process control.
The efficiency of the 4W method to distinguish
poorly sintered parts have been proved. We have also
shown that it is a valid technique to identify mechanical
properties like shear modulus.
An important issue is probably the precise positioning of the four probes on very small or on arbitrary
shaped parts.
REFERENCES
Batchelor, G. and R. OBrien (1977). Thermal or electrical
conduction through a granular material. Proceedings of
the Royal Society of London. Series A, Mathematical and
Physical Sciences 355(1682), 313333.
Bauer, T. (1993). A general analytical approach toward
the thermal conductivity of porous media. International
journal of heat and mass transfer 36(17), 41814191.
Borcea, L. (2002). Electrical impedance tomography. Inverse
problems 18, R99.
C. Andr (2007). Modlisation quantitative du procd
de frittage slectif par laser: Relation paramtres/
microstructure. Ph. D. thesis, EPFL.
C. Argento and D. Bouvard (1996). Modeling the effctive thermal conductivity of random packing of spheres
through densification. International Journal of Heat and
Mass Transfer. 39(7), 13431350.
Chermant, J., M. Coster, and J. Jernot (1984). Modlisation
des proprits morphologiques et physiques des matriaux
fritts en phase solide. Mmoires et tudes scientifiques
de la revue de mtallurgie 81(1), 518.
Chivel, Y. and I. Smurov (2010). On-line temperature monitoring in selective laser sintering/melting. Physics Procedia 5, 515521.
285
Craeghs, T., E. Yasa, S. Clijsters, F. Bechmann, S. Berumen,

and J. Kruth (2011). Real-time process monitoring of
layerwise laser melting. Optics and Lasers in Engineering.
Del Pedro, M., T. Gmr, and J. Botsis (2004). Introduction
la mcanique des solides et des structures. PPUR presses
polytechniques.
Dodd, C. and W. Deeds (1968). Analytical solutions to
eddy-current probe-coil problems. Journal of applied
physics 39(6), 28292838.
E. Tsotsas and H. Martin (1987). Thermal conductivity of
packed beds: A review. Chem. Eng. Process 22, 1937.
Jernot, J. and J. Chermant (1982). On the importance of
the contiguity and electrical conductivity for sintered
materials. Scr. Metall. 16(8), 943946.
Jernot, J., J. Chermant, and M. Coster (1982).A New Model to
Describe the Variation of Electrical Conductivity in Materials Sintered in Solid Phase. Phys. Status Solidi(a) 74(2),
475483.
J. Jhabvala, E. Boillat, and R. Glardon (2011). On the use of
EBSD analysis to investigate the microstructure of gold
samples built by selective laser melting. Gold Bulletin. To
appear.
Ke, Z., L. Cheng-Feng, and Z. Zhen-Gang (2007). Measurement of electrical conductivity of porous titanium
and ti6al4v prepared by the powder metallurgy method.
Chinese Physics Letters 24, 187.
Kohn, R. and M. Vogelius (1984). Determining conductivity by boundary measurements. Communications on Pure
and Applied Mathematics 37(3), 289298.
Matti, A. (Janvier 2008). Qualification par mesures rsistives
de pices mtalliques poreuses. Technical report, travail
pratique de master EPFL-DGM.
Moulder, J., E. Uzal, and J. Rose (1992). Thickness and conductivity of metallic layers from eddy current measurements. Review of scientific instruments 63(6), 34553465.
Mukhopadhyay, A., M. Ray Chaudhuri, A. Seal, S. Dalui,
M. Banerjee, and K. Phani (2001). Mechanical characterization of microwave sintered zinc oxide. Bulletin of
Materials Science 24(2), 125128.
Niu, H. and I. Chang (2000). Selective laser sintering of gas
atomized m2 high speed steel powder. Journal of materials
science 35(1), 3138.
Page, J. and R. McCulloch (1986). Ultrasound propagation in sintered metal powder: evidence for a crossover
from phonons to fractons. Physical review letters 57(11),
13241327.
Rao, S. and S. Rao (1990). Mechanical vibrations. AddisonWesley Reading, MA.
Spinner, S. and W. Tefft (1961). A method for determining mechanical resonance frequencies and for calculating
elastic moduli from these frequencies. In Proc. ASTM,
Volume 61, pp. 12211238.
Wexler, A., B. Fry, and M. Neuman (1985). Impedancecomputed tomography algorithm and system. Applied
optics 24(23), 39853992.
York, T. (2001). Status of electrical tomography in industrial
applications. Journal of Electronic Imaging 10, 608.
Zhu, H., L. Lu, and J. Fuh (2003). Development and characterisation of direct laser sintering cu-based metal powder.
Journal of Materials Processing Technology 140(13),
314317. Proceedings of the 6th Asia Pacific Conference
on materials Processing.
286

Influence of hygrothermal aging on the mechanical properties of

Nylon 12 composites processed by selective laser sintering
R. Seltzer & J. Segurado Escudero
Institute IMDEA Materials, Madrid, Spain
F. Martin de la Escalera Cutillas

Aernnova Engineering Solution Iberica, Madrid, Spain
ABSTRACT: This study aims to elucidate the combined effects of reinforcing particles and water on Nylon
12 processed by selective laser sintering (SLS). Nylon 12 is one of the toughest polyamides. However, Nylon
12 processed by SLS is rather brittle due to the presence of voids and the relative low molecular weight of the
polymers used in SLS. It is demonstrated that moisture gained by hydrothermal aging impoverishes the fracture
toughness of Nylon 12. Addition of hard particles counteracts these adverse effects.
1
INTRODUCTION
When intricate pieces in small series are needed to be

produced, rapid prototyping techniques are the ideal
option. Particularly, selective laser sintering (SLS)
allows for polymers, metals and ceramics to be processed from powder-like material into solid pieces.
Among polymers, PA12 is widely used for SLS processing since it exhibits a large difference between
melting and crystallization points (400 C) and a
large enthalpy of fusion (Tontowi & Childs 2001).
These characteristics allow for the temperature in the
sintering chamber to be held at a high level, resulting in
low shrinkage on solidification and good dimensional
stability of the sintered pieces. Furthermore, PA12 has
a very high toughness compared to other thermoplastics and even relative to other polyamides (Page 2001).
One of the main disadvantages of the PA12 processed
by SLS is its high porosity and low molecular weight,
which deteriorate its mechanical properties, especially
ductility and toughness (Ajoku et al. 2006).
Polyamides are hygroscopic, and eventually, they
take up moisture when exposed to humid air. In PA12
processed by melting techniques (e.x. injection, extrusion) this is advantageous in terms of toughness and
ductility which are increased due to the plasticizing
effect of the water (Ellis et al. 1984). For this reason,
water saturation is often accelerated by submerging
the polyamidic parts in hot water, i. e. hydrothermal
aging.
The objective of this research is to evaluate the
effect of hydrothermal aging in the fracture behaviour
of SLS processed PA12. Furthermore, different reinforcement particles will be used as fillers in order
to determine whether a combination of the particles and water moisture yields a tough and strong
material.
2.1
EXPERIMENTAL DETAILS
Materials
The two materials under study were neat Nylon 12

and ceramic short fiber reinforced Nylon 12. This
composite has 25 w/w % of fibers, as determined by
thermogravimetric analysis.
2.2
Fracture tests
Fracture tests were performed on dry (24 hs at 90 C)

and water saturated (water submerged for 3 weeks at
90 C) specimens.
The stress intensity factor Kq was detemined following the recommendation of the standard ASTM
D5045. The tests were carried out in a universal
testing machine INSTRON in three-point bending at
500 mm/min to induce brittle fracture.
2.3
Microscopy
To study the failure micromechanisms, the fracture

surfaces of the SLS materials were inspected using
electron scanning micrography (SEM). The fracture
surfaces were gold coated to enhance their conductivity and were examined using a SEM ZEISS EVO
MA15. The potential used was 15 and 20 kV.
2.4 Thermal analysis

Differential scanning calorimetry (DSC) was used to
study the crystallization characteristics of the SLS
materials. All samples were extracted from the center
of prismatic specimens. The DSC tests consisted of a
heating ramp at a rate of 10 C/min from 30 C to 250
287
Table 1. Stress intensity factors of Nylon 12 and its

composite.
Kq (MPa m0.5 )
Material
Dry
Water saturated
Neat Nylon 12
Nylon 12 composite
2.99
3.39
0.69
2.64
C, followed by a cooling ramp at a rate of 10 C/min

from 250 C to 30 C.
3
3.1

Fracture behavior
The fracture toughness of Nylon 12 is enhanced by

fiber reinforcement in both, dry and wet conditions
(see Table 1). Hydrothermal aging greatly deteriorates
the Kq of Nylon 12, while its composite is not as
sensitive.
The fracture surfaces of the dry and wet SLS materials are shown in Figure 1. Dry Nylon 12 failed
by microfibrillation, resulting in semi-brittle fracture
(Figure 1-a). This type of failure was not changed by
addition of particle reinforcement (Figure 1-c). The
degree of fibrillation was diminished by hygrothermal aging in the composite, while it was completely
suppressed in neat Nylon 12 (Figures 1-b and 1-d).
Figures 1-b and 1-d show that there is little compatibility between matrix and fibers, indicated by
their clean surface. The load transference occurred
by mechanical locking, thanks to the irregularity of
these particles. According to these figures, the failure
micromechanisms of the composite stemming from
the fibers were fiber breakage and pull-out, which
consume more energy than matrix fibrillation alone,
denoted by the increment in Kq (see Table 1).
3.2
Crystallization
The crystalline characteristics of the dry and wet SLS

materials were studied to aid the understanding of their
fracture response. Table 2 shows that all samples have
a single melting peak around 182 C belonging to
the -form crystals of Nylon 12 (Zarringhalam et al.
2006). This melting peak remained practically unmodified after addition of fibers or water conditioning.
Conversely, the degree of crystallinity Xm was
decreased by presence of fibers (see Table 3). Second
phases generally induce heterogeneous nucleation, and
hence, the crystallinity rate increases, and the crystals tend to be smaller and more imperfect (Fornes &
Paul 2003). Consequently, depending on the matrix
and the particles, crystallinity can either be increased
(Ishak & Berry 1994) or decreased (Fornes & Paul
2003, Quintanilla & Pastor 1994) by reinforcement
addition.
Figure 1. SEM micrographs the fracture surfaces of: (a)

dry Nylon 12; (b) water saturated Nylon 12; (c) dry fiber
reinforced Nylon 12; and (d) watr saturated fiber reinforced
Nylon 12.
Water conditioning resulted in further crystallization of both SLS materials, evidenced by Xc being
larger than Xm . This occurs because hydrolyzed
molecules are more easily folded, increasing crystallization degree and rate (Su et al. 2007). Hence, this
indicates that there has been noticeable degradation of
the matrix by molecule hydrolization (Fornes & Paul
2003).
Both, decreasing the molecular weight and increasing the degree of crystallinity, reduced the fracture
toughness in resins due to the reduction of the tie
288
Table 2. Melting (Tm ) and crystallization (Tc ) temperature

of the Nylon 12 and its composite.
Dry
Material
Water saturated
Tm ( C) Tc ( C) Tm ( C) Tc ( C)
Neat Nylon 12
183
Nylon 12 composite 181
144
145
182
182
148
152
Table 3. Crystallinity (Xm ) and crystallization (Xc ) degree

of the Nylon 12 and its composite.
Dry
Water saturated
Material
Xm (%)
Xc (%)
Xm (%)
Xc (%)
Neat Nylon 12
Nylon 12 composite
33.8
29.0
48.5
40.8
39.4
36.0
54.0
61.5
molecules density which act as crystals connectors

(Pecorini & Hertzberg 1993). Particularly, it has been
found that nylons present a ductile/brittle transition
with increasing crystallinity. For example, in the case
of nylon 6, this transition is around 35% (Bessell et al.
1975). Hence, although water conditioning at 90 C
provoked changes of a similar magnitude in the neat
Nylon 12 and the composite morphologies, neat Nylon
12 crystallinity fell above the ductile-brittle transition, while the composite remained in the ductile or
semi-ductile region.
4
CONCLUSIONS
The effects of water conditioning and short fiber

reinforcement on the fracture behavior of Nylon 12
produced by the SLS technique have been studied.
From fracture tests, DSC analysis and SEM imaging,
it was found that:
Unexpectedly, accelerated water conditioning led

to deterioration of fracture toughness in the tested
SLS materials. This occurs because the plasticizing
effect of water (which generally increases toughness and ductility in polyamides) is counteracted by
hydrolysis and increasing crystallinity.
Short fibers prevented the Nylon 12 to undergo
ductile-brittle transition upon hygrohermal aging.
In dry condition, discontinuous reinforcement

mildly enhanced the toughness of Nylon12. Toughening stems mostly from fiber breakage and
pull-out.
REFERENCES
Ajoku, U., Hopkinson, N. & Caine, M. 2006. Experimental
measurement and finite element modelling of the compressive properties of laser sintered Nylon-12. Materials
Science and Engineering: A 428(1-2): 211216.
Bessell, T.J., Hull, D. & Shortall, J.B. 1975. The effect of polymerization conditions and crystallinity on the mechanical
properties and fracture of spherulitic nylon 6. J.Mater.Sci.
10(7): 11271136.
Ellis, T.S., Jin, X. & Karasz, F.E. 1984. Water induced plasticization behavior of semi-crystalline polyamides. 25(2):
197198.
Fornes, T.D. and Paul, D.R. 2003. Crystallization behavior of
nylon 6 nanocomposites. Polymer 44(14): 39453961.
Ishak, Z.A.M. and Berry, J.P. 1994. Hygrothermal aging studies of short carbon fiber reinforced nylon 6.6. JAppl Polym
Sci 51(13): 21452155.
Page, I.B. 2001. Polyamides as engineering thermoplastic
materials. Shawbury: Rapra Technology.
Pecorini, T.J. and Hertzberg, R.W. 1993. The fracture toughness and fatigue crack propagation behaviour of annealed
PET. Polymer 34(24): 50535062.
Quintanilla, L. and Pastor, J.M. 1994. Structural analysis
of polyamide-6,6 reinforced with glass fibre by the use
of Fourier transform infra-red spectroscopy with photoacoustic detection and differential scanning calorimetry.
Polymer 35(24): 52415246.
Su, K., Lin, J. & Lin, C. 2007. Influence of reprocessing on
the mechanical properties and structure of polyamide 6.
J.Mater.Process.Technol. 192193: 532538.
Tontowi, A.E. and Childs, T.H.C. 2001. Density prediction
of crystalline polymer sintered parts at various powder bed temperatures. Rapid Prototyping Journal 7(3):
180184.
Zarringhalam, H., Hopkinson, N., Kamperman, N.F. & de
Vlieger, J.J. 2006. Effects of processing on microstructure
and properties of SLS Nylon 12. Materials Science and
Engineering A 435436: 172180.
289

Structural and functional properties of NiTi shape memory alloys

produced by Selective Laser Melting
H. Meier & C. Haberland
Ruhr University Bochum, Institute Product and Service Engineering, Bochum, Germany
J. Frenzel
Ruhr University Bochum, Institute for Materials, Bochum, Germany
ABSTRACT: Generally, processing of Nickel-Titanium shape memory alloys (NiTi) is challenging because
functional and structural properties of this material strongly depend on microstructural features. In the present
work we demonstrate that Selective Laser Melting (SLM) represents an attractive alternative to conventional
processing routes for NiTi. For the first time, we present mechanical data of SLM NiTi and identify similarities to
conventional NiTi. Both materials show a similar functional behavior during thermomechanical cyclingwhereas
fracture strength and fracture strains of SLM NiTi are slightly lower. Thus, SLM is a promising method for the
near-net-shape processing of NiTi shape memory alloys.
1
INTRODUCTION
Due to their outstanding functional properties shape

memory alloys (SMAs) are attractive materials for special engineering applications. Binary Nickel-Titanium
alloys (NiTi) with near-equiatomic compositions
exhibit both a thermal shape memory (pseudoplasticity) and a mechanical memory (pseudoelasticity).
Both effects are based on a reversible martensitic phase
transformation. On cooling from the high temperature
phase austenite (B2) the SMA starts to transform at
the martensite start temperature MS into the low temperature phase martensite (B19). At MF (martensite
finish temperature) the transformation is completed.
During subsequent heating, the reverse transformation
starts at AS and finishes at AF (austenite start and finish temperatures, respectively). The type of the effect
primarily depends on the transformation temperatures,
which can be adjusted through variation in the Ni/Ti
ratio. A slight increase in the Ni content of 0.1 at.-%
results in a decrease in transformation temperatures
for approximately 10 K (Frenzel et al. 2010).
Generally, all manufacturing steps of NiTi are
challenging. Due to the high reactivity the alloy
can easily pick up impurities during high temperature processing. This results in the formation of
Ti-rich phases, e.g. oxygen-rich Ti2 Ni and TiC accompanied with an increase in the Ni content of the
matrix material (Khalil-Allafi et al. 2002). Smallest compositional variances or microstructural defects
strongly affect or even fully impede the shape memory behavior. Additionally, machining of NiTi is by
no means easy. Stress induced martensitic transformation, work hardening, spring back effects, burr formation and adhesion can quickly degrade the work piece
quality and can cause considerable tool wear (Kahley

2010, Weinert & Petzold 2006, Wu 2002). Against this
background, near-net-shape technologies and especially additive manufacturing (AM) seem to provide
attractive potential for the fabrication of NiTi, although
they have only received little attention so far. There
are less than ten international research publications on
AM of NiTi to date (Bormann et al. 2010, Dudziak
et al. 2010, Meier et al. 2010, Clare et al. 2008,
Krishna et al. 2007, Shishkovskyet al. 2007, Chalker
et al. 2006, Malukhin & Ehmann 2006, Shishkovsky
2005). In a previous paper, we demonstrated that
pre-alloyed Ni49.7Ti50.3 powder material can be successfully processed by Selective Laser Melting (SLM)
and that the resulting material exhibits distinct shape
memory behavior (Meier et al. 2010). In the present
work, we focus on mechanical and functional properties of SLM NiTi carried out by compression testing.
The obtained results are compared to the behavior of
conventional NiTi.
EXPERIMENTAL WORK
2.1 Material and Processing

Experiments were performed using pre-alloyed
Ni50.2Ti49.8 powder materials that were prepared by
gas-atomization of as cast ingots by TLS Technik
GmbH. The particle size was ranging from 25 to
75 m, which represents a good compromise of flowability, powder bed density and impurity content.
All SLM NiTi samples were produced under Argon
atmosphere with a commercial SLM workstation of
type Realizer SLM 100 (MTT Technologies GmbH,
291
Figure 1. Micrographs of the same SLM NiTi sample

(arrow marks processing direction). Deep etch preparation
(a); color etch preparation (b).
Figure 3. DSC charts of SLM NiTi and conventional NiTi

(reference).
Figure 2. SLM NiTi compression testing samples on substrate after SLM processing (a); schematic illustration of
sample orientations (b).
now SLM Solutions GmbH) equipped with a 100 W

Ytterbium fibre laser. The layer thickness was 50 m
and a laser energy density of 85 J/mm3 was usedto
produce dense NiTi parts as identified in our previous
work (Meier et al. 2010). As a scanning regime, an
alternating x/y-movement of the laser was chosen. For
compression testing cylindrical samples (l = 9 mm,
= 6 mm) were used, which were machined by
turning from cylindrical SLM parts (l0 = 13 mm,
0 = 9 mm). SLM processing is generally associated with the evolution of layered microstructures
(Fig. 1a) and epitaxial solidification across several
layers (Fig. 1b), which can go along with mechanical anisotropy (Witt & Sehrt 2010, Meier & Haberland
2008). In order to investigate mechanical anisotropy in
SLM NiTi, compression test specimens were prepared
with different orientations (Fig. 2a). For the chosen
scanning regime, the number of possible orientations
can be reduced to one octant of the Cartesian coordinate system. Within this octant, five orientations were
selected, such that the specimen rotation axes form
angles of 0 , 45 or 90 to the coordinate axes (Fig. 2b).
2.2
Materials characterization
Microstructural analyses were carried out by optical andscanning electron microscopy (SEM; LEO
1530 VP). The phase transformation temperatures
were determined by differential scanning calorimetry
(Netsch DSC 204F1 Phoenix).
Figure 3 shows DSC charts of SLM NiTi and the
conventional material. All characteristic temperatures
of the SLM material are indicated. It is obvious that
the conventional material shows sharp transformation
peaks while those of SLM NiTi are less pronounced.
Nevertheless, the SLM material clearly shows phase
transformations for heating and cooling procedures,

too. It is important to note that the transformation temperatures of our SLM NiTi and the conventional NiTi
are nearly identical.
Mechanical characterization was performed using
a tensile-compression test instrument of type Zwick/
Roell Z100 at a temperature of 293 K which is 30 K
below AS of our material (see Fig. 3). Prior to testing,
the samples were cooled in ice water to ensure fully
martensitic material states (MF = 285 K, see Fig. 3).
The displacement rate was 0.5 mm/min.
For clarity, we describe the expected material behavior during mechanical testing (Otsuka & Wayman
1999). Figure 4a shows a schematic compression curve
of martensitic NiTi. Initially, the material has a twinned
martensitic structure which receives elastic deformation during loading in region 1. On further loading, detwinning occurs (region 2) and the detwinned
martensite is elastically deformed (region 3) before
dislocation plasticity and crack initiation/growth can
be observed in region 4. We note that detwinning refers
to the growth of favorably oriented martensite variants,
which allows for pseudoplastic deformation.
Figure 4b shows a different type of experiment
where shape recovery and cyclic stability were evaluated. Generally, microstructural processes mentioned
above are accompanied by irreversible plastic deformations. It has been shown in literature that these
irreversible effects depend on the microstructure
and thus on the processing history of the material
(Gromann et al. 2008). In the present work, we characterize the shape memory behavior on the basis of
the pseudoplastic strain (PP ), the elastic strain during unloading (el ) and the residual/irreversible strain
(irrev ).
Two separate types of cyclic experiments were
conducted where samples were loaded to two maximum stress levels (400 and 1200 MPa, respectively).
After reaching maximum stresses, the samples were
unloaded and heated to 393 K, which is 33 K above
AF (see Fig. 3). Heating results in the reverse transformation to austenite and thus in shape recovery.
Afterwards, the samples were cooled in ice water,
and the remaining/irreversible strain was determined.
292
Table 1.
Figure 4. Schematic illustrations of testing procedures.

Compression testing to failure (a); thermomechanical testing to characterize shape recoveryand cyclic stability of the
shape memory behavior (b).
Figure 5. Stress-strain plots for SLM NiTi and conventional

NiTi (reference).
Mechanical testing was repeated for 15 times in order

to evaluate cyclic stability of the shape memory effect.
3
3.1

Mechanical behavior up to failure
Figure 5 presents stress-strain curves of SLM NiTi

samples with different orientations (see Fig. 2b) and of
conventional NiTi (reference), which was hot worked
and straight annealed. It can be seen that all materials
show a very similar mechanical behavior. However, all
SLM NiTi samples fail at stress levels, which are up to
750 MPa lower than the stress level where first crack
initiation was observed in conventional NiTi. Anyway,
these high stresses are not required for shape memory
devices.
Table 1 summarizes mechanical data of the compression testing series. For each SLM sample orientation and for the conventional material, three tests
were performed. It is remarkable, that there is no significant influence of sample orientation on mechanical
behavior of SLM NiTi. Usually, the interfaces between
the stacked layers or adjacent hatches can represent
sensitive regions for failure due to localized defects
like small pores or particles. It has been shown for
SLM steel samples (316L) that fracture during tensile testing can occur between these regions (Meier &
Haberland 2008). One possible reason why this behavior was not observed in the present study could be that
Mechanical data of compression testing to failure.
Sample
detwinning
stress
MPa
fracture
stress
MPa
fracture
strain
%
Orient. 1
Orient. 2
Orient. 3
Orient. 4
Orient. 5
reference
155168
165166
157160
148160
159167
191210
32453343
32563349
32313403
32093366
34583469
39423975*
3940
3739
3940
3739
4042
47*
*first crack initiation.
Figure 6. Specimens prior to compression testing and after

loading to 100 kN (a). SEM micrographs showing fracture surfaces of SLM NiTi after compression testing (b-d);
regions withcracked bonding between adjacent hatches (b);
cleavage fracture (c); dimple fracture (d).
small pores collapse during compression and thus they

play a marginal role in contrast to tensile testing. In
addition, we note that the scatter of all characteristic
mechanical data of our SLM NiTi is very low (Tab. 1).
The relative deviation of both detwinning stresses and
fracture strains is about 12%, while the fracture stress
exhibits a narrow deviation of less than 8%.
An interesting result concerning deformation and
fracture mechanisms was obtained. Figure 6a shows
compression testing samples, which were loaded to
100 kN. The conventional material exhibits radial
bulging while SLM NiTi fails at almost the same load
without noticeable radial deformation.
Furthermore, fracture in all SLM NiTi samples
occurred with an angle of 45 to loading direction,
which represents the direction of highest shear stress.
This again indicates that the processing orientation
does not primarily influence macroscopic fracture
behavior during compression.In contrast to this result,
Figure 6b shows the fracture surface of an SLM NiTi
sample which was produced in orientation 3 (see
Fig. 2b). In this case, shear occurs parallel to layers.
The corresponding micrograph clearly shows cracks
between adjacent hatches. In passing, we note that no
porosity was detected for SLM NiTi produced using
our set of processing parameters. However, all samples
293
as specified in Figure 4b, were extracted from the

stress-strain plots presented in Figure 7.
Figure 7. Stress-strain plots of cyclic experiments. SLMNiTi, maximum stress 400 MPa (a); conventional NiTi,
maximum stress 400 MPa (b); SLM NiTi, maximum
stress 1200 MPa (c); conventional NiTi, maximum stress
1200 MPa (d).
show heterogeneous fracture surfaces with regions

where cleavage (Fig. 6c) and dimple fractures (Fig.
6d) were observed. In summary, our results (Figs. 5
6, Tab. 1) show that the layered structure in SLM
NiTi only plays a minor role during crack growth and
fracture.
3.2
Pseudoplastic behavior and cyclic stability
As known for conventional NiTi, shape memory

alloys havea limited functional stability. Microstructural defects, dislocations in particular, evolve and
accumulate during each phase transformation cycle.
As a consequence, degradation of the shape memory
effect and accordingly, an accumulation of irreversible
strain occur (Gromann et al. 2008).
Figure 7 shows the cyclic mechanical behavior
during functional testing (see Fig. 4b) where SLM
and conventional NiTi have been loaded to maximum
stress levels of 400 MPa (Figs 7a, b) and 1200 MPa
(Figs 7c, d), respectively. These stress levels were
chosen because 400 MPa corresponds to a stress (in the
first loading cycle) where detwinning is completed.
In contrast, 1200 MPa corresponds to a stress where
elastic loading of detwinned martensite is almost
exhausted (see Fig. 5).
The following key results were obtained: 1) Both
materials show an accumulation of irreversible strains.
2) Highest irreversible effects occur in the first cycle,
while they continuously decay with increasing cycle
number. 3) SLM NiTi has a more stable shape memory
effect than conventional hot worked NiTi because irreversible effects and changes in the stress-strain behavior appear less pronounced. 4) Irreversible strains
accumulate faster for higher maximum stresses.
Figure 8 demonstrates how maximum strains per
cycle (max , Fig. 8a), reversible strains per cycle (rev ,
Fig. 8b) and total/accumulated irreversible strains
(irrev , Fig. 8c) evolve during thermomechanical
cycling. The corresponding characteristic parameters,
Figure 8. Evolution of characteristic values during cyclic

testing: maximum strain max (a); reversible strain rev (b)
and total irreversible compressive strain irrev (c).
The maximum strains per cycle continuously

decrease with increasing cycle number (Fig. 8a) for
both stress levels. The decrease is stronger for conventional NiTi than for SLM NiTi. Similar results
were obtained for the evolution of reversible strains
per cycle (Fig. 8b). For both stress levels and forboth materials, a decreasing tendency with increasing cycle number was observed. SLM NiTi shows
highest reversible strains. E.g. the SLM sample provides almost 7.7% reversible strain (1st cycle with
1200 MPa), whereas conventional NiTi only shows
5.5% reversible strain.One striking result was found
for the accumulated irreversible strain (Fig. 8c).
In this case, SLM NiTi outperforms conventional
NiTi because irreversible effects are less pronounced.
E.g. 8.4% irreversible strain (referred to the initial
294
Figure 9. NiTi SLM actuator showinga shape memory effect. Elongation plot (left), photographs and thermal images (right).
Situation 1: elongated actuator, martensitic state prior to heating. Situation 2: contracted actuator, austenitic state during
heating. Situation 3: elongated actuator, martensitic state after cooling down.
sample geometry) have accumulatedduring 15 thermomechanical cycles (1200 MPa) for the conventional
NiTi, while only 5.3% irreversible strain were accumulated for SLM NiTi. One reason for the better
performance of SLM NiTi could be that the stress level
for detwinning of SLM NiTi (Figs 7a, c, Tab. 1) is
lower than for conventional NiTi (Figs 7b, d, Tab. 1).
As a consequence, lower dislocation densities might
evolve during cyclic loading. However, further work
is required to clarify the microstructural mechanisms
for this behavior.
In order to demonstrate that workable shape memory components can be produced by SLM, a meandering actuator was prepared and implemented in
a test rig for functional characterization (Fig. 9).
The setup allows for a detection of both elongation
through an ultrasonic distance sensor and temperature through thermal imaging. Further details on the
test rig are given elsewhere (Dilthey & Meier 2009).
The initial length of the actuator was 22.5 mm. Before
implementation into the test rig the sample was elongated by 40% to 31.5 mm in the martensitic state
which constitutes to the starting point for this experiment (Fig. 9, situation 1). During thermomechanical testing the actuator was constantly loaded with
10 N.
In a next step the actuator was heated by a direct
current (10 A) which results in a maximum temperature of approximately 440 K and hence the material
transforms to the high temperature phase (Fig. 9,
situation 2). This allows for an almost complete recovery of the initial shape. It can be clearly seen that
there is a significant actuator contraction of approximately 10 mm. After switching off the current, and
hence cooling down, the constant bias load (10 N) provides a reversible elongation of the actuator (Fig. 9,
situation 3).
SUMMARY AND CONCLUSIONS
In the present work, Selective Laser Melting was used

to process NiTi shape memory alloys. We focused on
mechanical and functional properties of the material.

The following results were otained:
1. SLM NiTi exhibits a similar stress-strain behavior
than conventional NiTi for moderate strain up to
10 % compression.
2. Fracture strains and stresses of SLM NiTi are lower
than for conventional NiTi (hot worked and straight
annealed). Nevertheless, as SMAs belong to the
class of functional materials, they usually are not
exposed to high loads where cracking can occur
during shape memory operations.
3. SLM NiTi shows a good functional stability during thermomechanical testing. In the present study,
higher reversible strains and less irreversible effects
were observed than in the case of conventional
NiTi.
4. We demonstrate that shape memory actuators can
be produced by SLM, which exhibit a distinct
pseudoplastic shape memory effect.
This study shows that SLM offers a promising method
for processing shape memory alloys. Moreover, this
technology can provide a new scope for design of
complex NiTi devices which hardly can be realized
by conventional processing methods.
ACKNOWLEDGEMENT
The authors acknowledge funding by the German
Research Foundation Deutsche (DFG) and the
state North-Rhine-Westphalia within the collaborative
research centre SFB459 (shape memory technology).
REFERENCES
Bormann, T., Friess, S., de Wild, M., Schuhmacher, R.,
Schulz, G. & Mller, G. 2010. Determination of strain
fields in porous shape memory alloys using micro computed tomography. In S. R. Stock (ed.), Developments
in X-Ray Tomography 7-Proceedings of the SPIE Vol.
7804, San Diego, 15 August 2010. Bellingham: Society
of Photo-Optical Instrumentation Engineers.
295
Chalker, P. R., Clare, A. T., Davies, S., Sutcliffe, C. J. &

Tsopa-nos, S. 2006. Selective laser melting of high
aspect ratio 3D Nickel-Titanium structures for MEMS
applications. In S. J. Bull et al. (eds), Surface engineering for manufacturing applications; Materials Research
SocietySymp. Proc. 890, Boston, November 27December
1 2005. Warrendale: Materials Research Society.
Clare, A. T., Chalker, P. R., Davies, S., Sutcliffe, C. J. &
Tsopanos, S. 2008. Selective laser melting of high aspect
ratio 3D Nickel-Titanium structures two way trained for
MEMS applications. International Journal of Mechanics
and Materials Design 4(2): 181187.
Dilthey, S. & Meier, H. 2009. Simulation-based design of
a rotatory SMA drive. Journal of Materials Engineering
and Performance 18(56): 686690.
Dudziak, S., Gieseke, M., Haferkamp, H.,Barcikowski, S. &
Kracht, D. 2010. Functionality of laser-sintered shape
memory micro-actuators. Physics Procedia 5: 607615.
Gromann, C., Frenzel, J., Sampath, V., Depka, T.,
Oppenkowski, A. Somsen, C., Neuking, K., Theissen, W. & Eggeler, G. 2008. Processing and property
assessment of NiTi and NiTiCu shape memory actuator
springs. Materialwissenschaft undWerkstofftechnik 39(8):
499510.
Frenzel, J., George, E. P., Dlouhy, A., Somsen, Ch., Wagner,
M. F-X. & Eggeler, G. 2010. Influence of Ni on martensitic phase transformation in NiTi shape memory alloys.
Acta Materialia 58: 34443458.
Kahley, F. 2010. Grundlagenuntersuchungen und simulations basierte Optimierung zur fnfachsigen Mikrofrsbearbeitung von NiTi-Formgedchtnislegierungen.
Phd-Thesis Technical University Dortmund. Essen:
Vulkan Verlag.
Khalil-Allafi, J., Dlouhy, A. & Eggeler, G., 2002. Ni4Ti3 precipitation during aging of NiTi shape memory alloys
and its influence on martensitic phase transformations.
ActaMaterialia 50(17): 42554274.
Krishna, B. V., Bose, S. & Bandyopadhyay, A. 2007. Laser

processing of net-shape NiTi shape memory alloy. Metallurgical and Materials Transactions A: Physical Metallurgy and Materials Science 38: 10961103.
Malukhin, K. & Ehmann, K. 2006. Material characterization of NiTi based memory alloys fabricated by laser
direct metal deposition process. Journal of Manufacturing
Science and Engineering 128; 691696.
Meier, H., Haberland, C., Frenzel, J. & Zarnetta, R. 2010,
Selective Laser Melting of NiTi shape memory components. In P. J. Brtolo et al. (eds.), Innovative Developments
in Design and Manufacturing; Advanced Research in
Virtual and Rapid Prototyping Proc. of VR@P4, Leiria,
610 October 2009. Leiden: CRC Press Balkema.
Meier, H. & Haberland, C. 2008. Experimental studies on
selective laser melting of metallic parts. Materialwissenschaft und Werkstofftechnik 39(8): 665670.
Otsuka, K. & Wayman, C. M. 1999. Mechanism of shape
memory effect and superelasticity. In K. Otsuka &
C. M. Wayman (eds), Shape Memory Materials, 2748.
Cambridge: Cambridge University Press.
Shishkovsky, I.V., Morozov,Y. & Sumurov, I. 2007. Nanofractal surface structure under laser sintering of Titanium
and Nitinol for bone tissue engineering. Applied Surface
Engineering 254(4): 11451149.
Shishkovsky, I. V. 2005. Shape memory effect in porous volume NiTi articles fabricated by selective laser sintering.
Technical Physical Letters 31(3): 1521.
Weinert, K. & Petzold, V. 2006. Micromachining of NiTi
shape memory alloys. Production Engineering Research
and Development 13(2): 4346.
Witt, G. & Sehrt, J. 2010. Static strength analysis of beam
melted parts dependent on various influences. In Proceedings of the 21thInternational Solid Freeform Fabrication
Symposium, Austin, 911 August 2010.
Wu, M. H. 2002. Fabrication of Nitinol materials and components. Materials Science Forum 394395: 285292.
296

Investigation on the inclusions in maraging steel produced by

Selective Laser Melting
L. Thijs & J. Van Humbeeck
Department of Metallurgy and Materials Engineering, Catholic University Leuven, Belgium.
K. Kempen, E. Yasa & J.P. Kruth

Department of Mechanical Engineering, Catholic University Leuven, Belgium
M. Rombouts
VITO nv, Mol, Belgium
ABSTRACT: Maraging steel is an iron-nickel steel alloy which achieves its superior strength, hardness and
toughness by aging the martensite phase. In earlier investigations, however, the toughness of selective laser
melted (SLM) and subsequent aged maraging steel was found to be low. Several authors attribute this to the
large oxides present in the finished product. In this work, the origin for these oxides is investigated by using
microscopy and spectroscopy. The analysis is done on specimens produced under different processing conditions
and the most important characteristics of the inclusions are discussed. It is found that during the SLM process,
an oxide layer containing Al and Ti is created on top of each layer. Upon melting the next layer, the oxide layer is
broken and dragged further. Therefore, accumulations of oxide material are found in the final product as large,
irregularly shaped inclusions.
1
1.1
INTRODUCTION
Selective Laser Melting (SLM)
Selective Laser Melting is a powder-based additive

manufacturing (AM) technique by which functional
parts are produced directly from a Computer Aided
Design (CAD)-model. Successive thin layers of metallic powder are melted locally upon the interaction
with a laser beam. The path of the laser is determined by the cross section of the CAD-model for the
layer under consideration. After the scanning of the
cross section of one layer, another layer of powder
material is deposited and scanned until the part is completely built. The schematic overview of the process is
illustrated in Figure 1. By using high intensity laser
beams, almost 100% dense, functional parts can be
produced. Moreover, the CAD-based and layer-wise
production allows a mass customization production
and high geometrical freedom. Interesting applications
are the production of patient-specific implants, tooling
inserts with complex conformal cooling channels and
light weight supporting structures. More information
about the process details, its benefits and applications
can be found in Kruth et al. (2007).
1.2
Maraging steel
The maraging steels are well known for combining good material properties like high strength,
high toughness, good weldability and dimensional
stability during aging heat treatment. Therefore,

maraging steels are mainly used in the aircraft and
aerospace industry and secondly in tooling applications. (Rohrbach & Schmidt 1990) In this research,
the maraging steel grade 300, containing 18 wt% of Ni
was used. The nominal composition of this maraging
steel grade is given in Table 1.
Maraging steels differ from high strength conventional steels in terms of the hardening mechanism. The
hardening in Maraging steel doesnt involve carbon,
but is achieved by a metallurgical reaction. The relatively soft body centered cubic martensite, which is
formed upon cooling, is hardened by the precipitation
of intermetallic compounds at temperatures of about
480 C (Rohrbach & Schmidt 1990).
1.3 SLM of maraging steel
The microstructure of maraging steel parts produced
by SLM differ significantly from the microstructure
obtained after conventional production techniques,
like forging, casting and milling. In the top views (see
Fig. 2 left), the individual scan tracks and their scanning direction can be recognized. In the side views (see
Fig. 2 right), the cross section of the melt pools can
be seen. The grains consist of fine cellular dendrites
(<1 m) and are mostly not confined to the melt pool
borders.
From the pictures (see Fig. 2), it can be seen that
the produced parts are almost dense. However, a lot
297
Table 1.
Elements
[wt]%
Chemical composition of 18Ni-300 steel according to specifications DIN 1.2709.

Fe
Bal.
Ni
1719
Co
8.59.5
Mo
4.55.2
Ti
0.60.8
Al
0.050.15
Cr
<0.5
C
<0.03
Mn, Si
<0.1
P,S
<0.01
Table 2. Overview of the applied scanning parameters

which were derived from Yasa et al. (2010).
Power
W
Speed
mm/s
Spacing
m
Island size
mm
Layer thickness
m
105
150
112*
55
30
* This is 62% of spot size d99% .
Figure 1. Schematic illustration of the Selective Laser

Melting (SLM) process.
of dark grey inclusions are present. These inclusions

were previously observed in SLM parts of Maraging steel 18Ni(300) by Stanford et al. (2008), Yasa
et al. (2010) and Kempen et al. (2011). In Stanford
et al. (2008), these inclusions are indicated as being
combined TiO2 :Al2 O3 oxides.
Since the presence of large oxide inclusions is
detrimental for the mechanical properties (Kiesling &
Lange 1978), a closer look into these inclusions will
be given in this work. The inclusions composition,
shape and location will be discussed and an insight
into the formation of these will be presented.
More information about the microstructure and
mechanical properties of maraging steel 18Ni(300)
parts produced by SLM can be found in Stanford et al.
(2008), Yasa et al. (2010) and Kempen et al. (2011).
2
2.1
EXPERIMENTAL PROCEDURES
Production
A Concept Laser M3 Linear machine was used to

build the specimens (Concept Laser GmbH). This
machine employs a diode-pumped Nd:YAG laser with
a wavelength of 1,064 nm and a maximum laser output
power of approximately 100 W measured in continuous mode. The laser beam diameter d99% at the powder
bed surface is about 180 m. The powder material was
supplied by LPW (M300-1) (Concept Laser GmbH).
The SLM samples were built with a set of process
parameters chosen in terms of maximal density which
were derived fromYasa et al. (2010) and which are presented in Table 2. The different layers were scanned
according to the island scanning pattern, which is
patented by Concept Laser GmbH. (see Fig. 3). The
parts were made under a N2 atmosphere containing
two different levels of residual O2 content. During
one build, the normal atmospheric conditions of the
CL machine, i.e. technical pure N2 , was applied. For
this atmosphere the residual oxygen content level

in the process chamber measured with a Greisinger
Electronic GMH 3691 digital oxymeter placed was
below 0.5 vol%. To obtain a N2 atmosphere with
higher oxygen content, ambient air was allowed to
leak into the machine by stopping the build and atmosphere flushing after a few layers. As a result, an
oxygen enriched N2 atmosphere containing 12 vol%
of O2 was obtained. In each build, a single melted part
(i.e. a part in which the layers are only scanned once)
and a remelted part (i.e. a part in which the layers were
scanned twice) were built.
2.2 Materials characterization
Light optical microscope (LOM) images were made
on an Axioskop 40 Pol/ 40 A Pol microscope after
immersing the polished samples for 10 s in a 10%
Nital solution, i.e. 10% HNO3 in ethanol. A Philips
Scanning Electron Microscope XL 40 equipped with
a LaB6 electron gun and Energy Dispersive X-ray
Spectroscopy (EDX) system was used to make the
secondary (SE) and back scattered (BSE) images and
the composition measurements.
For more accurate compositional measurements,
Electron Probe Micro Analysis (EPMA) was performed using a JEOL JXA-8621MX apparatus. The
chemical composition of the different phases was
determined using a wavelength-dispersive detector
(WDXA) equipped with crystals especially suited
for the detection of light elements like oxygen and
nitrogen.
3
RESULTS
First, the inclusions found inside the Maraging steel

parts produced with SLM are described. Then, the
oxides at the SLM parts top surface are studied.
Finally, the powder quality is checked.
3.1 Inclusions in SLM parts
The inclusions in the part produced under technical
pure N2 atmosphere can already be seen from the
pictures in Figure 2. In Figure 4, 5 and 6, examples
of inclusions produced under the oxygen enriched N2
atmosphere are shown as well. The inclusions that are
present in the SLM parts are big; the sizes range from
298
Figure 2. LOM micrographs at different magnifications of the SLM parts built with laser re-melting under technical pure
N2 atmosphere. Top (left) and side views (right) are showing the melt pool shapes and the dark grey oxides containing white
parent powder particles (indicated by the white arrows) and yellow TiN inclusions (indicated by the black arrows).
10 to 100 m. Some inclusions are spherical, but the

majority has an irregular and elongated shape.
In the LOM pictures using white light, these inclusions appear grey. Inside these grey inclusions, smaller
white and yellow/gold colored particles can be found.
The yellow particles are cubic and their size ranges
from 1 to 5 m. The inclusions also show cracks and
sometimes contain air bubbles (see Fig. 5). Due to
the cracking of the inclusions, some part of the inclusion material may be removed during grinding and
polishing.
At first sight, the location of the oxides seems random. But a closer look at the top views in Figure 2
point out that a higher concentration of inclusions can
be found near the island borders.
The composition for an inclusion in a part produced
under oxygen enriched N2 atmosphere determined by
Figure 3. Illustration of the islands scanning strategy.
EDX is given in Table 3. The measured inclusion is

shown in Figure 6.
It should be noted that a large electron beam size
was used and therefore influence of the surrounding
299
Figure 8. SE (left) and BSE (right) pictures of the top surface of a Maraging steel part produced with SLM in N2 atmosphere
and with single melting of each layer. The black and white arrows in the SE graph (left) are indicating the scanning direction
for different melt tracks in two different scanning islands. The white phase in the BSE graphs (right) is the Maraging steel
matrix; the dark phase which is present on top of the melt pools is an oxide layer. Maraging steel powder particles that are
attached to the top surface are encircled in black.
Figure 4. LOM image of the top surface of a Maraging

steel part produced with SLM in N2 atmosphere enriched
in oxygen and with single melting of each layer. The white
parent powder particle is indicated by the white arrow; the
yellow TiN inclusions by the black arrows.
Table 3. EDX measurement results on different features
inside an inclusion in a part produced under an oxygen
enriched N2 atmosphere with single layer melting. The measuring spots are indicated by the arrows in Figure 6. The
metal particle composition is given in wt%; the inclusions
composition in at%.
Metallic particle
[wt%]
Spot1
Fe
59
Ni
20
Co
12
Mo
6
Ti
3
Inclusions
[at%]
Spot2
Spot3
Ti
29
25
Al
2
0
O
63
34
N
5
41
Fe
1
/
Fe or oxide matrix could not be avoided. Furthermore,

this technique is not sensitive to light elements such as
oxygen and nitrogen. Therefore, these results should
only be considered indicatively. More accurate results
are obtained by using EPMA, see further.
Figure 5. SE image of the top surface of a Maraging steel

part produced with SLM in N2 atmosphere enriched in oxygen and with single melting of each layer. All inclusions are
heavily cracked. The lower inclusion contained a gas bubble.
Table 4. EPMA results of inclusions inside a single melted
Maraging 18Ni-300 steel part produced under an oxygen
enriched N2 atmosphere. The average and standard deviation
of the content is given in at%.
[at%]
O2 enriched N2
Ti
34.3 0.4
Al
3.1 0.4
O
62.6 0.3
From these results, it can be deduced that the

grey inclusions are Ti and Al containing oxides. The
white particles embedded in the oxide inclusions are
entrapped parent maraging steel powder particles. And
finally, the last type of particles found in the big inclusions is TiN particles. Being TiN explains the gold
color of this phase when viewed with LOM. Also the
composition of other inclusions in the same SLM part
as well in the parts produced under different conditions
was measured. All results confirmed these findings.
EPMA measurements on various inclusions were
performed to determine more accurately the exact
300
Table 5. EPMA results of inclusions at the top of single

melted maraging 18Ni-300 steel parts produced under technical N2 and oxygen enriched N2 atmosphere. The average
and standard deviation of the content is given in at%.
[at%]
N2
O2 enriched N2
Figure 6. SE (top and bottom-left) and LOM (bottom-right)

image of the top surface of a Maraging steel part produced
with SLM in N2 atmosphere enriched in oxygen and with
single melting of each layer. The arrows 1, 2 and 3 are indicating the corresponding EDX measuring points. In LOM,
the white parent powder particle is indicated by the white
arrow; the yellow TiN inclusions by the black arrow.
Figure 7. LOM image of the top surface in side view of

a Maraging steel part produced with SLM in N2 atmosphere enriched in oxygen and with single melting. The white
arrows are indicating the oxide layer on top of the SLM part;
the black arrows the yellow TiN inclusions.
composition of the oxide phase. This is illustrated in

Table 4 for inclusions inside single melted SLM samples produced under oxygen enriched N2 atmosphere.
The presence of TiN inclusions was also confirmed by
EPMA.
3.2
Oxides at top surface
A closer look at the top surface in the side view of the

sample produced under the N2 atmosphere enriched
in oxygen in Figure 7 reveals the presence of an oxide
Ti
Al
34.8 0.5
34.2 0.7
3.6 0.4
3.1 0.3
61.5 0.5
62.3 0.4
layer on top of the last melted layer containing the

yellow TiN particles as well.
Figure 8 shows the secondary electron (SE) and
backscattered electron (BSE) image of the top surface
of a single melted part produced under N2 atmosphere.
The individual scan tracks of the top surface can be discerned. The pictures are taken at the intersection of 4
islands. The black and white arrows in the SE graph
of Figure 8 indicate the direction and order of melting
the tracks of two neighbouring scanning islands. The
direction of each single scan track can be deduced by
the direction of the crescent-shaped solidification ripples. The order, in which the tracks were scanned, can
be reasoned by the part of the crescent-shaped ripple
which is covered by the neighbouring scan track.
The BSE graph in Figure 8 shows the presence of a
second material phase on the top surface which contains lighter elements than the Maraging steel matrix.
With EDX and EPMA analysis (Table 4 and 5), this
phase was identified to be a Ti and Al containing
oxide with the same composition as the oxides that
are present in the bulk of the SLM parts.
As illustrated by the end scanning vector (track 8
in Figure 8), the oxide layer forms at both the side
borders of the melt pool. For the melt pools which
have a neighbouring scanned track, only part of the
melt pool remains covered with the oxide layer due
to the about 40% overlap between scan tracks that
was applied. At the end of track 8, a circular oxide
phase is seen due to the small recline when the laser
beam moves to another position. Furthermore, some
Maraging steel powder particles are attached to the
top surface. These are indicated by the circles in
Figure 8.
In Figure 9, the BSE graphs of the top surface
of the parts produced under 4 different production
conditions (produced under technical pure N2 or oxygen enriched N2 atmosphere; single or remelted each
layer) are compared.
In the parts produced under the oxygen enriched
N2 atmosphere, a higher surface fraction is covered
by the oxide and the colour of the oxide material is
more uniform. The higher oxygen level results in a
higher surface coverage as well as thicker oxide layers. Thus, more oxide material is present on top of
the parts produced in the oxygen enriched N2 atmosphere compared to the parts under the technical pure
N2 atmosphere.
The shape of the oxide layer on top of the single melted scan tracks is irregular near the centre
of the melt pool. The shape of the inner border of
301
Figure 9. Backscattered Electron (BSE) graphs of the top surface of Maraging steel SLM parts in which the layers were single
melted (left) or remelted (right) and produced under a technical pure N2 (top) or an oxygen enriched N2 (bottom) atmosphere.
the oxide layer on top of the re-melted scan tracks

is straighter. Re-melting the just-scanned layer thus
produces a more uniform width of the oxide layer.
The more fluctuating width of the oxide layer of
the single scanned parts reflects the more fluctuating nature of melting powder particles compared
to the more steady nature of re-melting solid material. In melting a powder layer, the heat absorption
depends on the local particle size and the way of
stacking.
Due to the difference in expansion coefficient, and
differences in mechanical properties like E-modulus
and yield strength between the oxide and metallic
phase, the oxide layer on top of the melt pool is
cracked. This can be seen more clearly for the thicker
oxide layers produced under the oxygen enriched
atmosphere (see Fig. 9 bottom).
3.3
Figure 10. SEM picture of the recycled Maraging steel

powder that was used for building the SLM parts.
Powder material
To find a cause/origin for the presence of the oxides

and nitrides, the used powder material quality was
checked.
In Figure 10, a SEM micrograph of the recycled,
gas atomized powder is given. The powder particles
are nicely spherical and in between the powder particles no oxide material was found. The powder material
is not introducing exogenous contamination into the
powder bed.
DISCUSSION
In the layer-based manufacturing techniques, one of

the most important parameters is the condition of the
top layer of the part being created. During the process,
each top surface of the just added layer is the substrate
for the next layer to be built. From this research it can
be seen that the formation of an oxide layer on top of
the melted tracks results in the formation of big oxide
inclusion in the bulk of the maraging steel 18Ni(300)
SLM parts.
302
Figure 12. Schematic representation of the Maragoni flows

due to a temperature gradient at the surface of a melt pool for
a material with a negative (a) and a surface with a positive
surface tension gradient (b). (Mills et al. 1998).
Figure 11. Data points representing the titanium and

remaining oxygen content in the inclusions of single melted
SLM parts built in N2 and oxygen enriched N2 atmosphere,
assuming that all the aluminum in the inclusions is present
as Al2 O3 . The solid lines represent the theoretical content for
various types of oxides.
The oxide layer is formed by the oxidation of the

alloying elements with the highest affinity to oxygen.
In maraging steel 18Ni(300), Ti and Al have a the
highest affinity to oxygen (Hong et al. 2000, Hong
and DebRoy 2001), so it shouldnt be a surprise to find
these elements in the oxide phase created here. These
elements are oxidized by reaction with the solute oxygen inside the melt pool. Due to the lower density of
the oxides compared to the liquid melt, the oxide material will float. Although the oxide phase is generally
more stable than the nitride in the steel melt, part of
the Ti from the Maraging steel reacts with nitrogen to
form small, cubic, yellow TiN particles. It is seen from
Hong and DebRoy (2001) that at low temperatures
(i.e. 1700 K), growth of TiN inclusions is faster than
of the oxides. The growth of TiN is most likely encouraged by the high N2 supply from the atmosphere.
The atomic percent ratio of Ti and Al inside the
oxides found in the maraging steel 18Ni(300) SLM
parts is 90 to 10. This is comparable to the ratio of Ti
and Al in the Maraging steel alloy (see Table 2).
From various studies in literature, it is known that
the most stable Al oxide in steel is the Al2 O3 phase
(Hong 2000, Hong 2001, Babu 1995). Ti can bond
with oxygen in different ratios depending on the temperature and partial pressure of oxygen. Based on
the EPMA results of the oxide layer, the best correspondence was found with the Ti3 O5 phase. This is
shown in Figure 11 where the solid lines are the theoretical ratios of Ti and O in various oxides while
the data points represent the experimentally measured
composition.
The highest stability of Ti3 O5 in molten steels is
also reported in other studies (Hong et al. 2000). It
should be noted that for the oxygen enriched N2 atmosphere the results tend to deviate from Ti3 O5 towards
oxides with higher oxygen content, while the opposite deviation is observed for N2 atmosphere. These
results are in line with the expected trend as a function
of partial pressure of oxygen. These findings indicate that during SLM of Maraging steel 18Ni(300), a
combined oxide phase containing mainly Ti3 O5 and

Al2 O3 is formed. Furthermore, a trend towards more
Ti2 O3 when producing under a lower oxygen partial
pressure can be seen. This shift is also expected based
on the known Ti-O binary phase diagrams.
The oxide material on top of the melt pool is only
present at the sides of the melt pool surface. This is also
observed during Ar-O2 GTA welding of stainless steel.
(Lu et al. 2004) The melt pool is known to have a large
temperature gradient across the surface. The driving
force for oxide formation is higher at lower temperature, so higher at the periphery than in the centre of
the melt pool. Also, as shown in Hong and DebRoy
(2001), an equilibrium temperature exists for a given
metal oxide-steel system. Above this temperature no
metal oxides will grow and/or the existing ones will
dissolve. Due to the extremely fast cooling rates during SLM (i.e. more than 103 K/s), it is assumed that no
further significant oxidation takes place upon solidification and further cooling. Therefore, the oxides are
formed at the periphery of the melt pools.
In addition, at relatively low oxygen contents in the
melt pool an outward Marangoni flow exists at the surface of the melt pool (see Figure 12(a)). This flow may
destroy the oxide layer and drag it (further) towards
the sides of the melt pool. However, due to the oxygen
take up during SLM, the oxygen content in the melt
pool may increase and causes the flow in to become
inward (Figure 12(b)).
During selective laser melting of the maraging steel,
part of the previously melted layer is re-melted to
assure a good adhesion between the successive layers. During this re-melting, the cracked oxide layer is
broken up and taken away with the moving melt pool.
Like mentioned before, the oxide material will tend
to float on top of the new melt pool due to the difference in density. However, part of the material may
become entrapped and accumulates into large inclusions wrapped by the melt pools. Hereby, it is also
possible to trap powder particles or gas bubbles inside
the oxide inclusion as seen in Figures 4, 5 and 6.
CONCLUSIONS
During SLM of Maraging steel 18Ni(300) under technical pure N2 and oxygen enriched N2 atmosphere an
oxide layer containingTiN particles is formed on top of
the SLM part. This oxide layer is composed of Al2 O3
and mainly Ti3 O5 . The ratio between the Ti and the
303
Al oxides is about 85:15. The amount of oxide layer

is seen to increase with higher oxygen content in the
atmosphere and upon layer re-melting. When proceeding the SLM process by melting the next layer, the
oxide layer is broken up. Part of this oxide material
will float on top of the newly formed melt pools; part
of it will be trapped inside the SLM part. Especially
when the scan track changes direction, oxide material
is prone to be left as big oxide inclusions. The inclusions present in maraging steel 18Ni(300) SLM parts
are big (10100 m) and irregular of shape. The inclusions contain the same oxide material as the layers that
were created on top of the melt pools. Furthermore,
they also contain smaller yellow TiN particles, parent
maraging steel (powder) particles and gas bubbles.
In this article, an explanation for the formation of
the big oxide inclusions containing small TiN particles found in Maraging steel 18Ni(300) SLM part.
The atmosphere during the SLM of maraging steel
18Ni(300) turns out to be play an important role.
Further investigations should be performed by using
atmospheres low in oxygen and nitrogen like argon
or argon-helium mixtures or producing under vacuum
conditions.
ACKNOWLEDGEMENTS
The authors acknowledge the Catholic University Leuven (K.U.Leuven) for support through the
Project GOA/2002-06 and the Agency for Innovation by Science and Technology in Flanders (IWT)
for support through the DiRaMaP Project. The latter
institute is also thanked by Lore Thijs for providing a
doctoral bursary.
REFERENCES
Babu, S. S., David, S.A.,Vitek, J. M., Mundra,K. and DebRoy,
T. (1995). Development of Macro- and Microstructures
of C-Mn Low Alloy Steel Welds Inclusion Formation.

Materials Science and Technology, 11(2), 186199.
Concept Laser GmbH/ M3 Linear, Web-Based Data, Concept
Laser GmbH Co., Germany, http://www.concept-laser.de/,
as 29.04.2011.
Hong, T., DebRoy, T., Babu, S. S., David, S. A. (2000).
Modeling of inclusion growth and dissolution in the weld
pool. Metallurgical and MaterialsTransactions B, 31B(1),
161169.
Keene, B. J. (1998). Review of data for the surface tension of
iron and its binary alloys. International Materials Reviews
33(1), 137.
Hong, T. and DebRoy, T. (2001). Effects of time, temperature, and steel composition on growth and dissolution of
inclusions in liquid steels. Ironmaking and Steelmaking,
28(6), 450454.
Kempen, K., Yasa, E., Thijs, L., Kruth, J., Van Humbeeck, J.
(2011). Microstructure and mechanical properties of
Selective Laser Melted 18Ni-300 steel. In Proceedings of
the Lasers in Manufacturing (LIM) 2011, Munich, 2326
May 2011.
Kiesling, R., Lange, N. (1978) Non-metallic inclusions in
steel. Bournemouth: The Metals Society.
Kruth, J., Levy, G., Klcke, F., Childs, T. (2007). Consolidation phenomena in laser and powder-bed based layered
manufacturing. CIRP Annals-Manufacturing Technology.
56(2), 730759.
Mills, K. C., Keene, B. J., Brooks, R. F., Shirali, A. (1998).
Maragonif effects in welding. Philisophical Transactions: Mathematical, Physical and Engineering Sciences
356(1739), 911925.
Rohrbach, K. & Schmidt, M. (1990). Maraging Steels, Properties and Selection: Irons, Steels, and High-Performance
Alloys ASM Handbook, ASM International, Vol. 1,
793800.
Stanford, M. Kibble, K., Lindop, M., Mynors, D. and
Durnall, C., (2008). An investigation into fully melting a
maraging steel using direct metal laser sintering (DMLS).
Steel research int. Special Editions Metal forming Conference 9(2), 847852.
Yasa, E., Kempen, K., Kruth, J., Thijs, L. and Van Humbeeck,
J. (2010). Microstructure and mechanical properties of
maraging steel 300 after Selective Laser Melting. In Solid
Freeform Fabrication; Proc. Intern. Conf., Austin(Texas),
1214 August 2010.
304

SMARTerials for high performance buildings

N. Lazarovich & G. Capeluto
Faculty of Architecture and Town Planning, Technion Israel Institute of Technology
M.S. Silverstein
The Department of Materials Engineering, Technion Israel Institute of Technology
ABSTRACT: The dynamic nature of the environment has led to the need for adaptive solutions for which the
architect must possess a set of tools in order to cope, not just with extreme situations (such as midsummers,
frigid winters and more), but also with a wide range of interim climactic conditions. This paper will present the
significance of new advanced material exploration, with an emphasis on smart materials, and the complementary
application of technology to achieve high performance buildings and sustainable design. A new digital tool for
smart materials classification and selection for architects and designers will be presented. The application of
the tool will be shown through a case study with an emphasis on building facades that react to specific climatic
conditions (such as temperature change or solar radiation).
1
1.1
INTRODUCTION
Motivation
Over the last few decades, awareness of global environmental issues has grown, along with recognition
of the significant impact architecture can have on
global energy consumption (Kienzi, 2002). Recent
data published by the U.S. Department of Energy
shows that energy used by commercial and residential buildings in the U.S. is about half of the
total national energy consumption (Mazria, 2009). In
Europe and Israel the numbers are approximately the
same.
1.2
High performance buildings
Growing awareness of ecological issues, has given rise

to relevant architectural solutions; the dynamic nature
of the environment has led to the need for adaptive
solutions. Todays architect must have a set of tools to
cope not just with extreme situations (such as midsummer, frigid winters), but also with a wide range of in
between situations and requirements. Over the last few
decades, many attempts have been made in the field
of responsive buildings.
One of the first attempts to create a responsive
building faade was made in the 1980s by architect
Jean Nouvel. The facade of the building LInstitute
du Monde Arabe in Paris (Kolarevic and Malkawi,
2005), was based on a mechanical system of apertures
designed to respond to the changing amount of radiation hitting the facade throughout the day by opening
and closing accordingly. Due to the over complexity of the mechanical system, this system failed to
work after a short time, however, in the contemporary

architectural design of smart, high performance buildings, we see an evolution of this concept.
Smart buildings are buildings that have intelligence
in the context of their reaction to environmental conditions, creating optimal comfort conditions for the
users of the building (Ochoa, 2009). The term is
also connected to the ongoing management of the
building (Arkin et al., 1994). One such building is
the New York Times building in New York, designed
by R. Piano, where the contractors developed a new
tool to assess the use and efficiency of the shading and lighting systems. Like the Times building,
Smart Buildings set performance standards during the
commission and design process that are rechecked
during the construction stage and beyond (Lee et al.,
2007).
Today, most dynamic solutions are based on
mechanical, electrically-driven systems that monitor
and respond to the environment, as seen in the above
example. The current research shares the same basic
goal dynamic response to environmental conditions,
however, it seeks responsiveness inherent in the materials used, rather than as an outcome of complex
mechanisms and algorithms. The research investigates
the possible applications of smart materials systems,
based on advanced materials with inherent dynamic
properties, and examines how these materials respond
to changes in environmental conditions, and finally,
assesses the suitability of the materials for use in high
performance building facades.
The use of high performance building systems
based on smart materials enables optimal building
performance with reduced energy consumption and
enhanced user comfort.
305
2
2.1
BACKGROUND
2.3 Material classification systems
Smart Materials
Smart Materials and structures are those objects that

sense environmental events, process that sensory information, and then act on the environment (Kirk and
Othmer, 2006)
In the past two decades, there has been a growing interdisciplinary effort focused on the development of advanced materials and systems, affecting a
fundamental change in the way scientists, engineers
and designers approach materials (Ball, 1997). Smart
materials of today allow for great specificity, demonstrating changeable properties that are responsive to
transient needs (Addington and Schodek, 2005b).
Currently their use can be observed in a variety of
fields, including: Space and Aeronautics, Medicine,
Automotive and Technology. The use of smart materials in Architecture, however, remains very limited,
despite their tremendous potential for improving
building performance and enhancing architectural
design possibilities through their ability to respond
and thus adapt to internal and external environmental
stimuli (Addington, 2007).
While there are building products based on smart
materials that have already become commercialized
in different architectural applications, such as photovoltaic cells or smart glazing for windows, they are
upheld as commercial catalog products rather than as
materials with distinguished behavior and features that
should enter the architects awareness.
2.2
Limited architectural use
One of the major reasons for the limited architectural

use of advanced materials is the difficulty involved in
creating information flow between the field of material
engineering, where such materials are designed and
created, and the end users the architects and designers. It seems that there is a lack of a mutual platform
or common ground that would enable architects to
explore novel advanced materials, and material engineers to comprehend the needs and vast possibilities
that exist in contemporary architecture.
There is a need for the development of a comprehensive framework whereby the great potential
of new materials and technologies can be exploited.
An important question that has been raised is: Can
architects move beyond perceiving smart materials
as merely a replacement for conventional materials?
(Addington and Schodek, 2005a) In order to do so
they must first clearly define the building performance
that is required from the smart materials; they must
start thinking and looking at smart materials from a
more phenomenological perspective, considering their
effects on the luminous, thermal, acoustic and kinetic
environments. For this purpose there is a need for a
new architectural material classification system that
enables the architect a full understanding of smart
materials behavior.
Currently there are several material classification

systems that each describe different aspects of the
materials (Addington and Schodek, 2005b). The scientific classification system describes the connection
between atoms and molecules, the engineering classification system describes a materials mechanical
properties and performance, and the current architectural material classification system describes the
properties of different catalog products (windows,
concrete, brick, etc.), as can be seen in the ARCHITONICS material library (ARCHITONICS, 2011) or
the SWEETS NETWORK online catalog (Sweets,
2011), providing manufacturer and product information as well as free CAD files.
Much research has been undertaken throughout the
world in the area of material libraries, and there are
today a few web based material libraries enabling
classification of materials (and not merely products),
according to their properties. Numerous networks
and material libraries for designers, architects, engineers and industrialists were established in Europe
during the last decade, such as the Dutch material search engine MATERIA (MATERIA, 2009),
the Spanish material information center materiO
(http://www.materio.com/, 2001) and the material
library iMatter (iMatter, 2008) in the Holon design
museum in Israel. All of the material libraries listed
above provide a good general overview of materials
and existing products, regular updates of innovative
materials and a platform for communication with the
manufacturer, however they lack detailed information of materials properties and behavior and fail to
provide a specific method for smart materials classification. The aim of this study is to develop a
comprehensive framework for classification of smart
materials, which will enable both architects and material engineers to better define and understand existing
advanced materials, and those that are being created
today, within the framework of their potential needs
and uses.
3
METHODOLOGY
3.1 SMARTerials classification system

The smart classification algorithm of this application
allows the users to approach the search at each step of
the design, beginning with the criterion that matches
their specific needs the most. The proposed classification system for smart materials helps architects
identify and fit the best material for their goals from
a smart materials library. The web based digital tool is
called SMARTerials (www.smarterials.co.il).
3.2 The classification algorithm of SMARTerials
The proposed system is based on an algorithmic
network of questions (Table 1) that bridges the building performance requirements and the appropriate
306
Table 1. A set of questions brought from two distinctive points of view: architectural and materials engineering.
Architectural Point of View
Materials Engineering Point of View
What is the appropriate material characteristic?

What class of materials does it belongs to?
What is the required reaction type (property changing/energy exchanging)?
What is the relevant input (stimulus) for this material?
What is the required material behavior (output)?
What is the relevant range of action?
Is it a reversible material action?
What environmental phenomenon can act as an input?
What is the activation ambient trigger (temperature, radiation, sound, etc.)
What are the building system needs?
What kind of process occurs at the boundary?
What is the material molecular structure?
What form and sizes of the material are available today?
In which part of the building is this material supposed to be used?
What is the materials price?
What level of glossiness does the material have?
What is the materials visible transmittance?
What is the density of the material?
How do you describe the texture of the given material?
What is the material hardness?
What is the acoustic performance of this material?
What is the color/s of the material?
What is the fire resistance level of this material?
What is the UV resistance of this material?
What is the scratch resistance of this material?
What is the strength of the material?
What is the chemical resistance of this material?
What is the weight of this material?
Is this material renewable?
Where is this material manufactured today?
material properties. The algorithm for the smart material selection is based on two different points of view,
that of the architect/designer and that of the materials
engineer. Moreover, there are some questions that are
relevant to both disciplines, as can be seen in Table 1.
The classification algorithm of this tool (Figure 1)
was based on the answers to questions shown in Table
1. The answers to each of the questions are represented
by a tag (For example: Q: What class does the material
belong to? A: Polymers). In this manner, each material in the catalog is defined by a set of representative
tags enabling targeted searches. In the first iteration,
the users input results in the generation of a tag which
represents one of the material properties. The program
then searches through each of the materials in the catalog. When a match is found, the required tag for that
material is left open while the tags for the others are
hidden and visually removed from the users library.
The second and the third iterations work in the same
manner, adding more filtration tags according to the
users needs and desires, while the materials that do
not have the matching tags are being hidden visually
to the user. Finally, the program generates output in
the form of a list with one or more materials that have
varying levels of suitability.
3.3
Users interface
The classification algorithm of this application has two

major steps (Figure 2): The first search is the most
critical since the algorithm discards all the materials

that dont match the first choice. The most important
criteria should therefore be selected first. In the second
step, the designer can add 3 more criteria, resulting in
3 more matching levels; in other words, the order of
the next 3 searches is not important.
The questions that refer to the material matching
process were the basis for the 6 main classification
tabs:
1.
2.
3.
4.
5.
6.
General Info
Material Behavior
Detailed Info
Sensorial Data
Technical Info
Location
Each of the tabs refers to a different aspect of the

material. In each main tab there are two sub categories,
the secondary filtration tab and the third filtration tab.
3.4
SMARTerials and the design process
The main design process flow diagram is shown in

figure 3. It demonstrates how the proposed classification system is incorporated into the architectural
design process. The architect/designer may use the
classification tool in different stages of his work:
using the tool between the conceptual stage and the
schematic design is useful for exploring new material
307
Figure 1. SMARTerials web tool classification algorithm.
tool provides a more specified classification of materials for well defined needs thus enabling the user easy
communication with the materials manufacturers.
4
Figure 2. Graphical indication of the match percentage in
the data base.
possibilities without having a final specification of the

system needs; the material gallery of projects existing
within the classification tool may provide inspiration
for further development; between the schematic and
detailed design stage, the usage of the classification
CASE STUDY USING SMARTERIALS
This section presents a sample application of the

proposed classification tool, for a responsive shading faade element. Several basic requirements were
given:
308
Building System needs: Self responsive shading

envelope activated with zero energy
Location: Tel Aviv, Israel
Smart Material Reaction Type: Property changing
Figure 5. Second search path, 100% and 75% suitability

percentage. Material Behavior Response Deformation.
Figure 3. Flow chart of the design process integrating the

SMARTerials tool.
Figure 6. Third search path, 100%, 75% and 50% suitability percentage: Detailed Info Use in Building Thermal
Actuator (From www.smarterials.co.il).
Figure 4. First search path 0% or 100% suitability percentage: General Info Reaction Type Property Change.
Environmental Input: Temperature change (ambient

temperature)
Material Required Behavior: Geometry change
Use in Building: Actuator of a self responding
faade component, that responds to external stimuli
Figure 7. Fourth search path, 100%, 75%, 50% and 25%

suitability percentage: Detailed Info Use in Building
Thermal Actuator (From www.smarterials.co.il).
This demonstration shows one possible use of the

SMARTerials tool according to the given architectural definitions.The first critical search (Figure 4), the
reaction type of the smart material, is the most influential criteria defining the character of the potential
material system.
In the second step of the classification process
there were 3 more iterations made, resulting in 3 more
matching levels (Figures 57).
At the end of the search there is textual output
(Figure 8) showing a list of suitable materials, sorted
according to the most recent search. In this case the
most suitable material for the given goal, as shown
in the Found Info report (Figure 8), is the family of

Shape Memory Alloys. These materials have the ability to revert, or remember, a previously memorized
or preset shape (Addington and Schodek, 2005a). The
user can get more information about the desired materials by clicking on the relevant icon and can even
connect to the web site of the manufacturing company
(Figure 9).
The materials that have been found by the program
belong to the family of shape memory materials, which
can be polymers, metals or ceramics. For the prototype, the specific material being integrated into the
309
Figure 8. Found Info report shows a list of suitable materials, sorted according to the last search.
Figure 11. Shading component Activated by shape memory

alloy sprigs closed state.
Figure 9. Detailed information about one of the proposed

materials (SMA).
Figure 10. Shading component Activated by shape memory

alloy sprigs open state.
responsive shading device is Shape Memory Alloy.

This material was explored and used both as a sensor and actuator, and tests were done on prototype
components based on compression SMA springs that
are currently being developed in the research (Figures
1112). An initial evaluation of the shading potential
and radiation exposure on a glass faade is presented.
The results show that in the open state of the components there is no direct exposure to the sun. In
the semi-open state of the components (Figure 12)
the faade absorbs about 260 IRD [KW-h/m], half of
what is absorbed in the closed state of the components
Figure 12. Energy simulation of self-shading responsive

faade semi open state. (IR4SU (Irradiance for SketchUp),
Amir and Capeluto).
Figure 13. Energy simulation of self shading responsive

faade close state. (IR4SU (Irradiance for SketchUp), Amir
and Capeluto).
(Figure 13), when most of the faade is exposed to the

suns radiation.
DISCUSSION AND CONCLUSIONS
This article presents theoretical and practical platforms

for the integration of smart materials into structural
systems. A new classification system for families of
smart materials is proposed for integration into the
310
architectural design process. An examination of this

tool and its application is demonstrated in a case study
that examines building facades relevant for specific
climatic conditions (such as temperature change or
solar radiation). At this point, the research attempts
to design self-responding, shape memory alloy-based
shading components, while the next step is to simulate
the processing of the proposed shading component.
Smart materials and technologies, like those shown
in the case study of the responsive shading component, enable architects to animate their architecture
by making it more interactive and responsive to the
surrounding environment. In this way, smart materials
may replace more conventional building materials or
mechanical- electric based systems creating a new
architectural vocabulary.
Questions and issues that emerged during the process:
It is difficult to run digital climatic simulations on

the proposed smart components due to the lack
of data regarding these kinds of materials in the
simulation program base data (ex: EnergyPlus).
Todays smart materials are used mostly as replacements for conventional materials. Can architects
change this over time?
Todays smart materials are still not optimal for
architectural use; one of the major problems being
the range of the materials action scale.
This research presents the use of smart materials that already exist, materials which were chosen
using the SMARTerials web tool. In the future, we
expect architects and designers to take an active role in
designing new materials and in defining the required
properties according to building performance needs.
ACKNOWLEDGMENT
Noy Lazarovich is grateful to the Azrieli Foundation
for the award of an Azrieli Fellowship.
REFERENCES
Addington, M. (2007). For smart materials change is good.
Architectural Record, September 2007, 160162.
Addington, M., Schodek, D. (2005). Smart materials and
technologies. Architecture and Urbanism 5(3) 813.
Addington, M., Schodek, D. (2005). Smart Materials and
Technologies for the Architecture and Design Proffesions.
In Oxford. Elsevier (eds).
Architectonics. (2011). Architectonics Products and Materials, The independent resource for architecture and design:
www.architonic.com/team.
Arkin, H., Becker, R. (1994). Intelligent Buildings. In Haifa.
Technion-Israel Institute of Technology (eds).
Ball, P. (1997). Made to Measure: New Materials for the
21st Century. In N.Y. Princeton University Press-Israel
(eds). iMATTER. 2008. Imatter material library, Design
Museum Holon: www.imatter.org.il.
Kienzi, N. (2002) Evaluating Dynamic Building MaterialsThe Potential of Climaticlly Responsive Building Enclosure. In Cambridge, Harvard-Graduate School of Design.
PhD:220.
Polhill, R.M. (1982). Crotalaria in Africa and Madagascar.
Rotterdam: Balkema.
Kirk, R. E., Othmer, D. F. (2006). Encyclopedia of Chemical
Technology. New Jersey: John Wiley & Sons, Inc.
Kolarevic, B., Malkawi,A. (2005). PerformativeArchitectureBeyond Industrumental. In N.Y.C. Spon Press (eds).
Lee, E., et al. (2007). Daylight The New York Times
Headquarters Building. CA Lawrence Berkley National
Laboratory.
MATERIA. (2009). Web site material library, Amsterdam:
www.materia.nl/538.0.html
materiO. 2001. MateriO information centre: www.materio
.com
Mazria, E. (2009). The Building Sector: A Hidden Culprit.
Architecture 2030. From: www.architecture2030.org/
current_situation/building_sector.html.
Ochoa, C. E. (2009). Design of Intelligent Building Skins for
Hot Climates. In Haifa, Architecture and Town Planning,
Technion. PhD:181.
Sweets. (2011). McGraw Hill Construction Sweets Network:
products.construction.com
311

Porous ceramic filters through 3D printing

A. Withell, O. Diegel, I. Grupp & S. Reay
Centre for Rapid Product Development, Auckland University of Technology, Auckland, New Zealand
D. de Beer
Technology Transfer & Innovation, Vaal University of technology, South Africa
J. Potgieter
Massey University, School of Engineering and Advanced Technology, Auckland, New Zealand
ABSTRACT: This paper describes current and on-going work in adapting Z-Corp 3D printers to operate with
low-cost ceramic materials. The components produced with these clay-based ceramic powders can be fired to
produce strong, complex and lightweight ceramic parts. The final material properties, including the porosity of
the parts, can be controlled through the part design and, potentially, through additives to the material that burn
out during firing. The paper begins with a brief description of the 3D printing process and how it can be used
with clay powders. It then introduces a factorial design experiment initiated to explore the effect of ingredient
and parameter variations on the dimensional stability and material properties of green and fired ceramic parts.
It then explores the porosity properties of fired ceramic parts for use in filter applications.
1
INTRODUCTION
Many educational institutions use 3D printers as part

of their curricula, to build prototype parts for design
and engineering projects. One of the factors that can
restrict the use of these printers is the relatively high
costs of materials. This often means that students either
have to pay for the material costs of 3D printing, or the
number of 3D printed iteration of their designs may be
limited, restricting the ability for them to perfect their
designs.
This research project was initiated to find a material replacement for the Z-Corp 3D printer that was
sufficiently low-cost to allow students at educational
institutions to print using additive manufacturing processes without the current cost restriction. As the
project developed and began to produce successful
parts, it was realised that the low cost ceramic material
had potential for a wide range of applications, resulting
in the need for further research.
2 ADDITIVE MANUFACTURING
Additive Manufacturing is a term referring to a range
of layer-upon-layer manufacturing technologies, used
for both prototyping and manufacturing purposes.
It allows complex components to be manufactured
that, in most cases, could not be manufactured using
traditional manufacturing techniques.
In contrast to conventional subtractive manufacturing where material is removed from a solid block to
create the desired part, additive manufacturing (AM)

starts with nothing and builds the part layer by layer
until the part is complete (Chua and Leong, 2003).
Additive manufacturing is capable of manufacturing
complex parts, from a number of different materials,
without the need for tooling or machining. AM has
few restrictions on the complexity of the parts it can
produce, thus giving designers complexity for free.
All additive manufacturing technologies begin with
a 3D computer model of the component to be produced. This 3D model is then sliced into thin sections
typically ranging from a few microns thick to 0.25 mm
thick depending on the technology by the AM software.TheAM machine then makes the part by building
each consecutive slice of the model on top of the previous one until the part is complete. The different AM
technologies differ in how they build each slice of the
component and the material each technology uses.
Although a variety of AM technologies have been
used to create ceramic parts (including Selective
Laser Sintering (SLS) and Fused Deposition Modeling (FDM), this paper concentrates on the technology
developed by Massachusetts Institute of Technology
(MIT) that is commonly referred to as 3D Printing.
2.1 3D Printing
3D Printing is a powder-based AM technology that
was developed at MIT around 1993. The technology is
one in which a layer of powder is deposited on a buildplatform, and liquid binder is applied to the appropriate
313
Table 1.
Original powder recipes from Open 3DP.
Stoneware
Powder
Redart
Terracotta Slip
SPS Stoneware
Buff1000 units
SPS Redart
Terracotta
Slip1000 units
Powdered
Powdered
Sugar250 units
Sugar250 units
Maltodextrin250 Maltodextrin250
units
units
Figure 1. 3D Printing Process Diagram.
parts of the layer using an inkjet printer head. Wherever the binder is applied to the powder, the powder
solidifies. The build-platform is then lowered, another
layer of powder is spread on over the first, and the process is repeated until the part is completed within the
powder bed (Figure 1).
After completion, the build-platform is raised and
the loose powder that has not had any binder applied
to it is brushed away to fully expose the completed
green part. Options are then available to strengthen
components, or make them more flexible by applying
additional additives to the part.
Z-Corporation is a 3D Printing licensee that directly
addresses the AM market. They use the process to create conceptual models out of starch, plaster and other
types of powders. A Z-Corporation system was utilized for this study to explore the potential to convert
the system to be used with clay powders.
Clay is a natural mineral aggregate, consisting
essentially of hydrous aluminium silicates. It becomes
plastic when sufficiently wetted, rigid when dried in
the green state, and vitrified when fired to a sufficiently high temperature (Ceramic Tile Institute of
America, 2011). A variety of clays are available in
powder form, making them ideal to use in powderbased 3D printing systems. In essence, this study
explored the potential to deposit the clay material in
powder form, wet it with printed liquid to turn it into its
plastic form (without the kneading traditionally associated with working with clay), before it dries to its
green state, and is subsequently kiln fired to a final
vitrified state.
Although there is much literature on the 3D printing
of advanced ceramics for bone replacement (Lowmunkong et al., 2009), ceramic scaffolds for biomedical applications (Warnke et al., 2010), and with such
advanced materials as Ti3SiC2based ceramics (Nan
et al., 2011), including a seminal paper by Utela, et al.,
on printing a wide range of powder based materials
(Utela et al., 2008), there is little formal literature with
respect to the 3D printing of low-cost clay powders.
An examination of the literature available through a
variety of open-source forums revealed a large amount
of experimental research in the area of 3D printed clay
ceramics. The Open 3DP forum, run by the Solheim
Rapid Prototyping Laboratory in the Mechanical Engineering Department on the University of Washington
(Ganter et al., 2011), and the DIY 3D Printing and
Xtra White
Ceramic
SPS Xtra White
Cone 061000
units
Powdered
Sugar250 units
Maltodextrin250
units
Table 2. Composition of custom-made earthenware powder

and binder.
Custom-made Earthenware
China Clay
Potash Feldspar
Ball Clay
Silica
Talc
Powdered Sugar
Maltodextrin
30%
27%
19%
12%
12%
1000 g
250 g
250 g
Fabrication on Yahoo Forums (Yahoo Forums, 2011),

in particular, present a wide range of experimental
results and recipes for printing low-cost, clay-based
materials.
This literature served as an excellent starting point
for this research into developing a low-cost formulation that was suitable for low-cost student printing.
3
EXPERIMENTAL PROCEDURE
A range of experiments was conducted in order to

examine the effects of multiple variables and their
degree of interaction using samples with standardized
dimensions. The ceramic material can be influenced
at three different levels: the material composition, the
printing parameters, and the firing processes.
3.1 Composition
An in-powder binder system, based on recipes published at the Open 3DP forum, was chosen as an initial
starting point (Table 1). The intent was to trial these
as an initial experiment to observe how the ceramic
powders behaved in a Z-Corp 310 Plus 3D printer.
The first challenge encountered was that none of
the brands/powders listed in the Open 3DP forum
were available in New Zealand, so equivalents had
to be found. In this project, two different powders
were tested, which included a commercially available
mid-fire clay powder and a custom-made earthenware
powder (Table 2).
In contrast to the proprietary Z-Corp materials,
in which the liquid binder (table 3) consolidates the
powder, the above recipes use maltodextrin and sugar
314
Table 3.
XF1 Recipe
91% pure
Isopropyl
Distilled
Water
Food
Colouring
A fast a heating rate can cause cracks and an incomplete or inhomogeneous sintering of the clay particles.
The maximum temperature and holding time influence material strength, shrinkage and distortion. After
firing at low temperatures the material is, in general, more fragile and porous while vitrifying at high
temperatures adds rigidity.
Original binder recipe from Open 3DP.

XB1 Recipe
23%
77%
a few
drops
Denatured
ethanol
Distilled
Water
Food
Colouring
750 ml
1500 ml
a few
drops
4
in the powder material to act as an adhesive which
binds the powder when activated by the selectively
deposited printing liquid. The alcohol-based printing
fluid dissolves the hygroscopic maltodextrin and sugar
on contact, forming a matrix in which the clay particles are bound. During printing and curing the volatile
printing liquid evaporates leaving little residue. This
system is not specific to clay powders and can also
be used in combination with metals such as titanium
(Utela et al., 2008, Wiria, 2003).
The HP10 cartridge, which is the one used by the
Z-Corp 310 Plus 3D printer needs a surface tension of
45 dynes/cm, and a viscosity of 1.35 cps (Utela et al.,
2008). Water is the main component of the binder but,
as water has a surface tension of 72 dynes/cm this surface tension needs to be reduced in order to function
with the print heads. The addition of isopropyl alcohol
(IPA) as a surfactant reduces the surface tension to a
level suitable for the print head. The addition of food
colouring was optional and was used to make it easier to see the object being printed. As the colour is an
organic based material, it burns out during firing.
3.2
3D Printing parameters
The ZPrinter 310 Plus (Z-Corp) allows adjustment to

the following settings: saturation level (binder/volume
ratio) and layer thickness. The porosity and stability
of the green parts is influenced by these parameters.
The amount of rest time required before a part can be
removed from the build platform depends on the curing
time, and therefore on the saturation settings. In addition, the volatility of printing liquid plays an important
role. The saturation levels can be adjusted separately
for the outer shell and inner core of an individual part.
The ratio between the printing liquid and the material/binder mix influences the degree of consolidation
during printing. The layer thickness not only determines the final resolution of a part, but also affects the
required infiltration depth of the printing liquid.
3.3
Firing parameters
Firing transforms printed green parts into vitrified

ceramics. A debinding of the part takes place as the
binder burns out during the firing process and leaves
little residue. Critical parameters in the firing process are the heating rate, maximum temperature and
holding time.
RESULTS
The objective of this research was to create strong

ceramic parts. The green parts had to provide enough
strength to remove them from the 3D printer, and to
depowder. In addition, the dimensional accuracy is
an important attribute with respect to the usability
of a component produced using this process. While
homogenous shrinkage can be compensated for in the
CAD model, the unpredictable distortion of a part
may render it useless for the majority of applications. To analyse the overall suitability of parts created
with different parameter combinations their external
dimensions were measured and compressive strength
tests were performed using a Hounsfield compression
tester.
4.1
Effects of composition
Two different clay powders and two printing fluids

were used to create 4 series covering all possible
combinations based on the recipes provided by Open
3DP. It was possible to achieve strong ceramic parts
using each combination. The main difference observed
was in the saturation required for the shell and core
of a part. As IPA is more volatile than ethanol,
a higher saturation level was necessary to solidify
the powder into the green parts. In addition, the
custom-made powder was observed to produce parts
with greater strength than the commercially available
powder.
4.2
Effects of 3D printing parameters
The saturation settings are the most important factor

that influences the strength of green and fired parts,
but also influences the degree of shrinkage and distortion of a part. A high saturation resulted in the creation
of strong green parts but significant inhomogeneous
shrinkage occurred during the printing process. The
use of a high concentration of printing fluid resulted
in an immediate interaction with the binder which led
to a decrease in part volume. The resulting gaps were
subsequently filled with a new layer of powder, resulting in the formation of a higher packing density and
structural cohesion. As the liquid drained to the bottom of a part (due to gravity), this effect was intensified
resulting in a tapered shape of the final part. A smaller
amount of liquid affected the structural integrity of
the green part, but resulted in high porosity and good
dimensional accuracy (Figure 2).
315
Figure 2. Green part with saturation: a) Shell 50%, Core

50% b) Shell 100%, Core 100%.
Increasing the layer thickness required a higher saturation to ensure complete infiltration and cohesion
between layers.
After firing at high temperatures the level of saturation had an inverse relationship to the strength of
a part. Samples with lower saturation tend to result
in relatively dense parts, but parts were observed to
demonstrate a compressive strength comparable to
concrete.
4.3
Effects of firing parameters
The firing temperature affected the part properties with

respect to the degree of shrinkage and compressive
strength. Bisque firing at low temperatures resulted
in foam-like, extremely brittle parts without structural
cohesion. At this stage, most of the binder had already
burned out while the clay particles had not sintered
together properly. Parts printed with a higher saturation level were slightly stronger, most likely due to
the higher packing density. With an increase in temperature, greater vitrification took place transforming
the powder compound into a ceramic part (figure 3).
The vitrification causes significant shrinkage which
exceeds the shrinkage occurring when firing slipcast parts (figure 4). The shrinkage was sometimes
inconsistent between the top and bottom of the part.
This is one of the factors resulting in inhomogeneous
shrinkage that will be addressed in the future.
Holding time, or how long a part is kept at maximum temperature, was another parameter investigated
in this project. Longer holding times were observed
to allow the particles to rearrange more thoroughly.
Consequently, they were able to sinter into a stronger
part.
5
POROUS FILTER APPLICATION
A series of ceramic filters were produced in order

to test whether variations on printing parameters had
an effect on their porosity. These filters consisted of
discs which were designed in CAD to measure 100 mm
diameter 5 mm thick (figure 5). They were produced
with a number of different binder saturation levels,
and ingredient mixes to determine the effect of these
parameters on porosity.
In this section we present some preliminary results
in which the porosity of the filters is estimated from
an analysis of the microscopy pictures of the filters.
Close up pictures of the filters were taken with an
Figure 3. Effects of firing: a) green part, b) fired at 1200 C,

c) fired at 600 C.
Figure 4. Relationship between saturation and firing: a)

green part (saturation: Shell 100%, Core 100%), b) green
part (saturation: Shell 50%, Core 50%), c) the same parts
after firing at 1200 C.
Figure 5. Simple 100 mm diameter 5 mm thick filter.
Olympus BX60M microscope camera and the pictures

were analyzed with ImageJ.
A comparison was made between the surface area
of solid material and the pores in the material. For each
sample, at least 8 pictures were taken across the surface of the filter, and the porosity for each picture was
estimated using the ImageJ software, and the values
for each filter averaged.
It should be noted that, for this particular paper,
there was not enough time to undertake strength testing
on the filters in order to correlate changes in strength to
corresponding changes in porosity. This work will be
the focus of a future paper which will present the complete factorial design experiment and tabulate results
for a complete comparison.
As can be seen from the results below (Table 4 and
figures 6, 7, 8 and 9), a variation in the saturation levels ranging from 50% to 100% produces a difference
in porosity of around 8%. A change in the proportions of clay powder, sugar and maltodextrin produces
variations in porosity of around 10%.
316
Clay powder [%]
Sugar [%]
Maltodextrin [%]
Saturation [%]
Fire time [min]
Average [%]
Effect of print parameters on porosity.
Trial
Table 4.
4
5
6
17
23
66.7
66.7
66.7
66.7
50
16.7
16.7
16.7
0
25
16.7
16.7
16.7
33.3
25
100
75
50
75
75
360
360
360
360
360
28.4
29.2
22.1
31.4
21.0
Figure 8. Pores on trial 17 with 66.7% clay, no sugar, 33.3%

maltodextrin, a print saturation at 75% and a 31.4% porosity.
Figure 6. Pores on trial 5, with a print saturation at 75% and

a 29.2% porosity.
Figure 9. Pores on trial 23 with 50% clay, 25% sugar, 25%
maltodextrin, a print saturation at 75% and a 21.0% porosity.
Figure 7. Pores on trial 6 with a print saturation at 50% and

a 22.1% porosity.
US$1/Kg). The paper describes the effect of the various parameters and how they influence the quality of
the finished part.
The paper also describes some preliminary experimental work on producing porous ceramic filters
in which the porosity can be controlled by varying
the parameters such as print saturation, and material
composition.
Future work on this project includes analysing the
varying porosity and collating it with the effect of
parameter variations on strength, hardness, etc. Experimental work will also be done on testing filters in
which the geometric structure of the filter is changed
to include channels and other features to aid in the
filtration of specific contaminants.
REFERENCES
CONCLUSIONS
This paper has described current and on-going work in

adapting Z-Corp 3D printers to operate with low-cost
ceramic materials. The components produced with
these clay-based ceramic powders can be fired to produce strong ceramic parts at very low cost (around
Ceramic Tile Institute of America, Glossary of Terms,

http://www.ctioa.org/index.cfm?pi=GL&gaction=list&
grp=C, Accessed March, 2011.
Chua, C.K., Leong, K.F., Rapid Prototyping: Principles and
Applications. (2nd ed). World Scientific Publishing Co,
Singapore, 2003.
317
Lowmunkong, R, Sohmura, T., Suzuki,Y., Matsuya, S., Fabrication of freeform bone-filling calcium phosphate ceramics by gypsum 3D printing method, Journal of Biomedical
Materials Research Part B: Applied Biomaterials, Volume
90B, Issue 2, pages 531539, August 2009.
Ganter, M., et al., open 3D printing forum, http://open3dp.me.
washington.edu/, 03/2011.
Nan, B., Yin, X., Zhang, L., Cheng, L., Three-Dimensional
Printing of Ti3SiC2-Based Ceramics, Journal of the
American Ceramic Society, Article published online: Feb
2011.
Utela, B., Storti, D., Anderson, R., Ganter, M., A review of
process development steps for new material systems in
three dimensional printing (3DP), Journal of Manufacturing Processes 10 (2008) 96104.
Warnke, P.H., Seitz, H., Warnke, F., Becker, S.T., Sivananthan, S., Sherry, E., Liu, Q., Wiltfang, J., Douglas, T.,
Ceramic scaffolds produced by computer-assisted 3D
printing and sintering: Characterization and biocompatibility investigations, Journal of Biomedical Materials
Research, Volume 93B, Issue 1, pages 212217, April
2010.
Wiria FE, et al. Printing of Titanium implant prototype. J.
Mater. Design (2010). Ink-jet printing of highly loaded
particulate suspensions. MRS Bulletin 2003; 28(11):
8158.
Yahoo Forums, DIY 3D Printing and Fabrication, http://
tech.groups.yahoo.com/group/diy_3d_printing_and_
fabrication/, March, 2011.
318

Production of alumina parts through Selective Laser Sintering of

alumina-polyamide composite powder
J. Deckers, S. Khuram, J. Vleugels, J.-P. Kruth & S. Boury
Catholic University of Leuven, Belgium
ABSTRACT: A powder metallurgy (PM) process to fabricate alumina parts through indirect Selective Laser
Sintering (SLS) of spherical alumina alumina-polyamide composite powder particles is presented. The PM
process includes powder production, SLS, debinding and furnace sintering. In order to improve the density of
the final alumina parts, the density of the green parts are optimized through different strategies: remelting, cold
isostatic pressing (CIP), quasi isostatic pressing (QIP) and infiltration. Further, the geometrical quality of the
green parts and final alumina parts is assessed.
1
INTRODUCTION: RESEARCH STRATEGY
SLS of metal or ceramic components can be done

directly or indirectly. Indirect selective laser sintering
involves melting of an organic binder phase to produce green parts. After the laser sintering step, the
binder phase is removed and the density of the part is
improved by furnace sintering or infiltration. Direct
selective laser sintering does not involve a sacrificial
binder phase and the material is directly sintered or
melted to produce parts (Dewidar et al. 2008).
Different types of organic binders are already examined to fabricate ceramic parts via SLS: waxes (e.g.
stearic acid by Liu et al., 2007), thermosets (e.g.
phenolic by Evans et al., 2005), amorphous thermoplastics (e.g. PMMA by Subramanian et al., 1995), or
(semi-)crystalline thermoplastics (e.g. polyamide by
Gill T.J. et al., 2004). T.J. Gill et al. used SLS to produce SiC/polyamide composite parts. The polyamide
(PA) fraction was melted as a result of laser beam irradiation, gluing the SiC particles together. However,
the production of pure SiC parts by binder burnout
and subsequent sintering was not studied.
1.1 Strategy to improve final density
Indirect SLS with the use of a sacrificial binder phase
is a good method to produce crack free green parts, but
the final density after debinding and furnace sintering
is generally low. In order to improve the final density
of the parts, the green density should be as high as possible. In this paper additive manufacturing processing
strategies (remelting) and powder metallurgy processing techniques (CIP, QIP or infiltration) are applied to
improve the green density.
Remelting means to laser scan powder layers multiple times instead of only once. Remelting is already
vastly studied for Selective Laser Melting (SLM) of
metals. It is used to improve the part properties like
Figure 1. Schematic representation of a wet-bag coldisostatic pressing system.
surface roughness, hardness and density (Yasa E. et al,

2011). Also for SLS, remelting has the possibility to
improve the density of the fabricated parts.
Cold isostatic pressing (CIP) is commercially used
to produce homogeneous and high density (up to 60%)
green powder compacts with increased shape complexity by applying pressure from multiple directions.
However, CIP has some geometrical restrictions: it
cannot be used to densify complex geometries with
internal structures (these structures would collapse
during the CIP process). The CIP technology can be
divided into a wet-bag or dry-bag approach. During
wet-bag isostatic pressing, the powder is encapsulated
in a rubber sheath or bag (mould) that is immersed in
a liquid that transmits the pressure uniformly to the
powder (figure 1). Dry bag presses have a bag as an
integral part of the pressure vessel and are used when
many parts of simple geometry have to be made.
Since the shape of the parts is maintained during
wet-beg CIPing, combined indirect SLS and wet-bag
CIPing can be used as a net shaping technique to produce dense parts. Du et al. (2010)used CIPing in order
to produce AISI304 stainless steel parts by SLS. Liu
319
Figure 2. QIP installation which uses alumina as pressure

transmitting medium.
et al. (2010) used CIPing in order to produce dental

K2 O-Al2 O3 -SiO2 glass-ceramic parts by SLS.
The difference between QIPing and wet-bag CIPing, is that during QIPing powder particles are used
instead of a liquid as pressure transmitting medium
(see figure 2).It is also not needed to encapsulate the
samples in a rubber sheath or bag (mould) before
isostatically pressing. Moreover, QIPing can be performed at elevated temperatures, which enables the
binder to plastically deform during isostatic pressing.
On the other hand, since the pressure is only applied in
one direction during the QIP process, the concomitant
shrinkage is non-uniform.
Another way to improve the density of the green
parts is to infiltrate it with a ceramic suspension.
Subramanian et al., 1995 found that infiltration of
small quantities of alumina colloids improve the part
strength greatly during debinding and sintering.
1.2
Strategy to improve geometrical quality
In order to obtain high strength ceramic components

with a good final microstructure (i.e. small grain
sizes), submicrometer ceramic particles have to be
used as a ceramic starting material for SLS. However, such submicrometer powder is hard to deposit
in thin layers during the SLS process: interparticle
forces make submicrometer powder particles to cluster
when deposited by conventional SLS deposition systems, such as a roller or scraper system. The random
clustering of the particles decreases the flowability of
the powder and the deposited layers generally are inhomogeneous (Kolosov et al., 2006). In order to increase
the homogeneity of conventionally deposited layers,
the submicrometer ceramic powders need to be turned
into spherical agglomerates. In this paper, an innovative phase inversion technique is used to produce
spherical agglomerates of about 50 m consisting of
submicrometer alumina encapsulated in a polyamide
matrix. This agglomerate size is large enough to avoid
the unwanted clustering and fine enough to produce
parts with a good geometrical quality.
POWDER METALLURGY PROCESSING

FLOW CHART
Figure 3 schematically presents the ceramic processing

flow charts assessed in this work. The next paragraph
(3) of the paper describes the production of the composite starting powder. Paragraph 4 explains how the
SLS parameters are optimized to produce green parts.
It also explains the debinding (deb.) and furnace sinter
(FS) parameters used in the paper. In order to improve
the final density of the alumina parts, the density
of the green parts has to be optimized. Three startegies to improve the density of the green parts are
discussed in paragraphs 5, 6 and 7: remelting, CIP/QIP
and infiltration. Finally, paragraph 8 demonstrates the
geometrical achievements and limits of the proposed
powder metallurgy process.
The paper investigates the quality of the components during the different processing steps through
density measurements (obtained by using the geometrical method), geometrical assessments and/or
scanning electron microscopy (SEM) microstructural
analyses. In this way, the paper gives an objective view
on the feasibility of processing alumina parts through
SLS of spherical alumina-polyamide composite powder particles.
3
POWDER PRODUCTION: PHASE

INVERSION
A phase inversion (or dissolution-precipitation) technique is used to produce 40wt% alumina 60wt%
PA composite spheres (Yuan and Williams 2007).The
phase inversion technique involves the dissolution of
polymer in a suitable solvent by mixing, heating or
increasing the pressure. The polymer is then allowed to
precipitate from the homogeneous polymer solution by
cooling the solution, also known as thermally induced
phase separation (TIPS), reducing the pressure, evaporation of the solvent or adding a non-solvent (Van de
Witte et al. 1996).
High purity -alumina (grade SM8, Baikowski,
France) powder with a d50 0.3 m, together with
PA powders (grade Duraform PA, 3DSystems, USA)
are added to DMSO in a ratio 5/5/90 vol% alumina,
PA and DMSO (dimethyl sulfoxide). This suspension
is externally stirred in a 2 litre flask. The suspension
is heated to 140 C (above the dissolution temperature
of 135 C) for 15 minutes under N2 atmosphere to
dissolve PA in DMSO. While stirring, the suspension
is allowed to naturally cool down to room temperature. Cooling of the PA solution leads to a liquid-liquid
phase separation to form two phases consisting of a
DMSO rich phase and a PA rich phase. Upon further
cooling the PA rich phase solidifies while surrounding
the alumina submicrometer particles. Vacuum filtration is used to separate the PA-alumina precipitates
from the DMSO and around 80% of the DMSO could
be recovered. Precipitates are subsequently washed
multiple times with ethanol and dried in an oven at
80 C for 24 hours.
320
Figure 3. Powder metallurgy processing flow charts.
Figure 4. Particle size distribution of composite microspheres.
Figure 5. (a) Honeycomb pattern in pure PA powder. (b)

Alumina particles, filling the honeycomb pattern.
Fourier transform infrared spectroscopy (FTIR)

(Avatar 370, Thermo Optek, USA) confirmed that
the polyamide doesnt experience structural changes
during the powder production process. Differential
scanning calorimetry (DSC) and thermal gravimetric
analysis (TGA) (Model-2920, TA instruments, USA)
confirmed that the thermal properties of the composite
powder are more or less the same as the thermal properties of the PA starting powder. The particle size of the
composite microspheres is measured by laser diffraction (Mastersizer Plus, UK). The composite powder,
shown in figure 4 shows a monomodal distribution
with an average diameter of 53 m and a particle
size ranging from 2 to 105 m. The morphology of
the PA starting powder and produced composite powder is studied by scanning electron microscopy (SEM,
XL30-FEG, FEI, The Netherlands). As depicted in

figure 5a, the pure PA powder shows a honeycomb pattern. During the phase separation process, the channels
of this honeycomb pattern are filled with the submicrometer alumina particles (figure 5b). The result of
the powder production process are 53 m spherical
alumina-PA composite particles.
SELECTIVE LASER SINTERING,

DEBINDING AND FURNACE SINTERING
4.1 SLS of the produced powder

Green samples are fabricated using a Sinterstation
2000 machine (DTM Corporation/3DSystems, USA)
321
equipped with a 100 W CO2 laser (f100, Synrad, USA)

with a wavelength of 10.6 m and a laser beam diameter 1/e2 of 400 m. Powder layers are deposited by
a counter current roller and irradiated with the laser
beam. In order to avoid thermal oxidation, SLS is
performed in N2 atmosphere (Air Liquide, Belgium,
[O2 ] < 5%). In order to improve the laser sinterability
of the powder and to avoid thermal cracks, the parts
are produced at elevated temperature of the powder
bed (175 C). These powder preheating parameters
are the standard preheating parameters for SLS of pure
Duraform PA powder (without alumina addition). The
energy required to melt PA is partly supplied by preheating of the powder bed to a temperature slightly
below the melting point of 179 C (distributed cylinder heating and surface IR heating) and by extra laser
irradiation which locally raises the temperature above
the melting point.
According the Quasi Isothermal Theory (Drummer
D. et al., 2010), during the SLS process the composite powder is preheated to a temperature between the
crystallisation and melting temperature of polyamide.
The laser energy is used to melt the PA and the
PA-alumina remains at a melted state until all layers of the component are produced. The surrounding
composite powder that is not melted, supports the generated melt. During the final cooling down stage of the
process, the PA crystallizes and a solid alumina-PA
component is formed.
Besides tests to investigate the powder preheating
and cooling conditions, a parametrical study is performed to investigate other crucial SLS parameters.
In this parametrical study, parts of 15 15 10 mm3
are produced with a laser power P, scan speed v,
scan spacing sand layer thickness lvarying between
respectively 37 W, 3001200 mm/s, 100500 m and
80150 m. The laser energy density, e, combining
these parameters is an important parameter, and is
defined as:
When scanning with low laser energy densities, the

melted PA phase is too little to consolidate powder particles (figure 6a). When scanning with too high laser
energy densities, the PA degrades (figure 6b). Both
too low and too high laser energy densities result in
fragile green parts, which cannot be used for further
processing. During the SLS process, the scanned layers
can delaminate (figure 6c). Delamination especially
occurs when a too low layer thickness (e.g. 80 m)
is used: during layer deposition, the surface roughness
peaks hit the roller which drags along the parts being
produced.
After manual inspecting the strength and the amount
of delamination, four parts are selected to further
investigate the geometrical accuracy by using a vernier
caliper and the green density by using the geometrical
method. Finally, the following parameter set is chosen: 5W laser power, 600 mm/s scan speed, 150 m
scan spacing and 150 m layer thickness (table 1). As
Figure 6. Parts produced with too low (a) and too high (b)
laser energy densities. Delamination of a part with a relatively
low (80 m) layer thickness (c).
Table 1. SLS parameters of selected parts after manual
inspection (optimized parameter set in bold).
Part 1
Part 2
Part 3
Part 4
laser
power
scan
speed
scan
spacing
layer
thickness
mm/s
3
5
5
5
600
600
1250
600
150
150
150
300
150
150
150
150
Table 2. Geometrical and density analysis of the selected

parts (optimized parameter set in bold).
Part 1
Part 2
Part 3
Part 4
volumetrical
error
green
density
relative green
density
g/cm3
% of TD
11
9
11
11
1,18
1,20
1,16
1,17
54
54
52
53
depicted in table 2, this parameter set resulted in a

geometrical error of 9% and a relative green density
of 54% of the theoretical density TD of the composite powder (=2,204 g/cm3 ). The optimized set of
SLS parameters is used for all further parts produced
in this paper. The resulting parts have a mean green
density of 55% of the TD.
4.2 Debinding and furnace sintering
In a debinding step, the polymer is removed from the
green parts. A heating rate of 0.1 C/min is applied
with a 2 hours dwell time at 275 C and 600 C, followed by furnace cooling. The submicrometer alumina
particles are solid state sintered in a furnace sintering
step. In this step a heating rate of 5 C/min is applied
with a dwell time of 60 minutes at 1600 C, followed
by furnace cooling.
After the furnace sintering step, the parts produced
with optimized parameters have a mean density of
about 50% of the TD of alumina (=3.92 g/cm3 ).
SEM images of the final parts (figure 7) show that
322
Table 4. Green densities after remelting (bold, top) and final

densities after furnace sintering (bottom) of the parts which
are remelted twice. Densities are in % of TD. (density before
remelting = 55%).
Figure 7. SEM images of the final part.

Table 3. Green densities after remelting (bold, top) and final
densities after furnace sintering (bottom) of the parts which
are remelted once. Densities are in % of TD. (green density
before remelting = 55%).
Figure 8. SEM images of a part produced by remelting
twice. (P = 10W, v = 1280 mm/s).
Table 5. Relative densities during a PM process, which
uses a CIPing step with varying amount of pressure (100
200 MPa).
CIP pressure
100 MPa
the agglomerates do not collapse during the debinding cycle. Instead, they are sintered together during
furnace sintering.
5
after SLS
after CIP
after deb. and FS
SLS AND REMELTING
A first strategy to improve the densities of the green

parts is to laser scan every powder layer two or more
times instead of only once, i.e. remelting. During
the performed remelting tests, every powder layer is
scanned once with the optimized parameter set for
SLS and also one or two times with different remelting
parameters. The remelting parameters are the following: a laser power P of 510W, a scan speed v of
3201280 mm/s and a scan spacing s of 150 m.
As depicted in table 3 and table 4, applying the
remelting scan strategy can increase the green densities of the parts up to 63%. However, the final density
after debinding and furnace sintering is only increased
up to 51% of the theoretical density. This non significant increase in final density cannot be clearly
observed by comparing the SEM images of figure 8
with the SEM images of figure 7.
55%
74%
62%
150 MPa
200 MPa
55%
77%
53%
54%
78%
59%
COLD AND QUASI ISOSTATIC PRESSING
6.1 Cold isostatic pressing (CIP)

Another strategy to improve the final density is to CIP
the green parts before debinding and furnace sintering.
Samples, produced by SLSing with optimized parameters, are vacuum packed in a rubber sheat or bag. The
vacuum packed samples are placed in a pressure vessel (figure 1) and isostatically pressed at respectively
100 MPa, 150 MPa and 200 MPa.
CIP can increase the relative green density of the
parts from 55% to 76% (table 5), i.e. about 20%.
Final densities up to 62% are obtained. The increase
in final density can be observed by comparing the
microstructure of the produced parts with (figure 9)
and without (figure 7) the CIP step. CIPing at higher
pressures seems to give slightly higher green densities,
but no influence of CIP pressure can be observed on
the final densities of the parts.
323
Figure 9. SEM images of CIP parts after furnace sintering.

Table 6. Relative densities during a PM process which uses
a QIPing step with varying temperature (100 C and 160 C)
and varying amount of time (5 and 30 minutes).
QIP temperature-time
after SLS
after QIP
after FS
6.2
100 C5 min
100 C30 min
160 C5 min
160 C30 min
54%
64%
64%
56%
63%
54%
59%
81%
64%
55%
72%
57%
Quasi isostatic pressing (QIP)
The difference between QIPing and wet-bag CIPing, is

that during QIPing powder particles are used instead
of a liquid as pressure transmitting medium (see figure 2). Moreover, QIPing can be performed at elevated
temperatures. QIPing tests at 100 C and 160 C are
performed for 5 and 30 minutes.
Compared to QIPing at 100 C, QIPing at 160 C
clearly results in higher green densities (table 6). At
160 C PA has a higher plasticity and therefore the
green part can be compressed to higher densities. QIPing for a short period of time (5 minutes) seems to
result in higher green densities, compared to QIPing
for longer time periods (30 minutes). Finally, densities up to 64% of TD are reached by a PM process
which uses a QIP step. This increase in final density
again can be observed by comparing the microstructure of the produced parts with (figure 10) and without
(figure 7) the QIP step.
Figure 10. SEM images of QIP parts after furnace sintering.

Table 7. Relative densities during a PM process, which uses
an infiltration step with varying amount of time (212 hours)
and concentration of the ethanol suspension (10 vol% and
20 vol%).
after SLS
after infiltration
after FS
2 h10 vol%
4 h10 vol%
6 h10 vol%
8 h10 vol%
10 h10 vol%
12 h10 vol%
2 h20 vol%
4 h20 vol%
6 h20 vol%
8 h20 vol%
10 h20 vol%
12 h20 vol%
51%
53%
53%
53%
54%
55%
52%
54%
53%
52%
54%
53%
55%
58%
60%
59%
63%
62%
63%
65%
65%
64%
64%
65%
47%
48%
50%
48%
55%
47%
48%
48%
55%
49%
52%
51%
the ethanol suspensions. The infiltration behaviour of

two ethanol suspensions with different alumina contents (10 vol% and 20 vol%) is investigated. Moreover
the infiltration time is varied between 2 and 12 hours.
As illustrated in table 7, infiltrating with a 20 vol%
alumina suspension results in higher green densities
compared to infiltrating with a 10 vol% alumina suspension. Infiltrating for longer periods of time seems
to result in slightly higher green densities. Final densities up to 55% of the TD of alumina are reached
after furnace sintering. No clear correlation between
the final density of the parts and the alumina concentration in the suspension or infiltration time can be
observed. Moreover, no microstructural changes can
be clearly observed by comparing the SEM images of
figure 11 with the SEM images of figure 7.
INFILTRATION
A last method investigated to improve the green density is to infiltrate green parts with a suspension
containing submicrometer alumina particles (grade
SM8, Baikowski, France). As well water as ethanol
are used as suspension fluida, but only a significant
increase in green density could be established by using
GEOMETRICAL ASSESSMENTS
The investigated PM process enables the production

of freeform parts with a good geometrical accuracy
through indirect SLS (see figure 12). The production
of parts with a good geometrical accuracy is possible
324
Table 8.
Part 1
Part 2
Part 3
Part 4
Part 5
Part 6
Part 7
Part 8
Part 9
Part 10
Dimensional error of SLSed parts.

error in
x-direction
error in
y-direction
error in
z-direction
0
1
1
1
1
1
5
3
2
1
1
2
1
3
4
1
3
1
5
1
9
11
11
5
10
10
10
15
10
13
Figure 11. SEM images of infiltrated parts after furnace

sintering.
Figure 12. Complex shaped alumina parts produced by the

PM process.
since the particle size of the starting powder is relatively low (50 m). Moreover, due to the spherical
shape of the powder, the small particles can be well
deposited by the counter current roller system which
is used during SLS.
8.1
Figure 13. Seemingly uniform shrinkage of part during

debinding and furnace sintering.
means that there is a non-uniform deformation of the

part during the sinter process.
Dimensional error introduced during SLS
The dimensional error introduced during the SLS process is investigated by producing 10 identical parts
(15 1510 mm3 ) with the optimized SLS parameters. The the x-, y- and z-dimensions of the parts are
measured by a vernier calliper and used to calculate the
procentual error according to the following formula:
Table 8 shows that the procentual dimensional

errors in the scan directions (i.e. x- and y-directions)
have more or less the same magnitude, but are relatively small compared to the dimensional error of
about 10% in the build direction (z-direction). This
8.2 Dimensional change during debinding and

furnace sintering
Figure 13 represents the famous Belgian statue Manneken Pis as a green and final part. The figure
illustrates that the produced green parts undergo a
large but seemingly uniform shrinkage during debinding and furnace sintering.
8.3 Dimensional change during the whole
PM process
A benchmark part is produced in order to investigate the roughness, the dimensional accuracy and the
shrinkage in the x- and y-directions during the PM process. After the SLS step and after the debinding and
325
Figure 15. Discrepancy between designed x-y-dimensions

and x-y dimensions of green and final part.
Figure 14. Benchmark part.

Table 9. Mean roughness values of green and final benchmark part.
Green part: Ra [ m]
Green part: Rt [ m]
Green part: Rz [ m]
Final part: Ra [ m]
Final part: Rt [ m]
Final part: Rz [ m]
x-direction
y-direction
17 m
140 m
110 m
16 m
112 m
98 m
17 m
140 m
112 m
15 m
104 m
88 m
furnace sintering step, roughness values are obtained

by a Talysurf-120L device (Taylor-Hobson, UK).
Multiple dimensions of the green and final benchmark part (figure 14) are measured by a coordinate
measuring machine (CMM, FN905, Mitutoyo, Japan).
Different roughness values (Ra, Rt and Rz) are
obtained in x- and y-direction. A gaussion filter with a
lower (Ls) and higher (Lc) cutoff value of respectively
0,008 mm and 2,5 mm is used to process the measured
data.
As depicted in table 9, there seems to be no significant directional dependency between the roughness
values in the x- and y-dirction. On the other hand, the
roughness values of the final parts are somewhat lower
as the roughness values of the green parts. This may
be due to the shrinkage of part during debinding and
furnace sintering.
Figure 15 compares different dimensions of the
CAD design of the benchmark part with the corresponding CMM measurements of the green and final
part. The green part dimensions are slightly smaller
compared to the dimensions of the designed part, but
during debinding and furnace sintering a larger shrinkage occurs. The larger the size of the dimension, the
more the absolute amount of shrinkage.
From the measurements with the CMM, the procentual shrinkage of the parts during the SLS process
and during debinding and funace sintering are calculated, together with the 95% confidence intervals
(figure 16). A distinction is made between internal and external geometries. After the SLS process,
the x- and y-dimensions of the benchmark part are
Figure 16. Shrinkage of internal and external geometries

during the PM process (95% confidence levels are shown).
Table 10. Evolution of densities during different PM
processes.
mean
maximal
maximal
density after density after
final
SLS
post processing density
Only SLS
SLS & remelting
SLS & CIP
SLS & QIP
SLS & infiltration
55
55
55
55
63
78
81
65
50
51
62
64
55
generally some percentages smaller than the corresponding CAD dimensions. On the other hand, the
x- and y-dimensions of the final part are about 75%
80% of the original CAD dimensions. Finally, the 95%
confidence intervals are much smaller for the external
as for the internal dimensions. This means that the part
accuracies are better for external geometries.
9
SUMMARY & CONCLUSIONS
A PM process is presented to produce freeform

alumina parts through indirect SLS. An innovative phase inversion technique is used to produce
spherical alumina-PA composite starting powder of
326
50 m. Different strategies are investigated to

increase the green and final densities of the produced
parts: remelting, CIP, QIP and infiltration (table 10).
Applying a remelting strategy can increase the green
densities of the parts after SLS, but not significantly
the final densities. Using CIP and QIP can increase
the final density of the alumina part up to 62% or 64%
of the TD. By applying the infiltration technique, the
green and final density of the produced parts can be
increased up to respectively 65% and 55% of the TD.
Notice that the remelting and infiltration technique
can be applied on any geometry, whether CIPing cannot be applied on e.g. parts with internal structures.
QIPing has the disadvantage to deform the green part
non-uniformly. A suggestion for further research is
to investigate further the influence of remelting and
infiltration (or maybe a combination of both) on the
final densities of the parts. Also working with other
binder materials which have a lower viscosity during
the SLS process seems a promising strategy to improve
the green and final densities of the parts.
As depicted in figure 12 and figure 13, the presented PM process allows the production of freeform
and complex alumina geometries with relatively low
roughness values. During the SLS process, the dimensional error in the scan directions is more or less the
same, but much lower as the dimensional error of about
10% in the build direction. The produced parts seem
to undergo a uniform shrinkage of about 75%80%
during the debinding and furnace sintering step. Also
the assessed roughness values decrease during debinding and furnace sintering. Finally, it is observed that
external geometries can be produced more accurately
than internal geometries.
ACKNOWLEDGEMENTS
The authors acknowledge IWT for support through
the SBO-project DiRaMaP.
REFERENCES
sintering of metals, J. Mater. Sci. Technol., Vol. 24,

No. 2, pp. 227232.
Z.H. Liu, J.J. Nolte, J.I. Packard, G. Hilmas, F. Dogan and
M.C. Leu (2007), Selective laser sintering of high density alumina ceramic parts, Proceedings of the 35th
international MATADOR conference, pp. 351354.
Evans R.S., Bourell D.L., Beaman J.J., Campbell M.I. (2005),
Rapid manufacturing of silicon carbide composites,
Rapid Prototyping Journal, Vol. 11, No. 1, pp. 3740.
Subramanian K., Vail N., Barlow J., Marcus H. (1995),
Selective laser sintering of alumina with polymer
binders, Rapid Prototyping Journal, Vol. 1, No. 2, pp.
2435.
Gill T.J., Hon K.K.B. (2004), Experimental investigations into the selective laser sintering of silicon carbide
polyamide composites, Proceedings of the Institution of
Mechanical Engineers, Part B: Journal of Engineering
Manufacture, Vol. 218, No. 10, pp. 12491256.
Yasa, E., Deckers, J., Kruth, J. (2011). The investigation of
the influence of laser re-melting on density, surface quality
and microstructure of selective laser melting parts, Rapid
Prototyping Journal (unpub.)
Du Y.Y., Shi Y.S., Wei Q.S. (2010), Forming simulation and
experimental verification of combined formation of selective laser sintering and cold isostatic pressing, Journal of
Materials Science Engineering and Performance, DOI:
10.1007/s11665010-96852.
Liu J., Zhang B.,Yan C., ShiY. (2010), The effect of processing parameters on characteristics of selective laser sintering dental glass-ceramic powder, Rapid Prototyping
Journal, Vol. 16, No. 2, pp. 138145.
Kolosov S., Vansteenkiste G., Boudeau N., Gelin J.C.,
Boillat E. (2006), Homogeneity aspects in selective
laser sintering (SLS), Journal of Materials Processing
Technology, Vol. 177, pp. 348351.
Yuan Q., Williams R.A. (2007), Large scale manufacture
of magnetic polymer particles using membranes and
microfluidic devices, China Particuology, Vol. 5, Issues
12, pp. 2642.
Van de Witte P., Dijkstra P.J., Van den Berg J.W.A., Feijen
J. (1996), Review: Phase separation processes in polymer
solutions in relation to membrane formation, Journal of
Membrane Science, Vol. 117, pp. 131.
Drummer D. et al. (2010), Development of a charcterization
approach for the sintering behavior of new thermoplastics
for selective laser sintering, Physics Procedia, Vol. 5,
pp. 533542.
Dewidar M. M., Lim J.-K., Dalgarno K.W. (2008), A

comparison between direct and indirect selective laser
327

Material characterisation of Additive Manufacturing components made from

a polyetherketone (PEK) high temperature thermoplastic polymer
M.A. Beard, O.R. Ghita, J. Bradbury, S. Flint & K.E. Evans
University of Exeter, College of Engineering, Mathematics and Physical Sciences, Exeter, EX4 4QF, UK
ABSTRACT: Additive Manufacturing (AM) is a process capable of producing optimised engineered components with highly complex geometries that would be impossible by any other manufacturing method. Until
recently, Selective Laser Sintering (SLS) has been limited to using thermoplastics with relatively low glass
transitions and melt temperatures. This paper investigates some of the physical and material properties of a
high performance PEK polymer powder in the SLS process. The results reveal the significant improvements in
properties using this range of materials over traditional polyamide materials used in the SLS process. Thermal
analysis of the samples shows stable properties such as glass transition and entropy of melting across the build
chamber of the system.
INTRODUCTION
Additive Manufacturing (AM) offers unique advantages for the production of bespoke parts with highly
complex features. AM has opened up the possibility
creating new products both for old and new markets
and across many sectors (Wohlers 2009). The key benefits of manufacturing components this way include
the ability to fabricate parts that are impossible to
make by traditional methods (e.g. injection moulding
and milling) and the ability to optimise existing parts,
leading to reduced weight and material usage. Other
advantages include eliminating the need for expensive
tooling and reducing the time for a new product to
reach the market.
High performance thermoplastics such as polyamide
12 (PA12) have been used in AM for many years producing dense parts with relatively high mechanical
strength (Wendel et al. 2008). A PA12 matrix reinforced with glass beads, carbon or aluminium has
been used to improve mechanical properties. However
certain material characteristics, such as low melting
temperatures (Tm) and low glass transition temperatures (Tg), limit the possible applications it can
be used for. The high performance thermoplastics
of the polyaryletherketone family (Brydson 1982)
such as polyetherketone (PEK) have high melting
temperatures, excellent mechanical properties, high
wear and chemical resistance, along with high resistance to burning, smoke and toxicity performance.
These attributes enable PEK to be used across various industrial sectors with applications such as piston
components, bearing linings, wire insulation, cable
couplings and connections, prosthetics, pump casings
and filters. There are also opportunities for parts that
are biocompatible for use as load bearing implants
(Drummer et al. 2010). Until recently, thermoplastics
such as PEK have not been used for AM, primarily because of the process temperatures which are
too high for use in standard PA equipment. However,
recent developments mean that a commercial machine
is now available for use with high temperature polymer
powders.
Previous research into selective laser sintering of
high temperature polymers has focused mostly on
various grades of PEEK materials in modified standard equipment (Rechtenwald et al. 2005, Pohle et al.
2007). One of the main challenges in these studies was
achieving a much higher preheating temperature of the
powder than the standard PA12. The melting temperature of PEEK is around 370 C compared to 172 C for
PA12. Rechtenwald et al. (2005) modified an EOSINT
P380 by integrating an additional heating device into a
circular building platform to allow the system to reach
temperatures of up to 250 C. The effect of a still relatively low preheating temperature had a significant
effect on the relative density of the parts produced and
subsequently on the bending tests results. In a follow
on study (Pohle et al. 2007), the same group designed
a high-temperature inner process chamber called a
Heating Dome to increase further the preheat temperature to 345 C. Using this system, the preheating
was uniform over the entire building platform. Unfortunately, no mechanical testing was performed, and
only osteoblast proliferation in vitro was carried out.
In addition to problems of high the preheating
temperatures other studies (von Wilmowsky et al. 2008,
Pohle et al. 2007) have raised the concern of irregular geometry of PEEK particles and tried to overcome
it by introducing small percentages of fillers. Hence,
Wilmowsky et al. (2008) and Pohle et al. (2007)
both used 1% wt carbon black in addition to other
bioactive ingredients. In both cases, the repose angle
was reduced and the flow ability improved.
329
Figure 1. Diagram of the build position of parts within the

build chamber.
This paper presents an initial study into parts made

in a standard high temperature SLS machine (EOSINT
P800 [8]) using EOS PEEK HP3 material [9]. The
physical and material properties of the parts made
using PEK powder were examined and presented in
the following sections.
2
2.1
EXPERIMENTAL
Figure 2. SEM image of the surface of a PEK sample.
the lid. Dynamic heating runs were performed from

30 C to 400 C at a heating rate of 10 C min1 .
For each sample four repeat measurements were
performed.
2.5
Material and AM part production
Parts were manufactured using EOSINT P800 (EOS

Electro Optical Systems 2011a) using EOS Original Parameters (EOM). The system uses a preheating temperature of approximately 366 C. Parts
were built in dogbone geometries with dimensions
180 10 4 mm. The build bed positions are shown
in Figure 1.
The material used was EOS PEEK HP3 powder
(EOS Electro Optical Systems 2011b), supplied by
EOS Ltd. EOS PEEK HP3 is a powder specifically
designed for use in additive manufacture, optimised
to allow good particle sintering and parts formation.
Dynamic Mechanical Thermal Analysis

(DMTA)
DMTA tests were performed on the samples using

a Mettler-Toledo DMA SDTA861e. The experiments
were performed in three point bending mode at a frequency of 1 Hz. The storage modulus (E) and loss
factor (tan ) were recorded for each specimen having nominal dimensions of 75 mm 10 mm 4 mm.
The samples were heated during the experiment from
30 C to 250 C at a heating rate of 3 C/min. The strain
displacement was kept constant at 10 m to ensure
measurements were performed in the linear viscoelastic region. Three repeat measurements for each sample
were carried out.
2.2 Scanning Electron Microscopy (SEM)
The SEM examination of the SLS manufactured samples was performed using a Hitachi S-3200N scanning electron microscope. All samples were coated
with a 4 nm gold coating to reduce surface charging and secondary electron images were taken with
an accelerating voltage of 25 kV.
3.1 Topology and material structure
2.3 Micro Computer Tomography ( CT)

Computer Tomography was performed using a CT
160 Xi System. The CT 160 Xi system is fitted with a
high contrast x-ray image intensifier, 100 KV 5 micron
open type x-ray source to reconstruct the image.
2.4 Differential Scanning Calorimetry (DSC)
The DSC tests were performed using a Mettler Toledo
820e, under a nitrogen atmosphere. Small specimens
were cut from the SLS dog bone components and
placed in closed aluminum pans with a small hole in
RESULTS AND DISCUSIONS
Failings in surface finish and porosity of SLS samples

are often quoted as disadvantages in the parts produced. Optimisation of the SLS parameters (e.g. scan
speed, laser power, hatching spacing) is constantly carried out to allow manufacture of dense samples with
good surface finish. To help assess the degree of sintering between PEK particles, SEM analysis of the
samples was conducted. The first stage of the analysis
was to examine the surface of the samples. A typical
image is shown in Figure 2.
Laser sintering is known to give a porous surface
due to the partial sintering effect of powder around
the contour of the geometry. Individual particles can
be identified in Figure 2 with partial wetting of adjacent particles. This wetting effects produces a surface
that is porous but with partially melted particles that
resemble flattened particles. To help establish the bulk
structure of the sintered material the samples were
330
Figure 3. SEM image of the bulk structure of a PEK sample.

Figure 4. CT scan of a PEK sample.
fractured to examine the material below the surface.

The facture surface of a sample is shown in Figure 3.
The fracture surface shown in Figure 3 is typical for all built PEK samples and clearly shows the
layering effect throughout the XY build plane of the
sample. Distances between each layer is approximately
0.25 mm and is consistent throughout the entire sample
structure. The interface between layers was examined
using the SEM at various magnifications. No evidence
of voids or detrimental layer interaction was found.
As SEM is a surface analysis technique a CT scan
was used to analyse the sub-surface properties of the
part further. Figure 4 shows a cross-section through the
bulk of a dogbone. Differences in roughness between
the surfaces of part on the XY plane and the ZY plane
can be clearly observed. The CT image shows the parts
to be fully dense throughout the bulk of the part. Any
dark areas would indicate the presence of the voids or
defects within the sample.
3.2
Mechanical and thermal analysis
As described in the introduction, previous research on

high temperature laser sintering has focused mostly on
machine adaptation (Rechtenwald et al. 2005, Pohle
et al. 2007) and parameter selection (Rechtenwald
et al. 2005, Tan et al. 2003). Except SEM analysis,
to date no thorough analysis of the material properties
of the PEEK parts produced using SLS is described in
any research literature.
Figure 5 and 6, shows the E and tan results of
the DMTA temperature sweep for the three different
samples. The DMTA analysis clearly shows the typical
behaviour of a crystalline polymer passing through its
glass transition temperature. Within the glass region
at temperatures below 140 C the storage modulus and
tan remain constant until the brittle/tough transition temperature is reached at approximately 150 C.
As the sample continues to pass through its glass transition temperature at 180 C the modulus continues to
decreases sharply as the viscoelastic behaviour of the
sample reaches the rubber regime. The modulus of
the sample drops from 3.5 GPa to 0.5 GPa during this
Figure 5. Modulus values of PEK samples (only three

samples shown for clarity).
transition and is related to the mechanical performance

as well as the fracture mechanics of the samples as they
move from a glassy to a rubbery state.
The modulus results presented in Figure 5 show
that the modulus does not vary significantly between
the test specimens. Small changes in modulus
below the brittle tough transition temperature can be
observed but above this temperature the curves overlay each other very well. Small changes in the glass
transition temperature can be observed in Figure 6.
Due to the length of the samples tested in the DMTA
any slight differences in properties along its length
would be averaged out to smooth the data. Smaller
samples would help to identify if any of the mechanical
or thermal properties of the samples vary from smaller
specimens manufactured at different places within the
build chamber. Although this averaging effect may
be noticed along the x axis (see Figure 1), specimens
were built at different y positions and show little difference in glass transition and modulus between parts.
The DSC results show good repeatability and
consistency within the same sample and across
the range of samples tested. For example, sample
LX1.2 had an average Tg value of 178.2 C and
331
Figure 7. DSC thermograms of the PEK samples (only three

samples shown for clarity).
Figure 6. Tan values of PEK samples (only three samples

shown for clarity).
enthalpy of 56.6 J g1 , similar to samples LX1.6 which

recorded an average Tg of 180.7 C and enthalpy of
57.6 J g1 . The presence of only one melting peak
(see Figure 7) shows a relatively simple crystallisation
process, without any annealing or re-crystallisation
effects. DSC crystallisation investigations of PEEK,
a different polyetherketone derivative than the one
studied here, identified double melting peaks representing primary (first formed) lamellae at highest
melting point and crystallites amongst the lamellae at
low melting points (Bassett et al. 1988).
The morphology and thermal behaviour of of-theshelf polyaryletherketone materials (Lee & Porter
1988, Chu & Schultz 1990) or injection moulded parts
(Bakar et al. 2003) has been extensively explored.
However, this is the first study to attempt the analysis of PEK samples manufactured using standard high
temperature SLS machine and powder specifically
designed for this technology.
4
CONCLUSIONS
The results presented in this paper show some of the

improvements the new selective laser sintering process of PEK powder has over standard PA12 material.
Parts produced are fully dense with excellent thermal
and mechanical properties that do not appear to vary
significantly with build bed position.
ACKNOWLEDGEMENTS
The authors would like to thank Electro Optical Systems (EOS) for their technical assistance throughout
the research and Dr Wears for her assistance with the
CT imaging.
REFERENCES
Bakar, M.S.A., Cheang P. & Khor K.A. 2003. Mechanical properties of injection molded hydroxyapatite
polyetheretherketone biocomposites. Composites Science
and Technology 63 (34): 421425.
Bassett, D.C., Olley. R.H. & Al Raheil, I.A.M. 1988. On crystallization phenomena in PEEK. Polymer 29: 17451754.
Brydson, J.A. 1982. Plastic Materials. London: Butterworth
Scientific.
Chu, J.M. & Schultz J.M. 1990. The influence of microstructure on the failure behaviour of PEEK. Journal of
Materials Science 25(8): 37463752.
Drummer, D., Rietzel, D. & Khnlein F. 2010. Development
of a characterization approach for the sintering behavior
of new thermoplastics for selective laser sintering. Physics
Procedia 5(2): 533542.
EOS Electro Optical Systems 2011a. http://www.eos.info/
en / products / systems-equipment / plastic-laser-sinteringsystems/eosint-p-800.html
EOS Electro Optical Systems 2011b. http://www.eos.info/
en/products/materials/materials-for-plastic-systems/eospeek-hp3.html
Lee, Y. & Porter R.S. 1988. Effects of thermal history
on crystallization of poly(ether ether ketone) (PEEK).
Macromolecules 21(9): 27702776.
Pohle, D., Ponader, S., Rechtenwald, T., Schmidt, M.,
Schlegel K.A., Mnstedt, H., Neukam, F.W., Nkenke,
E. & von Wilmowsky, C. 2007. Processing of threedimensional laser sintered polyetheretherketone composites and testing of osteoblast proliferation in vitro.
Macromolecular Symposia 253(1):6570.
Rechtenwald, T., Eer, G., Schmidt, M. & Pohle, D. 2005.
Comparison between laser sintering of PEEK and PA
using design of experiment method. Proceedings of the
3rd International WLT-Conference on Lasers in Manufacturing (LIM 2005) 3:263267.
Tan, K.H., Chua, C.K., Leong, K.F., Cheah, C.M.,
Cheang, P., Abu Bakar, M.S. & Cha, S.W. 2003.
Scaffold development using selective laser sintering of polyetheretherketone-hydroxyapatite biocomposite
blends. Biomaterials 24(18): 31153123.
von Wilmowsky, C., Vairaktaris, E., Pohle, D., Rechtenwald,
T., Lutz, R., Mnstedt, H., Koller, G., Schmidt, M.,
Neukam, F.W., Schlegel K.A. & Nkenke, E. 2008.
Effects of bioactive glass and beta-TCP containing threedimensional laser sintered polyetheretherketone composites on osteoblasts in vitro. Journal of Biomedical
Materials Research-Part A 87(4): 896902.
Wendel, B., Rietzel, D., Khnlein F., Feulner, R., Hlder
G. & Schmachtenberg, E. 2008. Additive Processing of
Polymers. Molecular Materials and Engineering 293(10):
799809.
Wohlers, T. 2009. Wohlers Report 2009. Colorado USA:
Wohlers Associates Inc.
332

Material study of laser cladded Inconel 625

M. Rombouts, G. Maes & R. Persoons
VITO, Mol, Belgium
ABSTRACT: Laser cladding is a near-net shape technology, that has as unique capability to build up complex
features on existing components. The process consists in depositing different layers of metal on top of each other
by laser melting metal powder delivered in a gas stream to the metal substrate. In this investigation Inconel 625
powder is used as feedstock material. This superalloy is an attractive material for applications requiring good
mechanical strength and oxidation resistance at high operating temperatures. It is well known that properties
such as mechanical strength are dependent on the thermal and mechanical history of the material. During laser
cladding a rapidly solidified microstructure, which differs from the microstructure in conventional parts, is
formed. In this paper the microstructure and mechanical properties after laser cladding is studied. A mixed
cellular dendritic structure composed of a Ni-based phase and MC type of carbides is formed after laser
cladding. The tensile strength is higher while the elongation is lower for samples loaded perpendicular to the
built direction than for those loaded parallel to the build direction. Upon heating to 1000 C and 1200 C followed
by furnace cooling, M23 C6 and M6 C carbides are formed. In addition, a drop in hardness is observed.
1
INTRODUCTION
Laser cladding also called LENS or laser metal

deposition is an additive process wherein a
laser source is used to melt metal-based powder or
wire on to a metal substrate. Its main application area
in industry is the repair of high-value components like
turbine blades. The same principle can also be used
to apply thick wear and/or corrosion resistant coatings like inconel alloys, stainless steels, stellites and
metal matrix composites, . . . The background of this
paper is the use of laser cladding to produce functional components in an additive way or to build up
functional structures on existing parts. In this case,
not only the knowledge of the surface properties like
corrosion and wear resistance but also of the bulk
mechanical properties of the laser cladded material is
important. The microstructure of laser cladded components differs substantially from conventional material
due to the rapid solidification, which can lead to the
formation of non-equilibrium phases and element distribution. Furthermore, the rapid solidification results
in a fine grain-sized microstructure. Although these
microstructural features often bring about improvements in properties, rapidly solidified microstructures
in commercial alloys that were designed to be used
in conventional processed condition do not necessarily show superior properties. Moreover, due to the
different microstructural features the effect of heat
treatment of rapidly solidified materials may also
be different and bring about the rapid formation of
undesirable phases.
Inconel 625 is a Ni-Cr-Mo superalloy that is resistant to a wide range of severely corrosive environments
in combination with a good mechanical strength, also
at elevated temperatures. The material is used in

chemical processing, aerospace and marine engineering and nuclear reactors. Many of these components
have complex shapes and are expensive to produce
due to extensive machining. Inconel 625 achieves its
strength by solid-solution hardening of the austenitic
matrix by the presence of Mo and Nb. Precipitation hardening in this alloy at elevated temperature is
mainly derived from the metastable phase at aging
temperatures of 550650 C and phase at temperature above 750 C (Shankar et al. (2001)). It is also
observed that precipitation of MC, M23 C6 and M6 C
carbides occurs on subjecting the alloy to temperatures
of 760 C980 C (Shankar et al. (2001)).
In this paper the microstructure and mechanical
properties of laser cladded Inconel alloy 625 are
presented. The response of the laser cladded samples
upon high temperature exposure is studied.
2
EXPERIMENTAL
Laser cladding is performed using a 7 kW IPG fiber

laser with out-coupling fiber with a diameter of
600 m. The use of a focus lens of 250 mm and a collimator lens of 125 mm results in a laser spot diameter
of 1200 m on the substrate.
The parts are built on stainless steel AISI 316L flat
substrates with a thickness of 8 mm. The Inconel 625
gas atomized powder has following chemical composition: Ni (bal.) 21.5% Cr 8.8% Mo 3.71% Nb
0.96% Fe 0.47% Mn 0.41% Si 0.06% O 0.12%
N 0.02% Al 0.03% Ti. The particle size distribution measured by laser diffraction is: d10 = 65 m,
d50 = 91 m, d90 = 130 m. The powder is fed to the
333
Table 1. Process parameters of fill tracks applied to

produce tensile test specimens.
Laser Scan
Scan
Layer
Powder Transport
power speed
spacing thickness flow
gas
W
mm/min mm
mm
g/min
l/min
500
750
0.6
0.4
2.46
Figure 2. Laser cladded samples used to produce tensile

test specimens for investigation the effect of build direction
(corresponds to the vertical axis). The arrow indicates the
tensile loading direction.
The effect of heat treatment on the hardness and

microstructure is studied. Samples are held during
1 hour at 800 C, 1000 C and 1200 C under argon
atmosphere, followed by furnace cooling at a rate of
20 C/min.
Figure 1. Dimensions (in mm) of tensile test specimens.
coaxial cladding nozzle (Fraunhofer ILT) in an argon

gas stream by means of a Medicoat powder feeder. In
addition, argon shielding gas is provided.
SKM-DCAM software is used for the generation of
the CNC program with specified toolpaths. The parts
are built up by first scanning an outer contour. The
scanning direction of successive layers is rotated by
an angle of 90 . In one layer, a bidirectional scanning
pattern is applied. The process parameters applied for
the production of tensile test specimens are given in
Table 1.
For the production of tensile test bars (see
Figure 1), two different geometries are built. One set
of tensile bars are machined from laser cladded beams
built in standing orientation, which are 10 * 15 mm2
and 100 mm high (Figure 2(a)). The tensile loading
direction is parallel to the build direction. For the second set the loading direction is perpendicular to the
build direction. These specimens, which are built in
lying orientation, are 25 * 100 mm2 and 10 mm high
(Figure 2(b)). The specimens are milled to the final
shape prior to tensile testing.
The microstructure is studied by optical and scanning electron (Jeol JSM-6340F) microscopy and X-ray
Diffraction (Siemens D500). The samples are etched
using Vogels Sparbeize.
The mechanical properties are determined by tensile testing (Instron 5582) using an extensometer. At
least 4 specimens produced under identical conditions
have been subjected to tensile testing. The hardness is
obtained by Vickers indentation measurements using
a load of 1 kg.
3.1 Microstructure
The density of the parts measured using Archimedes
method is 99100%. Only some small spherical gas
inclusions are present in the samples (Figure 3).
Figure 4 shows the etched cross-section of the tensile test specimens near the top. The melt pool formed
at the top of a sample built in standing orientation
(100 mm high) is clearly larger (order of 40%) than
of a sample built in lying orientation (10 mm high).
This is attributed to the lower heat conduction in the
former case. Figure 5 shows the micrograph taken
by SEM. The microstructure is fine cellular/dendritic.
The structure is finer for the samples built in lying
orientation (dendrite spacing in the order of 4 m),
as expected. The phases present in the intercellular/dendritic region could not be resolved by SEM due
to the fine scale.
The main phase present after laser cladding is a
Ni-based (fcc) phase, as determined by X-ray diffraction (Figure 6). The lattice parameter of this phase in
the laser cladded sample calculated based on the first 3
reflections is 0.36000.3606 nm. The lattice parameter
of solution treated inconel 625 is reported in literature
to be 0.3600 (Dey et al. (1989)). A slightly higher lattice parameter could be attributed to the greater level
of supersaturation in the rapidly solidified structure. In
the laser cladded material also the presence of a phase
with reflections coinciding with those of (Nb, Ti, Ta)C
(Davis (2000)) primary carbides is observed. Earlier
334
Figure 3. Cross section of laser cladded Inconel 625 sample.
Figure 5. Etched scanning electron micrographs at the top

of laser cladded Inconel 625 samples built in standing (top)
and lying (bottom) orientation (see Figure 2). The build
direction is along the vertical axis.
Figure 4. Etched optical micrographs at the top of laser

cladded Inconel 625 samples built in standing (top) and
lying (bottom) orientation (see Figure 2). The arrow indicates the build direction.
work had indicated that solidification products in commercial Inconel 625 could include both MC or Laves
(hexagonal crystal structure) eutectic (Cieslak et al.
(1988), Formenti et al. (2005)). The formation of small
Nb-rich MC particles located primarily at the grain and
cell boundaries was also observed after rapid solidification of Inconel 625 (Dey et al. (1989)). The primary
carbides serve to control the grain structure and contribute to the strength. Besides and carbides also
Figure 6. X-ray diffraction data of laser cladded Inconel 625

sample.
phases like fcc (Ni3 (AlTi)), bct (Ni3 (NbAlTi))

and orthorhombic (after prolonged aging) phases
could be formed in Inconel 625. However, the detection of these phases is not straightforward without
prolonged aging and requires analysis techniques like
TEM (Rai et al. (2004)).
335
Figure 7. Tensile test bars after tensile testing.
Figure 8. Mechanical properties of laser cladded Inconel

625 components determined by tensile testing (see Figure 2).
Literature data corresponding to for cast material (CW-6MC)
(ASTM specification SA-494) and annealed sheet (ASM
Metals Handbook, Vol. 2) are given as reference.
3.2
Mechanical properties under tensile loading
Figure 7 shows some tensile test bars after loading. The

results of the tensile tests are summarized in Figure 8.
The yield and ultimate tensile strength are lower for
the specimens loaded parallel to the build direction
(standing orientation). The elongation on the other
hand is higher for these samples. The coarser grain
structure for the samples built in standing orientation
will contribute to this difference.The dendritic/cellular
structure formed upon solidification can clearly be
resolved on the fracture surfaces (Figure 9). Paul et al.
(2007) did not observe any effect of build orientation
on the tensile properties. It should however be noted
that sub-size specimens were applied in this study and
the difference in heat flow for samples built in different
orientations will not be as pronounced as in this study.
The properties for cast material (CW-6MC) and
annealed sheet are given as reference in Figure 8. It can
be observed that generally the strength values for both
sets of specimens are better than for cast and annealed
sheet material. The elongation is better than for cast
material but inferior compared to annealed material.
3.3
Effect of high temperature exposure
Figure 10 shows the Vickers hardness of laser cladded

samples in as-produced and heat treated state. The
hardness reported in literature for annealed Inconel
625 is 150200 HV. This is well below the values
Figure 9. Fracture surfaces after tensile testing for samples

built in lying (top) and standing (bottom) orientation.
Figure 10. Vickers hardness (1 kg) of laser cladded Inconel

625 components after heat treatments at 800 C, 1000 C and
1200 C during 1 hour in argon atmosphere.
measured for the laser cladded specimens (320 HV).

The hardness after rapid solidification by melt spinning of Inconel 625 is 250 HV, which is also lower
than after laser cladding (Dey et al. (1989)). The hardness is not significantly altered at 800 C but a drop
can be observed at 1000 C and 1200 C.
No marked difference in XRD profile is observed
for the sample heated to 800 C. The lattice parameter
of the phase is slightly decreased to 0,35960,0601.
336
than for those loaded perpendicular to the build

direction while the opposite trend is observed for
the elongation. The lower cooling rate during laser
cladding in the samples loaded parallel to the build
direction results in a coarser microstructure and
will contribute to a stronger, less ductile mechanical behavior. The difference in phase constitution
for both type of samples in on-going.
Upon heating to 1000 C and 1200 C followed
by furnace cooling, M23 C6 and M6 C carbides are
formed. The microstructural changes upon high
temperature exposure are accompanied by a drop
in hardness.
Further investigation of the effect of heat treatment
on the properties of laser cladded Inconel 625 is
planned.
Figure 11. X-ray diffraction data of laser cladded inconel

625 samples, in as-produced and heat treated (1 h at specified
temperature followed by furnace cooling) conditions.
After heating to 1000 C and 1200 C the presence of

secondary M6 C and M23 C6 carbides is revealed. The
former carbides can typically contain Mo, Ni, Cr, Nb,
Si, and Fe while the latter can contain Cr and some
minor amounts of Fe, Ni and Mo (Patel et al. (2001)).
The formation of M23 C6 carbides during aging of a
rapid solidified Inconel 625 microstructure is also
observed by Dey et al. (1989). Decomposition of MC
to M23 C6 and to M6 C after high temperature exposure is known to occur in superalloys (Shankar et al.
(2001)).The effect of MC degeneration on the mechanical properties of superalloys is debated in literature
(Qin et al. (2008)). The formation of secondary carbides is detrimental to corrosion resistance due to the
formation of a depletion zone (Davis (2000)). The lattice parameter of the phase is not largely affected: it
is 0,35970,3605 and 0,35990,3611 after heating to
1000 C and 1200 C respectively.
4
CONCLUSIONS
Three-dimensional, dense parts without defects,

apart from some small spherical gas inclusions, are
produced by laser cladding of Inconel 625 powder.
X-Ray diffraction reveals the presence of a Ni-based
phase and primary MC carbides after laser
cladding.
The laser cladded samples have a higher tensile strength than conventionally produced (cast
or annealed) Inconel 625. The elongation is in
the range of 2040%, which is lower than for
conventional produced and annealed Inconel 625.
The yield and ultimate tensile strengths for samples loaded parallel to the build direction are lower
REFERENCES
Cieslak, M.J., Headley, T.J., Kollie, T., Romig, A.D. 1988. A
melting and solidification study of Alloy 625, Metallurgical Transactions 19A, 2319.
Davis, J.R. 2000. Nickel, cobalt and their alloys. ASM
International.
Dey, G.K., Albert, S., Srivastava, D., Sundararaman, M.,
Mukhopapahyay, P. 1989. Microstructural studies on
rapidly solidified inconel 625. Materials Science and
Engineering A119, 175184.
Dinda, G.P., Dasgupta, A.K., Mazumder, J. 2009. Laser
aided direct metal deposition of Inconel 625 superalloy:
Microstructural evolution and thermal stability. Materials
Science and Engineering A 509: 98104.
Formenti, A., Eliasson, A., Mitchell, A., Frederiksson, H.
2005, Solidification sequence and carbide precipitation
in Ni-base superalloys IN718, IN625 and IN939. High
temperature Materials and Processes 24(4) 239258.
Paul, C.P., Ganesh, P., Mishra, S.K., Bhargava, P., Negi, J.,
Nath, A.K. 2007. Investigating laser rapid manufacturing
for Inconel 625 components. Optics and Laser Technology
39: 800805.
Qin, X.Z., Guo, J.T., Yuan, C., Chen, C.L., Hou, J.S.,
Ye, H.Q. 2008. Decomposition of primary MC carbide
and its effects on the fracture behaviors of a cast Ni-base
superalloy. Materials Science and Engineering A485,
7479.
Rai, S.K., Kumar, A., Shankar, V., Jayakumar, T., Rao, K.B.S.,
Rai, B. 2004. Characterization of microstructures in
Inconel 625 using X-ray diffraction peak broadening and
lattice parameter measurements. Scripta Materialia 51,
5963.
Shankar, V. Bhanu Sankara Rao, K., Mannan, S.L. 2001.
Microstructure and mechanical properties of Inconel 625
superalloy. Journal of Nuclear Materials 228: 222232.
337

A preliminary investigation on Selective Laser Melting of M2 high speed

steel
Z.H. Liu, C.K. Chua & K.F. Leong
K. Kempen, L. Thijs, E. Yasa, J. Van-Humbeeck & J.P. Kruth

Katholieke Universiteit Leuven, Belgium
ABSTRACT: Selective Laser Melting (SLM) is an Additive Manufacturing (AM) technique that is able to
process both metallic and ceramic materials in powder form. The main attraction of SLM is the ability to
produce near fully dense functional parts with high geometrical complexity. Parts can be fabricated with features
and designs conventional manufacturing methods cannot achieve. In this study, the feasibility of processing M2
High Speed Steel (HSS) with the SLM technique is investigated. The main focus is on understanding the material
properties to facilitate the optimization of process parameters in order to produce parts with good quality and
high density. Laser powers of 90 W and 105 W were used with scan speeds ranging from 100 mm/s to 850 mm/s.
Preliminary results show that parts exhibit warpage, cracking and partial separation from the base plate. In
addition, the degree of cracking and base plate separation is more pronounced at lower scan speeds. While
residual stresses mainly accounted for the problems occurred in this study, in depth microstructural analysis
might also explain the cracking at lower scan speeds. It was found that the microstructure was different for parts
produced at different scan speeds, providing some insights on suitable processing parameters for the family
of tool steels. In an attempt to reduce residual stresses, lower thermal gradient was achieved by preheating of
the base plate to 180 C. Indeed, cracking, warpage and base plate separation were greatly reduced upon close
visual inspection. It was also found that the chances of parts separating from the base plate were lower for AISI
1085 steel than for M2 HSS base plates. In conclusion, this study showed that SLM of M2 HSS is feasible with
preheating conditions.
1
1.1
INTRODUCTION
Background
SLM is a promising technique in the field of AM

(Campbell 2009) as it can process relatively a wide
range of materials ranging from metallic to ceramic
powders compared to other AM techniques. The ability and flexibility to produce parts without the need to
include fabrication and manufacturing post processes
such as milling and turning also give rise to its popularity for many different applications in areas such as
biomedical, automotive and aerospace (Kruth, et al.
2005; Wehmoller, et al. 2005; Mumtaz, et al. 2007;
Shishkovsky, et al. 2007; Vandenbroucke, et al. 2007;
Chua, et al. 2010).
Some commercialized materials that have proven
to work with SLM include Titanium alloys (TiAl6V4 ,
TiAl6 Nb7 and cp-Titanium), Aluminum alloys (AlSi10
Mg, AlSi12 and 6/7000 series), Stainless Steel
(316L and a few other series), Inconel (718 and
625), and Cobalt Chrome (3DSystems 2011; ConceptLaser 2011; EOS 2011; Louvis, et al. 2011;
MTTTechnologies 2011; PhenixSystems 2011). For
ceramics (Bertrand, et al. 2007), examples include

Alumina (Shishkovsky, et al. 2007), Alumina
Zirconium Zirconia and YttriaZirconia (Mumtaz, et
al. 2007). The list of SLM materials continues to grow
today as researchers explore into different materials
such as Gold (Khan, et al. 2010), Copper and Magnesium alloys (Nga, et al. 2010; Nga, et al. 2011). The
range of applications is therefore expected to increase
with the growth of SLM materials list.
1.2 Objectives
The main objective of this study is to investigate
SLM of a hardenable carbon steel, M2 HSS. In this
research, different SLM process parameters are investigated in an attempt to produce good quality parts
with high density and retaining properties comparable
to conventionally produced M2 HSS. Microstructural
analysis and microhardness tests are also carried out
to understand how different process parameters affect
the resulting microstructure of the SLM M2 HSS
part. This understanding will help to facilitate future
studies on processing other tool steels with the SLM
339
Figure 2. SEM of virgin M2 HSS powder.

Figure 1. Cumulative frequency distribution of M2 HSS
particles.
Table 1. Elemental composition of M2 HSS powder provided by LPW and compared with conventional M2 HSS
composition.
LPW
Conventional
Element
Carbon, C
Silicon, Si
Manganese, Mn
Chromium, Cr
Molybdenum, Mo
Vanadium, V
Tungsten, W
Iron, Fe
0.9
0.35
0.38
3.97
4.89
1.82
6.15
Balance
0.780.88, 0.951.05
0.200.45
0.150.40
3.754.50
4.505.50
1.752.20
5.506.75
Balance
technique which is in line with the objective of this

research.
2
2.1
EXPERIMENTAL
Material
The M2 HSS powder used in this research was bought

from LPW Technology (UK) and is produced by the
gas atomization process. Particle size is between 15
to 45 m as stated in the data sheet provided by
LPW. Tests carried out with LEICA QWin software
algorithm based on pixel counts from Light Optical
Microscope (LOM) pictures resulted in an average
grain size of 27.7 m. The cumulative frequency distribution of the powder is shown in Figure 1. Table
1 shows the composition of the M2 HSS in weight
percentage provided by LPW Technology compared
with the typical M2 HSS composition.
A SEM picture of the powder shown in Figure 2
reveals that the particles are spherical in shape as
expected from the gas atomization process. This will
provide excellent powder flowability that is necessary
for an ideal powder layer deposition, ensuring good
part density and quality.
2.2
Procedure
2.2.1 SLM machine and process parameters

The SLM machine used for this experiment is M3
Linear, from Concept Laser. Two laser powers of 90 W
Figure 3. Parts showing crackings, base plate separation

and warpage.
and 105 W were used with varying laser scan speeds

from 250 mm/s to 700 mm/s in a nitrogen atmosphere
where remnant oxygen level was kept less than 1%.
For some SLM parts, laser remelting of each layer was
applied to observe any improvements in surface roughness and part density. (Kruth, et al. 2008; Yasa, et al.
2009). It was found that better surface roughness and
part densities were achieved when applying laser remelting to AISI 316L Stainless Steel (SS) SLM parts.
In the current research of M2 HSS, the laser re-melting
was carried out with laser power 105 watt and scan
speed 700 mm/s. As for scan spacings, values of a1 =
62% (scan spacing between scan tracks), a2 = 35%
(specifying the distance of the first scan track to the
island border) and a3 = 50% (elongation of the scan
tracks in one island) of the spot size were set. These
scan spacings were optimized for 316L SS. Detailed
information on scan spacing factors for the Concept
Laser M3 Linear in island scanning for SLM is well
documented in (Yasa, et al. 2010).
2.2.2 Problems encountered
There were some problems encountered during the
processing of this material. The parts exhibited
warpage, crackings and base plate separation. Furthermore, the degree of warpage, crackings and base plate
separation was more pronounced in parts fabricated
with low scan speeds. Figure 3 shows examples of
cracking, warpage and base plate separation encountered. These problems seemed to suggest that high
thermal stresses were induced during the process.
Moreover, its high carbon content of 0.9% also results
in high brittleness compared to AISI 316L SS and
maraging 300 steel which have a carbon content of
0.03%.
340
Table 2. Comparison of part quality at 25 C and 180 C

preheating, based on visual inspection.
Process
25 C
Parameters
(On a scale of 0 to 10, good to bad)
90 Watt, 250 mm/s

90 Watt, 400 mm/s
90 Watt, 550 mm/s
105 Watt, 250 mm/s
105 Watt, 400 mm/s
105 Watt, 550 mm/s
105 Watt, 700 mm/s
*105 Watt, 700 mm/s
5
5
2
8
7
5
10
10
180 C
1
0.5
0
1
0
0
0
0
*With laser re-melting.

Figure 5. Preliminary density results.
Figure 4. Parts produced with (right) and without (left)

preheating.
2.2.3 Pre-heating
In the literature on laser welding, it is mentioned
that pre-heating is necessary for high carbon (above
0.6% carbon content) steels (Ion 2005). High amounts
of carbon are trapped within a body-centered cubic
(BCC) crystalline structure during very fast cooling rates forming matensitic structure (body-centred
tetragonal BCT), containing tremendous amount of
internal stresses (Totten, et al. 2002). As such, preheating is typically applied to reduce the tendencies
of thermal shock and residual stresses (Dawes 1992;
1997). In view of this, SLM of M2 HSS adopted preheating of the base plate before the process. The base
plate was heated to 100 C and 180 C in two separate experiments. It was clear that preheating of 180 C
reduced the cracking, warpage and base plate separation issues greatly. Table 2 shows the ratings of part
quality without (25 C) and with preheating (180 C)
based on visual inspection.
It is observed that preheating improves the part
quality by reducing the thermal stresses. Figure 4
shows the quality of parts produced with and without
preheating.
3
Figure 6. (a) Aligned porosities; (b) scan track irregularity

and discontinuity; (c) stable and regular scan track.
550 mm/s. While a scan speed of 250 mm/s might

yield higher density parts, they possessed cracks all
the time. Figure 5 shows the preliminary density
results obtained. Scan speeds above 550 mm/s without
re-melting produce parts with densities below 91%.
Under LOM, aligned porosity (Figure 6a) can
be observed on parts produced with scan speed of
700 mm/s. This suggests that the scan spacing values
that were set are not suitable at higher scan speeds
because of a smaller scan track width, higher scan track
irregularity and discontinuity which resulted from a
lower energy input. As shown in Figure 6b, the width
of scan tracks decreased with increasing scan speed.
This also explains the obtained low density of 91%.
RESULTS
3.2 X-Ray Diffraction (XRD)
3.1 Density
Using Archimedes principle (the absolute density
value of M2 HSS used is 8.15 g/cm3 ), it is found that
good quality parts (without cracks) of density between
96% to 97% can be achieved with laser power of
105 W and laser scan speeds between 400 mm/s and
XRD is carried out for three different M2 HSS parts:

(1) as cast M2 HSS without heat treatment; (2) SLM
produced M2 HSS part and, (3) as supplied M2 HSS
virgin powder. These are shown in Figures 7a, 7b and
7c respectively. All 3 graphs look similar under comparison and exhibit the FCC and the BCC phase peaks.
341
Figure 8. As cast M2 HSS part with controlled cooling rates

(Zhou, et al. 2010).
Figure 7. XRD graphs of (a) as cast M2 HSS part without

heat treatment; (b) M2 HSS SLM part; (c) as supplied M2
HSS virgin powder from LPW.
It can be seen that the bases of the peaks in Figure 7b

are comparatively wider than in Figures 7a and 7c.
This suggests that the crystal lattice structures of SLM
parts are experiencing certain level of internal stresses
that are thermally induced during the rapid solidification SLM process. There is also a shift of the peaks
between the as cast and SLM M2 HSS to lower angles
suggesting the presence of tensile stresses.
Peaks corresponding to expected carbides, M23 C6
and M2 C (where M represents V, Cr, Mo or W) in
SLM parts are also missing when compared to Figure 8. These are the carbides normally found in M2
HSS parts from similar processes such as laser surface melting (Kac, et al. 2003; Arias, et al. 2010; Arias,
et al. 2010). Other types of carbides (MC and M6 C)
usually found in annealed and conventionally hardened (Kim, et al. 1979) M2 HSS parts, were also not
found from three of the XRD tests. Preliminary XRD
results also suggested that M2 HSS SLM parts contain
80% martensitic phase and 20% austenitic phase after
Rietveld Refinement analysis.
with different scan speed exhibit different percentages

of dark and bright phases as compared in Figures 9b
and 9c. The percentage bright phase increases with
increasing scan speed which can be seen in Figure 9c
where the microstructure is dominated by the white
phases. This corresponds to the increasing speed of
solidification.
Upon close magnification under LOM, bright
phases revealed a needle like martensitic structure
(Figure 9d) whereas dark phases revealed a dendritic
(Figure 9e) and cellular (Figure 9f) like structure with
growth direction towards the melt pool center.
At the melt pool boundary, the region shown in Figure 10 is supposed to be the remelted fusion zone.
In this fusion zone, solidification of the melt material
proceeded spontaneously by the epitaxial growth of
remelted grains on the previously solidified material.
This solidification morphology is influenced by the
cooling rate, thermal gradient and growth velocity as
described in welding (David, et al. 2003).
The M2 HSS SLM parts were also subjected to
Groesbeck reagent to etch for carbides. Under LOM,
dark spots (Figure 11) that are suspected to be carbides
can be seen along the cellular and dendritic structure
which cannot be detected by XRD or SEM-EDS.
3.4 Electron backscatter diffraction (EBSD)

In EBSD tests, the dark phases can be indexed. However, the bright phases cannot be indexed and it remains
as an unknown phase from the EBSD results. Figure 12
shows the indexed results from a M2 HSS SLM part.
The white areas correspond to dark phases with high
confidence index (CI) with values ranging from 0.1 to
0.6. The black areas are with low CI. The phase composition was then analyzed with OIM Analysis software
version 5.31. The results show that within the dark
phase, 45%, 53% and 1% were martensite, ferrite and
austenite respectively.
3.3 Microstructure analysis

M2 HSS SLM parts (in the y-z plane as illustrated
in Figure 9a) were polished till 1 m and etched with
Vilellas reagent for 10 s to review the microstructure.
Under LOM (Figure 9b), it can be observed that two
distinct phases, a dark and a bright phase are present
within every melt pool. Furthermore, parts produced
3.5 Coulometry test

Coulometry tests were carried out to determine the
percentage of carbon content in the M2 HSS after
the SLM process. The results show that the carbon
content is lower in SLM parts produced with a lower
scan speed. Also, the SLM parts show a lower carbon
342
Figure 10. Fusion zones between melt pools showing

epitaxial grain growth along the boundaries. Dark phase consisted mainly martensite and ferrite whereas bright phase
consisted mainly martensite.
Figure 11. Suspected carbides observed from LOM.
Figure 9. (a) y-z plane illustration; (b) Dark and bright

phase microstructure; (c) Dominant bright phase microstructure; (d) Needle-like martensitic structure found in the bright
phase; (e) Dendritic structure found in dark phase; (f) Cellular
like structure found in dark phase.
content as compared to the as supplied virgin powder. A supposedly 0.9% C-content in M2 HSS was
only 0.7% in the supplied powder and as low as 0.15%
in the SLM part. The resolution of this technique is
0.1 ppm, (=0.0001 wt%C) and the reproducibility is
about 0.5% relatively.
3.6
Microhardness test
Microhardness tests were carried out on the M2 HSS

SLM parts and the results are shown in Figure 14.
A load of 100 g was used (Shi, et al. 1995; Niu,
et al. 2000; Benyounis, et al. 2009). The results are
compared between SLM parts produced with different
scanning speeds and they do not show any significant difference. However, measurements taken from
Figure 12. EBSD results showing dark phases (in LOM)

being indexed with high confidence level.
the dark and bright phase show a significant difference. As shown in Figure 14b, average hardness for the
dark and bright phases was about 700 HV and 870 HV
respectively.
343
the high heat energy input resulting from low scan

speeds have more effect than the C-content on inducing
cracks.
From the XRD results, no peaks responsible for the
carbides can be found in the SLM parts. Coulometry
test results also suggested at the carbon content after
the SLM process had decreased to as low as 0.15%
from 0.7%. However, under LOM observations, black
spots that fit the usual carbides segregation description along cellular structures are spotted. These spots
are suspected to be the carbides. There are two reasons
why the XRD could not detect carbides. First, the carbon contents are very low and the amount of carbides
formed is low. Secondly, due to the quick solidification nature of the SLM process, the dendrites and cells
are small and from Figure 10, the sizes of the black
spots are extremely small, in the range of 0.1 m to
0.5 m. As such, carbide peaks may be too weak to be
detected compared to the inherent background noises
from the XRD.
The microstructure of M2 HSS SLM parts mainly
comprises of dark and bright phases, indicating the
melt pools that arose from the SLM building and scanning strategy. From EBSD results, the dark phases
consist of mostly ferrite and martensite which explain
the average hardness value of 700 HV. While the bright
phases could not be indexed, LOM and microhardness
tests suggest that the bright phases are martensitic as
needle-like martensite and high hardness of 800 HV
is observed and measured respectively. Moreover, the
bright phases were in the melt pool locations where the
highest cooling rate (which would result in martensite
formation) would occur.
Laser re-melting of M2 HSS parts did improve
the density significantly by about 3%. Also, microstructure and hardness values of laser re-melted
parts remained largely similar. Additionally, surface
remelting consistently improved the surface quality by about 2 m and 3 m for Pa (unfiltered
2D roughness test) and Sa (3D roughness test)
respectively.
Figure 13. Coulometry test results of M2 HSS.
5
Figure 14. (a) Hardness values of parts produced with different scanning speeds; (b) Hardness values of dark and
bright phase.
DISCUSSION
In this preliminary investigation, it is observed that

cracks are very easily induced in M2 HSS SLM parts.
This is probably because M2 HSS is a brittle material
owing to its high carbon content compared relatively to
AISI 316L SS. The coulumetry results show that parts
produced with lower scan speeds have lower carbon
content and yet these parts are more prone to crack.
This means that the thermal stresses created during
CONCLUSION AND FUTURE WORK
5.1 Conclusion
In conclusion, the SLM of M2 HSS has been proven
to be feasible. While base plate separation cannot be
fully eliminated in this preliminary investigation, parts
without cracks can be produced with an achievable
density of 97%. Also in this research, preheating of
the base plate was found to have a positive effect on
the the parts, minimising cracks and warpage. Furthermore, parts produced directly from the SLM process
have high average hardness values between 800 HV to
900 HV which are comparable to conventionally produced cast parts. These conventionally produced M2
HSS cast parts have hardness values between 750 HV
to 830 HV depending on the heat treatment process
(Davis 1995).
344
5.2
Future work
The future work aims to address some shortcomings in this research. Firstly, laser re-melting process
will be investigated to further improve density values
and prevent base plate separation. Currently, surface
remelting scanning speed stands at only 700 mm/s.
Different scanning speeds can be examined to improve
the densities of parts.
Secondly, it was found that preheating of the base
plate to 180 C produced better parts (less prone to
cracks) compared to a pre-heating of 100 C as the
thermal gradient was greatly reduced, thereby inducing less thermal stress. Since 180 C is the preheating
limit of the heating system currently installed on the
M3 Linear, future modification (within manufacturer
specifications) will be carried out to increase the preheating temperature above 180 C. Similar to laser
welding, preheating temperatures of 200 C may be
applied to reduce the chances of cracking.
Thirdly, the evidence of carbides was not strong
enough in this paper. Future work includes SEM and
TEM with higher magnification to verify the presence
of carbides and to also identify the type of carbides
that are present. Additionally, more research can be
carried out to understand the reason in the decreasing
carbon content.
Lastly, heat treatment process for M2 HSS SLM
parts will be optimized. Microstructural studies and
material characterization will be carried out to understand the essential difference, between cast and SLM
M2 HSS parts. Material properties such as wear
resistance, hot hardness and toughness will also be
carried out.
REFERENCES
3DSystems. (2011). 3D Systems. Retrieved March 2010,
2010, from http://www.3dsystems.com/.
Arias, J., et al. (2010). Microstructural characterization of
laser surface melted AISI M2 tool steel. Journal of
Microscopy-Oxford 239(3): 184193.
Arias, J., et al. (2010). Modification of AISI M2 high
speed tool steels after laser surface melting under different operation conditions. Revista De Metalurgia 46(3),
206218.
Benyounis, K. Y., et al. (2009). Rapid solidification of M2
high-speed steel by laser melting. Materials & Design
30(3), 674678.
Bertrand, P., et al. (2007). Ceramic components manufacturing by selective laser sintering. Symposium on Laser
Synthesis and Processing of Advanced Materials held at
the E-MRS 2007 Spring Meeting, Strasbourg, France,
Elsevier Science Bv.
Campbell, I. (2009). The international journal for research
on additive manufacturing technologies and rapid product
development. Rapid Prototyping Journal 15(5), 315315.
Chua, C. K., et al. (2010). Rapid Prototyping Principles
and Applications. Singapore, World Scientific Publishing
Co. Pte. Ltd.
ConceptLaser. (2011, 16th July 2009). CONCEPT Laser
GmbH. Retrieved September, 2009, from http://www.
concept-laser.de/.
David, S., et al. (2003). Welding: Solidification and

microstructure. JOM Journal of the Minerals, Metals and
Materials Society 55(6), 1420.
Davis, J. R. (1995). Tool Materials, ASM International.
Dawes, C. (1992). Laser Welding A Practical Guide.
Cambridge, Abington Publishing.
Dawes, C. (1997). Weld Integrity and Performance. USA,
ASM International.
EOS. (2011). EOS e-Manufacturing Solutions. Retrieved
March 2010, 2010, from http://www.eos.info/en/home.
html.
Ion, J. C. (2005). Laser Processing of Engineering Materials: Principles, Procedure and Industrial Application.
Burlington, Elsevier Butterworth-Heinemann.
Kac, S., et al. (2003). SEM and TEM microstructural investigation of high-speed tool steel after laser melting.
Materials Chemistry and Physics 81(2-3), 510512.
Khan, M., et al. (2010). Selective Laser Melting (SLM) of
pure gold. Gold Bulletin 43(2), 114121.
Kim, Y.-W., et al. (1979). Laser melting and heat treatment of m2 tool steel: A microstructural characterization. Metallurgical and Materials Transactions A 10(7),
881886.
Kruth, J. P., et al. (2008). Experimental Investigation of Laser
Surface Remelting for the Improvement of Selective Laser
Melting Process. Proceedings of SFF Symposium, Austin,
TX, USA.
Kruth, J. P., et al. (2005). Digital manufacturing of biocompatible metal frameworks for complex dental prostheses
by means of SLS/SLM. Virtual Modeling and Rapid
Manufacturing - Advanced Research in Virtual and Rapid
Prototyping. P. J. Bartolo, A. J. Mateus, F. C. Batistaet al.
London, Taylor & Francis Ltd., 139145.
Louvis, E., et al. (2011). Selective laser melting of aluminium
components. Journal of Materials Processing Technology
211(2), 275284.
MTTTechnologies. (2011). MTT Technologies, Rapid
Manufacturing Technologies. Retrieved March 2010,
2010, from http://www.mtt-group.com/.
Mumtaz, K. A., et al. (2007). Laser melting functionally graded composition of Waspaloy((R)) and Zirconia powders. Journal of Materials Science 42(18),
76477656.
Nga, C. C., et al. (2011). Microstructure and mechanical
properties of selective laser melted magnesium. Applied
Surface Science 257(17), 74477454.
Nga, C. C., et al. (2010). Layer manufacturing of magnesium
and its alloy structures for future applications. Virtual and
Physical Prototyping 5(1), 1319.
Niu, H. J., et al. (2000). Microstructural evolution during laser
cladding of M2 high-speed steel. Metallurgical and Materials Transactions a-Physical Metallurgy and Materials
Science 31(10), 26152625.
PhenixSystems. (2011, 18th Sep 2009). Phenix Systems.
Retrieved 18th Sep 2009, 2009, from http://www.phenixsystems.com/phenix_en/home.htm.
Shi, G., et al. (1995). Microstructure and properties of laser
surface hardened M2 high speed steel. Acta Metallurgica
et Materialia 43(1), 217223.
Shishkovsky, I., et al. (2007). Alumina-zirconium ceramics synthesis by selective laser sintering/melting. Applied
Surface Science 254(4), 966970.
Totten, G., et al. (2002). Handbook of Residual Stress
and Deformation of Steel. Materials Park, Ohio, ASM
International.
Vandenbroucke, B., et al. (2007). Selective laser melting of
biocompatible metals for rapid manufacturing of medical
parts. Rapid Prototyping Journal 13(4), 196203.
345
Wehmoller, M., et al. (2005). Implant design and production - a new approach by selective laser melting. CARS
2005: Computer Assisted Radiology and Surgery. H.
U. Lemke, K. Inamura, K. Doi, M. W. Vannier and
A. G. Farman. Amsterdam, Elsevier Science Bv. 1281:
690695.
Yasa, E., et al. (2009). Investigation on Occurance of Elevated
Edges in Selective Laser Melting. Twentieth Annual International Solid Freeform Fabrication Symposium. Austin,
Texas, USA.
Yasa, E., et al. (2010). Investigation of Sectoral Scanning in

Selective Laser Melting. Biennial ASME Conference on
Engineering Systems, Design and Analysis (ESDA) 2010,
Istanbul, Turkey.
Zhou, X. F., et al. (2010). A Study on the Microstructure of
AISI M2 High Speed Steel Manufactured by Continuous
Casting. Advanced Materials Research 146147, 1211
1215.
346

Fatigue behavior of additive manufacturing parts. A preliminary analysis

J. Mungua, K.W. Dalgarno & R. Reid
School of Mechanical & Systems Engineering, Newcastle University, UK
ABSTRACT: The fatigue behavior of Additive Manufacturing (AM) parts has been scarcely studied in previous years. There have been recent attempts to characterize basic fatigue behavior especially of laser sintered
parts, however the different mechanisms of fatigue and the influence of various parameters such as frequency,
temperature, specimen design and rig configuration, have not been discussed for AM. As the mechanical properties of polymer based AM materials are key for their effective use, one factor that prevents their widespread
application for final products is this lack of long term fatigue information. This work presents the preliminary
results of a series of tests aimed at two different additive manufacturing processes: Fused Deposition Modeling
(FDM) and Selective Laser Sintering (SLS). Advances in the design of purpose built rigs for different forms
of fatigue testing are described along with issues encountered within different processes such as scan spacing
effects, internal hatching strategies or surface roughness. Studied methods include: tension/compression testing
on servo-hydraulic equipment, cyclic bending fatigue and Whler-like fatigue, both being developed as purpose
built fatigue rigs.
1
1.1
INTRODUCTION
Mechanisms of fatigue in polymers
The phenomenon of fatigue is generally described as

the loss of strength or other important property as
result of stressing over a period of time (Hertzberg
1980). The different processes that lead to fatigue
failure in polymers and their correlation with frequency, loading conditions, material structure and
manufacturing method should be of great interest
to scientists and engineers alike, however currently
available literature falls short in proving sufficient
information to this respect. Therefore the field of
fatigue studies for the newest Additive Manufacturing (AM) technologies represents a still fertile area for
research.
In order to design an appropriate testing protocol
there are three main elements to consider which are
intended to simulate as closely as possible the actual
service conditions (Brown 1999):
Selection of specimen geometry (flat, dumbbell,

sheet)
Definition of control variables (stress-strain)
Selection of test conditions (temperature, cyclic
frequency, loading rate)
A number of standards and testing protocols have
been developed for different forms of fatigue testing
(ASTM D4762 11, BS ISO 13003:2003, ISO 1352,
etc.) however it is recognized that most of the existing work is devoted to metals, therefore is common to
find examples of polymer testing that borrow some
of these procedures trying to adapt the general set-up
to the levels of stress, frequency, speed, temperature
and testing conditions that result more appropriate for

plastics.
Current efforts in setting up fatigue testing devices
for AM parts are presented, along with the description
of work and main findings aimed at contributing to the
state of the art of AM fatigue properties.
1.2 Previous studies on fatigue testing of polymers

Among the different testing conditions, the specific
type of load induction is the most relevant factor
for obtaining fatigue data. Main mechanisms include:
rotational bending, reciprocal bending, reciprocal
torsion and pulsating axial loads between many others. The former is one of the preferred methods for
engineers since the rotation of a symmetric cylindrical
specimen provides a uniform moment across the testing surface (Dimarogonas, 2001). The second method
on the other hand provides uniform flexural stress
along the test length. Weibull (1961) compiled a comprehensive collection of fatigue testing procedures,
mostly for metallic materials which can be adapted
to polymer testing, and most of which have remained
unchanged.
Few methods however have been specifically
designed for polymer testing; ASTM D671-71(1993)
outlines a method for flexural fatigue testing by constant amplitude force. The aim is to evaluate the material resistance to deterioration as a result of repeated
stress. Specimen shape is optimized for cantilever-type
bends and the bending stresses are applied by an unbalanced variable eccentric (figure 1). Some drawbacks
of this already withdrawn method include the need
for a special testing apparatus, with a designated
347
Figure 1. Fixed-Cantilever, Repeated-Constant-Load Type

Fatigue Testing Machine (ASTM D67171).
at 5 kg mean stress (Instron 2008) as well as limitations in speed control usually set in the interval
of 140010000 rpm. Existing standards also provide
specifications for metals including: the preferred
specimen size, surface finish, and test temperature.
Due to the versatility of this test it would be desirable
to adapt it to polymer testing, however little literature
reports on similar attempts.
A widely used method for testing fatigue behavior
of polymers is axial loading through servo-hydraulic
equipment. Such computer controlled systems are the
most common in operation today and are capable of
monitoring and controlling a desired cyclic pattern
based on: Load/Time, Strain/Time or Displacement/
Time. Most experiments performed on polymers
make use of this testing procedure as noted by past
research (Takemori 1984; Van Hooreweder et al 2010).
Some disadvantages arising from the use of axial loading in servo-hydraulic equipment are the limitations of
low frequencies (1 to 10 Hz for practical purposes) and
the need to ensure proper gripping on both ends which
becomes problematic especially for additively manufactured cylindrical samples for tension-compression
tests.
1.3 The need for further fatigue info
Figure 2. Schematic of a rotating bend fatigue machine (BS

ISO 1143:2010).
frequency of 30 Hz that requires changing the drive

unit and the spring mass system in order to adjust
the frequency required. This presents difficulties to
reproduce real service factors during the test.
Few studies are published following this method.
Berchem & Hocking (2006) developed a purpose built
multi-specimen flexural fatigue tester based on a sheet
metal test specimen with 3 different displacement levels based on shaft-mounted eccentrics. This provides
a good approximation on how to reconfigure testing
equipment around the specific fatigue specimen.
Rotating beam-bending fatigue tests provide the
advantage of completely reversed stresses on small
cylindrically shaped samples at variable frequencies to
understand the effect of different loading conditions
on the specimen. Figure 2 shows a generic diagram
of a four-point rotating fatigue tester. This test allows
observing the effect of load in the number of cycles to
failure, hence allowing plotting the now conventional
S/N graph. The main interest of developing similar
testing rigs is to find the optimum specimen shape and
stress level at which the material does not fail after any
number of revolutions.
Although the fatigue curves for metals such as steel
are known to be between 5 and 10 million cycles,
it is not possible yet to state such affirmation with
similar precision for polymers, specially AM-based.
Most of the existing rotating fatigue testers are developed mainly for metals, indeed the standard equipment
specifications prevent these machines from being used
for plastics, due to the high initial loads, some starting
Accurate fatigue life prediction of additive manufacturing parts is not currently possible, both for the
lack of relevant independent material information, and
secondly due to the lack of contrastable experimentation in the field. There are however some sources
that show a trend of interest in this subject. An early
example of high frequency fatigue testing for polymers
was outlined by Crawford et al. (1974) by testing an,
injection molded specimen for uni-axial and rotating
bending fatigue tests. Some of the observations highlight the fact that, as opposed to low frequencies for
uni-axial testing, in rotating bending the frequency can
be increased fivefold (25 Hz) and the same stress can
still be used with a relatively small temperature rise.
In the case of acetal co-polymer, at 25 Hz in rotating
bending, thermal failures only occur for stresses above
35 MN m2 . But more importantly, it was observed that
for specimens subjected to rotating bending fatigue,
temperature rises and either stabilizes at some value
until a conventional type of fatigue fracture occurs.
Bellemare et al. (2004) performed the characterization of PA6 non-filled and nano composite material
samples by means of load-controlled axial fatigue
(tension-tension) under two testing regimes. For the
high-stress regime approximately 78 MPa or 78% of
the UTS-, fracture occurred in a macroscopically ductile manner by necking. One of the few studies devoted
to AM was presented by Van Hooreweder et al. (2010)
In this study dynamic tension-compression cycles
were applied on laser sintered PA12 specimens on
a servo-hydraulic machine, at three different stress
amplitudes (23.6, 18.9 and 17.7 MPa). Results showed
that temperature rise is an issue at a frequency as
low as 3 Hz and when it is avoided the components
348
Figure 3. Four point bending schematic with the load

application and the sine wave stress pattern (Karwa, 2006).
Figure 5. Some standard sample fatigue specimen geometries.
Figure 4. Four point rotating bending rig schematic.

1) motor, 2) bearing boxes, 3) rotating specimen, 4) base plate,
5) flexible coupling or universal joint, 6) rotating support.
experience a relatively long fatigue life (>1.6 106

cycles). It is suggested by the authors that PA12
exhibits an acceptable fatigue resistance on stresses
below 20 MPa.
2
SETTING UP OF FATIGUE TESTS FOR AM
As additive manufacturing technologies become

increasingly used for the fabrication of end-use products, more information pertaining to long term behavior is needed. It is expected that more studies appear
in order to provide useful fatigue data, however test
must bear sufficient significance in relation to real
case loading. The following is a description of current efforts in setting up purpose-built fatigue rigs for
AM-base materials.
2.1
Four point Rotating Fatigue Bending (RFB)
Figures 3 and 4 show a schematic of the RFB being

developed. As opposed to cantilever-type fatigue
testers the system makes use of four point mechanism of fatigue (two support points two load bearing
points) so that the specimen is subjected to a constant
bending moment. As the beam rotates the specimen
undergoes maximum tension (when at the bottom)
and maximum compression when at the top position.
Similar rotating fatigue rigs apply stresses and the
end of the cylindrical sample in a cantilever-bending
manner however the conditions for pure bending are
not always successfully replicated by such methods.
Main features of this tester compared to similar rigs
for metal specimens include: a finer speed control by
means of a programmable ABB drive with adjustable
frequencies 150 Hz, 3-phase motor 0.37 kw, lighter
bearing housings on both ends, simpler specimen

clamping system by means of set screws, thus eliminating collets or chuck-based systems and their associated
weight. Stresses are applied by means of dead-weights.
Cycle counting is made by an inexpensive transmissive sensor (Honeywell HOA1887) with speed limit
3 kHz. A key element is the rotational motion transmission from the motor, which has been accomplished
by a needle roller universal joint capable of providing a
maximum speed of 4000 rpm and torque up to 220 Nm.
This set-up is under refinement so that unattended
testing can be performed over long time cycles.
2.1.1 Specimen design for rotating bending
The specimen geometry directly defines the mode of
breakdown for the testing procedure. Most rotating
fatigue specimen shapes are taken from typical metal
fatigue processes (ASTM E 60604, BS3518-3:1963,
etc.). The typical rotating specimen consists of two
ends of similar diameter and a middle cross section
of reduced cross section where the test conditions are
being simulated and failure is expected (Figure 5).
The determination of fatigue properties then is calculated on the basis of a four point loading to apply a
constant bending moment applied to a rotating specimen. Initial tests were performed in the range of
2530 Hz (1500/1800 RPM) in order to avoid heat
build-up and high vibration in the system. Loads are
applied by means of hanging weights. In order to
define a suitable range of loads the stress level at the
surface of the specimen is calculated using the elastic
beam equation.
where S is the nominal stress acting normal to the

cross-section, M is the bending moment, c is the distance of the surface from the neutral axis and I is the
moment of inertia. Table 1 shows sample calculation
for a given specimen design made of sintered PA12.
Currently trial test are being performed in order
to refine actual testing parameters. The aim is two
correlate this test with other forms of fatigue testing:
bending and axial tests.
349
Table 1. Test calculations for PA12 specimen.

Specimen data
Material name
E-modulus
Specimen length
Thin section diameter
Moments arm
Surface moment of inertia
Bending moment
Bending force
Units
PA12
1600
95
8
25
201.06
0.080
6.433
[MPa]
[mm]
[mm]
[mm]
[mm4]
N.m
N
Figure 7. a) Specimen geometry and dimensions b) test

samples made in FDM.
Figure 6. Bending principle for alternating bending
fatigue rig.
2.2 Alternating bending rig

As opposed to the cantilever and 4-point bending
machine types this rig is intended for the deflection
controlled fatigue testing of flat specimens (figure 6).
The main purpose being to characterize materials
whose mode of failure during normal operation is
more related to flat-type geometries with thicknesses
between 14 mm.
The basic principle follows specifications given by
ASTM D671-71 for frequency and deflection ranges,
however instead of using an eccentric wheel, motion
is transmitted by a crank-slider mechanism, where
the crank is a rotating disk with 3 different possible
eccentric distances 15 mm, 25 mm and 35 mm which
respond to the expected maximum deflection levels
for a number of AM materials whose deflection is
given by the cantilever beam bending equation:
where d is the displacement, P is the Load at point

x = Lo, Lo is the beam length, E is Youngs modulus and I is the moment of inertia of the rectangular
cross section. Then the maximum stress is related to
the specimen geometry by equation 2:
where _max is the maximum stress, o the corresponding deflection and H the specimen thickness.
2.2.1 Specimen design
Overall specimen geometry follows the recommendation ofASTM D671-71 where the design of a particular
specimen depends upon specimen thickness, and the
stress range of measurements to be made (figure 7).
Berchem & Hocking (2006) provide an example of
specimen modification following the original triangular form which provides uniform stress distribution
over the test span.
Recommendations of surface finish for standard

AM parts are not possible to fulfill since current standards are oriented to bulk materials machined from
stock, thermoformed or injection molded, therefore the
test response and accuracy will be conditioned by the
selected AM process and its roughness levels.
Table 2 shows sample design information regarding the specimen design and testing rig parameters
based on basic material data. From these calculations it will be possible to select different levels of
stress/displacement to be applied to the specimen.
2.2.2 Test rig main components
Alternating bending motion is achieved by a crankslider system as stated above. The Alpha testing rig
makes use of clevis eye rods and studding working
as slider, connected to a horizontal bar that travels
vertically along two rails on both ends (figure 8). Specimens are clamped to the moving bar and fixed to
a support bar on the wider section. The system has
to be adjusted according to the level of stress being
tested. As for the rotating fatigue rig a 3-Phase 0.37 kw
motor has been selected due to the nature of materials
being tested as the rated torque 47 Nm is considered
sufficient; however in case of migrating to higher maximum stresses or multiple specimen rig, the system
is upgradeable to motors providing higher torques.
A transmissive sensor is used for measuring linear
displacement along with a digital counter (Kubler type
130). Considerations for calibration and setting up
tests will be developed after further testing.
2.3 Axial load testing
The main interest of using a servo-hydraulic equipment for fatigue testing is due to the fact that most tests
identified in literature have been performed following
this procedure.
An Instron 8511 was used for cyclic axial loading. This machine is normally used for dynamic and
fatigue tests for different applications: biomaterials,
350
Table 2.
Speciem geometry design for flexural bending testing rig.

