High Performance Structures and Materials III PDF

High Performance
Structures and
Materials III
WIT Press publishes leading books in Science and Technology.

Visit our website for the current list of titles.
www.witpress.com
WITeLibrary
Home of the Transactions of the Wessex Institute.
Papers presented at High Performance Structures and Materials III
are archived in the
WIT eLibrary in volume 85 of WIT Transactions on
The Built environment (ISSN 1743-3509).
The WIT eLibrary provides the international scientific community with immediate
and permanent access to individual papers presented at WIT conferences.
http://library.witpress.com
Transactions Editor
Carlos Brebbia
Wessex Institute of Technology
Ashurst Lodge, Ashurst
Southampton SO40 7AA, UK
Email: carlos@wessex.ac.uk
WIT Transactions on The Built Environment

Editorial Board
E Alarcon
Universidad Politecnica de Madrid
Spain
C Alessandri
Universita di Ferrara
Italy
S A Anagnostopoulos
University of Patras
Greece
E Angelino
A.R.P.A. Lombardia
Italy
H Antes
Technische Universitat Braunschweig
Germany
D Aubry
Ecole Centrale de Paris
France
D E Beskos
Greece
J J Bommer
Imperial College London
UK
F Butera
Politecnico di Milano
Italy
P G Carydis
National Technical
University of Athens
Greece
J Chilton
University of Nottingham
UK
S Clement
Tranport System Centre
Australia
M C Constantinou
State University of New York at Buffalo
USA
G Degrande
Katholieke Universiteit
Leuven
Belgium
A De Naeyer
Universiteit Ghent
Belgium
W P De Wilde
Vrije Universiteit Brussel
Belgium
J Dominguez
University of Seville
Spain
F P Escrig
Universidad de Sevilla
Spain
M N Fardis
Greece
C J Gantes
National Technical
Greece
L Gaul
Universitat Stuttgart
Germany
Y Hayashi
Nagoya University
Japan
M Iguchi
Science University of Tokyo
Japan
L Int Panis
VITO Expertisecentrum IMS
Belgium
W Jager
Technical University of Dresden
Germany
C M Jefferson
University of the West of
England, Bristol
UK
D L Karabalis
Greece
E Kausel
Massachusetts Institute of
Technology
USA
K Kawashima
Tokyo Institute of Technology
Japan
A N Kounadis
National Technical
Greece
W B Kratzig
Ruhr Universitat Bochum
Germany
A A Liolios
Democritus University of
Thrace
Greece
J W S Longhurst
University of the West of England,
Bristol
UK
J E Luco
University of California at
San Diego
USA
L Lundqvist
Unit for Transport and Location Analysis
Sweden
M Majowiecki
University of Bologna
Italy
G D Manolis
Aristotle University of Thessaloniki
Greece
G Mattrisch
DaimlerChrysler AG
Germany
F M Mazzolani
University of Naples "Federico II"
Italy
K Miura
Kajima Corporation
Japan
G Oliveto
Universit di Catania
Italy
E Oate
Universitat Politecnica de
Catalunya
Spain
A S Papageorgiou
Rensselaer Polytechnic Institute
USA
G G Penelis
Aristotle University of
Thessaloniki
Greece
A M Reinhorn
State University of New York at Buffalo
USA
F Robuste
Universitat Politecnica de
Catalunya - LAMOT
Spain
C W Roeder
University of Washington
USA
J M Roesset
Texas A & M University
USA
M Saiidi
University of Nevada-Reno
USA
F J Sanchez-Sesma
Instituto Mexicano del
Petroleo
Mexico
S A Savidis
Technische Universitat Berlin
Germany
J J Sendra
Universidad de Sevilla
Spain
Q Shen
Massachusetts Institute of Technology
USA
A C Singhal
Arizona State University
USA
P D Spanos
Rice University
USA
C C Spyrakos
National Technical
Greece
H Takemiya
Okayama University
Japan
I Takewaki
Kyoto University
Japan
E Taniguchi
Kyoto University
Japan
J L Tassoulas
University of Texas at Austin
USA
M A P Taylor
University of South Australia
Australia
R Tremblay
Ecole Polytechnique
Canada
R van der Heijden

Radboud University
Netherlands
R van Duin
Delft University of
Technology
Netherlands
A Yeh
The University of Hong Kong
China
M Zador
Technical University of
Budapest
Hungary
R Zarnic
University of Ljubljana
Slovenia
THIRD INTERNATIONAL CONFERENCE ON

HIGH PERFORMANCE STRUCTURES AND MATERIALS
HIGH PERFORMANCE
STRUCTURES AND MATERIALS III
CONFERENCE CHAIRMAN
C.A. Brebbia
Wessex Institute of Technology, UK
INTERNATIONAL SCIENTIFIC ADVISORY COMMITTEE

K.S. Al Jabri
B. Alzahabi
W.P. De Wilde
G. Belingardi
M. Ingber
T. Katayama
S. Kravanja
H. Martikka
R.A.W. Mines
F. Romano
R. Schmidt
H. Takagi
I. Tsukrov
Organised by:
and
Free University of Brussels, Belgium
Sponsored by:
The High Performance Structures and Materials Book Series
This page intentionally left blank
High Performance
Structures and
Materials III
Editor:
C.A. Brebbia
C.A. Brebbia
Published by
WIT Press
Ashurst Lodge, Ashurst, Southampton, SO40 7AA, UK
Tel: 44 (0) 238 029 3223; Fax: 44 (0) 238 029 2853
E-Mail: witpress@witpress.com
http://www.witpress.com
For USA, Canada and Mexico
Computational Mechanics Inc
25 Bridge Street, Billerica, MA 01821, USA
Tel: 978 667 5841; Fax: 978 667 7582
E-Mail: infousa@witpress.com
British Library Cataloguing-in-Publication Data
A Catalogue record for this book is available
from the British Library
ISBN: 1-84564-162-0
ISSN: 1746-4498 (print)
ISSN: 1743-3509 (on-line)
The texts of the papers in this volume were set
individually by the authors or under their supervision.
Only minor corrections to the text may have been carried
out by the publisher.
No responsibility is assumed by the Publisher, the Editors and Authors for any injury and/or
damage to persons or property as a matter of products liability, negligence or otherwise, or from
any use or operation of any methods, products, instructions or ideas contained in the material
herein.
WIT Press 2006
Printed in Great Britain by Cambridge Printing.
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system,
or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, or
otherwise, without the prior written permission of the Publisher.
Preface
This book contains the edited papers presented at the Third International
Conference on High Performance Structures and Materials, held in Ostende,
Belgium in 2006.
The Conference was dedicated to honouring the distinguished career of
Professor Patrick De Wilde, from the Free University of Brussels, Belgium, on the
occasion of his retirement.
The Conference addressed issues involving advanced types of structures,
particularly those based on new concepts or new types of materials. This responds
to the need to develop a generation of new materials that are suitable for high
performance structures which can easily resist a wide range of external stimuli and
react in a non-conventional manner.
Professor De Wilde, whose achievements are summarised in the following
pages, was the originator and driving force behind this Conference series since it
started.
The Editor
Ostende, 2006
Dedication
Professor W. Patrick De Wilde
This book is in homage to the long and
distinguished career of my friend and
colleague Professor Patrick De Wilde from
the Vrije Universiteit Brussel (VUB).
I was fortunate enough to meet Patrick
for the first time many years ago when we
were both young researchers. This was
on the occasion of the Seminar on Finite
Elements that I organised in Southampton
in 1970. He was then a Research
Assistant at the Department of Applied
Continuum Mechanics of VUB and I had
just been appointed a lecturer at the
Department of Civil Engineering at the
University of Southampton. Since then we kept in constant contact and this led me
to a better appreciation of his outstanding scientific career and the depth of his
knowledge and imagination. Patrick always brought a new and fresh perspective to
the most challenging engineering and scientific problems.
His academic career has been centred at VUB where he obtained his PhD in
Applied Sciences in 1976, rising from lecturer to Head of the Laboratory of Structural
Analysis and to full Professor in 1984. He also served with distinction as Chairman
of the Research Council and Vice Rector of VUB, and is now about to retire from the
position of Head of the Department of Mechanics of Materials and Constructions.
Patrick has always been characteristically generous with his knowledge,
lecturing part time at many institutions in Belgium and abroad. An appointment
that reflects his range of interests was as a Professor at the Higher Institute of
Architecture Victor Horta, the link between architecture and engineering being
always close to his heart. He also contributed substantially to the enhancement of
training and research capabilities in South American and African countries to mention
but a few of his international activities.
Always prepared to share his time, he headed up and motivated several

generations of young researchers at VUB as well as contributing to the development
of many institutions around the world, including our own Wessex Institute of
Technology where he has served on the Board of Directors since its beginning.
Patricks written contributions comprise many papers in learned journals and
conference meetings in the field of Computer Aided Engineering, finite element
methods; composite materials, mechanics of materials and structures. He has
participated in many scientific committees and in the advisory board of different
publications and international conferences. He is permanent Co-chairman of the
international conference on High Performance Materials and Structures as well as
one of the most active members of the Scientific Committee of the Conference on
Structural Repairs and Maintenance of Historical Buildings.
In spite of his substantial research activities he has not neglected to collaborate
with industrial partners and as a result has carried out consulting work for different
companies and is currently Chairman of the Technical Board of SECO (Civil
Engineering Certification Company).
It has been a privilege for me as well as for many other colleagues around the
world to be associated with Patrick all the years he has been working at VUB. Soon
his teaching days will be over and we hope this will herald the beginning of a new
period for him: concentrating on unfinished research and original work which was
interrupted by the demands of university duties.
Carlos A. Brebbia
The New Forest, 2006
Contents
Section 1: Conceptual design and structural analysis
(Special session organised by W. P. De Wilde)
Conceptual design of lightweight structures:
the role of morphological indicators and the structural index
W. P. De Wilde ......................................................................................................3
Influence of dynamic loads on the optimum design of trusses
J. Van Steirteghem, W. Ponsaert, W. P. De Wilde & Ph. Samyn........................13
Modular grid-based design concept for fibre reinforced composite shells
E. De Bolster, H. Cuypers, W.P. De Wilde & J. Wastiels...................................21
Influence of stiffness constraints on optimal design of trusses
using morphological indicators
T. Vandenbergh, W.P. De Wilde, P. Latteur, B. Verbeeck,
W. Ponsaert & J. Van Steirteghem......................................................................31
Variations in form and stress behaviour of a V-shaped membrane
in a foldable structure
M. Mollaert, N. De Temmerman & T. Van Mele ................................................41
Section 2: Composite materials and structures
A new composite material based on natural fibres and a thermoset:
technology, applications and properties
G. Wuzella ...........................................................................................................53
Experimental test of threaded steel rods glued-in hardwood with epoxy
D. Otero, J. Estvez, E. Martn & J. A. Vzquez.................................................63
Predicting the mechanical behaviour of large composite rocket motor cases
N. Couroneau ......................................................................................................73
Micromechanical modeling of random or imperfect composites

M. ejnoha & J. Zeman.......................................................................................83
Flexural behaviour of ferrocement roof panels
A. S. Alnuaimi, A. Hago & K. S. Al Jabri............................................................93
Finite element modeling of actuated fibre composites
M. Martinez, A. Artemev, F. Nitzsche & B. Geddes..........................................103
CFRP strengthening of prefabricated timber panel walls
M. Premrov & P. Dobrila .................................................................................111
Evaluation of the structural integrity of a sandwich composite
train roof structure
K. B. Shin, B. J. Ryu, J. Y. Lee & S. J. Lee........................................................121
Measurement of the fiber stress distribution during pull-out test
by means of micro-Raman spectroscopy and FEM analysis
K. Tanaka, K. Minoshima & H. Yamada ..........................................................131
Section 3: Natural fibre composites
(Special session organised by T. Katayama and H. Takagi)
Effect of surface treatment to tensile static and creep properties
for jute fiber reinforced composite
K. Takemura......................................................................................................143
Effects of forming conditions on mechanical properties of
resinless bamboo composites
H. Takagi & H. Mori.........................................................................................151
Compression moulding of jute fabric reinforced thermoplastic composites
based on PLA non-woven fabric
T. Katayama, K. Tanaka, T. Murakami & K. Uno ...........................................159
Quality control of fibers end-milled from bamboo pipe
using spiral tool path
K. Ogawa, E. Aoyama, T. Hirogaki, Y. Tomioka & H. Nakagawa ...................169
Characteristic behaviors of CFRP and GFRP at cryogenic
temperature under static and cyclic loadings
S. Kubo, K. Okubo & T. Fujii............................................................................179
Mechanical properties of loosing natural fiber reinforced polypropylene
K. Mizuta, Y. Ichihara, T. Matsuoka, T. Hirayama & H. Fujita.......................189
Section 4: Material and mechanical characterisation

Identification of strain-rate sensitivity parameters of steel sheet by
genetic algorithm optimisation
G. Belingardi, G. Chiandussi & A. Ibba ...........................................................201
Mechanical characterisation of a viscous-elastic plastic material,
sensitive to hydrostatic pressure and temperature
V. D. Le, M. Caliez, M. Gratton, A. Frachon & D. Picart................................211
Identification of the material properties of composite beams:
inverse method approach
E. Euler, H. Sol & E. Dascotte..........................................................................225
Full-field optical measurement for material parameter identification
with inverse methods
J. Gu, S. Cooreman, A. Smits, S. Bossuyt, H. Sol, D. Lecompte
& J. Vantomme..................................................................................................239
Multiaxial characterization of the mechanical behaviour of
aluminium foam
L. Peroni, M. Avalle & P. Martella...................................................................249
Characterisation of the high strain rate properties
of Advanced High Strength Steels
J. Van Slycken, P. Verleysen, J. Degrieck & J. Bouquerel ...............................259
Evaluation of bond strength in Roller Compacted Concrete
under various normal pressures
M. Madhkhan & A. Arasteh ..............................................................................269
High performance fibres and the mechanical attributes
of cut resistant structures made therewith
S. Rebouillat & B. Steffenino.............................................................................279
The study of surface oxidation of tin(II) fluoride and chloride fluoride
materials by Mssbauer spectroscopy: to oxidize or not to oxidize,
that is the question
G. Dns, E. Laou, M. C. Madamba & A. Muntasar ........................................301
Study of double ionic disorder (cationic and anionic) and disorder
of two kinds of tin(II) (ionic and covalent) within the same material:
the incredibly complex Ba1-xSnxCl1+yF1-y solid solution
and its study by Mssbauer spectroscopy
S. Boufas, G. Dns, J. Kochuparampil, H. Merazig & A. Muntasar...............311
Study on fatigue and energy-dissipation properties of

nanolayered Cu/Nb thin films
Y.-C. Wang, T. Hoechbauer, J. G. Swadener, T. Darling, A. Misra,
R. Hoagland & M. Nastasi................................................................................323
Advances in computational modeling through the use of
higher-level microstructure characterization
M. Groeber, M. Uchic, D. Dimiduk, Y. Bhandari & S. Ghosh .........................331
Deformation of aluminum alloys AD-1, AMg-6 and D-16
at dynamic compression and temperatures of 25250oC
V. A. Pushkov, S. A. Novikov, V. A. Sinitsyn, I. N. Govorunov
& O. N. Ignatova...............................................................................................343
X-ray residual stress measurements on plasma sprayed
molybdenum coatings
K. Hirukawa, K. Akita, S. Tobe, T. A. Stolarski & S. Ohya ..............................351
Compressive residual stress generation process by laser peening
without pre-coating
H. Tanaka, K. Akita, Y. Sano & S. Ohya...........................................................359
The scale effect of roughness in contact problems
S. Mezghani, A. Jourani & H. Zahouani...........................................................369
Section 5: High performance concretes
Copper slag as fine aggregate for high performance concrete
K. S. Al Jabri .....................................................................................................381
Application of FRC in tunnel reinforcement
P. Prochzka .....................................................................................................391
Innovative procedure to produce high performance pretensioned
concrete girders combining high strength concrete and
normal or special concrete types
C. Vzquez-Herrero & F. Martnez-Abella.......................................................401
Properties of high performance concrete: the effect of cracks
E. Mahonkov, M. Jiikov & R. ern ....................................................409
Seismic upgrading of square and rectangular RC columns using
FRP wrapping
M. A. N. Abdel-Mooty, M. E. Issa, H. M. Farag & M. A. Bitar........................419
Arch bridge made of reactive powder concrete

D. Cizmar , D. Mestrovic & J. Radic ................................................................429
Elastic and inelastic seismic response comparison of
reinforced concrete buildings with normal resistance concrete and
with high resistance concrete
J. A. Avila & D. Rivera .....................................................................................439
The strength effects of synthetic zeolites on properties
of high performance concrete
P. Frontera, S. Marchese, F. Crea, R. Aiello & J. B. Nagy ..............................449
Behaviour model and experimental study for the torsion
of reinforced concrete members
C. E. Chalioris ..................................................................................................459
A study on use of blended ferrocement: a high performance material
for repair/strengthening of brick masonry columns
T. Kibriya ..........................................................................................................469
Section 6: Damage and fracture mechanics
Crush behaviour of open cellular lattice structures manufactured
using selective laser melting
M. Santorinaios, W. Brooks, C. J. Sutcliffe & R. A. W. Mines..........................481
Stress intensity factors for cracked cold-drawn steel wires
under tensile loading
B. Lin & G. Lu...................................................................................................491
Void growth and damage ahead of a crack in pressure-sensitive
dilatant polymers
H. B. Chew, T. F. Guo & L. Cheng ...................................................................501
An orthotropic damage model for crash simulation of composites
W. Wang, F. H. M. Swartjes & M. D. Gan .......................................................511
Ring-shaped crack propagation in a cylinder under nonsteady cooling
V. A. Zhornik, Yu. A. Prokopenko, A. A. Rybinskaya & P. A. Savochka...........521
Investigation of the hygrothermal performance of wooden beam ends embedded
in inside insulated outside walls
H. Stopp & P. Strangfeld ..................................................................................529
Bond repair of cracked beams

H. Cruz, J. Custdio & D. Smedley ..................................................................539
Microtexture and nanoindentation study of delamination cracking
in Al-Cu-Li-X alloys
R. Crooks, M. S. Domack & J. A. Wagner ........................................................549
A contribution to the rehabilitation of reinforced concrete structures by
non-destructive electrochemical methods
J. L. Rovira Santa Olaya & P. Pardo Trfach..................................................559
Section 7: Adhesion and adhesives
Timber specimens parametrized design for numerical analysis
E. Martn Gutirrez, J. Estvez Cimadevila, D. Otero Chans
& S. Muiz Gmez ............................................................................................571
Theory of the elasticity of the materials of the second order
V. Shorkin & V. Gordon....................................................................................581
Section 8: Optimal design
Optimal design of fibre reinforced tubular structures
H. Martikka & E. Taitokari...............................................................................593
MINLP optimization of steel frames
S. Kravanja & U. Klanek.................................................................................605
Optimization of timber trusses considering joint flexibility
S. ilih, M. Premrov & S. Kravanja..................................................................615
The HendrickxVanwalleghem design strategy
W. Debacker, C. Henrotay, W. P. De Wilde & H. Hendrickx...........................625
The optimization of a truss facade
B. Verbeeck,W. P. De Wilde & Ph. Samyn .......................................................635
MINLP optimization of the single-storey industrial building steel structure
T. ula, U. Klanek & S. Kravanja ...................................................................643
Genetically optimised placement of piezoelectric sensor arrays:
linear and nonlinear transient analysis
J. N. Rao, S. Lentzen & R. Schmidt ...................................................................653
Numerical analysis of the process of trapezoidal thread rolling

L. Kukielka & K. Kukielka ................................................................................663
Three-dimensional limit analysis of ancient masonry buildings
with rigid block models
A. Ordua..........................................................................................................673
Section 9: Reliability of structures
Analysis of diffusional stress relaxation in submicron Cu
interconnect structures using the model with enhanced
vacancy diffusivity in grain boundary region
I. Tsukrov, W. M. Grich & T. S. Gross..............................................................685
Application of fuzzy sets to structural reliability of existing structures
I. Mura ..............................................................................................................695
Non-linear response of combined system, 3D wall panels and
bending steel frame subjected to seismic loading
M. Z. Kabir, A. R. Rahai & Y. Nassira..............................................................705
Section 1
Conceptual design and
structural analysis
(Special session organised by
W. P. De Wilde)
High Performance Structures and Materials III
Conceptual design of lightweight structures:

the role of morphological indicators and
the structural index
W. P. De Wilde
Department of Mechanics of Materials and Constructions (MeMC),
Vrije Universiteit Brussel, Brussels
Abstract
Firstly, but in no way exclusively, this paper addresses architectural engineers
facing critical design decisions in the phase of so-called conceptual design.
The resulting design must yield a structure showing a sound behaviour in both
the serviceability limit state (SLS) and in the ultimate limit state (ULS), also
meaning that three essential criteria should be satisfied: strength, stiffness and
stability (the latter including, if relevant, an acceptable dynamic behaviour).
However, the question very often remains open as to which of the three criteria is
overruling the other ones. This paper, a synthesis of the work of a research group
headed by the author, tries to show that a conceptual design methodology can be
developed, hereby using the concept of morphological indicators (originally
developed by P. Samyn, subsequently P. Latteur and within the research group)
and a structural index (introduced by Shanley). This methodology also gives an
answer to the question that very often arises when designing lightweight
structures: design for strength or design for stiffness? Examples in
subsequent papers, presented during this Conference and thus included in these
Proceedings, illustrate the methodology nowadays used by our students in
architectural engineering when designing and analysing their structures.
Keywords: conceptual design of architectural structures, strength, stiffness,
structural stability, structural vibrations.
Morphological indicators
The so-called morphological indicators (MI) were introduced by P. Samyn [1] in

1999. As different papers, both within this conference and in other journals and
WIT Transactions on The Built Environment, Vol 85, 2006 WIT Press
www.witpress.com, ISSN 1743-3509 (on-line)
doi:10.2495/HPSM06001
4 High Performance Structures and Materials III

books (see e.g. a more recent book of P. Samyn [2]), have already or will
underline(d) the potentials of these dimensionless numbers, and thus we will
restrict ourselves to a very short introduction.
P. Samyn essentially introduced two indicators, one related to the minimum
volume of material required for a structural typology achieving a fully stressed
design, the other related to the maximum displacement in the same structure. The
combination of those two indexes, being the volume indicator W and the
displacement indicator allow the designer to select not only a (sub)optimal
typology but also its optimal aspect ratio.
The indicators are dimensionless numbers, function of very few parameters,
the most important being the so-called slenderness of the structure L/H, in which
L is the larger and H the lesser dimension of a window, framing the structure.
P. Samyn defined them as:
V
a) W=
, in which is the admissible stress (in practice we consider the
FL
allowable stress in the serviceability limit state (SLS), V the volume of material,
F the resultant of (static) forces, loading the structure, and L its span.
E
, in which E is the elastic modulus and the maximum displacement.
b) =
L
W = V/FL
MULTI-WARREN trusses
WARREN trusses
ARCH + TIE + SW
1,700
ARCH + TIE + PR
1,414
ARCH + TIE
ARCH + SW
1,155
ARCH + PR
0,816
ARCH
0,707
0,577
PR = push rods
SW = suspension wires
1,633
2,309
2,828
3,138
3,266
L/H
Figure 1.
He also shows that they can be expressed as a function of L/H: W = W(L /H)
and = (L /H). This allows one to draw, either analytically or numerically,
diagrams showing the values of the indicators in function of the slenderness, for
different typologies of structures. The recent book of P. Samyn [2], although not
exhaustive, contains an impressive amount of typologies that can be compared:
see fig.1 as an example of a comparison of different typologies through the
volume indicator: this diagram is taken from [9]. Similar graphs were developed
for the displacement indicator .
It is clear that P. Samyn see the title of his thesis hereby provided the
architect with a tool allowing to reach a suboptimal design at the stage of the
conceptual design. The fact that he is still using it today, and undoubtedly
producing designs of outstanding quality, proves the robustness of the tools he
introduced.
However, two major objections could be foreseen and they were very soon
subject of controversy. Indeed, the two indicators allow for a preliminary design,
achieving the required performances of strength and stiffness with a minimum
volume of material (a fully stressed design of statically determinate structures,
subject to classical load cases)...but what about (in)stability? It is clear that, in its
most simple form, the developed method does not take at least explicitly - into
account possible buckling phenomena. However it would be unjust to consider
that it completely overlooks the existence of this phenomenon: in both [1] and
(more detailed) in [2], P. Samyn shows that correction factors can be
computed, thus defining an increase of material consumption, and trying to
control the stability of the equilibrium.
Nevertheless, gradually we were convinced that conceptual design should
take into account the totality of the criteria to be satisfied by the structure:
the strength of the structural parts is controlled through the indicator of