Specimen design
Material
Moulded
PA
EOS Sintered PA12
Duraform
PA (Z)
Duraform
PA (XY)
FDM ABS
Moulded
ABS
Testing rig parameters
Material data
Length
Thickness
Max
deflection
Repeated stress MPa

(provided by fatigue rig)
Max stress MPa

(from equations)
Flexural
Modulus GPa
65
2.5
15
568.93
607
64.1
65
35
41.1
45
1.24
65
35
65.6
53.01
1.6
65
35
56.33
74.9
1.7
65
65
3
4
25
25
24.35
62.01
26.9
66
1.372
2.62
Figure 8. Alternating bending rig schematic and main components: 1) 3-phase motor 14003000 rpm, 2) specimen
support, 3) flat specimens & clamping bar, 4) slider (clevis
pins), 5) crank wheel.
Figure 9. Instron 8511 servo-hydraulic fatigue equipment.
implants, composites, rubbers, elastomers. In order to

benchmark the test being made with existing literature, specimens are being designed according to ISO
1352 (rotating bar bending fatigue testing). The rationale behind this selection is to apply the same type
of specimen for both, axial and rotating tests. Forces
are measured using the Instron 25 KN 2525 Series
Drop-through Load Cell (Instron, 1990).
Figure 10. Instron hydraulic wedge grips.
2.3.1 Preliminary results

Currently batches of 5 specimens are being tested
in order to validate the method. Efforts are being
addressed to identify the appropriate conditions for
layered manufactured parts to be tested under axial
loading, which has proved to be difficult for some test
regimens. Tests started by Reid (2011) identified unexpected inconsistencies for both, the behavior of the
equipment during testing (continuous stops and scatter data on dynamic plots) and behavior of the test
specimen mainly at the clamping area (deformations
due to clamping stress and misalignment due to specimen poor verticality). Grips being used (figure 10)
are the general purpose hydraulic-wedge grips for
fatigue testing (Instron 2011) which are capable of
accommodating either flat or round specimens for
static, dynamic, reverse stresses, etc.
During preliminary tests it was noted that if the
cylindrical specimen is lightly tightened in order to
avoid extra stresses, movement of the specimen relative to the grips occurs leading to a failed test as
the load cell senses the lack of gripping as a failed
specimen or early rupture. Main problems were identified when testing cylindrical samples on compression
or tension/compression regimes, where the servohydraulic grips exhibit irregular clamping pressures on
the specimen.One strategy applied to improve clamping accuracy was to lower the test frequency from
3 Hz to 2 Hz, which mitigated most of the clamping
351
Figure 11. S-N fatigue curve for FDM ABS, semilogarithmic scale (Reid, 2011).
issues; however for practical purposes it is desirable

to keep frequencies as high as possible without inducing significant heat-build up. Alternatives for gripping
layered manufactured specimens include the replacement of current grips for instance by using clevis
grips as used in fracture mechanics, or by redesigning
the specimen for optimal fitting. Reinforcement of the
samples clamping area by means of resin, adding a
brass shim or diameter reduction was ruled out since
this would add extra steps to the preparation of the sample and it was preferred to test a simple material piece.
Reid (2011) reported the first S/N graph for ABS
material produced with the FDM method in a Dimension Print machine (figure 11), however in order to
gain sufficient validity more samples are being tested
at different loading conditions.
Significant variability is found in specimen accuracy and roughness. It is noted that the low cost range
of FDM equipment tends to deliver poorer mechanical
properties as they are intended for concept modeling,
therefore at the moment there are not S/N fatigue curve
representative of the FDM technology. The same can
be stated for the rest of AM technologies.
2.3.2
Observations during uni-axial tension

compression tests of ABS cylindrical
specimen
Test have continued for ABS samples built with the
FDM process under tension/compression regime at
variable stress levels and frequency kept fixed at 2 Hz.
It is possible for most tests to observe a period of
adjustment for the load being applied and the displacement of the specimen which normalizes after 100
cycles. These tests correspond to the upper left part of
an S/N graph where specimens are meant to fracture
after a reduced number of cycles. At the moment
reproducibility has been achieved for stress up to
.17 kN at 2 Hz with specimens reaching the point of
rupture after 5000 cycles (figure 12).
Fracture of FDM samples being analyzed shows a
tendency to occur in the lower density areas coinciding with thin diameter sections (figure 13). The lack
of bonding and abundance of air gaps even for solid
specimens seems to reveal poor resistance for repeated
axial loading for specimens built in the XY direction.
Figure 12. Reproducibility of tension/compression test of

FDM samples at 2 Hz and 0.17 kN load.
Small voids present in the cross section of specimens have a direct relationship with density thus
reducing the overall strength of the sample and inducing fatigue rupture in areas with big air gap concentration. It is therefore reasonable to consider as control
parameters for fatigue testing the internal hatching
strategy, deposition width and orientation of FDM
samples. However it is expected to observe similar
behavior in other AM technologies.
An additional control parameter that must be considered for testing procedures is the specimen roughness, as this is a highly variable parameter dependant
on orientation, speed, temperature and other processing factors. It is necessary to establish a standard
352
Figure 15. Typical outer surface of FDM cylindrical samples and rupture area.
Figure 13. FDM samples for uni-axial tension/compression

fatigue test and their fracture surfaces.
Figure 14. Typical fracture faces for FDM samples where

air gaps and low density are evident indicators of tendency
to rupture.
post-processing procedure that provides adequate minimum roughness levels prior to testing. Sanding of
specimens with low grade (No 00) sand paper is recommended by testing standards (ASTM D671-71)
however due to the difficulties for sanding AM parts
other possibility is to determine an acceptable level of
coating to bring more stability to the test.
Additional testing sets continue to be tested however maximum frequency is limited up to 3 Hz. The
aim is to build a reliable fatigue curve, therefore at the
moment no conclusions can be drawn about the significance of preliminary fatigue data.
3
CONCLUSIONS
This work has presented initial efforts to build a

wide base of testing procedures for the characterization of the fatigue behavior of layered manufactured
materials.
The fatigue studies carried out as part of a wider
study to characterize the behavior of AM samples are
made from the approach of engineering design rather
than materials science. That is, the aim of this research
is to find the applicability of different modes of fatigue
testing for a specific scenario. One of the areas of
interest for the applicability of these procedures is the
fatigue life prediction of materials that are suitable
for the fabrication of foot and ankle foot orthotics. As

conventional materials currently used for this purpose
include PP, HDPE, CF, and other composites formulation it is important to understand the limitations of
alternative options offered by AM methods.
In order to achieve an appropriate correlation with
the final use it is important to identify the actual service
conditions and the levels of stress involved. Since most
of the existing standards and guidelines are intended
for composites and thermosets, research efforts are
also conducted in finding the optimum test design,
specimen configuration that provide a more useful
overall significance.
Specimen geometry across different procedures for
fatigue testing of AM must be revisited prior to establishing testing conditions and drawing conclusions
from such tests. State of the art makes use of unaltered
geometries taken from composite material testing, or
borrowed from metal testing procedures however need
for adapting such test is evident.
Establishing a desired optimal density and surface
roughness is critical for the test to be valid. Density
measurement procedures are equally prone to be
assessed mainly due to the highly porous nature of
layered manufactured parts. It has been observed that
crack initiation is highly influenced by changes in the
internal structure from hollow to pattern filled to solid
samples.
The generation of S/N curves usually starts with
higher load regimes in order to populate the higher
limits of fatigue graphs. At the moment no complete
S/N curves have been generated as scatter data from
testing variability is an issue.
Finally, lifetime estimation by means of existing fatigue information of AM materials should be
exercised with caution, as most methods tend to
analyze the phenomenon of fatigue from a generic
perspective without defining the initial service conditions to be assessed. Therefore it is desirable to
state an end-use application in order to propose
adequate testing procedures and generating relevant
information.
ACKNOWLEDGEMENTS
This work was funded through the European Commission Framework Seven Program (grant number
NMP2-SE-2009-228893) as part of the A-Footprint
project (www.afootprint.eu).
353
REFERENCES
ASTM D4762 11 2008 Standard Guide for Testing Polymer
Matrix Composite Materials.
ASTM D671-71 (1993) Standard Test Method for Flexural
Fatigue of Plastics by Constant-Amplitude-of-Force.
ASTM E 60604 Standard Practice for Strain-Controlled
Fatigue Testing.
Berchem, Hocking, (2006) A simple plane bending fatigue
and corrosion fatigue testing machine. Measurement
Science and Technology 17(10).
BS ISO 1143:2010 Metallic materials Rotating bar bending fatigue testing.
BS ISO 13003:2003 Fibre-reinforced plastics Determination of fatigue properties under cyclic loading conditions.
Dimarogonas, 2001 MAchine design. A CAD approach.
ISBN: 978-0-471-31528-5.
Hertzberg, RW & Manson, J.A 1980 Fatigue of engineering
plastics. Academic Press (New York) ISBN 0123435501.
Honeywell 2010 HOA1887Transmissive sensor instructions
sheet.
Instron 2008.SATECTM Series/Industrial Products Group.
000058-02-0604-EN Operating Instructions.
Instron 2011 Hydraulic Wedge Grips for Fatigue Testing.
(http://www.instron.co.uk/wa/acc_catalog) last visited
may 2011.
Instron, 1990, User reference Guide, Instron, Buckinghamshire.

ISO 1352 1977. Steel Torsional stress fatigue testing.
Karwa (2006). A Text Book of Machine Design. Laxmi
Publications ISBN 9788170088332.
Crawford, R. D. Benham, P. P. (1974). Journal of Materials
Science 9 12971304.
Brown, R.P. (1999). Handbook of Polymer Testing Marcel
Decker. ISBN 0-8247-0171-2.
Reid, (2011) Fatigue of Fused deposition modelled (FDM)
Acrylonitrile butadiene styrene (ABS). Stage three individual project MEC3098.Newcastle University School of
Mechanical and Systems Engineering.
Simon, C., Bellemare, Martin N., Bureau*, Johanne Denault,
and J. Ivan Dickson. (2004). Polymer Composites 25(4).
Takemori (1984) Polymer Fatigue. Ann. Rev. Mater. Sci. 14,
171204.
Van Hooreweder, B., De Coninck, F., Moens, D., Boonen,
R., Sas, P. (2010). Microstructural characterization of
SLS-PA12 specimens under dynamic tension/compression
excitation. Polymer Testing 29 (3), 319.
Weibull (1961). Fatigue testing and analysis of results. Advisory Group for Aeronautical Research and Development,
North Atlantic Treaty Organization, by Pergamon Press,
1961 Technology & Engineering 305 pages.
354

Characterisation of carbon fibre reinforced nylon-12 composites for

selective laser sintering process
Chunze Yan
College of Engineering, Mathematics and Physical Sciences, University of Exeter, Exeter EX4 4QF,
Devon, United Kingdom
Lin Xu
State key Laboratory of Materials Processing and Die & Mould Technology, Huazhong University of Science and
Technology. Wuhan, Hubei, China
Liang Hao
College of Engineering, Mathematics and Physical Sciences, University of Exeter, Exeter EX4 4QF,
Yusheng Shi
State key Laboratory of Materials Processing and Die & Mould Technology, Huazhong University of Science and
Technology. Wuhan, Hubei, China
ABSTRACT: The particle sizes and particle size distributions, thermal stability and impact strengths of three
kinds of CF/PA composites with the fibre weight content of 30%, 40% and 50% for selective laser sintering (SLS)
were investigated. The results reveal that the CF/PA composite powders have a wide particle size distribution of
0250 m. The volume average particle sizes of 30% CF/PA, 40% CF/PA and 50% CF/PA composite powders
are 51.0 m, 67.4 m and 68.5 m, respectively, which are suitable for the SLS process. Compared with the pure
PA-12, the CF/PA compo-sites have higher maximum decomposition temperatures to provide higher thermal
stability, which extends the applications of polymer SLS parts to some high temperature services. The impact
strength of CF/PA SLS parts decreases with increasing the carbon fibre content from 0 to 50 wt%, due to the
incorporation of brittle fillers.
INTRODUCTION
Selective laser sintering (SLS) is one of additive

manufacture (AM) technologies, which canmanufacture three-dimensional components according to a
computer aided design (CAD) model by selectively
sintering successive layers of powdered materials
(Kruth, et al. 2003). Polymers are widely used in
SLS process. However, only several thermoplastics
such as polycarbonate (PC), polystyrene (PS), and
nylon (PA) are proved to be suitable for SLS process.
Obviously, these materials can hardly meet the requirements of various applications. Many efforts have been
made to prepare the composites of these thermoplastics to enhance some properties of their SLS parts.
Nowadays, some polymer-based composite powders
have been developed for SLS materials. Chuang and
Das (2006)investigated the fabrication of functionally
graded materials (FGMs) by SLS of Nylon-11 composites filled with different volume fractions of glass
beads. Ho, et al. (2002) studied the SLS of PC/graphite
blended composites, and found that graphite powder
could improve the laser absorptance of the powder
system for SLS for graphite can absorb CO2 laser

more readily. Hao, et al. (2007) used HA to improve
the bioactivity of the polymer matrices for medical
applications, and formed porous structures by the SLS
of HA/high-density polyethylene (HDPE). Kim and
Creasy (2004) investigated the SLS characteristics of
nylon-6/clay-reinforced nanocomposites, and found
that clay increased the viscosity of nylon-6, and thus
sintered clay/nylon-6 nanocomposite showed lower
final density than the pure material. In our previous
work (Yan, et al. 2009), nanosilica was used to reinforce the SLS parts of nylon-12, and it was found that
3 wt% nanosilica could enhance the thermal stability,
tensile strength, tensile modulus and impact strength
of nylon-12 parts.
The applications of carbon fibre/polymer composites have been rising during the last decade, mainly in
the car and aerospace industry, due to the improvement of the electrical conductivity and mechanical
stiffness (Varelidis, et al. 1999). In this work, carbon
fibre/nylon-12 (CF/PA) composites were prepared,
and the properties of the composite powders and
sintered parts were investigated.
355
of 95 wt% ethanol, 4.5 wt% butanone and 0.5 wt%

distilled water.
The CF/PA composite powders were prepared
by the dissolution-precipitation process, which was
described in our previous work (Yan, et al. 2009a).
The CF/PA composite powders with the carbon fibre
contents of 30 wt%, 40 wt%, 50 wt% were prepared,
and denoted as 30% CF/PA, 40% CF/PA and 50%
CF/PA, respectively. A pure nylon-12 powder (PA) was
also prepared for comparison.
2.2 Selective laser sintering process
The laser sintering experiments were carried out using
an HRPS-III SLS system made by Huazhong University of Science and Technology (HUST) in China. The
SLS system was equipped with a continuous wave CO2
laser ( = 10.6 m), the maximum power of which was
50 W. In this study, the sintering parameters were as
follows: the laser beam speedwas 1500 mm/s; the scan
spacingwas 0.1 mm; the fill laser power was 22 W; the
powder layer thickness was 0.1 mm.
2.3 Measurements
The average particle sizes and particle size distributions of the composite powders were determined
by Master Min laser particle size analyzer (Malvern
Instruments Ltd.).The impact strength was determined
by Pendulum Impact Tester 5113.300, Zwick/Roell
Corporation in Germany, according to the standard
ISO1801993(E). The decomposition behaviors of
the PA and CF/PA composite powders were determined by Thermogravimetric Analysis (TGA). TGA
measurements were performed using a PerkinElmer
Instruments Pyris1 TGA at a heating rate of 10 C/min
over the range from room temperature to 600 C under
the protection of nitrogen gas.
3
3.1
Figure 1. Particle size distributions of (a) 30% CF/PA,

(b) 40% CF/PA and (c) 50% CF/PA composite powders.
2
2.1
EXPERIMENTAL
Materials and powder preparation
The carbon fibres used in the present work were supplied by Jilin Jiyan High-tech Fibre Co. Ltd., China.
The nylon-12 pellets were purchased from Degussa
Co., Germany. The nitric acid with a concentration
of 67% was obtained from CNPC Jilin Chemical Co.
Ltd., China, and used for oxidization surface modification of carbon fibres. The mix solvent was comprised

Particle size and particle size distribution
The particle size distributions and average particle

sizes of the CF/PA composite powders were measured from laser diffraction particles size analysis.
The particle size distributions of 30% CF/PA, 40%
CF/PA and 50% CF/PA composite powders are shown
in Figure 1(a), (b) and (c), respectively. The three
CF/PA composite powders have a wide particle size
distribution of 0250 m. In Figure 1(a), 30 particle
sizes of 4550 m. Figure 1(b) shows that the particle
sizes of 7585 m are dominant in 40 particle sizes of
2530 m and 9095 m are dominant in 50% CF/PA.
All the three CF/PA powders have a small amount of
particles with the size of more than 100 m, which
increase with the increase in the carbon fibre content. This is because there are some fibres with the
length of more than 100 m in the carbon fibre powder. In the preparation process of the carbon fibre
356
indicating that the CF/PA composite has higher thermal stability than the pure nylon-12. This result is due
to the much higher thermal stability of carbon fibres,
compared with the polymer matrix of nylon-12. The
high thermal stability property of CF/PA composites
extends the applications of polymer SLS parts to some
high temperature services.
3.3 Impact strength

The effect of carbon fibre content on the impact
strength of the CF/PA SLS parts is shown in Figure 3.
The impact strengths of aluminum/polyamide-12
(Al/PA) and glass beads/polyamide-12 (GF/PA) SLS
parts presented in our previous works (Yan, et al. 2009;
Yang, et al. 2010) are also provided in Figure 3 for
comparison. The impact strength of CF/PA SLS parts
expectedly decreases with increasing the carbon fibre
content from 0 to 50 wt%, due to the incorporation of
brittle fillers. As shown in Fig. 3, the impact strengths
of all the three composites decrease, but the least
reduction happens to CF/PA composites. For instance,
when the filler content is 50 wt%, the impact strengths
of CF/PA, Al/PA and GF/PA composites decrease by
38.4%, 74.4% and 59.2%, respectively. This result may
be attributed to the fact that the carbon fibres belong to
fibre fillers and have an aspect ratio, which can prevent
crack growth more effectively than nearly spherical
fillers like glass beads and aluminum powders. In addition, fibre pullouts can absorb more impact energy.
Figure 2. TGA differential curves of PA and 30% CF/PA.
4
Figure 3. Variations of the impact strengths of CF/PA,
nylon-12/aluminum (Al/PA) and nylon-12/glass beads
(GF/PA) SLS parts with the filler contents. The impact
strengths of Al/PA and GF/PA SLS parts are cited from Yan,
et al. 2009 and Yang, et al. 2010, respectively.
powder, 0100 m sized particles ere expected to be

obtained by the sieving method. However, because
the diameter of the fibres is only about 7 m, some
fibres with the length of more than 100 m could
perpendicularly go through the sieve net, and came
into the carbon fibre powder. Small quantities of
long fibres will be helpful to improve the mechanical properties of the composites. The volume average
particle sizes of 30% CF/PA, 40% CF/PA and 50%
CF/PA composite powders are 51.0 m, 67.4 m and
68.5 m, respectively, which are suitable for the SLS
process.
3.2 Thermal stability

The TGA differential curves of PA and 30% CF/PA
are shown in Figure 2. As can be seen in Figure 2,
the maximum decomposition temperature of the 30%
CF/PA is 487.34 C, which is higher than the pure PA12 maximum decomposition temperature of 458.68 C,
CONCLUSIONS
The particle sizes and particle size distributions, thermal stability and impact strengths of three kinds of
CF/PA composites with the fibre weight content of
30%, 40% and 50% for SLS were investigated. The
volume average particle sizes of the CF/PA composite powders are suitable for the SLS process. Compared with the pure PA-12, the CF/PA composites
have higher maximum decomposition temperatures to
provide higher thermal stability, which extends the
applications of polymer SLS parts to some high temperature services. The carbon fibre fillers caused the
reduction in the impact strength of the polymer matrix,
but with much lower percentage in comparison with
the glass beads and aluminum fillers.
REFERENCES
Chung, H., Das, S. (2006). Processing and properties of
glass bead particulate-filled functionally graded Nylon11 composites produced by selective laser sintering.
Materials Science and Engineering a-Structural Materials Properties Microstructure and Processing 437(2):
226234.
Hao, L., Savalani, M. M., Zhang, Y., Tanner, K. E.,
Heath, R. J., Harris, R. A. (2007). Characterization of
selective laser-sintered hydroxyapatite-based biocomposite structures for bone replacement. Proceedings of the
357
Royal Society a-Mathematical Physical and Engineering

Sciences 463(2084): 18571869.
Ho, H. C. H., Cheung, W. L., Gibson, L. 2002. Effects
of graphite powder on the laser sintering behaviour of
polycarbonate. Rapid Prototyping Journal 8(4): 233242.
Kruth, J. P., Wang, X., Laoui, T., Froyen, L. (2003). Lasers
and materials in selective laser sintering. Assem. Autom.
23(4): 357371.
Kim, J., Creasy, T. S. (2004). Selective laser sintering
characteristics of nylon 6/clay-reinforced nanocomposite.
Polymer Testing 23(6), 629636.
Varelidis, P. C., McCullough, R. L., Papaspyrides, C. D.
(1999). The effect on the mechanical properties of carbon/
epoxy composites of polyamide coatings on the fibers.
Composites Science and Technology 59: 18131823.
Yan, C. Z., Shi, Y. S., Yang, J. S., Liu, J. H. (2009a).

A Nanosilica/Nylon-12 Composite Powder for Selective Laser Sintering. Journal of Reinforced Plastics and
Composites 28(23): 28892902.
Yan, C. Z., Shi, Y. S., Yang, J. S., Liu, J. H. (2009b). Preparation and selective laser sintering of nylon-12-Coated
aluminum powders. Journal of Composite Materials 43:
18351851.
Yang, J. S., Shi, Y. S., Yan, C. Z. (2010). Selective laser
sintering of polyamide 12/potassium titanium whisker
composites. Journal of Applied Polymer Science 117:
21962204.
358

New material development for laser additive manufacturing

Liang Hao
College of Engineering, Mathematics and Physical Science, University of Exeter, Exeter EX4 4QF,
ABSTRACT: Selective Laser Melting (SLM) and Selective Laser Sintering (SLS) are the two main laserbased additive manufacturing (LAM) processes which can be utilised to make functional parts directly. The
layer-based SLM/SLS material processing approach needs to consolidate powder materials to form functional
parts. This posts restrictions on the material selection, but meantime provides opportunities to develop new
materials, especially new metallic alloy and composite materials, for the SLS and SLM. This presentation will
discuss the recent research on new SLM/SLS material developments carried at the University of Exeter. These
materials include new aluminium alloy, aluminium metal matrix composite, hydroxyapatite and stainless steel
bio-grade composites, austenitic and martensitic mixed stainless steels for the SLM process and carbon fibre
and nylon composite materials for the SLS process.
Keywords: Additive manufacturing; Selective laser sintering; Selective laser melting; Composites.
INTRODUCTION
Over the last decade, significant progress has been

made on developing new and advanced techniques for
additive manufacturing (AM). Selective Laser Melting
(SLM) and Selective Laser Sintering (SLS) are the two
main laser-based additive manufacturing (LAM) processes which can be utilised to make functional parts
directly. SLM/SLS seem to be the most versatile AM
processes to generate complex three-dimensional parts
by solidifying successive layers of powder material on
top of each other (Kruth, Levy et al. 2007). It should
be noted that while SLS uses sintering mechanism to
partially melt and then fuse polymer powders, SLM
melts metallic powders to directly fabricate functional
end-use products.
Currently, LAM has a limited range of material
choices available but it does at least offer some choices
in polymers, metals and their composites. Metal LAM
parts can be produced from CoCr alloys, steels, Ni
alloys and Ti alloys: recently, Al and Cu alloys have
been commercially released for LAM use. Polymer
LAM parts are primarily produced in DuraForm PA
(a nylon-like material); additionally polymers are typically reinforced with a filler material, an example
being DuraForm GF which is DuraformPA blended
with glass particles. Recently, Poly Ether Ether Ketone
(PEEK) has been released for commercial ALM use;
a significant advance since PEEK is considered a high
strength engineering polymer. As with any manufacturing process, research in materials is an ongoing
process for AM: potential composite material use is
also keenly investigated (Kumar and Kruth 2010).
The powder format requirement for SLM and
SLS processes has posted restrictions on the material
selection, but meantime provided opportunities to

develop new materials, especially new metallic alloy
and composite materials. This paper will present the
recent research on new SLM/SLS material developments carried at the University of Exeter. These materials include new aluminium alloy, aluminium metal
matrix composite, hydroxyapatite/stainless steel biograde composites, austenitic/martensitic mixed stainless steels for the SLM process and carbon fibre/nylon
composite materials for the SLS process.
2
SLM OF NEW ALUMINIUM ALLOY
The Al alloys currently available in SLM are based on

casting alloys due to the combination of weldability
and material molten flow behaviour; there is a growing
consensus that SLM may need aluminium alloys that
have not been considered viable before in conventional
processing. The research represented here examines
the possibility of using customised aluminium alloys in
the SLM process to produce a part that has the mechanical properties of the highest performing aluminium
alloys using customised powder mixtures. 6061 aluminum alloy was used as a base powder material due
to its well responding well to heat treatment/welding
and pure copper was used as an additive to produce
custom Al-Cu based alloys.
The microstructures of the SLM samples produced
from 6061 alloy and its copper additions are shown
in Figure 1. As seen, samples produced using only
6061 Al alloy show a dendritic structure; exposing the
powder to air prior to processing increases segregation between dendrites. The samples produced with Cu
additions show a dramatic change especially when Cu
359
Figure 2. Hardness comparison of 6061 Al alloy powder

and its mixtures after SLM processing.
growth of dendrites leading to a finer microstructural

characteristic. It also produces Al-Cu intermetallics
appearing inside matrix and increasing the strength
of the alloy. The positive influence of Cu on strength
by making Al-Cu intermetallics was more pronounced
especially in higher Cu percentages near eutectic
compositions.
Figure 1. Optical micrographs of consolidated and etched

6061 samples: (a) 6061 kept under inerted conditions;
(b) 6061 exposed to air prior to processing; (c) exposed
6061+6%wt Cu; (d) exposed 6061+30%wt Cu. All samples
were produced at a laser power of 100 W and a scan speed of
20 mm/s.
is increases to 30%wt; the microstructure is composed

of darker and brighter broken dendrites and droplets
as well as lamellar structure in this case.
The hardness results are shown in Figure 2 showing
a hardness of 200 HV for 606130%wt Cu which is
comparable to than that of high performance Al alloys.
The lowest hardness is 50 HV for exposed 6061. The
exposure to air slightly reduces the samples average
hardness by approximately 5 HV.
The 6061 Al alloy and its mixtures with Cu can
be more successfully processed by SLM, while the
addition of Cu has a positive influence on samples
hardness. The improvement in strength was related to
the ability of Cu to alter the resultant microstructure.
The microstructural analysis of samples supports the
positive influence of Cu. It revealed that Cu can restrict
SLM OF ALUMINIUM COMPOSITE
The SLM process has been investigated as a novel manufacturing technique to activate the in-situ reaction
in the mixture of Al/Fe2 O3 and directly produce netshape Al matrix composite parts. This new technique
(SLM process) provides opportunities to manufacture
Al composite parts with complicated geometries as
well as good particle distributions and reinforcement,
while other techniques are not (or hardly) able to do
so.
Figure 3(a,b) shows the progress of SLM (laser
power 40 W and scanning speed 0.07 m/s) in single
layer Al powder and Al/5 wt.%Fe2 O3 powder mixture, revealing the comparative appearance of the laser
sparks. As seen from Figure 3a, a small spark is
observed when laser scans the pure Al powder. This
small spark is generated by the instant laser-material
interaction which causes the melting of Al powder.
In contrast, in Figure 3b, the same SLM parameters
used to produce a layer specimen from Al/Fe2 O3 mixture generate a much larger spark and broader melting
pool. This spark is believed to be initiated by the
laser-material interaction and then further amplified
by the in-situ thermite interaction between Al and
Fe2 O3 powders.
A series of multilayer three-dimensional cubic parts
was successfully produced from the Al/5 wt.%Fe2 O3
mixture using SLM process, as seen in Figure 3(c).
These parts demonstrate that SLM process can be used
as a novel in-situ technique to produce Al net-shape
matrix composite parts.
Figure 4 shows the typical microstructural appearance of the sample, demonstrating the presence of
very fine particles inside grains and in boundaries.
The particles should be in the form of aluminium
oxides and Al-Fe(-O) combinations (Al, O, and Fe are
the main elements). The particles are very fine (perhaps 502000 nm) in size, have a strong bonding in
360
Figure 4. Typical SEM microstructural views of the SLM

part fabricated from Al/5 wt.%Fe2 O3 mixture showing fine
particles formed inside grains and boundaries (overall chemical composition was acquired by EDS).
Figure 5. The typical SEM micrograph of 316 L SS/HA

powders distribution.
extend the commercial applications of the Al matrix

composite in various industrial sectors.
Figure 3. A comparative visual observation of SLM process to produce single layer specimens from (a) Al powder
and (b) Al/5 wt.%Fe2 O3 powder mixture, (c) The typical
multilayer cubic parts produced by SLM of Al/5 wt.%Fe2 O3
powder mixture.
their interface with the matrix, and are distributed uniformly. These particles act as reinforcement phases,
with very high strength of ceramics and intermetallics.
This is highly desirable in Al matrix composites leading to desirable mechanical properties. Overall, the
results indicate that the particle reinforced Al matrix
composite has been produced through SLM activated
in-situ reaction process.
The study reveals that SLM can directly produce Al
matrix composite parts by an in-situ technique from
the Al/Fe2 O3 mixture using appropriate parameters.
The very fine particles (as reinforcements) mainly
are appeared to be minerals and ceramics such as
Fe2+ Al2 O4 and Al2 O3 and uniformly distributed in
the Al matrix, establishing a good interface with the
matrix. These well-distributed reinforcing particles are
expected to provide desirable mechanical properties.
This new development has great potential to make high
performance Al matrix composite components and
SLM OF HYDROXYAPATITE/STAINLESS
STEEL BIO-GRADE COMPOSITES
The study presented here investigate the SLM of

stainless steel (SS) and hydroxyapatite (HA) composite to develop load bearing customised implants
that can be bioactive to bond bone tissues. The addition of HA particles to SS powder could significantly
affect SLM processing parameters as well as resulting
microstructures and materials properties.
The HA particles and 316 L SS powders were mixed
with 5 vol. % HA using a bespoke made blender.
The mixture of the powders was examined by SEM
confirming a good distribution of the HA particles
(Figure 5).
Figure 6 shows an SEM microstructural image of
316 L SS/HA composite fabricated using laser power
of 42.2 W, scanning speed of 0.160 m/s, and scan line
spacing of 50 m. The SS/HA composite shown in
Figure 6(a) appears to have smaller grain size and
hence it has much more grain boundaries than that
of the pure 316 L SS sample (Figure 6(b)). The pure
316 L SS sample had an average grain diameter of
8.95 m, while the SS/HA composite had a less average grain diameter of 6.01 m. This phenomenon
361
Figure 6. Typical SEM view of (a) SS/HA and (b) 316 L SS

microstructure.
Figure 8. Cross sections of consolidated samples (top

surface of samples directly exposed to laser): (a) 0 wt-%
17-4 PH, (b) 25 wt-% 17-4 PH, (c) 100 wt-% 17-4 PH.
Figure 7. Typical Eng. stress-strain behaviour of 316 L

SS/HA specimens fabricated by SLM using laser power of
42.2 W and scan line spacing of 50 m in different scanning
speeds.
can be attributed to HA particles that work as nucleation sites to assist heterogeneous nucleation and leads
to a further refinement of the material.
The Figure 7 shows that the tensile strength of material alter by different scanning speeds, in such a manner
that reduction in scanning speed improves the tensile
strength until the scanning speed of 0.160 m/s and then
decreases it, so that an optimum of scanning speed may
appear at the highest tensile strength of 95 MPa. It is
interesting to note that cortical bone has an average
stiffness of 15 GPa, tensile strength of 60130 MPa,
and ultimate strain of 1 to 5% varied with people and
their ages. So, the average tensile strength of 95 MPa,
and homogenous strain of 2.7% for the SS/HA specimen made using optimum parameters are very close
to that of bone strength and elongation in human body.
Therefore, the fabricated composite specimen seems
to be not only suitable for load bearing application, but
also not too stiff to prevent stress shielding.
The present study indicates that an optimum condition is existed to fabricate good quality 316 L SS/HA
composite specimens using a duplicate scanning strategy during SLM process. Moreover, it was perceived
that 316 L SS/HA has a finer grain size than that of
316 L SS attributed to the role of HA particles as nuclei
to facilitate heterogeneous nucleation. The SS/HA
composite fabricated using the optimum parameters
exhibited a tensile strength of 95 MPa which is adequate for load bearing application. Accordingly, one
may conclude that the SS/HA composite manufactured
in this research is highly potent to be employed within
bioactive and load-bearing bone implant applications.
SLM OF AUSTENITIC/MARTENSITIC
MIXED STAINLESS STEELS
Some previous studies have used the SLM process to

consolidate 316 L and 17-4 PH steel powders separately and have investigated the effects of processing
parameters on the density and mechanical properties
of the consolidated parts. However, there has been no
investigation into the potential to tailor the microstructures and properties of parts by SLM of a mixture of
two stainless steel grade powder mixtures. The study
presented here has investigated the use of SLM to
produce metallic parts with tailored mechanical and
magnetic properties by the consolidation of SS powder
mixtures with varied 316 L/17-4 PH mixture ratios.
Figure 8 shows the cross section of the consolidated samples. As can be seen from Figure 8(a), the
0% 17-4 PH (100% 316 L) sample shows a fully dense
362
Figure 9. Optical etched microstructures taken from below

the lased surface of the 316 L steel samples with: (a) 0 wt-%
17-4 PH, (b) 25 wt-% 17-4 PH, (c) 50 wt-% 17-4 PH, (d)
75 wt-% 17-4 PH, (e) 100 wt-% 17-4 PH.
region until approximately half the depth of the sample with a less dense and porous region occupying the
bottom half. The porous region at the bottom half of
the samples is apparently liquid phase sintered by partial melting of particles. It is interesting to note the
much improved consolidation of the 25 wt-% 17-4 PH
samples over the 0 wt-% 17-4 PH samples. 316 L has
0% copper and 17-4 PH has 2.5 to 4.5 wt % copper. It
is thought that the full consolidation of 316 L samples
containing 25% 17-4 PH is due to the copper addition
that triggers the full melting and flow of the melted
steel powders. The copper could be flowing in a liquid
phase between powder particles due to capillary forces
which would create a region that contains heat more
effectively. The 100% 17-4 PH sample shows full consolidation throughout most of the cross section, with
porosity evident only on the bottom of the sample. This
also confirms that the copper elements in 17-4 PH steel
powders played an important role in the melting and
densification of the laser scanned powders.
The 316 L steel with 0 wt% 17-4 PH sample show
a very clear consolidated austenitic microstructure
in Figure 9(a). It is encouraging to see that the
SLM process can produce a widely recognised and
fully dense microstructure from powdered steel. The
25 wt-% 17-4 PH microstructure appears dendritic in
nature.The structure is still austenitic but demonstrates
more defined grain boundaries compared to the 0 wt%
17-4 PH samples. Also, the addition of copper element
via 25% of 17-4 PH into austenitic steel could cause
the precipitation in the grain boundaries to reinforce
the grain boundary. At 50 wt-% 17-4 PH the austenitic
316 L still dominates the microstructure but the apparent precipitation of -phase ferrite and cementite is
more apparent. The 75% 17-4 PH microstructure consists of elongated austenite grains laminar in nature
with small regions of martensite forming. It appears
that a critical value has been reached that now allows
martensite to form as grains. The 100 wt% 17-4 PH
sample shows a typical lath martensite microstructure
that has been tempered.
The graph in Figure 10 shows the initial value
of 316 L with 0% 17-4 PH to have a hardness of
approximately 170 HV, over 10% higher than if conventionally manufactured (148 HV). This indicates
that laser irradiation introduced hardening effects to
Figure 10. Graph of content of 17-4 PH versus mean

Vickers hardness and maximum mean magnetic adherence
(error bars represent standard deviation).
the 316 L samples due to rapid heating in a localised

area followed by a cooling rate that allows a small
amount of martensitic transformation. When the samples consist of 100% 17-4 PH, the hardness is approximately 337 HV, slightly lower than conventional values
(349 HV). This is due to the overlapping multiple laser
scanning passes which could lead to a tempering effect
in 17-4 PH and reduce its hardness to a small extent.
The magnetic values determined experimentally show
a striking similarity to the results found for hardness.
The 100% 17-4 PH samples show the strongest affinity for magnetic adherence whereas the 0% 17-4 PH
samples only have trace values. This matches with
the magnetic permeability of the materials (1.02 for
316 L, 95 for 17-4 PH). Just like the hardness values,
the increase in magnetic adherence does not follow
a linear trend throughout the whole range of values.
Given the evidence of the hardness and magnetic values in Figure 10, it appears that the samples consist
of primarily austenitic phases until a 75% martensite
fraction is reached at which the hardness and magnetic
adherence increase dramatically.
SLM is provided to be able to manufacture stainless steel composite parts via the powder mixture of
316 L and 17-4 PH powders. The addition of 25%
17-4 PH SS to 316 L SS powder dramatically changes
densification mechanisms to improve consolidation.
The composition ratios of the powder mixture directly
influence the microstructure evolutions which led to
changes in hardness and magnetic properties.
6
SLS OF CARBON FIBRE/NYLON

COMPSOITE
The application of carbon fibre/polymer composites

has been rising during the last decade, mainly in the
automotive and aerospace industry, due to the improvement of the electrical conductivity and mechanical stiffness. Recently, Goodridge et al. (Goodridge,
Shofner et al.) investigated the SLS of carbon nanofibre (CNF)/PA-12 composite powders prepared by
melt mixing and cryogenic milling. However, the
surfaces of CNF were not modified, and consequently the composite powders did not exhibit the
strong interfacial interaction between CNF and PA
matrix and the cryogenic fracturing method did not
363
with increasing the carbon fibre content. The flexural strengths of 30% CF/PA, 40% CF/PA and 50%
CF/PA sintered specimens are increased by 44.5%,
83.3%, 114%, respectively, as compared with the flexural strengths of the pure PA-12 SLS specimens; and
the flexural moduli are enhanced by 93.4%, 129.4%
and 243.4%, respectively. The significant improvement in flexural strengths and moduli demonstrates
that the carbon fibres effectively reinforce the PA-12
matrix due to their homogenous dispersion and strong
interfacial bonding.
Therefore, the developed CF/PA powders are considered to have suitable interfacial bonding and powder
morphology for SLS to manufacture high performance
components and broaden its application.
Figure 11. SEM micrographs of the 30% CF/PA composite

powder at the magnifications of (a) 300 and (b) 2500.
Figure 12. Variations of the flexural strength and flexural

modulus of the CF/PA SLS parts with the carbon fibre
content.
produce powder with suitable particle morphology

for SLS, resulting in the lower storage modulus to
the SLS composite parts than the SLS parts using
pure commercial PA-12 powders. In this sudy, carbon fibre/polyamide-12 (CF/PA) composite powders
were prepared by a dissolution-precipitation method.
The carbon fibres were surfacetreated and coated with
nylon-12 through dissolution-precipitation process to
form suitable CF/PA composite powders for SLS.
The SEM micrographs of the 30% CF/PA composite
powder (Figure 11) shows that the composite powder is comprised of the irregularly shaped particles
and fibres with different lengths. It can be found that
carbon fibres are coated with a layer of PA-12, indicating that good adhesion is established at the interface
between two materials and carbon fibre is well dispersed in the composite powder. powders used in the
SLS process.
As shown in Figure 12, the carbon fibres can
greatly enhanced the flexural strength and flexural
modulus of sintered components, which are increases
CONCLUSION AND FUTURE OUTLOOK
The research on new materials for SLM/SLS demonstrates the great potential to tailor the materials
microstructure, interface between composite materials
and improve their material properties of the resulting
parts. It shows that the compositions and format of
initial material powders and laser processing parameters can play an important role in the consolidation
performance during SLM/SLS process.
The outcome of our study indicates the great scope
to discover and identify more wide range and specific
material for SLM/SLS, enabling LAM to produce high
performance and functional components.
ACKNOWLEDGEMENT
This work has been contributed by Mr. Peter Jerrard,
Mr. Dadbakhsh Sasan, Dr. Chenze Yan.
REFERENCES
Goodridge, R. D., Shofner, M. L. et al. (2011). Processing
of a Polyamide-12/carbon nanofibre c omposite by laser
sintering. Polymer Testing 30(1): 94100.
Kruth, J. P., Levy, G. et al. (2007). Consolidation phenomena in laser and powder-bed based layered manufacturing. CIRP Annals Manufacturing Technology 56(2):
730759.
Kumar, S., Kruth, J.-P. (2010) Composites by rapid prototyping technology. Materials & Design 31(2): 850856.
364

Mechanical behavior of Epoxy-Aluminum composite for

rapid tools applications
G.V. Salmoria, F.A. Yaez-Villamizar, A. Sabino-Netto & G.M.O. Barra
ABSTRACT: Casting thermosets is also applied in the manufacture of functional prototypes such as rapid
tools. The thermal and mechanical properties are of great importance for rapid tools such as molds for injection
of thermoplastics, since these tools are subjected to particular conditions of work with temperature variations
and mechanical requirements. The mechanical properties of parts or tools manufactured by casting depend
on the degree of cure of the resin. Manufacture parameters, such as the moisture and casting technique during
construction, maximum density of cross-linking, rate and degree of cure, and post-cure method are very important
to define the structure and properties of casting samples. Knowledge of the relationship between resin structure
and properties is useful in the casting tools manufacturing process, application and quality control. This work
correlates the mechanical behavior with material properties and post-cure techniques. The investigation of
mechanical properties of epoxy-aluminum specimens post-cured by different routines showed that results are
related with the self controlled diffusion characteristic of thermoset polymeric systems. High temperature postcure routine result in higher values of modulus, higher stiffness and high values of glass temperature, Tg, to the
specimens. The fracture surfaces analyses showed the presence of defects, like empty spaces in the epoxy matrix
due to the moisture and casting process. The defect size and the specimen strength showed a direct correlation,
Weibull modulus was 7.95 to epoxy specimens and characterize a low toughness and a defect material as showed in
the fractography. To stress application higher than 38 MPa, the failure probability rise rapidly to 50 of probability.
1
INTRODUCTION
The casting process permits the rapid manufacture

of parts and objects by thermoset polymer curing.
The main applications of the casting process are the
production of prototypes and models used in the
automobile, electric, aerospace and biomedical areas.
Casting thermosets is also applied in the manufacture of functional prototypes such as rapid tools [13].
The thermal and mechanical properties are of great
importance for rapid tools such as molds for injection of thermoplastics, since these tools are subjected
to particular conditions of work with temperature
variations and mechanical requirements [46]. The
mechanical properties of parts or tools manufactured
by casting depend on the degree of cure of the resin.
Manufacture parameters, such as the moisture and
casting technique during construction, resin characteristics such as monomers, oligomers and curing agent
compositions, maximum density of cross-linking, rate
and degree of cure, and post-cure method are very
important to define the structure and properties of casting samples. Knowledge of the relationship between
resin structure and properties is useful in the casting
tools manufacturing process, application and quality control [57]. This work correlates the mechanical behavior with material properties and post-cure
techniques.
EXPERIMENTAL
2.1 Materials
A epoxy polymer filled with aluminum in 30% by
weight, supplied by Hunstman as RenCast 436 were
mixed with its curing agent, Ren HY 150 from the
same company. The material is an epoxy system base
in Bisphenol A and Phenol Novalac derivatives.
2.2 Mixture procedure
In order to determine the effect of the post-cure process on the mechanical properties of epoxy-aluminum
composite, the mixture of epoxy pre-polymer with
aluminum particles and curing agent were prepared
using mechanical stirring (85 rpm) for 2 minutes
at 20 mmHg. The mixture was poured into silicone
rubber models to fabricate test specimens.
2.3 Cure and post-cure procedure
Experiment was formulated in order to evaluate the
influence of the post-cure on the tensile strength
and the dynamic mechanical properties of the material. Specimens were cured at room temperature for
24 hours according to the procedure suggested by the
supplier, and post-cured using a conventional thermal
oven for every post-cure routine shown in Table 1.
365
Table 1. Post-cure routines applied in the

casting epoxy specimens.
Specimens
Post-cure routine
2 h at 70 C; 2 h at 90 C; 2 h
at 120 C; 2 h at 150 C
4 h at 90 C; 4 h at 150 C
4 h at 70 C; 4 h at 150 C
B
C
2.4 Tensile tests

Tensile test specimens according to the standard
ASTM D638 type I, were made using the epoxyaluminium composite. A number of five Specimens
for each post-cure cycle were tested in uniaxial tension
at room temperature using a EMIC universal testing
machine at displacement rate of 5 mm/min. Youngs
modulus was determined through the stressstrain correlation considering the linear region. The strain was
measured by an incremental extensometer.
2.5
Figure 1. Stress versus strain curves for the epoxy specimens post-cured using different routines.
Fractography
Fracture surfaces of the tensile specimens were

observed with a scanning electron microscope, SEM,
(Philips XL30) in order to investigate the fracture topography. Every specimen was coated with
gold in a Bal-Tec Sputter Coater SCD005. Fracture
image analysis was carried out using Analysis Pro
2.11.002Soft-imaging Software GmbH in order to
failure investigation.
2.6
Dynamic mechanical analysis (DMA)
Dynamic mechanical analysis were carried out in

specimens with a size of 10 mm 40 mm 3.5 mm
over a temperature range from 70 C to 200 C
using a DMTA analyzer (Elkin-Palmer Instruments)
in the three bending point mode geometry according to the standard ASTM D5023. They were scanned
isochronally at 3 C/min. Damping factor (Tan ) and
storage modulus (Log E) were recorded at 1 Hz.
2.7 Thermal tensile tests
Specimens according to the standard ASTM D638
type V were made using epoxy-aluminum resin mentioned previously. Specimens were post cured using the
cycle proposed by the supplier (A). Specimens were
tested in uniaxial tension at 23 C, 70 C, 120 C and
150 C, using a Instron universal testing machine at
displacement rate of 5 mm/min.
3
The Figure 1 presents the stress-strain curves for

the post-cured specimens. Table 2 shows the main
values of elastic modulus and tensile strength for
Figure 2. Storage Modulus E and Tan curves for the

epoxy specimens post-cured using different routines.
the post-cured specimens where is possible to see

that specimens postcured using routine A presented
higher values of tensile strength.
These results are related with the self controlled
diffusion characteristic of thermoset polymeric systems, presenting a more extensive crosslink density improvement when gradual thermo-curing ramp
routines are applied, i.e. in A routine.
The storage modulus E and Tan are properties
related to the stiffness and toughness of epoxy systems. The post-cure routine B result in higher values
of modulus and higher stiffness to the specimens. The
glass temperature, Tg, is obtained when Tan is maximum and reveled a higher value of Tg for B specimens
confirming high cure degree.
The fracture surfaces analyses showed the presence of small defect like empty spaces (bubbles) in
the epoxy matrix due provably to the moisture and
casting process. The investigation about the relationship between the defect size and the specimen strength
showed a direct correlation presented in the Figure 4.
366
Figure 6. The failure probability of epoxy specimens.

Figure 3. Fracture surfaces of specimens post-cured by
different routines.
Figure 4. Defect size and strength relationship to epoxy

specimens.
Figure 5. The Weibull modulus characterizing the material

fiability.
CONCLUSIONS
The relationship between resin structure and properties is useful in the casting tools manufacturing
process, application and quality control. The investigation of mechanical properties of epoxy-aluminum
specimens post-cured by different routines showed
that results are related with the self controlled diffusion characteristic of thermoset polymeric systems,
presenting a more extensive crosslink density improvement when gradual thermo-curing ramp routines are
applied reaching higher values of tensile strength and
lower defect presence due homogenous cure. High
temperature post-cure routine result in higher values
of modulus, higher stiffness and high values of glass
temperature, Tg, to the specimens. The fracture surfaces analyses showed the presence of defects, like
empty spaces in the epoxy matrix due to the moisture
and casting process. The defect size and the specimen
strength showed a direct correlation, Weibull modulus was 7.95 to epoxy specimens and characterize a
low toughness and a defect material as showed in the
fractography.To stress application higher than 38 MPa,
the failure probability rise rapidly to 50 of probability.
Epoxy-aluminum composite show interesting thermal
and mechanical properties to the tools manufacture,
but aluminum content and manufacture defects can
limit its use in critical conditions such as specific
geometries and high stress molding.
ACKNOWLEDGMENTS
The Weibull modulus characterizes the material fiability and the defects influence. The Figure 5 shows the
Weibull modulus is the inclination curve. It was 7.95
to epoxy specimens and characterize a low toughness
and a defect material as showed in the fractography
analyses independent of post-cure process.
The Figure 6 shows the probability of failure as
function of applied stress. To stress application higher
than 30 MPa, the failure probability rise rapidly. It
reaches 50 and 80% of probability when stress is
close to 38 and 42 MPa, respectively.

support.
REFERENCES
Jacobs, P. F. (1993). Rapid Prototyping and Manufacturing:
Fundamentals of Stereolithography, Society of Manufacturing Engineers, Dearborn, MI, USA.
Jacob, P. F. (1999). From Rapid Prototyping to Rapid Tooling,
ASME, New York.
367
Salmoria, G. V., Lafratta, F. H., Biava, M. M., Ahrens, C. H.

(2008). Rapid manufacturing and rapid tooling of polymer
miniaturized parts using stereolithography. J of the Braz
Soc of Mech Sci & Eng (30): 710.
Salmoria, G. V., Gonzalez, V. J., Ahrens, C. H., Soldi, V.,
Pires, A. T. N. (2005). Stereolithography somos 7110
photosensitive resin: Study of curing kinetic and thermal
degradation. J Mater Process Technol 168(11): 164171.
Salmoria, G. V., Ahrens, C. H., Fredel, M., Soldi, V., Pires,
A. T. N. (2005). Stereolithography Somos 7110 resin:
Mechanical behaviour and fractography of parts postcured by different methods. Polym Test 24(2): 157162.
Salmoria, G. V., Klauss, P., Pires, A. T. N. (2008). Investigations on cure kinetics and thermal degradation of
stereolithography Renshape 5260 photosensitive resin.
Polym Test 27: 698704.
Salmoria, G. V., Ahrens, C. H., Beal, V. E., Soldi, V. (2009).
Pires ATN. Evaluation of post-curing and laser manufacturing parameters on the properties of SOMOS 7110
photosensitive resin used in stereolithography. Materials
and Design 30: 758763.
368

Thermal characterization of laser sintering of nylon-12

T.T. Diller, M.M. Yuan, D.L. Bourell & J.J. Beaman
Laboratory for Freeform Fabrication, The University of Texas at Austin
ABSTRACT: Precise thermal control of laser sintering is desirable for improving geometric accuracy, mechanical properties, and surface finish of polyamide (nylon) parts. A modeling and measurement system was set up
to monitor thermal aspects of laser sintering of nylon-12 processing with a view to develop thermal control
strategies to improve these characteristics.
A critical component of understanding the thermal processes in laser sintering is to know the thermal properties
of the materials being processed. Measurements were made of the thermal conductivity and heat capacity of
nylon-12 powder using a transient plane-source method with HotDiskTM equipment. Thermal properties are
reported over the range 30 C170 C in a dry N2 environment similar to the build chamber of a laser sintering
machine. The powder was packed in the sample holder at a tap density of 0.547 g/cc (54.2% relative density).
Thermal conductivity was found to vary with temperature in the range 0.10.114 W/m-K.
This data was used in a computational model to understand the powder warmup process with respect to the
thermal transport processes between the part, the part bed, and the build chamber.
This research was funded by the Air Force Research Laboratory under Grant Number GRT00015778.
1
INTRODUCTION
Laser sintering is an important method of rapidly and

flexibly fabricating polymer parts for prototyping and
production of small batches or parts with geometries
that would otherwise not be manufacturable. The feedstock is a polymer powder, typically nylon although
other polymers are available. Successive layers of
powder are deposited in a build chamber, and the powder coalesced in a pattern through localized melting
by a laser.
Geometric details on the order of 0.5 mm are possible because melting can be localized, and diffusion
of thermal energy through the powder is slow. Careful
control of the thermal environment is critical to prevent
undesirable growth of the part after scanning by the
laser. Additionally, during the cooling process, thermal gradients in the part must be minimized to avoid
geometric distortion resulting from thermally induced
residual stresses. The need to reduce thermal gradients
in the cooling part bed leads to long cooling times and
reduced productivity of laser sintering machines.
In this study, the powder warm-up process is briefly
examined using a measurement and modeling system previously described (Diller et al. 2010). Thermal
properties for the powder used in the study were not
previously available over the appropriate temperature
range, so they were measured for use in the simulation
and analysis.
The most critical material property for thermal
processes is thermal diffusivity, th :
where kth is the thermal conductivity, is the density,

and c is the heat capacity. kth and c typically vary with
temperature; thus, th is also temperature dependent.
METHODS
2.1 Measurement of thermal properties

PA 650, a nylon-12 powder in common usage in polymer laser sintering and produced by Advanced Laser
Materials, LLC, was chosen as the sample material in
this study.
Materials, LLC, was chosen as the sample material
in this study.
Thermal diffusivity was measured using the transient plane source technique (Gustafsson et al. 1990,
Gustafsson 2009, Gustavsson et al. 2009, Maqsood
et al. 1994) using a TPS500 from HotDisk AB. To carry
out the transient plane source method, a measurement
probe was embedded in a sample of the powder. The
measurement probe consisted of a bi-flliar nickel spiral
in the shape of a disk. A controlled current was passed
through length of the spiral such that the power flux
into the surrounding powder was constant. Because the
relationship of resistance to temperature of the nickel
filament was well known, measurement of the voltage
drop across the filament yielded an accurate indication
of the temperature of the probe and, by inference, the
powder in its immediate vicinity.
An analytical relationship exists for the temperature of the sample as a function of time and its thermal
diffusivity (Gustafsson et al. 1990). The thermal diffusivity of the powder was found by fitting the observed
369
Figure 2. Build chamber of a SinterStation2500 laser

sintering machine.
Figure 1. Powder sample holder for transient plane source

measurements of thermal conductivity.
time-temperature curve to the analytical solution. This

method was repeated over a range of temperatures
relevant to laser sintering, from 40 C to 170 C.
The powder was held in a sample chamber, shown
schematically in Figure 1, that enabled purging of the
powder with N2 . This matched the conditions in the
build chamber, where purging with an inert gas is
necessary to avoid oxidation of the powder.
The sample holder consisted of a cylinder in two
halves. A shallow cutout in one of the cylinder ends
allowed penetration of the measurement probe into the
sample. An O-ring formed a gas-tight seal where the
cylinder halves joined together. The sensor probe element is shown edge-on in the center of the powder
sample where it penetrates the side wall of the sample
holder.
Caps at either end of the sample holder were similarly sealed with O-rings. A fitting and gas-permeable
filter allowed introduction of the inerting gas into one
end of the sample holder. A filter kept powder from
escaping through exhaust holes in the other end of the
sample holder. Pre-heated purge gas flowed through
the sample powder during temperature changes to protect the powder and speed the temperature change but
was shut off during the measurements.
Thermal conductivity was calculated by multiplying the thermal diffusivity by the measured density
and published heat capacity values (Mark 2007),
according to Eq. (1).
2.2 Simulation of thermal processes
A photograph of the build chamber of a 3D Systems SinterStation 2500, a typical example of a laser
sintering machine in use, is shown in Figure 2.
The build chamber consists of three bins at the bottom containing powder. The central part bin is where
the part is built and is empty at the beginning of a build.
Figure 3. Mesh of the computational domain for the simulation of the laser sintering build chamber.
The bins to either side are called feed bins and hold the
powder until it is needed. Radiant heaters above each
of the bins preheat the powder. The heater over the part
bin is controlled by feedback from an infrared sensor
that measures a spot on the powder bed surface.
Before a build is begun, the machine is preheated to
an initial temperature of around 80 C by means of the
radiant heaters. Then, once a build has been initiated,
there is a warmup period, in which successive layers
of powder are deposited in the build bin to form an
insulating layer against the piston at the bottom of the
bin. The powder is deposited in layers of increasing
temperature so that a thermal gradient is built into the
powder. At the end of the warmup, the powder is preheated to 170 C, just below the melting temperature
(180 C).
To deposit a layer of powder, a feed bin rises a
small increment to bring a measured volume of powder
above the surface. A counter-rotating roller passes over
the feed bin and pushes the powder in a smooth layer
over the part bin. After a pattern is melted into the surface, a piston under the part bin lowers by a small increment, and the process is repeated until the part is built.
A 2-dimensional representation of the build chamber was built for use in the Fluent computational fluid
dynamics software from ANSYS. The computation
domain mesh is shown in Figure 3.
The model included a build chamber based on the
geometry of the machine shown in Figure 2, with a
mezzanine structure containing heaters over powderfilled bins. The heaters were modeled as steel plates
with power flux subject to proportional control based
on the temperature at the center of the powder bed surface. The mezzanine was modeled as fiber-glassfilled
370
steel walled structure. The side and upper walls of the

build chamber were modeled as 50 mm thick fiberglass, with a 5 W/m2 K external convective coefficient
at 27 C surrounding temperature. The bottom wall of
the build chamber was modeled as 50 mm thick steel.
The powder bin walls were modeled as 3 mm thick
steel, and the part bin walls were modeled as 6 mm
thick fiberglass insulation. Thin steel baffles hang
from the mezzanine to block line-of-sight between the
heaters and the adjacent powder beds.
In the machine serving as the basis for the model,
the nominal power of the part bed heaters was 1815 W,
and the nominal power for the feed bin heaters was
1000 W. Examination of the machine heater control
logic revealed a proportional control system.
In the model, a software control was designed to
mimic the hardware control in the machine. Heater
power was set according to the following law:
where Pnominal was the nominal maximum power of

the heater, and b was the control variable set according
to
Figure 4. Thermal conductivity of nylon-12 powder packed

at 0.547 g/cc measured by the transient plane source method
in an N2 atmosphere. Error bars show the standard deviation
of the result over five measurements.
Thermal conductivity was found to increase

linearly with temperature from 0.1 W/mK at 40 C to
0.114 W/mK at 170 C.
where p was the proportional control coefficient and
was the difference between the set temperature and
the measured powder surface temperature.
p was set by observing the range of over which
the machine operated in open-loop mode. The software heater control was set to match the same openloop behavior. Because the heater effectiveness was
unknown, the nominal power was set with an iterative
process to match the surface temperature time histories
of the simulation with an observed surface temperature
in the machine.
The model was used to simulate a simplified
warmup process. That is, the transport and deposition of powder were not modeled. Rather, the thermal transport processes were modeled to understand
the convective, radiative, and conductive processes in
the build chamber. The evolution of temperature at the
powder bed surface and through the depth of powder
bed were examined, and the convective flow patterns
were predicted.
3
RESULTS
3.1 Material properties

The mass and volume of the powder sample were measured, and the density of the powder was found to be
0.547 g/cc. Using the measured density and the temperature dependent heat capacity published by Mark
(2007), thermal conductivity was calculated over the
range 40170 C, as shown in Figure 4.
3.2
Simulation of warmup
A 30 minute warmup period was simulated with

the CFD model, and the results are compared with
machine temperature measurements in Figure 5.
The heater controller in the machine operated in
open-loop mode, that is, the heaters were on full power,
until the temperature difference was about 10 C.
Therefore p was set to 0.1. In the machine, the part bin
heaters had a nominal power of 1815 W and a surface
area of 0.1282 m2 , so the nominal surface power flux
was 14.2 kW/m2 . At steady state, the machine heaters
operated at about 60% of the nominal maximum
level.
In the simulation, the part bin heaters were represented by lines of 0.443 m total length. Therefore, the
nominal power level for the perfectly effective heaters
in the simulation would be 6300 W/m. However, simulation with that power level for the heaters resulted
in too fast a rise time. The power level used for the
result shown in Figure 5 was 4025 W/m. Total radiative
heat transfer to the surface during the open-loop portion of the warmup was 1180 W/m. Thus, 12% of the
power to the heaters was transferred to the powder surface. At steady state, the simulated heaters operated at
3350 W/m, and radiant heat transfer to the surface was
500 W/m. The steady-state heater power was 83% of
the nominal maximum.
During the warmup, a stable flow pattern was established over the surface of the part bin powder surface.
The pattern was typical of convection over a heated surface and is shown as a set of vectors in the upper set of
371
Figure 5. Surface temperature during a warmup from 80 C

to 140 C. Simulation temperature is plotted with a solid line;
temperature measured by the IR spot sensor in a SinterStation
2500 is dotted.
axes in Figure 6. The length of each vector represents

distance traveled in one second at the instantaneous
velocity at the time snapshot. The limits of the x-axis
correspond to the edges of the part bin powder surface.
On the lower set of axes in Figure 6, a profile
of the part bin surface temperature is plotted against
x-position at the same time shown in the vector plot
above. At the edges of the powder surface, where the
tangential gas velocity is higher, the convective coefficient was higher, so the surface temperature was
somewhat lower. Near the center the gas was nearly
stagnant, and the convective coefficient was lower. The
convection pattern induced a warm spot in the powder
bed surface.
DISCUSSION
4.1 Thermal properties

Thermal conductivity of nylon-12 powder measured
with the transient plane source method was somewhat
higher than previous measurements by Xue and Barlow (1990). Differences in the analytical method are
not likely to cause substantial differences in results.
Many of the same underlying assumptions were made
for each set of measurements. Both methods rely on
essentially the same principle of fitting a measuredresponse to an analytical modeled transient thermal
response.
Rather, the difference in measured thermal conductivity was most likely due real differences in
the measurement samples. Xue and Barlow did not
specify the chemistry of their sample, nor did they
report the packing density, which would have had a
strong impact on the thermal conductivity Deissler
and Boegli, Gusarov, Laoui, Froyen, and Titov, Hall
and Martin, Tsotsas and Martin. Without knowing the
Figure 6. Final temperature profile in the powder bed and

convective pattern over the powder bed at the end of a simulated warmup. Outlines of the partbins heaters and mezzanine
structure are shown at about 20 cm above the powder surface.
details of the previous measurements, however, it is not

possible to conclusively account for the differences.
The thermal conductivities reported in this work
are for a commonly used powder at a packing density
that would be found in a part bin and at temperatures
relevant to the laser sintering process.
4.2 Modeling results

The simulation was necessarily a simplification of the
real problem. The simulation had enough complexity
to capture some of the important features of the heat
transfer processes in the build chamber during nylon
laser sintering.
However, some important details were not included.
Passage of the powder-deposition roller over the powder bed surfaces disrupts the flow patterns ina real
build in a way that was not accounted for in the simulations. Deposition of powder at different temperatures
was also not accounted for. The impact of the laser
itself on surface temperatures and material properties
was not examined, although it is of critical and central
importance to the process.
Nevertheless, some important insights were gained.
In Figure 5, it is seen that the slope of temperature
curve is higher in the simulation than in the machine
measurement. At around 4 minutes, there is a distinct
change in the slope of the line. This corresponds to the
time at which the feed bin heaters turned off. Thus,
in the simulation, the cross heating between the part
bin and feed bin heaters is greater than in the real
simulation. Given that the geometry of the machine
is closely matched, it is likely that the difference in
cross-heating is due to an inaccuracy in the radiation
372
characteristics of the heater surface. In the simulation, no account was made for the angular variance
in emitted radiation.
In the simulation, the surface temperature was
maintained with a steady-state heater power of 83%
of the maximum, compared with 60% in the machine.
This was likely the result of greater losses through
convection at the heater surfaces and then through the
walls. Future simulation could be improved by making
the wall boundary conditions more insulating.
Accuracy of the simulation was limited both by
the level of detail included and the amount of time
required to run. One important tradeoff in this simulation was the result of finite element simulation
involving large thermal gradients in an insulator. The
grid size needed to be kept small so that the expected
high thermal gradients near the powder surface could
be resolved. However, in order to avoid a numerical instability, the ratio of the time step to the grid
size needed to be constrained. Otherwise, non-physical
anomalies would appear in the powder surface temperature profile. Therefore, the time step was reduced
until the error due to this instability was bounded and
small. The available mesh resolution was limited both
by the number computations and the need to reduce
the time step to reduce numerical instabilities.
5
CONCLUSIONS
New measurements of the thermal conductivity of

nylon-12 powder provided previously unavailable data
for simulation and analysis of nylon laser sintering.
Measurements with the transient plane source method
yielded conductivities somewhat higher than earlier,
more limited, measurements. Thermal conductivity for
nylon-12 powder packed at 0.547 g/cc was found to
vary from 0.1 W/mK to 0.114 W/mK when heated
from 40 C to 170 C.
Simulation of a warmup and comparison with
results measured in a machine allowed assessment of
the thermal processes in the build chamber. Power to
the simulated heater was somewhat less than in the
real machine, implying that the effectiveness of the
machine heater was less than ideal. Radiative transfer to the powder bed surface was a relatively small
fraction (12%) of the power emitted by the heater, indicating that radiant heater configuration is inefficient
at transporting thermal energy to the powder bed surfaces. Differences between the simulated and actual
power consumption by the heaters at steady-state
implied that convective losses at the heater surfaces

and through the walls of the build chamber have a
substantial impact on the total power consumption of
the process.
In addition, examination of the convection pattern
that developed at the part bin surface indicated that the
formation of convective cells was an important source
of non-uniformity in the powder surface temperature.
REFERENCES
Deissler, R. and J. Boegli (1958). An investigation of effective
thermal conductivities of powders in various gases. Trans.
Amer. Soc. mech. Engrs, 1417.
Diller, T. T., R. Sreenivasan, J. Beaman, D. Bourell, and J.
LaRocco (2010, August). Thermal model of the build environment for polyamide powder selective laser sintering.
Twenty FirstAnnual International Solid Freeform Fabrication SymposiumAn Additive Manufacturing Conference.
Gusarov, A. V., T. Laoui, L. Froyen, and V. I. Titov (2003).
Contact thermal conductivity of a powder bed in selective
laser sintering. International Journal of Heat and Mass
Transfer 46(6), 11031109.
Gustafsson, S. (2009). Transient plane source techniques
for thermal conductivity and thermal diffusivity measurements of solid materials. Review of Scientific Instruments
62(3), 797804.
Gustafsson, S., B. Suleiman, and N. Saxena (1990). The transient plane source technique:experimental design criteria.
High TemperaturesHigh Pressures (UK) 23(3), 289293.
Gustavsson, M., E. Karawacki, and S. Gustafsson (2009).
Thermal conductivity, thermal diffusivity, and specific
heat of thin samples from transient measurements with
hot disk sensors. Review of Scientific Instruments 65 (12),
38563859.
Hall, R. and D. Martin (1981). The thermal conductivity
of powder beds. a model, some measurements on UO2
vibro-compacted microspheres, and their correlation.
Journal of Nuclear Materials 101(12), 172183.
Maqsood, A., N. Amin, M. Maqsood, G. Shabbir,
A. Mahmood, and S. Gustafsson (1994). Simultaneous
measurements of thermal conductivity and thermal diffusivity of insulators, fluids and conductors using the transient plane source (tps) technique. International Journal
of Energy Research 18(9), 777782.
Mark, J. (2007). Physical properties of polymers handbook.
Springer.
Tsotsas, E. and H. Martin (1987). Thermal conductivity
of packed beds: a review. Chemical Engineering and
Processing 22(1), 1937.
Xue, S. S. and J. Barlow (1990). Thermal properties of
powders. In Solid Freeform Fabrication Proceedings,
pp. 179185.
373

Metallurgical, accuracy and cost analysis of Ti6Al4V dental coping

fabricated by electron beam melting process
W.P. Syam, A.M. Al-Ahmari & M.A. Mannan
Princess Fatimah Alnijriss Research Chair for Advance Manufacturing Technology, King Saud University,
Saudi Arabia
Huda A. Al-Shehri & Khalid A. Al-Wazzan

Department of Prosthetic Dental Science, College of Dentistry, King Saud University, Saudi Arabia
ABSTRACT: The design and fabrication processes of a metallic coping for metal-ceramic crown (MCC)
restoration along with the accuracy, metallurgical analyses and the cost analysis are presented in this paper.
The metallurgical analysis exhibited that a 100% dense coping could be obtained by EBM fabrication. MicroCT scanning for accuracy analysis of complex shape parts fabricated by EBM process has been found to be an
efficient quality assurance method. The fabrication of a large number of coping in a single EBM fabrication
cycle makes this process competitive to the conventional process when complex dental prostheses and medical
implants of biocompatible material such as Ti6Al4V is considered.
1
INTRODUCTION
There are an increasing number of applications of additive manufacturing (AM) and rapid prototyping (RP)
in dentistry and medicine. In these fields, the parts
often have intricate shape, are highly customized and
low production volume is another significant characteristic for these parts. AM methods are the best choice
to fabricate these types of parts. In AM, material is
added in a layer-by-layer fashion following the 2D
sliced contour of a large number of 2D cross section
of a 3D object. The 3D model is in Stereolithography (STL) format, which is the de facto format for
layered manufacturing (Chen et al., 1999). Liu et al.
(2006) reviewed the advantages of AM methods in
dentistry. They showed that a number of AM methods can be used for several applications in dentistry
such as fabrication of dental devices, visualization
and diagnostic tools, surgical planning, customized
implant design, dental prosthetic, etc. AM methods are
closely related to reverse engineering technique (RE).
RE is a method to reconstruct 3D model of existing
part from digitized point data (Iuliano and Minelota,
2009). Since, in medical and dental application, most
of the parts are designed from an existing human body
part or teeth. Computed Tomography (CT) Scan and
Magnetic Resonance Imaging (MRI) are commonly
used in digitization process as well as laser scanning
process.
In prosthetic dentistry, one of the common methods
for dental restoration is metal-ceramic crown (MCC)
restoration as shown in Figure 1. In this method, there
are two layers that constitute the restoration restoration: first layer is a metal coping fitted on the prepared
tooth to support the artificial crown and the second

layer is the porcelain crown restoration fired on the
metal coping. The porcelain crown is a representation of the restored tooth. A biocompatible metal is
required for the metal coping fabrication. Titanium
alloy Ti6Al4V is undoubtedly the most common biocompatible material in the market (Polmear, 2006).
This alloy is a two-phase + titanium. Ti6Al4V
has many advantages such as high strength-to-weight
ratio, high corrosion resistance, high ductility, high
integration with bone and tissue. These characteristics
make Ti6Al4V compatible with human body. On the
other hand, processing of Ti6Al4V parts using conventional machining and casting is rather difficult.
The reasons are:
1. This alloy has low conductivity and high hardness.
Hence, it is difficult to machine.
2. Eliopoulos et al. (2004) observed porosities after
casting of titanium due to rapid oxidation with
oxygen in the process.
3. According to American Dental Association (2003),
the casting of titanium and its alloys needs a special vacuum-casting machine and a special cooling
cycle to avoid rapid oxidation.
AM machines based on selective laser sintering/
melting (SLS/M) have been predominantly applied
during the last decade for fabricating the parts using
metallic powder including Ti6Al4V. Laser beam is
used in SLM as a heat source.There are some studies of
SLS/M in dentistry. By using SLM, a dental framework
of Ti6Al4V had been successfully fabricated by Vandenbroucke and Kruth (2007). A satisfying marginal
and internal fit accuracy of metal coping fabricated
375
Figure 1. Schematic presentation of MCC restoration.
from cobalt chromium by SLM had been obtained by

Quante et al. (2008). Similar results were also obtained
by Ucar et al. (2009). They used SLS process instead
of SLM process.
Yang, et al. (2008) obtained satisfying quality of
NiTi parts by SLM process. They had selected an
appropriate laser mode and scanning strategy for
melting process by the laser. A high strength of
solid Ti6Al4V specimens had been produced by SLM
(Facchini et al., 2010). Beside the satisfying results,
some drawbacks have been reported in the fabricated
SLS/M parts of titanium and its alloys. Gu, et al. (2010)
reported the observation of thermal cracks in the final
part. Mercelis and Kruth (2006) observed high residual stress in the final parts. Both of them are due to high
thermal gradient since no preheating is implemented
in SLM process.
There is a relatively new AM process that is electron
beam melting (EBM). In this paper, fabrication results
of metal coping for MCC restoration by using EBM
are presented. The accuracy of the fabricated coping in term of volumetric gap between the fabricated
metal coping and the prepared die has been measured
using a micro-CT scanner and the microstructure of the
coping is studied through metallographictest. Finally,
cost analysis is presented tojustify the usage of EBM
for MCC restoration.
2
2.1

Electron beam melting
Electron beam is the heat source for melting process

in an EBM system. The electron beam has a very
high power density (Einstein et al., 1963). The beam is
resulted from accelerated electrons between two electrodes at high potential difference. Magnetic lenses are
used to focus, deflect, and control the electron beam.
Hence, a very fast scanning process can be achieved.
Thanks to high power density and very fast scanning speed, fabrication process is much faster in EBM
than SLS/M processes. Furthermore, the implementation of powder layer preheating improves the process.
Thus, thermal gradient within the part being fabricated is reduced. Since an EBM process takes place
in a chamber, the cooling rate can be controlled and
residual stresses in the part can be minimized. EBM
process is carried out in a near vacuum chamber at
2 103 mbar in order to minimize rapid oxidation.
ARCAM EBM Model A2 from ARCAM AB, Sweden
Figure 2. A schematic view of electron beam melting

process.
Table 1. Ti6Al4V powder composition (source:ArcamAB).
Element Al
%
Fe
Ti
6.05 3.99 0.006 0.056 0.13 0.009 0.002 balanced
was used in this study. Facchini et al. (2009) produced

specimen of Ti6Al4V by EBM and found that the specimen properties satisfied the requirement. Murr et al.
(2009) obtained similar or even superior mechanical
properties of Ti6Al4V parts in terms of tensile strength
and hardness. Figure 2 shows a schematic view of an
EBM system.
2.2 Ti6Al4V material
Ti6Al4V metallic powder used in our investigation
had grain size ranging from 45 to 100 m. Detailed
chemical composition of supplied Ti6Al4V powder
is shown in Table 1 and its properties are listed in
Table 2.
2.3 Metallographic test and accuracy measurement
Metallographic test and analysis were conducted to
study the microstructural properties of the specimen
produced by EBM. For micro structure observation light optical microscope (LOM) series BX51M
from Olympus and Scanning Electron Microscope (SEM) series JEOL JSM-6610LV were used.
376
Table 2. Ti6Al4V powder properties (source: Arcam AB).

Properties
Value
Yield Strength
UTS
Elongation
Reduction of Area
Fatigue Strength @600 Mpa
Rockwell Hardness
Modulus of Elasticity
Specific Density
Apparent Density
Melting Point Range
Boiling Point
Appearance
950 MPa
1020 MPa
14%
40%
>10,000,000 cyc.
33 HRC
120 GPa
approx. 4400 kg/m3
approx. 2500 kg/m3
1500 C to 1670 C
3287 C
Silver-Grey
X-Ray Diffractometer (XRD) was used to obtain elements composition spectra for the powder before
melting process and for the part after melting process to qualitatively study the chemical composition.
Shimadzu XRD-7000 was used for this purpose.
In addition, Energy Dispersive X-Ray (EDX) measurements were conducted to quantify the elements
composition in the fabricated metal coping.
A micro-CT scanner from Skyscan was used to measure the volumetric gap between the coping and the die.
The Measurement of accuracy of a metal coping by
micro-CT is relatively new method (Pelekanos et al.,
2009). The advantages of using micro-CT scanning
are the better accuracy of measurement and its nondestructive nature. The micro-CT scanner scans the
part and produces a large numbers of 2D sliced files
for each thickness which are predetermined before
the process. The 2D sliced files together reconstruct
a 3D image of the scanned part.
COPING DESIGN
During the designing processes, two different software

Magics 13 and Geomagic 11 were used to obtain the
3D model of the metal coping in STL format. Figure 3
presents the design steps of the STL 3D model.
The steps to obtain the 3D solid model of the coping
for fabrication are as follows:
1. A model of a coping die was obtained in STL format by reconstruction of digitized points (point
cloud) data probed by laser scanning process. Subsequently, a surface model of prepared die was
generated based upon the point cloud. The laser
scanner used was Dental Wing Scanner (Dental
Wing, Canada). This scanner is a specialized noncontact 3D measuring device for dental application.
2. In Magics 13, the surface model in STL format was
modified to remove unnecessary region of the surface model of the coping die so that a clean surface
model of the die could be obtained for the coping
design. Subsequently, from the STL model of the
coping die, sharpen operation in Geomagic 11 was
Figure 3. Preparation of coping 3D CAD model in STL

format.
used to extract the marginal line. NURBS surface of

the die was created after sharpening process. STL
file was generated from this NURBS surface model
and imported back to Magic 13. Finally, shell operation of Geomagic 11 was applied to add thickness
of 0.5 mm. Thus, a complete 3D solid model of
the coping was obtained. A file in STL format was
extracted from the solid model.
3. Furthermore, the resulted 3D solid model was
scaled up for shrinkage compensation and prepared for fabrication in Magics 13. In addition, STL
error check and repair of the coping model were
deployed. After model orientation and positioning
on the base plate, the model was imported to EBM
machine for fabrication.
4
4.1

Fabrication result
The fabrication process in EBM used process parameter from the previous study by Syam et. al (2011,
in press). In their study, scale up factors to compensate the shrinkage in solidification process and
process parameter of EBM have been discussed. In
their results, the scale up factor used for fabrication
was 1.06215 in X-Y-Z directions. After classifying and
determining range of parameters, they selected these
parameters within the specified range by usingTaguchi
orthogonal array method. The selected parameters
consisted of three factors and four levels each. Signalto-noise (S/N) ratio analysis was applied to select the
best parameters within the range. Subsequently,
the process parameter used for fabrication by EBM
377
Figure 4. EBM-manufactured of coping before polishing

(a) outer view, (b) inside view, (c) inside view after polishing,
(d) Assembly view of coping on prototyped-plastic die.
Figure 7. SEM image of etched surface on (a) position 1

side and (b) position 2 side, EDX analysis on (c) North side
and (d) South side.
Figure 5. Location of Microstructure analysis.
Figure 6. Microstructure image on (a) position 1 side and

(b) position 2 side.
were 800 mm/s beam scanning speed, 3 mA beam current, and 15 mA focus offset.After fabrication by EBM
process, manual polishing process was carried out to
reduce the surface roughness for inner surface since
this surface is the most important region of the metal
coping. The fabrication and polishing result are shown
in Figure 4.
4.2
Metallographic analysis
Two locations were selected for metallographic analysis of the fabricated coping. These locations were on
two diametrically opposite sides as shown in Figure 5.
Microstructure images as observed by LOM are
shown in Figure 6. It can be seen from this figure that
the fabricated parts were 100% dense and no porosities existed at either of the position 1 and position 2.
Basket-weave microstructures of Ti6Al4V ( + Titanium) having and phases were observed which
can be attributed to the slow cooling rate. In these
optical microscope images, -phase appeared light
(brown color) and -phase appeared dark.
Figures 7a and 7b show SEM images on position 1
and position 2 respectively. SEM images also confirmed the non-existence of porosity in the fabricated
part. In SEM images, -phase of Ti6Al4V appeared
dark and -phase of Ti6Al4V appeared light (white).
This is because the -phase was etched over -phase
Figure 8. XRD spectra for (a) the initial Ti6Al4V powder

and (b) EBM-melted Ti6Al4V.
that makes topographical difference between -phase

and -phase.
EDX analyses were conducted to verify the amount
of element composition on both locations. Figures 7c
and 7d show the EDX analysis results on position 1
and position 2. In the results, both locations show that
main elements, Ti, Al, and V, have dominant weight
percentage. Total element composition on position 1
of Ti, Al, and V was 82.85%, 5.84%, and 3.51%
respectively and 81.91%, 5.84%, and 3.49% on position 2. Polluter elements, such as Carbon (C), Natrium
(Na), Chlorin (Cl), and Fluorine (F), were not significant. They were observed due to excess of Krolls
etchant and particles of surrounding atmosphere. Oxygen (O2) element was not observed in the EDX spectra.
378
Table 3.
Material, EBM process, design, engineering, labor, and post-processing cost.
No.
Cost Type
1.
Material:
Raw powder Ti6Al4V price = US$295/Kg = US$ 0.295/gram.
Density: 4430 Kg/m3 = 0.00443 gram/mm3 .
Approximate model Volume = 168.806 mm3 .
Approximate mass of the coping (volume density) = 168.806 0.00443 gram
= 0.7478 gram. To consider waste powder, then the value is doubled.
Approximate material cost = 0.7478 gram 2 $0.295/gram = US$ 0.44
EBM Electricity:
a. Process time estimation
Approximate time to obtain required level of vacuum = 30 min = 0.5 h
Approximate start plate heating time = 30 min = 0.5 h
Approximate build time = 35 min = 0.583 h
Build time calculation for one stack (64) copings:
coping height 7.5 mm
layer thickness = 0.07 mm number of layer = 7.5/0.07 = 108 layers
layer time (raking + preheating + melting) 20 s
total build time = 35 min = 0.583 h
Approximate cooling time = 3 h (fixed for any number of parts fabricated)
Approximate total time = 4.583 h
b. Process cost calculation
EBM power consumption = 7000 VA
Energy consumption = 7 4.583 = 32.081 kWh
Approximate EBM process cost = 32.081 kWh SAR 0.26/kWh
= SAR 8.341
= US$ 2.22
Note: electricity tariff = SAR 0. 26/KWH (source: www.se.com.sa)
Design of Metal Coping:
Man-Hour for designing in Magic 13 = 1 h 1 person = 1 MH
Man-Hour in designing in Geomagic 11 = 0.5 h 1 person = 0.5 MH
Total MH = 1 (1.5 +1) = 1.5 MH
Cost per MH = SAR 20
Total Design Cost = SAR 20/MH 1.5 MH = SAR 30 = US$ 7.98
Engineering cost to prepare a STL file:
Man-Hour in design modifying in Geomagic 11 = 1 h 1 person = 1 MH
Man-Hour in preparing the file in Magics and EBM Build Assembler = 1 h 1
person = 1 MH
Total MH = 1 (1+1) = 2 MH
Total Engineering Cost = SA 30/MH 2 MH = SAR 60 = US$ 16
Labor:
Man-Hour for preparing (load) EBM machine = 1.5 h 1 person = 1.5 MH
Man-Hour for unloading (include PRS process for part) EBM machine = 1.5 h 1
person = 1.5 MH
Total MH = 1 (1.5+1.5) = 3 MH
Total Engineering Cost = SA 15/MH 3 MH = SAR 45 = US$ 11.97
Post-Processing (Polishing):
a. Polishing MH cost calculation
Number of employee needed = 1 person.
Total processing time = 15 minutes = 0.25 hours
Total Man-Hour (MH) needed= 1 0.25 = 0.25 MH
Cost/MH = SAR 20
Then, total MH cost = 0.25 MH 20 SAR/MH = SAR 5 = US$ 1.33
b. Polishing cost
Machining cost for polishing = SAR 0.429 = US$ 0.15
Then, total Polishing cost = US$ 1.33 + US$ 0.15 = US$ 1.48
2.
3.
4.
5.
6.
Amount in US$
Presence of oxygen is a disadvantage since it indicates

the generation of Titanium-Oxide in form of TiO2 .
A 2 mm thick, 20 mm wide and 20 mm long plate
was fabricated with identical process parameters as for
0.44
2.22
7.98
16
11.97
1.48
the metal coping. Subsequently, XRD analyses were

conducted for Ti6Al4V powder and the EBM fabricated plate. Figure 8 shows the XRD spectra of powder
and the melted specimen of Ti6Al4V. The material
379
Table 4.
Calculation of cost of spare parts.
No.
Part
Price
Maintenance Cycle/Requirement Value
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
Start plate 150 mm 150 mm

$95
400 hrs
Filament F120, 1 pc
$80.80
100 hrs
Rake Blades Long 370, 4 pcs
$62.70
100 hrs
Observe Window glass Coated, 1pcs
$64
400 hrs
Protection foil column, 1 pcs
$4.5
200 hrs
Ground ring for protection foil
$46
200 hrs
Thermo couple, 0.15 m
$4.30
100 hrs
Ground cable, 0.15 m
$2.50
100 hrs
Thermo plate
$28
400 hrs
Latex gloves, 100 pcs by weight
$7
2 pcs/running
EBM Maintenance
Man-Hour for maintenance of EBM machine = 1 h 1 person = 1 MH
Total Engineering Cost = SA 15/MH 1 MH = SAR 15 = US$ 4
Total
Table 5. Calculation of EBM machine ownership cost

(Equipment cost).
Cost Type
0.26
0.87
0.68
0.17
0.03
0.25
0.05
0.03
0.08
0.14
4
6.560
Table 6. Total fabrication cost summary.

No.
Cost Type
1.
2.
3.
Material
EBM Electricity
EBM Equipment cost/coping
(when a stack of 64 copings
is fabricated)
Metal coping design cost
Engineering of design
Labor
Maintenance & Spare parts
Post-Processing (Polishing)
Total Cost
Amount in US$
Cost in US$/hr
EBM Ownership (Investment) 22.983

Machine price = US$ 700000
Lifespan or Depreciation = 10 Years
Annual Depreciation = US$700000/10 year= US$70000/year
Basic maintenance and service cost = US$ 3360/month
= US$ 40320/year.
Annual Ownership cost (yearly depreciation + annual basic
maintenance and service contract)
= (US$ 70000 + US$ 40320)/year
= US$ 110320/year
Assumption: effective working days/year = 300 days
Ownership cost per day = US$ 110320/300 days
= US$ 367.73/day
Assumption: time used for machine operation is 2/3, and time
used for load-unload (preparation) of the machine is 1/3.
Ownership cost/hr = US$ 367.73/16 hrs = US$ 22.98/hr
composition of melted plate specimen is almost identical to Ti6Al4V powder before melting except for one
new peak of -Ti. In the XRD spectra of both powder and melted specimen, peak corresponding to TiO2
was not observed which indicates no reaction between
titanium and oxygen.
4.3
Cost/coping
build (1.083 hrs:
Start Plate Heating +
Build Time) in US$
Listing and numbering
The objective of this analysis is to have a basic understanding of different costs involved in the process and
conduct a comparison between fabrication costs of
metal coping fabricated by EBM process with that of
commercially produced one. The definitions of each
cost are as follow:
1. Material cost is the cost of Ti6Al4V powder used
for one metal coping. In order to take into account
the wastage of powder during handling, the amount
of powder used for one metal coping is doubled.
4.
5.
6.
7.
8.
0.44
2.22
1.65
7.98
16.00
11.97
6.56
1.48
48.30
2. EBM electricity cost is the electricity cost of how

long the EBM machine is switched on to produce
one stack of metal copings. One stack consists of
64 metal copings.
3. Metal coping design cost is the cost of total time
needed in designing one metal coping in the Magic
13 and Geomagic 11 to prepare master die model
and coping surface model. Man-Hour (MH) tariff is
calculated based on the working tariff for one hour
of a diploma holder.
4. Engineering cost. A STL file had to be prepared
from the 3D model of the designed coping and it
requires certain amount of engineering. Thus, the
engineering cost includes the salary of an engineer
to perform these tasks.
5. Labor cost is the cost of time needed to prepare,
start, load and unload the EBM machine. Labor cost
does not include cost of time needed to maintain
the EBM machine.
6. EBM maintenance and spare parts costs are the cost
of consumable parts of EBM machine that have to
be changed after certain period of operation hours
and the cost to maintain the EBM machine.
380
Figure 9.
Pie-chart of fabrication cost with different
contributing elements.
Table 7. Cost reduction by increasing the number of copings
per fabrication.
Numbers of Coping
Produced
32
64
128
192
384
640
1280
1984
Cost (US$)/coping
The cost: Material +
EBM Process (Electricity
and Investment cost)
3.80
2.12
1.39
1.13
0.89
0.8
0.73
0.7
Figure 10.
ume.
7. Post-processing (polishing) cost is the cost to polish the metal coping to reduce the inner surface
roughness of the metal coping.
8. EBM ownership cost (equipment cost) is the cost of
investment of the machine including the machine
depreciation per annum and basic maintenance and
service cost per year.
Table 3 shows the detailed cost calculation for
material, EBM process, design, engineering, labor,
and post-processing costs. Detailed cost calculation for EBM maintenance and spare parts costs is
shown in Table 4. Table 5 shows calculation of ownership (investment) cost. Finally, Table 6 summarizes
all the costs needed for fabrication of one metal coping by EBM process. Total cost for one metal coping
by EBM is US$ 48.30. As a rough comparison, the
price of commercial one metal coping is US$ 187.2.
Figure 9 presents a pie-chart of fabrication cost with
different contributing elements for one metal coping.
The cost for one metal coping that has been shown
was calculated under the assumption that a single stack
of copings (64 copings/stack) was produced in a single fabrication cycle. Usually, the total build volume
(150 150 350) is utilized to minimize the cost. By
using base plate of 150 mm 150 mm and chamber
height of 350 mm, as many as 31 stacks (having 64
Numbers of coping and stack in the build vol-
copings/each) can be placed on top of each other.

Hence, total number of copings that can be produced
in one fabrication cycle are 1984 copings. The gap
between the copings in each stack and the gap between
the stacks for clearance have already been considered.
Figure 10 shows the coping and the stacks in the build
volume.
Hence, the EBM process cost and material cost
per coping can be significantly reduced by producing
many coping in one fabrication cycle. To illustrate this
cost reduction, the cost calculation to produce 64 copings (1 stack) in one fabrication is presented. The cost
includes EBM Process cost (electricity cost), Ownership cost, and material cost. The calculation is as
follow:
1. Total Build time needed:
381
Start pump to start plate heating = 0.5 hr

Start plate heating time = 0.5 hr
Build time = 1 stack 35 = 35 minutes = 0.583 hrs
(The build time for one coping and 64 copings per
layer is identical since EBM capable to melt the
parts simultaneously. Thanks to the multi beam
feature of EBM).
Cooling time = 3 hrs
Total time = 4.583 hrs
Figure 11. (a) Result of Micro-CT scan of the metal coping and the prepared die and (b) marginal and internal gap.
2. Total electricity cost

= 4.583 7 KW 0.26 SR/KWH = SR 8.34
= US$ 2.22
3. Total ownership cost
= 4.583 US$22.983/hr = US$105.33
4. Total material cost (including waste material)
= 64 (0.7478 2) US$ 0.295/gram
= US$ 28.24
5. Total = US$ 135.79/64 copings
= US$ 2.12/coping.
With similar calculation, the cost reduction by
increasing the number of copings per a single fabrication cycle was calculated and is presented in Table 7.
Subsequently, the EBM process cost is competitive
with other processes.
4.4 Accuracy measurement
The accuracy of metal coping is represented either
by marginal and internal gaps or by volumetric gap
between the metal coping and the prepared die.
The lower the value means the better the accuracy.
A certain amount of clearance, which is called die

spacing, between the prepared die and the metal coping is needed. The clearance required by the dentist
is about 100 m. The purpose of this clearance is to
provide space for cementing the metal coping on the
prepared die. In our analysis, we used volumetric gap
measurement since this can represent the overall gap.
The measurement of the volumetric gap is conducted
in a micro-CT scanner. Figure 11a presents the measurement results of volumetric gap. In this figure, the
prepared die appears in white color. The material of
the prepared die is stainless steel. The metal coping is
in grey color. The function of the V-groove is to fix
the metal coping position on the prepared die.
The resulted volumetric gap was 50.797 mm3 . This
volumetric gap includes the marginal and internal gaps
as well (Figure 11b). Marginal gap in MCC restoration
is the gap at most coronal position of the continual
line of origin at or near gum line. Internal gap is the
gap at the inner surface of the coping. For comparison, the volumetric gap for metal coping produced by
standardized casting process is 13.807 mm3 . In this
382
result, volumetric gap of metal coping produced by

EBM is still larger than the one produced by casting method. The volumetric gap difference is about
36.99 mm3 .
The main factor that contributes to higher volumetric gap is manual polishing process. In manual
polishing, the amount of material of the inner surface removed can not be controlled. Obviously, amount
of material larger than the desired value has been
removed. Then, an accurate computer controlled finishing process, such as CNC 5-axis milling/grinding
can be the option for the finishing process so that
roughness can be removed as desired and the amount
of material thickness removed can be controlled.
The other contributors for higher volumetric gap
are the error in reverse engineering phase to reconstruct the solid coping model and the error of the model
of scaling up factors for shrinkage compensation
applied to the design for fabrication.
5
CONCLUSION
In this study, EBM process was used to fabricate

metal coping of Ti6Al4V for MCC restoration.
Before fabrication, metal coping design processes
were carried out. The solid model of metal coping in STL format was reconstructed from laser
scanned digitized point cloud data. After scaling
up the model and with apriori optimized process parameters, the metal coping was fabricated
by EBM. In the LOM and SEM microstructure
images, no porosity was observed. Oxide bonding
in form of TiO2 was not detected in XRD analysis.
The cost analysis clearly show that though the initial
cost of acquisition of an EBM system is significantly
larger than the equipment for conventional fabrication,
the fabrication of a very large number of parts in a single fabrication cycle reduces the equipment cost/part
remarkably and makes the EBM process very competitive. Finally, volumetric gap between the metal coping
and prepared die was measured by a micro-CT scanner. The volumetric gap obtained was 50.797 mm3 .
This value was still higher than the one obtained for
a coping fabricated by the standard casting method.
The main reason is due to uncontrolled manual polishing process of the inner surface. Hence, an accurate
CNC 5-axis milling/grinding machine is proposed to
be used as for finishing process.
REFERENCES
American Dental Association Council on Scientific Affairs.
2003. Titanium applications in dentistry. The Journal of
The American Dental Association 134(3): 347349.
Chen,Y. H., Ng, C.T., Wang,Y. Z. 1999. Generation of an STL
File from 3D Measurement Data with User-Controlled
Data Reduction. International Journal of Advance Manufacturing Technology 15: 127131.
Einstein, P. A., Harvey, D. R., Simmons, P. J. 1963. The design
of an experimental electron beam machine. Journal of
Science and Instrument 40: 562567.
Eliopoulos, D., Zinelis, S., Papadopoulos, T. 2004. Porosity of cpTi casting with four different casting machines.
The Journal of Prosthetic Dentistry 92(4): 377381.
Facchini, L., Magalini, E., Molinari, A., 2009. Microstructure and mechanical properties of Ti6Al4V produced
by electron beam melting of pre-alloyed powder. Rapid
Facchini, L., Magalini, E., Robotti, P., Molinari, A., Hoges,
S., Wissenbach, K. 2010. Ductility of a Ti-6Al-4V alloy
produced by selective laser melting of prealloyed powders.
Gu, D., Hagedorn, Y. C., Meiners, W., Wissenbach, K.,
Po-prawe, R. 2011. Selective laser melting of in-situ
TiC/Ti5Si3 composites with novel reinforcement architecture and elevated performance. Surf and Coatings
Technology 205(10): 32853292.
Iuliano, L., Minetola, P. 2009. Enhancing moulds manufacturing by means of reverse engineering. International Journal of Advance Manufacturing Technology
43: 551562.
Liu, Q., Leu, M. C., Schmitt, S. M. 2006. Rapid prototyping in
dentistry: technology and application. International Journal of Advance Manufacturing Technology 29: 317335.
Mercelis, P., Kruth, J. P. 2006. Residual stresses in selective laser sintering and selective laser melting. Rapid
Murr, L. E., Esquivel, E. V., Quinones, S. A., Gaytan, S. M.,
Lopez, M. I., Martinez, E. Y., Medina, F., Hernandez,
D. H., Martinez, E., Martinez, J. L., Stafford, S. W.,
Brown, D. K., Hoppe, T., Meyers, W., Lindhe, U., Wicker,
R. B. 2009. Microstructures and mechanical properties
of electron beam rapid manufactured Ti-6Al-4V biomedical prototypes compared to wrought Ti-6Al-4V. Materials
Characterization 60: 96105.
Pelekanos, S., Koumanou, M., Koutayas, S. O., Zinelis, S.,
Eli-ades, G. 2009. Micro-CT Evaluation of the Marginal
Fit of Different in-CeramAlumina Copings.The European
Journal of Esthetic Dentistry 4(3): 278292.
Polmear, I. J. 2006. Light Alloys: From Traditional Alloys to
Nanocrystal. Butterworth-Heinemann: United Kingdom.
Quante, K., Ludwig, K., Kern, M. 2008. Marginal and
Internal Fit of Metal-Ceramic Crowns Fabricated with
A New Laser Melting Technology. Dental Materials 24:
13111315.
Sercombe, T., Jones, N., Day, R., Kop, A. 2008. Heat treatment of Ti-6Al-7Nb components produced by selective
laser melting. Rapid Prototyping Journal 14(5): 300304.
Syam, W. P., Al-Shehri, H. A., Al-Ahmari, A. M., Al-Wazzan,
K. A., Mannan, M. A. 2011. Preliminary fabrication
of thin-wall structure of Ti6Al4V for dental restoration
by electron beam melting. Rapid Prototyping Journal,
Accepted, In Press.
Ucar, Y., Akova, T., Akyil, M., Brantley, W. A. 2009.
Internal Fit Evaluation of Crowns Prepared Using A
New Dental Crown Fabrication Technique: Laser-Sintered
Co-Cr Crowns. The Journal of Prosthetic Dentistry
102(4): 254259.
Vandenbroucke, B., Kruth, J. P. 2007. Selective laser melting of biocompatible metals for rapid manufacturing of
medical parts. Rapid Prototyping Journal 13(4): 196203.
Yang, Y., Huang, Y., Wu, W. 2008. One-step shaping of
NiTi biomaterial by selective laser melting. in Lasers in
Material Processing and Manufacturing III proceeding
of the international conference in Beijing, China, 1214
November 2007, SPIE, Bellingham, Vol. 6825.
383

Thermal and dynamic-mechanical behavior of Fullcure 3D Printing Resin

post-cured by different methods
L.F. Vieira, R.A. Paggi & G.V. Salmoria
Laboratrio CIMJECT, Department of Engenharia Mecnica, Universidade Federal de Santa Catarina,
ABSTRACT: The Fullcure720 resin is an epoxy-acrylic photopolymer used in rapid prototyping (Objet Polyjet
system). This research investigates the mechanical and optical properties of Fullcure720 specimens submitted
to two different post-cure treatments: by heating in a thermal oven and exposure in a UV chamber. Dynamical
mechanical analysis (DMA) was employed to monitoring the post-cure effect in the elastic modulus and in the
transition temperature. The yellowing of the samples was measured by spectrophotometer using the b* parameter
in the CIELAB color space, and the sample warping was observed. The post-cure treatments caused a significant
increase in elastic modulus and in the glass transition temperature. The thermal post-cure resulted in sample
warping and yellowing, what not occur in the UV post-cure.
INTRODUCTION
they might have the mechanical properties modified

by post-cure treatments.
The current consumer goods market is characterized

by intense competition and reduced product-life cycle.
The agility and quality in the design of new products
are vital to corporate performance. In most cases, the
pioneer in launching can take benefits for longer and
wider profit margins. This can be aided by the use
of Rapid Prototyping (RP). This technology is capable to produce objects directly of a three-dimensional
computer model. The evaluation of a prototype aids in
decision-making process and in the problems identification, especially with those reported to aesthetics and
ergonomics issues.
One problem with these techniques is the need to
purchase different types of materials, or even different
RP systems, for the construction of prototypes with
different properties. An alternative to this limitation is
the post-cure process, which modifies the properties
of the prototype for example, more or less flexible
without the need of large investments.
Most studies of post-cure treatments with rapid
prototyping resins focus in those used in stereolithography (SL). Colton and Blair [1] investigated the effect
of UV chamber post-cure and a conventional oven;
Salmoria et al [2], investigated these two treatments
and also the microwave oven treatment. Nevertheless
it is important to mention that the post-cure in samples
obtained by SL is often mandatory, because the part
is submerged in the liquid resin, resulting in an tacky
surface [3] which makes the handling dangerous.
FullCure 720 resin is used in the rapid prototyping
process by PolyJet rapid prototyping system (Objet
Geometries Ltd.). Due to the chemical similarity with
the resins used in stereolithography, it is expected that
2.1 Resin FullCure 720 and the PolyJet system

The fullcure 720 is a translucent resin, epoxy-acrylic
based, that supports up to 20% strain and can be handled directly after construction. The PolyJet system
consists of a print head that applies a photo-curable
resin layer with a UV lamp that cures each layer
immediately after its building. There are two modes
of construction according to the thickness of the layer:
HQ (High Quality) with 16 m and HS (High Speed)
with a layer of 32 m.
There are not many studies with this system, since
it is not so traditional like others equipments as the
SL. However, there are some studies with this system
in reproductive surgery [68] due to the high detailed
reproduction and transparency of the prototypes, and
some studies with others applications [4,5].
2.2 Experimental sequence
The samples were manufactured in a Objet Eden 250
3D Printer/FullCure 720 Resin, in rectangular prisms
shape with dimensions of 3,5 5 1,45 mm. Samples
were constructed with both 16 m and 32 m thick
layer in order to check also the influence of this factor.
A view of the experimental sequence is shown in
Figure 1.
The study was performed according to a factorial
design, using as the response variable the flexural
modulus.
385
Figure 1. Representation of the experimental sequence.

Table 1.
Experimental design
This factorial study based the definition of the optimum parameters, desiring the greatest possible change
in the mechanical properties. Specimens manufactured with these optimal parameters are tested with
dynamic mechanical analysis (DMA) and fatigue test.
2.3
Post-cure process
The factorial design was performed according to

Table 1.
2.4
Mechanical tests
All mechanical tests were performed on a DMA Q800

(TA Instruments) in the single-cantilever mode, with
a loading rate of 2 N/min. The flexural modulus was
calculated as the secant line between 2 and 3% strain.
Dynamic mechanical analysis were performed at the
frequency of 1 Hz, in a temperature range between
30 and 130 C and maximum deformation of 0,5%.
Fatigue tests were performed at 30 C with a frequency
of 1 Hz, with 1500 m of maximum deflection. The
analysis of variance was performed with a significance
level () of 5%.
2.5
Colorimetric methods
Post-cure processes can change the color of the material, which is an undesired effect when the part has aesthetic requirements. The color variation was checked
by a Spectra-guide 45/0 spectrophotometer (BYKGardner), in the CIELab color space (condition CIE
D65/10 ) which expresses the color on three parameters: the lightness L* (0100), the a* parameter, from
green (negative values) to red (positive values) and the
b* parameter, from blue to yellow [9].
Two samples of each configuration were measured
using four measurements for each sample. Measurements were made on a white background with
parameters L* = 94.26 0.08, a* = b* = 0.02 0.09
2.17 0.08.
3
Figure 2. Pareto chart (a) e response surface (b) for flexural

modulus Thermal post-cure.
The most influential factor is the post-cure temperature. The quadratic factor of post-cure temperature
indicates that this influence is not so proportional.
Figure 2 (b) shows a response surface estimated, the
regression coefficient R2 of 99.12% indicates a good
fit model.
3.2 Factorial Study: UV Post-cure
Figure 3 (a) shows the Pareto chart for standardized
effects in the study with UV post-cure. Likewise the
thermal post-cure, the energy supplied remains the
most significant factor (in this case exposure time).
Since the quadratic factor exposure time is significant, once again there is a non-proportional behavior.
However, in this case the response surface concavity
(Figure 3 (b)) is facing downwards, showing a saturation of the exposure time effect. The coefficient R2
of 98.73% again shows a good fit model.
3.3 Post-cure parameters optimization
From the factorial study are found the optimum
parameters, shown in Table 2.
The post-curing changed significantly the mechanical properties; controversy to what the resin commercial name suggests (FullCure).
3.1 A Factorial Study: Thermal Post-cure

The results of the analysis of variance are illustrated
in the Figure 2. The Pareto chart, in the figure 2 (a),
shows that three factors are statistically significant.
3.4 Evaluation of the visco-elastic properties and

fatigue behavior
Figure 4 (a) shows the storage modulus (E ) and the
Tan delta parameter as a function of temperature.
386
It was noted an increase in the values of E with

the post-cure. This increase in the values of E can
be associated with a higher degree of cure, resulting
in a more restricted molecular motion. The effect of
thermal post-cure is more prominent: the UV radiation comes from the activation energy for the curing
reaction, but the thermal agitation, besides this effect,
provides an enhanced mobility of the polymer chains
[2] increasing the likelihood of the reactive species to
react, thus increasing the cure amount.
In the three tan curves a peak related to the glass
transition temperature (tg) appeared. The treatment of
post-curing increased the tg temperature from 47 C
(without post-cure) to 62 C (UV) and 68 C (thermal). Some authors [12, 13] correlate the peak height
of tan with the impact resistance of the material.
This relationship indicates that the viscoelastic energy
dissipation factor is the dominant influence on the
properties of impact resistance [12]. The UV treatment
did not affect the peak height of tg, so the samples
tend to have the same tenacity as those without postcuring; although the thermal post-cure tends to cause
brittleness.
In all the cases, the stiffening provided by the postcure treatments was accompanied by a decrease in the
Figure 3. Pareto chart (a) e response surface (b) for flexural

modulus UV post-cure.
performance under fatigue, as shown in Figure 4 (b).

This effect was more pronounced in the thermal postcured sample.
3.5 Color changes and warping
In the Figure 5 it is shown the evolution of the b*
parameter of CIELab color space.
With the thermal post-cure the samples surface
became yellowed. On the other hand, in the UV cured
material, no visible color change appeared, with few
changes in the b* parameter. The thermal post-cure
also caused warping of the specimens, probably due to
the cooling; these effects are not found in the post-UV
cure. These differences are in agreement with studies
[13, 1] using other photo-sensitive resins.
Figure 4. Storage modulus (E ) and Tan (a) and fatigue

test (b) for optimized post-cured samples.
Table 2. Flexural modulus comparison for optimizided post-cured and no post cured
material.
Post-cure Treatment
Flexural modulus [MPa]
Enhancing [%]
Without post-cure (16 m layer)

Thermal post-cure: 180 C, 32 m layer
UV post-cure 10 h, 16 m layer
649,6 14,91
1130,5 19,9
920,3 1,34
74,04
41,67
387
Figure 5. CIELab b* colorimetric parameter measurements.
CONCLUSION
The post-cure treatments promoted an increase in the

flexural modulus. There were not found directly influences of the layer thickness. The analysis showed
higher glass transition temperature for the post-cured
material, and indicated a possible decrease of toughness with the thermal post-cure. This phenomena was
not noted for UV post-cure. The treatments, even
though increased the samples stiffness, decreased the
performance under fatigue. The thermal post-cure also
resulted in sample warping and yellowing, whereas it
did not occur for the UV post-cured samples.
ACKNOWLEDGEMENTS
support.
REFERENCES
Colton, J.; Blair B. Experimental study of post-build cure
of stereolithography polymers for injection molds. Rapid
Prototyping Journal v.5, p. 7281, 1999.
Salmoria, G.V.; Ahrens, C.H.; Fredel, M.; Soldi, V.; Pires,
A.T.N. Stereolithography somos 7110 resin: mechanical
behavior and fractography of parts post-cured by different
methods. Polymer Testing, v.24, p. 157162, 2005.
Cheah, C.M.; Fuh, J.Y.H.; Nee, A.Y.C.; Lu, L.; Choo, Y.S.;
Miyazawa, T. Characteristics of photopolymeric material
used inprototypes. Part II. Mechanical properties at postcured state. Journal of Materials Processing Technology,
v.67, p. 4649, 1997.
Zhu, H.; Zhang, B.; Bewer, B.; Popescu, F.G.; Nichol, H.;
Chapman, D. Field flatteners fabricated with a rapid prototyper for K-edge subtraction imaging of small animals.
Nuclear Instruments and Methods in Physics Research A,
v.588, p. 442447, 2008.
Pilipovic, A.; Raos, P.; ercer, M. Experimental analysis of
properties of materials for rapid prototyping. The International Journal of Advanced Manufacturing Technology,
v.40, p. 105115, 2009.
Ibrahim, D.; Broilo, T.; Heitz, C.; De Oliveira, M.; De
Oliveira, H.; Nobre, S.; Dos Santos Filho, J.; Silva,
D. Dimensional error of selective laser sintering, threedimensional printing and PolyJet models in the reproduction of mandibular anatomy. Journal of CranioMaxillofacial Surgery, v.37, 2009.
Faber, J.; Berto, P.M.; Quaresma, M. Rapid prototyping as
a tool for diagnosis and treatment planning for maxillary
canine impaction. American Journal of Orthodontics and
Dentofacial Orthopedics, v.129, n.4, p. 583589, 2006.
Salles, F. M.; Anchieta, P.C. Bezerra, M.L.G.M. Torres, E.
Queiroz, Faber, J. Complete and isolated congenital aglossia: case report and treatment of sequelae using rapid
prototyping models. Oral Surgery, Oral Medicine, Oral
Pathology, Oral Radiology & Endodontics, v.105, n.3,
p. 4147, 2008.
C.I.E. Colorimetrie. 2nd ed. Publication C.I.E. No. 15,2.
Viena: Central Bureau of the Commission Internationale
de LEclairage, 1986.
Nair, K.C.M.; Thomas, S.; Groeninckx, G. Thermal and
dynamic mechanical analysis of polystyrene composites
reinforced with short sisal fibres. Composites Science and
Technology, v.61, p. 25192529, 2001.
Sichina, W.J. Thermal Analysis for the Characterization of
Polymer Impact Resistance. Brochure Termal analysis
aplication note. Perkin Elmer, 2000.
Miyagawa, M.; Mohanty, A.K.; Misra, M.; Drzal, L.T.
Thermo-Physical and Impact Properties of Epoxy Containing Epoxidized Linseed Oil, 2. Macromolecular Materials and Engineering, v.289, p. 636641, 2004.
Karalekas, D.; Aggelopoulos A. Study of shrinkage strains
in a stereolithography cured acrylic photopolymer resin.
Journal of Materials Processing Technology, v.136,
p. 146150, 2003.
388

Rheological study of poly vinyl alcohol with two thermo-initiator

for use in rapid prototyping
L.L. Lima, C.B.L. Ulbrich & C.A.C. Zavaglia
State University of Campinas DEMa/FEM LABIOMEC National Institute of S&T in Biofabrication
(INCT-BIOFABRIS)
V.P. Bavaresco
State University of Campinas Deplas/CTC National Institute of S&T in Biofabrication (INCT-BIOFABRIS)
J.G.M. Coelho & C.G.B.T. Dias

Federal University of Par-National Institute of S&T in Biofabrication (INCT-BIOFABRIS)
ABSTRACT: This study evaluated the rheological behavior of poly vinyl alcohol (PVA) with two thermal
initiators for used in rapid protoyping (RP). The PVA has excellent biocompatible and lubrication properties,
it is a versatile biomaterial which presents studies for the use in the treatment, replacement of cartilage and
also as scaffold. The PVA is crosslinked to have different mechanical, chemical and physical properties. The RP
technology has been used and studied in medical field to production of customized medical devices for each
patient, and to production of biomodels for surgical plans. RP manufacturing technique allows the production of
products with high precision and a very fast production. The rheological study was made to determine the optimal
parameters for chemical crosslinking using the (PVA) aqueous solution in presence of potassium persulfate (KPS)
and 4,4 Azobis (4 Cyanovaleric acid) (Azobis) as thermo initiators.
INTRODUCTION
Hydrogels are a class of hydrophilic polymers containing covalent bonds, and weak bonds such as hydrogen
bonding, ionic and intermolecular bonds. This class is
able to retain a large amount of water without dissolving its structure. The hydrogel can be used in biomedical applications due to its elastic and hydrated nature
what minimizes irritation on the tissues (Kishida &
Ikada, 2002).
The PVA hydrogels has been use in the biomedical
and pharmaceutical industries due to: (a) biocompatibility, (b) the ability to swell and retain large quantities
of water or biological fluid, (c) the similarity to natural
tissue, (d) the low surface tension, (e) the high permeability of small molecules, (f) low toxicity, and (g) the
possibility of fabrication in many geometric shapes.
Due to this the PVA hydrogels are studied for use as
replacement and treatment of cartilage.
The cartilage has poor capacity for regeneration and
will not heal spontaneously because: (a) is avascularizated, (b) have characteristic aneural, (c) low cell
density, (d) low mitotic activity of cells, (e) and the
structural restriction that does not allow free migration of cells, therefore defects in cartilage represent a
major problem in orthopedics (Swieszkowski, Fray, &
Kurzydeowski, 2008).
One tool that can be used to produce of artificial
cartilage is the rapid prototyping (RP). The RP is a
technology that allows construction of complex models with internal details that could not be manufactured
by other manufacturing methods. The models are constructed from a virtual model obtained by 3D CAD,
computed tomography scan, and among others.The RP
has allowed construct the precise and detailed biomaterial to use in medicine, this method is characterized
by an additive process, where models are built layer
by layer (Souza & Ulbrich, 2009). The RP has been
developed to use in medicine because aims to produce
the specific devices implantable for each patient.
Wiria et al, 2008 produced scaffold to tissue
engineering (TE) using the PVA and hydroxyapatite
biomaterials.The three dimensional scaffold were constructed by Selective Laser Sintering (SLS) technique.
The result was favorable to process these biomaterials
by SLS, to produce scaffold to use in TE.
The PVA is crosslinked to have different mechanical, chemical and physical properties. The chemical
crosslinking in the hydrogel is the result of a chemical
bonding, what is irreversible at temperatures of human
body (Jagur-Grodzinski, 2010).
In this paper the rheological behavior was made
to determine the optimal parameters for chemical
crosslinking of PVA hydrogel with two thermo initiators. In the chemical crosslinking, functional groups
receiving radiation or crosslinker agent and form insoluble polymer chains. For the chemical crosslinking
is necessary that the polymer has reactive functional
389
groups, and requires functional crosslinker agents.

The chemical crosslinking is irreversible and does not
allow changes in the polymer after the reaction.
In the rheometer was used the scanning of temperature and the time to check the temperature and
the time that occur chemical crosslinking process. The
rheological analysis determines time and temperature
that occur change in the properties of material. The
rheology allows analyze the changes in the viscoelastic properties that occur in the material as a function
the application of stress and/or strain. The viscoelastic
properties in material depend of molecular structure.
In this paper the PVA was studied in different
compositions to observe the influence of water on
the behavior of chemical crosslinking process. In this
study was possible to observe the difference in temperature of transitions, and the values modulus using
Azobis or KPS as thermal initiator, what can confirmed that there are differences in the kinetics of
crosslinking using those two thermal initiators and
also, the presence of the water influences the temperature and the modulus values.
2
2.1
Solutions
In this study were produced 6 samples of PVA hydrogels. The samples were prepared dissolving the PVA
(MM 89,00098,000 Mark Sigma Aldrich) 10 and
15% (w/v) in distilled water at 85 C upon agitation.
After cooling, was added 1% thermo-initiators: Azobis
or KPS (w/v) agitating by one hour.
The concentrations of solutions studied were:
Figure 1. Spectra of G , G modulus and viscosity (PVA

10% (w/v) and PVA 15%(w/v)).
PVAl 10% (w/v)

PVAl 15% (w/v)
PVAl 10% + 1% KPS (w/v)
PVAl 15% + 1% KPS (w/v)
PVAl 10% + 1% Azobis (w/v)
PVAl 15% + 1% Azobis (w/v)
2.2
(SEM), Zeiss trademark, model Leo 430i, and optical

microscope (OM) with polarized light GX 51 model,
mettalurgical microscope Olympus trademark.
2.5 Thermal analysis
Rheological tests
In this study were evaluated the modulus and the viscosity. The equipment used was Rheo Haake, Stress
RS6000 model, with dispositive type cone/plate, the
distance was 0,5 mm. The tension mode was shear, the
frequency test used was 1Hz, the scanning was 1 Pa.
The assays were realized in ambient atmosphere.
2.3
Figure 2. FTIR spectra of PVA 10% without thermoinitiators, specimen analyzed G , G modulus and viscosity.
Spectroscopy of Infrared
The samples resulting of rheometer tests, were analyzed in infrared spectrophotometer, mark Thermo
Science Nicolet, IR 100 model, resolution 4 cm1 . It
was monitored the peak of ester group (Costa-Junior,
2009).
2.4 Scanning electron microscopy (SEM) and
Optical microscope (OM)
The samples resulting of rheometer tests also were
visual analyzed by scanning electron microscopy
The thermal analysis was realized by differential scanning calorimetric (DSC) technique, using equipment
of TA Instruments model Q100. The samples analyzed
were heated and cooling: initially heat 25 at 250 C,
cooling 250 C at 0 C and after heat 0 C until 300 C.
3
RESULTS AND DISCUSSIONS
3.1 Rheological tests

In the test (Figure 1) was analyzed the rheological
behavior of solutions of PVA 10% and 15% (w/v).
During the assay the solution PVA 15% presented more
viscosity and stability elastic than PVA 10%, what can
occur due to water acts as plasticizer.
The specimen PVA 10% (w/v) resulting of rheometer was analysed in the FTIR to monitored the peaks
and chemical groups formed. In the FTIR analysis, it
was possible to view the peak relative the group (C-O)
in 1142 cm1 and 1094 cm1 (doublet pair); the peak
of hidroxyl (OH) in 3349 cm1 ; the peak of carbonyl
(C = 0) in 1654 cm1 (Figure 2).
390
Figure 6. Image of specimen PVA 15% (w/v), specimen

analyzed G , G modulus and viscosity (a) OM (b) SEM.
Table 1. Temperatures of crossover point presented in the
rheometer ( C).
Figure 3. DSC curve (second heat) of PVA 10% (w/v),

specimen analyzed G , G modulus and viscosity.
Samples
Temperature of
crossover point ( C)
PVA 10% + 1% KPS (w/v)

PVA 15% + 1% KPS (w/v)
PVA 10% + 1% Azobis (w/v)
PVA 15% + 1% Azobis (w/v)
92
62
68
80
Figure 4. Image of specimen PVA 10% (w/v) analyzed G ,

G modulus and viscosity (a) OM and (b) SEM.
Figure 7. Spectra of G , G modulus and viscosity PVA

10% + 1% KPS (w/v).
Figure 5. FTIR spectra of PVA 15% (w/v), specimen

analyzed G, G modulus and viscosity.
In the thermal analysis by DSC of specimen PVA

10% (w/v), the curve obtained in the second heat indicates the presence of Tg in 90 C and the fusion
peak in 225 C, what confirm there is necessity of
thermo-initiator to occur the chemical crosslinking in
the biomaterial (Figure 3).
Specimen image of PVA 10% (w/v) shows orientation and the shaping of bubbles in the biomaterial
(Figure 4).
The spectra obtained of solution PVA 15% (w/v) is
similar the PVA 10% (w/v), but the peak referent carbonyl group (C = O) appear in doublet 1665 cm1
(Figure 5).
The image of specimen of PVA 15% (w/v) shows
the presence of hydrophilic domains (figure 6).
In the solutions that have thermo-initiators, is possible observed there are differences in the temperatures
and modulus values when is used the KPS or Azobis
as thermo-initiator. It was confirmed that there are
differences in the kinetics of crosslinking using those
two thermal initiators. The different concentrations of
PVA also presented changes on the crossover temperature, what can to confirm the presence of the water
influences the temperature and the modulus values.
Temperatures of chemical crosslinking are presented
in the table 1.
The PVA 10% + 1% KPS (w/v) presented the temperature of chemical crosslinked at 92 C in therheological tests (Figure 7).
The specimen PVA 10% + 1% KPS (w/v), resulting of rheometer was also analysed in the FTIR to
monitored the peaks and chemical groups formed.
It was possible to view the peak of group (C-O) in
1087 cm1 ; the peak of hidroxyl (OH) in 3364 cm1
appeared more broad; the peak of carbonyl (C = 0) in
1654 cm1 (see fig. 8).
10% + 1% KPS (w/v) resulting of rheometer, the
391
Figure 8. FTIR spectra of PVA 10% + 1% KPS (w/v),

speci-men analysed G , G modulus and viscosity.
Figure 11. Spectra of G , G modulus and viscosity specimen of PVA 10% + 1% Azobis (w/v) resulting of rheometer.
Figure 9. DSC curve (second heat) of PVA 10% + 1% KPS

(w/v), specimen analyzed G , G modulus and viscosity.
Figure 12. FTIR spectra of PVA 10% + 1% Azobis (w/v),

specimen analysed G , G modulus and viscosity.
Figure 10. Image of specimen PVA 10% + 1% KPS (w/v),

analyzed G , G modulus and viscosity (a) MO (b) SEM.
curve obtained indicates that occurred the chemical

crosslinking. In the second heat is possible to view the
Tg 95 C, dont have fusion peak (Figure 9).
The color of specimen PVA 10% + 1% KPS (w/v)
resulting of rheometer is brown, is possible to view
any bubles in the surface (Figure 10).
The specimen of PVA 10% + 1% Azobis (w/v)
resulting of rheometer presented the temperature of
chemical crosslinked at 68 C (Figure 11).
The specimen PVA 10% + 1% Azobis (w/v) resulting of rheometer was analysed in the FTIR to monitored the peaks and chemical groups formed. It was
observed that spectra is similar the PVA 15% (w/v)
(figure 5). It was possible to view a doublet peak relative of group (C-O) in 1094 cm1 and 1142 cm1 ;
the peak of hidroxyl (OH) in 3342 cm1 ; the peak of
carbonyl (C = 0) in 1665 cm1 (Figure 12).
Figure 13. DSC curve (second heat) of PVA 10% + 1%

Azobis (w/v), specimen analyzed G , G modulus and viscosity.

10% + 1% Azobis (w/v) resulting of rheometer, the
curve obtained in the second heat indicates the presence of Tg in 90 C and the fusion peak in
200 C, what can be due to chemical crosslinking
occurred only partially, therefore the azobis is a themoinitiators less eficacy in the chemical crosslinking
process (Figure 13).
392
Figure 14. Image of specimen PVA 10% + 1% Azobis

(w/v), analyzed G , G modulus and viscosity (a) MO
(b) SEM.
Figure 17. Image of specimen PVA 15% + 1% KPS (w/v),

ana-lyzed G , G modulus and viscosity (a) OM (b, c) SEM.

15% + 1% KPS (w/v).

15% + 1% Azobis (w/v).
Figure 16. FTIR spectra of PVA 15% + 1% KPS (w/v),

speci-men analysed G , G modulus and viscosity.
The image of specimen PVA 10% + 1% Azobis

(w/v) resulting of rheometer exhibition any formation
and orientation of crystals (Figure 14).
The samples with 15% of PVA obtained temperatures different in the chemical crosslinking. Using
the PVA 15% + 1% KPS the temperature of chemical
crosslinking was 62 C (Figure 15).
The spectra of specimen PVA 15% + 1% KPS (w/v)
obtained in the FTIR analysis, is similar to PVA 10%
(w/v) (figure 2). It was possible to view the peak referent to carbonyl group 1654 cm1 , doublet of C-O
group in 1094 cm1 and 1142cm1 , the hydroxyl in
3349 cm1 (Figure 16).
The image of specimen PVA 15% + 1% KPS (w/v),
exhibition slightly brown color, and any formation
bubbles and orientation of crystals (Figure 17).
Figure 19. FTIR spectra of PVA 15% + 1% Azobis (w/v),

specimen analysed G , G modulus and viscosity.
Using the Azobis as thermo-initiator the temperature of chemical crosslinking in the sample PVA
15% + 1% Azobis (w/v) was 80 C (Figure 18).
The specimen PVA 15% + 1% Azobis (w/v) resulting of rheometer was also analysed in the FTIR. It
was possible to view the peak of group (C-O) in
1095 cm1 ; the peak of hidroxyl (OH) in 3311 cm1
appeared more broad; the peak of carbonyl (C = 0) in
1662 cm1 (see Figure 19). This spectra is similar to
PVA 10% + 1% KPS (Figure 8)
393
Figure 20. Image of specimen PVA 15% + 1% Azobis

(w/v), analyzed G , G modulus and viscosity (a) OM
(b) SEM.

analysed shear time.
Figure 21. Spectra of shear x time PVA 15% and PVA

10% (w/v).
The image of specimen PVA 15% + 1% Azobis

(w/v), exhibition formation and orientation of crystals
(Figure 20).
Analyzing the viscosity of solutions is observed
that the PVA 10% have smaller strain than PVA 15%,
due to presence of water that act as Newton solutions
(Figure 21).
In the FTIR analysis of specimen PVA 10% ( w/v)
resulting of rheometer, it was possible to view the peak
of group (C-O) in 1095 cm1 ; the peak of hidroxyl
(OH) in 3367 cm1 appeared broad; the peak of carbonyl (C = 0) in 1653 cm1 (see Figure 22). In the
specimen of PVA 15% (w/m) it is possible to view the
same chemical group, but the positions are different:
the peak of group (C-O) in 1093 cm1 ; the peak of
hidroxyl (OH) in 3391 cm1 appeared broad; the peak
of carbonyl (C = 0) in 1656 cm1 (Figure 23).
4
CONCLUSIONS
The cross over point of the dynamic modules (G = G )

corresponds to crosslinking, the temperatures found
were: PVA 10% + 1% KPS (w/v) temperature at 92 C,
PVA 15% + 1% KPS (w/v) temperature at 62 C. With
the thermal initiator Azobis the crosslinking process occurs at different temperatures of KPS: PVA
10% + 1% Azobis (w/v) temperature at 68 C, PVA
15% + 1% Azobis (w/v) temperature at 80 C. Temperatures found in this study will be used in the rapid
prototyping machine to production of medical devices.

analysed shear time.
After the rheological tests some specimen resulting

of rheometer, were analyzed using the FTIR to check
the peak corresponding the chemical crosslinked.
This possible observe the differences in the form of
hydroxyl group, and the dislocated position on the
carbonyl group, also observe the presence the doublet
relative of C-O group.
The morphology presented orientacion and formation of crystal.
ACKNOWLEDGMENTS
The authors would like to acknowledge the Institute
National of Science & Technology in Biofabrication (INCT-BIOFABRIS), and the Capes agency of
foments.
REFERENCES
Costa-Junior, E.S.; Stancioli, E.F.B.; Mansur, A.A.P.;
Vasconselos, W.L.; & Mansur, H.S.2009. Preparation
and characterization of chitosan/poly(vinyl alcohol)
chemically crosslinked blends for biomedical applications.Carbohydrate Polymers, v. 76 p. 472481.
394
Kishida, A.; & Ikada, Y. 2002. Hydrogels for biomedical and

pharmaceutical applications. Japan: ed. Marcel Dekker,
p. 13.
Jagur-Grodzinski, J. 2010. Polimeric gels and hydrogels for
biomedical and pharmaceutical applications. Polymers for
advanced technologies, v. 21, p. 2747.
Souza, A. F.; & Ulbrich, C. B. L. 2009. Engenharia integrada
por computador e Sistemas CAD/CAM/CNC Princpios
e aplicaes. So Paulo: Editora Artliber Ltda, 332p.
Swieszkowski, W.; Fray, M.; & Kurzydlowski, K. J. 2008.

Biomaterials for cartilage reconstruction and repair:
Editora Taylor e Francis Group, 678p.
Wiria, F.E.; Chua, C.K.; Leong, K.F.; Quah, Z.Y.;
Chandrasekaran, M.; & Lee, M.W. 2008. Improved
biocomposite development of poly(vinyl alcohol) and
hydroxyapatite for tissue engineering scaffold fabrication
using selective laser sintering. Journal Material Science:
Material in Medicine, v 19, p. 989996.
395
Rapid Tooling & Manufacturing

A study on the mouldability of technical parts using

hybrid moulds and structural foams
A.A. Nogueira
Department of Polymer Engineering, University of Minho, Guimares, Portugal
P.G. Martinho
Polytechnic Institute of Leiria, School of Technology and Management, Leiria, Portugal
Institute for Polymers and Composites/I3N, University of Minho, Guimares, Portugal
A.M. Brito
Department of Polymer Engineering, University of Minho, Guimares, Portugal
A.S. Pouzada
Institute for Polymers and Composites/I3N, University of Minho, Guimares, Portugal
ABSTRACT: The production of large plastics parts in small series by injection moulding requires the development of a new concept of moulds that allows an economic production. This type of products is applied in
urban furniture, automotive, nautical and aerospace industries. The use of alternative methodologies for design
and manufacturing in a current stage in Portugal gave rise to the concept of hybrid mould. The control and
monitoring of the injection moulding process is essential to evaluate the reproducibility of the mouldings and to
obtain a product with good quality. Thus the injection mould, which has a significant influence on the final part,
must be carefully designed.
1
INTRODUCTION
Injection moulding is one of the major processes

used in the manufacture of plastics parts, with small
thickness. For plastics parts of large thickness, low
pressure structural foam moulding is a good alternative
(Osswald 1998).
One solution for obtaining thick parts using less
material structural foams (SF) are a leading to parts
with a low density core and a solid integral skin. The
average density of SF typically varies within 50 to 90%
of the density of the non-cellular polymer. SF are used
in large parts, namely in applications as gardening,
furniture, automotive, aerospace and nautical (Rosato
et al. 2000). Typically the stiffness of SF is four times
higher than the solid plastics with the same weight, and
the mechanical strength and the stiffness-to-weight
ratio are adequate to most load bearing applications
(Wong et al. 2008).
Hybrid moulds are a design solution for injection
moulds that use moulding elements (core, cavities or
other inserts) produced by rapid tooling and a conventional mould structure, being seen as an alternative
solution to prototype series or short production series
(Martinho et al. 2005; Martinho et al. 2009; Pouzada
2009). The hybrid mould concept was developed to
address current market challenges, such as shorter
design and manufacturing times and reduced costs.
The use of materials with low thermal conductivity

(like epoxy resins) tends to increase the cooling time
and consequently change the part properties (Lima
et al. 2003; Pouzada et al. 2009). To overcome this
problem, the epoxy resin is usually filled with metallic
charges (commonly aluminium powder) that improve
the heat transfer, reduce the wear of the tool (Ma
et al. 2007; Vasconcelos et al. 2006), and improve the
mechanical strength (Pouzada 2009). The manufacture
costs are about 40% lower than conventional machining and the lead time is typically between 24 weeks
(Bareta et al. 2006; Menges et al. 2001).
The rapid prototyping and tooling processes (RPT)
are very useful to produce moulding elements for
hybrid moulds and are usually classified into two
main groups: indirect RP (e.g. vacuum casting or silicone moulds) when the moulds are manufactured
from a pattern, and direct RP (e.g. direct AIM or
Selective Laser Sintering) when the mould is manufactured directly using the RP technique. Tools for short
production series are usually soft tools, while tools for
large production series are hard tools (Martinho et al.
2005; Martinho 2010; Pouzada 2009).
The vacuum casting of resin composites is a process that allows the manufacture of soft tools using
composites like the epoxy-based composites.The resin
composite is cast in vacuum over the pattern to produce
the moulding blocks. The pattern reproduces the part
399
or moulding insert with accuracy and good replication

of the pattern (Dunne et al. 2004). One of the advantages associated to this casting process is the short
time required to obtain freeform moulding geometries
(Ramos et al. 2003; Vasconcelos et al. 2004).
Monitoring the performance of the injection moulds
is important to control the quality of the mouldings.
The monitoring process follows the evolution of process variables as temperature, pressure, or clamping
force. The cavity pressure is a critical variable in the
injection moulding process, as it influences the material shrinkage and the ejection force (Kazmer and
Hatch 1999; Pontes and Pouzada 2004). Previous studies on the monitoring of hybrid moulds, showed that
the pressure drop in the impression is slower when
an epoxy composite is used, the maximum pressure
being always lower when compared with conventional
moulds (Pouzada et al. 2009). It was also observed
that the holding pressure was reduced in 40% and the
cycle time increased by 60 s when the injection processing conditions (pressure and cooling time) were
adjusted to the thermal and mechanical properties of
the epoxy composite moulding blocks (Martinho et al.
2009). The maximum pressure in hybrid moulds with
epoxy composite moulding blocks is lower and its drop
slower than in hard tools due to the slower cooling rate
(Martinho 2010).
Due to the low thermal conductivity of the epoxy
composites the hybrid moulds tend to run at a higher
temperature during the cycle time, especially during
cooling, when compared with conventional moulds
(Pouzada et al. 2009). This causes the cooling time,
and consequently the cycle time, being longer when
this type of moulds is used (Martinho et al. 2009).
In this paper the processability of structural foams
using monitoring a hybrid mould at a laboratory scale
is assessed having in view that the use of epoxy composite moulding blocks favour the expansion of the gas
during injection.
2
2.1
EXPERIMENTAL
Figure 2. Structure of the hybrid mould.
Figure 3. Location of temperature sensors at the moulding

surface.
Moulding
The design details and main dimensions of the plastics

part are shown in the Figure 1.
The part was designed to evaluate the processability
of SF using hybrid moulds. It features ribs, castles and
other details with different sizes and thicknesses that
are likely to exist in technical parts.
2.2
Figure 1. The plastics part.
Hybrid mould
A hybrid mould was designed to produce SF mouldings, and to monitor the injection moulding process,
namely the expansion force during the filling of the
impression. The structure of the hybrid mould is shown
in Figure 2.
This mould was equipped with a Kistler 9204B
load cell (L), three Kistler type N 4008 B 0.4 temperature sensors on the moulding cavity surface, and
two Resitec type J, TC.001 thermocouples (T4 and T5)

in the inner of moulding insert as shown in Figure 2.
The load cell is used to follow the force associated to
the injection of the molten material and its expansion
caused by the blowing agent. The temperature sensors
are used to measure the mould temperature in the ejection and injection sides, and also to monitor the arrival
of the melt at specific position in the mould.
The position of the temperature sensors on the
mould surface is shown in Figure 3. The sensor T1
is located close to the sprue, T2 at approximately
80% of the mould filling and T3 at the end of
moulding.
For expansion force was measured using a moving
systemdepicted in Figure 4. When the melt is injected,
the moving system can be displaced and the clamping force is measured by the load cell. This system
400
Table 1.
Properties of the epoxy.
Property
Biresin L74
Coefficient of thermal expansion

Hardness
Impact strength
Heat deflection temperature
Specific gravity
6.8 105 K1
85 shore D
29 kJ m2
160 C
1.1 Mg m3
Table 2.
2009).
Properties of the epoxy composite (Pouzada et al.

Biresin L74 + 60% Al
Properties
Specific gravity
Specific heat
Thermal conductivity
Thermal diffusivity
Coefficient of thermal expansion
Flexural modulus (20 C)
Table 3.
Figure 4. Scheme of moving system.
1.65 Mg m
1279.19 J kg1 K1
0.606 W m1 K1
0.286 106 m2 s1
6.00 105 K1
56 GPa
Injection moulding processing conditions.
Parameter
Aluminium
230 C
Moulding temperature
Mould temperature
Hybrid
Core
Cavity
Cooling time
Filling of mould [%]
Chemical blowing agent [wt%]
80 C
80 C
80 s
80 C
20 C
200 s
Figure 5. Assembly the load cell to the moving system.
80
85
90
95
100
3
4
consists of a plate that supports the moulding block.

The moulding block is assembled to this plate and
guided using four L-locators, and fixed with screws.
The guidance of this assembly is made through guides
and low coefficient of friction bushings to facilitate
the motion during the displacement of the load cell
caused by the injection and expansion pressures. On
the back side of this system there is a spindle which
contacts the load cell.
The Figure 5 shows the assembly of load cell in the
moving system.
2.3
Materials
Two moulding blocks were produced in a composite of Biresin L74epoxy resin filled with 60 wt%
of aluminium powder, and in aluminium. The epoxy
composite was vacuum cast to manufacture the soft
moulding block. The properties of the Biresin L74

epoxy are listed in Table 1.
The aluminium powder has an average particle
size of 50 m and a specific gravity of 2.43 Mg/m3 .
The thermal and mechanical properties of the epoxy
composite are listed in Table 2.
The mouldings were produced in polypropylene Domolen 1100 N, from Domo, with MFI of
12 g/10 min (230 C, 2.16 kg). The blowing agent of
the SF is a masterbatch with the chemical blowing
agent (CBA) Tracel PP 2200 SP, an endothermic compound fromTramaco, with decomposition temperature
in the range 135220 C.
2.4 Injection moulding

The mouldings were produced with an Engel Victory
Spex 50 machine. The parts were produced using two
configurations of the mould: a) steel injection side and
aluminium type 5083-H11 ejection side to emulate a
conventional mould; and b) steel injection side and
epoxy composite ejection side to work as a hybrid
mould.
The processing conditions were set according to the
mould material properties, as shown in Table 3.
401
Figure 7. Clamping force evolution in the hybrid mould

for 80% of mould filling and 4 wt% of CBA.
Figure 6. Monitoring data of hybrid mould with 90% of

mould filling and 3 wt% of CBA.
3
3.1
RESULTS
Monitoring
The mould monitoring data when using the epoxy composite moulding block at 90% of mould filling and 3
wt% of CBA are depicted as an example in Figure 6.
The thermocouples T4 and T5 are located close to
the cooling channels and measure the running temperatures of the moulding blocks. At the injection
side, the temperature is close to 80 C, 10 C above
the coolant temperature, whereas in the ejection side
it is around 25 C. However, due to the low thermal
conductivity of the epoxy, the surface temperature of
the moulding block is much higher. This temperature
is approximately 75 C when the moulding is ejected.
At the sensors placed at the moulding surface, when
the molten material contacts with the sensor, the temperature increases sharply and then decreases slightly
during the remaining cycle time. There is also a significant variation of the melt temperature during the
mould filling: the sensor T1 close to the spruemeasures a temperature higher than the others sensors,
and the sensor T3 at the end of fill measures a low
temperature.
Concerning to the force resulting from the injection
and subsequent expansion of the SF, there is a peak
during the injection phase followed by an abrupt drop.
It appears that during the expansion phase this force
the expansion force remains constant until the end
of the moulding cycle (Figure 7).
Upon increasing the mould filling percentage there
is a significant increase of the clamping force, resulting from the higher mass of injected material. The
amount of CBA in the polymeric matrix does not influence significantly the clamping force.The use of epoxy
composite moulding blocks results in lower clamping force. This I attributable to the epoxy composite
flexural modulus between 56 GPa, much lower than
the aluminium with an elastic modulus about 70 GPa.
The lower modulus enables more deformation of the
moulding block during the injection process, which
results in a smaller clamping force during the injection
process.
Figure 8. Clamping force for various processing conditions,

CBA percentage, and mould types.
When the filling starts, the clamping force increases

linearly until the required percentage of mould filling. The complete mould filling is promoted by the
CBA masterbatch, with a significantly lower clamping
force.
The variation of the clamping force with the percentage of mould filling for different processing
conditions is presented in Figure 8.
3.2 Moldflow analysis
The prediction of the injection moulding cycle of SF
is of interest to the mould design. In this case the
software Autodesk Moldflow Insight 2010 (Autodesc
Inc., Australia) was used. The clamping force along the
mould filling was simulated using the Microcellular
Injection Molding module (Figure 9).
This simulation suggests that a higher clamping
force results from the blowing agent expansion, differently from the actual mould monitoring (Figure 6).
The difference between the experimental data and the
Moldflow simulation may arise from the simulation
software being not prepared to simulate the moulding of structural foams with CBA. This CBA is not a
gas differently from the systems used in microcellular injection moulding, Mucell. Nevertheless, in this
402
There is a significant decrease of the clamping force

(approximately 40%) when hybrid moulds are used.
Different percentages of blowing agents have a negligible influence on the clamping force in the injection
of structural foams.
The use of Moldflow for the simulation of injection
moulding structural foams is accurate to in the injection phase but it is not consistent with the experimental
results in the expansion phase.
Figure 9. Autodesk Moldflow simulation for 90% of

mould filling and 3wt% of CBA masterbatch.
ACKNOWLEDGEMENTS
The authors acknowledge the support of the program QREN Concesso de Incentivos Financeiros
no mbito do Sistema de Incentivos Investigao e
Desenvolvimento Tecnolgico that funded the contract 2010/013307Project Hybridmould 21.
REFERENCES
Figure 10. Moldflow simulation of clamping force for

various injection moulding situations.
case of moulding SF it is of interest to retain the value

of the clamping force corresponding to the end of the
imposed mould filling, as pointed out in Figure 9.
Data from estimating the maximum clamping force
for various mould filling, CBA contents and mould
type (hybrid and conventional) using the Moldflow
software are depicted in Figure 10. It is worth mentioning that different percentages of mould filling
(80, 85, 90 and 95%) lead to a linear increase of the
clamping force, as expected, but that the use of hybrid
or conventional moulds is irrelvant for the Moldflow
simulation.
CONCLUSIONS
The injection moulding of structural foams is a process

that requires higher cycle times because the mouldings have greater thicknesses than a conventional part
and it is necessary to increases the mould temperature to promote the expansion of chemical blowing
agent (CBA). The cycle times are higher when hybrid
moulds are used.
The maximum clamping force increases with the
percentage of mould filling and does not depend on
the expansion of the CBA. Thus, the expansion force
of blowing agent is not critical in injection moulding structural foams, since it is lower than the force
resulting from the injection of the molten material.
Bareta, D. R., Zeilmann, R. P., Costa, C. A. & Pouzada, A. S.

2006. Application of alternative materials in hybrid mould
cores. In RPD 2006-Building the future by innovation;
Proc. Int. Conf., Marinha Grande, Portugal, November.
Dunne, P., Soe, S. P., Byrne, G., Venus, A. & Wheatley, A. R.
2004. Some demands on rapid prototypes used as master
patterns in rapid tooling for injection moulding. Journal
of Materials Processing Technology 150: 201207.
Kazmer, D. O. & Hatch, D. 1999. Towards controllability of injection moulding In 1999 ASME International
Mechanical Engineering Congress & Exposition: Materials Processing Symposium; Proc. Int. Conf., Nashville,
USA, November.
Lima, P., Ramos, J. & Pouzada, A. S. 2003. Thermal performance of hybrid injection molds with epoxy inserts. In
ANTEC 2003; Proc. Int. Conf., Nashville, USA, May.
Ma, S., Gibson, I., Balaji, G. & Hu, Q. J. 2007. Development
of epoxy matrix composites for rapid tooling applications.
Journal of Materials Processing Technology 192193:
7582.
Martinho, P., Brtolo, P. J., Queirs, M. P., Pontes, A. J. &
Pouzada, A. S. 2005. Hybrid Moulds: the Use of Combined Techniques for the Rapid Manufacturing of Injection Moulds. In P. J. Brtolo (ed.), Virtual Modelling and
Rapid Manufacturing-Advanced Research in Virtual and
Rapid Prototyping: 421, London: Taylor and Francis.
Martinho, P. G. 2010. Mechanical design of hybrid mouldsMechanical and thermal performance implications, PhD
thesis. Guimares: University of Minho.
Martinho, P. G., Bartolo, P. J. & Pouzada, A. S. 2009. Hybrid
moulds: effect of the moulding blocks on the morphology
and dimensional properties. Rapid Prototyping Journal
15 (1): 7182.
Menges, G., Michaeli, W. & Mohren, P. 2001. How to Make
Injection Molds, Munich: Hanser.
Osswald, T. A. 1998. Polymer Processing Fundaments,
Munich: Hanser.
Pontes, A. J. & Pouzada, A. S. 2004. Ejection force in tubular
injection moldings. Part I: Effect of processing conditions.
Polymer Engineering and Science 44 (5): 891897.
Pouzada, A. S. 2009. Hybrid moulds: a case of integration
of alternative materials and rapid prototyping for tooling.
Virtual and Physical Prototyping 4 (4): 195202.
403
Pouzada, P. S., Martinho, P. G. & Alves, J. L. 2009. A study of

the integration of an alternative molding block material in
a conventional structure and the influence on the moldings. In ICIT&MPT 2009; Proc. Int. Conf., Ljubljana,
Slovenia, October.
Ramos, A. M., Relvas, C. & Simes, J. A. 2003. Vacuum
casting with room temperature vulcanising rubber and
aluminium moulds for rapid manufacturing of quality
parts: a comparative study. Rapid Prototyping Journal 9
(2): 111115.
Rosato, D. V., Rosato, D. V. & Rosato, M. G. 2000. Injection
Molding Handbook, New York: Springer.
Vasconcelos, P. V., Lino, F. J., Batista, A. M. & Neto, R. J. L.

2006. Tribological behaviour of epoxy based composites
for rapid tooling. Wear 260: 3040.
Vasconcelos, P. V., Lino, F. J. & Neto, R. J. 2004. Importance
of the vacuum in rapid tooling of polymeric-based moulds.
In RPD 2004-Rapid Product Development; Proc. Int.
Conf., Marinha Grande, Portugal.
Wong, S., Lee, J. W. S., Naguib, H. E. & Park, C. B.
2008. Effect of Processing Parameters on the Mechanical
Properties of Injection Molded Thermoplastic Polyolefin
(TPO) Cellular Foams. Macromolecular Materials and
Engineering 293 (7): 605613.
404

Rapid tooling in metal forming processes using 3D-printed tools

S. Junk, R. Wagner, M. Trnkle & S. Ct
University of Applied Sciences Offenburg, Campus Gengenbach, Faculty of Business Administration and
Engineering, Germany
ABSTRACT: 3D-Printing is a widespread additive manufacturing process using plastic powder. Usually the
3D-printing process is followed by manual infiltration of resin in order to increase the strength and brilliance
of printed colours. This resin has a very small depth of penetration only a few millimetres under the surface
of the printed model. Due to the low strength of 3D-printed parts, this process is mainly used in the production
of presentation or demonstration models, i.e. it is limited to Rapid Prototyping. The use of 3D-printed parts
for Rapid Tooling is not possible today, because the strength of the product is insufficient for application to
heavily-loaded metal forming processes, such as deep drawing.
In order to drastically increase the strength of 3D-printed parts, a completely new infiltration method has
been developed. In addition, a new type of resin has been developed. Thus the resin is able to penetrate not
only the surface, but almost the complete volume of the 3D-printed part. These two new innovations create the
opportunity to apply 3D-Printing to parts that are heavily loaded.
To demonstrate the high strength of 3D-printed parts using this new infiltration process, a female die that is
part of a complete deep-drawing tool is 3D-printed and subsequently infiltrated with the newly developed resin.
The 3D-printed female die is then used in combination with two tool parts made conventionally of hard steel
(male die and blank holder) in a forming process. Thus a completely new hybrid deep-drawing tool is created.
A series of experiments using this hybrid tool show fairly good results in the deep drawing of metal sheets made
of aluminium, as well as mild steel.
Also, the economic aspects of this new technology, e.g. tool costs and tool production time, are comparable
to conventional tool making and other Rapid Tooling processes, e.g. laser metal sintering. In conclusion, this
new process for Rapid Tooling of metal forming tools offers a considerable reduction of production time, plus
competitive tool costs in comparison with deep-drawing tools made of steel. This new technology is therefore
particularly suitable for small batch production, and for quick changes in tool design.
1
INTRODUCTION
1.1 General characteristics of the additive

manufacturing technologies
Since the 1980s a number of different additive manufacturing (AM) technologies have been developed,
including, for example, stereolithography (SL), laminated object manufacturing (LOM), fused deposition
modelling (FDM) and also laser sintering (LS). These
technologies vary, on the one hand, in the materials
used, which range from paper sheets and plastic wires
or powder to metal powder [1].
The processed materials are also closely connected
to another important distinguishing feature of these
different technologies, notably the strength of the
manufactured models [2]. Models of little strength,
such as those created from plastic wire with FDM,
are therefore usually only suitable for demonstration
purposes or installation tests. However, LS can be
used to process metal powder into models of high
strength, which can even be used as tool components
in injection moulding tools. LOM can be used to
process layer of paper sheets or metal foils for the
manufacture, for example, of sample tools for deep

drawing [3].
On the other hand, completely different machines
and systems are used: with the SL, LS and metal
foil LOM technologies, lasers are usually used, which
increases the system costs considerably in comparison to technologies and machines which work without
lasers (e.g. FDM).
With FDM, for example, merely a heating nozzle is required to melt the plastic wires. Despite of
high material costs this simple and inexpensive system
equipment allows this technology to be used considerably more economically than other AM technologies,
which is why it is currently widespread [4].
1.2 Characteristics of conventional

3D-Printing Technology
As a relatively new additive manufacturing technology, 3D printing (3DP) was developed at the beginning
of the 1990s at the Massachusetts Institute of Technology MIT and afterwards introduced to the market.
3DP is distinguished by the fact that, instead of a laser,
405
FURTHER DEVELOPMENT OF 3DP
2.1 Improvement of the infiltration process
Figure 1. Comparison of different AM Technologies concerning material costs and compressive strength according [2].
merely a printing technique is applied which is very

similar to that of an ink-jet printer. Since no supporting
material is required, the building speed of 3DP is relatively high. Due to this simple and inexpensive system
technology, 3DP is relatively inexpensive.
The areas of application of 3DP range from the
manufacture of lost moulds for sand casting to the
implementation of drafts in the areas of architecture and art [5]. One can take advantage of the
fact that the layered structure of the models enables
complex geometries which cannot be implemented
with conventional manufacturing methods. With additional changes in CAD also functional models are
produceable. [6]. Furthermore, in comparison to all
other additive manufacturing technologies, 3DP has
the advantage that a huge number of colours can be
displayed, which is particularly important, in particular for the manufacture of presentation models for
new products as well as for both educational and
topographic models.
With regard to the strength of 3D printed models,
it must be mentioned that printed green bodies exhibit
very low strength. That is why green bodies are usually processed manually with an infiltrate (e.g. water
or single-component or double-component synthetic
resin) after printing (see Fig. 1).
This infiltrate penetrates the surface of the models
several millimetres and increases the strength of the
surface. It also improves the brilliance of the printed
colours. However, since the infiltrate only affects the
surface, the achievable strength is merely sufficient
to endure low loads, such as those which arise during
the thermoforming of plastic sheets [7]. Higher loads,
such as those which arise during deep-drawing of metal
sheets or injection moulding, cannot be implemented
with this conventional infiltration technology. Therefore a further development of the infiltration process
itself but also of the used materials (powder, binding
agent and resin) is needed.
The most important objective of the newly developed

infiltration technology is a considerable increase in
the strength of the printed models. This is achieved
by the further development of the infiltration processes as well as the improvement of the materials
used. With conventional infiltration, the infiltrate is
merely applied by hand (e.g. using a brush or by
manual immersion in a bath), whereas the improved
infiltration is an automatic process in several steps.
A green body, for example, is first heated in an
oven to approx. 60 C to extract the remaining moisture
from the powder. Furthermore, a vacuum is generated
to support the drying process. Finally, the infiltration
process is performed by lowering the body in a basin.
The resin can penetrate the green body considerably
deeper by means of the described pre-treatment. The
entire process runs automatically in a system developed for this purpose. The system can be programmed
to allow the times for the individual treatment steps to
be defined and optimised separately.
2.2 Improvement of the materials used
Besides the infiltration process itself, the powder and
resin used have also been improved. According to the
manufacturer, the powder used is of higher strength
than the powder previously used. Likewise, the resin
used in the automatic infiltration system increases the
strength of the models considerably in comparison to
the previous standard products, since it infiltrates the
components deeper and soaks them completely, not
only their surface. According to the manual instructions a change of the powder also necessitate a change
of the binding agent and an update of the printer
software.
Due to the additional system technology for the
automatic infiltration and the improved high-quality
resin, costs for the automated infiltration process
are expected to be higher than for the previous
manual infiltration. However, this disadvantage is
compensated by the higher strength and considerably better reproducibility of the improved infiltration
process.
3
CASE STUDY: RAPID TOOLING FOR

METAL FORMING
3.1 Tool design and manufacture

A testing tool has been manufactured by means of
3DP in order to test the considerably improved infiltration technology and the optimised materials. This
tool is used to cup-shaped parts (diameter of 54 mm
and height of 26 mm ) and represents the preliminary
stage for further forming operations (see Fig. 2). The
tool consists usually of a male die (punch), a female
die and a blankholder.
406
Figure 4. Initial blank and sequence of formed cup-shaped

parts with increasing depth of draw.
Figure 2. CAD-section of the hybrid tool (left) and

CAD-half-section of formed cup-shaped part (larger scale,
right).
Figure 3. Mounted hybrid tool in the press (left) and

3D-printed female die after usage (left).
The steel female die from an existing deep-drawing

tool was replaced by a 3D-printed female die in the
course of the tests (see Fig. 3). The drawing punch
(male die) and blankholder are still of steel. This means
a hybrid tool with tool components of steel and 3D
printing was used.
An allowance of approx. 0.5 mm was initially added
to all surfaces in the computer-aided design of the
3D-printed die in order to compensate for any potential
shrinkage during drying and the subsequent infiltration. This allowance is also required to allow the
infiltrated tool to be reworked by means of grinding
and polishing. The main purpose of this reworking is
to improve the tribological properties of the tool during the forming process. The drawing die was printed
on a ZPrinter450 3D printer from ZCorporation. As
powder zp130 was used and zb59 as binding agent.
The resin used was xlaForm.
3.2 Test procedure
When installing the 3D-printed tool components, make
sure they are fitted precisely in the existing fixture of
the forming press to transfer the arising loads during
the forming process to the fixture. The tests were conducted on a double-acting press. The drawing depth
was able to be varied in the process.
Furthermore, a lubricant is used for forming to

reduce the friction between the sheet and tool. The
residue of this lubricant can also be seen in the
illustration of the tool after the tests (see Fig. 3).
Round sheets with a diameter of 95 mm and thickness of 1.8 mm were formed during the tests. Both
aluminium AlMg3 (EN AW-5754, 3.3535) and deep
drawing steel DC03 (EN 10139) were used as forming
materials.
In order to test the load capacity of the tool, only
very low drawing depths were initially implemented
during the tests. The drawing depth was then increased
continuously up to the complete forming of the components. The arising stage sequence is illustrated in
Fig. 4. The tool endured the loads both with aluminium
sheets and steel sheets. Small cracks on the surface of
the female die did not affect the application of the
hybrid tool.
COMPARISION OF MANUCATURING
COSTS AND TIME
A distinction is made between the material costs and

manufacturing costs in the comparison of the costs.
It must also be taken into account that the machines
and systems employed vary considerably in terms of
their purchasing costs. These costs are included in the
comparison in the form of hourly machine rates. The
individual cycles and the required working time are
compared with each other in the comparison of the
manufacturing time (see Fig. 5).
4.1 Manufacturing costs

The costs for the raw material for 3DP (i.e. powder,
binding agent and resin) are relatively high in comparison to tool steel. However, only the actual component
volume of raw material is required for 3DP. At slightly
more than 50% of the entire raw material costs, the
costs for the newly developed resin are considerably
higher than the costs for conventional resin.
However, in conventional manufacturing, a steel
billet is used of which the volume is reduced by approximately 55% during machining. Waste is created in
the form of chips, whereas the excess powder can
407
technology, pressure and tensile tests should be conducted on suitable samples in the future. This determination of parameters should be extended to newly
developed types of powder and binding agents.
Furthermore, an attempt should be made to create
further tool components on the basis of 3DP in order
to fully benefit from the advantages of this technology. Moreover, it should also be tested whether this
new technology can also be applied to other forming
processes.
ACKNOWLEDGEMENTS
Figure 5. Comparison of manufacturing costs and time of
3DP and conventional machining.
be directly reused with 3DP. The high purchase costs

for conventional CNC machining centres in comparison to the simple 3DP system technology arise during
fabrication. The level of finishing required for the tools
in the form of grinding and polishing is similar in both
technologies.
4.2
Manufacturing time
The comparison of the manufacturing time revealed

that the actual printing process and the subsequent
infiltration take slightly less time than conventional
production. The additional advantage of 3DP is that
the process is fully automatic and no monitoring by
operating personnel is necessary.
5
CONCLUSIONS AND OUTLOOK
With conventional 3DP technologies, it is currently

not possible to manufacture components of sufficient strength for the forming of metal sheets. That is
why new infiltration technologies and new infiltration
substances are applied.
A number of tests have revealed that the application of this new technology allows components to
be printed which area by all means able to meet the
high requirements for metal forming. An economical
comparison with conventional tool manufacturing also
reveals that considerable savings are able to make with
this new technology with regard to the manufacturing
costs and time.
Since there are currently only sparse details on
the technological parameters of the new infiltration
We would like to express our gratitude to E. ERNST

Umformtechnik GmbH, Oberkirch-Zusenhofen and
KISTERS AG, Aachen for the generous support in the
production of tools and execution of tests.
REFERENCES
[1] Gibson, I., Rosen, D.W. & Struker, B. 2010. Additive
Manufacturing Technologies, New York: Springer
[2] Kim, G.D. & Oh, Y.T. 2008. A benchmark study on
rapid prototyping processes and machines: Quantitative
comparisons of mechanical properties, accuracy, roughness, speed, and material cost in: Proceedings of the
Institution of Mechanical Engineers, Part B (Journal
of Engineering Manufacture). Birmingham, AL (USA):
Professional Engineering Publishing
[3] Seidel, S., Otto, A. & Geiger, M. 2007. Contour generation and surface quality of parts manufactured by
automated metal foil LOM-technology, in: Proceedings
of 5th Laser Assisted Net Shape Engineering, LANE
2007, Vol. 1, Erlangen, DE. Bamberg: Meisenbach
[4] Choi, J.-W., Medina, F., Kim, C.; Espalin, D., Rodriguez,
D., Stucker, B. & Wicker, R. 2011. Development
of a mobile fused deposition modeling system with
enhanced manufacturing flexibility, in: Journal of Materials Processing Technology, 211 (2011)
[5] Utela, B., Storti, D., Anderson, R. & Ganter, M. 2008.
A review of process development steps for new material
systems in three dimensional printing (3DP). In: Journal
of Manufacturing Processes, 10 (2008)
[6] Junk, S., Smann-Sun, J. 2008: Application of 3D
colour printing for the rapid prototyping of functional
models, in: Innovative developments in design and manufacturing Advanced Research in Virtual and Rapid
Prototyping, London: Taylor & Francis
[7] Junk, S., Smann-Sun, J. & Niederhofer, M. 2009.
Application of 3D Printing for the Rapid Tooling of
Thermoforming Moulds, Proceedings of 36 th MATADOR Conference, Manchester (GB), London: Springer
408

Conformal cooling in moulds with special geometry

M.A. Garca & C. Garca-Pando
Fundacin Prodintec, Gijn, Spain
C. Marto
BBE Engineering, Marinha Grande, Portugal
ABSTRACT: Additive Manufacturing technologies (AM) are based on layer by layer additive fabrication
processes from a CAD design. There are market niches in all industrial sectors, it is in the field of injection
moulding which yields best results today. The main reason for this is that AM allows manufacturing of optimized,
highly complex and integrated conformal cooling channels, otherwise impossible to achieve by conventional
technologies in similar budget and time frames, or even impossible at all. This paper describes a case in which AM
technologies allowed to improve the quality and properties of an injected part, as well as dramatically reduce its
injection cycle time, thanks to an optimized design and manufacturing of a conformal cooling channel integrated
in the mould.
Keywords: Additive Manufacturing (AM), cooling channels, direct metal laser sintering, hybrid mould
1
INTRODUCTION
Unlike other conventional technologies (e.g. CNC

milling), Additive Manufacturing (AM) implies the
manufacturing of complex, freeform, solid objects
directly from digital models without the use of tools
(Mansour et Hague, 2003; Roscowski et Matuszak,
2000; Santos Costa et al., 2006; Dolisenk, 2005).
Tool-less manufacturing is key factor that will allow
industry to move from a resource base to a knowledge
base and from mass-produced, single-use products to
new mass production of customized, high value, short
life-cycle products. It is expected to revolutionize the
way in which parts are manufactured in the near future.
Despite the important improvements both in processes and materials, as well as in process speeds,
neither researchers nor industrialists have been able
to find the perfect synergy with traditional manufacturing processes so as to introduce AM as a common
technology in the production line of factories, in a wide
range of sectors. In the last 10 years, only a limited
number of high added value applications have been
developed in a non-systematic way, and usually as a
consequence of the demands of particular industrial
sectors, such as Formula 1 racing cars or prostheses
for hearing aids. Should this trend continue European AM activities and market share will soon fall
behind those of other competing regions at a global
scale.
One of the key strategies in order to fully implement
AM technologies in the European industry is working with industrial designers as close as possible. This
should lead us to new generation of well-trained design
professionals who are capable of thinking differently,
developing a new mind-set.
In this article we present the successful case of

a close collaboration between designers and AM
researchers that has led to manufacturing a high added
value hybrid mould which helped to dramatically
reduce production cost of a complex plastic part,
thanks to the optimized design of the internal cooling
channel.
EXPERIMENTAL
Mould parts have been manufactured in a hybrid

process. The lower part has been CNC milled from
European 1.2709 material; the upper part has been
grown layer by layer on top of the lower part by
means of a Direct Metal Laser Sintering (DMLS)
(Kruth et al., 2005) process, using an EOS EOSINTM270 machine (EOS GmbH; Germany), equipped
with a 200 W Ytterbium laser, giving a sport diameter of 0.1 mm. The selected raw material was EOS
MS1 powder (European 1.2709), using EOS standard
exposure parameters in 0.04 mm thick layers, with ultimate tensile strength of 1100 Mpa + 100 MPa in as
sintered condition. Simulations have been performed
with Moldex 3D. The selected material for the injection process is SAN (Luran 368 R Crystal Clear, from
BASF).
3
3.1

Design and simulation (CAE)
In the present case, the mould has been designed in

such a way that the core and the injection side are the
409
Figure 1. Part and section.
Figure 3. Cooling channel design.
Figure 2. Mould without cooling channels.
same (see Figure 2); while this configuration provides

a compact design it also brings an additional difficulty
to the cooling process of both the core and the part. The
part itself is difficult because it has a combination of
thin rims with thick and long parts, especially the long
central bar, as seen in Figure 1 This amount of material
inside a deep well in the mould implies a very long
cooling time in order to get a dimensionally accurate
and acceptable part.
First step in design process is determining the
geometry of a mould with no cooling channels. Simulation with Moldex 3D shows a cooling time of
152 seconds until the last point reaches the ejection
temperature.
In practice, customer will not wait for such a long
cycle time and will open the mould in about 60 seconds to help cooling down. This produces accuracy
problems, warpage and other defects, which in turn
create quality problems and rejected batches, as well
as a lot of handwork.
An iterative design-simulation process brings an
optimized result in terms of heat density, maximum
cooling time and cooling speed. This can be achieved
by positioning the amount of cooling fluid at the optimal distance of the part in relation to the local amount
of material and expected heat flow. The optimized

design is shown in Figure 3.
The distance between loops can also be precisely
adjusted to create an optimal heat transfer from the
part to the fluid.
Moldex 3D simulations give the following results
for this new mould design with optimized conformal
cooling channels, using water at 60 C as a coolant,
with standard injection conditions for the selected
material: average cooling time improves from 152 to
7.35 seconds, average cooling temperature improves
from 128 C to 67.22 C, and heat exchange rate also
improves while the gradient is reduced.
Regarding the cooling system we get a 0.135 C
thermal step at a 120 cc/s flow rate and a very low
pressure drop.
3.2 Manufacturing
It is quite obvious that making spiral conduits, as
those proposed in this case, inside a metal block is
not possible unless lost wax casting is an acceptable
solution from the materials and part accuracy points
of view. This solution is not acceptable for manufacturing injection moulding dies and cores. Layer
based Additive Manufacturing technologies (AM)
allow the manufacturing of these and even more
complex geometries, within new limits and constraints
never imagined before.
While Additive Technologies are good for complex
geometries they are slow and expensive alternatives
when it comes to build simple shapes, as cylinders,
squares or plain blocks, etc. So it is also true that its
enormous potential comes out when they can be combined with traditional technologies to compensate for
these known draw-backs: time and cost. This is called
Hybrid Manufacturing.
In this case, the lower part of the moulds has been
manufactured in a traditional way by CNC milling
410
Figure 5. Preforms fixed to the building platform.
Figure 4. Hybrid design. Top sintered, bottom CNC milled.
stock material (European 1.2709). We can see the top

plane of the so called preform as the line that separates the two differently hatched areas in the section
represented in Figure 4.
Preform is then fixed to the building platform of
the Laser Sintering machine. In this case four identical parts were built simultaneously on the same
production batch. Once the preforms are fixed to the
platform they are grinded to get an overall flatness
of less than 0.01 mm all over the building area. (See
Figure 5).
A measuring step is now required in order to get precise references for all parts that have to be sintered on
top of preforms, so these references can be translated
to the CAD/STL files and then to the laser optics. The
sintering machine has no on-board measuring add-ons
so a CMM is required. The building platform is then
carefully positioned on the platform carrier inside the
EOS EOSINT M-270 machine and properly levelled
to the recoater plane.
The machine is already loaded with the EOS MS1
powder (European 1.2709), and the properly oriented
and positioned STL files, according to the CMM measurements. The process is started and run with EOS
standard set of exposure parameters for this material,
in order to achieve the expected results, matching those
mechanical and chemical properties declared by the
manufacturer.
After the process ends, the building chamber is
opened, the excess powder is removed, and the building platform is unscrewed from the carrier (Figure 6).
Figure 6. Hybrid parts ready.
Once outside the machine the parts are removed from

the platform and ready for finishing operations.
Parts are put into normal operation after finishing
process and after been adjusted to the injection system.
As expected from the simulation the machine is able
to produce good parts at a 3.4 parts per minute while
it was hardly possible to get 1 part per minute with a
high quality rejection rate.
CONCLUSIONS
While it is impossible to manufacture a spiral conformal cooling channel-based cooling system by

traditional technologies, this proves to be a very efficient solution to dramatically reduce cycle times and
improve part quality. Laser Sintering empowers design
strategies and enables the manufacturing of highly
complex mould parts that yield great results in terms
of both part quality and injection process speed.
411
On the other hand, Laser Sintering process can be

successfully and accurately combined with conventional technologies to make parts that join the benefits
of both reducing the overall production cost and time
compared to Additive Manufacturing only and the
amazing geometries the Additive Manufacturing is
capable of. This advantages can also be beneficial
for the cost per injected part, mainly in small series
production.
In this case study the initial disadvantage of having
the core and the injection side together can be turned
into an advantage thanks to a careful design of an optimized conformal cooling channel which allows both
cold runners and parts are cooled down at the same
time, and in a very efficient way.
Thanks to this conformal cooling inserts we get:
Cooling time reduction from 152 to 7.35 s

Average cooling temperature reduced from 128 C
to 67.22 C
Temperature gradient during cooling is reduced,
reducing warpage.
The production rate rises from scarcely one part per
minute (with possibly 3 to 5 good parts out of every
ten) up to 3.4 per minute 10/10 good parts.
REFERENCES
Costa E., Shiomi M., Osakada K. & Laoui, T. 2006. Rapid
manufacturing of metal components by laser forming.
International Journal of Machine Tools & manufacture,
46, 14591468, 0890-6955.
Garca C., Gonzlez D. & Garca M. 2008. Personalized

tools manufactured by rapid manufacturing for medical surgery-a success case of a Spanish SME. In Igor
Drstvensec & Slavko Dolinsek (ed.), Additive Layered
Manufacturing: From Evolution to Revolution. Maribor,
2008.
Garca. M., Sastre I. & Garca, C. 2010. Fabricacin mixta
de rapid manufacturing y mecanizado. 18 Congreso de
Mquinas-Herramienta y Tecnologas de Fabricacin.
Noviembre 2010, San Sebastin.
Greses J & Weilhammer J. 2008. Rapid Manufacturing
of final dies and cores for production of large series
of sintered parts. EOS Whitepaper, EOS GmbH (ed.)
Krailling/Mnich, September 2008.
Hopkinson N., Hague R. & Dickens. P. 2006. Rapid
Manufacturing-An Industrial revolution for the digital
age. N. Hopkinson, R. Hague and P. Dickens (eds.), John
Wiley & Sons, Ltd, Chichester, UK. ISBN-13978-0-47001613-8.
Levy, G. N., Schindel, R. & Kruth, J.P. 2003. Rapid manufacturing and rapid tooling with layer manufacturing
(LM) technologies, state of the art and future. Manufacturing Technology Volume 52, Issue 2, 2003, Pages
589609.
Mayer S. 2009. Optimised mould temperature control
Using DMLS. EOS Whitepaper, EOS GmbH (ed.)
Krailling/Mnich, 2009.
Pandremenos J., Paralikas J., Chryssolouris G., Dybala B.
& Gunnink J.W. 2008. RM Product development: design
principles, simulation and tools. Proceedings of the International Conference on Additive Technologies (ICAT
2008), Ptuj, Slovenia, September 2008.
Shellabear M., Weilhammer J. 2007. Tooling Applications
with EOSINT M. EOS Whitepaper, EOS GmbH (ed.)
Krailling/Mnich, September 2007.
Wohlers T. 2011.Wohlers Report 2011. ISBN 0-97544297-X.
412

Thermoplastic Resin Transfer Moulding in a rapid manufactured mould

E. Atzeni, F. Calignano, L. Iuliano, P. Minetola & A. Salmi
E. Bassoli, L. Denti & A. Gatto

Universit degli Studi di Modena e Reggio Emilia, Department of Mechanical and Civil Engineering (DIMeC),
Modena, Italy
ABSTRACT: Resin Transfer Moulding (RTM) is a low-cost process for the production of composite parts
with thermosetting resins. However, the economic convenience is lost in the case of large components, such as
aeronautical panels, because of the high tooling cost. Step milling of a resin master for the electro-deposition of
a nickel shell is proposed in this study as a valid alternative for the fabrication of a resin transfer mould. This
solution allows overcoming the limitations regarding part size and cutting down tooling costs, but the expected
dimensional accuracy needs to be assessed. On the field of materials, innovative thermoplastic resins are now
available that can be used in the form of low viscosity oligomers for optimal mould filling and fibre impregnation.
Subsequent in situ polymerization provides high toughness composites. Cyclic Butylen Terephtalate (CBT),
which polymerizes into PBT (Poly-Butylen Terephtalate), is studied in this work. The research focuses on two
objectives: quantifying the dimensional accuracy of the technological chain and setting up the process of in situ
polymerization.
INTRODUCTION
Resin Transfer Moulding process (RTM) was developed in the 1980s and spread for the production of
fibre reinforced thermoset composites, expecially for
aeronautical and aerospace applications (Potter, 1999).
The distinctive capabilities of RTM include the production of complex parts with large surface areas, with
high product quality and close dimensional tolerances
(Palardy et al., 2008). A dry fibre perform is placed
within the mould cavity, the mould is closed and an
activated liquid thermoset resin is injected (Kendall
et al., 1992). The relatively low viscosity of thermosets
allows for a low injection pressure and a good wetting
of the reinforcement. Fibre size, content and orientation can be controlled in the perform to optimize
the mechanical properties of the final product (Kim &
Daniel, 2005; Norman & Robertson, 2003).
Recently, reactive thermoplastic resin compounds
have been studied, which can be processed in the
form of low viscosity mono-or oligomeric precursors, ensuring easy fibre impregnation before in situ
polymerization (van Rijswijk & Bersee, 2007). The
limitations of traditional melt processing of high viscosity thermoplatics can thus be overcome and the
advantages of thermoplastic composites over thermoset ones can be contemplated in the near future:
higher toughness, faster manufacturing, recyclability,
easier raw materials storage, weldability and ability to post-form moulded parts (Verrey et al., 2006).
The application field of thermoplastics in the form of
reactive oligomers can be expanded to liquid impregnation processes like RTM and infusion (van
Rijswijk & Bersee, 2007). Despite the potential benefits, the industrial application of reactive thermoplastic
systems is still limited by the need for better knowledge of the process set up and performances (Verrey
et al., 2006). The cyclic oligomer butylene terephthalate (CBT) has a water-like processing viscosity,
which can be as low as 0.02 Pa s, and can be transformed into linear high molecular weight PBT via
entropically-driven ring-opening polymerization in a
short time scale (Mohd Ishak et al., 2007). The reaction
occurs under isothermal conditions, which means that
curing can be avoided with a great cost and time saving
with respect to thermoset resins. The control of process
parameters (time, temperature, catalyst percentage,
stirring) are crucial to the kinetics of ring-opening
and polycondensation reactions, which determine the
molecular weight, melting point and mechanical properties, toughness above all, of the obtained engineering
polymer (Mohd Ishak et al., 2007).
RTM is typically an economical process, however
the product cost is directly related to the tooling cost,
that is highly affected by the part geometry, the production rate and quality requirements. For low production
volume of large size components the investment for
tooling could be considerable (Verrey et al., 2006).
Mould halves are made of aluminium, steel or nickel
alloys and are traditionally machined by milling operations. The material selection is a function of the
required production volume and surface finish. In the
413
Figure 1. Electro-deposition on the resin master manufactured by step milling (a) and final RTM mould (b).
case of aeronautic panels, whose overall dimensions

are typically large, the machining of the RTM mould
involves the removal of a large material volume resulting in long manufacturing times and high costs. Some
alternative techniques for mould making have been
introduced to lower tooling costs, such as Light RTM
or Vacuum Infusion (VI). In both cases one mould half
is substituted by a shell: a semi-rigid thin shell made
of composite material is used in Light RTM and a thin
flexible membrane in the VI process (Hutchinson et al.,
2011; Park & Saouab, 2009). A vacuum is created
inside the tool cavity to draw the shell down onto the
other mould half. However in both cases the injection
pressure must be lower than the atmospheric pressure while the part should be designed with a uniform
thickness, thus limiting the shape complexity.
In this paper an alternative method is proposed for
the production of a metal shell mould for the RTM
process. In fact, metal shells can properly replace the
mould halves because of the low injection pressure
(typically less than 7 bar), assuring the same potentialities in terms of shape complexity and thickness
variations. The method is based on the fabrication of a
nickel mould by electro-deposition on a resin master,
previously coated with a conductive paint (Gatto et al.,
2003; Ippolito et al., 1996). Since the working surface
of the nickel shell is the reverse shape of the resin master (Figure 1a), the latter must be a copy of the part
geometry, apart from the shrinkage of the thermoplastic material. In mould design additional surfaces must
be provided to ensure the correct alignment and closing
of the mould halves. The resin master is shaped by step
milling, which allows overcoming size limitations and
containing manufacturing costs (Iuliano et al., 2001).
Step milling is classified among Rapid Tooling technologies, since it combines traditional milling with the
philosophy of layer manufacturing. The part geometry
is sliced into parallel sections with a constant thickness
(about 15 mm ) that are machined on a 2 1/2 axis miller.
The mould shells are produced by electro-deposition,
separated from the masters, machine dto add runners
and vents and to adjust the mould parting plane, and
finally reinforced with a back filler (Figure 1b). Design
for electroforming guidelines must be considered to
ensure a uniform thickness distribution: sharp edges
must be radiused, deep recesses should be avoided,
and convex shapes should be used instead of concave
ones wherever possible. Moreover, the mould should
be designed to guarantee a correct ejection of the part

and to achieve a good product quality: draft angles
should be added to vertical walls, large variations in
cross section and rapid changes in geometry should
be avoided, and the parting line should be located to
reduce flashing and minimize the need for cores.
The proposed mould making procedure needs an
evaluation of benefits and limitations. In fact, even if
each single process (step milling, electro-deposition,
RTM) is not innovative, their combination into a technological chain, applied to the production of large
composite parts, is novel and requires an assessment of reliability, industrial feasibility and expected
dimensional accuracy.
A first aim of this research is the estimation of
dimensional deviations induced by each step of the
mould manufacturing sequence, so as to identify the
corrections that can be applied in advance to mould
design in order to meet the required part tolerances.
To this objective, a benchmark is selected and a procedure is proposed for dimensional control throughout
the mould production. The manufacturing sequence
is outlined in Figure 2, together with the procedure
proposed for evaluating the dimensional deviations
introduced by each step. Both traditional pointwise measurements, i.e. on a Coordinate Measuring
Machine (CMM), and contactless optical scanning are
used for dimensional inspection. The first are adopted
for measuring thickness and length on a limited number of cross-sections, whereas Reverse Engineering
(RE) technique is used for the entire working surface
of the as fabricated mould.
For the first step, that is the resin master manufacturing, the dimensional measuremnets are compared
with the corresponding original CAD model. The comparison can be limited to the active area of the mould,
that is to the surfaces defining the shape of the cavity plus the neighbouring closing plane. The results of
this inspection provide information about the errors
induced by the step milling operation and subsequent
assembly of the machined sections. For the second
step, that is the electrodeposition, the measurements
on the nickel shells are compared with the data previously acquired on the corresponding resin masters,
to assess the errors induced by the electrodeposition
step. Finally, the deviation of the whole mould cavity
before moulding is evaluated by comparing the original CAD model of the mould cavity with the virtual
414
Figure 2. Procedure for evaluating the deviations introduced by each step of mould manufacturing.
mould cavity, obtained by Boolean subtraction from

the scan data of the two electroformed mould halves.
All the described analyses of scan data are made by
means of a specific commercial RE inspection software (i.e. Rapidform 2006) after the alignment in the
same Cartesian reference system.
For the first step, that is the resin master manu. A
second focus of this work regards the kinetics of in situ
polymerization. Preliminary tests are conducted to
determine the optimal time and temperature set-up for
CBT melting, catalyst and reinforcement addition, and
isothermal polymerization within the mould. Short and
long carbon fibres are evaluated as fillers.
2
2.1

Mould manufacturing
A benchmark is selected representative of aeronautic

components, which resembles an H-shaped structural
component with a variable cross section along the
longitudinal axis (Figure 3). From the geometry of the
reference part, the resin master is designed on the basis
of the following considerations.
The active surface of the resin master is the
exact copy of the part shape and dimensions: the
allowance needed to compensate the shrinkage of
the thermoplastic material during RTM is disregarded in the first step and will be evaluated in the
Figure 3. 3D CAD model of the benchmark (overall dimensions 337.5 77 54 mm).
developments of the research. Also, shrinkage does

not affect the analysis of dimensional accuracy in
mould manufacturing.
The benchmark plane of symmetry is selected as
the mould parting plane and its normal direction is
labelled Z axis. The benchmark has no undercuts
for the considered orientation, so only two master
halves are required.
For each half of the part model, the resin master geometry is obtained by adding a base block
(400 140 10 mm ). Two side walls 10 mm high
are added to the block to define two side reference planes for the correct alignment of the two
shells.
No allowance is provided on the parting plane
for the adjustment operations, since its value
should be determined depending on electrodeposition tolerances, which will be a result of this
research.
415
Table 1.
Machining strategy and cutting parameters.
Machining
strategy
Cutting speed
Feed per tooth
Side step
Cutting time
Roughing
Finishing
Tool End mill

15 mm, 3 teeth 1.2
Zig-Zag
depth of cut: 2 mm
allowance 0.2 mm
140 m/min
0.17 mm/tooth rev
7.5 mm
65 min
Ball-end mill
5 mm, 2 teeth
Parallel planes
(45 )
95 m/min
0.05 mm/tooth rev
0.2 mm
510 min
All edges are radiused at 2.5 mm and draft angles

of 2 are added to planar surfaces parallel to the Z
axis.
Finally, each master model is sliced perpendicular
to the Z axis to a thickness of 16 mm , obtaining 3
sections to be machined by step milling. Holes are
added for assembly purposes.
Six blocks (400 140 16 mm ) of resin for moldmakers are used to step mill the slices. Toolpath is
elaborated through a CAM software (Vero Visi Series
v.17): the parameters adopted in the roughing and finishing step are detailed in Table 1. The blocks are
machined on the 2 1/2 axis CNC milling machine SHE
HONG-VMC 850 and reference holes are added using
a 4 mm diameter twist drill bit.
Machined sections are then assembled via pins and
a bi-component adhesive. Holes are filled with a resin
and manually finished to obtain the smooth surface
that will be copied by electrodeposition. Finally, the
masters are coated with a silver paint and 5 mm thick
nickel shells are built by electroforming. The mould is
then completed by the addition of the feeding system.
2.2
Dimensional inspection
After fabrication the master halves and the electroformed shell are inspected through the evaluation
procedure described in the introduction. Pointwise
measurements are carried out by a probe tip of 2 mm
on a DEA CMM, model GLOBAL Image 07.07.07,
that is equipped with an indexable swinging head and
a touch trigger probe. The volumetric length measuring uncertainty MPE of the machine, according
to ISO-10360/2, is 1.5 + L/333 m; where MPE is
the acronym for Maximum Permissible Error and L
is the measured length. Measurements are replicated
three times on each inspected component. The thickness of the arms of the H-shaped component are
measured on two levels (Z = 9 mm and Z = 17 mm)
on seventeen transversal cross-sections evenly spaced
every 20 mm (Figure 4). The length of the arms is
also measured on the same two levels, while the three
main planes (base plane, middle plane and top plane)
are checked for location (Z level) and planarity by
measuring 30 points on each of them.
In order to assess the whole real geometry of the
master and mould, inspection is also carried out by
Figure 4. Points on evenly spaced cross-sections measured

by DEA CMM.
optical scanning using an Atos (Advanced TOpometric Sensor) Standard device. It is a general purpose
structured light scanner that exploits binocular vision
with two built-in 0.4 MPixel CCD cameras, which
store images of the light fringes projected on the
scanned object. The projector, placed in the centre of
the sensor, projects a sequence of four interference
patterns (phase-shift technique) (Wang et al., 1996)
followed by six Gray coded binary images (Pages
et al., 2003; Perry & McKelvie, 1995). The nominal
accuracy is 0.05 mm. In less than 10 seconds the scanner retrieves about 360,000 points on a working area
of 200 160 mm.
2.3 In situ polymerization tests
Preliminary tests are planned in order to study the
kinetics of in situ polymerization of CBT into PBT
and to determine process set-up for obtaining high
molecular weight and toughness.
A first test is performed using the monomer formulation CBT 160 (Cyclics Corp., USA), already
containing the polymerization catalyst. This resin does
not require dosage and dispersion of the catalyst, but
the control of process time and temperature is critical since polymerization starts unrestrained. 10 g are
molten at 140 C under continuous stirring, then temperature is raised up to 190 C where polymerization
is expected to occur in 25 min.
All subsequent tests are carried out using the uncatalyzed formulation CBT 500 (Cyclics Corp., USA),
processed through the following steps.
1. Preparing of the catalyst solution: polyethylene glycol (PEG) is vacuum dried up at 150 C for 1 hour
and added with butylchlorotin dihydroxide (BuSnCl(OH)2) in the volume ratio 35:1. Complete
dissolution is achieved in 4 hrs at 130 C under
vigorous stirring.
2. Drying up under vacuum of CBT500 and of
fillers, when used, for 12 hrs at 110 C, to reduce
the risk of hydrolytic degradation.
3. Melting of 10 g of resin at 180 C under continuous
stirring. The monomer is stable up to 290 C in air
or to 370 C in nitrogen.
4. Addition of the catalyst solution to the molten resin,
in such a quantity to obtain 0.45 wt% of catalyst
over CBT. The blend is stirred and heated up to the
recover of temperature decrease.
5. Isothermal step for the polymerization and
simultaneous solidification of the resin. Several
416
Figure 7. One of the two nickel shells.
One of the two assembled master halves is shown

in Figure 6 and one of the electroformed nickel shells
in Figure 7. The desired thickness is obtained in about
10 days.
3.2 Dimensional inspection
Figure 5. 3D CAD model of the slices of one resin master
half (overall dimensions 400 140 48 mm).
Figure 6. One of the two resin master halves.
temperatures are evaluated in the range 160 to

200 C for times of 5 to 20 min.
6. Air cooling.
In some of the tests the following modifications of the
above protocol are introduced:
2b Addition of 5 to 10 wt% of Polycaprolactone (PCL)
to CBT as a toughening element.
2b Addition of 10 wt% of the monomer CL to CBT
in order to obtain a tougher copolymer.
4b Reinforcement of the resin by addition of carbon
short fibers or woven cloth.
1 Remelting of the resin at 250 C for 5 min to ensure
the achievement of a higher molecular weight and
avoid the risk of cold cristallyzation.
Differential scanning calorimetry (DSC) is carried out
on all the samples (710 mg; heating rate 20 C/min up
to 260 C; 2 min isothermal, same rate down to room
temperature).
3
3.1

Mould manufacturing
The geometry of one of the two symmetric resin master

halves is shown in Figure 5. The parting plane is set
as origin for the Z axis and it is addressed to as base
plane.
The results of pointwise measurements on the upper

and lower halves of the resin master are reported in
Tables 2 and 3. Mean value and standard deviation
are listed separately for the two arms and the different measuring levels. As to planes, the measured vs
nominal position is shown as well as planarity. From
CMM measurements, the deviation introduced by step
milling and assembly of the resin master is assessed to
be around 0.10 mm. The deviation of the arm thickness
is higher than the one of the arm length, but it should
be noticed that the value refers to a greater number of
measurements (seventeen cross-sections). Since deviations are positive in sign, the fabricated resin master
is bigger than the nominal CAD model.
The results of pointwise measurements for the electroformed mould halves are listed in Tables 4 and 5.
The deviation introduced by electrodeposition on the
arm thickness ranges from 0.24 to 0.35 mm, whereas
deviation on the arm length is around 0.50 mm. In both
cases the dimensions of the electroformed shells are
bigger than the ones of the resin master halves.
Planarity on the nickel shells is worse than on
the resin master. Such result suggests that the master
deforms within the high temperature electrodeposition
bath. The deformation is smaller for the middle plane,
because this feature is made stiffer by the presence
of the H-shape arms. In addition, it can be noticed
that the Z level of the base and top planes on the
electroformed shells is higher than of the same elements on the resin master. This result is in accordance
with the constraint imposed by the arms for the central
part of the H-shape and it can also partially justify the
higher deviation of arm thickness on the electroformed
mould.
Based on these results, the allowance required for
mould adjustment can be quantified in about 1 mm,
which can include the dimensional deviations induced
by both step milling and electroforming.
The optical scan data are processed according to the
same approach as CMM measurements. For the first
manufacturing step (master fabrication), the scan data
of the two halves of the resin master are compared with
the corresponding original CAD model. For the second step (electrodeposition), the scan data of the two
nickel shells are compared with the measures of the
417
Table 2. Deviation from CAD nominal values of arm length

and thickness on the resin master, measured in the transversal
cross-sections (All dimensions are in millimetres).
Arm/level
Thickness first/Z = 9
first/Z = 17
second/Z = 9
second/Z = 17
Length
first/Z = 9
first/Z = 17
second/Z = 9
second/Z = 17
Table 6. Dimensional deviation measured by optical scanning. (All dimensions are in millimetres).
Absolute Distance of
Scan Data
Average (SD)
Upper half
average (SD)
Lower half
average (SD)
Number
of points
0.07 (0.02)
0.08 (0.02)
0.09 (0.02)
0.10 (0.03)
0.01 (<0.01)
0.01 (<0.01)
0.01 (<0.01)
0.01 (<0.01)
0.09 (0.01)
0.11 (0.01)
0.10 (0.02)
0.11 (0.02)
0.02 (<0.01)
<0.01 (<0.01)
0.02 (<0.01)
0.01 (<0.01)
Resin master vs CAD model

upper half
127,434
0.06 (0.16)
lower half
134,579
0.10 (0.18)
Electrodeposited mould vs resin master
upper half
142,197
0.26 (0.23)
lower half
137,311
0.33 (0.23)
Table 3. Measurements of plane levels and planarity on the

resin master. (All dimensions are in millimetres).
Upper half
Plane
Lower half
Z level Z level
Z level
nominal meas. planarity meas. planarity
Base
0.00
Middle 3.10
Top
27.00
0.01
3.10
26.95
0.07
0.04
0.07
0.01
3.10
26.83
0.17
0.07
0.17
Figure 8. Colour deviation map for scan data comparison between the resin master and the electroformed shell
of the mould upper half. (Absolute deviation values are in
millimetres).
Table 4. Deviation of arm length and thickness of the nickel

shell with respect to the resin master, measured in the
transversal cross-sections (All dimensions are in millimetres).
Thickness
Length
arm/level
Upper half
average (SD)
Lower half
average (SD)
first/Z = 9
first/Z = 17
second/Z = 9
second/Z = 17
first/Z = 9
first/Z = 17
second/Z = 9
second/Z = 17
0.27 (0.05)
0.34 (0.05)
0.24 (0.05)
0.29 (0.06)
0.46 (0.01)
0.60 (<0.01)
0.43 (<0.01)
0.50 (<0.01)
0.28 (0.05)
0.35 (0.05)
0.25 (0.05)
0.32 (0.05)
0.51 (<0.01)
0.60 (<0.01)
0.49 (<0.01)
0.57 (<0.01)
Table 5. Measurements of plane levels and planarity on the

nickel shells. (All dimensions are in millimetres).
Upper half
Plane
Lower half
Z level Z level
Z level
nominal meas. planarity meas. planarity
Base
0.00
Middle 3.10
Top
27.00
0.20
3.12
27.17
0.59
0.18
0.34
0.17
3.11
27.08
0.62
0.18
0.22
corresponding halves of the resin master which they

have been built on. The results are shown in Table 6.
After alignment and superposition in a common reference system, the comparison of the scan data can
be viewed as colour deviation maps with the relative
deviation distribution, as the one in Figure 8. As to

the resin master, the mean distance of scanned points
from the nominal CAD values is lower than one tenth
of millimetre. Deviations of the electroformed mould
from the resin master, which can be appreciated in Figure 7, are much higher: this result is consistent with
CMM measures and confirms that deformations occur
during electrodeposition.
Finally, if the scan data of the mould halves are
aligned by selecting corresponding points on the
mould parting plane, the virtual volume of the mould
cavity can be extracted from a block by Boolean subtraction, performed on triangulated STL models. The
same operation can be carried out on the mould CAD
model, obtaining the nominal volume of the mould
cavity. Since the mould parting plane was not finished
or adjusted, flash along the virtual parting line is a consequence of the poor planarity of the base planes of the
electroformed mould. Comparing the virtual volume
with the nominal CAD one, Rapidform software computes an average distance of 0.24 mm (with a standard
deviation of 0.40 mm), calculated on 261,240 points.
Disregarding the thermoplastic material shrinkage, the
virtual volume is expected to be equal to the volume
of the final part that is manufactured by RTM.
3.3 In situ polymerization tests
CBT 160 shows many processing difficulties: due
to inhomogeneous heating the areas that hear more
quickly start polymerizing while adjacent portions are
418
Figure 9. DSC curve: CBT 500 polymerized at 170 C up

to solidification (black line) and after remelting at 250 C for
5 min (gray line).
still melting. Inhomogeneous samples are obtained

with a wax-like consistency. The tests on this formulation are thus abandoned.
CBT 500 guarantees a much easier process control. The resin melts completely and gets evenly transparent at 180 C. The addition of the catalyst solution
causes a milky appearance due to a temperature drop.
This step can be optimized by preheating the catalyst solution, but the degradation of PEG at 180 C
needs to be considered. The addition of PCL does not
require any set-up modification in the first process
steps, but the blend solidification does not occur above
130140 C. For these samples polymerization during
step 5 takes place in the molten phase and the resin
solidifies afterwards, during cooling.
All the samples undergo a consistent shrinkage and
show high fragility, which often cause cracks or complete failure. The remelting step yields to a good
toughness improvement for samples obtained with
low times and temperatures during step 5. As an example, the DSC curves in Figure 9 refer to a sample of
unblended CBT 500 polymerized at 170 C up to the
complete solidification (black line) and to the same
sample after remelting (gray line). The low melting
temperature (208 C) and the presence of a secondary
peak in the black curve indicate that polymerization
is not completed and proceeds during the analysis.
The sample treated at 250 C, instead, melts with a
well definite peak at 215 C, which is near to the theoretical melting temperature of high molecular weight
PBT. The samples relative to the curves in Figure 8
are shown in Figures 10a and b respectively: the first
breaks due to shrinkage stresses whereas the second
is solid and free from cracks. This result suggests
that chain mobility at 170 C is not high enough to
ensure the achievement of high molecular weights.
Since in situ polymerization occurs at a temperature
that is below the melting point of obtained polymer,
chain-lengthening and solidification are competitive
reactions. High molecular weight and good toughness
can be obtained if polymerization (step 5) is carried out
at a temperature of 190200 C maintained isothermal
Figure 10. CBT 500 polymerized at 170 C up to solidification (a) and after remelting at 250 C for 5 min (b). Same
sample as (b) poured on a woven carbon cloth (c). CBT
500 with 10 wt% of PCL polymerized at 190 C for 10 min,
solidified at 150 C (d).
Figure 11. DSC curve: CBT 500 polymerized at 190 C

for 10 min, solidified at 150 C.
for at least 10 min, without any subsequent remelting.

DSC on a sample produced with this set-up, shown in
Figure 11, proves that polymerization is complete and
leads to a PBT that melts at 215 C.
Fillers, both in the form of short fibers or woven
cloth, obstacle the material shrinkage and enhance the
formation of cracks, even for samples polymerized at
high temperature (Figure 10c).
The addition of PCL effectively improves toughness: 5 wt% only is enough to avoid through cracks
and improve samples solidity and optimal results are
achieved for 10 wt%. The DSC curve in Figure 12
is relative to a sample added with 10 wt% of PCL,
polymerized at 170 C up to complete solidification
and then remolten. The melting temperature of the
final polymer lowers of about 1.5 C for each percentage point of PCL in comparison with unblended CBT.
The absence of PCL melting peak at 55 C proves the
perfect miscibility of the two polymers.
419
dihydroxide catalyst solved in PEG. Key aspects for

successful polymerization regard:
Figure 12. DSC curve: CBT 500 with 10 wt% of PCL

polymerized at 170 C up to solidification and remolten at
250 C for 5 min.
Based on the results, the optimal material formulation and process set-up are determined. If the material
is to be used without fillers, CBT 500 can be successfully processed following the steps 1 to 6 in the
materials and methods, provided that in step 5 a temperature of 190 C is reached and maintained for 10
min. If carbon fibers are added, a higher toughness is
required and can be obtained by a blend of CBT 500
with 10 wt% of PCL, processed in the same conditions
of the unfilled polymer: one of these samples is shown
in Figure 9d.
CONCLUSIONS
An innovative procedure is applied for manufacturing

a mould for RTM of a fibre-reinforced thermoplastic
part. An H-shaped reference part is selected to benchmark the aeronautic sector. The mould is fabricated
by nickel electrodeposition on a resin master that is
previously machined by step milling.
The dimensional accuracy in tool fabrication is evaluated by contact and non-contact measurements that
are repeated after every step of the manufacturing
chain. This, together with the procedure defined for
the comparison of the acquired data, allows the assessment of dimensional deviations induced by each single
mould production phase. The comparison of the final
mould cavity with the nominal CAD model points
out an average deviation of 0.24 mm, mainly due to
the electroforming step. On one hand, the results of
this study suggest that electrodeposition on the resin
master is critical and that particular attention should
be paid in order to contain thermal deformations.
Dimensional stability of the resin, when in contact
with the bath reagents at the deposition temperature,
could be optimized. On the other hand, the information
about deviations can be used to compensate a priori
dimensional errors in the mould design phase.
From the point of view of materials, the process of in situ polymerization of CBT into PBT is
studied through a wide range of tests. Material formulations and process parameters are determined for
the achievement of high molecular weight and sufficient toughness. The best results are obtained with the
uncatalyzed monomer, which ensures an easier process control, added after melting with butylchlorotin
careful exsiccation of raw materials to avoid

hydrolytic degradation need;
carrying out of polymerization in the molten state,
avoiding simultaneous solidification, which means
at a temperature of 190200 C maintained for 10
min;
the addition of 10 wt% PCL if carbon fibers are
required to provide sufficient toughness and avoid
cracking during solidification.
Under the above conditions, a PBT is obtained that
melts at 215 C, or at 196 C if blended with PCL.
Fibre impregnation is very easy thanks to the low resin
viscosity.
As a conclusion, interesting industrial applications
can be envisaged for reactive thermoplastics, with
fiber reinforcements, to be processed by RTM, if a
careful process control is accomplished. Further investigations will regard the application of the knowledge
on these materials for the obtaining of a moulded part
in the rapid manufactured mould.
REFERENCES
Gatto, A., Iuliano, L., Ravera, M. & Violante, M.G. 2003.
Employment of step milling for the manufacturing of
moulds for resin tranfer moulding for infusion. Comp. 03:
Advances in Composite Technology, Proc. of 5th Int. Symposiumn on Advanced Composites, Corfu, Greece, 57
May 2003: 110.
Hutchinson, J.R., Schubel, P.J. & Warrior, N.A. 2011. A
cost and performance comparison of LRTM and VI
for the manufacture of large scale wind turbine blades.
Renewable Energy 36(2): 866871.
International Organization for Standardization, Geometrical
product specifications (GPS) Acceptance and reverification tests for coordinate measuring machines (CMM)
in: Part 2: CMMs used for measuring linear dimensions,
2009.
Ippolito, R., Iuliano, L. & Gatto, A. 1996. EDM tooling by
solid freeform fabrication and electroplating techniques.
Proc. 7th Solid Freeform Fabrication Symposium, Austin,
1214 August 1996: 199206.
Iuliano, L., Gatto, A. & Marchiandi, G. 2001. Step milling
for prototyping of large and precise part. 7th European
Conference on Advanced Material and Processes, Proc.
of Euromat 2001, Rimini, Italy, 1014 June 2001.
Kendall, K.N., Rudd, C.D., Owen, M.J. & Middleton, V. 1992.
Characterization of the resin transfer moulding process.
Composites Manufacturing 3(4): 235249.
Kim, S.K. & Daniel, I.M. 2005. Transient gas flow technique
for inspection of fiber preforms in resin transfer molding.
Composites Part A: Applied Science and Manufacturing
36(12): 16941699.
Norman, D.A. & Robertson, R.E. 2003. The effect of fiber
orientation on the toughening of short fiber-reinforced
polymers. Journal of Applied Polymer Science 90(10):
27402751.
Pages, J., Salvi, J., Garcia, R. & Matabosch, C. 2003.
Overview of coded light projection techniques for automatic 3D profiling. Robotics and Automation, 2003.
Proc. of ICRA 03. IEEE International Conference 1:
133138.
420
Palardy, G., Hubert, P., Haider, M. & Lessard, L. 2008.

Optimization of RTM processing parameters for Class A
surface finish. Composites Part B: Engineering 39(78):
12801286.
Park, C.H. & Saouab, A. 2009. Analytical Modeling of
Composite Molding by Resin Infusion with Flexible
Tooling: VARI and RFI processes. Journal of Composite
Materials 43(18): 18771900.
Perry, K.E. & McKelvie, J. 1995. Reference phase shift
determination in phase shifting interferometry. Optics and
Lasers in Engineering 22(2): 7790.
Potter, K.D. 1999. The early history of the resin transfer
moulding process for aerospace applications. Composites
Part A: Applied Science and Manufacturing 30(5):
619621. van Rijswijk, K. & Bersee, H.E.N. 2007. Reactive processing of textile fiber-reinforced thermoplastic
composites An overview. Composites Part A: Applied

Science and Manufacturing 38(3): 666681.
Verrey, J., Wakeman, M.D., Michaud, V. & Mnson, J.A.E.
2006. Manufacturing cost comparison of thermoplastic and thermoset RTM for an automotive floor pan.
Composites Part A: Applied Science and Manufacturing
37(1): 922.
Wang, Z., Bryanston-Cross, P.J. & Whitehouse, D.J. 1996.
Phase difference determination by fringe pattern matching. Optics & Laser Technology 28(6): 417422.
Mohd Ishak, Z.A., Leong, Y.W., Steeg, M. & Karger-Kocsis,
J. 2007. Mechanical properties of woven glass fabric reinforced in situ polymerized poly(butylene terephthalate)
composites. Composites Science and Technology 67(34):
390398.
421

A creative solution for rapid and inexpensive model-making in product

design processes
Fernando Veloso, lvaro M. Sampaio & Filipe Chaves
Polytechnic Institute of Cvado and Ave, Campus do IPCA, Barcelos, Portugal
Ricardo Simoes
Polytechnic Institute of Cvado and Ave, Campus do IPCA, Barcelos, Portugal
Institute for Polymers and Composites IPC/I3N, University of Minho, Campus de Azurm, Guimares, Portugal
ABSTRACT: Rapid prototyping (RP) is an approach for automatically building a physical object through solid
freeform fabrication. Nowadays, RP has become a vital aspect of most product development processes, due to
the significant competitive advantages it offers compared to traditional manual model making. Even in academic
environments, it is important to be able to quickly create accurate physical representations of concept solutions.
Some of these can be used for simple visual validation, while others can be employed for ergonomic assessment
by potential users or even for physical testing. However, the cost of traditional RP methods prevents their use
in most academic environments on a regular basis, and even for very preliminary prototypes in many small
companies. That results in delaying the first physical prototypes to later stages, or creating very rough mockups which are not as useful as they could be. In this paper we propose an approach for rapid and inexpensive
model-making, which was developed in an academic context, and which can be employed for a variety of objects.
1
1.1
INTRODUCTION
Framework and motivation
Rapid prototyping is an automated process for building

a physical object through solid freeform fabrication
(Chua et al. 1998). This technology has paved the way
to new paradigms of product development (Bartolo
et al. 2005). Nearly all product design and development
(PDD) processes require the production of models
or prototypes at some point. These can be analogic
or digital in nature, and their purpose may be visual or
functional. In addition, they may be used at different
stages of the product development process, such as in
concept development, detail design, etc.
Despite the positive aspects of digital models in current PDD processes, in most cases, a physical model is
extremely useful. However, creating accurate physical
models is not always a trivial task. Parts with relatively complex geometries usually require expensive
prototyping solutions. This also implies access to the
specific technologies that enable creating such prototypes. Thus, despite the large variety of solutions
available for producing quality (even functional) prototypes, this requires either investment in specialized
equipment in-house, or having to rely on subcontracting (added time, cost, and other constraints). For micro
or small companies, it may be very difficult to justify
these costs. Also, in the academic context of industrial
design degrees, it is nearly impossible for students to
support the costs of prototyping the different products
they develop during their training.
Thus, there is definitely room for simple, rapid and

inexpensive prototyping solutions, even if the resulting models are rough and not as accurate as standard
prototyping technologies. This paper presents such
an approach, which was developed by a design student
in cooperation with faculty. This solution was based on
the principle of vacuum-assisted thermoforming and
works effectively for thin-walled hollow parts, such
as containers. This solution can be used to produce
either transparent or opaque models. It enables obtaining fairly accurate and complex shapes that would be
extremely difficult to produce manually, and uses a
very simple home-made apparatus. Namely, a support
structure in wood, a plastic sheet heated in a typical
kitchen oven, a polyurethane model for providing the
negative mold shape, and a vacuum-cleaner to ensure
the sheet acquires the mold form.
1.2 Paper structure
In this paper, we start by discussing the solution concept. Subsequently, we exemplify how to employ this
solution in a step-wise approach, including the parameters that were found to work most effectively (heating
time and temperature, wall thicknesses, etc). Finally,
we demonstrate how it was successfully validated for
producing models for 3 novel products in an industrial
design project-based course. This solution is a single
example of the type of ingenuity that designers must
have in order to create useful models with very limited
resources.
423
RAPID PROTOTYPING
Materializing ideas into solid models has been very

useful in PDD. The ability to feel, touch and grab an
object rather than visualizing it in a screen or sheet of
paper is of the upmost importance when analyzing its
features and sometimes to compare with other options.
It is also important to obtain models for manufacture
processing, such as casting. Therefore prototyping is
an important technology, which has been evolving into
new horizons and applications.
Three known phases that lead to Rapid Prototyping
can be distinguished as Manual Prototyping, Soft or
Virtual Prototyping and Rapid Prototyping (Chua et al.
2010).
Traditional methods for obtaining models, such as
manual molding, wood molding and scale modeling
were used for many years requiring however special
skilled professionals, with manual abilities. This was
associated with no reproducibility and time consuming processes. With the computational development
of the last decades of the twentieth century, Computer
Numerical Control (CNC) machines played an important role in the industry and were also used to obtain
models from machining materials such as wood, steel,
polymers and clay. However CNC machining was still
an expensive and time consuming process for obtaining models in prototyping because of the tools, the
materials and also specialized workers. Also, geometry complexity should be avoided leading to limitations
in obtaining models.
Trying to overcome this limitations, and making available prototyping, several Rapid Protoyping
(RP) technologies have been developed since the first
Stereolitography Apparatus (SLA) was commercialized by 3D Systems Inc. in 1987 (Chua et al. 1998).
At least 30 different RP technologies were developed and commercialized since 1988 (Chua et al.
2010), such as Selective Laser Sintering (SLS), Solid
Ground Curing (SGC), Laminated Object Manufacturing (LOM), 3-Dimensional Printing (3DP), Fused
Deposition Modeling (FDM), Solid Creation System
(SCS), Solid Object Ultraviolet-laser Plotter (SOUP),
Selective Adhesive and Hot Press (SAHP), MultiJet Modeling system (MJM), Direct Shell Production
Casting (DSPC), Multiphase Jet Solidification (MJS)
and Ballistic Particle Manufacturing (BPM). Biomanufacturing (BM) based on RP is being perfected
in order to obtain alternatives for human tissues or
organs (Yongnian et al. 2009). Nowadays there are also
being developed web-based Rapid Prototyping systems, relying on the network environment that allows
spreading 3D models all over the World (Hongbo
2009).
This new technology (RP) brought a completely
different approach that, instead of removing, added
layers of material with a bi-dimensional drawing,
ultimately resulting in a three-dimensional object.
Despite the advantages in production time reduction, reproducibility and process simplification,
there are some drawbacks associated with some RP
technologies like the surface finishing and the price of

some materials like resins that can reach up to 1.000
per Kilogram (Alves et al. 2001).
In the future, more user-owned RP machines will
be used, as result of the reduction in costs of acquisition and operating this type of technology and two
distinct markets will be available: the workshop/modelmaking machine for production accurate and functional parts and the design-office 3 DP for easily
generating parts for design verification (Pham & Gault
1998). The technology shown with this paper takes this
second market idea to an extreme of Do It Yourself,
with a very simple design and affordable prototype
manufacturing, ideal for small budgets.
3 ACADEMIC CONTEXT
3.1 RP in the academic context
Creating a prototype is a crucial step in the design
process. One of the main reasons why prototyping is
important in industrial design is due to the possibility
of evaluating and testing the design itself.
Since ideas and drawings of a design are clearly
abstract representations, in a 2D representation almost
anything is representable and possible. By creating a
prototype we have a 3D physical model that allows
the designer to determine which aspects are valuable
and which features need to be improved, changed, or
rejected. In the process, it is also possible to find gaps
that, on paper, were not perceptible. In an initial stage
of the design process it is also used to communicate
ideas easier and faster.
Prototypes are also essential as a way to evaluate the
new concept through user trials. Since all user tests are
conduct with a prototype, the results are more accurate
and the data collected can improve the project before
the detail design and production phases.
3.2 Context for this project
The Industrial Design Project curricular unit is part
of 3rd year of the Industrial Design undergraduate
course of the School of Technology of the Polytechnic
Institute of Cvado and Ave.
This curricular unit is planned in a project-based
learning environment. Since, design students feel more
motivated when they identify the applicability of
the learning contents and while developing practical
projects (Simoes & Sampaio 2009).
The objective of this curricular unit is the accomplishment of design projects with intention to develop
new products based in a user-centered methodology, with focus in the market. Throughout this usercentered methodology, knowledge about, creativity,
innovation, market and the importance of validating
the solution with the user is valued.
These projects are developed by groups of four students. The Project is developed in phases, including:
Development of ideas; Selection of ideas; Development of concepts; Selection of concepts; Construction
424
of models, Test of models, User evaluation, Final

selection, detail design and prototype.
The workflow of these projects is very diverse. It is
performed in a divergent-convergent method as proposed by Pugh (1991). As projects were developed
by groups of four elements it was necessary to establish internal competition between the elements. In this
way groups were divided in two sections of two elements. When divergent tasks needed to be performed
(development of ideas; development of concepts), sections make them. Then for the convergent tasks in the
product development (several selections), the sections
group to select the best proposal for the next phase.
The main focus of the development project is center
in the evaluation of the product by users. After selecting the best concept, groups are encouraged to build
models of the solution. The purpose for this phase is to
develop a physical model with the real shape, size and
functionality of the solution for the project. The ultimate aim is to evaluate the interaction between the user
and the product, focus in three different aspects: effectiveness (being able to complete a task), efficiency (the
amount of effort required to complete the task) and satisfaction (the degree to which the user was happy with
his experience while performing the task). In this way
students can compare the state of the product being
designed to the desire state, and established relations
between tasks and properties in order to generate a
solution that fulfills the overall objective with the user
in the center of the development process.
Through this development approach students need
to develop capacities of construction models for the
evaluation needed. These models are constantly being
modified as the inputs from the users are proven
worthily. So, the importance of building a model is
paramount but the way to build it is also very important
since budgets are always tight.
Since several changes in the models could be necessary in the evaluation process, and some projects
are, by their shape, material or functionality, very difficult to build (difficulty is here seen as an economic
and time problem) students sometimes need to be creative in order to innovate in the task of building their
model.
4
4.1
PROPOSED APPROACH
Concept and developed tool
Vacuum forming was one of the first thermoforming

technologies, developed in the early part of the 20th
century (Throne 2008). This technology was developed by 1960s and applied industrially recurring to
expensive machines that can only be profitable for
large production sets.
Three different approaches are viable: (a) using a
female mold with holes to extract the air and drown the
plastic sheet towards its walls, named vacuum forming; (b) pressure forming that uses air pressure to force
the plastic sheet to the mold walls; (c) using a male
Figure 1. Key components of the developed prototyping

tool.
mold with holes to extract the air, pulling the plastic

sheet into the walls is known as drape forming (Chiles
et al. 1996). In this last case, the interior surface of the
part is in close contact with the mold and will have a
better surface finish and detail than the outer one.
This method requires materials with good hot melt
strength, allowing stretching when heated below the
melting point. Also, the corner radii should be at least
four times the material thickness and the ratio of draw
length to part diameter shall not exceed 2:1 (Chiles
et al. 1996). These limitations should be taken into
account to obtain good parts, but rapid prototyping
of curved surfaces with a variable geometry vacuum
forming mold is shown to be a feasible process by
Kleespies & Crawford (1998).
Based on the drape forming principle, simpler
designs have been developed and used in schools to
simulate thermoforming and demonstrate the technology. These designs are also suitable to obtain small
prototypes for evaluation and product development
purposes, requiring less detail and a sparse surface
finishing.
Recurring to home appliances such as a vacuum
cleaner to promote air extraction through a small box
with a punctured surface and a kitchen hoven to heat
the plastic sheet, a homemade vacuum forming is
easily implemented with very low costs.
The specific developed solution is represented in
Figusre 1. The different components indicated in that
Figure are:
1.
2.
3.
4.
5.
6.
7.
425
Melamine panel (16 mm thick)

Melamine panel with hole for vacuum cleaner
Melamine panel side panels
Perforated Platex hardboard
MDF plate
Polypropylene (PP) sheet
Melamine panel mold support
The materials used to build the prototyping tool

include:
Perfurated Platex hardboard with dimensions

0.610 m 1.220 m 3 mm
Melamine panel 1.22 m 2.44 m 16 mm
MDF plate, 3 mm thick
Polypropylene sheet 0.5 m 0.5 m 0.8 mm
Adhesive aluminum tape
Self-tapping screws
Aside from miscellaneous shop tools, an electric

drill, and a hot glue gun, used in the tool preparation
process, the prototyping process required a vacuum
cleaner (1800 W) and an electric oven (0.79 kWh).
In addition, a high density polyurethane foam block
(density 100), of size 0.5 m 0.5 m 0.10 m, was
used to produce the mold.
Figure 2. First stage of the prototyping methodology:

assembling the mold onto the mold support tool.
4.2 Tool assembly

The mold support is assembled first. Essentially, the
support is a melamine 400 400 mm width box with
16 mm thickness side panels. One of the sides should
exhibit a hole to connect the vacuum cleaner. The
bottom of the box is a 3 mm thick MDF plate, and
the top is 3 mm perforated hardboard (Platex ). The
mold support corresponds to components 1 through
5 in the bottom part of Figure 1.
Second, the forming sheet frame should be assembled. This frame has the same dimensions as the
mold support, but is open at the top and the base
(components 1 and 3 in the top part of Figure 1).
4.3
Prototyping methodology
For this procedure, a mold of the desired part is

required. This mold can be prepared by hand by filing and sanding a block of rigid polyurethane, and
should have two halves, dividing the part. The two
halves of the mold will have to be used individually
for prototyping the two respective half-parts and these
resulting half-parts must then be joined through localized heating or (typically) simply gluing the two parts
together.
The mold should be assembled onto the mold support tool described in the previous section, as shown in
Figure 2. Note that the mold is not placed directly on
the hardboard but on a 16 mm melamine plate which
is then attached to the hardboard. This plate keeps the
mold separate from the hardboard, enabling the plastic sheet to adhere to the entire external surface of the
mold.
Subsequently, the polypropylene (PP) sheet must be
attached to the frame, using self-tapping screws. The
frame is then placed inside the oven, with the sheet on
top, at approximately 3 cm from the heating element,
with the oven set for a temperature of 220 C. This is
shown in Figure 3.
The adequate exposure time and oven temperature depend substantially on the specific plastic
sheet employed. Thus, some trial and error is always
Figure 3. Attaching the PP sheet to the frame and heating

the frame.
required. In our case, we achieved the desired results

by heating for 3 minutes. We have found that softening
of the sheet can be controlled visually; for our sheet,
the center of the sheet deformed vertically by as much
as 5 cm compared to the original position.
It is very important to point out that for the step
of heating the plastic sheet, extreme care should be
used in three key aspects. First, personal protection
against fumes possibly released from the plastic sheet
should be worn at all times in the form of a facial
mask. Second, adequate venting must be ensured; the
oven should have an air extractor or, if not available,
the oven should be used in an open space. Third, heat
resistant gloves should be worn to prevent accidental
burns.
Once the sheet is sufficiently pliable, the vacuum
cleaner was turned on, the frame removed from the
oven, and placed as quickly as possible on top of the
mold, aligned with the mold support tool (as shown
in Figure 4). Some force should be applied to the top
edges of the frame, to improve contact with the mold
support tool. Once the plastic sheet adheres to the perforated hardboard around the mold, some vacuum is
achieved, and the sheet is immediately observed to start
conforming to the mold shape. Using a damp cloth,
some manual pressure can be applied at small features
of the mold, to achieve more accurate molding.
After this, the vacuum cleaner is turned off, and the
frame now with the formed plastic sheet is displaced from the mold support tool so that the mold
can be removed. This is shown in Figure 5. Using
rigid polyurethane for the mold prevents the plastic
sheet from sticking to it. The excess material is then
426
Figure 4. Placing the frame on the mold support tool and

applying vacuum to conform the plastic sheet to the mold.
Figure 7. Water container prototype.
Figure 5. Final stages of the prototyping procedure: extracting the conformed plastic sheet from the mold and removing
excess material.
Figure 8. Half water container prototype (raw from

molding).
Figure 6. Caption of a typical figure. Photographs will be

scanned by the printer. Always supply original photographs.
removed from the formed sheet with a scissors and, if

necessary for very small features, a scalpel.
The process is then repeated for the other half
of the mold. Finally, the two half-parts are glued
together. Any accessories or labels can then be applied
to produce the complete prototype.
The entire prototyping methodology is summarized
in Figure 6.
SAMPLE APPLICATIONS
Figure 9. Citrus squeezer (left) and produced prototype

(right).
5.1 Water container

One of the major objectives of the developed prototyping tool was to quickly produce a visual prototype for a
recently designed product, namely a 5 liter water container. Since the design is innovative and the product
features a series of small very important details, it was
necessary to obtain a prototype sufficiently accurate to
enable validation with different types of users. However, the geometry is quite complex and it should be
transparent, creating further prototyping difficulties.
The result obtained with the developed prototyping
tool is shown in Figure 7.
The model in Figure 7 already includes final manual operations (including adding the cap, gluing the
two molded halves together, and applying a label). In
Figure 8 it is possible to see a raw half as extracted

from the prototyping tool.
5.2 Other parts
In order to show the potential applications of this
simple prototyping approach, several household and
consumer products were used as molds for the proposed prototyping process. The results can be seen in
Figures 9 (citrus squeezer), 10 (soap holder), and 11
(several cell phones and PDA models).
It was found that parts which have only one surface are easier to produce and, provided they have no
extremely complex features, give good results. Parts
427
to be produced. However, it is not always feasible (for

time, availability, and even financial constraints), to
use those more complex techniques.
We have found that for academic purposes, the
proposed approach can be a very interesting solution.
ACKNOWLEDGEMENTS
Figure 10. Soap holder (left) and produced prototype
(right).
Foundation for Science and Technology, Lisbon,

through the 3 Quadro Comunitrio de Apoio and the
POCTI and FEDER programs.
REFERENCES
Figure 11. Front model of several cell phones and PDAs.
with holes also work very well, since these help the
vacuum process.
Finally, this process can also be used to prepare
hollow showcase models, for example, to place in
advertisement scenarios.
6
CONCLUDING REMARKS
The models obtained with the proposed prototyping

tool are perfectly adequate for the purposes of showcasing the products and even for usability trials. With
some manual tweaking, it is possible to obtain models
of similar quality to standard prototyping techniques.
However, the proposed methodology is much simpler, and the cost of the discussed prototyping tool
did not exceed 100 (including the PU foam block).
Subsequent prototypes have a cost of approximately
20 (including the PU foam block). If a real part
can be used as mold, such as the cases shown in
Figures 911, then the cost is merely that of the PP
sheet, which was about 3.
Obviously, the proposed methodology and tool do
not intent to replace standard prototyping techniques,
which enable much more complex and accurate parts
Alves, F., Braga, F., Simo M., Neto, R. & Duarte, T. 2001.
Protoclick. Edies Inegi.
Bartolo, P., et al. 2005. Virtual modelling and rapid manufacturing. Taylor & Francis.
Chiles, V., Black, S., Lissaman, A. & Martin, S. 1996.
Principles of Engineering Manufacture, 3rd Edition,
Butterworth-Heinemann.
Chua, C. K., Chou, S. M. & Wong, T. S. 1998. A Study of the
State-of-the-Art Rapid Prototyping Technologies. International Journal of Advanced Manufacturing Technology
14:146152.
Chua, C. K., Leong, K. F. & Lim, C. S. 2010. Rapid Prototyping: Principles and Applications, 3rd Revised edition,
WSPC.
Hongbo, L. 2009. Web-based rapid prototyping and manufacturing systems: A review. Computers in Industry 60:
643656
Kleespies, H. S. & Crawford R. H. 1998. Vacuum forming
of compound curved surfaces with a variable geometry
mold. Journal of Manufacturing Systems 17:325337
Pham, D.T. & Gault, R.S. 1998. A comparison of rapid prototyping technologies. International Journal of Machine
Tools and Manufacture 38: 12571287.
Pugh, S. 1991. Total Design-Integrated methods for successful product engineering. Addison-Wesley Pub. Co.
Simoes, R & Sampaio, A. M. 2009. The role of project-based
learning in product design and development education:
meeting the industry needs. 4th International Conference
on Higher Education Marketing, Guimares, 13 April.
Throne, J. 2008. UnderstandingThermoforming. 2nd Edition,
Hanser Verlag.
Yongnian, Y. et al. 2009. Rapid Prototyping and Manufacturing Technology: Principle, Representative Technics,
Applications, and Development Trends. Tsingua Science
and Technology 14, S1: 112.
428

Fabrication of a Biopsy Micro-Forceps prototype with Incremental

Sheet Forming
R. Perez-Santiago, M. Garca-Romeu & I. Bagudanch
Universitat de Girona-Departament dEnginyeria Mecnica i de la Construcci Industrial, Girona, Spain
ABSTRACT: The main goal of this work is the application of Incremental Sheet Forming (ISF) technology
tomanufacture one part of a biopsy micro-forceps prototype. Both, single point and two point ISF variants were
studied. The real device size is on the order of a few millimeters forcing the utilization of a multiplying scale
factor. As a consequence, the prototype is useful to assess the general characteristics of the design but not its
fitting inside the endoscope. The parts were manufactured using aluminum alloy and stainless steel. Process
simulations based on the finite element method allowed to decrease the number of experiments required to
obtain a successful component. The geometry of the produced part was compared with the target design in order
to evaluate the accuracy of the process. The preliminary results show the potential of ISF for the prototyping of
medical instruments manufactured with sheet metal processes.
1
INTRODUCTION
1.1 The biopsy micro-forceps

In order to identify a specific pathology, a physician
may examine the interior of a hollow organ or cavity
of the body with an endoscope. Additionally, a sample
of the tissue can be removed to be analyzed in the
laboratory. This operation is called a biopsy and can be
performed by the use of Biopsy Micro-Forceps (BMF)
inserted through the endoscope.
According to Yang et al. (2000): The ideal biopsy
forceps for Gastro-Intestinal (GI) endoscopy would
(1) provide adequate tissue specimens for histologic
evaluation, (2) have no mechanical problems, (3)
beeasy to use, (4) allow no possibility of crosscontamination to patients or healthcare workers, and
(5) be inexpensive.
The configuration of a typical BMF instrument
consists of three main parts: the handle, the manipulation flexible tube and the sharp fine blades which
are used to remove the tissue from the patient
(Fig. 1).
The set of blades consists of two half-shells, referred
as jaws in the rest of the study, and a system which
allows them to rotate as to facilitate opened-out and
closed functions.
The jaw set is made of metal, stainless steel in
the case of disposable instruments, in a size that
allows them to fit inside the endoscopy biopsy channel.
According to Danesh et al., (1985) GI biopsy channels
permit the use of forceps between diameter 1.8 and
3.4 mm.
Currently, a great deal of expense and long period
of time are required to manufacture the BMF because
complicated mechanical processing and manual operations are involved (Aoki et al., 1995).
Figure 1. Illustration of a commercial biopsy microforceps, showing its flexible tube and actuating handle.
At right, a close-up of the forceps is presented (Boston
Scientific, 2011).
Manufacturing metal processes for this kind of

components have relative advantages and disadvantages as manufacturing methods. In the work of Aoki
et al., they describe these pros/cons comparing etching,
discharge machining, machining (as cutting, grinding) and micro forming techniques. Their work is
focused on a combination of microforging and forming, obtaining the jaws of the BMF starting from a
wire.
There exist other techniques to obtain the BMF, for
instance the work of Aguirre et al., 2008 and Aguirre
et al., 2009. Their work is related to design and fabricate nanoparticulate enabled surgical instrumentsusing an iterative fabrication-design technique, being
this technique a modification of UV lithographytechnique. The application of this novel fabrication process and design optimization method for a
mesoscale forceps is also presented. In the same direction, Cohen et al., 2010 present the EFAB technology,
a patented method using a solid freeform fabrication process that is both additive and subtractive in
nature. The process involves three steps: selective electrodeposition of one metal, blanket electrodeposition
of a second metal, and planarization. This technology makes possible an unprecedented level of device
complexity at the sub-millimeter to millimeter scale,
including the creation of fully assembled mechanisms
429
Figure 3. SPIF Process map.

Figure 2. The two main ISF variants: (a) single point and
(b) two point incremental forming.
with independently moving parts, often obviating the

need for costly micro-assembly.
Despite the wide range of manufacturing processes
available for BMF manufacturing, the purpose of this
work is to use Incremental Sheet Forming process to
manufacture the jaws of a GI meso-forceps.The selection of the jaws has not been randomly done but is part
of a larger project aimed to develop, prototype and test
a new multiple BMF concept.
1.2
Incremental sheet forming
Incremental sheet forming (ISF) is a relatively new

technology useful to manufacture full or scale size
prototypes as well as small batch or one-of-a-kind
sheet metal products. Parts obtained by ISF may be
totally functional and have a geometrical complexity
which is difficult to attain with other forming processes (Jeswiet et al., 2005). Furthermore, it has been
proved that the localized deformation feature of ISF
extends the formability limits of the processed materials; in particular, deformations leading to failure in
conventional drawing can be widely overcome using
this technique (Shim & Park, 2001).
As described by Jeswiet et al. (2005), ISF technology is potentially available to any manufacturing facility having CNC machinery and CAD/CAM
software. Nevertheless, there is not yet CAD/CAM
software oriented for ISF. In consequence, several
design-manufacturing iterations are required before
attaining a satisfactory component.
The main components of ISF technology are
depicted in Figure 2: the blank or raw material in
sheet form is fixed by a clamping device. The blank
is progressively deformed using a hemi-spherical end
forming tool. The movement of the tool is driven by a
CNC machine or robot. Depending on the existence or
absence of a die, or support post, the two main variants
of the process are identified as Single Point Incremental Forming (SPIF, Fig. 2a) or Two Point Incremental
Forming (TPIF, Fig. 2b).
ISF can be applied in many fields, such as aeronautic and automotive industries and most recently
in the medical field (Ambrogio et al., 2005, Tanaka
et al., 2001). However, further development leading
to guarantee a good correlation of the part produced
with the design intent is required before ISF becomes

a widespread manufacturing process.
The remainder of this work will describe the
methodology followed to prototype the jaws of a
BMF. The gains obtained from the intensive use of
Finite Element Method (FEM) simulation allowed the
authors to decrease the number of experiments and in
turn to accelerate the prototyping process.
DESCRIPTION OF THE PROCESS CHAIN
The ISF basic process can be observed in Figure 3.

In case of complex shapes, the lack of accuracy
obtained with the basic process (SPIF)may force the
use of a partial or full die. Additionally, several iterations involving the modification of the tool-path may
be required before attaining a satisfactory part. FEM
simulation can be used to decrease the number of
iterations.As a result, an optimized procedure to manufacture complex parts with TPIF technology can be
obtained:
1. The process begins with a 3D model of the target
component.
2. A negative shape is obtained after subtracting the
jaw from a prismatic (stock) block.
3. The CAD/CAM software generates the contouring
tool-path.
4. Simulation of the process is used to foresee the
results obtained with the tool-path generated in Step
3 in combination with other process parameters like
blank dimensions and material.
5. Finally, if the results obtained in Step 4 are satisfactory, the actual manufacturing of the component
is realized.
The main steps are depicted in Figure 4. Further
details about the experimental and numerical methods
are described separately in the following section.
METHODS
3.1 Studied component and tool-path generation

The external dimensions of the 10 scaled jaw, in
mm, are presented in Figure 5. Experiments and simulations were conducted with 0.5 mm thick blank
sheet excepting an essay in 0.8 mm aluminum alloy.
430
Figure 4. TPIF optimized process map.

Figure 6. SPIF process experimental set-up.
sheets were 150 150 mm and the effective working

area was 120 120 mm. The forming was made with a
hemispherical Vanadis 23 steel tool of 6 mm diameter.
The spindle speed had a constant value of 600 rpm
and the feed rate was set to 1000 mm/min. The tool
depth step, z, was 0.2 mm. Cold forming oil basedlubricant was poured directly over the blank in order
to decrease friction effects.
3.3 Simulation setup
Figure 5. Dimensions of the 10 multiple BMF jaw (in
mm).
It should be remarked that the thicknesses employed

are not conformal to the scaling process.
A 3D model of the jaw was generated in Autodesk
Inventor. The jaw stock model was exported in a
neutral file readable by Goelan v5, the CAD/CAM
software utilized.
Inside Goelan v5, the neutral file stock model was
imported and the tool diameter and other process
parameterswere specified before generating the toolpath. Then, the trajectory of the tool was written in
CNC format using a post-processor compatible with
the Fidia control of the machining center.
Before moving forward with the physical and virtual processes, the correctness of the tool-path is
assessed with the simulation function embedded in the
CAD/CAM software.
3.2
All FEM models share the following features:An

explicit solver, LS-DYNA, running in a 4 CPU Windows PC was used.Symmetry simplifications were not
applied.In order to decrease the solution time, feed rate
and mass density scaling were utilized.
The mid surface of the blank was described by
a regular mesh of fully integrated linear shell elements using 5 integration points thickness-wise. Blank
element size is 1.5 1.5 mm.
External surfaces of forming tool were meshed with
shell elements in combination with a rigid material
defined with steel properties.
An elasto-plastic constitutive model was used to
simulate the behavior of the AA1050-H24 aluminium
alloy and the stainless steel SS304 used for the experiments. The elastic regime was based on Youngs
modulus and Poissons ratio while plasticity was modeled with the von Mises yield criteria and the Swift
hardening law (Eq. 1). The parameters of the hardening law were obtained from uni-axial tensile tests
conducted on both materials. The material properties
are summarized in Table 1.
Experimental setup
The experimental tests were conducted in a Kondia

HS1000 3-axis milling machine. A table-type
dynamometer Kistler 9257B was mounted in the worktable in order to measure the process forming forces.
The forces were acquired using a DaqBoard 505 data
acquisition card and the DaqView 9.0.0 software. A
fixture system compound by a bottom plate, four supports, a clamping plate and a top plate was bolted on
the force measurement system.
The blank sheet was placed between the clamping
and the top plate (Fig. 6). The dimensions of the blank
where K = strength coefficient of the material;

0 = initial strain; p = plastic strain; and n = strainhardening exponent.
Penalty based frictionless surface contact was
defined between tool and blank. Displacement of the
nodes defining the perimeter of the blank was blocked.
The x, y, z coordinates of the tool-path described
previously are translated into velocity curves for the
simulation using a self-made python routine.
431
Table 1. Parameters of the hardening law for the aluminum

alloy and the stainless steel.
Material
Young
Modulus Possons
(GPa)
ratio
K (MPa) n
AA1050 72.6
H24
SS 304
207
0.36
122.0
0.30
1557.5
0.027 0
0.594 0.055
Figure 8. Frustum of pyramid formed with SPIF ( = 65 ),
SS304, 0.5 mm thick, material.
Table 2. Comparison of the analytical (Eq. 2), maximum
experimental and numerical forces obtained in the formed
pyramidal frusta.
Figure 7. SPIF trials of the jaw using AA1050 in (a) 0.5 and
(b) 0.8 mm thick material.
4
4.1
RESULTS
Jaw manufactured with SPIF
4.1.1 Aluminum alloy tests

Initially, the jaw was manufactured in AA1050-H24
aluminum alloy using the single point ISF variant. Two
different blank thicknesses, 0.5 and 0.8 were essayed.
As shown in Figure 7a, the 0.5 component failed while
a sound component was obtained from the 0.8 mm
thick material (Fig. 7b).
4.1.2 Stainless steel tests
Before testing the stainless steel, it was necessary to
ensure that the forces developed during the process
respected the force limit of the ISF adapted machine.
Proposed by Aerens et al. (2009), Equation 2 can
be used to estimate the steady state axial force (Fzs )
of Uniform Wall Angle (UWA) conical frusta formed
with SPIF.
Forming angle
[deg]
Eq. 2
Fz [N]
Experimental
max. Fz [N]
Simulation
max.Fz [N]
45
55
65
890
859
734
1456
1593
1724
1747
2019
2284
where Rm = tensile strength of the material; t0 = initial

blank thickness; = wall angle; dt = tooldiameter;
and h = scallop height.
Although the authors (Perez-Santiago et al., 2011)
have studied the applicability of the same equation
in variable wall angle components, their accuracy for
arbitrary geometries like the BMF jaw is uncertain.
Thus, the forming force of three different UWA
pyramidal frusta was estimated using the tensile
strength of SS304 (669 MPa) in Equation 2. After
checking that the predictionswere below theforce limit,
the components were actually manufactured; the =
65 pyramid is shown in Figure 8.
The force was measured during the experimentand the results compared against FEM simulations.
Table 2 summarizes the force prediction (Eq. 2) and
the maximum experimental and FEM values.
Based on Table 2 results, it seems that Equation 2
is unsuitable for this material. On the other hand, the
comparison shows important discrepancies between
simulation and experiments that can be attributed to
the non-saturated hardening law utilized.
Beyond the differences, it can be noticed that the
FEM model overpredicted the experimental forces for
the three cases. Since the simulation of the jaw SPIF
process in SS304 material, 0.5 mm thick, showed a
peak force of 1219 N, the authors felt confident to
proceed with the physical testing.
Actually, the experimental forces were below 1000
N and a sound component was obtained as shown in
Figure 9.
4.1.3 Comparison between aluminum and stainless
steel tests
Figure 10 presents the FEM thickness distribution of
the jaws manufactured inAA1050 and SS304, 0.5 mm
432
Figure 9. Jaw manufactured with SPIF using SS304 in

0.5 mm thick material.
Figure 12. FEM results of the component obtained with the

TPIF single support die. (a) Capped cross section isometric.
(b) Longitudinal profile.
Figure 10. Comparison of thickness distribution for the

SPIF manufactured jaws in (a) AA1054 and (b) SS304, 0.5
thick, materials.
Figure 13. TPIF double jaw support die showing a close-up

of the modification in the geometry utilized to correct the
tool-path.
Figure 11. Longitudinal cross section of the formed midplane profile as obtained from the FEM simulation (cut for
clarity). Distance in different points indicate deviations from
the designed external profile.
thick, materials. Darker zones indicate higher thinning. The minimum thickness predicted by FEM is
0.17 mm (66% thinning) and 0.32 mm (36% thinning)
for the aluminum alloy and stainless steel respectively.
In both cases, the maximum thinning is found in the
frontal nose of the jaw; the same place where fracture
occurred in the AA1050, 0.5 mm thick,jaw (Fig. 7a).
In addition to maximum force and thinning, the
numerical results give valuable information about the
accuracy of the process. A comparison of the (midplane) formed to the (external) designed longitudinal
cross sections is presented in Figure 11.
Ideally, a gap of 0.25 mm, equivalent to the blank
half-thickness, should be measured between the two
profiles. However, it is evident that the SPIF process
fails to generate an accurate part especially at the ends
of the part where bending instead of stretching dominates. The former results led to the development of a
second iteration using TPIF.
4.2 Jaw manufactured with TPIF

4.2.1 First virtual process iteration
In order to eliminate the undesired bending in the
material, a blank supporting die, able to be allocated
between the bottom and top plate of the ISF fixture (Figure 6), was designed. Once again, the gains
obtained from the TPIF strategy can be assessed with
simulations before actual die manufacturing.
The first (virtual) TPIF iteration was based on a
die having a single cavity at the middle of the stock.
As depicted in Figure 12a, the die seems to limit the
deformation of the blank to the perimeter of the jaw.
However, the closer comparison of Figure 12b
shows that, although more uniformly distributed, the
geometrical deviations are of the same order of those
obtained from the SPIF process (Figure 11).
4.2.2 Second virtual process iteration
Based on the previous observations, a corrected toolpath was implemented in the second TPIF iteration.
Additionally, a double cavity die that allows the manufacturing of a pair of jaws in a single operation
was designed; both modifications can be observed in
Figure 13.
433
Figure 14. FEM results of the component obtained with the

TPIF double support die using a corrected tool-path.
Table 3. Geometrical deviations from the design for each
forming iteration. Results are based on FEM models after
subtracting the material half-thickness.
SPIF
TPIF 1st iter.
TPIF 2nd iter.
Zone 1
Zone 2
Zone 3
1.598
0.916
0.670
1.757
2.127
1.671
1.018
1.650
1.054
FEM results indicate a different thinning behavior

of the two materials, which leads to fracture in the case
of the aluminum alloy.
Our preliminary results show that SPIF is unsuitable
to accurately produce the studied component, forcing
the authors to implement the TPIF variant by means
of a supporting die.
Based on FEM simulation of the TPIF process, it
seems that the die limits the deformation of the material to the jaw perimeter. However, a correction of
the tool-path was required in order to close the gap
between the formed component and the design intent.
After manufacturing the die in aluminum alloy, the
actual TPIF process will be realized and the results
compared with those from the virtual study.
Future work will study the additional deviations caused by springback once the component is
unclamped from the fixture. Also, the miniaturization
process of the jaw will be initiated by scaling down
the jaw design, die cavity, material thickness and tool
diameter.
ACKNOWLEDGEMENTS
The authors gratefully acknowledge support from
the International Research Exchange for Biomedical Devices Design and Prototyping (IREBID-FP7PEOPLE-2009-IRSES-247476).
REFERENCES
Figure 15. TPIF double jaw support die.
Aimed to improve the geometrical accuracy of the

component, the correction procedure consists on modifying the 3D geometry of the negative jaw shape in
the stock. Specifically, two cylindrical shapes were
carved from the bottom of the cavity. The results of
the corrected trajectory can be observed in Figure 14.
Table 3 summarizes the geometrical deviations
obtained from each iteration.
A third TPIF iteration may be required to attain a
more accurate component. However, since the geometry modifications are required only for the tool-path,
the design of the die support can be completed. The
final die design assembled between the top and the
bottom plate is shown in Figure 15.
CONCLUSIONS AND FUTURE WORK
The capability of ISF to prototype medical instruments

has been tested in a BMF jaw. Both aluminum and
stainless steel allows were actually tested using the
SPIF variant.
Aerens, R., Duflou, J.R., Eyckens, P. & Van Bael, A. 2009,

Advances in force modelling in force modelling for
SPIF, International Journal of Materials Forming, vol.
2, no. 1, pp. 2528.
Aguirre M.E., Hayes G., Yuangvai C., Frecker M., Adair J.,
Antolino N.: Fabrication and Strength-Based Design of a
Meso Forceps. 33rd Mechanisms and Robotics Conference, Parts A and B ASME 2009 Proceedings Vol. 7: doi:
http://dx.doi.org/10.1115/DETC200987557.
Aguirre M.E., Hayes G., Frecker M., Adair J., Antolino
N.: Fabrication and Design of a Nanoparticulate Enabled
Micro Forceps. 32nd Mechanisms and Robotics Conference, Parts A and B ASME 2008 Proceedings Vol. 2: doi:
http://dx.doi.org/10.1115/DETC200849917.
Ambrogio, G., De Napoli, L., Filice, L., Gagliardi, F., Muzzupappa, M., 2005, Journal of Materials Processsing
Technology. Vol. 162163, pp. 156162.
Aoki I.,TakahashiT., Mihara S.,YamagataY., HiguchiT.:Trial
production of medical micro tool by metal deformation
processes using moulds. Proc IEEE: pp 344349(1995)
Radial Jaw4 Biopsy Forceps, 2011, Boston Scientific,
Retrieved from: <http://www.bostonscientific.com/Device.
bsci?page=HCP_Overview&navRelId=1000.1003&
method=DevDetailHCP&id=10005381&pageDisclaimer
=Disclaimer.ProductPage>.
Cohen A., Chen R., Frodis U., Wu M-T., Folk C.: Microscale
metal additive manufacturing of multi-component
medical devices. Rapid Prototyping Journal, 16/3
pp. 209215 (2010)
434
Danesh, B., Burke, M., Newman, J., Aylott, A., Whitfield,

P., Cotton, P., 1985, Comparison of weight, depth, and
diagnostic adequacy of specimens obtained with 16 different biopsy forceps designed for upper gastrointestinal
endoscopy, Gut;26:227231 doi:10.1136/gut.26.3.227,
vol 54/2, pp. 623650.
Perez-Santiago, R., Bagudanch, I., Garca-Romeu, M.L.,
2011, Force modeling in single point incremental forming of variable wall angle components, Key Engineering
Materials.Vol. 473, pp 833840.
Shim, M. & Park, J., 2001, The formability of aluminium

in incremental forming, J. of Mat. Proc. Tech., 113, 654
658.
Tanaka, S., Nakamura, T., Hayakawa, K., Nakamura, H.,
Motomura, K.,2005, Proceedings of the 8th International
Conference on Technology of Plasticity.
Yang, R., Serena, Ng., Nichol, M., Laine, L., 2000, A cost
and performance evaluation of disposable and reusable
biopsy forceps in GI endoscopy, Gastrointestinal
Endoscopy.
435

Using additive manufactured tooling in the fabrication of

poly (L-Lactide-co-Glycolide) implants
S.F. Khan & K.W. Dalgarno
School of Mechanical and System Engineering, Newcastle University, UK
School of Mechatronic Engineering, Universiti Malaysia Perlis, Malaysia
M.J. German
School of Dental Sciences, Newcastle University, UK
ABSTRACT: Rectification of defective bone structures due to trauma or deformation has benefited from the
introduction of mostly rigid metal prostheses. However, the high stiffness of metal implants can cause stress
shielding in bone restoration. In addition a second operation is required for the removal of the metal implants,
screws, plates, and rods once the bone has healed. This paper discusses an adaptable solution where a customised
implant is defined which is secured by adhesion and resorbable to enable tissue restoration without a second
operation. This paper describes research on methods of creating customised structures from poly (L-lactideco-glycolide) (PLGA). An existing 3D mandible model in stereolithography file format was used as the basis
for this research which a rapid tooling approach is used to create a mould for sintering PLGA particles. The
tooling was developed through a stl file of a mandible being imported into a NURBS-based CAD modeller.
The modeller was used to create a 3D solid patch model for use as an implantable fracture fixation component.
Based on the 3D patch, the tooling was designed and then fabricated by using Stereolitography. Granules of
PLGA were placed inside the mould and heated under low pressure to produce the scaffolds with three different
formulation of PLGA used in this study. Consolidation of the PLGA occurred under constant pressure with a
sintering temperature of 73 C for 2.5hours. Mechanical testing was performed on samples and the results were
comparable with conventional processing of PLGA. The geometrical shape of the fabricated PLGA implant
resembles the desired fixation component and fits accurately to the fracture site. This work demonstrates that
this approach is a viable route in fabricating customised PLGA structures for orthopaedic surgery.
INTRODUCTION
Permanent metalic fixations have been used as a standard despite the drawbacks. Debate on the use of
permanent metal based devices for internal fracture
fixation have been on going since the introduction of
internal fixation concept back in the mid-19th century. Associated drawbacks include migration of the
device cause by bone apposition to the device, palpability, bone atrophy due to stress shielding, infection,
thermal sensitivity, and general discomfort, increasing rate of infection due to colonization of implants,
permanent and rigid brought discomfort growth and
radiopacity of the material can interfere with proper
radiological evaluation. Often a secondary removal
operation is required once the fracture is healed and
causing a new surgical procedure with new risks.
All these drawbacks had driven the development of
bioresorbable materials as a possible replacement for
metallic devices. The use of a bioresorbable bone plate
would achieve its function until the fracture is fully
healed before being naturally eliminated. The degradation time of biopolymers ranges between 1 month
to greater than 24 months depending upon the formulation and composition of the biopolymers (Sabir
et al., 2009). Polylactide and polyglycolide, as well as
copolymers polyglycolide-co-polylactide and poly(Llactide-coD,L-lactide) (P[L/DL]LA) are most often
used in the manufacturing of bioresorbable plates and
screws which follow the conventional method of fracture fixation. The bioresorbable plates need to be bent
and set into the required shape at room temperature
or temporary heating to a temperature greater than the
glass transition for adaptation to the bone geometry
and and the screws are not self-tapping which render them unsuitable for thin bone regions (Laughlin
et al., 2007; Bell and Kindsfater, 2006; Ylikontiola et al., 2004). Furthermore, the various methods
used, such as solvent casting, phase separation, freeze
drying, polymerization, gas foaming, salt leaching,
sintering, injection moulding and compressing moulding, are suitable in producing standard shapes. In this
437
Table 1.
Ratio of PLA-PGA co-polymer.
PLGA materials
Co-polymer ratio
Average Molecular
weight Mw
P2191
P2066
430471
50:50
65:35
85:15
40,00075,000
40,00075,000
50,00075,000
study, fabrication of customised PLGA 3D structure

by sintering under constant pressure was proposed.
METHODOLOGY
The proposed study in the fabrication route for patient

specific PLGA fixation plate has been performed in
two parts. The initial part has been used to evaluate the
manufacturing parameter which was followed by rectangular specimens creation for assessing material and
mechanical properties. In the final part a mandibular
geometry has been created in order to assess the ability
to create physiologically shaped parts and the ability
to maintain shape definition throughout the process.
The application of bioresorbable apparatus has been
used in maxillofacial surgery to achieve the application of a biocompatible material. Reported of studies
have indicated that fractures of the mandible account
for 3670% of all maxillofacial fractures (Deogratius
et al., 2006, Sojot et al., 2001, Ellis et al., 1985,
Brook and Wood, 1983, ). The fixation plate fabricated
was customised to fit the geometry of the fractured
mandible.
2.1
Figure 1. PLGA granules in SLA mould.
Fabrication and assessment of PLGA

test specimens
Figure 2. Heating profile for sintering process.
Figure 3. Bending test diagram and the actual test on

universal testing machine.
PLGA materials were obtained from Sigma-Aldrich in

granule form and used as is during the experiments. 3
different ratio of PLGA with almost the same average
molecular weight were used as shown in Table 1.
In order to obtain suitable temperature for fabrication of different ratios of PLGA used in this
study, DSC analyses were performed using Mettler
Toledo 823e. The glass transition temperatures of the
PLGA materials were obtained from the DSC analysis. Mouldings of the testing specimens were carried
out using SLA moulds produced by Paragon Rapid
Technologies Limited UK. Mould release agent was
applied to the mould to ease the release of the moulded
test specimens. The granules of PLGA were placed
manually inside the SLA mould cavities (Figure 1).
Fan assisted oven was preheated together with compression weight to 73 C. The completely filled mould
was then put inside the oven and compression weight
was applied onto the mould assembly. It was then
heated to above the glass transition temperature of
PLGA and maintained at that temperature for 2.5
hours, before the heater was switched off and the oven
was allowed to cool down at its natural rate. The sintering under constant pressure of the PLGA specimens
follows the heating profile as shown in Figure 2. Once
the filled mould was at room temperature, the test
specimens were demoulded.
The mechanical properties of the test specimens
have been assessed through three point bending tests.
The bending test was conducted on an Instron 5567
universal testing machine using a load cell of 1kN with
loading speed of 0.5 mm/min. Testing was based on BS
EN ISO 178:2010. The range of thickness and width
for the specimens are from 1.869 mm to 2.924 mm and
5.027 mm to 5.267 mm respectively. The overall length
is 40 mm and the span length is 25 mm. The layout of
the test is as shown on Figure 3.
Motic microscopes with Motic Images Plus 2.0
software were used to capture and measure the type of
bending failure in the sintered test specimens. Further
DSC analysis were performed to ensure that the fabrication route does not significantly alter the property
of PLGA.
438
Figure 4. Step in generating solid plate (a) creating polynomial curves (b) solid plate from surface mesh.
Figure 6. CAD model of the moulding assembly.
Bioresorable material PLGA was use to fabricate

the fixation plate. The manner in which the plate would
be attached the fracture site is by the use of adhesive
bonding. Heiss et al. indicated one of the potential
usage of medical bone adhesive in traumatology and
operative orthopedics is in the area of joining together
fractures with a joint bridging fixation (Heiss et al.,
2006). The customised fixation plate would be fixed
to the fractured mandible using bio-adhesive and the
plate would be able fit tightly with adhesive to the
mandible.
Figure 5. FDM model of fixation plate.
2.2
Design and fabrication of fracture fixation

implantable device
A 3D computer model in stereolithography (STL) format of a mandible was obtained from 3D Systems. A
fixation plate/patch was design for a simulated fracture
on the body region of the mandible as this is indicated
to be the most likely region of fracture by Fridrich and
associates which showed that most fractures occur in
the body (29%), condyle (26%), and angle (25%) of
the mandible (Fridrich et al., 1992).
CAD modeler Rhino 4 with RhinoReverse was used
in designing the fixation plate. The surface geometry
of the site was reverse engineered using RhinoReverse.
It is then used to generate a solid CAD model of the
plate. The geometry of the plate which was customised
to the shape of the fractured site (Figure 4) was then
fabricated using FDM for form fitting purpose (Figure 5). The CAD model of the plate was translated
into STL format and imported into Stratasys Dimension SST768 fused deposition machine. CatalystEX
software processed the STL file into a CMB file that
was sent to the Dimension printer for printing. A layer
thickness of 0.33 mm was selected and commercial
ABS plastic was used during the fabrication.
Once the designed plate was satisfied to fit on the
fracture site, the CAD model of the plate was then
converted to STL format and imported into AutoDesk
Inventor for use in mould design stage (Figure 6).
The 3D model of the moulding assembly was later
translated into STL format and upload to an SL system for fabrication. The fabrication was out source to
Paragon Rapid Technologies Limited UK using materials from DSM Somos ProtoGen 18920 liquid
photopolymer.
The glass transition temperatures of PLGA 50:50,

65:35 and 85:15 range from 42 C to 65 C as shown
in Table 2. Based on this analysis, the SLA material selected for the mould was 18920 ProtoGen with
UV and thermal post-cure. This produced a moulding assembly which can withstand temperature of
up 97.5 C as the fabrication temperature of PLGA
3D structure investigated would be ranging between
65 C80 C. The sintering of PLGA microspheres to
produce 3D porous PLGA scaffolds was achieved
by using a mold and heated to 5 C above the glass
transition temperature (Tg)) for 24 h and cooled to
room temperature before de-molulding (Borden et al.,
2002). Hence in this study, attempt was made to
shorten the sintering time by applying pressure moulding system. The sintering of PLGA granules were
performed by pouring into a mould and heated at 73 C
for 2 hours with an applied pressure of 78 kPa to the
moulding system. Then it was allowed to cool to room
temperature before de-moulding.
In total, ten rectangular test specimens for each of
the different ratio of PLGA of co-polymer were measured as shown in Table 2. The average thickness and
width are in a range of between 2.102 mm to 2.568 mm
and 5.064 mm to 5.125 mm, respectively. The flexural
modulus rages from 1.9 GPa to 4.3 GPa. As reviewd
by Waris et al. (Waris et al., 2004), the usual reported
values of synthetic bioabsorbable polymers produced
by melt-extrusion, injection moulding or compression moulding for bending modulus is between 2
439
Table 2. Average value for properties of sintered test sample

of PLGA.
Ratio of
PLGA
copolymer
Average Average
Glass
Thick- Width
tranness
(mm)
sition
temper- (mm)
ature,
Tg ( C)
Average
Maximum
Flexure
Load
(N)
Average
Flexural
Modulus
(GPa)
50:50
65:35
85:15
4255
4460
4865
39.400
44.492
60.327
3.012
3.055
3.014
2.102
2.112
2.568
5.069
5.064
5.125
Figure 8. Comparison of virgin to sintered PLGA 65:35.
Figure 7. Brittle failure of PLGA.
and 7 GPa, while the tensile and bending strength is

between 40 and 140 MPa. Furthermore, these bioabsorable devices are usually brittle and flexible and only
used in areas of low stress.
The image shown indicated that the specimens
failure type is of brittle failure (Figure 7) where
the average distance from surface before failure is
0.5 mm. Different investigations on failure of meltmanufactured biopolymer also failed in the similar
trend (Engels et al., 2009; Grijpma and Pennings,
1994).
Comaparison of both virgin PLGA as received and
sintered PLGA for all three diferrent ratio were performed using DSC. Three cycle analysis weas carried
out in which the surface area under the curve of DSC
thermograms for both the 1st and 2nd cycle does not
show any significant variation in total energy between
virgin and sintered PLGA. The integral of all the thermograms in sintered PLGA show a difference of less
than 5% (Figure 8). This indicated that the processing route does not alter the properties of PLGA. It
can be concluded that the sintering temperature profile
with applied pressure had not altered the bioploymer
in which do have similarity with the finding of RothenWeinhold et al. using melt-manufcaturing by extrusion
or injection (Rothen-Weinhold et al., 1999).
Figure 9 shows the PLGA plate as fabricated using
the proposed fabrication route. The 2 mm fracture fixation plates have no loss of definition but was perhaps
a bit thick to attach to fracture jaw (Figure 10). However this is the same as if getting from off-the-shelf
in biometal fixation plate except the PLGA fixation
plate fabricated follows the exact shape of the patient
anotomy. The plate would be fixed to the fractured
jaw using bio-adhesive. An equivalent force needed
to hold the fracture together; as produced using conventional biometal plate and screw, can be achieved
by increasing a larger area of surface contact between
the PLGA plate and the patience mandible bone. This
Figure 9. PLGA before and after sintering in SLA mould.
Figure 10. PLGA 50:50(left) and PLGA 65:35(right) plate.
has been shown achieveable from investigation of

bioabsorbable P(L/DL)LA bone plate. The plates were
designed to less than half the volume of a titanium
strut-style plate and yet be as strong as when fixating fractures of the mandible body despite the polymer
material having only 6% of the stiffness of the titanium
(Gaball et al., 2011).
4
CONCLUSION
The integration of an AM tooling to assist in the

fabrication of customised PLGA 3D structure for
440
orthopaedic surgery proof to be feasible. A method

to fabricate bioresorble devices using a combined AM
tooling, sintering and applied pressure to granules of
PLGA in three diferrent ratiio of PLA and PGA was
developed. All three different PLGA co-polymer ratios
were sintered into fracture fixation plates without loss
of definition of the geometry and the property were not
altered. The mechanical properties were comparable to
standard bioresorble device.
The hardwares involved in the fabrication of the fixatiion plate are fairly fundamental and readily available
in most dental hospital except for the fixation plate
mould which can be out scource. This route can be
feasibily implemented in a clinical setting of a dental
hospital for fabrication of customised fixation plate.
REFERENCES
Bell, R. B. and Kindsfater, C. S. (2006) The Use of
Biodegradable Plates and Screws to Stabilize Facial
Fractures, Journal of Oral and Maxillofacial Surgery, 64,
(1), pp. 3139.
Borden, M., Attawia, M., Khan, Y. and Laurencin, C. T.
(2002) Tissue engineered microsphere-based matrices
for bone repair: design and evaluation, Biomaterials, 23,
(2), pp. 551559.
Brook IM, Wood N. Aetiology and incidence of facial
fractures in adults. Int J Oral Surg. Oct 1983;12(5):2938.
Deogratius BK, I. M., Farrid S. ( 2006) Epidemiology
and management of maxillofacial fractures treated at
Muhimbili National Hospital in Dar es Salaam, Tanzania,
19982003., Int Dent J., 56, (3), pp. 131134.
Ellis E 3rd, Moos KF, el-Attar A. Ten years of mandibular
fractures: an analysis of 2,137 cases. Oral Surg Oral Med
Oral Pathol. Feb 1985;59(2):1209.
Engels, T., Sntjens, S., Smit, T. and Govaert, L. (2009)
Time-dependent failure of amorphous polylactides in
static loading conditions, Journal of Materials Science:
Materials in Medicine, 21, (1), pp. 8997.
Fridrich, K. L., Pena-Velasco, G. and Olson, R. A. J. (1992)
Changing trends with mandibular fractures: A review of
1,067 cases, Journal of Oral and Maxillofacial Surgery,

50, (6), pp. 586589.
Gaball, C., Lovald, S., Baack, B. and Olson, G. (2011)
Minimally Invasive Bioabsorbable Bone Plates for Rigid
Internal Fixation of Mandible Fractures, Archives of
Facial Plastic Surgery, 13, (1), pp. 3135.
Grijpma, D. W. and Pennings, A. J. (1994) (Co)polymers
of L-lactide, 2. Mechanical properties, Macromolecular
Chemistry and Physics, 195, (5), pp. 16491663.
Heiss, C., Kraus, R., Schluckebier, D., Stiller, A. C., Wenisch,
S. and Schnettler, R. (2006) Bone adhesives in trauma and
orthopedic surgery, European Journal of Trauma, 32, (2),
pp. 141148.
Laughlin, R. M., Block, M. S., Wilk, R., Malloy, R. B.
and Kent, J. N. (2007) Resorbable Plates for the Fixation of Mandibular Fractures: A Prospective Study,
Journal of Oral and Maxillofacial Surgery, 65, (1),
pp. 8996.
Rothen-Weinhold, A., Besseghir, K., Vuaridel, E., Sublet, E.,
Oudry, N., Kubel, F. and Gurny, R. (1999) Injectionmolding versus extrusion as manufacturing technique for
the preparation of biodegradable implants, European
Journal of Pharmaceutics and Biopharmaceutics, 48, (2),
pp. 113121.
Sabir, M., Xu, X. and Li, L. (2009) A review on biodegradable polymeric materials for bone tissue engineering
applications, Journal of Materials Science, 44, (21),
pp. 57135724.
Sojot AJ, Meisami T, Sandor GK, Clokie CM. The epidemiology of mandibular fractures treated at the Toronto
general hospital: A review of 246 cases. J Can Dent Assoc.
Dec 2001;67(11):6404.
Waris, E., Ashammakhi, N., Kaarela, O., Raatikainen, T.
and Vasenius, J. (2004) USE OF BIOABSORBABLE
OSTEOFIXATION DEVICES IN THE HAND, The
Journal of Hand Surgery: Journal of the British Society
for Surgery of the Hand, 29, (6), pp. 590598.
Ylikontiola, L., Sundqvuist, K., Sndor, G. K. B., Trml,
P. and Ashammakhi, N. (2004) Self-reinforced bioresorbable poly-L/DL-Lactide [SR-P(L/DL)LA] 70/30
miniplates and miniscrews are reliable for fixation of anterior mandibular fractures: A pilot study, Oral Surgery,
Oral Medicine, Oral Pathology, Oral Radiology &
Endodontics, 97, (3), pp. 312317.
441

Developing Ti jewelry through additive manufacturing and

conversion technologies
B. Paiva & Rui J. Neto
INEGI, Porto, Portugal
Jorge Lino
FEUP, Porto, Portugal
ABSTRACT: Nowadays there are an enormous variety of titanium (Ti) jewelry artifacts. These products have a
high specific strength, are corrosion resistant, promote a warm feeling to the human touch and are easily surface
modified by temperature or anodization to acquire a sophisticated aesthetic colour. Ti is more durable than gold
or platinum, has a lower price and is a fashion material due to the successful use in challengeable applications.
This paper describes the importance of using additive manufacturing and conversion processes to develop
innovative jewelry titanium artifacts. Combining 3D modelling, stereolithography, conversion technologies with
silicone and low melting point metallic molds and precision casting in ceramic shells molds in controlled
atmosphere, we were able to manufacture complex shapes with shiny thin sections, that after an adequate heat
treatment exhibit a variety of attractive colors that are more difficult and expensive to obtain with existing
technology for manufacturing jewelry products (metal forming and machining).
INTRODUCTION
1.1 The importance of titanium

High specific strength (tensile strength/density), high
corrosion resistance and biocompatibility are the titanium main attributes. Space vehicles, satellites and
airplanes use it due to the high specific strength. It
is intensively employed in airplanes reactors due to
the creep resistance (till 550 C) and specific strength.
In oil rigs for Deep Ocean can be seen in the extraction tubes, due to the fatigue resistance (induced by
the waves), high specific strength and corrosion resistance to the salt water, debris and carried fluids. The
medicine sector currently uses it in distinctive medical
applications, due to the biocompatibility (is hypoallergenic) and corrosion to body fluids, and, when
compared with other potential materials, has the closest elasticity modulus to the human bone. It is also used
in components to process chemical products due to the
good corrosion resistance (Lutejering, et al. 2005).
With contributions in important human goods, parts
for penetration in the Deep Ocean and our galaxy and
push the medicine, science and industry to new high
levels of the human knowledge, Ti assumes itself as
a metal with a great impact on the recent humanity
history.
1.2 Titanium in jewelry
Jewelry is the art of producing jewels that in reality are
artifacts with ornamental characteristics, having as a
tenuous frontier the human body (traditional habitat).
Titanium manufacturing as a meaningful industrial

product only started after the second half of the XX
century. Its use in jewelry artifacts production rapidly
took place and the same happened with the world dissemination that was under deep changes in teaching
methodologies of decorative arts.
Until this moment, gold and silver, materials under
continuous development through the years, were the
base materials for the manufacturing of jewelry artifacts. These two noble materials were easily processed
by goldsmiths, artists or simple technicians due to the
lower melting point, simple foundry procedures, easily
welding, and much more important, its excellent cold
malleability. For Ti, the simple change of a ring size, it
is almost impossible. If, for any reason, gold and silver need to be cut, a simple plier can do it, whether
titanium demands specific and expensive cutting
equipment.
Seen from this perspective, Ti is in fact a nonfunctional material in terms of jewelry, but ably occupies a place among the precious metals. Besides
the well-known mechanical properties, its durability is
eternal (no erosion problems over time), is considered
inert, not reacting with water, or chemically with the
human body fluids and is hypoallergenic (skin reactions do not occur). Its resistance to corrosion is such
that its color and brightness last long periods of time
without change. On the other hand is synonymous
of modernity, which, although expensive is cheaper
than the gold or platinum. Presents in natural state a
silver color, very bright and has a high capacity for
surface oxidation (perhaps the more relevant property
443
perform single memorable artifacts or small usually

limited series. Only in the second part of the 70s,
beginning of the 80s Ti pieces began to appear, at common prices, in the jewelry sections of the big jewelry
stores. Nowadays, there are many artists to intensively
using Ti in their creations and diverse websites are
selling Ti artifacts.
Figure 1. Anodized Ti brooch with silver frame and gold pin
(www.photostore.org.uk/seFULLPF.aspx?WID=16561&
MAKERID=217).
Figure 2. Anodized Ti buckle for a belt, 1967 (www.

reflectionandfraction.co.uk/history.html).
in jewelry), allowing a surface with an endless range

of colors (Pietro 1999).
The first artistic approaches in Ti were done in
England in the 60s (The Art of Reflection and Refraction 2009) by a group of researchers led by J.B Cotton
from the corrosion section of the Department of
research and development of the Imperial Metal Industries (England), which presented several corrosion
studies in new materials, pioneering the development
of different anodization methodologies for Ti. In 1967,
jointly with Hayfield presented the publication entitled
Decorative Finishes on Titanium (Hayfield 1998).
This document was prepared with the intent of revealing and captivating at the same time decorative arts
teachers and students to the beauty of colors that
could be obtained using certain surface oxidation techniques, where each color was function of the oxide
layer thickness produced on its surface (Figure 1).
The oldest Ti jewelry piece (1967) was designed by
the student Ann Marie Shillito, in the School of Jewelry of the Institute of Arts and Design of Birmingham.
It is a buckle, machined from a Ti bar colored anodized,
for a belt (figure 2) (The Art of Reflection and
Refraction 2009).
In 1976 took place the first exhibition dedicated to
Ti, Electrum Gallery. This event had three young
artists (Ed de Large, Kevin Coates and James Brent
Ward) with already extensive work in Ti, all of them
from the Royal College of Arts (London) and with high
knowledge of how to work the Ti (Vads 2009).
Modernity in jewelry proclaims that all materials
must have the same weight, but in reality this does not
happened with Ti. Until 1975 it was seen as a very
special material being skillfully used by designers to
MANUFACTURING PROTOTYPES FOR

TITANIUM JEWELRY ARTIFACTS
What is necessary for the physical materialization of

a dream? Using Additive Manufacturing (AM), the
idea arises, the object takes shape quickly and the
machine makes it.AM is the term to describe the physical and fast manufacturing of objects directly obtained
from 3D CAD files. This manufacturing technology,
was established to quickly provide, in the developing
phase (product development), a better design communication (Warfel 2009). There is a wide range of AM
equipments dedicated to the world of jewelry, with
some machine manufacturers considering this a priority market. The jewelry, as an industry, is an economic
sustainable activity in continuous growth.
There are two different ways that can be used to produce final metallic prototypes using AM. The first way
uses dedicated equipment that works with Ti powders
that are deposited, layer by layer, until the prototype is
obtained in accordance with a .stl file generated from
a 3D CAD program. The second one is the AM (or
using a prototype obtained by any other process) of a
non-metallic prototype that is used to directly or indirectly convert it in the final metallic model(s) (Alves
et al. 2001, Vasconcelos et al. 2006a,b).
2.1 Direct conversion
In the direct conversion, a model, usually obtained
by an AM process, is directly converted into the
final metallic part. Processes such as Stereolithography (quick cast), Solidscape, SLS, or others, produce
a model that is lost during the conversion process.
This way, a metallic prototype can rapidly be obtained.
This procedure is very interesting, but only allows the
production of a single part because the model is lost,
which means that does not allows errors during all the
conversion stages (Alves et al., 2001). Of course that
several parts can be obtained if the same number of
prototypes are made.
2.2 Indirect conversion
In the indirect conversion, the prototype, obtained by
any traditional or AM process, is used to produce a
mold to cast a wax for investment casting. This way,
single or pre-series of parts can easily be obtained,
although, with a longer deliver time.
Another possibility is the direct production of the
mold where different parts could be injected. 3D Systems developed an epoxidic resin that has superior
mechanical properties to the common thermosets
444
Figure 5. Place the mother mold with gating system and

guides over the parting plane.
Figure 3. Casting of the silicone over the model, placed in a
box with the associated gating system (entrance of material,
gas escaping channels and raisers).
Figure 6. Casting the resin.
Figure 4. Cut the silicon mold along the parting plane to

remove the model.
resins. If necessary, metallic frames can be used to

reinforce the mold to withstand the high mechanical loads produced during the injection process. With
these molds short series of lost models in thermoplastic waxes with complex and accurate shapes can be
obtained. Although this seems very interesting, due to
the low thermal conductivity of these molds, only few
models can be injected during a working day, and so it
has only interest when speed is the goal.
In terms of jewelry, this process is very expeditious
for wax injection for production of small series of
models to be used in foundry by investment casting
(Alves et al. 2001).
2.2.1 Silicone molds
Any model can be used as mother model to produce silicon rubber molds, normally processed at
room temperature using vacuum systems.
This method allows the manufacturing of short
series of parts in rigid or flexible polymers. In the case
of jewelry, thermoplastic waxes are used to work as
lost models for the investment casting process.
There are two different silicon families for rapid
molds manufacturing. Addition silicones have the particularity of having a room temperature post-cure null
shrinkage, while condensation silicones are characterized by their lower price.
Usually, this is a manual process, although in the
last years several vacuum chambers were introduced
in the market with automatic mixing and casting in
controlled environment.The main stages of the process
are described in figures 3 and 4.
Figure 7. Casting the resin over the first half mold.
2.2.2 Resin molds

With this indirect process short prototypes series can
be obtained with excellent dimensional and geometrical reproducibility. This type of molds is usually made
with polyurethane or epoxy based resins filled with
metallic charges to promote better rigidity and thermal
conductivity. These molds are already similar to the
traditional metallic ones for thermoplastics injection.
The objective of these molds is to inject thermoplastics waxes to produce lost models for investment
casting. They follow the same principle for rubber
molds, with the exception of some particularities
related with the materials employed (processability)
(Vasconcelos et al. 2006a,b) and parting plane definition due to negative drafts. The main process stages
can be seen on figures 5 to 7.
The mother model, usually obtained by AM, is usually divided as a function of the previously selected
parting plane. Place the mother model (with the associated gating system considering the parting plane(s)
and guiding system) over the parting plane. Apply the
demolding film over the previous mount and cast the
filled resin. After the resin cure, remove the parting
plane to place the remaining mother model, gating
system (and demolding agent). Cast the resin and after
445
Table 1. Chemical composition and density of some low

melting point alloys (ASM International, 1998).
Alloy
Melting
point
Chemical composition (wt. %)

Bi
Pb
Sn
Cd
Density
(kg/m3 )
A
B
C
70
95
100
50,00
52,50
46,00
26,70
32,00
20,00
13,30
15,50
34,00
10,00
0,0
0,0
9755,01
9896,35
9252,20
Figure 9. Casting under vacuum the low melting point alloy
over the previous half mold.
Figure 8. Casting under vacuum the low melting point alloy.
curing, demold. Clean and post-cure in accordance to

the resin technical specifications.
The mold is now ready to receive the injection of
the thermoplastic waxes.
2.2.3 Metallic molds
The objective of these molds is to quickly inject a
large number of thermoplastic waxes parts for investment casting. With this indirect process, large series
of parts with excellent dimensional and geometrical
reproducibility can be obtained.
These molds are made in low melting point alloys
(see Table 1), usually cast under vacuum in a metallic frame that allows more rigidity of the assembly
and facilitates the connection to the structure of the
injection machine.
These alloys should have melting points among
70 C and 120 C, because the wax injection temperatures are around 6575 C, and the AM materials
for prototypes have working temperatures lower than
120 C. Higher temperatures can irreversibly damage
these masters.
The main manufacturing stages for these molds are
similar to the ones indicated for resin molds.
The mother models are usually obtained by AM
(frequently with paints containing metallic pigments).
The model, with gating and the guidance system is
placed on a surface containing the parting plane. The
demolding film is applied over the previous mount
and the low melting point alloy is cast under vacuum, and rapidly cooled. The procedure for the other
halve mold is then repeated (see Figures 8 and 9).
The following experimental part describes the
methodology employed to manufacture different jewelry artifacts, using an AM process associated with
indirect conversion technologies, to produce lost
Figure 10. 3D CAD images of the rings (Solidworks) and

elliptical shaped bracelet (Rhinoceros).
models for the investment casting of titanium. The

final Ti parts were submitted to different finishes.
EXPERIMENTAL PART
The manufacturing of prototypes followed the traditional methodology accepted in product design (chain
of product development) and was supported by AM.
So, in all development stages the use of these technologies allows the physical materialization of the
proposed prototypes with great speed and flexibility
(Ulrich and Eppinger 2003).
The process began with the creative phase, designing sketches of rings and bracelets. The sketches
represent simple shapes with complex attributes, considering their feasibility in terms of casting and
coloring processes.
The 3D CAD modeling (Figure 10) was done
with Solidworks (version 2009) and Rhinoceros (version 4.0) programs, selected in function of prototypes
complexity.
3.1 Prototypes
3D CAD files were converted into .stl files of 0.01 mm
resolution to produce the prototypes by AM, using the
stereolithography apparatus Viper 3D Systems. Two
different resins were used, a flexible one (Huntsman,
RenShape SL 7810) for bracelets, due to demolding
446
Figure 11. Prototypes of the ring with rectangular pyramidal

holes (left); ring with flat face with four elliptical grooves
(center) and bracelet with elliptical rings (right).
Figure 13. Assembly of elliptical rings bracelet (left) and

cluster of rings in Red and Orange Castaldo wax (right).
Figure 12. Parts of the metallic mold a central core in

aluminum and a cavity in a low melting point alloy.
considerations, and a more rigid one (3D Systems,

Accura 60 Plastic) for rings.
The final SL prototypes are presented in figure 11.
3.2
Molds for indirect conversion
The pattern of the rings were made using the indirect

conversion with silicon molds (translucent silicone
from Dow Corning, reference T4-0). After casting the
silicone, the mold was cut, along the selected parting
plane, to remove the SL model.
The pattern of the rings bracelets were made
together with the feeding system, to promote a more
effective filling of the wax and subsequently the
casting of Ti.
A metallic mold, using a low melting point alloy
(Alloy A; see table I), was obtained for wax injection
(the former only allow the filling of wax by gravity casting). The mold also includes an interior rod
and external frame in aluminum (Figure 12). With this
type of mold, injection pressures of 250 bar can be
used, which regarding the quality of previous waxes
prototypes is substantially better.
3.3
Casting and assembly the waxes
The waxes used in this study are identical to those

referenced in the traditional jewelry (Castaldo waxes,
reference Aqua, Red and Orange and Kerr waxes reference Aqua Green). The difference between the two
brands lies in the ability to reproduce intricate details,
Figure 14. Individual shells for bracelet (left) and rings

(right).
shrinkage factor and cooling rate. All these characteristics are important to achieve stable shapes and
sizes.
The wax was heated at 120 C and subsequently
casted into the silicon die. For the metallic mold, the
wax was heated at 70 C and injected at 15 bar.
The lost wax models were welded to the gating system. For casting the wax bracelet, individual bunches
were chosen. In the case of the rings multiple assemblies were made but only two of them were cast
(Figure 13). Although different waxes were used, for
these types of artifacts no differences were detected.
3.4
Manufacturing of the ceramic shells
The first two ceramic layers used a slip based on an

aqueous solution of colloidal silica (in a reduced %)
mixed with yttria flour, while the remaining layers
were made with water based colloidal silica mixed
with alumina flour. The sand for the first two layers
was yttria while for the others layers was alumina. A
total of nine layers were made (Duarte et al. 2007) Figure 14 shows the shells in the drying process between
layers deposition. The shells were then placed in an
oven at 900 C for 2 h for dewaxing (thermal shock and
calcination), followed by a pre-sintering at 1200 C for
1 h to improve the mechanical properties of the shell.
3.5
Casting Titanium
The Ti alloy used in this study was the Ti6Al4V, ASTM

B 348-06a. To facilitate and enable a good casting,
the shells were pre-heated (to mitigate thermal shock
447
Figure 17. Colors obtained with protection from radiation.

The rows indicate the surface state while the columns represent the groups indicated in table 2 (0 means with no heat
treatment).
Figure 15. Ceramic shell after Ti casting in the furnace for
reactive alloys of INEGI.
Figure 16. Titanium bracelets with golden color (7 minutes) (left), and heterogeneous color (14 minutes) (right), both
obtained by oxidation at 700 C.
with the ceramic shell walls) covered with a glass fiber

blanket at 1000 C. Ti was cast at 1700 C and cooled
under an inert atmosphere of argon, with a cooling
time around 30 minutes (Figure 15).
3.6 Finishing
The finishing involves the slaughter of the ceramic
shell (vibratory pneumatic hammer and water jet), and
blasting with glass beads (removal of ceramic debris
on the surface due to the usual tiny reaction between
metal and ceramic) and cutting of the gating system.
After this phase, a chemical cleaning with an
aqueous solution of nitric (7%) and hydrofluoric acid
(15%) was applied to remove any eventual tiny debris
still existent.
The final surface finish obtained is enough to
perform the staining by heat. If a polished finishing is demanded, it is necessary to perform a series
of sequentially-sanding grinding operation with SiC
papers, followed with abrasive pastes for metal.
3.7 Coloring
Tests were made in two bracelets to obtain colors by the
heating process. After chemical cleaning (solution of
hydrofluoric and nitric acid), the parts were cleaned
with acetone and placed in an oven at 700 C. The
golden color appeared after 7 min and heterogeneous
color after 14 min (Figure 16).
Staining by heat is the most widely used method

by jewelers because of the ease with which it is
held. For this, is simply necessary an oven with a
temperature controller, normal cleaning agents and
ordinary sandpaper for the surface preparation to color.
The anodization process is more complex and needs
dedicated equipment; such as electrolytes, chemical
compounds and a power source that allows the voltage
control and current intensity between an anode and a
cathode (electrolytic cell).
Due to the simplicity and color quality of the heating
process, this was the selected process for this experimental work. A commercial pure titanium sheet of
2 mm thickness, Grade 3 (ASTM B265), was cut in
samples of 20 15 mm. Samples in as supplied condition are designated as S3. S2 means that the surface
was chemically etched by emersion in an aqueous solution of nitric acid (7%), hydrofluoridric acid (15%), 1 g
dissolved titanium and 5% wetting agent, and ultrasound for 15 minutes at 40 C. S1 are polished state in
humid media with SiC papers #360, 600 and 1200.
After the first color test at 700 C a new test was
done at the same temperature with the particularity of
covering the samples with a glass fiber blanket to prevent heat from the fraction of radiation. The spectrum
of colors obtained increased from the former test. The
aim was to achieve a better control of the oxide film
formation with a better uniformity in color along the
entire surfaces of the specimen. The colors obtained
are indicated in figure 17, while table 2 presents the
experimental conditions. In this test series S1 and S2
are the specimens with the best coloration, while S1
samples exhibited more heterogeneous colors. These
results show that the surface state is very important for
the color quality and that good spectrum of colors are
easily obtained.
Subsequently, these samples were used to design
jewelry artifacts, namely; two bracelets and two necklaces. The rectangular sections were placed side by
side, and connected in different ways with a flexible
polymer tape, as shown in Figures 18 and 19).
The three bracelets showed excellent reproducibility in terms of shapes and surface finish, surpassing
the initial expectations. It is noteworthy that these
bracelets have complex geometry (see detail in Figure 20).
448
Table 2. Colors obtained after heat treatment with protection of a glass fiber blanket.
Group
1
2
3
4
5
6
7
8
9
Surface state
(Series)
Time
(minutes)
S1, S2; S3
S1, S2
S3
S1
S2
S3
S1, S2
S3
S1, S2; S3
S1, S2
S3
S1
S2, S3
S1, S2
S3
S1, S2
S3
10
14
15
16
17
14,5
15
15,5
16
Color
Yellow
Pink
Brown
Violet
Brown
Stain violet
Blue
Blue-Violet
Blue
Blue
Blue-Brown
Blue
Blue-Violet
Blue
Blue-Violet
Blue
Blue-Violet
Figure 21. Ring with different finishings; polished (left),

shot peened (center) and with just the frontal surface polished
(right).
The rings obtained also exhibit excellent reproducibility of shape and dimension. The surface quality
allowed the implementation of different finishes, all of
acceptable quality in jewelry. The cast assembly contained several rings of each type that were shot peened
with Corindom FEPPA 60 or polished with water SiC
paper of 320, 600 and 1200 (Figure 21).
Figure 18. Bracelets.
CONCLUSIONS
This work has shown that current 3D CAD tools enable

expeditiously modeling of geometric shapes starting
from a simple sketch.
The presented manufacturing methodology allows
the manufacturing of prototypes with the necessary
quality (shape and dimensional accuracy) for jewelry.
The flexible and rigid molds manufactured for
indirect conversion technology allowed obtaining different types of waxes with adequate quality for
investment casting.
The ceramic shells manufacturing methodology
allowed cast Ti alloys under controlled atmosphere
to produce jewelry artifacts with adequate quality and
different finishes obtained by colorization by heat, shot
peening and polishing.
REFERENCES
Figure 19. Necklaces.
Figure 20. Bracelet detail after chemical cleaning.
Alves, F., Braga, F., Simo. M., Neto, R. & Duarte, T. 2001.
Protoclick, Prototipagem Rpida. Porto: Protoclik.
ASM International (ed.) 1998. Casting ASM Handbook 15
9th Edition.
Crafts Council. 2009. Image Library/Explore Crafts/
Crafts Council. Crafts Council. [Online] Crafts Council, 2009. http://www.craftscouncil.org.uk/explore-craft/
image-library/.
Duarte, T., Neto, R.J., Flix, R. & Lino, F.J. 200. Optimization of ceramic shells for contact with reactive alloys.
Materiais 2007, FEUP, 14 April.
Hayfield, P.C.S. 1998. Development of the noble metal/oxide
coated titanium electrode. Platinum Metals Rev 42(1):
2733.
Lino, F.J., Ala, P., Neto, R.J., Paiva, B., Paiva, R. & Sousa, R.
2003. Indirect rapid tooling with investment casting and
ceramic moulding, Proceedings ofVRAP 2003, 1st International Conference on Advanced Research in Virtual
and Rapid Prototyping. Escola Superior de Tecnologia e
Gesto de Leiria (Eds.): 51724. Portugal, 14 October.
Lino, J.,Vasconcelos, P. & Neto, R.. 2009. Cincia, tecnologia
e design do produto. Tecnometal 180 (JanFev): 1018.
449
Lutjering, Gerd & Williams, J.C.. 2007. Titanium: SpringerVerlag. ISBN 978-3-540-71397-5.
Pietro, P. 1999. Titaniocromia (e Altre Cosa). Interlinea (ed.).
Italy.
The Art of Reflection and Refraction. 2009. http://www.
reflectionandrefraction.co.uk/history.html. The Art of
Reflection and Refraction. [Online] 2009.
Ulrich, K.T., & Eppinger, S.D 2003. Product design and
development. Mc Graw-Hill, 3rd Ed. ISBN 007-123273-7.
Warfel,Todd Zaki. 2009. Prototyping:A Practitioners Guide.
s.l.: Rosenfeld Media , 2009. 1-933820-22-5).
Vasconcelos, P.V., Lino, F.J. & Neto, R.J. 2004. Importance of the vacuum in rapid tooling of polymeric-based
moulds. RPD 2004, Agile Development for Productivity,
Centimfe, Marinha Grande, Portugal, 1213 October.
Vasconcelos, P.V., Lino, F.J. & Neto, R.J. 2006a. Design
epoxy resins based composites for rapid tooling applications. Mechanics & Materials in Design, M2D 2006,
5th Int. Conference, FEUP, Porto, 2426 July.
Vasconcelos, P.V., Lino, F.J., Baptista, A.M. & Neto, R.J.
2006b. Tribological behaviour of epoxy based composites for rapid tooling, Wear 260 (12): 3039.
450

Automation for building manufacturing

F. Craveiro, J.M. Matos, H. Brtolo & P.J. Brtolo
ABSTRACT: The performance of the construction sector is a key development issue for any countrys economy.
Traditionally the construction sector is very conservative, risk averse and reluctant to adopt new technologies
and ideas. There are substantial challenges and opportunities to develop modern and innovative methods, which
to be successful need to be integrated in a very demanding human and entrepreneurial environment.
This work proposes a new system for the construction automation of buildings, based on extrusion-based
technologies, which will enable to construct eco-efficient buildings with complex forms and geometries, based
on sustainable construction principles. It will enable to produce buildings with good integrated construction
processes and less waste. The proposed multimaterial deposition system will also allow the construction of more
efficient buildings regarding thermal, acoustic and structural conditions.
INTRODUCTION
The performance of the construction sector is a key

development issue for any countrys economy. It consumes a very significant proportion of the raw materials produced around the world but also provides
the fabric of the built environment on which society depends. The construction sector generates around
5.6% of Portugals GDP, according to INE (2008). Its
weight in Portuguese economy is overwhelming, as the
workforce related with the construction sector is huge
involving a wide range of stakeholders and companies.
The perceived image of the construction sector is
predominantly low-tech, which can undermine the
future development of the industry and its contribution for the countrys competitiveness and sustainable
development.
Romans already poured concrete in the same way
we currently do. Concrete was poured into temporary formwork, a simple method of moulding being
used for centuries. Conventionally, the construction
industry still relies on craft based methods and suffers
from poor performance and quality (Buswell et al.,
2007).
Several works has identified the need for radical
changes in the construction industry, which can lead
to dramatic improvements in a wide range of key
performance indicators for the construction industry
in a global market. Actually, the construction sector
faces a range of difficult issues that require to be
addressed through a concerted approach, as many are
interrelated.
Traditionally the construction sector is very
conservative, risk averse and reluctant to adopt new
technologies and ideas. On the other, it is fundamental to prevent construction site accidents and promote
safety for everyone involved in the construction sites,
converting the construction activity into a more sustainable one.

There are, therefore, substantial challenges and
opportunities to develop modern and innovative methods, which to be successful need to be integrated
in a very demanding human and entrepreneurial
environment.
The uniqueness of the construction sector constitutes a challenge for the direct adaptation of technologies that are used in many other industries. This work
proposes a new system for automatic construction
based on additive manufacturing technologies, which
will enable to construct eco-efficient buildings with
complex forms and geometries, based on sustainable
construction principles (Matos et al., 2009).
The use of automation processes constitutes the central strategy of the proposed construction system, as
it uses layer-by-layer principles and additive strategies
to fabricate buildings. This research work describes
this new construction process.
2 AUTOMATION FOR BUILDING
MANUFACTURING
Additive technologies, usually called rapid prototyping technologies, are one of the most quickly
developing manufacturing technologies in the world
(Wohlers et al., 2011). The first techniques became
available in the 1990s.
Additive technologies refer to a class of manufacturing processes, in which a part is built by adding layers
of material upon one another. Initially, they were used
to produce conceptual product models for aesthetics
and ergonomic purposes, identifying design flaws and
testing products during the design process.
Today, actual product manufacturing is increasing,
and toys, furniture, and medical equipment are among
451
Figure 1. The proposed system for the automatic construction of a building (real scale).
the many items being produced with the technology.

There is a wide range of these technologies applied in
a variety of industries, including the aerospace, architectural, automotive and medical fields. However, they
are less suitable for construction industry practices for
two main reasons: i) they can only be used for medium
or small-scale objects, and ii) the system does not
easily support the use of more than one material.
Khoshnevis (2004) developed a concept for the
automatic fabrication of a house called Contour Crafting, which consists of the automatic fabrication of its
walls layer by layer, until the creation of a formwork
filled, after the cure, by mortars mainly composed
by cement. Later, he developed some reinforcement
strategies through the adding of some rigid elements
(Khoshnevis et al. 2006). This system uses a device
to hold the material to be extruded by a piston.
From a mechanical point of view, this is a very easy
process, though it does not allow obtaining a good
material mixture whenever the processing involves
multimaterial aggregates.
Lim et al (2009) developed a concrete printing
strategy to produce 3D customized products. This is
a very versatile system, though it creates artifacts
with rough surface finishing. Concrete printing is also
based on the extrusion of cement mortar.
Dini (2011) developed the D-Shape process that
uses a poder deposition binder similarly to the FDM
process. In this process, each layer of build material
is laid to the desired thickness, compacted and then
the nozzles on the gantry frame deposit the binder in a
selective way. This process is very slow, uses a massive
material placement and removal of unused material.
The Centre for Rapid and Sustainable Product
Development is developing a novel 3D additive technology, using the know-how acquired by the production of structures for cell support (scaffolds), to be
applied in building construction, using materials like
mortars mainly composed by cement and clay.
The proposed fabrication system uses a computer
controlled mobile crane integrating multideposition
devices (a significant advance regarding other existing
Figure 2. Equipment for the automatic construction of

contour paths.
methodologies), with several degrees of freedom. This

platform uses fast curing tixotropic materials with low
shrinkage (Figure 1). This crane has two parallel rails
to facilitate its movements. To minimize the time and
costs spent for the transport and installation of the
crane, a trailer is used as rail support.
This additive manufacturing automation process
allows building the houses walls layer by layer, in a
continuous way, as well the whole building structure.
An extrusion head will continuously deposit material
until it approaches a window or door opening space,
then it slows down until stopping at these previously
selected points. This process is repeated layer by layer
until the top of each houses wall.
3
PROTOTYPE SYSTEM
To validate the proposed methodology and test strategies for material deposition, a prototype system was
developed. The equipment, illustrated in Figure 2,
comprises a building platform (1220 610 mm), a linear movement system and two extrusion heads, each
one with a screw system to enable the processing of
different materials. The controller is able to control 4
axes (3 axes with xyz linear movement and the rotation
of the extrusion head) and monitoring its movement
through each encoder.
The first extrusion head is able to produce contour
paths, smoothing the material on the lateral surfaces,
through 10 mm height parallel guide structures, with a
distance of 9 mm from each other, allowing creating a
9 10 mm profile. The second extrusion head, used to
create reinforced meshes, has neither rotational freedom, nor lateral guides. To adjust the reinforced mesh
diameter, nozzles with different sizes can be used.
The 3D construction process is similar to other
additive manufacturing processes (Figure 3), i.e.,
after creating the initial conceptual model, where the
built environment is represented by any commercially
available 3D CAAD (Computer-Aided Architectural
Design) system, the model is then tesselated (the surfaces are represented through a network of triangular
452
Figure 3. Information flowchart of the additive manufacturing process.
Figure 6. CAD Model of the simple wall element.
availability, as well its potential use for construction.

Clay and Portland cement (with additives) were the
selected materials.
Architects are enthusiastic regarding the use of clay
as a building construction material, even if the design
and the construction methods need to be adapted to
this type of material. This can be considered an ideal
material for a sustainable construction. It is an easily
processed and abundant material.
The selected materials have different behaviours
during and after processing. For instance, clay is better to keep its form and an easier processing though,
after processing has a high shrinkage and a much lower
strength.
Figure 4. STL model.
Figure 5. SLI model.

Table 1. Commonly used functions for the geometry building of the product.
Code
Description
G00
G01
G02
G03
G17
G21
G40
G90
Rapid Positioning
Linear Interpolation
Clockwise Circular Interpolation (CW)
Clockwise Circular Interpolation (CCW)
XY Plan selection
Programming using the International System (SI)
Cancel the compensation of the tool diameter
Absolut positioning
The additive manufacturing process presented in this

work was experimentally tested to produce two different building elements (Figure 6):
Construction of a simple wall element using a single

extrusion head;
Construction of double wall elements structurally
reinforced (sandwich structure), using two extrusion heads.
elements similar to a FEM mesh) to obtain the STL

model (Figure 4). This STL model is then sliced to
obtain the SLI model (Figure 5), where each slice
corresponds to a section of the built environment to
be physically produced (Alves, et al., 2006). This
information is stored as a g-code.
The deposition instructions are then followed by the
machine until the model is totally built. In Table 1,
the more frequently used functions for the geometry
building of the product are listed.
4
CONCEPT VALIDATION
MATERIALS
Two types of materials were selected for the initial

validation of the process, taking it account its cost and
The principle associated with the double wall case

is illustrated in Figure 7. In this case, the first extrusion
head aims to create the two resistant walls with a good
surface finishing. This extrusion head has rotational
freedom and guide structures, allowing smoothing in
a tangential direction all along the extruded path.
The second extrusion head extrudes material to
reinforce the previously built walls. This reinforcement prevents the use of a massive wall, increasing the
strength/weight relationship, and allowing space for
pipes and thermal and acoustic insulation. This head
has a conical nozzle and no guide structures, so it can
be moved with no restrictions when, for instance, we
want to define curve rays in order to ensure a good
adhesion between the interior mesh and the external
walls.
The diameter and the lay-down pattern need to be
adjusted whenever there is a need to use the space
between walls, namely for diverse type of pipes. This
mesh can be designed to support structural loads. In the
453
Figure 9. A simple wall element processed with clay.
Figure 7. Double deposition strategy.
Figure 10. A simple wall element processed with cement.
Figure 8. Non-structural wall.
cases there is not a structural wall, it must has spaces

to introduce metallic structural elements or function as
a formwork to obtain a reinforced concrete structure
(Figure 8).
These two last models are illustrated in Figures 9,
10 and 11. The processing conditions for these cases
were the following ones:
Scanning speed: 100 mm/min;

Screw rotational speed: 44 rpm;
Pressure: 4 bar.
These models enable us to conclude that it is possible to build complex structures with linear and
curved areas to be used in the construction of building
Figure 11. A clay double wall with reinforcements.
elements. Clay structures have good quality, keeping

its form after processing, while concrete structures
present a lower quality. To improve its quality, the
material characteristics need to be optimized (viscosity, mix additives to accelerate the curing speed,
etc.).
454
ACKNOWLEDGEMENTS
CONCLUSIONS
This work proposes a new system for the construction automation of buildings, based on extrusionbased technologies. This new system presents several
advantages over traditional approaches:
This project has been supported by the Polytechnic

Institute of Leiria through the Project Buid-it-Green.
REFERENCES
Increased geometrical freedom

Structural optimisation
Single material construction
Faster Construction
Cost Reduction
This system can be a step forward to change the

construction industry from a labour intensive activity
to an integrated one, through cutting-edge technologies and knowledge. It can be very important to
promote Portuguese competiveness helping companies to successful compete in global markets.
Modern Architecture is demanding more and more
from the design of buildings. This system will enable
architects to construct buildings with exotic and complex geometries difficult to build using conventional
construction processes. It will enable to produce buildings with a better integrated construction processes
with less waste. The proposed multimaterial deposition system will also allow the construction of
best-adapted buildings regarding thermal, acoustic
and structural conditions.
The optimization of this system requires further
investigation regarding the selected materials in order
to establish an optimal correlation between the initial material characteristics (viscosity, composition),
the processing conditions (pressure, scanning speed,
screw rotational speed) and the quality of the final
structure.
Alves N M and Brtolo P J, Integrated computational tool for

virtual and physical automatic construction, Automation
in Construction, Vol. 15, 2006, pp. 257271.
Buswell R A, Soar R C, Gibb A G F and Thorpe A, Freeform
construction: mega-scale rapid manufacturing for construction, Automation in Construction, Vol. 16, 2007,
pp. 224231.
Dini, L., D-Shap, www.d-shape.com (accessed on March
2011)
Khoshnevis, Hwang B D, Yao K, Yeh Z, Mega-Scale Fabrication using Contour Crafting, International Journal
of Industrial & Systems Engineering, Vol 1, No. 3, 2006,
pp. 301320.
Lim S, Le T, Webster J, Buswell R, Austin A, Gibb A and
Thorpe T, Fabricating construction components using
layered manufacturing technology, Global Innovation
in Construction Conference 2009 (GICC09), Loughborough University, Leicestershire, UK, 1316 September,
2009.
Matos J M, Craveiro F, Brtolo H M, Martins-Ferreira N
and Brtolo P J, Prototipagem rpida aplicada construo, 1 Forum Internacional de Tecnologia da Construo TECCON 2009: Tecnologias associadas ao
Processo do Empreendimento de Construo, Faculdade
de Engenharia da Universidade do Porto, Porto, 2009.
Wohlers T, Rapid Prototyping, Tooling and Manufacturing:
State of the Industry, Wohlers Associates, USA, 2011.
455
Additive and Nano Manufacturing Technologies

Optimization of Selective Laser Melting technology using design of

experiments method
M. Averyanova
Ecole Nationale dIngnieur de Saint-Etienne (ENISE), Laboratoire DIPI, Saint Etienne, France
E. Cicala
Mechanical Engineering Faculty, Politehnica University of Timisoara, Timisoara, Romania
Ph. Bertrand
Ecole Nationale dIngnieur de Saint-Etienne (ENISE), Laboratoire DIPI, Saint Etienne, France
Dominique Grevey
Laboratoire Interdisciplinaire Carnot de Bourgogne, quipe LTm, UMR 5209 CNRS-Universit de Bourgogne,
Le Creusot, France
ABSTRACT: The Selective Laser Melting (SLM) technology is an Additive Manufacturing process that
depends on a large number of different parameters: materials (powder size distribution, size, morphology, chemical composition, crystalline state etc.), laser (laser type, diameter and energy distribution in the laser spot, density
of energy etc.), machine (type of SLM machine, layer delivery system, protective atmosphere etc.). In order to
determine optimum process parameters setting for achieving maximum density in 3D parts manufactured by the
SLM technology, the method of Design of Experiments has been applied. The effect of four process parameters
(laser power, scanning speed, layer thickness and hatch distance) with four levels for each factor and interaction
between them has been studied.
A methodology in three steps has been applied. The first preliminary tests have been performed to analyze
a large operating domain. Then, the exploratory tests allowed reducing this large parametric window. Finally,
using the optimization tests the optimal operating parameters setting has been validated.
The results show that a link between the geometry of the weld track and function issued from the physical
parameters of the process, compared to operating parameters, has been identified. Thus, a stability criterion of
the weld track has been proposed. It let us to determine quickly the optimal process parameters window and to
manufacture 3D objects with a porosity rate below 1%. The values of mechanical properties of tensile specimens
of 17-4 PH powder manufactured by SLM are the same order of magnitude as the wrought material elaborated
by conventional methods.
Keywords:
Selective Laser Melting, Design of Experiments, Weld track stability
INTRODUCTION
Nowadays the moving from Rapid Prototyping to

Direct Manufacturing is a major challenge for the
Additive Manufacturing technologies. It requires a
robust repeatable process that makes possible to fabricate parts with desirable properties and microstructure. Selective Laser Melting (SLM) is one of such
additive manufacturing process of joining powder
materials layer by layer in order to fabricate 3D
parts of complex shape directly from its CAD data
(Wohler, 2010). Contrary to conventional technologies, this technique makes it possible to manufacture complex shape highvalue added parts without
dies or moulds. The development of this technology
was essentially driven by industrial need to fabricate
approximately 100% dense objects with mechanical

properties comparable to those of bulk materials.
The fields of application are varied and include
the fabrication of internal cooling channels, complex
weight-light structure, functional-graded materials,
functional-graded coatings etc. for aerospace, nuclear,
chemical and petrochemical uses (Kruth et al., 2004),
(Emmelmann et al., 2009), (Stamp et al., 2009), (Yadrtoitsev, Bertrand et al., 2007), (Hollander et al., 2003),
(Hao et al., 2009). Other applications can be found in
the medical field (implants, tissue engineering scaffolds), porous structure, dental restoration creation
of dental crowns etc. (Gibson, 2005), (Bartolo et al.,
2009).
The physical phenomena involved in SLM concern mainly the interaction between the laser and the
459
powder bed, the heat transfer, the rapid heating and

solidification, the phase transformation, the oxidation
etc. (Jhabvala, 2010). They are affected by the process
itself and by the material properties.
Usually in order to manufacture dense parts
with necessary level of mechanical properties the
researches use an empirical approach of search of
optimum process parameters setting. It is considered
that the laser power, the scanning speed and interaction between them (energy density) is presented as the
most significant input parameter that affects the quality of produced parts (Simchi et al., 2006), (Mumtaz
et al., 2009), (Kruth et al., 2004). The enhancement
of energy density up to an optimal value changes
the viscosity and surface tension of the melted material that finally results in a part with higher density,
(Simchi et al., 2003). As it has been shown before the
SLM process depends on an important number of controlled and non-controlled parameters (Rehme et al.,
2005). In order to determine rapidly the significance
of the parameters, the effect of their interactions, to
better understand the process and optimize the parameters setting, an Experiment Design Approach (DOE)
should be applied (Taguchi, 2005).
It should be noted that contrary to the SLM process a considerable number of investigations of the
SLS (Selective Laser Sintering) process has already
been done and reported (Kumar et al., 2003). An
overview of the published studies about the application of experiment design to SLS/SLM process
shows that researchers often use a Taguchi method
to find out the significance of the impact of process
factors on the mainly characterized properties manufactures samples such as density, porosity and hardness
(Liao et al., 2007), (Dingal et al., 2008), (Jailani
et al., 2009).
The objective of the present work is to determine optimum process parameters setting for achieving maximum density in 3D parts manufactured by
SLM process from martensitic precipitation hardening
stainless steel using the DOE method.
EXPERIMENTAL METHODOLOGY AND

EQUIPMENT
Experiments were carried out on PHENIX Systems

PM 100 machine. The PM 100 is equipped with IPG
Photonics fiber laser with a Gaussian profile, delivering a maximum power P = 50 W, at =1075 nm and
having a laser spot size 70 m. The parts are manufactured in a 100 mm diameter 304L steel substrate
under Argon protective atmosphere.
In the present paper, 17-4 PH stainless steel powder that belongs to precipitation hardened steels of
martensitic class and is widely used as a structural
material for petrochemical, chemical, tooling, surgical and other fields due to its good wear resistance,
corrosion resistance and mechanical properties, has
been used (Antony, 1963), (Nakagawa et al., 1999).
Table 1 gives the chemical composition of investigated
Table 1.
Chemical composition of 17-4 PH powder.
Cr,%
Ni,%
Cu,%
Mn,%
Si,%
15,017,5
Fe,%
bal.
3,05,0
C,%
0,07
3,05,0
P,%
0,04
1,00
S,%
0,03
1,00
Figure 1. Morphology of 17-4 PH powders: (a) Powder 1

(b) Powder 2.
17-4 PH powder that meets standard specification A

693-06.
In literature it has been shown that particle size
distribution of used powders for SLM technology
is between 20 and 50 m (Dai et al., 2006). The
previous research results of our laboratory show that
for industrial machine PM100 of Phenix Systems the
fine powders (D90 25 m) should be applied.
Two spherical gas atomised 17-4 PH powders
with slightly different particle size distribution
(D90 15 m (Powder 1) and D90 25 m (Powder 2) have been chosen for the present work.
A full analysis of 17-4 PH powder properties for
SLM application was reported earlier in (Averyanova
et al., 2009).
460
EXPERIMENTAL DESIGN METHODOLOGY
Firstly, the preliminary tests will be performed in

order to classify and analyze the significance of input
Table 2. Levels of each of the investigated SLM process

parameters.
Table 3. Orthogonal array. Main SLM process parameters

with their levels.
Experiment No
Laser power, W
Scanning
speed, mm/s
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
35
35
35
35
40
40
40
40
45
45
45
45
50
50
50
50
50
80
120
150
50
80
120
150
50
80
120
150
50
80
120
150
Experiment
No
Layer thickness,
m
Hatch distance,
m
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
20
30
40
50
30
20
50
40
40
50
20
30
50
40
30
20
100
120
140
160
140
160
100
120
160
140
120
100
120
100
160
140
Level of each parameter

N
Parameter
Level 1
Level 2
1
2
3
4
Laser power, W
Scanning speed, mm/s
Layer thickness, m
Hatch distance, m
35
50
20
100
40
80
30
120
Level of each parameter

N
Parameter
Level 3
Level 4
1
2
3
4
Laser power, W
Scanning speed, mm/s
Layer thickness, m
Hatch distance, m
45
120
40
140
50
150
50
160
parameters like laser power, scanning speed, layer

thickness and hatch distance on stability, dimensional
accuracy of tracks and roughness of single layers. Secondly, a number of experiments will be conducted
to determine an optimal domain of input factors.
Finally, 3D parts with necessary density, hardness and
microstructure will be manufactured.
In order to reduce the number of experiments the
fractional factorial analyses with an orthogonal array
has been chosen for the first step. It should be noted
that the choice of an appropriate orthogonal array suitable for a research task is one of the difficulties of
experiment design, (Taguchi et al., 2005).
At this research work, based on previous scientific
experience, four factors with four levels have been chosen for the experiment. The fractional factorial design
reduced the number of experiments (in our case to 16)
and took into account interactions between parameters (Taguchi et al., 2005), (Roy et al., 2001). The
levels of the main SLM parameters selected in the
present research for an orthogonal array are shown in
Table 2. The orthogonal array used in the present work
is presented in Table 3.
The output parameters are geometrical characteristics, stability (1D and 2D objects) and also the surface
layer roughness for a single layer (2D object) and final
parts density (3D objects). The analyzed parameters
are presented in Fig. 2.
Where for a single track 1D (Fig. 2(a)): w1 width,
w2 width of dilution zone, h1 height, h2 height
of dilution zone and 1 contact angle;
for a single fused layer 2D (Fig. 2(b)): h1 height,
h2 height of dilution zone, h3 height after 2d laser
passage, h4 height of dilution zone after 2d laser
passage, L1 = Ec-hatch distance, Ec/2 half of hatch
distance;
for a 3D part density.
Five single tracks and one single layer were fabricated
for each set of parameters. The obtained results have
made it possible to assess the effect of each influence
factor on geometrical characteristics. For single fused

layers (2D objects) two scanning strategies so called
1 zone and 2 zones techniques have been studied.
The description of the techniques has been reported
earlier (Yadroitsev et al., 2007).
4
4.1 Preliminary tests

During the preliminary tests, the main objective of
single tracks fabrication is to analyze the effects
of process parameters (input parameters) on output
parameters (geometrical characteristics, stability of
single tracks and single layers) and to classify them.
It should be noted that in order to analyze the significance of input parameter on studied output parameters, its average value is compared to its standard
deviation.
In Fig. 3 the main effect plots show the effect of laser
power, scanning speed and layer thickness on the width
461
Figure 2. Manufacturing strategy a) optimum single track, b) optimum hatch distance, c) resulting cube
manufacturing.
w1 and w2 of a single track. It could be concluded that

the width w1 and w2 mainly depend on laser power and
scanning speed. The effect of layer thickness is nonsignificant. It was also observed that for stable single
tracks the value of w1 and w2 are almost the same.
Due to a large process parameters domain analyzed at a first step of DOE method, single tracks of
various shapes, dimensional characteristics, stabilities
are obtained. The reproducibility of the results is also
different. Besides, from the preliminary experiments
the following inferences can be shown:
The effect of layer thickness (Ep) on the height of a

single track (h1) was significant for both powders.
The height of dilution zone h2 is affected by laser
power P. The factor of influence for Powder 2 is
more important.
The width w1 and w2 of the tracks depend on
laser power P, scanning speed V and interactions
between these parameters (energy density) for both
powders. The effect of layer thickness Ep on w1 and
w2 is non-significant, contrary of its effect on the
contact angle .
The same experiments have been conducted for
2D parts manufacturing. It is statistically proved that
the layer thickness has the most significant effect on
the manufactured layer. Layer thickness affects the
layer surface roughness. The roughness can be affected
by surface tension forces during SLM process and
depends on laser processing parameters and melt pool
control. Some industrial applications require a definite
roughness (low or high). In our case it was important
to find the set of parameters that would reduce final
Figure 3. Main effect plots: (a) Effect of process parameters on the mean value of width (w1) of a track with mean
square deviation S = 7 m2 and (b) Effect of process parameters on the mean value of width (w2) of a track with mean
square deviation S = 6 m2 manufactured from Powder 1
where P-laser power, V -scanning speed, Ep-layer thickness.
surface roughness. It is observed that using finer powder and finer layer thickness can lead to manufacture
parts with smooth surfaces.
4.2 Exploratory tests
During the exploratory tests a number of experiments were conducted to determine an optimal domain
of input parameters. The effect of Powder 1 and
Powder 2 on output parameters is approximately the
same. Besides, the mean square standard deviation for
Powder 1 is less than for Powder 2. The repeatability
of single tracks manufacturing is better for Powder1.
Generally, at the highest laser powers (50W), medium
scanning speed (70130 mm/s) and layer thickness
(3050 m) the balling phenomenon has not been
observed.
Once this optimum process parameters domain for
the best powder (Powder 1) is defined the second step
of the present work is to study more in detail the influence of input factors using the following Design of
Experiments: EFCP23 and central point (Fig. 4).
For this, 8 experiments will be conducted using the
max and min optimal values and 1 experiment using
the parameters that provide a medium optimal value
(Fig. 4). The previously obtained results show that in
order to manufacture regular single tracks and single
layers the laser power should be fixed at maximum
50 W. During the second step of DOE scanning speed,
462
Figure 4. Scheme of Design of Experiments: EFCP23 and

central point.
Table 5. Factors and interactions having a significant influence on the objective functions of tracks (starting with the
most influenced) where w-width, h1-height, h2-height of
dilution zone, -contact angle of the track and P-laser power,
V -scanning speed, Ep-layer thickness.
w
h1
h2
P, V,
Ep
Ep, EpV,
V
V, EpV, E,
P
V, Ep
Figure 5. Estimated effect (a) and estimated response surface (b) of process parameters on the width of the track w1
manufactured from Powder1 where (1) CV-scanning speed,
(2) CEp-layer thickness, 1*2-interaction between scanning
speed and layer thickness.
layer thickness and hatch distance will be varied and

laser power will be kept constant. It is considered that
the effect of laser power in this new domain is less
pronounced than in previous domain of first stage of
DOE method.
As a result, it is proved statistically that the width
of a scan (w) depends on laser power, scanning
speed, interaction between them and on layer thickness
(Table 5). Using little value of scanning speeds and big
value of layer thickness, unstable irregular tracks are
manufactured.
The width of the track w1 is particularly influenced
by scanning speed and layer thickness (Fig. 5). When
the layer thickness increases and the scanning speed
decreases, the width increases and becomes unstable.
The detailed description of the first two steps of
DOE method as well as all analyses and conclusions
are reported in (Averyanova et al., 2011).
Based on single tracks analyses obtained by second stage of experimental design method, the optimum
domain has been specified.
It has been demonstrated that track dimensional
characteristics depend on different process parameters and interaction between them. In order to choose
rapidly the optimal set of parameters a special function should be proposed. All dimensional characteristics explaining the physical phenomenon during
manufacturing of a track should be taken into account.
At the present time, it has been decided to develop
a Complex objective function (F) without units which
will describe single track parameters and identify

their stability:
where the first parameter presents the factor related to

the shape of the upper part of the fused:
where the h1 height of a track, w1 width of a track.

The second parameter presents the factor related to
the shape of the penetration depth of the fused track:
where the h2 height of a dilution zone of a track, w2

width of a dilution zone of a track;
The third parameter presents the angle of contact of
the fused track:
where 1 and 2 are the contact angles of a solidified

track.
463
Figure 7. Average values of F function and its parameters

tgA, tgD and tgP for 16 experiments.
Figure 6. Possible single track shapes manufactured from

17-4 PH Powder1 using different F values.
Figure 8. Porosity measurements by Image Analysis

method of 3D parts manufactured with optimum process
parameters setting.
4.3 Optimization tests

Based on our proper experience one should say
that the empirical results give that w1 = w2 and
1 = 2 = . Finally, the objective function F can be
defined as follows:
where h1 height of a track, h2 height of a dilution

zone, w width, angle of contact
The obtained function F determines the physical
shape of the track. If F is small (F < 1), the track is
spread. If F is too big (F < 2), the balling phenomenon
is observed. The notion of optimal shape enables the
optimized process parameters window. The possible
tracks shapes obtained with different values of F are
presented below (Fig. 6).
It was important to analyze the effect of each component of the FO function. In fact the obtained results
show that FO depends mainly on tgA. Among three
components tgP is less affected by changing process
parameters. Also it has been noticed that when the
fused track has an optimum shape and the value of FO
is about 1, the values of tgD and tgP are close to each
other (Fig. 7).
The optimization tests are performed in order to adjust

the process parameters setting chosen in previous
stage. It is important to analyze the influence of an
optimum track shape on 3D parts manufacturing. It
is supposed that a set of parameters that allows optimum track fabrication should provide 3D dense parts
manufacturing. The numerical value of function F
(minimum, maximum and middle values) has allowed
choosing appropriate process parameters. Two 17-4
Powders have been applied for 3D parts manufacturing. The values of chosen process parameters cannot
be disclosed for the sake of confidentiality.
The results obtained show that Fmax leads to a
porosity level of 25% and Fmin to a porosity level
up to 40%. Fmiddle leads to less than 1% of porosity (Fig. 8) in parts manufactured from two 17-4 PH
powders. It means that an optimum process parameters setting has been found and could be applied
for all 17-4 PH powders that fulfil necessary powder
specifications.
For mechanical tests (standard NF EN 10002-1)
rectangular samples have been manufactured, machined and analyzed. Two types of scanning strategies
have been used in order to study the influence of scanning direction from layer to layer on parts mechanical
properties. The results show that the layer alternated
464
Table 6. Influence of the scanning strategy on the mechanical properties of the SLM part manufactured from 17-4 PH
Powder 1.
Powder 1-type of scanning
strategy
Rm
(Mpa)
Rp0,2%
(Mpa)
A%
Powder 1 layer alternated

scanning strategy
Powder 1-non-layer
alternated scanning
strategy
880
614
25%
569
454
2,0%
This research work was supported by CETIM (French

Technical Centre for Mechanical Industry).
REFERENCES
scanning strategy provides more homogeneous melting conditions and consequently, gives a better values
of mechanical strength, elongation.
The values obtained correspond to those expected
for 17-4 PH samples manufactured by the SLM
process and could be improved using an appropriate
type of heat treatment.
5
ACKNOWLEDGMENTS
CONCLUSIONS
The method of Design of Experiments has been

applied in order to determine optimum process parameters setting for achieving maximum density in 3D
parts manufactured from 17-4 PH powders by SLM
technology using Phenix Systems PM100 machine.
The influence of input parameters such as laser
power, scanning speed, layer thickness, hatch distance
for two martensitic precipitation hardening stainless
steel powders on output parameters such as geometrical characteristics and stability of single tracks (1D),
geometrical characteristics and roughness of single
fused layers (2D) and density of 3D parts has been
studied.
The authors propose the development of a complex
objective function F taking into account all physical
and geometrical parameters. This function describes
the shape of a fused track and enables to determine
rapidly an optimum process parameters window. The
methodology applied to explore SLM process is the
following: on the first step of the Experimental Design
Approach, among a big number of process parameters
values the optimal set of parameters can be quickly
determined by the complex objective function. Once
the first parameter window is determined, the second step of the statistical approach leads to study
more in detail the influence of input factors using
the following EFCP23 and central point Design
of Experiments. Finally, using the optimization tests
(third step), the chosen parameter window can be
adjusted and optimized.
The optimum process parameters setting has
allowed manufacturing of dense 3D parts and samples. The mechanical properties values correspond to
those expected for 17-4 PH samples manufactured by
the SLM process and could be improved using an
appropriate type of heat treatment.
Antony, K.C. 1963. Aging reaction in precipitation hardenable stainless steel, Journal of metals 927, pp. 922927.
Averyanova, M.; Bertrand, Ph. 2009. Direct Manufacturing of dense parts from martensitic precipitation hardening steel gas atomized powder by selective laser
melting (SLM) technology, Proceedings of the 4th
International conference on advanced research and rapid
prototyping, pp. 343348, Taylor and Francis group,
Leiria, Portugal.
Averyanova, M; Cicala, E., Bertrand, Ph., Grevey, D. 2011.
Experimental design approach to optimize selective laser
melting of martensitic 17-4 PH, Rapid Prototyping Journal, vol. 18, issue 1, in press.
Bartolo, P.J.S., Almeida, H., Laoui, T. 2009. Rapid prototyping and manufacturing for tissue engineering scaffolds, International Journal of Computer Applications in
Technology, Vol. 36, Nb 1, pp. 19.
Dai, K., Shaw, L. 2006. Parametric studies of multi-material
laser densification, Materials Science and Engineering
A 430, pp. 221229.
Dingal S, Pradhan TR, Sundar S, Roy Choudhury A, Roy SK.
2008. Experimental investigation of selective laser sintering of iron powder by application of Taguchi method,
The International Journal of Advanced Manufacturing
Technology, Vol. 38, No. 910, pp. 904914.
Emmelmann, C., Petersen, M., Goeke, A. 2009. Laser
Freeform Fabrication for Aircraft application, Proceedings of the 5th International WLT-Conference on Lasers in
Manufacturing, pp. 171174.
Gibson, I. 2005. Advanced Manufacturing Technology for
Medical Applications, Wiley, p. 256.
Jailani, H.J., Rajadurai,A., Mohan, B., Kumar,A.S., Sornakumar, T. 2009. Multi-response optimization of sintering
parameters of Al-Si alloy/fly ash composite using Taguchi
method and grey relational analysis, The International
Journal of Advanced Manufacturing Technology, Vol. 45,
No.34, pp. 362369.
Jhabvala, J. 2010. Study of the consolidation process under
macro- and microscopic thermal effects in selective laser
sintering and selective laser melting, PhD Thesis.
Hao, L., Dadbakhsh, S., Seaman, O., Felstead, M. 2009.
Selective laser melting of a stainless steel and hydroxyapatite composite for load-bearing implant development, Journal of Materials Processing Technology 209,
pp. 57935801.
Hollander, D.A., Wirtz, T. et al. 2003. Development of
Individual Three-Dimensional Bone Substitutes Using
Selective Laser Melting, European Journal of Trauma,
Vol. 29, 2003, pp. 228243.
Kruth, J.-P., Froyen, L., Vaerenbergh, J. V. et al. 2004. Selective laser melting of iron-based powder, Journal of
Materials Processing Technology, 149, pp. 616622.
Kumar, S. 2003. Selective Laser Sintering: A qualitative
and objective approach, Literature review. Modelling and
Characterization, Journal of the Minerals, Metals and
Materials Society, pp. 4347.
Liao, Hsin-Te, Shie, Jie-Ren. 2007. Optimization on selective
laser sintering of metallic powder via design of experiments method, Rapid Prototyping Journal, 13/3, 2007,
pp. 156162.
465
Mumtaz, K., Hopkinson, N. 2009.Top surface and size roughness of Inconel 625 parts processed using selective laser
melting, Rapid Prototyping Journal, pp. 96103.
Nakagawa, H. and T. Miyazaki.1999. Effect of retained
austenite on the microstructure and mechanical properties of martensitic precipitation hardening stainless steel,
Journal of Materials Science, Volume 34, Number 16,
pp. 39013908.
Rehme, O., Emmelman, C. 2005. Reproducibility for properties of Selective Laser Melting products, Proceedings
of the third international WLT-Conference on Lasers in
Manufacturing 2005, Munich, pp. 16.
Roy, R.K. 2001. Design of experiments using the Taguchi
approach, John Wiley & Sons, Inc., New York.
Simchi A., H. Pohl. 2003. Effects of laser sintering processing
parameters on the microstructure and densification of iron
powder, Materials and Engineering, A359, pp. 119128.
Simchi, A. 2006. Direct laser sintering of metal powders:
Mechanism, kinetics and microstructural features, Materials Science and Engineering A 428, pp. 148158.
Stamp, R., Fox, P., ONeill, W., Jones, E., Sutcliffe, C.

2009. The development of a scanning strategy for the
manufacture of porous biomaterials by selective laser
melting, Journal of Material Science: Mater Med 20,
pp. 18391848.
Taguchi, G., Chowdhury, S., Wu, Y. 2005. Taguchiss Quality
Engineering Handbook,
John Wiley & Sons, Inc., Hoboken, New Jersey.
Wohlers, T. 2010. Wohlers Report: State of the Industry.
Annual Worldwide Progress Report, ISBN 0-97544296-1.
Yadroitsev, I., Bertrand, Ph., Laget, B., Smurov, I. 2007.
Application of laser assisted technologies for fabrication
of functionally graded coatings and 3D objects for International Thermonuclear Experimental Reactor components,
Journal of Nuclear Materials, 362, pp. 189196.
Yadroitsev I., Thivillon, L., Bertrand, Ph., Smurov, I. 2007.
Strategy of manufacturing components with designed
internal structure by selective laser melting of metallic
powder, Applied Surface Science, 254, pp. 980983.
466

From functional specifications to optimized CAD model: Proposition of a

new DFAM methodology
J.Y. Hascoet, R. Ponche, O. Kerbrat & P. Mognol
Institut de Recherche en Communications et Cybernetique de Nantes (UMR CNRS 6597), Nantes, France
ABSTRACT: Nowadays, due to rapid prototyping processes improvements, a functional metal part can be built
directly by additive manufacturing (AM ). It is now accepted that these new processes can increase productivity
while enabling a mass and cost reduction and an increase of the functionalities of a part. However, in order to
manage it, new design methods have to be developed to take into account the specificities of these processes, with
the Design For Additive Manufacturing (DFAM ) concept. In this context, a methodology to obtain a suitable
design of parts build by additive manufacturing is proposed; both design requirements and manufacturing
constraints are taken into account.
1
1.1
INTRODUCTION
dimensional modification (Sambu et al. 2004),

substrate determination (Kerninon et al. 2009).
Here, some modifications of the part CAD model
are proposed to improve manufacturability.
Design for additive manufacturing
The metal additive manufacturing processes, as the

rapid prototyping ones, build parts layer by layer. They
can be classified in three technologies: powder based
metal deposition (PBMD), selective laser sintering and
wire feed deposition (Ruan et al. 2006). Each one
has its own characteristics and enables to produce
functional metal parts.
Besides, the direct manufacturing proportion in
the additive processes applies increases every year in
the industry. Indeed, because Additive Manufacturing
(AM ) for production eliminates the need of tooling and
can generate free forms, many of the current restrictions of design for manufacturing and assembly are
no longer valid (Wohlers 2010). AM allows many new
benefices in terms of cost and functionality. Nevertheless, these processes have their own characteristics
and also their limits which have to be taken into
account in a general Design For Additive Manufacturing (DFAM ) approach. The definition of DFAM by
David W. Rosen (2007) is: Synthesis of shapes, sizes,
geometric mesostructures, and material compositions
and microstructures to best utilize manufacturing
process capabilities to achieve desired performance
and other life-cycle objectives. Some works have
been done in this way; they can be divided into two
objectives:
Firstly, manufacturability estimation with cost
estimation (Ruffo et al. 2006), process selection
between AM processes (Masood 2002) and between
additive and subtractive processes (Kerbrat et al.
2010). Assessment criteria of the manufacturability
are thus established.
Secondly, manufacturability improvement with
general build guidelines (Teitelbaum 2009), local
1.2 Partial approach versus Global approach

All these works enable to determine and to improve
the manufacturability of a part from its CAD model
for a given AM process. Because they start from
aninitial geometry (given by the initial CAD model),
these analysis can be qualified as partial approaches
(Fig. 1). In this case, it is difficult to determine the real
optimized characteristics for a given AM process while
fulfilling original functional specifications. Indeed,
the initial CAD model was thought to be manufactured
by an initial process manufacturing often very different from anAM process (for example machining which
is the most often used), moreover the proposed modifications are local and the result is never away from
the initial design. The CAD model that is obtained is
thus never really designed for the AM process that is
chosen.
In opposite, a global approach (Fig. 2) starts
directly from both the chosen manufacturing process
characteristics and the functional specifications of the
partto design. Thanks to this, the designer can determine the geometry which uses in the best way the
chosen AM process characteristics while meeting the
functional specifications.
The aim is that the geometry is not limited by an
initial idea of the part shapes but that it depends only
on the manufacturing process and the functional specifications. This new way of thinking is in opposite with
traditional Design For Manufacturing (DFM ) methodologies. However, it involves the capitalization of the
entire knowledge about the manufacturing processes.
A beginning of a global approach, based on topology
467
Firstly, in section two, the needed data are presented. Then in section three, the different steps of
the methodology based on the choice of MD are
explained. The methodology has been applied on a part
manufactured by PBMD process; this constitutes the
fourth section.
PROPOSED REQUIREMENTS
FOR A GLOBAL DFAM
In this section, the requirements for a global DFAM

are presented. They constitute the needed data for the
methodology.
2.1
Figure 1. The partial approach.
Functional specifications
The global objective is to propose a structured

approach which helps the designer to integrate all
knowledge of the chosen AM process in his design
to meet the functional specifications. The functional
specifications can be detailed as:
Functional surfaces (FS): type, dimensions and
position
Dimensional and geometric specifications,linked to
the FS
Mechanical requirements, they depend on the chosen material characteristics
Empty volumes: dimensions and position. They
correspond to the volumes outside the FS that must
contain no part, in order to be able to assemble the part designed into the system to which it
belongs.
Figure 2. The global approach.
optimization with manufacturing constraints, has been

applied to casting process (Harzheim & Graf 2005).
2.2 Context
1.3
Scope of the paper
Although the global approach seems very interesting,

no work has been done in this way in AM.Indeed,
because the AM process are quite young, still little
known and very different from the other manufacturing process, the psychological inertia phenomena can
prevent the designer from utilizing in the best way all
these processes characteristics.
The AM processes are based on a layer by layer
manufacturing. In addition to their own characteristics and whatever their technology, it involves
that of the characteristics of the manufactured volumes, in terms of mechanics, microstructure and
geometric possibility depend on the manufacturing
direction. In addition to the process characteristics,
the choice of the Manufacturing Direction (MD)
according to functional specifications is the key of a
global DFAM which would help designers in exploring new design spaces. That is why a new methodology
which starts directly from both the functional specifications and the process characteristics is proposed in
this paper.
The study context is an influential factor. Because it

can be translate into a concrete objective in terms of
mass, cost, manufacturing time, it has thus to be taken
into account too.
2.3 Manufacturing characteristics
Because the manufacturing characteristics are linked
one another and cannot be seen separately, a global
view which draws upon all the knowledge and experiences of the community is needed, it must be able
to take into account the main characteristics of the
machine(s) which are:
Kinematics
Maximal dimensions
Minimal dimensions
Capability in terms of accuracy
Required accessibility
but also physical phenomena involved in the manufacturing processes, which are decisive in terms of
final properties of parts and which are linked with the
manufacturing strategy.
468
Figure 4. The definitions of thethickness: (a) without finshing (b) with finishing.
step of the methodology enables to know if all the

surfaces can be merged with one another by the chosen AM process in a single part. A first geometrical
analysis is done; it takes into account the maximal
and minimal dimensions which can be obtained by the
process. If the dimensions are not suitable, the product
has to be modified or divided into different parts by
the designer and the functional specifications of these
new parts are then studied.
3.2
The functional volumes (FV ) are defined directly from

the FS which a thickness is added. Indeed, only the
tolerances in the normal direction of the surfaces are
significant in terms of functionality. The others are initially ignored. The thickness, denoted E, depends on:
Figure 3. The proposed DFAM methodology.
2.4
Finishing process characteristics
dimensional accuracy of the AM process denoted a

tolerances linked with the FS, denoted p
3 ALIKE, IF SPECIFICATIONS (GEOMETRIC,

DIMENSIONAL) CANNOT BE DIRECTLY
REACHED BY THE AM PROCESS THAT IS
CHOSEN, A FINISHING PROCESS IS
NEEDED. BECAUSE IT CAN INFLUENCE
THE FINAL GEOMETRY, IN
PARTICULARITY IN TERMS OF
OVERTHICKNESS, IT HAS TO BE TAKEN
INTO ACCOUNT. THE PROPOSED DESIGN
METHODOLOGY
There are two different possibilities: if a finishing

is not needed, (for example if p a), the thickness is
determined from Equation 1 which ensures the functional minimal thickness and from Equation 2 which
ensures the compatibility with the AM process.
From these data, a structured methodology can help

designers in taking into account the manufacturing
constraints while suggesting him an appropriate design
for AM. The methodology is presented in Figure 3. It is
divided into three main steps which enable to include
gradually manufacturing knowledge in the shapes and
the volumes of the part to design. The first step pertains
to global dimensions, the second one to the FS and
their specifications and the third one to the volumes
which merge them.
3.1
Step 2: Determination of the functional volumes
where t = the minimal thickness corresponding to

the local mechanical requirements; d = the minimal
dimension that can be obtained by the chosen AM process; = the overlap between two adjacent cladding
passes; and n is a positive integer.
The equation 1 is illustrated in Figure 4(a).
If a finishing is needed (for example if p < a), then
the thickness is determined from Equations 2 and 3
(Fig. 4(b)).
Step 1: Analysis
The FS are obtained from the functional analysis of

the product and are given by the designer. The first
where e = the overthickness which depends on the

finishing process and the surfaces geometry.
469
Figure 5. Global view of the studied system.
There can be different values of the parameters d,

and a (according to MD). In this case, the choice is
done according to the context of the study.
3.3
Step3: Determination of the volumes linking
The objective of this step is to merge the FV one

another to define the volumes of the part while taking into account at best both design requirements and
manufacturing constraints.
In the case of additive manufacturing, volumes are
obtained layer by layer. It involves that their geometries
strongly depend on the MD. Indeed,the choice of the
MD has a direct influence on material quantity (need
of supports) and build time. Moreover, because the
physical phenomena are not exactly the same in all
the directions during the process, the MD influences
also the accuracy, the roughness and the mechanical
characteristics of the manufactured volume. That is
why the determination of the MD is in the center of
the methodology.
The determination of the Linking Volume (LV ) is
divided into four points.
The first point is to determine the most critical MD,
which can be characterized by the shape or the number
of FV which can be manufactured in the same way.
It is done according to the study context and to the
capability of the process given by its kinematics and
the physical phenomena involved.
The second point is, in the chosen MD, to merge
the selected FV which can be manufactured from the
same substrate. It is done according to the empty
volumes, which must not be between two FV in
the MD.
In the third point, the shapes and the internal structure of the substrate(s) and of the support(s) are here
determined according to:
The selected FV
The other FV
The kinematics of the process
The accessibility required by the process
The physical phenomena that are involved
The mechanical requirements
Figure 6. The FS and the empty volumes of the part to

design.
The empty volumes

The study context
The last point is to determine if all the volumes
are merged. If it is not the case, the first three
points are repeated while taking into account the FS
that have not been yet analyzed and the substrates
obtained previously and so on until all the FV are
merged.
Then, the process is over and an appropriate design
for the chosen AM process is obtained. Indeed, the AM
process specifications are taken into account step by
step in parallel with the functional requirements. This
ensures that the most possible process-related knowledge is taken into account to obtain the final shapes of
the studied part.
4
EXAMPLE
As an example, a case for a robot hinge in stainless

steel has been designed (Fig. 5). The input data of the
study then the three steps of the design are detailed in
this section.
4.1 Input data
4.1.1 Functional specifications
The case is composed of twenty FS: four bearing
holder (hollow cylinders) and sixteen base planes,
which are shown with their nominal dimensions in
black in Figure 6. To enable the assembly of the case
with the other parts of the robot, some empty volumes
are defined. They are represented by the transparent
volumes.
The functional analysis of the robot has enabled to
determine the geometric and dimensional specifications linked with each surfaces. An extract is shown in
Figure 7.
The mechanical requirements are translated into a
final minimal thickness t for each surface. It is 5 mm
for the hollow cylinders and 3 mm for the based planes.
470
Table 2.
Definition of the thicknesses.

t
(mm)
a
(mm)
e
(mm)
d
(mm)
Surfaces
E
(mm)
Cylindrical
Flat
5
3
0.2
0.2
0.5
0.5
0.8
0.8
10
6
5.9
3.6
Figure 7. An extract of the specifications taken from

functional analysis.
Table 1.
Characteristics of the process CLAD.
required
maximal
accessibility dimensions
Kinematics (mm)
(mm)
5 axis
60
a
(mm) (mm)
500*650*700 0.8
0.3
0.2
4.1.2
4.1.3
Context
Because of the robotic context, the global
objective is to minimize the mass of the
studied part. Manufacturing characteristics
TheAM process that is chosen to manufacture this case
is CLAD. It is a PBMD process, based on the 3D layer
by layer deposition of laser melted powders to build the
profile of the requested part. Its main characteristics
are presented in Table 1.
In this example, the assumptions that parameters d,
and a are equal in all the directions is made.
The constraints due to the physical phenomena
linked with the process (in particularity the thermal
phenomena) are limited to a design rule: the thicknesses of the substrates must be at less equal to the
thicknesses of the FV that it enables to manufacture.
4.1.4
Finishing process characteristics
4.2 The finishing process that is chosen is high

speed milling; it involves a minimal
overthickness of 0.5 mm. Step 1: Analysis
According to the geometric analysis of the functional
surfaces (Figs. 67) and the maximal and minimal
dimensions of the manufacturing process (Tab. 1), the
result of the first step is that all the dimensions are
compatible with it.
4.3
Step 2: determination of the functional volumes
The thicknesses E are determined from the different

parameters of Equations 2 and 3, the results are given
Figure 8. The different design stage of the volumes: (a) the

FV (b) the LV in Z1 (c) the LV in Z2 (d) the LV in Z3 .
in Table 2. All the functional volumes are shown in

Figure 7(a).
4.4 Step 3: determination of the linking volumes

4.4.1 Linking volumes in Z1
The manufacturing directions are determined according to the context of the study, to minimize the mass
of the part. A geometric analysis of the FV enables
to determine two MD:Z 1 and Z2 (Fig. 8(a)). Because
the process has 5 axes work, the both can be studied
separately. Z1 enables to manufacture heighten FV (in
dark grey) and Z2 enables to obtain the two last (in
light grey), Z1 is thus preferred and studied in first.
To minimize supports in the empty volumes shown
in Figure 5 and thus to minimize the finishing operations, all the FV linked with Z1 must not be manufactured together but from two different substrates. The
substrates positions are determined (Fig. 8(b)) to minimize supports and time of finishing and to guarantee
accessibility for the powder feed nozzle.
In this example, their geometries are simple. They
are defined according to the FV position and thickness. The supports are then added (by a normal
projection of the FV on the substrates).
Because all the FV are not merged another MD is
analyzed.
471
The future works will focus on a local optimization

of the shapes and the internal structures of the linking volumes as it is done for example by Almeida &
Brtolo (2010).
5
Figure 9. The final shapes of the studied case.

Contrary to the substrates obtained previously, the two
hollow cylinders of the Z2 direction involve a single
substrate possibility. That is why the second studied
MD is Z2 .
As previously, because of the empty volumes, the
two cylinders cannot be manufactured together. It
involves two substrates which are shown in Figure 8(c).
They dimensions subsume the FV because of a lack
of accessibility which is, because of the other FV, lower
than that required by the process (Tab. 1).
As all the FV are not yet merged, the process is
repeated one more time.
A third MD Z 3 is determined from the previous substrates (common orthogonal).
The final substrate and supports are determined one
more time, due to the context, to minimize the material
quantity and to avoid as much as possible the empty
volumes.
The substrate dimensions are given by the volumes
that it merges and the supports are, due to the 5 axes
of the process, obtained by a normal projection of the
previous substrate on the last one.The result is shown
in Figure 8(d).
Finally, all the FV being merged, the methodology
is finished and the blank of the final part is obtained.
4.4.4 Final result
Thanks to the methodology that is proposed, each
shape of the part has been designed in order to utilize
the characteristics and capabilities of the AM process
CLAD to meet the functional specifications of the
study while taking into account its general context.
At the end, the finishing is done by milling; it
enables to meet the required specifications (geometric, dimensional) and to remove the material in the
empty volumes.The final part after finishing is shown
in Figure 9.
Because the global objective to minimize mass is
taking into account all along the design stage, the
geometry enables to obtain a final weight of only
550 g.
CONCLUSION
This paper described the initial stage of a promising research about a global DFAM approach. A new
methodology is proposed to obtain an appropriate
design forAM processes.It starts directly from both the
functional specifications and the AM processes characteristics in order to enable to best utilize these processes capabilities to achieve desired performances.
Indeed, thanks to its 3-steps structure, it enables to
prevent the psychological inertia phenomena which
can limit innovation while taking in consideration the
specificities of the additive manufacturing processes.
Nevertheless, our methodology still must be
upgraded, in particularity to be even more global
and general.Moreover, in order to obtain shapes and
structures always more appropriated to the AM processes, our works will focus on the capitalization of
knowledge about them.
REFERENCES
Almeida, H. D. A., & Brtolo, P. J. D. S. (2010). Virtual topological optimisation of scaffolds for rapid prototyping.
Medical engineering & physics, 32(7), 77582.
Harzheim, L., & Graf, G. (2005). A review of optimization
of cast parts using topology optimization. Structural and
multidisciplinary optimization, 31(5), 388399.
Kerbrat, O., Mognol, P., &Hascoet, J. Y. (2010). Manufacturability analysis to combine additive and subtractive
processes.Rapid Prototyping Journal, 16(1), 6372.
Kerninon, J., Mognol, P., &Hascoet, J. Y. (2009). Global
approach to design and manufacture Direct Parts.
Advanced research in Virtual And Rapid Prototyping,
111116.
Masood, S. (2002). A rule based expert system for rapid
prototyping system selection. Robotics and Computer
Integrated Manufacturing, 18(34), 267274.
Rosen, D. W. (2007). Computer-aided design for additive
manufacturing of cellular Structures. Computer-Aided
Design & Applications, 4(5), 585594.
Ruan, J., Sparks, T., Fan, Z., Stroble, J., & Panackal,A. (2006).
A Review of Layer Based Manufacturing Processes for
Metals. Solid Freeform Fabrication Symposium (pp. 233
245).
Ruffo, M., Tuck, C., & Hague, R. (2006). Cost estimation for
rapid manufacturing laser sintering production for low to
medium volumes. Journal of Engineering Manufacture,
220(9), 14171427.
Sambu, S., Chen, Y., & Rosen, D. W. (2004). Geometric Tailoring: A Design for Manufacturing Method for Rapid
Prototyping and Rapid Tooling. Journal of Mechanical
Design, 126(4), 571.
Teitelbaum, G. A. (2009). Proposed build guidelines for use
in fused deposition modeling to reduce build time and
material volume. Mechanical Engineering.
Wohlers, T. (2010). Wohlers report: Additive Manufacturing
State of the Industry. Fort Collins: Wohlers associates.
472

A novel approach to produce Functionally Graded Materials

for additive manufacturing
P. Mognol, P. Muller & J.Y. Hascoet
Institut de Recherche en Communications et Cybernetique de Nantes (UMR CNRS 6597), Nantes, France
ABSTRACT: Recent developments in additive manufacturing processes add possibilities to manufacture industrial fully dense metallic parts. Nowadays, customized parts can be obtained in short delay directly from digital
data. Moreover, one of major interests of these processes is the possibility to carry out parts with functionally
graded materials (FGM). The state-of-the-art in this field is limited to studies on manufacture of discrete FGM
which consist to assembly homogeneous materials layers with modification of material after one or several layers,
or nearly continuous FGM on samples. A method to produce morphologically complex parts with a continuous
manufacturing approach is proposed in this paper. This approach enables to have a continuous modification of
material throughout the fabrication of functional complex parts.
1
INTRODUCTION
FGM can be characterized by the variation in composition and structure gradually over volume (Fig. 1).
Results are corresponding changes in the properties
of the material like hardness, elasticity module, wear
resistance or biocompatibility. The concept of FGM
was first proposed in 80s to develop heat-resistant
materials for the propulsion system and airframe of
space planes (Niino et al., 1987). Some fields of
activitieslike aeronautical (Domack & Baughman,
2005) or biomedical industry (Balla et al. 2009)-are
concerned by FGM developments.
FGM can be produced by many manufacturing processes which are divided into three groups: gas phase,
liquid phase and solid phase processes (Kieback et al.
2003). Additive manufacturing processes of metal
components have the advantage of producing important size parts with complex morphology. So they
represent an ideal solution to develop and use these
materials. However, some scientific limits exist to
manufacture FGM parts.
Since first studies about manufacture of metallic
multi-materials structures in middle of 90s (Fessler
et al. 1997) and metal-ceramic structures in the beginning of 2000s (Pei et al. 2002), much research have
been done to develop FGM for various applications.
Even if metal-ceramic structures are limited by the
fabrication of few metal-ceramic layers in metallic
substrate, knowledge about this topic begins significant. Indeed, there are many studies about FGM analysis: micro-structure, material phases, macro-structure,
mechanical and biological characteristics (Banerjee
et al. 2003, Ocylok et al. 2010). They show that FGM
samples produced with additive manufacturing processes are interesting because they permit to have
various properties gradients. Papers concern influence
Figure 1. Homogeneous material, bonded material and

FGM.
of few process parameters on FGM and processes control (Yadroitsev et al. 2007, Majumdar et al. 2009).
Design, representation and process plan aspects has
been partially studied but majority of them dont take
into account specificities of processes or are limited to discrete multi-materials manufacture (Xu &
Shaw 2005, Shin et al. 2003). Digital chain becomes
adapted to additive manufacturing but not totally to
FGM (Bonnard et al. 2009).
Majority of FGM parts produced with additive manufacturing processes are simple morphology parts,
non-functionally and with small dimensions. It is necessary to have a global control of processes and to
develop methodologies to help designers and manufacturers to produce FGM complex parts. Methodologies
must enable to make appropriate choices concerning design and manufacturing and take into account
specificities of additive manufacturing processes.
2
METHODOLOGY FOR MANUFACTURING

FGM PART
The main objective of this methodology is to find the

best way to obtain a fully dense FGM part. The principle is to associate multi-material part design with
manufacturing strategies, given that the material repartition and the geometry of the FGM part are defined.
473
Figure 3. Example of a FGM part: material repartition,

domain DM and sub-domains DMi .
Figure 2. Principle of the methodology.
It consists in classifying all typologies of bi-material

gradients with mathematical description. This mathematical description gives information about gradient
direction and type of equipotential surfaces.According
to these data, geometry and process characteristics, the
design is associated with manufacturing strategies in
order to choose slicing and path strategies and global
control of process parameters (Fig. 2).
Strategies are proposed taking into account multimaterials design and process requirements. This step
of the methodology considers in all manufacturing aspects. Strategies parameters present in monomaterial manufacturing are significant for FGM parts
manufacturing (Kerninon et al. 2008):
Slicing strategy (3 axis, 5 axis slicing),
Path strategy (raster path, contour path).
Proposals are principally made in terms of material
distribution with mathematical data from the classification. But the geometry of part limits the possible
directions of deposition. That is why, the direction of
deposition and slicing strategy are chosen in terms of
the typology of gradient and also geometry of part and
requirements of process. These strategies are proposed
for discrete and continuous FGM manufacturing.
The classification of gradients and first tests-parts
are presented in this paper. FGM test-parts were manufactured and analyzed to begin the study of the global
control of an additive manufacturing process.
3
3.1
CLASSIFICATION OF GRADIENTS
Figure 4. Composition in mA and mB .
The function M represents the material composition

of FGM in space. It is defined in a domain DM R3 ,
corresponding to the part:
DM can be decomposed in several sub-domains DMi

(Fig. 3). In this case, methodology must be applied on
all sub-domains.
Material composition (Fig. 4) is defined by:
M(xyz) = 0 if part is made up of only material mA ,
M(xyz) = 0 if part is made up of only material mB ,
The gradient is the vector corresponding to the
variation of the function M in DM :
3.2 Gradient dimension

A gradient is a one dimensional gradient (Fig. 5) if
and only if there exists a constant vector, such that:
A gradient is a two dimensional gradient (Fig. 5) if

and only if there exist two constant vectors, such that:
Definitions and assumptions
For this methodology, parts must be made up of two

materials mA and mB . They consist of only one area
with exclusively material mA and only one area with
exclusively material mB .
474
Others gradients are three dimensional gradients.
Figure 5. Gradient dimension: representation of a one

dimensional gradient and a two dimensional gradient.
Figure 8. Representation of a closed gradient and equipotential lines in Pi (b).
An equipotential surface is a surface on which the

function M has a constant value. It is defined by:
A two dimensional gradient is an offset gradient

(Fig. 8) if and only if:
A three dimensional gradient is an offset gradient if

and only if there exist three real numbers a1 , a2 and
a3 , such that:
Figure 6. Representation of a two dimentional gradient and
representation of the spaces DA and DB .
where b ( M) is the orthogonal projection of M on

Pi (b).
Otherwise it is a complex gradient.
3.4 Representation of classification
FGM parts which check assumptions (Section 3.1) are
classified in one of typologies of gradient (Fig. 9).
3.5 Example
Figure 7. Representation of a closed gradient and representation of the space {DM DA DB }.
3.3
Equipotential surfaces
The spaces DA and DB (Fig. 6) are defined by:
The domain DM of the example of FGM part is decomposed in two sub-domains DM1 and DM2 (Section 3.1).
The material mA is in the middle of the part while the
material mB is on the outside. On the domain DM1 , the
function M is defined by:
A two dimensional gradient is an open gradient if

and only if the space {DM DA DB } is simply
connected. Otherwise it is a closed gradient (Fig. 7).
475
Figure 10. Example of a FGM part: material repartition on

the domain DM , material repartition on the domain DM1 and
representation of the space {DM1 DA1 DB1 }.
The gradient is an offset gradient on the domain DM1 .

The gradient is the same type on the domain DM2 .
It is possible to classify the different FGM and to
define them mathematically to develop a manufacturing strategy.
MANUFACTURING AND ANALYSIS OF

FIRST TEST-PARTS
4.1 Experimental procedure
Figure 9. Representation of classification.
The gradient is defined by:
There exist two constant vectors such that:
with d1 = (1,0,0) and d2 = (0,1,0), so the gradient is

a two dimensional gradient.
The space DM1 DA1 DB1 is simply connected,
so the gradient is a closed gradient (Fig. 10).
4.1.1 Process and powders

Laser direct metal deposition experiments were conducted with the CLAD process. This process is
based on the three dimensions layer by layer deposition of laser melted powders to build the profile of the
requested part (Fig. 11). The CLAD system consists
of a five-nozzle coaxial powder feed system, a 4000
W fiber laser and three powder feeders. Argon gas is
delivered to prevent the melt pool from oxidizing and
oxide contamination throughout fabrication.
Commercially stainless steel 316L powder with
particle size between 45 and 90 m and chromium
cobalt alloy Stellite 6 with particle size between 50
and 120 m was used in these studies (Tab. 1). Testparts were manufactured on a substrate of 10 mm thick,
commercially stainless steel 316 L.
Test-parts were deposited using the two powder
feeders, while the first contained elemental 316L powder, the second contained elemental Stellite 6 powder.
The composition of deposited powder was achieved by
controlling the flow rate of powders from each feeders and gas in pipes. Optimal process conditions to
manufacture 100% 316L or 100% Stellite 6 are known.
476
4.1.2 Continuous FGM manufacturing

The deposition started with only the first powder
feeder, containing 316L powder. After one millimeter,
the flow rate of the first powder feeder was reduced
continuously and flow rate of the second powder feeder
was increased continuously in the same proportion.
The deposition finished with only the second powder
feeder.
4.1.3 Material analysis
Parts were polished and analyzed. Composition and
micro-structure along compositional gradient has
been characterized with scanning electron microscopy
(SEM) and energy dispersive spectroscopy (EDS).
Samples without pore and crack were obtained

using additive manufacturing CLAD process. All
powder particles are melted (Fig. 13).
Composition analyses were made all 0.5 millimeters on all elements present in 316L and Stellite
6 (Fig. 14). The composition is determined by the
iron and cobalt rate, which are representative of the
presence of 316L and Stellite 6.
Analyzed composition is different from commanded composition. The gradient is not linear and
is focusing on four or five millimeters.
Moreover, the evolution of composition is more progressive after eight millimeters. The difference is the
consequence of inertia that exists between the control
order and effective action at the nozzle. This parameter
is now introduced into the control loop.
4.2 Analysis
First test-parts were manufactured with continuous
approach to control the process and identify possible technical problems. These studies were made on
material aspect: micro-structure, interface between
substrate and the first layer, interfaces between layers and material composition. Samples are thin walls
(Fig. 12) 15 mm height, 95 mm length and 0.8 mm
thickness with an one dimensional gradient. The
direction of the material gradient is perpendicular to
the substrate.
CONCLUSION
In this paper, a methodology to find best way to

produce FGM parts is exposed. A classification of
gradient is proposed from material repartition. Mathematical data about this gradient are obtained with
this classification. The general principle of strategies
choice and main issues are introduced. First test parts
were manufactured with continuous approach and analyzed. Samples with near pore and crack free were
obtained using additive manufacturing CLAD process. Material gradient is continuous but global control
of process must be improved.
Further researches will be conducted in two ways.
First, to develop the determination of manufacturing
strategies: strategies must be proposed for each type of
gradient with novel paths, criteria must be established
to choose best strategies in terms of material, geometry
and process.
Second to improve the global control of CLAD process for multi-materials manufacturing. Test parts and
Figure 11. Direct Laser Manufacturing system.

Table 1.
Material
Chemical composition (wt.%).

C
Co Cr Fe
Stellite 6 1.2 60
316L
0.03
Mn Mo Ni Si
29 2
17 64 2
2 1
4.5
12 0.75 0
Figure 12. Thin walls sample.
Figure 13. Microstructure of thin wall sample.
477
Figure 14. Composition (wt. %).
functionally complex parts will be produced to show

the interest of the methodology.
This work was carried out within the context of the
working group Manufacturing 21 which gathers 16
French research laboratories. The topics approached
are: the modeling of the manufacturing process, the
virtual machining and the emerging of new manufacturing methods.
REFERENCES
Balla, V., De Vas Con Cellos, P., Xue, W., Bose, S. &
Bandyopadhyay, A. (2009). Fabrication of compositionally and structurally graded Ti-TiO2 structures using laser
engineered net shaping (LENS). Acta Biomaterialia, 5(5),
18311837. Acta Materialia Inc.
Banerjee, R., Collins, P., Bhattacharryya, D., Banerjee, S. &
Fraser, H. (2003). Microstructural evolution in laser
deposited compositionally graded / titanium-vanadium
alloys. Acta Materialia, 51(11), 32773292.
Bonnard, R., Mognol, P. & Hascoet, J. (2009). Integration
of rapid manufacturing processes in a high-level numerical chain. Proceedings of 4th International Conference
on Advanced Research in Virtual and Rapid Prototyping.

Leiria, Portugal.
Domack, M. & Baughman, J. (2005). Development of
nickel-titanium graded composition components. Rapid
Fessler, J., Nickel, A., Link, G. & Prinz, F. (1997). Functional gradient metallic prototypes through shape deposition manufacturing. Proceedings of the solid freeform
fabrication symposium, 521528. Austin, USA.
Kerninon, J., Mognol, P., Hascoet, J. & Legonidec, C. (2008).
Effect of path strategies on metallic parts manufactured
by additive process. Proceedings of the Solid Freeform
Fabrication Symposium, 352361. Austin, USA.
Kieback, B., Neubrand, A. & Riedel, H. (2003). Processing
techniques for functionally graded materials. Materials
Science and Engineering: A, 362(12), 81106.
Majumdar, J., Manna, I., Kumar, A., Bhargava, P. & Nath, A.
(2009). Direct laser cladding of Co on Ti-6Al-4V with
a compositionally graded interface. Journal of Materials
Processing Technology, 209(5), 22372243.
Niino, M., Suzuki, A., Hirai, T., Watanabe, R., Hirano, T. &
Kuroishi, N. (1987). Method of producing a functionally
gradient material. Ausz. Eur. Patentanmeld. I, 3(30), 1555.
Ocylok, S., Weisheit, A. & Kelbassa, I. (2010). Functionally
graded multi-layers by laser cladding for increased wear
and corrosion protection. Physics Procedia, 5, 359367.
Pei, Y., Ocelik, V. & De Hosson, J. (2002). SiCp/Ti6Al4V
functionally graded materials produced by laser melt
injection. Acta Materialia, 50(8), 20352051.
Shin, K., Natu, H., Dutta, D. & Mazumder, J. (2003). A
method for the design and fabrication of heterogeneous
objects. Materials & Design, 24(5), 339353.
Xu, A. & Shaw, L. (2005). Equal distance offset approach
to representing and process planning for solid freeform
fabrication of functionally graded materials. ComputerAided Design, 37(12), 13081318.
Yadroitsev, I., Bertrand, P., Laget, B. & Smurov, I. (2007).
Application of laser assisted technologies for fabrication
of functionally graded coatings and objects for the International Thermonuclear Experimental Reactor components.
Journal of Nuclear Materials, 362(23), 189196.
478

Electrochemical micromilling with ultra short pulses

R. Zemann, F. Bleicher, C. Habersohn & R. Zisser-Pfeifer
Institute for Production Engineering and Laser Technology, Vienna University of Technology, Austria
ABSTRACT: The tendency to make progressively smaller and increasingly complex products is no longer
an exclusive demand of the electronics industry. Many fields such as medicine, biomechanical technology, the
automotive and the aviation industries are searching for tools and methods to realize micro and nanostructures
in various materials. The micro-structuring of very hard materials, like carbides or brittle-hard materials, pose
a particularly major challenge for manufacturing technology. For these reasons the Institute for Production
Engineering and Laser Technology (IFT) of the Vienna University of Technology is working in the field of
electrochemical micromachining with ultra short pulses. With the theoretical resolution of 10 nm, this technology
enables high precision manufacturing. [Kock M.]
1
INTRODUCTION
The machining technology of electrochemical micromachining (ECM) is based on the already wellestablished fundamentals of common electrochemical
manufacturing technologies. The enormous advantage of the highest manufacturing precision underlies
the fact of the extremely small working gaps achievable through ultra short voltage pulses. This describes
the main difference to common electrochemical
technologies.
Another big advantage of the electrochemical micro
machining technology is that the treatment of the work
piece takes place without any mechanical forces or
thermal influences. Therefore, no abrasive wear of the
tool occurs and aspect ratios of >100 are possible
which sets the basis for extremely sharp-edged geometries; so there is no unintentional rounding of edges and
no burring on the part.
At the moment for several nonferrous metals like
nickel, tungsten, gold etc., as well as alloys like noncorroding steel 1.4301, appropriate electrolytes have
already been found. Nevertheless, a main focus of
research for the Institute will be the search for new
material-electrolyte combinations to expand the field
of application for this technology and to enhance its
manufacturing productivity. This needs to be accomplished in order to fulfill the requirements of industrial
production because in industries such as the automotive sector, the rate of production is very important.
At the Nano-/Micro-Machining-Center of the IFT, an
excellent assortment of measuring devices are available, for example, a Zeiss F25 coordinate measuring
machine, high end optical measuring devices like
the Alicona Infinite Focus 4G or the Nikon Nexiv
VMR-3020 and a JEOL JCM-5000 scanning electron
microscope. Based on the technology of ECM and
by the use of high end measuring devices, specimens
and parts in the micrometer range and smaller are to
be manufactured and analyzed in order to investigate

material removal rates and the accuracy of resulting
work piece geometries.
Due to the multidisciplinary nature of this technology, intensive cooperation with other institutes of
the Vienna University of Technology in the fields of
electro-technical engineering, high frequency technology and electrochemistry are established. The goal
of this research will be to elevate this technology
to an appropriate level for possible industrial use by
enhancing the manufacturing accuracy and the process efficiency for current components. Therefore a
profound knowledge of material science, electrochemistry and production technology for extremely small
dimensions will be required. The necessary expertise
in these fields will be provided by the cooperating
institutes and interested companies.
To accomplish these improvements in the technology of electrochemical micromachining with ultra
short pulses it will be necessary to merge several
research projects which are already dealing with topics
of piezo driven nano-positioning devices or development of high precision machine structures for different
types of machines.
ELECTROCHEMICAL MICROMILLING
WITH ULTRA SHORT PULSES (ECM)
Similar to conventional electrochemical manufacturing methods the ECM process with ultra short voltage
pulses uses an oppositional electric voltage for the
work piece and the tool. At the phase boundaries
between tool and electrolyte and also between work
piece and electrolyte, an electrochemical double layer
is formed. [Kirchner V.] The functionality of which can
be understood principally as a kind of a double capacitor. In addition to the proper choice of the electrical
479
Figure 2. Time variable voltage curve in the electrochemical

double layer with (red) and without (black) pre-pulse.
Figure 1. Electrochemical double layers and schematic

illustration of the electrolyte as electrical resistor between
tool and work piece.
Figure 3. Tools after producing a groove with different tool

voltages.
process parameters like the amplitude of the pulses, the

pulse width, the voltages at the tool, the work piece,
and the backing electrode, the right choice of electrolyte is probably the most important aspect for this
process. The whole machining process takes place in
a basin filled with an electrolyte solution which has
to be adequately adapted to the work piece material
used. Even during the filling of the basin, the greatest
caution is required due to the fact that once in contact with the electrolyte, the surface of the material
immediately begins to react. To prevent the work piece
surface from the influence of the electrolyte-solution,
a cathodic protection-current is applied. With the basin
filled as needed, the process of work piece calibration
can be performed. This is an advantageous aspect of
this technology because neither the tool nor the work
piece has to be unclamped.
Then the manufacturing program, which conforms
to a standard CNC-program, is started. The tool
moves along the pre-programmed paths and selectively
ablates material due to the effect of the voltage transfer
in the electric double layer triggered by the ultra short
pulses. Figure 1 shows the schematic illustration of the
tool and the work piece in the electrolytic reservoir
and the electrochemical double layers. If the voltage
pulse width is very short, the erosion takes place very
close to the tool (red part in Figure 1), since the ohmic
resistance of the electrolyte prevents ablation at areas
further away of the tool (blue part) due to the double layer capacitor not being able to be sufficiently
recharged. [Zemann R.]
The pulse width depends on the choice of electrolyte
and the necessary precision of the machined structure.
Small working gaps between the tool and the work
piece of less than 1 m are produced with pulse widths
of less than 100 nanoseconds. Even more accurate
machining can be achieved with pulse widths of less
than 1 nanosecond and by separating the processing
pulse into a pre-pulse and a main pulse. In particular,

the investigation of the influence of targeted pre-pulses
is one of the main research topics to be elaborated at the
IFT (Figure 2). In order to elaborate on the research
work concerning the technology of using ultra short
voltage pulses, the relevant demands of industry, basically increasing the material removal rate, has to be
considered as a main goal. Subsequently, an increase in
the already excellent machining accuracy is regarded
as a main target.
Another major advantage of this technology is the
possibility to reverse the process electrically. This
means that not only the work piece can be machined,
but also the tool itself can be defined as work piece
and be machined to its ideal geometry without any
further set-up.
The existing technical installations also admit a
high precision calibrating process. Regarding all these
functionalities the requirements for precise micro
machining are met. Possible tasks that can be performed with this machining centre include: tooling,
milling, turning, sinking and measuring.
Characteristics of the ECM process:
High precision (theoretical resolution of 10 nm)

No thermal load
No mechanical process forces
High aspect-ratio (>100)
No tool wear
Small working gaps between tool and workpiece
(<1 m)
Manufacturing of hard materials
Very small edge-rounding
No burring
High quality measuring function
Besides the research work on the basics of ECM

machining itself, and the development of the pre-pulse
480
Figure 4. Effect between dwelling time and material

ablation.
technology, several application like experiments will

be explored. One of the main objectives of these experiments will be the specific processing of the surface
of cutting plates to manipulate the chip formation in
milling and turning operations. This surface structuring method will also be analyzed; on the one hand
considering the application for micro injection moulding tools and on the other hand for the production of
special surfaces with similar behavior like the lotus
flower- and sharkskin-effects
MANUFACTURING OF TUNGSTEN
The refractory metal, tungsten, has very interesting

characteristics in relation to use as tool and work
piece material in the process of electrochemical micromachining. For example, the hardness (Mohs) of 7.5
and the Youngs modulus of 411 GPa, which is around
twice as high as that of steel, is an advantage for micro
materials. Also, the possibility to take the caustic
metallic base, sodium hydroxide (NaOH), which is a
common and well established base in industry, for the
processing electrolyte is another advantage of tungsten. [Buhlert M.] The so produced electrolyte for the
manufacturing process would be minimally hazardous
and easily available; so far it is a preferable substance
for experimental and industrial use. A lye concentration of 2M NaOH assures a good ratio between
machining time and structure precision for most experiments. If the concentration was higher the precision
will decrease and the ablation rate will increase; for a
less concentrated electrolyte the effects are vice versa.
Due to tungstens characteristics it is the preferable material in micro manufacturing for products
with structural requirements. For example at the IFT,
tungsten is used for the engineering of the small
measuring calipers for the use in high precision measuring machines like the Zeiss F25. These styluses,
with diameters of less than 100 m, are machined by
electrochemical micromachining with a tungsten tool
from a tungsten work piece. The problem with having
the same material for the tool and the work piece is
that generally the electrolyte which is adapted to the
work piece material is also able to react with the tool.
So it is possible that in some ranges of the different
electrochemical parameters, for example the voltage
at the tool, the tool dissolves during manufacturing;
now the process would experience a kind of wear.
With other pairings, for example non-corroding steel
as tool material, tungsten as work piece material and
NaOH as electrolyte, there is chemically no danger
of any unwanted dissolution. But in some cases it is
necessary to manufacture the tool and the work piece

in the same electrolyte to avoid any time delays.
Figure 3 shows the results of incorrectly choosing
the electrochemical parameters. The voltage at the tool
should be set so that the tool experiences no wear
and that there is no traditional electrochemical ablation through a polarization of the work piece caused
by the voltage at the tool. The pictures in Figure 3
show two different tools with diameters of 250 m.
The tool in the left picture has no kind of wear. The
right picture shows a tool with wear through a positive
tool voltage. The resulting chamfer has the dimensions
of 42 34 m after milling a grove with a length of
1000 m and a depth of 40 m. With this information
it would appear that by setting the voltage at the tool
to the negative extreme, it would be possible to avoid
that dissolution at the tool, but there is another aspect
which narrows the negative voltage range. If the voltage at the tool is, for example, 500 mV you might
not have wear at the tool but the tool will positively
polarize the work piece. If this local polarization is
large enough to superimpose the cathodic protective
effect, some unwanted dissolution could again occur.
This time though, the work piece would suffer and traditional problems of electrochemical machining like
a strong edge rounding would be witnessed. Another
reason again for not setting the voltage at the tool very
negative is that the dissolved and positive ions coming
from the work piece may attach to the tool. So the tool
geometry will change, which is undesirable in regard
to the manufacturing precision of the process. There
is a specific range in which the voltage at the tool has
to be set within. In general, this range depends on the
materials to be used, the electrolyte and the concentration of the electrolyte. For tungsten in 2M NaOH
the voltage at the tool should be between 200 and
100 mV.
Another phenomenon of traditional electrochemical
machining technologies is that of conical shaped
flanks after, for example, drilling of a hole. To solve
this problem there are a number of different strategies. One possible solution is to change the form of
the tool from a simple cylinder to a tool with a thin
shaft and a disc at the end. This leads to a longer distance between the shaft of the tool and the work piece
so that only the disc surface of the tool is near enough
to recharge the electrochemical double layer formed
at the phase boundaries of the work piece. Such tool
geometry is used for better localization of the ablation of drilling processes. Another possibility to solve
this problem is to separate the drilling process in a
strong dissolving and a slight dissolving process part.
This is possible through different parameter sets and
does not need any preparation time at the machine.
The finishing part of the process could have an amplitude of 2200 mV and a pulse width of 80 ns instead of
2800 mV and 200 ns for the basic process. With this
strategy it is possible to outperform the single cycle
process with just one parameter set in case of machining precision. Experiments with such a strategy have
shown, for example, that the edge rounding decreases
481
and a pulse width from 70 to 200 ns is shown. This

process would be a rather more dissolving one. If the
precision requirements of the product are not met a
finishing cycle could help.
Finishing with an amplitude of 2200 to 2500 mV
and a low pulse width under 100 ns would create a
sharper structure.
4
Figure 5. Working gap depending on the pulse width at a
pulse amplitude of 2750 mV for a tool with a diameter of
75 m in 2M NaOH.
to under 4 m and the roughness of the produced

surfaces is about Ra 0.06.
Figure 4 shows the effect of the dwelling time during
the process. In this experiment a tool with a diameter of 250 m was positioned 1 m over the work
piece surface. The voltage of the tool was 100 mV,
the amplitude of the pulse was 2800 mV and the pulse
width was 200 ns. There are eight stop positions visible. On each position from the left to the right end the
dwelling time was doubled from 5 s at the first position
to 640 s at the last position. With the maximum depth
of 7 m after the dwelling time of 640 seconds this
experiment confirmed the relevance of the dwelling
time for the manufactured geometry. That is one of the
effects, which has to be controlled for an industrial use
of the ECM technology.
The most important parametrical relationship in
the use of electrochemical micromachining is that
between amplitude, pulse width and working gap.
[Hamann C.H.] The energy for the process is the integral over the amplitude and the pulse width. Increased
energy in the double layer at the work piece leads to
a rougher, more powerful process. In this case the
appeared working gap is large and the precision of
the process is no advantage for the ECM technology.
The size of the working gap is the most important
benchmark for the precision of the parts and the creation of sharp edged geometries. The working gap for
tungsten manufacturing with a cylindrical tool and a
diameter of 75 m in the electrolyte 2M NaOH, has
a range from 2 to over 30 m. In Figure 5, the graph
of the working gap for a pulse amplitude of 2750 mV
CONCLUSION
The technology of electrochemical micromachining

with ultra short pulses has successful displayed the
many applications especially for prototype building
or for the manufacturing of special products where
there is no other technology which can combine a very
high manufacturing precision for special materials
without any mechanical forces or thermal influences.
[Kirchner V.] Also the use of pre-pulse technology
and the applicable effects on process accuracy and
material removal rate of difficult to machine materials
offers a wide range of possible applications for ECM
technologies. The occurring electrochemical problems
are tradable and topics at the IFT, as well as the
micromachining of many other materials like nickel,
non-corroding steels or hard metals.
REFERENCES
Buhlert, M. 2009. Elektropolieren, Saulgau, Eugen G. Leuze
Verlag, ISBN 978-3-87480-249-9.
Hamann, C.H., Vielstich, W. 2005. Elektrochemie, 4.
vollstndig berarbeitete und aktualisierte Auflage,
Weinheim, WILEY-VCH Verlag GmbH & Co. KGaA,
ISBN 3-527-31068-1.
Kirchner, V. 2001. Elektrochemische Mikrostrukturierung mit
ultrakurzen Spannungsimpulsen, Berlin, Dissertation
Freie Universitt Berlin.
Kock, M. 2004. Grenzen der Mglichkeiten der elektrochemischen Mikrostrukturierung mit ultrakurzen Spannungspulsen, Berlin, Dissertation-Freie Universitt Berlin.
Zemann, R. 2010. Electrochemical Milling, Annals of
DAAAM for 2010 & Proceedings of the 21st International
DAAAM Symposium Intelligent Manufacturing &
Automation: Focus on Interdisciplinary Solutions, 20
23rd October 2010, Zadar, Croatia, B. Katalinic, DAAAM
International, Vienna, ISSN 17269679, ISBN 978-3901509-73-5, S. 843844.
482

Functionally Graded Rapid Prototyping

Neri Oxman, Steven Keating & Elizabeth Tsai
Mediated Matter Group, MIT Media Lab
ABSTRACT: Functionally Graded Rapid Prototyping (FGRP) is a novel design approach and technological
framework enabling the controlled spatial variation of material properties through continuous gradients in functional components. Such variations are traditionally achieved as discrete delineations in physical behavior by
fabricating multiple parts comprised of different materials, and assembling them only after the fabrication process
has been completed. Recent advances in Computational Topology Design (CTD) and Solid Free-Form Fabrication (SFF) are promoting the creation of building components with controlled micro-, and macro-architectural
features. The FGRP approach combines a novel software environment with a mechanical output tool designed
as a 3-D printer to allow computer control of material distribution within a monolithic structure. Inspired by the
integration of material, structure and form found in natural systems, this biologically inspired design approach
allows for physical prototyping of graduated properties in product and architectural design scales. FGRP introduces the potential to dynamically mix, grade and vary the ratios of different materials, resulting in continuous
gradients, and structurally optimized designs with efficient use of materials, reductions in waste and production
of highly customizable features. The paper presents the FGRP technology as part of an overall integrated design
approach to functionally gradient design fabrication. Two work-in-progress explorations of FGRP implementation are presented: a robotic arm able to 3-D print concrete with controllable density and a 3-D printer for
UV-curable polymers exploring variable elasticity. Research methods and processes devised for its development
are presented and design applications demonstrated. Current technological limitations and future directions are
discussed and their implications reviewed.
1
1.1
INTRODUCTION
Background
Current rapid prototyping technologies, specifically

additive manufacturing platforms, are limited in
their capacity to represent graduated material properties. Their basic strategy is typically to assign material
properties to pre-shaped building components such as
concrete columns or fiberglass panels (Sachs et al.
1993). Within the design process, this translates into
assigning a material property to predefined solids or
closed surface polygons (Sheng et al. 2003). Both
computer-aided design (CAD) tools and industrial
fabrication processes are thus not set up to represent
graduation and variation of properties within solids
such as varied density in concrete, varied elasticity
in rubber, or varied translucency in glass (Oxman
2011). As a result, the design process is constrained to
the assignment of discrete and homogeneous material
properties to a given shape (Oxman 2011).
The paper introduces a data-driven fabrication
approach for designing and prototyping materials with
graduated properties inspired by functionally graded
materials found in nature.
1.2
Problem definition
Functionally graded materials, which are materials

with spatially varying composition or microstructure,
are omnipresent in nature. A typical cross-section of

a palm tree will reveal radial density gradients corresponding to the bending stiffness instantiated across
its height (Rich 1987, Gibson et al. 2010); cancellous
bone exhibits sponge-like cellular tissue with dynamic
distributions corresponding to loading patterns
(Gibson 1985, Keaveny et al. 1993). Such natural
materials offer material and structural efficiencies at
various length scales (Ortiz & Boyce 2008).
In contrast to natural materials and biological
tissues, industrially fabricated constructions, such
as concrete pillars, are typically volumetrically
homogenous. While the use and application of
homogenous materials allow for ease of production,
many qualities such as improvements in strength,
weight, material usage, and functionality could
be obtained by the development and application of
functionally graded materials at the product and
architectural scales.
2 AIMS & OBJECTIVES
2.1 Developing a new fabrication approach
The mechanical response of materials designed and
engineered with spatial gradients in composition and
structure is of considerable significance in diverse disciplines such as biomechanics, fracture mechanics,
optoelectronics, geology, tribology, nanotechnology,
483
product engineering and even architectural design

(Miyamoto 1999). Damage and failure resistance of
surfaces to normal and sliding contact or impact
can be substantially controlled and modified through
such gradients. Graded materials hold a profound
place in the future of material engineering; the ability to synthetically engineer and fabricate products
with material gradients using additive fabrication
offers improvements in structural and environmental
performance, enhances material efficiency, promotes
material economy and optimizes material distribution
(Markworth et al. 1995). These processes however, are
known to exist predominantly at the micro scale, and
are well known in fields such as tissue engineering
(Fan et al. 2000, Yang et al. 2002, Yeong et al. 2004).
Given the significant potential of the ability to
design and fabricate building components with varied properties (i.e.density, elasticity, translucency,
etc.) supporting the integration of functions (i.e. loadbearing, natural ventilation, etc.), this work seeks to
develop modeling and fabrication environments for
functionally graded products of industrial application
and architectural scale.
The paper aims to introduce the Functionally
Graded Fabrication (FGF) process as a general
approach to the design of structural components
with graduated properties. In addition, the work will
demonstrate a novel deposition technology, coined by
the authors Functionally Graded Rapid Prototyping
(FGRP). This technological framework offers gradation control of materials within one 3-D print with the
aim of increasing mechanical efficiency and reducing
energy input in the products generation and lifespan.
We introduce the application of this approach in two
material contexts polymers and concrete with the
aim of enabling 3-D rapid prototyping of structural
components with variable properties corresponding to
structural, functional, and environmental conditions.
2.2
Developing novel design and fabrication

processes using FGRP
The aim is to produce an automated tool able to

dynamically mix and vary the ratios of different
component materials in order to produce complex
continuous gradients in monolithic structures. Two
separate examples of FGRP-based processes are being
developed: a variable-density concrete system and a
variable-elasticity polymer system. We report upon
work-in-progress for the utilization of this approach
in product-design scales, and speculate on its further
development and automation.
3
3.1
STATE OF THE ART

Comparison to existing RP technologies and
prior arts
Rapid Fabrication (RF) and Rapid Manufacturing (RM) technologies have emerged, since the
mid 1980s, as promising platforms for building

construction automation (Jacobs 1992). The FGF
approach and potential technologies defer profoundly
from current additive fabrication technologies in that
they aim to produce material organizations of varied
properties. Generally classified by the material phase
used in their extrusion whether liquid-based (e.g.
stereolithography), powder-based (e.g. selective-laser
sintering), or solid-based processes (e.g. fused deposition modeling) consistent to all such technologies
is the use of materials with homogeneous properties
for prototyping and fabrication purposes (Jacobs 1992,
Gibson & Shi 1997, Weiss et al. 1997). An expanded
review of prior arts is provided in (Oxman 2011).
3.2 Functionally graded materials
The general idea of structural gradients was initially
proposed for composites and polymeric materials in
1972 (Miyamoto et al. 1999), but it was not until
the 1980s that actual models investigating the design,
fabrication and evaluation of graded structures were
proposed.
Functionally Graded Materials (FGMs) are characterized by the gradual variation in composition and
structure over their volume, resulting in corresponding
changes in material properties. Such materials can be
designed and engineered for a specific set of functions
and applications. Various approaches based on particulate processing, preform processing, layer processing
and melt processing are used to fabricate FGMs.
4
PREVIOUS WORK & APPLICATIONS
4.1 Variable-property design fabrication

Previous work has been carried out setting up the theoretical, methodological, and technical foundations for
Variable Property Design Fabrication (Oxman 2010).
A variable-property 3-D printing prototype able to
dynamically mix and vary the ratios of different materials in order to produce a continuous gradient in a 3-D
printed part was developed. This project establishes a
novel nozzle design coupled with a mixing chamber
that can produce a continuous gradient, using colors
as a substitute for material properties (Oxman 2010,
Oxman 2011).
4.2 Design applications
Potential design applications for Functionally Graded
Rapid Prototyping occupy a vast range of possibilities
in medical device design, product design, and architectural design (Oxman 2007, Oxman 2009). We present
two examples facilitated by the FGF approach: Beast, a
prototype for a Chaise Lounge, is a 3-D printed chair.
It combines structural, environmental and corporeal
performance by adapting its thickness, pattern density,
stiffness, flexibility and translucency to load, curvature, and skin-pressured areas respectively. A single
484
Figure 3. Variable stiffness and elasticity as demonstrated

by Beast. Prototype for a Chaise Lounge. 2008. Boston
Museum of Science.
Figure 1. Graphic illustrations of discretely graded (top)

and multiphase graded (bottom) microstructures of functionally graded materials. Within FGMs, the different
micro-structural phases have different functions; the overall
FGMs attain their multifunctional status from their property gradation, thus enabling various multifunctional tasks
by virtue of spatially tailored microstructures.
Figure 4. Variable thickness and elasticity as demonstrated

by Carpal Skin. Prototype for medical splint. 2008. Boston
Museum of Science.
distribution of elastic modulus is applied. This painmap is then folded back to its 3-D form and 3-D printed
using photopolymer composites (Oxman 2010).
5
METHODOLOGY
5.1 Biologically-inspired rapid fabrication
Figure 2. Physical prototype of Variable-Property 3-D

printer.
continuous surface acting both as structure and as skin

is locally modulated to cater for structural support on
the one hand, and corporeal performance on the other.
Carpal skin is a process by which to map the
pain-profile of a particular patient its intensity and
duration and distribute hard and soft materials to
fit the patients anatomical and physiological requirements limiting movement in a customized fashion.
The formation process involves case-by-case pain registration and material property assignment. The 3-D
scan of the patients hand, including its pain registration, is mapped to a 2-D representation on which the
Previous work and current work reported upon

in this paper are developed and implemented following research into biological systems and natural
formation processes. Rather than a product-oriented
approach to fabrication, we devise a processoriented approach mimicking the formation of tissue
and natural material property variation as explored in
muscle (i.e. variable elasticity), or bone tissue (i.e.
variable density).
Material formation processes found in nature do
not operate by the same logic traditionally applied
in digital design processes and rapid fabrication
platforms, such as additive manufacturing (Sarikaya
1999). Nature constructs solid objects by means of
local processes using unformed locally available raw
materials (Vincent et al, 2006). Demonstrably sustainable, natural structures possess the highest level of
seamless integration and precision with which they
485
roughly uniform diameter along their height by thickening the cell walls in certain regions, producing radial
density gradients across the surface and volume area
of the stem. The density is highest at the peripheries
and lowest in the center, for example with densities
ranging from 1001000 kg/m3 in a single stem of the
Iriarteagigantea (Rich 1987).
6.3 Variable-density graded fabrication of concrete
and cement foams
Figure 5. Left: radial density in palm-tree stem; Right: variable density in spongy bone. Palm tree cross-section image
from the Food and Agriculture Organization of the United
Nations (FAO, 1996).
serve their functions (Oxman 2010). A key distinguishing trait of natures designs is the capability in
the biological world to generate complex structures
of organic or inorganic multifunctional composites
designed with variable properties observed, for example, in the way in which fibers are oriented in wood
or mechanisms by which calcium is distributed in the
bone as illustrated below (Benyus 1997, Gibson &
Ashby 1982,Freyman et al. 2001). Combined with
extracellular matrices, these structural biomaterials
form microstructures engineered to adapt to prearranged external constraints introduced upon them
during growth or throughout their life span (Vincent
1982). Such constraints generally include combinations of structural and environmental performance
(Gibson & Ashby, 1982, Vincent 1982, Vogel 2003).
Since all biological materials are made of fibers,
their multi-functionality often occurs at the micro
or nano scale and is typically achieved by mapping
performance requirements to strategies of material
structuring and allocation; material is concentrated in
regions of high strength and dispersed in areas where
maximal stiffness is not required. The shape of matter
is therefore directly linked to the influences of forces
acting upon it (Neville 1993, Vogel 2003).
6
3-D PRINTING VARIABLE DENSITY

CONCRETE
6.1 Variable density concrete foam fabrication

Presented in this section is work-in-progress for the
rapid fabrication of variable-density cement foams.
First prototypes are presented illustrating foams of
varying densities using aluminum powder admixtures.
The project is motivated by the hypothesis that density gradients in structural building components made
of concrete may increase the strength of a structural
component while reducing material waste.
6.2
Biological inspiration
The work is inspired by load-induced variable densities

found in cancellous bone and by radial-gradient densities found in palm tree stems. Palm trees maintain a
The aim is to design a novel rapid fabrication apparatus

able to dynamically vary cellular materials corresponding to any given structural constraints. The
following work focuses on cement and concrete foams
where the density is controlled through an admixture
of aluminum powder and lime, which react to produce
hydrogen gas bubbles.
Compression samples of Type III Portland cement
foams with varying density were produced and measurements of material density, pore size, and
mechanical properties were taken using mass measurements, micrograph image analysis, and compressive
strength testing, respectively. Measured foam data
agrees well with existing data on cement foams made
with a protein-based foaming agent (Tonyan 1991).
The porosity was able to be controlled by the proportion of aluminum in the mixture; past an optimum
amount the porosity decreased as the ratio of aluminum
to cement increased.The maximum porosity achieved
was 40% using a weight ratio of 1.2 g/kg aluminum
to dry cement mix.The compressive testing results
strongly agree with the exponential ceramic strength
model, with a found value of q = 5.4, compared to
the q value of 7 found in the literature (Rice 1976).
The implications of radial density gradients in concrete foams were explored and theoretical calculations
using the experimentally found mechanical properties
show that for a cylinder under bending stress, a graded
beam can have 9% less mass than a solid cylindrical
beam of the same dimensions and support the same
load. This is achieved by grading the radial density to
set the tensile strength equal to the tensile stress generated by the bending load (while maintaining a shear
stress lower than the shear strength to ensure failure in
bending). This creates a more efficient use of material
by optimizing bending stiffness relative to weight.
6.4 Work-in-progress: automating variable-density
rapid fabrication
Current work-in-progress looks into the controlled
automation of density gradients in concrete using a
robotic platform. Through the use of a dynamic mixing
chamber and extrusion head mounted on a robotic arm
(a 6-axes KUKA arm), we aim to 3-D print concrete
with controllable density. This demonstrate of FGRP
will use a rapid-set concrete mixture with either an aluminum foaming admixture or a protein-based foaming
agent. The use of a 6-axes robotic arm offers complete positional and angular control of the extruder
486
Figure 6. Left: Radial density gradient in a concrete sample produced by varying ratio of foaming agent (aluminum
powder). Top right: Linear density gradient in a concrete sample with the center of gravity highlighted by the pivot point.
Bottom right:Functional gradients of other various properties such as aggregate ratios (top), color/optical properties
(bottom), and other material properties are possible.Samples
made using Type III Portland cement and fly ash as the base
material and were made with assistance from Timothy Cooke
of the Building Technology program at MIT.
head, generating interesting fabrication possibilities

utilizing a setup similar to current fused deposition
modeling technologies. Additional material properties, such as aggregate ratios and optical properties
(Figure 6) can also be controlled through dynamic
mixing.
7
3-D PRINTING OF U-V CURABLE

POLYMERS WITH VARIABLE ELASTICITY
7.1 Variable-elasticity fabrication

Presented here is work-in-progress for the rapid fabrication of functionally graded monolithic polymer
products. The ability to 3-D print products with controllable gradients in material properties may allow for
more optimized designs as well as reductions in the
stresses and strains commonly present at the interface
between materials of different properties.
7.2
Biological inspiration
Biological structures are capable of transitioning

seamlessly between materials of different compositions, properties, and microstructures. Natural materials such as human cartilage and many types of bird
beaks exemplify how spatial variations in density and
stiffness can help optimize the response of structures to environmental and mechanical constraints.
The goal is to mimic these systems by developing
a 3-D printer that is capable not only of producing multi-material structures with an incredibly wide
range of mechanical properties, but also of stretching
the current boundaries of human design and product
manufacturing.
7.3
Biologically-inspired variable-elasticity
graded fabrication
In many biological systems, the physical properties of the materials are determined by the chemical
Figure 7. Samples generated by combining UV curable silicones and polyurethanes with various properties in a gradient
across each sample.
composition and microstructure of its matrix. In soft

collagenous tissues such as cartilage, for example,
the mechanical behavior of the matrix is determined
by the amount and crimp of the collagen it contains.
Experimentally, increased ratios of collagen to proteoglycan in the cartilage matrix correspond to higher
tensile moduli (Williamson 2003). We are currently
in the process of developing polymer mixtures that,
when combined in different ratios, produce blends
with broad ranges of customizable mechanical properties. Shown in Figure 7 are examples of silicone and
polyurethane samples containing gradients in hardness
produced through blended casting.
7.4 Work-in-progress: automating
variable-elasticity rapid fabrication
By controlling the ratios in which two or more polymers are mixed immediately prior to deposition and
UV curing, monolithic structures with functional gradients can be produced using additive fabrication
technologies. Current work-in-progress is focused on
automating the controlled mixing and deposition of
polymer layers using a 6-axes robotic arm, as well as
integrating the physical fabrication platform with user
design interfaces.
8
CONTRIBUTIONS
Functionally Graded Rapid Prototyping (FGRP) is a

novel approach and method introducing the ability to
dynamically mix and vary the ratios of component
materials in complex 3-D distributions in order to produce continuous gradients in a3-D printed part. This
487
ability expands the potential of prototyping, since the

varying properties allows for optimization of material
properties relative to their structural and functional
performance and for more accurate evaluations of the
intended final product, such as stress testing. FGRP
could also contribute to efficient conservation of material usage, high performance of integrated structures,
optimized response to mechanical stimuli, and overall
improved product life spans.
Finally, it is expected that in parallel to the emerging
capabilities of multi-material, freeform fabrication,
materials with a wide range of mechanical, electrical,
thermal, and optical properties will soon be seamlessly
3-D printed. Indeed as of today, traditional CAD programs are inadequate in efficiently utilizing this vast
design potential. In this research, we have outlined an
approach and demonstrated the first steps in the design
of a complimentary approach and potential technology acting as an additive digital fabrication platform
supporting continuous cellular and elasticity gradients within a structural component to meet high level
functional goals.
9
FUTURE WORK
Future work will focus on developing the robotically automated design fabrication platform to achieve
numerically controlled variable density and variable
elasticity in product scales. In addition, the development of a 6-axes 3-D printing platform (in contrast
to conventional 3-axes platforms) will offer opportunities to explore additional avenues of 3-D printing
such as automated embedded-part printing and integrating additional fabrication technologies with 3-D
printing (i.e. milling, painting, casting, etc.). In parallel, continued exploration and development digital of
form-generation environments will support the generation of 3-D forms incorporating material properties
and behavior with potentially real-time fabrication
feedback.
ACKNOWLEDGEMENTS
The Mediated Matter group at the MIT Media Lab
wishes to thank our network of colleagues and advisers. Specifically, we wish to thank Professor Carter,
Professor Gibson, and Professor Ortiz from the Materials Science and Engineering Department, as well as
Professor Wallace form the Mechanical Engineering
Departments for their insightful input and support.
We also with to thank our team of undergraduate
research assistants that have been involved in various stages of this work including Rachel Fong, Mindy
Eng, Tim Robertson, Cody Daniel. Special thanks to
Leah Nation, Lina Garcia,and Timothy Cooke for their
contributions on the concrete foam experiments.
REFERENCES
Benyus, J. M.(1997) Biomimicry: Innovation Inspired by
Nature, Quill. New York.
Fan, H., Y. Lu, et al. (2000). Rapid prototyping of patterned

functional nanostructures. Nature 405(6782): 5660.
Food and Agriculture Organization (1996). Coconut palm
stem processing: technical handbook, 1996, available online at: http://www.fao.org/docrep/009/ag335e/
AG335E02.htm
Freyman, T., I. Yannas, et al. (2001). Cellular materials
as porous scaffolds for tissue engineering. Progress in
Gibson, I. and D. Shi (1997). Material properties and fabrication parameters in selective laser sintering process.
Gibson, L. and M. Ashby (1982). The mechanics of threedimensional cellular materials. Proceedings of the Royal
Society of London. Series A, Mathematical and Physical
Sciences: 4359.
Gibson, L. and M. Ashby (1999). Cellular solids: structure
and properties, Cambridge Univ Pr.
Gibson, L. J. (1985). The mechanical behaviour of
cancellous bone. Journal of Biomechanics 18(5):
317328.
Gibson, L. J., M. F. Ashby, et al. (2010). Cellular materials
in nature and medicine, Cambridge Univ Pr.
Jacobs, P. (1992). Rapid prototyping & manufacturing:
fundamentals of stereolithography, Sme.
Jacobs, P. F. (1992). Rapid prototyping & manufacturing:
fundamentals of stereolithography, Sme.
Keaveny, T. M., R. E. Borchers, et al. (1993). Trabecular bone modulus and strength can depend on specimen
geometry. Journal of Biomechanics 26(8): 991995, 997,
9991000.
Markworth, A., K. Ramesh, et al. (1995). Modelling studies applied to functionally graded materials. Journal of
Miyamoto,Y. (1999). Functionally graded materials: design,
processing, and applications, Chapman & Hall.
Miyamoto, Y., W. Kaysser, et al. (1999). Functionally graded
materials: design, processing, and applications, Chapman
& Hall.
Neville, A. (1993). Biology of fibrous composites: development beyond the cell membrane, Cambridge Univ
Pr.
Ortiz, C. and M. C. Boyce (2008). Bioinspired structural
materials. Science 319(5866): 1053.
Oxman, N. (2007). Get Real Towards PerformanceDriven Computational Geometry. International Journal
of Architectural Computing 5(4): 663684.
Oxman, N. (2009). Material-based design computation:
Tiling behavior.
Oxman, N. (2010). Material-based design computation.
Oxman, N. (2010). Methods and Apparatus for Variable
Property Rapid Prototyping, Google Patents.
Oxman, N. (2010). Structuring Materiality: Design Fabrication of Heterogeneous Materials. Architectural Design
80(4): 7885.
Oxman, N. (2010). Structuring Materiality: Variable Property Fabrication of Heterogeneous Materials Architectural Design: THE NEW STRUCTURALISM: DESIGN,
ENGINEERING AND ARCHITECTURAL TECHNOLOGIES 80(4): 7885.
Oxman, N. (2011). Variable property rapid prototyping.
Virtual and Physical Prototyping 6(1): 331.
Rice, R. W. (1976). Microstructural Dependence of Mechanical Properties of Ceramics in RK McCrone. Treatise on
Materials Science and Technology. Vol 2. Academic Press,
NY.
Rich, P. M. (1987). Mechanical structure of the stem of
arborescent palms. Botanical Gazette 148(1): 4250.
488
Sachs, E., M. Cima, et al. (1993). Three-dimensional printing: the physics and implications of additive manufacturing. CIRP Annals-Manufacturing Technology 42(1):
257260.
Sarikaya, M. (1999). Biomimetics: materials fabrication
through biology. Proceedings of the National Academy
of Sciences of the United States of America 96(25): 14183.
Sheng, W., N. Xi, et al. (2003).Surface partitioning in automated CAD-guided tool planning for additive manufacturing, IEEE.
Tonyan, T. (1991) Mechanical Behavior of Cementitious
Foams. Cambridge, MA: MIT.
Vincent, J. F. V. (1982). Structural Biomaterials. London,
Macmillan.
Vincent, J. F. V., O. A. Bogatyreva, et al. (2006). Biomimetics: its practice and theory. Journal of the Royal Society
Interface 3(9): 471.
Vogel, S. (2003). Comparative biomechanics: lifes physical

world. Princeton, NJ, Princeton University Press.
Weiss, L., R. Merz, et al. (1997). Shape deposition manufacturing of heterogeneous structures. Journal of Manufacturing Systems 16(4): 239248.
Williamson, A.K., A.C. Chen, et al. (2003) Tensile mechanical properties of bovine articular cartilage: variations
with growth and relationships to collagen network
components. Journal of Orthopaedic Research.21(5):
872880.
Yang, S., K. F. Leong, et al. (2002). The design of scaffolds
for use in tissue engineering. Part II. Rapid prototyping
techniques. Tissue Engineering 8(1): 111.
Yeong, W. Y., C. K. Chua, et al. (2004). Rapid prototyping
in tissue engineering: challenges and potential. TRENDS
in Biotechnology 22(12): 643652.
489

Continuous contour printing versus layer by layer printing

in an Additive Manufacturing technology
G. Vallicrosa, J. Delgado, L. Seren & J. Ciurana
Grup dEnginyeria de Producte, Procs i Producci (GREP), Universitat de Girona, Spain
ABSTRACT: Additive Manufacturing (AM) technologies are changing how to manufacture workpieces.
Opposed to subtractive or conventional manufacturing, they join materials to make objects from 3D model
data, usually layer upon layer. This work compares conventional layer by layer approach to build objects with
a new path planning continuous methodology, on a thin walled model. The new methodology created better
finished figures, reduced significantly the calibration time and slightly the fabrication time, providing a reliable method to manufacture with AM technologies, though it has some limitations that need further research.
Nevertheless, it has some limitations that need further research.
1
INTRODUCTION
Additive Manufacturing (AM) is a relatively new

technology based on joining materials to make objects
from 3D model data, usually layer upon layer, as
opposed to subtractive manufacturing methodologies
like milling or turning. AM is also referred to as Additive Freeform Fabrication (AFF), Rapid Prototyping
(RP), Layered Manufacturing (LM) or 3D Printing
(3DP). Developed in middle 80s, AM became useful
for making prototypes, nowadays, with the improvements in control and quality made in the last years,
these additive techniques are also capable to make final
products [1, 2].
Within this context, some open hardware projects
have emerged, like Fab@Home and RepRap, allowing
particular users to print their own 3D models at a low
price [3]. Fab@Home is a project dedicated to develop
machines able to make almost anything, following a
layer by layer approach. Using syringes as tools, the
machine can print a wide range of liquid or extrudable
materials at room temperature, or any other materials
processed with heating or cooling equipments.
Several process parameters, involved in the Fab@
Home system, may influence the properties of the final
piece. Some of them are: deposition rate, layer height,
path width, pushout, suckback, and diameter of the tip.
Deposition rate is the millimeters plunger motion per
millimeters of path printed. Layer height is defined
as the distance between layers of deposited material.
Path width is considered the width of the stream of
material deposited. Pushout refers to the time for early
dispensing to start flow quickly. Finally, suckback is
the time of reverse plunger motion to stop the material
flow.
All of them are very important parameters, but difficult to define for each material, resulting in a waste
of material and time due to the calibration process.