V
, as it starts from a fully stressed design (at a stress
volume: W =
FL
level ). In the conceptual design stage, a minimal value of this
indicator is aimed at, thus trying to achieve a minimal consumption of
material. But, as we shall experience very quickly, a mere choice of the
slenderness L/H, corresponding to a minimum of the curve for the
chosen typology (see e.g. fig.1) does not solve the problem! The reason
is simple: this simple minimisation does not take into account the
stiffness, stability...and dynamic requirements of the design. In this
sense we could say that each of the three other criteria introduce kinds
of forbidden zones in the W=W(L/H) diagrams: a forbidden zone for
the excessively flexible structures, one for the unstable structures and a
last one if relevant for the unacceptable vibrations.
the stiffness of the structure is evaluated through the indicator of
E
. As one can see, it is proportional to the ratio
displacement: =
L
/L, which is generally limited by design codes (see e.g. Eurocodes). As
we will show later this indicator, originally introduced by P. Samyn will
be of primordial importance when evaluating the risk for resonance of a
structure.
through the buckling indicator =
, introduced by Latteur [3], in

qEF
which is the length reduction factor (due to end conditions of the
I
elements), and q =
in which I is the (minimal) inertia moment of
2
the section and its section, one can evaluate the sensitivity of the
structure to (local) buckling: the higher , the higher the risk. We are
convinced that this instrument should be used in conjunction with the
other (W, ,), as it gives a better estimate of the penalty in material
consumption than the correction factors used by P. Samyn. To be
mentioned is that J. Van Steirteghem shows in his thesis work [10] that
one can evaluate the risk for global instability, starting from this
indicator.
Figure 2.
Eventually one also has to consider an indicator describing the dynamic
behaviour of the structure. This is achieved through the use of the indicator of
L
1
= f , , extensively discussed by J. Van
the first natural frequency =
H
Steirteghem in [10]. As one sees, there is a direct link with the indicator of
displacement . This indicator is also directly dependent on the buckling
indicator, which is not surprising: both resonance and buckling phenomena in
elastic systems are related with transformation of energy (compression into
bending or torsion for buckling, potential into kinetic for resonance). An
important observation can be made here: if one accepts all the simplifications
introduced by the concept of morphological indicators and if one has a closer
look at the indicator , one will notice that it is independent on the volume
indicator and thus on the mass of the system. This could bring one to the
conclusion that at least, at the stage of conceptual design there is little hope
that one can improve the dynamic behaviour of the structure through simple
addition of mass. This has been confirmed when designing e.g. slender and
elegant footbridges: one quickly comes to the conclusion that the penalty in
mass, in order to achieve an acceptable spectrum of resonance frequencies is
(very) much higher than the one necessary to control its buckling behaviour (see
e.g. [10]).
A very well known example, proving this observation, is the 350m long
Millennium Bridge, designed by Sir Norman Foster in collaboration with Arup
Ltd. See e.g. http://www.arup.com/millenniumbridge/, site from which we quote:
...There are two fundamental ways to limit dynamic excitation:
* Stiffen the structure, so the frequency of the bridge and our footsteps no
longer match;
* Add damping to absorb the energy.
It was concluded that stiffening the bridge to change its frequency was not a
feasible option. The bridge would need to be at least tenfold stiffer laterally to
move its frequency out of the excitation range. The additional structure required
to do this would dramatically change the appearance of the bridge.
It was decided to adopt a damping solution, either active damping or passive
damping. Active damping uses powered devices to apply forces to the structure
to counteract vibrations. Passive damping relies on harnessing the movements of
the structure to absorb energy. (end of quote).
This is only one of the many examples proving that one has to give necessary
attention to the dynamic behaviour of lightweight structures, if one wants to
design an acceptable lightweight structure! But this also proves that solving the
problem with addition of mass and/or stiffness is very seldom the best option. A
new challenge thus for designers!
Design for stiffness versus design for strength
If one wants to consume a minimum of material it is logical that he follows a

strategy by which the indicator of volume W is set as the hierarchically most
important. Given a span L, loads F and material properties (among which
controls W and E controls ) one can select an acceptable slenderness L/H and
thus H for a chosen typology of structure. As P. Samyn very rightly points out in
[2], one can take into account the stiffness and even the stability requirements,
although one can argue that the latter requirement is addressed in a somewhat
empirical way.
However, from our experience and research in the Vrije Universiteit Brussel,
it appears that:
the strategy, putting forward the indicator of volume W as objective
function, although yielding performing and lightweight preliminary
designs, practically always produces solutions that have to be corrected
in order to satisfy the remaining requirements. As the objective is to
minimise volume, one very often achieves designs which are too
flexible, unstable...and, most of all, dynamically unacceptable.
it would be good to think about the strategy that consists in an a priori
selection of the indicator that should be selected as the hierarchically
more important: W versus , or strength versus stiffness. This

alternative is also linked to the two analyses in serviceability limit state
(SLS) and ultimate limit state (ULS). In fact, one quickly gets a feeling
that there are two extreme situations: on one hand the heavy and very
stiff structures (e.g. masonry bridges [11]), on the other hand the
lightweight and often rather flexible structures (e.g. the modern
footbridge, cf. Millennium Bridge in London). The experience acquired
during studio work with architectural engineers, who have to design
structures of all types, covering a wide spectrum of functionalities and
spans, has underlined that a measure of the danger for (= sensitivity to)
instability phenomena (both buckling and resonance) is the best guide
for the choice between design for strength and design for stiffness.
other empirical observations are important:
o when one foresees risks for instability phenomena it is better to
approach the problem through design for stiffness.
o when designing lightweight structures it is almost always the
design constraints in the SLS that prevail over those for the
ULS. In other words, these constructions have an enormous
reserve in strength, and they thus show a much higher value of
W than the optimal one!
o the penalty imposed by additional material consumption, in
order to avoid unacceptable vibrations, is much more severe
than the one controlling (in)stability.
one could thus look for a kind of separation line or a criterion
allowing the selection of the adequate strategy. It seems also that, in
order to take into account the scale effects, it is better to look for a
parameter which takes into account the absolute value of both span L
and forces F (hereby included the permanent forces on and self-weight
of the structure). Van Steirteghem, in his PhD work [10], suggests that
this parameter could be L / F , which in fact has been introduced by

Shanley [8], who called the inverse of its square (F/L2) the structural
index. Large values point to large span structures with relatively low
permanent loads, small values to the contrary. In another paper, here
presented, he will show the role played by the structural index in some
structural choices. This choice is by no means arbitrary as the indicators
for buckling and for the first resonance frequency can both be
linked to it: the former even explicitly, as it is proportional to the
structural index, the latter being an implicit function of .
This brings us to an important conclusion: at the stage of conceptual design
one should in the first place check which of the two design strategies, being
design for strength versus design for stiffness should be adopted. A
reasonable indication can be obtained by the analysis of the mentioned
morphological indicators, but most of all by the structural index. The latter will
direct you to a design in SLS rather than ULS, if risks for buckling and/or
resonance are to be dealt with.
Alternate strategies
Probably one of the most challenging and promising alternatives to the design
strategy described in the previous sections consists in the use of new
algorithms, allowing a search for improved typologies, especially in the case of
complicated load conditions. Instead of looking for the optimal typology of
structure, given the span L and the loads F, which yields an optimal slenderness
L/H for the catalogued typology (e.g. a Warren truss), one can open the search
space to yet unknown typologies. This can be achieved using e.g. genetic
algorithms, evolutionary strategies, etc. Verbeeck et al. show in [14] that genetic
algorithms can be very powerful tools in optimisation processes, but above all
have shown in [15] that they can yield new, often unsuspected, typologies, better
suited for the given problem. Important is to mention that the objective function
these algorithms try to minimize is the volume indicator W, the other
morphological indicators reducing the search space by inducing side constraints.
Preliminary conclusions
As will be shown by Vandenbergh et al. [12] in a subsequent paper, the

conceptual design of lightweight structures introduces the need for an a priori
selection of a design strategy based on the need to satisfy in the first place a SLS
or an ULS. This can be achieved through an adequate use of morphological
indicators, among which the structural index is also important: it is the only one
taking into account the so-called scale effects. J. Van Steirteghem has
convincingly shown in his Ph.D thesis [10] that it is of paramount importance to
include the buckling indicator as a full partner when designing structures subject
to dominant static loads. As the buckling indicator is directly related to the
structural index, it immediately gives indications about the limit state, either SLS
or ULS, in which the structure should be calculated. This choice also is linked to
the alternative between design for stiffness and design for strength.
The result of the minimisation of the volume indicator W, in an acceptable
design subspace, and in which the two other morphological indicators and
introduce side constraints, combined with the eagerness of modern architects to
show their ability to cross very large spans, very often yields designs of
structures showing an unacceptable dynamic behaviour. The penalty induced by
corrections on mass or stiffness practically always affects the optimum value of
W in a much more severe way than e.g. the need to satisfy buckling or stiffness
requirements. Ponsaert et al. [13] will show that the unacceptable vibrations can
be controlled through active, hybrid or passive damping, rather than through the
tuning of stiffness and/or mass.
Although it is perhaps a little early to make final statements, I strongly
suspect that the direction taken by my student B. Verbeeck in the field of new
numerical algorithms will yield interesting results. It is indeed quite logical to
think that, as long as you restrict your search for a minimum volume structure

within a given set of typologies, you could miss better solutions! I am thus
convinced that there is still an open field and want to explore it.
I am quite thrilled by the enthusiasm of the students who discover the
potentials of these indicators and indexes, when working for their studio work.
At least it proves that it can effectively be used at the stage of conceptual design.
And was that not the goal one tries to achieve as a teacher of structural design?
Acknowledgements
Scientific research is very rarely the output of a single brain, and it has not been
different with the present study. I thus want to pay due respect to colleagues and
students. In the first place, two very important contributors to my evolving
insights in structural design and analysis: Philippe Samyn, eminent architect and
engineer, and Pierre Latteur, a structural engineer with a real feeling for
structural behaviour. They triggered my interest for morphological indicators,
allowing me to direct several theses at both Masters and Doctoral level. All of
the others will recognise themselves, but I want to underline the outstanding
research work of Jan Van Steirteghem, architectural engineer, who sort of filled
some gaps left by the excellent work of P. Samyn and P. Latteur, and also
initiated work in the field of dynamics of lightweight structures: he allowed me
to come back to my real passion for structural vibrations.
During the last years I have been also collaborating with one of my former
students, Marijke Mollaert, who finished her Ph.D in 1984 on tensile structures
and, since then, has become a colleague and always has remained a close friend.
I still spend a lot of time discussing new ideas with her and her students, all of
them involved in this exciting field of form-active tensile structures. Although
the discipline of designing these structures is very much different from what we
are doing in my research group, the ideas one can pick up during those informal
meetings are refreshing and really exhilarating. Thank you, Marijke, for your
friendship.
WIT Conferences were also regular meeting points, confronting me with new
ideas and insights: first CADCOMP, later evolving into these HPSM
Conferences, OPTI, but also Structural Repair and Maintenance of Historic
Structures. In particular, I want to mention my friends Santiago Hernandez (U of
La Corua, Spain) and Jerry Connor (MIT, USA), with whom it is an
enlightening experience to discuss all kinds of matters, even structures!
Research needs funding and scholarships and I thus also want to acknowledge
the Institutions which helped this research financially: Fund for Scientific
Research in Flanders (FWO), Fund for Applied Research in Flanders (IWT), the
VUB and several industrial partners.
Last but not least a personal word of gratitude for Carlos A. Brebbia, always
showing more confidence in my skills than I deserve, and allowing me to meet
such fine persons within the Board of Directors of WIT. It is now about 35 years
that we met for the first time and it is always with extreme pleasure that we see
each other, either in the Lodge or at his Conferences.
11
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
Samyn, P., Etude compare du volume et du dplacement de structures