Moreover, when fabricating slender figures or very
thin walls, the conventional layer by layer approach
is not enough. Typical algorithms, like the one used
by the Fab@Home software, create toolpaths from 3D
models. Firstly, defining the contours of the layer and
then, filling them. When defining contours, exterior
and interior (holes) contours are generated. This solution has some disadvantages when working with thin
walls. For example, for a cylinder with a wall thickness
similar to the path width, two contours are created, one
for the exterior part of the cylinder and another for the
holes, being the minimum wall thickness two times the
path width.
On the other hand, the problems calibrating deposition rate, suckback, and pushout parameters on the
Fab@Home system, are well known for their users.
If the deposition rate is set slightly high, a residual
pressure is created inside the syringe. This pressure
depends on the length of the path. When the suckback parameter is used to stop the material flow, it
can do it correctly, excessively or slightly. If the flow
is stopped excessively, the suckback used to restart
the flow will not be enough, generating a gap in the
piece. On the other hand, if the flow is stopped slightly,
some extra material will be deposited, generating a
drop on the piece. All these problems may increase
when building thin wall pieces, where a gap cannot be
filled with path overlapping or with the filling contour
toolpath.
Regarding the drawbacks associated with the
Fab@Home system, this work presents a new methodology, named continuous path planning, used for
printing contour objects with a Fab@Home machine,
while reducing the number of variables involved and
analyzing the manufacturing limits.
491
Table 2. The Fab@Home machine parameters.

Parameter
Value
path width
path height
deposition rate
pushout
suckback
path speed
0.4 mm
0.3 mm
0.00093
0.26 s
0.27 s
10 mm/s
Figure 1. 3D model proposed.

Table 1.
Experimentation parameters.
Parameter
Values
Unit
Wall angle
Start diameter
Total height
25, 35, 45, 55, 65, 75, 85

2, 6, 10
20
mm
mm
EXPERIMENTAL WORK
The experimental setup consisted of two parts: the

path methodologies and the manufacturing of the three
models using the above mentioned approaches.
2.1
Programming
All path planning algorithms were done in Python

(www.python.org) code and exported to a vector file
able to be imported by the Fab@Home software as
a path already planned. The vector file structure was
divided in lines. Each line contained a single point
with X, Y and Z positions separated by a space. When
a line was left empty the machine interpreted that the
current continuous path was finished and had to start
a new one.
The 3D model selected to compare different path
planning and fabrication limits was an inverse truncated cone (Figure 1) with a wall thickness of one
path.
Three process parameters were studied: the wall
angle, defined as the angle from the worktable to the
cone wall, the diameter of the circumference of the
first layer, and the total height of the piece, which is
constant (Table 1).
The 3D inverse truncated cone was modeled creating circumferences from points spaced 10 from
each other. This procedure created a thirty-six polygonal structure resembling a circumference at the scale
worked. For the modeled points, three different path
planning techniques were tested: non-continuous layer
by layer, continuous layer by layer, and continuous
spiral.
2.2 Manufacturing
The AM machine used to carry out this work was a
Fab@Home model 1 machine with a syringe deposition tool of 10 mL of capacity. The syringe tip inner
Figure 2. Non-continuous layer by layer path planning.
diameter used was 0.25 mm. And, the material utilized

to print the model was common acetic silicone (Fischer
Iberica, Cambrils, Spain).
The default process parameters provided by the
machines software, detailed in Table 2, were used to
evaluate the three path planning techniques.
The three path planning techniques were tested using

the Fab@Home machine and common acetic silicone.
The results obtained are described below.
The first path planning technique tested was the
non-continuous layer by layer approach. In order to
simulate this method, a random point was chosen as
the starting point, forcing the Fab@Home to stop the
flow at the point prior to the starting point, move to a
new randomized point and then restart the flow again
(Figure 2). This approach is broadly used by the users
of the Fab@Home machine due to it is the usual
planning technique provided by the manufacturers
[4]. Figure 3 shows the part produced following the
non-continuous layer by layer approach.
Secondly, the continuous layer by layer technique
was developed, within this work, with the aim of avoiding pauses between layers when the machine needs to
stop and find a new random starting point. Following this new approach the start and end points were
considered the same for each layer, therefore behaving as a continuous path (Figure 4). Figure 5 shows the
part produced following the continuous layer by layer
approach.
Finally, the continuous spiral planning developed in
this work was considered a connected path, as opposed
to the two mentioned above. The spiral path planning
492
Figure 3. Non-continuous layer by layer model (45

10 mm).
Figure 6. Continuous spiral path planning.
Figure 4. Continuous layer by layer path planning.
Figure 7. Continuous spiral model (55 10 mm)

Table 3.
Manufacturing limits.
Minimum angle fabricated
Figure 5. Continuous layer by layer model (55 10 mm).
was proposed as a methodology capable to divide the

layer height by the circumference perimeter, distributing the growth of the layer height between all points.
This technique, which raises height at every point, is
able to make a much continuous path (Figure 6). Figure 7 shows the part produced following continuous
spiral approach.
The manufacturing parameters proposed initially
were tested for each path planning. The deposition
rate, suckback time, and pushout time were found the
most difficult parameters to calibrate, incrementing
the calibration time expected.
In opposition to the non-continuous layer by layer
planning, the new continuous methods do not need to
stop and restart the flow flux for every new layer. As
a result, no stop/start flow variables, such as pushout
and suckback, are needed, simplifying the calibration
process. The pushout parameter can be substituted by
Diameter
2 mm
6 mm
10 mm
Layer by layer
Continuous
Spiral
45
35
35
45
35
35
35
35
35
a 1020 mm path at the start approaching of the continuous path to have a homogeneous printing filament.
For the suckback parameter, a high value was applied
in order to stop the material flow correctly. The appreciate control of both manufacturing parameters in the
continuous approaches gave a good surface finish in
all fabricated models. The surface finish of the two
3D continuous models was appreciably smoother than
the one following the non-continuous layer by layer
approach (Figures 3, 5 and 7). As Figure 3 shows,
the cone manufactured with the non-continuous layer
by layer method exhibits drops and holes within and
between layers.
Concerning the manufacturability of the three techniques, there were no remarkable differences between
them, producing similar results on the geometries studied. Continuous layer by layer and continuous spiral
were able to print until 35 , without problems, in all
starting diameters. The non-continuous layer by layer
approach only reached 35 with the 10 mm diameter circumference, for the rest, it arrived until 45
(Table 3).
493
the buckling variable can be considered an important

parameter to take into account when printing with a
Fab@Home machine.
Tables 4 and 5 show some comparative advantages and disadvantages, based on experiments carried out, between the continuous and non-continuous
techniques.
4
The two continuous methodologies produce smoother

3D models compared with the non-continuous
approach usually used. The calibration and manufacturing time slightly decrease due to the possibility of
leaving out some calibration parameters such as the
pushout time and suckback time, this way reducing
the nonproductive fabrication time. One of the main
advantages of the continuous methods is the possibility of creating models with a wall angle of 45 ,
allowing the manufacture of complex thin 3D geometries. Future work is required to improve the limitations
associated with the design process.
Figure 8. Buckling on fabrication

Table 4.
Non-continuous path planning.
Advantages
Disadvantages
One algorithm for

all 3D models
Can be applied to
all geometries
Requires no knowledge
More material
calibration time
Worse surface finish
CONCLUSIONS
More fabrication time

More wall thickness
ACKNOWLEDGEMENT
Table 5.
Continuous path planning.
Advantages
Disadvantages
Better surface finish
Has to be coded for

every model
Hard to apply to
complex geometries
Requires a programmer
Less material calibration time

Less fabrication time
Less wall thickness
Another manufacturing limitation observed during

the experimental procedure was buckling. Buckling is
the deformation of the printed figure while it is fabricated. This variable can affect negatively the final
geometry (Figure 8). Moreover, when printing thin
and tall cylinders, buckling becomes more evident,
limiting the maximum length printable. Therefore,
This work was carried out with the grant supports

from the Spanish Government (project IBE-RM:
PS-0200002009).
REFERENCES
[1] N. Hopkinson, R.J.M. Hague and P.M. Dickens,
Introduction to Rapid Manufacturing, in Rapid Manufacturing: An Industrial Revolution for the Digital Age,
John Wiley & Sons Ltd., Chichester (UK), 2006.
[2] D.T. Pham, R.S. Gault, A comparison of rapid prototyping technologies, Int. J. Mach. Tools Manuf. 38 (1998)
12571287.
[3] E. Malone and H. Lipson, Fab@Home: the personal
desktop fabricator kit, Rapid Prototyping Journal 13/4
(2007) 245255.
[4] http://www.fabathome.org
494

Methodology for analyzing the depth of sintering in the building platform

J. Delgado, L. Sereno & J. Ciurana
Department of Mechanical Engineering and Civil Construction, Universitat de Girona, Girona, Spain
L. Hernandez
ABSTRACT: Additive Manufacturing (AM) comprises group of relatively new technologies for the rapid
fabrications of prototypes or parts without the need of moulds or special tools. Since the launch of the first
commercial equipments further machines have been developed and improved, being able to manufacture even
functional parts with complex geometries. Selective laser melting (SLM) is one of the AM technologies and
consists on building metallic parts by successively scanning specific zones of a metallic-powder bed or layer,
with a laser beam to re-melt and solidify the powder. Several process parameters can influence the development
of the melting process during every single scan-track, which could affect the grabbing of the first layer to the
building platform, causing problems during the operation. Furthermore, inadequate melting and joint of adjacent
layers may lead undesirable physical and mechanical properties of the manufactured part. This work proposes a
methodology to experiment in a single building platform a wide range of combinations of the process parameters
and observe the effect over the melting process. A stainless steel plate was used as the building platform was
used to carry out the SLM experiments on a CoCrMo powder which is considered a biocompatible alloy, suitable
for medical applications. Full factorial experimentation was designed changing three process parameters: power
laser, scan speed, scan overlap, and layer thickness. The methodology proposed in this work reduces the number
of experiments needed to study and define the appropriate process conditions to perform SLM of metallic
powders.
1
INTRODUCTION
Additive Manufacturing (AM) technologies are relatively new and permit to obtain prototypes and final
parts from three-dimensional (3D) images. Selective
Laser Melting (SLM) is one of the AM techniques
and consists on using laser scanning to achieve full
melting and solidification at specific sections of a
metal-powder bed, to form a two-dimensional (2D)
layer. Further 2D layers are successively molten and
solidified over their previous layers, to finally form a
3D solid part.
There is a big range of metals and alloys that
could be processed by this technology, but the process parameters depend on the specific material and
the application of the final part. To define the value
of these parameters is the focus of current investigations. For instance, the wavelength of laser source has
a big influence in the absorption of laser energy by
the metallic powder and, therefore, in the efficiency of
the process [Tolochko et al. 2000]. Details of the most
important aspects of SLS and SLM can be found on
[Kruth et al. 2007].
A lot of process parameters influence the properties of the final part. This influence can be studied
by evaluating output parameters such as mechanical,
physical, metallurgical and dimensional properties.
By studying the effect of process parameters on the

characteristics of single tracks [Yadroitsev et al. 2010]
and single layers [Yadroitsev et al. 2011] produced
by SLM, information can be generated to contribute
building the knowledge to control the final properties of parts produced by SLM. Therefore, the aim
of this study is to determine a methodology to gather
as much information as possible from a single SLM
experiment.
Material composition has big influence on metallic
powders for SLM technology. Powders can be composed by two different materials (one of them being
the binder) or only one material. For the latter, the
selection of the working conditions is more critical
because the metal must be fully molten, being much
more complicated to control in comparison with combined materials where one of the metals is always in
solid state, being structurally more stable. The condition where single-metal powders can be processed
by SLM must be such that full melting is achieved
but avoiding excessive heating of the molten material, which leads to irregularities of the final part.
[Tolochko et al. 2004].
When irregularities occur, such as the balling
defect, the deposition uniform fresh powder bed during the SLM process becomes difficult. The balling
refers to the solidification of metallic balls along the
495
Table 1.
Design of experiments in each strategy
Strat.
Exp.
Parameter
Min.
Max.
Increments
80
36
P [W]
SS [mm/s]
LT [ m]
P [W]
SS [mm/s]
O [%]
LT [ m]
25
33.3
40
100
33.3
5
40
500
83.3
500
300
83.3
25
500
25
16.7
continuous
100
16.7
10
continuous
2.4 Strategies
Figure 1. Image of CoCrMo powder used for the
experimentation.
laser-scan path, instead of a continuous material track,

and may tend to cause porosity of the final part [Gu
et al. 2009].
2
2.1
Materials
Both strategies were carried out over building platforms which consisted on steel plates grade AISI 1045
with dimensions of 250 250 30 mm. The size of
the platform aimed to dissipate the heat during the
experiments and maintain the temperature. Moreover,
the height of the building avoids the deformation of
the plate do to thermal stresses.
Python was used as a Computer Aided Manufacturing (CAM) software for programing the trajectory
of the laser in G-Code, which is the traditional for
Computer Numerical Control (CNC).
A CoCrMo powder was used for the experimentation. The powder has a chemical composition of
Co-31.62 Cr-8.14Mo in weight percent, as semiquantitatively determined by EDX. The shape of
the powder particles is predominantly spherical with
diameters mostly between 20 to 50 microns (figure 1).
This alloy is biocompatible and has been the focus of
previous studies on SLM process for manufacturing
medical devices [Vanderbroucke et al. 2007].
2.4.1 First strategy

The first strategy consisted on 80 experiments over one
building platform, where each experiment refers to a
straight laser scan. The 80 experiments come from the
combination of 20 different laser-power values and 4
different scanning speeds. Each scan was performed
over a powder bed with variable thickness, from 40
to 500 microns along a distance of 230 mm. The aim
of each experiment was to form a single track of
re-molten metal, with length of 230 mm.
2.2
2.4.2 Second strategy

The second strategy consisted on 36 experiments over
one building platform, where each experiment refers to
laser scans following a zigzag trajectory, with a vector length of 5 mm. The 36 experiments come from
the combination of 3 different laser-power values, 4
scanning speeds, and 3 overlap percentages. Similarly
to strategy 1, each zigzag scan was performed over a
powder bed with variable thickness, from 40 to 500
microns along a distance of 230 mm. The aim of each
experiment was to form a single layer of re-molted
metal, with dimensions of 5 230 mm.
Equipment
The experiments were carried out on a Kondia HS1000 vertical milling machine, whereYW30 Prodintec
welding head has been mounted and aligned with the
vertical (Z) axis. The welding head has a focal length of
125 mm, and the minimum spot size produced with this
configuration is 150 m. The source of laser radiation
is a FL x50sYtterbium fiber laser by Rofin, that obtain
a maximum power of 500 W in continuous wave and
operate at a wave-length of 1080 nm.
2.3
Design of experiments
Two strategies were used in the experimental methodology, as detailed in the design of experiments of
Table 1. For each of the experiments different parameters were applied in order to find their effect over the
SLM process.
The DoE used a full factorial design, changing from
3 to 4 process parameters for strategy 1 and strategy 2
respectively. The parameters varied during the experiments were: laser power (P), laser scan speed (SS),
powder layer thickness (LT) and scan overlap (O).
2.5 Methodology of the inclined plane

The variable thickness of the powder bed was possible thanks to a slope produced over the building
platform by grinding. This allowed evaluating the continuity of the re-molten tracks and layers under the
different combinations of parameters, plus a variable
thickness of the powder-bed layer. The inclination of
the plane is 0.115 degrees and produces a powder bed
with variable thickness from 40 to 500 microns, as
shown schematically in figure 2.
496
Figure 2. Schematic of the building platform with inclined

plane.
Figure 4. Graphic representation of the different levels
of energy density (J/mm3 ) over the building platform of
strategy 2.
Figure 3. Graphic representation of the different levels

of energy density (J/mm3 ) over the building platform of
strategy 1.
2.6
Energy density
Each combination of the process parameters results in

a value of energy density (ED), which represents the
amount of energy supplied by the laser per square millimeter of metallic powder. The ED value is presumed
to be the main factor affecting the melting process and
is determined by equation 1 [Meier et al. 2008]:
where P is the laser power [W], SS is the laser scan

speed [mm/s], LT is the thickness of the powder bed
[mm], and HS is the hatch scan distance [mm] and it is
equal to the spot size multiplied by the complementary
of the overlap (1-O).
Figures 3 and 4 show a graphic representation of the
different levels of energy density over both building
platforms, due to the different combination of parameters. Both experiments cover a wide range of energy
density values. Additionally, different zones with the
same value of energy density corresponded to different
combinations of process parameters.
3
Two building platform were obtained, containing a

total 116 SLM experiments, in the form of single tracks
or single layers of re-molten metal. Figure 5 shows
building platform 1, with its 80 single-track experiments, while figure 6 shows building platform 2, with
36 single-layer experiments.
Every position along the tracks and layers has different features that can be evaluated, such as the shape,
Figure 5. Tracks of re-molten CoCrMo over the building

platform from strategy 1.
and can be related to a value of energy density and to

a specific combination of process parameters.
Based on the appearance of the tracks and layers, the
different zones can be divided in the following three
basic categories:
Continuous: the tracks and layers show a regular shape with continuity on its length (figures 7a
and 7b).
Discontinuous: the tracks and layers may show deviations from its longitudinal axe, interruptions of its
continuity, or even balling defect, which consists
on the formation of independent balls of re-molten
metal along the original trajectory of the laser
(figures 7c and 7d).
No sinterization: zones of complete absence of
re-molten metal, where the original surface of the
building platform can be observed, without much
alteration (figures 7e and 7f ).
It is important to clarify that the re-molten layers
from the building platform 2 must be analyzed only
along the central line, at the middle of the width. The
condition of the re-molten metal on both sides of the
layers width are significantly influenced by acceleration and deceleration of the lasers scanning, causing
defects such as balling (figure 7f ). The changes of the
scan speed are derived from the nature of the CNC
equipment used for this study.
497
plane the amount of information generated is further

increased. Another advantage of the proposed methodology is the save of metallic powder due to the use of
only a single powder bed for the production of the
whole set of experiments.
The results from the proposed methodology has
the potential to provide information about features of
the re-molten tracks and layers, such as morphology,
transverse profile, microstructure, etc., which might
be further analyzed and related to their corresponding
process parameters (figure 8). Such kind of analysis could be used for the definition of process maps
focused on different requirements such as roughness,
mechanical properties, building speed, etc.
Among the work that is needed to improve the presented results, it is important to highlight the need
of performing the experiments under a controlled
inhert atmosphere, since it is presumed that the energy
released during the oxidation of the metal, due to the
presence of oxygen in the air, influences the melting
phenomenon and might lead to imprecise conclusions.
However, the proposed methodology has the potential to be applied to other metallic powders in order
to look for the process maps which aid to gain more
control of the final properties of parts produced by
SLM.
Figure 6. Banded layers of re-molten CoCrMo over the

building platform from strategy 2.
ACKNOWLEDGEMENT
This work was carried out with the grant supports
from the University of Girona BR09/04 and the
Spanish Government (project DPI200909852).
REFERENCES
Figure 7. Representative zones of the three basic categories

for the shape of the tracks and layers: continuous (ab),
discontinuous (cd), and no-sinterization (ef).
Figure 8. Schematic of the correlation between input and

output variable of the SLM process.
CONCLUSIONS AND FURTHER WORK
The methodology presented in this study allows the

generation of a large amount of experiments per building platform, and thanks to the use of an inclined
Gu D., Shen Y., 2009, Balling phenomena in direct laser

sintering of stainless steel powder: Metallurgical mechanisms and control methods. Materials and Design,
Number, Number 30 pp. 29032910.
Kruth, J.P., Levyt, G., Klocke, F. and Childs, T.H.C., 2006,
Consolidation phenomena in laser and powder-bed based
layer manufacturing. CIRP Annals Manufacturing Technology, Vol 56, Number 2, pp. 730759.
Meier H., Haberland Ch., Experimental studies on selective laser melting of metallic parts. Material science and
engineering technology, Vol 39, Number 9, pp. 665670.
Tolochko 2000 Absorptance of powder materials suitable for
laser sintering. Rapid Prototyping Journal, Vol 6, Number
3, pp. 155160.
Tolochko NK., Mozzharov SE., Yadroitsev IA., Laoui T,
Froyen L., Titov VI., Ignatiev MB., 2004, Balling processes during selective laser treatment of powders. Rapid
Prototyping Journal, Vol 10, Number 2, pp. 7887.
Vandenbroucke, B., Kruth JP., 2007, Selective laser melting
of biocompatible metals for rapid manufacturing of medical parts. Rapid Prototyping Journal, Vol 13, Number 4,
pp. 196203.
Yadroitsev 2010. Single track formation in selective laser
melting of metal powders. Journal of Materials Processing
Technology, Vol 210, Number 12, pp. 16241631.
Yadroitsev 2011. Surface Morphology in Selective Laser
Melting of Metal Powders. Physics Procedia, Vol 12,
Number 1, pp. 264270.
498

Influence of process parameters in the first melting layer of a building

platform in a SLM machine
J. Delgado, L. Sereno & J. Ciurana
Department of Mechanical Engineering and Civil Construction, Universitat de Girona, Girona, Spain
L. Hernandez
ABSTRACT: Selective Laser Melting (SLM) is an Additive Manufacturing (AM) technology that allows the
transformation of metallic powder into solid parts, without the need of molds or special tools. Usually, commercial
SLM machines are configured to use a specific combination of parameters for the processing of each metallic
powder. However, no information is provided about different strategies that could be used in order to vary
the output characteristics of the manufactured parts, such as mechanical properties, in order to meet different
requirements. In this work, Kondia vertical milling machine equipped with a Rofin laser has been used to carry
out SLM experiments on a CoCrMo powder to form single tracks, to study the influence of laser power, scan
speed and layer thickness over morphology features of single track of remelted CoCrMo alloy. It was found that
the relation between the maximum layer thickness that could be used and still form continuous tracks, and the
specific values of scan speed and laser power followed a linear behavior. The results obtained in this work form
part of the knowledge needed to reach a better control of the SLM process on CoCrMo, which can be used for
the production of biomedical applications.
INTRODUCTION
Additive Manufacturing (AM) is one of the Rapid

Manufacturing (RM) technologies mostly developed in recent years. AM includes stereolithography
(SLA), Layer Object Manufacturing (LOM), Fused
Deposition Modeling (FDM), 3-Dimentional Printing
(3DP), Selective Laser Sintering (SLS), and Selective
Laser Melting (SLM), amongst others.
The latter two consists on melting and solidifying
specific sections of a metal-powder layer by using a
laser, and repeating the process on further deposited
upper-layers. More details about these processes can
be found at Kruth et al. [2007]. The main difference between SLS and SLM is the amount of energy
supplied to the metal powder, which depends on
their different melting and consolidation mechanisms
[Kruth et al. 2005]. In the first one the melting occurs
on the outer layer of the powder particles in such a
way that the melted material of adjacent particles flow
and get mixed and joining the particles after solidification. A post-process is generally needed to increase
the density of the produced parts.
For SLM the powder particles must melt completely
to reach higher densities after solidification. SLM
technology is suitable to rapidly manufacture complex, unique, or low-volume metallic parts, without
the need of moulds or special tools. SLS and SLM

machines have been developed using low wavelength
lasers which increases the absorption of energy by the
metal powder [Tolochko et al. 2000]. The amount of
absorbed energy is determined by different process
parameters and influences the melting process, and the
stability of the molten metal, affecting, in turn, the grip
of the remelted powder to the substrate or under-layer.
The adequate combination of the process parameters
in SLM is critical to ensure consolidation of all the layers into a single piece [Yadroitsev 2010], and acquire
the desired properties.
SLM machines have been developed and studies
have been carried out to process different metals,
including biocompatible ones such as CoCr alloys
[Vanderbroucke 2007]. However, not much is openly
known about the influence of varying input process
parameters over output features. By gaining more
knowledge about this influence, different physical or
mechanical requirements could be obtained in parts
produced by SLM.
The influence of process parameters on physical
features of single tracks of remelted CoCrMo is studied in this work.A relatively wide range of values of the
process parameters could be studied by experimenting
on only one SLM building platform, thanks to the use
of a powder bed with variable thickness.
499
Figure 1. Schematic of the base plate for the SLM experiments.

Table 1.
Design of experiments.
Experiments
Parameter
Min
Max
Interval
80
P [W]
SS[mm/s]
LT [ m]
25
33.3
40
500
83.3
500
25
16.7
Continuous
2
2.1
Materials and SLM equipment
The processed material is a CoCrMo powder whose

particles have spherical shape with diameters
between 20 and 50 m. Its chemical composition
es Co-31.62 Cr-8.14Mo in weight percent, as semiquantitatively determined by EDX. The building platform, which is the substrate where the initial layer is
deposited on, consists on a plate made of steel AISI
1045 with 30 mm thickness.
The experiments were carried out on Kondia HS1000 vertical milling machine equipped with a laser
source mounted and aligned to the vertical axe of
the machine. The laser is a continues wave Nd:YAG
type with maximum power of 500 W, which operates at a wavelength of 1080 nm. The welding head is
equipped with a focal length and collimator of 125 mm
producing a minimum spot size of 150 m.
2.2
Figure 2. Building platform with 80 experiments divided in

4 groups.
Methodology of inclined plane
The experimental methodology used in this study

allowed evaluating the morphology of single straight
tracks of remelted CoCrMo, produced by a large
amount of combinations of SLM process parameters,
including the continuous variation of the layer thickness, from 40 to 500 m. It was possible thanks to the
use of a building platform, made of steel grade AISI
1045, with an inclined plane as schematically shown in
figure 1. All the experiments were carried out in only
one building platform. The summary of the design of
experiments (DOE) is shown in table 1.
2.3 Analysis of track morphology
The experiments were visually analyzed and the limit
of track continuity was determined. Additionally the
information from the experiments was statistically
analyzed in order to look for the relationships between
input and output variables. The evaluation was carried out by 5 persons to look for reproducibility of the
results.
The result of the SLM experiments is a building

platform containing 80 experiments produced under
different combinations of the main process parameters. Each experiment consists on a single track with
230 mm in length, where the powder layer thickness
was varied continuously from 40 to 500 m. Figure 2
shows the building platform with 4 families or groups
of experiments. Each family corresponds to a different scan speed value, and within each family the
straight tracks were produced under different laser
power values from 25 to 500 W.
3.1 Visual inspection
Figure 3a shows an example of what was categorized as a continues track, while figures 3b and 3c
show examples of discontinuity, categorized as irregular tracks and balling effect, respectively. There were
considered as continuous zones, those showing continuous width of the track, while at the discontinuous
zones the tracks showed irregular shape with variation
of the width, and even interruption of the track length.
Also, within the discontinuous classification, are the
zones where balling effect was present. The balling
effect consists on the formation of independent balls
of remelted metal along the track direction [Tolochko
et al. 2004].
3.2 Limit of continuity
Every track was evaluated visually and the length
where track was considered as continuous was measured. Those experiments that did not show track
formation, or the track exceeded limit of the building
platform, were not considered. Figure 4 shows one of
the four families (20 experiments) marking limit of the
continuity zone, while figure 5 shows the continuitylimit curves of the four families. Figure 5 shows that
as the scan speed was increased the continuity zone
resulted smaller.
500
where P is the laser power [W], SS is the laser scan

speed [mm/s], LT is the thickness of the powder layer
[mm], and HS is the hatch scan distance [mm], which,
in turn, is equal to the spot size multiplied by the complementary of the overlap (1-O). For the case of single
track, the overlap value (O) is 0 (no adjacent tracks
were produced).
This ED value can be plotted to see the distribution
of its different levels on the surface of the building
platform and observe the ED level where continuous
tracks are formed. Figures 6a to 6d show the mentioned
ED distribution, with the curves of track continuity
(figure 5) imposed over.
Part of the continuity-limit curves seems to fall
around an ED value of 150 J/mm3 . The value of ED
corresponding to the maximum layer thickness where
continuous tracks could be formed (LTmax ) at each
experiment, was determined and defined as EDmin .
When averaging the values of EDmin at every scan
speed family, after eliminating those values out of
a 3 standard deviations, it was found that the four
values fall between 153 and 158 J/mm3 . It is assumed
that around 155 J/mm3 is the minimum value of ED
which allows the formation of continuous tracks for the
SLM of CoCrMo.
Figure 3. Basic categories of track condition: (a) continuous, (b) irregular, and (c) balling defect.
3.4
The full factorial design of experiments allowed analyzing the effect of input process variables over the
length of the continuous tracks, and more specifically
over LTmax .
The length of continuous track was translated into
a value of layer thickness, according to the following
equation 2:
Figure 4. Example of continuity limit for a family of

experiments (SS = 66.7 mm/s).
LT(x) = LTmin + x tg()
Figure 5. Continuity zones for the SS families of

experiments.
3.3
Statistical analysis
Energy density contours
Every point on the main surface of the building platform can be translated into a value of energy density
(ED), which represents the amount of energy supplied
by the laser per cubic millimeter of metallic powder.
ED is determined by equation 1 [Meier et al. 2008]:
P
ED =
(1)
SS LT HS
where LT(x) is the layer thickness at the position x,

LTmin is equal to 40 m and refers to the layer thickness
at position 0 mm, x is the position along the scan vector
(from 0 to 230 mm), and is the inclination angle of
the building platform.
The data was validated by a normality test and then
an Analysis of Variance (ANOVA) was carried out to
plot the main effect of the input parameters (SS and P)
over LTmax . Figure 7 shows the main effect of SS and
P over the LTmax .
Then, the full factorial DOE was anlysed to find out
if the effect of the input parameters and their combination over LTmax was significant, and determine the
equation that relates them. According to the P-value
of the equations constants, the effect of P, SS and
their combination over LTmax is significant with a 95%
confidence level. The derived equation for the range
of parameter values used in this study is as follows
(eq. 3):
501
LTmax = 0.0017 P 0.0031 SS

8.883 106 P SS + 0.0871
(2)
Figure 7. Main effect plot of SS and P over LTmax .
The minimum energy density where continuous

tracks could be formed at each experiment was defined
as EDmin . It is assumed that around 155 J/mm3 is the
EDmin value for the SLM of CoCrMo.
It was found, by means of statistical analysis, that
the laser power and scan speed had a significant effect
over the value of LTmax , following a lineal behavior.
ACKNOWLEDGEMENT
This work was carried out with the grant supports from
the University of Girona BR09/04 and the Spanish
Government (project DPI200909852).
REFERENCES
Figure 6. Track-continuity limit for the four families of

experiments, over imposed on contour maps with the different levels of energy density: (a) 83.3 mm/s, (b) 66.6 mm/s,
(c) 50 mm/s, and (d) 33.3 mm/s.
CONCLUSIONS
A lean methodology was employed to produce a single building platform with 80 SLM experiments
performed under different process parameters. This
methodology includes the use of an inclined plane
to continuously varying the powder layer thickness at
each experiment.
Kruth J.P., Mercelis P. Van Vaerenbergh J. Froyen L., Rombouts M., 2005 Binding mechanisms in selective laser
sintering and selective laser melting. Rapid Prototyping
Journal, Vol 11, Number 1, pp. 2636.
Kruth J.P., Levyt G., Klocke F., Childs T.H.C., 2007, Consolidation phenomena in laser and powder-bed based
layer smanufacturing. CIRP Annals Manufacturing
Technology, Vol 56, Number 2, pp. 730759.
Meier H., Haberland Ch., 2008, Experimental studies on
selective laser melting of metallic parts. Material science and engineering technology, Vol 39, Number 9,
pp. 665670.
Tolochko NK., Laoui T., Khlopkov Y., Mozzharov SE., Titov
VI., Ignatiev MB., 2000, Absorptance of powder materials suitable for laser sintering. Rapid Prototyping Journal,
Vol 6, Number 3, pp. 155160.
Tolochko NK., Mozzharov SE., Yadroitsev IA., Laoui T,
Froyen L., Titov VI., Ignatiev MB., 2004, Balling processes during selective laser treatment of powders. Rapid
Prototyping Journal, Vol 10, Number 2, pp. 7887.
Vandenbroucke, B., Kruth JP., 2007, Selective laser melting of biocompatible metals for rapid manufacturing of
medical parts. Rapid Prototyping Journal, Vol 13, Number
4, pp. 196203.
Yadroitsev I., Gusarova A., Yadroitsava I., Smurova I., 2010
Single track formation in selective laser melting of metal
powders. Journal of Materials Processing Technology,
Vol 210, Number 12, pp. 16241631.
502

Surface finishing techniques for SLM manufactured stainless steel

316L components
W. Morton, S. Green, A.E.W. Rennie & T.N. Abram
Engineering Department, Lancaster University, Lancaster, United Kingdom
ABSTRACT: Selective Laser Melting (SLM) is an additive manufacturing (AM) process which builds components through melting layers of powdered material together.The surface finish on these manufactured components
typically requires additional manual finishing or post-processing. Traditionally, mechanical polishing techniques
utilise semi skilled labour and usually requires significant time input resulting in it being an expensive and inefficient process. This paper considers less labour intensive post processing techniques such as media blasting and
tumbling for SLM produced stainless steel 316L components. In particular, the change in the surface profile due
to the finishing processes employed was the primary issue for investigation. However, changes in the dimensions,
geometric features and overall form were also studied. Parameters of the build process (machine operation) that
ultimately impact on the final components surface finish or finishing process required, were noted and discussed.
It was found that a range of surface roughness and surface finishes can be achieved using the techniques outlined
above. It is possible to achieve a good surface finish using some of these techniques, although, it is obvious that
these techniques are not solely sufficient for the post processing of stainless steel 316L components.
1
INTRODUCTION
Using this SLM, very complex components can be created that would not be possible by conventional means.
However, due to the inherent nature of the process
of fabrication, the surface finish of these components
can require additional processing, particularly where
the removal of support structure is necessary. This is
a common problem with AM technologies and much
research has been carried out to try and reduce the post
processing that components require (Kruth et al, 2005;
Mellor et al, 2010; Ahn et al, 2007).
The machine that was utilised for this study
was a SLM-100 manufactured by Realizer Gmbh,
Germany. This machine is typically used for the
creation of smaller components having a build envelope of 123 mm diameter (platen) by 80 mm height
(Realizer Gmbh, 2010).As these components are small
(and sometimes complex in geometry), achieving the
desired surface finish through post processing, without
losing feature definition, can be challenging.
Traditionally, in order to achieve a good surface
finish, mechanical abrasion techniques are necessary.
These techniques usually need to be carried out manually by a semi skilled tradesmen resulting in a very time
consuming and expensive process (Groover, 2002). In
the manufacturing environment, this is problematic
since it takes time, resources and ultimately increases
production costs. Another problem is that manual processes usually result in variations between individual
components that have been manufactured as part of
a batch, hence repeatability is compromised. Santos
et al (2010) concluded that it is not possible to achieve
a good surface finish whilst retaining feature geometry and ensuring that excessive surface material is not
removed using traditional abrasive techniques.
Newer polishing techniques (such as laser polishing) have been investigated. Dadbaksh et al (2010)
found that surface roughness could be reduced to less
than 2 m without excessive material removal. This
was a decrease in roughness of over 80%, although,
this was carried out on single planes with no geometric
features.
Although there has been research into SLM technology, the majority of the research analyses the
process settings and the capabilities of the technology
itself, and not the post processing techniques usually
required. It seems that SLM technology does have the
capability to create complex components, however,
the accuracy and surface finish of these components
is usually substandard for function. Particular build
issues such as the stair step effect, curling, delamination and warping are the main factors that affect feature
definition on components. Kruth et al (2005) showed
the possibilities and limitations of SLM/SLS manufacturing and utilised a well designed sample component.
The design of the component incorporated many features that the SLM manufacturing process typically
struggles to build.
2
A series of experiments were established, separated

by their process type: media blasting and tumbling. In
order to generate sufficient data for analysis, it was
503
Figure 1. Sample component design with key features

highlighted.
determined that a minimum of ten SLM components

would be required to experiment with a adequate number of blast and tumble medias. A component was
designed in order to provide a sample piece for the
various post processing experiments.
This design incorporated many features all of which
had been purposely chosen and located to test both the
capability of the SLM build process and the effects
of the post processing technique. The dimensions of
the component measured approximately 47 mm
27 mm 12 mm (L W H). In order to reduce
weight, material usage, build time and ultimately, build
cost, the component was designed as a shell with a wall
thickness of 3 mm (Figure 1).
With reference to Figure 1, two key heights have
been integrated (items A & C) to investigate the material loss due to post processing. Chamfers and radii
were placed on the outside corners of the component to
investigate the effect of post-processing on both small
and large features. These small and large features are
marked as items E & G and D & F respectively. The
compound curved slope that can be seen in the centre
of the component was added to explore the common
AM problem known as the stair-step effect. This slope
is made up of a linearly extruded cut triangle which is
perpendicular to the dimension H. Markings and numbers on the top face of the sample component then refer
to the gradient of the slope at that point. These markings allow an approximate value to be read off where
the step effect begins to takes place. In the top left
corner of the upper face, the digits refer to the component identification number. Sections marked I & J in
Figure 1 are the sections of the component where
surface measurements were taken.
A similar methodology was utilised throughout all
experiments: the components were fully inspected
prior to any processing, were then used for testing and
were finally inspected again. This inspection included
measuring the key dimensions, photographing the key
geometric features, measuring various features of the
surface and recording all points of interest. Since tumbling experiments ran over many hours, inspections
were also carried out at intervals during processing in

order to record any changes varying with time.
Linear dimensions (lengths, angles and radii) were
measured using a standard set of Vernier callipers, a
Vernier protractor and Radii gauges. In order to measure surface parameters, a Mitutoyo SJ-400 surface
roughness tester was used. The SJ-400 is a stylus type
profilometer capable of measuring surface roughness
among several other values such as waviness and primary profiles. The following are key settings that were
used unless otherwise stated: Standard ISO1997;
Filter GAUSS; cut off length 0.8 mm; and N 5.
The media blasting experiments were conducted
at Guyson International (based in Skipton, North
Yorkshire, England) to utilise the various grades of
media on offer. Through guidance from Guyson, the
blast media that were chosen for experimentation were
soda lime glass bead and 95% virgin aluminium oxide
of two grades each (Guyson. 2011). The tumbling
experiments were all conducted in-house (Lancaster
University, Lancashire, England) and all blast media
types were bought in for experimentation. Through
research and availability, the tumbling medias that
were chosen for experimentation were ceramic shapes
(triangular), stainless steel shot, grade 30 aluminium
oxide, dry walnut shell and wet corn maize (UKGE,
2011).
RESULTS AND DISCUSION
As previously noted, the experiments were separated

by their process type: media blasting and tumbling.
Table 1 summarises the components, the component
build orientation, which post-processing operation
was used and details regarding the media. Unless
otherwise stated, media blasted components were
subjected to blasting by hand for approximately
2 minutes.
3.1 Blasting experiments

The blasting experiments at Guyson yielded several
interesting observations. The first was that whilst
blasting component 2A with grade 10 glass bead, the
surface layer of the component began to delaminate.
This may be due to the blasting process causing a
peening effect on the surface. Shot peening locally
plastically works the surface resulting in compressive
stresses in the surface material and tension stresses in
the inner material (Callister, 2007). One reason why
the surface delaminated in the way it did, is due to
the SLM manufacturing process itself. When building
a component, SLM starts a new layer by melting the
inner and outer edges of the component. Next, material is melted between these inner and outer edges thus
creating a solid structure. The layer that delaminated
from the component is the outer layer of the component. The fact that this outer layer has delaminated
from the rest of the component suggests that it was not
504
again using the glass bead which, although gave the

surface a shinier finish, showed most of the surface
faults.
Due to the resolution of the linear measuring equipments, the measurement results showed no significant
changes. However, the surface measurements yielded
some results to which conclusions could be drawn.
Figure 4 shows how the change in the arithmetic mean
deviation of the profile (Ra) changes due to the processing time.The value Ra was chosen since it provides
an average surface roughness and this allows a comparison to be made between components. Figure 4 shows
that component 1A and 2A resulted in a reduction in
Ra. Component 1A was only blasted with grade 13
glass bead (over sections I & J) whereas component 2A
was briefly (approximately 30 seconds) blasted with
grade 20 aluminium oxide and then grade 10 glass
bead. Interestingly, component 1B was blasted with
grade 36 aluminium oxide and then grade 13 glass
bead but did not seem to show the same Ra reductions as 1A & 2A. This implies that the grade of blast
media and the blasting time have a significant effect on
the surface profile. Another possibility, as suggested
by component 1B, section J (premeasurement), is
that a surface anomaly was measured resulting in
ambiguous data.
Figure 2. Surface delamination due to media blasting.
3.2 Tumbling experiments
Figure 3. Surface delamination following media blasting.
sufficiently bonded to the rest of the inner material.

This delamination can be seen in Figures 2 & 3.
Another interesting observation was that whilst
blasting component 2B with grade 20 aluminium
oxide, the support structure from the SLM fabrication
process could be removed purely using the blasting process. This is an interesting observation since
removing the support structure by hand is typically a
time consuming, labour intensive activity which could
result in damage of the component. Using the blasting process, the support structure was removed in a
matter of minutes. However, the component suffered
from significant feature loss due to the intensity of the
blasting process.
The final notable observation was how the blasting process could be used to disguise the surface
faults on components. By blasting a component with
the aluminium oxide, it was possible to give the components surface a more consistent and even finish. All
of the surface imperfections, such as the evidence of
porosity and build imperfections consistent with the
SLM process, were no longer observed. However, to
prove that these faults had not been removed in their
entirety by the blasting process, the surface was blasted
Component 2B (ceramic shapes) was tumbled for

a total of 24 hours in which measurements were
taken at 2, 4, 6, 12, 18 and 24 hours. During this
24 hour period, the component showed little visual
surface change which was backed up by roughness
results.
Component 2D (stainless steel shot) on the other
hand, showed significant surface change which was
evident from the surface roughness data. In addition
to the 24 hours tumbling, component 2D was tumbled for an additional 56 hours (a total time of 80
hours). This was undertaken to determine whether
extended periods of processing time had any significant effect on the surface profile. This turned out not
to be the case, with the most significant surface change
happening in the first two hours of tumbling. For
example, the root-mean-square deviation of the profile
(Rq) showed approximately a 25% reduction after two
hours as opposed to a 36% over the total eighty hour
period.
Components 3A & 3B surface measurement results
did not show any particular surface change, however,
there was a clear change in the cosmetic appearance of
the components. Figures 5, 6 & 7 show the change in
the colour of the surface which is caused by the media
itself remaining on the surface.
Similarly, component 3C (wet corn maize) showed
no notable changes in surface measurement, but
showed a significant change in the brightness of the
component. This is simply due to the Raytec polishing agent which is present in the corn maize.
This brighter surface can be seen in Figure 8.
505
Table 1.
Components properties and related post processing data.
Component ID
Build Orientation
Process
Media & General Comments
1A
Horizontal
Shot Blasted
1B
Horizontal
Shot Blasted
2A
Horizontal
Shot Blasted
2B
Horizontal
Tumbled
2C
Vertical
Shot Blasted
2D
Vertical
Tumbled
3A
Horizontal
Tumbled
3B
Horizontal
Tumbled
3C
Horizontal
Tumbled
3E
Horizontal
None
Half shot with grade 36 Aluminium Oxide and the other half shot
with grade 13 glass bead for comparison. Sections I and J (where
surface measurements were taken) were both in glass bead blasted
section.
Shot with grade 36 Aluminium Oxide THEN shot with grade 13
glass bead.
Shot with grade 20 Aluminium Oxide (approximately 30s) THEN
shot with grade 10 glass bead.
Ceramic triangular shapes. Support structure was not removed
prior to tumbling. Measurements were taken at 2, 4, 6, 12, 18 and
24 hours.
Heavily shot with grade 20 Aluminium Oxide. Support structure
was removed by the shot blasting alone.
Stainless steel shot. Support structure was not removed prior to
tumbling. Measurements were taken at 2, 4, 6, 12, 18, 24 and
80 hours.
Tumbled with grade 30 Aluminium Oxide. Support structure was
not removed prior to tumbling. Single 18 hour run.
Tumbled with dry walnut shell. Support structure was not removed
prior to tumbling. Single 18 hour run.
Tumbled with wet corn maize. Support structure was not removed
prior to tumbling.
This component was left for comparison.
Figure 4. Surface measurement results from blasting experiments grouped by component and process stage (pre/post).
4
4.1
CONCLUSION AND FURTHER WORK

General observations
Prior to any post processing, there were several interesting points that were noted following the SLM
component fabrication process. Firstly, several components showed signs of porosity throughout their
structure. In order to observe this more clearly, the
top face of component 1B was linished. This can be
seen in Figure 9.
Obviously, this porosity will have a considerable
effect on the achievable surface finish of SLM manufactured components. If precise or highly polished
surfaces are required, then the SLM technology needs
to be improved in order to reduce the resulting

porosity or a further post process will always be
required.
Another interesting build imperfection that was
present in all components were indented lines in several of the side surfaces (for example, as shown in
Figure 10). These faults seem to line up with the inner
shell thickness of the components. It is thought that
this imperfection is a wiper blade related issue, however, its exact cause is unknown but several points can
be noted about it:
1. It is present in both horizontally and vertically built
components suggesting that it is not a fault with
506
Figure 5. Comparison
component 3E.
Figure 6. Component 3A
(Note darker surface).
Figure 10. Build imperfection on the side wall of all
components.
Figure 7. Component 3B
(Note slightly brighter
surface).
material. Another possible cause for the difference in

surface finish is the presence of powdered material
around the side of a component. This could act as a
barrier keeping the molten material in place, whereas
on the top surface, material can escape due to the sputtering nature of the material melting (XYZ Innovation,
2011).
Finally, the stair-step effect was observed although
it was not particularly detrimental to the form or
appearance of the component. In the majority of the
components built for this study, the stair-step effect
was more noticeable at plane angles approximately
>30 .
Figure 8. Component 3 C
(Note bright surface).
4.2 Conclusions
Figure 9. Presence of porosity in component 1B.
one particular section of the sample component

design.
2. Components 1A and 1B were in fact built 3 times
larger than other components. These larger components still showed this indented fault implying that
the fault is not a resultant effect of the components
physical size.
3. The fault is not random since it appeared in the
same locations on all the components built for this
study.
One clear observation is the fact that the vertical build planes (relative to the SLM build platform)
tended to have a better surface finish than that of the top
surfaces. It is thought that this observation is related
to the delaminated component (noted earlier in this
paper) in that it has resulted from the SLM manufacturing process itself. The surface finish results suggest
that the separately produced outer edges and inner
material of the component have different surface finishes. Simply put, the outer edge material produces
a smoother finish than that of the inner (hatched)
The aim of this study was to assess the viability

of media blasting or tumbling as a post-processing
technique for SLM manufactured components.
It has been observed that media blasting had the
unexpected possibility of being a viable way to remove
the support structure from the component.
During tumbling experiments, it was noted that the
stainless steel shot had a significant effect on the
surface of component 2D reducing the roughness by
approximately 25% in the first two hours. Also during
tumbling, it was seen how wet corn maize resulted in a
polishing effect thus making the component brighter.
However, these positive observations also have
drawbacks. In the case of component 2B, although the
support structure was removed, the components features were significantly impaired. Although tumbling
using stainless steel shot showed significant improvement to the surface roughness, little was done to the
support structure or the rougher sections of the component. Similarly, the wet corn maize made significant
improvements to the brightness of the component but
made no change to the surface roughness or support
structure. For these reasons it can be concluded that
media blasting and tumbling as post-processing techniques for SLM manufactured stainless steel 316L
components are not viable solutions on their own.
However, some observations are noteworthy and
could be used in conjunction with other postprocessing techniques successfully. For example,
following support structure removal, blasting with aluminium oxide has the capability to disguise surface
507
Table 2.
Summary of advantageous blast and tumble media.
Process
Media
Benefits
Blasting
Aluminium Ability to disguise surface

oxide
flaws
Possibility to remove support
structure
Provides consistent surface
finish
Gives matt finish
Tumbling
Glass
bead
Generally brightens surface
Stainless
steel shot
Brightens surface
Can reduce surface roughness

Can reduce surface roughness
Removes small imperfections
Wet corn
maize
Ability to clean/polish surface
imperfections if the aesthetic/cosmetic appearance of

the component is the important factor. Table 2 summarises the positives of each media type from both
blasting and tumbling processes.
In addition to the observations made on the postprocessing techniques, several further observations
were made regarding the SLM manufacturing process itself. From these observations, several practical
design rules can be stated (in order of importance):
Orientate the component on the build platform in

order to minimise support structure or minimise
support structure through good use of software such
as Magics. This reduces post processing time and
labour.
On the sections of the component that will come in
contact with support structure, add additional material (>1 mm) to allow the removal of the support
without affecting the component itself.
Orientate the component so the support structure
joins an outside surface rather than (for example)
an internal shell. This makes the removal and post
processing of the component simpler, quicker and
generally achieves a better finish.
If there is a particular surface that requires a good
surface finish, then the component should be orientated to make this surface as close to vertical (or
90 to the build platform) as possible.
Avoid thick sections of material by utilising shell
features or otherwise. This minimises the possibility
of curling, delamination and warping.
4.3
Future work
The primary cause of distortion to the surface of SLM

manufactured components is the removal of support
structures. Throughout the majority of this study, the

support structure has been left on to see the effect of the
post-processing on the support structure itself. In all
cases the post-processing had no effect on the support
structure with the exception of component 2C (blasting with grade 20 aluminium oxide). However, this
process had a destructive effect on the feature definition. Further research could be conducted in order
to find possible ways to remove the support structure
by media blasting without a detrimental effect to the
component itself.
Similar to work carried out by Ahn et al (2009);
using the results from this study, an approximate linear equation can be developed to predict the surface
roughness (Ra) based on a plane angle from the SLM
build platform:
where R = arithmetic mean deviation of the profile,

and = surface plane angle relevant to the build
platform ( = 0). Further research would need to be
conducted to verify this equation and possibly develop
a more complex, but more accurate, model.
It has been shown that although some improvements
can be made to the surface finish of SLM manufactured components, tumbling or media blasting are not
suitable solutions on their own. However, this study has
only considered components manufactured in stainless
steel 316L. Additional research should be conducted
in order to determine whether these post-processing
techniques are suitable for other, particularly softer,
materials. Further work could also be expanded to
other AM processes.
As a result of the delamination of component 2A
and the way in which the surface broke away from the
main body of the component, further research should
be conducted to investigate the strength (or surface
strength) of SLM fabricated components.
Finally, the build imperfection that appeared on the
side of all components built in this study is a curiosity
which cannot be definitively explained without further
research.
REFERENCES
Ahn, D. Kim, H. & Lee, S. 2007. Fabrication direction optimization to minimize post-machining in layered
manufacturing, International Journal of Machine Tools
and Manufacture, 47 (34), pp. 593606.
Ahn, D. Kim, H. & Lee, S. 2009. Surface roughness
prediction using measured data and interpolation in layered manufacturing, Materials Processing Technology,
209 (2), pp. 664671.
Callister, W. D. 2007. Materials Science and Engineering.
New York: John Wiley & Sons, Inc.
Dadbaksh, S. Hao, L. & Kong, C.Y. 2010. Surface Finish
Improvement of LMD Samples Using Laser Polishing.
In: Lancaster Product Development Unit, Rapid Design,
Prototyping and Manufacturing, University of Lancaster,
England, 11 June 2010, Buckinghamshire: CRDM.
Groover, Mikell P. 2002. Fundamentals of Modern Manufacturing. New York: John Wiley & Sons, Inc.
508
Guyson International, 2011. Blast Abrasives Data Sheets.

[online] Available at: <http://www.guyson.co.uk/
consumables/media.html> [Accessed January 2011].
Kruth, J.P., Vandenbroucke, B., Vaerenbergh van, J. &
Mercelis, P. 2005. Benchmarking of different SLS/SLM
processes as rapid Manufacturing techniques. Int. Conf.
Polymers & Moulds Innovations (PMI). Gent, Belgium,
2023 April 2005.
Mellor, S. Hao, L. Zhang, D. 2010. An Overview of Business Models and Strategy of Addative Manufacturing.
In: Lancaster Product Development Unit, Rapid Design,
Prototyping and Manufacturing, University of Lancaster,

England, 11 June 2010, Buckinghamshire: CRDM.
Realizer, 2010. Selective Laser Melting; Visions become
Reality. [Brochure] Borchen: Realizer.
UKGE Ltd., 2011. Grits, Polish, Media; Lapidary Supplies. [online] Available at: <http://www.ukge.co.uk/UK/
Lapidary.asp>.
XYZ Innovation, 2011. Selective Laser Melting. MTT
Technology Group. [video online] Available at: <http://
www.xyz innovation.com.au/selective-laser-melting-slm>
[Accessed January 2011].
509

Variable Fused Deposition Modelling analysis of benefits, concept design

and tool path generation.
H.L. Brooks, A.E.W. Rennie, T.N. Abram, & J. McGovern
Lancaster Product Development Unit, Department of Engineering, Lancaster University, Lancaster, United Kingdom
F. Caron
IUP Gnie Mcanique et Productique, UFR Sciences et Techniques, Universit de Bretagne Occidentale,
Brest, France
ABSTRACT: Current Fused Deposition Modelling (FDM) techniques use fixed diameter nozzles to deposit
a filament of plastic layer by layer. The consequence is that the same small nozzle, essential for fine details,
is also used to fill in relatively large volumes. In practice a Pareto-optimal nozzle diameter is chosen that
attempts to maximise resolution while minimising build time. This paper introduces a concept for adapting an
additive manufacturing system, which exploits a variable diameter nozzle for the fused deposition of polymers.
The variable nozzle allows the print resolution and the build speed to become independent variables which may
be optimised. The paper discusses a concept design for the variable diameter nozzle to be fitted to a RapMan
3D printer and the software used to generate the tool paths for the extrusion head. The methodology involves the
use of existing software solutions to gather basic data from STL files and generate the tool paths. A method for
integrating the data and the deposition system is proposed. The challenges and possibilities of the technology
are discussed as well as future research.
1
1.1
1.2 Literature review
INTRODUCTION
Fused deposition modelling
Fused deposition modelling (FDM) is a common

additive manufacturing (AM) method for creating
polymer parts. FDM builds parts additively by depositing a small bead of molten plastic through an extrusion
head onto a work platform. The nozzle (or platform)
moves via computer control to lay down a plastic
pattern on the platform. Complete parts are built
by laying down successive patterns one layer at a
time [1].
While the lead time to produce AM parts via FDM
is relatively short, the per unit manufacturing time is
much longer than conventional mass production techniques. This is partly due to the fact the extrusion head,
which has a small diameter nozzle, has a low volumetric deposition rate [2]. In conventional FDM, the
nozzle diameter is small to maximize the resolution of
each layer.
Three main options exist to increase the build rate:
i) increase the speed of the horizontal movements;
ii) minimize the material of the part; or iii) increase
the diameter of the nozzle.
This paper introduces a concept for adapting an
existing AM system which employs a variable diameter nozzle for the fused deposition of polymers. The
variable nozzle allows the print resolution and the build
speed to become independent variables which may be
optimised.
Research into FDM can be split into three main

categories: novel applications, materials and system
improvements [3, 4]. The vast majority of system
improvements are software based and are related to
tool path generation, slicing algorithms and part orientation optimisation [58]. The basic mechanical
system used for FDM has changed little since its inception in the 1980s and consists of three Cartesian linear
actuators, an extruder(s) and temperature controls [1].
Multiple extrusion heads are used to deposit different
materials and some research has been carried out for
printing multi-material parts [9].
Very little research work has been published on
the advancement of FDM extruder technology. Tseng
patented two novel deposition techniques called adaptive filament deposition (AFD) and planar layer deposition (PLD) for the freeform fabrication of metals
and ceramics [10, 11].
AFD utilizes a conical spindle inside a deposition
head to vary the volumetric flow rate of liquids through
the extrusion orifice. After leaving the orifice the liquid jet is cooled by the ambient air temperature and
becomes a filament. Retracting the spindle away from
the orifice allows higher flow rates which in turn
increases the filament size. PLD utilises adjustable
planar nozzles and rollers to deposit uniform thickness layers of material. The most notable feature of
PLD is the fact that the whole layer can be deposited
in one pass.
511
Figure 1. Geometric error due to circular nozzle.
Figure 2. Graph showing minimum geometrical error of an

angle printed with a circular nozzle.
Yang et al patented a variable nozzle similar to PLD

which allows layers to be deposited in a single pass
[12].
Both of these deposition techniques are not well
suited to the printing of engineering plastics and
require significant alterations to conventional FDM
systems. It is also unknown whether these deposition techniques can practically rival the levels of detail
achieved by conventional FDM.
Due to the lack of research on FDM extruder design,
and the current limitations imposed by constant diameter nozzles, the authors believe variable diameter
nozzles to be a worthwhile research topic.
2 VARIABLE FUSED DEPOSITION
MODELING (VFDM)
2.1
Impacts on horizontal resolution
FDM nozzles utilise circular orifices to take advantage

of the axisymmetric properties in the horizontal plane.
The benefit of this is that the nozzle does not need to
be kept normal to the extrusion path. The disadvantage
is that all angles will have radii greater than or equal
to the radius of the nozzle as shown in Figure 1.
The geometric error can be calculated using
Equation 1 below:
Where r = radius of the extrusion orifice and =

angle of the modelled geometry.
Figure 2 shows the relationship between the geometric error and the intended angle for normalised
nozzle radii. The bottom line represents a radius of
one unit while the middle and top line represent radii
two and three times larger respectively. As can be seen
on the graph, the error is small for large angles but
grows significantly for angles less than 60 . For 60
angles the error is at least equal to the orifice radius.
In practice, extrusion paths like the one shown in
Figure 1 are prevented by the tool path generation
software due to the overlapping projections of the nozzle path as it enters and exits the corner, as shown in
Figure 3.
Figure 3. Geometric error due to nozzle size and tool path

generation software.
To verify this statement, an isosceles triangle with

a 15 vertex in the horizontal plane, was printed on a
Dimension 1200 BST FDM machine. The slicing and
tool paths were generated using Stratasys proprietary
software Catalyst EX. According to Equation 1, the
calculated minimum error is 0.67 mm while the measured error was 3.87 mm. Therefore it is evident that
the errors associated with printing high aspect ratio
features have been highly underestimated and the need
for small diameter nozzles is even greater.
VFDM allows for a larger diameter orifice to be
used for infill, where geometric error is less important, and smaller diameters for the exterior shell where
high resolution is desirable. The authors believe that by
using this process it is possible to improve the quality
of FDM parts whilst increasing build speed.
2.2 Impacts on vertical resolution
A common feature amongst most additive layer manufactured parts is visible discontinuities on surfaces
between 0 and 90 from the horizontal. This is known
as the staircase effect and is a consequence of the fact
the sloped surfaces are approximated by small vertical
steps [13].
512
Figure 6. Plan view geometry of 30, 60, 90 triangle for

extrusion time saving analysis.
Figure 4. Staircase effect in layer manufacturing using
variable layer thickness.
Figure 7. Graph showing the extrusion times as a function

of part volume for different nozzle configurations.
Figure 5. Reduced staircase effect by utilising adaptive
layer slicing and VFDM.
Much progress has been made in the development

of slicing algorithms which alter the layer thickness
depending on the slope of the external surfaces. In
general, near vertical slopes require fewer thicker
layers, while the inverse is true for near horizontal
slopes (Figure 4).
Sabourin et al developed an adaptive slicing technique to create high resolution parts with high build
speed interiors [8, 14]. This is achieved by having thick
internal layers and thin sub-layers in the shell. The
height of the shell sub-layers must add up to the height
of the corresponding interior layer. In this way the resolution of sloped surfaces can be increased without
increasing the build time of the interior.
Combining similar slicing software with VFDM
could improve the vertical resolution of even the most
curved, high aspect ratio features by reducing the nozzle diameter. Figure 5 shows how the layer thickness
could vary for exterior and interior layers.
Inevitably the nozzle will have a maximum and
minimum diameter setting so it may be necessary to
have more than one path width between the interior
layers and the exterior of the part. When considering
the resolution of the horizontal layers, it is evident that
the horizontal shell thickness will need to be large
enough to cover any underlying thick layers, so in
practice there would be more shell layers than shown
in Figure 5.
To avoid collisions between the part and the nozzle, the shell sub-layers will need to be deposited
before each corresponding interior layer; however
this is already a common-place printing strategy.
2.3
Impacts on extrusion time
For large parts, with a high percentage of infill, the

extrusion time during a build is a function of the nozzle
diameter and the geometry of the part.
For the purposes of an example, consider a 30, 60,
90 triangle which is 1 unit thick and has dimensions
as defined in Figure 3.
Assuming the part is 100% solid and the nozzle
diameters, feed rate and layer thicknesses are known, it
is possible to calculate the extrusion time as a function
of part volume (Equation 2).
where volume = the volume of the part (mm3 ), D = the

nozzle diameter of the part (mm), f = nozzle feed rate
(mm/s) and l = layer thickness.
If the nozzle is able to use a small diameter for the
outer shell(s) and a larger diameter for the infill then
Equation 2 becomes:
where volumeshell(s) = volume of the outer layer(s) of

the part, volumeinfill = volume of the inside of the
513
room required on the positioning rack. This problem is magnified for machines with multiple material
capabilities.
Unfortunately, adaptive slicing like that mentioned
in section 2.2 will not reduce the time required to
build parts with vertical sides, however, previous
research has shown average time savings, for a wide
range of part geometries, to be between 37 and 50%
[8, 15]. Therefore, combining VFDM with adaptive
layer slicing could save up to 75% on extrusion time.
2.4 Concepts for VFDM
Figure 8. Two stage nozzle with a) small diameter orifice

and b) large diameter orifice.
part, D1 = small nozzle diameter, D2 = large nozzle

diameter.
Using the following parameters, D1 = 0.5 mm,
D2 = 1.0 mm, f = 25 mm/s, and l = 0.5, a graph was
created showing the extrusion times for the 30, 60,
90 triangular prism over a range of volumes and variable nozzle diameter configurations. Configuration A
(the steepest line) shows the extrusion time required if
only a nozzle of diameter 0.5 mm is used. Configuration D (the shallowest line) shows the extrusion time
for a fixed 1.0 mm nozzle. C and B show the extrusion times for parts with one and two layer thick shells
respectively.
As can be seen by the small difference between configurations B, C and D, the amount of time required
to add high resolution outer shells is insignificant for
large volume builds. The analysis also indicates that,
as a rule of thumb, the savings in time for a two stage
nozzle is proportional to the different nozzle diameters. In the case of the 30 vertex, replacing the 1.0 mm
nozzle with the 0.5 mm nozzle reduces the theoretical
error from 1.43 mm to 0.72 mm. In practice, due to the
way the tool paths are calculated, the improvements
in resolution will be far greater. For extremely high
aspect ratio features it may be necessary to use the
small nozzle to fill in the gap between the infill and
the shell.
It should be noted here that the same extrusion
time savings could be made by utilising two separate
extrusion heads with different fixed diameter nozzles. However, the problem with this solution is the
time taken to switch extrusion heads and the extra
Possible concepts for VFDM nozzles can be placed

into two main categories: continuously variable nozzles (CVNs) and discretely variable nozzles (DVNs).
CVNs allow much more freedom in choosing the
print resolution and build speed and allow for the
possibility of size changes whilst printing. The main
disadvantage of CVNs is the mechanical complexity and the necessary design compromises that result.
Possible solutions reviewed by the authors include
mechanical irises, sliding jaws and smart materials.
All CVN solutions analysed required that the orifice be an approximation of a circle. This increases
the error relative to a circle of the same area and may
result in orthotropic extrusion properties. For example think of a square moving in a direction parallel
with its sides, and then imagine the square moving
diagonally. Obviously the track will be wider for the
diagonal movement. Another difficulty in designing
a CVN is scaling down the actuating mechanisms to
a size desirable for typical FDM applications. The
design and build of CVNs is likely to be the topic
of future research by the authors.
DVNs are mechanically simpler than CVNs and are
therefore relatively simple to manufacture at the submillimetre scales required. Possible solutions to DVNs
are interchangeable nozzles and multiple stage nozzles. Interchangeable nozzles could sit in a carousel
and automatically change via computer control. This
concept was abandoned due to the difficulties associated with the nozzles into and out of the polymer
flow stream. The multiple stage nozzle works by moving nested inserts inside the nozzle. The multiple stage
nozzle was found to be relatively easy to implement
while allowing circular orifices and uninterrupted
polymer flow.
The design for a two stage nozzle is presented in the
next section.
2.5 Two stage nozzle design

The two stage nozzle consists of an inner and outer
nozzle each with fixed orifice diameters. The inner
nozzle is actuated by a solenoid, and in its lowest
position is nested firmly against the inner surface
of the outer nozzle, restricting the flow through the
smaller orifice. When the inner nozzle is lifted, a secondary melt chamber is formed which allows multiple
514
Figure 10. Two stage nozzle and heating assembly.
Figure 9. Unmodified RapMan 3.0 from Bits for Bytes.
polymer flow streams to coalesce before leaving the

larger orifice as shown in Figure 5.
This concept achieves many of the objectives of
VFDM but also has a number of significant drawbacks.
Assuming the outer shell of each layer is deposited
first, there will be a small delay required before the
infill can start as the secondary chamber must be filled
with molten polymer. The second major drawback is
that polymer in the secondary chamber is likely to
be pushed out as the inner nozzle moves back into
the lower position. This not only wastes material but
will add extra build time as the nozzle moves to a
designated discharge point.
Notwithstanding these problems it is expected that
the two stage nozzle will reduce the build time
for large, solid, high resolution parts. The following section describes how the nozzle is integrated
into a RapMan 3.0 3D printer from Bits from Bytes
(BfB) [16].
4 TOOL PATH GENERATION

It is the authors intention to develop an advanced
tool path generation algorithm in parallel with future
developments in VFDM. The tool path generation
described in this paper is a more basic method which
will be used for testing the two stage nozzle concept.
4.1 Standard printing procedure
A summarised standard procedure for printing parts
on a RapMan machine is as follows:
3 TWO STAGE NOZZLE DESIGN

3.1
to the RapMan control board along with a thermistor for temperature control. The other power resistor
is connected to an external circuit to enable a base
heating load giving greater temperature control via the
thermistor, see Figure 7.
The filament guide also acts as the connection
between the solenoid and the inner nozzle. The filament feed mechanism will be moved slightly to allow
the solenoid to be positioned directly above the inner
nozzle.
The RapMan uses open source software, Skeinforge, to generate the required G-code from .STL files.
The next section describes the software tool chain
used in the proof of concept printing.
RapMan 3D printers
The RapMan 3.0 is a FDM machine based on the open

source Reprap project and is shown in Figure 6. The
low cost and open structure of the RapMan makes it
highly suitable for research related modification.
The VFDM extruder is designed to replace the
original RapMan extruder with minimal changes. The
standard RapMan heating system will be changed from
a fixed resistance wire and fire cement configuration
to a removable aluminium block which contains two
power resistors. One of the power resistors is connected
1. Generate an STL file from a CAD program.

2. Upload the STL file into Bits from Bytes Axon
(Skeinforge).
3. Generate the G-code with the .bfb extension.
4. Upload the .bfb file onto an SD card.
5. Insert the SD card into the RapMan card reader and
run the file.
The standard input file to most FDM systems is the
stereolithography or STL file. STL files describe the
surface geometry of parts by using a triangular mesh
[17]. The BfB Axon program is a user friendly frontend to the more complicated but versatile Skeinforge
515
Figure 11. Tool path generated with advanced Skeinforge

settings as shown in Skeinview.
open source software. Skeinforge is a tool chain composed in Python script that allows the user to control
many of the parameters that generated the tool path
G-code. G-code is the common name given to the computer numeric control programming language adopted
by many automated machine tools [18]. While generic
G-code generators exist, many manufacturing technologies use machine specific G-codes which require
post-processing in order to be used in other machines.
The RapMan uses its own form of G-code which is,
amongst other things, differentiated by the extension
.bfb.
The next section describes how the Skeinforge
settings were adjusted for the two stage nozzle and how
the BfB G-code was subsequently altered to allow for
the nozzle mode changes.
4.2
Printing procedure for the two stage nozzle
The STL file (or equivalent) is produced in CAD software via the usual method. Once the STL file is loaded
into the BfB Axon software, the Skeinforge program
can then be accessed by clicking on the advanced settings button. In the Skeinforge program it is possible
to alter the infill large diameter nozzle. In practice this
requires setting the Infill solidity (ratio) so that the
paths are spaced further apart, whilst simultaneously
setting the Infill width over thickness (ratio) to make
the bead width wider. For example: if the layer thickness is 0.25 mm, the width of the shell is 0.25 mm
and the desired width of the infill tracks is 0.5 mm,
then Infill solidity (ratio) should equal 0.5 and the
Infill width over thickness (ratio) should equal 2 (i.e.
0.25 mm 2 = 0.5 mm). A number of fill patterns can
be used, however for 100% solid parts it is better to use
rectangular or line fill options.
It is also possible to change the number of shells
desired. This is more important for parts with sloped
sides than parts with vertical sides.
Once the settings have been changed in Skeinforge
it is possible to produce the .bfb G-code file in BfB
Axon. The .bfb file can then be loaded in a module of
Skeinforge called Skeinview and analysed to see if the

settings had the correct effect, see Figure 8.
Once the G-code with the correct path spacing is
created, it is necessary to alter the code manually to
introduce steps required for the two stage nozzle. This
involves adding an extruder delay between the shell
and infill depositions to allow the secondary melt
chamber to fill up. The extruder flow rate and temperature may also be changed at this point. Once the
infill is complete, extra lines of code are required to
move the head to the designated purge point, after
which the extruder may return to the original flow rate
and temperature settings.
Finding the code lines where the machine changes
from extruding the shell to the infill is currently done
by viewing the code line-by-line in Skeinview. While
this manual manipulation of the G-code is acceptable for a proof of concept trial, the process will ideally
be automated for future versions of VFDM.
FUTURE WORK
The next step in the development of the two stage

nozzle is to run print tests to determine usable print
profiles for common FDM polymers. This will require
the systematic varying of process parameters as a
precursor to materials testing and metrological analysis. This will allow VFDM parts to be built and
the performance characteristics to be compared with
conventional FDM.
In order for VFDM to reach its full potential, the
current barriers to CVNs need to be overcome. This is
a matter of ongoing research by the authors. In parallel
with this research is development of VFDM tool path
generation algorithms which in combination with variable layer slicing programs could drastically reduce
build times and improve build accuracy and surface
finish.
CONCLUSIONS
This paper presented a concept for variable fused deposition modelling. The major findings are as follows:
The theoretical benefits for horizontal accuracy
were derived in terms of generic circular nozzles.
The potential improvement in accuracy for convex
angles less than 60 is considerable.
The time savings in extrusion time for a solid 30,
60, 90 triangular prism was calculated as a function of volume for different build strategies. The
results showed that for large build volumes and
constant layer thicknesses the savings in extrusion
time is approximately proportional to the ratio of
the maximum and minimum nozzle diameters.
The benefits in vertical resolution and build time
by combining adaptive layer slicing and VFDM
were discussed. The improvement of the vertical
resolution is restricted by the minimum nozzle
516
diameter and the sub-layer thickness. The reduction

in build times from adaptive slicing alone have been
reported in the range of 37 to 50%.
A concept design for a two stage nozzle was provided along with the methodology required to generate the G-code for manufacture. The advantages
and disadvantages of DVNs compared with CVNs
were also discussed.
Future work was discussed with the aim of improving the design of continuously variable nozzles and
the associated tool path generation.
By making clear the advantages of VFDM and
providing the methodology for a detailed concept, it
is hoped that further research will be stimulated in this
area.
REFERENCES
[1] S. S. Crump, Apparatus and Method for Creating
Three-Dimensional Objects, United States Patent, No.
5,121,329, 1989.
[2] N. Hopkinson and P. Dicknes, Analysis of rapid
manufacturingusing layer manufacturing processes
for production, Proceedings of the Institution of
Mechanical Engineers, Part C: Journal of Mechanical
Engineering Science, vol. 217, pp. 3139, 2003.
[3] S. H. Ahn, et al., Anisotropic material properties of
fused deposition modeling ABS, Rapid Prototyping
Journal, vol. 8, pp. 248257, 2002.
[4] I. Zein, et al., Fused deposition modeling of novel
scaffold architectures for tissue engineering applications, Biomaterials, vol. 23, pp. 11691185, 2002.
[5] P. M. Pandey, et al., Real time adaptive slicing for
fused deposition modelling, International Journal of
Machine Tools and Manufacture, vol. 43, pp. 6171,
2003.
[6] K. Thrimurthulu, et al., Optimum part deposition orientation in fused deposition modeling, International
Journal of Machine Tools and Manufacture, vol. 44,
pp. 585594, 2004.
[7] D. Chakraborty, et al., Extruder path generation

for Curved Layer Fused Deposition Modeling,
Computer-Aided Design, vol. 40, pp. 235243, 2008.
[8] E. Saburin, et al., Accurate exterior, fast interior layered manufacturing, Rapid Prototyping Journal, vol.
3, pp. 4452, 1997.
[9] W. K. Chiu and S. T. Tan, Multiple material objects:
from CAD representation to data format for rapid prototyping, Computer-Aided Design, vol. 32, pp. 707
717, 2000.
[10] A. A. Tseng, Adaptable Filament Deposition System and Method for Freeform Fabrication of ThreeDimensional Objects, United States Patent, No. US
6,251,340 B1, 2000.
[11] A. A. Tseng and M. Tanaka, Advanced deposition techniques for freeform fabrication of metal and
ceramic parts, Rapid Prototyping Journal, vol. 7,
pp. 617, 2001.
[12] D. Y. Yang, et al., Method for Manufacturing ThreeDimensional by Variable Deposition and Apparatus
Used Therein., United States Patent, No. 6,505,089
B1, 2000.
[13] A. Dolenc and I. Mkel, Slicing procedures for
layered manufacturing techniques, Computer-Aided
Design, vol. 26, pp. 119126, 1994.
[14] E. Sabourin, Adaptive High-Precision Exterior, HighSpeed Interior, Layered Manufacturing., Master of
Science in Mechanical Engineering, Mechanical Engineering, Virginia Polytechnic and State University,
Blacksburg, 1996.
[15] J. Tyberg and J. H. Bhn, FDM systems and local
adaptive slicing, Materials & Design, vol. 20, pp. 77
82, 1999.
[16] B. f. Bytes. Available: http://www.bitsfrombytes.com/
[17] C. K. Chua, et al., Rapid Prototyping: Applications
and Principles, 2nd Edition ed. Singapore: World
Scientific Publishing Co. Pte, Ltd., 2003.
[18] E. Oberg, et al., Machinerys handbook. New York,
USA: Industrial Press, Inc., 1996.
517

Innovative features in implants through Beam Melting

a new approach for Additive Manufacturing of endoprostheses
B. Mueller, T. Toeppel, M. Gebauer & R. Neugebauer
Fraunhofer Institute for Machine Tools and Forming Technology IWU, Chemnitz/Dresden, Germany
ABSTRACT: At the current state of the art, endoprostheses are predominantly manufactured by cutting,
forming or casting technologies. Another, rather new way of manufacturing endoprostheses and implants is
the Additive Manufacturing process called Beam Melting. In particular the customized production with no
need for any type of tooling, combined with the unique freedom of design, spark interest in this technology
to manufacture complex and patient-specific implants. The use of Beam Melting facilitates the fabrication of
endoprostheses with almost any design of inner and outer geometries. Recent developments and research activities were focused on structured surfaces or hollow structures with variable elasticity. In this paper, possibilities
for the integration of innovative features in endoprostheses through Additive Manufacturing with Beam Melting
will be introduced.
INTRODUCTION
Implants, so-called endoprostheses, do partially or

fully replace damaged body parts, like worn joints
or defective bone material, and remain permanently
inside the human body. In Germany alone, more than
400,000 artificial joints are getting implanted every
year. 90 percent of these implantations account for
artificial hip and knee joints. The remaining cases
concern the replacement of shoulder, elbow or finger joints as well as ankle and metatarsophalangeal
joints (Gradinger & Gollwitzer 2006, Veit et al. 2009,
N.N. 2011).
Main cause for the necessity of an artificial joint is
wear of the cartilage layer between two bone ends in
a joint, scrubbing against each other (arthrosis). With
the artificial joint, a pain-free, every day motility and
load-bearing capacity and by that the patients quality
of life are supposed to be restored (Fels 2007).
Figure 1 shows that in between five years only (2004
through 2008) the number of initial implantations
for hip endoprostheses has increased by 14 percent.
Especially the group of 50 to 59 year olds and 80 to
89 year olds are affected by that. Explanation can be
found on the one hand in the fact that today already
more than fifty percent of all Germans older than fifty
years are affected by arthrosis in at least one joint (N.N.
2011) and on the other hand, rising life expectancy in
Germany does lead to a higher demand in endoprosthetic care for the elderly.
In addition, Figure 1 shows that the number of
revisal surgery with hip endoprostheses is rising. This
is to be explained with the fact that endoprostheses
have a limited service life only and with respective
Figure 1. Hip endoprostheses initial implantations (divided

according to patients age) plus hip endoprostheses replacements and component replacements in Germany between
2004 and 2008.
young age of initial implantation, have to be replaced

during the patients life span by a revisal implant.
This results in a constantly rising demand in innovative endoprostheses with long service life and at
the same time low risk in post-operative complication,
e.g. wound infections (Veit et al. 2009).
519
Figure 3. Laser Beam Melting technology in process for

Rapid Manufacturing of three hip stem prostheses.
Figure 2. Schematic image of Beam Melting process.
2 ADDITIVE MANUFACTURING OF
IMPLANTS
2.1
Figure 4. Structured surface of an endoprosthesis (left:

exemplary structure type, right: transition from surface
structure to solid implant body).
State of the art in implant AM
At the current state of the art, endoprostheses are

predominantly manufactured by cutting, forming or
casting technologies (Sprung et al. 1987, Gradinger &
Gollwitzer 2006, N.N. 2008, Fladerer 2008, Kuttkat
2010). However, restrictions apply to these technologies, e.g. necessary accessibility for cutting tools
or draft angles for moulds and dies, which restrict
the geometric freedom in implant design or increase
manufacturing complexity with long process chains.
An innovative approach for manufacturing of endoprostheses can be found in the Additive Manufacturing
(AM) technology of Beam Melting. With that, the
implant gets manufactured by layer-wise, local melting of metal powder with a laser or electron beam,
based on 3D CAD data (Fig. 2).
Via Beam Melting technologies, different
biocompatible and medically accredited metals can
be processed, including titanium and titanium alloys
(e.g. Ti-6Al-4V or Ti-6Al-7Nb) or cobalt chromium
alloys (e.g. CoCrMo).
With Beam Melting technologies existing limitations of conventional manufacturing processes (e.g.
milling, forming or casting) can be overcome. Because
of the additive process approach with no need for any
type of tooling, virtually any inner and outer design
can be realized.
Thus, completely new geometries and functions can
be implemented in implants and being manufactured
with Beam Melting technologies. These include, for
example, surface structures (Sprung et al. 1987,
Laoui et al. 2004, Neuner 2008) for a better
bone ingrowth into the implant, or lattice-like volume structures (Christensen 2007, Neuner 2008,
Hltkemeier 2008, Challis et al. 2010) to adapt the

stiffness of the endoprosthesis to the bone.
Beam Melting technology is basically suited for
direct manufacturing of individual implants based on
computer or magnetic resonance tomography data
(Janssens & Poukens 2007, Mller 2009, Tppel 2010)
as well as mass production of standardized implants
(Cremascoli & Ohldin 2009, N.N. 2010) as to be seen
in Figure 3.
2.2 Macro-porous surfaces structures
Additive Manufacturing technologies allow an
implants surface to be structured in virtually any
desired design option. Figure 4 (left) shows an
exemplary implant surface structure representation.
The surface structure can be applied to selected surface
areas only as well as to the implants entire surface. The
structures depth (thickness) can be chosen arbitrarily for best possible support of bone ingrowth (Fig. 4,
right). Furthermore, active elements can be easily integrated in additively manufactured implant bodies for
better primary stability and prevention of loosening
(Brunig 2010).
2.3 Internal cellular structure
With Additive Manufacturing, regular lattice structures can be generated in various shape and size.
Applying this kind of lattice structure to the inner
volume of an endoprosthesis, the implants stiffness
and rigidity (Youngs modulus) can be adapted to
that of human bone. Furthermore, the implants dead
520
Figure 7. Inner channel design of hip stem implant to distribute bone cement or bio-resorbable filler to implant-bone
interface.
to support the vision of a life-long implant and to

come a significant step closer to this ultimate goal.
The following sub-chapters describe these potential
new functions and their benefits in detail.
Figure 5. Microsection of implant with solid shell and inner
lattice structure.
3.2 Better fixation of cement-free implants
Figure 6. Beam Melted hip stem prosthesis with functional channels and cavities, inner cellular structure and
macro-porous surface areas.
weight can be significantly reduced, which is especially useful for heavy endoprostheses made out of
cobalt chromium alloys. Figure 5 shows a close up of
an implants inner lattice structure as described above.
3
3.1
A better fixation of cement-free endoprostheses can be

achieved by targeted, local insertion of bone cement or
bio-resorbable filler through the inner channels to the
implant-bone interface after implantation. Thereby it
becomes possible to compensate fitting inaccuracies
after the implant has been tapped into the bone. In addition, the operation surgeon receives the opportunity
to backfill unexpectedly bad bone conditions, which
could not be identified in pre-surgical diagnostics, during surgery with minimal amounts of a compensational
substance.
Another intriguing opportunity is to do so not during
initial implantation but rather years later to counteract
loosening of an implant for preventive or correctional
purposes. Rather than a heavily invasive revision operation, replacing the implant with a revision type of
bigger size, this minimally invasive approach can spare
valuable sound bone structure and save the patient
from the inconvenience of implant replacement.
A potential channel design for this added value
described above can be seen in Figure 7.
3.3 Post-operative medical treatment through
drug depot
INTEGRATION OF NEW FUNCTIONS IN

ENDOPROSTHESES
Motivation
The geometric freedom of Beam Melting technology allows not only the manufacturing of implants
with cellular structures but also the integration of
complex channels and cavities for a variety of new
functions in endoprostheses. To this end, an implant
with inner, functional channels and cavities, which
can be designed to any desired requirement profile,
has been developed. A hip stem has been chosen as an
example for a typical implant with high potential to
add functionality to it (Fig. 6).
A whole variety of additional functions have been
identified to add value to the implant. Main goal of all
these added functions is the elongation of the implants
service life and the prolongation or even avoidance
of revisal surgery. These added values are supposed
Another added value of the presented endoprosthesis with functional cavities is the possibility of
post-operative medical treatment of the patient.
Thereby, the steady, regular release of medication
from a cavity inside the implant (drug depot) through
defined channels to the surrounding bodys own bone
and soft tissue (wound) becomes possible. A possible
cavity and channel design for this purpose is shown in
Figure 8.
This feature allows the aimed promotion of wound
healing as well as ingrowth behavior of somatic cells
into the endoprosthesis. In addition, medication for
pain relief and prevention of infections (antibiotics)
can be supplied through the inner channels to the
implant-tissue interface.
If necessary, the channels can also be used for
post-operative discharge of blood and wound ooze
521
Figure 8. CAD image of a hip stem prosthesis with drug

depot (light color) and distribution channels (medium and
dark color).
Figure 9. CAD image of a potential channel design aimed

for endoscopic inspection purposes in a hip stem prosthesis.
(drainage). With one central drain, connected to the

implants inner channel system, it becomes possible to
access many different locations around the implant for
drainage purposes, depending on the number of channel exits which can be numerous thanks to the design
freedom of Additive Manufacturing.
3.4
Endoscopic inspection through the implant
In another potential application the inner channels

can be designed in such a way, that they can be utilized for endoscopic inspection of the contact area
of implant and bone as well as surrounding body
tissue. Consequently, another immediate option for
post-operative, minimally invasive medical monitoring alongside imaging technologies like computer
tomography is getting created, allowing earlier detection of possible complications. Figure 9 shows a
possible channel design for endoscopic purposes.
3.5
Explantation support
Finally, the functional channels can be utilized in case

of a necessary explantation (revision operation) to distribute a medium for locally limited decomposition of
implant-bone bonding to ease and speed up the surgical intervention for implant removal with minimal
destruction of sound bone structure. This function can
play an important role in the process of convincing
surgeons to give up their hesitations to employ endoprostheses with structured surfaces due to the expected
complication of a potential explantation caused by the
intensive bonding and ingrowth.
Certainly, this medium has to be applied carefully
and well-dosed to avoid distribution beyond the direct
contact section of implant to bone.
Figure 10. Low volume Rapid Manufacturing of endoprostheses with special functions (three piece build job).
3.6 Application range

Functional channels and cavities can be applied to
individual, patient-specific implant geometries as well
as to series-type implants for volume production
using Additive Manufacturing technology like Beam
Melting in terms of Rapid Manufacturing (Fig. 10).
SUMMARY
Additive Manufacturing of implants by means of Beam

Melting technologies allows undreamt-of possibilities
in the geometric design of implants. In addition to
patient specific design and surface or volume structuring of implants, the integration of value-added features
in implants is a new field of application for Additive
Manufacturing technologies in arthroplasty. Internal
functional channels and cavities provide implants with
completely new and additional features and improved
performance which can result in less revision surgeries, less intra-and post-operative risks and therewith
in an overall increased patient comfort.
OUTLOOK
The previously presented example has demonstrated

the technical feasibility of implementing internal
channels and cavities with unique geometric freedoms
in endoprostheses. In subsequent projects, specific
requirements to these channels and cavities as well
as their respective benefits for the operating surgeon
and the patient need to be further worked out. Related
to this, it has to be determined for which types of
522
implants beside hip stems these additional features are

desirable.
Further research activity should also focus on insuring the effect of the above outlined potential extra
functions of endoprostheses in practice. One major
issue to be solved is to insure the channels functions
during the implants service life in the human body,
e.g. by avoiding ingrowing and congestion of the channels with body tissue (hard or soft). Another key topic
for future research is the proper medium distribution
for decompostion of implant-bone bonding to assure
its effect to stay limited to the targeted implant-bone
contact region only with minimal negative effect on
surrounding bone tissue that is supposed to be saved.
Last but not least, the question has to be answered, if
the outlined added value in endoprostheses can satisfy
somewhat higher manufacturing cost. Different health
insurance and cost coverage models in different countries might come up with divergent answers to this,
mainly depending on who the final decision maker is
the patient, the operating surgeon or the insurance
provider, and if potentially lowered follow-up costs are
considered.
REFERENCES
Brunig, J. 2010. AktiLocKonzept fr ein Implantat mit
aktiver Verankerung. In Jahresbericht 2009/2010: 47.
Chemnitz: Fraunhofer IWU.
Challis, V.J., Roberts, A.P., Grotowski, J.F., Zhang, L.C. & Sercombe, T.B. 2010. Prototypes for Bone Implant
Scaffolds Designed via Topology Optimization and Manufactured by Solid Freeform Fabrication. Advanced
Engineering Materials (12)11: 11061110.
Christensen, A. 2007. Additive Manufacturing In Metal: New
Option for Medical Applications. MoldMaking Technology (2007)12: reprint.
Cremascoli, P. & Ohldin, P. 2009. Series production of
CE-certified orthopaedic implants with integrated network structures for improved bone ingrowth. In Advanced
Research in Virtual and Rapid Prototyping; Proc. 4th
intern. conf., Leiria, 610 October 2009. London:
Taylor & Francis.
Fels, H.-G. 2007. Langfristige Nachuntersuchungsergebnisse
der zementfreien Zweymller-Hfttotalendoprothese. Eine
retrospektive klinische und radiologische Untersuchung
12,0 Jahre postoperativ. Dissertation. Mnster: Westflische Wilhelm-Universitt, Medizinische Fakultt.
Fladerer, F. 2008. Zerspanung im Dienste der Medizintechnik. In MaschinenMarkt online: http://www
. maschinenmarkt . vogel . de / themenkanaele/produktion/
spanendefertigung/maschinen/articles/107564(pageview:
20110113). Wrzburg: Vogel Business Media.
Gradinger, R. & Gollwitzer, H. 2006. Ossre Integration.
Heidelberg: Springer Medizin Verlage.
Hltkemeier, K.-U. 2008. Backe, Backe Knochen. In Konstruktionspraxis.de:http://www.konstruktionspraxis.vogel.
de/themen/werkstoffe/rapidtechnologien/articles/157406
(pageview: 20090205). Wrzburg: Vogel Business
Media.
Janssens, M. & Poukens, J. 2007. Rapid Technologies in
medicine: What can, cant be done and why. In Competitive Manufacturing (COMA); Proc. intern. conf.,
Stellenbosch, 31 January2 February 2007.
Kuttkat, B. 2010. Implantate von der Stange fordern Fertigung heraus. In MaschinenMarkt online: http://www.
maschinenmarkt . vogel.de / themenkanaele /produktion /
zerspanungstechnik/articles/293819 (pageview: 2011
0113). Wrzburg: Vogel Business Media.
Laoui, T., Santos, E., Osakada, K. & Shiomi, M. et al. 2004.
Properties of titanium implant models made by laser processing. In Laser Assisted Net Shape Engineering; Proc.
4th intern. conf., Erlangen, 2124 September 2004.
Mller, B. 2009. Generative Fertigungsverfahren
Mglichkeiten der Herstellung patientenspezifischer
Implantate und medizintechnischer Komponenten. In
Przisionsfertigung fr die Medizintechnik; Proc. workshop, Chemnitz, 13 May 2009. Chemnitz: Fraunhofer
IWU.
Neuner, M. 2008. Implantat mit Orangenhaut. Fraunhofer
Magazin (2008)4: 3031.
N.N. 2008. Implantat ist nicht gleich Implantat. Information
paper. Freiburg: Zimmer Germany GmbH.
N.N. 2010. Strukturen entscheiden. Werkzeug & Formenbau
(20)3: 4042.
N.N. 2011. Stand und Entwicklung des knstlichen Gelenkersatzes in Deutschland. Berlin: Bundesverband Medizintechnologie e.V.
Sprung, J. et al. 1987. Feinguss fr alle Industriebereiche.
Dsseldorf: Zentrale fr Gussverwendung.
Tppel, T. 2010. Praxiseinsatz der Laser CUSINGTechnologie im Formenbau und weiteren Anwendungsgebieten.
In Praxisseminar Laser CUSING, Lichtenfels, 16 September 2010. Lichtenfels: CONCEPT Laser GmbH.
Veit, C., Bauer, J., Dbler, K. & Fischer, B. (ed.)
2009. Qualitt sichtbar machen. In BQS-Qualittsreport
2008: 146160. Dsseldorf: BQS Bundesgeschftsstelle
Qualittssicherung gGmbH.
Wirtz, T.P. 2005. Herstellung von Knochenimplantaten
aus Titanwerkstoffen durch Laserformen. Dissertation.
Aachen: RWTH, Fakultt fr Maschinenwesen.
523

Investigating the influence of build parameters on the mechanical

properties of FDM parts
J. Giannatsis, K. Sofos, V. Canellidis, D. Karalekas & V. Dedoussis
Laboratory of Advanced Manufacturing Technologies & Testing, Department of Industrial Management and
Technology, University of Piraeus, Greece
ABSTRACT: As the field of Rapid Manufacturing (RM) is continuously gaining acceptance in the industrial
and manufacturing sector, the evaluation of the mechanical properties of RM parts is becoming crucial for
design and functional purposes. According to previous studies, the mechanical properties of RM parts depend
mainly on the properties of the materials employed, but also to a significant extent on the build (fabrication)
parameters selected for their production. Fused Deposition Modeling (FDM) is one of the most promising,
in terms of RM, technologies due mainly to the nature of the material commonly employed, namely ABS. In
this context an experimental investigation of the influence of build parameters on the mechanical properties
of parts fabricated with FDM was undertaken. The experimental work involved the fabrication of a series of
test specimens, which were built employing two different types of FDM machines, namely a commercial FDM
office printer and a desktop open-source 3D printer, and varying build parameters. The fabricated specimens
were subsequently subjected to tensile and bending loads. The results demonstrate noticeable differences in their
mechanical behavior.
INTRODUCTION
Since their first introduction, Additive Manufacturing

(AM) technologies have developed significantly, to
the extent that they can now be considered as alternatives for the production of actual manufacturing
tools or products, a concept usually referred as Rapid
Manufacturing RM (Ogando 2007). A significant
portion of the AM systems installed worldwide are
extrusion-type, i.e. the layers are formed through the
continuous extrusion and deposition of molten material supplied in filaments. Extrusion-typeAM systems,
such as these based on Fused Deposition Modeling
(FDM) technology, are among those that seem to
have the biggest potential for RM, due mainly to the
versatility of the materials that can be used (polymers, ceramics, even edible materials), and to their
respective properties (Grimm 2003). Furthermore,
extrusion-type systems seem to dominate another area
of increasing interest, namely the development of
low-cost and open-source 3D Printers, which is being
pursued by both the academia and a relatively large
base of amateur users for the production of final
products for home, personal or educational use.
Common to the above trends is the necessity for
increased functionality and quality of the AM products, a requirement that is directly connected with
their mechanical properties. As with other manufacturing technologies, the mechanical properties of AM
products depend to a large extent on the properties
of the corresponding raw material, but depend also

significantly on various fabrication (build) parameters (Kim & Oh 2008). In extrusion-type systems such
parameters can be the layer deposition style (controlled
by filling density and raster angle), the layer thickness,
the build orientation, the chamber and nozzle temperature, nozzle diameter and travelling speed (Ahn et al.
2002). Some of these parameters, like the chamber and
nozzle temperature, are machine dependent and cannot usually be altered by the operator, while others, like
build orientation, are defined by the operator/machine
user.
With the increasing use of AM technologies it
becomes interesting from a scientific and practical
point of view to investigate the effect of build parameters on the mechanical behavior of the final product.
These data could be subsequently used for modeling
the behavior of parts prior to fabrication and the optimization of their structure and form, as well as for
the optimization of their production in terms of cost
and quality. Furthermore, they could provide insights
on the nature of complex phenomena such as the
layer bonding mechanism and stress accumulation and
guide further development of the technologies. In the
present paper the preliminary results of an investigation in this subject are presented. The investigation
focused on two polymer extrusion-type systems, an
FDM 3D printer and a manually assembled low-cost
3D printer, both employing ABS as the main raw
material.
525
LITERATURE REVIEW
The effect of process parameters on the mechanical

properties of FDM parts has been the subject of several
earlier studies. Sun et al. (2008) studied the mechanisms controlling the bond formation among extruded
polymer filaments in FDM. According to their results
the envelope temperature and variations in the convection coefficient had strong effects on the overall quality
of the bond strength between filaments, as observed
in flexural tests. A study of the mechanical behavior
of ABS filaments was also conducted by Rodrguez
et al. (2001), who report that there is significant influence of the mesostructure of the filaments on their
stress-strain response of FDM tensile specimens.
According to Es-Said et al. (2000) layer raster
orientation/angle for the deposition of molten material has a significant effect on the tensile, flexural
and impact strength of FDM parts. The effect of build
direction/orientation on the FDM parts compressive
strength is shown in an experimental study conducted
by Lee et al. (2007). Similar effects of build orientation and deposition style on the tensile, flexural and
impact strength of FDM parts are reported in studies
by Bellini and Geri (2003) and Kim & Oh (2008).
The effect of layer thickness and deposition style on the
mechanical behavior of FDM parts is investigated in
a study by Lee et al. (2005), who employ the obtained
results in order to optimize the functional behavior of
a simple part.
In a recent study Sood et al. (2010) also report that
layer thickness, raster angle, raster width, raster gap
and build orientation have significant effect on the
strength of FDM parts. Finally, Zhang & Chou (2008)
study experimentally the effect of deposition parameters on part distortions and stress accumulation of
FDM parts and employ the Finite Element Analysis
method in order to model them.
EXPERIMENTAL SCHEME
The Laboratory of Advanced Manufacturing Technologies & Testing of the University of Piraeus is
equipped with two extrusion-type 3D Printers, one
FDM Dimension Elite machine and one open-source
educational RapMan machine (Fig. 1). The Dimension
Elite machine is a relatively low-end industrial AM
system that is mainly targeted at the design modeling
and prototyping sector, but is nevertheless capable of
relatively good product accuracy and quality.
The RapMan machine, on the other hand, is a commercial application of the Rep-Rap project, which
aims at the development of open-source and low-cost
3D printers (Jones et al. 2011). RapMan machines are
targeted mainly to the educational sector and they are
sold in the form a pre-compiled components kit, which
is assembled by the customer according to the provided
instructions. In both machinesABS is employed for the
production of specimens, ABSplus for the Dimension
Elite machine and Magnum 3904 ABS natural resin
for the RapMan machine.
Figure 1. The Dimension Elite (a) and RapMan (b) 3D

printers.
Figure 2. Schematic representation of the tensile specimen.
The two systems can be considered to some extent

complementary since they offer some distinct advantages. The Dimension Elite machine provides great
repeatability and accuracy that can be attributed to
the accumulated technical know-how regarding materials and process, as well as to the incorporation of
higher accuracy machine components (nozzle, drives,
belts etc.). In order to ensure this repeatability several
process parameters are embedded in the software of
the system and cannot be directly controlled by the
operator. This also makes the Dimension Elite system
quite simple to use and program, but does not leave
room for experimentation with process parameters.
Assembled 3D printers on the other hand are controlled by open-source software and give the operator
greater freedom in the selection of process parameters, but suffer from significant problems in accuracy
and repeatability (due to the nature of the components) and ease of use. Furthermore the absence of
support material deposition mechanism and software
makes the fabrication of complicated geometries quite
difficult and sometimes impossible.
In order to test to what extent variables in the
machine configuration and components affect the
mechanical behavior of parts a limited experimental investigation of tensile properties was conducted.
The tensile specimens were designed in CAD software
according to the ASTM D-412 C standard (Fig. 2) and
transferred in STL form to the pre-processing software
of the two machines.
In order to further investigate the effect of process parameters, besides the differences in the machine
configuration, the parameter of build orientation has
also been considered. Two alternative orientations
526
Figure 3. The two build orientations examined.
Figure 5. Load Travel graph of the tensile test specimens.
Figure 4. Deposition (filling) patterns employed in the

Dimension system.
were selected and they are shown in Figure 3. Hence,

a total of four specimens were fabricated, one for
each alternative orientation at each machine. Layer
thickness was 0.254 mm and filling density accounted
to Solid for the Dimension Elite machine and to
the equivalent 100% for the RapMan machine, so
that all specimens layers did not present any voids.
For the RapMan on-edge specimen a custom support
structure in ABS was also designed in order to ensure
accurate fabrication.
For an initial investigation of flexural properties
three-point bending test were performed. Simple slab
specimens were fabricated according to ISO 178: 2003
standard and with the following dimensions: 80 mm
10 mm 4 mm (length width thickness). Three
specimens were fabricated solely by the Dimension
Elite machine due to technical problems with the Rapman printer. In this experiment orientation and layer
thickness (0.254 mm) were considered fixed and the
specimens were fabricated with variable filling density, which obtained three values: Solid, Sparse-Low
Density and Sparse-High Density. These three values are the sole alternatives that are can be selected
through the controlling software (CatalystEX 4.1) of
the Dimension Elite machine. As mentioned earlier
Solidimplies full layer filling, while Sparseimplies
partial layer filling that can be of either low or high
density, as can be observed also in Figure 4. The material deposition follows a raster pattern of alternating
45 and 135 angle for successive layers. It should
be noted that sparse filling is performed only at the
eight middle layers (total number of layers is sixteen),

since both the first (bottom) four layers and the corresponding top four layers are deposited in full density
(solid).
For the determination of the mechanical properties a Zwick Z010 (10 kN) universal testing machine
has been used. Tensile testing has been performed
through extending of specimens at 5 mm/min speed,
while three-point bending was performed at 2 mm/min
speed.
4
In Figure 5 the results of tensile tests in the form

of force (load) exerted by the machine to the absolute extension (travel) of the specimen are presented.
The Rap Man specimens built in flat (a) and onedge (b) orientations are symbolized with Ra and Rb
respectively, while Da and Db are the corresponding
Dimension Elite specimens.
Some remarks that can be drawn from the graph are
the following:
The RapMan specimens fabricated in flat orientation (Ra ) exhibit less strength than the specimens
fabricated in on-edge orientation (Rb ).
The elastic behavior of RapMan specimens is not
significantly affected by the build orientation.
The orientation of the specimens fabricated with
the Dimension Elite seems to have a minor effect
on their tensile properties.
On-edge RapMan specimens (Rb ) exhibit comparable strength to that of Dimension specimens, in
terms of the afforded maximum load, but present a
more elastic behavior.
Possible causes for the observed differences
in elastic behavior between RapMan and Dimension
specimens could be the differences in the composition of the corresponding ABS, as well as the actual
527
Figure 6. Load Travel graph of the bending test specimens.
differences in storing and use of the corresponding

filament. It should be noted that filament for the
Dimension machine is provided in closed and sealed
cartridges, while RapMan filament in open reel which
is exposed to humidity and light effecta.
Regarding the effect of build orientation on tensile
strength it is interesting to note that it is noticeable
in the RapMan specimens, but almost insignificant
in Dimension specimens. The relatively low strength
of RapMan specimens in flat orientation could be
attributed to weaker bonding of the layers or the presence of void areas due to relatively high speed of
deposition or low ABS flow rate. Nevertheless, since
on-edge RapMan specimens seem to exhibit similar
maximum strength to that of the Dimension specimens, it could be assumed that detailed experimentation and fine-tuning of RapMan build parameters
could significantly improve the mechanical behavior
of the produced parts. In order to verify these observations and investigate possible causes for the observed
differences further more detailed experimentation is
required.
The results of bending tests are presented in
Figure 6 where it is obvious that the layer filling
density has an effect on the flexural elasticity and
strength of FDM specimens. The maximum strength is
exhibited by the specimens fabricated with Solidfilling, followed by specimens with Sparse-High filling,
while the worse in terms of strength are the specimens with Lowfilling. Since in this experimental test
only one parameter was considered variable it is relatively straightforward to conclude that for FDM parts
an increase in the density of the layers has a positive
effect on their maximum flexural strength.
CONCLUSIONS
The aim of the present study was to conduct a preliminary investigation of the effect of build parameters
on the mechanical properties of ABS parts fabricated by extrusion-type AM systems. According to

the experimental results, build orientation and filling density affect the mechanical behavior of specimens. The nature and extent of the effects seems
to vary depending on the specific mechanical properties investigated and the specific characteristics of
the machines and materials employed to fabricate
the specimens. An interesting observation that could
be made is that low-cost extrusion-type 3D printers
employing ABS seem to provide parts of comparable
strength to that of similar industrial systems, assuming
that proper build parameters have been selected. This
by no means implies that both systems are functionally
equivalent since the factors of accuracy, repeatability
and ease of use should be also considered. Furthermore, since the presented tests are quite limited, the
reported preliminary results should be verified by
further experimentation and detailed analysis.
Further investigation seems also necessary in
order to accurately identify the apparently complex
relationships between fabrication parameters (build
parameters, AM system characteristics, material properties) and the mechanical behavior of parts. This
investigation should be extended to the analysis of
other build parameters, such as layer thickness, and
should also involve the study of the combined effects
posed by different combinations of parameters.
REFERENCES
Ahn, S.-H., Montero, M., Odell, D., Roundy, S. & Wright, P.K.
2002. Anisotropic material properties of fused deposition
modelingABS. Rapid Prototyping Journal 8(4): 248257.
Bellini, A. & Geri, S. 2003. Mechanical characterization of
parts fabricated using fused deposition modeling. Rapid
Es-Said, O.S., Foyos, J., Noorani, R., Mendelson, M., Marloth, R. & Pregger, B.A. 2000. Effect of Layer Orientation
on Mechanical Properties of Rapid Prototyped Samples.
Materials and Manufacturing Processes 15(1): 107122.
Grimm, T. 2003. Fused Deposition Modeling: A Technology
Evaluation. Time Compression Technologies Magazine,
http://www.time-compression.com/articles/html/fused_
deposition.html.
Jones, R., Haufe, P., Sells, E., Iravani, P., Olliver, V., Palmer,
C., & Bowyer, A. 2011. RepRapthe replicating rapid
prototyper. Robotica 29: 177191.
Kim, G.D. & Oh, Y.T. 2008. A benchmark study on rapid
prototyping processes and machines: quantitative comparisons of mechanical properties, accuracy, roughness,
speed, and material cost. Proceedings of the Institution
of Mechanical Engineers, Part B: Journal of Engineering
Manufacture 222(2): 201215.
Lee, B.H., Abdullah, J. & Khan, Z.A. 2005. Optimization of
rapid prototyping parameters for production of flexible
ABS object. Journal of Materials Processing Technology
169(1): 5461.
Lee, C.S., Kim, S.G., Kim, H.J. & Ahn, S.H. 2007. Measurement of anisotropic compressive strength of rapid prototyping parts. Journal of Materials Processing Technology
187188: 627630.
Rodrguez, J.F., Thomas, J.P. & Renaud, J.E. 2001. Mechanical behavior of acrylonitrile butadiene styrene (ABS)
528
fused deposition materials. Experimental investigation.

Sood, A.K., Ohdar, R.K. & Mahapatra, S.S. 2010. Parametric appraisal of mechanical property of fused deposition
modelling processed parts. Materials & Design 31(1):
287295.
Sun, Q., Rizvi, G.M., Bellehumeur, C.T. & Gu, P. 2008.
Effect of processing conditions on the bonding quality
of FDM polymer filaments. Rapid Prototyping Journal
14(2): 7280.
Zhang, Y. & Chou, K. 2008. A parametric study of part

distortion in fused deposition modeling using three dimensional element analysis. Proceedings of the Institution of
Mechanical Engineers, Part B: Journal of Engineering
Ogando, J. 2007. Rapid Manufacturings Role in the Factory
of the Future. Design News August issue/13.
529

Porosity as a key to increase material properties of laser sintered parts

S. Rsenberg, L. Schmidt, H. Hosse & H.-J. Schmid
DMRC, University of Paderborn, Paderborn, Germany
Chair of Particle Technology, University of Paderborn, Paderborn, Germany
ABSTRACT: The energy density is an important parameter for the laser sintering process. By changing laser
power, scan velocity and hatch distance an influence on the properties of manufactured components, especially on
density, porosity and the structure of laser sintered parts will occur.The aim of this project is obtaining information
about porosity and to understand the creation of pores within sintered parts. By reducing the porosity, an increase
of mechanical properties is possible. The overall porosity is determined by using image analysis of thin sections
and X-ray computer tomography. The X-ray method is used to examine the 3D porous structure of test specimen.
It clearly reveals different porosities in one laser sintered part. A solid, shell with a thickness of approximately
0.5 mm is presented within the inner component volume. This phenomenon is observed by image analysis of
thin sections. An influence on mechanical properties is also examined.
1
1.1
STATE OF THE ART

Laser Sintering
Polymer Laser Sintering is an additive manufacturing

process based on polymer powder. With this technology it is possible to generate a product layer by layer
directly from a 3D file without using any tools. In this
work an EOSINT P390 laser sintering system, produced by EOS Company GmbH, Munich, Germany,
is considered.
In Figure 1 a schematic representation is shown.
After coating a layer of powder, the temperature within
the process chamber is increased close to the melting temperature. The laser beam applies the missing
energy for a local powder sintering by writing the
contour. After this process, the building platform is
lowered for another layer. The recoater applies new
material and the laser sintering process repeats.
Two supply bins (item 2 in Figure 1) for providing material are located to the left and right of the
process chamber. A fluidization compressed gas is
blasted within the supply bins incrementally to avoid
a aggregation of powder materials. The recoater (3) is
filled with material on both end positions by a dosage
notched roller, which defines the amount of powder.
The building platform (5) is lowered down by one layer
thickness and powder is coated by moving the recoater
from one side to the other. This system is possible to
produce layers that are 100 m, 120 m or 150 m
thick. Two overflow bins (4) collect excessive material. The surface is heated up by an infrared heating
system and the subsequent laser beam (1) is sintering
the powder locally.The depth of penetration of the laser
beam energy into the powder bed is greater than one
layer and subjacent layers are sintered together. These
steps will be repeated layer by layer until all parts are
generated. After a cooling down phase, the complete
Figure 1. Laser Sintering Process [Adam2010a].
product is separated from the powder cake. [Gebahrd

2007a].
1.2 Material
Polymer Laser Sintering is a high potential additive
manufacturing process. Theoretically it is possible to
treat all thermoplastic materials in powder conditions
[VDI3404 2009a]. Nylon 12 promises good properties
as related to powder distribution, flow ability, viscosity and thermal properties to manufacture products
with high requirements on strength. Approximately
8090% of all laser sintered parts are produced by
using nylon 12 [Schmid2010b]. In this study, material
from one batch of virgin material PA2200 (based on
nylon 12) produced by EOS Company is used. Grain
shape and grain size distribution are important factors for laser sintering material and measurements are
performed by a laser diffraction method. The average grain size is D0.5 = 55 m where the grain size
distribution ranges between 30 m and 100 m.
531
Figure 2. Particle Size Distribution measured by laser

diffraction.
Figure 4.
Table 1.
Combination of the exposure type mechanic.

Selected Parameter Variations.
Types of exposure
Parameter
unit
value
Speed
Laser Power
Beam delay
Energy density
mm/s
W
mm
mJ/mm2
3000
40
0.14
44.44
Speed
Laser Power
Beam delay
Energy density
mm/s
W
mm
mJ/mm2
1400
28
0.14
66.67
Speed
Laser Power
Beam delay
Energy density
mm/s
W
mm
mJ/mm2
2800
40
0.14
47.62
Filling
Figure 3. Cuboid with frame geometry.

Edges
2 TEST SETUP AND EXECUTION

This chapter describes the manufacturing of the used
test specimens and the methods used for the porosity
analysis.
Contour
2.1 Test specimens

For a determination of porosity with different methods
the following test specimens are created.
2.1.1 Cuboid
In order to determine the porosity and create thin section cuts, cuboids with an edge length of lk = 21 mm
are manufactured. They are enclosed by a case used
for storage and numbering (Figure 3).
Previous research papers discussing porosity inside
specimens discovered an outer solid layer using an
exposure setting type Mechanic. In this work, the
porosity based on the machine setting type of exposure
will be determined. Thus cubes with the same dimensions are manufactured using four different types of
exposure. Mechanic is the base type of exposure and
is a combination of the variations Edges, Filling and
Contour. A combination of these three exposure strategies is used as an attempt to determine the possible
optimal obtainable material properties. In figure 4 the
different strategies for a cube surface are displayed. At
the beginning of the exposure the variation Contour
(marked as inner dashed line) is used to sinter a rough
outline of the plane. This is done with a high laser
speed and power. The Filling (marked as inner surface) is done with a low laser power and a high speed.
Finally the Edges (marked as outer drawn through
line) of the part are trailed with slow speed and high
laser power to optimize the part properties and the
dimension accuracy.
The deviations represent a change of the parameters

speed, power and beam displacement (Table 1). The
hatch distance between the sinter lines is constant at
hs = 0,30 mm.
The different settings for Speed, Laser Power and
hatch distance have an effect on the energy density
calculated according to equation 1. It measures the
energy input at the surface of the powder bed [Kaddar
2010d].
With
EDm average energy density
PL
laser power
hs
hatch distance
vs
scanning velocity
The different energy densities for the variations Filling, Edges and Contour are presented in Figure 5. The
densities of Filling and Contour are similar at the same
level; the exposure type edges has a higher energy
density.
The left pillar is for filling, the central one for edges
and the right one for contour as an exposure strategy.
532
Figure 7. Outer pillar (left) and central pillar (right) for the
manufacturing of thin section cuts.
Figure 5. Energy densities for variations of the exposure
type (left pillar: filling; central pillar: edges; right pillar:
contour).
Figure 8. Polycut-S Microtome by Cambride Instruments

GmbH.
Figure 6. A pyramidal test specimen to determine the wall

thicknesses influence on the porosity. Thin section cuts are
made from the top to the bottom.
2.1.2 Pyramids
A pyramidal test specimen to determine the wall
thicknesses influence on the porosity is manufactured
with the exposure type Mechanic. It consists of a
square base with edge lengths of lk = 21 mm and the
height hG = 5 mm (for easy fixing into the microtome),
an overall height of h = 35 mm and is presented in
Figure 6.
2.1.3 Thin section cuts
For the microscopic analysis thin section cuts are used.
The thin and translucent compounds are created using
a microtome. In order to issue a thin section cut, it
is necessary to remove the excess pieces of the laser
sintering part. In this paper two slices are obtained
from each plastic cube and examined (Figure 7). For
each pillar, 3 measuring points are determined, which
are at height of 30 m thick and 1000 m by 2000 m.
Three thin section cuts are manufactured at each point,
so that the porosity for each test specimen is determined by 9 different positions. 3 positions are for the
outer solid layer and 6 positions are for the inner layers
which are more porous.
After fixing test specimens in the support of the

microtome, a blade creates a thin section cut with a
thickness of 30 m. It is mounted on a slide and is
pressed down with a cover glass (Figure 8).
The thin section cuts are performed by using a
Polycut-S Microtome produced by Cambridge Instruments GmbH. The start velocity was about 2 mm/s and
the cut thickness about 30 m with a hardness grade
of HK1.
3 APPLIED METHODS
Three different methods to measure porosity are used.
The results were compared
3.1
Computer tomography
The principle of the computer tomography consists

in general of the measurement of spatial distribution
of a physical property of the examined object and then
calculating images without overlap from the gathered
information. [Kalender 2006a].
The CT measures the weakened intensity I of the
x-rays behind the measurement object. The primary
intensity I0 has to be recorded since the weakened radiation of each beam is calculated from the focus of the
x-ray tube to the receiver. The attenuation coefficient
can be assumed for simple cases but can only be determined by tomographic imaging for inhomogeneous
objects. The projection value P (attenuation) is defined
533
Figure 9. Example of a thin-section 50 magnified, original (left) and binary image (right).
Figure 10. Example of a section used for the image analysis

(left; 30x magnification, right: 50 magnification).
as the natural logarithm of the primary intensity I0 to

the weakened intensity I [Kalender 2006a].
The CT-Data presented within this paper have been
created by the company CTM-do GmbH. The required
CT data specifications for this study are met by the
program VGStudio Max 2.1 from the company Volume Graphics where it is possible to conduct a defect
analysis. The overall porosity of the part is determined
in this process by calculating the total volume and the
imperfection volume. As part of the analysis the pore
size distribution of the specimen is created.
The voxel size in the current paper is about
27.6 m 27.6 m 27.6 m. Therefore the pores
volume is determined by the numbering of voxels.
All volumes with a minimum volume about 8 voxels
(Vmin = 0.1683 mm3 ) are detected.
porosity is the same as the average surface porosity, if

the pores are randomly arranged.
3.2
Scanning electron microscope
In the present paper the scanning electron microscope

produced by Phenom-World Company is used. It features a more durable CeB6 electron source and a
constant accelerating voltage of 5kV at 1nA.
3.3
Optical microscopy and image analysis
The examinations of sliced layers are done using an

optical-digital microscope produced by Keyence Company. The zoom lens VH-Z20 is used. It offers a
high resolution image for the standard magnifications
between 20 and 200. For this examination the 50x
magnification is used. The resolution of the digital
images is 1600 1200 pixels, one pixel corresponds
to a project area of 16.26 m2 .
The porosity is determined by digital image analysis of thin sections. The image is transferred from
a grayscale image with 256 grayscale levels into a
binary image. In order to transfer a grayscale image
into a binary image, the specification of a threshold
value ST is required, which defines for all pixels if
they appear in the binary image as black or white pixels. The threshold value ST is set at the beginning, and
is then applied to all images. Preliminary tests revealed
for ST = 144 is a good compliance between the original and the binary image. Figure 9 shows an exemplary
result of the image processing.
The surface porosity F of the test specimen is calculated, as shown in Equation 2, as ratio of black pixels
to the total pixels. Theoretically, the desired volume
In the present paper all image processing for the

determination of the porosity F are done with the freeware program ImageJ, version 1.42q [Image 2010e].
This program for particle analysis provides not only
the projected area AP , and offers also the option to
determine the form factors and other geometrical characteristics of the particles as well, such as roundness
r, Feret diameter dF and others.
For the image analysis microscopic images 50
magnified are used. Figure 10 shows a 30 magnification on the left side. The red marks show the area
of the thin-section that is used for the image analysis (50 magnification of the right side). Because of
the 50 magnified images, a representative area of
the thin-section can be captured. At the same time the
border area can be excluded from the measurement,
which would not be appropriate for the measurement
due to the preparation technique using a handsaw. For
a statistically representative result, three thin-sections
are surveyed for each operating point and summarized
into one average value.
The calibration of the conversion factor FP between
pixel and actual area AP is done by means of a scale
for microscopes.
4
4.1
RESULTS & CONCLUSIONS

Porosity in dependence of different exposure
types
Analysis of the data indicates that the porosity of laser

sintered parts do not seem to be affected significantly
by the type of exposure. The fluctuations of the results
are rather small and thus can be neglected. Figure 11
shows the porosity calculated using the computer
tomography for the four different types of exposure.
These results are consistent with the image analysis.
The area porosity F stays constant for varying types
of exposure.
However, it was discovered, that laser sintered parts
consisted of a porous core and an outer solid layer
which is non-porous. The solid layer detected during
the tests, which surrounds the edge of every part can be
534
Figure 11. Porosity for varying types of machine exposure

settings.
Figure 13. Comparison of the porosity for thin section cuts

(left pillar) and CT (right pillar).
Figure 12. Outer Solid Layer for varying types of exposure.
Figure 14. Pore deformation.
found for each type of exposure setting and has a constant thickness of approximately 500 m. Figure 12
shows this phenomena for all four examined settings
which is to be independent from the type of exposure.
The energy density of the outer layer shows a constant
thickness and porosity.
The overall porosity and the solid layer is unaffected
by varying the energy density.
microscope. For the purpose of finding proof for this

underestimation a thin section cut is created from a
pillar (see Figure 7) for laser sintered test specimens.
The resulting image is compared with an image of the
bottom side using a SEM which shows deformation of
the pores (Figure 14). Further on the cutting direction
and the pore deformation is visible as well.
It appears that no tearing of particles occur with the
thin section method, since each pore that is observed
in the thin section cut can also be found in the SEM
images. If tearing of the particles occurs, a pore could
be seen in the SEM image but not in the thin section
cut. Material is drawn into pores and provides therefore
a variation of pores. This explains the underestimation
of the pores and the differences of the results regarding computer tomography. Figure 18 is a sample for
all other exposure types but the deformations can be
detected for all thin section cuts regarding to different
exposure types which attend a quantitative statement
for this method.
Figure 15 presents a sample for the deformations
of the pores for the exposure type Edges. The cutting direction of the blade can be identified. It is cut
from top to bottom in this thin section. The top SEM
image shows two features. Both pores on the top left
are highly deformed, which can be determined from
the different pores of the thin section cut and the SEM
4.2
Challenges by comparison of thin section cuts

and Computer Tomography
Because of the differences between the area porosity F and the overall porosity , the accuracy of
the thin section method to determine the porosity is
examined. The results from chapter 4.1 clearly show
that the porosity determined by means of thin section cuts is lower than the porosity determined by
means of CT (figure 13). For each area porosity F
an underestimation of the pores can be detected.
The computer tomography is a non-destructive test
procedure and allows a very accurate determination of
the part properties. The conclusion is that the determination of the porosity by means of thin section cuts is
biased.
In order to find an explanation for this bias, the thin
section method is examined using a scanning electron
535
Figure 17. Pore sizes as a function of different exposure

strategies.
Figure 15. Deformation of the pores for the exposure type

edges.
Figure 18. Porosity for varying wall thicknesses.
Figure 16. from top left to bottom right: Laser Power PL =

20, 30, 40 and 50 W, respectively Hatch distance and Scanning
velocity identical, Magnification 200.
image. The pores on the right side show the undervaluation of the pores since material is drawn into
them.
4.3
to reduce the influence of different exposure strategies

which are used for Mechanic. Filling and Contour have
a similar energy density calculated by different scanning velocities and laser power. Edges have a much
higher energy density (see Figure 5).
The accumulated rate Q shows differences regarding the slope of the curve. For smaller volumes of pores
(<0,01 mm3 ) the curve slop for Edges is flatter than
for Filling and Contour. Further on the numbering of
pores is shown. Edges depicts fewer small pores than
all other exposure strategies but therefore the number
of bigger pores increases.
Pore size in dependence of different energy

densities
Figure 16 shows four thin sections of specimens with

different building parameters at a 200x magnification,
with increasing energy density from top left to bottom
right. At a laser power of PL = 20 W (corresponding
energy density of EDm = 22.22 kJ/mm2 ) not melted
particles can be clearly seen. The hollows run along the
sintering lines and indicate insufficient connection of
the single lines because of the low energy density. Even
for an energy density of EDm = 33.33 kJ/mm2 a few
insufficiently particles can still be detected. Bigger
pores show a roughly round shape, the sintering lines
can no longer be detected. With an increasing energy
density there seem to be fewer pores, but which have
a bigger cross section. Thin section cuts are showing
only a little extract, so that a pores analysis determined
by computer tomography is necessary to substantiate
this assumption.
In Figure 17 a pore size analysis as a function of
different exposure strategies is shown. The first step is
4.4 Porosity depending on the cross sectional area

The porosity based on the cross sectional area is
determined and has been analyzed. For this purpose
the pyramids described in Section 2.1.2 and the thin
section method are compared.
Figure 18 presents the determined porosity for varying wall thicknesses.The pyramids have a thick coating
layer just like the cubes. This has a strong influence on the overall porosity for low wall thicknesses.
The porosity for wall thicknesses of up to 3.5 mm is
below 1%. With a wall thickness of 5 mm a constant
porosity can be observed for the parts, since the ratio
of the coating layer thickness to the total thickness
decreases.
An absolute statement regarding the overall porosity cannot be made due to the reasons described in
chapter 4.2. A quantitative statement regarding the
porosity as a function of the cross-sectional area can be
made.
536
Figure 19. Elongation at break as a function of energy

density (by varying laser power).
It is clearly shown that the overall porosity is

reduced for thin-walled laser-sintered parts. However,
completely non-porous parts are not achievable. Further design guidelines in the area of laser sintering can
be evaluated with specific recesses for thick-walled
parts.
4.5
Influence on mechanical properties
The influence of the numbering of pores as well as the

pore size by changing energy density is shown. Further
on the energy density has an influence on mechanical
properties, representative for elongation at break determined by tensile tests and test specimen orientated in
z-direction (Figure 19).
By increasing the energy density, the elongation at
break also increases, as shown in Figure 19.
Tensile tests are heavily dependent on the type
of test specimen which has to be free of damages,
shrink marks and other blemishes [Hellerich2001a].
In contrast to injection molded parts test specimen
manufactured by using laser sintering have an inferiority quality regarding to surface finish and porosity.
Because of this the pore size as well as the numbering
of pores has an influence on mechanical properties.
Also there might be other influencing factors like temperature distribution, powder aging, etc. which have to
be determined.
5
CONCLUSIONS
In the present paper the porosity of laser sintering

part is determined and the analysis techniques are
discussed. In order to determine the porosity as a function of the type of exposure, the exposure strategy
Mechanic, a combination of the three different exposure strategies Edges, Filling and Contour, is analyzed.
Their distinctive features are laser power and speed,
which leads to different energy densities. For each type
of exposure a cube is manufactured via laser sintering
process.
For the determination and confirmation of the
porosity a non-destructive examination procedure
computer tomography as well as a thin section method

is used. By using the thin section method it is important examine the 30 m thin sections by using a
light microscope for a digital image. By means of
a computer program the image is transformed into
a binary image and an analysis of the area porosity
is done.
The results of the CT analysis were that the porosity
of laser sintering parts is unaffected by the type of the
four given exposure strategies and that it is constant at
about 5.5% to 6%. The computer tomography allows
making a clear statement regarding the pore size and
the numbering of pores. At low energy densities like
Filling and Contour the distribution can be described
as exponential. The number of the small pores is the
highest. The higher energy density for the exposure
type Edges results in a shift of pore size distribution. The number of small pores is lower than for other
exposure types. After the peak at a small pore size the
number decreases rapidly reaches a constant level for a
moment and after this plateau decreases exponentially
with increasing pore size.
The examination of the CT file and the thin section cuts revealed a solid layer for all test specimens
which has contrary to the center of the part a very low
porosity. This external shell can be found on every part
surface, which was determined by an analysis using the
computer tomography. The solid layer exists for every
type of exposure and energy density and has a constant
thickness of about 500 m.
Further on the examination of the porosity shows
that the value determined by the thin section method
is below the value determined by the CT. Therefore the thin section method is examined regarding
its accuracy. During the cutting process the pores
are deformed, which results in their undervaluation.
An examination with a scanning electron microscope
revealed that the pores are partially displaced. By
means of thin section cuts only a quantitative statement
regarding the porosity can be made, but no absolute
statement.
The porosity as a function of the cross-sectional area
is determined using pyramids. Since it was proofed
that the type of exposure has no effect on the result
those pyramids are manufactured using the exposure
type Mechanic. In order to make a quantitative statement the thin section method is used. Because of
the thick solid layer the porosity is below 1% for
wall thicknesses up to 3 mm. at a wall thickness of
5 mm and higher a constant porosity is reached. Using
these results a design rule can be developed that minimizes the porosity of thin-walled part. An exclusion
of the porosity for laser sintering parts with this is not
possible.
As the current paper confirm, the exposure strategy has no influence on the porosity. Another theory
which have to be examined is the cooling-down rate. In
additional works these influences have to be detected.
Regarding the mechanical properties its influence has
to be determined as well. Further on the influence of
porosity on mechanical properties has to determine
537
more exactly by a consideration of all other parameters

as well. The influence of energy density and therefore
of porosity on mechanical properties is tearing in this
paper. Further experiments have to be accomplished
as well.
REFERENCES
Hellerich 2001a: Hellerich, W.; Harsch, G.; Haenle, S.: Werkstofffhrer Kunststoffe Eigenschaften Prfungen
Kennwerte, 8. Auflage, Hanser Verlag, 2001.
Kalender 2006a: Kalender, W.A.: Computertomographie,
Publicis Corporate Publishing, 2006.
Gebhard 2007a: Gebhard, A. 2007a. Generative
Fertigungsverfahren Rapid Prototyping, Rapid Tooling,
Rapid Manufacturing. Mnchen: Carl Hanser Verlag.
VDI3404 2009a: VDI Richtlinie 3404 Generative Fertigungsverfahren Rapid-Technologien (Rapid Prototyping)
Grundlagen, Begriffe, Qualittskenngren, Liefervereinbarungen, Verein Deutscher Ingenieure; Dezember
2009.
Drummer 2009b: Drummer, D; Khnlein, F.; Rietzel, D.:
Selektives Lasersintern von teilkristallinen Thermoplasten; RTejournal; 2009; Abruf 02.08.2010;URL:http://
www.rtejournal.de/ausgabe6/2214/
Adam 2010a: Adam, G. Konstruktionsregeln fr additive

Fertigung in Lehre und Forschung, Berliner Kreis, Ausgabe 15, 2010, Seiten 2223, Wirtschaftliches Forum fr
Produktentwickelung e.V.
Schmid 2010b: Schmid, M.: Kunststoffe fr Selektives
Lasersintern; plastics.now!; Ausgabe Juni 2010.
EP 2010c: Vorrichtung und Verfahren zum schichtweisen
Herstellen eines dreidimensionalen Objekts aus einem
pulverfrmigen Aufbaumaterial; Europische Patentschrift; EP 2 026 952 B1; Patenterteilung 29.09.2010.
Kaddar 2010d: Kaddar, W.: Die generative Fertigung mittels
Laser-Sintern: Scanstrategien, Einflsse verschiedener
Prozessparameter auf die mechanischen und optischen
Eigenschaften beim LS von Thermoplasten und deren
Nachbearbeitungsmglichkeiten; Dissertation Universitt Duisburg-Essen; November 2010.
Image 2010e: ImageJ Image Processing and Analysis
in Java; Abruf: 12.02.2010; URL: http://rsbweb.nih.
gov/ij/
HTW 2011a:http://www.htwaalen.de/sgv/Materialographie/
forschng/wrkchar/rstrmikr/rstrmikr1.htm, Letzter Zugriff
17.03.2011.
Malvern 2011b: http://www.malvern.de/LabGer/technology/
laser_diffraction/particle_sizing.htm, Letzter Zugriff
30.05.2011.
538

Software tools for rapid Prototype as Design

A. Withell, O. Diegel & S. Reay
Auckland University of Technology, Centre for Rapid Product Development, Auckland, New Zealand
ABSTRACT: Design teams are expected to produce physical prototypes that demonstrate the working
principles of their designs. These projects may involve multiple areas of technology, such as industrial
design, electronics, mechanical engineering, software, and even marketing. The integration of physical, threedimensional prototypes into the new product development (NPD) process, i.e. Prototype as Designis an effective
way to evaluate form, ideas, testing function, and for optimising products for intended users. The advent of
the latest additive manufacturing and CAD/CAM technologies has transformed this process into a Rapid Prototype as Design (RPaD) methodology. This paper describes the RPaD methodology and presents a case study of
student product design projects, in which RPaD was used as a key design methodology. It also presents a software
toolbox that is used as a tool to facilitate the process, and use of the toolbox is demonstrated in the case studies.
INTRODUCTION
A prototype is an early model built to test an idea or

product so that it can be learned from. Prototyping
is one of the oldest product development techniques
and has been used by artisans for centuries. Artisans
created prototypes of their ideas, to ensure that they
worked, before making the planned primary artefact.
Prototyping falls into a number of different types,
ranging from proof-of-principle prototypes, to demonstrate some mechanical principle or basic idea, to
form study prototypes, used to explore shapes and
aesthetics, to user-experience and ergonomic prototypes, to test principles from a user perspective, to
functional models, to explore aspects of a product from
a functional point of view. Many of these prototyping
methods can take both virtual and physical forms in
which an idea might be tested on paper, or on a computer, all the way through to different levels of physical
models, ranging from simple card or foam models, to
elaborate CNC machined or rapid-prototyped models.
The integration of physical, three-dimensional prototypes into the new product development (NPD)
process, i.e. Prototype as Design has always been an
essential and effective way for evaluating form, ideas,
testing function of individual parts and for optimising products for intended users. It is, for example,
often impossible to precisely specify functional and
user requirements at the front end of the NPD process.
According to Mulenburg (2004), even if possible, it
may be undesirable to do so. Further to this Singh &
Vijayaraghavan (2001)], go on to discuss that this often
makes a strategy where the use of prototype as a
design critical for success, as it is a highly interactive, integrated process that allows multiple iterations
of complex aspects of a product to be quickly evaluated
and adapted into a functioning whole.
Physical prototypes play an essential role in NPD as

they are a means of demonstrating function, scale and
realism in a way that paper drawings and CAD models cannot. According to Broek, Sleijffers, Horvath, &
Lennings (2000), the translation from two dimensional
to three dimensional representations is a key stage
in NPD. The progression of prototypes can be seen
as going from two dimensional to three dimensional
on-screen, to three dimensional physical models. However there are large differences in perception between
a user seeing a CAD model and then seeing and, possibly more importantly, feeling a real physical working
model. According to Krar and Gill (2003), the additional tactile, haptic and true three-dimensional perception produces two completely different responses
in the user.
Prototypes are also useful in producing one-of-akind projects by eliminating some of the formality
of the traditional stage-gate engineering design processes. This philosophy has led to the development
of a design methodology referred to as Prototype
as Design. The traditional Prototype as Design technique, as used by the NASAs Ames Research Center
(Mulenburg, 2004), is very useful in creating unique,
one-of-a-kind research hardware for small, high-risk
projects. It is a methodology that encourages a much
more parallel design process than can sometimes
be achieved with conventional stage-gate techniques
(Fig. 1).
Where this methodology differentiates from conventional prototyping is that the final product is, in
fact, still seen as a prototype, that has been created
through a number of preliminary prototype stages.
Though the Ames Research Center uses the methodology for one-of-a-kind products, many eventually
mass-manufactured products can be thought of as
a one-of-a-kind product during their development
539
Figure 1. Comparison of Traditional and RPaD Processes.
process. This, with the addition of a manufacturing

consideration filter, can prove to be an effective technique for developing conventional products faster, and
with results that can better satisfy user needs.
With the increasingly complex nature of contemporary products, and the sophisticated expectations of
buyers and users, the use of prototype as design to
optimize products during the design process is becoming ever more important. The advent of the latest rapid
prototyping, computer aided engineering (CAE) and
manufacturing (CAM) technologies has added a new
dimension to the traditional Prototype as Design
methodology. The new generation of software and
hardware tools now allows engineers to perform complex finite element analysis (FEA) on their products,
to test for interference problems and thermal or structural problems, or to simulate how plastic may flow
through an injection molding tool during manufacturing. It is now evolving into a Rapid Prototype as
Design methodology.
RAPID PROTOTYPE AS DESIGN
The overall generic design process can now be

described as follows: Initial conceptual sketches
are still usually undertaken in 2D, both on paper and
on the computer. More advanced conceptual design
and engineering design models are then usually produced using 3D CAD software. This produces a virtual
model that can be rotated, zoomed in on, measured
and manipulated on-screen. From this 3D computer
model, a physical rapid prototype can be produced.
Traditionally, the only way to produce a real, physical
model was to either use a subtractive technology such
as Computer Numerically Controlled (CNC) machining or to produce expensive tooling into which the part
could be injection molded. Both these methods can be
time consuming and expensive.
The latest generation of rapid prototyping technologies such as stereolithography (SLA), Selective Laser
Sintering and Melting (SLS/SLM), Fused Deposition
Modelling (FDM) and 3D printing now allow physical prototypes to be produced within hours rather than
days (Chua & Leong, 2003).
The rapid prototyping process begins by taking a 3D
computer generated file and slicing it into thin slices
(commonly ranging from a few microns to 0.25 mm
per slice depending on the technology used). The rapid
prototyping machine then builds the model one slice at
a time, with each subsequent slice being built directly
on the previous one (Wohlers, 2009). Chua and Leong
(2003) present a good outline of how the technologies
may differ for each method in terms of the materials
they use to build the part, and the process used for
creating each slice of the model.
Some of the rapid prototyping processes which,
until recently, were only able to make plastic-like parts,
are now producing metal parts in a variety of metals
including aluminum, titanium, and stainless steel, and
ceramics. Not only is the choice of materials and processes increasing, but the last few years have seen
a significant reduction in the cost of these technologies. Systems are now also available for simulating the
behaviour and performance of electronic circuits, and
for rapid prototyping complex printed circuit boards
(PCBs).
These technologies mean that it is now possible to
construct highly advanced virtual prototypes, and then
fully working physical prototypes, including mechanical hardware, software and electronics, almost as fast
as they are designed, thus allowing many more iterations of a design within a shorter timeframe. This, in
turn, allows for products that are even better suited to
their intended users in even shorter times.
It is important to remember that a product prototype
includes more than just its mechanical parts. Many
products also include electronic and software components which must also be prototyped as part of the
process. It is vital to understand that the mechanical,
electronic and software systems are closely related to
each other and that the design of one should therefore
affect the others. This is why it is so important that all
disciplines work as a single unit rather than as simple
parallel activities. It is also vital to remember that the
RPaD process, as described above, is not intended to
be a linear or sequential process. This distinction can
easily get lost when trying to describe a process on
paper. To operate effectively, the RPaD process must
operate as a parallel tasking project in which the prototyping happens in parallel to any, and all, stages of
the process.
DESIGN METHODS TOOLBOX
To facilitate the use of RPaD, and a number of other

design processes and tools, an online software package was created to allow students and practioners easy
access to the tools as well as a range of examples and
applications of the tools.
The Design Methods Toolbox presents a conceptual model of a six stage, product design and
540
development process (see fig 1.). Within each of the

stages, six key design methods are presented (see
fig. 2). The goal is to provide a resource that gives
undergraduate product design students and practioners a simplified, but solid and practical framework for
learning about, and applying the design process in their
practical design work. The methods in the Design
Methods Toolbox are currently delivered to the students as part of the design studio teaching programme
through lectures and practical hands-on workshops,
while the online version of the software gives them
Figure 2.
1. Introduction and background to the method

2. Key steps in use the of the method
3. Examples based on best practices by other students.
4. Variations links to other related methods
5. References
Front page of the Design Tools Online software.
Figure 3. Conceptual model of the design process.

Table 1.
easy access to refresh themselves in the use of the

methods.
To give the students a good grounding in each design
method the following is included:
Figure 4. Introduction Page.
Design methods provided by the toolbox.
1.0
Project
Planning
2.0
Research
3.0
Creativity &
Exploration
4.0
Concept
Development
5.0
Concept
Testing
4.1 3D
Models
4.2 3D
Prototypes
4.3 Computer
Aided Design
(CAD)
4.4 Design for
the Environment
(DFE)
4.5 Rapid
Prototyping
as Design
4.6 Matrix
Selection
5.1 Task
Analysis
5.2 In Situ
Placement
5.3 CAD
Simulation
1.1 Design
Process
1.2 Goal
Setting
1.3 SWOT
Analysis
2.1 Lit
Review
2.2 Survey
2.3
Interviews
3.2 Role
Playing
3.2 Mind
Mapping
3.3
Brainstorming
1.4 Gantt
Chats
2.4
Observation
3.4 Lotus
Blossom
1.5 PESTLE
2.5 Trends
Analysis
1.6 Project
Brief
2.6 Personas
3.5 Six
Thinking
Hats
3.6 Concept
Ideation
541
5.4 Life Cycle

Analysis
(LCA)
5.5 Product
Costing
5.6 IP
Protection
6.0
Communication
6.1 Reports
6.2 Story
Boards
6.3
Presentation
Boards
6.4
Photography
6.5 Moving
Image
6.6 Reflective
Practice
6. Key links to more detailed resources and examples

are provided for students independently further
explore
7. Templates and/or other resources to aid in use.
For example the Rapid Prototyping as Design
(RPaD) method is presented as a key part of the
Concept Development phase of the product design
and development process. The basic concepts of rapid
prototyping are communicated through simple diagrams, each technology is then described, and most
importantly examples of best practice student work are
included. Figure 4 shows an example from the RPaD
method pages.
learn to understand and independently select the most

appropriate prototyping methods for a given context.
To achieve this at AUT, a number of key strategies
are used to teach and integrate RPaD into the programme. This includes the development and delivery
of key lectures and discussions to engage students in
a discourse around the broader issues of the use of
prototyping, hands-on workshops with a variety of prototyping processes from low-tech to high-tech, case
studies of professional projects, physical examples
previous students project outcomes, and the use of
other resources such as videos, site visits and access
to online resources.
To illustrate the use of RPaD at AUT, a number of
successful case studies are presented.
PRODUCT DESIGN AT AUT

6
The three-year undergraduate product design programme at AUT University is a relatively new programme, developed in 2007 and launched with the
first intake of students in 2008. In 2011 the programme has 80 students across the three years and
8 postgraduate students. The programme is centred on
project/problem-based learning in which students are
given studio space to work in, and access to workshops and prototyping facilities. Over the three years
of undergraduate study the students work through a
number of projects ranging from short i.e. two week,
to full semester i.e. twelve week projects. Expectations range from conceptual outcomes i.e. blue sky
projects through to product outcomes as close to
realization as possible.
While the development of a new academic programme provides many organisational and operational
challenges, it also presents a unique opportunity to
develop new approaches to teaching and learning without the constraints of institutional history and tradition.
An innovative pedagogical approach to product design
is currently being developed in the product design programme at AUT that focuses on integrating emerging,
contemporary design methodologies and processes.
The concept of Rapid Prototype as Design is seen as
a key methodology for the product design programme.
As part of their studio programme students were asked

to design a set of medical products with the goal to
meet/exceed the needs of intended key users and
to improve the overall experience of using the product. The students had twelve weeks in which to design
the product from research, initial ideas to proof-ofconcept prototype and put on an exhibition and create
a product plan and report.
In order to achieve this goal within such a tight
time-frame, the students were advised to specifically
use RPaD as the key design methodology. In addition the students were asked to use clearly identified
prototyping methods that best suited their idea or concept context. In some case this was undertaken through
virtual prototyping and the use of CAD models. With
other students, relatively crude but quick card or foam
mock-ups were used. When more complex ideas were
being tested, students used laser cut or rapid prototyped models. They were strongly encouraged to use
their prototypes as a way of thinking about the problems they needed to overcome to reach their project
goals. Overall students utilised a very wide range of
prototyping processes.
6.1
INTEGRATING RAPID PROTOTYPE

AS DESIGN
Traditionally product design schools have focused on

the specific use of drawing and CAD as the primary
creative methodologies, and a model or prototype was
something that was usually created at the end, not
something generated throughout the process. Given
the traditional use of design methodologies, it is
essential for the successful integration of new methodologies such as Rapid Prototype as Design to get
student buy-in and a culture change away from the
traditional approaches and to a more hands-on process,
using quick, effective and many generations of prototypes to test and evaluate ideas. Students must also
CASE STUDIES
Moon-boot
The goal of this project was to design and develop

an innovative moon-boot cast for people with broken
ankles. The students first undertook a detailed analysis of key users and, from the information gathered,
identified that current moon-boots were unwieldy and
bigger than they needed to be for a large part of a users
convalescence. After brainstorming to generate a number of design concepts, most of which were quickly
prototyped, they came to a final design for a modular moon-boot in which sections could be removed as
the user progressed through their recover, thus making
the user more comfortable and therefore more likely
to recover faster.
From the start of the project, the students tested all
of their concepts and ideas with prototypes. Relatively
crude card prototypes where first used to visualize
542
Figure 5. Examples of student prototypes used in the

project.
Figure 6. CAD Model and Prototype of Moon-Boot.
initial ideas, then after starting to virtually prototype in

CAD and physically prototype in parallel, the students
progressed to laser-cut polypropylene prototypes leading to final, 3D printed plastic prototypes produced on
a Dimension FDM machine.
One of the challenges faced by the students was in
learning to identify which method of prototyping was
most effective in achieving the purposes of a particular
challenge, be it communicating and idea or testing an
engineering or manufacturing principle.
The prototype moon-boot shown in figure 3 is comprised of a mix of plastic 3D printing and laser cutting
and is a fully functional proof-of-concept model.
6.2 Ambulatory blood pressure monitor

In this project an ambulatory blood pressure monitor
was redesigned and improved. An ambulatory blood
pressure monitor is a blood pressure monitor that is
worn continuously by the patient for 24 hours and
which takes readings at preset intervals. Current models are worn on the belt and have air pipes leading up
to the cuff which is wrapped around the upper arm.
This makes it difficult for patients to wear at night as
the tubes get in the way, and can stress patients to the
extent of affecting their blood pressure.
After prototyping a number of different concepts,
the team settled on a design in which the entire monitor was worn on the arm. The electronics and pump
Figure 7.
Monitor.
Prototype of Ambulatory Blood Pressure
became an integral part of the cuff. One of the prototyping methods used by this particular team was
in the reuse of existing components, a very useful prototyping method that often gets ignored. All
543
Figure 8. Prototype of Respiratory Humidification System.
internal components, pump, solenoid, circuit boards

were reused from an existing blood pressure monitor. This reuse of components drastically shortened the
teams development time on the technology front, and
relatively easily allowed them to construct a working
prototype.
6.3
Respiratory humidification system
This project was focused on a respiratory humidification system in conjunction with Fisher and Paykel
Healthcare, a major NZ manufacturer of healthcare
products. The student was asked to redesign an existing product based on the findings of extensive user
research in hospitals. The focus was to improve the
experience of both staff and patients and to create
the next generation of the product. In this case the
student also challenged the existing humidification
technology and proposed an alternative and potentially radical method of both generating and recycling
humid air.
The project involved prototypes in the form of early
concept form studies in foam, through to working prototypes for the development and testing of systems to
produce humidity. This involved setting up alternative
methods for generating humidity, creating air flow and
the testing and comparing of each of the methods. A
number of presentations of the prototypes were made
to the client and to users for feedback. Once the overall form factor was developed and the complete system
designed, the final form was refined using CAD. This
culminated in the production of a final rapid prototype
using FDM technology for high level communication
and display purposes.
CONCLUSIONS
As newer virtual and physical rapid prototyping technologies emerge, the way in which they are used to
more effectively manage the NPD process must evolve
in tandem. Further to this, the traditional NPD processes must evolve into Rapid New Product Development processes. The combination of rapid prototyping
technologies, not only in the mechanical area, but also
in the electronic and software areas can be used to
reduce the product development cycle if they are used
effectively. Not only can the project time be reduced,
but more desirable products can often eventuate as
more design iterations can be gone through, thus more
closely meeting the needs of the users.
It is essential that design and engineering programmes also engage students with new and emerging
design methodologies and processes such as RPaD,
and that they are embedded deeply into programme
curriculums.
This paper has presented how at AUT, the product design programme is integrating RPaD into the
teaching and learning programme, and has showcased
a number of student design projects that have utalised
RPaD as the core design methodology. All student
teams came up with innovative solutions which were
not only optimised for the users needs, but were also
relatively easy to manufacture. Most of the teams created between twenty to thirty prototype iterations for
their projects (ranging from crude cardboard and foam
concept models, to CAD prototypes, to highly polished plastic or laser-cut final product prototypes),
which allowed them to develop their ideas in an effective and efficient manner. As they prototyped every
idea they had, the idea was automatically tested for
validity through the prototype. The AUT students have
demonstrated that RPaD, as an excellent emerging professional NPD methodology, can also be effectively
utilised by design students to develop innovative, new
products.
REFERENCES
Broek, J. J., Sleijffers, W., Horvath, I., & Lennings, A. F.
(2000). Using physical models in design Paper presented
at the CAID & CACD 2000, Hong Kong. from.
Chua, C. K., & Leong, K. F. (2003). Rapid Prototyping:
Principles and Applications. (2nd ed). Singapore: World
Scientific Publishing Co.
Krar, S., & Gill, A. (2003). Exploring Advance Manufacturing Technology. New York: Industrial Press Inc.
Mulenburg, G. (2004). Design by Prototype: Examples
from the National Aeronautics and Space Administration.
International Association for Management of Technology.
Singh, D. P. K., & Vijayaraghavan, R. (2001). Applications
of Op-timization in Metal Casting. International Journal
of Vehicle Design, 25(1), pp. 4052.
Wohlers, T. (2009). Worldwide progress report on the rapid
prototyping, tooling, and manufacturing state of the industry Wohlers Associates.
544

Direct Manufacturing Design Rules

D. Zimmer & G. Adam
Chair for design and drive technology, Paderborn, North Rhine Westphalia, Germany
Direct Manufacturing Research Center, Paderborn, North Rhine Westphalia, Germany
ABSTRACT: Design rules for additive manufacturing processes support the manufacturing orientated designing for additive manufactured products. Thus they contribute an important influence on the acceptance of these
processes at a broad spectrum of users. The research project Direct Manufacturing Design Rules (DMDR)
was initiated by the chair for design and drive technology (KAt, University of Paderborn) and is handled in
collaboration with the Direct Manufacturing Research Centre (DMRC, University of Paderborn). Design rules
for additive manufacturing processes will be developed. The laser sintering, laser melting and fused deposition
modeling processes are considered. Project seminars support the practice orientated review of the developed
design rules. As well further research needs can be identified during the project seminars.
1
INTRODUCTION
Design rules for additive manufacturing processes are

important for the acceptance of these technologies and
mainly required by the industry. Furthermore design
rules are necessary to provide and teach the new design
freedoms to users of these technologies as well as
to students. But at time there are no comprehensive
design rules given which could be used for training
and teaching. Therefore the project Direct Manufacturing Design Rules (DMDR) has the aim to develop
design rules for additive manufacturing processes and
to publish them to a broad spectrum of users from
scientific and industry as well as to students. Laser
sintering, laser melting and fused deposition modeling
processes were examined at first to determine which
process parameters take influence on the design of
additive manufactured products.
Parallel standard elements were defined which
often reoccur by designing technical products. Each
standard element can be characterized by different
properties and the properties can ingest different
values.
By manufacturing these standard elements using
different additive manufacturing processes, design
rules are to be developed. These design rules provide recommendations to design engineers how standard element properties should be expanded for a
manufacturing orientated design.
In the further proceeding of the project DMDR
additive manufacturing processes will be compared
with conventional manufacturing processes like molding, machining or forming. Differences in geometry,
surface, value of needed energy and emission will be
examined.
The priority of the DMDR project is the transfer of knowledge by published literature, lectures
and seminars. In addition the courses offered at the
University of Paderborn will be expanded by a series

of seminars on the subject additive manufacturing and
by the expansion of established lecture contents.
2
MANUFACTURING PROCESSES
In the beginning of the DMDR project additive manufacturing processes were analyzed to indicate process
parameters and characteristics which influence the
design of additive manufactured parts. The focus is on
the parameters which take influence on the geometry
and manufacturability, while parameters which influence the mechanical properties are not considered at
this place.
Additive manufacturing processes have the common characteristic that parts are created of firmly
bonded part layers. All part layers have the same thickness and are shaped into the building plane which is
parallel to the building platform. Thus part contours
which are parallel to the building platform can be
shaped continuously and contours orthogonally to the
building platform are influenced by the stair stepping
effect (Gebhardt 2007a).
Depending on the additive manufacturing process,
part layers consist of hedge lines or filaments. Hedge
lines are created by an energy beam (laser or electron)
which scans over initial powder material. The widths
of hedge lines or filaments influence the design and
are important processing parameters.
In addition to the previously mentioned characteristics, hedge lines and filaments need to be underpinned,
at least partially, by support material. Support materials have the purpose to create a stable platform for the
material above and to conduct heat out of the actually
manufactured part layer. Depending on the manufacturing process different, types of support material can
be used. Low temperature powder based processes
545
like the plastic laser sintering can use the surrounding

powder in its initial state or the partially completed sintered part. Powder based processes with a much higher
temperature like the metal laser melting have to transfer the heat out of the actually manufactured part layer.
Because the heat conducting coefficient of the powder
is too small in its initial state it cannot be used as support material, only melted solid structures can conduct
enough heat. Thus hedge lines need to be supported by
the partially completed part or by additionally manufactured solid support structures which need to be
removed after manufacturing. Additive manufacturing processes which lay down filaments can use the
partially completed part or additionally manufactured
structures of different materials as support material,
likewise these structures have to be removed after
manufacturing.
All powder based processes have a common method
of support material (powder and/or solid support
structures) which needs to be removed after manufacturing. Whereas the filament based processes
are able to manufacture parts without any support
material.
Analyses of additive manufacturing processes indicate a relationship between the design and process
parameters. Layer thickness influences the part dimensions in vertical and hedge line or filament width in
horizontal directions. Depending on the manufacturing process different types of support material are
used to stable parts which influence the orientated
manufacturing design as well.
3
Figure 1. Schematic display of a building platform with

the platform coordinates xB -yB -zB , the directrix coordinates
xL -yL -zL , the directrix curve and the profile coordinates
u-v-w.
To set the profile and the directrix curve in a relation

to the additive manufacturing machine, a reference
point in the machine is necessary. The most suitable
component of the machine as a reference is the building platform. Each standard element holds its position
relative to the building platform during the manufacturing process, while the platform moves relatively
to the machine. The reference point onto the building platform is defined with the platform coordinate
system xB -yB -zB (Figure 1). The directrix coordinate
system xL -yL -zL (Figure 1) is used to connect the directrix curve and the profile with the platform coordinate
system. This coordinate system enables a positioning of standard elements relatively to the platform
coordinates. The following definitions apply for the
coordinates and the directrix curve:
STANDARD ELEMENTS
Platform coordinate system xB -yB -zB (Figure 1):
Parallel to the process characterization, standard elements are defined which often reoccur in designing
of technical products. These standard elements are
divided into three groups. The first group Basic elements contains all elements which are necessary to
form basic geometries. Into the group Aggregated
elements firmly bonded basic elements are combined. With these elements the direct joining zone of
firmly bonded basic elements shall be examined as
well as relative positions of basic elements towards
each other. The third group Primary mounted elements contains elements which are form fitted joined
by the manufacturing process.
All standard elements have belonging properties.
With the values of these properties geometries of
standard elements can be set and characterized.
The geometry of standard elements and their connection to the additive manufacturing machine are
defined mathematically. Therefore the following definitions have to be set first.
3.1
Definitions
Each basic element is created by transferring a profile

along a directrix curve. Thereby the profile is defined
in the u-w-plain of the profile coordinate system
u-v-w (Figure 1) which is an accompanying tripod of
the directrix curve.
The origin correlates with the centre of the building

platform.
The xB -axis runs parallel to the user sided building
chamber opening. The positive counting direction
is defined from left to right from a users point of
view.
The yB -axis runs orthogonal to the user sided
building chamber opening with a positive counting
direction away from the user side.
The zB -axis runs orthogonal on the xB -and yB axis with positive counting direction contrary to the
gravity direction.
Directrix coordinate system xL -yL -zL (Figure 1):
Conforms at the parallel translated platform coordinate system.
Directrix curve (Figure 1):
In the directrix coordinate system the positions of
the points that form the directrix curve are defined
by the position vector rLk (, ).
The azimuth angle determines the angle between
the projection of the position vector rLk onto the
xL -yL -plane and the positive xL -axis.
The angle is measured with a positive counterclockwise direction of rotation.
0 applies for all .
546
Figure 2. Profile of a standard element into the u-w-plain

of the profile coordinate system.
Figure 3. Schematic display of a double curved element

with its basic properties.
The polar angle () defines the angle between the

position vector rLk and its projection on the xL -yL plane.
0 360 applies for all angles ().
The projections of directrix curves on the xL -yL plane create concave curves whose centers of curvature are located on the side which is oriented towards
the origin of coordinates.
The pitch angle Gp describes the angle between the
tangent of the directrix curve and the xL -yL -plane.
Profile coordinate system u-v-w (Figure 1):
Table 1. Examples for different basic geometries created

with a double curved element.
rLk
ra
ri
Element
mm
mm
mm
Sphere
Torus
Ellipsoid*
Hyperboloid*
0
> ra
0
> ra
>0
>0
>0
> ri
0
0
0
>0
360
360
360
360
0
0
0
0
90
0
90
90
270
360
270
270
* ra ( = 90 ) = ra ( = 270 ) > ra ( = 180 )
The origin is located on the directrix curve at the

end of the position vector rLk .
The v-axis is tangential to the directrix curve.
The u-axis runs horizontally and is orthogonal on
the v-axis.
The w-axis is orthogonal on the u-and v-axis.
Figure 2 displays the profile of a basic element. A
profile is an area which is defined into the u-w-plain
of the profile coordinate system as follows:
The dimensions of the area are limited by an inner
and an outer boundary line and two connection lines.
Boundary lines are concave curves with their center
of curvature located on the side which is oriented
towards the origin of coordinates.
Each boundary line has a starting and an ending
point.
Connection lines pairwise connect the starting and
ending points of outer and inner boundary lines.
The points of the boundary lines are described by
the position vectors ra (,rLk ) and ri (,rLk ).
ra > ri 0 applies for the position vectors.
The angle is measured from the positive uaxis with a positive counterclockwise direction of
rotation.
The start and end angles of the boundary are termed
as 1 and 2 .
0 1 <2 360 applies for all angles .
Standard elements can be defined in a mathematical
way with the set definitions. According to the different
groups of standard elements, the basic elements will
be defined first.
3.2 Basic elements

The purpose of the basic elements is to describe basic
geometries which often reoccur by designing technical
parts. The general basic element is the double curved
element. Special forms of this general basic element
are simple curved elements and plates.
3.2.1 Double curved elements
Double curved elements arise by transferring a profile
along a directrix curve. Figure 3 shows the general
form of a double curved element. The position vectors
rLk , ra and ri as well as the angles and , 1 and 2
are the properties of these elements.
By setting the properties in different relations to
each other different basic geometries can be created.
Table 1 exemplarily lists some double curved elements.
3.2.2 Simple curved element
Simple curved elements are a subquantity of double
curved elements. They arise if the position vector rLk
tends to infinity and the polar angle tends to 0 .
Figure 4 shows a simple curved element in its general form. The position vectors ra and ri as well as the
angles and , 1 , 2 and Gp are properties of these
elements. The pitch angel Gp defines the orientation
of the simple curved element relatively to the building
platform.
By setting the properties in different relations to
each other different basic geometries can be created.
Table 2 lists examples of those elements.
547

with plates.
l
b(v = 0) b(v = l) d(v = 0) d(v = l) Gp
Element mm mm
mm
mm
mm
Plate
>0 l
wedge > 0 l
pyramid > 0 l
Figure 4. Schematic display of a simple curved element and

its basic properties.
with a simple curved element.
Element
ri
l
ra
mm mm mm
1 2
Gp
Cylinder
>0 >0 0
0 0 0
Hollow cylinder > 0 > ri > 0 0 0 0
Cone *
>0 >0 0
0 0 0
360 0
360 0
360 0
* ra tends to cero as a function of the length l
b(v = 0) l
b(v = 0) l
0
l
d(v = 0) 0
0
0
0
0
consist of properties of the involved basic elements

and properties that occur within the aggregation.
Aggregated elements are classified into three
groups. The first group Joints contains elements
whose direct joining areas are considered. Elements
of the second group Relative positioned elements
enable the examination of advisable relative positions
of basic elements towards each other.
In addition to the both mentioned groups, aggregated elements contain the third group Complex
aggregated elements. These elements have a superior technical relevance and are eminently complex
to manufacture. The complexity arises because of the
interdependence of their properties. The manufacturing orientated expansion of the basic element property
values are contrary to the individually needed expansion of the technical properties. For example font
characters are complex aggregated elements.The complexity is given because font characters are created
of a number of basic and aggregated elements on a
tiny space. Otherwise the technical properties cognizability and legibility (DIN 1450) have to be
fulfilled.
3.4 Primary mounted elements
Figure 5. Schematic display of a plate and its properties.
3.2.3 Plates
Plates are a subquantity of double curved elements
which occur if the position vectors rLk , ra and ri tend
to infinity, the angle tends to 0 and the angles 1
and 2 tend to 90 . Figure 5 shows a plate in its general
form.
The plate width is a function b(v) = f(1 ,2 ) and
the thickness is set by the difference d(v) = ra -ri . So
the length l, width b(v), thickness d(v) and orientation
Gp are properties of a plate furthermore, l b(v) and
l d(v) applies for all plate dimensions.
Different basic geometries can be created by
variation of the properties. Table 3 lists examples of
plates.
Additive manufacturing processes enable the designer

to create form fitted joined elements during the manufacturing process. According to the definition of basic
elements, primary mounted double curved elements,
primary mounted single curved elements and primary
mounted plates are realized. All these elements are
placed into the group Primary mounted elements.
Exemplarily primary mounted double curved elements are explained in detail here. These elements
consist of two double curved elements whose directrix coordinate systems are identical. Figure 6 shows
schematically the general form of a double curved element. The geometry is defined by the properties of the
included basic elements. For instance, the gap width
between both elements is defined by the difference
of the position vectors ri,1 ()ra,2 (). Thus, primary
mounted double curved elements can be used as a ball
joint.
3.5 How to use standard elements
3.3 Aggregated elements

Aggregated elements are created of firmly bonded
basic elements. The properties of aggregated elements
Standard elements have been developed by dividing parts into separate elements which often reoccur
by designing technical parts. They consider basic
geometries, joints (firmly bonded or form fitted) and
548
Figure 8. Real thickness d(is) of a plate manufactured with

the laser sintering and fused deposition modeling process
assigned to the ideal thickness d(ideal).
Figure 6. Schematic display of a primary mounted double

curved element.
Figure 7. A bored wall consisting of a plate and a simple

curved element. The plate supports the outer surface of the
simple curved element.
Therefore standard elements were manufactured

with different additive manufacturing processes and
with different property values. It was examined how
property values influence the quality of standard elements. Based on the results first design rules for
additive manufacturing processes were derived.
The hereinafter explained design rules exemplarily
show results of the work package Development of
design rules. They describe the manufacturing orientated design of standard elements in a qualitative, not in
a quantitative way. Thus a certain degree of generality
will be reached, because concrete numeric values can
only apply for individually used materials, processes
and process parameters.
4.1 Thickness of a plate
positions relative to each other separately. To create

technical parts, standard elements have to be combined. Thereby standard elements interact with each
other. For example, one standard element can work as
support material for another standard element.
Figure 7 exemplarily shows a wall with a bore
inside. The wall itself can be considered as a plate.
The bore inside the plate is formed by a simple curved
element. The plate material works as support material
for the simple curved element.
This example shows in an easy way how parts can be
created by combining standard elements. In addition
it is necessary to know, how the property values of
the involved standard elements have to be expanded
manufacturing orientated.
DESIGN RULES
Design rules shall provide recommendations to design

engineers how parts should be designed in a proper
manufacturing orientated way. Assuming that parts are
made of combined standard elements, design rules
have to provide recommendations how standard elements should be designed for manufacturing. Thus
they must provide recommendations for advisable
property value expansions of standard elements.
Because of its technical relevance the thickness of a

plate is one of the most important properties of this
standard element. It is examined how thin a plate can
be. There for plates with different ideal thicknesses
d(ideal) from 0.2 mm to 2.0 mm were designed and
manufactured. The real thickness d(is) was measured
and compared to the ideal thickness. Figure 8 shows the
results. The measured thickness is displayed assigned
to the ideal thickness.
It becomes obviously that the real thickness has a
minimum value which it does not fall under. This minimum value is based on the dimensions of the energy
beam or filament of the used additive manufacturing
process. An energy beam cannot produce thicknesses
which are smaller than the beam itself. The FDMmachine always tries to lay down a contour around the
part layer and to fill the part layer with raster filaments.
Because of that it makes sense to recommend a minimum thickness for FDM-produced plates not smaller
than 3 filament widths; Two for the contour and one for
the raster. The derived design rule is shown in Figure 9.
4.2
Length of a plate
Another important property of a plate is its length.

Tests showed that plates can be manufactured with any
length if they are supported. If they are not supported,
549
Figure 9. Design rules for plates.
Figure 11. Design rules for plates.
Figure 10. Test specimen to examine a plates advisable

length related to its thickness.
which is possible by using the FDM-process, form

defects might occur if a plate is too long. Figure 10
shows the test specimen which was manufactured
to examine the advisable length. The test specimen
contains several plates of the same length but with
different thicknesses.
This test shows that the manufacturable length of a
plate has to be related to its thickness. During the manufacturing process forces are generated by the tip and
the plates dead weight. They effect on the plate. Comparatively a plate reacts on these forces like a bending
beam loaded with a force. The maximum bending is
influenced by the stiffness and the length. The stiffer
a beam is the smaller is the maximum bending. The
stiffness is influenced by the thickness mainly. And
the smaller the length of a beam is the smaller is its
bending, too. So the maximum bending is addicted
to the ratio of the length and the thickness. To keep
the maximum bending into a proper manufacturing
orientated range, an advisable ratio of the length and
thickness should be complied. Tests showed that the
ratio of the length l and thickness d should be smaller
than 10. Figure 11 shows the belonging design rules.
4.3
Layer cycle times of an FDM-process
In the FDM-process filaments need to be heated up

before they can be laid down. There for the tip of an
FDM-machine is very hot and can be considered as a
heat source. By designing parts for the FDM-process
it is necessary to minimize the influence of the heat
source on the manufactured parts. If the influence of
the heat is too big and the cooling time is too short form
defects occur. Figure 12 shows a test specimen from
Figure 12. Test specimen in the top view with bulges which
occurred because the tip heated the part too much.
the top view. It is obvious to see that bulges exist which

cause form defects. The bulges occurred because the
cooling time was much too short.
One way to enlarge the cooling time is to enlarge the
layer cycle time of an FDM-process. The layer cycle
time is the sum of the times in which the tip lays down
filaments and the times in which the tip moves between
positions.
The recommended layer cycle time t(cycle) shall be
related to the local building time t(local). The local
building time is the time in which the tip works above
a local part region and heats it.
The layer cycle time was measured indirectly with
the manufactured part layer size. Different test specimens with defined part layer sizes of 100 mm2 and
200 mm2 and a quadratic cross section were built.
To minimize the percentage of the time in which the
tip moves between positions the test specimens were
placed as near as possible to the tips home position.
For the tests at first only one test specimen with a
part layer size of 100 mm2 was manufactured. Here the
layer cycle time was as long as the local building time.
The next test included two test specimens, each with a
part layer size of 100 mm2 . So the layer cycle time was
enlarged to the double local building time. The layer
cycle time was further increased up to the fivefold
of the local building time. To rate the cooling times
influence on the form defects the occurred dimensional discrepancy was measured. Figure 13 shows the
results.
550
Figure 13. Form discrepancy assigned to the ratio of the

layer cycle time and the local building time.
Figure 14. Design rules for relative positions.
It becomes obviously that the value of the dimensional discrepancy is directly addicted to the ratio
between the layer cycle time t(cycle) and the local
building time t(local). To eliminate the form defect
the layer cycle time should be four times as long as the
local building time.
Layer cycle times can be increased by a proper relative positioning of involved standard elements. If it is
not possible to design a part in a manufacturing orientated way sacrificial elements can be used to enlarge
the layer cycle time.
Figure 14 shows the design rules which gives recommendations regarding the length of the layer cycle
time.
5
developed and published. Therefore different additive

manufacturing processes have been examined at first.
The layer thickness, hedge line and filament width as
well as different types of support material have been
identified as important process parameters which take
influence on the design of parts.
Parallel standard elements were developed. They
are classified into three groups: Basic elements,
Aggregated elements and Primary mounted elements. To the basic elements belong Double curved
elements, a Simple curved elements and Plates.
With these three elements basic geometries can be
created. Aggregated elements contain elements which
enable examinations of joining zones and relative positions of firmly bonded basic elements. The third group
classifies elements which are form fitted joined during the manufacturing process. In summary every
technical part can be designed by combining standard elements. These are defined mathematically. A
clear delimitation between them was achieved and
a clear definition of their properties and property
values is possible. Furthermore the standard elements create a common foundation for the development of design rules for additive manufacturing
processes as well as for conventional manufacturing
processes.
Based on the standard elements first design rules
were developed. They provide recommendations for
proper allocations of property values.
In the further proceeding the development of design
rules will be continued. Parallel additive manufacturing processes will be compared with conventional manufacturing processes. Therefore standard
elements, built with additive manufacturing processes,
will be compared with standard elements which were
built with conventional processes. Further the effect
of these manufacturing processes on the environment
will be examined comparatively. Therefore the potential to safe resources will be figured out as well as
the value of the generated emissions. The developed
knowledge will be transferred by seminars, lessons and
publications.
REFERENCES
SUMMARY AND OUTLOOK
In the project Direct Manufacturing Design Rules

design rules for additive manufacturing processes are
Gebhard, A. 2007a. Generative Fertigungsverfahren

Rapid Prototyping, Rapid Tooling, Rapid Manufacturing:
1113. Mnchen: Carl Hanser Verlag.
551

A priori process parameter adjustment for SLM process optimization

S. Clijsters, T. Craeghs & J.-P. Kruth
Katholieke Universiteit Leuven, Department of Mechanical Engineering, Division PMA, Belgium
ABSTRACT: Selective Laser Melting (SLM) is a layerwise production technique enabling the production of
complex metallic parts. In the SLM process parts are built by selectively melting subsequent layers of powder
by a laser beam. Nowadays a SLM machine is provided with a fixed scan strategy (laser power, scan velocity
and scan pattern) throughout the full build process. However, the parts geometry has a large influence on the
stability of the process and therefore the quality of some features like for instance thin walls, sharp corners,
down facing layers (layers above powder), is often poor. This problem can be overcome by using knowledge
of the geometry a priori. This paper presents a methodology to detect critical features in the model of the part,
based on the slicing data. In this way these critical features can then be processed with optimized parameters.
As a proof of concept this a priori parameter adaptation methodology is applied on production overhang.
INTRODUCTION
Selective laser melting (SLM) is Additive Manufacturing technique enabling the production of complex
metallic parts in a layerwise manner (Kruth, et al.
2007a). Figure 1 shows a schematic overview of a typical SLM machine. In this process a thin layer of metal
powder is deposited on a base plate by means of a powder deposition system. After the layer is deposited, a
laser will selectively melt the powder layer according to a predefined scanning pattern. Such scanning
pattern typically consists of a set of subsequent scan
vectors. After the layer is scanned, the build platform
moves over a fixed distance equal to the thickness
of one layer (in SLM typically 20 to 40 m) and a
new layer is deposited. This cycle is repeated continuously until the last layer is scanned. The whole melt
process takes place in a process chamber filled with
a protective gas, typically nitrogen gas for processing steels and argon for processing reactive materials
e.g. titanium.
The SLM process has a huge potential for a large
range of applications. Due to the almost infinite
geometrical freedom, there is no need to design or
manufacture dedicated tools for production. Since
material properties of SLM parts are nowadays
comparable to the properties of the corresponding
bulk material (Thijs, et al. 2010, Rombouts, et al.
2006), applications of the process can be found in
domains such as the medical sector, e.g. dentistry
(Kruth, et al. 2004, Vandenbroucke, 2005), in tool
making industries for the manufacturing of tools (Abe,
et al. 2001, Klocke, et al. 1996, Berger, 2001 and
Voet, et al. 2005), the general manufacturing industry (machine construction, automotive, etc.) while
the potential in production of lightweight structures
Figure 1. Schematic overview of a typical SLM set-up.
(Rehme & Emmelmann, 2006) is investigated for

aerospace applications.
Until today, all existing additive manufacturing
technologies (like SLS and SLM) use a fixed set of
process parameters (laser power, scan speed, hatch
spacing, scan strategy) during scanning of a part.These
parameter sets are mainly selected first of all to obtain
the highest possible material density and second the
highest possible production speed. However, using
constant process parameters throughout the build process does not ensure constant quality throughout the
part since in SLM very different geometrical features
have to processed, which have a different influence on
the melt pool behavior. An example of such feature
are overhanging structures, which are zones in a layer
that are completely build on loose powder. When such
(critical) features are scanned with constant process
parameter sets, this will result in a non-constant melt
pool size and shape due to the different heat flow situations during processing. For instance during scanning
553
of an overhang structure, the melt pool becomes very

elongated, which leads to balling. This is not beneficial for the quality and therefore it is necessary to
keep the melt pool dimensions under control during
the process.
To solve the problem of variations in thermal behavior of the melt pool, two different solution strategies
can be considered. The first strategy is to monitor
the melt pool dimensions continuously throughout the
build process, using melt pool sensors as a high speed
NIR camera and a photodiode (Kruth et al. 2007b,
Craeghs et al. 2010,). With such a monitoring system
the surface of the melt pool can be monitored accurately at during the process (on-line) in real-time and
at high frequency: the use of FPGA image processing enables real-time image processing up to 10kHz
(Craeghs & Kruth, 2010). If the measured melt pool
dimensions deviate from the desired melt pool dimensions, feedback is given to the process input parameters
to control the melt pool dimensions towards the desired
dimensions. However, in the SLM process for processing of each different geometrical feature the desired
melt pool dimensions may differ. Therefore the reference values for the feedback system are function of
the processed feature. When only one single reference
value is used for all different geometries, the efficiency
of feedback control is very low.
The second solution strategy to overcome the problem is to define the desired changes in process parameters and scan strategies a priori (off-line), at the level
of job preparation. This can be done by extracting the
information out of the geometry data of the part. By
using this method it is possible to choose the process
input parameters such that the melt pool has the desired
shape and size for all different heat flow situations.
This paper will explain how this a priori adaptation
has been implemented and will show a case study to
prove its utility and applicability.
METHODOLOGY
To implement a priori parameter adaptation in a SLMprocess, current process strategy should be extended
with two aspects as shown in figure 2. An accurate
job preparation is the first important aspect in a priori parameter adaptation. It is crucial to know where
parameters should be adapted. Parameter adaptation
should occur when the heat flow situation is different
from the nominal situation (i.e the situation for which
the standard process parameters lead to optimal process quality).These locations need to be extracted from
the geometrical (CAD) model. In such way the scan
vectors can be classified in different vector classes.
Each class will have a certain vector identity, which
reflects its heat flow situation. The algorithm for scan
vector classification will be discussed in section 2.1.
The second essential aspect in is the choice of
process parameter sets for each vector class.These process parameters must be optimized for different heat
flow situations and induce an optimal melt pool for
Figure 2. Methodology for a priori scan strategy adaptation.
different heat conductivities. Since these parameter

sets are material dependent and are currently only
known for nominal/standard heat flow situations, optimizing such a parameter set is a time consuming
process with a lot of trial and error experiments.
Linking these two aspects, vector classification and
optimized parameter sets for different vector identities
will optimize the process for each heat flow situation.
In the following paragraphs the implementation for
both vector classification and parameter optimization
will be explained into further detail.
2.1 Scan vector classification
As mentioned before, it is crucial to detect heat transfer changes a priori to adapt the parameters. In this
paragraph the scan vector generation procedure for
SLM has been extended to classify vectors according
to their processing behavior, as shown in Figure 3. The
procedure consists of four steps:
(1) Slicing
In this phase, the job preparation converts the parts 3D
model into sequential slices with 2D contours. The to
build CAD part is converted into a .stl-file (which is
a standard in additive manufacturing). After orienting
this .stl-model in the preferred building direction, the
model is sliced. The output of this slicing is a file
containing all the layers (slices) with their contours.
(2) Offsetting
Once the slices of each layer are available, the contours
of the slices are offsetted. Offsetting is needed to compensate for the finite dimensions of the laser beam:
without offsetting the dimensions of the part would be
biased outwards. The procedure is very similar to offsetting in milling: A mill has a certain radius, therefore
the tool path in contour milling has to be offsetted to
554
Figure 3. Scan vector classification procedure.
mill the correct contour. However in this case the generated melt pool is the mill of the process. More details
on the offsetting algorithm can be found in Moesen,
et al. 2011.
(3) UMD-splitting
In the standard state-of-the-art job preparation a lot
of geometrical information gets lost. This information
on the part can be very useful for detecting differences in heat transfer situations. The main different
heat flow situations are illustrated in Figure 4. The vertical direction of the table distinguishes between huge
differences in heat flow, while the horizontal direction refers to small heat conductivity variations due to
neighbor scan tracks.
The goal of the UMD (Up, Middle and Down
facings) splitting is to import the geometrical 3D
knowledge and to recognize three different zones in a
slice: up facing areas (areas on which no layers will be
built), down facing areas (areas build on loose powder) and middle areas (layer above and beneath). To
recognize these different zones and changes in heat
conductivity, the slices are compared to eachother.
A relatively simple algorithm is used to detect these
zones.
The algorithm used for this identification is
extracted from the open source CGAL library
(CGAL). This algorithm of CGAL ascribes separate
values to areas defined by the contours in a layer
(e.g. Figure 5: areas in layer i have value 1 and the
areas in layer i + 1 have value 2). To detect 3D information out of these two layers, the layers will be placed
on top of each other and the values of the areas will
be added up, resulting in areas with calculated values (e.g. areas with value 1 and 2 indicate respectively
only in layer i and i + 1, value 3 indicates the common areas of the two layers). By filtering the right
values the down facing, up facing or middle areas
can be easily distinguished: in Figure 5 value 1, value
2 in (layer i) + (layer i + 1) indicate respectively the
Figure 4. Classification of vectors by their heat

conductivity.
Figure 5. UMD-splitting: detection of down facing.
up facings of layers i and the down facings of layer

i + 1, the middles are represented by a value 3. Once
these different areas are distinguished, the classifying
of the scan vectors can be applied during the generation of the scanning pattern. Vectors of a scanning
pattern can be classified based on the zone in which
they are located.
(4) Hatching
The next step in SLM job preparation is filling the
2D contours with scanning patterns. Such a scanning
pattern typically consists of a set of subsequent linear scan vectors representing the tool path of the laser
beam. Commonly these scan strategies are generated
by the SLM machine and the specific hatching algorithms are in general IP of a certain SLM machine
vendor. Therefore little information is available on this
topic in literature. However it is noticed that these scan
555
strategies all look similar. Sectorial scanning (Yasa,

et al. 2010) is commonly used by all the SLM companies. In this research simple zig-zag scanning patterns
are generated and used.
This four step procedure is implemented in an
in-house tool path generator. The final result of the
new vector generation tool is a scanning pattern with
geometrical knowledge included: each subsequent
scan vector of the scanning pattern, when generated,
receives an identity which is dependent on its geometrical location. The next step is to define optimal
process parameters for all these different identities. In
this work only optimization of down facing surfaces
will be discussed.
2.2
Parameter optimization for vector identities
Once all different vectors zones are classified according to their process behavior, the parameters (scan
speed, laser power) for each identity and zone must
be optimized (second branch in Figure 2). To greatly
reduce the amount of trial and error experiments, a
numerical model of SLM has been used which allows
estimating the process parameters for different geometries. This way the number of experiments to optimize
the processing of each vector class can be reduced
significantly in comparison with fully experimental
parameter optimization.
Once an estimation of the process parameters is
determined for a specific geometrical situation, further detailed optimization has to be performed with
experiments. These experiments aim to find the definite optimized parameter sets for the specific situation. As an example, parameter optimization for
processing of downfacing surfaces will be discussed
in section 4.2.
3
3.1
EQUIPMENT & MATERIALS

SLM machine of KULeuven PMA
In this research experiments have been performed

on a home-made SLM machine of KULeuven-PMA.
This machine is equipped with a 300W IPG fiber
laser (wavelength 1064 nm) and a spot size of 80 m
(99%). The central control unit of the machine is
a National Instruments PXI system, equipped with
two field programmable gate arrays (FPGA) to control the scanner and to process the data of the melt
pool sensors in real-time. The use of in-house developed software opens opportunities to implement and
experiment with own developed hatch strategies and
parameters sets.
3.2
3.3 Material
All experiments are executed with Ti-6Al-4V powder,
since the behavior of this material has been studied
extensively (Thijs, et al. 2010). With the already available knowledge, conclusions on melt pool behavior
can be interpreted better.
3.4 Numerical model of SLM
To limit the amount of trial and error efforts in optimizing the SLM process, simulation models are being
developed in many research institutes in the last 5
years. The process is a complex combination of heat
flow, fluid dynamics, optics and mechanics. A total
model of the production process has therefore not yet
been developed and is not expected in the near future.
Depending on the aim of the simulations, different
models are developed. Each of the models tries to predict a certain aspect of the process (e.g.: thermal stress
and deformation (Zaeh & Branner, 2009)). The most
interesting models for this research are models which
try to predict the melt pool behavior.
By modeling the process at micro level and implementing phase transitions the model should be able
to estimate the melt pool size, shape and/or behavior.
The literature shows that among others the numerical models from Gusarov (Gusarov, et al. 2009) and
Verhaeghe (Verhaeghe, et al. 2009) have been able to
predict the melt pool behavior.The model ofVerhaeghe
et al. will be used in this research to predict the process
parameters for certain geometries.
4
OPTIMIZATION OF OVERHANG
STRUCTURES
To show the applicability and utility of the developed methodology (Fig. 2), processing of overhang
geometries will be discussed. As a reference for
these overhang geometries a more specific square
shaped overhang/bridge structure (Fig. 6a) was investigated. Building such a bridge with standard parameters results in big deformations and dross formations
(Fig. 6b); sometimes even resulting in process abortion. Optimizing the parameters should result in a
stable (constant melt pool dimensions in process)
and controllable melt pool in every separate heat
flow situation of the bridge to minimize the dross
formation.
Scan vector generation software tool
An in-house developed software tool has been written

which enables to implement different scanning strategies and the vector classification methods as described
above.
Figure 6. Bridge structure: (a) Reference bridge with

dimensions, (b) dross formation without parameter
adaptation.
556
Figure 7. (a) Simulation of overhang structure, (b) Melt pool behavior during simulation.
To implement the a priori methodology the two

aspects, vector classification and parameter optimization should be considered.
4.1
Scan vector classification of overhang

structures
In overhang structures mainly two different heat flow

situations can be distinguished: scanning on loose
powder (Fig. 4. Overhang heat flow) and scanning on a
powder layer with standard heat flow conditions (solid
substrate beneath the powder layer, Fig. 4. Standard
heat flow). To classify the vectors into the right heat
flow situations it is crucial to detect which area will
be scanned on loose powder (down facing). This down
facing and standard (middle) areas as mentioned are
already detected by the UMD-splitting. Therefore no
further classification needs to be done.
4.2
Parameter optimization
Since a set of parameters is already available for scanning under standard heat flow conditions (a laser power
of 42W and a scan speed of 225 mm/s), only a parameter set for scanning the down facing needs to be
determined. The estimation of this parameter set is
done with a numerical model to minimize the amount
of experiments. The further optimization is done by
trial and error experiments.
(1)
Parameter estimation by a numerical model of

SLM
The use of a numerical model reduces the intensive
trial and error to find the correct process parameters.
The goal of these simulations is to find an estimation
of stable parameters to control the melt pool volume
and shape in the down facing heat flow condition of
the overhang. Scanning on a lower heat conductive
powder results in larger extended melt pool. Figure 7
depicts the simulated melt pool size when scanning
an overhang using fixed parameters, optimized for
standard heat flow conditions. During the numerical
experiment, in the first 240 m the laser scans on
solid foundation. Once reaching the bridge overhang,
the heat flow situation switches and the simulation
shows the transformation of the melt pool. The melt
pool will take a half barrel shape (depth = width/2) as
the conductivity is about equal in all the directions

perpendicular on the scan direction (since a single
line track is simulated). Another notable phenomenon
is the increase in melt pool length, since heat can
only transfer backwards throughout the solidified scan
track.
A melt pool will be called stable when it is not
deeper as the melt pool of the scan line in standard
conditions, to avoid dross formations, and on top of
that its length should be constant. To create such a stable melt pool it is obvious that the laser power and scan
speed have to be lowered to avoid a deep melt pool and
to give the melt pool time to solidify.
A set of stable process parameters for melting on
loose powder could be determined after simulating
an amount of parameter sets. The simulations indicated that very low power (4 W) with low scan speed
(70 mm/s) were preferable to create a stable process to
bridge the gap between the two pillars in this heat flow
situation.
During the melt pool simulation it is noticed that
the melt pool behavior changes before reaching the
border of the overhang (Fig. 7). The heat conductivity
drops when reaching the border (Fig. 4 Transition heat
flow). This results in an enlargement of the melt pool.
However this zone of 40 m is too small to define an
own set of parameters. Therefore to avoid a big melt
pool at the start of the overhang which results in dross
formation, an offset is induced on the down facing zone
to cover this transition zone. This offset enlarges the
down facing area to induce a smooth transformation
of the standard melt pool to a down facing melt pool
and to avoid the dross formation at the border.
Once an estimation of the parameters is defined, it
is useful to estimate the influenced zone of this overhang (i.e. the amount of solid layers that is needed to be
in standard process conditions.) Simulations showed
that with 150 m solid material above an overhang
(5 layers of 30 m) the process is back in standard conditions (Fig. 8). Therefore after scanning 5 layers with
the optimal defined parameters for down facing layers,
the build can be continued with standard parameters.
(2) Experimental optimization
By means of the estimated parameters, experiments
have been performed to study and optimize the parameters more into detail. First the estimated parameters
557
Figure 10. A bridge with a span of 40 mm is produced
Figure 8. Melt pool behavior scanning a powder layer with

x m solid beneath to conduct the heat (standard parameters
42W 225mm/s).
Figure 11. Tunnel with a span of 10 mm and a length of

40 mm.
Figure 9. Microscope image of a cross section of the bridge

deck.
where optimized to avoid deformation, this resulted in

optimal parameter set (laser power 5W and scan speed
85 mm/s).
These optimized parameter set was used to build
an overhang structure. After scanning 5 layers at the
optimal parameters with a hatch spacing of 20 m,
the standard parameters, as indicated in the simulations, were used. Dross formation occurred, since the
first 5 layers at low power and scan speed were not
dense enough to conduct the heat towards the pillars.
Therefore two extra zones of 5 layers are put on top
of the first 5 overhang layers of which the parameters incrementally raise towards the standard parameters. A microscope image of the cross section of a
bridge deck is displayed in Figure 9. This figure shows
the 3 increasing parameter sets before the standard
parameters are reached. Reaching the standard parameters (Fig. 9: above the top line) the part is almost
full dense again. The density of the 10 first layers
however is low. This explains why dross formation
occurred, after switching back to standard heat conductivity parameters after 5 layers. These 5 layers are
not able to conduct the heat to the pillars since they
are not dense.
Sandblasting these completed bridges erodes these
first layers. These sacrificial layers however are necessary to create mechanical resistance against deformation induced by thermal stress and to create a heat
flow towards the pillars of the bridge.
Figure 12. The case study house built in Ti-6Al-4V.
4.3 Extrapolation of geometry

The methodology of a priori scan strategy adaptation seems very promising on the reference bridges.
Therefore this building strategy was applied on different bridge dimensions. To check the possible span
width, a long geometry with a gap of 40 mm was
built (Fig. 10). This was so to date not possible with
state-of-the-art job preparation, without any support
structure.
The next test-geometry built, was a long tunnel with
pillars of 3 mm and a span of 10 mm (Fig. 11). This
tunnel with a length of 40 mm was built successfully,
however at the entrances of the tunnel deformation
occurred by the lack of mechanical resistance (Fig. 11).
To solve this issue it is necessary to use support
structures at the entrances.
CASE STUDY
To prove the improvements of this a priori parameter

adaptation a house with carport was built as a case
study (Fig. 12). This house was built twice, once with
standard fixed parameters and once with the a priori adapted scan strategy. There was a clear difference
558
Figure 13. Circle cavities build in a 3 mm thick wall (a) with

fixed parameters, (b) with parameter adaptation for down
facing layers.
CONCLUSIONS
The geometric boundaries of current SLM processes

are limited by the static character of the parameter
sets chosen in the SLM process. Adjusting parameters to the appropriate heat conductivity would push
the boundaries of the SLM process further and would
introduce new possibilities.
As was shown for overhang structures, support
structures can be minimized and quality can be
improved by optimizing the job preparation. To execute this optimization it is necessary to use the
knowledge of the CAD model to predict the heat
conductivity.
Many other features which influence the heat flow
can be investigated. However a lot of simulations and
trial and error work should be done to find the correct
parameter sets for different situations.
ACKNOWLEDGEMENTS
Figure 14. Cross-section of the case study.
between those two parts. To show these differences

two features of the house will be compared. The circular attic window and roof of the house are compared
further into detail.
The authors acknowledge the financial support from

the IWT SBO-project DiRaMaP and the KULeuven
IOF-project IOF-KP/06.
REFERENCES
5.1
Circular attic window
To show the utility of the a priori methodology, it is

worth to take a closer look at the circular attic window (diameter 6.35 mm), which is shown in Figure 13
(a) using fixed parameters and (b) using the a priori
scan parameter adaptation. It shows that the circular
cavity with a priori parameter adaptation has a more
circular shape than the one with fixed parameter sets.
Measuring the circularity with an optical microscope
shows that the circularity with parameter adaptation
(0.163 mm) is much better than with fixed parameters
(0.318 mm). This optimized scan strategy for overhang structures had a remarkable improvement for this
circular cavity.
5.2
Roof of the house
In this case study one challenging overhang feature

was designed, namely the top roof of the house.
The size of this overhang is 40 mm by 22 mm. This
feature cannot be built properly without support structures with fixed scan parameters. The scan strategy
had a major influence on the quality of this roof.
The roof of the house built without a priori scan
strategy adaptation was rough and deformed by dross
formation and the thermal stresses. The house with
parameter adaptation was built successfully with no
visible deformation and dross formation was avoided
successfully. The surface was even smoother than any
top layer. In Figure 14 a cross section of the roof of
the case study is illustrated. It shows the overhang of
40 mm by 22 mm which was gapped successfully with
a minor amount of dross formation and deformation.
Abe, F.; Osakada, K.; Shiomi, M.; Uematsu, K. & Matsumoto,

M. 2001. The manufacturing of hard tools from metallic
powders by selective laser melting. Journal of Materials
Processing Technology, 111(13):210213.
Berger., U. 2001. Rapid tooling and computertomography
for aluminium casting of automotive components. In
uRapid 2001 International users conference on rapid
prototyping & rapid tooling & rapid manufacturing.
CGAL, Computational Geometry Algorithms Library,
http://www.cgal.org
Craeghs, T.; Bechmann, F.; Berumen, S. & Kruth, J.-P. 2010.
Feedback control of layerwise laser melting using optical
sensors. 6th international conference on Laser Assisted
Net Shape Engineering (LANE), 2124 September 2010.
Erlangen, Germany.
Craeghs, T. & Kruth, J.-P. 2010. Developing a Quality Inspection Method for Selective Laser Melting of Metals with NI
Hardware and Software. National Instruments case study,
http://sine.ni.com/cs/app/doc/p/id/cs-13103
Gusarov, A. V.; Yadroitsev, I.; Bertrand, P. & Smurov, I. 2009.
Model of radiation and heat transfer in laser-powder interaction zone at selective laser melting. Journal of heat
transfer, 131(072101):, 110.
Klocke, F.; Wirtz, H. & Meiners.,W., 1996. Direct manufacturing of metal prototypes and prototype tools. In
Proceedings solid freeform fabrication symposium, august
1996, Austin.
Kruth, J.P.; Froyen, L.; Kumar, S.; Rombouts, M. & Van
Vaerenbergh, J. 2004 Study of laser-sinterability of ironbased powder mixture. In 10mes Assises Europennes de
Prototypage Rapide. 1415 September 2004, Paris.
Kruth, J.-P.; Levy, G.; Klocke, F. & Childs, T. 2007a. Consolidation phenomena in laser and powder-bed based layered
manufacturing. CIRP Annals-Manufacturing Technology.
56: 730759.
Kruth, J.-P.; Mercelis, P.; Vaerenbergh, J. V. & Craeghs,
T. 2007b. Feedback control of selective laser melting.
559
Proceedings of the 3rd International Conference on

Advanced Research in Virtual and Rapid Prototyping,
2007:521527.
Moesen, M., Craeghs,T., Kruth, J., Schrooten, J. 2011. Robust
beam compensation for laser-based additive manufacturing. Computer-Aided Design, 43 (8), 876888.
Meiners, W. 1999, PhdThesis, Direktes Selektives Laser Sintern einkomponentiger metallischer Werkstoffe, Aachen.
Rehme, O. & Emmelmann, C. Rapid manufacturing of lattice
structures with selective laser melting. In Proceedings of
SPIE Photonics West, LASE 2006 Symposium, LBMP-III
conference, January 2006, San Jose, California, USA.
Rombouts, M.; Kruth, J.; Froyen, L. & Mercelis 2006. P.
Fundamentals of Selective Laser Melting of alloyed steel
powders. CIRP Annals-Manufacturing Technology. 55:
187192.
Thijs, L.; Verhaeghe, F.; Craeghs, T.; Humbeeck, J. V. &
Kruth, J.-P. 2010. A study of the microstructural evolution during selective laser melting of Ti-6Al-4V. Acta
Materialia, 58: 33033312.
Vandenbroucke, B. 2005. Rapid manufacturing of dental prostheses by means of sls/slm. In 11e Assises
Europeennnes du Prototypage Rapide, 45 October, 2005,
Paris-Maison de la Mecanique.
Verhaeghe, F.; Craeghs, T.; Heulens, J. & Pandelaers, L. 2009.
A pragmatic model for selective laser melting with evaporation. Acta Materialia, 57: 60066012.
Voet, A.; Dehaes, J.; Mingneau, J.; Kruth, J.-P. & Van Vaerenbergh, J., 2005; Study of the wear behaviour of conventional and rapid tooling mould materials. In International
Conference Polymers & Moulds Innovations PMI, April
2023 2005, Gent, Belgium.
Yasa, E.; Deckers, J.; Kruth, J.-P.; Rombouts, M. & Luyten, J.
2010. Investigation of sectoral scanning in Selective Laser
Melting. Proceedings of 10th Biennial ASME Conference
on Engineering Systems, Design and Analysis.
Zaeh, M. F.; Branner, G. 2009. Investigations on residual
stresses and deformations in Selective Laser Melting
(SLM). Production Engineering 4(1): 3545.
560

Surface roughness analysis in Selective Laser Melting

G. Strano, L. Hao, R.M. Everson & K.E. Evans
College of Engineering, Mathematics and Physical Sciences, University of Exeter, Devon, EX4 4QF, United Kingdom
ABSTRACT: Selective Laser Melting (SLM) is an increasingly employed additive manufacturing (AM) system
in the production of medical, aerospace, and automotive parts. Despite progress in material flexibility and
mechanical performance, relatively poor surface finish still presents a major limitation in SLM. This study
analyses the effect of the stair step and particle bonding effect, to the surface roughness of SLM processed parts, at
different orientations. Surface characterisation conducted for the surfaces at different inclination angles by surface
profilometer and scanning electron microscope (SEM) revealed that the presence of partially-bonded particles
on the top surfaces might affect the surface roughness significantly when the layer thickness is comparable to
particle size. Classic models for roughness prediction, based on pure geometrical consideration of the stair step
profile, fail to describe the trend of the experimental data observed. The paper presents key contribution factors
influencing surface morphology and set the basis of a theoretical model for roughness prediction.
1
1.1
INTRODUCTION
SLM and surface quality
Selective Laser Melting (SLM) is an emerging additive manufacturing system for the production of end
use parts with complex shape, for medical, aerospace,
automotive applications. Its diffusion in recent years
has been permitted by an increasing availability of
processing materials and mechanical performance of
the resulting SLM parts (Levy et al. 2003). The SLM
process enables the direct melting of powders of a
number of metals, such as Titanium, Steel, Chrome
Cobalt, Aluminium alloys, and building of net-shape
parts through a layer by layer approach; for each
layer a scanning laser beam supplies the energy to
locally melt a layer of deposited metal powder and
fuse it onto previously melted layer.SLM still faces
an apparent limitation in terms of surface quality if
compared to some alternative metal manufacturing
processes such as machining process. Surface quality
is greatly influenced by the stair step effect, which
is a stepped approximation of the edges of curve and
inclined surfaces.
1.2 Theoretical models based on geometrical
considerations
In the past a number of studies have been conducted
to predict the surface roughness of parts processed
through different AM platforms.a theoretical model
was presented to predict the surface roughness for
Stereolithography (SLA) parts by introducing two different expressions to predict the roughness of upward
and downward-facing surfaces considering the layer
thickness, surface angle and layer profile (Reeves &

Cobb 1997). Campbell (Campbell et al., 2002) presented a comparison between theoretical roughness
obtained from a trigonometrically derived equation,
on the stair step profile, and empirical roughness measured on several different AM platforms. Luis Perez
(Perez et al., 2001) proposed a geometrical roughness model to predict the average surface roughness of
AM parts.
All the previous models based the prediction of surface roughness on the geometrical description of the
stair-stepped profile of sloped surfaces. Such models
could fail to accurately predict the surface roughness of AM parts, because surface roughness might
be influenced also by other process parameters as
shown in (Bacchewar et al., 2007). An alternative
approach to roughness prediction has been used in
(Ahn et al., 2009); the model interpolated data from
empirical observations of test samples; theoretical and
real distributions were compared through the fabrication of test parts manufactured by SLA. Strano
(Strano et al., in press) developed a model for the prediction of roughness of parts manufactured by SLS,
based on a phenomenological approach. Experimental roughness data were collected and interpolated
for a range of deposition orientations, and a model
for the evaluation of roughness objective part was
formulated.
This study analyses the effect of the stair step and
balling effect, to the surface roughness of SLM processed parts, at different orientations.A comparison
between the measured roughness and the roughness
predicted through classic models,based on pure geometrical consideration of the stair step profile, is
investigated.
561
2
2.1
Surface roughness
The surface roughness, according to ISO 4287 standards (ISO 1997), is defined as the average deviations
of the roughness profile from the mean line (Figure 1).
Considering the surface profile of length L, the definition of surface roughness Ra is set in following
Equation:
Supposed to have N values of measured surface

roughness on the surface profile, y1 , y2 , . . . yN , the
surface roughness is numerically calculated as:
The definition of surface roughness can be extended

to a 2D surface profile of area A (Figure 1), is the
average deviations of the roughness profile from the
mean plane obtained through levelling of the mean
square plane of the measured surface.
Suppose to have NxM values of measured surface roughness on the surface profile, y1 , y2 , . . . yNxM ,
the surface roughness of a 2D surface profile is
numerically calculated as:
Figure 1. Representation of 1D (top) and 2D (bottom)

surface profiles.
Following is presented an analysis of the contributions, namely stair step and balling effect, to the
surface roughness of SLM processed parts, at different
orientations. For this purpose surface characterisation
on manufactured samples was conducted at different inclination angles by surface profilometer and
scanning electron microscope (SEM).
2.2
Equipment and sample fabrication
In order to characterise the actual surface roughness

distribution of SLM processed part, a sample with the
truncheon geometry (Figure 2) has been fabricated,
using Steel 316L alloy. The sample was designed to
measure roughness of all the angles in the range from
0 to 90 by a 5 step. The sample was fabricated
through a EOS M270 machine; the process parameters
were, layer thickness 20 micrometers, hatch spacing
100 micrometers, beam scan speed 900 mm/s, beam
spot diameter 100 micrometers and laser power 195 W.
The analysis of sample surface was first carried
out using a surface profilometer (Talyscan 150, Taylor
Figure 2. Sample geometry.
Hobson Ltd). For each inclination, surface roughness

was collected through stylus gauge on a surface of
10 mm 1 mm, with a scanning spacing fixed at 5 m
and scanning speed at 2500 m/s. Measurements on
bottom surfaces were taken only for surface without
support (range from 45 to 90 sloping angle). Secondly, to further investigate the surface morphology at
microscopic scale, the sample was treated using isopropanol and surface analysis was carried out using a
SEM (S-3200N, Hitachi).
562
Figure 3. Experimental roughness in SLM process at different sloping angle; upward-facing (top), and downward-facing
(bottom) oriented data.
3 ANALISYS OF SURFACE MORPHOLOGY

3.1
Effects of sloping angle on surface roughness
Roughness measured on downward oriented surfaces

was (Figure 3, bottom) observed to have higher values of surface roughness if compared to top surfaces (Figure 3, top). A similar trend was observed
in (Vandenbroucke & Kruth 2007) and it can be
explained by the presence of stalactites formed during
the solidification due to gravity.
As expected roughness of upward oriented surfaces
presents a minimum value at 0 (flat surface); that
is mainly caused by the rippling effect due to scan
tracks, which occur on top surfaces. As the inclination angle increases from 0 , higher surface roughness
has resulted from the introduced stair-step effect. In a
prospective of surface roughness minimisation, it is
important to notice that roughness of flat surfaces
is possible to be eventually reduced through surface
remelting process (Kruth et al., 2010). Nevertheless, as
the surfaces are inclined, the laser remeltingon the top
of the inclined surface is not possible, due to the SLM
process limitation to sinter material only horizontally, hence the importance of a prediction of surface
roughness of parts surfaces inclined at certain angle.
Figure 4. Surface profiles of a) 5 sloped surface, b) Profile

of 10 sloped surface, c) 65 sloped surface
The trend of measured roughness is mainly constant

in the range of 5 45 , and presents a slight decrease in
the range 50 90 . A further investigation of the
upward surfaces reveals the presence of patterns with
vertically aligned reliefs for surfaces at low sloping
angle. Figure 4a shows a number of surface profiles
at different inclinations of 5 , 10 (white lines in Figure 4a, Figure 4b) and 65 . In each pattern, vertical
lines occur regularly at distance equal to the distance
between step edges, which is determined by the stair
step between consecutive layers.
563
spacing (100 m) and laser beam diameter (100 m).

At the bottom of Figure 5, it is shown the profile of
90 inclined surface. It confirms an high number of
spare and grouped particles is present on the surface;
particles can be considered the main cause of surface
roughness at 90 , since no stair step effect occur,
when layer consecutively overlapped(however a minor
residual roughness might be expected, because of the
limits on the repeatability of the process accuracy to a
perfectly matching overlap between layers).
4
DISCUSSION
4.1 Comparison between measured and

theoretical roughness
SLM processes start with a CAD model of the object
that has to be built; slicing the geometry involves a
level of approximation, described by the stair step
effect (6). The surface roughness Ra for the inclined
surface represented schematically in Figure 6, can be
defined as:
Figure 5. SEM picture of flat surface (Top); presence of
particles on 90 sloped surface (Bottom).
Figure 6. Schematic representation of sliced profile.
At higher sloping angles (65 ), no vertical lines pattern war observed; surface was characterised by the
presence of high peaks (Figure 4c). In order tocharacterize the surface morphology at higher sloping angles,
and to investigate the contribution of other effects to
surface roughness, a surface morphology analysis was
carried out by SEM.
At the top of Figure 5 it is shown the profile of
flat surface, normal to the build direction. When flat
surfaces are sintered, there is not presence of spare
unsintered particles on the top of the surface; this is
because the low layer thickness (20 m) and the high
power (195 W) supplied by the laser beam, fully melt
the powders and fused them into a relatively smooth
and uniform layer. The effects of scan direction and
strategy (highlighted by the spotted arrows) are visible in the figure; for each scan line it is noticeable
cusp tracks oriented in the sense of the moving laser
beam. The distance between cusps with same orientation is of about 200 m, as results of the chosen hatch
with Lt layer thickness and surface sloping angle.

From Equation (5) is evident that, an improved surface finish it is expected through either the choice of
low layer thickness or through increasing the surface
inclination angle, according with the cosine term. The
prediction of surface roughness expressed in Equation (5), has been formulated considering only the
effect of the stair steps on inclined surfaces. Previous models in literature present similar expressions,
deriving the expression of the surface roughness from
trigonometrical considerations on the stair step profile; nevertheless, often these model were able to
predict roughness for a partial range of surface inclinations with respect to the build directions (Campbell
et al., 2002), (Luis Prez et al., 2001). Following is
proposed a comparison between our experimental data
collected on SLM platform, and the theoretical roughness predicted considering only the stair-step effect.
The comparison shows a mismatching for almost
the entire range of inclination angles; in particular,
experimental roughness does not decrease as much as
expected by the cosine function.When surfaces are low
inclined (in the range of 5 15 ), the width of each step
(228 m74 m) is bigger than the average diameter
of the particle (20 m), therefore the surface conserves the characteristic of a zig-zag profile, due
to the stair-step,and roughnesscan be well predicted
by Equation (5). As the surface become more sloped,
the width of each step get smaller, and it is comparable
with the particle size; therefore the presence of spare
particles influences the surface profile. As example,
we can observe the roughness at very high sloping
angles; at 90 the roughness due to the stair step effect
is expected to be null, according to cosine term of
Equation (5). It might expect that the predicted zero
564
predicted by classic models, experimental observations

showed an high concentration of particles which
justify the high roughness measured at that sloping
angle.
REFERENCES
Figure 7. Comparison of experimental roughness, and

roughness predicted considering only the stair-step effect.
roughness well interpreters the situation when layers

are overlapped on the top of each other. In reality, the
measured roughness presents is much higher (14 m),
caused by presence of particle on the surface.
5
CONCLUSIONS
An investigation of surface roughness and morphology has been conducted for Steel 316L alloy part
made by Selective Laser Melting (SLM). In order to
characterise the actual surfaces at different sloping
angles, a truncheon sample has been produced and an
analysis has been conducted at different scales, by surface profilometer and scanning electron microscope.
Roughness measured by profilometer was different
from the one predicted through the classic model based
on pure geometrical consideration on the stair step
profile due to the presence of spare particles on the
surface. When layer thickness size is comparable to
particle diameter, partially-bonded particles present
at top surface, can fill the gaps between consecutive layers, thus affecting the actual surface roughness. This has been particularly evident at surface
sloped at 90 ; although minimum roughness would be
Ahn D., Kim H., & Lee S., 2009. Surface roughness prediction
using measured data and interpolation in layered manufacturing. Journal of Materials Processing Technology
209(2): 664671.
Bacchewar P. B., Singhal S. K., & Pandey P. M., 2007. Statistical modelling and optimization of surface roughness
in the selective laser sintering process. Proceedings of the
Institution of Mechanical Engineers, Part B: Journal of
Engineering Manufacture 221(1): 3552.
Campbell, R. I., Martorelli M., & Lee H. S., 2002. Surface roughness visualisation for rapid prototyping models.
Computer-Aided Design 34(10): 717725.
ISO 1997. 4287, Geometrical product specifications (GPS),
Surface texture: profile method terms, definitions and
surface texture parameters.
Kruth, J. P., Badrossamay M., Yasa E., Deckers J., Thijs L.
& Van Humbeeck J., 2010. Part and material properties
in selective laser melting of metals. 16th International
Symposium on Electromachining (ISEM XVI). Shanghai,
China.
Levy G., Schindel R., & Kruth J. P., 2003. Rapid Manufacturing and Rapid Tooling with Layer Manufacturing (LM)
Technologies, State of the Art and Future Perspectives.
CIRP Annals-Manufacturing Technology 52(2): 589609.
Luis Prez C. J., Calvet J. V. & Prez M. A. S., 2001. Geometric roughness analysis in solid free-form manufacturing
processes. Journal of Materials Processing Technology
119(13): 5257.
Reeves, P. E. & Cobb R. C. 1997. Reducing the surface deviation of Stereolithography using in-process techniques.
Rapid Prototyping Journal 3 (1): 2031.
Strano G., Hao L., Everson R.M. & Evans, K., 2011.
Multi-Objective Optimisation of Selective Laser Sintering
Processes for Surface Quality and Energy Saving, Proceedings of the Institution of Mechanical Engineers,
Part B: Journal of Engineering Manufacture, in press.
Vandenbroucke, B. & J. P. Kruth 2007. Selective laser
melting of biocompatible metals for rapid manufacturing of medical parts. Rapid Prototyping Journal 13(4):
196302.
565

AsTeRICS a rapid prototyping platform for Assistive Technologies

Christoph Wei & Christoph Veigl
Institute of Embedded Systems, University of Applied Sciences Technikum Wien, Vienna, Austria
ABSTRACT: A vast number of people with disabilities in Europe and worldwide are supported by Assistive
Technologies (AT). Often AT solutions do not fit the specific requirements of a user out of the box making costly
adaptations necessary. AsTeRICS aims to change this situation by employing rapid prototyping technologies
in AT. A construction set of building blocks to create flexible prototypes for each individual user is provided.
This set is configured by a visual modeling software application which allows to connect all building blocks as
needed tailoring the prototype to the users abilities. Building blocks include sensors (such as switch inputs but
also vision systems and brain computer interfaces), data processing elements (mathematical and flow control)
and actuators (such as mouse/keyboard replacement, building automation and mobile phone access).
1
INTRODUCTION
Table 1.
A vast number of people with disabilities in Europe

and worldwide are supported by Assistive Technologies (AT). More than 2.6 million people in Europe have
problems with their upper limbs and therefore many of
them depend on AT (EC 2005). As the potential of the
individual user is very specific, adaptive, ICT-based
solutions are needed to let this population group participate in modern society. Such solutions are rarely
available on todays market.
The different abilities of each user make assistive
technologies an ideal field for the application of rapid
prototyping methods. Each use case needs its own
individual application; thus every use case is a new
prototype.
AsTeRICS aims to provide a flexible and affordable
rapid prototyping toolbox for building assistive functionalities which can be highly adapted to individual
users needs. The scalable and extensible system allows
integration of new functions without major changes.
AsTeRICS provides people with severe motor disabilities with access to a standard desktop computer
but also to embedded devices and mobile services,
which have not offered highly specialised user interfaces before.
MOTIVATION
The last decades brought a considerable amount of

information and communication technology based
Assistive Technology (AT) devices to people with
disabilities. However, adaptation of software and hardware to the individual users abilities is often needed
before the technology can be used successfully. The
vast difference between the types of application for
each user often makes assistive technologies which
Distribution of disabilities in user sample.
Disability
Tetraplegia
Paraplegia
Spastic tetraparesis
Hemiplegia
Other type of mixed motor disabilities
Percentage
54.5%
21.5%
9.1%
6.1%
8.8%
have been optimised for one specific use case not

applicable out of the box in other situations. Moreover,
the comparatively small target group for AT solutions
makes AT devices expensive. This in combination with
the limits of adaptability of devices excludes some
people in need of assistance from being supplied with
the appropriate kind of assistive technology.
According to the methodology plan of the project,
a user requirement collection involving 33 persons in
three users sites was composed of a different subset of
primary (people with disabilities) and secondary users
(professional care givers). The sample of primary users
was composed of males and females, aged between
18 and 91 from Spain, Austria and Poland. The user
samples disabilities were distributed as indicated in
Table 1.
For the evaluation of primary users in Spain, Austria
and Poland, a common extensive questionnaire was
used. All participants were informed about the project
and signed a consent form for their participation in the
study.
Results from technology used from the user requirement collection involving these persons concluded
that:
567
97% of the users are able to use a PC regardless of their disability (email exchange 72%, chatting 57.6%, reading and watching films 48.5%,
general internet use 48.5%, videogames 36.4%,

work 21.2%).
32.3% of the PC users could not use the standard
mouse at all and 35.5% could use it with difficulties.
81.8% of the users could use a mobile phone, but
only 21.2% of those could properly use the numeric
pad. 57.6% could send SMS messages, but only
24.4% without any difficulty.
72.7% of the queried users had never used before
Smart Home facilities (e.g. to open doors, control
lights, blinds, temperature or home entertainment
devices) and would like to do so.
3
METHODOLOGY
AsTeRICS will provide a flexible and affordable rapid

prototyping toolbox for user driven Assistive Technologies or assistive functionalities. Sensors, processors and actuators are combined via an embedded
computing platform (or alternatively a platform with
higher processing power such as a laptop or desktop computer). A graphical and accessible configuration software enables to create prototypes out of
functional building blocks for sensors, processing elements and actuators. Therefore, the assistive solutions
realised with AsTeRICS facilities can be tailored to
the users abilities and needs without high adaptation and configuration effort and without additional
costs.
This approach allows to apply the following basic
principles:
Reusing and integration of standard sensors and

actuators.
Combination of components via the AsTeRICS
embedded platform, which processesssensor signals and controls actuators.
Reduced effort of adapting the platform to the users
needs and requirements through easy configuration
of the components.
Planned as an open source project (apart from components provided by commercial partners). Thus,
extensions with new sensors and actuators are possible for every developer once the results of the project
are made available under an open source license.
Home control integration to facilitate the users
interaction with his/her environment in the daily life.
Integration of modern and cutting edge technologies like vision systems for eye tracking and brain
computer interfaces (BCI).
Overall, the goal of the project is to enable more

people with disabilities to participate in the digital
society and living a more independent life. The idea is
to provide a system that not only can be tailored to an
individual user, but which also can be adapted to the
users change of abilities on a daily or long-term basis.
In terms of project implementation, the AsTeRICS
project follows the user centered design methodology:
The requirements for the technical implementation of
the AsTeRICS architecture were deduced from the user
requirement findings. After the development of the
first system prototype (May 2011), user evaluations in

three countries (Austria, Poland and Spain) will provide valuable feedback for improvements of the final
prototype.
4
STATE OF THE ART
Rapid prototyping which in the case of this project

resolves to visual modeling of processing chains is
known in many other fields of research and application. However, it is a new strategy for prototyping
Assistive Technologies. The following section will
give an overview of existing solutions for graphical
modeling, with special focus on features which are
important for the AsTeRICS project, such as user interaction capabilities or support of embedded runtime
environments.
Visual modeling of signal processing configurations allows building virtually any type of signal
processing chain without the need to know a particular programming language. A design is created and
adjusted visually by connecting graphical design elements via channels and editing properties of those
elements. The design flexibility is only limited by the
available processing elements and the supported inputand output modalities (e.g. hardware support for measurement equipment, sensor interfaces, interfaces to
control hardware or software support to display data
on the computer screen in an appealing way).
Especially in the area of scientific computing,
visual design has become very successful and is a
de facto standard for mathematical modeling in engineering and design of control systems today. Featurerich application packages like Matlab/Simulink or
LabView have contributed significantly to the success of visual design for signal processing, as their
use can reduce prototyping times for new scientific
approaches and even time-to-market of new product
solutions tremendously.
The idea of visual design for dynamic modeling is
quite simple: Many different problems (ranging from
simple simulations to complex systems) can be composed out of atomic functional blocks. Drawing a block
diagram or a flow chart diagram is a common method
in engineering to reduce the complexity of a system
or to illustrate its structure. By providing functional
blocks via graphical design software, creating even
large models becomes possible in an intuitive way,
without having to know about implementation details
of a programming language or operating system. The
block diagram or flow chart not only illustrates the
structure of a system, it also defines its functional
properties and implementation. Furthermore, the reuse
of well-tested functional units decreases development
time and increases robustness of the resulting model.
A visually designed signal processing chain is composed out of following fundamental types of design
elements:
568
Signal sources, having one or more output ports.

A signal source could be a signal generator, a file
reader, a measuring device like a volt-meter, a thermal sensor etc.

Mathematical operators, having at least one input
and one output port, e.g. digital filter, correlation/
coherence, average, threshold-level, integrator, FFTtransformation
Signal sinks (targets) with one or more input ports.
Possible sinks include sound-output, graphical output, device control, file archives etc.
In the field of assistive technologies, rapid prototyping has not been employed widely, however there
are some areas of research touching the same topics as
the AsTeRICS project, with Brain Computer Interface
(BCI) and Biofeedback applications being the most
prominent ones.
The BioExplorer (see www.cyberevolution.com)
application focuses on opto-acoustic feedback of
online EEG and ECG measurements. Key features are
interfacing to various biosignal acquisition hardware
and low-cost EEG devices, online signal processing and monitoring of signal features, recording and
playback of archive files and offline processing of
biosignal data. BioExplorer offers a small number
of feature-rich signal processing elements, mostly
dedicated to EEG and ECG analysis.
BioEra (see http://www.bioera.net/) by Proatech
LLC is a software application for signal analysis
and presentation with focus on biofeedback. BioEra
started as an open source project but has been commercialised after two years of development. A BioEra
design contains a signal flow model from input to output. The signal flow is passed through visually edited
objects (elements). Usually, BioEra is used together
with a biofeedback device which sends digitised bioelectric signals (EEG, ECG or other biosignals) to the
host computer via USB, serial port, wireless or other
media.
BrainBay (Veigl, 2006) is a Windows-based software application for graphically supported design of
biosignal processing configurations, targeted to online
biofeedback and to the development of alternative
human computer interfaces (HCI). BrainBay provides
several design elements dedicated to assistive functionalities like face detection from live camera images,
mouse-control or sending keyboard input to other running applications. A designer manual describes the
integration of new processing elements, making the
system open to future extensions like novel output elements for device control or interfaces to other assistive
technologies.
BrainBay allows adjustment of element-parameters
via respective dialog windows during system operation, the effects of parameter changes can be evaluated
in the running configuration. This is very useful compared to other approaches, where the running design
has to be stopped before changes can be applied.
Apart from visual modeling systems, several
camera based AT systems can be found, mostly
based on eye tracking (Betke 2002, Grauman 2000).
Furthermore, the projects BCI2000 (Mellinger 2007)
and OpenVibe (Renard 2010) provide flexible and
Figure 1. Schematic concept of AsTeRICS.
configurable environments for Brain Computer Interfaces, which mainly targeted for research purposes.
5 TECHNICAL CONCEPT
The basic concept of the AsTeRICS project is of the
combination of functional building blocks (sensors,
processors and actuators) to create working prototypes of assistive technology. The AsTeRICS Runtime
Environment (ARE) hosts Java OSGi plugins for
these building blocks. The OSGi framework enables
dynamic plugging and un-plugging of 3rd party components (bundles), enabling a scalable and modular
system architecture.
As shown in Figure 1, sensors and actuators are
connected to an embedded computing platform running the ARE by wire (USB, Ethernet) or wirelessly
(Bluetooth, ZigBee). The system can be configured
remotely using the AsTeRICS Configuration Suite
(ACS), which provides a graphical user interface to
set up and tailor the components to the specific needs
of the primary user.
Furthermore, the AsTeRICS project incorporates
a set of libraries intended for external developers.
These libraries aim to help developers to build or
adapt their application for people with motor disabilities. The libraries are delivered as native Microsoft
Windows DLLs, and they work independently of ARE.
Libraries already available are: Phone Library, Keyboard Library, Tremor Reduction Library and 3D
Mouse Library.
Currently, the first prototype of the AsTeRICS
platform is presented to users in a user evaluation
phase. In this prototypical implementation an embedded platform based on a Pico-ITX board with an Atom
Z530 processor is used. This hardware platform offers
six USB ports for the connection of sensor-and actuator modules used in a specific model as well as
standard input devices. The board has been extended
with a control panel featuring a graphical display and
navigation keys to provide an interface to the system.
This panel can be used to switch parameters of the
building blocks in specific use case scenarios and to
load different prototype models stored on the platform.
569
It also contains four general purpose inputs and outputs which can be connected to switches or provide
input to an actuator.
partner SENSORY (www.clarosoftware.com) serves

as a configurable user interface which allows selecting
letters, navigating a user menu or choosing functions
supported by a particular system configuration.
The ARE supports communication and exchange
of system models with the graphical configuration
suite via a TCP connection. Theoretically, the ARE
could run on different computing systems supporting
Java/OSGi, but is optimised for the AsTeRICS embedded computing platform (Intel Atom CPU, 16Ghz,
2MB RAM, 32GB SSD drive).
Apart from streaming data, which can be manipulated within the signal processing chain, a second
type of communication between components is supported by the AsTeRICs middleware: An event system
which allows sending sporadic messages between plugins with event inputs and outputs. Examples of such
events include a binary switch being pressed, a signal reaching a given threshold value, or an interesting
feature being detected in an EEG signal.
6 THE ASTERICS RUNTIME ENVIRONMENT
7 THE ASTERICS CONFIGURATION SUITE
The AsTeRICS Runtime Environment (ARE) consists

of a JAVA middleware which provides the OSGi interface and infrastructure for the software components of
the system (sensor elements, processing elements and
actuator elements).
Sensor elements connect classic AT-devices such as
binary switches and special joysticks but also emerging sensor techniques like computer vision for headand eye-tracking and Brain Computer Interfaces (BCI)
for EEG-driven computer use and writing. Electrophysiological data is acquired via the Enobio wireless
EEG/ECG/EMG sensor by AsTeRICS parnter Starlab
(see http://www.starlab.es).
Processing elements provide means to modify or extract data according to the applications
requirements. For the analysis of electrophysiological
signals, an extensive set of signal processing components has been implemented, including filtering,
decimation, FFT, threshold comparison, dissimilarity,
cross-correlation, epoch averaging or more complex
algorithms like eye blink detection.
Actuators are software interfaces to devices like
simple relais switches, digital-to-analogue converters but also more complex modules like keyboard-,
mouse- or joystick emulation, generic infrared remote
control, a Konnex(KNX)/EIB interface to existing
building automation systems or mobile phone access
via Bluetooth.
Channels connect inputs and outputs of the components in a model. Channels are used for representing
the data flow between the components. An example for
a functional AsTeRICS model can be seen in Figure 2.
For adjusting parameters of the running system or to
receive graphical feedback of system parameters and
sensor values, an optional touchscreen can be attached
to the Personal Platform. Via the touchscreen, the
OSKA on-screen keyboard developed by AsTeRICS
The AsTeRICS Configuration Suite (ACS) provides an

accessible graphical user interface to set up the systems processing chain and tailor the components to
specific needs of the primary user. Sensor and actuator modules are represented in the ACS by graphical
symbols with input and output connectivity (see Figure 2). Module-parameters like threshold level, gain,
sensitivity etc. define the mapping of sensor values
and the responsiveness of the assistive configuration.
Users and caretakers are offered means to modify key
parameters in a simplified version of the configuration
suite.
Figure 2. Configuration suite model for mouse control via

web camera.
USE CASE EXAMPLES
In the following, three use cases for mouse and

keyboard replacement will be described using block
diagrams created in the ACS. Mouse and keyboard
replacement are the most important use cases to
facilitate access to the digital society.
The ARE supports components for keyboard and
mouse replacement both on the same platform (e.g.
ARE running on a laptop) or on an external host
via USB connection (e.g. ARE running on personal
platform connected to desktop computer).
The first very simple use case is a proof of concept for creating mouse movement by tracking head
movements on a web camera. The block diagram (see
Figure 2) holds four components: a web camera sensor
component which detects the face of the user, an averager component for the face-coordinates which avoids
spikes in the mouse movement and the mouse actuator converting the coordinates to mouse movements
on the screen. This prototype is applicable for users
who have good control of their head movements. The
model does not include mouse click capacities.
570
Figure 3. Configuration suite model of six switch mouse control.
Figure 4. Configuration suite model of sip/puff switch controlled keyboard replacement.
The second model (see Figure 3) supports users who

can operate a set of binary switches such as push buttons or bend switches. The model uses an external
input module for binary switches with six switches
attached. Four of these switches are used for navigation just as in a joystick.The other two switches provide
two click buttons for the mouse replacement. Events
(like a switch press by the user) are passed via event
channels (see the red connections in Figure 3). Timer
elements are used to gradually increase the speed of
the mouse displacement, thus realizing an accelerated
mouse movement. In this model, the data sink is a
USB HID mouse emulator which represents a standard
mouse when connected to another computer.
Figure 4 shows a model for keyboard replacement
via a sip/puff switch. The user can use sip or puff
actions by using a mouthpiece connected to a pressure
sensor to operate a virtual keyboard.
The OSKA plugin controls the On Screen Keyboard
Application, which operates in a row/column scanning
method, allowing the user to first select the column in
which the desired key is located by moving a marker
via sip-and puff actions and in a second step select
the key in this column with the same actions (see
Figure 5).
The pressure sensor provides an analogue voltage
output representing the strength of the sip or puff.
Figure 5. OSKA in scanning mode with highlighted

column.
This voltage is read by an external ADC module and

the input is compared to the standard pressure value
the sensor is providing in default position. If the input
deviates far enough from the default, the value is
passed to an integrating element. This element either
increases or decreases its output according to the sip
or puff input, thus moving the highlighted marker on
the keyboard forward or back. Selection is performed
by dwell timing (longer period of sip/puff inactivity
selects column or key).
The AsTeRICS project aims to provide a new flexible approach in the field of assistive technologies,
571
using rapid prototyping methods for creating tailored, adaptable AT solutions. This approach keeps
the costs low through the use of standard components and the removal of development work for
adaptations.
The project is divided into two development-and
two user test phases. At the moment of writing, the
first prototype is undergoing several tests with users
at three test sites. The scenarios evaluated in these sessions include (among others) the examples described
in Section 8. The very first reactions by users have
been positive but no quantitative results can be named
at this moment. Results and findings of the user
tests will provide input for the second development
phase.
ACKNOWLEDGEMENTS
The AsTeRICS project is partially funded by the
European Commission under the Seventh (FP7
20072013) Framework Programme for Research and
Technological Development, Grant Agreement number 247730.
REFERENCES
Betke M., Gips J. & Fleming P. 2002. The Camera Mouse:
Visual Tracking of Body Features to Provide Computer
Access For People with Severe Disabilities. In IEEE
Transactions on Neural Systems and Rehabilitation Engineering. IEEE.
European Commision (EC). 2005. Eurostat: Population
and Social Conditions: Percentual Distribution of Types
of Disability by Sex and Age Group. http://epp.eurostat.cec.eu.int (accessed on June 13, 2011).
Grauman K., Betke M. et al. 2000. Communication via Eye
Blinks Detection and Duration Analysis in Real Time. In
IEEE Computer Society Conference on Computer Vision
and Pattern Recognition (CVPR). Hilton Head: IEEE
Computer Society.
Mellinger J. & Schalk G. 2007. BCI2000: A General-Purpose
Software Platform for BCI Research, In G. Dornhege et al.
(eds.), Toward Brain-Computer Interfacing, MIT Press.
Renard Y., Lotte F. et al. 2010. OpenViBE: An OpenSource Software Platform to Design, Test and Use BrainComputer Interfaces in Real and Virtual Environments,
In Presence: Teleoperators and Virtual Environment
19(1), MIT Press.
Veigl, C. 2006. An Open-Source System for Biosignal- and
Camera-Mouse Applications. In Miesenberger et al. (eds),
Computers Helping People with Special Needs. Linz:
Springer.
572

Comparison of five rapid prototype techniques

(SLS/FDM/DLP/3DP/Polyjet)
P. Inforatti Neto, M.F. Oliveira & J.V.L. Silva

Renato Archer Information Technology Center, Division of 3D Technology, Brazil
ABSTRACT: The rapid prototyping technology is an effective tool in making models for use in medical
applications. This paper proposes the study of rapid prototyping processes through the analysis of the technical
features measuring a standard model. The processes evaluated are SLS (Selective Laser Sintering), FDM (Fused
Deposition Modeling), DLP (Digital Light Processor), PolyJet and 3DP (Tridimensional Printer). This article
aims to describe the characteristics of materials used in these processes, such as roughness, hardness, surface
finish and dimensional analysis (using scanning inspection). This study also brings a brief description of the
concepts involved in each process and what parameters should be observed during processing.
1
1.1
INTRODUCTION
Rapid Prototype and Scanner Inspection
Rapid Prototype (RP) is a process that builds physical objects using additive manufacturing technology
(Wohlers, 2008). A virtual model can be created using
CAD/CAM system. The model shape should represent
a closed volume.
Scanner Inspection is a process that uses a noncontact device to capture a point cloud with millions
of points to inspect a model (Souza & Ulbrich, 2009).
1.2 The five RP technologies investigated
The five RP technologies investigated are described by
Volpato (2007) and Chua (2010) as following.
The SLS (Selective Laser Sintering) process builds
a model by sintering a CO2 laser beam hit in a thin layer
of powdered material.The interaction of the laser beam
with the powder raises the temperature of the powder to
the point of melting, fusing the particles to themselves
to form a solid.
The FDM (Fused Deposition Modeling) process
builds a model when the material in filament (in cartridge or spool) form is melted in a specially designed
head, which extrudes the nozzle. As it is extruded, it is
cooled and thus solidifies to form the model.
DLP (Digital Light Processor) is a projection
technology to solidify a liquid photopolymer which the
key device is the DMD (Digital Micromirror Device),
the producer of image. Each DMD chip has hundreds
of thousands of mirrors arranged in a rectangular array
on its surface to steer the photons with great accuracy. Each mirror is represented as one pixel in a
projected image.
The 3DP (Three Dimensional Printer) process creates parts by a layered printing process and adhesive
bonding, based on sliced cross-sectional data. A layer
is created by adding another layer of powder. The
powder is selectively joined where the part is to be
formed by ink-jet of a binder material.
In Objects Polyjet process the material is printed
on the tray and it is cured by the UV light which is
integrated with the jetting head. Parts are built layer
by layer, from a liquid photo-polymer. Both the part
material and support material will be fully cured when
they are exposed to the UV light.
This article aims to compare theses processes and
describes some characteristics of each one of that.
2
2.1 RP Samples
Figure 1 shows three-dimensional CAD model and
five prototypes created: SLS (Sinterstation HiQ
3D Systems), FDM (Vantage I Stratasys), DLP (Z
Builder Ultra Z Corporation), 3DP (Z Printer 310
Z Corporation) and Polyjet (Connex 350-Objet).
2.2 Visual analysis
The samples resulting were visual analyzed by optical
microscope (OM) with polarized light GX 51 model,
Olympus (Japan) trade mark.
elemental microanalysis
The samples resulting were visual analyzed by scanning electron microscopy (SEM), Leo 440i and
573
Table 3.
DLP (Digital Light Processor)
Machine
Z Builder Ultra (Z Corporation)
Process time 06h41

Model material (SI500)
Support material (SI500)
Table 4.
28,047 cm3
11,220 cm3
Total material cost
3DP (Three Dimensional Printer)
Machine
Z Printer 310 (Z Corporation)
Process time 01h00

Model material (ZP131)
Binder (ZB60)
28,047 cm3
32,300 ml
Total material cost
Figure 1. CAD model and the five prototypes used in this

case: SLS, FDM, DLP, 3DP and Polyjet.
Table 1.
Machine
28,047cm
$6,75
Total material cost
$6,75
Process time 13h47

Model material
(P400 MM ABS)
Support material
(P400 SR ABS)
$27,00
59,0 grs
$15,90
Total material cost
$42,90
34,345 cm3
$11,60
23,131 cm3
$8,79
Total material cost
$20,39
The samples resulting were digitalized by 3D Scanner

Orcus, trade maker 3D Spatium (Brazil) and analyzed
by Geomagic software.
3
3.1 Process time and material cost
Hardness
The samples resulting hardness were analyzed by

Micro Hardness Tester, trade maker Shimadzu.
Roughness
The samples resulting roughness were analyzed by

Perthometer, trade maker Mahr and MarSurf XR 20
software, V1.40-3 (SP1) version.
2.6
57,5 grs
2.7 Scanner Inspection
Vantage I (Stratasys)
EDS: 6070, LEO Electron Microscopy/Oxford trade

maker (Cambridge, UK).
For these samples, it was used accelerating voltage 20 kV and beam current of 150 pA to obtain the
micrographs and 600 to obtain the X-ray spectra.
Metallic coating was doing with a Sputter Coater
POLARON, SC7620 model, VG Microtech trade
maker (Uckfield, UK).
2.5
Process time 03h50

Model material
(Fullcure 720)
Support material
(Fullcure 705)
Connex 350 (Objet)
FDM (Fused Deposition Modeling)
Machine
2.4
$1,80
$4,95
$6,75
Polyjet
Machine
Sinterstation HiQ (3D Systems)
Process time 04h07

Model material
(Duraform PA)
Table 2.
Table 5.
SLS (Selective Laser Sintering)
$10,87
$4,35
$15,22
Dimensional analysis
The samples resulting were analyzed by Measuring

Microscope 100/50, Toolmakers Digital.
Material costs that were consumed, process time and

model machines for each RP technology are describing
in Table 1, 2, 3, 4 and 5.
The faster process for the execution of prototype
was 3DP with 01h00, followed by processes PolyJet
(03h50), SLS (04h07) and DLP (06h41). The FDM
was slower with 13h47, this is because each filament
is built by extrusion.
The processes at lower costs were SLS and 3DP
with $6,75 each, followed by DLP ($15,22) and FDM
($20,39). The most expensive was $42,90 with PolyJet.
This study considered only the material costs.
Investment in equipment and skilled worker were not
considered.
The values shown refer to the acquisition of material for research institution in Brazil (excluding taxes).
3.2 Visual analysis results
Figure 2 illustrates the result of the comparisons
using optical microscope (OM). Five regions were
analyzed: sphere, cone, rectangle, hole and thickness.
Magnifications range from 6, 7 to 30.
574
In the region of the sphere can be emphasized the

effect produced by the layers formed in the FDM
prototype. The 3DP model shows grains of resin in
post-processing. PolyJet is slightly transparent. DLP
presents a smooth shape followed by SLS.
In the region of the tip of the cone can be emphasized again the layers of filaments produced by the
FDM prototype. SLS was the one that came closest
to the final height created in the CAD model. 3DP
was the most rude and it is possible to see the grains
of resin. PolyJet is slightly transparent and presents a
smooth finish surface as DLP model.
In the square region can observe the behavior of
prototypes that need support in their creation. In the
FDM model is possible to observe the filaments and
the irregularity of shape due to removal of internal
support. The DLP model also needs support, and it
was removed. DLP has a slight deformation. In the
model 3DP can observe the additive layers of powder.
Polyjet and SLS present a regular basis.
In the region of the holes can be emphasized the
smooth finish of the SLS and DLP models. FDM
strategy highlights the fill and the outline of the hole.
3DP has highlighted the grain boundary and PolyJet
presents a salient and transparent boundary.
In the region of thickness can be analyzed the behavior of the thickness and the small details creation such
as the formation of the letter C. FDM strategy highlights the filling and formation of filaments extruded
layers. 3DP has highlighted the grains. SLS, PolyJet
and DLP have a regular shape.
elemental microanalysis results
3.3.1 SLS Technology
This model was created on the machine Sinterstation HiQ (manufactured by 3D Systems) and using
DuraForm PA material.
In Figure 3 it can be seen in (a) the detail of a corner
model that shows a contour irregularity (50 magnification). The detail (b) illustrates the outer surface
(400). A polymerized region is shown in (c) with
1.000 magnification and in (e) with 4.000 magnification. The detail (d) shows empty regions with
4600 magnification. Grain detail is shown in (f) with
magnification 12.000. Inner region is shown in (g)
with 50x magnification and in (h) with 1.000 magnification. The detail (f) and (g) show the inner region
of the prototype.
Figure 4 illustrates SLS elemental microanalysis
that shows the predominance of C (65,8%), followed
by O (28,82%). Other elements with significant concentrations are: Si (2,65%), Na (1,01%), Ca (0,74%)
andAl (0,6%). Elements as K (0,16%) and Mg (0,21%)
appear in low concentration.
Figure 2. Visual analysis by Optical Microscope (OM).
3.3.2 FDM Technology

Model was created on Vantage I (manufactured by
Stratasys) and using P400 MM ABS (as a model
material) and P400 SR ABS (as a support material).
575
Figure 3. SLS Contour detail: (a) 50 magnification;

sample outside: (b) 400, (c) 1.000, (d) 600, (e) 4.000,
(f) 12.000; cross section: (g) 50 and (h) 1.000.
Figure 5. FDM Contour detail: (a) 50, (b) 100, (g)

600; filaments overlap: (c) 50 and (e) 100 magnification; cross section: (d) 50; outside: (f) 400 and one grain
in (h) 7.000.
Figure 4. SLS Elemental Microanalysis.

Figure 6. SLS Elemental Microanalysis.
Figure 5 shows (a), (b) and (g) that are extruded filaments in sample outside with 50, 100 and 600
magnification, respectively. Items (c) and (e) illustrate
filaments overlap seen from the side with 50 and
100. Item (d) illustrates a cross section and the profile seen by an increase of 50. Item (f) illustrates
outside (400) and (h) illustrates a grain (7.000).
Figure 6 illustrates FDM elemental microanalysis
by O (22,01%). Other elements with significant

concentrations are: Na (1,53%), Ca (0,85%).
3.3.3 DLP Technology
Model was created on Z Biulder Ultra (manufactured
by Z Corporation) and using SI500 material to build
model and support parts.
576
Figure 9. Polyjet Sample outside: (a) with 50 magnification; surface: (b) 1.000 and (c) 7.000; cross section: (d)
50 amd (e) 1.000; grain details: (f) 10.000.
Figure 7. DLP Sample outside: (a) and (b) with 50 magnification; surface: (c) 400 and (d) 1.000; cross section:
(e) 1.000 and (f) 7.000; grain details: (g) 4.000 and (h)
10.000.
Figure 10. Polyjet Elemental Microanalysis.
by O (36,65%). Other elements with significant concentrations are Ca (0,43%) and Ti (0,44%).
Figure 8. DLP Elemental Microanalysis.
Figure 7 shows sample outside in (a) and (b) with

50 magnification. A polymerized region is shown in
(c) 400 and (d) with 1.000 magnification. Items
(e) and (f) illustrate a cross section by an increase of
1.000 and 7.000, respectively. Items (f) 7.000,
(g) 4.000 and (h) 10.000 illustrate details of grain.
Figure 8 illustrates DLP elemental microanalysis
3.3.4 Polyjet Technology

Model was created on Connex 350 (manufactured by
Objet), using Fullcure 720 (as a model material) and
Fullcure 705 (as a support material).
Figure 9 shows sample outside in (a) with 50 magnification. External surface is in (b) 1.000 and (c)
7.000. Items (d) and (e) illustrate a cross section by
an increase of 50 and 1.000, respectively. Item (f)
shows details of grain by an increase of 10.000.
Figure 10 illustrates Polyjet elemental microanalysis that shows the predominance of C (61,95%),
followed by O (37,65%) and Ca (0,40%).
577
Figure 13. 3DP Sphere: Elemental Microanalysis.
Figure 11. 3DP Contour detail and sample outside: (a)

50; cross section: (b) 100x magnification; sphere element:
(c) 150 and (d) 3.000 magnification; internal details in
cross section: (e) 1.000 and (f) 4.000; square element: (g)
3.000 and (h) 4.000.
Figure 14. 3DP Square: Elemental Microanalysis.

Table 6.
Prototypes Hardness
Hardness Hv (Vickers)
SLS
FDM
DLP
3DP
Polyjet
Figure 12. 3DP External: Elemental Microanalysis.
3.3.5 3DP Technology

Model was created on Z Printer 310 (manufactured by
Z Corporation). Materials used were ZP 131 (powder),
ZB60 (binder) and Z Bond (resin).
Figure 11 shows contour details in (a) with 50
magnification. In this image it is possible to see the
resin shell applied in post-processing. Items (b) 100
and (e) 1.000 also illustrates the inner part (in a cross
49,9
18,5
26,9
69,5
14,9
41,2
24,8
24,9
39,4
13,7
50,7
22,3
25,1
46,5
12,9
52,4
24,0
26,7
76,9
12,7
Average
44,5
25,4
29,5
110
13,7
47,7
23,0
26,6
68,5
13,6
section) where the resin finish is not infiltrated. Grains

with a sphere shape are showed in (c) 150 and (d)
3.000. Grains with a square shape are showed in (f)
4.000 and (g) 3.000. Crystals covering the grains
can be observed in (f) 4.000, (g) 3.000 and (h) by
an increase of 4.000.
Three types of materials have been found in these
samples and are described below.
Figure 12 illustrates 3DP external elemental microanalysis (Figure 11a). It shows the predominance of C
(36,93%), followed by O (30,29%), Ca (18,57%) and
S (14,21%).
Figure 13 illustrates 3DP elemental microanalysis
for the sphere element (Figure 11d). It shows the predominance of C (79,56%). Followed by O (19,91%),
Al (0,33%) and Ca (0,21%).
578
Table 7.
Prototypes Roughness
Ra (m)
Rz (m)
Rmax (m)
Rk (m)
Rpk (m)
Rvk (m)
Mr1 (%)
Mr2 (%)
A1 (m2 /mm)
A2 (m2 /mm)
Vo (mm3 /cm2 )
Wa (m)
Wq (m)
Table 8.
SLS
FDM
DLP
3DP
Polyjet
8,673
48,378
67,773
21,288
12,070
19,853
14,107
82,956
851,368
1691,959
0,169
14,661
16,690
3,798
22,579
35,883
8,677
2,211
13,530
3,685
78,649
40,746
1445,212
0,145
2,624
3,259
1,659
11,671
13,232
5,695
2,012
2,236
6,590
90,310
65,953
107,820
0,011
1,104
1,237
9,149
47,249
54,725
32,816
9,825
9,252
8,618
92,621
423,741
341,201
0,034
5,482
6,298
1,366
4,166
5,635
2,377
5,110
0,273
27,938
93,775
714,290
8,709
0,001
10,659
12,002
Polyjet
Dimensional analysis (mm)
Nominal Dim.
SLS
FDM
DLP
3DP
F1
7,000
F2
5,000
F3
3,000
F average
C1
7,000
C2
5,000
C3
3,000
C average
E1
0,700
E2
1,000
E3
3,000
E4
3,000
E5
1,000
E average
L1
56,000
Ht
28,000
7,223
5,233
3,180
0,212
6,905
4,978
3,032
0,028
1,219
1,189
2,949
2,759
1,085
0,100
55,463
28,047
6,950
4,952
2,867
0,077
6,799
4,728
2,902
0,190
0,739
0,959
2,910
2,772
0,856
0,093
55,841
27,963
6,870
4,960
3,022
0,049
6,875
5,003
3,060
0,021
0,683
0,927
2,906
2,986
0,890
0,062
55,826
27,537
6,120
4,150
2,534
0,732
7,549
5,518
3,473
0,513
1,222
1,539
3,454
3,447
1,553
0,503
56,333
27,945
Figure 14 illustrates 3DP elemental microanalysis

for the sphere element (Figure 11g). It shows the predominance of C (63,35%), followed by O (32,57%),
K (2,10%), Ca (1,12%) and S (0,86%).
3.4
3.6 Dimensional analysis

Table 8 shows dimensional average analysis. F1, F2,
F3 are cylinders. C1, C2, C3 are holes. E1, E2, E3,
E4, E5 are thicknesses. L1 is width and Ht is height.
The largest deviation was 3DP (0,732 mm).
Hardness
Table 6 shows the average of hardness prototypes. The

highest hardness was 3DP (68,65Hv), followed by SLS
(47,7Hv), DLP (26,6Hv), FDM (23,0Hv) and Polyjet
(13,6Hv). However, 3DP process presented a considerable variation in measurements because of surface
roughness in the act of measuring. Parameters used
were 0 m 5 kg in 15 s of indentation.
3.5
6,781
4,799
2,761
0,220
6,985
4,993
2,945
0,026
1,219
1,189
2,949
2,942
1,085
0,137
55,965
27,991
Roughness
3.7 Scanner inspection

Table 9 shows the dimensional errors that were found
in these prototypes. SP1, SP2 represent sphere diameters and Dist1 is the distance between the spheres.
Cil1, Cil2 represent cylinder diameters and Dist2 is the
distance between the cylinders. Processes presented
deviations between 0,072 mm (Polyjet) and 0,267 mm
(SLS and DLP).
4
Table 7 shows that highest roughness was 3DP

(9,1249 m), followed by SLS (8,673 m), FDM
(3,798 m), DLP 1,659 m) and Polyjet (1,366 m).
CONCLUSION
This paper has done an analysis comparative of five

rapid prototyping technologies using a CAD model.
579
Table 9.
Scanner Inspection analysis (mm)
Nominal Dim.
SLS
FDM
DLP
3DP
SP1
SP2
Dist1
Cil1
Cil2
Dist2
average
6,832
4,480
9,951
6,739
4,670
10,727
0,267
6,846
4,762
9,985
6,722
4,818
10,930
0,156
6,714
4,720
10,002
6,344
4,874
10,746
0,267
7,274
5,312
9,986
7,264
5,190
10,956
0,164
7,000
5,000
10,000
7,000
5,000
11,000
Parameters such as material cost and process time

show that 3DP was faster (01h00) and cheaper ($6,75).
The slower process was FDM (13h47) and PolyJet
was more expensive ($42,90). Visual analysis showed
especially filaments in FDM process and external finish in 3DP. Scanning Electron Microscopy showed
external and cross section images emphasizing the
formation of layers and structural forms, with emphasis on the filaments of the FDM and the different
3DP shapes. Elemental microanalysis results showed
that five processes are composed mainly of C, O
and Ca. The highest hardness was 3DP (68,65 Hv)
and highest roughness was also 3DP (9,1249 m).
Dimensional analysis showed largest deviation was
3DP (0,732 mm) and scanner inspection showed deviations between 0,072 mm (Polyjet) and 0,267 mm
(SLS and DLP).
Polyjet
6,872
4,866
9,990
6,840
4,978
11,022
0,072
ACKNOWLEDGMENTS
Authors would like to acknowledge INCT BIOFABRIS and CNPq for financial support.
REFERENCES
K. Chua; C.K.; Leong K.F. & Lim C.S. 2010. Rapid Prototyping: Principles and applications. 3rd Edition. Singapore:
World Scientific Publishing Co. Pte. Ltd., 2010. 512p.
Souza, A.F. & Ulbrich, C.B.L. 2009. Engenharia integrada
por computador e Sistemas CAD/CAM/CNC Princpios
e aplicaes. So Paulo: EditoraArtliber Ltda, 2009. 332p.
Volpato, N. 2007. Prototipagem rapida: Tecnologias e aplicaes. So Paulo: Edgard Blucher, 2007. 244p.
Wohlers, T. 2008. Wohlers Report 2008. State of the Industry,
Annual Worldwide Progress Report.
580

Droplet impact dynamics in ink-jet manufacturing

Wenchao Zhou, Drew Loney, Andrei G. Fedorov, F. Levent Degertekin & David W. Rosen
The George W. Woodruff School of Mechanical Engineering, Georgia Institute of Technology, Atlanta, GA, USA
ABSTRACT: Ink-jet printing enables a more efficient, economic, scalable manufacturing for a wide variety of
materials than other traditional additive techniques. The impact of droplets onto a substrate is critical for accuracy
control and optimization of the droplet deposition process. However, most previous research about droplet impact
focused on the spreading radius of the droplet, which does not provide enough information for manufacturing
purposes. This paper presents new methods to model and characterize droplet shape change during impact so that
the droplet deposition process can be optimized to build desired geometries. A validated numerical model is used
to study the shape change of the droplet impingement on a solid surface. A dimensional analysis is conducted to
reduce the number of parameters of the impact conditions by matching Reynolds, Weber, and Froude numbers as
well as possible. A wide range of impact conditions is simulated with the validated numerical model using both
physical and dimensionless quantities. In addition, a new method of characterizing droplet shape is presented
that measures its similarity to a desired shape. Using this measure, it is possible to track the evolving shape of a
deposit formed by multiple droplets.
INTRODUCTION
Ink-jet printing is an emerging technology that provides a more efficient, economic, scalable way of
manufacturing than other additive techniques. A good
understanding of the underlying physics of droplet
impingement onto a substrate is critical for accuracy
control and optimization of the deposition process.
Although the interaction of drops with surfaces has
been extensively and intensively investigated for over
a century (Worthington 1876), the development of
comprehensive, predictive models has been difficult
due to the complexity of the process and the interactions among many physical phenomena, including
fluid mechanics, phase change, and surface chemistry.
Nevertheless, numerous research progresses have been
reported during the past century. The contact angle
hysteresis phenomenon was also observed (Dussan
1979) and analyzed (Carey 1992; Nagai and Carey
2001). Experimental studies have been performed on
the time evolution of droplet shape on a solid surface
(Rioboo, Marengo et al. 2002) and on impact regimes
based on the combination of driving forces and resisting forces to simplify the problem (Schiaffino and
Sonin 1997). Splash phenomena have also been studied, different causes for splash have been found
(Worthington 1876; Xu, Zhang et al. 2005; Xu, Barcos
et al. 2007; Pepper, Courbin et al. 2008), and suggestions for avoiding splash have been proposed, all of
which are essential for ink-jet manufacturing. Different theoretical models (McHale, Rowan S. M. et al.
1994; Astarita 1997) and numerical models (Harlow
and Welch J.E. 1965; Sussman, Smereka et al. 1994;
Zhao, Poulikakos et al. 1996; Bussmann, Mostaghimi

et al. 1999;Yue, Zhou et al. 2006) have been proposed.
However, most of the previous research focused on
studying the change of a single variable the droplet
spreading radius, which cannot provide enough information for manufacturing purpose.This paper presents
a new measure to characterize droplet shape change
during impact so that the droplet deposition process
can be optimized to build desired geometries. With
the new measure of droplet shape characterization, a
dimensional analysis is first conducted to reduce the
number of parameters of the impact conditions. Then
a wide range of impact conditions are simulated with a
validated numerical model (Zhou, Loney et al. 2010).
Results show that it is possible to track the evolving
shape of a deposit formed by multiple droplets.
The rest of the paper is organized as follows. In
section 2, a measure of droplet shape characterization
is presented. Section 3 gives a dimensional analysis of
droplet impingement dynamics. In section 4, droplet
impact under a wide range of impact conditions are
simulated and the droplet size effects are presented.
Section 5 is devoted to studying the shape evolution
of successive droplets impact. Conclusions are given
in section 6.
2
DROPLET SHAPE DEFINITION
In order to build the desired geometry, it is desirable to

optimize the droplet impact conditions so that desired
geometries can be fabricated. To accomplish this, the
shape evolution of droplets during impingement needs
581
Figure 1. Half of the droplet cross section at one moment

during the impingement.
Figure 3. Change of R-coefficient during the droplet

impingement.
R-Coefficient (RC) can be defined to measure how

close the droplet shape is to the desired shape:
Figure 2. Droplet included by a rectangle.
to be characterized and compared with the desired

geometry. The many methods of shape representation
can be categorized generally into contour-based methods and region-based methods (Zhang and Lu 2004).
In this paper, the contour based shape representation
method is used. To characterize the droplet shape, the
contour of the droplet is obtained from the simulation
using image processing techniques; a resulting droplet
cross section is shown in Figure 1.
With the contour, we can get the coordinates of the
points on the contour, which are (x( j), y( j), where
j = 1,2,N). The centroid distance is used to represent
the shape of the droplet as a shape signature, which is
defined as the distance between the boundary points
and the centroid of the droplet shape. The centroid is
calculated by:
The centroid distance is defined as:
The R-coefficient ranges from 0 to 1 and 1

means the perfect match between the droplet shape
and the desired geometry. Figure 3 shows how the Rcoefficient is correlated with the droplet shape and
how it changes with time. It can be seen that the
R-coefficient correctly characterized the similarity
between the droplet shape and the desired geometry.
DIMENSIONLESS ANALYSIS
To understand the effects of different fluid parameters on droplet impingement, a dimensionless analysis
is performed on the numerical model (Zhou, Loney
et al. 2010). The phase-field method is used and the
governing equations are:
To overcome the problem of noise sensitivity, a Fourier

transform is applied to the centroid distance vector r:
The vector R can be used to characterize the droplet

shape. In the same way, we can define a vector RR
to characterize the desired geometry. In this paper, the
desired geometry is set to a uniform film (i.e., a rectangle in 2D). For uniformity, we define a rectangle that
includes the droplet as shown in Figure 2.
The similarity between the droplet shape and the
desired geometry can be evaluated by the Euclidean
distance between the two shape vectors R and RR,
representing the simulated and the desired droplet
shapes, respectively. Therefore, a similarity coefficient
where is the fluids density (kg/m3); u is the velocity vector (m/s); p represents the pressure (Pa);
denotes the dynamic viscosity (Pas); is the mobility (m3s/kg), is the mixing energy density (N),
is a dimensionless phase parameter and (m) is the
interface thickness parameter.
582
Table 1. Dimensionless
parameters.
parameters
and
physical
Dimensionless Parameters
Physical parameters
u = 1
=1
ink
air
= 0.0012
ink = 1
air = 0.0017984
= 1
g = 1
D=1
Re = 50
We = 69.4734
Fr = 2.0387e5
ink = ink /Re = 0.02
air = air /Re = 3.6e-5
= / We = 0.0144
g = g /Fr = 4.905e-6
u = 10 m/s
ink = 1000 kg/m3
air = 1.225 kg/m3
ink = 0.01 Pa*s
air = 1.7894e-5 Pa*s
= 0.07 N/m
g = 9.81 m/s2
D = 50e-6 m
Figure 4. Comparison of spreading radius change between

physical domain and dimensionless domain.
The physical parameters can be nondimensionalized with respect to the reference physical parameters:
where * stands for dimensionless, the subscript ref

denotes the reference parameter.
Then equation (5.a) can be rewritten as:
Figure 5. Comparison of R-coefficient change between

physical domain and dimensionless domain.
The simulation results are compared as shown in

Figure 4 and Figure 5. It can be seen that the results in
dimensionless domain match well with the results in
the physical domain.
Rearranging equation (6.a), we have
4
Based on this analysis, the solutions of the governing equations under different impact conditions
will be the same as long as these three dimensionless
numbers match with each other (The contact angle
between droplet and solid surface is limited to /2 in
this paper). Therefore, instead of dealing with 6 physical parameters, we can focus on these 3 dimensionless
numbers. A numerical simulation is performed using
this numerical model with both the physical parameters and their corresponding dimensionless numbers,
which are listed as in Table 1.
DROPLET IMPACT UNDER DIFFERENT

IMPACT CONDITIONS
This section is devoted to studying the droplet impact

behavior under different conditions, i.e. different combinations of the dimensionless numbers, so that we can
know how to control the impact conditions to make
the droplet shape closer to the desired shape. According to (Schiaffino and Sonin 1997), the droplet impact
behavior is mainly determined by the driving force and
resisting force of the impact and can be divided into
four different impact regimes as shown in Figure 6. The
Weber number can be used to characterize the driving
force and the Ohnesorge number, obtained from the
rearrangement of Reynolds number and Weber number (Oh = (We)1/2 /Re), can be used to characterize the
resisting force.
Another way to define Weber number and Ohnesorge number is to define them as a ratio of different
583
Figure 8. Change of Spreading velocity and R coefficient

with time in Regime I.
Figure 6. Regime map for the spreading (redrawn based on

(Schiaffino and Sonin 1997)).
Figure 7. Change of spreading factor and R-coefficient

with time in Regime I.
timescales. There are three different timescales during droplet impingement: spreading timescale spr ,
oscillation timescale osc and viscous timescale vis :
Weber number can be defined as a ratio of oscillation timescale to spreading timescale while Ohnesorge number can be defined as a ratio of oscillation
timescale to viscous timescale:
The impact behaviors are simulated in the four different regimes. In regime I, spreading is driven by the
impact pressure and resisted by inertia. In the simulation, the Weber number and the Ohnesorge number
are 100 and 0.01 respectively. The change of spreading
radius and the R-coefficient with time are plotted as
shown in Figure 7. As can be seen, the aspect ratio of
the optimal droplet shape is very low, which would
result in a lower manufacturing resolution than the
droplet size. In addition, oscillation is expected due to
the small Ohnesorge number (i.e. oscillation timescale
is smaller than viscous timescale), and oscillations are

across one dimensionless time unit because the Weber
number is larger than 1 (i.e. oscillation timescale is
larger than spreading timescale).
Another important observation that can be made
is the maximum R-coefficient is achieved after the
maximum spreading velocity is achieved as shown
in Figure 8 and before the droplet achieves its maximum spreading radius as shown in Figure 7. This
relationship holds true for all the regimes as verified in
the following. One possible reason is that the droplet
goes away from its equilibrium shape (i.e. sphere-like
cap or sphere) after it hits the substrate and before it
reaches its maximum spreading velocity because the
inertial force dominates the spreading at this stage.
After the droplet reaches its maximum velocity, the
droplet shape keeps going a little bit further away
from the equilibrium shape and starts to relax back
to its equilibrium shape before it reaches its maximum
spreading radius since at this stage the capillarity force
dominates the spreading.
In the second regime, spreading is mainly driven
by the capillary force imbalance at the contact line
since the impact velocity is so slow that its effects
can be neglected. Inertia resists the spreading, which
is followed by under-damped oscillation. In the simulation, the Weber and Ohnesorge numbers are 0.01
and 0.01 respectively and the results are shown in Figure 9. As can be seen, the aspect ratio of the optimal
droplet shape is relatively high, which could lead to a
better manufacturing resolution than the droplet size.
This suggests that the aspect ratio of the droplet shape
can be changed by changing the Weber number while
keeping the Ohnesorge number the same. Besides,
oscillation exists because the Ohnesorge number is
smaller than 1 and oscillations disappeared after one
dimensionless time unit because the Weber number
is smaller than 1. Similar to regime I, the maximum
R-coefficient occurs between the times of maximum
spreading velocity and maximum spreading radius, as
shown in Figure 9 and Figure 10, although in this
regime the three points are very close to each other.
In the third region, spreading is driven by the capillary force and is over-damped by the viscous force so
584
Figure 9. Change of spreading factor and R-coefficient with

time in Regime II.

with time in Regime II.

with time in Regime III.
that the inertial oscillations are absent. We simulated

droplet impingement with a Weber number of 1 and
an Ohnesorge number of 10. The results are shown in
Figure 11. As can be seen, the droplet shape starts to
change once it hits the substrate while the spreading
radius does not change much until a much later stage.
Once again, this confirms that R-coefficient can show
smaller and more detailed changes during the impingement process. In addition, the optimal droplet shape
does not go very far away from its equilibrium shape

with time in Regime III.

with time in Regime IV.
due to the high viscous force. Of possible concern is

that the sphere-like shape is likely to cause voids in the
parts during the manufacturing process and thus is not
a desirable shape. Tthe impingement process also goes
much slower which would lead to longer manufacturing times. In addition, the viscous timescale is smaller
than oscillation timescale (Ohnesorge number is larger
than 1) and thus the oscillation is damped out. Similar
to the other regimes, the maximum spreading velocity occurs first and then the maximum R-coefficient
and the maximum spreading radius occur, as shown in
Figure 11 and Figure 12.
In the fourth region, the impact pressure drives
the spreading and the viscous force over-damps the
spreading so that no oscillation happens. The Weber
number and the Ohnesorge number are both 100. The
simulation results are shown in Figure 13. Similar to
regime III, the high viscous force and the low inertial
energy significantly increase the spreading time and
thus the manufacturing time and thus this may not be
a desirable regime for ink-jet printing. This regime also
gives the same relationship between maximum spreading velocity, R-coefficient and maximum spreading
radius.
From the above simulation results, we can see that
the spreading factor gives us less related information
about the droplet shape for manufacturing purposes
while the R-coefficient can better characterize the
585
Table 3.
Impact conditions with different droplet sizes.
Droplet
Diameter
Corresponding
Impact velocity
Corresponding
Viscosity
50 m
200 m
600 m
1 mm
10 m/s
5 m/s
2.8868 m/s
2.2361 m/s
0.1060 Pa*s
0.2120 Pa*s
0.3672 Pa*s
0.4740 Pa*s

with time in Regime IV.
Table 2.
Material properties of the simulation.
Medium
Density
Dynamic
Viscosity
Surface
Tension
SR351
Air
1.109e3 kg/m3
1.1839 kg/m3
0.106 Pa*s
1.8616E-5 Pa*s
0.0361 N/m
droplet shape and give us more information about

how to get the desired shape by controlling the impact
conditions.
In the regime map in Figure 6, the Froude number derived from the dimensionless analysis was not
included. Therefore, a set of simulations was conducted with various Froude numbers and the same
Weber number and Reynolds numbers to see the effects
of the Froude number on droplet impact behavior.
Another reason for doing this is for the purpose of
experimentation since it is difficult to conduct experiments at the micro-scale and it is not easy to match up
all the three dimensionless numbers at the same time
when experiments are conducted at larger size scales.
Our interested material is SR351 resin (a photopolymer with fast cure response and numerous applications) (Sartomer) and the material properties at 25C
are listed in Table 2.
We are interested in the impact behavior of SR351
droplets at micro-scale for potential ink-jet printing
application. The Weber and Ohnesorge numbers are
set to be 153.6 and 2.37 respectively. The material
properties are as listed in Table 2 and the droplet
diameter is 50 m. The Weber and Ohnesorge numbers are matched up when the droplet size is scaled
up to 200 m, 600 m and 1 mm by changing the
impact velocity and material viscosity correspondingly as shown in Table 3. The simulation results are
compared as shown in Figure 15 and Figure 16.
As shown in the figures, the Froude number does
not affect the droplet impact behavior much (the discrepancies in the late stage of the impingement do
not matter much since we care about the time range
around the maximum R-coefficient), which suggests
Figure 15. Spread Factor change with time for different

droplet sizes.
Figure 16. R-coefficient change with time for different

droplet sizes.
that the regime map is effective in characterizing

droplet impact behavior and it is possible to conduct
the experiment in macro-scale to get an estimate of
the droplet impact behavior in micro-scale. This can
be further confirmed by comparing the Froude number with the Weber and Ohnesorge numbers. We can
see that in the above simulated impact conditions, the
Froude number is over 500 (for droplet size of 1 mm),
which is much higher than Weber number (153.6)
and Ohnesorge number (2.37). Referring back to the
dimensionless analysis in Section 3 (Equation (6.b)),
the contribution of Froude number is much less than
that of Weber number and Ohnesorge number.
586
Figure 17. Change of R coefficient with time for two

successive droplets impact.
SUCCESSIVE DROPLETS IMPACT
Besides studying single droplet impact behavior on

the regime map, we can also explore multiple droplet
impact behaviors with the defined droplet shape coefficient (R-coefficient) to see if there is a way to
improve the droplet shape without changing the impact
conditions. In this section, an ideal situation of successive droplets impact is investigated. The ideal situation
is where one droplet impinges on the solid surface and
is frozen instantly at the moment when its shape is
closest to the desired shape. Then the next droplet
hits on top of the previous frozen droplet and is
frozen instantly. This process is repeated for successive
droplets.
The impact conditions are set to be Weber number
equal to 71.42 and Ohnesorge number equal to 0.169
(or Reynolds number equal to 50) and the Froude number equal to 2.0387E5. The first droplet impingement
is shown in Figure 3 and is frozen instantly when its
R-coefficient reaches maximum. The second droplet
impingement on top of the first frozen droplet is plotted in Figure 17. Comparing Figure 17 with Figure 3,
we find that the best achievable shape is improved
(i.e. the maximum R-coefficient for two successive
droplets impact is larger than that of the single droplet
impact).
To further test this idea, the multiple successive
droplet impact simulations are conducted and the
R-coefficient is plotted against the number of droplets,
as shown in Figure 18. As can be seen, the shape of the
droplets is getting closer and closer to the desired shape
as the number of droplets increases, which suggests it
is possible to optimize the manufacturing process by
controlling the droplet deposition and solidification
process.
CONCLUSIONS
A shape coefficient is defined to provide more

relevant information of droplets impingement for
Figure 18. Change of best achievable shape coefficient with

number of droplets.
manufacturing purpose, which enables the optimization of the manufacturing process of ink-jet printing to
build the desired geometry. The definition of the shape
coefficient has been tested and is able to find the closest droplet shape to the desired geometry during the
droplet impingement. Then a dimensionless analysis
of the droplet impact is conducted and has revealed that
three independent dimensionless numbers determine
the droplet impact behavior. The comparison between
the simulation results in dimensionless domain and
physical domain validated the dimensionless analysis. A wide range of impact conditions on the regime
map that is composed of two of the three independent
dimensionless numbers (i.e. the Weber number and the
Ohnesorge number) have been simulated and provided
more information of the droplet impact behavior in
terms of the shape coefficient. The examination of the
effects of the third independent dimensionless number
(i.e. the Froude number) on the droplet impact behavior is performed and its shown that Froude number
does not affect the droplet impact behavior very much
under interested impact conditions, which makes it
possible to conduct the droplet impact experiments
in macro-scale to estimate the impact behavior in
micro-scale. Other than the investigation of the single droplet impact behavior on the regime map, the
multiple droplets impact has also been explored. An
ideal situation of successive droplets impingement is
examined, which shows that the droplets shape can be
improved toward the desired shape by increasing the
number of droplets under this ideal situation.
ACKNOWLEDGEMENT
We gratefully acknowledge the U.S. National Science
Foundation, through award DMI-0900322. Any opinions, findings, and conclusions or recommendations
expressed in this publication are those of the authors
and do not necessarily reflect the views of the National
Science Foundation.
587
REFERENCES
Astarita, G. (1997). Dimensional analysis, scaling, and
orders of magnitude. Chemical Engineering Science
52(24): 46814698.
Bussmann, M. & Mostaghimi, J., et al. (1999). On a threedimensional volume tracking model of droplet impact.
Physics of Fluids 11: 14061417.
Carey, V. P. (1992). Liquid-vapor phase change phenomena,
an introduction of vaporization and condensation processes in heat transfer equipment. Taylor and Francis,
Series in Chemical and Mechanical Engineering.
Dussan, E. B. (1979). On the spreading of liquids on solid
surfaces: static and dynamic contact lines.Annual Review
of Fluid Mechanics 11: 371400.
Harlow, F. H. & Welch J.E. (1965). Numerical calculation
of time dependent viscous incompressible flow of fluid
with free surface. Physics of Fluids 8: 21822189.
McHale, G. & Rowan S. M., et al. (1994). Frenkels method
and the spreading of small sperical droplets. J. Phys. D
27: 2619.
Nagai, N. & Carey, V. P. (2001). Assessment of surface wettability and its relation to boiling phenomena. Proceedings
of the 2001 ASME International Mechanical Engineering
Congress and Exposition, New York, NY. USA.
Pepper, R. E. & Courbin, L., et al. (2008). Splashing on elastic membranes: the importance of early-time dynamics.
Phys. Fluids 20: 082103.
Rioboo, R. & Marengo, M., et al. (2002). Time evolution of
liquid drop impact onto solid, dry surfaces. Experiments
in Fluids 33: 112124.
Sartomer. Sartomer Product Detail. 2011, from http://
www.sartomer.com/prodselectdetail.asp?apid=1&plid=
1&sgid=4&prid=SR351.
Schiaffino, S. & Sonin, A. A. (1997). Molten droplet

deposition and solidification at low Weber numbers.
Phys. Fluids 9(11): 31723187.
Sussman, M. & Smereka, P., et al. (1994). A level set
approach for computing solutions to incompressible twophase flow. J. Comp. Phys. 114: 146159.
Worthington, A. M. (1876). On the forms assumed by drops
of liquids falling vertically on a horizontal plate. Proc.
R. Soc. London 25: 261271.
Xu, L. & Barcos, L., et al. (2007). Splashing of liquids:
Interplay of surface roughness with surrounding gas.
PHYSICAL REVIEW E 76: 066311.
Xu, L. & Zhang, W. W., et al. (2005). Drop splashing on a
dry smooth surface. Phys. Rev. Lett. 94: 184505.
Yue, P. & Zhou, C., et al. (2006). Phase-field simulations
of interfacial dyanmics in viscoelastic fluids using finite
elements with adaptive meshing. J. Comp. Phys. 219:
4767.
Zhang, D. & Lu, G. (2004). Review of shape representation
andd escription techniques. Pattern Recognition 37(1):
119.
Zhao, Z. & Poulikakos, D., et al. (1996). Heat transfer and
fluid dynamics during the collision of a liquid droplet on
a substrate I. Modeling. International Journal of Heat
and Mass Transfer 39: 27712789.
Zhou, W. & Loney, D., et al. (2010). Impact of polyurethane
droplets on a rigid surface for ink-jet printing manufacturing. Solid Freeform Fabrication Symposium 2010,
Austin, TX.
588

Rapid Prototyping techniques for individualized medical prosthesis

manufacturing
A. Fiorentino, G.P. Marenda, R. Marzi & E. Ceretti
University of Brescia, Brescia, Italy
D.T. Kemmoku & J.V. Lopes da Silva

Center for information technology Renato Archer, Campinas, So Paulo, Brazil
ABSTRACT: Incremental Sheet Forming (ISF) process consists in the local deformation of a metal blank
enabling a flexible production of complex parts. It uses a punch moved by a CNC machine and it is suitable when
customized parts or small batch productions, short realization time and low cost are required. Moreover, since
forming forces are low, the geometrical accuracy of formed parts can be improved using full or partial dies. Those
can be made of resin, wood or realized by rapid prototyping techniques (RP). RP techniques realize parts adding
the material layer by layer. It can be a photocurable resin in form of liquid, a polymeric or metallic powder, a
wire or a solid sheet as a raw material. Products, such as medical prosthesis characterized by high customization
represent an interesting market for ISF. In this work, ISF Titanium and PCL prostheses were manufactured,
compared and discussed in order to study and optimize a reverse engineering and recreation process for human
body anatomy.
INTRODUCTION
Incremental Sheet Forming is a suitable process when

customized parts or small batch production quantities
are required in a very short realization time and at a low
cost. The process basic concept is based on the forming of a blank: a hemispherical tool locally deforms
a metal sheet following a predefined path (Jeswiet
2001).
Today there are many processes based on the local
deformation of metal sheet enabling the flexible production of complex parts, but the incremental forming
process is attractive also because it can be easily
accomplished by a three axis CNC milling centre
(Jeswiet et al. 2005).
There are two kinds of incremental sheet forming (ISF): Single Point Incremental Forming (SPIF)
with free sheet deformation and Two Point Incremental Forming (TPIF) where a die is present under the
sheet. In this case, the presence of the die improves the
part dimensional and geometrical accuracy. To reduce
the tooling cost, the die can be made of resin, wood,
low carbon steel or by Rapid Prototyping.
If traditional manufacturing consists in the subtraction of material from a block, Rapid Prototyping (RP)
allows manufacturing of models and components by
adding of liquid, powder, wire or rolled materials layer
by layer starting from a mathematical tridimensional
model defined in a 3D CAD.
RP processes are flexible, quick, and high automated and the model design can be evaluated on all
aspects with a 3D CAD. Moreover, the prototype can
be realized in a short time, in some cases hours, in
different types of materials and studied in its technical,

functional and aesthetic features.
The increasing of product complexity associated to
the reduction of the product life cycle and delivery
times have leveraged the development of RP techniques: a delay of several months of market entry time
of a product can also lead to loss of profits, while an
increase of development costs is acceptable if product introduction in the market is quicker and higher
quality is expected. The RP processes represent the
link between product design, equipment design and
pre-series models manufacturing.
The biomedical sector has an important social and
economic role in the innovation system because it
involves manufacturing areas with intensive research
integrating scientific and technological advances
developed in high-tech industries. Key fields for the
health industry are the product development, manufacturing and services related to medicine that help in
increasing the effectiveness and efficiency of health
system. The interesting fields are several, such as diagnostics, consumable equipments (surgical tools, prosthesis), therapy and rehabilitation. In the last years, the
biomedical sector is rapidly increasing, also because
of the population average age has increased and the
personal services demand, related to the high quality
of life, has grown. Therefore products, such as medical
prosthesis characterized by high customization, represent an interesting market for ISF (Ambrogio et al.
2005).
The basic cocepts involved in this work is to
apply a reverse engineering process for human body
anatomy in order to realize high customized prosthesis
589
(Wehmoller et al. 2004). In particular, it is developed

a technological process that leads to the RP of a die to
be used in ISFed implants. In order to determine the
geometry of the patients prosthesis, medical image
processing based on Computed Tomography or Magnetic Resonance have been used. (Wehmoller et al.
2000).
In the medical field the process of reverse engineering is necessary, in fact each human body is different,
so the concept of individualization becomes a key
factor.
2 THE PROCESS
The developed process can be divided in three macro
areas:
Figure 1. Segmentation of image dataset.
Medical image processing and 3D geometrical

model;
Die realization using RP processes;
Sheet incremental forming of prostheses.
2.1
Medical image processing and

3D geometrical model
Image processing is a noninvasive method to acquire

internal information about a patient. This technique
was first discovered by Roentgen more than a century
ago with the use of planar X-ray. Since then much more
advanced techniques have been developed. Today,
three-dimensional techniques such as Computerized
Tomography (CT) and Magnetic Resonance Imaging
(MRI) are of common use for diagnoses, surgical
planning and interventions. Basically an anatomical
image is obtained of internal structures in form of a
bi-dimensional (2D) dataset of images and exported
commonly in DICOM (Digital Imaging and Communication in Medicine) an international Standard (ISO).
The 2D dataset is processed by means of specialized software to separate the region and structures of
interest and reconstruct the anatomy in a 3D model
(Fig. 1).
In this case the InVesalius (www.softwarepubligo.
gov.br), a freeware software, was utilized for this purpose. After the model (Fig. 2) is generated in the
STL format (a standard for Rapid Prototyping communication) and exported to the RP machines to be
automatically produced. Before, some pre-processing
has to be done preparing the file and machine for
production.
2.2
Die realization using RP processes
Rapid Prototyping is a class of additive processes that

permits the construction of a physical model directly
from a virtual model. It is achieved by means of a
controlled deposition of layers of material. Nowadays,
Rapid Prototyping is being called Additive Manufacturing according to ASTM subcommittees. This
technology came up more than two decades ago to
Figure 2. 3D model of a palate.
speed up product development shortening the prototyping cycle. Since then, many applications have
been developed in other domains like in the medical,
archeological, and veterinary among many. In medical applications it provides an accurate replica of the
patients anatomy facilitating diagnosis, planning and
complex surgery interventions.
The patients anatomy consisting of a virtual model
obtained from medical image processing software
is automatically sliced by software and sent to the
machine that deposits a material as a sequence of layers. There are many processes commercially available
using polymeric materials in different forms such as
liquids, powders, sheets, and filaments. Solid materials such as metals, ceramics, and paper can also be
used. Each layer is glued on top of the other based on
different chemical and physical processes such as laser
material interaction, binder gluing, and heat extrusion,
as examples.
Theoretically any shape or geometry can be reproduced without limitations due to the deposition of
layers. There most common processes are: Selective Laser Sintering (SLS) based on the melting of
a fine polymeric powder by means of heating IR
lasers; Stereolitography (SLA) that is based on the
polymerization of a liquid resin with an specific wavelength initiator using the same wavelength lasers;
Fused Deposition Modeling (FDM) where a filament
is extruded and deposited continuously in thinner filaments of polymeric materials; and Tridimensional
Printing (3DP) where each layer of powder is glued
to the upper by means of ink jetting a binder. In this
research the SLS technology was used because of
590
Figure 4a. ISF equipment: working table.
Figure 3. A palate die produced by SLS process.
the speed to produce the dies, its good mechanical

properties and reasonable costs (Fig. 3).
2.3
Sheet incremental forming of prostheses
Since the high customization required by prosthesis

leads to production costs increasing, it is necessary to
adopt a flexible and low cost manufacturing process to
be efficient and competitive. In this work, ISF will be
evaluated, because it meets those characteristics and
moreover it enables to use different materials: metals,
non metals, polymer (Franzen et al. 2009) and to guarantee the accuracy control (Fiorentino et al. 2010). The
case study refers to palate forming tests performed on
a self designed incremental forming device mounted
on a CNC milling machine (Kopac et al. 2005). The IF
device (Attanasio et al. 2008) is equipped with a semispherical punch mounted on the machine spindle, an
instrumented table fixed on CNC table and a blank
holder which connects the palate die and the instrumented table and holds the sheet in the right position
(Figs. 4a, b).
3
MATERIALS AND EXPERIMENTAL TESTS
In this work, a reverse engineering process is developed in order to manufacture a palate prosthesis
made of biocompatible material (Schiller et al. 2004).
That is a material that has the ability to perform with
an appropriate host response in a specific application (Williams 1999). Prostheses can be manufactured
in a variety of biomaterials, including metals, polymers, ceramics and their composites (Teoh 2004).
Biomedical materials can be divided in:
Polymeric: Ultra High Molecular Weight Polyethylene (UHMWPE), polymethylmethacrylate
(PMMA), polyethyletherketone (PEEK), silicon,
Figure 4b. ISF equipment: deforming tool.
polyurethane (PU), polytetrafluoro-ethylene (PTFE),

polycaprolactone (PCL)
Metallic: stainless, cobalt-based alloy, titanium,
titanium alloy, nitinol, tantalum, gold, platinum
Ceramic: alumina, zirconia, carbon, hydroxyapatite, tricalcium phosphate, bioglass, calcium aluminate
Composite: carbon fibre (CF)/PEEK, CF/
UHMWPE, CF/PMMA, zirconia/silica/BIS-GMA
(Ramsden et al. 2007)
In this paper, the attention is focused on PCL
and Ti, gr 2. Titanium has been used extensively as
an implant material in different medical applications
since 1960 for its biocompatibility and properties such
as resistance to corrosion, high strength-density ratio,
low density and ability to join with the bone and
other tissues (osseointegration). The low melting point
(5964 C), exceptional blend-compatibility, high flexibility, medium Youngs modulus at room temperature
and blow or cast processability of PCL has stimulated
extensive research on its potential application in the
biomedical field since 1980 (Woodfuff et al. 2010;
Spevcek et al. 2007).
591
Table 1. Biopolyesters Properties and Titanium grade 2

Properties.
Melting Point, K
Glass Point, K
Cristallinity, %
Modulus, MPa
Elongation
at break, %
Tensile stress, MPa
Biodegradation, %
Water permeability,
kg/m2 /day
Physical Properties
PLA
PHBV PCL
PEA
PBSA
389
331
01
2050
426
278
51
900
338
212
64
190
385
244
33
262
387
228
41
249
15
100
100
>500
14
100
420
17
100
>500
19
100
0.177
0.680 0.33
0.172 0.021
Density
Mechanical Properties
Hardness, Rockwell B.
Hardness, Vickers
Tensile strength
Elongation at break
Modulus of elasticity
Fatigue strength
Shear modulus
Figure 6. Tool path strategy and CAM software path.
4510 kg/m3
80
145
344 MPa
20%
105 GPa
300425 MPa
45 GPa
Figure 7. Experimental equipment: Die, sheet in blank

holder and tool.
Table 2.
Process parameters.
Tool path
Punch
Lubrication
Spindle speed
Feed
Sheet dimensions
Repetitions
Constant Z movement,
Step depth z = 0.1 mm,
Spiral strategy
Head radius
Rs = 3 mm
Material
C45
Absent
06001000 rpm
200800 mm/min
1 mm thick,
140 mm width,
140 mm length
2
Figure 5. 3D Palate geometry.
The adopted punch has a semispherical head with a

radius of 3 mm, it was chosen to be minor than minimum radius measured on 3D geometry of the palate.
(Fig. 5).
The tool path was defined using CAD CAM techniques (Attanasio et al. 2008); after simulation and
preliminary tests, it was chosen a punch movement
from outside to center of final palate geometry, following a spiral path with constant z equal to 0.1 mm
(Hirt et al. 2004) (Fig. 6).
A die support is used to house the RP die. The sheet
to be deformed is positioned and fixed between the die
support and the frame (Fig. 7).
The sheet is 1 mm thick, 140 mm width and 140 mm
length (Table 2).
The final part profile was measured along 2 main
sections (Fig. 8) using a CMM machine and then it
was compared with the CAD profile.
A qualitative comparison between the CAD and part
profile is reported in Figure 9. In order to quantify the
geometrical Accuracy of the part (A), it was chosen
to measure the total area included between the ideal

(CAD) and actual (geometry) profile (Fig. 9).
Final parts were measured and evaluated. In particular PCL preliminary results showed the need of an
additional frame (Fig. 10) in order to reduce the sheet
elastic strain during the forming process as it is shown
in Figure 11.
According to table 2, tests at different feed and spindle speed were performed and the geometrical accuracy (A) results, calculated as described in paragraph
2, are reported in Figure 12.
For comparison in Figure 12 it is reported the accuracy (A) when it is considered a part profile which
has 0,25 mm offset form ideal one, this results in an
accuracy (A) of 20 mm2 .
It is possible to observe that a low feed and a high
spindle speed lead to an improvement of the accuracy.
592
Figure 11. Additional frame: elastic spring back effect.
Figure 8. Measuring section profiles.
Figure 9. Comparison cad and real geometry.
Figure 12. PCL test campaigns results.
Figure 10. Additional frame.
A possible explanation is a local increase of temperature, which is generated by an increase of friction at the
punch sheet interface and of time that punch remains
on the deformation point.
With regard to Ti sheets, different tests were performed. The final accuracy is significantly improved
with respect to PCL. (Fig. 13)
For example, in case of f = 800 mm/min and S = 0
rpm the Ti sheet accuracy (A) of profile 1 is equal to
20.3 mm2 .
Figure 13. Ti vs PCL accuracy.
This is due to the higher spring back that PCL sheet

undergoes.
Unfortunately, part finishing problems (Fig. 14) do
not allow obtaining acceptable prosthesis parts. In fact
the realized chipped surface could release Ti chips and
expose new and non oxidized material that is no more
compatible so creating problems to the patient.
593
Researcher Exchange for Biomedical Design and

manufacturing) project.
REFERENCES
Figure 14. Galling phenomena.
CONCLUSION AND FUTURE

DEVELOPMENT
In this work, it is proposed the use of anatomical

digital image processing, rapid prototyping and ISF
techniques to reverse engineering and manufacturing
a plate prosthesis. In order to realize a personalized
part, reverse engineering process has been applied to
a human palate and titanium and PCL prostheses were
manufactured. Results showed the consistency of the
process, and, moreover, that in PCL Incremental Sheet
Forming it is possible to improve the accuracy acting
on the process parameters (low tool feed, high spindle
speed). A possible explanation is given by the local
heating of the sheet due to friction action.
Ti showed a better accuracy but severe surface
finishing problems have to be solved.
In the future, the process will be optimized and
analyzed, also under the biological and clinical point
of view.
Initially, other test campaigns will be conducted
in order to discover the relations between parameters
(feed and spindle speed) and the accuracy for the PCL.
As far as titanium is concerned, process parameters
such as punch material, surface treatment, feed, spindle speed, lubrication will be investigated in order to
find a relation between the process parameters and the
sticking phenomena, and the surface roughness and
the dimensional accuracy.
In order to evaluate the capability of ISF technology to produce plate prosthesis for other application
(for example inner prostheses), biocompatible tests,
such as cellular growth and substrate strength, will be
performed.
Moreover, the developed process will be applied
to cranial or other plate prostheses. The technological
cycle presented in this work has a great potential for
customization of prosthesis in form of plates with a
very promising cost-benefit.
ACKNOWLEDGEMENT
This work was performed thanks to the FP7 Marie
Curie EU program and the IREBID (International
Ambrogio, G., De Napoli, L., Filice, L., Gagliardi, F. &

Muzzuppa, M. 2005. Application of incremental forming process for high customized medical product manufacturing. Journal of Materials Processing Technology.
Vol. 162163: 156162.
Attanasio, A., Ceretti, E., Giardini, C. & Mazzoni L. 2008.
Asymmetric two points incremental forming: Improving
surface quality and geometric accuracy by tool path optimization. Journal of Materials Processing Technology
Vol. 197, 13: 5967.
Attanasio, A., Fiorentino, A., Mazzoni, L., Ceretti & E.,
Giardini, C. 2008. Design of an equipment for forces
measurement in TPIF process. Proceedings of 9th ICTP:
17831788.
Fiorentino, A., Attanasio, A., Marzi, R., Ceretti, E. &
Giardini, C. 2010. On Forces, Formability and Geometrical Error in metal Incremental Sheet Forming.
Journal of Materials Processing Technology. Accepted for
publication.
Franzen, V., Kwiatkowski, L., Martins, P. A. F. & Tekkaya,
A. E. 2009. Single point incremental forming of PVC.
Journal of Materials Processing Technology Vol. 209, 1:
462469.
Hirt, G., Ames, J., Bambach, M. & Kopp, R. 2004. Forming
strategies and Process Modelling for CNC Incremental
Sheet Forming. CIRP Annals Vol. 53, 1: 203206.
InVesalius: public free software. www.softwarepubligo.gov.br
Jeswiet, J. 2001. Incremental single point forming. SME
Technical Papers. MF01246.
Jeswiet, J., Micari, F., Hirt, G., Bramley, A., Duflou, J. &
Allwood, J. 2005. Asymmetric single point incremental
forming of sheet metal. Annals of CIRP Vol. 54/2: 88114.
Kopac, J. and Kampus, Z. 2005. Incremental sheet metal
forming on CNC milling machine-tool. Journal of Materials Processing Technology Vol. 162163: 622628.
Ramsden, J. J., Allen, D. M., Stephenson, D. J., Alcock, J. R.,
Peggs, G. N., Fuller, G. & Goch G. 2007. The Design
and Manufacture of Biomedical Surfaces. CIRP AnnalsManufacturing Technology Vol. 56, 2: 687711.
Schiller, C., Rasche, C., Wehmoller, M., Beckmann, F.,
Eufinger, H., Epple, M. & Weihe S. 20047. Geometrically
structured implants for cranial reconstruction made of
biodegradable polyesters and calcium phosphate/calcium
carbonate. Biomaterials Vol. 25, 78: 12391247.
Spevcek, J., Brus, J., Divers T. & Grohens Y. 2007. Solidstate NMR study of biodegradable starch/polycaprolactone
blends. European Polymer Journal Vol. 43, 5: 18661875.
Teoh, S. H., 2004. Engineering Materials for Biomedical
Applications. Singapore: World Scientific.
Wehmoller, M., Lemke, H. U., Vannier, M. W., Inamura, K.,
Farman, A. G. & Doi K. 2000. TICC-Tomography, Image
processing, CAD, CAM: an advanced processing chain for
individual implant prefabrication. CARS 2000-Computer
Assisted Radiology and Surgery: 1057.
Wehmoller, M., Weihe, S., Rasche, C., Scherer, P. &
Eufinger H. 2004. CAD/CAM-prefabricated titanium
implants for large skull defects-clinical experience with
166 patients from 1994 to 2000. International Congress
Series Vol. 1268: 667672.
Williams, D. F. 1999. The Williams dictionary of Biomaterials
Woodruff, M. A. & Hutmacher, D. W. 2010. The return of a
forgotten polymer-Polycaprolactone in the 21st century.
Progress in Polymer Science Vol. 35, 10: 12171256.
594

Towards surface topography: Quantification of Selective Laser Melting

(SLM) built parts
A. Diatlov
Chair for Laser Technology (LLT), RWTH Aachen University, Aachen, Germany
D. Buchbinder, W. Meiners & K. Wissenbach

Fraunhofer-Institute for Laser Technology (ILT), Aachen, Germany
J. Bltmann
Department of Ferrous Metallurgy (IEHK), RWTH Aachen University, Aachen, Germany
ABSTRACT: Reduction of surface roughness is one of the key research issues within the additive manufacturing technique Selective Laser Melting (SLM) since one of the major cost factors is the post processing of
surfaces by means of milling, turning, etc. Time and resources needed for this post processing are most often
directly related to the surface roughness. In order to effectively reduce roughness during the SLM-process,
without reduction of melt pool size and therefore increase in production time, it is necessary to understand
the physical processes responsible for roughness emergence. First necessary step in understanding the physical processes responsible for roughness emergence in Selective Laser Melting (SLM) is obtaining the surface
topography of SLM built parts as exact as possible. Since the roughness of these parts can differ greatly between
Ra 1080 m with steep slopes partly >80 obtaining the topography data is not trivial. Measurement methods
are discussed with regard to the ability of mapping topography, using SLM built parts. Furthermore measuring
specifications are discussed including measurement area and necessary sample geometry and finally the concept
of roughness spectrum as an alternative to roughness value Ra is introduced and described. From this basis two
possible measurements devices are chosen for further work, requirements for sample geometry are defined and
first experimental results on single tracks and area measurements are shown.
INTRODUCTION
The market competition originated from countries

with low-cost work forces exerts pressure on companies worldwide and leads to a focus on innovation. Besides this, the increasing competition is also
compelling industries to improve the efficiency of
production processes, e.g. increasing the automation
level, improving process productivity or reduce post
production processes. Considering industrial production in high-wage countries today, these trends can be
cut down on two dilemmas that are closely related to
each other (see Fig. 1) [Schuh 2007]. The first dilemma
refers to the value-oriented vs. planning-oriented
production. The former approach focuses on valueadding processes (without consideration of planningpreparation-, handling- and transport processes) while
the latter focuses on extensive planning in order to
optimize value-adding (modeling, simulation, information gathering). The second dilemma is related to
the scale-scope dimension. Either the production
system is designed for high scale output without variances in the pro-duct design (critical masses, business
and manufacturing process decomposition, mastered
processes) or it is designed for individual products

down to a production batch of a unique product (onepiece-flow, complex and highly integrated processes).
The resolution of this production-related polylemma is
the main target of the Cluster of Excellence Integrative Production Technology for High-Wage Countries
(see Fig. 1).
Especially the scale-scope dilemma is boosted by
global trends like mass customization and open innovation which result in a demand for highly individualized products at costs matching or beating those
of mass production. One of the areas of greatest
potential for the resolution of this dilemma is rapid
manufacturing (RM) technologies due to their almost
infinite geometrical variability and freedom of design
without the need for part specific tooling. Selective
Laser Melting (SLM) is one of the RM technologies for metallic parts that additionally provides series
identical mechanical properties without the need of
downstream sintering processes, etc. which predestines it for individualized manufacturing. However,
the costly finishing operations make the method often
not cost efficient for series production and thus have
to be improved by reducing the surface roughness.
595
Figure 2. Schematic representation of the SLM process.
Figure 1. Polylemma of Production [Schleifenbaum 2009].
To achieve this the variables influencing the physical processes and effecting/causing the emergence of
roughness have to be better understood.
SLM PROCESS
The Fraunhofer ILT-developed SLM process is an

additive manufacturing process that fabricates metallic components layer by layer directly from
3-D-CAD data. This process enables the production
of nearly unlimited complex geometries. The starting
material used in the SLM process is a metal or ceramic
powder which is deposited as a thin layer (approximately 50 m) on a substrate plate. The powder is
selectively melted under an inert atmosphere by a laser
beam according to the CAD model (see Fig. 2).
Subsequently, the substrate plate is lowered by one
layer thickness and a new powder layer is deposited
above. Again, this layer is selectively melted and
metallurgically bonded to the layer below. The scan
direction is alternated after each layer in order to
deter imperfections, which may occur during the melting process, from growing throughout several layers.
Hence, the final component is built of many single
layers. The use of standard metallic powders and the
complete melting enables a density of approximately
100% which in turn assures mechanical properties that
match or even beat those of conventionally manufactured parts. All in all the SLM process enables a single
component to combine the benefits of high geometrical freedom and functional integration with series
suitable mechanical properties.
SURFACE ROUGHNESS
Generally there are three main aspects responsible for

properties of a surface: chemical composition, morphology and topography. All are important for quality
and technical characteristics of a part, but the role
played by topography is often underestimated. Former
researches into additive manufacturing mostly focused
Figure 3. Two perthometer measurements of a same

SLM-built part.
on the qualification of new materials and their industrial application. However, only little, if any research
concerning surface roughness of SLM-built parts has
been conducted yet. In order to come to a better
understanding of the variables influencing the physical processes and effecting/causing the emergence
of roughness on SLM-built parts it is necessary to
measure the surface topography of SLM-built parts as
exact as possible. While a tactile measurement device
for ex. perthometer can be used to receive 2 dimensional measurement data from a surface like Rz and
Ra, using the function z(x), three dimensional measurements on basis of function z(x, y) are needed to
achieve accurate information, since two linear profiles
on the surface of the same SLM-built part can differ
greatly (see Fig. 3).
3.1 Amplitude surface roughness parameters

Surface roughness parameters often used in showing the quality of a given surface are often amplitude parameters and are solely height descriptive (Ra ,
Rmax, Rz , Sa , Sq ).
Amplitude parameters:
Definition of Roughness parameter Rq :
It is the root mean square of the values of all points of

the profile also called root mean square.
596
Figure 4. Ra: arithmetic average of absolute values.
Definition of Roughness parameter Ra:
It is the arithmetic average of the absolute values of all

points of the profile, is also called centre line average
height.
Definition of Roughness parameter Rz :
It is the arithmetic average of the maximum peak to

valley height of the roughness valuesY1 toYi , i usually
equals five, of consecutive sampling sections over the
filtered profile.
Definition of area Roughness parameter Sa:
It is an arithmetical average height of the surface,

which can only be measured by optical measurement
devices and due to application of a significant area
has far smaller standard deviation, than measurement
methods using a line (see Fig. 5).
The difference in standard deviation between Ra and
Sa measurement is significant. By using a sample area
of 1,5 5 square millimeters the standard deviation of
Sa remains less than 5% for all reviewed samples (see
Fig. 5).
But while using optical roughness measurement
devices it is possible to receive a precise measurement
for surface roughness Sa , it is still a purely amplitude
parameter and does not express anything about form
or type of surface structure. While it conveys certain
information about the quality of the surface it does not
help process understanding or optimization.
3.2
Hybrid surface roughness parameters
Other measurement values are necessary. Roughness

parameters are scale-dependent and will depend on
the sampling interval and measurement scale. It was
demonstrated that variance of the height distribution
of a profile is linearly related to the measured distance
along the surface [Tholatha 1999]. Surface topographies often used in engineering practice and produced
Figure 5. Comparison between standard deviation of five

Ra measurements and five Sa measurements of the same SLM
built part.
by SLM frequently can be completely defined by two

characteristics: one related to amplitude of the waveform (roughness height distribution), other related to
spacing or wavelength [Sayles 1978].
Power spectra or autocorrelation function of the surface give information related to wavelength [Spragg
1970] and both are based on Fourier transform. Wavelength dependence can also be seen in optical measurement devices using cut-off filtering (both infinitely
sharp and one-or two-stage Gauss filters). These are
used in optical surface metrology to separate waviness
and form from surface roughness [Whitehouse 1994].
Experimental measurement and evaluation of surfaces
often must overcome three difficulties:
1) Measurement of topographies across the whole
range of interest often extends from nanometre to
the millimetre range. In such ranges the measurement is impaired by distortion of true surface profiles caused by envelope system [Tholatha 1999],
limited resolution (lateral and or vertical), accessibility of surface features, artefacts (optical in
non-contact type measurements because of microgeometry, reflectivity of surface [Windecker 1993],
in case of contact type measurements artefacts
emerge because of deformation of surface [Seewig
2010]). The consequence is that choosing the correct method to acquire topography of a given part
is a nontrivial and often challenging task.
2) 2-D surface profiles z(x) are comparatively easier
to obtain, but often do not represent the surface adequately, SLM built parts having irregular surface
structure are a good example of this (see Fig. 3).
Three dimensional data z(x, y) is therefore more
realistic approach to evaluating topography related
parameters [Seewig 2010].
3) Evaluation of topographical data in terms of usual
amplitude roughness parameters such as Rz, Ra
or Sa is often of limited value considering real
597
surfaces. [Whitehouse 1994], [Seewig 2010] It is

for example quite possible for two surfaces with the
same Ra values to behave completely differently in
a given situation [Seewig 2010], [Tomas 1999]. The
cause of such behaviour differences lies in random
elements on surfaces and fine roughness features
(usually low micron to submicron range), which can
be important for some applications but are often
hidden by the coarser contributions to roughness.
To avoid this, it is necessary to describe roughness
with wavelength dependent functions and not in
integral parameters [Whitehouse 2003].
Figure 6. Schematic representation of layer wise building

technique.
This paper presents the process of selecting an optical

topography measurement device capable of mapping
>90% of measuring points on a given part and exporting the data through application of a function z(x,
y) to a 3D data matrix to be processed further and
through application of filtering techniques address the
third point and extract wavelength dependent surface
topography in different wavelength ranges. For this
purpose roughness spectrum (Ra spectrum) will be
introduced.
4
4.1
EXPERIMENTAL METHODS
Experimental set-up
The preliminary experiments are conducted on an

ILT-constructed test rig that allows fast change and
adaption of components like laser source, F-Theta
lens, etc. in order to obtain a high degree of flexibility.
The laser source used in this setup is a 1000W
IPG fiber solid-state laser with Gauss beam profile
at the machining plane. The defocused imaging of the
optical fiber to 200 m at the machining plane still
exhibits a Gauss beam profile. So it is possible to
enlarge the beam diameter by means of defocusing.
During the experiments the beam diameter measures
200 m. Investigations are conducted using stainless
steel (1.4404) powder with a grain size distribution of
1030 m and aluminium (AlSi10Mg) powder with a
grain size distribution of 2545 m.
Figure 7. Sample geometry.
Figure 8. SLM built sample.
4.2 Test methodology

For the manufacturing of multi-layer components each
single layer is molten according to the 3-D CAD
model. The scan line direction is rotated 90 after each
layer. In order to smooth the components outer surface, the contour of each layer is scanned after the
hatchure is completed (see Fig. 6).
To measure all relevant areas of a SLM built part a
sample geometry covering all necessary attributes was
determined, those are: different surface wall angles,
top area, edges, corners.
All surfaces are big enough to inclose an area
of 1.5 5 square millimetre, necessary for measurement.(see Figs. 7, 8).
It is necessary to vary the surface wall angles

of the sample because roughness is a function of
different surface wall angles. However this will not
discussed or shown further in this paper, it will be
under investigation in the next paper.
After determining the sample geometry and deciding that 3D topography information is necessary for
further roughness analysis, it is now necessary to
approach the first difficulty and determine a measurement device capable of measuring the topography of
SLM built part in great detail and to reproduce results
as an x-y-z matrix for further analysis.
598
Figure 10. Topography of sample area acquired utilising

confocal laser scanning microscopy.
Figure 9. Overview of used topography measurement

methods.
MEASUREMENT METHODS
Since 3D topography data of SLM built parts is

required to analyze surface structures it is necessary
to use contactless optical topography measurement
device. However the surface of SLM built parts can
have steep slopes with angles greater than 80 and surface structures higher than 200 m, therefore achieving the necessary image quality with required lateral
resolution of less than 2 m/pixel and vertical resolution of less than 500 nm/pixel while at the same time
preventing quality loss due to optical artefacts and
surface reflectivity is far from trivial. An extensive
empirical comparison between different measurement
methods was conducted to determine the most suitablemethod (see Fig. 9). Several typical SLM-built
parts were used with the goal to acquire at least 90%
topography of sample field of 1.5 5 mm2 .
1) White light Interferometry
(New View 7300, Zygo)
Due to significant influence of optical artefacts it
was not possible to acquire even 70% of surface
topography, the method is not applicable.
2) Confocal white light microscopy
( surf explorer, nanofocus)
Due to influence of optical artefacts it was possible to acquire 70% of target area topo-graphy,
this method is only applicable for very approximate
measurements.
3) Chromatic confocal sensor
(Chrocodile X, Precitec)
Due to insufficient lateral resolution of 10 m/pixel
it was not possible to achieve satisfactory results.
4) Confocal Laser scanning microscopy
(ICM 1000 TCS SP2 TCS SL, Leica)
It was possible to acquire >95% of topography
using this method with 20x objective, but time interval necessary to achieve this was prohibitive, up to
6hours for 1.5 5 mm2 area (see Fig. 10).
5) Focus variation
(Infinite Focus, Alicona)
Best results: >95% topography of the sample area
were acquired using 20x objective the necessary
time for 1.5 5 mm2 is up to an hour (see Fig. 11).
Figure 11. Topography of the same area as in Fig. 10

acquired using Focus Variation.
From this point on all measurements are done using

Infinite Focus from Alicona, topography measurement device employing Focus Variation measurement
method.
RA SPECTRUM
Having addressed the first both problems brought up

in section 3.2 it is now necessary to think about another
possibility to measure roughness besides widely used
amplitude surface roughness parameters like Ra or Sa,
because on the basis of these parameters it is only
possible to compare surfaces that are manufactured
with the same manufacturing process. A comparison
of SLM built parts with machined or molded parts is
possible only with great caution, since no information
is given on the profile shape and it is so entirely possible to have two machined parts with the same Ra/Sa
value but very different mechanical properties.
For further development of SLM process the Ra or
Sa values are also not sufficient, since accurate quantitative evaluation of surfaces is not possible due to the
lack of information about the profile shape. For a systematic analysis of the experimental results, it would
be helpful to be able to segregate the surface profile
according to its wavelength sin short, medium and long
wave components. This allows identification of typical
surface structures, which is necessary for systematic
analysis.
To clarify the usefulness further it is necessary to
show an example.
599
Figure 14. Example of SLM built part from aluminium

powder (AlSi10Mg).
Figure 12. Sample of a metal structure with mill marks,
separation distance of milling gaps is approx. 240 m.
Figure 15. Ra and Ra spectrum of Aluminium SLM-built

part with different layer thickness.
7
Figure 13. = 160320 m: roughness maximum conforms with separation distance of milling gaps.
The topography of shown area with mill marks (see

Fig. 12) is acquired using Infinite Focus. The x-y-z
matrix is further separated in short, medium and long
wave components based on ISO 11562 with a phase
correct profile filter. The Ra values of the different
separated wavelengths shown as a graph results in Ra
spectrum. Figure 13 shows Ra spectrum of milled surface shown in Figure 12, measured at three different
positions.
A local maximum of roughness lies in area of
= 160320 m, which corresponds with separation
distance of milling gaps of approx. 240 m. Thus
using roughness spectrum in addition to roughness
also profile shape is shown.
The short-, middle-, and long wave surface structures can be analyzed separately and correlation of
process parameters and observed surface structures
becomes possible [Willenborg 2006]. This can further
lead to targeted optimization of SLM process.
FIRST EXPERIMENTAL RESULTS ON

SURFACES OF SLM BUILT PARTS
To show benefits of Ra spectrum as an alternative to

amplitude roughness parameters Ra and Sa, it will
be necessary to compare these values analyzing SLM
built aluminium samples (see Fig 14).
Two aluminium samples built as shown in (Fig. 14)
will be compared to each other. In this paper only
the marked with the arrow, 90 wall surfaces will be
compared, the difference between parts being the layer
thickness ds = 30 m in first case and ds = 50 m in
the other. The scan velocity of vs = 2500 mm/s and
laser power of 1 kW being identical (see Fig. 15).
Scalar values Ra and Sa can only directly be compared to each other and offer little extra information,
being very similar to each other in this case. Ra spectra, while both roughness maxima remain at = 320
to 1280 m and curve progression being very similar
to each other,do show the tendency of roughness maximum of the part built with ds = 50 m to move to the
right towards longer wavelengths.
Known parameters are grain size of the powder
being 20 to 45 m, layer thickness of powder being
30 and 50 m respectively, and beam diameter being
600
Figure 16. Parts built at commercial EOS SLM-system

ds = 30 m, beam diameter 80 m.
200 m. Still there are periodical structures with wavelengths = 320 to 1280 m. The dimensions of melt
pool are not exactly known, but due to heat propagation could be put in comparable order of magnitude
to the observed periodical wavelength. So the cause
of these surface structures can be hypothesized to be
in melt pool dynamics. This would also fit in the
observation of the slight shift to the right for the
50 m layer Ra spectrum, because of thicker powder
layer a bigger melt pool is created and bigger surface
structures amount to longer wavelengths. In order to
check this hypothesis a Ra spectrum of a part built
on commercial EOS SLM system with smaller beam
diameter of 80 m and consequently smaller melt pool
is calculated (see Fig. 16).
Ra spectra of both materials, Aluminium
(AlSi10Mg) and Nickel alloy (In718) show roughness maxima at comparably shorter wavelengths,
maxAl = 160 to 640 m and maxIn = 80 to
320 m which also checks out, considering that melting temperature of AlSi10Mg being 550 bis 575 C
is only a third of Inconel melting temperature of
ca. 1400 C. This would lead to bigger melt pool for
aluminium and consequently longer wavelengths of
Ra spectrum, this observation seems to reinforce the
hypothesis that melt pool dynamics are in large part
responsible for roughness maxima at longer wavelengths. But another examination can be undertaken,
since if the surface structures originated from melt
pool dynamics it should be possible to discover the
periodicity analyzing single tracks (see Fig. 17).
Periodicity confirmed at = 320 to 1280 m for
Aluminium AlSi10 Mg ds = 30 m, which corresponds exactly with the values measured in Fig. 15.
The hypothesis that these surface structures emerge
on the basis of melt pool dynamics seem to be viable.
For a Ra spectrum of stainless steel 1.4404 with
ds = 50 m we see two local maxima (see Fig. 18).
While the second maximum with = 160 to 640 m
probably emerges due to melt pool dynamics the first
maximum with = 5 to 10 m most probably has
another origin. This wavelength corresponds well with
partly molten powder grains which were originally
Figure 17. Single track Aluminium ds = 30 m.
Figure 18. Two localroughness maxima, at = 5 to 10 m

and at = 160 to 640 m Steel 1.4404, ds = 50 m.
1030 m in diameter. It is possible to at least detect

different kinds of surface defects using Ra spectra.
8
CONCLUSION AND OUTLOOK
The concept, that Ra spectrum offers more possibilities

to evaluate and to determine surface structures than
amplitude surface roughness parameters like Ra and
Sa has been demonstrated.
The strong possibility that a lot of surface defects
originate from melt pool dynamics was shown.
Possibility to be able to correlate different wavelengths to process parameters and observed surface
structures seems probable, as well as targeted optimization of SLM process as soon as correlation is
achieved. The possibility to separate surface structures depending on their wavelengths and discuss them
separately seems to held a lot of potential.
In the following paper the experimental results
which are only beginning to accumulate will be
discussed and analyzed in far more depth.
ACKNOWLEDGEMENT
The authors would like to thank the German Research
Foundation DFG for the support of the depicted
601
research within the Cluster of Excellence Integrative

Production Technology for High-Wage Countries as
well as the BMWi (Bundesministerium fr Wirtschaft
und Technologie) for the support of the depicted
research within the project GenSat (50YB1017) we
also would like to thank the project executor DLR
(DeutschesZentrum fr Luft-und Raumfahrt).
REFERENCES
Schuh, G., Klocke, F., Brecher, C., Schmitt, R. Excellence in
Production, 1st edition, Aachen: Apprimus-Verlag, 2007.
High Power Selective Laser Melting: A New Approach for
Individualized Series Production 2009 Aachen HenrichSchleifenbaum, Wilhelm Meiners, KonradWissenbach,
Christian Hinke.
DIN EN ISO 4287, ASME B46.1.

ISO 25178.
R.S. Sayles T.R. Thomas, Nature 271 1978.
Spragg, Whitehouse Proc. Inst. Mech Eng. 1970.
Whitehouse Handbook of surface Metrology, Institute
Physics Publishing, Bristol 1994.
Three-dimensional filtering of engineering surfaces using
envelope system Jackson Tholatha and V. Radhakrishnan
Department of Mechanical Engineering, Indian Institute
of Technology, Madras 1999.
Windecker, Tech. Mess. 60 1993.
Seewig Aachen Seminar Current and future norms of surface
measurement 2010.
T. Tomas, Rough surfaces Imperial college press 1999.
Whitehouse, Handbook of Surface and Nanometrology Inst
of Physics Pub 2003.
Edgar Willenborg. Polishing of Tool Steel with laser radiation. thesis RWTH Aachen, Shaker Verlag 2006.
602

An Additive Manufacturing method based on xerography

Y.E. Tan & C.K. Chua
ABSTRACT: A new powder based additive manufacturing (AM) technique based on the principles of electrophotography, called three dimensional xerography (3DX), is described. According to the principles of this
process, 3D objects are built layer by layer by selectively binding loose powder together using a binder powder.
The transport and deposition of the binder powder are achieved through the use of electric fields akin to that used
in electrophotography printing. The new method differs from an existing technique proposed by the University
of Florida in that binder powder is transferred onto the powder bed by electrostatic repulsion whereas the latter uses electrostatic attraction which necessitate generating increasing levels of potential difference across the
printer and the build platform as the height of the 3D part increases. A simple experiment, using mostly manual
steps, was designed and in the process of being constructed as a platform to investigate the viability of the new
technique.
1
INTRODUCTION
Additive manufacturing (AM) is an umbrella term

referring to a group of technologies used for building
physical models, prototypes, patterns, tooling components, and final production parts from computer
data, three dimensional scanning systems, or video
games (Wohlers 2008). AM form objects layer by layer
through the joining of liquid, powder, or sheet materials as oppose to traditional manufacturing techniques
like machining which are subtractive in nature. AM is
used to produce parts that are difficult or impossible
to fabricate using other techniques (Chua 1994).
Some of these technologies include stereolithography (Hull 1986), selective laser sintering (SLS)
(Bourell et al. 1992), three dimensional printing
(3DP) (Sachs et al. 1992), and laminated object
manufacturing (LOM) (Feygin 1994). Each technology has its own strengths and weaknesses (Chua et al.
1998 & Kochan et al. 1999). The number of AM
technologies is consistently growing as awareness,
acceptance and application of these technologies by
designers, engineers and other professionals increase.
The tissue engineering field is one area where AM has
been successfully applied (Bartolo et al. 2009). This
paper describes a novel AM method based on the principles of electrophotography called three dimensional
xerography (3DX).
Electrophotographic printing was invented in the
1930s (Carlson 1942). It consists of five main steps
that begin with the charging of a photoconductive substrate. The charges on the substrate are then selectively
dissipated according to data to be printed to create a
latent electrostatic image which serves to direct toner
particles to areas for deposition. The developed image
is a duplicate of the data to be printed and is transferred
to a second substrate like paper to be permanently

affixed with heat and/or pressure. Electrophotographic
printing has achieved a high level of productivity and
maturity in mainstream two dimensional (2D) digital printing but has limited application in AM (Jones
et al. 2010).
The use of electrophotographic printing in AM
has been investigated before. Some of the processes
are electrophotographic rapid prototyping (ERP)
(Kumar & Dutta 2003, Kumar et al. 2004) and
layer manufacturing using electrostatic imaging and
lamination (Liu & Jang 2002). ERP is a powder
based system that forms laminar cross-sections of
desired three dimensional (3D) parts using electrophotographic printing and involves five main steps with
the first four being identical to that used in 2D digital
printing. In the final step, the toner layer is transferred
onto a build platform (first layer) or a preceding layer
(subsequent layers) that was deposited on the build
platform to form parts by a layered printing process.
For conductive materials, toner transfer is carried out
by inducing an opposite charge on the top layer of the
part by applying an appropriate voltage to the build
platform. Thus, the toner layer, which is carrying a
first charge, becomes attracted to the top layer when it
is brought into its vicinity. For non-conductive materials, the build platform is similarly charged, but the
maximum height that parts can grow to is limited as
the effects of the electric field are not extended to
the top layer of the part. So far, parts approximately
1 mm in height have been reported (Kumar & Dutta
2003). To overcome this limitation, direct charging of
the top layer of the part was proposed. However, initial trials using this method revealed that the charges
on the top layer disrupt the image developing process and resulted in background printing. The second
603
process, layer manufacturing using electrostatic

imaging and lamination, also describes an AM technology that relies on electrophotography but did not
address the problems encountered by ERP. An alternate process combining the use of electrophotography and elastomeric materials had limited success
(Jones et al. 2010).
DESCRIPTION OF 3DX
3DX is a powder based AM technology that forms

3D parts layer by layer (Fig. 1). To begin, a uniform
layer of powder material is deposited on a substrate.
Then, a binder material is added to areas on the layer
of powder corresponding to cross-sections of 3D parts
being built to cause bonding (Figs 27). Background
areas are not modified and the powder there remains
loose and may be removed once the building process is
completed. Subsequent layers of the 3D parts are built
by repeating the process of deposition and bonding
on the topmost layer. Bonding is achieved between
the powder particles as well as the individual layers.
3D parts are separated from the loose powder upon
completion.
Key innovations of 3DX lie in the manner binder
material is added to the topmost layer of the part(s)
being built.
Formation of binder layers is achieved with electrophotography. A latent electrostatic image representing the cross-sections of 3D parts is generated on an
imaging member such as a photoconductive drum or
belt and developed with charged binder powder. The
binder powder is then heated before transferring to
the layer of loose powder in an image-wise manner to
cause bonding. The transfer process differs from existing technologies as the binder powder is repelled from
the imaging member as opposed to being attracted
away. In 2D electrophotography, the toner layer formed
on the imaging member is transferred to paper by
charging the later to a higher polarity than that existing
on the surface of the imaging member so that a resultant force in the direction of the paper acts on the toner
layer. However, previous experiments have shown that
this method is not suitable for 3D printing. Thus, a new
method is needed.
In 3DX, transfer of the binder layer is performed
using one of three methods.
Method 1 The first method creates an unstable
condition on the photoconductive surface of the imaging member by discharging it with a strong light when
a charged binder layer is still adhered to it. As the
majority of the binder particles share the same charge
polarity, a large portion of them are ejected from the
surface of the imaging member owing to mutual repulsion and deposit on the build platform. The repulsion
of the charged binder particles may be improved by
connecting the imaging member to a voltage source
equal in polarity to the binder particles rather than
merely grounding it. In this manner, the repulsion
force emanating from the imaging member is added
Figure 1. Schematic of 3DX.
Figure 2. Step 1 Charging. The charge applicator consists

of a corotron that ionizes the air above the photoreceptor so
that as the ions pass their charge to nearby areas of lower
potential, the photoreceptor becomes uniformly charged.
Figure 3. Step 2 Forming of latent electrostatic image.

A laser is shined across the photoreceptor to selectively dissipate the charge thereon to form a latent electrostatic image
corresponding to a cross-section belonging to the 3D parts
being fabricated.
to the force of mutual repulsion between the filler particles, thereby propelling them into the topmost part
layer with greater velocity and uniformity. This may
be accomplished by operating a controllable voltage
source.
604
Figure 4. Step 3 Development of latent image. The developer delivers a stream of binder material to the latent image
on the photoreceptor. The binder material is charged to the
same polarity as the background of the latent image and is
prevented from depositing in those areas.
Figure 6. Step 5 Heating of binder material. Binder material is heated with a heat source so that it is at least partially
melted.
Figure 5. Step 4 Transfer of binder material. The developed image is transferred to the donor member by applying a positive electric bias to the donor member while
simultaneously neutralizing the charges on the photoreceptor.
Method 2 The second method is similar to the first

except that charged binder particles are not directly
transferred to the topmost part layer or build platform.
Instead, a donor means, illustrated in figure 1 as a plate,
is used to transport the charged binder to the topmost
part layer. The charged binder particles are transferred
from the imaging member to the donor plate by electrically biasing the plate to a polarity opposite that of
the charged binder while concurrently exposing the
photoconductive imaging member to a light source
to discharge it. After the binder material has been
transferred to the plate, the plate is aligned with the
build platform and the biasing is reversed, so that the
charged binder is deposited on the topmost part layer
or build platform like in method one.
Method 3 To transfer conductive binder material from the imaging member to the topmost part
layer, method one and two cannot be followed. The
photoconductive surface becomes conductive when
exposed to light and will cause the charged binder particles to lose their charges as well. Once that happens,
the particles are neutralized and the repulsive force
existing between them is also lost. Consequently, the
binder particles may continue to adhere to the imaging
member due to van der Waals forces.
Thus, to transfer conductive binder material, the
imaging member is electrically biased to the same
polarity as that of the binder material and to a magnitude greater than both the image and non-image areas
Figure 7. Step 6 Transfer of binder material to loose part

powder layer. The electric bias applied to the donor member
is reversed from positive to negative causing the negatively
charged binder material to be propelled towards the loose part
powder layer.
of the electrostatic latent image on the imaging member, so that the binder particles are electrostatically
propelled to the topmost part layer. After the transfer
is completed, the imaging member is exposed to light
for discharging.
The binder material may cause bonding of the part
powder in several ways.
One way is to heat the binder material until it is
partially or completely molten. Also, the part powder is selected so that its melting point is the same as
or less than the boiling point of the binder. Accordingly, when the molten binder comes into contact with
the part powder, it permeates into the gaps existing in
between the particles thereof via capillary action. As
the temperature of the molten binder is the same as
or higher than the melting point of the part powder,
the outermost layer of the particles of the part powder melts, and upon solidification, coalesces at their
contacting surfaces. The binder material is incorporated into the final part. For this method, the bonding
of the part layer is dependent on the melting and subsequent coalescence of the part material. As a result,
the range of materials that can be selected is limited to
those with a melting point that is less than or the same
as the boiling point of the binder material.
605
To expand the range of materials that can be readily processed by 3DX, another method was developed
whereby no melting of the base material is required.
In the second method, the inter-particle and interlayer bonding is carried out using only the binder.
This method operates in a manner similar to the previous method whereby the binder material is either
partially or completely melted and allows to permeate
the topmost part layer. However, the part material is
not melted and the bonding of the layer is provided
solely by the solidification of the binder.
3
EXPERIMENTAL SETUP
Plans to investigate the viability of 3DX are underway.

Potential challenges in implementing 3DX have been
identified and they include the use of electric fields to
transport binder particles from the imaging member to
the topmost part layer.
Studies on the detachment and attachment of small
particles like toner have been done before as such
phenomena are fundamental to the electrophotographic printing process (Nagayama & Takeuchi 2000,
Yasushi & Suda 2007). However, most of the studies conducted were done on solid particles and very
little were on molten ones. An experimental setup,
using mostly manual steps, has been designed and
procurement for the necessary hardware has begun.
The objective of the experiment is to simulate the
actual 3DX process with an emphasis on determining
the factors that influence the detachment and attachment of molten non-conducting particle using electric
fields.
3.1
Equipment
An amorphous selenium photoconductive plate

(unknown manufacture) was selected as the imaging member. A Simco Pinner bar charge applicator
(Simco-ion, Pennsylvania, U.S.A) powered by a Spellman 23020R high voltage power supply (Spellman,
Valhalla, NewYork, U.S.A) was used to charge the selenium plate. A Trek 884 non-contacting electrostatic
voltmeter (Trek Inc., Nedina, New York) monitors the
surface potential on the selenium plate. Low density polyethylene powder (Innotek Powder Coatings
LLC., Big Spring, Texas, U.S.A) is used as both
the part material and binder material. The powder is
deposited on the selenium plate with a Parker Ionics GX7000CS powder coating unit (Parker Ionics,
Westland, Michigan, U.S.A). Heating of the LDPE
powder is performed by an induction heater (unknown
manufacturer).
3.2
Experimental procedure
Part of the experiment is carried out in a dark room.The

selenium plate is positioned on a grounded conductive
substrate and charged with the charge applicator by
making end-to-end sweeps in the direction of the long
edge of the selenium plate. At the end of each sweep,

the surface potential of the selenium plate is measured
with the non-contact electrostatic voltmeter. The selenium plate is charged to around 500 to 1000 V. To
develop an electrostatic latent image on the selenium
plate, the plate is covered with a sheet of paper printed
with a negative image of a cross-section of a sample 3D
CAD model. A line laser is shined over the entire paper
to selectively dissipate the charges on the selenium
plate according to the negative image. Since a negative
image is used, the background remains charged while
the cross-section is neutralized. Low density polyethylene powder (LDPE), which is pre-disposed to being
negatively charged, is deposited on the charged selenium plate using the powder coating unit by a method
called discharge area development (DAD).The powder
coating unit charges the LDPE powder to a negative
polarity which is the same polarity as the background
areas on the selenium plate. Since like charges repel,
the LDPE powder deposits in neutral areas but does
not deposit in background areas. Thus, an image of the
cross-section in LDPE powder on the selenium plate
is obtained.
A Teflon coated metallic substrate is aligned with
the developed image and placed 500 m above the
selenium plate. The selenium plate, together with the
developed LDPE layer, is exposed to light. Charges
remaining on the selenium plate are completely dissipated and an unstable condition on the plate is created
which releases the LDPE powder from the plate.
Simultaneously, a positive electric bias is applied to
the substrate to attract and transfer the LDPE powder
from the selenium plate to the metallic substrate. The
metallic substrate is then transferred to the build platform containing loose part powder and heated with the
induction heater. As the metallic substrate heats up, the
LDPE powder adhering to it melts. The electric bias
applied to the metallic substrate is then reversed from
positive to negative causing the negatively charged
LDPE powder to be propelled towards the loose part
powder layer on the build platform. Areas deposited
with binder material join together upon solidification
of the binder to form the desired cross-section. The
entire process is repeated as necessary to create more
layers.
Results are imaged by a microscope.
CONCLUSION
Electrographic printing is well-established in 2D printing but its application in AM is limited up to now. 3DX
is a new powder based AM technology based on the
principles of electrophotographic printing that forms
3D parts layer by layer. The new process provides an
alternative to existing powder based AM techniques
like SLS or 3DP. The use of electrophotography in
the AM process has the potential to contribute to
a decreased reliance on mechanical devices such as
the ink-jets in 3DP systems which need to be maintained or replaced on a regular basis. Additionally,
606
the printer is able to transport and deposit a variety

of binder materials including those that are not
amenable to being jetted by ink-jet printing devices.
Also, as electrophotography is a mature technology
with many components available off-the-shelf, the
cost of development and ownership will be minimized. 3DX seeks to address some of the challenges encountered by earlier pioneers in this field,
namely the transfer of binder material from the
photoconductive imaging member to the part powder. Potential challenges in implementing 3DX have
been identified and a simple experiment consisting of mostly manual steps has being planned.
Through the experiment, the authors of this paper
hope to determine the detachment and attachment mechanics of molten particles which will be
instrumental in the understanding and implementing
of 3DX.
ACKNOWLEDGEMENT
The authors would like to express their gratitude
towards SPRING Singapore for providing financial
support to the project.
REFERENCES
Bartolo, P.J., Chua, C.K., Almeida, H.A., Chou, S.M. &
Lim, A.S.C. 2009. Biomanufacturing for Tissue Engineering: Present and Future Trends. Virtual and Physical
Prototyping 4(4): 203216.
Bourell, D.L., Marcus, H.L., Barlow, J.W. & Beaman, J.J.
1992. Selective laser sintering of metals and ceramics.
The International Journal of Powder Metallurgy 28(4):
369381.
Carlson, C.F. 1942. Electrophotography. US Patent No. 2297691.
Chua, C.K. 1994. 3-Dimensional Rapid Prototyping
Technologies and Key Development Areas. Computing
and Control Engineering Journal 5(4): 200206.
Chua, C.K., Chou, S.M. &Wong, T.S. 1998. A Study of the

State-of-the-Art Rapid Prototyping Technologies. International Journal of Advanced Manufacturing Technology
14(2): 146152.
Kochan, D., Chua, C.K. & Du, Z.H. 1999. Rapid Prototyping
Issues in the 21st Century. Computers in Industry an
International Journal 39(1): 310.
Feygin, M. 1994. Apparatus and method for forming an
integral object from laminations. US Patent No. 5354414.
Hull, C.W. 1986. Apparatus for production of threedimensional objects by stereolithography. US Patent No.
4575330.
Jones, J., Wimpenny, D., Gibbons, G. & Sutcliffe, C. 2010.
Additive manufacturing by electrophotography: Challenges and successes. 26th International Conference on
Digital Printing Technologies: 549553.
Kumar, A.V. & Dutta, A. 2003. Investigation of an electrography based rapid prototyping technology. Rapid
Kumar, A.V., Dutta, A. & Fay, J.E. 2004. Electrographic
printing of part and binder powders. Rapid Prototyping
Journal 10(1): 713.
Liu, J.H. & Jang, B.Z. 2002. Layer manufacturing using
electrostatic imaging and lamination. US Patent No.
6376148.
Nagayama, M. & Takeuchi, M. 2000. Particle adhesion force
measurements by electric field detachment method. 2000
International Conference on Digital Printing Technologies: 731735.
Sachs, E., Cima, M., Williams, P., Brancazio, D. & Cornie,
J. 1992. Three dimensional printing: Rapid tooling and
prototypes directly from a CAD model. Journal of Engineering for Industry 114: 481488.
Wohlers, T.T. (ed.) 2008. Wohlers Report 2008: State of the
Industry. Fort Collins: Wohlers Associates Inc.
Yasushi, H. & Suda, K. 2007. Toner jumping phenomenon
in an electric field. Particulate Science and Technology
25(2): 147162.
607

Preliminary investigation on cellular support structures using SLM process

Ahmed Hussein, Chunze Yan, Richard Everson & Liang Hao
College of Engineering, Mathematics and Physical Science, University of Exeter, Exeter EX4 4QF,
ABSTRACT: Selective Laser Melting (SLM) is an additive manufacturing (AM) technique in which 3D metallic parts are manufactured in a layer-by-layer fashion, typically in small series. To build complex geometries with
overhanging and undercut surfaces, SLM has to solidify expensive materials into sacrificial support structures in
order to control curling and shrinkage of the part. The majority of previous studies on support structures concentrate on non-metallic processes. This study aims to explore the potential of using cellular structures to support
metallic parts manufactured by SLM. A key advantage offered by the cellular structures is their low volume
fraction which provides opportunities to greatly reduce the volume of support materials and build time as well
as allow for easy removal of raw powders which are trapped inside these sacrificial supports. The preliminary
experimental results reveal that these cellular support structures have the potential to be used for supporting
metallic parts and prevent part deformation.
Keywords: Additive manufacturing; Selective laser melting (SLM), Cellular structure, Support structure
INTRODUCTION
Over the last decade, significant progress has

been made on developing new and advanced techniques for additive manufacturing (AM). Both practitioners and academicians in the field have been
striving in developing ways in which sustainable
and economical products can be additively manufactured. Selective Laser Melting (SLM) is an
additive manufacturing process which has excellent prospects in the manufacturing of complex
lightweight metal components and cellular lattice
structures.
Support structures are essential for AM metallic
parts to build the overhanging and undercut surfaces
in complex geometries and to control part curling and
deformation during layer-by-layer fabrication. These
supports affect the material use, build-time, surface
finish, energy consumption and post-processing time
of the manufactured parts. Minimising the amount of
support structures could reduce the time needed to
remove the part from the platform and therefore help to
reduce manufacturing costs (Frank and Fadel, 1995).
A suitable part deposition orientation can improve
part accuracy and surface finish as well as reduce
the production time and support structure required for
building a part (Pham and Demov, 2001). Therefore,
researchers have developed methods to minimize the
support volume by changing the orientation of the
supported part (Frank and Fadel, 1995; Allen and
Dutta, 1995; Pham, et al. 1999).
It is worth mentioning, that the majority of the previous studies on support structures were specifically
developed for non-metallic processes such as Stereolithography (SLA) and Fused Deposition Modelling
(FDM). The current solutions for support generation
are either automatic or manual. The later depends
on the experience of an operator to manually design
and customize the support structure depending on the
requirement for a particular job. Whether automatic
or manual, the efficiency and reliability of the generated support is always under question, as it is difficult
to determine if the selected support is excessive or
inadequate for the part.
In SLM, support requirement is even more complicated. The thermal stresses cause distortions and
cracks in the part if inadequately supported Therefore, metallic support structures should be able to
support and take heat away from the part. Additionally, the removal of metallic supports from the part
and base plate is a tedious job sometimes requiring
machining. Particularly, large amount supports for delicate parts would increase the difficulties of support
removal and even destroy fine details on the downward
facing surfaces
In order to utilize the full potential of AM technology, it is important to investigate the opportunities
in developing new and improved designs for metallic support structures. This study aims to explore
the potential of using cellular structures to support
metallic parts manufactured by SLM. A key advantage offered by cellular structures is their low volume, which offers opportunities to reduce the support
material volume and increase the ease with which raw
powders trapped inside these sacrificial supports can
be removed.
609
Figure 1. Unit cell types used for support structure

generation.
Figure 3. SLM manufactured parts with support structures.
Stainless steel (316L) powders with average particle

size of 45 10 m were used to build the test samples. A SLM MCP-Realizer machine with 100 w CW
Ytterbium fibre laser, operating within 10681095 nm
was used. TalySurf from Taylor Hobson Ltd was used
to measure the deflection of test parts. This is the 3D
surface topography profiling instrument that moves
the workpiece under a stationary gauge head.
Figure 2. Design of support structures with the part.
MATERIALS AND EQUIPMENT
DESIGN OF CELLULAR SUPPORT

STRUCTURES
Figure 1 shows the unit cell types used for the generation of the support structures. These unit cells are
generated through the software provided by Simpleware Ltd. These are periodic lattice structures which
could assist in reducing the time and material volume
needed for building support structures. The volume
fraction defines the relative solid volume that is inside
the generated support. The lower the volume fraction
is, the more open the support structure will be. If the
value of volume fraction is very low, it may result in
loss of connectivity between adjacent cells in the support; very high values may result in closed support
volume.
To investigate the deformation of the parts supported by the cellular structures, specimens with
dimensions of 30 20 0.3 mm3 were built and the
deflections in z-direction were measured. As is shown
in Figure 2, the support structure designs were used
with unit cell sizes of 6 mm and 2.5 mm and with a
volume fraction of 12% and 15% respectively. The
height of the support was 5 mm and built on the steel
base plate bolted into the machine platform.
As shown in Figure 3, the specimens were fabricated

with the Schoen gyroid and Schwartz diamond type
cellular support structures. When a cell size of 6 mm
was used, the support structure did not connect well to
the edges of the specimen to restrict deformation, as is
shown in Figures 3(a) and (b). The supports connected
to the specimens were able to hold the areas above and
prevent the deformation of the specimens. In addition,
supports act as a heat sink and take the heat away from
newly melted regions, so without support the temperature gradients cause high thermal stresses which lead
deflection to the specimens.
Figures 4(a) and (b) show the obtained deflections
of the two types of support structure in the z-axis.
The measured profiles show higher deflections on the
edges for both support types with a 6 mm cell size.
Support structures with the smaller cell size of
2.5 mm were used to remove the edge deformation.
However, with a 12% volume fraction and the smaller
cell sizes, structures were weakly connected and thus a
610
more heat was taken away and edge deformation was

constrained. Even with the cell size used, there were
no issues with removing the raw powder trapped inside
the support structures.
There are no obvious differences between the two
types of cellular support structures, the gyroid and diamond cells, in terms of the resultant part deflection.
Both can be built in SLM if the cell size and volume
fraction are chosen properly. Concerning the powder
removal, the gyroid cell type is relatively easier due to
the structure formation as it is more open compared to
the diamond type. Comparisons based on other aspects
such as ease of removal of the support from the part
and base-plate are the subjects for future investigation.
The presented results demonstrate that smaller cell
sizes can restrain deflection. However, more contact
points on the downward facing surfaces might result
in a rougher surface. Lowering the volume fraction
can minimise the material used for the support, but
very low values may result in weaker supports and failure. A well designed support structure has to take into
consideration the manufacturability, support removal,
support volume, build time and surface quality of the
supported regions.
Figure 4. Measured deflection profiles in z-axis (a) Gyroid,

cell size = 6 mm, the volume fraction = 12% (b) Diamond,
cell size = 6 mm, volume fraction = 12% (c) Gyroid, cell
size = 2.5 mm, volume fraction = 15% (d) Diamond, cell
size = 2.5 mm, volume fraction = 15%.
volume fraction of 15% was preferred. As in Figures 4

(c) and (d), significant reduction in part deflection
can be observed. As the number of contact points on
the downward facing surfaces of the part is increased,
The majority of previous studies on support structure

concentrated more on non-metallic processes. Since
the support requirements of metallic laser processes
like SLM are more complicated, due to high thermal
gradients involved in the phase transformation process
resulting in thermal stresses and distortion of the part,
proper study on support structure design and manufacturing is required. This work explores the potential
of using cellular structures to support metallic parts
manufactured by SLM. Two types of cellular structures
were tested for their suitability for support structures.
The results reveal that these cellular support structures
can be used for supporting metallic parts. As the cell
size of the structure was made smaller, less deformation of the part occurred. The loose powder trapped
inside these support structures can be removed easily
due to the structures porous nature.
Smaller cell size on the interface of support and
the part should be used to restrain distortion without compromising the down facing surface quality. To
minimise the support material, lower volume fractions
should be used without compromising the ability of the
support to withstand loads.
Future work will investigate the effects of using
different volume fractions on the building time, the
quality of down-facing surface and removal of these
cellular support structures.
ACKNOWLEDGEMENT
This work has been supported by the UK Technology
Strategy Board (TSB) Research Project (BA036D).
611
The TSB funded project is entitled SAVING Sustainable product development via design optimisation
and AdditiVe manufacturING and is a collaboration
between the Simpleware Ltd, Delcam PLC, University of Exeter, 3T RPD, Crucible Industrial Design
Ltd, EOS Electro Optical Systems Ltd and Plunkett
Associates Ltd.
Frank D., Fadel G. 1995. Expert system-based selection of

the preferred direction of build for rapid prototyping
processes. J. Intll. Manufg 6: 339345.
Pham D.T., Demov S.S., Gault R.S. 1999. Part orientation in
SLA. Int J. of adv. Manufg. 15: 677682.
Pham D.T., Demov S.S. 2001. Rapid manufacturing: the technologies and applications of RP and RT. Springer-verlag,
London limited.
REFERENCES
Allen S., Dutta D. 1995. Determination and evaluation of
support structure in layered manufacturing. J. Design
manufacturing 5: 153162.
612

Microchannels fabrication in Direct Metal Laser Sintering (DMLS)

A.R.R. Bineli, A.L. Jardini, A.P.G. Peres, L.F. Bernardes & R. Maciel Filho
Laboratory of Optimization, Design and Advanced Process Control (LOPCA), School of Chemical Engineering,
UNICAMP, Campinas, So Paulo, Brazil
ABSTRACT: In recent years there has been a significant growth in the interest in technologies for additively
manufacturing parts in metal and also in micro manufacture technology, including microreactors. The additive
manufacturing directly in metal favors a rapid development of products and the microreactors improve chemical
processes due to reduced dimensions. Various types of micro-structured reactors have been developed in the last
decade, among them, the most promising concept considers stacked systems of channeled metallic platelets.
This systems are fabricated using traditional tooling, but in this work a DMLS (Direct Metal Laser Sintering) is
proposed to construct a plate with microchannels which are assembled in such way that leads to a microreactor
housing structure. This process achieved a high precision and showed a feasible way for a rapid development
and study of microreactors.
1
INTRODUCTION
In recent years there has been an enormous growth

in the interest in technologies for additively manufacturing parts in metal. This technology was born
out so-called rapid prototyping (RP) technologies in
the 1990s and were driven by the vision of a future
where additive manufacturing could be as widespread
and accepted as subtractive manufacturing methods
are today (Shellabear & Nyrhil, 2004).
Traditional material subtractive manufacturing
technologies such as, milling, tapping and turning create 3D physical models by removing material using
cutting tools. Although the advance in Computerized Numerical Control (CNC) and High Speed (HS)
milling technologies has brought significant improvement, the major disadvantage is the dependence on the
geometric complexity. Some features as small holes
inside a block are hard to manufacture due to the process constraints and when the sample size is small, the
time for process planning can constitute a significant
portion of the time needed to manufacture the part (Yu,
2005).
Differently of traditional subtractive machining
processes, Rapid Prototyping (RP) is a group of
techniques used to quickly fabricate prototypes or
assembly models using 3-D computer aided design
(CAD) data using layer-by-layer fabrication process.
There are several RP machines that utilize different
building methods and materials, such as 3-D printing,
fused deposition modelling, laminated object manufacturing, selective laser sintering (SLS), selective
laser melting (SLM) and 3-D laser cladding (Santos
et al. 2006). Every RP technique has its advantages and
disadvantages. Among the available technology SLS
is an optimal method to directly manufacture metal
parts because of its wider material range and better

flexibility (Yang et al. 2010).
Similar to SLS, Direct metal laser sintering
(DMLS) enables the quick production of complex
shaped three-dimensional (3D) parts directly from
metal powder. This process uses a laser that is directly
exposed to the metal powder in liquid phase sintering
(Khaing et al. 2001) and creates parts by selective fusing and consolidating thin layers of loose powder with
a scanning laser beam process. Additionally, due to
its flexibility in materials, shapes and control of process parameters it is also possible to produce porous
metallic components (Gu & Shen, 2008; Bineli et al.
2011).
Many groups around the world have developed
ideas, prototype methods and commercial systems
using different ways of creating metal parts. In particular, several companies have recently started developing
variations of the Direct Metal Laser Sintering (DMLS)
process (Shellabear & Nyrhil, 2004). This was developed by EOS GmbH of Munich, Germany, and has
been available commercially since 1995 (Khaing et al.
2001).
The production of end-use parts made of metal is
one of the most promising applications among additive manufacturing methods. The fabrication of metal
parts by these methods are especially suitable a small
number of pieces and mass customization. Furthermore it is possible achieve metal parts with excellent
mechanical properties (Santos et al. 2006).
1.1 Microrector technology
Currently microreactor device has attracted attention
in chemical processing due to their superior advantages to improve mass and heat transfer, smaller size,
613
energy and cost saving (Men et al. 2007), and additionally, safety. The benefits of this technology are within
the focus in world research and several microreactors
have been proposed (Kolb et al. 2007).
These devices offer high heat and mass transfer
coefficients due to high surface to volume ratio, lower
pressure drop compared to classical packed bed reactors, good structural and thermal stability and more
precise control of the process conditions leading to
higher product yields. They are generally also referred
to as microstructured reactors, microchannel reactors
or microreactors (Cai et al. 2010). The main characteristic of these systems are the dimensions of open paths
in the submillimeter range and mostly of microreactors
have multiple parallel channels ranging from 10 to
several hundred micrometers (Kiwi-Minsker &
Renken, 2005).
Generally, this devices are generated by numerous
techniques such as micro milling, electro discharge
machining (EDM), wet chemical etching, punching,
embossing, laser micro machining (ablation) and sintering (Hessel et al, 2005) and recently has also been
proposed by additive manufacturing methods (Jardini
et al. 2008; Bineli et al. 2011).
Microreactor technology creates opportunities for
the development of miniature chemical devices, in
which several unit operations are integrated with capabilities considerably exceeding those of conventional
macroscopic systems (Hessel et al. 2000; Jensen 2001;
Pierik et al. 2004). Furthermore, excellent controllability, low safety risks and low environmental impact
are achieved (Holzwarth et al. 2000; Delsman et al.
2004).
Although the microreactor technology has been
applied in numerous works, a rapid production of these
components by means of additive manufacturing can
facilitate the development, tests, and evaluation of
operational conditions of chemical process allowing
to minimize number of experiments and costs, which
lead in a optimized micro device, ready to large-scale
production and reduced time to market.
Thus, the objective of this paper is to propose the
construction of microchannels plate by additive manufacturing and assembles it with microreactor housing
structure.
Figure 1. Schematic diagram of the DMLS system.
Figure 2. Exposure strategies and process parameters.
Then, the head scan move the laser beam through twodimensional cross section and is precisely switched
on and off during exposure of designated areas. The
absorption of energy by metal powder will generate the
cure and sinter of the already solidified areas below.
This process proceeds layer-by-layer until all parts in
a job are completed (EOS, 2009).
MANUFACTURING METHODOLOGY
2.2 Expousure strategies
2.1
DMLS Principle
A schematic diagram for construction of any part

by DMLS machine is presented in Figure 1. First,
the building and dispenser platform are lowered by
one layer thickness for that the recoater blade can
move without collision. When the recoater stands in the
right position the dispenser platform rise to supply the
amount of powder for the next layer. Then, the recoater
moves from the right to the left position; in this way the
metal powder is spread from dispenser to the building
area and the excess metal powder falls into collector.
Figure 2 shows the laser scanning the top surface of a

thin powder layer to form the area enclosed by crosssections of the sliced object. Initially, all contour of
the layer structure is exposed with a laser power and
contour speed (Csp ) and after by hatching (Hsp ). As
the diameter of the sintered zone is usually larger
than the laser diameter, the effective laser diameter
(Senthilkumaran et al, 2009) it is necessary to compensate the dimensional error by use of Beam Offset (BO)
parameter for a good correspondence to the original
dimensions of the part.
614
Table 2. DMLS process parameters for the components

construction.
Value
Parameter
Laser power (W)
Scan speed (mm/s)
Hatching spacing (mm)
Layer thickness (m)
Beam offset (m)
Laser beam diameter (m)
Effective laser beam (m)
Scanning pattern
Figure 3. Microreactor assembly.

Table 1.
Hatching
Contouring
117
225
0.18
58,5
400
Not applied
30
20
40
200
Stripes (alternating 45 )
Dimensions of microchannels plate.

Value
Parameter
Parallel Channels
Wall thickness (mm)
Space between channels (mm)
Height (mm)
Length (mm)
Plate width (mm)
Case 1
Case 2
30
0.5
0.5
0.25
0.75
0.25
50
30.5
30.25
In the hatching step, the laser beam moves line after

line several times to assure that the sintering process
can unroll completely, because it maintains the temperature for a long period. The distance between the lines
is called hatch spacing (Hs) and is set about one quarter of the laser beam. Here the Beam Offset value is
again defined with respect to the edge of the boundary,
and if this value is higher or less than the correct value,
the particles of the irradiated region may be either not
sintered or over-sintered.
The layer thickness parameter ensures adhesion
between the layers and if this value is too high, there
will not be a optimal adhesion between single layers
because the curing depth is not high enough; furthermore, mechanical tension can be generated through
this layer which can lead to detachment of the layer
below. If the selected value is smaller, a tearing-off of
a structure can happen during the recoating process,
since the sintered particles get struck between it and
recoater blade (EOS, 2009).
2.3
Microreactor design and DMLS parameters
The proposed microreactor assembly are presented

in Figure 3, which is composed by housing structure, heaters compartment and microchannels plates.
Two microchannels plate using different dimensions
were constructed and the values are specified in
Table 1.
In this work, DMLS machine (EOSINT M270)
was used for microchannels fabrication. The EOS
Aluminium powder (AlSi10Mg) with particles ranging from 25 m to 45 m was used. The aluminium
Figure 4. Microchannels surface morphology. (a) case 1;

(b) case 2.
alloy have good thermal conductivity properties, light

weight and high strength.
The maximum power available withYtterbium laser
in the DMLS machine is 200 W and for the mentioned
material just 58.5% (117 W) and 29.25% (58.5 W) of
the maximum power was used. Scan speed from 225 to
400 mm/s was chosen for a shorter construction time.
The laser beam diameter and energy density used are
40 m and 2.328 1010 W/m2 , respectively. The process parameters used in the present study are given in
Table 2.
615
Microchannels plate obtained by DMLS technology

are shown in Figure 4. As can be seen, the process
allows an easily modification of wall thickness and the
spacing between microchannels. Theses modifications
are produced by a CAD software and quickly produced
by machine. Moreover, the adjustment of parameters
hatch spacing, hatch and contour speed and laser power
determines the amount of energy that the laser beam
deposits and fills the two-dimensional layers resulting
in a dense or porous structure.
In a previous work Bineli et al (2011) presented
a porous surface structure, since those metal powder
particles were not completely sintered. Nevertheless,
a simple application of a sand blast treatment in the
parts was able to remove the porous layer, resulting in
a smoother surface.
The machine is enabled to construct microchannels
plate accurately, however the challenge to manufacture
such parts comes when it is very thin and the thermal
stress accumulation leads to a warpage of the parts.
In order to minimize this effect the plates were produced assuming a angle between the base and the
microchannels plate.
The manufacture of the plates was done perfectly and both were assembled with the proposed
microreactor, as shown in Figure 5. According to
the design, two microchannel plates with dimensions
of the Case 1 were chosen for placement inside the
Figure 5. Final microreactor assembly containing microchannels plates.
compartment, which was sealed by a graphite gasket.

From Figure 5 it is also can be seen a vaporization
unit on the reactor and the heating cartridges that are
responsible to vaporize a mixture and controlling the
reactor temperature.
4
CONCLUSIONS
In this paper, the construction of microchannels plate

was proposed by using DMLS technology. This system
achieved a high precision and the sandblast treatment
was able to remove the porous layer.
This process showed a feasible way for a rapid
development and study of microreactors.
ACKNOWLEDGMENT
We acknowledge with gratitude the Brazilian financial
support of CNPq and FAPESP.
REFERENCES
Bineli, A.R.R, Peres, A.P.G, Bernardes, L.F, Jardini, A.L,
Maciel Filho, R., 2010, Design of microreactor by integration of reverse engineering and direct metal laser sintering process. Proceedings of 5th international workshop
on hydrogen and fuel cells. Campinas.
Bineli, A. R.R, Barbosa, M.I.R, Peres A. P., Jardini, A.L,
Maciel Filho, R., 2011, Surface morphology of a catalytic wall microreactor constructed by direct metal laser
sintering process, Chemical Engineering Transactions,
24, 223228
Cai, W., Wang, F., van Veen, A., Descorme, C., Schuurman,Y.,
Shen, W., and Mirodatos, C., 2010, Hydrogen production
from ethanol steam reforming in a micro-channel reactor,
International Journal of Hydrogen Energy, 35, 11521159
EOS, 2009, EOSint M 270 User Manual.
Gu, D., Shen, Y., 2008. Processing conditions and
microstructural features of porous 316L stainless steel
components by DMLS. Applied Surface Science, 255,
18801887.
Hessel, V., Lwe, H., Mller, A., Kolb, G., 2005, Chemical Micro Process Engineering: Processing and Plants,
Wiley-VCH, Weinheim.
Jardini, A.L., Costa, M.C.B., Bineli, A.R.R., Romo, A.F.,
Filho, R.M., 2008. Operability analysis and conception
of microreactor by integration of Reverse Engineering
and Rapid Manufacturing. Computer Aided Chemical
Engineering, 853858.
Khaing, M.W., Fuh, J.Y.H., Lu, L., 2001. Direct metal laser
sintering for rapid tooling: processing and characterization of EOS parts. Journal of Materials Processing
Kiwi-Minsker, L., Renken, A., 2005, Microstructured reactors for catalytic reactions. Catalysis Today, 110, 214.
Kolb, G., Schrer, J., Tiemann, D., Wichert, M., Zapf, R.,
Hessel, V., and Lwe, H., 2007, Fuel processing in integrated micro-structured heat-exchanger reactors, Journal
of Power Sources, 171, 198204.
Men, Y., Kolb, G., Zapf, R., Hessel, V., and Lowe, H.,
2007, Ethanol steam reforming in a microchannel reactor, Process Safety and Environmental Protection, 85,
413418.
Santos, E.C., Shiomi, M., Osakada, K., Laoui, T., 2006.
Rapid manufacturing of metal components by laser
616
forming. International Journal of Machine Tools and

Manufacture, 46, 14591468.
Senthilkumaran, K., Pandey, P.M., Rao, P.V.M., 2009. Influence of building strategies on the accuracy of parts
in selective laser sintering. Materials & Design, 30,
29462954.
Shellabear, M., Nyrhil, O., 2004. DMLS-Development
History and State of the Art, Laser Assisted Net Shape
Engineering (LANE), Erlangen, Germany.
Yang, J., Ouyang, H., Wang, Y., 2010. Direct metal

laser fabrication: Machine development and experimental
work. International Journal of Advanced Manufacturing
Yu, N., 2005. Process parameter optimization for direct metal
laser sintering (DMLS), Ph.D. thesis, National University
of Singapore, Singapore.
617

Knowledge based process planning and design for

Additive Manufacturing (KARMA)
B. Singh & N. Sewell
University of Exeter, Exeter
ABSTRACT: Additive Manufacturing (AM) has the potential to provide great advantages over traditional
subtractive manufacturing. One of the most important benefits is in cutting costs related to building parts due to
a significant reduction in material waste. In addition, certain other manufacturing constraints are removed such
as those related to the way in which the part is actually fabricated using a traditional process. This gives designers
the opportunity to create their products in ways which were previously considered impossible to manufacture, for
example, by defining the internal geometries of a component. However, this flexibility brings a new challenge;
there is great skill required to make effective use of AM technology, and given the wide range of systems and
processes available, expert knowledge is often in short supply. Without this knowledge, attempts to use AM often
result in disappointment for the end user as the products may fail to deliver what is expected in terms of form,
fit or function. This paper describes the creation of an easy-to-use tool which enables a wide range of users to
access and assess the strengths and weaknesses of AM process for manufacturing products.
1
INTRODUCTION
The Additive Manufacturing industry is growing

rapidly. There has been a remarkable increase in
both general interest and innovations in the AM field
since its inception over twenty years ago. AM processes are now increasingly used by the manufactures
especially for custom and short-run part production
(Wohlers 2010). There is a wide range of applications for AM techniques in the industry ranging
from aerospace, automobile and military applications
to highly specialised artistic artefacts and medical
implants.
The main reason for the rapid growth of the AM
field is the advantages that it has to offer when compared to traditional manufacturing processes. Its most
important feature is to be able to produce complex
freeform shapes that were earlier considered impossible to manufacture using the traditional manufacturing
techniques. In addition, there are significant reductions in material waste, tooling costs and time to
market. However, there are certain inherent limitations
associated to the AM technologies. Most importantly
the anisotropic properties (mechanical and thermal)
of the resulting parts that are exhibited by many AM
systems. Unlike conventionally manufactured parts,
AM parts do not behave identically in all directions
and are often relatively stronger in the manufactured
layer direction (XY plane) than in-between the layers (Z-direction) under tensile load. In addition, there
are limitations on the sizes of the parts that can be
produced, the types of materials that can be processed,
repeatability and accuracies that can be achieved using
the present AM processes.
The long term success of the AM field is still down

to how well the advantages and limitations of the technology are conveyed to the potential users (Laar 2008,
Wohlers 2010, Munguia et al. 2011). A number of AM
technologies are now available, however, the expert
knowledge available to make effective use of these is
still in short supply.
BACKGROUND
Traditionally, the successful application of an AM

technology for producing parts has been dependent
upon the knowledge and expertise of a technician.
Figure 1 shows a typical AM process. The process
starts with the design requirements which are normally
specified by a customer. These requirements dictate
what is expected from the finished part in terms of
geometry, surface roughness, tolerances and required
mechanical and thermal properties. The customer also
provides a 3D CAD model of the part, usually in the
STL file format, which is so far an industrial de-facto
standard for all the AM technologies.
The technician, who is experienced in using different AM processes, analyses the customer requirements
and suggests the most appropriate AM process and
a material that has potential to meet these requirements. The technology parameters are selected next.
These parameters relate to the machine settings such
as layer thickness, laser power, scan speed, powderbed temperature etc and are vital for a successful build
of the part. The final manual step carried out by the
technician is selecting the most appropriate orientation of the part in the build direction. The orientation
619
Figure 2. KARMA AM process Courtesy: KARMA project

consortium.
Figure 1. Conventional AM process Courtesy: KARMA

project consortium.
selection is a critical step in AM process planning as

it not only affects the resulting part properties and the
build quality but also dictates the cost of build. The
building cost is largely associated with the building
time and volume of the support structure used, both
of which are directly affected by the orientation of the
part in the build chamber.
After the completion of the manual process planning stages, the CAD model is fed to the AM system
that slices the model into layers according to the
layer thickness specified by the technician. The actual
manufacturing of the part is then commenced using
the selected AM technology. Once completed a post
machining process may be carried out in order to
remove the supporting structure that was used to support the part geometry. Finally, the manufactured part
is checked against the initial customer requirements. If
the requirements are not met, the process is repeated
by redefining some of the process parameters such
as part orientation, material and even in certain cases
different AM technology until the final part fulfils the
customer requirements.
In the conventional AM process, there is no definite way to predict if the customer requirements will
be fulfilled at the end as the initial AM process planning stages are still carried out solely based upon the
expertise of a technician. This experience takes time
to develop and is often the result of learning from
many successful and failed builds. However, there has
been some attempt in the past to help the users of
AM technology through assisted decision making process. These research efforts can be broadly classified
into two categories; the AM (/RP) selection systems
(Bibb et al. 1999, Massod & Al-Alawi 2002, Byun &
Lee, 2005, Munguia et al. 2011) and AM process planning systems (Hur and Lee 1998, Kulkarni et al. 2000,
Pande and Kumar 2008). The aim of the AM selection
systems is to advice users of the most relevant technology and the materials that can satisfy user needs.
The AM process planning systems help in performing
tasks related to a certain AM process such as determining optimal orientation, support structure estimation,
slicing and path planning. However, these two system types are complementary and should be provided
as a unified system for prediction of certain factors
like cost, manufacturability of critical features, surface
quality and build time estimation. These factors not
only depend on the AM technology and material combination but also on the part orientation and machine
parameters.
This paper describes the development of the
KARMA tool, a knowledge base driven system to
address some of the issues outlined. This tool enables
a wide range of users to specify their requirements and
helps in not only finding relevant AM technology and
material combination but also in AM process planning
in a way that is most likely to fulfil user requirements.
3
KARMA PROJECT
The aim of KARMA project is to develop a system to

assist end-users of AM technologies during the initial
technology and material selection and process planning stages. The KARMA system process is shown in
the figure 2. The process starts with a user specifying
the design requirements. The system then assists the
user in the selection of the most appropriate technology, material, machine parameters and part orientation
that is likely to fulfil the design requirements. Once the
user makes a selection, the part model in the proposed
orientation is sliced and the actual process of building the part starts. There is no loop in this process as
the user is informed beforehand regarding the overall
manufacturability and the resulting part properties.
3.1 KARMA system architecture
There are three main components to the KARMA system including Knowledge based engineering (KBE)
620
3.1.3 KBE tool

The KBE tool provides a user friendly, web-based
interface for end-users to access the information stored
in the KBE database. It also utilises various algorithms
at different stages of the process to analyse critical
issues such as surface finish, build time, costs and
material usage. A wizard type interface is used to capture user needs and to identify the most relevant AM
technology, materials and build scenarios that fulfil
user requirements. Various types of charts and graphs
are employed to display the results to the users.
Figure 3. KARMA System architecture.
3.2 Using KARMA system
database, Algorithms and a KBE tool (figure 3).

Additionally, the resulting (reoriented) part from the
KARMA system along with the material properties
can be exported to an external FEA package for the
testing of mechanical and thermal properties. These
components are briefly discussed in the following
sections.
3.1.1 KBE Database
The KARMA system utilises an underlying KBE
database to store all the relevant information regarding
various AM technologies. Currently in the prototype
system, there are four AM technologies included in the
database. These technologies include two laser based
technologies, one for metal powder (laser CUSING)
and one for polymeric material (Stereolithography).
The database is designed in way that it can accommodate more technologies as these become available to
the users. The database contains information regarding
previously tested build scenarios (various combinations of machines, materials and machine parameter)
and the resulting part properties (mechanical and thermal). There is also a provision for the users to create
and test their own build scenarios. The KBE database
serves the following three purposes:
a design guide for the end-users

a processing materials guide
machine parameters study
3.1.2 Algorithms
KARMA system incorporates five main algorithms.
These algorithms include:
algorithm for critical features and overall manufacturability estimation

algorithm for surface quality prediction
algorithm for support structure prediction
algorithm for build time estimation
algorithm for costs estimation
The design guide contained in the KBE database

specifies the limits and ranges of the given AM technologies and these values are used as input to these
algorithms.
Figure 4 shows the various stages in the KARMA system.The user starts by uploading a CAD part (step 1) in
the form of an STL file. The first stage in this system is
to check the file for any inconsistencies (step 2). After
validating the STL file some pre-processing (step 4) is
done on the file to ascertain if the given part size can be
manufactured on the available AM technologies. If the
part size is too big to be manufactured the user is notified to scale/resize the part (step 5). In the next step
(step 6), the user is asked to specify his preferences
regarding the AM technologies and the application of
the part to be manufactured. There are three possible
routes (scenarios) that a user can follow namely Application driven, Process driven and No preference. These
scenarios are discussed in the following sections.
3.2.1 Application driven scenario
The user selects this scenario (step 6-scenario A) when
he knows about the application of the part to be manufactured and is interested in knowing and comparing
the capabilities of various available AM technologies, materials and build scenarios. The application
of the final part can be to serve Form (concept models), Fit (assembly models) and Function (functional
prototypes) or can be used as a production part. For
example a user may be interested in the Form category when the appearance (feel, colour, aesthetics
and surface finish) of the part is most important. The
user at this stage selects the most appropriate category and is asked to select the most important part
properties/parameters that are relevant to his needs
under that category (e.g. scale/size, roughness, tactility, colour and transparency for Form category). The
user is then asked to rank the desired part properties
in an order of importance. The KARMA system, then
queries the underlying KBE database and presents the
user with the range of achievable values to some of
these properties with the available AM technologies.
Finally, the user specifies the most appropriate values
to the desired properties from the values shown.
3.2.2 Process driven scenario
Process Driven Scenario (step 6-scenario B) is selected
by a user who already has an AM process or a material
in mind and wants to know about the capabilities and
constraints of that process or the material. Here the
user can either select an AM process and will then be
621
Figure 4. Stages in the KARMA system.
622
presented with the list of possible materials that can be

processed or must specify a material and the possible
AM processes that can process that material are listed.
of appropriate AM technology, material and machine

parameters.
3.2.3 No preference scenario

This scenario (step 6-scenario C) is suitable for a user
who has no preconceived idea about the application
of the manufactured parts, available AM processes or
materials. The user is just interested in AM. It is envisaged that this option may be chosen most often by first
time or novice AM users.
Once the user has specified his requirements, the
KARMA system performs a multi-attribute search and
queries the underlying database to see which technologies can fulfil the user requirements. The initial results
are presented to the user in the form of Build Scenarios (step 7) (In the case of No Preference scenario,
all the Build Scenarios are presented). Each Build Scenario presented to the user contains an AM technology,
machine parameters and a material that are likely to
produce the required part properties. Additionally, the
user is given an estimate of the building time, resulting surface roughness and the building cost associated
with that particular building scenario. These scenarios are presented with a decreasing order of relevance
to the user preferences. User here can compare different building scenarios and select (step 8) the most
appropriate one to his needs.
In the next stage, the KARMA systems calculates
the optimal orientation according to the user criteria.
Initially, the user is prompted to select the most important feature (or surface) (step 9) on the part that should
have the best surface finish. The user then specifies the
most important factor among build speed, cost and surface finish for orientation calculations. The KARMA
system then utilises the underlying algorithms to calculates n-build directions, each being most favourable
regarding certain manufacturability aspects (cost, surface roughness or build time).The results are presented
to the user in a matrix form and the system highlights
the orientation that fits best according to the most
important factor indicated by the user. The user can
also compare and select other orientations suggested
by the system.
In step 11, the user has an option to export the
part geometry along with the material properties to
an external FEA package to analyse the mechanical/
thermal properties of the resulting part in the proposed
orientation. The results of FEA analysis are presented
to the user in step 12. If the user is not satisfied with the
results, there is an option to increase the resolution of
search space for the orientation calculations and steps
10, 11 and 11a are repeated until a satisfactory solution
is found.
In step 13, the system checks if there are any critical features such as long thin features present in
the part that cannot be manufactured using the given
technology. If so the user is informed to modify the
part accordingly. Finally, the user is presented with
an STL file of the original part, re-oriented in the
optimal build direction along with the information
DISCUSSION
KARMA system is being developed to cater for the

needs of different types of users ranging from novice to
experienced users. There are different routes that users
can follow. For example, experienced users paths may
be short and simple, however, for an inexperienced
user who may be seeking the best possible option, the
system provides a more in-depth guided route. The
KARMA system also takes a more holistic approach
by providing support not only for choosing the relevant
AM process and material combination but also for the
process planning tasks so that the design requirements
can be met in a best possible way. Some of the potential
benefits of using the KARMA system include:
Support for AM design and process planning

Increase efficiency of the AM process
Optimisation of AM part, reduced time to market
and costs
Insurance of all part requirements
This paper describes the development of the KARMA

system to assist the users of Additive Manufacturing
technology for design and process planning. The system incorporates a KBE database to store all relevant
information regarding various AM technologies and
materials which provides a technology and processing
materials guide for the users. In addition, the database
also contains the information regarding the previously
tested build scenarios and properties of the resulting parts. This information is then utilised by various
algorithms for the AM process planning tasks. The
KARMA system is useful for different users with any
level of experience with AM technology and adapts to
their needs.
The KARMA system is currently under development and once completed, will be a web-based
system capable of running on any platform (computer
operating system). The KBE database is both flexible and expandable. In future, further technologies
will be added to the system making it a comprehensive tool for both AM selection and process planning
tasks.
ACKNOWLEDGEMENTS
The research leading to these results has received
funding from the European Communitys Seventh
Framework Programme managed by REA-Research
Executive Agency http://ec.europa.eu/research/rea
([FP7/20072013] [FP7/20072011]) under grant
agreement n [2436312].
623
REFERENCES
Bibb, R.Taha, Z., Brown, R. & Wright, D. 1999. Development
of a rapid prototyping design advice system. Journal of
Intelligent Manufacturing 10: 331339.
Byun, H. S. & Lee, K. H. 2005. A decision support system
for the selection of a rapid prototyping process using the
modified TOPSIS method. Int J Adv Manuf Technol 26:
13381347.
Hur, J. and Lee, K. 1998. The development of a CAD environment to determine the preferred build-up direction for
layered manufacturing. Int Jour of Adv Manuf Technol
14(4): 247254.
Kulkarni, P., Marsan, A. & Dutta, D. 2000. A review of
process planning techniques in layered manufacturing.
Laar, M. 2008. Realizing the Business Potential of Rapid
Manufacturing. Additive Layered Manufacturing: From
Evolution to Revolution: 3339.
Masood, S. H. & Al-alawi, M. 2002. The IRIS

rapid protoyping system selector for eduational and
manufacturing users. Int. J. Engng Ed 18(1):
6677.
Munguia J., Bernard, A. & Bernard, A. 2011. Proposal and Evaluation of a KBE-RM Selection System. Rapid Prototyping Journal 17(4): Retrieved from
http://www.emeraldinsight.com/journals.htm?issn=1355
2546&volume=17&issue=4.
Pande, S. S. & Kumar, S. 2008. A generative process planning system for parts produced by rapid prototyping.
International Journal of Production Research 46(22):
64316460.
Wohlers, T. 2010. Wohlers Report 2010-Additive Manufacturing State of the Industry. Annual Worldwide Progress
Report, Wohlers Associates Inc., Fort Collins, CO, USA
2010.
624

A comparison of laser additive manufacturing using gas and

plasma-atomized Ti-6Al-4V powders
M. Naveed Ahsan & Andrew J. Pinkerton
Laser Processing Research Centre, School of Mechanical, Aerospace and Civil Engineering,
The University of Manchester, Manchester, United Kingdom
Laiq Ali
Engineering Support Division, DESTO, Chaklala Cantt Rawalpindi, Pakistan
ABSTRACT: This research presents a comparative study of the characteristics of laser additive manufacturing
(LAM) using two types of Ti-6Al-4V powder. Ti-6Al-4V powders prepared using the gas-atomization (GA) and
the plasma rotating electrode (PREP) processes were first analyzed using laser diffraction, scanning electron
microscopy and microcomputed tomography. A 1.5 kW diode laser with a coaxial deposition head was then used
to deposit a number of thin-wall structures at a range of processing parameters from each of the powders. The
deposited structures were characterized using optical microscopy, scanning electron microscopy, x-ray diffraction
and microcomputed tomography (MicroCT). In both cases, deposits of Ti-6Al-4V exhibit a unique epitaxial prior
beta grains microstructure that transforms to alpha lathes and retained beta during cooling. X-ray diffraction
results show that the overall microstructure is + . The lamellar + phase spacing (S+ ) increases with
laser power but seems unaffected by variation in the powder mass flow rate. Micro hardness of the laser deposited
Ti-6Al-4V is dependent on the lamellar + phase spacing (S+ ). The results show some potential benefits
of using PREP powder in laser additive manufacturing. PREP powder has a higher deposition rate and deposits
show lower intralayer porosity and lower surface roughness. However, PREP powder deposits show lower micro
hardness than GA powder deposits.
INTRODUCTION
Laser additive manufacturing (LAM) is a well-known

technique in the advanced manufacturing industry.
Typical uses of LAM are rapid manufacturing, rapid
repair and surface coating applications. LAM is a
layer-by-layer additive manufacturing technique that
involves gas-assisted metal powder transfer to a melt
pool created by a laser on the metallic substrate in
order to build a deposition layer. The addition of
multiple layers can produce 3-dimentional near net
shape parts. Gas-atomized (GA) powders are mostly
used for the LAM process; gas-atomization is very
common process for metal powder production, but
has the inherent weakness of entrapped gases during
rapid solidification (Rabin et al. 1990). The use of
the plasma rotating electrode process (PREP) powder is increasing; the PREP process is an advanced
powder production technique, able to produce particles with near perfect spherical morphology (Alagheband & Brown 1998). The GA and PREP powder types
dominate; other powder production methods exist
(e.g. the Hydride-DeHydride process) but these currently do not have significant market share.
LAM has been successfully employed for aerospace
(Pinkerton et al. 2006) and biomedical applications
(Ahsan et al. 2011, Dinda et al. 2008) using Ti-6Al-4V,
an aerospace grade Ti alloy having high strength to

weight ratio, extraordinary corrosion resistance and
relatively low density. A deposited clad/coating can be
equally good, and sometimes even better, in mechanical properties than the parent material depending on
microstructure. However, intralayer porosity defects
are sometimes found in laser additive manufactured
materials. These are highly undesirable because they
severely degrade mechanical strength and fatigue
resistance. The availability of alternative powder types
and demand for high quality LAM parts manufacture
and repair necessitates a comparative study using different types of Ti-6Al-4V powder to see if either gives
advantages in terms of geometric accuracy, surface
finish, microstructural and mechanical properties or
prevalence of defects such as intralayer porosity.
Comparative studies of LAM using GA and wateratomized (WA) metallic powder of 316L stainless steel
and H13 tool steel have been conducted previously
by Pinkerton & Li (2003a, 2005a) to characterize
microstructure, surface finish and deposition rate.
There have been previous investigations on Inconel
718 (Ni alloy) GA and PREP powder deposition (Qi
et al. 2009, Zhao et al. 2008). The studies showed that
PREP powder deposition structures exhibited minimal porosity (Qi et al. 2009, Zhao et al. 2008) and
superior mechanical properties to those of the GA
625
powder structures (Zhao et al. 2008), but this research

did not provide any comparative information on the
microstructures of the deposited samples. Wang et al.
(2007) studied the microstructure and mechanical
properties of samples produced by hot isostatic press
(HIP) forming of Ti-6Al-4V using GA and PREP powders at one set of parameters. The microstructures were
similar, consisting of platelet and fine transformed
structure, with fine equiaxed distributed uniformly,
but the strength of the GA samples were higher than
that of the PREP samples. In summary, there have been
few published studies in this area, particularly regarding laser additive manufacturing using GA and PREP
Ti-6Al-4V powders.
In this work, a systematic comparative investigation of LAM using GA and PREP Ti-6Al-4V powders
has been carried out. LAM characteristics in terms
of layer geometry, surface finish, microstructure and
micro hardness and internal porosity are compared
under similar process conditions. An advanced porosity assessment method using microcomputed tomography is used to quantify internal porosity in the powders
and final LAM samples because it is a volumetric
method capable of performing analysis on a designated volume, and is hence a more reliable method
than microscopy which considers only discrete planes.
The work shows a new way to quantitatively assess the
differences between the two types of powder that are
dominant in the production of high quality titanium
alloy components by LAM.
Figure 1. Schematic diagram of laser additive manufacturing (LAM) system.

Table 1. Processing parameters used in laser additive manufacturing experiment
Parameters
Laser power (W)
Mass flow rate (gs1 )
Level 1
Level 2
Level 3
Level 4
Level 5
800
1000
0.033
0.066
0.089
0.112
0.135
2.2 Laser additive manufacturing (LAM)
50 mm 50 mm 10 mm were used as substrates.

The blocks were grit blasted to enhance their absorptivity and degreased with ethanol. Two levels of laser
power and five levels of mass flow rate for each powder type were tested at a constant scanning speed of
5 mms-1, which gave a total of twenty experimental
runs as shown in table 1.
A 1.5 kW diode laser, producing light at dual wavelengths of 808 nm and 940 nm, coupled with a coaxial
deposition head was used for the GA and PREP powder
deposition. The laser was focused to a spot diameter
of 1.7 mm on the substrate with a standoff distance of
7.5 mm between the deposition nozzle and deposition
point.
The movements of the substrate in the x-y plane
were controlled using an Isel C116-4 CNC motion
control table and each track was deposited with the
substrate moving in same direction. A SIMATIC OP3
dual hopper disc powder feeder with Argon conveyance gas with a flow rate of 5 litres per minute
was used for powder mass delivery. Another flow of
Argon gas, with a rate of 6 litres per minute, through
the central passage of the coaxial deposition head was
used to shield the objective lens and to minimize oxidation. The combined, targeted flow of 11 litres per
minute of the Argon was sufficient to provide a local
inert atmosphere to minimize oxidation in the deposition process. For each laser power and powder type,
thin-wall samples were produced with 15, 12, 10, 8
and 8 deposition layers at 0.033, 0.066, 0.089, 0.112,
and 0.135 gs1 mass flow rate respectively.
A schematic diagram of the laser additive manufacturing (LAM) system used in the experiment
is shown in Figure 1. For laser additive manufacturing, Ti-6Al-4V blocks with a nominal size of
All deposition samples were characterized using optical microscopy, scanning electron microscopy and
2.1 Powder characterization

GA and PREP Ti-6Al-4V powders with a particle
range of 45105 m were used in the experiment.
Microcomputed tomography of the powder samples
was carried out using an Xradia microXCT-400 system
that operates in point projection geometry. It features
an x-ray source having a 7 m spot size with a tungsten target, and operates with an accelerating voltage
in the range 40150 kV. Scans of the GA and PREP
powder samples were carried out using the 10 objective with 1001 images being taken over 184 degrees
for source voltages of 80 and 65 kV respectively. The
final reconstructed voxel (3D pixel) dimensions were
2.3450 m and 2.3120 m for the GA and PREP
samples respectively.
A Malvern Mastersizer laser diffractometer was
used to perform particle size analysis of the powders
samples. A Hitachi S-3400N scanning electron microscope was used to examine the surface morphologies
of the powder samples and to reveal any irregularities
and/or pores.
2.3 LAM samples characterization
626
Surface roughness on the top of the thin-wall samples was measured using a white light interferometer
(Wyko NT1100 Optical Profiling System). In the
focused position, 1000 m 500 m of the clad top
surface area was scanned by the equipment to find
surface variations. Micro hardness measurement was
also employed on deposition samples using a Buehler
Hardness Testing Machine. Micro hardness was measured from top to bottom of thin-wall sample at
discrete locations 2 mm apart from each other along
a straight line. An average value of all the discrete
readings was taken as representative value.
For MicroCT sample preparation, the thin-wall
samples were cut in two traverse planes in the middle of the thin-wall in order to make their widths
and breadths approximately equal. Micro computed
tomography of the deposition samples was carried out
using a Metris XTH 225 CT machine. It features a
225 kV high energy micro focus x-ray source with a
3 m spot size and has a 5-axis sample manipulator
stage as shown in figure 2. Scans of each sample were
performed with 2880 projections over 360 degrees of
rotation. The final reconstructed resolution was in the
range of 5.5 m to 10.8 m depending upon the field
of view in each sample. A quantitative porosity analysis was performed on the tomography data from all the
deposition samples using image segmentation tools.
Figure 2. Microcomputed tomography setup; (a) Metris
XTH 225 CT machine; (b) MicroCT schematic diagram.
microcomputed tomography, white light interferometry and micro hardness testing. For microscopy, all
deposition samples were initially sectioned in two traverse planes near the middle of the wall about 10 mm
apart and then hot mounted in conductive mounting
resin (Struers ConduFast) using a Struers Labopress-3.
The cross-sectioned planes were ground using a Presi
Mecatec-334 polishing machine and then polished
with diamond suspensions of 6 m, 3 m and 1 m
diamond grain sizes using the same machine. The samples were then loaded on an oxide suspension polishing
(OSP) machine for two hours to produce a scratchfree and deformation-free surface. Krolls Reagent
was used to etch the fine polished samples to reveal
grain boundaries. An Olympus optical microscope was
used to characterize the microstructure of the deposition samples. Scanning electron microscopy was performed on selected samples using a Hitachi S-3400N
SEM. Average microstructure grain sizes were measured using an intercept method. Because of the relatively large grains, for each sample an average value
of seven consecutive grains along a straight line, with
random sampling, was taken as a representative value.
X-ray diffraction was employed at the top and
bottom positions of a thin-wall sample. Samples for
x-ray diffraction were prepared by cutting the thin-wall
parallel to the deposition direction, 2 mm and 6 mm
above the substrate level for the bottom and top sample respectively. The samples were ground and then
polished using oxide polishing suspension for four
hours to remove any deformation introduced during
the cutting operation.
3.1 Powder analysis

The laser diffractometer results show that the PREP
powder has, on average, smaller particles than the
GA powder. The mean particle diameter was found
to be 94 m for GA powder and 72 m for PREP
powder. Figure 3 shows the surface morphologies of
GA and PREP powders obtained by scanning electron
microscopy. The GA powder has an approximately
spherical morphology and generally rough surface,
with satellite particles and some surface pores also
visible, as shown in figure 3(a). The PREP powder
exhibits a highly spherical shape and fine surface with
no surface porosity, as shown in figure 3(b).
Comprehensive image processing analysis was carried out on the tomography data to reveal the structure
inside the powder particles. Figure 4 shows the powder
particles images of the particular tomography slices in
the whole powder volume under analysis. It is found
that the PREP powder sample has some internal porosity. The quantitative porosity results of these samples
are listed in table 2. The volumetric porosity in the
PREP powder is found to be three times less than the
GA powder.
Micro computed tomography analysis of both
powders shows that PREP powder has internal porosity; despite the fact that PREP powder sample does
not show any surface porosity in scanning electron
microscopy characterization. The PREP technique can
produce high purity spherical powder particle, but still
with this technique it is not possible to get rid of all
627
Figure 4. MicroCT images of the selected slices; (a) GA;

(b) PREP.
Figure 3. SEM images of the powder showing surface
morphologies; (a) GA powder; (b) PREP powder.
Table 2. Quantitative porosity results obtained by microcomputed tomography analysis
pores inside the powder particles. The larger mean particle size of GA powder than PREP powder leads to
the particles interacting differently with the laser beam
above the melt pool. Firstly, it is established that laser
beam attenuation decreases with increase in powder
particle size (Pinkerton 2007) and this will result in
more laser energy available at substrate level in the case
of GA powder deposition. Secondly, smaller particles
heat up more quickly in a laser beam and are more
likely to become fully molten in the powder stream
(Pinkerton & Li 2003b). Thus, there are more likely to
be molten or near-molten particles in the stream in the
case of PREP powder deposition.
3.2
LAM samples analysis
3.2.1 Geometric dimensions

A comparison of mean layer thickness of the GA
and PREP deposition thin-wall samples at two laser
power levels is presented in figure 5. As expected,
the layer thickness increases with mass flow for both
types of powders. PREP powder deposition shows a
little higher layer thickness at each mass flow rate and
laser power level than that of GA powder deposition.
The maximum difference of 0.22 mm in average layer
thickness is observed between GA and PREP powder
depositions. Comparison of mean layer width shows a
slight decrease with mass flow rate at both laser power
levels, as shown in figure 6. PREP deposition samples
Sample
Powder
volume (mm3 )
Pores
volume (mm3 )
%
Porosity
GA
PREP
3.877
2.329
0.002117
0.000399
0.055
0.017
generally found to have a lower layer width than GA

deposition samples.
As discussed in section 3.1, the greater energy at
the substrate with GA powder leads to more of the
initial substrate being melted and a wider and deeper
melt pool than with PREP powder. Powder is assimilated over the width of the pool, giving wider deposits
with GA powder than PREP powder (Fig. 6). It is
the amount of powder assimilated along the length of
the melt pool that gives the final track height. This is
initially greater for the GA powder, but the higher temperature of the PREP powder means it is more easily
assimilated towards the rear mushy zone of the pool,
thus increasing the pool length and the deposit height
for this type of powder. So, the melt pool generated during PREP powder deposition will be slightly shallower
and longer and the deposition layer thicker than during
GA powder deposition at the same parameters (Fig. 5).
3.2.2 Surface roughness
Figure 7 shows the surface roughness (Ra ) trends for
GA and PREP powder deposition as a function of mass
628
Figure 5. Comparison of mean layer thickness of GA and

PREP deposition samples as a function of mass flow rate and
laser power; (a) laser power 800 W; (b) laser power 1000 W.
flow rate and laser power levels. Surface roughness

tends to increase with mass flow rate; this increase is
quicker at 800 W than 100 W. There is also a slight
decrease in roughness when laser power increases
from 800 W to 1000 W. The surface roughness of the
PREP powder deposition increases from 4.4 m to
18.12 m as mass flow rate changes from 0.033 gs1 to
0.135 gs1 at 800 W laser power. PREP powder deposition samples show low surface roughness than the
GA deposition samples at all mass flow rates and laser
power levels.
The lower surface roughness (Ra) of PREP deposition samples can be attributed to the particles being
both smaller and hotter and there thus being fewer
cases of only partially-assimilation into the substrate
surface, which is in agreement with previous studies
(Pinkerton & Li 2003a, Pinkerton & Li 2005b).
3.2.3 Intralayer porosity generation in LAM
Form MicroCT data image analysis; porosity is found
at discrete locations of the sample and it is mostly
present in the first deposition layer. The GA deposition
samples also show porosity on subsequent deposition
Figure 6. Comparison of mean layer width of GA and PREP

deposition samples as a function of mass flow rate and laser
power; (a) laser power 800 W; (b) laser power 1000 W.
layers. The intralayer porosity data obtained from the

analysis is presented in figure 8. The results show that
the porosity percentage is highest at 0.033 gs1 mass
flow and porosity tends to decrease with mass flow
rate and reach its minimum value at 0.066 gs1 . Further increase of mass flow beyond this value increases
the intralayer porosity. Overall, results show that PREP
deposition samples have lower intralayer porosity
than GA deposition samples at all mass flow rates and
laser power levels.
GA powder shows more porosity than PREP powder (Fig. 4 and table 2) and there is more intralayer
porosity in GA deposition samples than PREP deposition samples (Fig. 8), evidencing a clear positive
relationship between initial powder porosity and deposition porosity. The effect of processing parameters on
intralayer porosity generation is more complex: strong
Marangoni flows at high laser power, melt pool instability at lower mass flow rates (Shah at al. 2010) and the
effects of these on pore nucleation, growth and movement within the pool may be influential factors. Overall, porosity can be minimised by PREP powder, high
laser power and a moderate powder mass flow rate.
629
Figure 7. Comparison of surface roughness Ra of GA and

PREP deposition samples as a function of mass flow rate and
laser power; (a) laser power 800 W; (b) laser power 1000 W.
3.2.4 Microstructure
Samples produced by laser additive manufacturing of
Ti-6Al-4V exhibit a microstructure of large columnar
prior beta grains, as shown in figure 9, which compares
fifteen-layer, thin walls deposited from GA and PREP
powder.
Approximately equally spaced layer bands are visible through the whole thin-wall except the last three
deposition layers. The prior beta grains boundaries are
continuous across the layer bands. Figure 10 shows the
changes in average prior beta grain sizes for GA and
PREP deposition samples with respect to mass flow
rate and laser power. It can be seen that the size of
prior beta grains tends to increase with laser power
and decrease with mass flow rate. Average grain size
increases from 277 m to 386 m when laser power
changes from 800 W to 1000 W at 0.066 gs1 for GA
deposition samples. PREP powder deposition samples
exhibit larger prior beta grains size than that of GA
powder deposition samples at all mass flow rates and
laser power levels.
The x-ray diffraction of a thin-wall deposition sample at top and bottom positions reveals that the sample
Figure 8. Comparison of intralayer porosity measurement

of GA and PREP deposition samples; (a) laser power 800 W;
(b) laser power 1000 W.
contains both alpha and beta phases (Fig. 11). At the

top, the sample contains more beta phase than at the
bottom. A previous study (Al-Bermani 2010) showed
that the absence of the beta phase might be an indication of a martensitic microstructure. The presence of
beta phase at both positions of the sample thus confirms that microstructure is + rather than a form
of martensitic microstructure. The deviations of peak
intensity from that on an ideal structure indicate that
the material analysed is texture; this is the case for
both alpha and beta.
Scanning electron microscopy was employed to
characterize inside the prior beta grains which exhibit
a fully lamellar + microstructure, as shown in
figure 12. The microstructure exhibits basketweave
like Widmansttten lathes outlined in retained
grains. Figure 12 compares lamellar + phase spacing (S+ ) in GA and PREP deposition samples. The
lamellar + phase spacing (S+ ) tends to increase
with laser power. PREP powder deposition samples
show slightly higher phase spacing than GA powder deposition samples produced at equivalent process
parameters.
The long columnar prior beta grains, which
grow across the layer bands, are mainly inclined
630
Figure 9. Ti-6Al-4V thin-wall structures produced by laser

additive manufacturing showing columnar prior beta grains
and layer band morphology at 0.033 gs1 mass flow rate; (a)
GA, power 800 W; (b) PREP, power 800 W.
perpendicular to the plane of laser scanning direction.