bidimensionnelles, sous charges verticales entre deux appuis vers un
outil dvaluation et de prdimensionnement des structures (4 vol.), Ph. D.
thesis, Universit de Lige, 1999.
Samyn, P., tude de la morphologie des structures laide des indicateurs
de volume et de dplacement, Acadmie royale de Belgique, Classe des
Sciences (2004), ISBN 0365-0952.
Latteur, P., Optimisation des treillis, arcs, poutres et cbles sur base
dindicateurs morphologiques application aux structures soumises en
partie ou en totalit au flambement (3 vol.), Ph. D. thesis, Vrije
Universiteit Brussel, 2000.
Schlaich, J., Bergermann, R., Leicht Weit, Light Structures, Prestel
(Mnchen, Germany), ISBN 3-7913-2918-9, 2005.
Latteur P., Samyn P. et De Wilde P., 2001, Optimization des arcs
paraboliques et en chanette aide la conception sur base dindicateurs
morphologiques. Revue franaise de Gnie Civil, Vol. 5, n 1.
Latteur P., Samyn P., De Wilde P., 2000, Comparaison des treillis
classiques type Warren, Pratt et Howe : optimization et
prdimensionnement sur base dindicateurs morphologiques. Revue
franaise de Gnie civil, Vol. 4, n 4.
Shea K., 1997, Essays of Discrete Structures: purposeful design of
grammatical structures by direct stochastic search, PhD Thesis, Carnegie
Mellon University, USA.
Shanley FR. Weight-Strength Analysis of Aircraft Structures. New York:
Dover, 1960.
Samyn P., Latteur P., Van Vooren, J., Volume of structures: application to
classical and harmonic structures, International IASS symposium on
Lightweight structures in Architecture, Engineering and Construction,
1998 October 5-9, Sydney, Australia.
Van Steirteghem J., A Contribution to the Optimisation of Structures
Using Morphological Indicators: (In)Stability and Dynamics, Ph.D. thesis,
Vrije Universiteit Brussel, Mechanics of Materials and Structures, W.P.
De Wilde, supervisor, 2006.
Audenaert A., Peremans H., De Wilde W.P., Static determination of the
internal forces and displacements in arch bridges, Masonry Society
Journal, vol.22, n1, pp.97-111, September 2004.
Vandenbergh Th., Van Steirteghem J., De Wilde W. P., Samyn P.,
Influence of stiffness constraints on optimal design of trusses using
morphological indicators, HPSM 2006, Wit Press, 2006.
Ponsaert W., De Wilde W. P., Samyn P., Van Steirteghem J., The use of
tuned mass dampers in beam structures, HPSM 2006, WIT Press, 2006.
Verbeeck B, De Wilde WP, Samyn Ph, Van Steirteghem J., The Use of
Genetic Algorithm and Morphological Indicators in the Optimization of
2D Trusses, HPSM 2004, Wit Press, 2004.

[15]
Verbeeck B., Van Steirteghem J., De Wilde W.P., Samyn Ph., The Need
of Numerical Techniques for the Optimization of Structures Using
Morphological Indicators, Proceedings OPTI2005, WIT Press, 2005.
13
Influence of dynamic loads on the optimum

design of trusses
J. Van Steirteghem1, W. Ponsaert1, W. P. De Wilde1
& Ph. Samyn1,2
1
2
Vrije Universiteit Brussel, Belgium

Samyn and Partners, Belgium
Abstract
The Theory of Morphological Indicators allows a preliminary optimisation of
structures at the stage of conceptual design. Samyn and Latteur developed the
Indicator of Volume to determine the efficiency of structures at early design
stages. The main advantage of this approach is that we only need a very limited
number of parameters. Samyn establishes efficiency curves, with respect to
minimum volume of material, for trusses in which he neglects buckling. Latteur
establishes efficiency curves in which he accounts for buckling. The
displacements are usually checked afterwards to verify if the normative
constraints are not exceeded. In this paper we argue that for trusses, loaded
dynamically and with large spans, dynamics become the dimensioning criterion.
We use the Indicator of the First Natural Frequency to determine the first natural
frequency of trusses. We find that for fully stressed trusses this natural frequency
is usually near to the excitation frequencies of man induced and wind induced
vibrations. Therefore, we need to include dynamics in the optimisation
procedure. We show that for trusses with important spans very large stress
reductions are necessary to obtain acceptable natural frequencies. This stress
reduction comes at the cost of a very important and unacceptable increase of the
volume of material. We determine which typology (Warren, Howe, Pratt) is the
most efficient (minimal volume) with respect to dynamics. Moreover we show
that when dynamics is the dimensioning criterion, the influence of buckling on
the optimum design is negligible since an important stress reduction is necessary.
Finally we propose a work scheme that allows considering dynamics in the
Theory of Morphological Indicators and we provide an example.
Keywords: conceptual design, morphological indicators, trusses, dynamics.
doi:10.2495/HPSM06002
Introduction to the theory of morphological indicators
Morphological Indicators are design tools allowing the optimisation of structures

for a chosen criterion (volume, stiffness) at the stage of conceptual design using
a limited number of parameters [1]. The Indicator of Volume W = V FL
allows the comparison of the volume of material used of different structural
systems subjected to the same load case. The Displacement Indicator
= E L compares the displacement of different structural systems for a
given stress level.
Samyn [1] and Latteur [2] established the analytical expressions of both W
and for trusses, beams, arches, cables, cable stayed structures, masts and
frames subjected to a limited number of (simple) load cases and support
conditions.
For statically determined structures these Morphological Indicators are only
function of the geometrical slenderness L H if instabilities, self weight and
second order effect are neglected. Latteur [2] introduced the Indicator of
Buckling = L
qEF
to take into account the additional volume of
material due to buckling. is a global measure of the sensitivity of a

structure to buckling as a function of the material used. Herein, is a reduction
factor that determines the effective buckling length of the compression member,
q = I 2 is a form factor (with I the moment of inertia and the section
area) that determines the disposition of the material with respect to the center of
gravity of the compression member (buckling efficient sections have large value
of q ).
Furthermore Samyn [1] and Latteur [2] introduced the Indicator of Self
Weight = L to take into account the influence of self weight on the total
volume of material. Herein, is the specific weight of the material used.
In [3, 4] the authors show that dynamics can become the dimensioning criterion
for structures with large spans and/or small stiffness/strength ratio. Therefore, an
Indicator of the First Natural Frequency = 1
is introduced [1, 3, 4].
Indicator of the First Natural Frequency
2.1 Classification of loads

For a structure spanning a length L , the total resultant of the loads F consists
of:
The external live loads Fl
The external permanent loads F p
The self weight F0
15
The external live loads are subdivided in non co-vibrating live loads Fl1 = cFl
and co-vibrating live loads Fl 2 = (1 c )Fl , with c the share of the non covibrating live loads in the total live loads.
Furthermore, we define the following combinations:
FP = F0 + F p , the sum of all permanent loads
FE = F p + Fl , the sum of all external loads
We can now express the resultant of all loads as follows:

F = F0 + FE = FP + Fl
(1)
When calculating the first natural frequency of a structure, we take into account
the co-vibrating loads as vibrating mass. We express the total co-vibrating load
FD as follows:
FD = F0 + F p + Fl 2 = FP + (1 c )Fl
Fl1
Non co-vibrating part
of live loads
cFl
Fl2
Co-vibrating part
of live loads
(1-c)Fl
FD =
Fl2 + Fp + F0 =
zF
all co-vibrating
loads
(2)
Fl =
Fl1 + Fl2
live loads
FE =
Fl + Fp
external
loads
F=
FE + F0
total
loads
Fp
external permanent loads
FP =
Fp + F0 =
F0
self weight
These loads refer to the Ultimate Limit State (ULS) calculation of the structure.
For a dynamic analysis, we must consider loads in a Serviceability Limit State
(SLS). Hence, we can express FD in SLS
F
F
(3)
FD* = P + (1 c ) l
1.35
1.50
And we express the vibrating mass m *D of a structure as:
m *D =
FD*
g
(4)

Finally, we express the total mass m of a structure as:
F
m=
g
(5)
2.2 The first natural frequency of a SDOF system

By approximating the structure by an undamped single degree of freedom
(SDOF) structure, with m *D the vibrating mass in SLS of the structure and k the
static stiffness of the structure, the first natural frequency can be expressed as:
c
k
(6)
f 1 = cor
2 m *D
By reducing a continuous system to a SDOF system, some approximations are
made. c cor is a correction factor that counters these approximations. For beams,
it can be determined analytically; for trusses on the other hand, it must be
determined numerically.
We can express the vibrating mass as a share of the total mass, with z * the
ratio of the co-vibrating load FD* in SLS and the total load F in ULS:
m *D = z * m
(7)
By substituting eqns (7) and (5) in eqn (6), the expression of the first natural
frequency becomes:
c
gk
(8)
f 1 = cor
2 z * F
With = F k the static displacement, eqn (8) becomes:
c cor
g
2 z *
or as a function of the Displacement Indicator:
c
gE
f 1 = cor
2 z * L
f1 =
(9)
(10)
The parameter indicates the stress level of the structure.

The Indicator of the First Natural Frequency becomes:
=
2f 1
c cor
z * L
gE
(11)
When we calculate the natural frequency we need a larger number of parameters

the span ( L )
the material ( , E )
the stress level ( )
the ratio of co-vibrating load in SLS and total load in ULS ( z * ).
17
2.3 Determination of the correction factor c cor

An approximate method to calculate the first natural frequency of a structure,
yielding very good results, is Rayleighs energy method [5]. By using this
method we derive the first natural frequency by equating the maximum potential
energy to the maximum kinetic energy. For a structure with concentrated loads
the fundamental frequency becomes, with j the static displacement under the
concentrated load F j :
2
j
1
f =
2
(12)
We checked the results of this method numerically with Robobat. The results
obtained using eqn. (12) underestimate the numerical ones marginally
(maximum 10% for small slendernesses).
2.4 Indicator of First Natural Frequency
We can use the Indicator of the First Natural Frequency to determine the first
natural frequency of structures.
=1
L = 90 m
1,5
L = 70 m
L = 50 m
Z=1
= 20
0,5
=0
1
L[m]
5
f [Hz ]
11
7f [ Hz 12
]
f [Hz ]
16
17
f [Hz ]
15
14
10
11
Figure 1:
13
10
12
11
10
concrete
timber
10
ccor =1,10
f1 =
S355
20
z * = 0,44
= 1,0
S235
9
15
c cor ,1
2
E
z*
1
L
Graph allowing the determination of the first natural frequency of

Warren truss with 10 panels.