The layer bands, which appears from the first deposition layer and disappear before the last three deposition
layers, are the results of the complex thermal history experienced during multiple layers deposition, as
explained by Kelly & Kampe (2004). The increase in
prior beta grains size with laser power can be related
to the lower cooling rates at higher laser powers. The
prior beta grains of PREP powder deposition samples
being larger than those of GA powder deposition samples at the same process parameters can be explained
by considering the longer melt pool generated during
laser additive manufacturing of the PREP powder. A
longer melt pool decreases the temperature gradient at
the solidification front, resulting in decreased cooling
rates (Hofmeister & Griffith 2001).
The analytical-numerical model developed for laser
additive manufacturing track and microstructure formation by Ahsan & Pinkerton (in press) is used here
to explain the relationship between cooling rates and
solidification microstructure. The model found that
there is a slight, but not steep, increase in cooling rate
with mass flow rate. Conversely, an abrupt decrease in
the grain size with mass flow rate is shown in figure 10.
This suggests that there is another important factor,
along with cooling rates, controlling microstructural
scale and this factor is most likely nucleation density
during microstructure formation. It is possible that an
increased mass flow rate changes solidification conditions in the upper melt pool portion by increasing
Figure 10. Comparison of average prior beta grain size of

GA and PREP deposition samples as a function of mass flow
rate; (a) laser power 800 W; (b) laser power 1000 W.
Figure 11. X-ray diffraction results of a thin-wall deposition sample at two positions i.e. 2 mm and 6 mm above the
substrate level.
nucleation density. The increased nucleation density

then changes the microstructure morphology from
long columnar to small equiaxed grains, as is evident
from figure 10 (Gumann et al. 2001, Pinkerton et al.
631
Table 3. Variations of micro hardness (HV0.5) with lamellar

+ phase spacing (S+ ).
Powder
type
Laser power
(W)
+ phase
spacing (m)
Mean micro
hardness (HV 0.5)
GA
PREP
GA
PREP
800
800
1000
1000
0.95 0.15
1.20 0.22
1.32 0.26
1.56 0.25
363
354
354
343
Figure 12. SEM images of the deposition samples showing

variations in lamellar alpha+beta phase spacing at 0.033 gs1
mass flow rate; (a) GA, power 800 W; (b) PREP, power 800 W.
2006). So, there are two reasons why the grain size
tends to reduce with increasing mass flow rate: the
first is the increase in cooling rates due to less energy
being available at the substrate and the second is the
increase in nucleation density.
3.2.5 Micro hardness
Multiple factors can affect a materials microhardness,
including phase, solute concentration and microstructural scale. However, most of these factors were
common the two types of powder and the powders
and substrate were of the same material, negating
any dilution effects. Therefore, analysis focussed on
quantitative metallographic measurements of lamellar
+ phase spacing (S+ ); its effects on mean micro
hardness are given in table 3. The table shows that
mean micro hardness tends to decrease with increase
in lamellar + phase spacing (S+ ). Vickers micro
hardness (HV 0.5) decreases from 363 to 343 as the
lamellar + phase spacing (S+ ) changes from
0.95 m to 1.56 m. The effects of mass flow rate and
laser power on mean micro hardness of GA and PREP
deposition samples are shown in figure 13. Micro hardness tends to decrease with laser power for both GA
and PREP deposition samples; mass flow rate does not
have a significant effect in either case.
Figure 13. Mean micro hardness (HV 0.5) as a function of

laser power and mass flow rate.
The result presented in table 3 show that the micro

hardness of the deposited samples is primarily related
to lamellar + phase spacing (S+ ). The less the
lamellar + phase spacing (S+ ), the greater the
phase boundary reinforcement, thus increasing the
hardness of the material. The larger lamellar +
phase spacing (S+ ) in the case of PREP powder
deposition may be attributed to the slower cooling rates. Micro hardness results indicate that the
lamellar + phase spacing (S+ ) is not affected
significantly by mass flow rate.
632
CONCLUSIONS
A comparative study of GA and PREP powder

deposition has considered the potential benefits of
using PREP powder in laser additive manufacturing.
MicroCT analysis of the powders shows that PREP
powder has three times less porosity, and PREP powder deposits show lower surface roughness, lower
intralayer porosity and a higher deposition rate. Laser
deposited Ti-6Al-4V exhibits long columnar prior beta
grains with a fully lamellar + microstructure.
Prior beta grain size in the deposits increases with
laser power and tends to decrease with mass flow
for both types of powders. Nucleation density also
plays an important role during microstructure formation and at higher powder mass flow rates increased
nucleation density in the melt pool tends to change
the microstructure morphology from long columnar
to small equiaxed grains. The lamellar + phase
spacing (S+ ) increases with laser power, but seems
unaffected by variation in mass flow rate. Micro hardness of the laser deposited Ti-6Al-4V is dependent
on the lamellar + phase spacing (S+ ): Vickers
micro hardness (HV 0.5) decreases from 363 to 343 as
the lamellar + phase spacing (S+ ) changes from
0.95 m to 1.56 m.
ACKNOWLEDGEMENT
Thanks to the staff of the Laser Processing Research
Centre (LPRC) for their help in the experimentation.
We are thankful to Dr Robert Bradley, Dr Richard
Moat and Ms Judith Shackleton of the University of
Manchester for their help in MicroCT and x-ray
diffraction analysis. The first author would like
to acknowledge the financial assistance of the
Government of Pakistan in the present work.
REFERENCES
Ahsan, M. N. & Pinkerton, A. J. (in press). An anlyticalnumerical model of laser direct metal deposition track
and microstructure formation. Modelling and Simulation
in Materials Science and Engineering.
Ahsan, M. N., Paul, C. P., Kukreja, L. M. & Pinkerton, A.
J. 2011. Porous structures fabrication by continuous and
pulsed laser metal deposi-tion for biomedical applications; modelling and experimental investigation. Journal
of Materials Processing Technology 211(4): 602609.
Al-Bermani, S. S., Blackmore, M. L., Zhang, W. & Todd,
I. 2010. The Origin of Microstructural Diversity, Texture, and Mechanical Properties in Electron Beam Melted
Ti-6Al-4V. Metallurgical and Materials Transactions
a-Physical Metallurgy and Materials Science 41A(13):
34223434.
Alagheband, A. & Brown, C. 1998. Plasma Atomization goes
commercial. Metal Powder Report 53(11): 2628.
Dinda, G. P., Song, L. & Mazumder, J. 2008. Fabrication of

Ti-6Al-4V Scaffolds by Direct Metal Deposition. Metallurgical and Materials Transactions A-Physical Metallurgy and Materials Science 39A (12): 29142922.
Gumann M, Bezenon, C., Canalis, P. & Kurz, W. 2001.
Single-crystal laser deposition of superalloys: processingmicrostructure maps. Acta Materialia 49(6): 10511062.
Hofmeister, W. & Griffith, M. 2001. Solidification in direct
metal deposition by LENS processing. JOM, 53(9):
3034.
Kelly, S. M. & Kampe, S. L. 2004. Microstructural evolution in laser-deposited multilayer Ti-M-4V builds: Part
I. Microstructural characterization. Metallurgical and
Materials Transactions A-Physical Metallurgy and Materials Science 35A(6): 18611867.
Pinkerton, A. J. & Li, L. 2003a. Rapid prototyping using
direct laser deposition the effect of powder atomization type and flowrate. Proceedings of the Institution
of Mechanical Engineers Part B-Journal of Engineering
Pinkerton, A. J. & Li, L. 2003b. Effects of powder geometry and composition in coaxial laser deposition of 316L
steel for rapid prototyping. CIRP Annals-Manufacturing
Technology 52(1): 181184.
Pinkerton, A. J. & Li, L. 2005a. Direct additive laser
manufacturing using gas- and water-atomised H13 tool
steel powders. International Journal of Advanced Manufacturing Technology 25(56): 471479.
Pinkerton, A. J. & Li, L. 2005b. Multiple-layer laser deposition of steel components using gas- and water-atomised
powders: the differences and the mechanisms leading to
them. Applied Surface Science 247(14): 175181.
Pinkerton A. J, Karadge, M., Syed, W. U. H. & Li, L.
2006. Thermal and microstructural aspects of the laser
direct metal deposition of waspaloy. Journal of Laser
Applications 18(3): 216226.
Pinkerton A. J, 2007. An analytical model of beam attenuation and powder heating during coaxial laser direct metal
deposition. Journal of Physics D: Applied Physics 40(23):
73237334.
Qi, H., Azer, M. & Ritter, A. 2009. Studies of Standard Heat
Treatment Effects on Micro-structure and Mechanical
Properties of Laser Net Shape Manufactured INCONEL
718. Metallurgical and Materials Transactions a-Physical
Metallurgy and Materials Science 40A(10): 24102422.
Rabin, B. H., Smolik, G. R. & Korth, G. E. 1990. Characterization of entrapped gases in rapidly solidified powders.
Materials Science and Engineering: A 124(1): 17.
Shah, K., Pinkerton, A. J., Salman, A. & Li, L. 2010. Effects
of Melt Pool Variables and Process Parameters in Laser
Direct Metal Deposition of Aerospace Alloys. Materials
and Manufacturing Processes 25(12): 13721380.
Wang, L., Lang, Z. B. & Shi, H.P. 2007. Properties and forming process of prealloyed powder metallurgy Ti-6Al-4V
alloy. Transactions of Nonferrous Metals Society of China
17: 639643
Zhao, X. M., Chen, J., Lin, X. & Huang, W. D. 2008.
Study on microstructure and mechanical properties of
laser rapid forming Inconel 718. Materials Science
and Engineering: A Structural Materials Properties
Microstructure and Processing 478(12): 119124.
633

Thermoplastic filament extruder head for desktop Additive

Manufacturing machines
P. Inforatti Neto, A.L. Lixandro Filho, F.D.A.S. Pereira & J.V.L. Silva
Centro de Tecnologia da Informao Renato Archer CTI, Campinas, Brasil
Z.C. Silveira
Escola de Engenharia de So Carlos, Departamento de Engenharia Mecnica,
Universidade de So Paulo, So Carlos, Brasil
ABSTRACT: Most of the additive manufacturing (AM) equipments are patented commercial ones, but few
are open-source systems (Fab@Home, RepRap, CandyFab), enabling to disseminate the technology and to
accelerate their development for specific needs. A machine called Fab@CTI was developed based on adaptations
of Fab@Home. This paper shows the development of a printer head to enable the use of thermoplastics polymers
in Fab@CTI and Fab@Home machines by means of filament deposition modeling technology. Major parts of
the extrusion head were built using AM machines and, in near future, the machine will be capable of building
its own parts. It is also possible to adapt the extruder head to different filament diameters and different melting
temperatures (room temperature to 400 C). Polycaprolactone was successfully used to build porous scaffolds
with minimum pore size of 200 microns. Acrylonitrile Butadiene Styrene parts were built with an average layer
thickness of less than 300 microns.
INTRODUCTION
The American Society for Testing and Materials

(ASTM) defined additive manufacturing by the process of joining materials to make objects from 3D
model data, usually layer upon layer, as opposed to
subtractive manufacturing methodologies, such as traditional machining" (ASTM F2792-10, 2010). With
the focus on industry, AM has been used for optimize products production quality (Silva et al. 1999).
Currently this technology is present in various fields
of knowledge, ranging from entertainment to health,
like a surgery aid tool, diagnostics tools and also for
customized prostheses built (Silva et al. 2004). Tissue engineering is still incipient and is seeking in the
AM better ways to reproduce highly complex geometries such as structures to support cell growth and
regeneration called scaffolds.
AM commercial machines generally needs huge
quantities of materials to construct parts, and still offer
limited operational settings to use copyrighted materials from the manufacturers. These limitations are an
issue when the machine needs to be used mainly for
research in materials and processes. To overcome that
some initiatives are emerging to fulfill the researchers
and hobbyists technological needs.
As an example may be cited the RepRap project created in 2004 by Dr. AdrianBowyer and Ed Sells whose
main objective is create a AM machine with low cost,
but also to be able to print their own pieces. Another
example to mention is the open-source Fab@Home
project (Malone & Lipson, 2007) developed at

Cornell University in 2006 by a doctoral candidate
Evan Malone and Hod Lipson his supervisor. This
project was designed to be versatile and able to deal
with various kinds of materials that can be extruded
through a syringe (Lipson & Kurman, 2010).
Within the strategic planning of the Centro de
Tecnologia da Informao Renato Archer (CTI) in
2006, a proposal was introduced to develop an AM
equipment with open-source project to work, initially,
with biomaterials research.
Given the limitations of commercial equipment and
the fact that the Fab@Home project presents similar
features to those targeted for the strategic planning, a
machine was built and modified by CTI researchers
(Fig. 1). This new branch from Fab@Home was called
Fab@CT. (Inforatti Neto & Silva, 2007).
Fab@CTI machines (Fig. 1) has undergone several modifications to the original design, especially
with the development of an interchangeable platform which enables to use different technologies
to structure 3D geometries with varied process and
materials in the same equipment, such as ceramic,
metal (Lobovsky et al. 2008), thermoplastics, among
others.
This work will detail the process of developing a
filament extrusion head adapted to be used, mainly,
in Fab@CTI machine. The needs will be detailed
using engineering project methodology by functional
diagram and morphological matrix to optimize the
construction of the filament extrusion head. It will also
635
Figure 1. Fab@CTI machine equipped with thermoplastic

filament extruder head.
Figure 2. Filament extruder head partial functional diagram

for a generic 3D printer.
be presented that the extruder can build complex 3D

models using thermoplastics.
DEVELOPMENT
The design of the filament extrusion head followed

the proposal of the original Fab@CTI project which
is dealing with different materials and in small quantities. Thus, for its development, the information was
obtained from some commercial equipment as well as
the characteristics of the materials to be used, such
as the filament diameter to be used, melting point,
viscosity, among others.
Engineering project methodology was used as the
survey of the main functions using a functional diagram (Fig. 2) to identify most important parts of the
system to be built. Then a morphological matrix was
created (Fig. 3) to explore all the parameters involved
to seek better global solutions, in order to improve and
increase the success rate in the first development.
Fab@CTI equipment imposed some requirements
like the size of the extruder head, its final weight,
thermal insulation and design. To improve and speed
up the development time, AM commercial machines
available in CTI research center were used to create a
functional model, based on the solutions presented in
Figure 3.
For the development of the heating system were
made initial tests of thermal conduction in different
materials like stainless steel, brass e and aluminum.
The last was chosen because it provided faster feedback to heating and cooling process, combined with
low weight. Some thermal properties of the materials
can be compared using Figure 4. It is important to highlight that even aluminum needing more heat to increase
its temperature, the larger thermal conductivity it is
more relevant for a lower thermal gradient. This makes
Figure 3. Morphological matrix with the best solution

enhanced.
Figure 4. List of possible materials to be used as heated part

of the extruder head. The best material with greater efficiency
and better thermal conductivity is the aluminum (Reference
values from: http://www.engineeringtoolbox.com).
the temperature control for cooling, heating with better feedback comparing with the other materials in the
Figure 4.
Initially, the thermoplastic filament enters the system through a plastic tube (Fig. 6b) is pulled by
two toothed pulleys (Fig. 6c) to a stainless steel tube
to the region of heating (Fig. 6e) where it is cast
and extruded through a nozzle (Fig. 6g). The heating system consists of resistors, aluminum blocks,
and thermocouple as temperature sensor. In the aluminum block (Fig. 6e), two electric heaters (60 Watts
cartridges) were used making possible to reach up
636
Figure 5. Heating test comparative using one (circle) and

two heaters (square).
Figure 7. Scaffolds of PCL built by Fab@CTI. a) Side view;

b) Top view; c) Condyle shape; d) MEV showing adhesion
between layers and porous in Z axis; e) MEV of a porous in
xy axis.
stainless steel tube, preventing heat from spreading in

the region of melting material for traction and entry of
the filament.
Figure 6. Fab@CTI thermoplastic filament extruder head.

a) Stepper motor; b) Plastic filament tube guide; c) Toothed
pulley; d) Fan; e) Resistive heat zone; f) Thermocouple; g)
Extrusion tip; h) PID temperature control.
to 400 C as shown in Figure 5. For the control system, we used a J-type thermocouple (Fig. 6f) and an
external PID (proportionalintegralderivative) temperature controller (Fig. 6h).
One advantage of this new head that is not present
in any other is that all parts were designed to be configurable according to the thickness of the filament input,
which can vary from 1.8 mm to 3.0 mm (Fig. 6i). In
addition, the nozzle used can also be easily replaced
to easily change its diameter (Fig. 6g).
For the thermal insulation of the heating region, a
ceramic blanket was used avoiding deformations and
degradations in the extruder head support.
Finally, to avoid jam in the system due to fusion and
solidification of the filament in an inappropriate area,
a cooling system (Fig. 6d) was inserted at the top of the
Scaffolds of Polycaprolactone (PCL) were built with

pores from 0.20 mm to 1.0 mm and minimum layer
thickness of 0.30 mm (Figs. 7a, b). Also, PCL was used
to mimics the human body structures such as a condyle
(Fig. 7c). In this application, the PCL filament had
a diameter of 2.0 mm input and 0.3 mm output to a
deposition rate of 3 mm/s.
The electron microscope images seen in Figure 7d
show good adhesion between the deposited layers and
the presence of micropores formed during the extrusion process, which are beneficial for cell adhesion in
culture medium.
Note that a pore size greater than 300 microns provides good tissue migration, transport of nutrients and
vascularization (GE et al. 2008). Recent surveys report
more specifically the best range of pore size for different types of cells or tissues. For example, according to
Oh et al. (2007), the pore size of about 5 m to neovascularization, 515 m for growth of fibroblasts,
about 20 m for growth of hepatocytes, 20125 m
for the regeneration of skin, 70120 m for growth
of chondrocytes, 40150 m for the regeneration of
liver tissue, 100300 m for bladder smooth muscle,
100400 m for the regeneration of bone and osteoconductive to 200350 m, depending on the material
used in the scaffold. Therefore, the scaffold model generated by Fab@CTI, Figure 7e, produced pores with
average diameter of 200 microns (simple average of
the distance between the parallel side of each pore)
and can be applied to bone regeneration.
Another material used for 3D structure, was
Acrylonitrile Butadiene Styrene (ABS) filament with
diameter of 1.8 mm. To structured objects with
layer thickness of 0.300 mm (Fig. 8d), the linear
deposition speed was 8 mm/s while for structured
objects with layer thickness of 0.400 mm, the linear
637
especially in small quantities required for research

purposes. This makes the Fab@CTI an extruder head
machine that can handle a wide range of polymers in
the form of filament, including biomaterials.
The filament extrusion head was developed for the
Fab@CTI platform, tough it can be easily adapted to
any additive manufacturing desktop platform.
REFERENCES
Figure 8. Structures of ABS built by Fab@CTI. a,b) Tower

with inside structures; c) Printing process d) Layer thickness
measure.
deposition speed was 15 mm/s. These speeds are limited by the diameter of the nozzle. Smaller diameter
of the nozzle requires greater pressure and eventually
fail due to the geometry of the teeth of the pulley that
damages the filament.
It was observed that the layer thickness is related to
two factors: speed of the extrusion head and deposition
flow rate. Thus the layer thickness is usually less than
the diameter of the filament extruded by the nozzle,
however, the width is greater.
4
CONCLUSIONS
The results were expected and are according with the

original proposal. It was built a fused filament deposition head which is modular, flexible and stable for
using thermoplastics polymer and is able to effectively
be used for building parts for industry (within certain
limits) and research, mainly with biomaterials.
Even with lower resolution (layer height around
0.3 mm) none of the commercial available devices
can handle materials with so different properties,
Ge, Z.; Jin, Z. & Cao, T. 2008. Manufacture of degradable polymeric scaffolds for bone regeneration. Biomed.
Mater. Vol. 3, n 022001, 11.
Inforatti Neto P. & Silva J. V. L. 2007. Mquina bsica
de prototipagem rpida de projeto aberto de hardware
e software para uso de pesquisas em bioengenharia,
desenvolvimento, inovao e aplicaes de prototipagem
rpida. Trabalho de Graduao, Faculdade Independente
do Nordeste.
Lobovsky M., Lobovsky A., Behi M., Lipson H. 2008.
Solid Freeform Fabrication of Stainless Steel Using
Fab@Home. Proceedings of the 19th Annual Solid
Freeform Fabrication Symposium, Austin TX, August
2008.
Malone, E. & Lipson, H. 2007. Fab@Home: the personal
desktop fabricator kit, Rapid Prototyping Journal, Vol. 13.
n 4, 245255.
Oh, S.H.; Park, I.K.; Kim, J.M. & Lee, J.H. 2007. In vitro and
in vivo characteristics of PCL scaffolds with pore size gradient fabricated by a centrifugation method. Biomaterials.
no 28, 16641671.
Silva J. V. L., Saura C. E., Bergerman M., Yamanaka M.
C. 1999. Rapid Prototyping Concept, applications, and
potential utilization in Brazil.
Silva J. V. L., Meurer E., Meurer M. I. 2004. Prototipagem
biomdica Aplicaes cirrgicas.
Volpato, Neri (Ed.). 2007. Prototipagem Rpida Tecnologias e aplicaes. So Paulo: Blcher: 910.
Lipson, H. & Kurman, M. 2010. Factory@Home: The
emerging economy of personal manufacturing. December
2010.
ASTM F2792-10. 2010. Standard Terminology for Additive
Manufacturing Technologies, copyright ASTM International, 100 Barr Harbor Drive, West Conshohocken, PA
19428. A copy of the complete standard can be obtained
from ASTM International, http://www.astm.org.
638

Manufacturing of defined porous metal structures using the

beam melting technology
J.T. Sehrt & G. Witt
University of Duisburg-Essen, Institute for Product Engineering, Manufacturing Technology, Duisburg, Germany
ABSTRACT: For the manufacturing of metal parts with defined porous structures the shaping and sintering
processes are usually used which press the metal powder material to a green body and heat-treat it afterwards.
However, disadvantages of this method can be compensated by applying the beam melting technology. The
characteristic additive build-up at the beam melting technology provides the opportunity to freeform porous and
defined structures at specific areas in one part. Here the process parameters are adjusted in a way to combine
fully dense areas and porous areas in a part individually depending on their needs without any joining process
afterwards. It turns out that specific filter characteristics such as density, permeability, pore size, porosity and
shear strength are comparable to conventionally made porosities. These positive results combined with the
advantages of this technology also lead to new fields of application in future.
INTRODUCTION
fluids. Thereby they fulfill main filtering functions

such as [Sehrt 2010 & Schatt et al. 2007]:
Todays market trend requires innovative thoughts and

further development of manufacturing processes such
as additive manufacturing. Especially with regard to
the increasing diversity of different product versions
with decreasing batch sizes at the same time the additive manufacturing technologies become more and
more important. Generally additive manufacturing differs from conventional technologies by its layerwise
and additive joining together material to a physical part
instead of removing or forming material. [Sehrt 2010]
For the manufacturing of metal parts with defined
porous structures (functional porosity) the manifold
methods of the powder metallurgy are commonly used
today. However, some disadvantages of this method
are the inefficiency in individual production due to
high costs for pressing tools, the restrictions with
respect to press-formability (undercuts) and the fairly
high shrinkage at the sintering process. In this paper
feasibility studies for producing defined porous metal
structures using the beam melting technology are
investigated.
SPECIFIC OBJECTIVES
One focus of this investigation is to illustrate new

development and application potentials for AM. The
main aim is to develop and establish a specific strategy
for the beam melting technology which can be applied
for the manufacturing of open porosities.These porosities are used for the perfusion of liquid or gaseous
protecting (flame arrester)

filtration and separation
dispersion
throttling and silencing
equalizing
sparging and fluidizing
The adjustment of the functional porosity of beam

melted parts made of stainless steel is the focus of this
investigation; therefore the variation and determination of the preferred process parameters is essential.
BASICS ON BEAM MELTING AND POROUS

STRUCTURES
3.1 Beam melting

Among the large number of additive manufacturing
processes the direct manufacturing of metal parts in
a powder bed using an electron or a laser beam is
called beam melting. This neutral term is comparatively new and was introduced and defined in the VDI
Guideline 3404 for the first time. Here beam melting is
also referred to as Laser-Cusing, Direct Metal LaserSintering (DMLS), Selective Laser Melting (SLM) or
Electron Beam Melting (EBM). Using this technology
usually 3D solid parts are made layer by layer from
fine atomized metal powder which is locally molten
by a focused laser or electron beam and connected to
the underlying solid layer (cf. Figure 1) [Sehrt & Witt
2010].
639
Figure 1. Principle of beam melting [Sehrt 2010].
Figure 3. Lateral view of single walls.
Figure 2. Possible porous structures via beam melting.
3.2
Porous structures
For the production of porous structures via beam

melting different strategies can be considered. One
possibility is the 3D design or specification on the
computer with a defined geometry (cf. Figure 2a) followed by the beam melting process itself. This strategy
is mainly used for solid parts. But for filter applications the pores inside are usually too big in this case.
The other two possibilities utilize the parameters of the
beam melting process. One possibility for producing
porous structures is the sinter neck formation known
from conventional sintering among individual particles (cf. Figure 2b). In reality these structures are very
difficult to manufacture due to different particle sizes,
the particle size distribution and the bulk density of the
metal powder. The investigated strategy in this work is
a combination of a defined geometry and specific process parameters (cf. Figure 2c). Here individual hatch
lines in a plane do not touch one another laterally. In
practice it is not possible to manufacture this ideal
line grid structure. Due to the lack of heat dissipation
to the underlying layer, the melt pool of the hatch lines
grows in height and touch one above the other. Nevertheless small holes remain in these areas by adjusting
the parameters. Generating the individual hatch lines
is indeed defined but the pore size and the pore size
distribution are statistical.
EXPERIMENTAL SETUP
DETERMINATION OF PROCESS WINDOW
The energy required to raise the temperature to the

melting temperature of the metal powder is delivered
by the laser. When coupling the laser radiation into the
powder bed not the total amount of the energy is converted into heat. Depending on the material and the
grain size a large portion of energy is reflected and
another portion is transmitted. Only the portion of the
radiation absorbed by the material is converted into
heat. The energy input of the laser into a powder bed is
dependent on the laser power, the hatch distance, the
layer thickness and the scan speed [Meiners 1999].
The layer thickness is kept constant for these studies
to their standard of 20 m in order to build delicate
structures. Also the hatch distance is initially set to
300 m, because a single weld line of GP 1 material is
about 150 m in width (standard parameters). So the
spaces between the grid structures have the same size.
For the further determination of the optimal process
window there are only two parameters left for variation. For the investigation of the two parameters laser
power and scan speed a uniform wall structure must
be found with many small and regularly distributed
holes for the air or gas to perfuse. As a test part for
subsequent assessment a single wall of the ideal line
grid structure is used (cf. Figure 2c). In consequence of
the selected hatch distance and the alternating xy-scan
direction of the laser only a single line is molten in
the y-direction whereas the x-direction shows 3233
short tracks. To evaluate the test results the single walls
are carefully separated from the building platform and
evaluated afterwards under a microscope (backlight
mode). The lateral view of the single walls enables
the visualization of the size, number and distribution
of the resulting holes (cf. Figure 3). Ideally the grid
consists of a stable structure with many small holes
evenly distributed. For that reason test walls with only
a few or no holes are not considered at all (too much
energy). Also test walls with large or irregularly distributed holes are not considered due to their instability
(too little energy).
A pre-selection of good parameters is framed in
Figure 3. A closer look from the top view and a further evaluation of the process window result in the
best parameter combination which is 1000 mm/s scan
speed and 100 W laser power. These parameters are
selected for further investigations. Without altering
the structure in principle the hatch distance is then
reduced in order to decrease the porosity afterwards.
640
The chosen hatch distances are 300 m, 250 m,

200 m and 150 m.
5
RESULTS AND DISCUSSIONSSPECIFIC

FILTER CHARACTERISTICS
With regard to the construction and the use of beam

melted parts and in order to reduce the disadvantages
of former processes a comprehensive material understanding is essential. Potentials for beam melting can
be shown by appropriate design and correct manufacturing of these parts. For this reason specific filter
characteristics of beam melted parts made of stainless
steel are investigated according to different standards
for filter systems.
Figure 4. Round test cylinder and different hatch distances.
5.1
Specific permeability according to

DIN EN ISO 4022
The DIN EN ISO 4022 describes the determination of

fluid permeability of sintered metal materials. The specific permeability is expressed as a friction-dependent
permeability coefficient (laminar flow) and an
inertia-dependent permeability coefficient (turbulent flow). These two coefficients to be determined are
dependent on the pressure drop, the volumetric flow
rate, the viscosity and the density of the test medium.
The following equation shows the relation between the
variables for the calculation.
with
Figure 5. Permeability coefficients and .
where = viscosity coefficient; = inertia coefficient; V = flow rate; e = filter thickness; = dynamic
viscosity air; A = filter surface; p = revised pressure drop at the filter; p = differential pressure;
and ( p ) = fluid density dependent on differential
pressure.
To determine the specific permeability the following round test cylinder was designed (cf. Figure 4a)
and manufactured with the chosen 4 hatch distances
(cf. Figure 4b).
The test cylinder has an outer diameter of 30 mm,
a height of 20 mm and a thickness of 3 mm. The
compressed air runs through the porous structure of
the cylinder. Each cylinder is manufactured 3 times
on order to statistically ensure the results. To determine the flow rate a drum gas meter is connected
at the downstream side of the measuring apparatus.

Due to the pressure drop of the measuring apparatus the measurement is first carried out without the
porous structure. By subtracting these values from
each recorded measurement the best approximation
can be achieved according to the pressure drop only
caused by the porous structure. The resulting values for
x and y can be calculated according to Equation 2 and 3
after all parameters are detected. These values in each
group are shown in Figure 5a. Afterwards regression
lines for each group are drawn into the diagram. The
intersection of these lines with the ordinate (y-axis)
represent the reciprocal of the permeability coefficient
(1/) and the slope of these lines represent the reciprocal of the permeability coefficient (1/). The values
for and in each group can be seen in Figure 5b.
641
Table 1.
Pore diameters and smallest retained particles.
Hatch distance
yS
[m]
Pore diameter
d
[m]
Smallest retained
particle (=0.2x d )
[m]
150
200
250
300
14.77
21.38
89.97
145.97
2.95
4.28
17.99
29.19
shape the apparently largest pore can be calculated

according to the following equation.
Figure 6. Flow rate.
The results for are very similar compared to conventionally sintered metal filters of the same permeability.
Striking are the negative values of according to the
negative slopes of the regression lines. No value of
conventionally sintered metal filters is negative here.
This difference can be justified by the fact that the
pressure drop increases disproportionally to the flow
rate at the conventionally sintered filters. In contrast
to this the pressure drop of the beam melted filters
is not disproportional. On the contrary, the flow rates
increase slightly disproportional to the pressure drop.
This can also be seen by the progressive increase in the
course of the lines in 5.2 in which the flow rate is illustrated dependent on the pressure drop. The dynamic
(turbulent) portion of beam melted filters is smaller
compared to conventionally sintered filters.
Despite of these differences the individual values of
the flow rates and the pressure drops of beam melted
filters are comparable to those which are conventionally sintered. It can be expected that a further increase
in the flow rates (higher speeds) lead to increased inertia effects of the particles which counteract the flow
direction. This would also result in degressive flow rate
lines which asymptotically reach a maximum value
(pressure drop increases exponentially with the flow
rate).
5.2
Pore size according to DIN ISO 4003
According to DIN ISO 4003 the so called bubblepoint test provides a simple method of determining
the size of the apparently largest pore of conventionally sintered filters. The porous specimen to be tested
is clamped and immersed in a liquid whose surface
tension is low and known (e.g. isopropyl). Hereby the
saturation of its open porosities is ensured. Also it is
important to immerse the test specimen as low as possible in the test liquid so that the subsequent escape of
the bubbles out of the specimen can be well observed.
Now slowly increased pressurized air is applied into
the test specimen and the liquid is pressed out of the
pores until the first bubble appears. At this point the
applied pressure corresponds to the size of the capillary tension of the test liquid. Assuming a circular pore
where d = apparent pore diameter; = surface tension isopropyl; and p = pressure drop at filter.
Here the apparent pore diameter relates to a circular
pore whose circumference equals that of the real irregularly shaped pore. By the real shape of the pores even
particles are retained in the filter which are smaller
than the apparent pore diameter. This is why empirical values are given which can be multiplied with the
apparent pore diameter to get an impression of the
smallest particle to be retained in the filter. As specimen for this test also the round test cylinder is used.
The results of this investigation can be seen in Table 1.
According to these values the desired pore diameters
can be adjusted by varying the hatch distance.
5.3 Porosity
The main feature of pressed or sintered parts like filters
is their porosity, which varies greatly depending on its
application. The porosity itself decides significantly
on all properties of a porous structure. Thus the porosity influences the mechanical properties of a part, the
soaking grade (oil), the machinability and the permeability for fluids. Put simply the porosity describes the
ratio of the cavities (pore volume) to the total volume
of a part. The total porosity consists of the open and
closed porosity. The open porosity is characterized by
the fact that all pores are connected to one another
and also connected to their environment so that they
form a continuous pore system (important for filters).
However the closed porosities are not interconnected.
A simple method to determine the overall porosity of
a part is the calculation of the ratio of the porous density of the part to the theoretical density of the solid
material (cf. Equation 6) [Schatt et al. 2007].
where Pt = total porosity; pK = density of porous

part; and kM = density of solid material.
Here porous cubes with an edge length of 15 mm
serve as test specimen (cf. Figure 7a). These cubes
642
Table 2.
Hatch
distance
yS [m]
150
200
250
Figure 7. Test cube and total porosities.
300
Orientation
Mean shear
strength
[N/mm2 ]
Type shear
strength
[N/mm2 ]
0_0
90_0
90_45
0_0
90_0
90_45
0_0
90_0
90_45
0_0
90_0
90_45
402.16
431.35
447.96
276.82
307.85
333.76
203.36
209.19
244.50
151.15
153.33
182.87
427.16
306.15
219.02
162.45
plane. Each round plate is again manufactured 3 times

for testing with the intention to statistically ensure the
results. The mean values of the calculated shear
strengths are summarized in Table 2.
When looking at the names of the orientation in
Table 2 the first number describes the polar angle
whereas the second number is the azimuth angle of the
disc axis. So there are the following three orientations.
Figure 8. Porous beam melted round plates.
are beam melted according to the same parameters for

porous structures described above. After being investigated the following porosities can be calculated for the
different parameters (cf. Figure 7b). As expected, the
porosity increases with the hatch distance from 17.2%
(hatch distance 150 m) to 51.7% (hatch distance
300 m).
5.4
Shear strengths.
Shear strength according to DIN 30911 part 6
The strength of filter material is investigated according to DIN 30911 part 6. In detail the shear strength
is determined. For the shear test a smaller disc with
a diameter d 14 mm is stamped out of a filter element with a specific thickness s (cf. Figure 8b). For
this purpose the required stamp force is measured. The
subsequent calculation of the shear strength is given
in Equation 7.
where = shear strength; F = stamp force;A = surface;

d = diameter; and s = thickness.
For this investigation porous, round plates are used.
They have a diameter of 20 mm and a thickness of
3 mm (cf. Figure 8a).
In addition to the different hatch distances the round
plates also differ by 3 different orientations on the
building platform. At the first orientation the flat circular surface of the plates is parallel to the xy plane.
The two other orientations of the plates are positioned
with their circular surfaces perpendicular to the xy
0_0: flat lying

90_0: standing vertical and parallel to x-axis
90_45: standing vertical and 45 to x-axis
The mean shear strengths of beam melted porous
structures in Table 2 increase from 162.45 N/mm2
(at hatch distance 300 m) to 427.16 N/mm2 (at hatch
distance 150 m). The explanation is the smaller the
hatch distance the higher the number of walls inside the
porous structure and thus the more resistant material
exists. According to this it is also noticeable that the
shear strength is more dependent on the selected hatch
distance than on the orientation.
6
CONCLUSIONS
The tested values of porous beam melted structures

??are consistent with the three representative values
which are given in DIN 30910 part 2 for sintered
stainless steel. According to this the viscosity coefficients of type 150 and type 200 are in the area
of the sintered stainless steel named Sint-AF 40-3. In
addition the filter porosity of type 150 corresponds
to the Sint-AF 40-3 structure. The third value given
in DIN 30910 part 2 is the shear strength. The values
of beam melted porous structures (162427 N/mm2 )
are far above the values of the sintered stainless steels
with 70190 N/mm2 . The very good variety of different porosities and pore sizes in combination with other
specific filter properties and the freedom of design at
the beam melting technology lead to very complex
filter systems in future. Also the beam melting of filters made of other metallic powder materials lead to
643
additional properties. Using this technology materials can be processed and filters can be manufactured
which can not be produced by conventional methods
at all. This is very important for specific applications
such as gas turbines. For this reason also a patent was
applied in cooperation with the industry. One aim is to
higher the cooling efficiency of blades. Recent work
with regard to porous structures such as filters is done
according to new scan strategies of the laser.
REFERENCES
DIN EN ISO 4022. 2006. Permeable sintered metal material
Determination of fluid permeability
DIN ISO 4003. 1990. Permeable sintered metal materials
Determination of bubble test pore size
DIN 30910 part 2. 1990. Sintered metal materials
Werkstoff-Leistungsbltter
(WLB)
Sint-material
specifications part 2: materials for filters
DIN 30911 part 6. 1990. Sintered metal materials Sintetesting procedures part 6: Determination of filter
properties
Meiners, W. 1999. Direktes Selektives Laser Sintern einkomponentiger metallischer Werkstoffe. Aachen: Shaker Verlag, ISBN: 3-8265-6571-1
Schatt, W., Wieters, K.-P. & Kieback, B. 2007. Pulvermetallurgie Technologien und Werkstoffe. Berlin Heidelberg:
Springer-Verlag, ISBN-13 978-3-540-23652-8.
Sehrt, J.T. 2010. Mglichkeiten und Grenzen bei der
generativen Herstellung metallischer Bauteile durch
das Strahlschmelzverfahren, Dissertation, University of
Duisburg-Essen, ISBN 978-3-8322-9229-4
Sehrt, J.T. & Witt, G. 2010 Part Management by Direct Integration of RFID Tags into Beam Melted Parts, RAPID
2011 Conference, Minneapolis, MN, USA, 23th26th
May 2011 (accepted)
VDI Guideline 3404. 2009. Additive fabrication - Rapid technologies (rapid prototyping) Fundamentals, terms and
definitions, quality parameters, supply agreements.
644
Virtual Environments and Simulation

An ontology-based automotive troubleshooting configuration system

development
J.S. Liang
Department of Vehicle Engineering, Yung-Ta Institute of Technology and Commerce, Taiwan
ABSTRACT: With the objective of enabling efficient generation of automotive troubleshooting configuration
by reusing configuration knowledge, an ontology-based approach to building troubleshooting configuration
knowledge is presented in this study. The utilization of ontologies as metadata descriptions of the information
sources is a possible approach to offering an integrated view of multiple elements libraries. A configuration metamodel is defined. The meta-concepts have explicit ontological semantics, so that they assist to identify domain
concepts consistently and structure them systematically. Modeling ontologies of real automotive standard braking
system is taken as an example task to show how to use the meta-concepts for configuring the troubleshooting
information.
1
INTRODUCTION
An automotive troubleshooting is planned and created using ready-made skills, components and tools
from multiple suppliers, e.g. motor company, material
supplier, and appliance supplier. Hence, the technicians at repair shop want to locate suppliers easily and
effectively evaluate the complementary elements. In
addition, because a number of recent developments in
information technology have opened up a vast potential for new electronic forms of procurement, users
have the possibility for searching for up-to-date data
all over the world and for selecting the most favorable offer. Previous research effort focused mainly
on the actual configuration process for solving product configuration problems, such as the rule-based
approach (Li et al., 2009) and the constraint satisfaction problem (CSP) approach (Wang et al., 2009;
Yuan et al., 2009). Besides, attention has been directed
toward the study of conceptual modeling of customizable products, namely product configuration models
(Wang et al., 2009). Product configuration systems
are able to deal with the problems of configuring complex products under mass customization. Meanwhile,
the reusability of configuration models can effectively
reduce the time of developing product information
systems. Ontology, which is defined as the conceptualization of terms and relations in a domain, provides
a means to structurally represent and reuse domain
knowledge. As an explicit specification of a conceptualization, ontology plays a significant role in
these aspects: sharing information, integrating different applications, implementing interoperability and
reusing knowledge.
In this study, we address the modeling of automotive troubleshooting configuration system with an
ontology-based approach in which structural knowledge is formalized ontology web language (OWL)
(McGuinness and van Harmelen, 2010; Grau et al.,

2008), and constraint knowledge in semantic web rule
language (SWRL) (Parsia et al., 2010), a rule language
based on OWL. Through the transformation of configuration knowledge into jess facts and jess rules, actual
configuration processes are executed with the support
of jess (Friedman-Hill, 2010), a rule engine for the java
platform.
2
BACKGROUND AND RELATED WORKS
2.1 Automotive troubleshooting configuration

Given a set of predefined elements, the task of troubleshooting configuration is to find a configuration
solution satisfying individual needs of technicians
without violating all constraints imposed on elements
due to technical and economical factors. Information configuration models describing all legal combinations of elements include knowledge about the
structure of troubleshooting and knowledge about
technical and economical criteria. Using a problemsolving technology, configuration engines perform
actual inference processes with both configuration
models and user requirements as the inputs and then
create a configuration as the output. A configuration
solution consists of the element individuals, the assignment of values to properties of these individuals and
the connection relations among elements such that
all criteria are satisfied. The architecture of an automotive troubleshooting configuration is illustrated in
Figure 1.
2.2 Related work in ontologies for product
configuration
Since ontology can facilitate cooperation among
participants (e.g. customers, technicians and sale
647
Figure 1. The architecture of automotive troubleshooting

configuration.
Figure 2. Hierarchical organization for automotive troubleshooting configuration.
engineers, etc.) by sharing common terminologies and

vocabularies in a domain, research regarding the application of ontology to the product configuration domain
has attracted a lot of attention in many fields recently.
Soininen et al. (1998) proposed a general ontology
of configuration, which integrated the concepts port,
function, constraint, attribute, and so on. Encoding
ontologies in a rigorous way is crucial to constructing
knowledge bases and making inference on knowledge
bases. For ontology codification, there are two mainstream approaches for formalization of ontologies
(Kalinichenko et al., 2003) one is based on first-order
logics (such as Ontolingua) and another is to employ
a description-logic approach (e.g. OWL). Although
Ontolingua is the most expressive of all the languages
that have been used to represent ontologies, its high
expressiveness results in difficulties in building a reasoning mechanism for checking the consistencies of
knowledge bases. A configuration ontology was developed (Altuna et al., 2004) in the OBELIX project to
support collaborative configuration of products and
services. The aim of the OBELIX project is to configure multi-products or services where constraints
are relatively simple and ports used to receive or
provide services are much more complicated. The representation of constrains in the OBELIX configuration
ontology is rather simpler. Typical constraints are those
restricting the number of components or services, the
parameters of components or services. Meanwhile,
ontologies are encoded in RDF. The representation of
complex constraints such as chaining of properties is
not supported in RDF.
As for the other applications, the ontology-based
assembly design was addressed (Kim et al., 2006)
to support collaborative product development such
that design intent could be well understood by other
designers, and the applications could reason about
assembly knowledge without any semantic ambiguity. An ontology for engineering design was developed
for knowledge sharing among engineer to assist engineers in indexing, searching, and retrieving design
knowledge (Ahmed et al., 2007). Cho et al. (2006)
realized that seamless integration of current digital
part libraries was impeded by semantic heterogeneity. They developed meta-ontology for part libraries to
provide an integrated view of multiple part libraries.
3 AN APPRROACH FOR MODELING

AUTOMOTIVE TROUBLESHOOTING
CONFIGURATION KNOWLEDGE
3.1 Architecture for automotive troubleshooting
configuration
Figure 2 illustrates the hierarchical organization for
troubleshooting configuration knowledge. At the first
level, namely the representation level, the aim is to
choose an ontology representation language to formalize configuration models. Typical knowledge representation languages involve OWL, UML meta-model,
KIF (Knowledge Interchange Format), etc. the second
level from up to bottom is the meta-model level that the
configuration meta-ontology is defined. An ontology
in a level plays the role of a meta-ontology of those in
an upper level. That is, an ontology gives a controlled
vocabulary and structure for modeling consistently
and systematically ontologies of the upper levels. Consequently, configuration meta-ontology serves as a
common semantic framework on which configuration models for specific configuration domain can
be built through the inheritance or reuse of concepts
and relations in the meta-model. The configuration
models for specific configuration domains, such as
appliance configuration, material configuration, etc.,
lie in the third level. The bottom level is the instance
level, denoting a concrete configuration result that
conforms to configuration models without violating
any criterion or constraint. By using the architecture, knowledge share is ensured due to the fact that
meta-ontology can be reused in various configuration
domains. Meanwhile, the presented approach is flexible in choosing the knowledge representation language
to formalize troubleshooting configuration models.
3.2 Meta-ontology for automotive troubleshooting
configuration
Meta-ontology for automotive troubleshooting configuration serves as the meta-model defining general
and common terms and relations common to the
configuration domain and thus is independent of specific configuration domains such as appliance and
material configurations. Meta-ontology involves the
identified concepts and relations among them. Metaontology is built by using the methodology presented
648
Figure 3a. The hierarchy of classes in the meta-ontology.
Figure 4. Meta-ontology for automotive troubleshooting

configuration.
which is indicated by the relationship Has_criterion

between component and criterion. Figure 4 shows
all identified classes and the relations among them
in the meta-ontology for automotive troubleshooting
configuration.
Figure 3b. The relations in the meta-ontology.
(Silva and Rocha, 2002; Ohgren and Sandkuhl, 2005).

This following section depicts identified concepts and
relations within the meta-ontology model.
The definitions, possible terms, and concepts within
the troubleshooting configuration domain are investigated by author. The terms investigated involved tool,
material, port, resource, attribute, value, etc. An ontology consists of classes, properties and axioms. A class
defines a concept while a property between two classes
indicates a relation holding between two instances of
these two classes. Based on the terms investigated,
classes and relations among them are extracted. The
hierarchy of classes in the meta-ontology for automotive troubleshooting configuration is illustrated
Figure 3(a).
In addition to the identified classes in the metaontology for automotive troubleshooting configuration, some relations among these classes are defined,
as illustrated in Figure 3(b). There are two properties defined Object and Data. The former shows
that ranges of these relations belong to classes. The
latter means that its range is integer, float, and
so on. The Has_attribute defined between Material and Attribute indicates that a material may have
attributes. The Has_node that exists between component and node means that a component may have
nodes. A material may have some criteria attached to it,
IMPLEMENTING A TROUBELSHOOTING
CONFIGURATION SYSTEM BASED ON
THE RULE ENGINE
This study applies an ontology-based approach for

generating troubleshooting configuration in automotive standard braking system. The building blocks
of the ontology-based troubleshooting configuration
system architecture mainly include ontology, service,
interface and user four layers, as illustrated in Figure 5(a). To implement a troubleshooting configuration, a forward-chaining rule engine is employed
in this study to perform actual inference processes.
We choose jess, a rule engine for the java platform, which is a corresponding java version of the
CLIPS expert system (Jess rule engine, 2010). The
jess uses the forward-chaining reasoning method and
implements the efficient Rete algorithm to process a
number of rules. The architecture of ontology-based
troubleshooting configuration system is shown in Figure 5(b). The developed troubleshooting configurator
is composed of three portions: the jess rule engine,
the OWLTojess and the SWRLTojess transformers.
The jess rule engine acts as the inference module for
matching the head of rules in the rule base with the
facts in the fact base. These rules whose premises are
satisfied according to matched facts are activated and
649
(deftemplate Component extends Material)

(deftemplate Part extends Component)
(deftemplate Caliper extends Part)
(deftemplate Brake_pad extends Part)
(deftemplate owl:Thing (slot name_t))
(deftemplate Tool extends owl:Thing)
(deftemplate Wrench extends Tool)
(deftemplate Screwdriver extends Tool)
The deftemplate construct in jess is employed to
define the type of slots in a fact. The extends construct indicates the inheritance relationship between
two templates defining facts. owl:Thing is a most
top level class in OWL from which all other classes
are directly or indirectly subclassed. The instances of
the OWL classes in configuration knowledge base
are actually involved in a troubleshooting configuration solution. For example, two instances, named
InsHigh_performance_pad and InsOpen-end_wrench,
which are the instances of the OWL class Brake_pad
and Wrench respectively are transformed into the
following facts using the OWLTojess.
Figure 5a. Conceptual diagram for developing troubleshooting configuration.
Figure 5b. Architecture of ontology-based troubleshooting

configuration system.
Jess facts
wait to be fired. To carry out inference processes,

OWL-based configuration knowledge and SWRLbased configuration criteria are required to be mapped
onto jess facts and jess rules respectively, which are
performed by developing corresponding transformers,
the OWLTojess and the SWRLTojess.
4.1 The interface layer
The interface layer provides a graphic front-end such as
navigation structures, knowledge visualizations, and
semantic browsers to the users at the user level and
supports the incremental development of user queries
in a graphic and semantic way. The interface layer
also offers a back-end to access various ontologybased web services such as semantic annotation,
ontology registration, aggregate directory, ontology
transformation, ontology reasoning and inference several services at the service layer to facilitate knowledge
consumption, reuse and supply on the semantic web.
4.2 The transformer of OWLTojess
The classes in OWL knowledge base are mapped
onto the jess templates that define the types of jess
facts. For example, in the troubleshooting configuration knowledge base for the case study, the classes
Caliper, Brake_pad, Wrench, Screwdriver as well
as their parent classes are transformed into these
following jess templates
(assert (owl:Thing (name InsHigh_performance_pad)))

(assert (Material (name InsHigh_performance_pad)))
(assert (Component (name InsHigh_performance_pad)))
(assert (Part (name InsHigh_performance_pad)))
(assert (Brake_pad (name InsHigh_performance_pad)))
(assert (owl:Thing (name InsHigh_performance_pad)))
(assert (Tool (name InsOpen-end_pad)))
(assert (Wrench (name InsOpen-end_pad)))
The assertconstruct in jess is used to declare a fact,

namely that InsHigh_performance_pad is an instance
of the Brake_pad type. The InsOpen-end_wrench is an
instance of the Wrench type. The declared facts will
be saved in the jess fact base. Meanwhile, some relationships between instances or between instances and
data values exist in the configuration. It is essential
to transform the relationship facts into jess facts. For
example, the relation that InsDisk_brake_assembly
(an instance of the class Disk_brake_assembl) has
InsHigh_performance_pad (an instance of the class
Brake pad) as its constituent part is transformed into
the jess fact.
Jess facts
(assert (has_part InsDisk_brake_assembly
InsHigh_performance_pad))
(assert (has_brakepad InsDisk_brake_assembly
InsHigh_performance_pad))
Jess templates
4.3 The transformer of SWRLTojess
(deftemplate owl:Thing (slot name_p))
(deftemplate Material extends owl:Thing)
The SWRL-based criteria knowledge is transformed

into jess rules by generating the SWRLTojess
650
transformer. For instance, the demand criterion C1 is

transformed into jess rules, as depicted below.
Jess rules
(defrule C1
(Automotive stand braking system(name ?u))
(Has_diskbrakeassembly ?u ?v) (Diskbrakeassembly
(name ?v))
(Has_caliper ?v ?w) (Has_brakepad ?v ?p)
(Caplier (name ?w)) (Brakepad (name ?p))
=>
(assert(dema_C1 ?u ?w ?p)))
The transformation of the node criterion can be
handled in a similar way, as shown below.
Jess rules
(defrule C3
(Automotive stand braking system(name ?u))
(Has_diskbrakeassembly ?u ?v) (Diskbrakeassembly
(name ?v))
(Has_caliper ?v ?w) (Caplier (name ?w))
(Has_ns_node ?w, ?x)
(Hydraulicunitassembly (name ?p)) (Has_valve ?p ?q)
(Valve (name ?q)) (Has_combinationvalve ?q ?r)
(Combinationvalve (name ?r)) (Has_ne_node ?r ?s)
(Cns1 (name ?p)) (Cne1 (name ?s))
=>
(assert(node_C3 ?u ?p ?s)))
In these jess rules, the assertions (e.g. dema_C1 )
that corresponding criteria are put into the jess fact
base when the SWRL rules are transformed. Both two
transformers are created by using the Extensible Stylesheet Language (XSLT) to perform the transformation
from XML documents to jess files. The XML documents are handled through Document Object Model
and Simple API for XML specifications
4.4 Implementation of jess inference engine
Importing jess files including jess rules and facts
derived by the transformers OWLTojess and SWRLTojess, troubleshooting configuring processes are implemented by the inference engine. The engine applies the
rules to facts in the fact base, and activates the rules
whose premises are satisfied, fires the activated rules
and executes the actions specified in the head of these
activated rules. The process is terminated until no rules
are in the rule base, and then the troubleshooting configuration results are derived and displayed. For the
troubleshooting configuration in automotive standard
braking system, there is a total 700 facts in the fact
base and of 36 rules in the rule base.
4.5 Communication architecture
A meaningful communication in a distributed architecture is possible only in the case that the communicating
Figure 6. An inform example of FIPA-ACL message.
agents achieve their knowledge interoperability. Agent

Communication Languages (ACLs) provide agents
with a means of knowledge interchange format. The
implementation described in this study adopts FIPAACL (FIPA specifications, 2010). The format of
FIPA-ACL mainly consists of the message type, the
identities of the sender (e.g. client end) and receiver
(e.g. server), the language to express the content of
messages exchanged between agents, the ontology to
offer terms for giving a meaning to the symbols in the
content expression, the protocol to specify the interaction protocol that the sending agent employs, and the
content to represent the domain knowledge exchanged
between agents. The query request for troubleshooting ontology at the ontology level can be transformed
from FIPA-ACL messages into OWL-QL format
[6263], while the troubleshooting configuration
ontology with OWL format can be encapsulated into
FIPA-ACL messages to facilitate communication and
sharing among agents. Figure 6 illustrates an example
of FIPA-ACL messages in which user agent informs
expert agent that a symptom of excessive pedal travel
with bent lining and shoe cause. The troubleshooting configuration is retrieved. Finally, the system
will display the table of troubleshooting configuration
solutions, including fields such as Solving Step Tool,
and Solving Note as shown in Figure 7. The technician can follow each process step by step and verify that the troubleshooting configuration solves the
symptom.
651
Figure 7. Table of troubleshooting configuration solutions.
CONCLUSION
We adopt an expressive OWL ontology language and

a SWRL rule language to model automotive troubleshooting configuration knowledge in which structural knowledge is represented in OWL and constraint
knowledge is described in SWRL in this study. The
concept ontology helps ontology modelers to distinguish troubleshooting concepts consistently and structure them systematically, so that ontology mismatches
are confined to manageable mismatches. We applied
the method for knowledge systematization to modeling the ontologies of real automotive troubleshooting
configuration. The benefit of OWL-based automotive
troubleshooting configuration models is that the reuse
of configuration models can be ensured since OWL,
as an ontology language, supports knowledge reuse
and sharing. The actual automotive troubleshooting
configuring process is implemented using a jess rule
engine by mapping OWL-based configuration knowledge and SWRL-based criteria knowledge into jess
facts and jess rules, respectively. Meanwhile, making
use of an existing rule engine, namely jess, facilitates
rapid development of an automotive troubleshooting configurator. However, the presented study only
deals with finding feasible solutions for troubleshooting configuration. In the future work, an algorithm
to obtain an optimal configuration in terms of some
objective is required to be developed when several
configuration solutions exist.
Although the authors work is illustrated in the troubleshooting application domain of automotive standard braking system, the approach can be extended for
more general engineering applications on the semantic
web. In the future work, we will look into developing and publishing more diverse domain-specific
engineering ontologies, e.g. design, manufacturing,
process planning, maintenance, etc. in OWL format
in order to capture an extensive set of annotations of
general engineering knowledge.
manufacturing, International Journal of Computer Integrated Manufacturing, 22 (3): 210224.

Wang L, Ng WK, Song B (2009) Constraint satisfaction
approach on product configuration with cost estimation,
Lecture Notes in Computer Science, 55(79): 731740.
Yuan QK, Zhang MT, ShiYP (2009) An eCommerce-oriented
product configuration design system based on Web, Proceedings of the 2009 International Symposium on Web
Information Systems and Applications, Nanchang, China,
May 2224: 414418.
Wang X, Zheng J, Zeng Q (2009) A design of product collaborative online configuration model, Lecture Notes in
Computer Science, 57(38): 359366.
McGuinness DL and van Harmelen F (201) OWL web
ontology language overview [online], Available from:
http://www.w3c.org/TR/owl-features//. [Accessed in 23
Aug. 2010]
Grau BC, Horrocks I, Motik B, Parsia B, Patel-Schneider P,
Sattler U (2008) OWL 2: the next step for OWL, Web
Semantics: Science, Services and Agents on the World
Wide Web, 6(4): 309322.
Parsia B, Sirin E, Grau BC, Ruckhaus E, Hewlett D
(2010) Cautiously approaching SWRL [online], Available from: http://www.cs.uwaterloo.ca/gweddell/cs848/
SWRL_Parsia_et_al.pdf. [Accessed in 25 Aug. 2010]
Friedman-Hill E (2010) Jess, the rule engine for the java platform, [online],Available from: http://www.jessrules.com/.
[Accessed in 30 Aug. 2010]
Soininen T, Tiihonen J, Mannisto T, Sulonen R (1998)
Towards a general ontology of configuration, Artificial Intelligence for Engineering, Design, Analysis and
Manufacturing, 12(4): 357372.
Kalinichenko LA, Missikoff M, Schiappelli F, Skvortsov N
(2003) Ontological modeling, In: Proceedings of the fifth
Russian Conference on Digital Libraries, St.-Petersburg,
Russia, Oct. 2931.
Altuna A, Cabrerizo A, Laresgoiti I, Pena N, Sastre D (2004)
Co-operative and distributed configuration, Lecture
Notes in Computer Science (LNCS), 3263: 6980.
Kim KY, Manley DG,Yang H (2006) Ontology-based assembly design and information sharing for collaborative
product development, Computer-Aided Design, 38(12):
12331250.
Ahmed S, Kim S, Wallace KM (2007) A methodology for
creating ontologies for engineering design, Journal of
Computing and Information Science in Engineering, 7(2):
141150.
Cho J, Han S, Kim H (2006) Meta-ontology for automated
information integration of parts libraries, Computer-Aided
Design, 38(7): 713725.
Silva N and Rocha J (2002) Merging ontologies using
a bottom-up lexical and structure approach, In: Seventh International Society for Knowledge Organization
Conference, Granada, Spain, July 1013.
Ohgren A and Sandkuhl K (2005) Towards a methodology
for ontology development in small and medium-sized
enterprises, IADIS International Conference on Applied
Computing, Algarve, Portugal, Feb. 2225.
Jess rule engine [online], Available from: http://herzberg.ca.
sandia.gov/. [Accessed in 20 Oct. 2010]
Foundation for Intelligent Physical Agents, FIPA specifications [online], Available from: http://www.fipa.org/
specifications/ [Accessed in 1 Nov. 2010].
REFERENCES
Li N, Liu D, Zhang J, Chen Y (2009) A method of rule
match conflict resolution for product configuration in
652

A tactile display for texture perception in virtual environments

M. Mengoni, M. Germani, B. Colaiocco & P. Morichetti
Department of Mechanics, Polytechnic University of Marche, Ancona, Italy
ABSTRACT: The paper presents a novel tactile display to simulate material properties such as roughness and
texture while exploring Virtual Prototypes surfaces in Virtual Reality environments. The final aim of the system is
to immerse users into an involving and enhanced product experience stimulating emotional and affective reaction.
The developed display adopts a selective stimulation technique to differently activate skin mechanoreceptors and
hence to generate surface texture sensations. The stimulation is realized by both an electric component made
of a pin matrix through which the current flows and by a mechanical one obtained by a shaker keeping the
tactile pad vibrating. Main development stages are described and experimental results discussed. They show that
almost 65% of users are able to discriminate different signals. When acoustic cues are added, this percentage
increases.
INTRODUCTION
The study of human interaction by means of virtual

prototypes (VP) has drawn a growing attention in
the last years due to their role in supporting product
design. Novel human-computer interfaces have been
introduced to simulate users experience and to predict
the perceived product qualities before the artefact has
been realized. Their integration in product development aims at fostering user-centred design (UCD) to
better answers to customers demands and at reducing
the number of costly physical prototypes necessary for
design solutions validation.
The creation of a realistic feedback during VP
exploration is influenced by the usability of the
adopted user interface, by the level of provided immersion and presence in the virtual environment, by the
number of involved human sensory channels and the
way they are stimulated. The paper focuses on the last
aspect of VP interaction to enhance product experience
into simulated environments. Numerous researches
confirm that the product experience is as similar as
the real one if the VP technology is able to simultaneously simulate visual, sound and tactile cues (Duffy,
2007; Hale et al., 2009). In this context, a previous
research deepened the role of sensorial modalities in
product experience, revealing that although vision is
considered the main perceptual channel, touch is the
mean to trigger out emotional and affective response
that influence user decision-making (Mengoni et al.
2010). It allows users to assess synaesthesia qualities such as significance and pleasure (Schifferstein,
2006). However, while visualization technologies are
quite mature, tactile feedback is still poorly reproduced
and scarcely simulated. Haptic interfaces allow users
to touch and feel the simulated objects they interact with. They can implement two basic perceptual
modalities: stimulation of muscles, tendons and joint

sensory receptors (kinesthetic) and stimulation of skin
receptors (cutaneous). The first modality commonly
adopts force feedback devices (FFD), while the second
one exploits cutaneous and thermal feedback (tactile feedback). FFD devices are generally concerned
with the reproduction of contact and non-contact sensations and their combination with force feedback
(Solazzi et al., 2010; McMahan et al., 2010; Bordegoni
et al., 2010) to carry out tasks such as parts assembly, clay modelling, product parts movement, etc. The
second type of systems aims at providing the cutaneous feedback by stimulating the different skin units
and reproducing tactile sensation, thermal impression
and skin deformation (Shinoda et al. 2004). These
sensations allow users to reach deep sensations and
emotional reactions during VP exploration. As a consequence, an effective virtual environments dedicated
to UCD cannot neglect the simulation of material properties such as roughness-smoothness, stick-slippery,
texture, etc. Due to the complexity of the perceptual mechanisms and of their reproduction, research
in tactile display is still far away to find a feasible
solution.
The proposed work focuses on tactile devices for
surface texture perception to fill the lack of available devices in roughness and texture simulation. The
developed tactile display, here described, adopts a
selective stimulation method to properly elicit skin
mechanoreceptors. It consists of a touch pad kept
vibrating by a shaker and headphones reproducing the
emitted sounds when the finger moves on a material
surface. Both electric and mechanical transmitted signals directly derive from the analysis and elaboration
of real material properties. The pad aims at reproducing different materials texture sensation during VP
exploration. It represents an innovation if compared
653
to current systems, especially for the following

aspects:
The correlation with material properties. The signal
waveforms adopted directly derive from processing of material surface. Synthesized signals are
obtained by properly elaborating surface texture
profiles;
The combination of mechanical and electrotactile
stimulations according to the frequency ranges of
human tactile units.
Main development stages and implementation
results are presented in detail. Preliminary experimentations have been carried out on 10 sample users
without any cutaneous sensitivity problems to measure
both signal discrimination and systems latency and
reliability. Testing results are illustrated and discussed.
2
RELATED WORKS
The development of a tactile display able to simulate

material properties sensations requires the investigation of:
How products are perceived by humans and which
type of information is transmitted while touching a
surface;
The tactile perception mechanisms involved in surface exploration from psychophysical and neurological perspectives. It allows the identification of
which receptors characteristics stimulate material
properties sensations;
Drawbacks and potentialities of digital enabling
technologies that are classified into mechanical, vibrotactile, eletrotactile and force feedback
devices.
The first issue was tackled in a previous research work
(Mengoni et al., 2009) that led to the identification
of the human capabilities supported by touch such
as sense of limb movement, perception of own position, improvement of visual information, perception
of hands speed movement on a surface, etc. The analysis of the touch role in decision-making highlights
that consumers are particularly attracted by materials
through which they achieve an active exploration of
products meanings, mainly affective and emotional
(Mengoni et al. 2010).
The perceptual process starts with the stimulation of skin mechanoreceptors by a physical media
that generate a sensation. The sensorial feedback is
processed by the neuphysiologic system and elaborated by the brain cortex. This demonstrates that a
key role to explore and perceive objects with bare
fingers is played by the mechanoreceptive afferents
innervating the glabrous skin. They can be classified into slowly adapting (SAI and SAII) and fast
adapting (FAI and FAII). Different receptive fields,
sensitive frequencies and amplitude thresholds characterize them (Knibestol, 1975; Johansson, 1987,
Muniak et al.,2007). FAI units have small receptive
fields (6208 mm2 ) with relatively distinct borders,
an amplitude threshold less than 0,5 mm and they are

sensitive to a low range (864 Hz) at very low levels of
amplitude. On the contrary, FAII receptors have larger
fields (444 cm2 ) with a more sensitive central area,
lower thresholds (50250 m) and higher sensibility
to vibratory stimuli (above 64 Hz, with a maximal sensitivity at 128400 Hz). SAI units are sensitive at low
stimulus frequency (232 Hz) while SAII units have
maximal sensitivity below 8 Hz. SAII contribution to
perception is smaller than the other units influencing more than 97% of the variance in ratings of the
perceived intensity.
An explorative study of psychophysical mechanisms allows to correlate the perception of materials
surface finishing (i.e. course and fine texture, roughness and friction) and peripheral neural response
(Hollins at al., 2002). SAI and FAI units are connected with the perception of smooth surfaces and
coarse-fine texture, FAII with roughness variation
and friction. Different types of stimuli elicit different mechanoreceptors. Two studies have been useful to
identify correlations among material properties, receptors characteristics and types of vibrating and electrotactile stimuli (Cascio et al. 2002; Bensmaia et al.,
2005). In particular SAI and FAI units can be stimulated by electric signals at low frequencies (5150 Hz)
while FAII by a mechanical vibration at 200250 Hz.
Literature overview points out the importance of
involving the stimulation of all recognized tactile channels according to the identified means to create an
effective roughness simulation display. This leads to
the need to adopt a selective stimulation approach.
2.1 Tactile displays: types, limits and potentialities
Tactile displays aim at reproducing tactile stimuli and
recreating the final sensation of touching a physical object. The present research mainly investigates
cutaneous feedback devices. In literature, two main
approaches can be identified to reproduce tactile sensations: mechanical and electrical stimulations (Figure 1). Several mechanical devices use a matrix of
moving piezoelectric actuators to generate mechanical skin deformation (Pasquero et al., 2003; Allerkamp
et al., 2007). They showed promising results in terms
of displaying the shape of an object, reproducing
the contact-non-contact transition, the manipulation
of an object among fingers. On the contrary they
have a low spatial resolution and a limited temporal
frequency range and they are less portable and comfortable to use. For these reasons they are not able to
realistically reproduce material roughness and surface
finishing properties. Another class of tactile displays
uses acoustic and ultrasound waves to stimulate skin
surface (Nara et al., 2001). Main problems concern
the use of a slide as a contact surface and the lack of
a clear relationship between the parameters modulating the acoustic waves and the ones characterizing the
real surface.
Electrical devices use current flow or electric
field producing an electric-electrostatic stimulation
654
Figure 1. Some examples of tactile displays.
(Kajimoto et al., 2004; Yamamoto et al., 2006). In

Smart-Touch the tactile display is an array of linear electrodes in contact with the fingertip skin.
The device stimulates different receptor responses by
applying electrical current pulses. Test results demonstrated the limitations in the discrimination among
different textures and in the correlation with the provided sensation of real materials. The electrotactile
display proposed by Yamamoto et al. (2006) showed
79% correct answers in texture discrimination but
the generation of a voltage waveform to represent a
specific texture resulted time consuming.
By comparing the two approaches, it is possible to
infer that electrotactile displays present some advantages when compared to the mechanical ones: simpler
actuator structures, lower power consumption, absence
of any temporal frequency band limitation, less expensive and wearable. However the analysis of electrical
approach-based displays shows some implementation
problems: high latency of the system and difficulty
to stimulate the receptors lying in deeper skin regions
(Kajimoto et al., 2003).
The reported analysis of perceptual mechanisms
points out the importance to adopt a selective stimulation approach by merging mechanical and electric
stimuli. Kuroki et al. (2007) developed a display
where a pin in contact with the fingertip carries out
a pulse current and is moved vertically by linear actuators. SAI, FAI and FAII units are actually stimulated.
The problem regards with the mechanical vibration
that affects also the threshold of the electrical stimulation used, decreasing the sensation of fine-coarse
texture. One of the main problems of selective stimulation concerns the nature of the adopted signals,
that are often impulse trains, designed according to
mechanoreceptors characteristics but that overlooks
real materials properties. As a consequence, research
Figure 2. The concept of the tactile display.
should address new methods of materials elaboration

to develop effective signals.
3 THE TACTILE DISPLAY
Two main issues emerge from related works analysis:
The adopted stimulating signals should derive not
only from the intrinsic characteristics of mechanoreceptors but also from the elaboration of real material properties in order to create a correspondence
between the achieved perception and the resulted
sensation.
The use of a selective stimulation approach combining both electrical and mechanical stimulations can
properly elicit different receptors located at different skin regions. This is confirmed by experiments
carried out by Konyo et al. (2007). They demonstrated that a selective stimulation allows to achieve
a complex impression derived from the integration
of different elementary sensations.
Both issues have been tackled: the first by the definition and application of a signal processing method
based on materials elaboration and the second by
the design of a device enabling both mechanical and
electric stimulation (Figure 2).
3.1 The selective stimulation method based on
material signal processing
Five sample materials have been selected to generate the tactile signals on the basis of real material
data. They are metal, paperboard, soft rubber, textile
and wood. Metal is a low carbon steel for cold forming, paperboard is taken from a common paperboard
655
packaging, rubber is a soft rubber made of Technogel, textile is a high resistance fabric with no water
absorption, wood is a wooden chipboard piece of furniture. For each material, a 20 20 mm square area was
realized. An optical scanning system (CHRocodile E
by Precitec) was used to obtain the sampled surface
profiles (z dimension). Two optical probes (RB 200
050 and RB 200 031) having a different measurement range (from 3 mm to 300 m) have been adopted.
They have different sensitive thresholds to the material reflective properties. The sampling frequency was
regulated according to the material typology (1 kHz
or 2 kHz). The profiles obtained depict amplitude in
z direction (where xy plane is parallel to the material
surface plane) versus acquisition points (1000 sample
points). As the material sampling frequency is given,
it is immediate to draw the amplitude versus time. The
Fast Fourier Transform (FFT) of this signal is then
computed. A Butterworth low-pass filter cleans the
signal from the high frequency components in order to
electrically stimulate FAI and SAI mechanoreceptors
that are sensitive to low frequencies. High frequency
components (>200 Hz) are used to setup the value of
the mechanical vibration to elicit FAII units. A proper
amplitude threshold is applied to extract the most significant contributions in frequency. As a result, for
each material class a set of base frequencies and the
corresponding amplitudes A() have been identified.
A synthesized signal S(t) is reconstructed considering
all base contributions as sinusoidal components of a
Fourier series as suggested by (1):
where c is a regulating factor determining the signal maximum amplitude, n are the base frequencies,
An are the corresponding amplitudes, k is the number of the identified base frequencies. In order to have
a simple function, the amplitude threshold is properly selected for each material class to have k = 6.
The c coefficient takes into account two factors: the
user finger impedance, determined a priori in the
device calibration phase (see 5), and the pressure
sensor feedback. S(t) is a simplification of the original scanning signal, which is more suitable to be used
as current signal stimulus due to its dynamics. The
last elaboration concerns the conversion of S(t) into
a current-based signal. The biggest absolute current
value the tactile device developed can drive is 15 mA.
Using a sample and hold algorithm, S(t) are sampled
to give Sd (t). The sampling parameters depend on the
adopted electronics. For instance, the time interval for
sampling is influenced by the switch speed. Comparing the Sd (t) of each material, the highest amplitude
value corresponds to the highest current value, at a
given fingertip impedance and at a maximum allowed
pressure value. The remaining amplitude values of
Sd (t) are proportionally converted in current values
in the range [15 mA; 15 mA]. The result is a current
Figure 3. The main module of the tactile display (a) and the
multiprocessor platform schema (b).
signal that is carried out by the electrodes of the tactile

display.
3.2 The design of the tactile display
The proposed tactile device implementing the selective
stimulation method is composed of three main parts:
a tactile unit, that provides the electro-tactile stimulation. It is made up of a board filled with electrodes that are in contact with the fingertip and the
hardware components driving it (Fig. 3a);
two force sensors evaluating the normal force
applied by the users fingertip. The measured force
is used to calibrate the electrical signal amplitude at
run time;
a shaker implementing the mechanical stimulation
working at the side of the electrical one.
256 pin-electrodes are arranged as a 32 8 grid
to spatially distribute current flow and to better support finger lateral movement. This dimension is a
multiplier of an 8 8 base module. The number of
electrodes is defined to allow the free movement of
the fingertip during surface exploration and hence
to enhance fine texture perception. The pin diameter
is 1.0 mm while the distance between the centers is
2.5 mm. The electrodes are employed as points of stimulation. Electric current flows from a pin electrode to
the surrounding ones stimulating mechanoreceptors.
Two low profile force sensors (FSS-SMT Series of
Honeywell) are positioned under the board to acquire
the force amplitude applied by users during surface
exploration. They are placed at the extremes of the
board. This allows a symmetrical distribution of the
force applied. The pressure contribution is computed
by varying the regulating factor c of equation (1) to
allow the modulation of the electrical signal amplitude at runtime. As pressure falls, a decreasing signal
energy is used. It also avoids fingertip overheating. A
shaker (SmartShaker by PCB group) is used to generate a mechanical vibration. This device integrates
the necessary solid-state power amplifier, involving
compactness, weight reduction and simplification of
driving software. A multiprocessor platform drives the
board. It consists of one master and four slaves, whose
number actually depends on the electrode matrix size.
The system is configured in a scalable way therefore
a matrix size increase simply entails a proportionate
656
4 TACTILE UNIT IMPLEMENTATION
Figure 4. Schema of the main blocks of the tactile unit.
growing of the number of the slave modules (Fig. 3b).

The master interfaces the tactile device with the PC
using a serial communication and it coordinates the
activities of the slaves. Each slave unit implements the
signals driving the relative electrodes.
The HW architecture is illustrated in Fig. 4. Each
slave (named Shift Register Control) controls a bench
of 2 8 Shift Registers driving an array of 8 8
Mosfets. They are used to implement the state of each
electrode (i.e. ground, insulate, signal). The Slave, the
sixteen Shift Registers and the sixty-four Mosfets constitute the so-called Slave Board. To drive the whole
board four Slave Board have been implemented. The
Supervisor is the microcontroller that is used as a
master. It carries out the following tasks: to communicate with the PC via a serial bus, to supervise the
slave actions and to regulate the output of the Current
Mirrors.
Acoustic cues are added to improve the sensation
of physically touching virtual objects whose perceptual effect has been demonstrated by several research
works (Kyung et al., 2006). In a semi-anechoic chamber, a microphone (a free-field microphone by Larson
Davis, Model 2541) is used to record the emitted
sounds by a finger moving on the different material
samples. The adopted velocity was maintained constant at 3 m/min. The recorded sound can be played
in the headphones during the tactile simulation of the
corresponding material. Wireless headphones (K930
by AKG) are used to let users freely move in the
workspace. At this implementation stage, the variability of the fingers velocity and the three dimensionality
of the auditory experience have not been taken into
account.
A tactile simulation for the virtual explored surface
is defined in a given instant by setting the state of all
the board electrodes. A single byte stores all necessary
information to determine the electrode state and if
needed the level of current that has to be carried out.
In a certain time interval ( t), the explored surface is
described by a data package. The package is composed
by a number of bytes equal at least to the number of
electrodes. It is received and elaborated by the master
and then distributed to the slaves for actuation. During
the simulation the package is created on PC by a custom application, developed in CVI/LabWindows 9.0
by National Instruments.
In order to reduce time to develop the tactile unit, two

simplifications have been introduced: the setting of the
finger exploration velocity at 0.0 m/s and the integration of a single current mirror. Due to the first one, the
finger movement is simulated by continuously shifting the signal across electrodes. The shift time is set
to enable the simulation of a 3 m/min finger velocity
that is similar to the one effectively adopted during
the lateral movement along the board. Due to the second one, two distinct electrodes cannot have different
current values at the same time. If multiple current
mirrors are implemented, the finger will be stimulated
by the synthesized signal spatially distributed among
the electrodes according to the amplitude values. In
this case, three consecutive electrodes carry out three
different values of current at the same time. In order
to compensate the use of only one current mirror, the
state of the three electrodes are kept constant for a certain interval and then shifted in the succeeding T in
order to simulate the finger sliding on the material surface. An additional scan velocity has been introduced
to simulate three different current values during the
time. It aims at compensating the adoption of a single current mirror, assuming that the finger acts as an
integrator in T. The two adopted scan velocities and
the fingertip integration should be sufficient to reduce
drawbacks connected to the assumed simplifications.
Moreover, this second scan velocity avoids a too high
stimulation intensity. In a certain interval T1 of T
( T1 T), the mirror acts on the first electrode by
setting its current value and the other two are settled to
ground electrodes. In the next T2 the mirror switches
to the second electrode while the other two turn into
ground ones. At the remaining T3 the mirror acts on
the third electrode and the others are connected to the
ground. This switching process is looped for the whole
T interval that has been set to 100 ms. The intervals T1 , T2 and T3 are set equal at 500 s. This
value depends on the velocity of the switches (HW
component) and the control SW latency.
By periodically changing the current values of the
selected electrode according to the synthesized signals, the system is able to produce an electrotactile
stimulation at any point and in any distribution. The
main advantages of the introduced simplifications are:
reduction in SW complexity (i.e. master-slave communication and signal synchronization are easier) and
absence of cross currents between source electrodes.
User safety will be implemented in the HW logic
(Zener diodes), independently from control software,
in order to assure energy interruption in case of danger.
Figure 5 shows the implemented tactile unit.
4.1 The control software
A custom application has been developed in CVI/LabWindows 9.0 to automate the following functionalities:
To initialize the different tactile signals;
To calibrate the signal amplitudes for tactile stimulation according to user fingertip sensitivity;
657
configure the current distribution along the neighboring electrodes. The below figure shows a source pin
(red) with all surrounding pins connected to ground
(blue). White pins are insulated electrodes. It has not
been demonstrated yet if the electrodes configuration influences the perceived sensation. This is the
main reason for the development of a SW module to
eventually change the default one.
PRELIMINARY EXPERIMENTATIONS
Experimentations have been carried out on the preliminary prototype of the tactile device. They had two
main objectives:
Figure 5. The tactile display prototype.
To verify the perceived tactile sensations;

To calibrate the c parameter of equation (1) in order
to improve the perception of one single user;
To point out the main improvements to carry out
in terms of adopted signals and technological solutions.
Figure 6. The software user interface.
To direct the data packages to the master-slave

platform on the basis of the material class;
To receive the pressure sensor feedback to adjust the
signals amplitude at runtime;
To regulate the tactile data packages by considering
user finger position during surface exploration;
To pilot the vibration frequency for the shaker;
To monitor the simulation conditions and progress.
To allow the interfacing with a turnkey platform
able to provide both graphic and sound rendering
to enhance surface experience.
A graphic interface has been developed to manage
and control the tactile unit. It consists of three sections:
one containing the selected material information, one
to set the terminal status and the photogram sequence
and the last to control the execution (Fig. 6). In particular in the second section a graphic tool allows to
At this preliminary stage, the undertaken testing

activities have mainly supported the definition of the
optimal parameters for roughness/texture perception
and the validation of adopted approach. The testing
session was carried out in three phases: preliminary
training, experimental tests, and data analysis. The
methodological approach used to schedule experimentations refers to well known psychophysical methods
proposed by Ehrenstein et al. (1999). The experimental
tests stage is divided in two steps: Signal Detection and
Dissimilarity Rating (Yoshioka et al. 2007). During the
second step the NullAdaptation technique proposed by
Hollins et al. (2006) is used to allow the tactile memory
deletion between two subsequent stimuli.
In the preliminary training phase, users were trained
to use the tactile interface. They had to keep the index
finger of their dominant hand on the display and the
contact area roughly constant. During the second phase
two tests were performed:
1) Signal Detection: it aims to detect the tactile stimulus threshold perceived by the user. The method of
constant stimuli was used, as the most appropriate and
safe due to the nature of the electrical signal. It consists
of the following steps:
The electric signal is increased step by step until the
user perceives a sensation;
The user responds to the perception of the stimulus
in terms of YES-NO;
The results are averaged over all users to infer the
average threshold.
2) Dissimilarity Ratings: it aims at understanding if
users differently perceive the generated tactile stimuli
for each material. The user is submitted to four different pairs of virtual stimuli presented in a random
sequence. He/she has to say if the stimulus is the same
or not (Ex. A B, B C, B B, A B, A A, A C).
Between two pairs, the user is asked to touch a perfectly smooth material (i.e. a glass) to implement the
658
null adaptation technique (to unload the tactile memory). Data are elaborated to verify the correctness of
user responses.
Data analysis stage regards with the collection of
testing results and with their elaboration.
The experimental set-up consists of the preliminary
tactile device prototype without any graphic rendering
of the displayed material. A single configuration of
current distribution has been considered: all electrodes
neighboring the one that carries out the signal are set to
ground. The shaker frequency has been set to 230 Hz,
assuming a constant mechanical frequency independent from the simulated material. Experimentations
have been carried out with and without the integration
of sound rendering.
Ten sample users have been involved in the tests.
They differ for age and gender distribution. They are
2336 years old, 5 male and 5 female. They did not
suffer from particular cutaneous sensitivity problems.
Users were asked to place their dominant hand on
the tactile display and to keep the finger in the same
position without applying torque movements. In this
condition, the scan velocity is set to 0 m/s and the
finger impedance can be calculated only once.
Each user was first submitted to the signal detection
experiment. A square wave at 30 Hz in the frequency
ranges of both SAI and FAI units was used as electric
signal. Its amplitude had been varied for subsequent
increases till the user detected it. The identified amplitude represents the user sensitivity threshold to the
electric signal. It was used to calibrate signal amplitudes during the second phase of the experimentation.
The impedance contribution and the signal detection
were used to carry out a proper amplitude shift of the
electric signals that are obtained from material surface properties processing. The detected thresholds are
reported in Table 1.
The next experimental stage regards with the dissimilarity ratings. Users were submitted to four different pairs of stimuli in a random order according
to relative material pairs: metal-wood, metal-metal,
wood-metal, wood-wood. At this preliminary stage,
these materials are chosen as they are characterized
by very different surface textures. Users were asked
to touch a glass surface between two subsequent pairs
and to repeat the test for five trials. The results for
a single trial are shown Table 2. The last row gives
an account of the correctness of signal discrimination
responses in percentage. The achieved mean is 65%.
The same discrimination ratings test was repeated by
adding acoustic cues during surface exploration. In
this case the mean value reaches the 75% of success.
5.1
Results discussion
The Signal Detection phase clearly shows that the

Detection Thresholds are quite smaller for women
than for men and then that the fingertip sensitivity
is greater. A similar trend is found in the user age:
young users seem to have a lower threshold than old
ones. The discrimination of different material signals
Table 1. Achieved detection threshold.

Signal detection
User
Gender
Age
Detection Threshold (mA)
1
2
3
4
5
6
7
8
9
10
Mean V.
M
M
M
M
M
F
F
F
F
F
23
26
30
30
35
28
29
33
35
36
2,9
3,2
3,6
3,4
3,5
2,1
2,5
2,8
2,7
2,8
2,95
Table 2. Achieved dissimilarity ratings

Signal Discrimination
User
M-W
M-M
W-M
W-W
USD
1
2
3
4
5
6
7
8
9
10
Mean V.
Y
Y
N
Y
Y
Y
Y
N
Y
Y
80%
N
Y
Y
Y
Y
Y
Y
N
Y
N
30%
Y
Y
N
N
N
Y
Y
Y
Y
N
60%
N
N
Y
Y
Y
N
N
N
Y
Y
50%
50%
75%
50%
75%
75%
75%
75%
25%
100%
50%
65%
*Users successful discrimination
is simpler than the same signal presented twice. In

particular, users are able to better recognize the WoodWood pair than the Metal-Metal pair. This is due to
the nature of relative materials and then of stimulating signals. Experiments actually show that users
found easier to identify those materials characterized
by a greater roughness value (Ra) than materials by
a lower one (Rawood = 8,95 m Rametal = 2,34 m)
and to recognize an anisotropic material (wood) than
an isotropic one (metal). This means that the developed display is quite effective to simulate anisotropic
and highly rough materials (i.e. laminated wood, plastic sheet, textile, etc.) while it must be improved in
case of glass, metal and rubber. Finally adding acoustic cues to the experience can significantly enhance
the material discrimination ability, increasing positive responses from 65% to 75%, confirming other
experimental results carried out by Kyung et al. (2006).
CONCLUSION
A novel tactile display to enhance product experience

into virtual reality environments is presented. It represents an innovation if compared with current research
659
achievements. It actually adopts a selective stimulation

method based on real materials properties processing
and elaboration according to skin mechanoreceptors
characteristics. Preliminary experimentations show
promising results in terms of materials discrimination and systems latency. However, additional testing
sessions are needed to validate the adopted approach
and technical solutions. Numerous parameters must
be combined to verify the system performance and to
identify an optimal setting. Next steps will be focused
both on texture properties perception to correlate the
user judgment about the simulated material and its
physical properties, on the collection of further data
and finally on the integration of visual cues during VP
interaction.
REFERENCES
Allerkamp, D., Bottcher, G., Wolter, F.H., Brady, A.C.,
Qu, J. & Summers, I.R. A vibrotactile approach to tactile
rendering. The Visual Computer, 23: 97108.
Bensmaia, S.J., Hollins, M. & Yau, J. 2005. Vibrotactile frequency and intensity information in the Pacinian system:
a psychophysical model. Perception & Psychophysics, 67:
828841
Bordegoni, M., Ferrise, F., Covarrubias, M. & Antolini, M.
2010. Haptic and sound interface for shape rendering. Presence: Teleoperators and Virtual Environments,
19(4):341363.
Cascio, C. & Sathian, K. 2001. Temporal cues contribute
to tactile perception of roughness. J.Neuroscience, 21:
52895296.
Ehrenstein, W.H. & Ehrenstein, A. 1999. Psychophysical
methods. In Windhorst and Johansson (eds), Modern
Techniques in Neuroscience Research. Springer, Berlin,
12111241.
Hollins, M., Bensmaia, S.J. & Roy, E.A. 2002. Vibrotaction
and texture perception. Behavioral Brain Research, 135
(1): 5156.
Johansson, R.S., Landstrm, U. & Lundstrm, R. 1982.
Responses of mechanoreceptive afferent units in the
glabrous skin of the human hand to sinusoidal skin
displacements. Brain research, 244 (1): 17-25
Kajimoto, H., Kawakami, N. & Tachi, S. 2003. Psychophysical evaluation of receptor selectivity in electro-tactile
display. Proc. of ISMCR, Madrid, Spain.
Kajimoto, H., Kawakami, N., Tachi, S. & Inami, M. 2004.
SmartTouch: Electric Skin to Touch the Untouchable.
IEEE Computer Graphics and Applications, 24 (1):
3643.
Knibestol, M. 1975. Stimulus response functions of slowly
adapting mechanoreceptors in the human glabrous skin
area. J. Physiol., 245: 6380.
Konyo, M., Tadokoro, S., Yoshida, A. & Saiwaki, N. 2005. A
tactile synthesis method using multiple frequency vibrations for representing virtual touch. Intelligent Robots
and Systems, 2005. (IROS 2005). 2005 IEEE/RSJ International Conference, 39653971.
Kuroki, S., Kajimoto, H., Ni, H., Kawakami, N., Tachi, S.
2007. Proposal for Tactile Sense Presentation that
Combines Electrical and Mechanical Stimulus. 2nd Joint
Eurohaptics Conference and Symposium on Haptic Interfaces for Virtual Environment and Teleoperator Systems
(World Haptics), Tsukuba, Japan.
Kyung, K.U. & Kwon, D.S. 2006. Multi-sensory Perception

of Roughness: Empirical Study on Effects of Vibrotactile
Feedback and Auditory Feedback in Texture Perception.
Lecture Notes in Computer Science, 4282: 406415, 2006.
M. Hollins, F. Lorenz, D. Harper, Somatosensory coding of
roughness: The effect of texture adaptation in direct and
indirect touch. Journal of Neuroscience, 26, pages 5582
5588, 2006.
McMahan, W., Romano, J.M., Rahuman A.M.A. &
Kuchenbecher, K.J. 2010. High Frequency Acceleration
Feedback Significantly Increases the Realism of Haptically Rendered Textured Surfaces. Proc. IEEE Haptics
Symposium 2526 March, Waltham, Massachusetts, USA.
Mengoni, M., Colaiocco, B., Germani, M. & Peruzzini, B.
2010. Design of a novel Human-Computer Interface to
support HCD application, Proc. International Design
Engineering Technical Conferences & Computers and
Information in Engineering Conference (ASME IDECTCIE 2010), Montreal, Canada.
Mengoni, M., Germani, M. & Peruzzini, M. 2009. Products
experience: how canVirtual Prototyping improve usability
testing?. In P.J. Bartolo et al. (eds), Innovative developments in design and manufacturing Advanced Research
in Virtual and Rapid Prototyping, Taylor & Francis
Muniak, M.A., Ray, S., Hsiao, S.S., Dammann, F.J. &
Bensmaia, S.J. 2007. The neural coding of stimulus
intensity: Linking the population response of mechanoreceptive afferents with psychophysical behavior. J. Neuroscience, 27 (43): 11.687-11.699.
Nara, T., Takasaki, M., Maeda, T., Higuchi, T., Ando, S. &
Tachi, S. 2001. Surface Acoustic Wave Tactile Display. IEEE Computer Graphics and Applications, 21 (6):
5663.
Pasquero, J. & Hayward, V. 2003. STReSS: A Practical Tactile
Display System with One Millimeter Spatial Resolution
and 700 Hz Refresh Rate. Proc. of Eurohaptics 2003.
Dublin, Ireland.
Schifferstein. N.H.Y. 2006. The perceived importance of
sensory modalities in product usage: a study of selfreports. In Acta Psychologica, 121: 4164
Shinoda, H., Iwamoto, T., Asamura, N. & Makino, Y. 2004.
Two Methods of Hi-Fidelity Cutaneous Display: MultiPrimitive Stimulation and Stress Reproducing. Proc. of
IEEE/RSJ International Conference on Intelligent Robots
and Systems, Sendai.
Solazzi, M., Frisoli, A. & Bergamasco, M. 2010. Design of a
Novel Finger Haptic Interface for Contact and Orientation
Display. Proc. IEEE Haptics Symposium, 2526 March,
Waltham, Massachusetts, USA.
Yamamoto, A., Nagasawa, S., Yamamoto, H. & Higuchi, T.
2006. Electrostatic Tactile Display with Thin Film Slider
and Its Application to Tactile Telepresentation Systems.
IEEE Transactions onVisualization and Computer Graphics, 12 (2): 168177.
Yoshioka, T., Bensmaia, S.J., Craig, J.C. & Hsiao. S.S. 2007.
Texture perception through direct and indirect touch:
An analysis of perceptual space for tactile textures in
two modes of exploration. Somatosensory and Motor
Research, 24, pages 5370, 2007.
660

Virtual Reality technology as a support to the maintenance activity

of buildings
A. Zita Sampaio & A. Rita Gomes
TU Lisbon, Department of Civil Engineering and Architecture, Lisbon, Portugal
ABSTRACT: A Virtual Reality model was created in order to help in the maintenance of exterior closures in
a building. It allows the visual and interactive transmission of information related to the physical behavior of
the elements, defined as a function of the time variable. To this end, the basic knowledge of material most often
used in walls, anomaly surveillance, techniques of rehabilitation, and inspection planning were studied. This
information was included in a data base that supports the periodic inspection needed in a program of preventive
maintenance. The results are obtained interactively and visualized in the virtual environment itself. This work
brings an innovative contribution to the field of maintenance supported by emergent technology.
1
INTRODUCTION
The main aim of a research project, now in progress

at the Department of Civil Engineering of the Technical University of Lisbon, is to develop virtual models
as tools to support decision-making in the planning
of construction management and maintenance. A first
prototype concerning the lighting system has already
been completed (Sampaio et al. 2009). A second prototype concerns the maintenance of the closure of
interior walls. This paper describes the VR of faades
maintenance (Gomes, 2010).
The interactive model integrates Virtual Reality
(VR) technology, the EON system (EON, 2010), and
an application implemented in Visual Basic (VB)
language. The model allows interaction with the 3D
geometric model of a building, visualizing components for each construction. It is linked to a database
of the corresponding technical information concerned
with the maintenance of the materials used as exterior finishes. The principal objective of the interactive
VR prototype is to support decision-making in the
maintenance domain.
The present project aims to bring important contributions to this domain, through the implementation of
virtual models able to relating the behaviour of materials, their characteristics, anomalies and repair work
to each other. During this work the basic knowledge
of the topics involved, such as aspects related to the
materials, the techniques of rehabilitation and conservation and the planning of maintenance is outlined and
discussed in addition, methods of interconnecting this
knowledge with the virtual model are explored.
The prototype for walls was trialled in a concrete
project. This kind of building element has a continuous
lifestyle, so requires the study of preventive maintenance (the planning of periodical local inspections)
and of corrective maintenance with repair activity
analysis. The model facilitates the visual and interactive access to results, supporting the definition of
inspection reports. It will be possible in the future to
apply this to other building situations whether in new
constructions or those needing rehabilitation.
2
INTERACTIVE AND COLLABORATIVE

MODELS
Virtual Reality technology can support the management of data that is normally generated and
transformed or replaced throughout the lifecycle of
a building. This technology becomes an important
support in the management of buildings allowing
interaction and data visualization. At present, the management of building planning can be presented in a 3D
form and various materials can be assigned to the fixtures and furnishing enabling the user to be placed in
the virtual building and view it from inside as well as
outside. This study contemplates the incorporation of
the 4th dimension, that is, time, into the concept of
visualization. The focus of the work is on travelling
through time: the ability to view a product or its components at different points in time throughout their life.
It is envisaged that the incorporation of the time dimension into 3D visualization will enable the designer/user
to make more objective decisions about the choice of
the constituent components of the building. In maintenance the time variable is related to the progressive
deterioration of the materials throughout the buildings
lifecycle.
The present prototype incorporates interactive techniques and input devices to perform visual exploration
tasks. To support this system a data base was created
which included a bibliographic research support made
in regard to the closure materials used in the interior
and exterior walls of a building, anomalies concerning
661
Figure 1. 3D model of the house: coating elements of

exterior and interior walls.
different kinds of covering material, and corrective

maintenance. Repair activities were also studied. The
programming skills of those involved in the project
had to be enhanced so that they could achieve the integration of the different kinds of data bases needed in
the creation of the interactive model.
The 3D model linked to a data base concerning
maintenance produces a collaborative virtual environment, that is, one that can be manipulated by partners
interested in creating, transforming and analysing
data in order to obtain results and to make decisions. For example, inspection reports can be defined
and consulted by different collaborators. The process of developing the prototype interface considers
these purposes. The developed prototype associates
the characteristics of the coating component of the
exterior walls to activities concerning the maintenance
of buildings (Fig. 1).
3
FAADES MAINTENANCE PROTOTYPE
Faade coatings play an important role in the durability of buildings, since they constitute the exterior
layer that ensures the protection of the wall against the
aggressive actions of physical, chemical or biological
nature.
Naturally, they should also give the faade the
required decorative effect. Since this building component is exposed to bad atmospheric conditions
it frequently shows an evident degree of deterioration, requiring maintenance interventions. To perform
maintenance activities a survey of failures in the building must be conducted in order to arrive at the best
solution for repair and maintenance.
In order to better understand the operation of faade
coating, bibliographic research of materials usually
applied to faade coatings was carried out and a table of
characteristics of these was drawn up. Subsequently, a
survey was made of anomalies, probable causes, solutions and methods of repair for each of the coatings
studied.
Figure 2. Different type of materials applied as faade

coatings.
The visualization of the maintenance data of a building and the impact of time on the performance of
these exterior closure materials require an understanding of their characteristics (Gomes & Pinto, 2009)
(Fig. 2):
Types of material: painted surfaces, natural stone
panels and ceramic wall tiles;
Application processes: stones (panel, support
devices, adherent products, ); ceramic tiles (fixing mechanism, procedures, ); painted surfaces
(types of paint products, prime and paint scheme
surface, exterior emulsion paints, application processes);
Anomalies: dust and dirt, lasting lotus leaf effect,
covering power, insufficient resistance to air permeability or weatherproof isolation, damaged stones or
ceramic tiles, alkali and smear effect, efflorescence,
fractures and fissures ;
Repair works: surface cleaning, wire truss reinforcing, cleaning and pointing of stonework joints,
removing and replacement of ceramic wall tiles,
removing damaged paint and paint surface, preparing and refinishing stone panels
662
Table 1.
Anomaly
Example of anomalies and the associated repair solution.

Specification of
the anomaly
Repair solution
Repair methodology
Fall in areas with

deterioration
of support
Replacement of the
coat (with use of a repair
stand as necessary)
1o Removal of the tiles by cutting grinder

with the aid of a hammer and chisel;
2o Timely repair of the support in areas
where the detachment includes material
constituent with it;
o
3 Digitizing layer of settlement;
4o Re-settlement layer and the tiles.
Failure of the
support (wide
cracks with
well-defined
orientation)
Replacement of the
coat (with repair of
cracks in the support)
1o Removal of the tiles by cutting grinder;

2o Removal of material adjustment in the
environment and along the joint;
3o Repair of cracks, clogging with
adhesive material (mastic);
4o Settlement layer made with cement in
two layers interspersed with glass fibre;
5o Re-settlement layer and the tiles.
Figure 3. The main interface of the interactive application.
3.1
Characteristics of the materials used in

faades
Depending on the role that the faade coatings play on

the wall as a whole they can be classified as finishing,
sealing or thermal insulation. The most frequent materials used as coating finishes are painting, tiling and,
as sealing coating of the natural stone:
Paint coating contributes to the aesthetic quality of the building and its environment and also
protects the surface of the exterior wall against
corrosion, deterioration and penetration of aggressive agents (Lopes, 2008). In order to obtain a
good performance as an exterior coating, several aspects must be considered, such as covering
power and resistance to water, sunlight, chemical products and to the development of micro
organisms;
The ceramic coating consists essentially of tiling

panels, cement and adhesive and the joints between
the slabs. The application of ceramic tiling to building faades has considerable advantages particularly as some degree of waterproofing is afforded by
the glazed surface along with a great resistance to
acids, alkalis and vapour. Other advantages are that
it does not need painting and that it can be easily
applied or substituted during repair work (Ferreira
et al. 2009);
The use of natural stone in the coating of faade
surfaces is a good solution both technically and aesthetically. The stone coating is composed of slabs of
stone attached to the wall by a support system. The
principal characteristics of the stones are: reduced
water absorption, sufficient mechanical resistance
to bending and impact, abrasion and shearing parallel to the face of the slabs (Veiga & Malanho, 2009).
663
3.2 The database

The most frequent anomalies that occur in the coated
faades were analysed in order to create a data base
linked to the virtual model that could support planning
of inspections and maintenance strategies in buildings.
The database contains the identification of anomalies that can be found in each type of material used
in faades and the corresponding probable cause. For
each kind of anomaly the most adequate repair solutions are also selected and included in the data base.
The following example concerning deficiencies in tiles
presents the methodology implemented in this virtual
application (Table 1).
The characteristics related to anomalies, causes,
repair solutions and rehabilitation tasks were included
in a database of each type of material and linked to
the 3D model of the building. Thus, the virtual model
gives users the ability to transmit, visually and interactively, information related to the closure properties
of exterior walls, allows them to analyse the anomalies
observed in an inspection of the real building and to
predict the corresponding repair work. The 3D virtual
model can be seen, therefore, as an important tool for
anomaly surveillance in structures and for supporting decision-making based on the visual analysis of
alternative repair solutions.
3.3 The interface

The implementation of the prototype system makes use
of graphical software programming, Visual Basic 6.0
Microsoft, software to establish an adequate database,
Microsoft office access, graphical drawing system,
AutoCAD Autodesk and VR technology based software, EON Studio.
Human perceptual and cognitive capabilities were
taken into account when designing this visualization
tool so the model is easy to use and does not require
sophisticated computer skills: many potential users are
not computer experts. It uses an interactive 3D visualization system based on the selection of elements
directly within the virtual 3D world.
Furthermore, associated with each component,
there are integrated databases, allowing the consultation of the required data at any point in time. The
interface is composed of a display window allowing
users to interact with the virtual model, and a set
of buttons for inputting data and displaying results
(Fig. 3).
For each new building to be monitored the characteristics of the environment (exposure to rain and sea)
and the identification of each element of the faades
must be defined. The data associated to each element
are the building orientation, the type of exterior wall
(double or single), and the area and type of coating.
Once each monitored element has been characterized, several inspection reports can be defined and
recorded and thereafter consulted when needed. An
inspection sheet is accessed by the main interface
(Fig. 4).
Figure 4. Inspection sheet interface.
Using the drop-down menus allowed by the interface, the user can associate the characteristics of the
observed anomaly to a faade element: the type of
anomaly, the specification, details and the probable
cause of the anomaly, an adequate repair solution and
pictures taken in the building. After completing all
fields relating to an anomaly, the user can present the
report as a pdf file.
664
Figure 5. Steps of the geometric modeling process.
Figure 7. Inspection sheet report.
All coatings studied were considered in this casestudy. Thus it was assumed that the main faade is
covered with tile and the remaining faades are painted
while hall faades are of natural stone. Figure 6 shows
how to identify a faade in the virtual model of the
building.
The figure 7 includes the inspection report of the
anomaly considered in Table 1.
The developed software is easy to handle and transport for on-site inspections and comprises information
of the causes, solutions and methods for repairing.
Figure 6. Identification of a faade element.
3.4 The case study

First, the 3D geometric model of a building case was
created. The building consists of a ground-floor, a 1st
floor and an attic with dwelling space shown. The
coating elements of the walls were then modelled as
independent geometric objects (Fig. 5). In this way,
each element can then support characterization data of
the applied material and different kinds of information
related to maintenance.
CONCLUSIONS
A VR model to support the maintenance of faades in

a building was developed within a research project.
It enables the visual and interactive transmission of
information related to the physical behaviour of the
elements, defined as a function of the times variable.
The model shows the characteristics of each element of
the building in the model and the information related
to inspection, anomalies and repair works. As the 3D
model is linked to a data base in an interactive environment and has a friendly interface to deal with this
knowledge, it allows a collaborative system.
With this application the user may fully interact with
the program referring to the virtual model at any stage
of the maintenance process and analyse the best solution for repair work. It can also support the planning
of maintenance strategies.
665
ACKNOWLEDGMENTS
The authors gratefully acknowledge the financial support of the Foundation for Science and Technology,
a Governmental Organization for the research project
PTDC/ECM/67748/2006, Virtual Reality Technology
applied as a support tool to the planning of construction maintenance, now in progress.
REFERENCES
Ferreira, L., Coroado, J., Freitas, V. & Maguregui, I. 2009.
Causes of the fall of tiles applied in faades, Patorreb, 3th
Meeting of Pathologies and Rehabilitation of buildings,
Proc. intern. symp., Porto, Portugal, March 1820.
Gomes, A.M. & Pinto, A.P. 2009. Didactic text of construction materials, Technical University of Lisbon, IST,
Lisbon, Portugal.
Gomes, A.R. 2010. Virtual Reality technology applied to the

maintenance of faades, Integrated Master Degree Thesis
in Construction, TU Lisbon, Portugal.
EON 2010. Introduction to working in EON Studio. EON
Reality, Inc.
Lopes, C. 2008. Anomalias da cor em revestimentos por
pintura em paredes exteriores. Integrated Master Degree
Thesis in Construction, TU Lisbon, Portugal.
Sampaio, A.Z., Ferreira, M.M. & Rosrio, D.P. 2009.
Interactive virtual application on building maintenance:
The lighting component, IRF2009, 3rd International
Conference on Integrity, Reliability and Failure: Challenges and Opportunities, Symposium Visualization and
human-Computer Interaction, Proc. intern. symp., Porto,
Portugal, July 2024, abstract pp. 221222, paper 11 pgs.
Veiga, M. & Malanho, S. 2009. Coating of faades in
natural stones. Diagnostic methodology and rehabilitation processes. Patorreb, 3th Meeting of Pathologies and
Rehabilitation of buildings, Proc. intern. symp., Porto,
Portugal, March 1820.
666

Some studies on dislocation density based finite element modeling of

Ultrasonic Consolidation
D. Pal & B.E. Stucker
University of Louisville, Louisville, KY, USA
ABSTRACT: A dislocation density based constitutive model has been developed and implemented into a crystal
plasticity quasi-static finite element framework. This approach captures the statistical evolution of dislocation
structures and grain fragmentation at the bonding interface when boundary conditions pertaining to the Ultrasonic
Consolidation (UC) process are prescribed.
Hardening is incorporated using statistically stored and geometrically necessary dislocation densities (SSDs
and GNDs) which are dislocation analogs of isotropic and kinematic hardening respectively. Since the macroscopic boundary conditions during UC involves cyclic sinusoidal simple shear loading along-with constant
normal pressure, the cross slip mechanism has been included in the evolution equation for SSDs. The inclusion
of cross slip promotes slip irreversibility, dislocation storage and, hence, cyclic hardening during UC. GND is
strain-gradient dependent and thus the model is scale-dependent. The model is calibrated using experimental
data from published refereed literature for simple shear deformation of single crystalline pure aluminum alloy
and uniaxial tension of polycrystalline Aluminum 3003-H18 alloy.
The model also incorporates various local and global effects such as friction, thermal softening, acoustic
softening, surface texture of the sonotrode, mating surface roughness and presence of oxide-scale at the mating
surfaces. Surface oxides contribute significantly to the grain substructure evolution at the interface and to the
anisotropic bulk deformation away from the interface during UC in general. The model results have been predicted
for Al-3003 H-18 alloy undergoing UC. A good agreement between the experimental and simulated results has
been observed for the evolution of linear weld density and anisotropic part strength. Similarly, microscopic
observations seen by multiple researchers, including embrittlement due to grain substructure evolution and
broken oxide layers at the UC interface have also been demonstrated by the simulation.
1
INTRODUCTION
As a direct result of ongoing research efforts in ultrasonic consolidation (UC) worldwide, it has become
apparent that a new approach to modeling of UC bonding is needed. A model which provides a better understanding of the effects of process parameter changes
on grain refinement, plastic deformation and bonding during UC will better enable researchers to predict
which materials will bond, how the mechanical properties of UC-produced parts can be improved, and how
to better design the next generation of UC equipment.
The continuum properties of parts made using UC
are strongly dependent upon the micromechanics of
the bonded interface (Johnson 2008). Interfacial-scale
microstructures can be studied fundamentally using
electron microscopy and can be used to correlate
atomic and mesoscopic mechanisms of deformation
to their continuum counterparts.A dislocation densitybased crystal plasticity finite element model (DDCPFEM) can capture the statistical distribution of dislocations, partials and various deformation mechanisms at
the bonding interface as inputs to predict macroscopic
deformation and mechanical property profiles as a
function of energy input characteristics. These input
characteristics are a function of the process parameters

used in a UC machine, namely vibration amplitude,
normal force, ultrasonic frequency, welding speed,
sonotrode geometry and temperature.
2
PROBLEM FORMULATION
It has been shown that material sheets subjected

to UC undergo inhomogeneous plastic deformation
through their thickness (Johnson 2008). Classical
continuum plasticity theories do not fully explain this
phenomenon (Bate 1999). Therefore, a study of strain
localization and grain refinement at the material interfaces during UC bonding is required. The following
steps lead to the calculation of these localized strains
and their effects.
2.1 Large deformation quasi-static formulation
The deformation map in space and time is described
by the total deformation gradient tensor F (Figure 1).
Applying the Kroner-Lee assumption, F is decomposed into elastic Fe and plastic gradient Fp tensors
using multiplicative operator theory
667
The plastic deformation gradient Fp includes constant volume plastic deformation without disturbance
of the crystal lattice. Elastic distortion and rigid
rotation of the lattice are described by a unique
intermediate configuration free of local stresses.
and the passing stress, pass

, caused by parallel dislocations can be found using:
and the cutting stress, cut

, at 0K caused by forest
dislocations can be found using:
2.2 The non-local dislocation density motivated

material model
The flow response for dislocation density motivated
crystal plasticity modeling in a given slip system
is given by (Ma 2006):
where the pre-exponential variable 0 is the upper limit

of the shear rate for the case where the Boltzmann
factor is equal to 1, which can be found using:
where Qslip is the effective activation energy for

dislocation slip.
The incompatibility in plastic deformation gradient
and non-local geometrical non-linearity is introduced
using GND
which computes the geometrically necessary dislocations required to maintain continuity
throughout the material.The evolution law for GND

is:
Material hardening at an integration point is both a
function of GND
and SSD
(statistically stored disloca
tion density). The evolution laws for SSD
are generally
linear in shear rate (Eqn7).
The various constants used in this modeling can be

attributed to physical phenomena as shown in Table 1.
2.3 Global and local solution strategies
Figure 1. Multiplicative decomposition of the total deformation gradient, F = F e F p .The rotation and stretching of the
lattice are taken into account through the elastic deformation
gradient Fe (Ma et al. 2006).
Table 1.
The global and local solution strategy developed for

solving the crystal plasticity constitutive model in
section 2.2 is shown in Table 2.
First, the applied loading is applied incrementally
as load steps. At the beginning of each load step
(k + 1), various nodal and integration point variables
Physical Interpretation for various constants used in the Constitutive model
Material Parameter
Physical Meaning
Prescribed Value
Qslip
Qbulk
c1
c2
c3
c4
c5
c6
c7
c8
c9
Energy barrier for slip

Energy barrier for climb (activated at higher temperatures)
constant for passing stress (due to in-plane dislocations) (Equation 3)
constant for jump width (Equation 4)
constant for obstacle width (Equation 4)
constant for lock forming rate (Equation 7)
constant for athermal annihilation rate (Equation 7)
constant for dipole forming rate (Equation 7)
constant for thermal annihilation rate (Equation 7)
constant for non-linear climb of edge dislocation (Equation 7)
constant for energy scaling at the interface
3.0 1019 J
2.4 1019 J
0.1
2.0
1.0
1.5 107 m1
10.0
1.0 1030 m1
1 107 m5 c8
0.3
103
668
such as displacement (u), plastic deformation gradient

(Fp ), SSD and GND are initialized to their converged
counterparts obtained in the previous load step. After
initialization, an iterative procedure is used to solve
for the local integration point variables. In every iteration (i), a local integration is performed via the B-Bar
scheme to compute the stress, plastic deformation gradient and the hardness due to immobilized dislocation
content (SSD). The stress is then resolved on a slip
system to obtain the resolved shear stress (RSS,) followed by plastic shear strain rate which further leads
to the calculation of the stiffness matrix (C) at the
integration point followed by a nodal updated tangent
modulus (K) as a function of displacement. The tangent modulus is a measure of local stress with respect
to local strain. This tangent modulus in uniaxial tensile plasticity is lower than the Youngs modulus (E)
since the force required to glide dislocations is less
than the force required for moving perfect atomic
planes. After the calculation of nodal tangent modulus, the internal nodal force based on stresses at
the integration point and the integration algorithm
(B-Bar) is used to calculate the force residual (Q).
Table 2. Solution strategy for solving the crystal plasticity
constitutive model with Nonlinear FEM.
Solving the matrix-inversion problem posed in B (II)

in table 2, an increment in the displacement ( u) is
obtained for the current iteration (i). The increment
in displacement is added to the current displacement
for updating its value for the next iteration (i + 1),
if required. The increment in GND is calculated at
this stage. The procedure is illustrated in Figure 2.
First, the plastic deformation gradient (Fp ) is linearly
extrapolated from local integration points to the corresponding nodes. Next, the nodal value of is averaged
out since the node under consideration is connected to
other elements. The curl ( FpT m0 ) is calculated
and returned to the integration points from their corresponding nodes in step 3. The curl ( FpT m0 ) at
integration points is multiplied by appropriate material variables to obtain GND for screw and edge
dislocations in their normal and tangential directions
respectively. To ensure that the local integration point
has converged, the increments in , SSD and GND
are compared against their respective threshold upper
bounds. This is followed by global convergence, where
the absolute value of the force residual (Q) is compared against a global threshold. Once both the local
and global convergences are ensured, various nodal
and integration point variables are predicted for the
next load step; and the same sequence described here
is repeated for the next load step (k + 1).
2.4
Process boundary conditions
The model formulations and solution strategies in sections 2.1, 2.2 and 2.3 have been validated for simple
boundary conditions such as simple shear and normal
compressive loading (Pal 2011). The model results
such as stress evolution, GND or inhomogeneous
plastic deformation evolution and SSD or homogenous plastic deformation evolution are in good agreement with the experiments (Ma 2006) and (Siddiq
2007) respectively. UC process requires a simultaneous application of the abovementioned boundary
conditions along with local interfacial contact based
boundary conditions. Therefore based on the accuracy
of the model established by validation of the results
in the simple cases, more complex boundary conditions representing the UC process have been applied
to the model and are discussed in this section. The
model predictions for these complex boundary conditions and their comparison with the experiments will
be discussed in section 3.
Figure 2. Procedure to determine the slip-rate gradients

within 3D trilinear elements with B-bar integration. The
interpolation of the curl{i } is performed via an eight-noded
solid element with (2 2 2) integration.
2.4.1 Global process boundary conditions

The schematic of the UC process in Figure 3 illustrates
the global boundary conditions, namely normal compressive force, oscillation amplitude, weld speed and
initial conditions like foil microstructures and surface
roughness at the interface.
The schematic in Figure 3 has been converted
into mathematical boundary conditions as shown in
Figure 4.
The normal compression in UC is around 1200 N
2000 N for Aluminum alloy. The applied shear
amplitude is of the order of 1 m25 m with foil
669
microstructure has been assumed. The bottom foil

and the top foil are comprised of 12 randomly generated crystals each, with the top foil having the same
granular microstructure as the bottom foil.
2.4.2 Interfacial contact boundary conditions
For interfacial contact with friction, the contact boundary conditions can be formulated as follows:
where w and w+ are displacements of the nodes on

the top of the bottom and bottom of the top foils in the
vicinity of the interface respectively. n denotes the normal vector perpendicular to the interface. The normal
vector is in the y-direction for this case.
In addition to equation 8, the normal traction components on the bottom of the top foil and top of
the bottom foil should be compressive and tensile
respectively. Also, from the point of view of force
equilibrium, both of these traction components should
be equal and opposite to each other as shown in
equation 9.
Figure 3. Schematic of the UC process.
Equations 8 and 9 further lead to the Kuhn-Tucker

conditions (Dolbow et al. 2001) for the normal traction
and normal displacement components as demonstrated
in equation (10)
Further, similar to the normal traction components,

the tangential components should be also in equilibrium (equation 11).
Figure 4. Schematic showing Ultrasonic Consolidation
global boundary conditions. The bottom of the model is held
rigid. This assumption is only true for 2 foils on a large plate
or a foil laid over a large plate.
thicknesses between 50 m200 m (a typically used

foil thickness is 150 m) and initial mating surface roughness varying from smooth (Ra 0.1 m)
to heavily surface damaged (Ra 5 m). The mentioned surface conditions are only observed in the top
part of the bottom foil because it has been freshly
deformed by the sonotrode during deposition of the
previous layer. The time period for one oscillation is
about 50 microseconds (corresponding to a sonotrode
frequency of 20 kHz). Since the travel speed of the
sonotrode horn varies from 1050 mm/s, the material
point under UC loading experiences a maximum of
3000 material deformation cycles. Foils are typically
composed of polycrystalline microstructures with random orientation as shown in (Johnson 2008). The
initial SSD has been assumed to be 5 1013 m2
(Pal 2011). The microscopic parameters required for
simulations can be extracted from Table 1. Since, orientation imaging microscopy (OIM) data for the foils
were not available, a random cubic polycrystalline
where PT is the tangential projection operator, given

as:
where denotes the cross product of two vectors.

For friction idealized by the Coulomb law, the maximum frictional force supported by the interface is:
where is the coefficient of friction. The equations to

be satisfied for the tangential component of interfacial
tractions and displacements are as follows:
670
Figure 5(a). GND evolution at T = 0 (i) Pure sticking

(ii) Frictional sliding.
3
3.1
Figure 5(b). GND evolution at T = 1/4th cycle (i) Pure

sticking (ii) Frictional sliding.

Effects of friction at the UC interface
Two cases have been modeled to understand the effects

of friction at the UC interface, one with a pure stick
condition and the other with an appreciable amount
of friction, replicating the friction behavior when the
constituent foils are in contact with each other. The
initial mating surface roughness has been fixed at
an average roughness, Ra 5 m with foil thickness
being 150 m. The boundary conditions applied are
(a) normal compression of 1800 N and (b) oscillatory shear amplitude of 16 m. The initial GND has
been assumed to be 0 throughout the microstructure
since the grain boundary area is negligible compared
to the entire volume. The GND evolution in the bulk
for one complete deformation cycle has been shown
in Figure 5. Figure 5(a) illustrates the initial configuration for both the pure sticking and frictional sliding
boundary conditions. During progressive loading in
the pure sticking case, the mating interface experiences no resistance to the shear deformation.This leads
to equal tangential displacements of the top and bottom foils in the vicinity of the interface. Therefore,
the GND starts developing at the top surface of the
top foil. On the contrary, in the frictional sliding case,
the bottom foil in the vicinity of the interface experiences significant resistance due to friction.This further
leads to much smaller tangential displacement of the

bottom foil compared to the top foil in the vicinity of
the interface. Henceforth, the GND is higher at the
interface compared to the top surface of the top foil in
this scenario. The GND evolution in both pure sticking
and frictional sliding scenarios has been demonstrated
in figures 5(b) through (e) corresponding to (1/4)th,
(1/2), (3/4)th and 1 full deformation cycle during UC.
It is also observed that the maximum GND after one
complete cycle (figure 5(e)) in the frictional sliding
case is greater than the pure sticking case by a factor of 2 suggesting more inhomogeneous deformation
in the former. The motivation behind demonstrating
these particular instants (figures 5(a)(e)) comes from
the fact that the applied shear loading either changes
its direction or its sign at these instants. Also, the
extent of GND evolution in the frictional sliding case
limits itself to about 20 m below the top surface
of the bottom foil in the normal loading direction
(y-axis). Therefore, this region experiences the maximum inhomogeneous plastic deformation, and kinematic hardening since the inhomogeneous plastic and
kinematic hardening is directly proportional to the
(Ma 2006). Moreover, continued GND evolution at
the interface means a higher production of dislocations
with either a positive or a negative sign. Therefore, a
671
Figure 5(c). GND evolution at T = 1/2 cycle (i) Pure sticking (ii) Frictional sliding.
very high number of these GNDs lead to their annihilation and formation of new relaxed equiaxed subgrains.
This phenomenon is termed dynamic recrystallization
(Rittel and Osivski 2010). The emergence of recrystallized subgrains can be confirmed from experiments
(Johnson 2008) where it has been observed that significant grain fragmentation has taken place near the
interface on the top part of the bottom foil with a good
amount of plastic flow beneath the region, as shown in
figure 6.
This continued evolution of GND at the interface
in the friction assisted case results in grain fragmentation (subgrain formation) by the end of 3000
material deformation cycles, since the average GND
in the 20 m strip below the top surface of the bottom
foil reaches a value 6 1015 dislocations/m2 in the
top part of the bottom foil (Figure 7). The sub-grain
diameter corresponding to this dislocation density is
2.28 m (Pal 2011).
3.2 LWD evolution with cyclic deformation

Linear Weld Density measures the amount of bonded
length to the total interfacial length, which is inversely
proportional to the average gap between the mating
Figure 5(d). GND evolution at T = 3/4th cycle (i) Pure

surfaces at the UC interface. LWD generally increases

with the amount of cyclic deformation. If the gap
becomes zero, both the surfaces adhere and start
plastically deforming together. Since the applied shear
loading is cyclic during UC processing, the interface
may lead to a fatigue type of failure and dynamic
recrystallization at the interface when subjected to
higher normal loads. Although the gap closure phenomenon has been included in the model, the prolonged fatigue phenomenon has not been included in
the current set of simulations and is left for future work.
To quantify the effects of processing parameters
such as oscillation amplitude and normal compressive load on LWD evolution, a design of experiments
(DOE) approach was used by Janaki Ram et al.
(2007) to systematically evaluate the effects of process parameters and to identify the optimum parameter
combination. Specific levels for each of the parameters were selected based on preliminary experiments,
machine setting limits, and available published information. A Taguchi L16 orthogonal array was utilized
in his study to determine the effects of individual
process parameters. Table 4 lists the parameter combinations used for the experiments, and Figure 8 shows
the experimental procedure.
672
Figure 7. GND evolution in the top part (20 m region)

of the bottom foil for a Ra 5 m in the top surface of the
bottom foil.
Figure 5(e). GND evolution at T = 1 complete cycle (i) Pure

Figure 6. Ion beam induced secondary electron micrograph

of DBFIB-etch result for UC interface showing extent and
depth of sub-grain size and morphology change as a result of
the plastic deformation (Johnson 2008).
3.2.1 Effect of oscillation amplitude

The L16 Taguchi experiments conducted by Janaki
Ram et al. (2007) are an ideal DDCP-FEM model
validation tool for UC. To investigate the effect of
oscillation amplitude on LWD, four distinct simulations have been carried out with an applied normal load
of 1750 N for 3000 deformation cycles and applied
ultrasonic amplitudes of 10, 13, 16 and 19 m. The foil
thickness was assumed to be 150 m, with an average
roughness of Ra 5 m for the top part of the bottom
foil. Figure 9 shows the average gap evolution between
the mating surfaces as a function of increasing amount
of cyclic deformation. It can be seen that for most
amplitudes, the average gap no longer decreases after
approximately the 10th cycle. For smaller amplitudes,
the average gap decreases with increasing amplitude
up to 16 m. Above 16 m the gap increases with
increasing amplitude.
LWD has not been explicitly predicted in the simulation since the model treats gaps between the top and
the bottom foils as 2 dimensional quantities whereas
LWD is a one dimensional, linear quantity. However,
the average gap can be used as an effective parameter
for comparison with LWD, as LWD and average gap
should show all of the same trends. A quantitative calibration between the average gap and LWD is left for
future work.
As can be seen from figure 10, the average gap
shows the same trends shown by Janaki Ram et al.
(2007), first increasing and then decreasing as a function of the applied ultrasonic amplitude. The optimum
amplitude where the average gap is found to be the least
was 16 m for the applied boundary conditions. This
is in good agreement with the experiments since the
peak LWD has been observed at the same oscillation
amplitude of 16 m.
3.2.2 Effect of normal force
Similarly, to understand the effects of increasing normal load on gap closure/LWD, four applied normal
loading scenarios have been considered. The applied
normal loads were 1450, 1600, 1750 and 1900 N.
The applied ultrasonic amplitude was 16 m. The foil
673
Table 3.
2007)
Parameters and levels selected for UC experiments(Janaki Ram et al.
Parameter
Level 1
Level 2
Level 3
Level 4
Oscillation Amplitude (m)

Welding Speed (mm/s)
Normal Force (N)
Temperature ( F)
10
28
1450
75
13
32
1600
150
16
36
1750
225
19
40
1900
300
Table 4. Taguchi L16 experimental matrix(Janaki Ram et al.

2007)
Run#
Amplitude
(m)
Welding
Speed
(mm/s)
Normal
Force
(N)
Temperature
( F)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
10
16
19
13
13
19
10
16
13
10
19
16
19
16
13
10
28
40
32
36
40
28
32
36
28
36
40
32
36
28
32
40
1450
1600
1750
1900
1750
1900
1600
1450
1600
1750
1450
1900
1600
1750
1450
1900
75
75
75
75
150
150
150
150
225
225
225
225
300
300
300
300
Figure 9. Gap evolution (in m) against number of deformation cycles with varying ultrasonic shear amplitudes.
Figure 10. Effect of processing parameters on linear weld

density, LWD vs. oscillation amplitude from (Janaki Ram
et al. 2007).
Figure 8. Schematic of Experimental UC Deposit Consisting of Four Layers. Welding occurred along the 100 mm
direction. Metallographic sample locations are shown in the
picture (Janaki Ram et al. 2007).
thickness and average roughness on the top part of

the bottom foil were kept at 150 m and Ra 5 m
respectively.Again, it can be seen (Figure 11) that there
is no appreciable change/decrease in the average gap
as a function of increasing cyclic deformation past the
end of the 10th cycle for most cases.
It has been observed by Janaki Ram et al. (Figure 12)
that when increasing the applied normal force, LWD
increases up to a load of 1750 N. A further increase in
normal load, however, lowered LWD. The most probable reason behind this LWD decrease is increase in the
Figure 11. Gap evolution (in m) against the number of

deformation cycles with varying normal loading scenarios.
inhomogeneous plastic deformation with load near the

top surface of the bottom foil. This leads to very high
amounts of dynamic recrystallization which leads to
dynamic failure of the bonded region at the interface.
674
Figure 12. Effect of processing parameters on linear weld

density, LWD vs. Normal force from (Janaki Ram et al. 2007).
Therefore, in order to predict the experimental results,

it will become necessary to incorporate fatigue crack
nucleation and growth phenomena within the current
DDCP-FEM.
The inclusion of fatigue phenomena is non-trvial
for the following two reasons:
4.2 Future work

The current model is able to predict the deformation
response during UC processing but requires additional
features to be realized as an efficient tool for material and parameter optimization in UC. The future
accomplishments required in this paradigm are:
The model assumes that once two nodes from the

top and bottom surfaces are in contact with each
other, the surfaces will be metallurgically bonded.
A traction-separation law to mimic crack propagation with increasing cyclic deformation is not
available from the literature.
In future work, a traction separation law will be

formulated as a function of involved fatigue damage
and further crack propagation to try to capture the bond
degradation seen during experiments at higher loads.
In conclusion, the effects of oscillation amplitude
and normal force on LWD evolution have been correctly modeled in this section with the exception of
decreasing LWD with increasing normal force beyond
a certain magnitude. Though the simulation parameter
(average gap) characterizing the bonded region has not
been calibrated quantitatively with LWD, it provides
a good heuristic for bond characterization since the
interfacial void closure is a 2 dimensional area based
phenomenon.
CONCLUSIONS AND FUTURE WORK
4.1
Conclusions
A DDCP-FEM model has been formulated and tested

based on UC boundary conditions comprised of simultaneous simple shear and normal compressive force
applied to polycrystalline Al 3003 alloy at ultrasonic
frequencies. The model was validated as follows:
A higher normal force causes better bonding at the

mating interface; though beyond a certain magnitude it results in a very high amount of inhomogeneous plastic deformation at the top surface
of the bottom foil causing severe grain fragmentation and dynamic fatigue failure. This is reflected
in LWD evolution, where LWD first increases with
increase in normal force and decrease beyond a
certain magnitude (Janaki Ram et al. 2007).
A higher amplitude enhances better bonding at

the interface. Although beyond a certain magnitude, it enhances the rate of gap relaxation, which
overcomes the rate of gap closure leading to lack
of fusion at the interface. Gap closure and gap
relaxation are directly and inversely proportional to
LWD evolution respectively. Therefore, it has been
predicted that the LWD would first increase with
the oscillation amplitude and thereafter decrease
beyond a certain magnitude. This prediction is consistent with experimental results (Janaki Ram et al.
2007).
To incorporate welding speed as a 3rd simultaneous

applied load alongwith normal force and oscillation
amplitude.
To incorporate more number of foils being deposited
on top of each other and their sequential consolidation. While doing this, the virgin foil for deposition
will be assumed to have minimal initial GND and
standard initial SSD content whereas for the previously deposited foils all the mesoscopic deformation variables such as SSD and GND will be stored
from the previous deposition instant and applied as
an initial condition for the fresh deposition.
To formulate a traction-separation law which can
quantify the delamination of foils as a function of
applied normal and simple shear loading. This is
important from the point of view of optimizing the
weld speed required to avoid crack propagation due
to fatigue after bond formation.
To perform Orientation Imaging Microscopy (OIM)
of the as-obtained and post-consolidated UC-foils
to generate realistic microstructures required for
realistic simulations.
To formulate a homogenized model based on current DDCP-FEM to predict deformation during
processing and in-service conditions. This is important since it will help with optimizing the initial
orientation for the product build.
To use the homogenized crystal plasticity model
and non-destructive engineering metrics for
closed-loop control (feed-forward and feed-back)
of the UC machine.
In its current form, the model is still under development and has to be validated for other crystalline
types such as single and polycrystalline variants of
body centered cubic (BCC) materials such as Molybdenum and Tantalum, hexagonal close packed (HCP)
materials such as Titanium and mixed alloys containing both HCP and BCC counterparts such as
Ti6Al4V. These alloys have been already demonstrated
as potential materials for UC practice, though the
biggest challenges to model these existing and new
675
alloys which may serve as potential materials for UC

are as follows:
Determination of the active slip systems. For HCP

and BCC, because of their limited slip system
activity, the active slip systems are a function of
the Schmid factor (projection of loading axis on
the slip system) and temperature. Therefore, the
active slip systems can either be obtained from
experiments and theoretical explanations provided
in the literature, phase field modeling or insitu
TEM experiments in which the resolved shear strain
evolution can be monitored.
Incorporation of mesoscopic deformation mechanisms, for example Nickel based superalloys such
as Inconel 718, which demonstrate significant twin
activity (Uchic et al. 2006) and BCC materials
exhibiting Peierls resistance to mobile dislocation
motion (Groger & Vitek 2009).
Determination of mesoscopic material parameters
(Qs and cs shown in Table 1).
UC fabricated simple parts to validate model
predictions.
REFERENCES
Bate, P. 1999. Modelling Deformation Microstructure with
the Crystal Plasticity Finite-Element Method. Philosophical Transactions: Mathematical, Physical and Engineering Sciences 357: 15891601.
Dolbow, J. et al. 2001. An extended finite element method for
modeling crack growth with frictional contact. Computational methods in applied mechanics and engineering,
190:68256846.
Friel, R.J. et al. 2010. The effect of interface topography

for Ultrasonic Consolidation of aluminium. Materials
Science and Engineering A 527: 44744483.
Groger R., & Vitek V. 2009. Stress Dependence of the Peierls
Barrier in BCC Metals. Proceedings of the 5th conference of the Asian Consortium in Computational Materials
Science, Hanoi, Vietnam.
Janaki Ram, G. D. et al. 2007. Effects of Process Parameters
on Bond Formation During Ultrasonic Consolidation of
Aluminum Alloy 3003. Journal of Manufacturing Systems
25: 221238.
Johnson, K. 2008. Interlaminar subgrain refinement in ultrasonic consolidation. Wolfson School of Mechanical and
Manufacturing Engineering, Loughborough University,
UK.. PhD Thesis.
Ma, A. et al. 2006. A dislocation density based constitutive
model for crystal plasticity FEM including geometrically
necessary dislocations. Acta Materialia, 54: 21692179.
Pal, D. 2011. Dislocation density based finite element
modeling of ultrasonic consolidation. Graduate School
of Mechanical and Aerospace Engineering, Utah State
University, USA. PhD Thesis.
Rittell, D. and Osivski, S. 2010. Dynamic failure by adiabatic shear banding, International Journal of Fracture
162: 177185.
Siddiq A., & Ghassemieh E. Theoretical and FE analysis of
Aluminum Alloy 3003. Journal of Manufacturing Science
and Engineering, 131:041007041011.
Uchic M.D. et al. 2006. Application of micro-sample testing to study fundamental aspects of plastic ?ow. Scripta
Materialia 54:759764.
Zhang, C.B. et al. 2006. A Study of Friction Behaviour in
Ultrasonic Welding (Consolidation) of Aluminium. Proceedings of the AWS Conference: Session 7: Friction and
Resistance Welding/Materials Bonding Processes
676

Virtual prototyping of force-feedback robotic instruments for surgery

J.M. Gmez-de-Gabriel & V.F. Muoz
University of Mlaga, Spain
W.S. Harwin & A. Barrow

University of Reading, United Kingdom
ABSTRACT: The design of surgical instruments is under strong constraints on the size, the kind and the
number of sensors and actuators which have to be kept to a minimum. This paper addresses the development of
a virtual prototyping tool for the assessment of the performance of different robotic instruments for tele-robotic
surgery, before building them. The telerobotic system should be able of performing intra-operative diagnostic
assessment in minimally invasive surgery by means of palpation. The full system is described in terms of its
design and implementation issues. Also, experiments with some virtual instrument prototypes, with different
mechanical and sensorial configurations have been evaluated.
1
1.1
INTRODUCTION
Minimally Invasive Surgery
Minimally Invasive Surgery techniques (MIS), where

interventions are made trough small incisions in the
patient using specialized instruments, allow for a
reduction on clinical complications and a reduced
hospitalization time.
Today assistant surgical robots are more specialized and miniaturized, working collaboratively with
the surgeon, stimulated by the new MIS procedures:
SILS (Merchant et al 2008) and NOTES (Malik and
others, 2006). Those techniques aim to minimize the
number of input ports to the abdominal cavity in order
to reduce the infection risk, morbidity, and a reduction of the patient recovery time. Specifically SILS
techniques (Single Incision Laparoscopic Surgery) are
using a single multi-port trocar to insert laparoscopic
instruments, while the NOTES methods (Natural OrificesTransluminal Endoscopic Surgery) are defined as
a mixed endoscopic-laparoscopic procedure that reach
the abdominal cavity through a natural orifice as the
mouth, vagina, urethra or anus. Both methods are considered less invasive than conventional laparoscopic
procedures.
Force or tactile feedback can be necessary for many
manipulation tasks, but the standard MIS training
provides the surgeons the skill to do manual tasks
(suturing, grasping, lump detection, etc.) through special instruments with video feedback only, without the
sense of touch. However there are some surgical tasks
where touch/force feedback is necessary.
Robotics and computers can enhance the performance of MIS techniques by means of scaling, filtering and other aids. However, all those systems have a
common deficiency: the lack of touch feedback.
Also, the growing interest of the surgeons in recovering the sense of touch (Snchez-de-Badajoz 2009)
to feel when grasping and touching the patient tissues
(King et al 2009) in MIS surgery seems to be against
the miniaturizing required for endoluminal robotics.
1.2 Haptic feedback
Haptics generally describes touch feedback (Okamura
2009), which may include kinesthetic (force) and
cutaneous (tactile) feedback.
Some reports (Ortmaier et al 2007) found that haptic feedback reduced unintentional injuries during a
dissection task, but operating time was longer.
1.2.1 Tactile feedback
Useful for exploratory tasks such as palpation, in
which distributed pressure or deformation information
can be used to identify hard lumps in surrounding soft
tissue.
It has been shown how tactile feedback induces
reduced grasping force in robot-assisted surgery
(King 2009), but it remains difficult to design both
tactile sensors and displays that are compatible with
the surgical environment.
1.2.2 Force feedback
Easier to implement and process than tactile feedback
by means of haptic devices, which are special robots
handled by the human operator.
Impedance type controlled haptics are the most
popular because of the lower cost (there are no force
sensors) and high responsiveness to human inputs.
Haptic techniques have been used for biomedical applications such as needle insertion control
(Zarrad, 2007), dentistry Learning (San Diego, 2008),
teleoperated palpation for calcified vessel detection
677
(Gwilliam, 2009), or force controlled telerobotic

grasper (Rosen, 1999) which can identify automatically different kinds of animal tissues among others.
Forces provide information about the stiffness of
the tissues which is closely related to its health. In this
sense, some special probes have been developed for
the identification of soft tissues abnormalities (Liu,
2009).
1.3
Purpose of this work
This paper is focused on the development of a virtual

prototyping tool for the assessment of the performance
of different robotic instruments for telerobotic surgery
by means of virtual prototyping. This way it should be
possible to find a minimum suitable configuration for
a robotic instrument for force feedback telesurgery.
After some information about the task model and
preliminary experiments, the system design, and then
its implementation will be described. Next, the virtual prototype model will be presented and later the
experiments and its results will be discussed. Finally,
conclusions on both the system and the experiments
can be found.
Figure 1. Tool for tactile feedback cancellation.
passive tool for tactile feedback cancellation (see Figure 1) and interaction with a virtual environment have
been used to answer the same question: Is it possible to properly locate the inclusions without tactile
feedback?
The passive tool showed the feasibility of a force
feedback control system without tactile feedback.
However rigid fixed inclusions (bones) translate closing motion into tool displacements.
2 TASK MODEL
The proposed task that will be used to compare the
performance of the instruments consists in finding a
lump inside an organ. In laparoscopic surgery many
organs can move and can be overlapped, so the desired
tool needs some organ handling capabilities.
2.1
Soft tissue mechanics
It is not easy to find a simplified model for the environment elasticity. Experimental measures (Rosen, 1999)
showed that it is possible to identify different types of
organs in terms of the variations of the tissue elasticity
with the displacement.
On the other hand, analytic models have been developed based on a dual Maxwell model with nonlinear
functions (Liu, 2009).
2.2
Simulators are commonly used for surgery and medical applications (Halvorsen, 2005) and preoperative
planning with models obtained from data from the real
patient (CT, MRI, etc). Laparoscopic simulators used
for training on the use of laparoscopic instruments
rarely use force feedback, however force feedback
commonly used in orthopaedic surgery training.
A full model of the environment mechanical properties is difficult to build. Moreover surgical procedures
as cutting or suturing will not be considered here.
Haptic interaction with the virtual environment has
been made through the multifinger haptic workstation
(Barrow 2006) as shown in Figure 2.
Four different experiments have been done with
moving/fixed inclusions and with visible/hidden conditions (see Figure 3).
Surgeon grasping mechanics
Interesting results have been obtained (Brown, 2004)

from a sample of surgeons of varying skill performing
different surgical tasks in terms of force and frequency:
mean force applied to the tool handles during tissue grasps was 8.52 N; maximum force was 68.17 N;
ninety-five percent of the handle angle frequency content was below 1.98 Hz, mean frequency percentage
below 5 Hz, during tissue grasps was 99.35%.
3
3.1 Interaction with virtual environments
PRELIMINARY EXPERIMENTS
Some preliminary experiments have been done in

order to check if the proposed task can be performed
with force feedback only (no tactile feedback): a
SYSTEM DESIGN
The virtual prototyping system is composed of a

physical part needed for the real-time haptic human
interaction and a software simulation of the proposed
tool. In this case, only the instrument kinematic constraints and sensorial system will be simulated. The
sensor readings are not simulated, but obtained from a
real sensor on a general purpose slave manipulator.
The different prototypes can have different kinematic constraints and force sensor configuration, that
are software components as can be seen in Figure 4.
With that system, the human operator can move the
two fingertips of the slave robots (with the motion
constraints imposed by the current prototype) while
678
Figure 4. General scheme of the virtual prototyping system.
Figure 2. Users point-of-view of the haptic master workstation with display for hand-eye coordination.
Figure 5. Classical implementation of a bilateral ForcePosition control system.
Each finger is modelled by its centre position and

its radius so just three axis cartesian forces for each
finger are considered. In that case each haptic device
needs three active axes, plus a passive spherical joint.
Each haptic has its own local coordinate frame, so
the global positions and forces have to be converted
between their local and the global reference frames.
Most commercial haptics provide a programming
toolkit with the direct and inverse kinematic models
that simplify programming.
Figure 3. Views of the virtual environments for preliminary

assessment.
feeling both the measured (with the sensorial constraints imposed by the current prototype) interaction
forces and the kinematic motion limitations of the
prototype, through the master haptic workstation.
The system also includes real video feedback, so
the user can associate its finger movements with the
robot movements.
4.1 The master workstation

The master system will be based on at least two haptic devices. Impedance control haptics are the most
suitable for this application. They receive torque (or
cartesian forces) references and return its current
position (or speed) as output. These devices dont
need force sensor, but instead, they need low inertia
back-drivable joints, while keeping enough torque and
speed.
Moreover, low latency is needed for stability, and a
high update rate is necessary to have a good stiffness
feedback. In general, less than 1ms latency, and 1 KHz
to 40 KHz update rate is recommended.
One haptic device per finger will be used. Although
there can be any number of them, only two have been
considered in this work.
4.2 Slave system

The slave system is composed by two positioncontrolled three-DOFs robots, with a round tip and
a three axis force sensor.
Position controlled system provides a straightforward way of following the masters positions. The
admittance control receives position references and
returns the sensed forces.
Each joint will have local joint position control,
with at least 1 KHz of reference update rate.
Control of the tip orientation is not available, so
the tips need to be spherical, and sensor readings will
have to be aligned to the global reference frame at the
slave site.
The sensor electronics should provide also a 1 K
samples/s per axis with minimum latency.
Also, global cartesian positions will have to be
translated to the local reference frames of each finger and the inverse kinematics model will be needed
to obtain the joint coordinates.
As the global workspace of the master workstation
is different to the slave workspace, some considerations have to be done in order to avoid unreachable or
singular positions, while avoiding fixed obstacles.
679
Figure 6. Dynamic modeling of the master/slave system.
4.3
Bilateral control loop
A bilateral position-force controller has been implemented. The control system without including the
virtual prototype constraints can be shown in Figure 5.
The force-position bilateral control loop is better than standard position-position control scheme in
non-contact motion.
However it is still a basic control scheme with a
trade-off between transparency and stability. Finding
a suitable force feedback gain depends on the overall
dynamics of the system (See Figure 6).
It is possible to find the dynamic parameters of the
master and the slave, but in surgery, the changes in the
environment stiffness Ke can be abrupt and large in
magnitude (e.g. touching a muscle, an organ, a bone,
another instrument or a rib for cardiothoracic surgery).
Also stiffness estimation can be difficult due to organ
motion (breathing or displacements)
Possible improvements to be implemented are the
adaptive position-position control system (Poignet,
2009; Zarrad, 2007) and environment stiffness estimator based on an active observer and a Kalman
filter.
However the effects of the changes of the dynamic
gain during a teleoperated palpation task are still to
know.
Figure 7. Dual haptic system with two Phantom Premium

1.0.
Figure 8. First version of the slave robotic system
IMPLEMENTATION
5.1 The multi-finger haptic workstation

The multi-finger haptic workstation is a previous
development of the University of Reading (Barrow
2006) for the multi-finger interaction with virtual
environments (See Figure 7) which features:
2 to 3 Phantom Premium 1.0 haptics (3 active DOF

each).
2 or 3D Stereographics immersive display
C++ environment with haptic control, maths and
graphics libraries.
The special arrangement of the haptics provides a
large collision-free workspace.
5.2
Slave system
A new custom multi-finger slave site has been developed. This special hand allows free cartesian motion
of each fingertip and also provides three axis force
measurement at each round fingertip.
A rigid frame holds the fingers upside down, so
the experimental tissues can rest on a flat surface at a
fixed distance.
Two versions of the slave robot have been implemented. The first one showed some resolution and
compliance problems solved with the second one.
5.2.1 Robotic fingers first version
The first version of the fingers was (See Figure 9)
implemented with the set of robotics servos and brackets from the Dynamixel AX series from Robotis. The
AX-18F model is an enhanced version of the classic
AX-12+ with a Maxon coreless motor that offers more
torque and speed while keeping the same electronics
and mechanics.
This servo has a TTL level semi-duplex UART
bus interface. Each AX-18F has proportional position control system with programmable parameters.
Additional information such as current position and
load can also be read.
They have a potentiometer for absolute position
feedback with 10 bits resolution that gives an angular resolution of about 0.3 degrees. This can be too
low resolution, especially for the two first joints.
These servos are lightweight, and low cost but
exhibit high backslash and high unmeasured compliance due to the low resolution.
680
Figure 9. Slave system v1.1 and organ simulator.
All the servos are linked by the same bus that provides communications and power as well. A high speed
(1 Mbps) and multicast messages are used in order to
obtain a maximum update rate of 1.1 KHz.
The same main computer for the master workstation is used for the control of the fingers in order
to reduce latency. A free C-language library is also
available from the manufacturer
The plastic frame set of the Bioloid robot system
has been used with poor results due to its flexibility
and tolerances that increased the finger compliances.
5.2.2 Robotic fingers 2nd version
In order to obtain a lower compliance finger with better
resolution, a different set of servos has been tried.
The EX-106 servos from Dynamixel offer 4 times
more angular resolution (12 bits) with a combination
of incremental encoder and potentiometer and higher
torque, but the resulting robot is no longer low-cost.
A set of metallic brackets and ball bearings make a
solid structure with less compliance and backlash. A
new kinematic configuration has also been adopted.
These new actuators use the same protocol and
programming, but the bus shows physical and electrical differences.
In order to reduce the total cost per finger, a
RX-64 servo can be used for the last joint, with 10 bits
resolution and less torque but the same other features.
5.2.3 Force sensing
The force sensor has been built with three 750 g microload-cells with a basic Wheatstone bridge which
outputs 0.7 mV/V, so a signal conditioning and data
acquisition system next to the sensors has been used for
each finger. The data acquisition systems, previously
developed (Tse 2010), are powered and communicated
by a USB port and provide 10 bits resolution readings
with a 1KHz sample frequency for each channel.
The three micro-load cells have been arranged in
a very sensitive serial configuration (See Figure 11)
that offers uncoupled cartesian measurements, without
moving parts, and with good sensitivity. This sensor
body is also the last link of the finger.
Figure 10. Second version of the slave fingers with metal

links and ball bearings.
Figure 11. Detail of the micro-load cells arrangement for

the three axis force sensor.
Zero setting is achieved automatically at start-up

and gravity compensation hasnt been found necessary, due to the low weight of the sensor.
Noise filtering was not found necessary as the high
frequency was filtered by the dynamics of the master
haptics, but when the forces are around zero, the noise
changes the sign of the signals and a high audible noise
appears at the haptic actuators. A 5 Hz low-pass filter
has been programmed to solve this problem.
5.3 Data logging
In order to analyze the results of the experiments with
the virtual prototypes, full information about the task
execution (positions, and unfiltered measured forces)
is always been recorded.
681
Figure 12. Schematics of the two virtual surgical instruments. a) Para lel grasper, b) Parallel symmetric gasper.
The information is archived with time-stamped file

names. The files have text format with 13 columns for
the time and positions and forces for the two fingers.
Data is recorded each 10 ms by a low priority task.
6 VIRTUAL PROTOTYPES
Virtual prototypes are the software component that
adds constraints to the motion and sensing. Positions
of the haptics are changed by the nearest position of
the virtual instrument, and then sent to the slave robot.
In order to make the user feel these constraints and
to keep the position of the master near the position
of the virtual tool, virtual forces, proportional to the
distance are calculated.
Force readings are also translated taking in count
the number and position of the sensors in the virtual
tool. Finally the sum of the virtual and sensed forces
is sent, to be displayed, to the haptic system.
Four different virtual prototypes for robotic
graspers have been tested.
6.1
General unconstrained
This is not a tool model. It is an unconstrained mode

for testing purposes that implements a transparent telepresence system. There arent position constraints and
all the three axis forces per finger are fed to the haptic
system.
This is the ideal case, and had been used to test all
the system components and to tune the force feedback
gains. In that mode the operator can move its fingers
freely and feel all the contact forces.
6.2
Parallel grasper
The parallel grasper (See Figure 12a) is a tool with

displacements in Z axis only and with fingers opening
independently in the X axis. The Y coordinate of the
fingers is always zero.
The gripper has two independent sensors, one on
each finger, for the forces along the X axis.
The calculated feedback for the first finger of this
virtual instrument F1 can be written as (1):
Figure 13. Virtual prototype experiments performing a

palpation/manipulation task.
where f1 is the measured force along the x axis from

the sensor in slave finger 1, y1 is the position of the
master finger 1 along the Y axis, z1 and z2 are the z
coordinates for the two master fingers, and K is the
elasticity constant for the virtual spring that keeps the
two fingers in the y = 0 plane and with same z. Similar
formulation applies to finger 2.
6.3
Parallel symmetric grasper
The parallel symmetric grasper (See Figure 13b) is a

tool with displacements in Z axis only and with both
fingers opening the same way in the X axis. The Y
coordinate of the fingers is always zero.
In this case, the gripper has a single actuator and a
single sensor for the grasping forces along the X axis
which is calculated as the sum of the two.
The calculated feedback for the first finger of this
virtual instrument F1 can be written as (2):
where f1 is the measured force along the x axis from

the sensor in slave finger 1, y1 is the position of the
master finger 1 along the Y axis, x1 , x2 , z1 and z2 are
the x and z coordinates for the two master fingers, and
K is the elasticity constant for the virtual spring that
keeps the two fingers in the y = 0 plane and with same
z. Similar formulation applies to finger 2.
7
EXPERIMENTS
The experiments are based on performing an

exploratory task consisting on finding a rigid object,
a crystal ball of 16 mm diameter, inside a partitioned
foam organ as seen in Figure 13.
Accuracy and completion time are the main parameters considered, but also qualitative information is
682
also gathered, like a rating for the ability for organ

handling and guessing the approximate size/shape of
the hidden object.
Currently, a trial with volunteers and different runs
for each one of the virtual instruments is running with
the following preliminary results:
7.1
analyze the performance of the grasping and palpation

tasks.
This tool can provide information about the quality
of the task execution for objective prototype evaluation.
7.5 Summary
General unconstrained
This telepresence model is evaluated in order to compare its results with the other prototypes. As expected,
the average results showed that it was easy and fast in
finding the inclusion, easy to guess the size but with
just fair organ manipulation ability.
A comparative summary of the results can be found

in Table 1, where it can be seen how the performance
is degraded when reducing the number of degrees of
freedom and sensors.
8
7.2
The average results for this three degrees of freedom

parallel grasper model showed that it was just fair to
find inclusion, no shape guessing at all, and limited
organ manipulation ability.
7.3
Symmetrical parallel grasper
The average results for this two degrees of freedom

parallel grasper model showed that it was very hard to
find the inclusion, no shape guessing at all, and very
poor organ manipulation ability.
7.4
CONCLUSIONS
Parallel grasper
Data processing
Processing of the experiments information is done

off-line using different tools. In that case (See Figure 14) Python mathematical libraries are being used
to obtain and plot finger distance and grasping force to
The design and the development of a system for the

study of the performance of different manipulation
instruments has been presented here. This tool is useful
for the evaluation of different tool alternatives without a physical implementation. A software mode of
the proposed instruments is combined with a physical
telerobotic system.
Also, the analysis of the obtained information during the trials provides a better understanding of the
use of the instruments, so further optimizations of the
prototype can be made.
This system is currently being used to evaluate the
performance of the designs of surgical tools at the University of Reading, although it can be used for the
prototyping of human manipulation devices in general.
The processing of the sensorial information for
dynamic force feedback gain and/or tissue identification will improve the performance of the system.
Although it may not be found necessary in some
cases, we think that a tactile sensor/display system can
widen the range of applications of this tool.
ACKNOWLEDGEMENTS
This work is fund by the Education Ministry of
Spain trough the Programa nacional de Movilidad
de Recursos Humanos del Plan nacional de I-D+i
20082011, and the Spanish national research project
Robotic platform for NOTES/SILS assistance. DPI
2010-21126-C03-01.
Special Thanks to Dr. V. Becerra, B. Tse and B.
Janko for his valuable help.
Figure 14. Data analysis of the experiment results.
Table 1. Preliminary comparative results of the experiments with the virtual

prototypes.
Features
Results
Prototype
DOFs
Sensors
Location
Shape
Handling
Unconstrained
Parallel Grasper
Symm. Grasp.
6
3
2
6
2
1
Easy
Fair
Hard
Easy
No
No
Fair
Limited
Poor
683
REFERENCES
Barrow, A. and Harwin, W.S. 2006. A Dynamic Virtual
Environment for Haptic Interaction Proceedings of
Eurohaptics Conference pp. 377382.
Bogado-Torres, J.M. 2007, Control bilateral de robots teleoperados por convergencia de estados, PhD Thesis,
Universidad Politcnica de Madrid.
Brown, J.D. 2004. Quantifying Surgeon Grasping Mechanics in Laparoscopy Using the Blue DRAGON System,
Studies in Health Technology and Informatics Medicine
Meets Virtual Reality, Newport Beach, CA.
Gwilliam J.C., Mahvash M., Vagvolgyi B., A. Vacharat, D
D. Yuh, and A. M. Okamura, 2009. Effects of Haptic
and Graphical Force Feedback on Teleoperated Palpation, IEEE International Conference on Robotics and
Automation, Kobe, Japan, May 1217.
Halvorsen, F.H. et al. 2005, Simulators in Surgery, Minimally Invasive Therapy 14:45; pp. 214223.
King, C.-H. Culjat, M.O. Franco, M.L. Lewis, C.E. Dutson,
E.P. Grundfest, W.S. Bisley, J.W. 2009, Tactile Feedback Induces Reduced Grasping Force in Robot-Assisted
Surgery, Haptics, IEEE Transactions on, vol. 2, no. 2,
pp. 103110.
Liu, H. Elhage, O. Dasgupta, P. Challacombe, B. Murphy, D.
Seneviratne, L. Althoefer, K., 2009. A haptic probe for
soft tissue abnormality identification during minimally
invasive surgery, Reconfigurable Mechanisms and

Robots, ASME/IFToMM International Conference on,
2224 June 2009.
Okamura, A. M., 2009, Haptic Feedback in Robot-Assited
Minimally Invasive Surgery, Curr. Opin. Urol.
Ortmaier T, Deml, B, Kuebler, B, et al. 2007. Robot assisted
force feedback surgery. In: Ferre M, Buss M, Aracil R,
et al., editors. Advances in Telerobotics, Springer Tracts in
Advanced Robotics (STAR) Vol. 31. Springer; New York:
2007. pp. 341358.
Poignet, P, et al. 2009. Some control related issues in minirobotics for endoluminal surgery, 31st Annual Intl. Conf.
Of the IEEE EMB, pp. 6850-6855
Rosen, J. Hannaford, B. MacFarlane, M.P. Sinanan, M.N.,
1999. Force controlled and teleoperated endoscopic
grasper for minimally invasive surgery-experimental
performance evaluation, Biomedical Engineering, IEEE
Transactions on, 46:10.
B. Tse, W. Harwin, A. Barrow, B. Quinn, J. S. Diego and
M. Cox, 2010. Design and Development of a Haptic
Dental Training System hapTEL Lecture Notes in
Computer Science Volume 6192, SpringerLink.
Zarrad, W. et al 2007, Stability and Transparency Analysis
of a Haptic Feedback Controller for Medical Applications, 46th IEEE Conference on Decision and Control
New Orleans.
684

A comparison between BioCAD and some known methods for finite element
model generation
P.Y. Noritomi, T.A. Xavier & J.V.L. Silva
Three-Dimensional Technologies Division, Center for Information Technology Renato Archer CTI, Campinas,
So Paulo, Brazil
ABSTRACT: The finite element analysis has been used as an important tool for several health area studies.
In face of the challenges related to the complex geometry of biological structures, there are many methods for
obtaining FEA models. The objective of this work was to compare the BioCAD protocol used by the DT3DCTI (Campinas, Brazil) with some other known methods and to suggest which one could be the most suitable
depending on the objective of the analysis. The BioCAD protocol can generate versatile, standardized FEA
models, which can be compatible with the data processing capacity of any laboratory, while providing coherent
and accurate results. Such protocol is suitable for analysis of general phenomena, while the use of medical images
to generate high detailed models are promising for individual cases, although, in some situations, the BioCAD
models of the database generated previously can be edited and used in such FEA simulations.
INTRODUCTION
The finite element analyses (FEA) have been used in

several researches which aim to better understand the
performance of anatomical structures and to develop
or improve surgical techniques, prosthesis or implant
designs and restorative materials for health area.
In order to obtain accurate and coherent results by
the analyses, so they can be extrapolated to explain real
situations, the FEA models must be representative of
the biomechanical solicitation which is the object of
the study. In the case of simulations of biological structures, there are many challenges related not only to the
complexity and variability of material properties, but
also to the geometry, full of intricate details.
With respect to the FEA models features, some
could be mentioned as important to make the analysis
practicable and to provide precise and reliable results,
such as:
representativeness of the real problem: it must contain the main structures which indeed participate
in the problem, modeled with an adequate level of
details according to the objective of the analysis,
allowing the application of loads and boundary conditions for simulating the biomechanical situation
the more realistic as possible;
mesh refinement at a level that obtains accurate
results;
an amount of information compatible with the data
processing capacity of the laboratory structure,
which sometimes requires the elimination of less
relevant details and the mesh refinement only in
the area of interest.
There are many methods for obtaining FEA models

described in the literature, ranging from the generation
of very simple geometric entities by computer-aided
design (CAD) tools, to more complex procedures, such
as those which use medical images for creating models
with a higher level of details.
In face of so many methods to generate biological models for FEA, the objective of this work is to
compare the BioCAD protocol, used by the ThreeDimensional Technologies Division (DT3D) of the
Center for Information Technology Renato Archer
(CTI, Campinas, SP, Brazil) with some other known
methods. Based on this comparison, it is intended to
suggest which method is the most suitable for obtaining a FEA model in function of the objective of the
analysis.
SOME KNOWN APPROACHES FOR

OBTAINING BIOLOGICAL MODELS
FOR FEA
In face of the difficulty in modeling complex biological structures, some studies build models based on
medical images, such as those obtained by computed
tomography (CT) and magnetic resonance, in order to
obtain more easily many details and the relative spatial
positioning of the structures to be simulated.
That is the case of the researches made by Gautam
et al. (2007) and Gautam et al. (2009), which aimed
to study the effects caused by orthopedic devices for
rapid maxillary expansion and a headgear, respectively, on the entire skull. The model was built based
685
Figure 1. FEA model of a skull generated with basis on a

CT image used by Gautam et al. (2007) and Gautam et al.
(2009).
on an image obtained by computed tomography scanning of a seven-year-old child dry skull, excluding
the mandible, with intervals of 2.5 mm between each
slice. The individual sections were traced in a CAD
program, which served as reference for generating
the surfaces joining the lines, and they were used to
generate volumes, simulating the skull (cortical and
trabecular bones). The teeth and periodontal ligament
were not simulated, because of the difficulty in representing them accurately by CT images. The image
(Fig. 1) of the model used in these studies allows noticing that the surfaces created to join the traced slices
did not respect the continuity of the curvatures of the
skull.
After the three-dimensional (3D) reconstruction of
an image taken by magnetic resonance, CT (or, in
the case of smaller objects with many details, microCT), it can be generated a STereoLithography (*.STL)
format file, which represents the 3D object by a triangulated (2D) surface mesh. Some authors create a
volumetric mesh from the surface triangulations in
order to use it in the FEA program. By means of such
a method, the mesh is generated without the existence
of a previous geometrical entity.
Boryor et al. (2009) published a paper describing the sequence for constructing highly detailed
mesh for a FEA model representing a left mandibular
canine tooth. The model comprised the dental structures (enamel, dentin and pulp) and the surrounding
tissues (periodontal ligament, cortical bone, trabecular bone and nerve), and it was built based on an
image taken by micro-CT. According to the authors,
the model construction included the integration of
many (and some expansive) software programs for
editing the mesh of the STL file after the segmentation of the micro-CT image. It was necessary a
considerable manual work in order to eliminate noise
and correct the triangles which encountered problems.
It took a couple of months to construct the model,
and it were obtained two sizes of model with different numbers of elements, which could be chosen by
the researchers and edited according to the desired
level of details and number of nodes. The level of
Figure 2. Cervical region of the left mandibular canine FEA

model, built by Boryor et al. (2009).
details required a high number of elements to represent them (Fig. 2), and even the lowest detailed model
had a large amount of data. The two meshes generated are available for free download from http://www.
uni-ulm.de/uni/intgruppen/fem/, inasmuch as the
authors were concerned about the necessity of a very
detailed model to obtain accurate results and were
trying to encourage other researchers to share their
detailed models with other groups too. Besides the
possibility of comparing results obtained by different
researchers by using the same model, the sharing of
models could reduce the time spent in constructing
them. The authors of the study made aware of the time
spent on the data loading and on the time needed for
the FEA simulation with such models, depending on
the machine used.
The so called CT-Based Finite Element Method has
been employed for generating FEA models individualizing both the anatomy and the heterogeneous material
properties (simulating the bone density distribution).
Bessho et al. (2007) suggested the possibility of using
such simulations as a non-invasive method for predicting the fracture risk of the proximal femur in patients
with osteoporosis so that, based on the results, it could
be planned a preventive intervention. For this study,
the cadaver femora used in the experimental test were
scanned by CT, for providing the anatomical shape and
the local material properties to the FEA model. The
heterogeneous properties of the bone were imputed
for each tetrahedral element of the FEA mesh according to the local average bone density obtained by the
CT image.
The use of geometrical entities built in a CAD
program instead of a STL file surface mesh provided by medical images can be an alternative for
simplifying the model, including only the most important details for simulating the biomechanical situation.
Two studies can be mentioned as illustration of using
geometrical entities to construct biological models:
Baek et al. (2008) and Chang et al. (2004), which
analyzed orthodontic movement in inferior and superior teeth, respectively. Without a medical image as
a basis, but based on measures available in the literature, they generated simple geometry representing the teeth, bones and the orthodontic devices.
686
Figure 3. The 3D FEA model generated by Baek et al.

(2008) of the mandibular dentition, consisting of teeth, periodontal ligament, bone and the brackets of the orthodontic
device.
Figure 5. One of the models built for the study of Field et al.
(2009) using a CAD program with NURBS representation of
the geometry.
Figure 4. The 3D FEA model generated by Chang et al.

(2004) of the maxillary dentition, comprising the teeth,
periodontal ligament, bone and brackets.
The periodontal ligament was simulated by 2D elements within the FEA program. The images from these
two works (Fig. 3,4) allow noticing that the complex
curvatures of the anatomical structures were not well
represented, with sharp contours of bones and teeth,
besides the considerable change in the shape of the
bone basis.
Sun et al. (2005) published a work in which it
is highlighted the suitability of the use of CAD
tools able to generate complex geometries (such as
those tools which work with Non-Uniform Rational
B-Spline NURBS functions) for modeling biological structures for several applications, which was
called Bio-CAD modeling. The use of such CAD tools
for building more detailed models based on medical images (Li et al. 2005, Sun et al. 2005) has been
employed as an interesting alternative for generating
biological models for FEA application.
The study made by Field et al. (2009) analyzed
the effects provoked by orthodontic loading applied to
inferior teeth, for what the authors used a CT image of
an adult mandible as a basis for generating a geometrical model in a CAD program by means of NURBS
representation of the geometry. In the same program,
it was modeled the orthodontic device according to
the manufacturers information. The Figure 5 allows
noticing that, while simple, the geometry follows the
natural complex curvatures and has a level of details
Figure 6. FEA model built by Li et al. (2005) representing

the two-unit cantilever dental bridge.
more coherent to the real structures than those used by

Baek, S. H 2008 and Chang YI 2004.
Li et al (2005) built a model using a program with
NURBS representation of the geometry, containing a
two-unit cantilever dental bridge. The model was built
based on a cloud of points provided by CT image of the
restorative (composite pontic, reinforced polyalkane
?ber and adhesive layer) and dental (enamel, dentin
and pulp) structures. In the CAD program, curves
were traced demarking the outer contour of each crosssectional slice provided by the CT scan, so the surfaces
were generated blending the curves and, after that, the
surfaces delimited the solids. According to the authors,
this method of geometry generation benefited the
meshing process during FEA. Taking the geometrical
entities into account, it was made a major mesh refinement in areas of greatest stress concentration: Figure 6
shows a simple mesh, but it can be noticed a coarser
mesh in the dental root. The authors also mentioned
the possibility of modifying the previous geometric
687
Figure 8. FEA model used by Lai (2010), in which it can

be seen the mesh refinement in specific regions.
Figure 7. Geometrical model obtained by Silva (2008)
representing an upper first premolar and its surrounding tissues constructed by BioCAD protocol. It can be seen lines
demarking some main landmarks.
model according to individual dimensions for future

simulations. In this work, it was suggested that the
method used for model generation, based on medical
images of each patient, can provide FEA results which
can serve as a basis for planning the prosthesis design
individually during the clinical procedures.
Since 2005, during the work of technical support to
postgraduate research on bioengineering, it was developed, in DT3D-CTI, its own BioCAD protocol, in
order to standardize the generation of models of biological structures for FEA. Such a protocol requires
the use of a CAD software able to generate complex geometries integrated with the knowledge about
anatomy and biomechanics by the researcher. Instead
of drawing all the external details of an anatomical
structure indiscriminately, the BioCAD protocol proposes the choice of the most universal landmarks,
only those which are important to well represent
the anatomy under an acceptable simplification. The
sequence consists in the generation of lines demarcating these main landmarks, which serve as reference
for the surfaces and they, in turn, delimitate the solids
to be imported into the FEA program. In addition to
the demarcation of the main landmarks, the protocol
proposes the generation of other lines and surfaces
in strategic locations for mesh refinement only in the
areas of interest and to facilitate the application of
loads and boundary conditions during the FEA.
One of the first studies (Silva 2008) developed with
technical support given by DT3D-CTI, employed the
BioCAD protocol for analyzing, by FEA, the stress
distribution in an upper first premolar provoked by
physiologic and non-physiologic occlusal contacts. In
addition to this objective, this work aimed to indicate which were the most important landmarks to be
modeled in this case. Without a medical image as reference, the model was generated in a CAD program
with NURBS representation following the average
measures of the main landmarks found in the literature, including the dental structures, periodontal
ligament and the surrounding bones. The geometrical

model (Fig. 7) was imported into the FEA program,
and the mesh was generated. The results obtained by
the simulation showed a stress distribution which can
explain the occurrence of enamel injuries in the case
of non-physiological occlusal contacts.
Another work (Lai 2010) developed with the help of
DT3D-CTI group used the BioCAD protocol during
the generation of a FEA model in order to study a kind
of medical device used for rapid maxillary expansion.
As the anatomic structures were more complex than
those simulated by Silva (2008), it was used a medical image of a skull obtained by CT, which is part of
the database of DT3D-CTI. For reducing the amount
of data in less interesting areas in order to make the
model compatible with the data processing capacity of
the DT3D-CTI structure, it was not modeled the entire
skull as made by Gautam et al. (2007) and Gautam et al.
(2009), but only the regions nearest the maxillary bone.
The model included the expander, cortical bone (without trabecular bone), the left and right first molars and
premolars (in which the expander was anchored) and
the palate mucosa only in the side where it was touched
by the expander device. I was not simulated the periodontal ligament, in order to decrease the amount of
elements in the FEA mesh. The lines were generated
in a CAD program by NURBS representation, passing
by the most representative anatomic landmarks, and
delimiting small surfaces in order to refine the FEA
mesh in some specific areas. Some lines were positioned on the alveolar process delimiting the local of
each tooth, even those teeth which were not modeled in
this case, so it could facilitate the positioning of them
in future studies. The details that did not correspond
to general aspects of the population were not included
in the model, because they would also increase unnecessarily the amount of mesh elements during the FEA.
Still aiming to reduce the number of elements needed
to build the FEA mesh, as this case involved a large
model, even the chosen details to make part of the
model (as the teeth cusps and the incisive foramen)
had their contours smoothed. The presence of lines
and surfaces in the model allowed a manual control
of the mesh refinement in the surfaces of the teeth,
688
expander and in the bone closer to the teeth (Fig. 8).

They also served as reference for the application of
the displacement (type of load applied in this case)
and the boundary conditions, eliminating the difficulties related to the application of such conditions by
using element faces. During the FEA, some boundary
conditions were also applied in order to compensate
the absence of the non-modeled structures.
3
DISCUSSION
In the studies made by Gautam et al. (2007) and

Gautam et al. (2009), the attempt to include the details
of the skull in the model found limitations because of
the large interval between each slice of the CT image.
These intervals were filled by surfaces which did not
follow the typical curvature of the anatomic structures,
so the details were not simulated accurately, besides
the inclusion of sharp angles that do not exist in real
skulls, which could influence on the results of stress
dissipation. In these two studies, the simulation of the
entire skull is justified because of the goal of making an exploratory study about the effects of maxillary
movements aided by biomedical devices on the entire
skull. Based on the results obtained by these studies,
which indicate which structures are the most affected
by the movement, or those which cause more influence on the movement pattern, it is possible to select,
for future models (which aim, for instance, to compare
different devices or techniques), only the known most
important features to be included in the model, as made
by Lai (2010). In the case of general studies which do
not analyze asymmetrical problems (on the contrary
of that analyzed by Lai 2010, which studied a modified asymmetrical expander), it could be modeled only
half of the geometry, which would be compensated
by the application of a boundary condition indicating
the symmetric problem during FEA. This procedure
would allow representing the geometries of interest
with more details.
The use of medical images for obtaining models
with a high level of individualization, as those made
by Bessho et al. (2007) and Boryor et al. (2009), can
generate results which do not correspond to a general
pattern, sometimes making them almost impossible to
be extrapolated to a population.
The use of a STL format file for generating a
mesh to be imported into the FEA program, as made
by Boryor et al. (2009), is an apparently convenient
method for obtaining easily accurate details, but in
many times, there is a necessity of much manual labor
to correct the mesh before using it. When this method
is employed, the number and size of elements is controlled mainly by the size growth rate from the surface
toward to the interior of the solid, while the surface
mesh refinement is defined by the STL file mesh. A
STL mesh with a high level of refinement, beyond
the necessary, can increase the time for the analysis
or even exceed the capacity of data processing of the
machine. On the other hand, a coarse mesh in areas
of interest can obtain results without accuracy. The

absence of geometrical entities further complicates
the application of loads and boundary conditions and
even the use of such models in future works which
need to include restorations, deformities, fractures or
biomedical devices in the model.
Whereas some authors show a concern about which
method is the most suitable for acquiring images for
constructing biological models (Li et al. 2005, Mehta
et al. 1997, Sun et al. 2005), the first study mentioned
here which employed the BioCAD protocol (Silva
2008) showed that it is not indispensable a previous
high detailed image of the real object to serve as a basis
for generating a model that best represents its structure. A CAD tool able to generate complex geometries
and the knowledge of anatomy and biomechanics to
choose the main landmarks are enough to generate a
representative model. A medical image can make the
modeling easier related to the structures dimensions
and relative spatial positioning among them, as long
as the modeler takes care to not include individual
details that could interfere on the results and make
them impossible to be extrapolated to general cases, if
the study does not aim to analyze a particular one.
The models built for the studies made by Baek et al.
(2008) and Chang et al. (2004) are examples of higher
level of simplification of the geometry than it is proposed by the BioCAD protocol. If, on the one hand,
a high simplification reduces the analysis time and
can produce results that can explain general phenomena, on the other hand, the lack of some landmarks
and the inclusion of unreal sharp angles during the
model generation may influence on the stress field
obtained by the analysis, making them impossible to
be extrapolated too.
The results obtained by Lai (2010) and Silva (2008)
allow saying that a model built with average measures,
in a level of details including only the main landmarks, followed by the application of adequate FEA
hypotheses, can decrease the amount of data to be processed while obtaining coherent and accurate results
in the interesting areas. Such results can be used as
a basis for a better comprehension of the mechanism
of injuries occurrence, as well as for the development
and improvement of preventive, restorative or surgical
techniques and biomedical devices for the health area.
Comparisons among the results made by different
researchers depend on standardized models, which is
a concern mentioned by Boryor et al. (2009), who suggested the sharing of them. Sharing of models is not the
unique way, neither is the easiest, to make comparisons
possible among researchers. The BioCAD protocol is
a way to generate standardized models related to the
kind and level of details to be included, according to the
biomechanical situation which is object of the study
and the area of interest within the model.
One of the most advantageous aspects of the BioCAD protocol is the possibility of the construction of a
database containing versatile, general models that can
be used in future works. It is possible to add lines and
surfaces to them aiming to make a mesh refinement or
689
to apply loads and boundary conditions in different

areas of a structure modeled previously. There still
is the possibility of editing the solids for simulating
deformities, fractures or biomedical devices, without
the necessity of beginning from the start whenever
someone wants to have a new model.
4
CONCLUSION
Among several possible methods of creating FEA

models of biological structures, the BioCAD protocol
demonstrated to be able to generate an amount of data
compatible with the data processing capacity of any
laboratory. Such simple models can be representative
of the real situations and generate coherent and accurate results. At the same time, they can be considered
standardized models related to the details included,
which makes it possible to compare results among different researchers. Such a protocol decreases the time
used to construct the models during the CAD and the
FEA steps, since the geometrical entities serve as reference for applying load and boundary conditions and
refining the mesh only in specific regions, besides the
possibility of editing some models in the database in
order to obtain new others for future works.
Models built by the BioCAD protocol are suitable
for studies which aim to explain some general phenomena or general patterns, when the results obtained by
individual models cannot be extrapolated. The methods which use medical images for obtaining FEA
models with a high level of individualization (related
to the anatomy and material properties) are promising as non-invasive method ones for predicting failure
risks or planning biomedical devices individually or
still, in future works, for simulating tissue remodeling
according to the history of stimuli of each patient. In
some situations which require individual details, the
BioCAD models of the database generated previously
can be edited and used in the FEA simulations.
REFERENCES
Baek, S.H.; Shin, S.J.; Ahn, S.J. & Chang, Y.I. 2008. Initial
effect of multiloop edgewise archwire on the mandibular
dentition in Class III malocclusion subjects. A threedimensional finite element study. European Journal of
Orthodontics 30: 105.
Bessho, M.; Ohnishi, I.; Matsuyama, J.; Matsumoto, T.; Imai,

K. & Nakamura, K. 2007. Prediction of strength and
strain of the proximal femur by a CT-based finite element
method. Journal of Biomechanics 40: 174553.
Boryor, A.; Hohmann, A.; Geiger, M.; Wolfram, U.; Sander,
C. & Sander, F.G. 2009. A downloadable meshed human
canine tooth model with PDL and bone for finite element
simulations. Dental Materials 25: e5762.
Chang,Y.I.; Shin, S.J. & Baek, S.H. 2004. Three-dimensional
finite element analysis in distal en masse movement of the
maxillary dentition with the multiloop edgewise archwire.
European Journal of Orthodontics 26: 33945.
Field, C.; Ichim, I.; Swain, M.V.; Chan, E.; Darendeliler,
M.A.; Li, W. & Li, Q. 2009. Mechanical responses
to orthodontic loading: a 3-dimensional finite element
multi-tooth model. American Journal of Orthodontics and
Dentofacial Orthopedics 135: 17481.
Gautam, P.; Valiathan, A. & Adhikari, R. 2007. Stress
and displacement patterns in the craniofacial skeleton
with rapid maxillary expansion: a finite element method
study. American Journal of Orthodontics and Dentofacial
Orthopedics 132: 5 e111.
Gautam, P.; Valiathan, A. & Adhikari, R. 2009. Craniofacial displacement in response to varying headgear forces
evaluated biomechanically with finite element analysis. American Journal of Orthodontics and Dentofacial
Orthopedics 135: 50715.
Lai, K.P. 2010. Anlise pelo mtodo de elementos finitos
das tenses resultantes da disjuno rpida de maxila
utilizando um disjuntor modificado [thesis]. Campinas:
Centro de Ps-Graduao So Leopoldo Mandic.
Li, W.; Swain, M.V.; Li, Q. & Steven, G.P. 2005. Towards
automated 3D finite element modeling of direct fiber
reinforced composite dental bridge. Journal of Biomedical Materials Research. Part B, Applied Biomaterials 74:
5208.
Mehta, B.V.; Rajani, S. & Sinha, G. 1997. Comparison of
image processing techniques (magnetic resonance imaging, computed tomography scan and ultrasound) for 3D
modeling and analysis of the human bones. Journal of
Digital Imaging 10: 2036.
Silva, A.O. 2008. Protocolo de modelagem tridimensional
do primeiro pr-molar superior para o mtodo dos elementos finitos e anlise das concentraes de tenses na
regio cervical do esmalte [thesis]. Araraquara: Universidade Estadual Paulista Julio de Mesquita Filho, Faculdade
de Odontologia de Araraquara.
Sun, W.; Starly, B.; Nam, J. & Darling, A. 2005. Bio-CAD
modeling and its applications in computer-aided tissue
engineering. Computer-Aided Design 37: 10971114.
690

Towards simulation of a bioreactor environment for biofabricated tissue

maturation
R.A. Rezende, C.A.R. Laureti & J.V.L. da Silva
Renato Archer Information Technology Center CTI, Campinas, Brazil
V. Mironov
Medical University of South Carolina MUSC, Charleston, USA
V. Kasyanov
Riga Stradins University, Riga, Latvia
R. M. Filho
School of Chemical Engineering, University of Campinas, Brazil
ABSTRACT: Bioprinting of tissues and organs can be defined as layer-by-layer additive robotic biofabrication of three-dimensional functional living macrotissues and organ constructs using tissue spheroids as building
blocks. The microtissues and tissue spheroids are living materials with certain measurable, evolving and potentially controllable composition, material and biological properties. Closely placed tissue spheroids undergo tissue
fusion, a process that represents a fundamental biological and biophysical principle of developmental biologyinspired directed tissue self-assembly. After the tissue spheroids structuring, the tissue/organ newly made is then
carried out into a bioreactor which should play an important role of providing an adequate environment to the
growth and maturation of the bioproduct. Bioreactors are used to accelerate tissue maturation through the control
of their mechanical, biochemical and electrical conditions. First of all, they should maintain the viability of the
engineered tissue. Following, they are many times employed as equipment to the cell seeding and can be also
applied to experimental and monitoring of maturation processes. The creation of a representative environment
inside the bioreactor is too complex since it can enclose a large range of variables. Simulating this scenery is
essential to the study. The success of tissues and organs bioprinting is straight linked to a set of an appropriate
environment in the bioreactor that assures the feasibility, maturation, biomonitoring, tests, storing and transport
of the involved elements on the generation of the new tissue such as the deposited cells and nutrients. As an
example, the perfusion and fluid diffusion phenomena within the organs in maturation process in bioreactor is
fundamental for understanding of the phenomenon. On the other hand, computational fluid dynamic software
packets have been increasingly developed during the past decade and are powerful tool to calculate flow fields,
shear stresses and mass transport within and around 3D constructs, including a bioreactor environment. This
work presents a preliminary study that reproduces elements included in the bioreactor approach with some
variables considered at the simulations based on the finite element method running on Ansys CFD software.
1
1.1
INTRODUCTION
Bioprinting or human organ printing
During the First International Workshop on Bioprinting and Biopatterning held at the University of
Manchester, UK, in September 2004, the term bioprinting was defined as the use of material transfer
processes for patterning and assembling biologically
relevant materials molecules, cells, tissues, and
biodegradable biomaterials with a prescribed organization to accomplish one or more biological functions
(Mironov et al. 2006a).
Organ-printing technology using self-assembled
tissue spheroids in certain aspects is also conceptually
very close to the recently invented concept of digital

printing (Hiller and Lipson, 2007, 2009).
It is obvious that large-scale biofabrication of tissue spheroids; development of continuous and digital industrial bioprinters; robotic bioprinting of 3D
macrotissues and living organ constructs using microtissues; bioengineering of a perfusable intraorgan
branched vascular tree built-in these 3D macrotissue constructs; development of microfluidic-inspired
irrigation dripping tripled perfusion bioreactor with
removable non-biodegradable porous tubes and accelerated tissue maturation technologies are serious technological challenges that need to be systematically
addressed in order to facilitate successful development
691
Figure 1. Concept of irrigation dripping tripled perfusion

bioreactor (Mironov et al. 2009).
of a microtissue-based approach. (Mironov et al.

2009).
Figure 2. A needle involved by a volume of hydrogel.
1.2
Bioreactor
With the rapid development of Tissue Engineering for

the past twenty years and more recently the organ bioprinting, bioreactors have gained more attention as an
indispensable tool to establish suitable conditions to
the new tissue maturation
A bioreactor may be defined as a fluidic system
that simulates physiological or desired environment
for the creation, maturation, physical conditioning,
biomonitoring, and testing of viable tissue engineered
constructs and/or isolated organs in vitro (Barron et al.
2003; Mironov et al. 2006a).
The term bioreactor refers to a system in which
conditions are closely controlled to permit or induce
certain behavior in living cells or tissues. The concept
of bioreactors is neither new nor restricted to tissue
engineering (Korossis et al. 2005).
Bioreactors can be designed to maintain physiological parameters at desired levels, enhance mass transport rates, and expose cell-seeded three-dimensional
scaffolds to specific biochemical or physical stimuli
(Korossis et al. 2005).
2
2.1
STRUCTURING A BIOREACTOR
ENVIRONMENT
General roles of a bioreactor
Bioreactors have many important functions. First, they

must maintain the viability of the engineered tissues.
Second, they are often (but on always) used as a cellseeding device. Often they can be used for accelerating
tissue maturation using mechanical, biochemical, and
electrical stimuli conditioning. They also can be used
for testing and monitoring maturation processes in
tissue-engineered constructs. Finally they can serve
for storage, preservation, and transportation of tissue
engineered constructs (Mironov, 2006b).
A bioreactor has to be able to operate over long periods of time under aseptic conditions since maturation
of a functional tissue may take up to 34 months. Providing three-dimensional tissues with nutrients may
rely on passive diffusion, or may be more actively
delivered by direct perfusion. However, direct perfusion introduces a new level of complexity when scaleup is encountered, and the engineering challenges may
be significant. Tissues that have been manufactured to
date have relied on diffusion, although tissues envisioned for future products will require a more active
delivery process (Korossis et al. 2005).
2.2 Irrigation dripping tripled perfusion bioreactor
A biofabricated 3D macrotissue constructs must
undergo a rapid process of tissue maturation or fluid
solid transition in order to maintain their shape,
composition and integrity (Mironov et al. 2009). The
development of irrigation dripping tripled perfusion
bioreactor with temporal nonbiodegradable removable
porous tubes (Fig. 1) can also help to maintain fragile
tissue constructs and buy time necessary for postprocessing tissue fusion, remodeling and maturation,
and also can be used for delivery of maturogens
(Mironov et al. 2008). Many needles will be displayed
in the bioreactor inside a hydrogel environment as
shown in Figure 2.
Bioreactors also allow studies of mechanical stimuli on 3D tissue structures. Mechanical stimuli, such
as shear stress due to flow characteristics, have been
shown to have a great effect on the development of
tissues (Ellis et al. 2005). For example, cardiac muscle in vivo encounters strong pulsatile flows, whereas
bone constantly encounters mechanical stress and
compression (Langer et al. 2008).
Concerning the use of bioreactor for construct, the
only mechanism by which nutrients and waste can
move when a scaffold is in static culture is by diffusion. It has been shown that despite homogeneous
cell seeding, after long periods in culture, more cells
are found on the periphery of constructs (Cartmell
et al. 2003) leading to peripheral encapsulation which
hinders nutrient and waste exchange from the centre, resulting in core degradation of tissue engineered
constructs. This is of major concern in the field of
tissue engineering, and is a major obstacle in the formation of a viable tissue in vitro (Partap et al. 2010).
692
Table 1. The needle geometry parameters adopted.

Non-newtonian system
Needle length
Uniform pore diameter
External needle diameter
Internal needle diameter
1500
40
470
420
All the units in m.
Figure 3. Model of needle with 28 rings.
Obviously, in the case of Biofabrication, there is not

a cell condition on periphery but the perfusion is of
major importance.
2.3
Proposal for the bioreactor environment
As an initial step aiming at constructing a suitable

bioreactor to maturate bioprinted organs, in this work,
it was highlighted a study on the diffusion of a fluid
through a porous laterally needle. The fluid was established to flow through the porous. Outside the needle
there is a volume containing a hydrogel (Fig. 2). The
proposal is to verify how a lesser viscous fluid flows
through a needle with certain conditions and also how
this fluid interact physically when it is spread outside the needle being immersed into a more viscous
material.
2.4
In this work, it was chosen an intermediary alginate

concentration of 3%, where k = 6 Pa.s and n = 0.84
with density equals to 1.4 g/cm3 and a shear rate range
from 0.01 to 100 s1 and an average viscosity in the
range of 2.8 to 12.5 Pa.s (Rezende et al. 2009).
Materials properties for the simulation

scenario construction
Concerning the environment composed of tissue

spheroids (or cell aggregates) and a kind of hydrogel, in this case the alginate, it was considered as the
material for simulations.
In a first moment, alginate is the material inside and
outside the needle and water being injected into the
needle with velocity equals to 1 mm/s trying to push
alginate out (Part I). On a second approach, without
alginate, water was considered inside and outside the
needle as a unique material (Part II).
Alginate is an anionic copolymer composed
by homopolymeric regions of 1,4-linked -Dmannuronic (M blocks) and -L-guluronic acid (G
blocks), interspersed with regions of alternating structure (Draget et al. 2006; Eiselt et al. 2000). According
to Rezende et al. (2009), alginate is a non-newtonian
material or a shear-thinning system since its viscosity varies in terms of the shear rate and which is
represented by the Power-Law model (Equation 1):
SIMULATIONS CONDITIONS
The general idea in our simulations of a bioreactor

is to evaluate the needles positioning and distribution
in the bioreactor environment including the distance
between the needles, the number of pores in each one,
the pores diameter, in order to optimize those parameters for a better tissue perfusion. This present and
preliminary study concerns the evaluation of just some
parameters at the needle such as the pores diameter,
the number of rings at long the needle and the number of pored per ring. Future studies can include the
association of many needles with controlled temperature, nutrient and oxygen taxes, stress and strain, for
example.
The simulations were carried out in ANSYS CFX.
The needles are porous and in all the cases the
geometry distribution of the pores is hexagonal. No
comparison about this distribution has done in this
paper.
In order to discrete the governing equations the software ANSYS CFX makes use of an element-based
finite volume method, which firstly involves to discrete the spatial domain using a mesh. The mesh is
used to construct finite volumes, which are used to
conserve relevant quantities such as mass, momentum,
and energy. A control volume is constructed around
each mesh node and these equations are integrated over
each control volume (ANSYS CFX Guide, 2006).
The instantaneous equations (Eqs. 2 to 4) of continuity and momentum can be written as follows:
The Continuity Equation:
The Momentum Equation:

1
where is the shear stress (Pa), is the shear rate (s ),

k is the consistency index (Pa.sn ) and n is the powerlaw index (dimensionless). The systems classification
is:
n < 1: shear-thinning system.
n = 1: Newtonian system.
n > 1: shear-thickening system.
where the stress tensor, , is related to the strain rate

by:
693
Figure 4. Volume Fraction: 28 rings around the needle

(Case 1).
Figure 7. Volume Fraction 8 pores per ring (Case 3).
Figure 5. Model of needle with 14 rings.
Figure 8. Volume Fraction 4 pores per ring (Case 4).
Figure 6. Volume Fraction half of the original rings

(Case 2).
The resultant flow of a liquid through a needle with

many holes or pores laterally and with the final tip
completely closed. Some important variables are the
diameter of each pore, the number of pores per ring and
the distance between the rings. All of these parameters
were considered in this work.
4
RESULTS
The results are divided on two parts and each part

subdivided into the sections linear and logarithmic
settings, respectively, which will be explained ahead.
PART IInteraction Water-Alginate

Linear setting
First of all, it was considered a condition in which
the needle was composed with 28 linear rings (Fig. 3)
containing 30 pores each one along the length of the
needle (Case 1) (Figure 4).
This simulation represents a real period of 12 seconds. What it can be also noted is that the water volume
leaves the needle practically through the pores at the
first ring of pores and the water does not get to push the
alginate content which is more dense along the needle
length. The exit at first ring is the easiest way the water
finds to escape to outside. Figure 4 shows that it is possible to subtly observe the effect of gravity (according
to y axis) having the middle line as a reference.
The number of rings (Fig. 5), in the Case 2 (Fig. 6),
was reduced by half keeping the diameter of the pores.
At the end, the water content reached less volume than in the Case 1 and therefore more water
escaped through first ring then the case before. Case 2
represents the same real period of 12 seconds.
694
Figure 9. Volume Fraction with logarithmically pores distribution per ring (Case 5): (a) in xz axis view and (b) in yz
axis view.
It looks like the field created outside the needle

difficult the exit through other pores or, on an opposite
reason, In Cases 1 and 2, the geometry distribution of
the pores was hexagonal (Figs. 3, 5).
Figure 7 shows the resizing of number of pores per
ring. Instead of 30 pores, now the scenario considers
8 pores. This case and the next were simulated for
real 5 seconds, so less than the cases before.
In the Case 3 (Fig. 7) and Case 4 (Fig. 8), after 5
seconds (much less than in Case 2), the water content
achieved the centre of the fifth pore and between fifth
and sixth pores, respectively. This demonstrates that as
less as the amount of porous per ring, more the water
content is spread along the needle.
It is interesting to note that as less as the number of
pores per ring much more the spread outside the needle
and also more pores are crossed, that is, in Case 4, the
field of water volume generated outside the needle is
larger and is feed now from new pores.
Figure 10. Velocity of alginate leaving the needle (with logarithmically pores distribution per ring) (Case 5): (a) in xz
axis view and (b) in yz axis view.
Logarithmic setting
As it has been observed that in the previous cases
the water content trends to leave the needle through the
initial pores, it was though a way to get more difficult
this and facilitate the flow through next pores. Then, a
logarithmic distribution was considered for the number of pores at long the needle increased as shown in
Case 5 and also for the pore diameter varying at long
the needle (being constant at the same ring). The logarithmic scale was chosen since there is a fast increase
on the beginning in order to compensate the high flow
escaping from the first ring (Table 2 column: number
of pores per ring).
Figure 9ab presents the volume fraction in two
views while Figure 10ab shows, also in two views, the
alginate velocity spreading from the needle to outside.
Figure 11 presents a distribution where the pore
diameter dimensioning was also logarithmic according to the Table 2 (column: pore diameter), where all
the 14 rings are composed from 5 to 10 pores each
one progressively (Case 6). In this case, there is an
695
Table 2. Number of pores per ring (non-discrete and discrete) and pore diameter adopted values.
Figure 11. Volume Fraction (with logarithmically pores

diameter at long the needle) (Case 6): (a) in xz axis view
and (b) in yz axis view.
Number of pores per ring
Pore Diameter
Ring
Non-discrete
Discrete
(mm)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
5.00
7.77
8.43
8.82
9.11
9.33
9.49
9.66
9.79
9.91
10.01
10.11
10.19
10.27
5
8
8
9
9
9
9
10
10
10
10
10
10
10
10.2
11.8
13.6
15.3
17.2
19.2
21.3
23.5
25.8
28.2
30.9
33.7
36.7
40.0
Figure 13. Volume Fraction 8 porous (Part II) (Case 8).
Figure 14. Volume Fraction with pores logarithmically

distributed per ring (Part II) (Case 9).
Figure 12. Volume Fraction 4 porous (Part II) (Case 7).
inversion about the exit flow profile. The larger diameters into the end created a way with fewer barriers
to water spread itself. Now, all the pores are used
flow water.
PART II Interaction Water-Water

The alginate properties were replaced by of the
water ones. It is the configuration where the needle
is already filled with water. Then, when the simulations are started, the internal volume of needle is
completed with water. In this case, it is evident that
the water is much more spread into outside the needle
696
Figure 15. Volume Fraction with logarithmically pores

diameter at long the needle (Part II) (Case 10).
first rings. When compensation in terms of distribution and diameters of pores is incorporated different
results are found being possible to check that the part
of the flow crosses the needle length reaching the last
rings.
The water to water interaction with no viscosity and
density differences shows an extreme. No mechanical
resistance was offered to the water as in the initial cases
and this got possible the entering water being spread
until the end of the needle length. Anyway, it was
not the ideal situation since the expectation is to have
the most uniform distribution in all the pores of the
geometry. It is clear that equilibrium must be achieved
between the two parts considered in this work.
As an initial work, it highlights the necessity to validate initial simulations and then go to a simulation of
the whole structure to check the optimized behavior
of perfusion. Many other phenomena are intrinsic in
a bioreactor and as soon as other parameters can be
incorporated so faster the development of a suitable
bioreactor for tissue maturation will be.
ACKNOWLEDGEMENTS
Figure 16. Gravity effect for an 8 pores needle as an

example.
This work was sponsored by So Paulo Research

Foundation (FAPESP), The Brazilian Institute of Biofabrication (INCT-BIOFABRIS), The National Council for Scientific and Technological Development
(CNPq) through CTI/PCI program and NSF R-II grant
South Carolina Project for Organ Biofabrication,
US.
more uniformly. All the next simulations (Figs. 1215)

run equivalently to real 5 seconds.
REFERENCES
Linear setting
Effect of gravity
In some previous figures the effect of the gravity
was not easily perceived. However, Figure 16 shows
clearly that the gravity influences the final results and
should be considered during the simulations. Eventually, this effect can be suppressed if the needles will
be displayed vertically inside the bioreactor.
5
CONCLUSIONS
Preliminary analyses on the phenomenon of diffusion

through a needle within the bioreactor show the interaction between two fluids, water and a hydrogel, and
how the lesser dense fluid is spread into this hydrogel.
Different geometric settings were adopted as such as
variations on number and distribution of porous and
the pores diameter.
Regarding the simulations in which the needle
contained hydrogel initially, it was verified that the
resistance the hydrogel offers to the water, lesser viscous, forces the flow to leave the needle already by the
Ansys CFX, 2009, Users Guide.

Barron V, Lyons E, Stenson-Cox C, McHugh PE, Pandit
A., 2003. Bioreactors for cardiovascular cell and tissue
growth: a review. Ann Biomed Eng. 31(9): 101730.
Cartmell, S.H.; Porter, B.D.; Garcia, A.J. & Guldberg, R.E.
2003. Effects of Medium Perfusion Rate on Cell-Seeded
Three-Dimensional Bone Constructs in Vitro. Tissue Eng,
9, 11971203.
Draget, K.I., Braek, G.S., Stokke, B.T., 2006. Similarities
and differences between alginic acid gels and ionically
crosslinked alginate gels, Food Hydrocolloids, 20, 170
175.
Ellis, M., M. Jarman-Smith, and J.B. Chaudhuri. 2005. Bioreactor systems for tissue engineering: A four-dimensional
challenge. In Bioreactors for Tissue Engineering, edited
by J. Chaudhuri and M. Al-Rubeai. Dordrecht: Springer.
Eiselt, P., Yeh, J., Latvala, R.K., Shea, L.D., Mooney, D.J.,
2000. Porous carriers for biomedical applications based
on alginate hydrogels, Biomaterials, 21, 19211927.
Grove, A.T. 1980. Geomorphic evolution of the Sahara and
the Nile. In M.A.J. Williams & H. Faure (eds), The Sahara
and the Nile: 2135. Rotterdam: Balkema.
Hiller J, Lipson H. 2007. Methods of parallel voxel manipulation for 3D digital printing. In: Proceedings of the 18th
solid freeform fabrication symposium, p. 20011.
Hiller, J., Lipson, H.. 2009. Design and analysis of digital materials for physical 3D voxel printing, Rapid
Prototyping Journal, Vol. 15, No. 2, pp. 137149.
697
Korossis, S.A., Bolland, F., Kearney, J.N., Fisher, J. and Ingham, E., 2005. Bioreactors in Tissue Engineering. In:
Topics in Tis sue Engineering 2005, Volume 2. Eds. N.
Ashammakhi & R.L. Reis.
Langer, R., Shukla, A., Moore, M., Yang, F., Neeley, W.,
Karp, J., 2008. Tissue Engineering [Internet]. In: Nair
L, Laurencin C, organizadores. Nanotechnology and
Tissue Engineering. CRC Press, p. 332. Available from: http://www.crcnetbase.com/doi/abs/10.1201/
9781420051834.sec1
Mironov, V., Reis, N., Derby, B., 2006-a. Bioprinting: a
Beginning. Tissue Engineering. Volume 12, Number 4.
Mironov, V., Kasyanov, V.A., Yost, M.J., Visconti, R., Twal,
W., Trusk, T., Wen, X., Ozolanta, I., Kadishs, A.,
Prestwich, G.D., Terracio, L. Markwald, R.R., 2006-b.
Cardiovascular Tissue Engineering I. Perfusion Bioreactors: A Review. Journal of Long-Term Effects of Medical
Implants, 16(2):111130.
Mironov V, Kasyanov V, Drake C, Markwald RR., 2008.

Organ printing: promises and challenges. Regen Med,
3:93103.
Mironov, V., Visconti, R.P., Kasyanov, V., Forgacs, G.,
Drake, C.J., Markwald, R.R.. 2009. Organ printing:
Tissue spheroids as building blocks. Biomaterials, 30,
216474.
Partap, S., Plunkett, N.A., OBrien, F.J., 2010. Bioreactors
in Tissue Engineering. In: Tissue Engineering, edited by
Daniel Eberli MD, In-Teh, Ch.16, pp 32336, India.
Rezende, R.A., Brtolo, P.J.S., Mendes, A., Maciel Filho,
R., 2009. Rheological Behavior of Alginate Solutions for
Biomanufacturing. Journal of Applied Polymer Science,
v.113, 386671.
698

Prototyping process of a virtual-reality treadmill system for

exploration of real world panoramic environments
J. Hu & G.M. Fadel
Department of Mechanical Engineering, Clemson University, Clemson, USA
I. Wood, P. Napieralski & S. Babu

School of Computing, Clemson University, Clemson, USA
ABSTRACT: This paper describes the prototyping process of a virtual-reality (VR) treadmill for exploration
of a panoramic environment. The VR treadmill is a non-motorized treadmill outfitted with automatic controls
so that its incline can be adjusted according to the terrain elevation data, and the head tracked imagery of a
panoramic environment (Google Streetview) can be displayed on a commercial head mounted display (HMD)
and updated as the person walks on the treadmill. This implementation of VR devices is an interdisciplinary effort
requiring the cooperation of mechanical, electrical and computer science researchers. The system is decomposed
into three subsystems: the mechanisms, the control devices and the VR components, with the approach to solve
the problem described. This paper discusses the motivation for building such a prototype. Further, the limitations
of the design are exposed and future research directions are presented.
1
INTRODUCTION
As an integral part of the design process, the prototyping of a product allows designers to explore design
alternatives, test theories and confirm performance
prior to manufacture. Thus, the quality of the prototype
to some degree impacts the final production design.
In this paper, we present the prototyping process
of a VR treadmill that integrates: a HMD with inertial orientation tracker; a treadmill simulating natural
travel; and a Google Streetview browser interface for
panoramic environment exploration. This system has
been built in order to provide a framework in which we
can investigate how humans coordinate between the
visual simulation in VR and the physical simulation
on the treadmill, and the effects of immersion within
a panoramic environment on human spatial cognition
and navigation.
Panoramic navigation enables the exploration of
virtual environments based on real world data in real
time. Although there are significant limitations and
challenges in exploring such environments in a fully
continuous manner due to physical restraints in data
gathering or limitations on bandwidth and processing power, there is a need to develop such systems
and investigate the human response to an immersive
environment such as Google Streetview. It enables
the internet-faring public to explore large regions of
the globe at ground level. This method of exploration
allows the user to achieve a greater degree of familiarity with the layout of unfamiliar places than if they
used traditional maps for navigation. However, the
value of immersion for human path-finding ability and
cognitive mapping of an environment is still the subject

of ongoing studies.
This paper is divided as follows: Section 2 describes
related work on prototyping technologies, human
way-finding and natural travel techniques; Section 3
describes the prototyping details of the three subsystems; Section 4 explains the limitations of the current
prototype and identifies what needs to be upgraded in
the new version of the prototype to satisfy the users;
Section 5 summarizes the value of the prototyping
process and discusses the initial conclusions of our
work.
RELATED WORK
Prototyping techniques are frequently referred to in

design and manufacturing. For example, Buchenau &
Suri (2000) defined prototypes as representations of
a design made before artifacts exist, ranging from
sketches and different kind of models at various levels to explore and communicate propositions about
the design and its context. They proposed that designers can improve prototypes by referring to users
experiences. Liou (2008) also proposed prototyping
as an important tool to evaluate design solutions and
a visual to help communicate the products purpose
and feel. Cooper (2001) introduced prototyping techniques in manufacturing, while Bao et al. (2002) and
Zhang et al. (2007) introduced computer-aided-design
(CAD) prototyping in conceptual design.
As for travel techniques in virtual environments,
Bowman et al. (1997) developed taxonomy for
699
classifying virtual travel techniques and an experimental method to evaluate and compare different
techniques. They reported that decoupling steering
from gaze direction to pointing can result in time
benefits for tasks involving relative motion, and that
travel by jumping from one waypoint to another
decreases a users sense of presence. Such jumping is typically used by panoramic environments,
although some, including Google Streetview, have
motion effects to lessen discontinuity. While Ruddle
et al. (1999) reported that using a HMD to navigate an expansive virtual environment is faster and
gives the user a better sense of straight line distance
than a desktop interface. Zanbaka et al. (2005) found
that participants who explored a small virtual environment with head tracking displays, natural travel
and exploration techniques answered questions related
to understanding, application and higher mental processes correctly more often than those who used a
monitor and joystick. They also report that real walking in a space as large as the virtual environment
allows participants to form better spatial awareness
of the environment, demonstrated through accurate
responses of the participants in a sketch map based
questionnaire. Suma et al. (2010) found that real walking and moving where looking allowed participants to
complete the maze faster and with fewer collisions
than moving where pointing. One possible explanation for their results may be attributed to the fact that
participants may have felt that moving where pointing
was less intuitive than moving where looking.
3
3.1
PROTOTYPING PROCESS
Objective
The aim of the project was to develop a virtual reality treadmill to study virtual environments. A design
process was followed to modify a non-motorized treadmill to accomplish the required tasks. The first step in
such a process is to identify the objective, and then
apply some methodology to obtain the required result.
The objective of this project is: Design a treadmill that
can adjust its platform incline automatically within
a range according to terrain data downloaded from
Google Streetview and connect VR glasses to display
the panoramic photographs from Google Streetview,
updating them as the person walks on the treadmill,
simulating the real environment. Thus, the total system should realize the simulated integration of both
the visual virtual reality and the physical movements.
Having defined the objective, next a list of requirements has to be specified to be used in the decision
process.
3.2
Requirements list
The requirements list is compiled based on the methodology proposed by Pahl et al. (2007) as seen in
Table 1. Demand indicates the requirement that must
be satisfied; otherwise, the design fails to achieve
its objective. While Wish means the expected
Table 1.
D/W
Partial requirements list of the VR treadmill.