In the first quadrant we plot the Indicator of the First Natural Frequency as a
function of the overall slenderness. In the second quadrant we plot 1
function of = 1
L as a
. Finally in the third quadrant we plot the frequency
abscissa for different material, stress level and value of z * . Since we include the
material, stress level and value of z * in the frequency abscissa the first and
second quadrant must never be changed for this structure.
In this example we determine the first natural frequency of a steel Warren truss
with 10 panels and a span of 50m in which we can neglect buckling ( = 0 ) . For
this example we obtain a first natural frequency of 1,7Hz.
Trusses dynamically loaded
3.1 Scope
We can now study the behavior of trusses that are subjected to dynamic loads.
We consider three truss typologies (Warren, Pratt and Howe):
Warren
Pratt
Howe
We consider three truss typologies (Warren, Pratt and Howe) with:
Three different spans (L = 20m - 50m - 70m)
Three number of panels (n = 4 - 8 - 12)
Three values of z* (z* = 0,15 - 0,44 - 0,74)
Two different materials (timber - steel).
Moreover, we impose that the natural frequency of the structure must be
larger than 5Hz because the man and wind induced frequencies are usually
between 0Hz and 5Hz.
For trusses in which we neglect buckling we can express the constraint on the
first natural frequency as a constraint on geometrical slenderness. First, we use
eqn. (10) compute an upper bound on :
2
gE
c
= dyn = max
0 5 Hz < cor
2 5 z * L
(13)
In [1] we find the expression of the Indicator of Volume of Warren trusses

(eqn. (14)) and Howe/Pratt (eqn. (15)) trusses:
dyn 2dyn (n + 2 )
2
L dyn + dyn (n + 2 )
<
<
H
(n + 2 2n )
(n + 2 2n)
dyn 2dyn (n + 1) L
<
(n + 1 2n )
H
<
19
(14)
dyn + 2dyn (n + 1)
(15)
(n + 1 2n )
On the other hand, if we consider buckling through this interval is calculated

numerically.
3.2 Results and comparison
When we select a target slenderness of L/H = 10, the large majority of
combinations (typology, L, n, z*, material) need a large stress reduction to obtain
a first natural frequency larger than 5Hz. This stress reduction increases when we
use:
Pratt or Howe truss instead of Warren truss
A large span
A large number of panels
A large value of z*
Howe-Pratt ( = 0 ): Necessary stress reduction for target
slenderness L/H = 8 for different materials, spans, number of panels
and z*.
Table 1:
f > 5Hz
=0
co-vibrating load
Material
z* = 0, 74
timber
Number of panels
L = 20m
L = 50m
L = 70m
=
=
=
Number of panels
L = 20m
L = 50m
L = 70m
=
=
=
=
=
=
steel
Timber
0,20
0,08
0,05
0,30
0,12
0,08
0,20
0,08
0,05
0,30
0,12
0,08
steel
timber
0,18
0,07
0,05
0,27
0,11
0,07
steel
4
0,34
0,13
0,09
0,88
0,35
0,25
0,34
0,13
0,09
0,88
0,35
0,25
0,31
0,12
0,09
0,81
0,32
0,23
12
0,16
0,06
0,04
z* = 0, 15
8
0,18
0,07
0,05
Number of panels
L = 20m
L = 50m
L = 70m
z* = 0, 44
4
0,18
0,07
0,05
L H =8
1,00
1,00
0,29
8
12
1,00
1,00
0,29
12
0,93
0,37
0,26

The stiffness/strength ratio of steel and timber is comparable. Table 1 gives an
indication of , the necessary stress reduction to obtain f 1 > 5Hz for a
Howe/Pratt truss.
Conclusions and further research
4.1 Conclusions
In this paper, we showed that we can include dynamics into the Theory of
Morphological Indicators. This comes at the cost of a larger number of
parameters to consider. However, we believe that in order to be useful we must
include the dynamic behaviour of (lightweight) structures. Moreover, we argue
that the number of parameters to consider are still small and are all available at
the stage of conceptual design.
We developed the Indicator of the First Natural Frequency, , to compute
the first natural frequency of structures. The expression of clearly shows that
the problem is not scale independent anymore.
We found that trusses with large spans, large values of z* and small
stiffness/strength ratio are sensitive to dynamic loads and need important stress
reduction to obtain an acceptable natural frequency.
4.2 Further research
In order to make the Theory of Morphological Indicators more powerful in terms
of dynamic behaviour, the following topics can be worked out in the future:
Stress reduction due to fatigue
Higher natural modes and frequencies
External damping devices (passif, actif, hybrid)
References
[1] Samyn, Ph., Etude compare du volume et du dplacement de structures
bidimensionnelles, sous charges verticales entre deux appuis vers un outil
dvaluation et de pr-dimensionnement des structures, PhD thesis,
Universit de Lige, 1999
[2] Latteur, P., Optimisation des treillis, arcs, poutres et cbles sur base
dindicateurs morphologiques Application aux structures soumises en
partie ou en totalit au flambement, PhD thesis, Vrije Universiteit Brussel,
2000
[3] Van Steirteghem, J., A contribution to the Optimisation of Structures Using
Morphological Indicators: (In) Stability and Dynamics, PhD thesis, Vrije
Universiteit Brussel, 2006
[4] Ponsaert, W., Het gebruik van trillingsdempers in balkconstructies, Master
thesis, Vriie Universiteit Brussel, 2005
[5] Fertis, D.G., Mechanical and Structural Vibrations, John Wiley & Sons,
New York, 1995
21
Modular grid-based design concept for

fibre reinforced composite shells
E. De Bolster, H. Cuypers, W. P. De Wilde & J. Wastiels
Department of Mechanics of Materials and Constructions,
Abstract
When designing a modular system, two main aspects need to be considered: the
aspect of one building stone and the aspect of the overall structure. Both will be
studied here as a function of geometry, mechanical behaviour and historical
background (architectural design, engineering science, etc.).
The state-of-the-art of design principles, used in architectural designs over the
latest decades, will provide the basic knowledge for the generation of a new kind
of modular construction. Combining both the artistic point of view (aesthetics)
and the engineering point of view (calculations, FEM), a new design concept
will be generated: a system with modular hyperbolic paraboloid building stones,
connected to one another through hinged connections.
Keywords: architecture, modular construction, design concept, hyperbolic
paraboloid.
Introduction
In contemporary architecture, one can notice the tendency towards modular

lightweight structures. These designs can easily be assembled and disassembled
and they require a minimal amount of materials, which makes this type of design
very advantageous. Another tendency that can be observed is the frequent use of
freeforms. These designs are usually lightweight as well as aesthetically
pleasing. The designs that combine the advantages of modular structures and
freeform structures are usually covered with textiles. However, composites
(textile reinforced cementmatrix composites, for example) could present a good
alternative; especially when used as faces in a sandwich construction. Therefore,
it would be interesting to perform a feasibility study of such a modular, freeform
doi:10.2495/HPSM06003

construction made of sandwich panels with faces of textile reinforced
composites.
Hyperbolic paraboloids
When designing a modular system, two aspects need to be considered: the aspect
of one building stone and the aspect of the overall structure, consisting of several
building stones. Since a modular system is only as strong as its components
(building stones and the connection between them), it is important to select them
with care.
A hyperbolic paraboloid (also called: hypar) building stone (see figure 1) has
many advantages over other types of building stones. Firstly, a hypar surface can
be applied over any foundation shape (rectangular, triangular, circular,).
Secondly, a hypar surface is a curved surface and has therefore better bearing
capacities than flat surfaces, because the forces can more easily be introduced in
an efficient way (i.e. through higher normal forces and lower bending moments).
Due to its anticlastic doubly curved shape, instabilities are less likely to occur:
the convex curve will stiffen the behaviour of the concave curve and vice versa
(see figure 1, on top).
Z
Y
Figure 1: Hyperbolic paraboloid (parabolas, parametric representation and the

two distinct independent families of straight lines).
In rectangular coordinates, with the origin in the saddle point of the surface
(as indicated in figure 1, in the middle), the parametric equation of a hypar
surface can be given as:
Z = k XY
where: k = a measure for the distortion of the hypar surface.
(1)
23
Even the fabrication of a hypar surface is not as difficult as one would expect
from a three dimensional shape: a hypar surface is a doubly ruled surface and can
therefore be formed by two distinct independent families of straight lines (see
figure 1, below), which makes the formwork much easier. When manufacturing
hypar-sandwich panels with composite faces, an adaptable and reusable cable net
could be used as a reconfigurable mould. The net, which is largely based on the
ADAPTENT-concept of Hebbelinck [1], then provides an easy, fast, re-usable
and economical methodology in fabricating hypar-sandwich panels that are to be
used in modular structures (see figures 2 and 3).
Figure 2:
Figure 3:
ADAPTENT-concept: asymmetrical hypar surface [1].
Laminating a hypar sandwich panel on top of the ADAPTENT.
Design principles
There are many basic concepts that underlie the field of design. All basic tenets
can be grouped into design elements and design principles. They are applicable
to all of the visual arts, amongst which architecture. The design elements are the
components or objects of the design: a line, a point, a surface, colour, etc. The
design principles on the other hand govern the relationship between the design
elements, giving them meaning. There are many additional concepts that are
related to the principles of design. The grid-concept is one of these techniques: it
provides a framework for all basic principles.
3.1 Grid
Geometric patterns have always been a fundamental aspect of human culture. In
nature, geometric order is rarely found. It is rather mans way to understand his
surroundings.
Science uses the grid as a technique to cope with complexity. The grid is the
basis for most techniques which sample and quantitatively analyse the real
world. In archaeology for example, archaeologists use three dimensional grids
as a baseline for complex digs. Excavation takes place within trenches or areas of
about 1 metre square, which are located within a site survey grid so that every
point on the excavation site can be given a horizontal coordinate. The
information, revealed through laborious excavation can then be gathered and
ordered in a rather simple manner.
Ecological sampling methods also make use of the grid-method. If one would
want to know what kind of plants and animals are in a particular habitat, and
how many there are of each species, it is usually impossible to go and count each
and every one present. Samples are taken instead. A method of random sampling
is used to map the area under survey and to overlay it with a numbered grid. In
short, the fundamental technique for filtering the complexity of the observable is
based on the quadrat and transect: small square or cubic areas (quadrats) are
examined in detail along a line or across a grid of evenly distributed sample
points (transect).
For reasons difficult to explain, the significance of the grid in modern and
contemporary architecture (and art) is just as important.
3.2 Modern and contemporary architecture
The grid is one of the oldest architectural design tools. A grid can help a designer
control the positions of built and space elements, making the layout task more
systematic. Especially in laying out plans for new towns and cities, the use of
grids enables the designers to make decisions at the urban scale, nevertheless
providing relative freedom at the block and lot scale for individual developers
and house designers.
Taking it one step further, the grid can also be applied to a single building.
This formal and structural approach to design was seen in the Swiss International
Style. The term International Style was first used in New York City in
25
conjunction with the 1932 Museum of Modern Art exhibition of contemporary

European architecture, entitled The International Style: Architecture Since
1922. The term is often used to refer to all Modern architecture.
According to Philip Johnson and Henry-Russell Hitchcock (who wrote the
catalogue that accompanied the exhibition of the International Style), Modern
architecture is characterized by three main principals [2]: architecture as volume,
regularity and avoidance of applied decoration. The rigid, simple, geometric and
asymmetrical style of the International Style arose from the work of Walter
Gropius, Ludwig Mies van der Rohe and of course Le Corbusier and lived on in
the work of for example Skidmore, Owings & Merrill. Many architects
retained the earlier mentioned modernist philosophies, aiming to integrate
modern technology and formal elements derived from the basic grid. Mies used
minimalistic, linear forms, evoking the language of the idealistic International
Style philosophy, called less is more. He embraced the glass and steel
skyscraper that showed its building materials as a form of ornamentation. His
buildings were carefully laid out on a grid, totally embracing the concept of
regularity. This careful grid system lined up everything including the tiles in the
plaza and even the lighting in the ceiling of the lobby. One of his well known
accomplishments is the Barcelona Pavilion in 1929. The grid-system is
undeniably present.
The ideal modernistic point of view became so stringent that a rupture
between architects and this architectural style was inevitable. Postmodernism
rebelled against the purity and rigidity of forms with lots of excess and
exuberance. Nowadays, in the rise of spirituality and fundamentalism however,
architecture has thrown off the ornamentation of postmodernism and returned to
some of the values of the International Style. In his book Supermodernism:
Architecture in the Age of Globalization [3], Hans Ibelings defines the postpostmodernism as a tech-inspired aesthetic movement that reacts against
postmodernism and adopts the philosophy of computer product design.
Architects are drawn towards the idea of tabula rasa; the idea that each new
design starts as a pure, empty plane. Structures appear portable and therefore
disconnected from their surroundings. Similar to a computer, all the details are
inside, while exteriors are neutral and unassuming. High-tech knowledge is used
to create spaces that are valued for their visual and spatial sensations only. Since
high-tech architecture strongly relates to science, the same techniques of
handling complexity are likely to be used. It seems evident that yet again the
grid-system is a widely-spread designing tool.
Overall design
Many have said grid-systems can suffocate creativity, as can modularity.