Requirements
1. Geometry:
D
The elevating mechanism must not interfere the
operating zone of the user;
W
The elevating mechanism should fit in the space
under the platform;
W
The number of components of the mechanism
should be as few as possible;
2. Kinematics:
D
Gradient adjustment range must = 5 0 ;
W
Gradient range should = 10 5 ;
3. Force:
D
Must carry a person of 250 lbs and an additional
200 lbs impact load;
D
Must be steady, with the error rate 3% when
loaded;
4. Energy:
D
Must use grid power;
D
Must be clean, steady, quiet and powerful;
5. Safety:
D
Must obey the OSHA standards;
6. Cost:
D
Must cost less than 800 $;
7. Others:
W
Display frequency 24 fps;
W
Resolution of VR glasses 640 480;
W
Error of synchronicity 1 s; . . .
requirement may or may not be achieved, but it can be

used to identify the better designs. The requirements
list was obtained by interviewing potential users, and
Table 1 shows a subset of the entire list.
The requirements list can be used to formalize the
prototyping objective from an abstract statement to
a specific set of technical criteria. Thus, an idealfinal-result (IFR) model (Altshuller, 1984, 1996) is
proposed assuming that all the demands and wishes
are realized in it. The IFR model will be applied as a
reference to compare the candidate solutions in latter
sections.
3.3 Workflow and decomposition
The system can be decomposed into three subsystems: mechanical, controlling and VR. Furthermore,
the components under each subsystem are detailed to
satisfy the requirements, with the signal flow stringing
them up as shown in Figure 1.
In a single loop of the signal transformation, the
workflow proceeds as follows:
Step 1: the sensor collects data from the rotating belt
of the treadmill to calculate the participants moving
speed and displacement;
Step 2: the collected data are transferred to the
computer;
Step 3: the computer uploads the data to the Google
Streetview server to identify the position of the
participant;
700
Table 2.
Power devices comparison.
Criteria (Weight)
Figure 1. The general workflow and decomposition of the

system.
Figure 2. Function structure of the system. EE: electric

energy; ME: mechanical energy; HE: human energy.
Step 4: the computer downloads the terrain and photographic data according to the identified position;
Step 5: the computer supplies the photographic data
to the VR device to display; meanwhile, it computes
elevation change from the terrain data to and sends
it to the controller;
Step 6: the controller translates the terrain elevation
data to controlling signals and transfers the signals
to the elevating mechanism;
Step 7: the mechanism adjusts the incline of the
treadmill.
To illustrate more clearly how the energy and signal
flows are transformed in the system, the systematic
function structure can be derived as seen in Figure 2.
3.4
Mechanical subsystem
For this project, a non-motorized treadmill was purchased from the market. This treadmill does not
include any motors to control the users speed or adjust
its incline. Its working principle is based on the difference between the friction coefficients on the two
opposite surfaces of the belt. The friction coefficient
of the top surface contacting with the users shoes must
be larger than that of the bottom one contacting with a
supporting platform, and hence when the user steps on
the belt and walks forwards, the force applied on the
top surface can overcome the reversed friction force
Hydraulic Pneumatic Electric IFR
Cost (9)
1
Complexity (4)
1
Thrust (9)
9
Noise (4)
9
Responding speed (4) 1
Anti-interference (4)
9
Total
30
(170)
Ratio to IFR (%)
56
(56)
4
1
1
1
4
9
20
(105)
37
(34)
9
4
4
4
9
1
31
(189)
57
(62)
9
9
9
9
9
9
54
(306)
on the bottom, driving the belt to move backwards

simulating normal walking motion. Note that the decision to use a non-motorized treadmill is specifically
to allow the user to stop and look around in the virtual
environment.
For the elevating mechanism, the power devices
drive it should be selected. Three options are available, hydraulic, pneumatic devices or electric motors.
According to the requirements list, the IFR of power
devices should be cheap, simple, powerful, silent,
quickly responding, and not easily interfered with.
Such an ideal solution is proposed as a reference and
compared with the three options through weighing
with quantitative scales (1, 4, 9: 1 = low, 4 = moderate,
9 = high) (Maier et al., 2009) in a decision matrix
(DM) as seen in Table 2.
It is necessary to introduce that a DM is used to
roughly evaluate the items in the column according
to the criteria in the row, and the quantitative scales
(1, 4, 9) only represent the hierarchy of the three levels of quality rather than the real differences among
the levels. Therefore, options obtaining near equal
final scores mean that they are equivalently acceptable. For example, in Table 2 the comparison indicates
that using either an electric motor or a set of hydraulic
devices to drive the mechanism is applicable. However, the hydraulic is graded with three low scores on a
high-priority criterion: cost, and two moderate-priority
criteria, complexity and responding speed; while the
electric solution obtains only one low score on a
moderate-priority criterion: anti-interference. Hence,
the electric motor is preferred in this project because
its performance is preferred to that of the hydraulic
solution according to the six criteria.
Once the actuation energy source is selected, the
mechanism that induces the elevation change has to
be designed. It should be driven by a motor and
transform the rotating power of the motor to the elevating force applied on the platform of the treadmill;
meanwhile, it should be installed preferably beneath
the platform of the treadmill in order not to protrude and accidentally injure the user. Based on this
functional description and the requirements list in
Table 1, the IFRs characteristics are identified: cheap,
steady, accurate, simple, small, quickly responding,
anti-interference and driven efficiently. These characteristics actually constrain the selection scope of
701
Figure 4. Problem of the four-bar linkage.
Figure 3. Sketches of six alternative mechanisms.

Table 3.
Mechanisms comparison.
Criteria (Weight)
Cost (4)
1
Strength (9)
9
Accuracy (1)
9
Anti-interference (4)
9
Occupied space (9)
1
Complexity (9)
1
Driven efficiency (9) 4
Responding speed (4) 1
Total
35
(188)
Ratio to IFR (%)
49
(43)
1
1
4
4
1
1
1
9
22
(96)
31
(22)
4
1
1
1
4
4
4
4
23
(154)
32
(35)
1
4
9
4
4
1
1
9
33
(155)
46
(35)
4
9
1
9
4
1
1
1
30
(192)
42
(44)
IFR
4
9
4
9
9
9
9
9
9
9
4
9
1
9
9
9
49 72
(259) (441)
68
(59)
the mechanism. Although numerous mechanisms can

perform the desired elevation, the more components
contained in the mechanism, the more difficult is
the mechanism to manufacture within the requirements constraints. Using brainstorming, six candidate
mechanisms are sketched as seen in Figure 3.
Comparing the IFR with the six alternatives in the
decision matrix is shown in Table 3.
The comparison in Table 3 indicates that the fourbar linkage is much closer to IFR than other mechanisms. However, this mechanism still obtains a low
score on driven efficiency because when the tongue
jack tows the horizontal bar from the farthest end to
the direction shown in Figure 4.
Since the angle approaches 0, the component
force T12 may not be large enough to lift up the platform with a user on it. It is even worse when the user
is standing on the AB part of the platform, because the
torque arm of the users weight is longer than that of
the supporting force at B.
Concept affordance is introduced to help evaluate
and resolve this problem. Affordance is proposed by
Gibson (1979) in ecology and applied by Maier &
Fadel (2009) in mechanical design to represent the
potential interactions between user and environment.
In this project, the platform is the user-interface and
one of its main expected affordances is to afford
Figure 5. Mechanical analysis and comparison.
supporting the user walking and the impact force

when changing the slope of the treadmill. However,
according to the previous analysis, the affordance
walk-on-ability fails when the user is walking on AB.
So one solution to guarantee the whole platform offering the walk-on-ability is to extend the torque arm of
the supporting force, for example, change the position
of joint from B to A as shown in Figure 5.

After the modification, T12
is significantly larger
thanT12 , because:
A prototype implementing this mechanism has therefore been built.

3.5
Control subsystem
This subsystem includes three devices, a magnetic, a

digital counter, and a controller. The magnetic sensor is installed on the frame of the treadmill platform,
pointing from a certain distance at a flywheel concentric with the front axle of the belt. On the side of the
flywheel facing the sensor, a small iron patch is fixed
that can pass through the detectable zone of the sensor once per rotation. Hence when a participant walks
on the treadmill, the belt drives the rotating axles, and
consequently the patch triggers the sensor to generate a high-level voltage signal once per rotation. The
digital counter connected to the sensor can receive the
signal and display the total number of the signals in
decimal format on its screen. Meanwhile, the signal is
702
transferred to a C++ server in the computer through

an RS232 COM connection. The number of signals
and the perimeter of the patch orbit can be calculated
together, with the result indicating the moving distance
of the participant on the treadmill in a certain time,
i.e. the average moving speed in this distance. Then
these data can be uploaded from the C++ server to the
Google Streetview server to download the panoramic
photograph and altitude data from Google.
The gradient is calculated by comparing the altitude
change from one waypoint to the next with the distance
between the two neighboring waypoints. Next, the gradient is translated to the motor-control command. In
this project, the DC motor is controlled through pulsewidth-modulation (PWM) commands. To be specific,
the relationships of the control parameters can be
derived by a hypothesis:
Suppose two neighboring waypoints P1 and P2 in
Google Streetview have corresponding altitudes A1
and A2 . The distance between the two waypoints is S.
If simulating a user moving from P1 to P2 , the gradient
can be obtained as:
Table 4. Experiment to test the relationship between PWM

values and the speed of the tongue jack.
The exact function between the control command

PWM values and the motors rotating speed R cannot be derived, so we use a general form to represent
the function as:
*N/A means under this PWM the tongue jack cannot work.
Similarly, the unknown function between the R and

the elevating angular velocity of the mechanism can
be expressed as:
Thus,
Hence, the motor needs to run for t seconds to reach
the new altitude,
If the user needs to spend tr seconds walking through

the distance S on the treadmill, when t tr + 1, the
synchronicity of the simulation is acceptable.
However, the angular velocity W is difficult to measure directly because of its nonlinearity. One way to
solve this problem is to measure the linear extending and retreating speeds of the tongue jack, and then
derive the W. Thus, as experiment was created to test
for this.
Objective of the experiment: according to different PWM values, test how much the time the
tongue jack needs to extend and pull back 50 mm.
The tongue jack is installed under the treadmill and
works under the load of the treadmill and a 200 lbs
person walking on it. The results are shown in Table 4
and Figure 6.
PWM (%)
Counted time
100
90
80
70
60
50
40
30
20
15
10
10
15
20
30
40
50
60
70
80
90
100
6.7
7.9
8.8
11.1
12.1
16.3
19.7
34.6
50.8
N/A*
N/A
N/A
N/A
49.3
30.4
19.3
15.4
12.3
10.4
8.8
7.7
6.8
6.7
7.8
8.9
10.8
12.4
16.1
22.1
35.8
84.3
N/A
N/A
N/A
N/A
57.4
31.0
20.5
15.2
12.3
10.3
9.0
7.7
6.8
Average time
6.8
7.9
8.9
10.4
21.7
71.3
N/A
N/A
N/A
N/A
57.6
20.1
10.4
8.8
7.8
6.8
6.7
7.9
8.9
10.8
10.3
16.2
21.2
35.2
68.8
N/A
N/A
N/A
N/A
54.8
30.7
20.0
15.3
12.3
10.4
8.9
7.7
6.8
Based on the results of the experiment, it can be

concluded that the larger the PWM value is, the longer
time the motor can run in full speed and thus the faster
the tongue jack can work through the 50 mm. Furthermore, when the PWM value is under 50%, the
performance of the tongue jack becomes not steady
and is easily affected by the load; in contrast, when
the PWM value is over 60%, the tongue jack works
steadily and is not easily affected by the load.
After installation, the tongue jack can elevate the
platform of the treadmill from 8 to 0 , then the tongue
jack needs to extend about 66.2 mm, and the corresponding average angular velocity W to the different
PWM values can be calculated as seen in Table 5.
In practice, a person walking in speed 1 m/s needs
10 s to move through 10 m, the distance between
the two neighboring waypoints in Google Streetview.
According to the synchronicity requirement discussed
above, t tr = 1 = 11 s, and thus the PWM value
should be set over 80%.
3.6 VR subsystem
The VR subsystem created provides an interface
between a HMD; a Nintendo Wiimote or treadmill;
and Google Maps API. This subsystem is originally
based on Aki Mimotos open source project, Stationary Cycling through Google Streetview (Mimoto,
2001). Though this subsystem has been heavily modified so that very little of Aki Mimotos original code
remains.
3.6.1 User experience
The interface allows a user to feel more like they are
really in the street captured in the panorama by looking
703
Table 5. Calculate the average W from-8 to 0 , according

to the different PWM values from 60% to 100%.
PWM (%)
60
70
80
90
100
Time t (s)
Average W
16.3
0.5
14.3
0.6
11.8
0.7
10.5
0.8
8.9
0.9
Figure 7. A picture of a user interacting with the system

using the VR920 HMD and treadmill.
Figure 6. The relationships between the working time

extending or pulling back 50 mm of the tongue jack and PWM
values.
around naturally via HMD, rather than dragging the

panorama with a mouse. Figure 7 shows a picture of a
user interacting with the system.
If the user employs a Nintendo Wiimote to navigate,
then movement is as simple as pushing a button to
move forward in the direction they are viewing, rather
than double clicking or finding the arrow to click to
tell the panorama where they want to go. If the treadmill is used for navigation, then the number of clicks
from the magnetic sensor that counts the number of
rotations of its front axle is used to make travel seem
more natural. The magnetic sensor connects to the VR
subsystem server over an RS232 COM connection.
If the number of rotations multiplied by the distance
of the axiss circumference at a given waypoint exceeds
the distance to next waypoint in view, the user is moved
forward. It is hypothesized that this natural locomotion in a panoramic environment could add a greater
sense of immersion and contribute to the way users
comprehend the space within which they are traveling.
The user then explores Streetview in Full Screen
mode on the HMD in a first person perspective. Fullscreen mode is important to creating a higher sense of
presence by immersing the users view almost entirely
in the panorama, allowing the user to make out greater
detail in the image.
The user can move to a new panoramic environment if the system first determines that there is a
waypoint within 60 of their current view direction,
or, alternatively, if there is a waypoint within 30 of
the users walking direction as illustrated in Figure 8.
704
Figure 8. This diagram illustrates how the system determines potential paths for moving forward from the middle
of an intersection. The dots represent waypoints linked to
the current panorama. The light gray wedge represents 60
about the users current view direction, while the black wedge
represents 30 about the users walking direction.
3.6.2 High level software system design

The VR subsystem is divided into two main components: a C++ backend server that handles communication with the input devices, and an HTML/Javascript
frontend that provides the view of the panoramic environment and interacts with Google Maps API. The
input devices include, in one condition virtual travel
with a Nintendo Wiimote, and in another condition
natural travel facilitated via the treadmill, and a HMD
with inertial tracker. Figure 9 shows a diagram of the
software framework.
3.6.3 Head mounted displays
Two HMDs are currently compatible with the VR subsystem: the Vuzix iWear VR920, and the eMagin z800.
The VR920 was chosen as it is relatively inexpensive and commercially available. Its price will make
it easier for other users to access and use our system,
keeping in spirit with Googles Terms of Use policy requiring public availability. However, the VR920
requires frequent calibration. The eMagin z800, while
more expensive, does not require calibration on startup
and consistently gives better head tracking performance. Both HMDs provide head orientation tracking
data with three degrees of freedom given as yaw, pitch,
and roll. Both HMDs have a resolution equal to or
greater than 640 480.
3.6.4 Backend server
The C++ backend is organized as follows: One thread
handles communication with the HMD, tracking its
orientation as frequently as recommended by the displays documentation. Another thread communicates
the trackers position to the frontend through a UDP
multicast. Yet another thread listens for TCP connections, accepting one connection so that the user can
issue commands to the backend during execution, such
as to center the head mounted display. Finally, a dedicated thread of communication listens to two possible
travel interfaces, either the Wiimote button presses, or
Figure 9. High-Level System Architecture. The C++ Server

manages input devices, communicates with the Webpage
through the JNEXT TCP/UDP browser plug-in, and the Webpage communicates with Google Maps API to provide the
user output.
the treadmill motion sensor and maps them to virtual

travel in Streetview.
The HTML/Javascript webpage interface was
developed and tested on Mozilla Firefox 4.0.1, Apple
Safari 4.0.5, Internet Explorer 7.0.6001.18000, and in
a limited capacity on Opera 9.64. The webpage is the
graphical user interface that the user sees and uses
to navigate between Google Streetview panoramas of
real world locations connected by links. When a new
panorama is displayed, typically by loading the webpage or following a link, connections to the current
waypoint are requested from Googles API. For each
link, the linked panoramas metadata is requested, from
which the distance between the requested and current
waypoints may be calculated.
The webpage client uses Google Maps API version
2 instead of the newer version 3 API. We chose to
remain with version 2 since version 3 was released
mid-development, and version 2 has a number of benefits over the newer version. For some browsers, version
3 relies on two vanishing points with every panorama,
giving the image a warped perspective that we believed
would diminish the immersive quality of our system.
Version 2 also has a motion effect or blur when moving
between waypoints, giving the user an impression of
natural travel from one location to another, rather than
jumping between locations. Version 3, on the other
hand, just changes panoramas, with square portions
changing at different rates, which we also believed
decreased immersion (Bowman et al. 1997).
3.6.5 Networking and optimization
Whenever users enter a virtual environment, real-time
display updates and minimal lag in motion tracking are
essential to maintain immersion and a high sense of
presence. Lag between tracker information and camera manipulation has been thought to cause severe
disorientation and simulator sickness, although there
are mixed reports. While Draper et al. have found
only a significant decrease in performance without an
increase in simulator sickness with increased visual
lag (Draper et. al. 2001), DiZio and Lackner have
found simulator sickness to increase with visual lag
(DiZio et. al. 1997). To reduce this lag, we focused
705
on optimizing communication between the backend

server and the webpage. Although no direct measurements on performance time have been taken currently,
perceptually our system appears to run in real-time.
UDP multicast was chosen as the method to communicate the head trackers orientation updates to the
Javascript frontend webpage for a number of reasons.
Originally, we tried using an embedded http server in
the backend to provide dynamically returned pages to
the frontend. However, this forced the Javascript components to run in a loop rather than use a naturally
handled event driven structure. The data that has to be
communicated to the frontend was also to be sent frequently in small packets, where the reliable sending
of every packet was not important to the user experience. Therefore, the overhead incurred through full
http transfers was determined to be unnecessary.
To enable Javascript to handle sockets communication, a plugin was created based on the JNEXT
(Javascript Native Extension) framework. A plugin
using this framework was already available, but missing the ability to access multicast options for UDP
communication. In order to avoid the increased local
network traffic that UDP broadcasting would create
while maintaining the convenience of one way communication in which the backend emitter has no need
to know of the receivers address, we extended the
JNEXT sockets plug-in algorithm to include multicast
capability.
The webpage communicates with the backend
server through the JNEXT plugin. Panorama viewpoint updates are received by the webpage as JSON
strings through the UDP multicast from the backend
server, and commands to the backend are sent through
an opened TCP connection. Currently the only commands sent to the backend are to center the HMD to
zero yaw, pitch, and roll (reset heading direction) or
to change the color of the HMDs to view anaglyphic
stereograms. However, this framework is extensible,
so commands can be added to communicate with
hardware added to the system in future extensions,
such as novel methods of travel. Google Maps APIs
Streetview service handles updating panoramic viewpoints directly from yaw and pitch data, which our
webpage passes to it after applying appropriate offsets
to set the users initial position to zero.
4
DISCUSSIONS ON FUTURE PROTOTYPE
Due to the financial and technical constraints, there

are still various limitations in every subsystem of this
prototype. First of all, the treadmill is non-motorized,
working based on the difference of coefficients of friction between the two surfaces of the belt as seen in
Figure 10:
As introduced previously, the difference between
the reversed friction forces on the two sides of the
belt, i.e. f2 f1 , drive the belt to move backward.
Therefore, when simulating going uphill, since a component N12 of the force N1 exerted by gravity on the
user is parallel with the belt motion, that force helps
Figure 10. The illustration shows in this prototype why the

user walking downhill cannot be simulated and why walking on horizontal plane is more difficult than uphill. N1 : the
pressing force on the platform caused by the users gravity;
N11 and N12 : two component forces of N1 ; f2 : the friction
force between the users shoes and the upper surface of the
belt; f1 : the friction force between the lower surface of the
belt and the supporting board.
drive the belt backwards. The user is thus fooled in

finding it easier to walk on the belt going uphill than
walking on a horizontal plane. This feeling, however,
sharply contradicts with the reality that the user should
exert more effort when walking uphill than walking on
horizontal ground. Similarly, going downhill cannot be
simulated by this prototype since under this condition
the component force N12 is in the same direction with
f1 and N12 + f1 may exceed f2 . To solve this problem,
the next generation of the prototype should be installed
with a motor to control the rotation of the axle, however, the start and stop issue will be more difficult to
resolve then.
In addition, a normal DC motor is used to drive the
tongue jack, with its rotating velocity and direction
controlled by PWM commands. Both the structure of
the motor and the PWM method are inaccurate, so the
error in the elevation can be as high as 1 . A stepping or
servo motor can solve this problem; however, since a
user can hardly feel the 1 error, whether it is necessary
to change the motor is still open to discussion.
As for VR programming, a new software feature currently in development is the integration with
WebGL graphics. Adding WebGL enables the system
to display two or three dimensional models and annotations on top of the panorama to augment the VR
exploration of real world places. Potential applications
include the addition of virtual agents to act as tour
guides, annotations placed in three dimensions to aid
in navigation and wayfinding, or architectural models
of planned building projects to preview them as they
would appear in their physical location. This feature
currently has only been tested to work in recent versions of Chrome and Mozilla Firefox, since WebGL
is a new technology and not yet implemented widely.
New versions of the system currently have the capability of displaying models with WebGL, but models
706
for specific uses with our application needs are still in

development.
Last but not the least, since the visual information
perceived by the users eyes is different from the reality,
it is necessary to do research on two aspects concerning
how the future prototype affords the user to coordinate
between the two worlds. First, the user cannot see the
slope change in the VR environment, but can feel the
change in the physical world. Such an incoordination
may result to the users dizziness and even falling down
from the treadmill. So the prototype should afford the
user to keep balance in the experience. For the current
prototype, two handles on the treadmill can offer such
an affordance; but also constrain the natural walking
gestures of the users arms and make the user uncomfortable. The future prototype might enable to display
the 3D VR images that can represent the visual information of the slope change to the user; or, it can be
equipped with a camera that projects the users body
images into theVR so that the users brain can deal with
the coordinated information from the sensory organs.
Second, when the user is about to select a route in a
crossing in the VR environment, the VR environment
can afford the users vision to turn to a new direction but the treadmill cannot offer the turn-ability to
the users body in the physical world. The user is thus
suggested to use a Wiimote to select the route or turn
the head to the desired direction and the tracker in the
HMD can guide to the route. However, when the separation angle between two neighboring routes is smaller
than a certain degree, it is difficult to select the desired
route efficiently via either of the two methods. Therefore, adding more degree of freedom to the treadmills
belt may be a way to realize the turn-ability in the
physical world.
5
CONCLUSIONS
This paper presents the detailed prototyping process of

the VR treadmill, ranging from identifying the objective and the requirements list to designing the three
specific subsystems. In this process, prototyping techniques act a vital role in modeling designers ideas.
Specifically, for the mechanical subsystem, by evaluating and comparing the prototypes, we clarified the
advantages and disadvantages of different plans and
then integrated the advantages to the final solution,
realizing the optimization. Also, a VR subsystem is
discussed that incorporates a server backend that sends
messages from the treadmill or Nintendo Wiimote and
HMD to the client webpage. The client webpage then
retrieves and incorporates data from Google Maps API
to give the full user experience. Finally, the affordance
based approach is used to analyze the limitation of the
current prototype and predict the evolving direction of
the new version.
REFERENCES
Altshuller, G.S. 1996. And suddenly the inventor appeared:

TRIZ, the theory of inventive problem solving. Worcester:
Technical Innovation Center.
Altshuller, G.S. 2000. The innovation algorithm: TRIZ, systematic innovation and technical creativity. Worcester:
Technical Innovation Center.
Bao, J.S., Jin,Y., Gu, M.Q.,Yan, J.Q. & Ma, D.Z. 2002. Immersive virtual product development. In: Journal of materials
processing technology: 129 (2002) 592596. Elsevier.
Bowman, D.A., Koller, D. & Hodges, L.F. 1997. Travel in
immersive virtual environments: an evaluation of viewpoint motion control techniques, Virtual Reality Annual
International Symposium: vol., no., pp. 4552, 215,
15.
Buchenau, M. & Suri, J.F. 2000. Experience prototyping.
ACM 158113-2190/00/0008.
Cooper, K.G. 2001. Rapid prototyping technology, selection
and application. New York: Marcel Dekker.
DiZio, P. & Lackner, J.R. 1997. Motion sickness side effects
and aftereffects of immersive virtual environments created
with helmet-mounted visual displays, NATO Research and
Technology Organisation Human Factors and Medicine
Panel Meetings Proceedings: 59 Dec. 1997, vol. 54, no.,
pp. 11-1-11-3.
Draper, M.H. & Viirre, E.S. & Furness, T.A. & Gowron, V.J.
2001. Effects of image scale and system time delay on simulator sickness within head-coupled virtual environments,
Human Factors: vol. 43, no. 1, pp. 129146.
Liou, F.W. 2008. Rapid prototyping and engineering applications: a toolbox for prototype development. Boca Raton:
CRC Press.
Gibson, J.J. 1979. The theory of affordances. In: The ecological approach to visual perception. Houghton Mifflin,
Hopewell: 127143.
Google group. 2011. Google Maps JavaScript API V2:
http://code.google.com/apis/maps/documentation/javascript/ v2/
Maier, J.R.A. & Fadel, G.M. 2009. Affordance-based design
methods for innovative design, redesign and reverse
engineering. In: Res Eng Design: 20: 225239, DOI
10.1007.
Maier, J.R.A., Sachs, R. & Fadel, G.M. 2009. A Comparative Study of Quantitative Scales to Populate Affordances
Structure. IDETC/CIE 2009. DETC200986524.
Mimoto, A. 2009. Stationary Cycling Through Google
Streetview: http://bako.ca/streetview-riding/
Pahl, G. & Beitz, W. 2007. Engineering Design, a Systematic
Approach. London: Springer.
Ruddle, R.A., Payne, S. & Dylan, J.M. 1999. Navigating
large-scale virtual environments: what differences occur
between helmet-mounted and desk-top displays? Presence: Teleoperators and Virtual Environments: vol. 8, no.
2, pp. 157168.
Suma, E.A. & Finkelstein, S.L., Reid, M., Babu, S.V., Ulinski, A.C. & Hodges, L.F. 2010. Evaluation of the cognitive
effects of travel technique in complex real and virtual environments, Visualization and Computer Graphics: IEEE
Transactions on, vol. 16, no. 4, pp. 690702, JulyAug.
Zanbaka, C.A., Lok, B., Babu, S.V., Ulinski, A.C. & Hodges,
L.F. 2005. Comparison of path visualizations and cognitive measures relative to travel technique in a virtual
environment,Visualization and Computer Graphics: IEEE
Transactions on, vol. 11, no. 6, pp. 694705, Nov.Dec.
Zhang, R., Noon, C., Winer, E., Oliver, J.H., Gilmore, B.
& Duncan, J. 2007. Immersive product configurator for
conceptual design. IDETC/CIE 2007. DETC200735390.
Altshuller, G.S. 1984. Creativity as an exact science. India:

Gordon and Breach Science Publishers.
707

Numerical simulations applied in a protocol for virtual prototype

of a femoral prosthesis stem fatigue life test
C.A.R. Laureti, D.T. Kemmoku, P.Y. Noritomi & J.V.L. Silva
Renato Archer Information Technology Center (CTI), Campinas, So Paulo, Brazil
ABSTRACT: This paper proposes the use of numerical simulations to a protocol for virtual prototype, reducing
the amount of mechanical fatigue tests for certification of femoral prostheses stem. A commercial prosthesis
was obtained from manufacturer and the assembly of standard experimental apparatus was modeled. Construction material of the prosthesis and load applicator was assumed as annealed steel as ISO 5832-1 for medical
application. Fixing structure was assumed as acrylic resin. Static analysis was linear due to the mechanical
behavior expected. Generated mesh on the prosthesis was structured tetrahedral triangular face element with ten
nodes. Stress tensor was applied to S-N model fatigue solver to obtain the life and damage distribution along the
prosthesis. Data leads to conclude that numerical simulation protocol provides similar results than mechanical
test provided by the manufacturer. Despite of it, there is no conclusive comparison to testify if the numerical
simulation is fully compatible with the experimental essay.
INTRODUCTION
The use of computational simulation has improved

in engineering applications, especially in the design
development field. The availability and growing up
computational power provided by recent IT (Information Technology) developments are among the main
reasons for this trend. Finite elements analysis is one
of the most applied methods to simulate mechanical behavior, with many different commercial and
research solutions available.
Finite elements method is based on mathematical models, which are developed to fit some specific
mechanical behavior within restrict variation limits
and subjected to prescribed and well known boundary conditions. These characteristics make such kind
of model what is called a deterministic model, which
means that any response can be directly accessed by
solving the governing mathematical equations under
correctly imposed boundary conditions set. While
it is very interesting for conventional engineering
mechanical systems development, due to the intrinsic control applied to this kind of application, it is
too much restrictive for more uncontrolled fabrication
processes, like biologic structures generation.
Usually, engineering products are designed to be
easy to fabricate and reliable in use, provided it is
demanded that all the products have similar performance in working activities. These characteristics
leaded the engineering processes to be more controlled
and predictable, allowing the use of deterministic
mathematical models as good representations of the
system behavior, considering their result as mean
values of a very small deviation distribution.
Despite the fact that this is a very useful approach,

not always it is possible to assure the hypothesis of
small dispersion of the system variables. Especially for
highly complex systems, it is very difficult to control
or predict the behavior of critical variables, due to their
dependency of many parts of the system, each one
subjected to its own variability (Wolfram et al., 2009).
This behavior is very well noticed in the experimental engineering, when simplified systems are put
to work monitored by sets of sensors, in order to reproduce some specific behavior of a more complex system
under controlled environment. The experimental setup
is idealized to reduce variability while keep the conditions similar to the real operation conditions of the
whole system. With this kind of experimental approach
it is possible to see the influence of variability and if
it is small enough to enable the use of mean values
as representative values for the system behavior (Lee
et al., 2005).
Even in very well controlled experimental setups
there is a possibility to see the rising of variability
of some variables and the results start to become too
much different from their mean values. In these cases,
it is very difficult to use computational simulations
based on deterministic mathematical models to predict
or study any occurrence, due to the lack of variability
in the behavior of these mathematical models.
In order to include some variability in finite elements analysis models, it has been proposed the use of
statistical variability approach to be applied to some
selected input variables (Fazilat et al., 2002). This
variability has some statistical distribution and can
be applied to selected input variables using different
techniques. Design of experiments (DOE) is one of
709
these techniques and has advantages when combined

to finite elements method approach.
One of these advantages is the capability of automation, due to the availability of implemented DOE
software as well as finite elements software that can
be put to work together in a automated software implementation, allowing the design of sets of experiments
and their execution through the substitution of each
condition prescribed by the sets to the finite elements
model in order to generate a specific result.
Other advantage is the availability of multiple
DOE approaches, in order to generate the sets of
conditions to be varied (Eriksson, 1999). They can
be as direct and computational demanding as
the full factorial approach or something more
mathematically simplified as Taguchis distribution,
with less sample points base in some special
hypotheses.
The reduction of the amount of simulations,
based on the use of special sample points choices,
can be very interesting provided the simulations can
be very time consuming. In many cases, the time
demanded by a computational simulation model to
run is much less than the time demanded to its
counterpart experimentally, but provided the computational resources are limited, there is always room for
reduction.
Recent advances in DOE software and their integration with finite element applications have increased
the possibilities of use of statistical approach mixed
with standard computational mathematical model simulations (Kadirgama et al., 2009). This combination
intends to combine real world variability with numerical mathematical model precision, aiding to bring the
deterministic models more close to the experimental results, especially when there is great dispersion
in fundamental characteristics of the model (Novak
et al., 2005).
MATERIAL AND METHODS
In order to develop the work, high power computational resource was demanded as well as software
applications for finite elements modeling (FEM) and
design of experiments (DOE) approach.
The high power computing was provided by an
application server that is part of DT3D/CTI (Threedimensional Technologies Division of the Renato
Archer InformationTechnology Center) infrastructure,
which is a double processor six (6) core per CPU, 32GB
RAM memory, with Microsoft Windows 7 64 bit
Professional Edition installed.
The finite elements modeling software was provided by NEiSoftware company, which solution is
called NEiNastran , working with pre and post processing by FEMAP . This solution is fully expansible
and capable to work starting from the simplest models like linear isotropic model until more complex
and timing consuming ones like nonlinear anisotropic
dynamic models.
The fatigue solver was provided by ThorusScisoft

company and is called Fatigo , which solution is capable to calculate life in fatigue based on a variety of
calculation methods and capable to handle with fatigue
with or without considering plasticity. This solution is
fully configurable and integrated with finite elements
results files.
The DOE software was provided by Esteco
company and is capable to handle with multiple
variables for multipurpose optimization and variable
sensibility analysis procedures. This software is called
Mode Frontier and is very flexible and easily adaptable in a way that is possible to use it as a command
center to start multiple finite elements analysis processes automatically and then collect their results,
grouping them in a more suitable way for the analysis.
The proposed method was to develop a standard
FEM analysis of an experimental setup for fatigue
essays in order to obtain the stress distribution. Based
on this stress field, using a specific fatigue solver,
obtain a fatigue failure profile showing the life expectation in number of cycles. Both the solvers were
integrated using mode Frontier capabilities and the
resulting data was collected by mode Frontier in
order to make it possible to organize it and gather
information.
The post processing of finite elements using
FEMAP was used in order to show color maps for
the stress distribution and fatigue life. Post processing
for statistical data evaluation was provided by mode
Frontier.
The analyzed prosthesis was developed and
produced by INCOMEPE Materiais Cirrgicos, a
Brazilian company which is manufacturer of surgical
and medical devices. The company provided the CAD
model and the material properties for a femoral stem
that was previously approved for commercialization
and has passed through the fatigue tests. The company
needs to make the CAD model to produce the forging
mold and the mechanical tests on the materials used
to assuring quality of the products.
The CAD model was imported to Rhinoceros to
build the other structures like the head of the prosthesis, supporting material and load applicator. Other
structures could be substituted by boundary conditions in finite element analysis. The final assembly
(Figure 1) was exported to FEM analysis software.
The simulation can be separated in two big areas.
The first is a FEM static analysis of the assembly
according to the standard. The second analysis is the
evaluation of the fatigue using Fatigo.
The test configuration for the first step respects the
ABNT NBR ISO 7206 Implants for surgery Partial
and total joint prostheses standard for the test configuration, settings and materials. This is a Brazilian
standard based on ISO standards and is separated in
six parts. The parts used are the Part 4 and the Part 6
because they are related to endurance of the stemmed
femoral components that is the focus of the research.
The standard defines the material to be used in
the prosthesis (ABNT NBR ISO 7206-1) and the
710
Figure 2. Constraints and load on experimental apparatus

for static structural finite element analysis simulation.
Figure 1. Experimental assembly based onABNT NBR ISO
7206-1 standard.
material where the prosthesis will be mounted in as far

as its mechanical properties. The load cycle must be a
sinusoidal with a frequency between 1 Hz and 10 Hz.
The minimum magnitude is 300 N with a maximum
of 2300 N. The position of the prosthesis in relation to
loading mechanism can be found in the figure 2 and
three values must be respect to position the prosthesis: alpha = 10 1 degrees; beta = 9 1degrees and
D1 = 67.64 mm for this case. The test must stop when
one of those events arise:
Runoff or loosening
Sample crack
One million cycles
The fatigue analysis was made using static analysis

and then taking the stress tensor of the elements to
calculate the life and damage of the structure.
The NEiNastran of NEiSoftware was used to
make the finite elements analysis. The simulation consisted in applying the load over the head of the prosthesis with a completely fixed basis. The load applicator
device must reduce non vertical loads. In order to represent this condition, the cylinder of the base was fixed
in a cylindrical restrain, which means constrained radials and avoid cylinder rotation around its long axis.
The top was fixed in vertical direction, allowing the
movement at other directions. The load (2300 N) was
applied in the bottom of the cylinder compressing the
structure. This configuration of boundary conditions
proved to be the right one, provided the use of the
cylindrical restrain over the applicator would not allow
the lateral movement of the head that must be allowed
according to the standard.
The interaction between the structures was considered glued between the support material and stem, and
simple contact which allows separation and sliding
between the structures in the stem and. The sliding contact between the mounting and the prosthesis
is due to high difference between the steel and the
Figure 3. Schematic flow diagram of modeFRONTIER

routine.
polymer. It was considered the axial and tangential

sliding for low mechanical interaction to allow sliding
when high load is achieved.
The model had 230.029 elements using 10 nodedtetrahedrical elements. The six values of the
stress tensor, associated with the prosthesis, were
exported to Fatigo to evaluate the damage and
life of the structure. The software takes into account
the endurance of the structure until the initialization of a crack using two calculation methods: stress-based (S-N) and strain-based (-N). The
choice of which method to use depends upon
the magnitude of the stress found in the static
analysis.
These steps were manually executed and lead to
specific values of life. In order to study the influence of variations in input values on the output results,
it is necessary to automate this process in order to
make it easy to execute multiple analyses based on
multiple combinations of variations in the input
values. Slinding
This automation was provided by mode-FRONTIER
features which made capable the design of a graphical
integration between the applications and the retrieval
of results and its visualization in a more organized way.
The sequence of steps performed in mode Frontier
can be summarized by the following flow diagram
shown in figure 23.
711
Figure 6. Overall Student pie chart.
In this fatigue analysis the results showed the

expected behavior where the minimum life is where
the model has higher values of maximum principal
stress. So the main result of the fatigue analysis is the
life contour plot. It is possible to realize that even the
lower values of life on the stem achieved the objective
of the project that was to be higher than 5000000 of
cycles according to ABNT standard.
Provided the experiment has intrinsic statistical
variability it was applied the DOE approach to try to
model it. One of the results of DOE application is the
information about the influence of each variable in the
variability of response (Figure 6).
Figure 4. Maximum Principal Stress on stem prosthesis.
Figure 5. Fatigue Life on stem prosthesis.
RESULTS
The most important result of finite element analysis is

the complete stress tensor that will be used as an input
to routine of fatigue calculus. The maximum principal
stress values are use as well to evaluated the critical
tension region. In the Figure 4, the maximum principal
stress contour is shown for the femoral stem.
The maximum principal stress contour on the stem
showed a typical behavior of a beam in flexion where
there is a tension in one side and compression on the
other side. This is due the geometry of stem and its
constraints. The maximum and minimum values for
maximum principal stress were 250 MPa of maximum
and 50 MPa of minimum and both were found on the
edge between the stem and the beginning of cylinder.
With respect of the simulation of fatigue, it gives as
results damage and life for each element of the mesh
based on signaled Von Mises obtained from the complete stress tensor. Below, in figure 5, it is shown the
life contour plot for the femoral stem.
CONCLUSION
Based on the results obtained so far, it can be concluded

that this numerical simulation procedure for substituting one of experimental essay for certification of a
femoral stem showed applicability. The evolution of
this procedure can lead, in the near future, to the development of a standard for certification authority. The
main issue that prevents the direct use of this numerical
analysis results to the certification is the lack of information regarding to the mechanical properties of the
prosthesis and other involved materials. The fatigue
analysis needs specifically materials properties that
even material supplier of the prosthesis do not have.
So for more accurate results and then the certification
use it is necessary to make fatigue experiments using
stem prosthesis to calibrate the numerical analysis.
One way to reduce the effect of lack of information
is to use DOE approach. The results show large influence of the variation of Youngs Modulus followed by
reliability correction factor on fatigue calculation.
Due to the requirement of the standard, it is common that the only result provided is a report that the
stem does or does not fail after 5000000 load cycles.
Then the fatigue experiment must be made until stem
fails and then provide the number of cycles that the
prosthesis can handle with.
The final conclusion is that finite element analysis itself cannot represent the real experiment due
lack of precision in the input variables. In order to
712
reduce this limitation, DOE technique can be applied.

The result is a statistical model which can be used
to study the sensibility of the input variables as well
as to develop multiple analyses trying to reproduce
variability.
REFERENCES
ABNT NBR ISO 7206-1. 2008. Implants for surgery Partial and total hip joint prostheses Part 1: Classification
and designation of dimensions. Associao Brasileira de
NormasTcnicas. 2nd ed. Brasil. 17p.
Eriksson, M. 1999. Finite Element Analysis and Design of
Experiments in Engineering Design. Division of Machine
Design, Department of Design Sciences, Lund Institute of
Technology, Lund University, Lund, Sweden. 122p.
Fazilat H.D., Judith R.M. and Richard M.A. 2002. Statistical methods in finite element analysis. Journal of
Biomechanics. 35. pp. 11551161.
Kadirgama, K.; Noor, M.M.; Rahman, M.M.; Harun, W.S.W.

and Haron, C.H.C. 2009. Finite Element Analysis and
Statistical Method to DetermineTemperature Distribution
on Cutting Tool in End-Milling. European Journal of
Scientific Research. Vol. 30. No. 3. pp. 451463.
Lee, W.C.C. and Zhang, M. 2005. Design of Monolimb
Using Finite Element Modelling and Statistics-Based
Taguchi Method.Clinical Biomechanics. 20(7). Elsevier.
pp. 759766.
Novk, D.; Vorechovsk, M.; Lehk D.; Rusina R.; Pukl,
R. and Cervenka,
V. 2005. Stochastic nonlinear fracture
mechanics finite element analysis of concrete structures.
Millpress. Rotterdam. pp. 781788.
Wolfram, U.; Schwen, L.O.; Simon, U.; Rumpf, M. and Wilke,
H-J. 2009. Statistical Osteoporosis Models UsingComposite Finite Elements: A Parameter Study. Journal of
Biomechanics. pp. 19.
713

BioCAD techniques: Example on maxilla for rapid expansion simulation

D.T. Kemmoku, C.A.R. Laureti, P.Y. Noritomi & J.V.L. Silva
Renato Archer Information Technology Center (CTI), Campinas, So Paulo, Brazil
ABSTRACT: This paper aims to show an overview of a technique used to model the bone structures of the
central region of the face, which involves the maxilla, some teeth and an expander which were used for FEM
analysis. This protocol intend to represent these structures in a more adaptable and intelligent way, in order to
enable their use in a second case where similar structures are demanded, thus making the job more quickly and
efficient. Modeling by this method, it is essential to use software with NURBS representation due to ability to
represent the anatomical curves in biological structures. The BioCAD modeling is initiated by the generation
of curves located in the main landmarks, indicated by the researchers of the area. The positions of these lines
should produce a clean geometry which facilitates the mesh control. Finished the maxilla, others structures are
modeled using the same technique having the maxilla as reference.
1
INTRODUCTION
This work presents the use of BioCAD modeling developed at the Center for Information Technology Renato
Archer, located in Campinas-Brazil.This methodology
has been used in the models developed to supporting
dentists and physician researchers.
The advancement in technology has increased the
ability to study biological structures using computational tools. It is therefore necessary to represent these
structures in an appropriate manner. Differently of
a mechanical component found in the industry, biological structures have landmarks that identify and
segment regions according to their functionality and
anatomy. The modeling became one of the most important steps towards the development of these kinds of
studies, since the considerations and simplifications
are determined at this step (Sun et al., 2005).
The amount of anatomical details presented in a
structure and their high geometrical complexity, make
necessary to use a technique of representation with
some level of simplification but still representative and
loyal to the geometry. That technique of CAD modeling of those structures that allows the identification of
these landmarks and is enabled to be used in finite element analysis software was developed in conjunction
with researchers in the areas of health and resulted in
a technique named BioCAD.
Using this method, the curves used to model the
structure promotes the recognizing of the main landmarks and turns the mesh control easier. The basic
reference to this kind of modeling comes usually from
three ways: engineering images of three views of the
structure imported to the software and used as a background; three dimensional models reconstructed from
a CT-scan and handmade modeling (Correia et al.,
2009; Tawhai et al., 2004; Silva, 1999; McInerney
et al., 1994). For this paper, the model was made
over a stereolitography mesh based on a CT-scan

reconstruction of a healthy maxilla used in a simulation
of a fast expansion in the treatment of a posterior cross
bite (PuiLai, 2005). The modeling was made using the
software Rhinoceros from Robert & McNeel where
the NURBS (non uniform rational b-splines) representation method has shown very flexible to represent
such structures.
2
MATERIAL AND METHOD
The first step of the modeling is to understand the

goals of the analysis that will be performed. At this
case, the objective of this analysis is to understand
the stress distribution resulted from a rapid maxillary
expansion when the posterior cross bite is observed.
One of the treatments to this abnormally is to insert
a maxillary expander in order to force the disjunction of the palatine bone. This expander is fixed at the
first premolar and the first molar of each side of the
maxilla and it is activated by a screw that increases
the distance between the halves of the maxilla. The
expander is not symmetric in order to apply different
loads to each side of maxilla. One side of the expander
has an acrylic polymer supporting part of the palate to
force an unsymmetrical growth of the involved bones
(Figure 1).
With the purpose to reduce the quantity of variables present in the analysis, the maxilla is symmetric
and the teeth are a mirror of the other side. In this
way, the stress distribution will be due only to the
unsymmetrical loads from the expander and not from
the unsymmetrical bone. With all the considerations
discussed, the modeling is started importing the threedimensional reconstruction obtained from computed
tomography scan (Silva, 1999). The segmentation of
the desired region was placed close to the origin of
715
Figure 1. Model to be transformed into BioCAD.
the coordinate system and the plane of symmetry is

aligned with some plane of the coordinate system.
The next stage is to place the reference curves over
the STL mesh. There are three types of curves which
each level characterizes the importance in the model
representation. The primary curves are placed over
the main landmarks such as bone sutures, process
and foramens. It helps to visualizing and recognizing
structures when the solid is imported to FEA software.
The secondary curves is used for assist the primary
curves to determine the surfaces edges and the limits
of the model. Spreading in a way to form quadrangular and triangular pattern, it will delimit the boundary
of the surfaces. The third type of the curves named
auxiliary curves intend to standardize the surfaces
reducing the complexity of the surface created. This
kind of pattern is better to form the surfaces using the
Rhinoceros and makes easier to control the transition
and the size of the meshes. These lines form a threedimensional silhouette of the structure under study,
and it is the basis for the creation of areas that united
form a solid provided they do not have free edges.
The previous study of which structures will be present
interferes where the secondary curves will be inserted.
The secondary curves must respect some rules to create a clean geometry i.e. avoiding the formation of
short edges, sliver surfaces and high order complexity
surfaces. This pattern can be observed in the figures 2
and 3. The primary curves are represented in continuous lines, the secondary curves are represented in
dashed lines and the auxiliary curves are represented
in dotted lines. The primary and secondary curves are
essential in the model because they define the contour
of the created surface then essential to represent the
geometry.
The lines over maxilla spread out in a way that the
main lines is placed delimiting the contours of the alveolar process, palate edges and the outline of the of
the zygomatic arch process. The secondary lines starts
from the symmetry line of the palate, passing through
the region of alveolar process demarcating the region
corresponding to each alveoli teeth and moving toward
Figure 2. Bottom view of the distribution pattern of curves

over maxilla.
Figure 3. Front view of the distribution pattern of the curves

over maxilla.
the eye orbit. The final external surfaces of the model

can be observed in the figure 4.
The number of landmarks to be represented depends
of the computational capacity and the focus of the
study. At this case, the limit number of the element
to be used is around 160 thousands elements. Therefore, very thin structures like periodontal ligament,
enamel and pulp are not modeled and are substituted
by adjacent structures. In the tooth, the pulp and the
enamel are substituted by dentin because it has a bigger volume. If the focus of the study was a dental
restoration, surely these structures would be modeled
in the analysis and the portion represented of maxilla
reduced.
For the representation of structures not present in
the reference mesh, it can be modeled applying surface offset from the created solid. It is used to modeling
the trabecular bone and the alveolus of the tooth in the
maxilla. Using the information found in the literature,
716
Figure 6. Teeth, gingival tissue and alveolus of the cancellous bone modeling.
Figure 4. BioCAD of the maxilla.
Figure 7. Final model with all the structures.
Figure 5. Cancellous bone modeling using offseted curves

from the cortical curves.
or making a second segmentation with the cancellous

bone but with loss of other structures, the cancellous
bone is modeled. If the information used is from the
literature, this information is based on the thickness
of the cortical bone at each region. This thickness can
be found making an offset of the original surface indicating the desired distance. Using the same pattern of
curves, the cancellous bone is modeled and added in
the model (Figure 5).
The offset of the surface is used in many cases
where it is need to model structures like dental alveoli
and the gingival tissue where the expander is supported (Figure 6). Finishing the maxilla, the next step
is the modeling of the involved teeth: first premolar and the first molar where the expander is fixed.
The teeth are simplified once the aim of the analysis is the stress over the maxilla. Details like valleys
and cusps at the teeths crown are represented without
effort in order to saving elements. The teeth are represented only by dentine because other components like
enamel and pulp just do not have an expressive volume

and or representative mechanical properties for justify
its representation. The last component modeled is the
expander and is used a real expander as a template. So
it is possible to copy thickness of the metallic support
and the positions of the fixation over the tooth. Conventional technique to model is enough once it is not
a highly complex structure. The resulting surfaces are
low order complexity surfaces, thus easy to mesh controlling. Some loose of details in the tridimensional
reconstruction is a consequence of the smoothing over
the reference lines like observed.
After finished the model, the other half of the model
is created mirroring the maxilla, tooth and the metallic
part of the expander (Figure 7). For an easier declaration of the contact pairs, the concerned surfaces are
splitted. Places where the type of contact is defined
as welded, we can split the surfaces exactly over the
regions where the structures get in touch. It reduces the
file size and complexity of the analysis generated. For
finalize the BioCAD model, more marks are needed
on surface to implement the boundary condition of the
problem.
717
element analysis had proved the unsymmetrical

stresses distributions on the jaw due to acrylic part
of the expander. The authors realized that the STEP
214 CC2 is the extension with fewer problems when
exporting models to FEA software.
4
CONCLUSION
This modeling method allows obtainment of models quite versatile and applicable in different studies.
As an example we have an expansion of the maxilla
using Le-Fort osteotomy (Figure 9) procedure using
the same model as base for analysis. The only difference is the insertion of cuts in the maxilla to simulate
the Le-Fort technique. Some fault of modeling can be
found in regions of intersection of structures modeled
separately, for example: the cancellous bone when is
modeled separately from the alveoli of the teeth. In this
case, a non clear geometry is obtained (Figure 10). At
some cases, it can be fixed using specific commands
found in the FEA software like virtual topology in
ANSYS or combine surfaces in Nastran .
The quantity of curves used to form the model
depends on the purpose of the analysis and the total
size of the model, but it is better to have more lines
than less, since the FEA software can disregard some
curves easier than create.
Figure 8. Mesh created over the maxilla and teeth.
REFERENCES
Figure 9. Use of the model in another case study of maxilla
expansion with LeFort osteotomy.
Figure 10. Surfaces with sliver edges and thin sections.
RESULTS
The model reached the goals of the proposal of the

BioCAD protocol as a model with an easy control of both mesh size and its distortion (Figure 8).
The pattern of the curves allows the use of mapped
mesh using tetragonal elements, but it is rarely used
due to complexity of the solid. The results of finite
Correia, A., Piloto, P., Campos, J.C.R. and Vaz, M. 2009.

Finite element analysis of the mechanical behavior of a
partially edentulous mandible as a function of cancellous
bone density.
McInerney, T. and Terzopoulos, D. 1994. A Dynamic Finite
Element Surface Model for Segmentation and Tracking
in Multidimensional Medical Images with Application
to Cardiac 4D Image Analysis. Journal of Computerized
Medical Imaging and Graphics, pp. 125.
PuiLai, K. 2010. Anlise das tenses resultantes das disjuno rpida de maxilla em estruturas esqueletais e dentais utilizando um disjuntor modificado para mordidas
cruzadas unilaterais posteriores verdadeiras pelo mtodo
de elementos finites. Thesis. So Leopoldo Mandic Post
graduation Center.
Rev. odonto cinc., 24(1), pp. 2227.
Silva, J.V.L. 1999. Prototipagem Rpida: Conceitos e Aplicaes. Campinas: Centro de Pesquisas Renato Archer
CenPRA.
Sun, W., Starly B., Nam J. and Darling A. 2005. BioCAD modeling and its applications in computer-aided
tissue engineering. Computer-Aided Design. 37. pp.
10971114.
Tawhai, M.H., Hunter, P., Tschirren, J., Reinhardt, J.,
McLennan, G. and Hoffman, E.A. 2004. CT-based geometry analysis and finite element models of the human and
ovine bronchial tree. J Appl Physiol. 97. pp. 23102321.
718

Finite Element Synthesis

N. Oxman
Mediated Matter Group, MIT Media Lab
ABSTRACT: Finite element applications have customarily been used as a means for the analysis of constantproperty prototypes prior to their physical fabrication. To a lesser extent they have been made instrumental for
purposes of design generation and fabrication of variable material-property prototypes. The work explores the
potential of applying finite element methods in the early stages of the design process and suggests a Finite Element
Synthesis (FES) approach to the design of physical prototypes fabricated with variable material properties. The
approach seeks to unify between analysis protocols and computational routines for design generation. With the
aim of achieving micro structural material property variation across the surface area and volume of a fabricated
component, the FES software environment is implemented in a design exploration for building skins modeled
after various natural tissues that demonstrate the variation of physical properties as a function of their performance
criteria. Two classes of explorations are suggested. The paper introduces the FES design approach, illustrates its
virtual methodological set up, and demonstrates the approach through design explorations recently exhibited at
the Museum of Science (Boston) and the Museum of Modern Art (NY). Future work into material-based analytic
routines for variable-property fabrication is suggested and its implications on the various fields of design are
reviewed.
1
INTRODUCTION
Virtual design generation for physical rapid prototyping, particularly for additive fabrication platforms,
typically assumes a given homogeneous material from
which the prototype is fabricated (Kruth, Leu et al.
1998; Hague, Campbell et al. 2003). Correspondingly,
most additive and subtractive digital fabrication technologies assume the use of materials characterized by
consistently uniform properties across their volume
or surface area (Zhang, Xu et al. 2002). Inspired by
formation processes in the biological world, where
properly variation corresponds with environmental
stimuli, this research promotes the integration of the
finite-element method in a generative design context,
supporting the controlled variation of micro-structural
material properties as part of form-generation and
digital fabrication processes. Material variations are
computed as functions of structural and environmental performance criteria through the development of
a material-based object-oriented finite-element software environment able to compute and assign graded
physical properties as continuous property gradients
of a functional component.
1.1
Background
The basic iterative algorithm for finite element optimization is based on reducing material concentration
where it is not required for purposes of structural or
mechanical performance, as defined by a given objective function (Zienkiewicz and Morice 1971; Johnson
1987; Hughes and Hughes 2000; Zienkiewicz and
Taylor 2000; Zienkiewicz, Taylor et al. 2005). A simulated solid block of any given material will undergo
formation changes during optimization resulting in
structural efficiency across various product-, or building scale (Figure 1). In this iterative process, stress
distribution is iteratively calculated and elements
with minimum stress values are removed. Such routine is implemented to cater for a specific objective
function (i.e. structural optimization) while assuming a relatively homogeneous material distribution
(Zienkiewicz and Morice 1971; Zienkiewicz and
Taylor 2000; Zienkiewicz, Taylor et al. 2005). The optimization algorithm may be combined with generative
routines such as Genetic Algorithms (GA) to assist
with fitness evaluation by implementing binary functions (retain, or remove, material) at the scale of the
element (Zwieniecki, Boyce et al. 2004).
Simplified, the general iterative algorithm for finite
element structural optimization operates as follows:
(1) Begin with a solid block of material and specified
load conditions; (2) Run finite element analysis to calculate stress distribution; (3) Check for any elements
at less than maximum allowable stress; (4) If such do
not exist, then terminate, otherwise; (5) Remove some
of the lowest-stressed elements; (6) Go to 2.
The effect, basically, is to remove redundant material until the minimum amount necessary to perform
the structural task is left-at which point, all remaining
material should be fully stressed; the algorithm then
terminates. Such typical algorithm assumes materials
of homogeneous properties (Figure 2). The only possible element conditions, then, are full of material
or empty.
719
material properties (Oxman 2011). Central to this

condition is the partitioning between design analysis, generation and fabrication, limiting the possibility of fast iterations between virtual finite element
environments to digital fabrication platforms.
Traditional design processes are instead streamlined, and generally devoid of platforms supporting property-variation based analysis and fabrication
(Oxman 2007; Oxman 2009).
2 AIMS & OBJECTIVES
2.1 Aims: FEM to FAB
This research explores the potential of [1] applying
the finite element method in the very early stages of
design generation and fabrication and, [2] incorporating heterogeneous material properties into FEM and
FAB environments. Coupled with generative protocols, this work suggests a hybrid, synthesis-analysis
approach to rapid fabrication platforms of heterogeneous material properties. The approach seeks to unify
between analysis protocols and computational routines
for rapid-fabrication, specifically additive fabrication
platforms.
Figure 1. Examples of various possible implementations

of FEM in product, and building scale. Analysis protocols
are traditionally devoid of generative-design and fabrication
capabilities.
2.2
Figure 2. The Mercedes-Benz Bionic Car Project. Image

illustrating the Soft-Kill Optimization (SKO) method
implemented on a solid block of material to form-find
the automobile optimal shape. The fabricated material is characterized by homogenous properties. Source:
http://www.carbodydesign.com.
However, the designer may decide to vary the material properties of a functional element (e.g. variabledensity metal foam), such that it can continuously vary
from 0 (empty, no material) to 1 (full of the strongest
available material), thus generating controlled heterogeneous material organizations both in the virtual
and physical domains. Such design possibilities are
at present considerably limited (Oxman 2010).
1.2
Problem definition
Applications of the finite element method (FEM)

are customarily used for the analysis of prototypes
prior to their physical fabrication. The finite element
method is a numerical technique for analyzing a computer model of a given homogeneous material that is
stressed or analyzed for specific results (Norrie and De
Vries 1973; Szabo and Babuka 1991; Moaveni 1999;
Zienkiewicz, Taylor et al. 2005; Moaveni 2007). To a
lesser extent has this method been made instrumental for purposes of design generation or fabrication,
particularly when the design involves heterogeneous
Objective [1]: Generative synthesis
The objective is to apply the logic and computation of finite element approaches to the problem of
design generation by considering the synthesis of form
as composed of finite-element units, each analyzed
according to spatial and material constraints.
2.3
Objective [2]: Variable-property finite element

fabrication
Implementing the ability to synthesize and control

material distribution as part of digital form-generation;
this research investigates design forms that become
possible when we can continuously vary material
properties. This becomes particularly significant in
considering not only structural optimization, but also
insulating properties, transparency, and other architecturally relevant properties of materials, which can
be optimized against multiple performance criteria
(Oxman 2010).
We propose a general approach to the problem of
computational form-generation of shapes with continuously varied material properties satisfying prescribed
material conditions on a finite collection of material
features and global constraints.
3
METHODOLOGY
3.1 Distribution and property-driven finite element

synthesis
The challenge of inverting an analytical routine to a
synthetic one requires the redefinition of the analytical
unit and mesh components as synthetic cellular entities
720
Figure 3. Left: distribution-driven finite element synthesis;

Right: Material-property driven finite element synthesis.
which can be further connected, combined, grown

or generated to form a surface, or a volume unit.
We propose two cases for finite element synthesis
within a fabrication context: [1] Distribution-driven
Finite Element Synthesis: this case corresponds with
objective 1, resulting in the coupling of generative
design routines and FEA using additive fabrication technologies. Finite elements are spatially distributed corresponding with external constraints. [2]
Property-driven Finite Element Synthesis. This case
corresponds with objective 2, implementing the first
objective within a variable-property material context,
using additive fabrication technologies. The two cases
are conceptualized in Figure 3 below.
3.2
Object-oriented finite element analysis
We implement and further develop the Object Oriented Finite Element Analysis (OOF) environment as
the basis for these experiments. OOF was developed
by Prof. Craig Carter at MITs Department of Material Science and Engineering in collaboration with
the National Institute of Standards and Technology
(NIST), for analyzing the effects of microstructure on
material properties (Cannillo and Carter 2000; Langer,
Fuller Jr et al. 2001). It serves to predict material
behavior under a range of objective functions defined
by the user. For any given 2-D image of material specimens, one can analyze its physical behavior based
on a hypothetical assignment of physical properties
to geometrical attributes (Carter, Langer et al. 1998;
Carter 2010).
The computation is performed using an imagebased finite element application. Physical properties
are imposed onto the image after which a computational mesh is created which includes the imageproperty information. The computation produces
various data sets including stress and strain data, heat
flow, stored energy, and deformation due to applied
loads and temperature differences. The results are
spatially analyzed and converted to a constructible
data structure using Mathematica (Cannillo and Carter
2000; Langer, Fuller Jr et al. 2001).
The following research utilizes OOFs ability to
integrate a finite element approach with material
Figure 4. Methodological set-up implementing OOF in

3-D modeling, analysis and fabrication context.
inputs. Further development of the software to implement 3D data for fabrication was carried out as
demonstrated in the design explorations (Figure 4).
The input files include micrographs and simulations using all available micro-structural data with
no mean-field approximations. Constitutive relations
translating stresses into strains using Youngs modulus
are defined by the user. OOF converts an image, or
a micrograph, of a heterogeneous, multi-component
material into a finite element mesh with constitutive
properties specified by the user.
3.3 Variable-density fabrication
With the aim of achieving micro-, and macro-structural
material property variation across the surface and
volume of a fabricated component, the software environment is implemented in a design exploration for
building skins modeled after various natural tissues
that demonstrate the variation of physical properties as
a function of their desired performance criteria. Two
classes of explorations are presented that demonstrate
distribution-driven and property-driven finite element
synthesis and fabrication.
4
DESIGN EXPERIMENTS
4.1 Distribution-driven finite element synthesis &

fabrication
Imagine the case in which the size of a mesh-free
particle, applied for the purpose of form-generation
informed by light performance, precisely matches the
721
element provides for a lower limit material definition establishing the degree of granularity required to
manifest the 3-D details of the design. From here, it
is relatively easy to imagine the implications of using
finite-synthesis elements as the units used for calibrating voxels and 3-D printing powder. The designer
is generating 3-D form using the precise units applied
to describe its physical manifestation.
Raycounting is a prototype for a product skin
designed as a variable-translucency surface. The
doubly-curved 3D model is generated by integrating a
3D generative modeling environment (Rhinoceros 4)
with a finite-element application to determine surface
thickness and material distribution as a function of a
desired array of light effects.
4.2 Property-driven finite element analysis &
fabrication
Figure 5. Raycounting, Museum of Modern Art, 2008.

3-D printed prototype embedded with color-coded lightperformance data defining material distribution and surface
thickness.
Mesh discretization processes allow the designer to

subdivide a continuous mathematical domain into
a set of discrete sub-domains referred to as elements and represented as singular geometrical entities.
Lattices and triangulations are common rationalization
discretization techniques, where quadrant and triangulated elements may respectively wrap the surface area
or volume of the object. Such structural meshes are
often implemented by engineers in order to simulate
structural loads, analyze their distribution and predict
any potential displacements that may arise.
Fatemaps is a prototype demonstrating the integration of variable-property material data with modeling and fabrication routines. This process allows the
designer to import any physical tissue data sampled
from the physical world, and embed its properties as
part of a 3-D modeling and fabrication environment.
The study explores natural tissues reconstruction by
calibrating the size, shape and proximity of each element comprising the tissue to the size and shape of the
material unit from which the form is to be fabricated
(Figure 7).
5
Figure 6. Raycounting, Museum of Modern Art, 2008. 3-D

printed prototype demonstrating variable surface thickness
informed by the Finite Element environment.
size of an imaginable powder molecule, or more

realistically speaking a material aggregate providing
for the substance of the 3D printing process. Such is
the design motivation behind Raycounting the form
of which is mediated by environmental and structural
constraints.
The element unit can be thought of as an intermediary representation linking the digital form to its
physical manifestation, particularly when rapid fabrication processes are considered. In this respect, the
CONTRIBUTIONS
Finite element analysis methods and their practical applications are primarily applied to evaluate a
given design relative to some objective function. Currently advanced applications exist that support the
integration of form-generation and evaluation from
a structural perspective. Such are, for instance, tools
developed for automobile shape optimization routines
based on the types of loads considered, their magnitudes and directions. However, despite their generative
advantage, such tools have yet to incorporate variablematerial data as part of the form-generation and
fabrication processes.
The Finite Synthesis Method (FSM) was developed
as a theoretical approach and methodology supporting the integration of modeling and analysis routines
in the process of digital fabrication. It affords the
722
undertaken in this research demonstrate an alternative

approach favoring material distribution over design
strategies of composition and assembly.
Figure 7. Fatemaps, Museum of Science Boston, 2010.

Right: illustration of the OOF environment further developed to include 3D data. Left: physical prototype reconstruction of a natural tissue (butterfly wing); the physical
model is generated within OOF and printed using an OBJET
multi-material. An integrated software environment support
modeling, analysis and preparation for 3-D printing.
Figure 8. Fatemaps, Museum of Science Boston, 2010.

Detailed view of 3-D printed tissue.
designer the ability to consider analysis tools in their

generative capacities while corresponding to variableproperty material data. In order to further implement
this approach within a design environment, the concept of finite element synthesis has been introduced,
whereby each element as defined by the FE application
may contain, in addition to its structural data, information regarding other performance criteria that are
of interest to the designer. In this regard, each material element is regarded as a tensor element defined by
indices negotiating various objective functions. This
method supports the distribution of properties across
the entire surface area of the design object relative to
the various architectural and engineering performance
criteria addressed.
Beyond the notion of performance-driven interpretations based on computational geometry methods,
this work has also engaged with the notion of computational analysis as a source for strategizing material
distribution. Rather than breaking down the design into
a series of componentized elements aiming at straightforward and simplified assemblies, the experiments
FUTURE WORK
The Finite Element Synthesis approach assumes and

facilitates the distribution of multiple material properties as a function of site-specific constraints of various
types. In other words, for each material element, the
size and shape of which is defined within the FE
software, there exists an array of related properties
specifically defined for that element. Future work will
focus on inventing new fabrication platforms able to
support variable material-property fabrication in the
scale of the finite element application. In this case,
each element may be computed as a physical entity
both virtually and physically.
Currently there exist no fabrication technologies
that allow for the production of objects with gradually varying structural properties at the scale of
product design and architectural construction. While
this work has demonstrated that such platforms are
beneficial given the potential advantages in terms
material and mechanical efficiencies, there are many
technical challenges yet to overcome. It is clear however, that an integrated approach to form-generation
where fabrication processes play an active role in the
form-generation and analysis process, promotes more
efficient products and building parts to be fabricated
allowing the designer to include evaluative functions
early in the design process.
ACKNOWLEDGEMENTS
The author wishes to thank Professor Craig Carter,
from the Department of Material Science and Engineering at MIT for his collaboration in the design
projects, his insightful comments, and for providing the initial methodological set up for this work
through the OOF software, which has been made available online. In addition, the author wishes to thank
Professor Gibson, and Professor Ortiz from the Materials Science and Engineering Department, as well as
Professor Woodie C. Flowers from the Mechanical
Engineering Department, for their constant insightful
input and support. The design explorations were supported by Paola Anonelli, chief curator at the Museum
of Modern Art in NY. The models are now part of the
permanent collection at MoMA and the Museum of
Science, Boston.
REFERENCES
Cannillo, V. and W. Carter (2000). Computation and simulation of reliability parameters and their variations in
heterogeneous materials. Acta materialia 48(13): 3593
3605.
723
Carter, C.-o. s. a. J. (2010). OOF 2: The Manual.

http://www.ctcms.nist.gov/langer/oof2man/ChapterOverview.html.
Carter, W., S. Langer, et al. (1998). The OOF manual: version 1.0. National Institute of Standards and Technology,
Gaithersburg, MD, NISTIR 6256.
Hague, R., I. Campbell, et al. (2003). Implications on design
of rapid manufacturing. Proceedings of the Institution
of Mechanical Engineers, Part C: Journal of Mechanical
Engineering Science 217(1): 25.
Hughes, T. J. R. and Hughes (2000). The finite element
method: linear static and dynamic finite element analysis,
Dover Publications.
Johnson, C. (1987). Numerical solution of partial differential equations by the finite element method, Cambridge
university press Cambridge.
Kruth, J. P., M. Leu, et al. (1998). Progress in additive
manufacturing and rapid prototyping. CIRP AnnalsManufacturing Technology 47(2): 525540.
Langer, S. A., E. R. Fuller Jr, et al. (2001). OOF: an imagebased finite-element analysis of material microstructures. Computing in Science & Engineering 3(3): 1523.
Moaveni, S. (1999). Finite element analysis, Prentice Hall.
Moaveni, S. (2007). Finite element analysis theory and
application with ANSYS, Prentice-Hall, Inc. Upper Saddle River, NJ, USA.
Norrie, D. H. and G. De Vries (1973). The finite element
method: fundamentals and applications, Academic Press.
724
Oxman, N. (2007). Get Real Towards PerformanceDriven Computational Geometry. International Journal
of Architectural Computing 5(4): 663684.
Oxman, N. (2009). Material-based design computation:
Tiling behavior.
Oxman, N. (2010). Material-based design computation.
Oxman, N. (2010). Structuring Materiality: Design Fabrication of Heterogeneous Materials. Architectural Design
80(4): 7885.
Oxman, N. (2011). Variable property rapid prototyping.
Virtual and Physical Prototyping 6(1): 331.
Szabo, B. A. and I. Babuka (1991). Finite element analysis,
Wiley-Interscience.
Zhang, H., J. Xu, et al. (2002). Fundamental study on
plasma deposition manufacturing. Surface and Coatings
Technology 171(13): 112118.
Zienkiewicz, O., R. Taylor, et al. (2005). The Finite Element MethodIts Basis and Fundamentals, volume 1,
Butterworth-Heinemann.
Zienkiewicz, O. C. and P. Morice (1971). The finite
element method in engineering science, McGraw-Hill
London.
Zienkiewicz, O. C. and R. L. Taylor (2000). The finite element
method: Solid mechanics, Butterworth-Heinemann.
Zwieniecki, M. A., C. K. Boyce, et al. (2004). Functional design space of single-veined leaves: role of
tissue hydraulic properties in constraining leaf size and
shape. Annals of Botany 94(4): 507.

Comparison of bone remodeling algorithms for hip implants

J. Frazo, H. Almeida, P. Brtolo & N. Alves
Center for Rapid and Sustainable Product Development, Polytechnic Institute of Leiria, Leiria, Portugal
ABSTRACT: The long term success of the utilization of biomaterials with structural function in the bone is
many times limited due to the bone mass loss. Several bone remodeling algorithms were developed that, combined
with the Finite Element Method, enable to predict how the bone may react to different implant designs, as well
to different biomaterials. Three of these existing algorithms were investigated. They were applied to a realistic
tridimensional model of a femoral bone with a hip implant stem. The comparison of results for each algorithm
enabled to get a better understanding on which could be the more appropriate algorithm to be used in a particular
situation.
1
INTRODUCTION
An essential aspect of the biomaterials development

process may react. The importance of this subject is
easily understood regarding biomaterials with structural function in the bone, which is the case of the hip
implant.
The unique characteristic of the bone tissue, both
cortical and trabecular, allow it to play a structural
role as the load-bearing structure of the body (Hill &
Orth 1998). Since the midst of the 19th century,
mostly due to Wolff s observation between the trabecular bone structure, e.g. its apparent density, and the
loads applied to the bone (Fyhrie & Carter 1986). The
bone, to continuously preserve this relationship, is in
a constant remodeling process. This is a complex process that involves bone resorption in a certain surface,
followed by a bone formation stage (Hill & Orth 1998).
Considering these bone tissue characteristics, it
can be easily understood that the introduction of
biomaterials into the bone leads to changes in the
stress distribution within the bone, which will consequently induce bone adaptation until a new equilibrium
state is reached. How the bone adapts, increasing
or decreasing bone mass, influences the biomaterials
performance and often limits their long-term success.
Considering the bone mass loss in the proximal femur,
when in the presence of an orthopedic implant, it can
lead to bone fracture, implant displacement or reduce
the available bone for a revision surgery (Huiskes et al.
1989).
In the last decades, several works were conducted
to mathematically reproduce the bone remodeling process and relate the bone adaption with the biomaterials
design and characteristics. These bone remodeling
algorithms are a useful tool for implant design, as they
can help to predict bone failure.
Quantitative predictions of the bone adaptation are
only possible by calculating the mechanical loads
inside the bone, in terms of stress and strain. The Finite
Element Method (FEM) is an efficient tool for this

purpose. Combining the mathematical algorithms for
bone remodeling with finite elements models, quantitative predictions on bone formation or resorption, in
realistic bone structures, are possible (Weinans et al.
1992).
Taking in consideration the existence of several
bone remodeling algorithms, it is important to study
their behavior and compare them in order to understand which could be more appropriate to a certain
case of bone implant.
In this work, a comparison between three different
bone remodeling algorithms is conducted, giving a hip
implant scenario. The three selected algorithms represent different approaches to bone remodeling. They
are not the most recent ones, though they are still
between the most widely used. In section 2, these bone
remodeling algorithms are presented. In section 3, it is
explained how these algorithms were applied, and the
results discussed. Finally, conclusions are presented in
section 4.
2
BONE REMODELING ALGORITHMS
2.1 Huiskes model

The first algorithm was presented by Huiskes et al.
(1987) and later Weinans et al (1992) applied it
to different bidimensional models, investigating its
behavior. In this model, bone is considered to be an
isotropic material, in which the Youngs Modulus (E)
is calculated through the bone apparent density ():
In this work, the values used were the ones presented

by Carter & Hayes (1977), being E0 equal to 3790 and
p 3. The values for the apparent density were limited
between 0.01 and 1.74 gcm3 , which means that, for
725
the maximum apparent density, the material presents

a Youngs Modulus near to 20 GPa.
In this algorithm, the remodeling stimulus considered is the Strain Energy Density (U), as follows:
The Strain Energy Density value is used in equation

(3) that describes the bone adaptation, obtaining the
apparent density adaptation velocity, as follows:
where C is a constant and s represents half of the stage

width in which the stimulus is considered insufficient
to lead to bone remodeling. In this work, the value
chosen for s was 30%, since this can lead to more
stabilized and realistic results.
The reference value (k) is empirically obtained
by changing its value until a more realistic result is
reached (Weinans et al. 1992). This value represents
the metabolic cost of bone maintenance so, if the k
value is higher, the resulting densities would be lower
and vice-versa.
2.2
Beaupr model
In 1990, Beaupr et al. (1990 a, b) wrote two articles

where a new bone remodeling algorithm was proposed.
These papers emphasize the importance of the time
factor in this model.
Similarly to the previous algorithm, this one considers the bone as an isotropic material where the
Youngs Modulus is based on the apparent density, but
the relationship between the Youngs Modulus and the
apparent density are different. The apparent density
may vary between 0.01 and 1.92 gcm3 . The modulus is presented in the equation (1), but the values of E0
and p are 1949.032 and 2.5, respectively, for apparent
densities bellow 1.2 gcm3 , and 1684.511 and 3.3 for
other density values.
In this algorithm, the bone remodeling stimulus is
based on the daily tissue level stress stimulus (), as
follows:
where ni is the number cycles of the load case I, and

the stress exponent m is an empirical constant that
reflects the weight of stress values over the number of
cycles. In this work, the value chosen for m was 6.
The stimulus for bone remodeling is the difference
between the daily stress stimulus and the attractor state
stress stimulus (AS ). The term attractor state refers

to the principle that most physical and biological systems are attracted to certain goal or attractor states, in
spite of these states can never be actually achieved.
Differently from the previous model, in this one
the apparent density adaptation velocity is not directly
calculated. Firstly, it is calculated the bone surface
apposition (or resorption), defining the values for the
slope (c) and for the half width (w) stage, analogous
to the s value in the previous algorithm, as follows:
This model authors refer that the bone surface apposition, or resorption, is physiologically limited. This
can be easily understood, since the number of cells
taking part in these processes are limited, as well its
velocity. Thus, there is an upper limit to how much
bone surface can be apposed, or resorpted, daily.
In this work, the recommended value used was
4 m/day, for both the maximum daily apposition rate
and the maximum daily resorption rate, in spite of the
last one can physiologically be higher.
After the bone surface change rate is obtained, this is
used to calculate the bone adaptation velocity. In this
calculation, it is taken into account the surface area
density (Sv ), which is different for different apparent
density values (Martin 1984). The apparent density
adaptation velocity is given by the following equation:
where t is the real bone density, which was considered

equal to 1.92 gcm3
2.3 Fernandes model
The last bone remodeling algorithm was proposed by
Fernandes et al. (1999). This model is, in many ways,
different from the other two, as it is based on a topology
optimization criterion.
The bone is modeled as a porous material, with a
periodic microstructure, obtained from the local repetition of cubic cells with holes of dimensions a1 , a2 and
a3 . Then, the relative density in each point is a function
of these parameters, calculated by = 1-a1 *a2 *a3 ,
where ai can be between 0 and 1. This formulation
leads to an orthotropic material, enabling the identification of the cells orientation and, consequently, the
simulation of the bone as an oriented material.
However, it was decided to make some simplifications to this algorithm in this work to better compare it
with the other ones. One of the major simplifications is
precisely to consider the bone as an isotropic material,
i.e. a1 = a2 = a3 = a.
The correlation between the bone Youngs Modulus and the relative density is the same as, in the first
726
algorithm, it corresponds to a relative density value of

1 (maximum value) and a apparent density value of
1.74 gcm3 (also the maximum value).
Using the results of the Finite Element analysis, it is
firstly calculated the direction vector (D) according to:
where v represents the volume of each cell and k

represents the metabolic cost of bone maintenance
(similarly to the first algorithm).
After obtaining the direction vector for each cell, it
is possible to calculate the new hole dimensions for
that cell:
Figure 1. Tridimensional model of a femoral bone with a

hip implant stem.
where represents the optimization step and defines

the upper and lower limits, to get a better understanding
of how much the dimensions can change in a single
step (Ruben, 2009).
3
COMPARISON OF THE BONE

REMODELING ALGORITHMS
3.1 Tridimensional model of femoral bone with

prosthesis
For the application of the bone remodeling algorithms, a tridimensional realistic model of the bone
was prepared.
A real model for the left femur, based on imaging
data, was obtained and then prepared to incorporate
a titanium hip implant stem (Fig. 1). The bone model
was then divided into 6587 tetrahedrical elements.
Initially, the bone remodeling algorithms were
applied to the elements, though in this work they were
instead applied to the nodes.
This node-based approach is useful to avoid the
numerical instabilities, e.g. the appearance of the
checkerboard effect in the results (Jacobs et al. 1995).
Three load cases were applied to this model: the
first load case corresponds to the one-legged stance,
the second and third ones to the abduction and adduction of the leg, respectively. Each load case consists
of two concentrated forces: one applied to the proximal end of the implant that is related to the joint
reaction force (Fh), and the other one is applied to

the greater trochanter related to the actuation of the
abductor muscles (Fa).
For the first load case, the Fh value is 2317 N
directed 24 with the vertical line, and Fa of 703 N
28 from vertical. The second load case consists of
a Fh value of 1158 N and 15 from vertical and Fa
of 351 N 8 from vertical. Finally, for the third load
case, Fh is 1548, N 56 from vertical, and Fa is 406 N
35 from vertical.
Since the second model, based on Beaupr et al.
(1990a, b), takes into account the number of daily
cycles for each load case, then the daily cycles for the
first load case was defined as 6000, and the second
and third load cases as 2000.
3.2
Discussion
Before the discussion of the results regarding the application of the bone remodeling algorithms, there are
some considerations to be made about the algorithms
themselves.
The first algorithm is the simpler one. It is easy to
implement and the one with fewer parameters to be
tuned, which does not imply that the results are worse.
This algorithm is a useful tool for simple applications with classical biomaterials, to investigate long
term results on the introduction of implants in the bone,
as long as the parameters are carefully chosen,
The second algorithm gives much more emphasis
to the time factor. This can be noted as the remodeling
stimulus is based on the daily stress stimulus, the daily
bone resorption, or apposition, which can be confined
727
Figure 2. Result of the application of the first bone remodeling algorithm to the tridimensional model.
to certain values, being the remodeling velocity for

each step dependent upon the apparent density. This
time factor emphasis allows the application of this
algorithm to cases where the bone remodeling evaluation need to be a function of time. Those cases are
mainly related to the use of degradable biomaterials,
to which the previous algorithm cannot respond.
The third algorithm differs in many ways from
the previous ones. This algorithm was simplified for
this work, the major remaining difference being the
approach to the bone remodeling process the topological optimization approach. This approach allows
some investigation the others algorithms do not, especially if the bone is considered an orthotropic material.
For example, the study of osteointegration around
the implant stem (Ruben 2009) or obtaining the best
scaffold designs for each bone region (Coelho et al.
2009).
Results obtained after the application of the bone
remodeling algorithms to the tridimensional model are
illustrated in Figures 2-4, where black represents the
minimum density values and white the maximum ones.
The results show that, in this case, all simulation
results are close to expected ones. The proximal region
of the femur would loose bone mass, and the region
above the implant stem would present maximum density values, due to the stress shielding effect. This is
exactly what the results show.
The result for the third algorithm, although presenting the same tendency regions, is different from the
other ones, which can be explained by having not yet
reached the best parameters values.
Some minor differences between the first two algorithms results can be explained through parameter
choices. Choosing each parameter to obtain similar
results is not an easy task. The second algorithm, based
on Beaupr et al. (1990a, b), shows a little difference
in density distribution, which can be easily explained
Figure 3. Result of the application of the second bone

remodeling algorithm to the tridimensional model.
Figure 4. Application results of the third bone remodeling

algorithm to the 3D model.
taking into account the number of daily cycles for

each load case, which can lead to a larger contribution
of the first load case regarding the bone remodeling
calculation.
CONCLUSIONS
This paper results show that all the three algorithms

can be useful tools, for the case study presented,
where the goal was to investigate the long-term bone
remodeling results and the used biomaterial, which is
non-degradable. If any of these parameters change,
then some models may not be suitable. The second
algorithm, based on the Beaupr model, emphasizing
the time factor, is suitable for the cases in which the
728
result need to be a function of time. The third algorithm, based on the Fernandes model, can be used in
more complex analysis, especially when they deal with
a oriented material.
REFERENCES
Beaupr, G.S., Orr, T.E. & Carter, D.R. 1990a. An Approach
for Time-Dependent Bone Modeling and Remodeling
Theoretical Development. Journal of Orthopaedic
Research 8: 651661.
Beaupr, G.S., Orr, T.E. & Carter, D.R. 1990b. An Approach
for Time-Dependent Bone Modeling and Remodeling Application: A Preliminary Remodeling Simulation.
Journal of Orthopaedic Research 8: 662670.
Carter, D.R. & Hayes, W.C. 1977. The Behavior of Bone as a
Two-phase Porous Structure. J. Bone Jt Surg. 59: 954962.
Coelho, P.G., Fernandes, P.R., Rodrigues, H.C.,
Cardoso, J.B. & Guedes, J.M. 2009. Numerical modelling of bone tissue adaptation A hierarchical approach
for bone apparent density and trabecular structure. Journal
of Biomechanics 42: 830837.
Fernandes, P.R., Rodrigues, H. & Jacobs, C. 1999. A Model

of Bone Adaptation Using a Global Optimisation Criterion Based on the Trajectorial Theory of Wolff. Computer
Methods in Biomechanics and Biomedical Engineering
2:125138
Fyhrie, D.P. & Carter, D.R. 1986. A Unifying Principle
Relating Stress to Trabecular Bone Morphology. Journal
of Orthopaedic Research 4: 304317.
Hill, P.A. & Orth, M. 1998. Bone Remodelling. British
Journal of Orthodontics 25: 101107.
Huiskes, R., Weinans, H., Grootenboer, H.J., Dalstra, M.,
Fudala, B. & Sloof, T.J. 1987. Adaptive Bone-Remodeling
Theory Applied to Prosthesis-Design Analysis. Journal of
Biomechanics 20: 11351150
Jacobs, C.R., Levenston, M.E., Beaupr, G.S., Simo, J.C. &
Carter, D.R. 1995. Numerical Instabilities in Bone
Remodeling Simulations: the Advantages of a NodeBased Finite Element Approach. Journal of Biomechanics
28(4): 449459.
Ruben, R. 2009. Projecto ptimo de Prteses de Articulao
da Anca, Phd Thesis, Technical University of Lisbon.
Weinans, H., Huiskes, R. & Grootenboer, H.J. 1992.
The Behavior of Adaptive Bone-Remodeling Simulation
Models. Journal of Biomechanics 25(12): 14251441.
729

Computer modelling and simulation of reaction injection moulding:

Filling and curing stages
R.T. Dias, A. Mateus, G.R. Mitchell & P.J Brtolo
ABSTRACT: Reaction injection moulding involves simultaneous flow and exothermal curing reactions. The
optimisation of these events enables to produce parts with better quality and improved properties. This work
proposes the combined use of ANSYS CFX together with appropriate mould filling and curing kinetic models.
The inhomogeneous multiphase flow model was considered together with both the second order backward Euler
transient scheme and the high resolution advective scheme. The simulation approach was experimentally verified.
A good agreement between experimental and numerical data was obtained by using a diffusion-controlled curing
kinetic model.
INTRODUCTION
Reaction Injection Moulding (RIM) is an important process for the rapid generation of complex
polyurethane (PU) objects (Mateus, 2011; Dias, 2011;
Macosko, 1989; Fan et al, 1997). This is a process
in which two low viscosity, highly reactive chemical
components are mixed and injected into a mould (Figure 1), allowing large complex and relatively thin parts
to be produced without the need for high clamping
pressures. RIM is also characterised by low injection
pressures and mould temperatures, so low cost and
large scale moulds can be used and important energy
savings may be achieved.
The operation cycle comprises the following main
phases (Figure 2) (Mateus, 2011):
Metering and mixing of two or more low-viscous

liquid reagents (phase A);
Mould filling (phase B);
Curing reaction to solidify the liquid polymeric
mixture (phase C);
Part extraction (phase D);
Post-curing operation to complete the solidification
process and improve the mechanical properties of
the produced parts (phase E).
The RIM process involves simultaneous flow and

exothermic curing reactions (Macosko, 1989). This
reaction controls the resin morphology that will determine the physical, electrical and mechanical properties
of the final product. Thus, during the operation cycle,
several physical and chemical phenomena occur influencing both the cycle time and the final properties of
the injected parts. Hence, it is fundamental to optimise
the processing conditions in order to produce high
quality parts in short production cycles. Computer
simulation plays here an important role.
Figure 1. Schematic representation of the RIM machine.
2 THE CURING REACTION

The curing reaction is a polymerization process characterized by chemical cross-linking reactions that
create an infusible, insoluble and highly cross-linked
three-dimensional network (Brtolo, 2001). The curing process starts from a formation and linear growth
of a chain. The chain rapidly starts to branch and
then cross-link. As the curing proceeds, the molecular weight rapidly increases and several chains are
linked together forming a network of infinite molecular weight. This reaction is characterized by two
main events: gelation and vitrification (Mateus, 2011;
Brtolo, 2001).
Gelation is a non-reversible event that corresponds
to the incipient formation of an infinite molecular network, which is associated to an increase in viscosity
and a decrease in processability (OBrien and White,
731
Figure 2. RIM operation cycle.
2003; Brtolo, 2001; Yousefi et al, 1997). After gelation, as the reaction further progresses, the amount
of solid material increases and the polymer becomes
more cross-linked, increasing its stiffness and strength.
From an injection view, gelation is a very important
event as, from gelation on, the resin no longer flows.
Vitrification corresponds to the formation of a
glassy solid material, due to an increase in both the
cross-linking density and the molecular weight of the
polymer being cured (Brtolo, 2001; Yousefi et al,
1997). The onset of vitrification occurs when the
increasing glass transition temperature reaches the
cure temperature. The rate of the reaction will undergo
a significant decrease after vitrification and the reaction becomes very slow, as it is controlled by the diffusion of the reactive species. This diffusion-controlled
effect produces a slowdown of the solidification process and determines the final degree of conversion.
Momentum
where = density; and = dynamic viscosity.

The heat transfer is described through the energy
conservation equation, which includes an exothermic
heat generation term given by an appropriate kinetic
equation:
where h = specific enthalpy of the resin; k = thermal

conductivity; H = total heat released during the curing
reaction; and /t = reaction rate.
Viscosity changes, associated with the cure process,
are modelled by Yousefi et al (1997), as follows:
3 THERMO-RHEO-KINETIC MODEL
The simulation of RIM comprises both 3D fluid flow
and heat transfer (Dias, 2011; Mateus, 2011). The
fluid flow is modelled by both the mass conservation
equation and the momentum equations.
Mass conservation
The curing kinetics is usually described through

developed phenomenological models, assuming that
only one reaction can represent the whole cure process. These models can be represented by the following
equation (Brtolo, 2001; Hu and Zhao, 2007), as
follows:
732
where f() is a function of conversion () and Kc (T),

the chemical controlled rate constant, is a function of
temperature.
Two kinetic models were considered in this research
work: the Kamal model and a diffusion-controlled
model based on Brtolos work (Brtolo, 2007, 2006,
2001; Kamal et al, 1973).
Kamal model:
where = pre-exponential factor of the rate constant;

E = activation energy; R = gas constant; + T =
absolute temperature; and m and n = reaction orders,
with the sum (m + n) being the overall reaction order.
Equation (6) do not explicitly includes the effects of
resin composition on the rate of cure and, consequently,
the kinetic parameters must be recalculated after each
change in the resin formulation. This model is also
not able to predict the diffusion control effects after
vitrification.
Diffusion-controlled model:
Figure 3. Parameters considered for the simulation and

respective correlations.
Figure 4. No-slip and free-slip boundary conditions.
where = diffusion coefficient; v = critical value

of fractional conversion corresponding to the onset
of diffusion-controlled effects over the curing reaction; = pre-exponential factor of the constant rate;
and [I] = isocyanate concentration. All the kinetic
parameters are function of temperature and isocyanate
concentration.
The following assumptions were considered for the

mould filling simulation:
Newtonian fluid
Laminar flow;
Constant inlet velocity with a parabolic profile;
Incompressible material with constant density and
thermal properties.
Neglect entrance length (L >> H);
Neglect side walls (W >> H);
Heat conduction at the walls of Kres = 0.82
[W/mK]).
COMPUTER IMPLEMENTATION USING

ANSYS CFX
The numerical simulation of both filling and curing stages of RIM was performed using Ansys CFX
software package, using the Finite Volume Method.
Figure 3 illustrates the correlation between parameters required to simulate both the filling and curing
stages.
First the filling stage was simulated during 3 sec.
The polymerization process was after simulated during 1800 sec. The no-slip condition on the walls was
applied to the resin, while the free-slip condition was
considered for the air (Figure 4). The other imposed
boundary conditions were: a parabolic velocity profile with an average velocity at the inlet and relative
pressure at the outlet equal to zero.
The second order backward Euler transient scheme

and the high resolution advective one were considered
to solve the thermo-rheo-kinetic model (Dias, 2011).
CASE STUDY
A rectangular L-shaped cavity, placed vertically with a

full gate at the bottom, was considered as a case study
(Figure 5).The main dimensional and filling characteristics considered for simulation proposes are indicated
in Table 1. The model was approximated using 1190
tetrahedrical elements with 1394 nodes.
733
Figure 5. Mould geometry.

Table 1.
Mould dimensions and filling conditions.
Symbol
Description
Value
L
W
H
Uin
Tw
Air
Air
Resin
Resin
kResin
Length
Width
Thickness
Inlet velocity
Wall Temperature
Air density
Resin viscosity
Resin density
Resin viscosity
Resin thermal conductivity
25 mm
2.5 mm
0.4 mm
0.0158 m/s
25 C
1.185 kg/m3
1.831 105 Pa/s
1135 kg/m3
2.44 Pa/s
0.108 W/mK
Figure 6 shows the volume fraction variation of the

resin during the mould filling phase. A mixture containing 40% of isocyanate and an inlet temperature of
30 C were considered. Results show an entrapment of
the air in the corner of the 90 curve, which can be
minimized by reducing the filling velocity.
To study the effect of both resin composition and
inlet temperature on the curing processes two different measurement points were considered (Figure 7).
Figure 8 shows numerically obtained contour plots of
the variation of fractional conversion with both isocyanate concentration and curing time. The results
were obtained using both the Kamal and the diffusioncontrolled kinetic model and an inlet temperature of
30 C. No significant differences were observed for
points 1 and 2. The experimentally obtained profile
is indicated in Figure 9 (Mateus, 2011). The results
show that a better agreement between numerical and
experimental results is obtained with the diffusioncontrolled curing kinetic model.
Figure 10 shows the numerically obtained contour
plots of the variation of fractional conversion with
inlet temperature. Results were obtained using both
the Kamal and the diffusion-controlled kinetic model
and resin samples containing 40 wt% of isocyanate.
Values were considered only for point 2. The experimentally obtained profile is indicated in Figure 11.
Results show that an increase in the inlet temperature increases the fractional conversion values. A
better agreement between numerical and experimental
results was obtained with the diffusion-controlled
curing kinetic model.
Figure 6. Resin volume fraction variation during the filling

phase. a) 1 sec; b) 2 sec.; c) 3 sec.
Figure 7. Position of the measurement points.
734
Figure 8. Numerically obtained contour plots showing the

variation of fractional conversion as a function of isocyanate
concentration for two different kinetic models. a) Kamal
model; b) Diffusion-controlled kinetic model.
Figure 9. Experimentally obtained contour plots showing the variation of fractional conversion as a function of
isocyanate concentration.
Figure 10. Numerically obtained contour plots showing the

variation of fractional conversion as a function of inlet temperature for two different kinetic models. a) Kamal model;
b) Diffusion-controlled kinetic model.
Figure 11. Experimentally obtained contour plots showing

the variation of fractional conversion as a function of inlet
temperature.
735
CONCLUSIONS
The computer modeling and simulation of the RIM

process presented in this work is an important tool to
study and optimize mould and part design. Two curing kinetic models (Kamal and diffusion-controlled)
were considered and the simulations validated against
experimental data. The new model describing the
diffusion-controlled effects that determines the curing
reaction after vitrification, shows a good agreement
with the experimental results. The differences between
the two curing kinetic models are more significant
for low inlet temperatures and resin samples with
compositions different than the stoichiometric mixture
(50/50 wt%).
REFERENCES
Brtolo, P.J. 2001. Optical approaches to macroscopic
and microscopic engineering, Unpublished PhD Thesis,
University of Reading, UK.
Brtolo, P.J. 2006. Computer simulation of stereolithographic
curing reactions: phenomenological versus mechanistic
approaches, Annals of the CIRP, 55(1), 221226.
Brtolo, P.J. 2007. Photo-curing modeling: direct irradiation,

Int J Adv Manuf Technol, 32, 480492.
Dias, R.T. 2011. Computer modeling and simulation of
reaction moulding processes: filling and curing stages,
Unpublished MSc Thesis, Polytechnic Institute of Leiria,
Portugal.
Fan, L.H., Hu, C.H., Pan, Z.Q., Zhang, Z.P. and Ying, S.K.
1997. Polymerization of polyurethane and vinyl ester
resin interpenetrating polymer networks during reaction
injection moulding process, Polymer, 38, 36093616.
Hu, X. and Zhao, L. 2007. A variable reaction order model for
prediction of curing of thermosetting polymers, Polymer,
48, 61256133.
Kamal, M. R., Sourour S. and Ryan, M. 1973. Integrated
thermo-rheological analysis of the cure of thermosets,
SPE Technical Paper, 19, p. 187191.
Macosko, C.W. 1989. RIM Fundamentals of Reaction
Injection Moulding, Hanser, Munich.
Mateus, A.J. 2011. Modelling and optimizing of reaction
injection moulding, Unpublished PhD Thesis, University
of Reading, UK.
OBrien, D.J. and White, S.R. 2003. Cure kinetics, gelation and glass transition temperature, Poly. Eng. Sci., 43,
863873.
Yousefi, A., Lafleur, P.G., Gauvin, R. 1997. Kinetic studies of
thermoset cure reactions: a review, Polymer Composites,
18, 157168.
736

Warped hexahedral meshing of an ellipsoidal inclusion for design of

composite material
L. Podshivalov
Laboratory for CAD and LCE
Computational Biomechanics Laboratory Faculty of Mechanical Engineering, Technion, Haifa, Israel
A. Fischer
Laboratory for CAD and LCE
P.Z. Bar-Yoseph
Computational Biomechanics Laboratory Faculty of Mechanical Engineering, Technion, Haifa, Israel
ABSTRACT: Composite material can be designed and modeled as a matrix with ellipsoidal inclusions of several
materials. These multiple ellipsoidal inclusions are randomly distributed in a unit cube volume according to the
material parameters, e.g. density, size, orientation. These models can be meshed and then used for mechanical
finite element analysis. Rapid meshing can be achieved with structured grids. However, regular structured grids
are considered sensitive to orientation and not robust, since the mesh quality strongly depends on the orientation
and spatial position of the geometric model. In this paper, we describe an automatic warped hexahedral meshing
algorithm that produces deformed structured meshes for composite materials with ellipsoidal inclusions. The
proposed algorithm successfully meets the above challenges and provides high quality meshes.
INTRODUCTION
Automatic high quality hexahedral meshing is considered a very complex task. Hence, despite extensive
research, currently there are no robust methods that
can handle grain-based geometry. The meshing process becomes more challenging when dealing with
inclusion models that consist of different materials.
The entire process is conducted on the Representative Volume Element (RVE). The RVE is defined as a
unit cube with material properties called matrix and
multiple inclusions with different material properties
(Zaoui, 2002). In such cases, following issues needs
to be addressed:
(a) Continuity must be achieved on the inclusion
boundaries;
(b) The meshing needs to be robust to dimensions,
location, size and orientation of the inclusions;
and
(c) The mesh needs to approximate the original
geometric model with high accuracy.
Rapid meshing can be achieved with structured
grids. Such algorithms are easy to implement and they
provide high quality elements. However, regular structured grids are considered sensitive to orientation and
not robust (Blacker, 2000). As a result, the mesh quality strongly depends on the orientation and spatial
position of the inclusions.
In this paper, we describe an automatic warped hexahedral meshing algorithm that produces structured
meshes for composite materials with ellipsoidal inclusions. This algorithm successfully meets the above
challenges and provides high quality meshes.
The process begins with a geometric model in which
the inclusions are represented by quadric surfaces. The
surface model is then converted into a volumetric
model based on a regular structured mesh. The mesh
is then warped according to the shape of the inclusions. The warping procedure involves applying an
inductive vector field on each node of the structured
grid. The user can control the influence exerted by
the field on the mesh by means of weights. Increasing the weights can improve the meshs conformity
to the inclusions, but can also reduce mesh quality.
Thus, optimal weights need to be selected. Grid resolution can also be controlled by the user. A dense
grid provides a better and more geometrically adaptive
mesh, but subsequently requires more computational
resources for finite element analysis. A sparse grid
is easier to solve, but its similarity to the original
inclusions is less precise.
The paper is organized as follows. In Section 2
details of the proposed meshing algorithm are provided, beginning with the geometric surface model,
proceeding with inductive field generation and concluding with the mesh warping. In section 3, example meshes are presented and mesh quality analysis
737
Restricting the geometry to spherical or prolate

ellipsoids we set a, b and c parameters as:
Each inclusion can have a random position and orientation. Thus the final geometry for each inclusion is
defined with rotation and translation transformations
as follows:
Figure 1. Algorithm flow for warped hexahedral meshing

of composite material with ellipsoidal inclusions.
is conducted. Section 4 summarizes the paper and

provides a discussion of future work.
In order to avoid intersection between adjacent

inclusions, a collision detection algorithm based on
Object Oriented Bounding Boxes (OOBB) (Jimnez
et al., 2001) is utilized in the process of generating
each inclusion.
Two examples of such models are presented in
Figure 2.
Next, the surface model is converted into a volumetric structured grid model where each element is a
voxel. Initially, the RVE is embedded into a structured
grid with a predefined resolution. Afterwards, each
voxel is tested and classified according to its location
inside or outside inclusions of the geometric model.
This test is carried out by utilizing the exact analytical model of the ellipsoidal inclusion. The point p that
is located inside inclusion must satisfy the following
condition:
2 WARPED HEXAHEDRAL MESHING

Warping the structured grid for surface reconstruction was proposed in (Azernikov and Fischer, 2006).
This algorithm achieved high quality reconstruction
from a variety of inputs, e.g. cloud of points and CT
scans. The same algorithm also produced volumetric
hexahedral elements. However, these elements have
been used only for visualization purpose and their
suitability for finite element analysis was not verified.
Moreover, this algorithm dealt with a single geometric
model rather than with a volume incorporating multiple embedded sub-models. Therefore, the issue of
continuity on the boundaries between the matrix and
the inclusions has not as yet been addressed. In this
paper, the warping concept is utilized for developing a
new automatic warped hexahedral meshing algorithm.
Schematic flow of the proposed algorithm is presented
in Figure 1.
The mathematical basis behind mesh warping is
briefly explained hereinafter.
The model of the Representative Volume Element
(RVE) consists of a union set of quadric surfaces where
each of these surfaces represents a different inclusion.
Moreover, that the void between inclusions is also
material that needs to be meshed. In case of ellipsoidal
inclusions the surfaces are defined as follows:
where f1 and f2 are the focus points of the ellipsoid

and a is the coefficient defined in (1) and (2).
Resulted model for a single inclusion is depicted in
Figure 3. It can be noticed that the orientation of the
ellipsoid does not coincide with the direction of the
grid.
After the geometric model was generated, a scalar
field D can be evaluated. For each voxel in the volumetric model the shortest distance from the closest
inclusion is computed. The computation is carried out
utilizing kd-tree data structure (Samet, 1990). Visual
representation of such field on a single slice is depicted
in Figure 4.
Given a scalar field D in a RVE, an inductive vector
field F can be found. For this aim, the scalar field is
differentiated, and then normalized it by the value of
the largest gradient.
In Figure 5, which depicts an example of a scalar

field, it is evident that the gradient initially increases
proportionally to the distance between the voxel
and the inclusion boundary. Note that the gradient
decreases to zero at the midpoint between each two
adjacent inclusions. As a result each inclusion defines
a volumetric cell that is known as a Voronoi cell. For
738
Figure 4. Scalar field representation: Dark colors stand

for large distances from the inclusions boundary and bright
colors for short distances.
Figure 5. Gradient inductive field where voids represent

locations of the inclusions.
Figure 2. Geometric models of composite material with
inclusions: (a) one type of inclusions, and (b) two types of
inclusions.
Thus, the new position of node P is defined by the

following expression:
3
Figure 3. Ellipsoid model: (a) surface representation, and
(b) structured grid representation with the surface.
each grid node P a directional vector is then defined,

making it possible to change the nodes position and
by that warp the mesh. It is also possible to control the
influence exerted by the field by means of weights.
EXAMPLES
In this section, several examples for warped hexahedral meshing are presented. Meshes in these examples
were created from the same geometric model by applying inductive vector field with different weights. The
selected weights depict three representative cases:
Example 1 (Figure 6a): = 0 This case is equivalent to the original structured mesh without any
influence exerted by the field. The mesh quality is optimal, but mesh poorly adapts to the original geometric
739
Table 1. Mesh quality analysis according to the criterion of

normalized Jacobian
Elements quality
Figure 6. Warped hexahedral meshes with different

weights: (a) = 0, (b) = 0.2 and (c) = 0.4. Upper image
depicts mesh warping according to the scalar field and lower
image shows the received mesh.
Weight
1.00.6
0.-5-0
<0
0
0.2
0.4
100%
98.9%
96.3%
1.1%
0.4%
3.3%
model. Moreover, in such configuration the mesh is

highly sensitive to orientation of the inclusion and
hence less robust.
Example 2 (Figure 6b): = 0.2 In this example,
the distortion of the mesh in the direction of the field
is evident. The correlation between the mesh and the
geometric model is very good and the mesh takes on
the shape of the inclusions.
Example 3 (Figure 6c): = 0.4 Even though the
mesh has adapted itself almost completely to the inclusions, illegal elements have been created that are not
suitable for finite element analysis.
The cases above indicate that the influence of the
field should be restricted and controlled. Such restriction can be determined either in pre-processing, by
estimating geometrical properties of the surface model
e.g. curvature or during the process, by verifying
quality of each element.
From these three examples it is evident that plays
important role in the meshing process. By selecting
an optimal weight it is possible to create a mesh that
imitates the geometric model and also produces high
quality elements. Currently the weight parameter is
defined by the user. In the extended algorithm this
parameter will be set automatically.
For each of the presented examples mesh quality
analysis is performed. The normalized Jacobian criterion (Bar-Yoseph et al., 2001) was utilized in the
current work. This parameter indicates how much
the given element resembles a cube. The maximal and
the optimal value is one. Values above 0.6 indicate high
quality elements. Negative values, which occur when
an element self intersects, are unacceptable since finite
element analysis fails in such cases. According to this
criterion, all the elements in the initial grid have values equal to one. Thus, it is interesting to detect the
elements whose quality decreases. Comparison of the
analysis is presented in Table 1. It can be seen that high
weight creates a significant amount of illegal elements.
Such elements do not exist in the case = 0.2.
Finally, the warped mesh proximity to the original
geometry was visually verified. Figure 7 depicts three
ellipsoidal inclusions meshed with different weight
parameters. The difference between the regular structured grid and the warped grid is visible. Warped
meshes better describe the underlying geometry and
thus produce better results.Although the largest weight
produces the best results from the geometrical aspect,
from the computational aspect such mesh currently is
discarded, due to appearance of illegal elements.
740
to regular structured. Moreover, the continuity on the

boundaries between the matrix and the inclusions is
guaranteed, due to structured grid topology. The presented method is automatic, robust and rapid. It is
also independent on the number of inclusions in the
model. In this paper, the proposed method deals with
two types of inclusion materials. However, the method
is not limited to any number of materials.
In the future work, the influence of the weight
parameter on the mesh quality will be further investigated and adaptive weight application on the grid will
be developed.
ACKNOWLEDGMENT
This study was partially supported by the TechnionHutchinson R&D collaboration and by the Samuel and
Anne Tolkowsky Chair at the Technion.
REFERENCES
Figure 7. Mesh approximation to the original geometric

model for three respective weights (0, 0.2 and 0.4).
SUMMARY AND FUTURE WORK
In this short paper warped hexahedral meshing

algorithm of composite material with ellipsoidal
inclusions was presented. This method allows utilizing a structured grid and at the same time produces
meshes very similar to the original geometry contrary
Azernikov, S., Fischer, A., 2006. A New Volume Warping

Method for Surface Reconstruction. Journal of Computing and Information Science in Engineering 6, 355363.
Bar-Yoseph, P.Z., Mereu, S., Chippada, S., Kalro, V.J., 2001.
Automatic monitoring of element shape quality in 2-D
and 3-D computational mesh dynamics. Computational
Mechanics 27, 378395.
Blacker, T., 2000. Meeting the Challenge for Automated
Conformal Hexahedral Meshing, 9th International Meshing Roundtable, New Orlean, Louisiana, pp. 1119.
Jimnez, P., Thomas, F., Torras, C., 2001. 3D collision
detection: a survey. Computers & Graphics 25, 269285.
Samet, H., 1990. The Design and Analysis of Spatial Data
Structures. Addison-Wesley, Reading, MA.
Zaoui, A., 2002. Continuum Micromechanics: Survey.
Journal of Engineering Mechanics 128, 808816.
741
Applications

Rapid manufacturing of removable complete denture components

Md Saidin Wahab & Noraniah Kassim
Faculty of Mechanical and Manufacturing Engineering, Universiti Tun Hussein Onn Malaysia,
Batu Pahat, Johor, Malaysia
Zainul Ahmad Rajion

School of Dental Science, Health Campus, Universiti Sains Malaysia, Kelantan, Malaysia
ABSTRACT: This paper presents the development of rapid manufacturing for complete denture components.
The aim is to develop a new processing route for the removal complete denture and its assembly by integrating
the techniques used in the advanced manufacturing technology. The current process is a labour intensive which
involves many procedures and requires a long time to complete. Most important, it does not preserve any
quantitative information for future retrieval. The 3D-ATOS scanning system was used to obtain the surface
data of the denture teeth sets, the edentulous models and the rims for patient. Some of these components have
been prepared in dental clinic laboratory. The surface quality of the digitized images was further refined by
using Geomagic Studio 10 and the complete model of the denture components were prepared using Solid Work
CAD application tool. A set of library for teeth was developed in the CAD system which includes the reference
lines and points for setup and also for assembly process. The master models of the complete denture were
produced using MJM rapid prototyping system. The complete denture was fabricated using acrylic resin (Vertex
Castavaria) material on silicon mold by vacuum casting machine. The finish denture was tested on patient and
found satisfactory with good fitting accuracy as compared to the conventional denture. The use of vacuum casting
procedure also has improved about 30% flexural strength as compared to the compression flask technique.
INTRODUCTION
In recent years, the used of Additive Manufacturing

(AM) techniques in the manufacturing of medical
products have expended together with the advancement of technology. Nevertheless in dentistry area,
denture components are still being produced manually and fabricated using conventional methods. The
compression flask process is one of the examples
which most commonly used in denture manufacturing. The process requires many steps and related to
labor intensive work. The good quality of the complete denture was rated by prosthodontists at 2 to 6
months to be produced [1]. Besides that, information
related to personalize denture cannot be retrieved and
high skills personal are needed in wax carving and
teeth arrangement, which depends mostly on the technicians subjective judgment [2]. In addition, in the
compression method, the material was not accurately
estimated and some of the material will be treated as a
waste. This gives a reason why this study is important
in denture manufacturing research. The introduction of
advanced manufacturing technologies (AMT) such as
3D scanning, CAD, Rapid tooling technique are able
to make complicated, optimized material, customerspecific denture product more faster and accurate than
the conventional process [3,4]. Rapid tooling describes
those applications that are aimed at making tools and
moulds for the production of prototypes and pre-series

products by using the same processes as those used
in rapid prototyping (RP) [5]. The RT used in this
process had classified as indirect tooling, in which
a rapid prototype master is converted into a mould
using a secondary process. Indirect tooling also had
been classified based on their expected mold life as
soft tooling and hard tooling [6]. Besides that the soft
tooling method had been used in this study to produce
soft moulds such as silicone rubber [6]. Vacuum casting process has been used to produce silicone moulds
and denture fabrication.
2
3D CAD DENTURE DESIGN
2.1 Digitized 3D images

CAD denture design starts with the denture teeth sets,
the edentulous models and the occlusion rims as shown
in Figure 1(a) which had been prepared in denture lab.
The edentulous model and rims were taken from the
patients impression as shown in Figure 1(b). By using
3D-ATOS scanning system, these components were
digitized to capture the surface data for 3D images
as shown in Figure (cd). Then the surface quality
of the complete digitized images was further refined
by using Geomagic Studio 10 for better finishing and
solid modeling preparation.
745
Figure 1. (a) edentulous models, occlusion rims and denture teeth, (b) Impression wax (c) 3D Scanning (d) Digitized
image.
2.2
Denture design in CAD software
2.2.1 Creating middle line and setup curve

Solid Work ver. 2010 was used in CAD design. The
denture part can be moved in 3D, rotated and displayed
along any coordinate axis. It also provides many interface file for further development. The middle lines
and setup curves has been created on the edentulous models for teeth arrangement. The process starts
by merging the occlusion rims with the edentulous
models. Then, both models were assembled together
between maxilla and mandible rims to get the occlusion plane. On the occlusion plane, a symmetric line
had been created to define the centre of setup curve.
Spline curves drown on opposite bridge of edentulous models by using the symmetric line as a reference.
Then, it was offset by 5 mm to create the setup curve,
as shown in Figure 2a.
The Cartesian world coordinate system had been
setup at the crossing point of the centre line and the
setup curve was set as XY axis. Then, the middle line
was created vertically to the facial midline and set
as Z axis, as shown in Figure 2b. This will be used
as a reference coordinate for the teeth arrangement
process.
2.2.2 Established teeth reference lines and points
Figure 2c shows the reference line for maxilla anterior,
in which the horizontal and vertical axes were designed
following teeth setup point procedure. The reference
lines and points of each tooth were stored in CAD
library as database for denture setup preparation.
2.2.3 Setup maxillary teeth
The imaginary plane had been created to each tooth
base on reference lines. Following that, all of the teeth
had been assembled to denture setup curve which
begins at the middle line for central incisor tooth
as shown in Figure 2d. There were also presents the
Figure 2. (a) Symmetry line and curves for setup teeth (b)
Facial midline. (c) Posterior maxilla teeth reference lines (d)
Setup maxilla teeth (e) Setup mandible teeth by Collision
Detection (f) Trim and fill new surface in gingival creation. (g) Freeform new surface. (h) Completed designed
the denture.
mesial imaginary edge of central incisor with coincident to the crossing point of teeth setup curve and
facial midline. Then the distal imaginary edge had
positioned to setup curve with PathMate. After that
the vertical axis line from teeth parallel to facial midline. The next tooth follows with the steps and the
complete assembly of maxilla teeth was symmetry to
both sides.
2.2.4 Setup mandible teeth
Mandible anterior teeth setup was similar to maxilla teeth with established the reference lines, points
and imaginary plane. However for mandible posterior teeth were assembled by moving and-rotating the
Collision Detection in CAD application, as shows
in Figure 2e. This method propose to positioning the
critical area between each mesial and distal of maxilla
and mandible posterior teeth.
2.2.5 Design artificial gingival and base plate
The gingival and base plate had been created by trimming the unnecessary surface at the occlusion rim.
Then boundary line was created between the teeth and
the trimmed area at the occlusion rims. Then the new
surface had been filled at the boundary region as show
in Figure 2f. The new surfaces have been modified by
746
Freeform. Then, gingival and base plates have been

created, as shows in Figure 2g. The surface model has
been transformed to solid by using tool called Knitt
and the final design shown in Figure 2h.
3
3.1
FABRICATION IN RAPID CASTING

PROCESS
Fabricated the master pattern
The CAD model for denture assembly was converted

to STL file which is the standard format used in the
rapid prototyping process. The physical model of the
denture assembly was produced on acrylic material
by using Rapid Pro Jet SD3000 machine, as shown in
Figure 3a.
3.2 Vacuum casting process

Silicon mold was made from silicone rubber VT750
and prepared in vacuum casting machine. The artificial teeth were set in upper and lower in the denture
silicone mold as shows in Figure 3b. Cold cure resin
(VERTEX Castavaria, Netherland), the acrylic based
polymer was used with the ratio of 0.95 g polymer and
1.7d monomer. From the ratio, the total amount of
acrylic resin used was calculated. The mixed acrylic
resin was degassed at 40 second in the vacuum casting machine and then poured into silicone mould, as
shown in Figure 3cd. Then, the resin was cured in
the oven with temperature set at 55 C for 30 minutes.
Figure 3ef shows the final denture that had been postprocessed by trimming, cleaning and polishing and
tested on patient for fitting and accuracy test.
4
Figure 3. (a) Denture model (b) Silicone mold (c) Setting the teeth (d) Complete denture assembly after molding
(e) Denture after trimming and polishing (f) Test on patient.
MECHANICAL PROPERTIES
Samples for flexural test were successfully produced

by two different processing methods; compression
flask and vacuum casting, as shown in Figure 4.
Then, 3-point bending test was conducted to measure the flexural properties by using universal testing
machine (UTM) conducted at room temperature and
the crosshead speed was 1 mm/min. The test result
(Fig. 5) shows that the flexural strength of samples produced from the compression flask process is 57.7 Mpa.
Whereas, for the vacuum casting process, the flexural
strength is 84.9 MPa which is about 30% higher that
compression flask.
The Morphological of the fracture surface
was observed using scanning electron microscopy
(SEM). Samples were coated with a layer of platinum/palladium at 5 nm thick, using an Agar high
resolution sputter coater (Auto fine coater, JFC-1600,
JEOL, Japan). The SEM was operated with the electron gun set at 10 kV. The microstucture for the vacuum
casting samples shows more brittle surfaces as compared to compression flask. This may due to quick
solidification time as compared to compression flask.
Figure 4. (a) Compression flask process (b) Vacuum casting

process.
Figure 5. Flexural Strength for sample produced from

compression flask and vacuum casting process.
747
ACKNOWLEDGEMENT
This research is supported under Graduate Incentive
Grant (GIS) of University Tun Hussein Onn Malaysia,
the author would like to thank USM dental school,
dental clinic Parit Raja and all team members for
supporting this research.
Figure 6. SEM images of fracture surface from flextural test

for (a) compression flask (b) vacuum casting sample.
The following conclusions are drown from the study;

The 3D profile of the denture teeth sets, the edentulous models and the occlusion rims were successfully
digitized by 3D imaging technique.
A CAD model of the teeth assembly was successfully designed according to standard practices in
dental technologies procedure.
The denture component was successfully produced
using vacuum casting technique, tested and found
satisfactory by the patient.
Flexural properties of the material produced by
vacuum casting process found to be higher that
the conventional process (compression flask technique).
REFERENCES
Y. Kawai, H. Murakami, B. Shariati, E. Klemetti,
J. Blomfield, L. Billette, J. Lund, J. Fein. (2005). Do traditional techniques produce better conventional complete
dentures than simplified techniques. Journal of Dentistry.
33(8). 659668
Schmitt, S., Rapid prototyping custom dental implants,
Rapid Prototyping Report, February 2001.
S. E. Eden, W. J. S. Kerr and J. Brown. (2002). A clinical trial
of light cure acrylic resin for orthodontic use. Journal of
Orthodontics. 29(1). 5155
Yuchun Sun and Yong Wang (2009). Study on CAD&RP
for removable complete denture. Journal computer
methods and programs in biomedicine. 93. 266
272.
Rosochowski A. and A. Matuszak. (2000). Rapid tooling:
the state of the art. Journal of Materials Processing
Technology. 106(13). 191198
Nagahanumaiah, K. Subburaj and B. Ravi. (2008). Computer
aided rapid tooling process selection and manufacturability evaluation for injection mold development. Computers
in Industry. 59(23). 262276.
The future work will be focusing on development

for CAD features in denture design and assembly. This
is the most difficult part of the design process and by
having the dentures features and tools in CAD system
for assembly will reduce the skill required in denture
design process.
748

Investigation of wear behaviour of FDM fixtures

P. Minetola & L. Iuliano
Politecnico di Torino Department of Production Systems and Business Economics, Torino, Italy
ABSTRACT: In the industrial field, the fabrication of custom fixtures was historically one of the first applications of Rapid Prototyping technologies for end-product manufacturing. The advantage of using additive
manufacturing custom fixtures is well assessed in literature in terms of production times and costs, whereas little
attention was given to the life time and wear behaviour of additive manufacturing fixtures. A specific apparatus
was designed for a novel wear test, whose results provide information about the durability of fixtures fabricated
by fused deposition modelling to be used for the inspection of sheet metal parts.
1
INTRODUCTION
Fixtures are a key element for pursuing high quality

standards in manufacturing. They are used whenever a component or a sub-assembly must be aligned
and held accurately and securely during different
operations (machining, assembling, welding, inspection, testing, etc.) along the manufacturing cycle.
Therefore they play a crucial role in components
quality.
Generally, several different fixtures have to be
designed for the same component to fulfil the locating and holding requirements of each single operation.
Dedicated fixtures, that are specifically designed for
a single component, are used in mass productions.
When the manufacturing volumes decrease, the loss of
economies of scale, justifies the use of general purpose
fixtures that can be adjusted to different components.
It has been estimated that fixturing cost can account
up to 20% of the total cost of a manufacturing system
(Bi & Zhang 2001).
General purpose fixtures can be divided into
two groups: conformable fixtures and modular ones.
The formers have internal elements (pins) that are
adjustable to mate different features of workpieces.
The latter show a higher flexibility because the fixture
elements are interchangeable: they can be arranged in
different combinations, so that several configurations
of the fixture are possible. Modular fixtures are easily
adaptable to product changes and ideal for intermittent
or one-time use, especially in the case of low volume
productions wherein the cost of the fixture cannot be
amortized over a large number of parts.
Over the past decades, attention of researchers was
driven to define iterative, semi-automated and automated procedures to ease the design of fixtures with
the aid of the computer (Boyle et al. 2001, Gaoliang
et al. 2010, Hunter Alarcn et al. 2010, Li et al. 2002,
Ryll et al. 2008, Wang et al. 2010, Wu et al. 2008,
Zheng & Chen 2010). It should be noticed that most
of the works related with Computer Aided Fixtures

Design (CAFD) refers to manufacturing applications,
but not to welding, assembling or quality control. During manufacturing the fixture has to accurately locate,
support and clamp firmly the workpiece that undergoes high stresses. In addition to this, the fixture can
be exposed to temperature changes, chemical agents
and mechanical loads.
On the contrary, in assembling and inspection operations, high clamping forces are not required. For
instance, during pointwise measurements performed
on a coordinate measuring machine (CMM), the part
is subjected to very low contact forces and it is not
placed in hostile environment.
As regards the use of modular fixtures in quality
control, an interesting innovation was introduced some
years ago for the customisation of fixture elements
through the use of Rapid Prototyping (RP) technologies (Violante et al. 2007). In the industrial field, the
fabrication of custom fixtures was historically one of
the first applications of RP processes for end-product
manufacturing, that are better known as Additive Manufacturing (AM) techniques nowadays. Materialise
was the pioneer with the release of RapidFit, a fixture design software to combine custom elements with
modular bases and beams.
Once the STL file of the part to be inspected is
imported into the software, the definition of the fixture is quite simple. First of all the base plate of the
modular fixture is inserted. Then the position of a supporting element can be easily selected by clicking on
the desired contact zone on the part surface.
In such a location, RapidFit software generates
automatically the custom element tuning it to the part
geometry (Fig. 1). Standard beams can be mounted
between the base plate and the RapidFit fixture in
order to reduce the length or the height of the custom element to be manufactured. The extent of the
contact zone of the custom element can be adjusted,
but it is not a single point, so the part weight or load is
749
Figure 1. Detail of a custom element matching part feature.
distributed over a certain area reducing contact pressure and deformations (Fig. 2).
When their shape is simple, custom elements
can be fabricated by traditional milling processes.
However AM techniques are preferred for several
reasons.
First of all, layer by layer manufacturing is free from
any geometrical constraint, so any desired shape can
be produced. The only limitation is related with the
overall dimensions of the part that cannot exceed the
building volume of the RP machine. The fixture elements have a small size, so several different custom
elements are fabricated at one job.
The use of AM is particularly worthwhile when
the complexity of the custom element increase. Today
freeform surfaces and complex geometric features are
widely used on large consumption products to enhance
the aesthetics and the attractiveness. Therefore AM
custom elements are an emergent requirement for
manufacturing industries.
Secondly, unlike traditional machining, additive
fabrication does not require any tooling, any tool-path
computation and no scrap is produced. The overall
manufacturing time and cost are reduced as well as
the ecological footprint.
Finally the cost of the AM custom elements mainly
depends on the material used and on the building
time of the RP machine (some hours), but not on the
complexity of the geometry. For applications in the
automotive sector, it was proven that using AM custom elements costs can be reduced up to 66%, while
the fixture lead time is almost cut by half (Eyers &
Dotchev 2010).
Of course the cost of an AM fixture also depends
on the fabrication material. Almost all the literature
about RapidFit fixtures is focused on savings related to
production times and costs. Little attention was given
to their durability.
As regards inspection applications, friction helps a
fixture to hold the inspected part without the use of a
clamping system. On one hand the loading and unloading operations are faster because no clamping device
has to be operated. On the other hand, as a consequence
of friction, fixtures undergo wear during part loading
and unloading and after some time the AM custom
Figure 2. Modular fixture with AM custom elements.
elements have to be replaced. The life time of an element depends on its material and several materials are
available for AM, mainly plastics and metals.
This work focuses on the use of ABS (i.e. Acrylonitrile Butadiene Styrene) custom fixture fabricated
by Fused Deposition Modelling (FDM). The aim is to
evaluate the durability of ABS fixtures assessing it in
terms of number of inspected sheet metal parts. To this
purpose, a specific wear test is proposed and a novel
apparatus was designed to resemble the real wearing
conditions on an FDM fixturing element with a standard geometry. Using the apparatus, the AM element is
subjected to repeated blocks of sliding cycles through
contact with an hole in a small sheet metal part. After
each repetition, wear is evaluated by measuring the size
reduction of the ABS specimen by means of a CMM.
2 WEAR TEST
2.1 Standard wear test methods
Wear test methods are aimed at testing different types
of wear in different conditions for a specific material. In the case of plastic materials, standard methods
for sliding wear evaluation are of the type block-onring (ASTM G137-97 and ASTM G176-03). Other
methods, not specific for plastics are the pin-on-disk
(ASTM G99-05) and the reciprocating sliding (ASTM
G133-05). They are not representative of real wearing
conditions of the fixture, because they do not simulate
the real process sufficiently well (Blau & Budinski
1999).
When a sheet metal part is loaded (and unloaded) on
the supporting fixture for inspection, the AM elements
undergo uniaxial sliding wear in the contact zone with
of part. The contact surface of the sliding part is very
small, because normally it is the lateral wall of a circular or shaped hole of the sheet metal part whose
thickness may reach some millimetres. The allowance
between the fixturing element and the part feature
has to be tight. The fabrication accuracy of the AM
techniques can be as high as 0.05 mm, but the custom
element can be manually adjusted and finished to fit.
750
Figure 3. Apparatus for wear testing AM fixture.
The friction force is low because it is related with the

component of the part weight distributed on the single fixturing element and sheet metal parts are rather
light.
In addition to this, it is difficult to express or convert
the result of a standard wear test in terms of number
of inspected parts. Such value is useful to evaluate the
life time of the fixture and to forecast when the custom
element should be substituted because of the excessive
wear. For its practical implication, this indication is
indeed very important for manufactures, that are not
interested in the tribology of custom fixture surfaces.
Thus standard wear test methods would not provide
satisfactory wear data for this specific application.
For this reasons, a novel apparatus was designed for
a specific wear test, aiming to evaluate the life time
and durability of AM custom fixtures under operating
conditions similar to real ones.
2.2
Novel apparatus
The apparatus provides an alternating uniaxial motion

to an AM fixture element that slides inside an hole
located in a sheet metal part. The part is constantly
kept in contact with the AM element by means of a
spring.
A commercial OWIS Limes 122 motorized positioning system with a step motor was selected to move
the AM specimen during the test. The OWIS linear
stage is placed horizontally on the working table and
an L-shaped support is connected to its carriage. In this
way, an alternating uniaxial horizontal motion is provided to the specimen that is mounted on the L-shaped
support (Fig. 3).
The apparatus is then completed by a vertical device
made of steel that can be addressed as reverseguillotine (Fig. 4). The reverse-guillotine consist of a
base which is connected to the OWIS linear stage and
two vertical guides for the guillotine. The guillotine
is a small block that is limited to move vertically by
its two side wings that are inserted into the grooves of
the vertical guides. A traverse is placed at the top of
Figure 4. Detail of the reverse-guillotine.
the vertical guides and the guillotine is connected to

it by means of a spring. When the spring is extended
from its unstretched position, it forces the guillotine
to move upwards. This is the reason why the vertical
device is called reverse-guillotine: a guillotine normally moves downwards. A sample of a sheet metal
part is mounted on the guillotine, so the spring is
used during the wear test to keep the sheet metal part
constantly in contact with the specimen.
The spring was dimensioned to apply a low contact force to the specimen. An extension spring with
German hooks was used. It has a wire diameter of
0.80 mm, an external diameter of 6.40 mm a free length
of 36.40 mm and a rigidity of 0.59 N/mm.
An additional guiding block manufactured by FDM
of ABS material completes the reverse-guillotine system. The ABS block is connected to the lower border of
the sheet metal part and has two wings that are inserted
into the grooves of the vertical guides. Together with
751
Figure 5. Geometry of the AM specimen.
the guillotine, it limits the vibration of the sheet metal

part during the test. The overall dimensions of the
apparatus are 420 mm 210 mm 135 mm (length
width height).
2.3 Test specimen
The AM specimen for the wear test was designed
referring to a typical RapidFit custom element. A
RapidFit fixture has a square base to connect it to
a modular beam. The specimen has a square base
with four holes to connect it to the L-shaped support
mounted on the OWIS stage. The custom geometry of
a generic RapidFit fixture was simplified to a cylinder,
which is a measurable shape allowing for wear evaluation in a simple way. A 20 mm long round pin of 8 mm
in diameter was used. A supporting cylinder 16 mm in
diameter and 20 mm in height was inserted between
the square base and the pin (Fig. 5).
Three specimens of ABS material were fabricated
by FDM on a Stratasys Dimension Elite machine.
They were manufactured from base to pin with a layer
thickness of 0.178 mm in about 6 hours.
1 mm thick zinc-plated steel sheets with a 10 mm
hole were used as samples of the sheet metal part to
be inspected. For the hole, it was chosen to adopt a
diameter larger than the one of the specimen pin, so
that the contact between the specimen and sheet metal
part is limited to a specific zone located in the lower
half of the pin diameter.
2.4 Methodology
Before starting the wear test, the specimen geometry
was assessed by pointwise measurements performed
on a CMM. Since the wear test length was set to 10 mm
centred on the pin height, the diameter in the middle
of the height was measured on seven sections evenly
spaced every 2 mm (Fig. 6).
The origin of the Cartesian reference system during
the part alignment on the CMM was set at the centre of
the top face of the pin. The Z axis was aligned with the
specimen symmetry axis and the seven sections were
positioned at the following Z levels: 4 mm, 6 mm,
8 mm, 10 mm, 12 mm, 14 mm and 16 mm.
With such a reference system, the wear test zone
extends from Z = 5 mm to Z = 15 mm. This means
Figure 6. Measurement strategy for evaluating pin diameter.
that the diameter of the first section (Z = 4 mm)

and the one of the last section (Z = 16 mm) were
measured outside the worn area.
In each section the pin diameter was measured by
sixteen points (Fig. 6). Five points are located in the
upper half of the pin and are evenly spaced on the semicircle every 45 degrees. The others are located on the
lower half and are evenly spaced every 15 degrees.
Thereby the point density is increased in the wear test
zone to evaluate and monitor diameter changes due to
pin consumption.
After assessing the pin geometry in the unworn
state (zero wearing cycles), the specimen was securely
mounted on the L-shaped support of the OWIS linear
stage. The stage is manually moved forward towards
the reverse-guillotine until the length of pin is centred
through the hole of the sheet metal part. During this
operation the spring was extended 1 mm to keep the
sheet metal in contact with the lower part of the pin.
Secondly, the OWIS stage was actioned to travel
automatically 5 mm forward, then 10 mm backward
and again 5 mm forward to complete one wearing cycle
at a linear speed of 1.20 m/min. Under these conditions, each cycle that is imposed to the ABS specimen
752
by the apparatus represents one loading and unloading operation of the sheet metal part on the supporting
fixtures.
Several wearing cycles were repeated and the pin
diameter was measured again on the CMM every five
hundred cycles by using the same measurement strategy described afore. Since measurement fixtures are
important aspects of the quality assurance of a product, in order to estimate the life time of the fixture,
the maximum wear limit of the pin diameter was set to
0.10 mm. Normally a minimum of four fixtures is used
(Markenscoff et al. 1990), so an increase in the play of
0.10 mm on each fixture compromises the replicability of fixturing and positioning during inspection. For
this reason, blocks of five hundred wearing cycles and
subsequent CMM measurements were repeated until
the diametral wear exceeds 0.10 mm, then the wear test
is complete.
3
RESULTS
The three ABS specimens are addressed with the letters A, B and C for convenience in the following. There
was no difference in the production process: the same
FDM parameters and layer build direction was used,
so theoretically the specimens are identical. Nevertheless, because of FDM process tolerances, the actual
pin geometry of every specimen as fabricated will differently differ from the one of the STL model. Thus,
the unworn pin geometry was assessed by a preliminary measuring phase on the CMM before starting the
wear test.
3.1
Measurements
Every specimen was inspected by a probe tip of 4 mm

on a DEA CMM model GLOBAL Image 07.07.07,
that is equipped with an indexable swinging head and
a touch trigger probe (Fig. 7). The volumetric length
measuring uncertainty MPE of the machine according
to ISO-10360/2 is 1.5 + L/333 m, where MPE is the
acronym for Maximum Permissible Error and L is the
measured length. The pointwise measurements of the
specimens were replicated three times at each measuring phase. The results of the measurements replication
for the unworn A specimen are listed in Table 1. For
each of the seven sections, the average pin diameter
was computed as mean value of the three replications.
Then the wear test was started and measurements
were repeated every five hundred wearing cycles. An
example of the results of worn pin measurements is
shown in Table 2 for the A specimen.
3.2 Wear evaluation
Every five hundred cycles, the wear of the pin was
computed for each of the seven sections by the difference between the average value of the initial unworn
diameter and the one of the current diameter. An example is given in Table 3 for wear evaluation on specimen
A after 500 cycles.
Figure 7. Pointwise measurements of a specimen on the

CMM.
Table 1. Results of the measurements of the A specimen at
0 cycles (unworn pin).
Pin
Diameter (mm)
Average
Section
Meas1
Meas2
Meas3
(mm)
Z = 4 mm
Z = 6 mm
Z = 8 mm
Z = 10 mm
Z = 12 mm
Z = 14 mm
Z = 16 mm
7.928
7.926
7.936
7.946
7.936
7.945
7.948
7.928
7.298
7.937
7.947
7.935
7.944
7.949
7.927
7.926
7.936
7.948
7.935
7.944
7.948
7.928
7.927
7.936
7.947
7.935
7.944
7.948
Table 2. Results of the measurements of the A specimen

after 500 wearing cycles.
Pin
Diameter (mm)
Average
Section
Meas1
Meas2
Meas3
(mm)
Z = 4 mm
Z = 6 mm
Z = 8 mm
Z = 10 mm
Z = 12 mm
Z = 14 mm
Z = 16 mm
7.919
7.909
7.920
7.929
7.915
7.930
7.954
7.918
7.910
7.919
7.929
7.914
7.929
7.955
7.920
7.911
7.920
7.929
7.916
7.930
7.952
7.919
7.910
7.920
7.929
7.915
7.930
7.954
A photograph of the worn pin surface of Specimen

A after 2500 wearing cycles is shown in Figure 8. It is
possible to visualize a rough profile of the pin geometry at a specific number of wearing cycles by plotting
the value of the average diameter of each section versus
its Z level or position. An idea of the pin consumption
and variation of the diameter of each section can be displayed by superposing the profiles of the pin geometry
corresponding to different numbers of cycles (Fig. 9).
753
The vertical dashed lines in Figure 9 represent the

limits of the wear zone, which ranges from Z = 5 mm
to Z = 15 mm. Profiles referring to 1000 and 2000
wearing cycles are omitted to increase figure readability. It can be noticed that the value of the average
diameter of the two sections located outside the wear
Table 3. Results of the wear evaluation for the A specimen
after 500 cycles.
Pin
Average Diameter (mm)
Diametral
Section
0 cycles
500 cycles
Wear (mm)
Z = 4 mm
Z = 6 mm
Z = 8 mm
Z = 10 mm
Z = 12 mm
Z = 14 mm
Z = 16 mm
7.928
7.927
7.936
7.947
7.935
7.944
7.948
7.919
7.910
7.920
7.929
7.915
7.930
7.954
0.009
0.017
0.016
0.018
0.020
0.014
0.006
Figure 8. Pin surface of Specimen A after 2500 wearing

cycles.
area does not changes significantly with the number

of wearing cycles. This is not surprising, since the
two sections lie outside the wear zone and should not
undergo wearing during the test. Nevertheless it is
important to observe that in some cases, the dimension of the average diameter of the external sections
on worn pins is greater than the corresponding one on
the initial unworn specimen. Consequently the diametral wear has a negative sign because it is computed by
difference from the dimension of the unworn diameter. For example, this was the case of last section
(Z = 16 mm) of the A specimen after five hundred
wearing cycles, whose values are listed in the last line
of Table 3.
As a consequence of the sliding action, the ABS
material of the specimen is consumed and removed.
However due to its high plasticity, it also partially flows
to the ends of the wear zone because it is deformed by
the sheet metal part and pushed ahead in the sliding
direction. The plastic deformation is responsible for
the enlargement of the pin diameter in the two sections
which lie outside the wear zone. The material consumption is predominant and the plastic deformation
can be disregarded inside the wear area.
In addition to this, as the number of wearing cycles
increases, the profiles of the pin geometry in the wear
zone do not remain parallel (Fig. 9). As a matter of fact,
over 1500 cycles the pin wear is greater in those sections (Z = 14 mm and Z = 16 mm) that are nearer
to the supporting cylinder of 16 mm. This fact suggests that during a wearing cycle the pin behaves like
a cantilever supported only on the side of the 16 mm
cylinder.The contact force imposed by the spring to the
sheet metal part causes the pin bending. The bending
moment and the consequent deflection are larger for
sections that are far from the fixed end of the pin. As
wear cycles increase, the material consumption grows
and the pin cross section decreases. Thus the bending
resistance of the pin is lower and the higher deflection
Table 4. Results of the wear evaluation for the three ABS

specimens.
Number
of cycles
0
500
1000
1500
2000
2500
3000
Figure 9. Profiles of pin geometry at a different numbers of wearing cycles.
754
Average Wear (mm)

Specimen
0.000
0.017
0.025
0.033
0.037
0.085
0.114
0.000
0.018
0.022
0.031
0.048
0.072
0.105
0.000
0.011
0.023
0.035
0.042
0.093
0.131
Figure 10. Profiles of pin geometry at different numbers of wearing cycles.
restraints the material consumption because the pin

bends away from the hole. Deflection is partially compensated by the action of the spring and is allowed
by the 2 mm play between the dimension of the pin
(8 mm) and the one of the hole (10 mm). These considerations and the results shown in Figure 9 for the A
specimen are similar for the others.
In order to have a unique reference value for the pin
wear at each number of wearing cycles, the average
wear of the pin was evaluated by the mean value of
the diametral wear of the five sections that are inside
the worn zone. The external sections located at the
positions Z = 4 mm and Z = 16 mm were not considered because they are not affected by material
consumption.
For instance, the average wear of the A specimen at
500 cycles is 0.017 mm, that is the mean of the values
from the second to the sixth row in the last column of
Table 3.
The value of the average wear at different numbers
of cycles for the three specimens is reported in Table 4
and the corresponding graph is shown in Figure 10.
The average wear of each of the three specimen at
3000 cycles exceeds the defined limit of 0.10 mm. It
can be estimated that they reach such a limit at different
numbers of cycles ranging from 2500 to 3000.
CONCLUSIONS
Referring to the inspection of sheet metal parts, the

life time and wear behaviour of AM fixture of ABS
material were investigated in this paper.
Since standard tests for wear evaluation are not representative of real wearing conditions of the fixture, a
specific apparatus was designed for a novel test. The

apparatus provides an alternating uniaxial motion to
an AM fixture that slides inside an hole located in a
sheet metal part. During the test, the hole is constantly
kept in contact with the fixture by means of a spring.
The specimen for the wear test was designed with a
standard geometry represented by a 20 mm long circular pin of 8 mm in diameter. The wear test length was
set to 10 mm centred on the pin height and a sliding
speed of 1.20 m/min was used. Wear tests were conducted on three ABS specimens that were fabricated
by FDM.
Each cycle that is imposed to the AM specimen by
the apparatus represents loading and unloading operations of the sheet metal part on the supporting fixture.
In such a way, wear was evaluated in terms of number
of inspected components before the fixtures should be
replaced.
Several wearing cycles were repeated and the pin
geometry was measured by a CMM every 500 cycles.
The pin consumption was evaluated on seven diameters evenly spaced every 2 mm over the pin height.
The maximum wear limit of the pin diameter was set
to 0.10 mm.
The three ABS specimens reached such a limit in
different times, corresponding to numbers of wearing
cycles between 2500 and 3000. Thus their life time
can be conservatively assessed in two thousand five
hundred inspected sheet metal parts.
During the test, the AM fixture underwent wearing
under worse than normal conditions. Because of the
action of the spring and the 2 mm play between the pin
diameter and the size of the hole, wearing was always
applied on the same zone of the pin surface. On the
contrary, in real use, the fixturing tolerances are tight,
755
so the contact zone between the fixture custom element

and the mating feature can vary from inspected part to
inspected part. For this reason, the life time of the AM
fixture was underestimated by the specific wear test
proposed in this study.
The proposed methodology and research activity
can be extended to other materials available for AM
of custom fixtures to compare their durability.
ACKNOWLEDGEMENTS
The assistance of Mr. Alex Borghesio during the
research activities is gratefully acknowledged. Thanks
are also extended to Mr. Giovanni Marchiandi who
looked after the fabrication of the specimens and wear
test apparatus.
REFERENCES
Bi, Z.M. & Zhang, WJ. 2001. Flexible fixture design and
automation: review,issues,and future directions. International Journal of Production Research 39(13): 2867
2894.
Blau, P.J. & Budinski, K.G. 1999. Development and use of
ASTM standards for wear testing. Wear 225(2): 1159
1170.
Boyle, I., Rong, Y. & Brown, D.C. 2001. A review and analysis of current computer-aided fixture design approaches.
Robotics and Computer-Integrated Manufacturing 27:
112.
Eyers, D. & Dotchev, K. 2010. Technology review for

mass customisation using rapid manufacturing. Assembly
Automation 30(1): 3946.
Gaoliang P., Gongdong W., Wenjian L. & Haiquan Y. 2010. A
desktop virtual reality-based interactive modular fixture
configuration design system. Computer-Aided Design 42:
432444.
Hunter Alarcn, R., Ros Chueco, J., Prez Garca, J.M. &
Vizn Idoipe, A. 2010. Fixture knowledge model development and implementation based on a functional design
approach. Robotics and Computer-Integrated Manufacturing 26: 5666.
Li, W., Li, P., & Rong, Y. 2002. Case-based agile fixture
design. Journal of Materials Processing Technology 128:
718.
Markenscoff X., Ni L. & Papadimitriou C.H. 1990, The
geometry of grasping. International Journal of Robotics
Research 9(1): 6174.
Ryll, M., Papastathis, T.N. & Ratchev, S. 2008. Towards
an intelligent fixturing system with rapid reconfiguration and part positioning. Journal of Materials Processing
Violante, M.G., Iuliano, L. & Minetola, P. 2007. Design and
production of fixtures for free-form components using
selective laser sintering. Rapid Prototyping Journal 13(1):
3037.
Wang, H., Rong, Y., Li, H., & Shaun, P. 2010. Computer aided fixture design: Recent research and trends.
Computer-Aided Design 42: 10851094.
Wu, Y., Gao, S. & Chen, Z. 2008. Automated modular fixture
planning based on linkage mechanism theory. Robotics
and Computer-Integrated Manufacturing 24: 3849.
Zheng,Y. & Chew, C.M. 2010. A geometric approach to automated fixture layout design. Computer-Aided Design 42:
202212.
756

The application of Laser Sintering for archaeological model-making

S.P. Soe, D.R. Eyers & A.T. Potter
Cardiff Business School, Cardiff, UK
T. Jones
Newport Medieval Ship Project, Newport, UK
N. Nayling
School of Archaeology, History and Anthropology, University of Wales Trinity Saint David, Lampeter, UK
ABSTRACT: Model-making is widely used in archaeological research allowing the creation of replica artefacts
at varying scaled sizes. These models are important tools for constructing visualisations of relics, and (in some
cases) for exploration of their physical and mechanical characteristics. This paper demonstrates the use of
Rapid Prototyping technologies in the practice of archaeological model-making, with particular emphasis on
the production of model ships. As these intricate models are typically constructed from a large number of
components, each must be individually fabricated and assembled to form the overall artefact. By examining the
traditional hand-crafted approach to model-making, and considering more recent virtual approaches, this paper
highlights opportunities to employ Rapid Prototyping technologies, focusing particularly on the capabilities of
the Laser Sintering process. Through a detailed examination of two case studies, attention is given to the design
and manufacture phases of model-making. For design, the application of reverse-engineering technologies to
achieve an accurate model of the artefact is discussed, along with a consideration of limitations arising from this
approach. For manufacturing using Laser Sintering, the specific challenges for the application are presented,
including a detailed consideration of issues arising from the production of a scale-item.
1
INTRODUCTION
The techniques of model-making have long been

practiced for archaeological applications in order
to provide tangible physical copies of historic artefacts. Models can be employed in research activities,
whereby archaeologists may interact with the physical
items to understand the aesthetic and functional characteristics in order to explain the artefact, or can be
used as tools in teaching suitable for providing illustration and visualization for the audience. Crucially,
models are required to achieve a high degree of accuracy in order to be useful tools for research, rather than
being merely pretty pictures. (Sims 1997). If details
are omitted or are incorrectly reproduced in the model,
future researchers who are reliant on the model may be
adversely affected, and therefore archaeological models must be faithful and accurate reconstructions of the
original item.
The production of archaeological models has
received relatively little research attention from the
perspective of enhancing manufacturing processes.
Normally produced at very low volumes (single models are commonplace), the manufacturing process has
remained a labour-intensive craft process for many
years, using traditional materials such as wood and
paper in the fabrication of replica items. More recent
forays into virtual modelling have provided an alternative to the craft approach, though these have often
been shown to be expensive, and do not enjoy universal
acceptance in the archaeological community.
In other industries the requirement for very low
volume, accurate, production of models is widely
found in the production of prototype products, for
which the technologies of Rapid Prototyping (RP)
are widely employed. For example, in the architecture industry RP models allow for visualization of
design ideas, the examination of processes, and also
the testing of concepts (Gibson et al. 2002). Similarly in medicine, physical models are used in both
teaching and the preparation and practice of surgical procedures (Hieu et al. 2005). RP processes have
been demonstrated as particularly suitable approaches
to mitigate many of the trade-offs associated with
such low volume/customized production, particularly
in terms of time and cost (Eyers and Wong 2010).
This paper therefore presents the application of the
Laser Sintering (LS) RP technologies in the production of archaeological models. Already RP processes
have received application in the production of skeletal
replicas (e.g. Fantini et al. 2008; Zhang et al. 2000),
however as yet little research has considered archaeological model-making with explicit focus on LS. In
this paper we demonstrate a three s approach to
757
the formation of an archaeological model through the

activities of scanning (including reverse engineering),
scaling, and sintering, and for each of these consider the specific challenges for the archaeological
application.
2
CONVENTIONAL APPROACHES TO
ARCHAEOLOGICAL MODEL-MAKING
There are multiple definitions of the modelconcept in

archaeology, from the creation of replica items, to the
simulation of historical events through the production
of modelling simulations. This paper is concerned only
with the production of replica three-dimensional models, and therefore employs the definition of a model
as a three dimensional representation of a person or
object, typically on a smaller scale than the original
(Soanes and Stevenson 2005).
2.1
Lock (2003) provides a more recent evaluation, highlighting the progress made in the combination of
reconstruction modelling and virtual reality as a result
of increases in both computing capabilities, and also in
the archaeology disciplines awareness of the capabilities of technology in model-making. Looking forward,
it is likely that continual improvements in modelling
processing and computing technology will increase
acceptance of virtual modelling in archaeology in
much the same manner as has been shown in virtual product prototyping. However, just as product
designers value the opportunity to interact with physical models, so too does the argument exist for a
continuation of the production of physical models
for archaeological applications. In the next section
we therefore consider how 3D computer modelling
may also be employed in the construction of physical
artefacts.
Craft approaches
In order to construct a physical model of an archaeological artefact, craft producers need to achieve a
good understanding of the nature of the original item.
Geometric properties such as size and shape may be
achieved using measurement equipment (varying from
simple measurement with rulers and tapes through to
the application of precision measurement technology).
Tracings, rubbings, sketching, and photography may
also be used to achieve 2D records of the original piece.
Using these information resources, the craft modeller produces their replica artefact, typically in a
skilled, manual process. High quality items are possible in this approach, though the achievement of
this necessitates highly skilled craft-workers, expert
guidance on materials, and often extended periods of
time for the model-making to be achieved. Hand-made
items can therefore be costly to produce, and are also
subject to potential variation as a result of the manual
processes.
2.2 Virtual modelling
Increasing computing capabilities in the 1980s saw a
significant increase in the popularity of virtual modelling, whereby artefacts are recreated in a digital form
on the computer screen using either surface or solid
modelling. Reily (1992) overviews the earliest developments in this area, focusing on the (then) relatively
emergent application of solid modelling, from which
significant limitations in the technologies and techniques are readily evident. Forte and Siliotti (1997)
provide a range of examples of the application of virtual modelling during the 1990s for archaeological
applications.
For earlier implementations, the need to build models by manually keying data into the computer (in
order to describe object), combined with inadequacies in computer capabilities (especially processing
powder) significantly constrained modelling attempts.
PRODUCTION OF LASER SINTERED

ARCHAEOLOGICAL MODELS
3.1 Scanning & reverse engineering

To generate the STL input files necessary to facilitate the LS build, two principal options are available:
manual 3D design using an application (such as SolidWorks or Pro Engineer), or the reverse engineering of
an existing item using a 3D scanning tool. Reverse
engineering systems may be classified as contact or
non-contact; the latter laser-based approaches tend
to be significantly faster in data acquisition, though
may experience difficulties when scanning shiny surfaces. Considerable research in the integration of
RP and reverse engineering has already been performed; the reader is directed to Raja and Fernandes
(2008) for direction in current state-of-the-art reverse
engineering.
For archaeological applications for which there
exists an existing physical artefact, the reverse engineering approach is typically most sensible on the
grounds of model creation time and potential for
higher levels of accuracy in reproducing the item. General guidance for the conduct of this activity may be
identified in the five-step process provided by Dotchev
et al. (2009): 1) estimation of geometric configuration,
2) digitization, 3) data pre-processing, 4) generation of
the CAD model, and finally 5) an examination of the
model to ensure accuracy within acceptable tolerance
levels.
3.2 Scaling
One of the main benefits of model-making for archaeological applications is the ability to alter the scale of
the item to suit its intended audience. Although lifesize replicas such as King Tutankhamen by Materialise
(2010) are desirable, in many applications smaller representations are more appropriate on practical grounds
such as storage and cost. Similarly, increasing the
758
scale of items can be beneficial where the original

artefact is miniature. Whilst craft-based approaches
would require human calculation of these changed
geometries, for virtual and RP modelling techniques
it is a relatively straightforward task to instruct the
computer to quickly re-compute the model to the
desired scale. However, whilst scaling if often beneficial, it is necessary to consider the implications
of performing this transformation. Whilst the computational requirements to rescale original articles to
model size are largely performed arithmetically, representing these in physically manufactured products
can become problematic due to manufacturing constraints. In a reduction scaling, important features and
detail from the original artefact can be lost since they
may not be adequately replicated in the manufacturing
process. Similarly, increasing the model to beyond the
size of the original may result in distortions of fine
details.
The extent to which a RP technology may accommodate a reduced scale product will be dependent on
a combination of process choice, material selection,
and machine configuration. For LS, typically the minimum operational thicknesses which can be produced
is 1mm, (though the machines can be tuned to produce parts at the 0.5 mm range). However, at this scale
important features (such as walls) may become very
small (<1mm), making their fabrication difficult and
highly susceptible to damage in the post-processing
activities of power removal.
3.3
Sintering
The processes of sintering in LS have received considerable research attention, particularly for issues
of material characteristics and selection (Schmidt
et al. 2007), build processes (Pham and Wang 2000),
and material recycling (Dotchev and Yusoff 2009).
Although each RP process has its own merits, one distinct advantage of LS over other processes such as
Stereolithography and Fused Deposition Modelling is
the negation of support structures for the build process.
By removing this requirement, much time is saved
in both build preparation, and also in part building,
where the laser does not need to be exposed on crosssectional areas of support geometry in each layer. This
also makes better use of the overall build chamber
capabilities: instead of using support materials which
will be discarded, parts can be nested together to
fill gaps in the chamber. However, whilst LS is able
to produce highly complex parts such as inlet manifolds and lattice structures, there are a number of
disadvantages particularly in terms of accuracy and
repeatability. The thermal nature of the LS process can
lead to part curling and warping (particularly for larger
parts), which can have negative effects for the overall product quality. Similarly LS surface finishing can
often be poor, particularly when recycled materials are
used. Whilst material recycling reduces costs and has
environmental benefits, the degradation in the material as a result of prior heating operations may lead
to the production of parts with an unwanted surface

texture called orange-peel, and also some deviation from the normally expected material performance
characteristics.
4
CASE EXAMPLES
4.1 The Newport Ship

The Newport Ship is a 15th-century medieval ship
recovered next to the River Usk in South Wales (UK)
between 2002 and 2003, and is one of the largest
and most complete ships from this period which
has been excavated and raised in Western Europe.
Believed to be French in origin, the ship was constructed from several thousand timbers, nailed together
in an overlapping manner (known as the clinker shipbuilding approach). Over a six month period in 2002,
archaeologists raised nearly 2,000 timbers of various
dimensions, each of which was placed in protective
storage for future documentation, conservation, and
eventual display.
The traditional approach to archaeological modelmaking for ships focuses on the production of a
scaled reconstruction of the original hull-form, which
includes removal of deformation caused by factors such as compression in the burial environment
(Crumlin-Pendersen and McGrail 2006). Often, these
models are constructed from cardboard or wood, and
are often hand-made from careful measurements of
the original artefact. Thomsen (2010) describes the
process employed in a Danish ship reconstruction,
reporting detailed prints of each timber being glued
to pieces of cardboard, which in turn were affixed to
plywood frames.
A number of limitations are evident from this
approach. The nature of the cardboard material, whilst
inexpensive and easily available has highly limited
functional characteristics (i.e. strength, durability etc).
The requirement for manual measurement and cutting of many hundreds of cardboard timbers is also
time consuming, and therefore from a labour utilization perspective, a costly process. Aesthetic quality of
this approach may be low; Thomsen (2010) notes in
the Danish example the cardboard models, whilst adequate as research tools would not necessarily be as
suitable for museum exhibition. Jones (2009) makes
a further important observation that many conventional techniques tend to be employed in applications
which reduce the original three-dimensional artefact
to a two-dimensional model, where much important information cannot be conveyed and hence is
lost; an eventuality which Sims (1997) identifies as
undesirable.
4.1.1 Scanning & reverse engineering
The Newport Ship is built of three timber components: keel/stem/stern, a series of frames, and a number
of planks (Figure 1). Every timber recovered from
the ship is unique: length, width, holes (for fastening by iron nail or wooden treenail) and curvatures are
759
Figure 1. CAD model of Newport Ship components.
Figure 3. Model demonstrating planks and screws.
Figure 2. FaroArm scanning of ship timbers.
all individual, requiring each artefact to be carefully

recorded. In traditional approaches, this would have
been achieved by scale drawings or full-size tracings
of the timbers, accompanying these with careful direct
measurements. However, in the Newport Ship project,
FaroArm co-ordinate measuring machines and noncontact laser scanning systems were used by a team
of project staff (numbering fifteen at peak scanning
times), who spent two years meticulously recording
the timbers of the ship (Figure 2) in order to achieve
a digital representation, with Rhino 4 being used to
create and assemble the data as a 3D representation
(Jones 2009). This application of FaroArm technologies to attain a 3D image of the timbers has been
considered pioneering in the field of reconstructive
ship archaeology (Thomsen 2010).
4.1.2 Scaling
The Newport Ship was originally over 30 metres in
length, for which a 1:10 scale representation of the ship
was desired and each timber proportionally scaled.
The scaled model plank parts are long, reasonably flat
pieces ranging from 200 mm to 450 mm in length with
an average width of 25 mm and average thickness of
3 mm.
For some original timbers which were particularly
thin it was identified that scaling would produce parts
which would have issues in manufacturing. As the
already thin pieces were reduced to a tenth of original
size, it was apparent that detail at the edges would be
lost, and the edges would gain a sharp feathering which
would be awkward in post-processing and to detriment

of the finished piece. Fortunately in this application
the consequences of this limitation were eliminated
in the assembly (where planks are overlapped due to
the nature of clenched nail building style), however
it is clear other application scenarios might not be as
fortunate. Hence, whilst LS can work at millimetre
accuracy, how scaling affects manufacturing is still
a limitation which must be acknowledged for other
applications.
A second scaling consideration issue concerned the
clinker nail holes, through which hundreds of nails
would have originally held the ship together (Figure 3).
To understand how these nails held and shaped the ship,
the model intended to recreate the holes in exactly
the same position, using screws to pull the timbers
together. However, at 1:10 scaling the holes diameter
became 1 mm, which while easily fabricated in the
LS process, sourcing appropriate screws and conducting assembly were impractical. Resultantly, the clinker
holes were produced at a larger scale on the grounds
of functional practicality. This was determined to not
significantly affect the properties of the ship, but is an
acknowledged difficulty which is likely to be relevant
for other studies.
4.1.3 Sintering
The model Newport Ship was fabricated in Polyamide
12 using an EOS P700 LS machine equipped with dual
50W CO2 lasers. Having a build chamber capacity
of 700 mm 380 mm 580 mm, the EOS machine
could build any of the parts in any build orientation
(without the need for splitting and re-joining in postprocessing). Although the build chamber could have
produced all components in a single build, the project
team exploited the opportunity to perform incremental fabrication through a series of 10 smaller batches.
This was motivated by the relative bottleneck of the
Rhino 3D scanning: whilst the LS machine could have
supplied all the parts within 48 hours, the process of
3D scanning was comparatively slow, with each timber
taking many hours to scan and prepare for building.
Compared to the conventional cardboard craftapproach to model ship-making, the materials
employed in this example are significantly more
expensive (total LS build cost 4000), and when
760
combined with the time-consuming 3D scanning process it is evident that the current LS approach is
costly. However, the increased accuracy of the model,
combined with the material characteristics of the
polyamide material has provided robust end product for both public viewing archaeological use (BBC
2009).
4.2
Pre-christian stones
A second example of archaeological model-making in

LS is the production of Medieval stones for display at
Ulster Museum, through the replication of the original
pieces held at Armagh Cathedral. Each stone is 0.5
metres tall, and weighs 2030 kilos.
4.2.1 Scanning
The scanning of the stones was performed at Armagh
Cathedral using a contactless Konica-Minolta VI-900
scanner over a twelve hour period (Deri Jones &
Associates 2010). Weighing approximately 11kg, this
portable unit is able to scan a surface in 2.5 seconds,
converting this recorded stone surface to a lattice of
over 300,000 vertices.
4.2.2 Scaling
For this application the replica stones were fabricated
at life-size, therefore requiring no scaling. The choice
of manufacturing material, and the intentional hollowing of the parts did result in a 90% weight reduction,
though this does not affect the appearance of the
finished objects.
4.2.3 Sintering
One of the major challenges of this application was to
achieve a realistic reproduction of the medieval stones
for display in a museum environment, necessitating
much consideration of the aesthetic qualities of the
artefact. This was particularly challenging given the
highly faceted nature of the part surface. As each stone
would be professionally painted in post-processing, at
the manufacturing stage it was necessary to conduct
a series of trial builds to evaluate the best approach
with regards to build orientation and the initial postprocessing activities of powder removal. Several samples were sent to the model-maker before the final
configuration was decided; subsequently it was possible to build two model stones in a single LS build. The
EOS P700 machine employed in this task operated at
10 mm per hour; the completed stones took 40 hours to
build at a combined cost of approximately 3,000. On
build completion excess powder was removed to reveal
the white stones before shipping to the model-maker
to be hand-painted (Figure 4).
5
CONCLUSION
This paper has considered the challenges facing

archaeologists and manufacturers in the production of
replica artefacts suitable for research and teaching.
Through case studies of archaeological reconstructions it has demonstrated the application of laser
Figure 4. LS model of a stone representation of a dog.
sintering to the production of very different archaeological items using the same RP production processes.
For the Newport Ship, RP was shown to produce
thousands of model timbers, which when assembled
allowed archaeologists to gain an accurate understanding of the shape of the ships hull. By comparison, the
pre-Christian stones are larger single-piece items for
which accuracy in overall geometry and surface detail
were most important. For each example, the paper has
shown how reverse engineering combined with LS
may reproduce the details of historic artefacts in robust
polyamide material for examination and investigation,
and highlighted the challenges for each approach.
ACKNOWLEDGEMENTS
The case study artefacts presented in this paper were
all manufactured at the Manufacturing Engineering
Centre, Cardiff University 20092010 in conjunction
with other projects listed below
Newport Ship
The ShipShape: solid modelling and visualisation
of the Newport Medieval Ship from 3D digital
record project was funded by the Arts and Humanities Research Council with support from Newport
Museum and Heritage Service. Ship and model photos
courtesy of Newport Museum and Heritage Service.
Archaeological Stones
This project was commissioned by National Museums
Northern Ireland, and was fulfilled by Deri Jones &
Associates Ltd and Geospatial Survey Solutions Ltd.
Painting of the finished items was conducted by
Specialist Models Ltd.
REFERENCES
BBC. 2009. Newports medieval ship model is taking shape
[Online]. Available at: http://news.bbc.co.uk/local/ southeastwales/hi/people_ and_places/history/newsid_ 8421000/8421130.stm
Crumlin-Pendersen, O. and McGrail, S. 2006. Some principles for the reconstruction of ancient boat structures.
International Journal of Nautical Archaeology 35(1):
5357.
761
Deri Jones &Associates. 2010. 3D stone replication [Online].

Available at: http://www.djaweb.co.uk/index.php/userstories/heritage/94-3d-stone-replication
Dotchev, K. and Yusoff, W. 2009. Recycling of polyamide
12 based powders in the laser sintering process. Rapid
Dotchev, K. D., Popov, I. E. and Eyers, D. R. 2009. Rapid
Manufacturing and Reverse Engineering: Key issues for
Mass Customisation. In: Nenov, P. ed. International Conference on General Machine Design. Rousse, Bulgaria,
1516th October 2009. University of Rousse.
Eyers, D. R. and Wong, H. W. 2010. Addressing trade-offs
in Mass Customisation using Additive Manufacturing
technologies. In: Forza, C. ed. 17th Annual EurOMA
Conference. Porto, Portugal, 69 June 2010.
Fantini, M., de Crescenzio, F., Persiani, F., Benazzi, S. and
Gruppioni, G. 2008. 3D restitution, restoration and prototyping of a medieval damaged skull. Rapid Prototyping
Journal 14(5): 318324.
Forte, M. and Siliotti, A. eds. 1997. Virtual archaeology:
Great discoveries brought to life through virtual reality.
London: Thames & Hudson.
Gibson, I., Kvan, T. and Ming, L. W. 2002. Rapid prototyping
for architectural models. Rapid Prototyping Journal 8(2):
9199.
Hieu, L. C., Zlatov, N., Vander-Sloten, J., Bohez, E., Khanh,
L., Binh, P. H., Oris, P. and Toshev, Y. 2005. Medical
rapid prototyping applications and methods. Assembly
Automation 25(4): 284292.
Jones, T. 2009. Three-dimensional recording and digitial
modelling of the Newport Medieval Ship. In: ACUA
Underwater Archaeology. Toronto.
Lock, G. 2003. Using computers in archaeology: towards

virtual pasts. London: Routledge
Materialise. 2010. Materialise builds 3D replica of King
Tuts mummy on Mammoth stereolithography machine
[Online]. Available at: http://www.materialise.com/
materialise/view/en/3256368
Pham, D. T. and Wang, X. 2000. Prediction and reduction of build times for the selective laser sintering
process. Proceedings of the Institution of Mechanical
Engineers, Part B: Journal of Engineering Manufacture
214: 425430.
Raja, V. and Fernandes, K. J. eds. 2008. Reverse engineering:
an industrial perspective. London: Springer.
Reily, P. 1992. Three-dimensional modelling and primary archaeological data. In: Reily, P. and Rahtz, S.
eds. Archaeology and the information age. London:
Routledge.
Schmidt, M., Pohle, D. and Rechtenwald, T. 2007. Selective
Laser Sintering of PEEK. CIRP Annals-Manufacturing
Technology 56(1): 205208.
Sims, D. 1997. Archaeological models: Pretty pictures or
research tools? IEEE Computer Graphics and Applications 17(1): 1315.
Soanes, C. and Stevenson, A. eds. 2005. Oxford Dictionary
of English. Oxford: Oxford University Press.
Thomsen, C. H. R. 2010. Reconstructing the lines of
the Princes Channel ship. University of Southern
Denmark.
Zhang, G. M., Tsou, Y. C. and Rosenberger, A. L. 2000.
Reconstruction of the Homunculus skull using a combined scanning and stereolithography process. Rapid
762

Novel concept for manufacturing lightweight centrifuge rotors for

laboratories
U. Klaeger
Fraunhofer Institute for Factory Operation and Automation IFF, Magdeburg, Germany
V. Galazky
carbonic GmbH, Haldensleben, Germany
ABSTRACT: Centrifuges have multiple uses in medicine and in laboratories, most notably to separate sub
stances such as blood plasma and serums. A centrifuges rotors hold sample containers. Centrifugal forces
equaling 25,000 times the force of gravity act on rotors and samples at speeds of 20,000 r/min.
Unlike present conventional solutions (rotors made of special aluminum alloys), the load-bearing structures of
the lightweight rotors presented here are made of carbon fiber-reinforced polymers (CFRP). The complex rotor
and mold geometries are von selective laser sintered. Given the fibers extremely high load-bearing capacity,
lightweight rotors manufactured with this novel technology set new standards for weight, stability and service
life compared to current concepts. Moreover, this can reduce the warm-up time typical for aluminum rotors by
at least 50%.
INTRODUCTION
Centrifugation is the process most frequently applied

to separate materials in liquids. Most notably, they
are used to separate materials, e.g. blood plasma
and serums, and produce genetically engineered substances.
A centrifuge consists of a housing, a drive unit with
controller, a rotor, a safety enclosure and, frequently,
a cooling system. The rotor holds sample containers.
There are rotors for different sizes of samples (from the
microliter range to one liter) and numbers of sample
containers (depending on the task). They are subjected
to extreme mechanical loads. Technically this makes
them core centrifuge components.
During centrifuging, a solutions solid constituents
precipitate under the effect of a stronger gravitational
field produced by the centrifugal forces generated
by rapid rotation. In this stronger gravitational field,
constituents with greater mass displace lighter particles, which are thrust closer to the axis of rotation.
The gravitational force increases in proportion to the
distance from the axis of rotation (radius). Superior
centrifuges operate at high speeds of frequently more
than 20,000 r/min. This produces gravitational fields,
which exceed normal gravitation several thousands of
times over. Conventional rotors are made of special
aluminum alloys. They are relatively easy and cheap
to manufacture but have drawbacks in terms of their
stability and attainable rotational speeds. In addition,
undesirable imbalances frequently appear.
Therefore, lightweight rotors are alternatively made
of fiber composite materials. A variety of methods
exist but, at present, only rotors with resin transfer

molded bodies and wound highly stressed annular
shells are commercially available.
Admittedly, lightweight rotors made entirely or partially of carbon fiber-reinforced polymers or mixed
are more expensive than conventional aluminum rotors
since they are predominantly manufactured by hand.
However, in addition to having a substantially lower
density, they weigh far less and are approximately six
times more stable than aluminum rotors.
Lower weight cuts centrifugesenergy consumption
and ramp-up times and also reduces an overall centrifuge systems mechanical loads. In addition, lighter
CFRP rotors simplify handling since aluminum rotors
often weigh 20 to 25 kg.
2
MOTIVATION AND AIMS
The Fraunhofer IFF and carbonic GmbH jointly developed a method of manufacturing lightweight rotors
from pre-molded woven carbon fibers. Carbon fibers
have high tensile strength, provided the fibers are
aligned with the direction of load. They are thusly
processed relatively easily when profiles are long and
shapes are flat or cylindrical.
However, tapering and freeform surfaces like those
of centrifuge rotors are more complicated since the
fibers are unable to adhere to these surfaces and slip
easily. Hence, rotors are only wound at present. The
drawback of this is that the fibers cannot be aligned
with the direction of load. Further, the manufacture of
wound rotors requires extremely expensive multi-axis
763
winders and wound surfaces are never really smooth.