However, this doesnt have to be the case. Grid-systems can facilitate creativity
by providing a framework and some predefined answers to questions such as
What span to use?, Which module to use?, etc. The grid simplifies decisionmaking by limiting the placement of elements to certain places, but if the grid

doesnt permit a sufficiently rich range of variation, the designer must react and
redesign it.
4.1 Design concept
In this paper a design concept is formulated in which a grid is used as an
integration system (through hinged line connections) of the individual hypar
surfaces in the overall roof-design. A two-dimensional grid is formed that
defines the horizontal coordinates of the four corner points of each hypar surface.
The vertical coordinate of the corner points depends on the formal design of the
overall structure and is thus insofar independent of the grid. Once the position of
four non-coplanar corner points are chosen, a hypar surface can be created. With
this systematic method, a variety of shapes can be made modular in five steps:
1.
2.
3.
4.
5.
Choose a mathematical function z=f(x,y), representing the formal

aesthetics you - as designer - wish.
Choose/note the dimensions of the horizontal rectangular base that
should be covered.
Divide this base in a number of elements (each element having the same
dimensions in the x- and y-direction).
Calculate the x- en y-coordinates of each corner point (on the base).
Use these coordinates and the mathematical equation of the threedimensional function to calculate the z-coordinate of each corner point.
y
z
Figure 4:
Generating a modular hypar-system, based on a grid.
The two-dimensional horizontal grid not only defines the overall design, but
also facilitates the prefabrication of each hypar surface (see section 2). In figure
4, it was chosen to work with a single regular grid to which each building stone
is related.
27
Since any mathematical function can be chosen as initial aesthetic design, a

variety of freeform constructions can be created without the fear of losing
oneself in an infinite amount of choices. Due to the grid and the specific shape of
the basic building stone, one is better capable of handling the complexity of
freeform design: a framework has been set that still leaves room for ones own
creativity and originality.
4.2 Modularity science
The abovementioned design idea has similarities with the finite element method.
In FEM, a complex object is simplified by dividing it in areas (finite elements)
and reducing each finite element to his corner points. Calculations of the stresses
are done in specific, discrete Gauss points. In the presented modular design
system, an overall shape is divided into pieces according to a horizontal grid and
approximated by the discrete corner points of each piece. The design is dictated
by the hypar surfaces that are drawn through the discrete points.
The similarities between the finite element method and designing method are
clarified in figure 5: a geodesic dome of Buckminster Fuller is shown. An
aluminium sphere with radius of 34.14m is used as roof for the former aviation
museum at the airport of Schiphol in the Netherlands. Starting from the formal
design (figure 5, top left), calculation methods are used to mesh the object
under study (figure 5, middle right). Each mesh is then extracted from the
overall object and studied individually: the module is born (figure 5, bottom left)
to which all calculations are related.
MESHING
MODULARISING
Figure 5:
From global concept to discrete solutions (National Aviation

museum Aviodome-Schiphol, Buckminster Fuller, 1971).

As with FEM, the approximation of the original object is best when the griddimensions (or in case of FEM: the mesh-sizes) reduce to zero. Similar to FEM,
it is expected that with a certain grid(mesh)-size a convergence can be reached.
This convergence limits the amount of hypar surfaces that needs to be created
and therefore limits the time needed for design, fabrication and erection of the
modular structure.
In the given analogy between finite elements and the modular design system
under study has a slight discrepancy at the end. When the division of an arbitrary
object in finite elements has occurred, calculation is performed in discrete points.
When the division of an arbitrary roof design in modular elements has occurred,
calculation is performed on each module. In other words, in finite elements
calculation is done in discrete points, while in the presented modular designconcept calculation is said to be done on a surface. In reality, calculation of each
module will be done through the use of finite elements. The modular hypar
surface building stone becomes the arbitrary object that will be divided in a
certain amount of finite elements and calculated in the discrete Gauss points of
each finite element (see figure 6).
Division
Modules
FEM
DESIGN SYSTEM
Object
Finite element
Figure 6:
Meshing
Object
Level-wise calculation of an arbitrary design shape.
Conclusion
The grid-system has always been an important design principle: not only in
science, art, etc, but also in architecture. Modern and contemporary architecture
have turned it into an undeniable part of building concepts. The evolution of
scientific knowledge has turned mankinds urge of controlling his surroundings,
and thus the grid, into an unmistakable part of building realisations.
29
The presented design concept of modular hypar surface-structures originates

from the combination of a two-dimensional, horizontal grid and a chosen,
aesthetically pleasing, three-dimensional design-shape. The concept shows a lot
of similarities with the finite element method: approximating continuous objects
(the overall design shape) by discrete objects (the modular building stones,
which are hyperbolic paraboloid surfaces). On a second level, it even makes use
of the finite element method: calculations are done in discrete points and no
longer on continuous surfaces.
Acknowledgment
Financial support from the Institute for the Promotion and Innovation by Science
and Technology in Flanders (IWT-Vlaanderen) is gratefully acknowledged.
References
[1] Hebbelinck, S., A generating system for temporary, adaptable and reusable
nets and tensile structures, PhD-thesis Vrije Universiteit Brussel, Belgium,
2002
[2] Leland, M., About the modern Style, http://www.michael.leland.name/
modern/index.html
[3] Ibelings, H., Supermodernism: architecture in the Age of Globalization, Nai
Publishers, Rotterdam, 1998
31
Influence of stiffness constraints on

optimal design of trusses using
morphological indicators
T. Vandenbergh1, W. P. De Wilde1, P. Latteur 2, B. Verbeeck1,
W. Ponsaert1 & J. Van Steirteghem1
1
Faculty of Engineering Sciences,

Department of Mechanics of Materials and Structures,
Vrije Universiteit Brussels, Brussels, Belgium
2
Agronomic Faculty of Gembloux and Setesco NV, Belgium
Abstract
Within the framework of sustainable development we strive for structures with a
minimum volume of material. When we only consider criteria on resistance and
buckling, Samyn and Latteur prove that even at the stage of conceptual design a
clear hierarchy among the different truss typologies can be established. Up to
now, stiffness constraints - such as the upper limit on static displacements - were
not considered. However, an optimum obtained by minimising the volume, only
considering the strength criterion, often results in solutions which violate the
stiffness constraint(s). To avoid large displacements a stress level reduction can
be imposed. However, this comes at the cost of a significant volume increase.
With an optimisation process that involves the stiffness constraints at the stage of
conceptual design, an optimum can be obtained without the necessity to alter the
structure drastically afterwards, which partly annihilates the main objective of
minimal use of material. This approach compares the different truss types on a
new priority scale, generating new optima. This implicates a non-negligible
change in the truss choice at conceptual design stage. The solutions are logically
depended on the displacement criterions. This approach forms a first step to a
new design philosophy that considers all the stiffness constraints (static
displacements, resonance, local and global buckling) at conceptual design stage
and is called design for stiffness.
Keywords: morphological indicators, stiffness, strength, truss, steel, static
displacements, optimisation.
doi:10.2495/HPSM06004
Introduction
1.1 Morphological indicators

Morphological Indicators (MI) are design tools allowing the optimisation of
structures for a chosen criterion (volume, stiffness) at the stage of conceptual
design using a limited number of parameters [1]. The indicator of volume
W = V FL allows the comparison of the volume of material used of different
structural systems. It is the volume of an isomorphic structure with unit span L,
with at least one section dimensioned on its unit allowable stress , subjected to
a system of loads with unit resultant F. The displacement indicator = E L
compares the displacement of different structural systems. It is the maximum
displacement of an isomorphic structure with unit span L in a material with unit
Youngs modulus E, with at least one section dimensioned on its unit allowable
stress , subjected to a system loads with unit resultant F.
The analytical expressions of both W and have been established by
Samyn [1] and Latteur [2] for trusses, beams, arches, cables, cable stayed
structures, masts and frames subjected to a limited number of (simple) load cases
and supports.
For statically determined structures those MI are only function of the
geometrical slenderness L/H if instabilities, self weight and second order effects
are neglected. Efficiency curves depicting the geometrical slenderness with
respect to minimum volume material can be established (Figure 1).
The indicator of buckling = L qEF is developed by Latteur [2] to
take into account buckling in compression elements. It is the image of the
buckling tendency of the compression elements in a structure with span L,
composed of bars with a form factor q = I 2 (with I the moment of inertia and
the section area) in a material with Youngs modulus E, with at least one
section dimensioned on its allowable stress with a system of load with total
resultant F. is the proportion of the buckling length of the compression bars
over their geometrical length (which depends on the connection type).
1.2 Stiffness constraints
Up to now, stiffness constraints have never been considered in the use of MI.
Only the resistance criterion was met with volume as the objective function. This
strategy is called design for strength and often results in lightweight structures
with a problematic lack of stiffness, which implicates a non-negligible volume
increase to meet the (imposed) stiffness criteria. In this paper we only consider
one stiffness constraint: the upper limit on the static displacement. We develop
an optimisation process that involves this stiffness constraint at the stage of
conceptual design and obtain an optimum without the necessity to change the
structure drastically afterwards.
Taking stiffness into account compares the different truss types on a new
priority scale, generating new optima. This implicates a non-negligible change in
33
the truss choice at conceptual design stage. As expected, we show that the
solutions are dependent on the displacement criteria.
Figure 1:
Indicator of volume as a function of the slenderness for some

classical structure typologies, neglecting buckling, Samyn [1].
Upper limit on static displacements
In Eurocode 3 [4] we can find the normative constraint on static displacements.

This is usually expressed as an upper limit on L . is the maximum total
displacement of the structure and L the span of the structure. Those are only
indicative values; the commissioner of the project can impose sterner limits.

There are 2 types of constraints: those on the displacements due to the total
loads and those on the displacements due to the live loads only. The
determinative criterion depends on the ratio of live loads to the total loads (self
weight included). We can demonstrate that when this percentage is smaller than
74% we must consider the constraints on the static displacements due to the total
loads. Since this is almost always the case, from now on we only consider the
displacement constraints due to the total loads.
Considering the nature of the displacement constraints, those can very easily
be translated to an upper limit on the displacement indicator, . Imposing an
upper value on L results in imposing an upper value on . However the
constraint on becomes material dependant. The larger the proportion of
Youngs modulus over the allowable stress, the larger the allowable value of
and the less severe the constraint. Table 1 shows the basic material properties of
the classic structure steels.
Table 1:
Material properties of (common types of) structure steel.
Name
Specific
weight
S235
S275
S355
(N/m)
78500
78500
78500
Youngs
modulus E
(GPa)
210
210
210
Allowable
strength
(MPa)
235
275
355
E
894.6
763.6
591.5
(m)
2993.6
3503.2
4522.3
Finally it is important to mention that, according to the Eurocodes, a

calculation of static displacements is a Serviceability Limit State (SLS)
calculation as opposed to a strength calculation which is an Ultimate Limit State
(ULS) one. The difference between, SLS and ULS resides in the partial safety
coefficient imposed on the loads: in ULS we typically use 1.5 for live loads and
1.35 for permanent loads, in SLS, on the other hand 1.0 for both load types. At
conceptual design stage, we can accurately transform loads in ULS to SLS by
dividing the total loads in ULS by 1.4 to obtain the total loads in SLS. Since the
static displacements are linearly proportional to the loads, we must divide the
displacement indicator calculated in Samyn [1] and Latteur [2] by 1.4.
Indicator of volume of structures subjected to stiffness

constraint
3.1 Strategy
We violate the displacement constraint for a large interval of the slenderness and
for common typologies of trusses. Figure 2 depicts the indicator of displacement
for Warren trusses with minimal W. We observe that for common values of
L ( 1 200 and 1 500 ) an upper value for exists which we cannot exceed.
This value of determines a validity interval for the slenderness.
35
4.5
= region of classic upper limits on , corresponding

to relative displacements between 1/500 en 1/200
3.5
3
2.5
2
1.5
1
= 20
= 30
0.5
= 10
L/H
0
0
Figure 2:
10
15
20
Displacement indicator of the lightest Warren trusses and common,

displacement constraints for S355 steel.
The designer is presented with different solutions:

Change the slenderness. This often results in a smaller slenderness.
However, in most cases a maximal height is imposed which results in a
minimum slenderness
Change the typology (number of meshes and/or truss type). This solution is
also subjected to technological and architectural limitations.
Introduce an initial camber: this can compensate the displacement caused by
the permanent loads.
Reduce the stress level: by doing so we increase the stiffness of the
structure. However this comes at the cost of an increase in volume of
material.
The scope of these solutions and their respective impact on the increase of
volume of material must be considered at conceptual design stage. A possible
approach is the calculation of W and for every structure. Then we must
compare the calculated indicator of displacement to its imposed upper limit.
When the calculated displacement is allowable we use this structure. If, on the
other the hand we violate the constraint we impose a stress reduction to increase
the stiffness and meet the criterion on static displacements. We use to denote
the necessary stress reduction and define as the ratio of the applied stress level
to the maximum allowable stress level.
Finally we determine the indicator of volume for this reduced stress level.
This leads to a relatively accurate prediction of the volume of the lightest
structure considering the strength requirement and the constraints on static
displacements. The new optimum curves depend on these stiffness constraints
and therefore on L and the material of the structure. The number of parameters
increases from 2 ( and L H ) to 4 ( , L H , L and material). The main
disadvantage is that we cannot draw general design curves valid for all structures

as those set up in Samyn [1] and Latteur [2]. Nevertheless, the number of
parameters remains small being the main advantage of the MI.
Figure 3:
Optimum indicator of volume and corresponding slenderness as a

function of the indicator of buckling for Warren truss loaded on
their lower chord [2].
3.2 Results
In [2] Latteur selects the optimal (i.e. the lowest) indicator of volume W and its
corresponding slenderness L H opt for every value of the indicator of buckling
. We plot W and L H as a function of (Figure 3). This allows us to
determine the optimal value of W and L H if is known. However, the main

disadvantage is the absence of sensitivity! Those curves only show the optimal
values. For another value of the slenderness, no information is available.
Considering that most optima are found for slendernesses between 2 and 12 [2]
and that usually trusses have values of the slenderness between 8 and 18 [3],
working at the optimum values is not always possible. Therefore we clearly
divide the input information in variables and parameters. The material
parameters and the displacement constraint are considered as being input
parameters, since their values are fixed. For this set of parameters, we plot
37
efficiency curves for the indicator of volume as function of the slenderness for
= 0,10, 20, 30. = 0 corresponds to the theoretical value of Samyn [1], in
which buckling is neglected. Common values of vary between 10 and 30.
Between the curves of = 10, 20 and 30 linear interpolation is allowed,
certainly at conceptual design stage.
10
9
without constraint on displacement

with constraint on displacement
8
7
= 30
= 20
6
5
= 10
=0
3
2
1
L/H
0
0
10
12
14
16
18
20
0.9
=0
= 10
0.8
L/H
= 20
= 30
0.7
0.6
0.5
18
10
12
14
16
18
20
n opt
16
= 30
14
12
= 20
10
8
6
4
2
= 10
Figure 4:
L/H
=0
0
2
10
12
14
16
18
20
Design curves of Warren truss of S355 steel loaded on their lower

chord, with (bold) and without the upper limit on static
displacement of 1 300 .

We also provide the optimal stress level reduction as a function of the
slenderness for the same values of , illustrating the impact of the stiffness
constraint on the exhaustion level of the elements.
To limit the amount of curves those calculations are repeated for values of the
slenderness between 0 and 20 and n (number of meshes) between 2 to 18. Only
the trusses with optimal (minimum W) number of meshes nopt , are selected.
Finally, for every slenderness, the corresponding nopt is plotted, representing the
optimal truss.
This yields 3 design graphs for each set of input parameters (Figure 4, bold
curves). The noise is the result of numerical calculations. A corresponding trend
line indicates clearly the tendency.
10
without constraint on displacement

with constraint on displacement
9
8
= 30
= 20
7
6
5,46
5
4
3
4,04
3,75
3,16
2,42
2
1
=0
0
0
Figure 5:
10
12
14
16
L/H
18
20
Impact of design for stiffness versus design for strength on

optimisation of volume.
Practical example
We design a S 355 steel Warren truss with a span of 50m and a maximal height
of 4m. We must limit the maximal static displacement to 1 300 . If we only
consider the resistance criterion [1] the slenderness and the mesh number are the
only variables and we obtain a truss with a slenderness of 12,5 and 2 meshes
when we limit the volume of material. In the strategy of Latteur [2], in which
buckling is considered explicitly, more input information is needed to determine
the value of : the selected material, the section type of the bars and the total
load. S355 steel, tubes with thickness/diameter ratio of 0.04 and a total load of
2500kN , result in an indicator of buckling of = 22.62. We obtain an optimum
truss with 11 panels and a slenderness of 12,5. Finally, the constraint on the
static displacement does not influence the slenderness, on the other hand the
number of panels decreases to 5. For this truss we obtain W = 3.75. Since we
include more phenomena (buckling, static constraints), the indicator of volume
obtained by Samyn [1] increases (from 2.42 to 5.46). The solution obtained by
39
Latteur [2] also increases due to the constraint on the static displacement (from
3.16 to 4.04). Those results are shown in Figure 5 and clearly illustrate the
importance of considering stiffness constraints at conceptual design stage. Here
we obtain a gain in volume of 8% in comparison with Latteur [2] and up to 46%
with the result of Samyn [1].
Conclusions
We found that trusses with a large stiffness are composed of a small number of
panels. Even though a larger number of panels reduces the buckling length of
compression members, we observe that when we consider the constraint on static
displacements we obtain an optimal solution with a smaller number of panels [1].
Moreover, stiff trusses are usually stocky (typically L H between 1.5 and 5),
which confirms the results obtained by Samyn [1], guiding us towards small
slendernesses. We observe that these stiff trusses are often very light (especially
when the truss is not sensitive to buckling).
The more buckling sensitive the truss, the smaller the necessary stress level
reduction, since the material against buckling also provides stiffness against
displacements. This can be noticed on the curves by the larger s at larger
' s for the same slenderness. On the other hand, a stress level reduction against
displacements means a less buckling sensitive truss. This explains why at larger
slendernesses (the most compliant geometry), the curves of the different s
join. For very large L/Hs and/or very strict displacements constraints, we can
assume that the optimal structures become independent of .
The optimal stress level is not always = 1 . This clearly shows that fully
stressed design doesnt always give the lightest structure.
Finally, Howe and Pratt trusses do not present any advantage in comparison
with Warren trusses, since they not only need more material but are less stiff.
Further research
The above mentioned strategy, which considers the constraints on static

displacements, can be applied to other truss topologies (K, Long, Smith,) and
to different typologies (arches, beams). Hence, we can consider a new design
hierarchy with respect to the minimisation of volume
This approach constitutes a first step to a design philosophy that considers all
stiffness constraints (static displacements, resonance, local and global buckling)
at conceptual design stage, called design for stiffness.
Finally, we must consider the influence of the connections (fixed or pinned)
on the stiffness and on the volume of material.
Acknowledgement
I would like to acknowledge the Institute for the Promotion of Innovation
through Science and Technology in Flanders (IWT Vlaanderen), which helps this
research financially.
References
[1] Samyn P., Etude Compare du Volume et du dplacement de Structures
Isostatiques Bidimensionnelles sous Charges Verticales entre Deux Appuis.
Vers un outil dvaluation de prdimensionnement des structures (Tome 1
4), PhD thesis, Universit de Lige, Belgium, 1999.
[2] Latteur P., Optimisation et Prdimensionnement des Treillis, Arcs, Poutres
et Cbles sur Base dIndicateurs Morphologiques (Tome 1 4), PhD thesis,
Vrije Universiteit Brussel, Belgium, 2000.
[3] Orton A., The way we build now (Chapter 2). Building Structures, E & FN
Spon , pp. 42-44, 1994.
[4] Eurocode Steel/3, NBN ENV 1993-1-1:1992 pp. 49-52, 1992.
41
Variations in form and stress behaviour of a

V-shaped membrane in a foldable structure
M. Mollaert, N. De Temmerman & T. Van Mele
Department of Architecture, Vrije Universiteit Brussel, Belgium
Abstract
Adaptable tensile structures are often considered to be either completely opened
or completely closed. The current study is part of a research project studying
adaptable tensile structures which demonstrate stable behaviour within a wide
range of opened positions.
In this paper a simple V-shaped membrane is studied during the unfolding
process. Starting from an initially flat folded membrane, which is not
pre-tensioned, a slight curvature is obtained when it is unfolded due to the fact
that along the folding line a curved section is cut out of the fabric. The tension
introduced in the transverse direction implies a tension in the longitudinal
direction too.
Two cases are analysed: one with a high curvature in the diagonal cable (~5%
sag) and one with a low curvature (~1.3% sag). Based on computer simulations
the form and the tensions are verified for different opening angles. The
deformation under loading is checked for the shape with a low curvature of the
diagonal cable at an opening angle of 70. The results indicate that the membrane
could be used as a fabric roof. Further refined analysis is needed to be able to
implement the presented concept for real applications.
Keywords: tensile structure, formfinding, coated textiles, adaptable shelter,
foldable structures.
Introduction
Tension structures, made of high-strength coated fabrics, are well-suited to create

adaptable buildings that can react (open/close) to changing environmental
conditions [1].
Several solutions are possible, but only a few techniques keep the membrane
tensioned while folding or unfolding. A first possibility is to start from a wave
doi:10.2495/HPSM06005

form. The chosen reference equilibrium form is characterised by an intermediate
opening angle (Fig. 5) and an appropriate pre-tension. When the wave form
opens (Fig. 6) or closes (Fig. 4), the membrane tension does not change too
much as long as the variation in the curvature is not too important and the
membrane can be used as a roof in the intermediate positions [2, 3].
Figure 1:
Parallel
sliding.
Figure 2:
Central
sliding.
Figure 3:
Circular
sliding.
Figure 4: Folded
Figure 5: Reference
Figure 6: Unfolded
configuration.
configuration.
configuration.
Figure 7: Folded rhombus Figure 8: Opened

Figure 9:
shape.
rhombus shape.
Curved
diagonal.
In the current study the reference equilibrium form is a folded configuration

without pre-tension. To test the basic idea a small model of a rhombus shape
with straight edges and covered by fabric was made (Fig. 7, Fig. 8 and Fig. 9).
The longitudinal diagonal divides the rhombus in two isosceles triangles with an
apex angle of 120 (see Fig. 13 for the definition of the angle). The longitudinal
diagonal is the folding axis. At this folding axis a part of the fabric was cut out
along an arc that represents the cutting line. As a consequence the longitudinal
diagonal is slightly curved. During the unfolding process the membrane is
43
tensioned in the transverse direction, and due to the bi-axial behaviour of the
material, the longitudinal direction is tensioned too.
The opening angles ranging from 60 to 80 are the most appropriate (in terms
of covered area: see Fig. 11, Fig. 12) if one wants to use the unfolded tensioned
membrane as an architectural roof.
Figure 10: Elevation for Figure 11:

opening
angles
ranging from
50 to 90.
Covered
Figure 12: Covered area
area for an
for an opening
angle of 80.
opening
angle of 70.
The cladding material is a coated fabric and the longitudinal diagonal is

reinforced by a belt or a cable. The membrane itself is modelled by a mesh of
25cm x 25cm. The boundaries are considered to be stiff beams and modelled as
fixed points on the boundary lines. Starting from a tensionless folded state,
called the reference equilibrium form, the stress behaviour is analysed, while
unfolded along the longitudinal diagonal.
Defining the reference equilibrium form
The reference equilibrium form is characterised by an opening angle of 1 and an

initial tension in the membrane of 0.02kN/m, considered as a tensionless state.
Calculations are made with the EASY software of Technet, based on the
Force Density Method [4]. The formfinding based on the force-density method
calculates an equilibrium form that is independent of the stiffness of the elements
in the net. Different force values in the diagonal cable can be considered. The
force values depend on the ratio of the force-density set for the longitudinal cable
to the force-density set for the internal net. Five cases have been calculated to be
able to estimate the curvature of the diagonal cable for different values of the
force in the diagonal cable. The curvature is characterised by the ratio of the
height H divided by the span S (see Fig. 13), the span being 6m.
For the numeric modelling a rhombus of 6.0m x 3.464m is considered. The
following figure illustrates the geometry and basic parameters:
For two cases the unfolding has been analysed:
Case 1, having a higher curvature in the diagonal cable, will be considered with
a low stiffness of 1kN of the internal net elements, which represents the situation
of a stretchable net
Case 5, having a lower curvature in the diagonal cable, will be considered with a
more realistic stiffness of 150kN of the internal net elements, both in the
longitudinal and in the transverse direction