This affects a rotors running smoothness adversely.
The new methods of positioning fibers aligned with
the direction of load on conical surfaces employs carbon spiral tapes that exact match the winding of a
rotors tapered surface geometrically. In a first step,
a base body that holds sample containers and a hub
with a conical exterior shape are manufactured. Then,
the spiral tape is placed around the base body. Since
the geometry of the spiral tape and base body corresponds, the spiral tape stays in place and only the two
ends of the tape have to be secured with some spray
adhesive. The base body layered with spiral tape is
placed in a second mold and impregnated with resin
by RTM. This design reduces the moment of inertia by
more than half.
The complicated design principal considered by
these authors requires suitable forming tools that
reproduce the complex geometries (e.g. undercuts).
Therefore, generative (laminate) methods of geometry
generation are used to make molds. Given their practically unlimited freedom of design, these technologies can, for instance, already produce close-contour
cooling channels during mold making. Furthermore,
selective laser sintered, geometrically complex inserts
reduce the weight of rotor bodies.
When this novel concept is successfully implemented, a lighter weight CFRP rotor will be at least
10% more stable than a lightweight wound rotor. At
the same time, it can be expected to have a permissible speed that is at least 10% higher than that of
lightweight rotors in the same class. First, comprehensive physical models and new calculation algorithms for the FEM analysis were developed, which
ensure that lightweight rotors can be manufactured
reproducibly for future implementation in practice.
Figure 1. Rotor design calculation and simulation.
Figure 2. Critical rotor geometry zones.
CONCEPTUAL DESIGN OF THE

MANUFACTURING TECHNOLOGY
First, the CAD models of the two sizes of CFRP rotor

analyzed (14 50 ml and 6 500 ml) were generated.
They served as the basis for subsequently calculating
the shell design and potential failure criteria (maximum stress, maximum strain, etc.) based on the finite
element method (FEM), Figure 1.
The ANSYS analysis package was used for the calculations since it contains tools that are especially
effective for the calculating the structures of fiber
composite materials.
This tool was also used to identify and calculate
the principal stresses produced during centrifugation
and the resultant critical zones in the rotor geometry
(Figure 2).
Taking the properties required of the rotor as the
point of departure, the energy of three different rotor
geometries was analyzed in order to draw conclusions
about the rotational energy of each. The goal targeted
for the overall system was 94,000 Nm. Only the variant Core 2 met this goal (see Table 1). Since the
Table 1. Energy analysis of a 6 500 ml CFRP rotor

compared with an aluminum rotor.
Rotational energy [Nm]
Type
Aluminum
Monolithic CFRP
Core 1-foam
Core 2-foam SLS
Rotor
154,246
158,913
135,208
86,426
Target
94,000
94,000
94,000
94,000
Difference
60,246
64,913
41,208
7,574
geometry had not been optimized, designing the rotor

geometry as a monolithic CFRP block even worsened
the initial values.
The simulation determined that CFRP rotors weigh
up to 44% less because they have a lower density
(approximately 1.5 g/cm3 at a content of approximately 60% fiber by volume) than aluminum (ca.
2.7 g/cm3 ) rotors.
764
Figure 3. Selective laser sintered master models displacers

as the basis for silicone molds.
Figure 4. Rotor variant Core 2 with 14 inserted SLS cores.
This weight advantage shortens acceleration and

deceleration times while retaining a centrifuges performance. Consequently, cycle times are shorter. In
addition, higher speeds can be run, thus increasing the
relative centrifugal acceleration. Moreover, this weight
reduction makes such rotors easier to handle. Their
resistance to corrosion and enhanced fatigue strength
are additional advantages.
4
FORMING TOOL DESIGN
Taking the calculation results as the starting point, socalled displacers (as inserts) for the rotor casting mold
were designed and laser sintered (Figure 3).
In addition to cutting weight, such laser sintered
inserts significantly reduced the rotational energy.
Arranged in position, the inserts are intended to assure
Core 2 has the rotational energy desired.
The prototyping technologies of selective laser sintering and vacuum casting were combined in order to
deliver the large quantities required. The overall design
was simultaneously optimized for fiber composites.
5
LABORATORY PROTOTYPE AND MOLD

MAKING
The results of these tests entered into the development

of the manufacturing technology to properly design
the future forming tool for casting.
Taking the theoretical calculations as the starting
point, initial rotor prototypes were subsequently produced to verify the variance analysis. This entailed
producing laminating molds to assure the reproducibility of the manufacturing in certain quantities. One of
the first rotor prototypes is pictured in Figure 4.
The complete lightweight rotor consists of nine different components, including the aerosol ring, hub,
rotor and six filling elements. The filling elements are
hidden in the rotor housing between the sample container holders and are made of a lightweight plastic,
thus reducing the lightweight rotors weight. The hub
Figure 5. Elements of the casting mold for the first rotor

prototypes black (1) and gray (2) were selective laser
sintered.
is force fit with the lightweight rotors while it is being

manufactured/layered. The use of two different hubs is
planned at present. The aerosol ring is separately made
of plastic and subsequently bonded to the rotor once it
has been manufactured.
The greatest challenge during development was reconciling the design of forming tools with the layering
technology to be developed. The forming tool must
have the rotors geometric complexity and its design
765
must facilitate the defined fiber layering, which is crucial to facilitating full impregnation and a uniformly
high content of fibers by volume.
These are essential for the manufacture of extremely
stable rotors. Therefore, the development partners
employed simulation methods to optimize the layering
technology and to design the forming tools.
Furthermore, the most complicated layerings were
field tested. Parts of the forming tool were in turn
retrofitted with laser sintered components.
The tests executed made it possible to implement design modifications in a matter of hours, thus
enabling the development partners to rapidly find and
test solutions to the most complicated layering steps.
The first proposed solution for the casting mold is
depicted in Figure 5.
6
CONCLUSION AND OUTLOOK
The extensive work to design and develop the layering technology delivered findings that enabled carbonic GmbH to completely engineer its manufacturing
processes without having to modify the molds design.
Rotors manufactured with the new methods combine the smooth surface of aluminum rotors with the
advantages of wound rotors, e.g. lower weight and better fracture characteristics. The new lightweight rotors
weigh up to 50% less than aluminum rotors and can
withstand up to 20% higher loads.
Compared with wound rotors, the new methods

can produce smaller quantities more cost effectively.
Further, rotors manufactured with the new methods
are more stable and have a smoother surface, which
enables rotors to operate smoothly.
ACKNOWLEDGEMENTS
Support for this project came from the Investitionsbank of Saxony-Anhalt, for which the authors express
their sincere thanks.
REFERENCES
Beckwith, S.W.; Hyland, C.R.: Resin Transfer Moulding: A
decade of technology advances. In: SAMPE Journal, Vol
34, No. 6, November 1998, p. 719.
VDI-Richtlinie 3404: Additive fabrication Rapid technologies (Rapid Prototyping). Berlin: Beuth Verlag GmbH
2009.
DIN EN 3783: Aerospace series; fibre composite materials;
normalization of fibre dominated mechanical properties
Berlin: Beuth Verlag GmbH 1992.
DIN 58970-1: Laboratory centrifuges Part 1: Definitions,
testing, marking. Berlin: Beuth Verlag GmbH 1996.
DIN 589702: Laboratory centrifuges; centrifuge tubes for
relative centrifugal acceleration up to 4000. Berlin: Beuth
Verlag GmbH.
766

Use of rapid prototype techniques for large prosthetic cranioplasty

G.H.L. Paschoal & J.V.L. Silva

Renato Archer Information Technology Center, Division of 3D Technology, Brazil
J.F.D. Zullo
State Hospital of Sumare, State University of Campinas (UNICAMP), Brazil
ABSTRACT: This work describes a set of procedures for the creation of a prosthetic 3D geometry based on
virtual modeling from medical images. This technique uses engineering CAD tools and rapid prototype systemsto
create the best shape structure for filling a bone gap cranioplasty in a patient involved in a car accident.This
application shows an engineering solution to create a prosthetic cranioplastic device according to the following
procedures: creation of a virtual model using CAD systems; creation of physical model for the surgical planning
using rapid prototype techniques; creation of moulds for the end model using rapid prototype techniques; and,
finally, the creation of a prosthetic device using the appropriate biomaterial.
1
INTRODUCTION
Computer-assisted design and modeling of cranioplastic materials has improved the cosmetic outcome as
well as minimized the procedure time needed for plate
insertion.
It is necessary when the patient presents a bone
defect that occurs from different kinds of traumas or
diseases.
Theses implants are hand crafted during surgery or
before it with the help of preoperative models. It would
be done creating patient-specific implants by using 3D
modeling and rapid manufacturing methods (Salmi,
2009).
Despite being in use for a long time, decompressivecraniectomy remains an established procedure to lower
intracranial pressure due to malignant brain swelling
(Jagannathan, 2007).
The surgical correction of skull defects has 2 main
purposes: protection of the brain and a satisfying
cosmetic result (Cabraja, 2009).
This article shows an engineering solution to create a prosthetic cranioplastic device according to the
following procedures:
1) creation of a virtual model using CAD systems;
2) creation of physical model for the surgical planning
using rapid prototype techniques;
3) creation of moulds for the end model using rapid
proto-type techniques;
4) and, finally, the creation of a prosthetic device using
the appropriate biomaterial.
2.1

Patient
Our 17-years-old female patient was injured in a

car accident. This is a case of a bifrontal decompressive craniectomy, because the patient had presented intracranial hypertension refractory to clinical
measures (mannitol, hi-hyperventilation, barbiturate
coma, hypothermia). ICP (intracranial pressure) may
not hold up to 20 mm Hg for a period high and when
this happens is indicated decompression. Currently,
the patient presents with a condition that is able to
understand commands, interact with other people, eats
well, but still has difficulty speaking.
The option for the use of three-dimensional technologies for the project and production was due to
complexity of the case and the lack of available solutions in the market capable to assist the patients
specific needs.
2.2
Creating a virtual model using CAD systems
The patient was submitted to exam of computerized

tomography multislice (Figure 1) in two different
moments, being the first accomplished previously to
the surgical procedure and second accomplished after
the retreat of the affected bone structure. For the
exams, an acquisition protocol was used using 1 mm
for increment between slices and 1 mm of thickness,
zero degrees of gantry inclination of the and axial orientation in equipment GE Medical System Hi-Speed.
767
Figure 4. Virtual model for the surgical planning.
Figure 5. Physical model for the surgical planning.
Figure 6. Boolean subtraction for prosthesis generation.
The STL file were edited using Magics 15.0 software (Materialise, Belgium), in order to minimize
surface imperfections. This manipulation enabled a
softening of the model in the upper skull region that
containing more widely spaced slices (Figure 3).
Figure 2. InVesalius for soft tissue segmentation.
Data obtained in the exams (DICOM format) were

converted in three-dimensional using InVesalius software (CTI-ProMED, Brazil). The software made possible to isolate woven of the bone structure through the
segmentation for threshold to export them in a STL file
as showed in Figure 2.
2.3 Creating a physical model for the surgical

planning using rapid prototype techniques
We created a 3D model from computed tomography
2D slice images that is shown in Figure 4.
A physical model was rapid manufactured using
rapid prototype that is shown in Figure 5. This model
768
Figure 7. Moulds generated from the prosthesis.
Figure 9. Mould made of polyamide and final model.
was used for material cure, where the mould was

immersed in water to 70 C for thirty minutes and later
high to 100 C for an hour and a half.
Figure 8. SLS Machine and polyamide prototype.
was created on Z Printer 510 (manufactured by Z Corporation). Materials used were ZP 130 (powder), ZB58
(binder) and Z Bond (resin).
This paper showed a manufacturing method for a large

prosthetic cranioplasty that enables exact fitting to surrounding tissues. This may result to better and more
natural recovery. Creating an implant before the surgical operation can reduce duration of surgery, improve
its accuracy, decrease patient morbidity and hence
improve quality of surgery.
It showed an engineering solution to create a prosthetic cranioplastic device according to the following
procedures: creation of a virtual model using CAD
systems; creation of physical model for the surgical
planning using rapid prototype techniques; creation of
moulds for the end model using rapid prototype techniques; and, finally, the creation of a prosthetic device
using the appropriate biomaterial.
2.4
Creating moulds for the end model using rapid

prototype techniques
For prosthesis generation, a subtraction Boolean was

accomplished among the two models generated starting from the exams (before the first surgery and after
that). This result bases prosthesis creation and fitting
borders to refine seeking as better anchorage between
prosthesis and bone that is shown in Figure 6.
The molds were developed using a specify module offered by Magics 15.0 (Figure 7), where retentive
areas were eliminated producing a split into two simple
structure. Reinforcement structures were inserted into
the model to increase its mechanical resistance. Four
locking points were inserted in the process. Creating
prosthetic device using the appropriate biomaterial.
Figure 8 shows SLS machine that created these
moulds: Sinterstation HiQ (manufactured by 3D Systems) and using DuraForm PA material. The whole
process took 48 hours.
2.5
Creating a prosthetic device using the

appropriate biomaterial
The material used for prosthesis (Figure 9) was

polyMethyl-Meta-Acrylate (PMMA), polymeric
broadly due to its biocompatibility and high mechanical strength. Moulds used for the resignation of the
material were covered with ceramic material to avoid
the adhesion of PMMA in the walls of the mold during pressing process. The cycle of slow polymerization
CONCLUSIONS
ACKNOWLEDGMENTS
Authors would like to acknowledge (INCTBIOFABRIS) and CNPq for financial support.
REFERENCES
Cabraja M., Klein M. & Lehmann T. N. 2009. Longterm
results following titanium cranioplasty of large skull
defects. Neurosurg Focus 26 (6):E10, 2009.
Jagannathan J., David O. Okonkwo D. O, Dumont A. S.,
Ahmed H., Bahari A., Prevedello D. M., Jane J. A. Sr.
& Jane J. A. Jr., 2007. Outcome following decompressive craniectomy in children with severe traumatic brain
injury: a 10-year single-center experience with long-term
follow up. J Neurosurg 106 (4 Suppl):268275, 2007.
Salmi, M. 2009. Masters Thesis: Design and Rapid Manufacturing of Patient-Specific Implants, Helsinki University of
Technology, Faculty of Engineering.
769

A-footprint: Ankle and foot orthotic personalization via rapid

manufacturing
J. Mungua & K.W. Dalgarno
School of Mechanical & Systems Engineering, Newcastle University, UK
J. Pallari
Materialise NV, Leuven, Belgium
S. Cook
Peacocks Medical Group, Newcastle upon Tyne, UK
ABSTRACT: Disabling ankle and foot pain affects a large proportion of the worlds population and has major
cost implications for healthcare systems. In Europe alone, the estimated annual cost of providing foot and
ankle orthotics used to treat many of these conditions reaches 312million. Personalized orthotics are usually
prescribed as indicated treatment, however production costs and delivery times are significantly high, and their
production remains craft-based with techniques such as impressions casts, templates and hand fabrication. The
A-FOOTPRINT project aims to develop novel orthotics for the foot and ankle by exploiting the latest scanning,
design and manufacturing technologies. This initiative is expected to act as a catalyst for the adoption of new
design and manufacturing technologies by companies across Europe and the UK. One of the main outcomes is
the set-up of a Rapid Manufacturing pilot facility run by Peacocks Medical Group, making it one of the first
orthotic manufacturers in the world to embrace digital fabrication. This technology is intended to reduce to
48 hours the complete process of customised orthotic supply.
1
INTRODUCTION
Disabling foot and ankle pain is a common condition

that it impacts negatively the overal health and quality
of life, with major cost implications on health systems
across Europe. The estimated prevalence in Europe is
approximately 200 million citizens and this is set to
rise in an ageing society with increasing chronic long
term conditions (Garrow et al. 2004). Ankle Foot and
Foot orthoses are commonly prescribed to alleviate
pain related symptoms and to provide enough stability
for the improvement of gait patterns affected by different conditions (Lehmann 1979). The cost of providing
this number of users with the appropriate orthoses
is currently estimated at 312 million per annum
across European health services which represents a
challenge specially when customized orthotics which
are tailored to individual users are to be prescribed
(A-FOOTPRINT 2009).
Nearly 70% of Ankle Foot Orthotics (AFO) and
Functional Foot Orthotics (FFO) are made using traditional techniques which rely on impressions casts, templates and hand fabrication (A-FOOTPRINT 2009).
Personalised devices tend to provide more effective
outcome in terms of symptom reduction, fit, comfort
and aesthetics, but are more costly and time consuming to manufacture. A minimum product lead time
of 10 days has been observed for the production of
these devices, from the initial patient assessment to

final delivery, however total time can extend up to
28 days when centralized manufacturing is used and
major modifications must be made on the final orthotic
(Munguia & Dalgarno 2010).
As a result of being highly dependent on manual work, the functional form of orthoses is difficult
to verify and chances of exact repeatability are low,
thus leading to prescriptions that can be inaccurate.
Additionally, hand manufacturing using thermoplastics limit design choice and limits personalised function to simple parameters such as cushioning and range
of motion.
1.1 About the A-FOOTPRINT project

The recently EU funded project A-FOOTPRINT
(Ankle and Foot Orthotic Personalisation via Rapid
Manufacturing) coordinates research, analysis and
pilot testing in 5 universities, 2 research organisations and 6 SMEs across Europe. The objectives of
the project are to explore the use of additive manufacturing and integrate it into the orthotics supply
chain. It also seeks to improve the accuracy of clinical
prescriptions for customised foot and ankle orthoses;
to improve the fit and functionality; to significantly
decrease manufacture time to 48 hours and finally
771
to develop a cost-effective, fully integrated orthotic

manufacturing solution.
2.2 Functional Foot Orthoses
STATE OF THE ART OF ANKLE AND FOOT

ORTHOTICS MANUFACTURING
2.1 Ankle Foot Orthoses

Ankle Foot Orthoses (AFO) are assistive devices that
are used to maintain ankle joint stability in the anteriorposterior and medial-lateral directions and also to
permit and stabilize motion (Lehman 1979). In clinical practice,AFOs are frequently prescribed to treat
gait related problems. These problems include limited
foot clearance in the swing phase, poor foot placement at initial contact, and a reduced stability in the
stance phase (Ridgewell et al. 2010). In deed the
study and development of AFOs is a wide subject at
the crossroads of several disciplines which go from
clinical assessment. (gait lab analysis, inverse dynamics, etc.) and mechanical studies (design, materials
selection, manufacturing, biomechanics analysis) to
performance and quality assesment (Chu 2001).
Custom made Ankle-Foot orthoses are normally
made from a positive model of the patient leg, obtained
by means of a negative cast, directly applied to the
area of interest (ankle-foot). The overall process for the
manufacturing of AFOs (figure 1) is highly standardized. It starts with the appropriate patient assessment,
casting and rectification of the positive cast impression. Once the casting is rectified the manufacturing
steps will vary according to the material to be used. In
the case of plastic AFOs the adopted process is vacuum
forming, being well documented by standards (BS
EN ISO 22675:2006 ) and official guidelines (ICRC
2006). These suggest a definition of appropriate thicknesses, type of material to be vacuum formed, as well
as the optimum material sheet size according to the
AFO type; rigidity needed and patient weight.
Functional Foot Orthoses (FFO) are used for a wide

range of applications; they can be prescribed to treat
chronic long term medical conditions such as rheumatoid arthritis and to maintain the health of the feet of
diabetic patients (Crabtree 2008). Other activities that
involve the prescription of orthotic insoles are sports,
for function improvement hence enhancing the biomechanical performance of the user and the overall gait
related patterns (Crabtree 2008).
Although recently CAD/CAM systems and a range
of scanning technologies have enabled the automation of insole fabrication, especially by means of CNC
milling equipment, some traditional methods based
around plaster cast mould manufacturing are still being
applied by podiatrists, thus the main fabrication methods for customized FFOs are: thermoforming, CNC
milling, and hand-layup.
2.2.1 Thermoforming
Thermoforming uses a plaster model or a pre machined
master pattern as mould, usually made of wood,
polypropylene or machinable plaster (figure 2). Thin
layers of thermoplastic or composite material sheets
are conformed around the predefined shape. Common
materials include ABS, PP, HDPE, HIPS and PC.
2.2.2 CNC milling
In milling/machining, material is removed from the
original block across 3 axes until the desired shape
is achieved. Currently available systems mill insoles
from a range of EVA blocks with variable hardness,
and also are capable of milling increased-hardness
EVA to act as mould for vacuum formed orthoses.
Although maximum CNC capacity is given by the
machine size and speed, production rates can typically
reach an approximate of 7 pairs of insoles per hour,
with an average of 815 min to mill a pair, getting
5060 pairs per working day, as operation must be
partially attended(Munguia & Dalgarno 2010).
2.2.3 Hand lay-up
Fabric/foam layup consists in the layering and bonding
of additional cushioning layers to a substrate of hardened thin material, such as woven polyester reinforced
resin (figure 3). This process is based in standardized
sizes of insole base, and standardized thicknesses and
hardness for additional cushioning layers and add-ons.
Figure 1. Typical stages for PP AFO manufacturing. a)

Cast rectification b) PP sheet placement c) Edge cutting d)
Add-ons.
Figure 2. Thermoforming stages. a) Plastic layer pressing.

b) Bonding of add-ons.
772
Figure 3. Layer deposition method. a) bonding of additional

foam layers. b) polishing final edges images: Junquera y Diz
product catalogue.
All of the previous methods for AFO and FFO manufacturing share the component of being a highly manual tasks whose overall quality depend on individual
skills of the technitians and the precision of craftman operations through the different stages of product
development. As these processes dictate the quality,
productivity and possible design alternatives iInnovation in the sector is limited and market competitiveness
is considered relatively weak.
What is clear from SME orthotic companies
is that hand-crafted devices can no longer remain
state-of-the-art. Indeed across the European health
care systems customised orthoses are handcrafted in
SME factories for approximately 75% of patients
(A-footprint 2009).
Current trends show an increasing adoption of
CAD-CAM systems with various centralised and
distributed models ranging from complete officebased solutions to SME factory-based manufacturing.
Herein lies a market opportunity to asses the feasibility
of integrating additive technologies which the orthotic
SME partners for A-FOOTPRINT project are aiming
to explore.
3 THE A-FOOTPRINT APPROACH FOR
ORTHOTIC DEVELOPMENT
3.1
General aim
The A-FOOTPRINT project aims at an integrated

approach for the development of next generation of
orthotic products. The new process will take advantage of existing 3D scanning capabilities for patients
geometry adquisition, and will integrate those with relevant personalized gait parameters based on plantar
pressure, EMG activity and gait pattern identification. Along with custom built CAD capabilities this
information will be processed to generate a highly customized design which not only improves shape, but
also function based on relevant biomechanical data.
Static and dynamic analysis will be then performed
on real time on the resulting design so that the optimal material and processing parameters are defined
prior to manufacturing. From this stage, rapid manufacturing methods will be applied which comply with
the overall product requirements. One of the main
research interest is to match an optimal orthotic design
with the capabilities of well stablished systems such

as Selective Laser Sintering (SLS).
Additionally rapid manufacturing techniques will
be used to develop novel customised orthotic components such as living hinges, variable stiffness and fine
resolution cushioning to enable better personalised
function. Setting new industry standards, prototype
devices will be evaluated by medical industry level
controlled trials to further improve product knowledge. This highly integrated, multidisciplinary project
aims at making a significant impact on the healthrelated quality of life and well-being of EU citizens.
3.2
Description of specific research
Given the current orthotics manufacturing scenario,

the A-FOOTPRINT strategy seeks to develop a comprehensive automated system for the manufacture of
highly customized ankle and foot orthoses. The main
elements of the system and their interaction are briefly
described.
3.2.1 Patient information system
The design process involved SME partners and other
stakeholders to capture relevant input/output data, user
interface features, access and storage. Alpha version
functionality includes input data relating to the patient
and its dossiers, containing symptoms, questionnaire,
measurements, diagnosis, treatments, orthotic design
and evaluation. This data is being compiled in an
application built within an Oracle database platform.
3.2.2 Development of scalable kinematic and
kinetic model of the foot
This part of the project deals with the definition of the
kinematic links between the bones including the centre
of rotation and orientation axes which were optimised
using the subject-specific motion capture kinematic
data. Data collection permits simultaneous measurement of the individual bone and joint kinematics as
well as ground reaction forces and plantar pressure
and EMG muscle activity for major muscles of the
lower limb and foot.
3.2.3 Computer Aided Design (CAD) software for
personalised ankle and foot Orthotics
The A-FOOTPRINT platform will rely on 3D CAD
software with the ability to interpret point cloud data,
translate it into various file formats (open system)
and manipulate the object with the aim of replacing
the manual-lengthy labour of cast rectification and
moulding with automated solutions. It is foreseen that
this practice will improve the overall fit tolerance of
personalised ankle and foot orthoses.
3.2.4 3D scanning technology
The acquisition of the patients 3D scanned data will be
crucial to provide digital models of surface anatomy
with enough accuracy. A digital file format will enable
orthothist migrate all orthotic designs from plaster
casts, templates and blueprints to digital information.
773
Table 1.
Summary of requirements for new RM facility.
3D product data types

different digital data formats as input information
open system accepting data from various 3D capture
devices/CAD modelling and file conversion systems.
Geometry and product size
Figure 4. The A-FOOTPRINT approach for orthoses development (Pallari et al. 2011).
Although a number of off-the shelf scanning solutions

exist, localized scanning for ankle and foot plantar
surface will be implemented based on proprietary
technology from the consortium members.
freeform surfaces with variable geometries as defined by

digital information
modelling accuracy must equal or surpass that of manual
lay-up methods and CNC milling
thin section thicknesses (24 mm range) and variable
thickness
it must be able to allocate one or multiple AFOs of the
maximum size: 300 100 600
it must be able to allocate one or multiple FFOs of the
maximum size: 300 100 50
capability to provide variations in designs within the same
build
personalized add-ons and additional comfort features must
be allowed for each individual part
Materials and properties
3.2.5 Co-creation
Orthoses have been traditionally manufactured in
absence of the patient, thus making it difficult to
receive feedback in the moment of creation. Integrating co-created digital design enables the practitioner to
obtain direct feedback from the user during the design
of the orthoses, in terms of shape, comfort, aesthetics
and functionality.
localized/variable stiffness possible

cushioning elements within the same part with the same
process
material/structure properties should range from hard (carbon fibre) to soft (EVA foam)
ability to tailor different properties within multiple parts
of the same build-vat
Productivity
3.2.6 Embedded sensing

Although a wide spectrum of sensing elements are
used in laboratory and research environments, the
capacity to collect and transmit real time information on the patient for parameters such as temperature,
pressure, contact pattern as well as indicator of daily
use (compliance) can bring unprecedented benefits.
A number of sensing technologies are being explored
in order to achieve reliable-cost effective embedded
sensing that provides important parameters for the
early detection of symptoms such as those exhibited
by diabetic patients.
3.2.7 Rapid manufacturing
The paradigm of RM is likely to change the current
practice of labour intensive orthotics craft-like manufacture. This will bring along the necessity of a
new production planning based on a dedicated production facility for high volume manufacturing, coupled
with localized desktop manufacturing solutions.The
first case corresponds to high capacity RM equipment
such as Selective Laser Sintering which is capable
of producing batches of several hundreds of units on
a weekly basis, but that must be run in a dedicated
production floor space. On the other hand, desktop
manufacturing refers to a different segment of rapid
manufacturing equipment which is low profile, i.e.
does not require a dedicated production floor and
the running cost, including the cost of ownership are
48 hour lead time instead of 1014 days

for in clinic manufacture, the lead time needs to be an hour
or less if a return visit is to be avoided
Productivity benchmarks: PP AFO, 34 per person per
day/FFO, 3060 per person per day
Costing
manufacturing cost should enable selling price of 200
for AFO and 70 for insole
enhanced performance parts and extra features at no extra
cost for manufacturer
system must allow room for cost reduction by optimizing
product design and production planning
added automation to compensate labour hours
the system must provide an added advantage regarding
normal plant operation:
savings in cast rectification
digital part storage vs. warehouse space
significantly lower. This will bring along two complementary business models. In the first case (localized
production) information is sent to a dedicated facility
and sent back to the practitioner, while in the second case (desktop manufacturing) the orthothist will
have the option to have the orthoses designed and
manufactured in-house.
Figure 4 illustrates the integrated approach for
the different elements of the A-FOOTPRINT project
774
Figure 6. Mechanical testing at UNEW: a) four point bending b) 3-point bending of PA12 samples c) Rotating fatigue
rig schematic and d) beta-rig.
Figure 5. From left to right: The SPro SLS equipment,

breakout station room, and new CAD office created as result
of the adoption of digital fabrication.
4.3 Set-up of pilot demonstration facility at

Peacocks Medical Group (Newcastle,UK)
research, aiming at a robust model for the development

of advanced orthotic products.
4
4.1
ONGOING WORK AND PRELIMINARY

RESULTS
Establishment of manufacturing facility
requirements
Work at Newcastle University is focused on analysing

the technical and economical feasibility of using rapid
manufacturing systems as an option to be integrated in
the orthotic development supply chain for customized
ankle foot and functional foot orthoses.
Early work focused on establishing the basic performance requirements for the manufacturing facility
in terms of required product dimensions, mechanical
properties, accuracy, surface finish and productivity.
These specifications were given in the form of a product design specifications. Table 1 is a summary of
relevant requirements.
4.2
Characterization of new and existing RM

technologies
A benchmark between high-end RM equipment (centralized facility) and lower-end alternatives (office
based) was assessed so that SME partners could select
the most convenient technologies for the trial manufacture of orthotics. Table 2 illustrates a benchmark
between basic product specifications and the general
capabilities of rapid manufacturing methods be it dedicated or desktop based. Although a virtual winner is
selective laser sintering there are additional factors to
be considered before selecting an alternative such as
range of materials and availability, equipment cost,
economic performance (economic batch, productivity,
etc).
As a result of equipment benchmark and talks to

manufacturers the sPro-SD SLS Center, by 3D systems proved to be within budgetary expectations and
offered the flexibility to produce both FFOs and AFOs
which was the main interest of the SME partner.
The newly acquired SLS equipment will comprise the
pilot demonstration facility (figure 5) which will be
used by the consortium during the project length for
the design, manufacture and testing of new orthotic
product concepts.
4.4 Long term material testing (Newcastle

University school of mechanical engineering)
Laboratory testing performed by UNEW includes tensile, flexural and fatigue test of laser sintered samples
and other RM processes such as FDM and other low
cost 3D printing methods (Figure 6).
In the particular case of fatigue, this is a long term
dynamic testing where the polymer material behaviour
is simulated on a high speed rotating machine in
order to calculate a reliable estimation of the service
life. Currently UNEW is adapting an existing rotating fatigue bench to be used as a dedicated machine
for testing RM specimens along the project. An additional fatigue test for polymer specimens is based on a
flexural bending machine, which will help assess the
repeated bending capacity of different formulation of
materials.
4.5 Design optimization for laser sintered orthotics

The aim of mechanical testing is to fully characterize
the behaviour of SLS snd other RM materials when
used for lower limb orthotics. For this reason optimal CAD design and simulation must be performed
before comiting to the production of end-use orthotics.
Analysis also include the economics of AFO and
775
Summary benchmark table for RM technologies.
Max AFO size

300 150 450
(mm)
Max FFO size
300 150 30
(mm)
Variable
thin
section(wall
thickness)
Productivity (34
AFO/day)
Accuracy
Surface finish
Multi parts (AFOs
& FFOs) in one go
Overall material
properties
Sum Y
Sum N
Objet
Vflas
Low capacity
3DP
SLS
FDM
Abbreviation
SLA
High capacity
uPrint
Table 2.
Y Y Y
N N N N
Y Y Y
Y Y Y Y
Y Y Y
Y Y Y Y
Y Y Y
N N N N
Y Y Y
Y N Y
Y Y Y
Y Y Y Y
N N Y Y
Y Y Y Y
N Y Y
N Y N Y
7
1
4
4
7
1
8
0
5
3
5
3
Figure 8. Concept AFOs fabricated by SLS.
6
2
Machine
Material
No. of
parts
Powder Total
base
build time
sPRO
DuraForm
Polyamide
Powder V3.45
DuraForm
Polyamide
Powder V3.45
10.69
18:18:07
10.69
18:45:07
SLS HiQ
Figure 7. Concept AFOs incorporating free geometry features.
FFO manufacturing according to the production levels afforded by the selected RM alternatives (figures
79).
Figure 9. Productivity estimation for variable series of

AFOs for SLS.
CONCLUSIONS
The A-FOOTPRINT project aims at enabling SMEs to

become international leaders with strong competitive
advantages as a result of the adoption of state of the
art digital technologies.. A-FOOTPRINT will directly
benefit SME partners in the following ways
Result in the development of cost-effective personalised ankle and foot orthoses which will have a
significant impact on health related quality of life
and wellbeing.
Produce through advanced technology and new
knowledge creation orthotic devices.
Impact the high value added global market for personalised ankle and foot orthoses; in a market with
high growth potential due to aging populations,
increased incidence of foot and ankle problems, and
strong consumer focus on personalised foot care
products.
The present work has also introduced the concept of personalized orthotic fabrication by means of
RM. This set of technologies does not only include
a migration from manual-craft to digital manufacture
but also involves the implementation of a number of
key technologies that must be harmonized in order to
accomplish this paradigm change.
Several barriers have been identified during the earlier stages of the project. One of the main obstacles
776
is the limited use of digital information by podiatrist

and orthotics alike. Since the exploration, diagnose
and acquisition of patient information will be the first
step in the newly proposed digital system, it will be
important to firstly introduce, the whole range of 3D
scanning, file conversion and digital information management into the practitioners office. Only then, will it
be significant to continue with the second link which is
the implementation of digital fabrication technologies
in the production floor.
Medium to short term objectives for the A-Footprint
project include: the analysis and comparison of rapid
manufacturing technologies to be used as dedicated
and desktop fabrication techniques, complete the relevant mechanical testing to ensure the material and
technologies fulfil the orthothist requirements, as well
as present the first prototypes for patient evaluation
in terms of comfort and effectiveness to treat the
identified symptoms.
ACKNOWLEDGEMENTS
Crabtree, P., V. Dhokia, et al. 2008. Design and manufacture of customised orthotics for sporting applications.
Engineering of Sport 7, Vol 1: 309317.
Garrow AP, Silman AJ, Macfarlane GJ 2004. Pain 2004:110:
37884.
ICRC 2006 Ankle-Foot Orthosis. Manufacturing Guidelines.
International Committee of the Red Cross. www.icrc.org.
September 2006.
Lehmann JF 1979. Biomechanics of ankle-foot orthoses:
Prescription and design. Arch Phys Med Rehabil.
1979;60:200207.
J. Munguia, KW Dalgarno 2010 D6.4 Delivery of intermediate report describing evaluation of existing and
new manufacturing processes for orthosis manufacture.
A-FOOTPRINT Grant agreement number: NMP2-SE2009228893.
JHP Pallari1, J Munguia2, J Oosterkamp3, T Putseys1, F
Holtkamp3, KW Dalgarno2 2011. Design systems for
additive fabrication of foot and ankle-foot orthoses. SCIENTIFIC TESTING OF ORTHOTIC DEVICES, Aix les
Bains, FranceMarch 2011, 23rd26th 2011.
Ridgewell, Dobson, Bach & Baker 2010. A systematic review
to determine best practice reporting guidelines for AFO
interventions in studies involving children with cerebral palsy. Prosthetics and Orthotics International. 34(2):
129145.
This work was funded through the European Commission Framework Seven Program (grant number
NMP2-SE-2009228893) as part of the A-Footprint
project (www.afootprint.eu).
REFERENCES
A-FOOTPRINT-Annex IDescription of Work 2009 Ankle
and Foot Orthotic Personalisation via Rapid Manufacturing Grant agreement no.: CP-TP 2288932.
BS EN ISO 22675:2006 Prosthetics-Testing of ankle-foot
devices and foot units-Requirements and test methods.
Chu TT 2001. Biomechanics of ankle-foot orthoses: past,
present, and future. Top Stroke Rehabil. 2001;7(4): 1928.
777

Physical prototypes in cross-functional team collaboration:

A study of the Model-T2 concept car project
S. Datta
School of Built Environment, Curtin University, Perth, Australia
S. Hanafin
School of Architecture and Building, Deakin University, Geelong, Australia
B. Rolfe & T. de Souza

ITRI, Deakin University, Geelong, Australia
ABSTRACT: Virtual and physical prototypes play a significant role in the design of artifacts.This is particularly
true in cross-functional team collaboration, where diverse disciplines come together to contribute to the concept,
design and development of an artifact. While the role of prototyping in design and cross-functional collaboration
has been widely studied independently, little is known of the role that physical prototypes play during crossfunctional team collaboration. This paper reports on a study of virtual and physical prototyping strategies used
by a cross-functional team comprising engineers and designers working on a concept car project. The paper
provides a background of the collaborative design process and details the role of physical prototyping methods
during the design of the concept car.
1
INTRODUCTION
that emerges is one of clear functional separation,

supplemented by osmosis and mediation
Teamwork is of considerable importance in normal

professional design activity, and is becoming of even
greater importance given the increasing complexity of
design problems and specialized nature of functional
and domain expertise. This is nowhere more evident
than in the design disciplines, whose practitioners
spend substantial portions of their professional lives
designing things in formal and informal collaboration.
These include engineers (mechanical, electrical, civil,
software) and architects.
In the use of digital tools and processes, cognitive behaviour and social interaction, we identify
three common principles underlying virtual and physical prototyping in design collaboration: authorship,
islands of automation and the management of complexity. In the Model T2 project, we observed that
design authorship is loose-coupled, with each functional team making strong contributions in their
respective domains of specialist expertise. As the
participants work together towards a solution of a
shared problem, attribution of authorship to singular
individuals became blurred.
Second, the role of knowledge sharing and interoperability of the cross-functional teams did not rest
on a predefined ontology, a shared digital model or
even a common platform. Instead, the teams worked in
islands of automation, only sharing and interoperating
on a need-to basis. The picture of the knowledge and
information sharing across the design supply chain
BACKGROUND
2.1 Model-T2 concept car project

The project to design a concept vehicle was motivated
by the Model for the 21st CenturyUniversity Challenge. Deakin Universitys submission, Model T2 ,
responded to the challenge by creating a zero emission concept car. A university-wide design research
team comprising staff, researchers and students drawn
from multiple disciplines from across the university (Engineering, Architecture, Intelligent systems,
Material and Fibre Innovation) with support from
Ford Australia was formed. The design research
team comprised a project steering committee and
a number of cross-functional design teams working over a period of 16 weeks. The team, inspired
by Henry Fords original vision, embraced simplicity in the manufacturing of components and
assemblies, combine component functionality into
single systems and introduced advanced materials
into the design. The result of the collaboration was
a prototype tri-wheel vehicle, driven by compressed
air high torque hub mounted motors in the two
front wheels complemented by natural gas, steering
by differential wheel speeds and use of advanced
materials.
779
2.2 Teamwork in design collaboration

Studies of interdisciplinary collaboration in design
model teamwork as a technical, cognitive and social
process. The use of a shared representation of knowledge (language and vocabulary) to facilitate communication between specialists is another strand of
teamwork research. Researchers argue that collaborators need formally defined protocols and agreements
that are useable both by people and their machines.
The information modeling view (Olsen et al, 1995)
of teams offer technics, models and methods such as
interoperable standards, formal ontologies and virtual
design studios (Wojtowicz, 1995) to facilitate knowledge interchange. In studies of teamwork, particularly computer-supported cooperative work (cscw),
researchers have shown that social and cognitive
aspects of team dynamics play a significant role in
facilitating technical collaboration. For example, studies of cognitive behaviour during collaboration [Kvan,
2000] show that design teams may have loose or tight
coupling, depending on the design task. The view of
designing together as a social process is explored in
protocol studies of team-based design [Cross et al,
1995]. The conjoint role of digital convergence
between research, education and practice is addressed
by Burry, who traces how disciplinary boundaries and
roles fluctuate in large teams working on complex
projects [Burry, 2005]. Burry shows how, in projects
such as the Sagrada Familia, trans-disciplinary design
collaboration, is the norm rather than the exception.
Disciplinary roles and responsibilities shift and evolve
and the traditional boundaries between disciplines and
individuals begin to blur.
3 TEAM COLLABORATION IN MODEL-T2
3.1 Team formation
To meet the challenge two teams were established,
a Project Steering Committee and a Design Team
with technical stream leaders. In addition, crossdisciplinary teams were developed by inviting participation from the many Faculties across Deakin
University. This included: Architecture and Building,
Business and Law, Science and Technology and the
Centre for Material and Fibre Innovation. The project
also received external support, advice and feedback
from Ford Australia.
The Project Steering Committee was responsible
for the overall coordination of the project including
the budget marketing and promotion, provision of
resources and for the successful and timely completion of the project. The Design Committee was the
coordinating group for all the key design tasks from
styling, structures, power train, interior, electrics, and
the human-machine interface.
3.2
Planning and information gathering
Design targets were clarified through a brainstorming

session with participants from all over the Asia/Pacific
region and highlighted some important key features for

inclusion in the design, these included: city driving,
flexibility in use, agility, space saving, need for low
emissions, and easy to part.
The target market for the Model T2 concept car is
the growing cities of China and India. Market research
was undertaken with city drivers in India through the
use of surveys and interviews and provided anecdotal
advice. The persons interview ranged from university
students and young professionals to family bread winners and home makers. Similar advice was received
from colleagues and associates who had grown up in
China. Utilising the survey data and the teams intuition about what a non-traditional consumer may want
a set of consumer needs were established.
A brainstorming session was also undertaken within
the University and included participants from different Faculties. This brainstorming session focused on
two specific areas; Overall Project Concepts and Subsystem Breakdown. Within these two areas different
topics were consider. In Overall Project Concepts: consumer needs, revolutionising the automotive industry,
vehicle concepts / platforms / coolness. In Sub-system
Breakdown: body / structure, power train, vehicle
dynamics, interior / human interface.
3.3 Requirements analysis

Once the consumer needs were established, measures
or metrics were determined so that the developed concepts could be assessed or ranked against each other
and against the competitors. A benchmarking exercise
was undertaken to determine how the competitors cars
compared against the metrics. This included; Smart
Car2008,TatNano,Hyundaii10 2008, ToyotaYaris
2008, Toyota IQ 2009, Mitsubishi i 2005, Chevrolet Spark, Chery QQ3, Maruti 800 Duo, Daihatsu
Trevis.
Given the needs for a small, nimble and agile vehicle, well suited to cities, this prompted the decision for
a three-wheel vehicle platform. Three wheel vehicles
are currently becoming more popular throughout the
world and are quite prominent in many Asian based
countries. Often termed a tri-car or cycle-car due to
its commonalities between a car and a motorcycle,
have the scope to be extremely light weight, fuel efficient and yet very versatile for many urban or city
residents. This platform lends itself to the Model T2
design philosophy of simplicity and total integration,
whilst allow for excellent manoeuvrability, through it
steering system, which utilises two hub based motors
as its steering device.
PROTOTYPING IN TEAM COLLABORATION
This section will discuss the various types of virtual

and physical prototyping methods undertaken during
the design of the vehicle and the role of prototypes in
cross-functional design team collaboration.
780
Figure 2. Plasticine model.
Figure 1. Model T2 vehicle dimensions.
4.1
Concept generation
With the initial framework of the concept established

two concurrent design lines were then explored. The
first was that of the engineering aspects of the design,
and this was the primary focus of the project and considered four key areas: powertrain, vehicle dynamics,
body and structure and control systems.
While the engineering systems were being developed, the styling of the vehicle was also being undertaken in parallel. The styling at this point was restricted
by only three limits that of the 3 wheels, two doors, and
two occupants.Various different options were explored
within the styling team, using both hand drawing and
clay modelling. During this process the exact package geometry of the vehicle was still being developed
by the engineering teams, therefore these explorations
were more about a design style and the overall look
and feel of the car rather than a final form.
During this process both the steering committee
and the design team met weekly. Two members of
the steering committee were also key members of the
design team and allowed for the concept to be discussed and feedback given about the current direction
of the concept. The design team meetings allowed for
the various subgroups to present current developments
and problems. These weekly meetings were critical in
setting goals for the next week and advancing the initial
concept of hurdles.
The initial concept was then present in a design
review. This review consisted of people not directly
involved in the project. These included lectures from
the School of Engineering and the School of Architecture and Building, researchers within the Centre for
Material and Fibre Innovation. Various external guests
were invited and included people from Ford Australia
and Defence Science and Technology Organisation
(part of the Australias Department of Defence).
Figure 3. Resulting laser scan of the clay model.
4.2 Concept prototypes

The first concept prototypes were form models made
of plasticine (figure 2). These were manually built
by the design and styling team to explore the formal
options for the concept car. The vehicle package geometry was laser cut out at 1/10th scale in order to allow
for a clay model to be produced around it to develop
the final shape of the vehicle.
The clay model was then laser scanned (figure 3)
and used as base geometry to develop the CAD model
of the exterior form of the car in Microstation. This
exterior geometry was combined into the engineers
master model, developed within Solidworks, which
contained all of the engineering details of the vehicle. The Solidworks model was used to undertake
engineering analysis and testing of the concept.
DESIGN DEVELOPMENT
The design of the Model T 2 vehicle body and structure

concept is focused around distinct upper and lower
vehicle assemblies. The lower assembly (figure 4) can
be considered essentially as a rolling chassis, providing
a large portion of the vehicles strength, stiffness and
energy absorption capabilities, whilst the light-weight
upper assembly (figure 5) forms most of the exterior
components of the vehicle.
781
There are significant advantages associated with

this two-assembly concept, as compared to the current monocoque approach. Firstly, the fact that the
structural members are independent of the stylized
exterior components allows the use of much simpler
and more structurally efficient components, thus providing large weight savings. The simpler shapes also
allow the employment of lower cost manufacturing
processes and higher strength materials. These combining factors result in a significant reduction in the
overall cost and weight of the vehicle, whilst still
maintaining excellent structural performance. Another
advantage of the separate rolling chassis is the degree
of modularity and flexibility of exterior shape that it
provides.
5.1
ties in with the Model T2 design philosophy of total

integration. Using embedded low cost devices, a modular system has been developed which allows for a
plug-and-play style system of independent interconnected networked elements, which control and monitor
dedicated sub-systems under the control of a master
processor, achieving a level of total integration. Like
the original Model T which had a production life span
of twenty years, the Model T2, through the central
computer system, allows the ability to plug-and-play
hardware and update of control software with ease,
turning the vehicle into a flexible and customisable
appliance.
5.2 Detailed design prototypes
Control systems
This Model T2 uses an electronic drive-by-wire system that will be used to control the various systems
throughout the vehicle. Using a complete drive-bywire system, allows precise control over all systems,
including the powertrain, braking, steering system
and other Human Machine Interfaces (HMI). This
Two different models were initially produced for the

submission to Ford Global Technologies in Detroit
USA. The first of these was a remote controlled driveable 1/10th scale model. This model had to parts to
it, the first a basic chassis which contained all the
remote control equipment and drive train. This was
constructed manually using the key dimensions from
the final CAD model. In order to provide an accurate representation of the exterior model a 3D print
(figure 6) was produced of the exterior geometry of
the vehicle. This model was then used to vacuum form
a plastic shell for the chassis.
Figure 4. Lower assembly.
Figure 6. 3D Print of the exterior geometry.
Figure 5. Light-weight upper assembly.
Figure 7. 1/8th scale structural model.
782
The second model developed was a 1/8th scale

structural model (figure 7), and was used to show the
engineering details of the vehicle and the layout of the
interior space. The model was broken down sections to
allow for it to fit within the print camber and provide
simply assembly once printed. These different sections
were exported separately from Solidworks to ensure
that the connections between sections were accurate,
with the print software only being used to apply colour.
FULL SCALE PROTOTYPE
After the completion of the competition a full scale

display model was desired. While the exterior geometry had been previously 3D printed, in order to develop
it at full scale a CNC foam router was used. The model
was broken down into sections that would accommodate the scale of the router, and was then glued together
before being finished (figure 8). In order to give the
model a realistic finish the foam form was fibre glasses
and then painted. The finished form was then fitted
with vehicle tyres and mirrors to complete it (figure 9).
6.1 Conclusions
A successful multi-disciplinary design research
project, such as the Model-T2 concept car outlined
in this paper, provides many lessons. In the use of
digital tools and processes, cognitive behaviour and
social interaction, we identify three common principles underlying design collaboration: authorship,
islands of automation and the management of complexity. In the model T2 project, we observed that
design authorship is loose-coupled (Kvan, 2000),
with each functional team making strong contributions
in their respective domains of specialist expertise.
The role of knowledge sharing and interoperability of the cross functional teams did not rest on a
predefined ontology, a shared digital model or even
a common platform. Instead, the teams worked in
islands of automation, only sharing and interoperating
on a need-to basis. The picture of the knowledge and
information sharing across the design supply chain
that emerges is one of clear functional separation,
supplemented by osmosis and mediation.
As the participants work together towards a solution of a shared problem, attribution of authorship to
singular individuals became blurred. Therefore, to initiate deep collaboration, the idea of design authorship
in architecture remains problematic. The projection
of individual expertise runs counter to the lessons of
cross-functional collaboration, particularly in technical disciplines and design research. Authorship is an
impediment rather than an instrument of change to
address the complexity of the problems of faced in
collaboration. The management of complexity within
the project shows the use of near decomposability
(Simon, 2005) Using modularity, the cross functional
teams moved seamlessly between conventional design
practice and innovative break throughs, relying on convention where necessary and innovating or working
around complexity as needed.
ACKNOWLEDGEMENTS
Figure 8. CNC foam model.
We gratefully acknowledge the support of the following people who contributed to the Model-T2 project.
The Steering Committee of Bernard Rolfe, Noel
Miller, Chee Jin Yap, Tim de Souza, Stuart Hanafin,
Sophie Torrens, Tim Hilditch, Ian Burston, and Emeritus Prof. Ian Duncan. The design team members, Tim
de Souza, Greg Pitts, Stuart Hanafin, Steve Agius, Ben
Adamson, Michael Pereira, Tim Black.
REFERENCES
Figure 9. Completed full scale prototype.
Burry, M. (2005). Digitally Sponsored Convergence of

Design Education, Research and Practice. Computer
Aided Architectural Design Futures 2005, B. Martens and
A. Brown, eds., Springer Netherlands, 322.
Cross, N., and Clayburn Cross, A. (1995). Observations of
teamwork and social processes in design. Design Studies,
16(2), 143170.
783
Kvan, Thomas (2000), Collaborative design: what is it?,

Automation in Construction, 9: 4, pp. 409415.
Olsen, Gregory R., Mark Cutkosky, Jay M. Tenenbaum,
and Thomas R. Gruber. (1995) Collaborative Engineering Based on Knowledge Sharing Agreements Concurrent
Engineering June 1995 3: 145159.
Simon, H.A. (2005) The Structure of Complexity in an Evolving World: The Role of near decomposability. Foreword
in Werner Callebaut, Diego Rasskin-Gutman (eds). Modularity: Understanding the development and evolution of
natural complex systems. The Vienna Series in Theoretical
Biology, MIT Press.
Wojtowicz, J.(ed), 1995, Virtual design studio, Hong Kong,
Hong Kong University Press.
784

Production of functional parts using SLM Opportunities and limitations

A.B. Spierings & G. Levy
Inspire AG institute for rapid product development irpd, Switzerland
L. Labhart
AMZ Formula Student racing team, ETH Zurich, Switzerland
K. Wegener
Inspire AG fr mechatronische Produktionssysteme, Zurich, Switzerland
ABSTRACT: Different Additive Manufacturing Technologies have already a certain stage of maturity, allowing
a wide range of applications. In the field of metal processing, Selective Laser Melting (SLM) is already growingly
used in different applications such as tooling (Conformal Cooling) and for the production of small series of smaller
sized complex functional parts in metal. However, the technology still sets limitations for the use in applications
and sectors where a high material integrity is required, such as medical implants or parts for aerospace etc.
Nevertheless, the overall options using this technology e.g. for lightweight applications for the automotive and
aerospace industry are recognized. Exemplary, the development of specially designed brackets for the ETHZformula student racing car and their production using SLM are presented. The process specific limitations
regarding the material integrity are discussed, pointing out the needs for a further development of the technology
in order to allow the future use of SLM in high performance technological applications and sectors.
INTRODUCTION
The Selective Laser Melting (SLM) technology is a

freefrom fabrication process that allows the production of parts in diverse metallic materials such as
diverse stainless steel types, hot-work steel, Titanium,
Aluminum, Nickel-based alloys, Cobalt-Chromium
and others ... (Lohmeier 2005, Abe et al. 2003, Santos et al. 2004, Shellabear 2008, Uckelmann 2006,
Mumtaz et al. 2008, Averyanova & Bertrand 2010,
Childs et al. 2005, Gu & Shen 2008, Lopez et al. 2008).
It is a layer-wise process where a high energy laser
beam scans on a powder bed successively the crosssections of a sliced CAD data file of the required part.
This leads to the consolidation of the metallic powder particles in the scanned area, resulting in a nearly
fully dense layer of the part being built. Successive
lowering the build platform, generating the next powder layer and scanning the cross-sections leads to the
final physical part.
The properties of additive manufactured materials are already widely analyzed by many researchers:
Density (Spierings & Levy 2009a), microstructure
(Klingbeil et al. 2004, Kruth et al. 2004, Simchi 2006)
and the static mechanical performance (Sehrt & Witt
2009, Spierings et al. 2011, Yasa et al. 2010) are well
studied, pointing out that the static mechanical properties are typically in the range of wrought conventional
materials, although a specific anisotropy of several
percent can be observed. The dynamic mechanical
properties however are so far only partially analyzed

for some specific materials andAM processes (Sehrt &
Witt 2010, Spierings et al. submitted, Christensen et al.
2007). An analysis of stainless steel 316L and 15
5PH shows that the fatigue limit typically is lower
compared to conventional materials, even for polished
samples (Starr & Spierings 2010). Interestingly, the
parts surface quality plays only a minor role within the
dynamic lifetime range (Spierings et al. submitted). As
a result, if the dynamic strength fits to the applicationand design-specific requirements, the way for a usage
of SLM for diverse industrial applications would be
open, allowing a real Additive Manufacturing (AM)
instead of simple Rapid Prototyping.
However, besides the description of the material
microstructure and some mechanical properties, the
todays knowledge about the correlation between the
microstructure of SLM materials and corresponding
properties are still poorly analyzed. In this respect,
only little work has been put in the analysis of e.g.
the fracture toughness, the dynamic impact strength
(Yasa et al. 2010) or further parameters where the
microstructural situation of SLM materials plays an
important role.
This is at least one reason hindering the SLM
technology evolving in industry, especially in hightechnological areas: Companies are careful in cases
of high technological requirements on the material
side and as they typically have limited knowledge
and access to the material specific property databases.
785
This holds especially for the subsuppliers. A further

reason might be the production costs, which can be
high especially for high-volume parts due to a limited
process-productivity and auxiliary required support
structures that have to be removed manually. As a
consequence, SLM is still mostly used for the production of prototypes although in many cases the
technology would fulfill the application-specific
requirements.
The fact that the industry requires metal AMproducts is supported by Mungua (Munguia 2008)
who performed a survey in the industry about their real
needs. The results show that about 37% of all answers
required diverse metal materials. In this respect, SLM
has already been used for the production of parts for
diverse applications: The production of tools containing complex cooling channels (Conformal Cooling)
is already state of the art and allows the reduction
of cycle time and the improvement of the part quality in plastic injection molding (Gravet 2008, Herzog
2008, Spierings & Levy 2009b, Villalon 2005) and
the additive production of dental cups and bridges
in Cobalt-Chromium is already performed on a daily
basis by several industrial service providers. Further fields of application are in medical engineering
for the production of instruments and implants and
there is a wide potential for lightweight structures
using internal grid structures (Rehme & C 2006) with
potential application in the automotive and aerospace
industry. No other process allows the production
of structures with a comparable high complexity
(Figure 1).
In order to motivate industry to use additive manufacturing technologies, in this case SLM, the current
publication shows a case study about the development,
results and potential of the use of SLM by producing brackets for the Formula Student Car 2010 and
2011 of the team of ETH Zurich, first place winner in Silverstone 2010. On this basis opportunities
and limitations of SLM are discussed pointing out
the need for a specific further development of the
technology.
Figure 1. Example of a grid-structure without (left) and with

(right) a shell; produced by SLM at inspire-irpd.
CASE STUDY
2.1 About Formula Student

(Formula_Student_webpage 2011): Formula Student (FS) is Europes most established educational
motorsport competition, run by the Institution of
Mechanical Engineers. Backed by industry and high
profile engineers such as () Ross Brawn, the competition aims to inspire and develop enterprising and
innovative young engineers. Universities from across
the globe are challenged to design and build a singleseat racing car in order to compete in static and
dynamic events, which demonstrate their understanding and test the performance of the vehicle
2.2 Brackets of the FS-car of ETH Zurich
The brackets are the key elements of the connection
between suspension and chassis. All the static and
dynamic loads occurring during a race are lead through
the wishbones to the brackets and finally into the
monocoque itself. As one can see easily, high dynamic
forces are applied on these elements. In total, eight
different types of bracket geometries, each in a right
and a left version, are needed to attach the beams of
the front-and rear suspension to the frame. Additionally, brackets for the attachment of the spring-damper
elements are necessary.
The aim of the case study was to use SLM as an
innovative method for the production of the complex
brackets, to investigate the usability of additive manufactured parts in under real loadings and to identify
the potential and limitations of SLM regarding its use
in an industrial environment.
3
3.1 Production of the brackets with SLM

The brackets were produced using a Concept Laser M2
machine, which is equipped with an Nd-YAG laser and
a fiber is used for beam delivery The reported M2 of
Figure 2. AMZ Formula Student Car 2010 & 2011.
786
Table 1. Static mechanical properties of SLM produced and

conventional stainless steel 316L.
Material 316L
SLM
Handbook
Yield
strength
Ultimate
strength UTS
Elongation
at break
(MPa)
(MPa)
(%)
640
170310
760
480620
30
3050
(Spierings et al. 2010, Spierings et al. 2011); (GRANTAsoftware 2011)
the laser source is 1.02. The maximum laser power at

the building platform is about 190W. The scan strategy used to produce the brackets is a chessboard-like
structure, where 5 5 mm2 -squares are scanned. More
details are described in (Spierings & Levy 2009a,
Badrosamay et al. 2009).
The brackets were produced in 30 m layers using
the maximal laser power and a scan-speed of 1050
mm/s. The orientation of the brackets during the build
process, affecting the mechanical properties due to the
anisotropy, was selected according to the requirement
for a stable and easy process. As typical in industry
orientation dependent properties of the material were
not taken into account.
3.2
Table 2.
External loads on bracket Nr. xx.
Load case
Static load
Cornering outside
Acceleration
Brake
Fx
FY
FZ
(N)
(N)
(N)
698
560
413
1077
1338
1115
822
2143
473
380
280
730
There are no torques acting on the brackets.
Material
Stainless steel 316L was selected as a suitable material for the brackets. This material is a well-known
material for SLM with good mechanical properties
(Table 1). The powder granulation plays a significant
role regarding the mechanical properties and the processing characteristics. Details about the powder are
described in (Spierings & Levy 2009a). The material is
characterized by both static (Table 1) and dynamic (3)
mechanical measurements (Spierings et al. submitted).
It can be seen that for dynamic strength the measured
() stress amplitude values for R = 0.1 (surface as
built) are higher compared to literature values and
are even in the range of R = 1 for standard materials
(GRANTA-software 2011).
The observed anisotropy in the UTS between the
vertical and the horizontal build orientation is typically between 5% and 15% (Spierings & Levy 2009a,
Sehrt & Witt 2009), depending on material type and
powder granulation.
3.3
Figure 3. Comparison of SLM-316L to literature values.

Wrought, annealed values by (GRANTA-software 2011).
Load case for the most critical bracket
All bracket types were analyzed regarding their acting

forces and mechanical stresses using a FEM calculation with ANSYS. Exemplary, the static load case for
most critical bracket (front suspension, 4) is presented in Table 2. Breaking and cornering (inside,
outside) lead to dynamic loads on the brackets. For
1000 laps (around 1000 km) and 25 curves per lap, a
total of 75,000 cycles occur.
Figure 4. Examples of mounted brackets (Version 2010).

Upper left, right: rear suspension; lower right: spring-damper,
front suspension and anti-roll bar; lower left: front
suspension.
RESULTS
4.1 Brackets 2010

The brackets were directly assembled to the chassis of the car (different composite structures) without
any additional supporting structures (Figure 4).
4.2 Bracket design and FEM analysis
The FEM analysis for the front suspension-bracket
showed that the part was highly over dimensioned.
However, due to the lack of experiences, fatigue and
fracture toughness data of this material, the brackets
787
Table 3.
Bracket geometries for FS-car 2010 and 2011
Figure 5. Examples of mounted brackets Version 2011.
Figure 6. FEM analysis of the front suspension bracket:

static load case.
Although the stiffness and the maximum stresses

are similar compared to the ones of last year, the total
volume of the brackets is distinctively smaller. The
brackets of 2010 have a total volume of 115.3 cm3 and
those of 2011 have a volume of 66.5 cm3 (Table 3).
After mechanical finishing the brackets, a total weight
reduction of 34% was achieved.
The bracket redesign 2011 consider AM adequate
options in complex geometry and light weight. A further optimization in the style shown in 1 was not
considered. This was not performed as the whole chassis of the car 2011 remained the same, defining the
positions of the brackets etc. A further improvement
of bracket design and weight will be considered in
future.
were successfully produced and mounted on the car
(Figure 4). During one race-year, no problems were
reported regarding these brackets. On the basis of the
positive results, the overall need for a further reduction of the cars weight and additional fatigue results
(Spierings et al. submitted), the brackets for 2011 were
re-designed with special emphasis on mass reduction
(Table 3).
The FEM analysis showed again that the Von-Mises
stresses of most of the part structure were smaller than
100 MPa. In the critical point, the maximum stress was
373 MPa, which is about 1/2 UTS (Table 1), indicating a
high margin of safety. The dynamic evaluation showed
a mean stress of 300 MPa (static load) and a stress
amplitude of about 140 MPa (R 0.36). Although this
R-value is somewhat higher compared to measured
values, the maximum stress of 440 MPa is acceptable
(Figure 3).
4.3 Bearing bolts

The bearing bolts 2010 were made in Aluminum
7075-T6, whereas the 2011 versions were made in
steel. Although steel is around 2.8 times heavier than
Aluminum, design optimizations finally lead to the
same weight. However, the benefit was a significant
increase in stiffness.
5
LESSONS LEARNED
5.1 Challenges for a future use of SLM in industry

SLM at present is industrially used for the production
of either series of small parts (e.g. dental products)
or for the production of complex medium sized metal
parts. In low quantity such as tools with integrated conformal cooling channels. However, for many industrial
788
Figure 7. Bearing bolts Version 2010 (left) and 2011 (right).
applications, typical serial sizes required are in the

range of 10 to 100 pieces or more. For such cases,
economical limits are dominant. SLM is still an expensive process due to its low productivity, and followed
by intensive manual work post-processing operations.
Therefore and against the background of real industrial
needs (Munguia 2008) there are several aspects that
require a further development (Bourell et al. 2009) in
order that SLM can find its way to a wider industrial
acceptance and well-established production technology. Main emphasis should be put on the following
aspects (Figure 8).
a) Material database
Development of a comprehensive database covering all relevant materials and industrial applications. The database should contain all relevant
information needed to successfully dimension
structural parts. This requires information about
their mechanical properties: anisotropic static
and fatigue behavior, impact strength, compression strength, fracture toughness, etc.
Moreover a description of recommended process and post-process information should be
included: influences of different surface finishing operations (Spierings et al. submitted) and
other post processes like HIP etc.
Development of suitable modeling routines
allowing to successfully simulate e.g. complex
grid structures (Figure 1) and to predict their
mechanical behavior.
b) SLM process
Optimization of the SLM process in order to

minimize needed support structures and corresponding finishing operations as well as increase
the productivity of the process. This opens the
way for new design concepts.
c) Production equipment
Development of an SLM-production environment with increased automation regarding

machine handling, part unloading and separation
from the base plate as well as part finishing.
d) TQM system
Implementation of a comprehensive TQM system, as proposed by (Levy et al. 2007) ensuring
Figure 8. Ishikawa diagram: Fields of needed development.
reproducibility and reliability of the whole process chain.

e) Education
Development of university courses at graduate

and undergraduate levels and industrial training
programs.
Such measures allow widening the fields of suitable

application for Additive Manufacturing. Demonstrators and exemplary applications like the presented
results will open the way to implement the new design
possibilities into more real parts and raise the SLM
usability and confidence level.
CONCLUSION
The aim of the case study was to demonstrate exemplarily that SLM design features and materials can be
suitable for structural, load-bearing parts. It demonstrates that the static and dynamic mechanical properties already can meet the specific requirements of
many applications. This principally allows the use of
SLM for the production of end use parts in many
industrial sectors in the sense of Additive Manufacturing. The results encourage further evolving the
presented brackets towards extreme lightweight structures as shown in Figure 1. Furthermore, the very wide
design possibilities of SLM allow creating lightweight
structures e.g. in steel materials with around the same
weight, but a considerably higher stiffness, strength
(Figure 7) and fatigue strength compared to conventional structures in Aluminum.
However, there are also a number of challenging
limitations that hinder a wider use of SLM for the production of functional parts, especially for sectors with
high material requirements like aerospace, automotive or medical. Therefore the aim is to close the gap
between the current technological state and required
process-and production needs. An on-going effort with
possibly increased support by national and international funding is crucial on the way to a wide-ranging
success.
789
REFERENCES
Abe, F., Santos, E.C., Kitamura,Y., Osakada, K. & Shiomi, M.
2003. Influence of forming conditions on the titanium
model in rapid prototyping with the selective laser melting
process. Proceedings Of The Institution Of Mechanical
Engineers Part C-Journal Of Mechanical Engineering
Science, 217, 119126.
Averyanova, M. & Bertrand, P. 2010. Direct Manufacturing
of dense parts from martensitic precipitation hardening
steel gas atomized powder by Selective Laser Melting
(SLM) technology. In Advanced Research in Virtual and
Rapid Prototyping, ed. P. J. e. a. Brtolo, 343348. Leiria,
Portugal: CRC Press/Balkema.
Badrosamay, M., Yasa, E., Van Vaerenbergh, J. & Kruth, J.P.
2009. Improving Productivity Rate in SLM of Commercial Steele Powders. In RAPID, 113. Schaumburg, IL,
USA.
Bourell, D.L., Leu, M.C. & Rosen, D.W. 2009. Roadmap
for Additive Manufacturing Identifying the Future of
Freeform Processing. Austin: The University of Texas at
Austin-Laboratory for Freeform Fabrication.
Childs, T.H.C., Hauser, C. & Badrossamay, M. 2005. Selective laser sintering (melting) of stainless and tool steel
powders: experiments and modelling. Proceedings Of The
Institution Of Mechanical Engineers Part B-Journal Of
Engineering Manufacture, 219, 339357.
Christensen, A., Lippincott, A. & Kircher, R. 2007. Qualification of Electron Beam Melted (EBM) Ti6Al4VELI for Orthopaedic Implant Applications. In Materials
& Processes for Medical Devices (ASM International),
4853. Palm Desert, California: ASM International.
Formula_Student_webpage. 2011. www.formulastudent.com.
CES EduPack.
Gravet, D. 2008. Hipermoulding: efficiency of conformal
cooling. In International Conference on Polymers and
Moulds Innovations PMI. University College Ghent,
Belgium.
Gu, D. & Shen, Y. 2008. Processing conditions and
microstructural features of porous 316L stainless steel
components by DMLS. Applied Surface Science, 255,
18801887.
Herzog, F. 2008. LaserCusing technology optimises die
casting dies. In Casting Plant & Technology, 1820.
Klingbeil, N.W., Bontha, S., Brown, C.J., Gaddam, D.R.,
Kobryn, P.A., Fraser, H.L. & Sears, J. 2004. Effects
of Process Variables and Size Scale on Solidification
Microstructure in Laser-Based Solid Freeform Fabrication
of Ti-6Al-4V. In Solid Freeform Fabrication SFF. Austin
TX.
Kruth, J.P., Froyen, L., Van Vaerenbergh, J., Mercelis, P.,
Rombouts, M. & Lauwers, B. 2004. Selective laser melting of iron-based powder. Journal Of Materials Processing Technology, 149, 616622.
Levy, G., Schindel, R., Schleiss, P. & Spierings, A.B. 2007.
Quality Management Representation and Application
for Rapid Manufacturing with SLS (Selective Laser
Sintering). In 15th International Symposium on Electromachining (ISEM XV).
Lohmeier, K.M. 2005. Stahl-und Aluminiumteile: Praktische Erfahrungen mit generativem Prototyping. In 3. Swiss
Rapid Forum. St.Gallen.
Lopez, G., Williams, S., Miranda, R.M., Quintino,
L. & Rodrigues, J.P. 2008. Additive manufacturing of
Ti-6Al-4V based components with high power fiber
laser. In Virtual and rapid manufacturing, ed. B. et.al,
369374.
Mumtaz, K.A., Erasenthiran, P. & Hopkinson, N. 2008. High

density selective laser melting of Waspaloy. Journal of
Materials Processing Technology, 195, 7787.
Munguia, J. 2008. Pursuing successful Rapid Manufacturing in a standards-less industry: a best practice approach.
In Virtual and rapid manufacturing, 477481. Leira
Portugal: Bartolo et al. (eds)
Rehme, O. & C, E. 2006. Rapid manufacturing of lattice
structures with Selective Laser Melting. In Laser-based
Micropackaging, Proceedings of SPIE
Santos, E.C., Osakada, K., Shiomi, M., Kitamura, Y. &
Abe, F. 2004. Microstructure and mechanical properties of
pure titanium models fabricated by selective laser melting.
Proceedings Of The Institution Of Mechanical Engineers
Part C-Journal Of Mechanical Engineering Science, 218,
711719.
Sehrt, J.T. & Witt, G. 2009. Auswirkungen des anisotropen
Gefges strahlgeschmolzener Bauteile auf mechanische Eigenschaftswerte. In Rapid.Tech. Erfurt Germany:
Messe Erfurt AG.
Sehrt, J.T. & Witt, G. 2010. Dynamic strength and fracture
toughness analysis of beam melted parts. In International
MATADOR Conference. Manchester, Lancashire, United
Kingdom
Shellabear, M. 2008. Direct Metal Laser-Sintering of
Titanium. NASA Tech Briefs May.
Simchi, A. 2006. Direct laser sintering of metal powders:
Mechanism, kinetics and microstructural features. Materials Science and Engineering: A, 428, 148158.
Spierings, A., Starr, L. & Wegener, K. submitted. Fatigue
performance of Additive Manufactured metallic Parts.
Rapid Prototyping Journal, submitted.
Spierings, A.B., Herres, N. & Levy, G. 2010. Influence
of the particle size distribution on surface quality and
mechanical properties in additive manufactured stainless
steel parts. In Proceedings of the Annual International
Solid Freeform Fabrication Symposium, ed. D. L. Bourell.
Austin, Texas.
Spierings, A.B., Herres, N. & Levy, G. 2011. Influence of the
particle size distribution on surface quality and mechanical properties in additive manufactured stainless steel
parts. Rapid Prototyping Journal, 17, 195202.
Spierings, A.B. & Levy, G. 2009a. Comparison of density of
stainless steel 316L parts produced with selective laser
melting using different powder grades. In Proceedings
of the Annual International Solid Freeform Fabrication
Symposium, ed. D. L. Bourell, 342353. Austin, Texas.
Spierings, A.B. & Levy, G. 2009b. Rapid Manufacturingauch mit SLM. In Rapid.Tech. Erfurt, Germany: Messe
Erfurt AG.
Starr, T.L. & Spierings, A.B. 2010. Tensile and Fatigue Performance of Laser Sintered Stainless Steels. In Proceedings
of the Annual International Solid Freeform Fabrication
Symposium, ed. D. L. Bourell. Austin, Texas.
Uckelmann, I. 2006. Generative Serienfertigung von individuellen Produkten aus CoCr mit dem Selektiven
Laser Schmelzen. In Fakultt fr Maschinenwesen
der Rheinisch-Westflischen Technischen Hochschule
Aachen, 110. Aachen: Technische Hochschule Aachen.
Villalon, A.V. 2005. Electron Beam Fabrication of Injection Mold Tooling with Conformal Cooling Channels 68.
Raleigh: North Carolina State University
Yasa, E., Deckers, J., Kruth, J.P., Rombouts, M. & Luyten, J.
2010. Experimental investigation of charpy impact tests
on metallic SLM parts. In Advanced Research in Virtual
and Rapid Prototyping, ed. P. J. Brtolo, 207214. Leiria,
Portugal: CRC Press/Balkema.
790

The integration of solutions that evolved by nature, into innovative

products of the future
R. Becker
Fraunhofer Institute for Manufacturing Engineering and Automation (IPA), Stuttgart, Germany
ABSTRACT: This paper presents sample products that were developed with additive manufacturing in mind
and were inspired by bionic archetypes. The applicable bionic archetypes being investigated and used were
mainly actuation and locomotion principles. One of them is the well known as the elephants trunk like robot,
that was developed in cooperation with the German company Festo. This product was awarded with the German
future award 2010 (Deutscher Zukunftspreis des Bundesprsidenten), the highest award for technical innovations
in Germany and is presented by the Federal President. This illustrates the possibilities of Additive Manufacturing
which are widely accepted in the scientific community and increasingly accepted in the industry as certainly
leading towards astounding new products.
1
INTRODUCTION
Additive manufacturing gives a geometric design

space for the product development that is almost without limits. More complex shapes can not only be
thought of and designed but also manufactured. With
the new manufacturing processes we can move faster
and more direct towards solutions for technical problems initially being found in nature. This includes not
only complex outer shapes but also functional inner
structures which can be integrated into the challenging
lightweight products of the future.
The Fraunhofer IPA concentrates on the research
and systematic development of possibilities that will
arise by these new manufacturing processes introduced as processes in industrial scale and quality. One
of the main goals of these efforts is to find and define
solutions and functions with the possibility to be integrated directly in products. Achieving this goal will
also increase the value of the manufactured products.
Our recommendation in cases alike is to take a close
look to the solutions presented by nature, which is
permanently testing these solutions in very long and
extensive test series.
2 DESIGN FOR ADDITIVE

MANUFACTURING BASICS
Additive Manufacturing is an uprising and very
promising production technology that has some advantages compared to the existing technologies. It will be
established in an industrial scale in the near future for
a range of applications that could clearly be defined
and that will have also clear economic advantages.
But do not forget that there are some limiting factors at the moment that need to be improved in the
near future to get an even broader range for application of Additive Manufacturing. The limiting factors
at the moment are the surface quality, the limited
size of economic products, a limited number of available materials and the output volume per hour. At
the IPA, we are normally using the laser sintering of
Polyamide 12 that has the best ratio of price and properties of all standardAdditive Manufacturing materials
for designs that have to withstand mechanical loads in
industrial environments.
2.1 Geometric freedom in the design of objects
In the laser sintering of plastics you have a total geometric freedom in designing objects with the only
limitation that you have to be able to get your ideas in
a 3-D CAD system. Another logical limitation is that
totally closed hollow structures make no sense because
it is impossible to extract the remaining powder from
there.
2.2 Complexity of geometries and functional
lightweight structures
In the additive manufacturing production process it
does not make any difference in time and price if a
build part is somehow very simple or if it is really
complex. So especially quite small but very complex
parts are the preferred ones to be built with Additive Manufacturing. As you can see in the following
example the way of designing a part changes from
design for manufacturing towards the combination of
all needed mechanical functions with a structure that
is as lightweight and material saving as possible but
also as stiff as needed.
The shown gripper is already established as an
original part delivered with packaging machines of
791
Figure 3. Robot gripper system with flexible and solid areas

built in one single piece. The flexible zones are the bellow
actuator and the deformation zones behaving as living hinges
for a limited angle (Source: IPA).
Figure 1. Concept study of an individually produced splint

with improved air ventilation, shower stability and less thick
walls than a normal splint made out of plaster. It is one single
part that could be opened and closed by integrated fasteners
(Source: IPA).
built and is increasingly solid when used in more compact volumes. These properties of the material can
be utilized to integrate very stiff and very flexible
areas in one single part without any mounting like
screwing or gluing. Such structures have been demonstrated by the Fraunhofer IPA for a few years. One
of the early examples are the well known internationally patented grippers from IPA (shown below) with a
bellow as actuator combined with a solid framework
and adaption flange, all built in one single piece.
3 AUTONOMOUS CRAWLING
ROBOTS MECHANICAL STRUCTURE
BUILT ENTIRELY WITH ADDITIVE
MANUFACTURING
Figure 2. Vacuum gripper for automated high speed handling applications (Source: IPA).
an Italian producer many times. Since the Interpack

2011 in Dsseldorf, Germany they are also available
as a catalog article from a large German company
producing and reselling vacuum components and
vacuum handling systems.
The air ducts leading from the connection point of
the tube towards the thread holes where the suction
cups will be mounted are integrated in the framework
of the gripper. Internal structures like honeycombs or
three dimensional wireframes are another possibility
to push the design to new lightweight construction limits. This offers the advantage to copy bionic structures
much more direct than possible with the conventional
manufacturing processes, for example an adapted size
and wall thickness of the honeycombs referring to the
distribution of the stresses in the part. Polyamide 12
has a quite high elasticity when thin structures are
Crawling autonomous robots are a field of constant

scientific development. Such robots are compared to
wheel based robots (e.g. modern vacuum cleaners)
normally used to move on uneven ground or in rough
terrain. The movement of crawling robots with multiple legs with many degrees of freedom is normally
slower than the wheel based ones but not stopping
so easily in front of obstacles. The complexity of the
mechanics of the body and legs and the complexity
of the control are the main reasons that make it so
difficult to establish them a standard machine and to
develop economic solutions.
3.1 State of the art in crawling robots
Crawling, biped or quadruple legged robots are known
from different projects around the world. All of them
combine a highly sophisticated mechanical design
and are normally electrically driven. One exception
is the so called Big Dog from Boston Dynamics that
at the moment definitively sets the standard for the
792
Figure 4. Example of a conventionally made crawling robot

showing the complexity of the mechanical system (Source:
Pfeiffer F., Philosophical Transactions of the Royal Society.
A 2007;365:109131).
control, stabilization and bioincally inspired ultra-fast

handling of irregular situations. There are videos on
YouTube about this project.
In the picture below you can see a random example
of a crawling robot that obviously shows the complexity of the mechanics of such a moving platform. Every
hinge normally has an own drive, a gearbox and a position sensing sensor. To reach a good quality of the
estimation about the stability and the position, all the
systems are built as stiff as possible. This fact includes
that with moving the legs on their circular path,
nearly all other hinges have to move synchronized to
compensate the not directional movement.
4 THE FRAUNHOFER IPA CRAWLING ROBOT
The Fraunhofer IPA crawling robot was designed with
some, maybe unusual, boundary conditions in mind.
All of the discussed crawling robots are made with a
huge amount of parts, a high assembly effort and with
high precision and very complex control units. This
prevents from being sold very cheaply and being used
very often. The idea behind the IPA crawling robot was
to build a mechanical platform for multiple applications as simple as possible, avoiding overengineering
in ways of precision and durability, with minimized
assembly time and a cost structure that allows to establish it as a so called single use cheap robotic platform
in hazardous areas like atomic power plants or to sell
them as technical toys for nerds. All these features are
very well supported by the use of Additive Manufacturing as the preferred production method. The design
of the crawling robot can be made systematically for
Additive Manufacturing to use all the possibilities that
are included there.
4.1
Bionic predecessor The spider
Bionic is a scientific discipline that is not clearly

defined yet. From the Fraunhofer point o view it
means something like . . . a variation of the innovation process, which has as a key component the
transfer of information from the biology to the technical field with the goal to develop technical products
or processes based on biological prototypes.
Figure 5. Spider Cupienniussalei (Source Barth-2001,

page 412).
Figure 6. Spiders leg (Source Foelix-1992, page 21).
Starting from the Spider as one crawling, multi

legged animal that is very well adapted to the movement in rough terrain we analyzed the mechanical
system very much in detail.
4.2 Movement of the spiders legs
A spider has commonly four pairs of legs hat could be
controlled separately. All the mechanics for the drive
of the legs are arranged radial around the front body
part.
As shown in the following picture all eight legs are
built up to a large extend identical in their mechanical
structure. They consist from seven sections that are
connected by different types of hinges with different
degrees of freedom. The front legs are pulling the body
and the rear ones are pushing forward in a way that
the movement of the legs in one step is oppositional
(See figure 7). Spiders can make really impressing
movements with the available degrees of freedom like
climbing, jumping, running and changing the direction
in an edge
After studying the real spiders legs we reduced the
degrees of freedom to the really needed functions to
move on nearly flat surfaces. In the first prototypes
we used one joint for moving the whole leg radial, one
major joint for lifting the outer area of the leg and an
additional joint as a flexible balancing element. An
interesting fact is that the spiders drive their legs by a
combination of muscle contraction and hydraulic force
as shown in the picture below.
793
Figure 7. Pushing (a) and pulling (b) of the opposite legs

for moving forward (Source: Ehlers-1939, page 398).
Figure 9. Leg and joint section in comparison between the

biological archetype and the leg designed for additive manufacturing (Sources: Reuenzehn-2010, page 10 and Barth
et al-1970, page 343369 and page 350351).
Figure 8. Natural bellow structure of a spiders legat the top

of the page (Source: Blickhan et al.-1985, page 127) compared to a bellow structure made by Additive Manufacturing
shown below (Source: IPA).
The joints have an integrated bellow made from

flexible skin like structure that expands when set under
pressure and move the extensions around a virtual axis.
Using bellow structures as a mechanical driving mechanism is perfectly fitting for Additive Manufacturing
and is well developed at the Fraunhofer IPA. The bellows are driven by pneumatic force and the supporting
channels are all integrated into the leg that is made out
of one piece.
4.3
Crawling robot designed for additive

manufacturing
Additive Manufacturing enables us to implement all

the mechanical functions of the locomotion apparatus
in only a few parts, eight legs and the one body that
can be easily assembled without using any tools. The
complexity of the whole assembly is reduced drastically and is shifted to the 3-D CAD in the data model.
Things like the air ducts, the joints and for example
Figure 10. First prototype in real size made out of 9 single

pieces (Source: IPA).
the pressure reservoir are fully integrated without extra

costs and with an outstanding lightweight result.
The robot made by Additive Manufacturing has an
elasticity of the thin walled sections that will prevent it
from damages in crash situation or when falling down
a staircase. The amount of needed energy to move the
robot will be low and it is planned to use it as an
autonomous platform with an electrically driven compressor that can carry out additional tasks with some
payload for additional equipment.
The assembly time and the needed qualification
therefore are reduced to a minimum that can never
be reached by a conventional manufactured design.
794
Production costs are relatively low and depending

on the scaling factor of the whole robot. The scaling
factor is not totally free but can be changed in a wide
range.
All these factors together will give an access to several new applications and for example to the market of
technical toys.
5
CONCLUSIONS
New and fascinating possibilities arise from the use

of Additive Manufacturing as an industrial production
process. It enables us to get much closer to biological archetypes and their highly developed solutions
for special tasks. The recommendable use of an integral design concept includes connectors and all kinds
of small parts like cable fixations in very few parts and
reduces the complexity of large assemblies, parts management and assembly time. As shown in this paper it
is for example possible to develop and produce the
whole mechanics of crawling robots out of only very
few parts that could be mounted without tools in a very
short time. Last but for sure not least it is possible to
build very lightweight and very stiff 3-D structures
that save energy for the movement and expand the
autonomy of the system.
Crawling robots made by Additive Manufacturing

will revolutionize the market for low cost robots in
many fields of applications
REFERENCES
Barth, Friedrich G. 2001: Sinne und Verhalten aus dem Leben
einer Spinne. Berlin, Heidelberg: Springer
Blickhan, Reinhard; Barth, Friedrich G.1985: Strains in the
exoskeleton of spiders. Journal of Comparative Physiology A: Neuroethology, Sensory, Neural, and Behavioral
Physiology, Volume 157.Berlin, Heidelberg: Springer,
Page 115147
Ehlers, Max1939: Untersuchungen ber Formen aktiver
Lokomotion bei Spinnen. Zoologische Jahrbcher, Volume 72. Jena: Fischer, Page 374499
Foelix, Rainer F.1992: Biologie der Spinnen. Stuttgart:
Thieme
Ondratschek, Kai 2011: Development and design of a bionic
spider robot based on additive manufacturing, Stuttgart:
Fraunhofer IPA
Reuenzehn, Stefan 2010: Mechanical design of the legs of
Dolomedes aquatics, Masterthesis. Dunedin: University
of Otago, Page 10
Barth, Friedrich G.; Libera, Wolfgang 1970: Ein Atlas der
Spaltsinnesorgane von Cupiennius salei. Zoomorphology,
Volume 68. Berlin, Heidelberg: Springer, Page 343369
and Page 350351
795

Comparative analysis between a CAD model design and physical models

obtained by manufacturing additive technologies using optical scan
F. de Alencar
Design Department, FAAC/UNESP, So Paulo State University, Bauru SP, Brazil
P.J. Brtolo
CDRsp, Centro de Desenvolvimento Rpido Sustentado de Produto, Instituto Politcnico de Leiria, Portugal
ABSTRACT: This article presents a detailed study of the application of different additive manufacturing technologies (sintering process, three-dimensional printing, extrusion and stereolithographic process), in the design
process of a complex geometry model and its moving parts. The fabrication sequence was evaluated in terms
of pre-processing conditions (model generation and model STL SLI), generation strategy and physical model
post-processing operations. Dimensional verification of the obtained models was undertook by projecting structured light (optical scan), a relatively new technology of main importance for metrology and reverse engineering.
Studies were done in certain manufacturing time and production costs, which allowed the definition of an more
comprehensive evaluation matrix of additive technologies.
1
INTRODUCTION
More recently one of the most revolutionary applications targeted for the production of physical models,
are related to the additive manufacturing processes
(Rapid Prototyping-RP). The term relates to a class of
technologies that can automatically construct physical
models by adding three-dimensional material. Basically, a computer interprets information from CAD or
CAE environment of the object geometry to be built
and converts this information into the CAM environment, in which a rapid prototyping machine builds the
object physically (Barbosa, 2009).
The additive processes have revolutionized the
design, engineering and manufacturing approach in
segments as diverse as aerospace, agricultural, construction, automotive, educational, healthcare, high
technology, industrial, marine and consumer goods.
Devoted primarily to the delay in constructing prototypes, is now used throughout the design process, and
extended its scope of application solutions for tooling
and production.
The RP was developed with the aim of reducing a
product development time, in order to minimize Engineering and Design problems, increase product life,
creating changes in the initial design, reducing a prototype building time compared to conventional methods.
Another application for models obtained by additive
technologies is regarding visual marketing, testing and
simulation.
It is also useful to validate conceptual models with
speed and low cost, tuning feasible what was before
only on paper or on computer screen. It is therefore
a resource to materialize ideas, which facilitates a
Figure 1. Physical models and its part details made by additive technologies. Source: Company CODI/Leiria/Portugal.
product visualization and generating a report of a

particular piece, allowing the fittings checking, interference with electronic components or internal parts,
dimensions and ergonomics (Figure 1).
There are a number of terms to designate, such
as free-form fabrication, manufacturing, solid freeform self fabrication, automated manufacturing in
a free, digital fabrication, 3D printing, prototyping,
laser manufacturing by-layer manufacturing addition
and a solid display. The multiplicity of terms and
definitions may confuse a discussion or description
of the technology.
The development of processes and materials currently used in additive technologies has allowed the
production of customized parts and shapes/designs,
which would not be possible by means of conventional
manufacturing resources, especially those involving
the injection of polymeric materials. Unlike the goals
and purposes that characterize the additive processes,
the current concern is go beyond the production of
physical models with the visual purpose only, but most
importantly, the possibility of building up models for
the purpose of direct application in a product system,
797
eliminating some specific procedures and restrictions

from conventional manufacturing processes.
The additive technologies for rapid manufacturing eliminate strategies and design requirements when
obtaining objects/products with a high level of geometric complexity, unlike the objects produced by injection molding in which the systems use conventional
tooling, using drawers or negative tooling angles,
requiring considerable technical effort to obtain.
However, as we will see, from the dimensional
point of view, additive manufacturing processes are
less accurate than the conventional subtractive processes. Some of those errors are: a) Resolution errors.
b) Errors due to the existence of an insufficient number
of triangles in the STL file. c) Errors associated with
an inadequate mesh of triangles. d) Errors associated
with the manufacturing process (warping, contractions, distortions etc.). e) Errors due to thick layers
(Brtolo, Mateus, 2002).
2
HYPOTHESIS
The hypothesis in which this research was the possibility of increasing the participation of additive processes
in product development and rapid manufacturing by
means of a broader understanding of the limits, similarities, technical characteristics and quirks of those
technologies. To achieve that, it was developed the
design of a CAD model, transforming it into STL file,
producing physical models by SLS, PolyJet, FDM and
3D Printing technologies, and through optical scanning, checking errors and deviations presented by the
constructive system of each analyzed technology.
3
DESIGN AND PRODUCT DEVELOPMENT
Like other areas of knowledge, product design was

heavily influenced by the technological revolution of
recent years, the arrival of new technologies and new
production processes, all associated with new design
methodologies adapted to this new context. Moreover,
due to the fact that currently, development of industrial
products is part of a complex and competitive technological universe, in which quality and efficiency are
paramount requirements, there is a call for designers
and engineers to use those new technologies and computer languages in order to minimize the risks involved
in the process (Alcoforado, 2008).
Segundo Gross and YI-Luen (2004), designers are
concerned with three different types of tasks: the usage
of drawings and models to communicate with other
members of staff; the predictions about the development and implementation of the project that they will
run, and the division of a range of activities in parallel
with the design process, aiming to generate new ideas,
testing and evaluating new alternatives of design, and
communicating with professionals and clients.
To achieve these goals, the designer adopts a
series of representation techniques, design languages,
mainly based on drawings and digital technologies,
which transmit information in a summarized form.

The representations used in design aims to facilitate
the information communication among stakeholders
(project team, suppliers and customers), to integrate
knowledge involved in the process, and to assist the
decision making, facilitating the product development,
minimizing the doubts around the process (Volpato,
Ferreira & Santos, 2007).
Thus, physical models are of great importance
because they allow to check and verify the design solutions for different purposes, ranging from perception
to objective analysis, dimensional and functional analysis or even more subjective analysis including aspects
regarding form and semiotic.
All types of representations are important, however,
the three-dimensional representation allows an easier
and more comprehensive design analysis and product
diagnosis before its production. This kind of representation includes a model, a prototype or a mockup, as
long as each one can be useful in a particular phase
of the design process (Barbosa, 2009). According to
Volpato, Ferreira & Santos (2007), most professionals,
regardless of age, sex or social class, tends to assimilate
the design information more quickly if it is transmitted
through a physical model instead of two-dimensional
drawings or illustrations on paper or computer. This is
because the physical model allows a interaction based
on synestesia, reducing the cognitive effort to interpret
words and two-dimensional images and to imagine
how the final product is going to be.
DESIGN AND RAPID MANUFACTURING
The DFM-Design for Manufacturing is a design philosophy applied since the early stages of a product
development, aiming to design parts and products that
can be produced more easily and in a more economic
way (Hague at al, 2003).
The direction of the Design for Rapid Manufacturing is already seen in some cases, as a real possibility
considering the current development stage of additive
technologies, although there is still huge restrictions
involving either economic or technological reasons.
The Rapid Manufacturing is still emerging, but there
are strong indications that many design leaders and
major manufacturers are using additive technologies to manufacture parts or components for specific
applications.
Wohlers, T, Grimm, T. (2010) states that the market
for additive fabrication increased 14.6% between 2004
and 2005 (from $707,200,000 to approximately $808.5
million). These companies have been accumulated
benefits by adopting Rapid Manufacturing, either by
reducing time-to-market, eliminating tools and their
costs, for example, or by manufacturing products that
could not be done otherwise. This last point is certainly
the most liberating feature of Rapid Manufacturing
which its freedom provides a fast production in terms
of design capability without the restrictions usually
imposed by traditional production process.
798
The design concept for Rapid Manufacturing can

be greatly aided by additive technologies, taking into
account the fact that the prototypes obtained by this
way give to designers and engineers an adequate number of visualizations, assemblies, functional tests and
perception about the possibility of production. The
major contribution is the production of parts or products which cannot be produced in series and by any
conventional process. This allows an increase in the
level of the designers perception about what is possible or not possible to produce, analyzing where the
complexity of the product can be reduced to facilitate their production (Smith, 2002). For Kaminski
(2004), in the product development process, there are
two distinctive moments: the design itself and the
design development. This means that when developing a product is necessary to define which applications,
users, processes and product specifications that has to
be followed. The establishment of a concept for this
product would be the design, whereas the final project
characteristics involving the materialization of that
design concept through manufacturing and its steps
and deadlines would be the design development.
By these definitions, it can be assumed, considering
a natural order for a product development process, that
and the additive technologies can make the difference
between what can be considered a product or a mere
a concept of it. Therefore, the Rapid Manufacturing
can be applied in these two stages of the process of
developing a product: for conceptual analysis of the
product or to understand the already designed product
(Selhorst, 2008).
5 THE ADDITIVE PROCESS FOR PROTOTYPE

PRODUCTION
The structure in all additive processes includes the
transformation of a CAD model in an intermediate
model in which the surfaces are described by a network of triangular elements in STL format. The greater
the number of triangles used in STL format and the
smaller the thickness of each layer, the greater the precision. In this regard we must consider an important
parameter called Problem Dimension which quantifies
the influence of the number of triangles and number of
layers in the time of creation of SLI files. The greater
the Problem Dimension the greater the time required
for SLI file generation (Brtolo, Mateus 2002).
The existing additive processes can be classified
according to the initial state of raw material, where the
main processes is organized into three groups: those
based on liquid, those based on solid and those based
on material in the form of powder (Volpato, Ferreira &
Santos, 2007).
obtained in the Assembly/Assembly environment. We

opted for a design consisting of two moving components characterized by the impossibility of production
by any conventional manufacturing process.
After converting the file to STL, it was sent to make
physical models using the using the additive technologies: Polyjet, SLS (Selective Laser Sintering), FDM
(Fused Deposition Modeling) e 3D Printing/Zcorp.
7
PROCEDURES FOR OPTICAL SCAN
For scanning the model, the COMET5 equipment was

used, and for dimensional analysis we use the COMET
software Inspect 4.5, both from CDRsp/IPL Laboratory in Leiria/Portugal. All models were generated the
same way to capture the clouds of points, and the deviation analysis was obtained by correspondence between
Nominal and Actual model. It was noticed in the CAD
system that it was impossible to keep the concentricity
geometrical relationship between the two components
after the production of its models. During the material
support removal, even smallest ones, inevitably, would
shift its positioning, causing, as a result, a scanning
error. For scanning purposes, it was necessary to fix
both components in order to catch the clouds of points
more appropriately. However, due to the inner component moving, it was not considered in the Nominal and
Actual models analysis.
8
DIMENSIONAL ANALYSIS
In Figure 3, is shown the deviation analysis was

obtained by correspondence between Nominal and
Actual model. The color standard indicates the deviations occurred in the current model. In that correspondence, the scaling factor used was of 0.8 mm for more
or for less.
Figure 2. Physical
technologies.
models
obtained
by
additive
The Design of the CAD model took advantage from

the additive technologies resources, in particular those
Figure 3. COMET5 scanning equipment and performed.
799
Figure 4. A = Model views SLS, B = Model views Polyjet,

C = Model views 3D Printer ZCorp, D = Model views FDM,
E = Legend of the scale of deviations in millimeters.
However, smaller values than this, initially set at

0.3 mm, in the SLS technology, presented none indication of such deviation, which means that just few
areas in the model were affected by deviation analysis
making difficult to measure the differences occurred.
9
Analyzing the data from scanning process and also

from Inspect software in all analyzed physical models,
the FDM technology presented the smallest deviation,
followed by PolyJet, Zcorp and SLS. This superior performance of the FDM technology is demonstrated in
comparative tables, in which it is also possible to see
that it does not present a higher production or material
cost, nor in material for building support. On the other
hand, its production time is critical resulting in low productivity also demanding the maximum occupancy of
its building area. Although the FDM has been the technology with the smallest number of layers (479) with
thickness of 178 m, it was necessary over 34 hours
of processing. This leads to the conclusion that it is the
slowest of all analyzed technology, even considering
the possibility to compensate this through more efficient building design procedures such as the reduction
in the number of layers and the form and deposition
of both base and support materials. In the case of the
PolyJet technology with 5988 layers and thickness of
16 m required half the time spent by the FDM, but
not enough to produce its superior results. However,
as it was not possible to apply Shell on their walls as it
was possible for the SLS and the FDM, it resulted in
models with thick walls that lead to more tensions and
warping. Therefore, due to what was exposed above in
association with the data of the deviation analysis, it
is possible to conclude that there are heat concentration and tension involved in the PolyJet technology
process, although not significantly. In addition, its
building system based on an excessive processing in
slicing layers, and materials used, justifies its high cost.
The 3D Printing/ZCorp technology presented the lowest cost of all compared technologies. With a number
of 812 layers with 100 m of thickness, it ends up

requiring the lowest processing time (about four hours)
which is extremely good for the rapid manufacturing
scenario. With no significant loss in terms of surface
quality and dimension, it can be stated that such results
are due to the fact that the 3D Printing/ZCorp is the
only technology among all which does not involve
heating in its building process.
Beyond the peculiar weakness of the models built
by this technology which ends up not allowing its application in many systems and products, it also has
another drawback: the high weight. However it has
proved to be a very adequate process with a vast range
of possibilities in the production of masters for some
types of molds as well as for small series of products.
Even though the SLS technology is widely used
for industrial applications, the models did not produce the best results, very probably also by recycling
the powder used. Irregular-shaped depressions and
excavations were visually detected on their surfaces
and also confirmed in the deviation analysis. All the
obtained and analyzed results classify the compared
technologies in a qualitative order already described.
However, it is important to note that the performance
rating of this study was determined by the volume and
geometry of each model, as well as their thicknesses.
In that sense, we can not necessarily guarantee the
same technology classification if the models were built
from a different volume/geometry and especially with
smaller thicknesses.
REFERENCES
Alcoforado, M.G. 2008, O Poder Comunicativo dos Prottipos. In: 8 Congresso Brasileiro de Pesquisa e Desenvolvimento em Design, SENAC, So Paulo.
Barbosa, R.T. 2009, Design e Prototipagem: Conhecimento
e Uso da Prototipagem Rpida no Design Brasileiro.
Universidade Estadual Paulista Julio de Mesquita Filho,
Bauru.
Brtolo, P.J., Mateus,A. 2002, O Estado daArte dos Processos
Aditivos de Prototipagem Rpida. Revista O MOLDE.
Gross, M. Yi-Luen do, E. 2004. The Three Rs of Drawing
and Design Computation, First International Conference
on Design Computing and Cognition. Kluwer, at MIT, July
1921.
Hague, R.; Mansour, S.; Saleh, N. 2003. Design Opportunities with Rapid Manufacturing. Assembly Automation;
23, 4; ABI/INFORM Global pg. 346.
SELHORST, A. J. 2008. Anlise comparativa entre os processos de prototipagem rpida na concepo de novos produtos: um estudo de caso para a determinao do processo
mais indicado.125f. Ps-Graduao em Engenharia de
Produo e Sistemas da Pontifcia Universidade Catlica
do Paran, Curitiba.
Volpato, N., Ferreira, C.V. & Santos, J.R.L. 2007, Prototipagem Rpida: tecnologias e aplicaes. Edgard Blcher,
So Paulo.
Wohlers, T, Grimm, T. 2010. Is CNC Machining Really Better
Than RP?. Perspective. http://www.timecompression.com
800

Rapid prototyping for original design

N.G. Harris & T.J. Coole
Buckinghamshire New University, UK
ABSTRACT: This paper looks at the benefits that rapid prototyping is able to offer when applied to the
verification of original design work;it includes a brief review of the available 3D CAD systems and the various
rapid prototyping processes and how they each fulfil a necessary role within this process.
The paper then looks at how rapid prototyping can be utilised for various aspects of design validation; highlighting any potential pitfalls that could be encountered, reviewing the comparative costs with more conventional
manufacturing methods and time to process; but overall looking at the benefits that that can be realised for both
academics and engineers.
The concepts are supported by the use of a case study using rapid prototyping to produce a number of parts
for a concept model submitted as part of the authors dissertation at Bucks New University, UK.
BACKGROUND
1.1 3D CAD Systems

To realise the many possibilities of rapid prototyping,
there will be an inherent requirement to use a 3D CAD
system. There are many 3D CAD systems available on
the market ranging from very basic entry level programs through to higher end systems. The multitude
of systems all vary in their capabilities, from their user
friendliness and levels of user support, through to their
variety of feature content, robustness in operational
use, capability and what ultimately can be achieved
with the systems.
Inevitably the myriad of 3D CAD systems also have
a wide variation in their cost implications, with some
of the higher end commercial packages commanding very high setup costs and sometimes what can be
considered as prohibitive annual maintenance costs.
The high end packages due to their cost implications
are therefore predominantly used by medium to large
companies and also by the larger academic institutes.
The companies who prefer the more expensive
higher end systems do so for the more positive benefits that those higher prices command; e.g. the systems
often have to be used by multiple teams of engineers
and designers who need to share data easily. Industry
companies need to ensure minimum down time due
to system problems and the robustness coupled with
more instant support levels of the higher end systems
offer better assurances of stability and continuity of
working capabilities [1].
Equally many academic institutes also prefer the
use of the higher end systems, just like industry they
often require multiple seats for deployment and also

prefer the good levels of robustness offered with the
higher end systems; but also they can often have
mutual working relationships with companies who
sponsor students or who employ students post study.
Companies often already utilise the high end systems,
therefore it can be mutually beneficial if university students are exposed to the 3D CAD systems that they will
invariably use in industry. This package synergy with
industry also supports mature students returning to
part time education where a common interface of
known systems again is beneficial.
Also, with the constant advances in computing
hardware, software programming and the more affordable nature of such technological advances; there is
growing scope for the more basic entry level to the
mid-range 3D CAD systems to share a place in the market. Whilst the basic and mid-range systems will
probably never replace their higher end equivalents,
they do offer a viable and more affordable alternative for individual designers, small companies and
hobbyists.
No matter which 3D CAD system is used, to be
compatible with rapid prototyping the system needs
to be able to produce STL file formats (IGES for
DMLS), being the primary method of transferring the
digital data. The selection of various systems in table 1,
all offer this capability; but the different systems can
show some variation of data integrity. This variation
of STL file integrity can be due to the conversion
process, where the data export capabilities are generally affected by the complexity and robustness of
the original 3D model [2].
801
Table 1.
Examples of 3D CAD systems.
System
Class
Cost
Student*
Cost**
Alibre
Catia
Inventor (Lite)
Powershape
Pro-E
Rhino
Solidworks
VariCAD
L/M
M/H
M/H
L/M
M/H
Med
M/H
L/M
620
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
110
1,200
92
Free
80
195
110
80
1,500
N/A
995
4,000
445
*A cut down or student version is available.

** Cost is given in Euros and was approximate at time.
This paper does not advocate any one system

in particular. Generally end users will make their
own judgements of the different 3D CAD systems
available; based on their circumstances, requirements
and expectations.
1.2
Rapid prototyping processes
The term 3D printing has recently become more

noticeable in the general consumer marketplace with
the advent of 3D desktop printers that can be purchased
and utilised by individuals or companies; but the terminology has been around for several years now and it
basically encompasses rapid manufacturing processes
where an object is constructed layer by layer, by a process utilising digital data; another more generic term
is that of additive manufacturing [3].
Rapid prototyping has developed several ways of
being able to produce parts, processes such as:
Direct Metal Laser Sintering (DMLS), utilises

metallic powders, sintered by laser.
Electron Beam Melting (EBM)
Fused Deposition Modelling (FDM), a process that
utilises molten thermoplastic polymers [4].
Laminate Object Manufacturing (LOM), a less used
process where thin layers of materials are laminated
and then cut to a final shape [5].
Selective Laser Sintering (SLS) & Sterolithography
(SLA), use polymer powders, which like DMLS, are
sintered by laser to produce parts.
3D Printing, a process which uses inkjet thermal
printing to form parts.
With the increasing availability of desktop based 3D

printers, inkjet technology is being used to produce
parts in a similar fashion to SLA but offering amuch
cheaper and quicker alternative route to obtaining
parts. Whether through individuals and small companies with their own desktop print capability or
via larger commercial sites who offer the service;
demand will increase as more companies realise the
value of utilising conceptual models early in the design
phase [6]
When designing parts at the outset some thought
should be given to the fact that the use of rapid prototyping process can be limited in certain areas; such
as attainable size of parts, available choice of materials, dimensional constraints and feature complexities,
there may even be a need to factor in some additional
support structuring to allow for the process to be able
to assimilate the required structure. If the final part
is going to be assembled in any way for visualisation
purposes, some lateral thought may have to be given as
to how the rapid prototyped parts could be assembled
together.
The primary reason for using rapid prototyping is
to aid 3D design and whilst a complete part could
be made, it could also mean that a partial part or an
area of interest could be made instead. So long as the
part/s produced benefit the design process and aid the
conceptual visualisation.
2
POSSIBILITIES
2.1 Academia
Universities and colleges that offer technological or
design based courses can benefit immensely from
the use of rapid prototyping. Already many academic
courses are tailored for students that include rapid
prototyping; but those courses are often limited to
awareness of the processes and where their use fits
within design and industry. However, there is great
potential to develop and utilise rapid prototyping more
within academic curriculums giving students a much
better appreciation of their capabilities and their value
to design visualisation [7].
Some educational institutes will still require the
ability to fully offer the more conventional technical
resources for certain course requirements.However,
by using rapid prototyping more as an educational
medium both by students and staff then some institutes
could scale back or even remove the need for having
the more conventional technological resources. While
others could possibly extend the courses they are able
to offer, without excessive additional expense.
Whilst the use of the more commercial rapid prototyping bureaus may be cost prohibitive, due to
their evolved setups to service the needs of industry. It should be noted that some educational institutes
already have close working relationships with rapid
prototyping bureaus, either through such facilities
being set up and based on institute grounds or through
specific research functionality for the institute. It could
follow that one option would be to develop a network
of facilities shared between academic institutes; but
this may prove logistically awkward and would incur
possible time penalties in obtaining parts.
With the growth of desktop 3D printing capabilities there is now a greater potential for academic
institutes to either possess their own 3D printing capabilities or to use those of emerging bureaus setup to
service hobbyists and model makers. Whilst desktop 3D printing cannot offer the ultimate range of
possibilities found with the more conventional rapid
prototyping processes, it can offer a cost effective route
to 3D visualisation capability.
802
But why is 3D visualisation so important? For

young students who may not have been exposed
to industry, there is a risk that visualisation can
be misconstrued. It is easy with the 3D CAD systems to design conceptual parts. Where the ability to
work freely with zoomed in detail and to easily create
design features without prior or full understanding of
manufacturing methods, can often be misleading and
may lead to flawed designs.
Without the underpinning knowledge of different
manufacturing processes and what is feasible, then 3D
CAD systems can be likened to using calculators without understanding the mathematics. What is produced
virtually on the 3D CAD system is believable when
that may not be the case; but the fact that students
can design a part and then easily take their idea from
screen to reality using a rapid prototyping method has
great benefits.
Often when the actual part is held in the hands of
the creator the realisation of scale, proportion and feature complexity, when seen for the first time can often
be quite surprising. Occasionally what seemed to be a
good idea on the screen doesnt translate as easily to
real world parts or inversely what could have been hard
to visualise fully on screen can sometimes make more
sense when seen in reality. Either way, from the students perspective a quick and cost effective route has
been used to aid design visualisation [9].
Students could follow up actual rapid production of
a part by discussing part features, complexity of the
concept and potential manufacturing methods, materials and cost analysis taking the part forwards to real
world production. These are all valuable lessons that
can begin to give the student a more grounded migration to subsequent careers in engineering or other areas
of design related work from furniture to jewellery and
many others.
2.2
can be made where manufacturing setup costs need

to be adjusted later or where parts have to be scrapped
due to errors or feature problems only found when
manufacturing has initiated or during later assembly.
Like students who can suffer from lack of real world
perception, industry based designers and engineers,
even with their wealth of knowledge and experience,can suffer similar visualisation problems.The use
of rapid prototyped parts to validate designs could be
utilised much more, offering companies the scope to
instil better confidence in their proposed designs.
By utilising rapid prototyped parts people involved
in design reviews could have a better grasp of the
design intent through the improved visualisation; especially for people who dont work directly in design or
engineering, as creative ideas can come from many
people with a multitude of different backgrounds.
Greater use of rapid prototyping could potentially initiate ideas at earlier stages in the design process and
offers the ability to either reduce the timescales or to
add additional design iterations within given project
timelines, offering better design prospects and possible
quicker time to market.
Engineers could utilise rapid prototyped parts to
ascertain working relationships with other parts, OEM
(Original Equipment Manufactured) parts, to ensure
particular areas of assembly processes are possible and
to plan for subsequent production manufacture including the provision of jigs and fixtures, which can all lead
to planned and better managed timelines and release
dates.
A greater use of conceptual parts allows other
departments to consider their own areas of responsibility much earlier.When a design is close to final
release marketing departments could utilise rapid prototyped parts for customer faced meetings, trade shows
and increasing product awareness; Customers, executives, sales and service people, as well as designers,
Engineering
The use of rapid prototyping within engineering has

been around for many years, but as a design resource
is still not used as extensively as it could be. Most
design based engineering companies have been using
3D CAD systems since their outset and have grown
their design capabilities along with the software developments. By utilising 3D CAD systems companies are
able to ensure that a lot of design work is carried out
in advance, often incorporating better design reviews
before committing to actual part production.
This all fits well with the ethos of getting the design
right in the early stages to avoid later problems, such
as the subsequent overheads of engineering changes
and what could be unnecessary cost implications due
to upgrades, corrective measures or possible product
recalls [8].
Despite companies putting in a lot of early design
work, they can become so convinced by the 3D CAD
work they can also be prone to migrating straight
from the 3D CAD model to 2D drawings and subsequent manufacture. However, expensive mistakes
Figure 1. The Itrative Structure of the Design Process

(Roozenburg & Eekels 1995).
803
need to touch, hold and feel physical conceptual model

to better understand each design [9].
With the on-going advances in rapid prototyping
there is even scope for engineers to plan for the actual
incorporation of rapid prototyped parts within final
designs. Some of the polymer materials used and the
now growing portfolio of metals allow for final end
use of real parts. There are implications that need to
be considered, such as material strength properties,
size of parts, surface finish and additional post process work. But there is sufficient potential that the
utilisation of rapid prototyping for final end part use
in certain cases should not be discounted and should
actually be a regular design consideration in amongst
the more conventional methods [10].
CASE STUDY
This case study looked at utilising rapid prototyping in

order to obtain some complicated parts both quickly
and cost effectively, to be able to build a working prototype model. This allowed for a better understanding of
the various rapid prototyping processes available and
gave an insight into the use of rapid prototyping. It
could be used more frequently and more productively
for original design work and concept visualisation.
Looking into the various rapid prototyping bureaus
available it became apparent that three tiers have
evolved to offer rapid prototyping services:
Research and development based

Commercial industry based
Hobbyist and model making based
The research and development based services were

often linked directly with academic institutes or were
geared around specific processes such as micro
moulding. The companies set within the commercial industry market were more readily viable as a
potential resource but often commanded high prices
for their services. The ideal solution for the case study
requirements was the utilisation of a bureau within
the hobbyist and model making arena, where low volume, competitive costing and a quick turnaround were
offered.
The chosen company offered a variety of processes
with both polymer and metal materials being available.
The whole process was managed remotely through the
companys web site; from uploading the STL files
to placing the order, submitting payment through to
monitoring order progression. All of which took place
smoothly and without any glitches.Bureaus that offer
this ease of access, coupled with responsiveness and
low costs could realise a growth area in promoting and supporting the potential for original design
validation.
Rapid prototyping was used due to the complexity
of the parts required. The parts could have been manufactured using more conventional methods but this
would have incurred additional costs and would have
added extra time to the progression of the project.
Although not implemented in the case study, by

opting to use rapid prototyping also offered the potential to manufacture various iterations of the parts.
For instance a variety of impellers could have been
made with differing profiles, pitches or numbers of
blades, leading to a quicker optimisation of the design
supported by relevant testing. To obtain a similar variety using conventional manufacturing methods would
mean differing machine set ups and would be much
less cost and time effective.
3.1 Design images
The following images are used to convey the complexity of the parts that were manufactured for the case
study.
The case study looked at the concept of placing
small turbines in the wing mirror support posts of
electric vehicles [11]. The small turbine driven by the
transient air flow could produce small amounts of electrical power with minimum drag. The use of rapid
prototyped parts allowed for both quick and potentially
iterative design, coupled with a possible potential to
utilise production parts made via a rapid prototyping
process.
Although other more complex parts are often made
using rapid prototyping it should be remembered that
the case study was originated for educational reasons.
Within that context the part complexities are sufficient
that their manufacture using conventional methods
may have proved prohibitive within available educational resources and full 3D visualisation may not have
been possible.
Figure 3 showstwo fluted end caps both made from
PA2200 material (also known as Nylon 12). Other
materials were available for selection, but due to the
parts requiring a central hub supported by three radial
spokes integral to the outer case. PA2200 material was
the preferred choice as it offered reasonable strength
properties for the design. If the parts had been intended
Figure 2. Case Study Concept.
804
for visualisation only then other materials could have

been used; but the requirement to assemble a model
and conduct basic tests dictated the better material
choice.
The impeller in figure 4 was designed to be mounted
on a central shaft supported by jewelled bearings. Its
intended function to transfer transient air flow into
rotational movement of other components attached to
the shaft.
In a final design the required strength would most
probably be realised by utilising more conventional
engineering plastics; but for the purpose of the case
study a rapid prototyping material and process was
more than adequate to be able to offer the basic level
of functionality required for the conceptual model.
The full suite of parts manufactured using the rapid

prototype process can be seen in figure 5 showing both
the range of complexity and sizes. At the time of manufacture the overall incurred costs were minimal, being
no more than the average cost of an engineering text
book.
Generally it could be said that the more complex
the parts design then the greater the % variance on cost
between conventional methods and rapid prototyping
methods of manufacture. Taking the impellor design
and looking at comparable costs for a part made using
more conventional methods, then a variance of about
+50% on cost would be realised. When working with
original designs, the ability to be able to manufacture,
visualise and manipulate various parts or conceptual
ideas in such a cost effective way can only be conducive
to the design process.
The close up in figure 4 shows in slightly better
detail the type of surface finish realised from a basic
level of rapid prototyping, it is possible during its manufacture to carry out additional finishing processes and
achieve better final surfaces, but that level of extra
work and its resultant additional cost implications was
deemed as unnecessary for the case study parts.
Figure 3. Fluted End Caps.
Figure 4. Impeller.
Figure 5. Full Set of Rapid Prototyped Parts
CONCLUSION
The part and assembly knowledge gained from using

rapid prototyping so early in the design process is critical for a quick and progressive flow of the design
cycle:
Invariably a design will go through several cycles
before being finally approved and using rapid prototyping in the case study supported the fact that early
visualisation is key to either reducing the number of
iterative steps or reducing the cyclical timeline of the
required steps [12].
The use of rapid prototyping in the many areas
of design and engineering both within academia and
industry is one that not only offers a cost effective and
quick route to part production; it is also a process that
is critical to key understanding of original concepts
and as such should feature more as an integral part of
the design process.
Figure 6. Close-up Detail.
805
BIBLIOGRAPHY
Kalpakjian S, Schmid SR, 5th Edition 2006, Manufacturing Engineering and Technology, Pearson
Education Inc.
REFERENCES
[1] Wolfe Stephen L, (2009), 9 Criteria for Choosing a
3D CAD System, Solidworks, White Paper, p3 para 4
[2] Chowdary Boppana V, Sahatoo Divesh R, Bhatti
Raj, (2007), Some STL File Generation Issues
in Rapid Prototyping, 5th Latin American and
Caribbean Conference for Engineering and Technology, 4B.12
[3] Sequin Carlo H, Rapid Prototyping a 3D Visualisation Tool Invades New Domains, Berkeley University.
[4] Bailey Mike, (2005), Layered Manufacturing for Scientific Visualisation, Communications of the ACM,
Vol 48 No 6 Pages 4445.
[5] Rapid Prototyping: LOM. http://www.efunda.com/
processes/rapid_prototyping/lom.cfm (accessed May
8th, 2011).
[6] Boehm Trevor, Carnasciali Maria-Isabel, Douglas
M Elizabeth, Fernandez Marco Gero, Williams
Christopher, (2004), Rapid Prototyping Industry
Analysis, Georgia Institute of Technology.
[7] Johnson Wayne M, Coates Cameron W, Hager Patrick,

Stevens Nyrell, (2009), Employing Rapid Prototyping
in a First Year Engineering Graphics Course, ASEE
Southeast Section Conference.
[8] Nichita Gabriela G, Using Rapid Prototyping Technologies in Concurrent Engineering Approach, Faculty of Management and Technological Engineering,
University of Oradea, Romania.
[9] Wolfe Stephen L, (2010), Physical and Digital Prototyping Belong Together, Z Corporation, White
Paper.
[10] Hague Richard, Tuck Chris, Raymond George, Rapid
Manufacturings role in Design Optimisation and Customisation, RMRG Loughborough University UK.
[11] Harris Neil G, (2011), Miniature Turbine Generator,
Paten Application, Application Number 1104099.5
[12] Roozenburg, N and Eekels, J (1995) Product Design:
Fundamentals and Methods, Chichester: Wiley.
RP WEBSITES ACCESSED (2011)

www.3Dcreationlab.co.uk
www.3TRPD.co.uk
www.CRDM.co.uk
www.prototypezone.com
www.shapeways.com
806

Integrated strategy for sustainable product development

A.L. Santos, H. Almeida, H. Brtolo & P. Brtolo
ABSTRACT: The implementation of sustainable methodologies can contribute for the implementation of more
ecological manufacturing practices. It is essential to develop systems and processes aiming at product reuse and
recycling. A computational tool, based on eco-design principles, was developed supporting more sustainable
decisions for the conceptual design phase of a project. This tool allows designers to investigate the best product
life-cycle solutions, as well to integrate the products physical data with the CAD model. This work presents the
life cycle assessment of a product through the developed system, including its dismantle, the 3D digitalization
using a reverse engineering technology and the analysis of the redesigned model.
1
INTRODUCTION
Growing global concerns on environmental issues

have been fostering sustainable manufacturing
approaches towards a more sustainable development.
Companies are becoming more concerned with environmental and social issues, to develop methodologies
and solutions minimizing environmental impacts as
a key element for a long-term product innovation
strategy. The development of competitive new products is a prerequisite for its success. Industrialisation,
open markets, higher (quality) customer requirements
and an increasing competitiveness created a serious
demand for a more structured process towards product
innovation.
It is fundamental to apply more eco-design principles in the development of a product, through its
production and consumption. To reduce the environmental problems associated with the production and
consumption of products, it should be considered the
environmental impact in the product design or redesign
process using, for instance, a reverse engineering
technology (Santos et al., 2011).
In this research work, a redesign strategy for the
environmental impact reduction of a product is presented.
2
D4E COMPUTATIONAL TOOL
A computational tool called Design for Environment

(D4E) was developed and presented in a previous
work (Santos, 2011; Santos et al., 2011). This tool,
based on eco-design principles, can be easily adapted
to all product design practices within an industrial
environment. The D4E initial window is shown in
Figure 1.
D4E tool support design decisions at early product design stages, and assist designers/users to obtain
more sustainable solutions aimed at minimizing the
Figure 1. D4E software starting window.
environmental impact of products during its whole life

cycle. It allows investigating the products life cycle,
as well to integrate the products physical data with the
CAD model (Santos, 2011; Santos et al., 2011).
The D4E system uses a Life Cycle Assessment
(LCA) approach to evaluate the environmental impact
of a product, supporting design decision-making as it
enables to quantify all inputs and outputs in a qualitative and quantitative way at all life cycle stages. These
life-cycle phases can be described as pre-production,
production, distribution, consumption and disposal,
according to Finnvedena et al. (2009).
This tool uses eco-indicators to quantify the environmental impact for each LCA product. These ecoindicators allow measuring the environmental impact
of a material or process throughout its life cycle, considering the materials used along the manufacturing
process, its energy consumption, transport and final
destination (Goedkoop & Spriensama, 2001).
Figure 2 shows D4E flowchart. It explains each step
the user needs to follow in order to obtain a products
life cycle assessment.
807
Figure 3. Hairdryer selected for the case study.
Table 1. The external product components and its mass.

Components
Mass (g)
Ruber Casing
Plastic housing
Rear Casing
Narrow opening concentrator
Switch
21,003
83,211
19,938
17,458
6,807
3.1 Product
The original product is an electrical device used to dry
and style hair. It uses an electric fan to blow air across
a heating coil, as the air passes through the dryer it
heats up. When the warm air reaches wet hair it helps
evaporate the water.
The materials and design of this product main target is to capture the consumers attention and increase
market share. The hair dryer selected is available in the
market, has a simple form and is very functional.
Figure 2. Flowchart of D4E tool.
2.1
Reverse engineering
Reverse engineering is one of the major technologies

contributing to a rapid technology change. Reverse
engineering initiates the redesign process, where a
product is observed, disassembled, analyzed, tested
and documented regarding its functionality, form,
physical principles, manufacturability and assemblage
(Otto & Wood 1998). The design process starts with
an abstract description of a target and via a succession
of refinements, produces a design that can be directly
implemented. Reverse engineering is the process of
investigating a system to identify its components and
relationships, to create a representation of the system
in another form or at a higher level of abstraction,
so that it can be considered the inverse of the design
process (Sobh et al., 1999).
This technology enables designers to make productmodels. In this work, it was necessary to digitize
a product using a reverse engineering technology in
order to obtain its geometric information (3D CAD
model), since there was no drawings, documentation
or computer model for this product.
CASE STUDY
A case study was performed in order to test the D4E

tool. An ordinary hairdryer was the selected product,
which can be observed in Figure 3.
3.2 D4E Life Cycle Assessment

The LCA methodology was performed with the D4E
tool in order to calculate the environmental impacts
associated to the hair dryer for each stages of its life
cycle (production, use and disposal).
The first phase comprised the actuals product
experience in terms of both function and form, its
full disassembly, design for manufacturing analysis
and functional analysis. This step corresponds to the
production stage.
For this analysis, it was necessary to identify the volume data for each component, which was determined
by either the 3D product analysis or the measure of
each component mass using an analytical balance. The
mass obtained for the external product components,
relatively to the original model product, is illustrated
in Table 1.
A visual inspection was carried out for each component, in order to identify the materials of which it
is made from. To differentiate the types of materials
that constitutes each product components is essential
to evaluate each life cycle, as each material and manufacturing process is associated with an eco-indicator
value. This products main material is polypropylene,
as shown in Figure 4.
808
Figure 6. Environment impact value window.
Figure 4. Percentage of materials for the hair dryer.
Table 3.
Products use data and distribution.
Transport of the goods from

Republic of China
Oceanic liner (km)
Distribution in Portugal
28Ton truck (km)
Useful lifetime (Years)
Power (W)
Voltage (V)
Usage Frequency (per week)
Usage Time (minutes)
Life cycle stages
Environmental impacts
(mPt)
Life Cycle
Production
Use
Disposal
10639,909
265,582
10413,580
39,253
The LCA products results on its environmental

impacts are also illustrated in Table 3.
This products use stage has excessive environmental impacts, so it was decided to decrease this value
through an optimized product design. The design of
external product components has a strong influence on
the value of the environment impact in the production
stage. All product components have different forms,
which aggravates the environmental impact during
the production phase as each component requires a
different mould.
It was assumed that it was possible to minimize this
value with the redesign of the external product component, so it was necessary to digitize the product with
a reverse engineering technology in order to obtain its
3D CAD model.
Figure 5. Products production phase (D4E tool).

Table 2.
Environmental Impacts.
18500
297.5
8
1000
220230
2
5
3.3
A representation of the products Production phase
using the D4E tool is shown in Figure 5.
For the use stage, in particular the transport and
consume, it was known that this product was produced
by the Republic of China and subsequently delivered
to Portugal, so it was necessary to introduce the estimated information regarding travel distances from the
products origin to distribution point, as well its usage
during its whole life cycle, as shown in Table 2.
The disposal stage refers to the final destination of
each product component. For this step, it was considered that each product component was recycled.
Afterwards, all data was inserted in the D4E system.
The hairdryers environmental impacts are displayed in Figure 6.
Hair dryer 3D Digitization
The acquisition of the products geometric information

is the first step and its most important work for reverse
engineering (Hsiao and Chuang, 2003).
The process used for the product digitization is
then described to obtain the 3D CAD model of the
hairdryer, as it is essential to support the redesign process model. A system called Steinbichler COMET 5
was used, as shown in Figure 7, for the scanning of this
3D CAD model.
The Steinbichler COMET 5 is a structured light
3D scanning system combining a unique performance
with perfect handling, making it the ideal solution for a
demanding industrial quality control requirement and
product development process. This system uses a single camera and a projector alongside with a patented
809
Figure 9. External product components: a) rear rubber, b)

plastic housing, c) rear casing, d) narrow opening concentrator, e) switch.
Figure 7. Digitalization
COMET 5.
with
systems
Steinbichler
Table 5. Weight of redesign product components and weight
reduction.
Components
Mass
redesign
model (g)
Mass
reduction (%)
a)
b)
c)
d)
e)
00,0
61,40
5,95
00,0
6,81
100,00
26,22
70,14
100,00
00,0
Figure 8. Applying whitening spray onto the product surface.

Table 4.
Hairdryers configuration.
Camera Resolution (pixel)

Measuring volume (mm3 ) 400
3D Point Distance (m) 400
Shortest Measuring time (s)
Automatic Object Positioning
Sensor Positioning
PC
1360 1024
480 360 250
350
0,6
Rotation Table
Tripod or sensor stand
with manual turn and
tilt axis
1 Intel Xeon Quadcore,
12 GB upgrade up to
24 GB
external light source. The combination of these and

other innovative features gives the system an high
dimensional stability, accuracy and repeatability (Silva
et al., 2011).
This system requires to follow a few steps to obtain
a satisfactory 3D product digitization: the selection of
lens, framing the product in the rotating platform, a
good selection of lighting, a previous use of whitening
spray to prepare shiny, translucent or very dark objects
for a 3D scan, as shown in Figure 8.
The configuration of the Steinbichler COMET 5
system for the 3D hairdryer digitizing is illustrated in
Table 4.
After the building of the 3D model, it was used to

generate new shapes helping the designer to develop
new creative design ideas.
This stage included the development and building of design models, the model calibration and
experimentation.
3.4 Product redesign
In this stage, the product redesign was initiated using
the reverse engineering results. This redesign was
performed to reduce the value of the products environmental impact, as stated before. Using the 3D CAD
model, it was possible to redesign the external product
components. The 3D CAD model of external product
components is represented in Figure 9.
Firstly, the functional description of the product was
updated by comparing the predicted function structure
to the actual.
It was decided to reduce a number of external components for the product redesign. The plastic housing
was jointed with the narrow opening concentrator into
a single object and the rubber casing was eliminated.
The thickness of the components was reduced. In general, the morphological matrix was simplified. The
same materials were used for its production.
Table 5 shows the mass weight of the redesign product components and the value of respective weight
810
After running the LCA of the original product and

its redesign model with the D4E tool, the redesign
model has less environmental impact than the original product, so the new model is more environmental
friendly.
At the production and disposal stages, the environmental impact value is minor because the number of
external components and the thickness was reduced.
The D4E computational tool enables to obtain data on

the products environmental impact at every stage of
its life cycle. LCA results obtained with the hairdryer
can be considered ecological data once they can contribute to optimize the design product and reduce the
product environmental impact during its life cycle.
D4E system allows a fast and efficient way to analyse
the life cycle of real products, as well virtual models,
generating solutions for the production, use and final
disposal phases. This eco-design tool intends to influence decision-making by creating products able to be
recycled and reused, thereby contributing to reducing
the impacts and costs associated at all stages of product
development.
Reverse engineering associated with sustainable
tools can provide new opportunities to assess the lifecycle of a product or component with no drawings
or CAD model. This system can be a step further
towards a more sustainable manufacturing product
development, facilitating the environmental analysis,
optimization and comparison of the product life cycle.
The proposed tool is an effective and structured
system, useful for eco-design and sustainable product
development, allowing not just experts but also others
to swiftly understand the environmental impact of their
decisions. The use of this system will allow companies to take a step forward towards a more sustainable
economic development. Once business is aware of the
importance of implementing sustainable policies, this
type of tool can give an important contribution for its
wide use in manufacturing.
Figure 10. Redesign model.
Figure 11. Overlap of the 3D original model of product with

its redesign model.
Table 6.
Environmental impact values.
Environmental
impacts
Original
product
(g)
Redesign
model
(g)
Mass
reduction
(%)
Life Cycle
Production
Use
Disposal
10639,909
265,582
10413,580
39,253
10624,841
239,331
10413,590
28,069
0,14
9,88
00,0
28,49
CONCLUSIONS
REFERENCES
reduction components. Figure 10 represents the 3D
CAD model of the redesign hairdryer.
Figure 11 showed the overlap of the 3D original
model of product with its redesign model.
After, LCA methodology with the D4E tool was
applied in order to calculate to environmental impacts
associated to redesign hair dryer in every stages of its
life (production, use and disposal). Were used the same
parameters in the use stage. It was supposed that each
product component was recycled.
Table 6 makes a comparison between the environmental impact values of the original product and
the redesign model, as well the percentage of mass
reduction obtained.
Finnvedena, G., Hauschildb, M., T. Ekvallc et al., 2009.

Recent developments en Life Cycle. Journal of Environmental Management, 92 (1), 121.
Goedkoop, M. & Spriensama, R, 2001 The eco-indicator
99: a damage oriented method for Life Cycle Impact
Assessment. Methodology Report: Neetherlands.
Hsiao, S.W. & Chuang, J.C., 2003. A reverse engineering
based approach for product form design. Design Studies,
24, 155171.
Otto, K. & Wood, K., 1998. Product Evolution: A Reverse
Engineering and Redesign Methodology. Research in
Engineering Design, 10, 226243.
Santos, A. L., 2011. Eco-design Computational Tool, based
on Eco-Indicators. Leiria: Unpublished MSc Thesis Polytechnic Polytechnic Institute of Leiria, Portugal.
811
Santos, A. L., Almeida, H. Brtolo, H. & Brtolo, P.,

2011. A ecodesign tool for sustainable manufacturing. In
H. Brtolo et al (Eds), 1st International Conference on
Sustainable Intelligent Manufacturing, Leiria, 29June
1July, IST Press: Portugal.
Silva, C., Domingues, F., Brtolo, P., Alves, N., & Almeida,
H., 2011. Reconstruo e Controlo Dimensional de
Formas 3D atravs de Luz Branca Estruturada. O Molde,

Associao Nacional da Indstria de Moldes, 22(88),
2832.
Sobh, M. T., Elleithy, M. K & Patel, S., 2007. Reverse Engineering of VLSI: A Roadmap. Journal of Engineering and
Applied Sciences, 2(2), 290298.
812

3D scanning and digital manufacturing technologies applied in

research projects in archaeology
J.R.L. Santos
Instituto Nacional de Tecnologia, Ministrio da Cincia e Tecnologia
Diviso de Desenho Industrial, Laboratrio de Modelos Tridimensionais, Rio de Janeiro, Brazil
Pontifcia Universidade Catlica do Rio de Janeiro, PUC Rio
Departamento de Artes e Design, Laboratrio de Modelagem e Simulao 3D, Rio de Janeiro, Brazil
S.A.K. Azevedo & S. Belmonte

Museu Nacional, Universidade Federal do Rio de Janeiro
Laboratrio de Processamento de Imagens Digitais, Rio de Janeiro, Brazil
A.B. Junior
Museu Nacional, Universidade Federal do Rio de Janeiro
Setor de Arqueologia, Rio de Janeiro, Brazil
R.C. Fontes
Instituto Nacional de Tecnologia, Ministrio da Cincia e Tecnologia
Diviso de Desenho Industrial, Laboratrio de Modelos Tridimensionais, Rio de Janeiro, Brazil
ABSTRACT: The project draws on modern techniques of 3D laser scanning and non-invasive image medical
technologies as CT and micro CT scanners in order to create replicas of rare and fragile parts related to collec-tions
as in archaeology and paleontology. This method allows the interchange of 3D files between museums, helping
researchers to manipulate structures as well the creation of physical models for public didactic exhibi-tions. The
3D virtual files obtained are transformed into physical accurate replicas through the use of additive manufacturing
technologies. The work involves a partnership between the Museu Nacional (National Museum) which is part
of the Federal University of Rio de Janeiro (UFRJ), and has the most important collection of natural history in
South America and the Instituto Nacional de Tecnologia (National Institute of Technology) equipped with a
cutting edge 3D modeling laboratory.
INTRODUCTION
The concept of 3D Reverse Engineering has its roots

in ancient cultures as the ancient Egyptian and Greek
civilizations, especially on sculpture works, through
conventional and mechanical devices, methods that
can still be seen in some contexts where it is necessary to acquire dimensions and geometries related
to the surface of objects by using non expensive
solutions.
According to the Egyptologist Nadia Tomoun, the
ancient sculptor began by transferring mechanically
a large number of predetermined measurement points
(Fig. 1) to the stone. In doing this, he drove the chisel
straight into the measurement points, instead of removing thin layers of stone from all around the statue, as
he had done previously. In this process of pointing,
a large number of points are used to transfer the exact
shape and measurements from the scale model to the
large block of material, thus enabling a fairly exact
copy of the original.
Renaissance sculptors also used the same principle,

called pointing up, meaning to recreate the sculpture point by point. Many sculptors used it to represent
their initial ideas in drawings or sketch models for their
patrons approval.
To work up large-scale sculptures from smaller
models, they used the pointing method, to enlarge
the form from reduced clay or plaster models.
This process required a mechanical device such as
a plumb line, drill or callipers to transfer relative distances between high points on the original model to
the corresponding points on the block of material.
The given verticals and the horizontal distances are
simply copied at the exact points from that chosen
exterior reference, and the entry is drilled repeatedly until reaching the correct depth according to the
model.
In the Rodin plaster model of 1913 (Fig. 2) we
see that the original bozzetto (hand made physical
model in general created in reduced scale) served as
a three-dimensional pattern, exhibiting many dots on
813
Figure 1. Ancient Egyptian scale model of a Sphinx (about

300 BC) where it is possible to see the squared lines done
in order to allow the enlargement transferring method. Altes
Museum Egyptian collection Berlin, Germany. Photo:
Jorge Lopes.
Figure 3. Technician scanning a rare funerary urn from the

Museu Nacional Amazon tribes collection using a portable
Viuscan Colour 3D laser scanner. Photo: Jorge Lopes.
Figure 2. Auguste Rodin plaster model with transferring

points (maquette avec points de repere pour la traducion
en marble) of Lady Sackville, 1913. The Rodin Museum.
Photo: Jorge Lopes.
the surface, which functioned as spatial coordinates for

the final enlargement. This method was widely used
up to the 19th century to copy spatial geometry from
physical scale models; the technique was abandoned,
however, by the vast majority of 20th century sculptors, such as Constantin Brancusi and Henry Moore,
who preferred to carve the raw material in-stead of
translating the idea from scale models by the use of
transferral devices.
Nowadays, new digital technologies for 3D Reverse
Engineering as 3D laser scanners, white light 3D scanners (Figs. 3 and 4) and micro CT scanners are helping
researchers in different areas of science to visualize
and analyze external and internal structures of rare
and fragile objects.
The 3D scanning project of the Museu Nacional1 ,
sponsored by the Rio de Janeiro research foundation
1
The Museu Nacional , is the first scientific institution

in Brazil, created by the Portuguese emperor in 1818 and
still one of the most important scientific institutions in
Latin America, internationally recognized by the important
collections
Figure 4. Ancient Egyptian royal stela being scanned using

a portable ARTEC white light 3D scanner. Photo: Jorge
Lopes.
FAPERJ and technologically supported by the Instituto Nacional de Tecnologia also in Rio de Janeiro
is focused on the 3D virtual geometry of important
pieces of the museum collection.
The 3D file capture is done through the use of
only non contact/non invasive image technologies,
including also equipments as CT and Micro CT scanners adopted when internal visualization of structures
turned necessary.
814
Figure 5. Ancient Roman terracotta lamp and physical

replica built in plaster on the Z Corp additive manufacturing equipment. Photo: Jorge Lopes.
Figure 7. 3D virtual model of the funerary urn.
Figure 8. 3D Graphic detail of the funerary urn.
Figure 6. Milling CNC machine during the construction

process of a stone replica of the detail of an ancient Egyptian
royal stela. Photo: Jorge Lopes.
A multidisciplinary team consisted of paleontologists, archaeologists, biologists, geologists, designers

and engineers are responsible for the researches and
selection of pieces to be analyzed and eventually also
replicated through digital manufacturing technologies
as additive technologies and CNC milling machines
(Figs. 5 and 6) for different purposes as itinerant
exhibitions and research exchange among museums
around the world.
RESULTS
The main objective related to this Project is the

implantation of a methodology do obtain the geometries of the most important pieces of the collection
(Figures 714), that varies from closed coffins from
the Egyptian Collection, marble and terracotta sculptures from the Greek and Roman collection, stone
Figure 9. 3D virtual model of Ancient funerary urn from

the Museu Nacional Amazon ancient tribes collection.
815
Figure 10. Detail of the reflective points positioned on the

restored parts of the ancient funerary urn. Photo: Jorge Lopes.
Figure 13. 3D virtual model of an ancient Egyptian royal

stela scanned on the ARTEC white light 3D scanner.
Figure 11. 3D shaded model to highlight the geometric

visualization of the urn patterns.
Figure 12. 3D virtual model of an Ancient Roman terracotta

lamp scanned on Next Engine 3D laser scanner.
Figure 14. Replica made of stone of the detail of an ancient

Egyptian royal stela. Photo: Jorge Lopes.
816
enclosed skeleton parts from archaeology finds, vertebrate fossils (dinosaurs) of the paleontological collection to meteors from the geological collection.
3
CONCLUSION
The technical processes to obtain the 3D files is done

through the use of 6 different la-ser and white light
scanners that belongs to the Instituto Nacional de
Tecnologia, part of the Brazilian Ministry of Science
and Technology and the Museu Nacional which is
part of the Universidade Federal do Rio de Janeiro.
CT scanners are used through a partnership with a
medical image clinic when necessary the internal
visualization of parts related to the collection as the
human bodies of the Egyptian and Pre Colombian collection and vertebrate extinct fossils. The accuracy
and quality of the digital data obtained are related
to the equipment selected which varies according to
technical aspects as the portability since the objects
are rare and the positioning of the objects can be
difficult to reach and manipulate which can also
reflect on the quality of the physical model to be
replicated.
REFERENCES
Ferguson, E.S. (ed.) 1992. Engineering and the minds eye.
Cambridge, MA: MIT Press.
Manchanda, C. (ed.) 2006. Models and prototypes. St. Louis,
MO: Mildred Lane Kemper Art Museum, Washington
University in St Louis.
Tomoum, N. 2003. The sculptors models of the late and
Ptolemaic periods: a study of the type and function
of a group of ancient Egyptian artefacts. PhD Thesis,
Cairo: National Center for Documentation of Cultural and
Natural Heritage and The Supreme Council of Antiquities.
Trusted, M. (ed.) 2007. The making of sculpture: the materials and techniques of European sculpture. London: V&A
Publications.
817

Fabrication of a soft morphing structure using a Shape Memory

Alloy (SMA) wire/polymer skeleton composite
J.S. Kim & S.H. Ahn
School of Mechanical and Aerospace Engineering, Seoul National University, Korea
ABSTRACT: The most common method of fabricating a smart structure using a shape memory alloy (SMA) is
to create an SMA-embedded structure. However, if the structure is too thick, actuation is decreased significantly.
Hence, SMA-embedded structures and robots are usually thin, leaving no space for additional parts. In this
research, an SMA-embedded soft morphing structure with large thickness and deformation was developed. A
skeletal structure and hinges were used to increase the overall actuation by maximizing the actuation in a specific
area. A prototype of the design was fabricated via rapid prototyping (RP) and casting. The performance of the
prototype was evaluated, and large deformation and actuation force were demonstrated. A cell phone robot was
suggested as an application, and the resulting fabricated prototype exhibited crawling actuation.
INTRODUCTION
Shape memory alloys (SMAs) are materials that have

the shape memory effect (SME), caused by martensitic transformation. Because of this effect, a deformed
SMA can recover its original shape when heat is
applied.
Nitinol, developed by the U.S. Navy in the 1960s,
is one of the most popular SMAs because of its stability, repeatability, pseudo-elasticity, and SME. The
SME and pseudo-elasticity allow an SMA to be used
as an actuator. In particular, since actuators made from
SMAs can be any desired shape, they can be used to
fabricate actuators with a simple structure and much
smaller volume and mass than classical actuators.
Compared to other smart materials used as actuators,
such as IPMC or PZT, SMAs are the nearest to practical usage, thanks to their high energy density and
good repeatability. Moreover, other functions, such
as precision control and vibration suppression, are
also possible, thereby facilitating research in various
fields.
An SMA-embedded composite was invented by
Rogers to suppress vibration. Rogers attempted to
apply additional stress to a structure by embedding
SMA wire in a composite and actuating the wire.
SMA-embedded composites have attracted the interest of many researchers because of their large damping
capacity and large deformation recovery.
SMAs are also used in small-scale robots with a
simple structure. SMA wire requires very little space
and no additional parts (such as gears), since linear
actuation is induced by direct application of heat or
electrical power.
However, to produce large deformation, SMAembedded smart structures and small-scale robots
must be thin, since the force generated by an SMA wire
is small compared to that of classical actuators. Thus,
space for additional parts (such as batteries or control circuits) is generally lacking, and these parts must
either be attached externally or omitted altogether.
Moreover, the assembly process for SMA-based smallscale robots is difficult, and their performance is
fraught with uncertainty.
In this study, an SMA-embedded smart structure
with a polymer skeletal structure was proposed. The
skeletal structure ensured the availability of space for
additional parts. To minimize the mass, volume, and
thickness of the skeletal structure, fused deposition
method (FDM) rapid prototyping (RP) was employed
for its fabrication. To increase the overall actuation, the
deformation was focused in specific areas by hinges,
which were designed with a low degree of stiffness.
Bending actuation was induced, and the amount of
actuation increased, by applying eccentricity.
A manufacturing process consisting of RP and casting was proposed. The skeletal structure was created
via RP. The casting process allowed the empty space
to be filled with a soft material that bonded with the
SMA wires and skeletal structure.
As an application, a cell-phone robot was constructed. Figure 1 shows the basic mechanism of the
robot. The lower structure was composed of polydimethylsiloxane (PDMS) (with a large coefficient
of friction) and acrylonitrile butadiene styrene (ABS)
(with a small coefficient of friction) to change the
bending actuation into a crawling motion by controlling the coefficient of friction.
819
Figure 2. Design of the skeletal structure and hinges.
Figure 1. Basic mechanism of the cell-phone robot.
2
2.1
DESIGN AND MANUFACTURE

Design of the smart structure
The smart structure consisted of two parts: the SMA

wire actuator and the actual structure. Thus, there were
two main factors that determined the amount of deformation: the design of the structure and the positioning
of the SMA wires.
2.1.1 Skeletal structure and hinges
The smart structure was composed of two kinds of
material, one stiff and the other soft. There was no
empty space and the volume of the SMA wires could be
neglected. Hence, the stiffness of the overall structure
was determined by the shape of the skeletal structure.
To increase the actuation, eccentricity was applied.
This was accomplished by fabricating the upper and
lower sides with differing degrees of stiffness, thereby
inducing bending actuation. In this way, a small contraction of the SMA could cause a large bending
actuation of the structure. Factors determining the
amount of deformation were the stiffness of each side
and the gaps between the upper side, lower side and
SMA wires.
When designing the structure, space for an embedded part had to be considered. We assumed that a
battery would be installed inside the structure. To
decrease the mass and stiffness of the overall structure and skeletal structure, a thin filament had to be
used.
However, even though the stiffness of the skeleton
was small, if the embedded part had been too stiff,
deformation of the stiff area would have been difficult
to achieve. Moreover, if the depth of the embedded
part and the gap between the SMA and the upper side
had been too thick, the bending deformation would
have been small. Hinges were used to maximize the
deformation.
The skeletal structure was designed to be thin
compared to the embedded part. Thus, hinges were
installed in the skeletal structure. Each filament was
located in the lowest area, and the SMA wires were
Figure 3. Detailed design of the prototype showing SMA

placement.
Table 1.
Specifications of the prototype
Properties
Value
Size
120 mm 65 mm 3.6 mm
(L W D)
37 mm
4.5 mm
10
Gap between hinges

Length of hinges
Number of SMA wires
placed just below the filaments. Figure 2 shows the

design of the skeletal structure and hinges.
2.1.2 Detailed design of the prototype
Figure 3 shows the detailed design of the prototype
and Table 1 lists the design specifications. The size of
the prototype was chosen to match the size of a cell
phone.
The depth of the prototype and the gap between the
hinges were based on the size of a Li-po battery. A
Li-po battery was selected because an SMA requires a
considerable amount of current. The size of the battery
was 31 mm 29 mm 2.9 mm (L W D), and
the capacity and voltage were 160 mAh and 3.7 V,
respectively.
2.2 Materials
To fabricate the smart structure, SMA wire and two
other materials were employed.
The SMA used in this research was FLEXINOL (Ni: 55 wt%, Ti: 45 wt%, Dynalloy, US).
The transformation temperature was measured using
820
Table 2.
Specifications of the SMA wire
Properties
Value
Diameter
Pre-strain
Recommended current
Density
Resistance
Specific heat
300 m
4%
1.5 A
6.45 g/cm3
12.2 /m
0.2 cal/g* C
Figure 4. Fabricated prototype.
Table 3.
Mechanical property of materials
Properties
ABS
PDMS
Density
Tensile modulus
Yield strength
Working temperature
1.05 g/cm3
1.627 GPa
22.0 MPa
2080 C
1.05 g/cm3
0.0018 GPa
6.2 MPa
20200 C
a differential scanning calorimeter (DSC). Measured transformation temperature is As = 37.25 C,

Af =90.16 C, Ms = 60.43 C, Mf = 8.23 C. The specifications of the SMA wire are listed in Table 2.
The material used for the skeleton was ABS, which
is a stiff, thermoplastic polymer that is often used in
structures. The ABS used in this research was P400
(Stratasys, US), a model material for the FDM rapid
prototyping machine (SST768, Stratasys, US).
The material used for the matrix was PDMS, which
is flexible and has good thermal resistance. The PDMS
used in this research was Sylgard 184 elastomer (Dow
Corning, Korea).
The mechanical properties of these materials are
listed in Table 3
2.3
Manufacturing process
First, the rapid prototyping machine (SST 768, Stratasys, US) was used to fabricate the skeletal structure.
The skeletal structure was placed in a prefabricated
mold, and the SMA wires were inserted in the planned
positions. The wires were then tightened and fixed in
the mold. Resin (PDMS) was then poured into the
mold and allowed to harden. The curing conditions
were 100 C for 3 h. Once the PDMS had cured, the
fabricated smart structure was removed from the mold.
3
PERFORMANCE EVALUATION
The fabricated prototype is shown in Figure 4. To evaluate the performance of the prototype, the deformation
and actuating force were measured.
3.1
Deformation
In the experiment to measure the deformation of the

smart structure, a laser displacement sensor and digital
camera were used. A current of 1.2 A was applied for
Figure 5. Results of the experiment to measure deformation.
8 sec. The experimental setup and results are shown in

Figure 5.
3.2 Actuating force

To measure the actuating force of the smart structure, a
force sensor (dynamometer) was used. The experiment
was performed with the tip of the specimen in the initial
position, deflected by 2 mm, and deflected by 5 mm.
A current of 1.2 A was applied for 8 sec. A support
block was used to control the distance through which
the specimen could actuate freely. The experimental
setup for measuring the actuating force is shown in
Figure 6.
4 APPLICATION: CELL PHONE ROBOT

4.1 Concept and basic design of application
A practical application of the smart structure developed in this research involved a cell phone case type
robot.
821
Figure 7. Motion of the cell phone case type robot.
Figure 6. Experiment to measure actuating force.
Cell phones are extremely popular electrical devices

with the highest communication/process performance.
If a cell phone with mobility were developed, the ripple
effect would be large, because such a cell phone could
be a small, high-performance robot.
When the SMA wires were contracted and released,
the coefficients of friction of the front and bottom parts
of the robot converted the bending motion of the structure into a crawling motion of the robot. The lower
structure was designed to change the coefficient of
friction of each side.
The size was set at 120 mm 65 mm 6.6 mm
(L W D). The depth of the lower structure was
3 mm, and 18 SMA wires were used.
The lower structure consisted of ABS (with a low
coefficient of friction) and PDMS (with a large coefficient of friction), and additional SMA wires and an
additional hinge were installed. Using this mechanism,
the area of the lower surface in contact with the surface
on which the robot rested could be changed, and the
frictional force between the surface and the front and
back parts of the structure could be controlled.
4.2 Actuation of prototype

The fabricated prototype exhibited a crawling action.
The distance covered in one stroke was 5 mm, and
about 18 sec were required for one stroke. The linear
velocity of the prototype was 0.3 mm/sec. The motion
of the prototype is shown in Figure 7.
CONCLUSION
To induce large deformation in a thick smart structure,

a structure consisting of a skeleton and hinges was proposed and tested. The resulting smart structure design
was fabricated via a rapid prototyping and casting process. The performance of the prototype was evaluated,
and visible actuation and a large actuation force were
measured. The cell phone robot was suggested as an
application, and the fabricated prototype of the device
exhibited a perceptible crawling motion.
ACKNOWLEDGEMENT
This research was supported by Basic Science
Research Program through the National Research
Foundation of Korea (NRF) funded by the Ministry of
Education, Science andTechnology (No.20110001572).
REFERENCES
J. G. Boyd & D. C. Lagoudas, A thermodynamical constitutive model for shape memory materials. Part.I. The
monolithic shape memory alloy, International Journal
of Plasticity, Vol.12, No.6, pp.805842, 1996.
Craig A. Rogers, Active vibration and structural acoustic
control of shape memory alloy hybrid composites: Experimental results The Journal of the Acoustical Society of
America, Vol.88, No. 6, pp. 28032811, 1990
A. Baz, T. Chen, and J. Ro, Shape control of NITINOLreinforced composite beams Composites: Part B, Vol. 31,
pp. 631642, 2000.
B. S. Jung, M. S. Kim, Y. M. Kim, W. Y. Lee & S. H.
Ahn, Fabrication of smart air intake structure using
Shape Memory Alloy wire embedded composite Physica
Scripta, accepted, 2010.
822
A. A. Villanueva, K. B. Joshi, J. B. Blottmanm & S. Priya,

A bio-inspired shape memory alloy composite (BISMAC) actuator, Smart Materials and Structure, Vol.19,
pp. 117, 2010.
J.S Koh & K.J Cho, Omegabot: Biomimetic Inchworm
Robot Using SMA Coil Actuator and Smart Composite Microstructures (SCM), International Conference
on Robotics and Biomimetics, December 1923, 2009,

Guilin, China
M.S. Kim, W.S. Chu, J.H. Lee, Y.M. Kim, B.S. Jung &
S.H. Ahn, Manufacturing of inchworm robot using
Shape Memory Alloy (SMA) embedded composite structure, International Journal of Precision Engineering and
Manufacturing, accepted, 2011.
823

Design and development of large rigid plastic packaging for Olives

Manuel F. Castro, Carla I. Martins & Antnio J. Pontes
Institute for Polymers and Composites/I2N, University of Minho Guimares, Portugal
ABSTRACT: The main function of packaging in the context of the food industry is to store, to promote an
increase in the lifetime of products and to facilitate their transport. In recent years there has been a growing
demand for plastic packaging for food applications, leading to the optimization of existing packagings. In order
to keep products with the desired quality they need to display good barrier properties to gases, vapors and
smells. Currently, the barrier in plastic packaging is achieved by the combination of several materials, so called
multi-layer packaging or by the use of rigid packaging with preserves (examples brine and oil solutions). This
work aims at developing a rigid plastic packaging for food industry, with the following requirements: extend the
lifetime of the product, enable the sterilization process, impermeable to oxygen, and avoid the use of preserve
solutions which often display high salt and pH levels.
1
INTRODUCTION
Todays packagings can be classified into three main

types: rigid, semi-rigid and flexible. They can be made
of metal, glass, plastic, paper or combination of more
than one type of material, however with the recycling
issue associated. Table I summarizes the main types
of packagings available on the market and its main
applications.
In the rigid packaging market good barrier properties can be achieved but either for economical or for
environmental reasons, they are only justified for small
capacity containers, i.e., up to two liters. Larger capacity containers (greater than 5 liters) are generally made
of metal. This type of packages has two main disadvantages, such as high cost of raw materials and metal
oxidation after a short period of time (about one year).
These drawbacks led to the need of creating a plastic
packaging with barrier and sealing characteristics at a
competitive cost.
This work aims at developing a rigid packaging
for olives, with 5L capacity. The requirements of the
companies of this type of packaged products are essentially: a container to store food with good oxygen
barrier properties, which can be sterilized, adaptable
to existing filling lines and guaranteed of low cost.
Also, taking into account that the main functions of the
packaging are: protection, conservation, transportation and functionality during handling. The package
to be developed should be able to meet all these
requirements.
2
PROJECT DEVELOPMENT
2.1 Specifications
The most important specifications for this product
were:
Production of components by injection molding.

low cost (less than 2). Easy opening. Possibility of
steam sterilization. Good looking and good surface
finish. Impermeability to oxygen. Weatherproofing.
Easy dcor. With security seal. Optimization of space
on the pallet. Stackable components. Simplicity in
closure. Easy handling. FDA approved materials.
2.2
Concept development
According to the abovementioned specifications a

rigid plastic packaging was developed composed of
three main parts: the body (bucket), the cover (lid)
and the wing (see Figure 1).
Other important aspects related to the handling of
the package were considered, namely: the stacking of
the cap and body and stacking of the final container.
In figure 2 it is possible to observe the stacking of
body and final container.
In order to assure an efficient and permanent seal
between the body and the cover, a sealing material,
a clamp and a tape with teeth, were integrated in the
cover, as it is shown in Figure 3 and 4.
The opening of the package is made by removing
the tape. When this tape is removed, much of the tooth
that ensures the closure is also removed, allowing an
easy opening. Thus, the tape facilitates the opening
of the lid and also works as a security seal: i) with
tape works as an inviolable packaging; ii) with tape
removed it indicates that the package has been opened
at least once.
In figure 5 it is possible to observe the cover with a
transparent tape
As this work aims at developing a rigid packaging with better barrier properties, two solutions
were considered: In Mould Decoration (IMD) using
films composed of barrier materials such as EVOH
825
Table 1. Types of packaging available on the market
Rigid
Semi-rigid
Flexible
Metallic
Glass
Plastic
Paper
Cans in tinplate
and aluminum
Aluminum trays
Bottles and jars
Aluminum foil,
laminated structures
Trays, bottles,
jars, crates and boxes
Polystyrene trays,
bottles, glasses
and pots
Films and laminated
structures
Cardboard
boxes
Cardboard boxes
and cartridges. Trays
and alveoli molded pulp
Sheet of paper
and laminated structures
Figure 3. Detail of the lid with seal.
Figure 1. Concept developed for the rigid plastic packaging.
materials available. The packaging material selected

should be able to meet design specifications.
The most important specifications considered are:
Tue-resistance brine (pH and salinity).Being thermoplastic material. Recyclable. Approved by the FDA
regulations (Food and drugs administration). Costeffective. Resistance to UV; resistance to autoclave.
Impact resistant. Good surface finishing. Good barrier
properties to Oxygen.
For the selection of materials it was used the
OMNEXUS database and IDES Plastics (for the selection of materials by grade). The most suitable material
for the bucket, lid and wing was the Polypropylene
PP HJ320MO from Borealis A/S. A thermoplastic
elastomer TPE (Santoprene) was selected for the
seal. The grade selected was EZPrene55A-400 produced by Polyprime.
2.4 Solution for cover with incorporated sealing
Figure 2. Stacking of bucket and final container.
(Ethylene Vinyl Alcohol) or using specific additives

to promote barrier properties.
The smooth surfaces achieved by injection molding
allows us to consider IMD as an alternative to traditional painting or labeling, saving cycle time during
the processing of the part. The IMD film can have
dual functionality, including decorating the packaging
making it more appealing, and increase the impermeability to gases whenever the film is produced using
materials with barrier properties [2].
2.3
Selection of the materials
The definition of specifications for the selection of

materials is one of the major milestones of the work,
because through them limits can be set on the range of
To increase the tightness and sealing in the area of

union between the bucket and the lid, as shown in
Figure 3, the solution adopted was to use a TPE
sealing material. The incorporation of such material,
with a specific cross-sectional area, in the lid, can be
developed through several processes:
a) Overmoulding In this process one of the materials
is inject over the other, in order to obtain a single
piece with the two materials. It is essential a good
adhesion between both materials [1].
b) Deposition of the sealing material in this solution,
after the production of the cap, the sealing material
is deposited in the molten state in the joining area.
c) Production of cap and seal independently. Cap and
seal are mounted on top of the each other.
Another way to ensure a tight seal, waterproof and
impermeable to gases in the junction area of the body
and the cover consists of thermosealing the package
with a film with barrier properties, just under the
cover. In this case, no seal is needed for the cover,
as represented in Figure 3.
826
Figure 7. Cutaway view for uncertain to production of seal.
Figure 4. Detailed view of a closed package bucket and

lid seal.
Figure 8. Cutaway view of the uncertain to cover production.
Figure 5. Detail of the cover with the transparent tape.
Figure 9. Cover prototype with sealant.
Figure 10. Bright field microscopy of PP/TPE interface.
Figure 6. Side view of the mold for the production of the

cover and sealant material by overmoulding.
In the present work, the overmoulding technique

was adopted for the production of the cover with the
sealing material incorporated. In this process, the rotation of the mold is used to carry the seal (pre-molded
shape in the first stage) to a second printing, which
will run the injection of the cover material [2, 3].
After starting the process, there is the simultaneous
injection of two materials: the seal material in the first
printing and the cover material on the preform obtained
with the seal material [4]. Figure 6 displays the layout
of the template developed for the production of the
cover with the seal.
The adopted constructive solution consists of the
use of two hot runner feeders in the fixed part of the
mold and another in the moving part. The feeder of

the fixed part injects the TPE sealing material and the
feeder of the moving part injects the cap. This mold
has two plugs in the nozzle of the Hot Runner which
are on the moving part to ensure that only one material
is injected into each cavity
2.4.1 Production and testing of sealing material
To test the adhesion between the material of the seal
and the lid, a prototype mold was developed. The
overmoulding technique has been adopted for the production of the cover. The drawings of the false cast can
be seen in Figures 7 and 8.
This mold consists of two cavities and a bushing.
The first step of the production phase consists of the
injection molding of the sealant material. The sealant
is then replaced by the hollow cavity of the lid and
the cover material is injected over the sealant material.
827
ACKNOWLEDGEMENTS
The authors acknowledge the financial support given
by Agency for Innovation to the Agropack Project in
which the Embalnor Company is the promoter.
REFERENCES
Figure 11. Polarized light microscopy of PP/TPE interface.
Figure 9 shows the cover prototype, with the sealant

material incorporated.
Figure 10 and 11 illustrate the results of the crosssectional area of the sample in the sealing region, as
observed by optical microscopy.
The results show a good adhesion at the interface
of the two materials, no defects were observed.
3
CONCLUSIONS
Gomes, M., Martino, D., Pontes, A. J. & Viana, J. C.

Co-injection Molding of Immiscible Polymers: Skin-Core
Structure andAdhesion Studies, Polymer Engineering and
Science.
Cunha, A. M., Pontes, A.J., Non-conventional injection
molds. In Injection molding: Fundamentals and application, Chapter 4. M. R. Kamal, A. Isayev, and S.-H.
Liu (eds), Hanser publishers, Munich, 2009. 21 Chapters,
p. 926, ISBN: 978-1-56990-434-3.
Pontes, A.J., Brito A.M. Eds, Manual do Projectista para
Moldes de Injeco de Plsticos. Centimfe: Marinha
Grande. 2004. 10 Fasciculus, 670 pg., ISBN: 972-988721-7
Dondero, M. Adhesion control for injection overmolding
of polypropylene with elastomic ethylene copolymers,
Polymer Engineering and Science, Society of plastics
Engineers, 2009.
Results show that the proposed concept meet the initial

specifications of the project. The selected materials seem to have a good adhesion and compatibility,
enabling the use of a cover simultaneously working
as a sealant. A rotational mould was developed for
the production of the lid, improving the production
efficiency
828

The potential for additive manufacturing in jewellery design

T. Ferreira, P. Brtolo & R.I. Campbell
Centre for Rapid and Sustainable Product Development, Polythecnic Institute of Leiria, Portugal
ABSTRACT: In a world economy of global challenges, additive manufacturing is increasingly playing a key
role in product development. In jewellery there is a constant need to innovate and optimize processes, as this is a
very demanding sector regarding products, materials and designs. There are some limitations in jewellery design
regarding new geometries, to produce differentiated products, as conventional methods do not support sustainable
production with controlled costs. This work highlights the advantages and disadvantages of traditional jewellery
production through the method of lost wax casting and proposes a new production method called Direct Metal
Laser Sintering (DMLS). A ring geometry was developed and produced by these two processes. Some limitations
occurred in the conventional process, as the ring was not produced with the desired requirements, while through
the DMLS process the ring geometry achieved a high quality level. The DMLS process can be an innovative
tool for jewellery production facilitating the designers creativity regarding new geometries and designs, while
reducing time-to-market.
Keywords:
Design, Jewellery, Additive Manufacturing, Direct Metal Laser Sintering.
INTRODUCTION
This study about jewellery aims to optimize the

jewellery manufacturing process by:
It is important to understand the role of Additive Manufacturing in an area still clinging to tradition, though
needing process changes in the short term to remain
economically competitive. In this context, this work
seeks to highlight the points where changes can result
in a greater economic efficiency and less environmental damage. The designer is the engine behind these
changes, taking into account the production processes
and the geometric freedom that can result from these
changes.
In the European Union it is estimated that the
processing of precious metals and jewellery-related
sectors involve more than 3700 million Euros per year,
while employing about 250 thousand people, without taking into account the gems and diamonds sector
(Eurostat, 2006). However, excluding the large exporting companies in Europe, it is an industry with a craft
family heritage tradition, still operating essentially
with manual methods.
There are some shortcomings in the energy and
waste treatment aspects of the industry, as well as a
lack of training of craftsmen to gradually implement
CAD processes and new working techniques. As a
result, there is an opportunity to decrease the production of waste and recycling costs, labor hours per unit
and energy consumption, while allowing the designer
to have more freedom regarding geometries, without compromising the viability of a particular piece
(Lemos, et al., 2004). Nowadays, specifically in Portugal, jewellery companies are still highly dependent
on the craftsman at all times of production.
Optimizing competitiveness;
Studying new fabrication processes;
Minimizing environmental damage;
Increasing the designers role as an engine of change
in manufacturing methods;
Introducing freeform and conceptual innovation.
To pursue this aim, a comparative study between

two manufacturing processes was undertaken:
Investment Casting;
Direct Metal Laser Sintering DMLS.
Both were used to produce ring designs created by

the author.
2
2.1

Materials
One of the fundamental properties of the metals used

in the jewellery industry is ductility, meaning a great
capacity to accepting modulation and support a significant deformation without rupture. Another quite
important property is resilience, which is the ability of
the material to support a force and return to its original
state without plastic deformation. Therefore, metals
and alloys used for jewellery should ideally have these
two properties combined with a relatively low melting point, which is essential for the use of certain
techniques.
829
Table 1.
Physical properties of Precious Metals.
We can consider that the precious metals must

include these essential features which distinguish them
from other metals:
Brightness;
Rarity;
Ductility;
Resilience;
Relatively low melting point;
Less reactive than other metals.
In the following table (Table 1) some physical

properties of these metals are described (Smith, 1998).
More recently, stainless steel is being used for jewellery. Despite being difficult to work, it is highly
resistant to corrosion (due to its constitution metals
such as chromium, nickel, silicon, and others that are
present) resulting in a prolonged polished appearance
(Brady, et al., 2002). For the case study rings production the materials used were silver and stainless
steel, for the investment casting and DMLS processes
respectively.
2.2
Figure 1. Framework of this research work.
Methods
Experimental work was carried out to understand the

use of additive technologies for jewellery. Different
stages ware considered in this research:
Five different rings were designed using the Rhino

CAD software.
Three different additive manufacturing technologies (EOS M270, Envisiontec Prefactory and Connex500) were utilized. These technologies enabled
either the direct produce of the final rings or the
production of masters for indirect fabrication.
Five masters produced by additive manufacturing
were used to create the final rings through the Lost
Wax Investment Casting technique.
One ring design was produced by direct manufacturing using Laser Sintering.
The framework of this research work is presented
in Figure 1.
Figure 2 illustrates the five different designs for the
rings considered in this research work. Ring design
12 rails was produced using both routes.
2.3
Fabrication of rings using additive

manufacturing.
Three different additive technologies were used to produce the rings indicated in Figure 2. Tables 2 to Table 6
indicates both the main characteristics of the additive
Figure 2. Rings Geometries: a) Peoplerow; b) Organic

holes; c) Eternity Line; d) Helix; e) 12rails.
technologies and materials used. Figures 3, 4 and 5

show the physical models of the rings produced on
the three different system. Five of them are polymeric
models and one is a metallic one. Tables 2, 3, 4, 5 and
6 gives details of the equipment and materials used to
build the models.
830
Table 2. Machine used to produce models for centrifugal

cast. Polyjet technology.
Table 3. Machine used to produce resin models by Stereolitography.
Figure 3. Rings with support material, pruduced by

Connex500 - 3D Printing System.
Table 4.
Material used in Stereolithography process.
Table 5.
DMLS technology.
Figure 4. 12Rails ring, model in resin produced by EnvisionTec Prefactory.

Table 7. Working time for each task performed.
Table 6.
Material used on EOS M270.
2.3.1 Investment casting

For this technique, it was decided to cast the ring in
silver 925, because in jewellry and through this process
the materials used are silver and gold. Two different
ovens were used to produce the final rings by the Lost
Wax Casting technique. The first ring was obtained by
a centrifugation oven and the second one by a vacuum
system.
2.3.2 Casting by centrifugation

The traditional casting process takes longer than an
AM based process, because it requires additional steps
between the model construction and the construction
of the wax tree, which in this case were not necessary.
These steps are normally required, because most jewelers do not have AM systems, so they need to make
rubber moulds of the jewellery pieces.
831
Figure 5. 12Rails ring, with supports and made in Stainless

Steel, produced by EOSINT M270.
Figure 7. Ring 12rails - Model in resin and final piece in

Silver 925.
is faster, the production steps are basically the same

(Costa, et al., 1992).
Model production by Stereolithographic process;

Model post cure;
Wax tree construction;
Construction and drying of cylinder master;
Heating the cylinder in oven to melt and burn the
wax and resin;
Metal cast, Silver 925;
Remove the piece from the plaster mould.
2.3.4 Ring produced by laser sintering
In this process, a 12Rails ring was produced directly in
stainless steel (Figure 8). It was first of all necessary
for the STL file to be checked for errors using the
Magics software, the others rings produced were also
checked in this way.
DMLS Process steps:
Machine preparation and heating;

Study of the size and location of the model supports;
Choice of model orientation, in this case the vertical
orientation was chosen because the supports could
be easily removed in this way;
Building of the model.
Figure 6. Final result of the four rings casted.
Although they can save some time in the manufacture of models, the initial prototype production
creates constraints, because it is often done by hand,
in materials such as silver or even directly in wax.
In this work, the reduction of working time was considerable, through skipping some steps, as the process
started directly with the construction of the tree casting (Cheah, et al., 2005). (need to refer to Table 7) The
work of separating rings from the rest of the tree, and
all the finishing was done manually. It was a lengthy
process because it required great care and thoroughness (Figure 6). Note that this process requires a greater
amount of process material that is not in the piece,
which serves only as a support and casting path. In
this case, the molten material that is not in the piece
exceeded 60%.
2.3.3 Casting by vacuum system
The ring cast by the vacuum system was the 12Rails
design. The ring model was made in resin by the Envisiontec AM machine (Figure 7). Although the process
Removal of supports by hand.
It was considered essential to manufacture a ring with

the same geometry, for a correct evaluation of the two
different processes of production, and hence to understand the most appropriate method of manufacture for
that same geometry. A 12Rails ring was produced in
Birmingham City University by the method of lost wax
casting, with models manufactured by stereolithography using an Envisiontec machine. The production by
DMLS was done by the 3TRPD company based in UK,
using an EOS M270 machine.
After the choice of geometry and its production,
it was essential to investigate the objects in order to
understand the differences and possibilities of the two
832
Figure 8. Stainless steel ring produced by DMLS with

supports.
production methods. For the ring modeling, it was necessary to obtain geometries capable of causing the
same constraints in the manufacturing through the two
cases, but not at all impossible to produce. Therefore,
a geometry that was apparently simple, created some
difficulties in the act of production, as follows:
For the Investment Casting:
Figure 9. The image shows the location of the three inner

rings to strengthen the structure.
The paths through which the molten metal would

have to travel, were long;
Very small thicknesses, causing incomplete molten
metal filling, in this case, the model had the minimum thicknesses of 0.64 mm;
The geometry was too flexible; the resin could not
keep its shape at the end of processing;
Because the resin was too flexible for the desired
thickness and geometry, the final geometry was
quite deformed.
Figure 10. The images show the supports. The red circle shows the small internal supports that are impossible to
remove.
For Sintering:
The ring produced, even after the application of the

three inner rings, was quite distorted compared to
the initial geometry, verifiable in full-scale.
Production difficulties using DMLS:
The difficulties were mainly related with the need

of several supports for production, the artisan took
more time in the finishing work to avoid damage in
the piece.
During the ring 12rails production, difficulties have

arisen in both the process of lost wax casting, and
DMLS, as described below:
Production difficulties in Investment Casting:
Geometry thickness although it is possible to

build geometries down to 0.5 mm thick, and the
ring 12rails thickness was 0.64 mm, the path that
the molten metal had to follow was too large for the
existent thickness;
The production of resin model by stereolithography
was found deficient the resin is a material too flexible for the geometry shown, causing deformation
in the model both at room temperature and when
handling;
Due to the deformation of the resin model, it was
necessary to modify the internal geometry of the
piece, putting three inner rings in order to get more
resistance, and less deformation (see Figure 9).
833
Placement and quantity of supports for processing

the piece geometry forced the placement of many
supports, some impossible to remove (internal to
the piece). It is also important to refer to the plates
needed to support the five arches of the ring bottom
(see Figure 13).
These supports needed to be correctly designed to
be easily removed with the minimum of marking.
Nevertheless, the hand removal of the support adds
cost and presents finishing limitations. The supports must also be accessible for the manual removal
process, which places practical restrictions on the
geometry.
Investment Casting steps:
For the model produced with the Envisiontec stereolithography machine, it was necessary to remake
the CAD file, because the resin production resulted
in a geometry deformation (Figure 10);
Construction of tree in wax and resin;
Construction of the plaster drum;
Burn out of the resin and the wax in oven;
Metal Casting in a plaster mould with a vacuum
system.
Figure 13. Geometry comparison of the processed rings.
Figure 11. The image shows the surface roughness in both

cases of production.
Figure 12. Images a, b and c, are from the sintered ring,

while images d, e and f, are from the casted ring.
It was also important to make an analysis of surface

roughness. This feature is very important in jewellery,
as the level of roughness dictates the time of the finishing work, in this case, hand-done by a craftsman. This
may cause high costs in the manufacture of jewellery
pieces.
The result was different roughnesses because the
production methods were different. In the sintered
ring, it is clear that the roughness is the result of powder fusion by laser, causing the small metal particles to
join together with a balling phenomena, an unfavorable
defect associated with direct metal laser sintering (see
Figure 11). In the ring produced by casting, the roughness is caused by the addition of different aspects, such
as: the finishing level of the resin model, the plaster
mold precision, the rail that the metal need to fill, (the
mould was too long), and watermarks (Ott, 1997)
(Grande, et al., 2004) (see Figure 12).
In this work, microscope photographs were taken,

with different magnifications of the two rings being
studed. Figure 13 shows the differences in the piece
surfaces, the left side shows the ring produced by
DMLS, the right is the ring produced by casting. Magnifications have not exceeded 200x (enlargement).
The sintered ring has a surface marked by laser
marks. The substantial process defect associated with
DMLS is the balling phenomenon, which may result
in the formation of discontinuous scan tracks. Laser
sintering is performed line by line and the energy is
absorbed by powder particles, consequently the molten
materials in the irradiating zone tend to aggregate
into an individual coarsened sphere of approximately
the diameter of the laser beam (Forrest, 2009). This
defect can be minimized though increasing the laser
power, lowering scan speed, or decreasing the powder
layer thickness, which this could alleviate the balling
phenomena (Gu, et al., 2009). The rings featuring a
casting roughness, are essentially because of watermarks. In general, this defect is independent of the
alloy composition, the cause is an incorrect investing
process and the casting method has little influence.
The surface imperfection of the resin model is also
one of the causes to obtain surface roughness (see
Figure 13). After mixing, the investment slurry contains an excess of free water. When the setting process
begins, the water is taken up by the binder. If the slurry
is left motionless for several minutes before the setting
starts, the water and solids can separate. Some of the
separated water can collect on the surface of the wax
models and may run together, forming the characteristic watermarks. Such defects occur frequently when
the slurry is made with too much water (Ott, 1997).
In the two rings that were produced is also visible
that the cast ring has a large geometry deformation.
In the sintered ring it is not perceived any significant
change in geometry (see Figure 13).
Therefore, a 3D scan of the object was made with a
Comet5 machine from Steinbichler to evaluate if there
was any geometry change during the ring process.
Due to the geometrical characteristics of the object it
was not possible to make a complete capture of the
shape, as (i) Geometry and thickness of the object
was very small; (ii) Geometry had multiple layers that
were difficult to scan. However, some deflections were
observed in areas with higher thicknesses, areas where
it was possible to make some scans, as illustrated in
Figure 14.
834
Table 8. Values related to geometry deflection.
Figure 14. The geometry deflection of the sintered ring

compared with the original CAD geometry.
In Table 9, it can be observed that the point with

more geometry deflection is point A (nearly 1 mm),
however it is also noteworthy that there is only geometry deviation on one side of the ring, while the other
side corresponds well to the initial CAD. These deviations may result from the contraction of the material
and piece weight not being symmetrical relative to its
axis.
After the rings were produced, there was still a need

for finishing, though the two processes achieved very
different levels of quality, culminating in different finishes for the two rings produced. The ring made by
the casting method required essentially a craftsman
working to improve the geometry, removing the sprue
casting and polishing, while the ring made by DMLS
just required supports removal, and because its geometry agreed well with the initial CAD model it only
needed metal plating to regulate the surface.
It can be concluded that the DMLS method for
manufacturing is an important step for improving the
jewellery industry in terms of time-to-market, while
bringing greater freedom in terms of the concepts and
geometries, as well as achieving high levels of quality,
all basic prerequisites for sustainability.
CONCLUSION
REFERENCES
After the experimental work carried out with rings

produced by different methods, it can be concluded
that not only complex and intricate geometries cause
production constraints in the current methods used in
jewellery. There are more factors that can invalidate the
production using the process of Investment Casting,
such as the thickness of the geometry, the dimension
of the path the molten metal must go through in the
mould part,and the material quality used for casting.
The designer must consider all these factors so that the
final object can be reproducible.
The ring produced by DMLS, 12rails showed higher
quality geometry, took less working time, and fewer
process steps, and produced less melted material in the
final piece, resulting in an increased competitiveness
compared to the Investment Casting process.
On the other hand, when the ring was produced
by lost wax casting, the 12rails ring had a greater
deformation geometry and surface roughness. This
deformation and roughness is the sum of several errors
and flaws at each stage of the process, which were
impossible to reduce without some changes in geometry, resulting in the difficulty of the molten metal
filling the mould cavity. The DMLS process also
presented another significant advantage when compared with the conventional process. While in the first
method the molten metal that is not in a piece is about
30% (in many geometries it can be reduced to almost
0%) in the method of the lost wax casting the molten
metal that is not in a part is about 60 %. This value can
only be reduced through the piece ratio of castings, but
never achieving values of less than 30%.
Brady, G.S., Clauser H.R.,Vaccari, J.A., Materials Handbook,

U.S.A. McGraw-Hill, 2002.
Cheah, C.M., Chua, C.K., Lee, C.W., Feng, C., Totong,
K., Rapid prototyping and tooling techniques: a review
of applications for rapid investment casting. London,
Springer-Verlag, 2004.
Cheah, C.M., Chua, C.K., Lee, C.W., Feng, C., Totong, K.,
Rapid Prototyping and tooling techniques: A Review of
Applications for Rapid Investment Casting, Int. J. Additive
Manufacturing Technology. UK, Springer-Verlag, 2005.
Vol. 25.
Eurostat European Business Facts and Figures, European
Communities, 2006.
Forrest, M. Mass customisation, massclusivity and mass
individualisation. The Jewellers feel the change? The
Goldsmith. The Goldsmith, 2009.
Grande, A.M., Ugues, D., Pezzini, D., Controls and Quality
Demands in the Jewellery Investment Casting Process,
Metalli Preziosi. Alessandria,2004.
Dongdong, G., Yifu, S., Balling phenomena in direct laser
sintering of stainless steel powder: Metallurgical mechanisms and control methods, Materials and Design. China,
Elsevier, 2009. Vol. 30.
Lemos, R., Canellas K.V., Economia Criativa: o Design
como Valor Estratgico de Inovao no Cenrio Mundial
Contemporneo. Universidade Regional de Blumenau,
2004.
Ott, D., Handbook on Casting and Other Defects, In Gold
Jewellery Manufacture. London, World Gold Council,
1997.
Smith, W.F., Princpios de Cincia e Engenharia dos Materiais. Portugal, McGraw-Hill de Portugal, Lda, 1998.
835

Customization tool for people with special needs

B. Providncia
Mechanical Engineering Department, University of Girona, Campus de Montilivi, Girona, Spain
Textile Engineering Department, University of Minho, Campus de Azurm, Portugal
J. Ciurana
Mechanical Engineering Department, University of Girona, Campus de Montilivi,Girona, Spain
J. Cunha
Textile Engineering Department, University of Minho, Campus de Azurm, Portugal
ABSTRACT: There are people with limited skills that are considered people with special needs (PSN). A large
number of persons have body pain due to the contact with support surfaces for a long time. This paper aims to
design and implement one tool to help design and manufacture customized devices in order to reduce the pain.
User-centered design (UCD) methodology in combination with psychophysical user data generates a 3D
geometry that can be produced in a rapid manufacturing environment. One tool is developed in Labview composed
by three modules: a) import and selection of psychophysical data, b) data analysis and reconfiguration into new
geometries, and c) build point cloud data that considers materials deflection which can be used by CAD systems
for rapid manufacturing. Results show that this is a tool capable of reducing/delaying the onset of pressure ulcers.
INTRODUCTION
The evolution of design methodologies based on the

user places user needs as the central point of product
design.
As a response to the improvement in technology and
production supported by a booming economy, there
was an increase in mass production and competition,
which forced smaller markets to rediscover new ways
to relate with clients. Customers who are increasingly
demanding, become the focus of attention during the
design process.
In lates 90s, Kerckhove (1997) stated that the relationship the user has with a product is a continuous
communication process, meanwhile in the computing
area, Human Computer Interaction (HCI) works were
developed on graphical interface relationship with the
computer, often addressing ergonomic issues, or other
areas likely to be quantitatively assessed.
Yong-Ki Lee (2009) in turn, speaks of the products
semantic relations as signs of value in a perspective
of language domains as a qualitative value. Black
(2008) of the Design Council, explores User Centered
Design based on the central premise of understanding
user needs and obtain user feedback, which results in
better designed products. The product use feedback
is obtained by the analysis of a set of exercises in
which the user expresses the experience in product
interaction.
The emotional product design introduced a new
way of being with the design, as shown by Donald
Norman (2005), in a more psychological perspective

based on three assumptions relating to our emotional
state:
a visceral level that makes an immediate judgment;
a behavioural level that relates to our most rational
part, the body control;
and finally, a reflexive level that reflects our interpretation, comprehension and reasoning part.
Desmet (2008) addresses the evaluation of tangible and intangible emotional design where a product
experience process goes from a response to emotional
stimuli, in a similar case of the Borg CR10 and other
works.
Killi (2007) refers the precious help in customizing
products, such as in the case of Adidas and Nike, where
the user has the possibility to participate in setting
up its own shoes through the Web, for example by
choosing color.
In the case of the Freitag (Providncia, 2008 and
Santarm et al., 2005), some bags made from recycled
truck canvas are studied, and the user has the possibility to interact/participate in defining the look for a
bag, from a set of canvas on display.
Not less interesting, is the use of rapid prototyping technologies for the production of hearing aids by
Siemens that involves the 3D scan of a mold of the
hearing cavity for, from the CAD treatment, allowing
the construction of a more discreet prosthesis adapted
and customized for the user. Or even the exercise
done by students who developed a pair of sunglasses
837
Borg CR10, a tool that works with a set of questions

posed to a patient during a particular process in time
and that are used to measure the evolution of the
emotional state.
Mocap, a tool that allows monitoring a patient
movements, from strategic points such as joints
where marks are placed to be recorded by a set of six
video cameras. The information is then processed
by a specific software.
based on the 3D scan of their face, contributing to an

approach that explores more the emotional relations
based in aesthetic factors (Killi, 2007).
These cases were the starting point for thinking
about design methodologies, not based in what users
express, but if possible, in their emotions, enabling
solutions where the probability of the products being
developed would be more certain both in terms of
acceptance as in terms of solving the individual needs
of each user.
4
2
DOMAIN OF APPLICATION
People having limitations in their motor skills, either

consciously or unconsciously, are defined as people
with special needs (PSN). This group also comprises
people who have post-surgical problems. A significant number of these persons experience problems
concerning the contact of body areas with support surfaces. Due to their lack of capacity for autonomous
movements, these persons are subjected to excessive
skin pressure and restricted blood circulation, resulting
in the development of pressure ulcers (PU).
PUs are skin wounds of ischemic origin usually
localized over a bony prominence, that can vary in
intensity either being superficial or deep (Blannes et al,
2004). The PU classification is obtained by visual
assessment of the level of tissue damage, in stages
ranging from I to IV (Bates-Jensen et al, 2008).
The main causes of PU are of extrinsic nature, and
among these the most relevant are pressure, humidity,
friction and shearing. Pressure is the main factor contributing to PU and the pathological effect on the skin
can be attributed to pressure intensity and duration as
well to the skin tolerance (Providencia and Ciurana,
2010).
METHODOLOGY
The use of a set of tools for acquiring and monitoring

psychophysical data has been a practice in the works
studying the discomfort often associated with product use. This is the case of the F-Scan systems of the
Tekscan company, that allow the mapping and monitoring of foot pressure on the insole of the shoe, or the
case of the pressure forces exerted on a bicycle seat
during the cycling activity where the crossing of the
pressure data with the movement capture during the
pedal cycle allows data crossing in a more complete
study. (Wilson et al, 2007). Similar studies address
issues associated with UCD and emotion, as the ones
previously described.
This paper addresses three psychophysical data
tools (Providencia and Ciurana, 2010).
Tekscan, a tool based on a film with pressure sensors that allows the mapping of pressures in a given
area. This technology allows for data acquisition by
a set of frames over time.
DEVELOPMENT AND TESTING OF THE

CORE SYSTEM
The core system is a software developed in Labview

that is part of a solution for the development of a tool to
support special needs, in this case people with physical
impairment that need to use wheelchairs.
The software is subdivided in a set of sub-modules
that perform the processing and interpretation of psychophysical data. A virtual three-dimensional surface
is obtained based on this data that can be exported to
a CAD system. This surface will provide the desired
pressure distribution for reduced PU risk.
4.1 Sub-module 1 Import and selection of
psychophysical data
In the first sub-module, a text file generated by the
Tekscan system, representing a series of measured
pressure maps, is imported. Pressure values (mmHg)
are given for each of the coordinates of the map. The
resolution of the maps is 32x32 cells (fig. 1)
The pressure values collected refer to values taken
perpendicularly to the film, in each grid cell.
The choice of a time interval for averaging of the
pressure, or a single frame to be considered in the
development process, is based on the choice of a specialized technician who makes the interpretation of the
intersection of the pressure data with the Borg CR10
scale (where the patient is induced to verbally express
feelings regarding discomfort in function of time),
and the movement interpretation is obtained through
motion capture (MOCAP).
The software presents in this first phase a set of
two graphs figure 2, a 2D-graph using a color scale
for representation of the pressure values and a second graph representing the pressure maps in 3D, that
allows a more visible interpretation of the pressure
values.
At this stage the selection of frame or frames, in
the timeline, to be considered during the study, is still
possible.
4.1.1 Relationship between pressure areas and
ulcers
Assuming that the highest pressure values are the propitious values to the emergence of UPs, the possibility
of reducing the pressure values was studied, especially
in the areas where UPs normally arise, through a better
pressure distribution.
838
Figure 1. Distribution of pressure sensors and obtained pressure data.
Figure 2. 2D and 3D simulation of the pressure map.
An approach was developed to reduce the values of higher pressure, which can be a solving
part of the problem, as in terms of morphology it
would create areas of lower contact at critical points
and minimize the pressure zones based on a better
distribution.
4.1.2
Relationship between the pressure areas and

the seat base morphology
In an initial stage, the pressure values where related
with the seat base morphology, creating a direct relationship between pressure values vs. surface cavity
values.
A maximum depth for the depressions was defined
in basis of the height of the coccyx.
This first approach was the starting point that provided the basis for the information processing of the
developed tool, in accordance with the experiments
carried out by the psychophysics working team.
4.1.3 First prototype
In the first prototype, the values of the pressure
were transferred to the CAD (Solidworks) in order to
prepare a first test.
Based on the pressure values a set of lines parallel
to each other was created. After the application of a
loft (solidworks tool to draw surfaces from lines), a
virtual surface was created (seat base). Loft tool draw
the surface in a such way that results poor surface morphology due to assumed vector values from Teckscan.
The surface drawn was then explored and a solid was
839
Figure 3. First prototype developed in MDF.
Figure 4. Pressure smoothing graphs.
built from a parallelepiped volume, a little longer and

wider than the pressure map and with a height of
6 cm (a higher value than the expected one calculated
through the difference in depressions).
After the construction of the virtual solid, the data
for milling was processed based on the CAM GOELAN tool. This tool allows, based on the CAD file and
considering the milling machine, to prepare a numeric
code for the milling, including information of drills
to use and axial speed and the design that the tools
perform during the milling process. This software also
allows testing design errors when machining as well
as having process time estimation. After processing all
this data, the software generates a code file that can be

read by the machine controller.
From the first prototype developed, figure 3, the
following conclusions were drawn:
840
The methodology used to construct the surface

in solidworks (loft), did not allow a continuous
smooth surface along the width and length as can
be seen in figure 3;
Although it is not entirely perceptive in the photo of
figure 3, the 8 mm diameter spherical head cutter
chosen was too small and therefore created marks
along the surface.
Figure 5. Prototype evolution made with MDF.
Figure 6. Deflection Modelling.
In conclusion, the obtained morphology had not

only a poor surface treatment but also the depressions
and transitions among areas were too sharp.
The analysis of the first prototype highlighted the
following questions:
How to control the excessively sharp depressions;
How to control the demarcation by the way in which
the loft was processed,
How to avoid the texture marks caused by the
milling.
4.2 Sub-module 2 Data analysis and

reconfiguration
To answer these questions, the sub-module 2 of the
core system was then developed following three steps
Step A): Flattening the pressure map for a more
distributed map.
In a first step, a new pressure map is created, which
represents the original map, but smoothed. This is
achieved through an average or local median in each
point. Each point is replaced by the average/median
841
study the foam behavior and the shape of the surface

when using a flexible material (fig. 7).
In Point 3, the viscoelastic foam to be used in
the cushion is characterized by compression tests
using a dynamometer, in order to determine the
force/deflection curves. Three different thicknesses
of the same foam were tested and it was found that
the relation between percentual deflection and force
applied is equal for all thicknesses, up to about 75%
of compression (fig. 8). At this point the foam starts
to saturate. A polynomial interpolation is performed
on the measured points and the resulting coefficients
are input into the core system, that generates the
pressure/deflection curve (fig. 9).
At this point the system has three pieces of information:
Figure 7. Schematic representation of the two layers of

the seat base. A) Initial state of the cushion; B) Cushion
representation when in use.
between itself and N of its neighbours in x and ydirection. The number of neighbouring points N allows
controlling the degree of surface smoothness.
The operation of local averaging results in a sum
of pressures that is different from the original sum.
To correct this point (because the sum of all pressures
must be equal to the original, for it results of body
weight) a scale factor must be used.
Step B): Smoothing of the surface produced from a
loft
In order to improve the surface, a tool was used for
the conversion of a point cloud into a surface.
This tool permitted to ensure that the morphological
relation between points was identical regardless of the
points acquisition order.
Step C): to avoid the texture marks made by the
milling process, a spherical cutter of 12 mm diameter
was used.
In step A, illustrated in figure 4, pressure smoothing, we can see on the left side the mean pressure
values. These values can be averaged or median and the
values of the neighborhood can be calculated, thereby
controlling the degree of surface smoothness.
The graph on the right side of figure 4 illustrates
the smoothing in 3D giving a better perception of the
results.
In Point 2, deflection modelling, the deflection
maps are generated for the given desired maximum
deflection (fig. 6).
For this purpose, the user imputes the maximum
deflection desired. This value is related to physical
factors, namely the height of the coccyx. The program
applies an exponential to the original pressure map in
order to fit the pressures between 0 and the maximum
deflection desired. Then a smoothing is applied to the
deflection map. Usually, the same degree of smoothing
is used for pressure smoothing.
4.3
Sub-module 3 Build point cloud data that

considers materials deflection
This sub-module aims to incorporate issues of material

deflection, since the objective is to build a seat base
in two layers, a rigid bottom layer and a top layer of
a flexible material. To do so, it became necessary to
The desired smoothed pressure map

The desired smoothed deflection map
The properties of the foam
For each pressure value of the smoothed map, and
assuming that the behavior of each cell of the cushion
is independent of the neighbours, it is now possible
to compute the percentual deflection that the foam
exhibits at the pressure applied. Considering that the
desired absolute deflection is also known, it is possible
to compute the height of the foam that should be used
to match the percentual deflection 1.The system is now
able to generate a surface representing the height of the
cushion to best approximate to the expected behavior
(fig.10). This method is an approximate method and
the success of the end product depends on the choice
of the foam. If the foam is too soft, for instance, the
system will compute values of percentual deflection
exceeding 100%, The software limits the maximum
deflection, but in practice this means that at those
points the foam will be excessively compressed and
pressure values will be higher than those computed
form the pressure/deflection curve. This means that the
desired smoothed pressure map will not be achieved.
The selection of the viscoelastic foam was based on
a good elastic recovery, allowing the non-deformation
of the seat base. This type of material is used in medical
areas for two main reasons: it is an open-cell foam
allowing better breathing of the skin when in contact
with it, and for its antibacterial characteristics.
On the left and right sides of the figure 9 the results
of the final shape of the cushion can be visualized,
considering the maximum and minimum thickness of
foam.
The values obtained by the time data was exported,
are automatically converted to get the final shape of
the desired foam, where a better distribution of contact
zones will minimize the pressure in critical points.
The exported file, then results in a seat base, which
can be machined or printed from the rapid prototyping
tools.
The file downloaded to the CAD system considers
the design of the base, dependent upon its the thickness
and the expected deflection, in order to obtain for the
patient the best desired shape.
842
Figure 8. Graph of extension (%) vs. force (N).
Figure 9. Foam Properties.
Figure 10. Relative deflection computed on basis of the desired pressure map, cushion height map that matches the computed
percentual deflection for the desired absolute deflection.
843
REFERENCES
CONCLUSIONS
The work presented here was designed based on two

premises:
to study a methodology based on the user, that
would allow the development of product design in
order to meet user needs;
to develop a methodology based on the rapid prototyping or rapid manufacturing technologies to
produce a customized product, responding to user
needs.
Analysis and evaluation techniques, in combination
with teams of psychophysics technicians, allowed during the evolution of the process to refine the model as
well as the technologies designed.
This interaction gave an increasingly broad perception of the relationship of the user with the product
experience, especially based on the emotional and
discomfort relation.
A core system was developed, a tool for aiding
people with special needs, which allows working with
psychophysical data to create a text file that can be
read by a CAD system, and use this data to build a
customized product.
In the tests that were done, according to the psychophysical teams, the result was quite interesting,
with a better distribution of pressure and thus an
expectable minimization of pressure ulcers and consequently a better quality of life of patients.
In future work, it seems that there is a large field
study in the area of perception and interpretation of
psychophysical data and in the development of more
effective technology in supporting patients with special needs, either as an aid in the interpretation of data
as in creating custom manufacturing solutions.
Similarly to the technological developments and its
democratization through an increasingly great industry offer, as well ever more accessible values can
be investigated to understand which technologies
should be chosen to respond to custom fabrication
solutions.
Kerckhove, Derrick De, 1997, The skin of culture: investigating the new electronic reality, Kogan Page Publishers,
London.
Lee, Yong-Ki, Lee, Kun-Pyo. User-Generated Product
Semantics: How People Make Meaning From Objects in
the State beyond Saturation, DesForm 2009, International
Workshop on Design & Semantics of Form & Movement,
Taipei, Taiwan
Black, Alison. http://www.creative-net.co.uk/About-Design/
Design-Techniques/User- centred-design-/, 20 October
2008
Norman, Donald A. 2005, Emotional design: why we love (or
hate) everyday things, Basic Books, New York
Desmet, Pieter M.A., Ortz Nicols, Juan C. and Schoormans,
Jan P. Product personality in physical interaction , Design
Studies (2008),Volume: 29, Issue: 5, Publisher: Elsevier,
Pages: 458477
Killi, S. Custom design: more than custom to fit! I: Virtual and rapid manufacturing: advanced research in virtual
and rapid prototyping : proceedings of the 3rd International Conference on Advanced Research in Virtual and
Rapid Prototyping, Leiria, Portugal, 2429 September,
2007 (s. 777783). London: Taylor & Francis.
Providncia, B., Boom, Sirkel, 2008, Issue 3; Publisher: Kem,
Pages 315 Oslo.
Santarm, B., Guedes, M. G., Cunha, J. Designium. Cultural
and Communication Common Grounds in Art Craft and
Design, Pride and Predesign CUMULUS LISBON 2629
May, 2005, IADE, Lisbon, Portugal.
Providncia B., Ciurana J., Interface tool for human communication to integrate psychophysical inputs with rapid
manufacturing technologies; International Journal of
Computer Integrated Manufacturing; Volume 23, Issue 8
& 9, 2010, Pages 777790.
Wilson, Chisom, Bush, Tamara R. Interface forces on the
seat during a cycling activity 2007 Clinical Biomechanics
Volume 22, Issue 9, Pages 10171023.
Blanes, L., Duarte, I. S., Calil, J. A., Ferreira, L. M. (2004).
Avaliao clnica e epidemiolgica das lceras por presso
em pacientes internados no hospital So Paulo. Revista Da
Associao Mdica Brasileira, 50(2), 182187.
Bates-Jensen BM, McCreath HE, Pongquan V, Apeles NCR.
Subepidermal moisture differentiates erythema and stage I
pressure ulcers in nursing home residents. Wound Repair
Regen. 2008;16:189197.
844

Rapid prototyping in collaborative architectural design process

Regiane Pupo, Regina Ruschel & Max Andrade
UNICAMP, Campinas, Brazil
ABSTRACT: This paper describes an experience in architecture design process where a SLS prototyped
model was useful during conceptual stage as well as for a final presentation. The aim of the research focused the
performative architecture throughout collaborative design process. The short time and detailing reached with this
experiment proved the efficiency and the need of such technologies throughout architecture design, especially
in the academic world.
INTRODUCTION
The use of rapid prototyping techniques throughout

architecture design process has been a challenge for
most of schools that has recently set up their labs.
The introduction of digital technologies for making
models, prototypes or even building parts is a reality in most schools. It seems that nowadays it isnt a
matter of if this introduction is relevant, but how
were going to introduce it in the curricula. The term
Rapid Prototyping (RP) was formally introduced to
architecture researches in 1991, by Streich (Sass &
Oxman, 2006, p. 331) as a transformation method of
3D CAD models in prototyped models, in a faster
way, in particular with Steriolitography techniques.
From that point on several experiments, researches
and even new specific techniques have been brought
to discussion in order to establish the most efficient
way to combine RP techniques and architecture design
process.
The paper describes an experiment that was a
semester activity for the discipline entitled Collaborative Design, which is officially a regular subject in the Architecture and Urban Design Course at
UNICAMP (State University of Campinas), in Brazil.
As a technical support, the facilities of the Automation and Prototyping forArchitecture and Construction
Laboratory (LAPAC) were used, a Z Corp 3DPrinter
and a laser cutter, as well as the collaboration of CTI
Renato Archer, an Information technology center in
Campinas.
The importance of this academic experience in
architecture design is discussed where a prototyped
model was useful during the process as well as for
a final presentation. The short time and detailing
reached with this experiment proved the efficiency and
the need of such technologies throughout architecture
design, especially in the academic world.
COLLABORATIVE DESIGN IN
ARCHITECTURE
2.1 Collaborative design

In the recent years the architectural design has experienced the use of algorithms by means of scripting and
programming in CAD environment. Instead of these
experiments have focused on a more intense collaboration with engineers (in order to multiple performances)
it has been used in the exploration of new formal
experiences and spatial relationship.
Design strategies for engineers and architects are
different. One major reason is the different nature of
the problem of the two professions. On the one hand,
problems related to engineering are treated with a scientific rigor through the procedures and predefined
methods; on the other hand, the architects work with
many non-quantifiable aspects that rely on subjective
criteria (Tessmann, 2008).
Traditionally, the architectural design is driven by
a complex process of negotiation of several parameters without the rigor of a data structure that contribute
to a robust optimization. In contrary, in several areas
of engineering many approaches are based on scientific rigor, but that didnt embody the complexity of
architectural design.
The idea to work in a collaborative design is to join
the robust methods of engineering and the complex
relationship of dependence of the architectural design
processes.
The popularization of the use of algorithms in
the field of architecture, almost 20 years after
the Computer-Aided Design (CAD) have become
widespread in this area, can contribute to a rapprochement of the processes of engineering and architecture.
The computer-numerically controlled fabrication and
the scripting language opened the door not only to a
845
digital continuity (from design to manufacturing), but

also allowed the use of evolutionary algorithm, agent
systems and adaptative strategies (Tessmann, 2008).
These formal systems has brought a new field for
architecture design based on well-defined strategies,
but not necessarily predictable.
The architectural design strategies, based on computational concepts made possible, and at the same
time, has the condition, a deep collaboration between
the various design disciplines. The consequence of the
use of the collaborative systems is that the project
can incorporate performance driven the form generation process (Oxman, 2008), using as through various
techniques and computational tools.
In other cases, the performance does not appear
driven the form generation process, but as the instruments for automatic evaluation of the design solution.
Therewith it possible, from an initial model, establishes assessments, as thermal, acoustic luminous,
functional, cost, structural, etc. While these assessments do not interfere directly in the form generation,
contribute to its modification; generate performance
results for solution proposed by the designs, which in
turn alters the solution due to these assessments.
3
RAPID PROTOTYPING FOR

ARCHITECTURE
Rapid prototyping and digital fabrication are respectively the automatic production of physical models and
final objects directly from a computer model, with the
use of computer-numeric controlled machines. The
definitions and the different uses of these expressions in the literature have been discussed in Pupo,
Celani & Duarte (2009). During the past decade,
rapid prototyping laboratories have been installed in
many architecture schools throughout the world. The
establishment of such labs has been, in most of the
cases, accompanied by structural changes in the curricula, in regards to the use of information technologies
in the design proves. Some authors, such as Mark,
Martens & Oxman (2001), have presented proposals
for integrating the new technologies into the curriculum. However, not much has been published about the
operational and administrative aspects of these labs.
3.1
Models
In sciences, to model means to represent reality in

a simplified, schematic, abstract way, showing just
the elements that are strictly necessary to understand
specific aspects of the phenomenon being studied.
Real-life situations are often too complex to be studied, so a model can be seen as a tool that supports our
cognitive process (Mc Millan, 1992). The higher the
complexity of a phenomenon, the higher the need for
modeling it.
There are different types of models and different
modeling techniques. According to Mitchell (1975),
the three modeling methods most used in architecture
are the analogue, the symbolic and the iconic.
In analogue models one set of properties (. . .) is

used to represent another analogous set of properties
of the item being designed (Mitchell, 1975, p.130).
The representation produced by Gaudi, with wires and
sand bags, of the Sagrada Familia, is a good example of an analogue model. In this representation the
tension vectors are represented by the wires, whose
shape and direction are analogue to the physical phenomenon they represent. As a result, the architect was
able to find the (inverted) ideal shape for the vaulted
structure.
Symbolic models use symbols, such as words,
numbers and mathematical operators. In architecture,
symbolic models are used for simulations and evaluations of structural, acoustical, lighting and thermal
performance. Symbols are typically displayed as mathematical formulae, tables and arrays. More recently,
though, advances in computer graphics have allowed
the three-dimensional display of quantitative information on top of geometric models, using gradients of
colors. This type of visualization allows architects to
make quick visual, qualitative evaluations (Kolarevic,
2007, p. 197).
Iconic models are more literal. Typical examples of
their use in architecture are plans, elevations and scale
models. These models involve scale (enlargements and
reductions) and projection (3D to 2D and 2D to 3D)
transformations. In fact, architectural models are usually miniature representations of buildings. Mitchell
(1975) emphasizes the role of this type of model in the
generative process; according to him, in iconic models
a particular state of the system actually looks like
the potential solution which it represents (p. 130).
3.2 Prototypes
Prototypes are a special type of models. According to
the Marrian-Webster dictionary, a prototype is a first
full-scale and usually functional form of a new type or
design of a construction. Prototypes are usually built
during the design process in order to preview how the
building or building part will look like and function.
It is also used for planning the production process.
According to Liou (2008) prototyping is an
approximation of the product along one or more
dimensions of interest (. . .) ranging from concept
sketches to fully functional artifacts. According to
this definition, prototypes are not necessarily fullscale models. In architecture, it is often unfeasible to
produce full-scale prototypes of entire buildings. For
example, when Frank Gehry designed the Walt Disney
Concert Hall, a 1:10 prototype of the main music hall
was built to test its acoustics (Fig. 1).
Usually, full-scale prototypes are used in architecture only when a building or a building part will be
reproduced many times. For example, a small part
of the faade system used in Gehrys Guggenheim
Museum in Bilbao was produced in full scale to test
its performance (Fig. 2).
Therefore, the concepts of prototype and scale
model in architecture overlap, especially when we refer
846
Figure 3. Virtual Simulation.

Figure 1. 1:10 scale model for Disney Concert Hall.
Figure 2. 1:10 scale model for Guggenheim Museum in

Bilbao.
to working models. Working models are one of the

three types of models defined by Hetchinger & Knoll
(2008). The other two are conceptual and presentation
models.
Since, as seen above, architectural scale models can
sometimes be considered prototypes, the expression
prototyping can also refer sometimes to making scale
models.
4 THE EXPERIMENT
The experiment took place in a discipline entitled
Collaborative Design, which is officially a regular subject in the Architecture and Urban Design
Course at UNICAMP (State University of Campinas),
in Brazil. As a technical support, the facilities of the
Automation and Prototyping forArchitecture and Construction Laboratory (LAPAC) were used: (1) a Z
Corp 3DPrinter and (2) a laser cutter, as well as the
collaboration of CTI Renato Archer, an Information
technology center in Campinas.
Throughout the second semester of 2010, the
activities were divided into three phases. On the
first phase, the activities were related to information
and discussions concerning collaborative architectural
design, Building Information Modeling (BIM) and
its application around the world. The students had
the chance to research, learn and practice about the
ways collaborative architectural design and BIM can

improve and facilitate all design processes stages, turning decision phases much faster. In a second phase,
the research went towards study cases where the performative architecture was an important issue for
the final building design. According to Kolarevic &
Malkawi (2005), the increasing emphasis on building
performance from the cultural and social context to building physics is influencing building
design, its processes and practices, by blurring the
distinctions between geometry and analysis, between
appearance and performance. These topics were discussed and fully exemplified with all the group of
students, through many topics such as sustainable
design, performative environments and efficiency and
optimization.
For the last phase and final project, the students were instructed to create a wrap on the
existing building where the Architecture and Urban
Design Course is located. The proposed design should
improve thermal and/or acoustical conditions of the
building as well as its energetic efficiency, in a collaborative design process. In order to perform a virtual
simulation, a full digital 3D ArchiCAD model of the
existing building was generated and software such
as Ecotect, Ecodesign and Energy Plus were helpful
during decision-making process (Figure 3).
Besides, a physical simulation was also required
aiming the use of the wind tunnel and heliodon
solar simulator to confirm their first results reached
with virtual simulations. Rapid prototyping techniques
were very important in this phase where a physical
model was required to test their solutions. To do so, a
1:250 scale model was prototyped in SLS technique
(Figure 4), gently cooperated by CTI Renato Archer.
The students then had to produce their proposed
solutions using LAPACs facilities and the SLS model
as a matrix. As seen in Figure 5, the teams of students
came up with different solutions for the same building.
The use of the laser cutter and the 3D Printer were useful to materialize the presented solutions. Among the
used materials were acrilic, wood and card board paper.
DISCUSSIONS
The discipline results show that the use of computational techniques for evaluation and form generation,
847
showed that it is possible, and very important, to perform simulations on models with direct impact on
design decisions.
Another major obstacle was the reuse of the digital model. The digital model of the existing building
was created in ArchiCAD BIM authority software.
This model, which was made available to students,
has been less used in the evaluation stage for the following reasons: students had difficulties in exporting
the model into a format that could be used with other
software (such as EcotectgbXML); some evaluation
software used a format that is not interoperable, which
required the reconstruction of the building model; in
some cases, in which the model was exported, part of
the information (such as properties of objects) were
lost, requiring a partial rework.
On the other hand, with the building digital
model, it was easily possible, after minor changes in
some parameters, to export to the rapid prototyping
machines. That is, throughout the information model
(in this case just the geometry) generated in BIM
authoring tools, a model for rapid prototyping could
be generated.
Figure 4. SLS prototyped model.
ACKNOWLEDGMENTS
The authors would like to thank the following research
agencies for their collaboration in this research:
FAPESP, CAPES, CNPq, FAPEAL and CTI Renato
Archer.
Figure 5. Prototyped proposed solutions.
REFERENCES
combined with rapid prototyping can serve as an

important mean for the generation and transformation of architectural form. The use of some digital
tools requires the definition of a series of guidelines
and design rules. These are based on systematic and
explicit knowledge. Rapid prototyping allows the students to view models and discuss deeper tectonics
aspects. The discussions over these models allow the
exploration of subjective aspects of architecture, using
implicit and creative processes.
The combination of analytical techniques based on
computer models with digital modeling techniques
presented in this rich experience as a tool for design
practice, has allowed students to explore the process
of performance-based form generation.
Despite the advantages envisioned in this experience, some obstacles hindered its vigor. The first major
obstacle was observed in the training of students in
the use of software modeling and analysis. Being a
software requiring specialized knowledge, it was difficult to carry out systematic assessments and that
could effectively contribute to the transformation of
architectural form as a whole. Even so, the experience
Kolarevic, B. 2007. Architecture in the Digital Age: Design

and Manufacturing. Oxford: Taylor & Francis.
Kolarevic, B. & Malkawi, A. 2005. Performative Architecture: Beyond Instrumentality. New York & London: Spon
Press.
Liou, F. 2008. Rapid Prototyping and engineering applications: a toolbox for prototype development. Boca Raton:
Taylor & Francis Book.
McMillan, J. 1992. Strategies and Managers How managers
can use game theory to make better business decisions.
Oxford: Oxford University Press.
Mitchell, W. 1975. The theoretical foundation of computeraided architectural design. Environment and Planning B
volume 2: 127150.
Oxman, R. 2001. Digital architecture as a challenge for design
pedagogy: theory, knowledge, models and medium.
Design Studies, volume 29, n. 2: 99120.
Oxman, R. 2006. Theory and design in the first digital age.
Design Studies, volume 27, n. 3: 229265.
Tessmann, O. 2008. Collaborative Design Procedures for
Architects and Engineers. 1st. ed. [S.I.]: Books on Demand
GmbH, 176.
Sass, L.& Oxman, R. 2006. Materializing design: the implications of rapid prototyping in digital design. Design
Studies. volume 27: 325355.
848
Bartolo FINAL 08-09-11 13:47 Pagina 1
Brtolo
et al.
Paulo Brtolo is Professor of Advanced Manufacturing Processes at the

Polytechnic Institute of Leiria (IPL), Director of the Centre for Rapid and
Sustainable Product Development of IPL and President of the Scientific Council
of the Institute for Post-graduation Studies and Research of IPL.
Paulo Brtolo is Editor-in-Chief of the Journal Virtual & Physical Prototyping,
published by Taylor & Francis (ISSN 1745-2759)
an informa business
Prototyping
Innovative Developments in Virtual and Physical Prototyping presents essential

research in the area of Virtual and Rapid Prototyping. The volume contains
reviewed papers presented at the 5th International Conference on Advanced
Research in Virtual and Rapid Prototyping, hosted by the Centre for Rapid and
Sustainable Product Development of the Polytechnic Institute of Leiria, Portugal,
from September 28 to October 1, 2011. A wide range of topics is covered, such
as CAD and 3D Data Acquisition Technologies, Additive and Nano Manufacturing
Technologies, Rapid Tooling & Manufacturing, Biomanufacturing, Materials for
Advanced Manufacturing Processes, Virtual Environments and Simulation,
Applications of Virtual and Physical Prototyping Technologies.
Innovative Developments in Virtual and Physical Prototyping is intended for
engineers, designers and manufacturers who are active in the areas of
mechanical, industrial and biomedical engineering.
Prototyping
Edited by
Paulo Jorge Brtolo et al.

Innovative Developments

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Innovative Developments

Hochgeladen von

Copyright:

Verfügbare Formate

Bartolo FINAL 08-09-11 13:47 Pagina 1

Paulo Brtolo is Professor of Advanced Manufacturing Processes at the

Innovative Developments in Virtual and Physical Prototyping presents essential

Paulo Jorge Brtolo et al.

INNOVATIVE DEVELOPMENTS IN VIRTUAL AND PHYSICAL PROTOTYPING

This page intentionally left blank

PROCEEDINGS OF THE 5TH INTERNATIONAL CONFERENCE ON ADVANCED RESEARCH

Innovative Developments in Virtual

Henrique de Amorim Almeida

Centre for Rapid and Sustainable Product Development

ISBN: 978-0-415-68418-7 (Hbk)

Innovative Developments in Virtual and Physical Prototyping Brtolo et al. (eds)

Additive Manufacturing-assisted scaffold-based Tissue Engineering

Individual contour adapted functional implant structures in Titanium

New approaches to prototype 3D vascular-like structures by additive layer manufacturing

Fabrication and characterization of biodegradable composite scaffolds for Tissue Engineering

Heterogeneous tissue scaffolds for spatiotemporally controlled release kinetics

Medical application of rapid prototyping in orthopedics surgical planning

Polycaprolactone-based scaffold plus BMP-2 in a sheep thoracic spine fusion model

Automation design and simulation of a stent

Production and in-vitro characterization of micro-structured implant surfaces

A study of mechanical and biological behavior of porous Ti6Al4V fabricated on EBM

Enabling technologies for robotic organ printing

The semi-automated design & manufacture of patient-specific intervertebral disc implants

Additive manufacturing of soft tissue geometries for reconstruction purposes

Development of patient-specific implants using Direct Metal Laser Sintering in Titanium

CAD and 3D Data Acquisition Technologies

Reverse engineering of casting equipment for process simulation

Patient specific parametric geometric modelling of cementless hip prosthesis

Cross-section morphological study

Cellular structure design for lightweight components

Multi-material blending for complex-shaped heterogeneous objects

Application of contour tracing algorithm for assisting non-contact data acquisition

Image-based direct slicing of a single line drawing for rapid prototyping

Design and additive manufacturing of cellular lattice structures

Robust STL processing for extrusion-based manufacturing

Influence of hygrothermal aging on the mechanical properties of Nylon 12 composites processed

Investigation on the inclusions in maraging steel produced by Selective Laser Melting

SMARTerials for high performance buildings

Porous ceramic filters through 3D printing

Production of alumina parts through Selective Laser Sintering of alumina-polyamide

Material characterisation of Additive Manufacturing components made from a polyetherketone

Material study of laser cladded Inconel 625

A preliminary investigation on Selective Laser Melting of M2 high speed steel

Fatigue behavior of additive manufacturing parts. A preliminary analysis

Characterisation of carbon fibre reinforced nylon-12 composites for selective laser

New material development for laser additive manufacturing

Mechanical behavior of Epoxy-Aluminum composite for rapid tools applications

Thermal characterization of laser sintering of nylon-12

Metallurgical, accuracy and cost analysis of Ti6Al4V dental coping fabricated

Thermal and dynamic-mechanical behavior of Fullcure 3D Printing Resin post-cured by

Rapid Tooling & Manufacturing

Rapid tooling in metal forming processes using 3D-printed tools

Conformal cooling in moulds with special geometry

Thermoplastic Resin Transfer Moulding in a rapid manufactured mould

Fabrication of a Biopsy Micro-Forceps prototype with Incremental Sheet Forming

Using additive manufactured tooling in the fabrication of poly (L-Lactide-co-Glycolide) implants

Developing Ti jewelry through additive manufacturing and conversion technologies

Automation for building manufacturing

Additive and Nano Manufacturing Technologies

A novel approach to produce Functionally Graded Materials for additive manufacturing

Electrochemical micromilling with ultra short pulses

Functionally Graded Rapid Prototyping