Used symbols
H
arc AA
arc BB
Figure 13:
Table 1:
Case
Geometry and used symbols.
S
=
AA
apex angle of
triangular
module
opening angle
of triangular
module
height of the
longitudinal
arc
arc in
longitudinal
direction
(along Xaxis)
arc in
transversal
direction
(along Yaxis)
span of the
longitudinal
arc
The height to span ratio of the diagonal for different forces in the
diagonal cable.
Elevation
Force
[kN]
Height
[m]
H/S
[%]
2.1453
0.0775
1.29
1.7173
0.0952
1.59
1.2895
0.1233
2.06
0.8622
0.1750
2.92
0.4359
0.3013
5.02
45
2.50
Force [kN]
2.00
1.50
1.00
0.50
0.00
1.29
1.59
2.06
2.92
5.02
H/S of the diagonal arch [% ]
Figure 14:
Force in the diagonal cable (at 1 opening angle).
It is obvious that the smaller the force in the diagonal cable, the greater the
curvature of the diagonal arc will be (for the same pretension in the membrane of
0.02kN/m).
But on the other hand, the greater the curvature of the diagonal cable, the
higher the stresses in the transverse direction will be when unfolding.
Tension in the membrane for different opening angles
3.1 Case 1: Higher curvature in the diagonal cable

The influence of the opening angle on the forces in the membrane and the
diagonal cable is verified for a low stiffness of the internal net (1kN) and a
stiffness of 5000kN of the diagonal cable. The first form given in Table 2 is the
reference equilibrium form (with an opening angle of 1). The other
configurations have been obtained by a static analysis, starting from the
reference equilibrium form, with the fixed nodes displaced according to the
considered opening angle.
Tension [kN/m]
0.50
0.40
0.30
0.20
0.10
0.00
10
20
30
40
50
60
70
80
Opening angle []
Figure 15:
Tension in the transverse direction due to variation of the opening

angle.
From the calculations with a stiffness of 1kN in the internal net it can be
concluded that
- in the longitudinal direction the tension in the membrane remains at a
value of about 0.02kN/m
- in the direction perpendicular to the folding axis the tension increases
from 0.02kN/m to 0.44kN/m for larger opening angles.
6.00
3.40
5.00
3.30
3.20
4.00
H/S [%]
Force [kN]
3.00
2.00
3.10
3.00
2.90
2.80
1.00
2.70
0.00
10
20
30
40
50
60
70
80
10
20
Opening angle []
Figure 16:
30
40
50
60
70
80
Opening angle []
Force in the diagonal Figure 17: H/S of the diagonal cable

cable due to variation of
due to variation of the
opening angle.
the opening angle.
When unfolding, the force in the diagonal cable increases up to a maximum

value. Then the force in the diagonal cable decreases for larger opening angles.
Since the force in the cable is proportional to the resultant of the membrane
tension at both sides of the cable, two different facts interact:
- for larger opening angles the resultant decreases if the
tension in the membrane remains the same
- for larger opening angles the membrane tension in the
transverse direction increases
Figure 18: A higher resultant force Figure 19: A lower resultant force
with higher membrane
with
lower
membrane
tensions at 70
tensions at 60.
3.2 Case 5: Lower curvature in the diagonal cable
The influence of the opening angle on the stresses in the membrane and the
diagonal cable is verified for a more realistic stiffness in the internal net of
150kN. The stiffness for the diagonal cable remains 5000kN. The 150kN per
0.25m corresponds to a value of 600kN/m valid for a typical PVC coated
polyester membrane.
From the calculations where the opening angle varies between 60 and 80
can be concluded that:
- in the longitudinal direction the tension in the membrane
varies from a value of about 0.6kN/m up to 2.2kN/m for
larger opening angles
- in the direction perpendicular to the folding axis the tension
increases from 0.4kN/m to 20.0kN/m for larger opening
angles
Similar to the membrane tensions in the longitudinal direction, the forces in
the diagonal cable increase up to a maximum value and then decrease due to
variation of the opening angle.
Table 2:
47
Changes in form and stresses by changing opening angle [] for

case 1 low stiffness.
Elevation
Front view
40
2.50
Tension [kN/m]
Tension [kN/m]
80
2.00
1.50
1.00
0.50
0.00
60
65
70
75
80
25.00
20.00
15.00
10.00
5.00
0.00
60
Opening angle []
65
70
75
80
Opening angle[]
Figure 20: Tension in the longitudinal Figure 21: Tension in the transverse
direction due to variation
direction due to variation of
the opening angle case 5.
of the opening angle
case 5.
Table 3:
Changes in form and stresses by changing opening angle [] for

case 5 higher stiffness.
Elevation
65
80
Front view
30.00
25.00
20.00
15.00
10.00
5.00
0.00
5.00
4.00
H/S [%]
Force [kN]
3.00
2.00
1.00
0.00
60
65
70
75
80
60
65
Figure 22:
Force in the diagonal

cable due to variation of
the opening angle
case 5.
80
Front view
Equilibrium form at an opening angle of 70.

Plan
3D-view with loading
Elevation
Front view
Figure 25:
75
Figure 23: H/S of the diagonal cable

due to variation of the
opening angle case 5.
Plan and Elevation
Figure 24:
70
Opening angle []
Opening angle []
Structure under snow load (0.4kN/m2).
Tension in the membrane at an opening angle of 70 under

external loading
Still for case 5, with a stiffness of 150kN in the internal net and 5000kN in the
diagonal cable, external loading will be applied for an opening angle of 70. Two
representative load cases have been chosen [5]: a uniform snow load of
0.4kN/m2 and an upward wind load of 0.6kN/m2. For the equilibrium form with
an opening angle of 70 without external loading the average tension in the
membrane is 2kN/m in the longitudinal direction and 4kN/m in the transverse
49
direction. The static analysis under external loading combines the displacement
of the fixed nodes according to the opening angle of 70 with the application of
load vectors for the selected load case on the equilibrium form without external
loading. A uniform snow load of 0.4kN/m2 has been applied. In this case the
maximum deflection is 2cm. In the transverse (load bearing) direction the
average tension increases from 4 up to 5kN/m.
Next an upward wind load of 0.6kN/m2 (suction) has been applied. In this
case a maximum deflection of 4cm occurs. In the longitudinal direction the
average tension increases from 2 up to 3.1kN/m.
Plan
3D-view with loading
Elevation
Front view
Figure 26:
Structure under upward wind load (0.6kN/m2).
General remarks
The simulations only give a first estimation of the structural behaviour: the
model does not take into account the shear stiffness of the coated fabric, the
material behaviour is considered to be linear and independent of the stress level
and frequently folding and unfolding will also influence the material behaviour.
For more precise results a refined model is required. Moreover, an optimised
configuration could be found by choosing a different stiffness or selecting
another boundary geometry in plan view.
Conclusion
For the same external loading on a membrane roof a more expressive double
curvature will imply lower tensions in the membrane. Nevertheless more and
more applications use a flat appearing curvature to cover box-like spaces. The
rhombus-shaped membrane elements will be used as cladding components in a
mobile foldable shelter consisting of articulated bars as the primary load bearing
structure [6]. Based on the simulation of a rhombus shaped membrane it is
shown that the unfolding of flat membrane pieces can, within a certain range of
opening angles, create a slightly curved tensioned membrane which can resist, in

an acceptable way, representative external wind and snow loads. Several
parameters should still be optimised (size, geometry, stiffness) for a specific
application of the V-shaped foldable membrane.
References
[1]
[2]
[3]
[4]
[5]
[6]
F. Otto, B. Baier and S. Meyer-Miethke, IL 14 Adaptable Architecture,

Karl Krmer Verlag, Stuttgart (1975).
M. Mollaert, N. De Temmerman, T. Van Mele, Ph. Block Philippe, F.
Daerden, Adaptable Tensioned Coverings, Proceedings of the IASSAPCS International Symposium on New Perspectives for Shell and
Spatial Structures, Taipei, Taiwan pp.204-205, (full paper on cd-rom)
(2003).
M. Mollaert, T. Van Mele, N. De Temmerman, Kinetic Structures:
Architectural Organisms as a Design Concept, Proceedings IASS
Symposium Shell and Spatial Structures from Models to Realization,
Montpellier, p.410, (full paper on cd-rom) (2004).
E. Moncrieff, Extreme Patterning: Lessons From the Cutting Pattern
Generation of the Mina Valley Tent City and the Federal Chancellery
Projects, TensiNet Symposium: Designing Tensile Architecture, pp.6681, Brussels (2003).
B. Forster, M. Mollaert, European Design Guide for Tensile Surface
Structures, TensiNet (2004).
De Temmerman, N., Mollaert, M., Decorte, W., Parametrization and
development of a foldable mobile shelter system, Vrije Universiteit
Brussel, Research Working Paper No. 2, (2005).
Section 2
Composite materials
and structures
53
A new composite material based on

natural fibres and a thermoset:
technology, applications and properties
G. Wuzella
Kompetenzzentrum Holz GmbH c/o FunderMax Industrie GmbH,
Klagenfurter Strae 87-89, A-9300 St. Veit/Glan, Austria
Abstract
The aim of this work was the development of a new composite material based on
natural fibres as reinforcement and an acrylic thermoset as binder to investigate
the material properties for various applications. All applications were made by
forming and curing the composite material in a hot forming press. In a first step
the technology for formation of such a composite material was tested: Based on
the nonwoven-process natural fibre mats were made, only composed of natural
fibres without binder and with different area weights. After that, different
technologies of application of binder and additives like hydrophobizing agents
were tested. In the second step the material properties (flexural and impact
strength, water uptake and moisture expansion, testing for climatic extremes and
sound-absorbing properties) were investigated for different applications as
function of the density of the pressed parts, the mixing ratio of binder and
additives. The following properties can be given as very special features of the
material: In comparison with composites made of usual thermoset binders the
fabricator has not to apply the resin immediately before processing. Secondly the
material can be stored easily, because the curing is started beyond 100C, at
lower temperatures the binder is thermoplastic. In comparison with composites
made of thermoplastic binders, a lesser fraction of binder is necessary and the
pressed parts have a very high flexural stiffness. Finally the critical properties
like the water uptake and moisture expansion were improved by the addition of
hydrophobizing agents.
Keywords: natural fibres, nonwoven, composites, acrylic resin, binding
technology.
doi:10.2495/HPSM06006
Introduction
Actual industrial demand for natural fibres has developed over the past few
years. Today, the use of natural fibres has become common in some applications;
the most important customer is the automotive industry. Technically speaking,
the use of natural fibres in automotive applications involves primarily pressmolded composites, produced by the pressing of a nonwoven mat with a binder.
Typical uses are in door panels, hat racks, and trunk liners. Two production
technologies are commercially employed:
a blend of natural and polypropylene fibres is processed into a nonwoven
mat and pressed into the desired shape under heat ("thermoplastic matrix");
nonwoven mats are coated with thermosets, such as epoxy resin or
polyurethane and moulded. The ultimate material is generated by
polymerization and hardening of the resin ("duromeric matrix").
The main reasons for the use of natural fibres are:
Weight reduction of 10 to 30%
Good mechanical and manufacturing properties
Possibility to manufacture complex structural elements from one
material in a single pass
Good performance in accidents (high stability, no splintering)
Superior environmental balance during material and energetic use
Occupational health advantages compared to glass fibres
No emissions of toxic substances
Overall cost advantage compared to conventional construction
The aim of this work is the development of a new composite material, based
on natural fibres as reinforcement but instead of a thermoplastic binder only a
thermoset is used to realise a new range of applications. Instead of the commonly
used thermosets like epoxy resins, polyurethanes or aqueous phenolic resins,
another thermoset is used, which can be processed easier and which corresponds
better to the proposition of an ecological awareness.
Due to the machines of the partner company the production of fibre mats is
based on the nonwovens technology: dry-laying staple fibres (the fibres are 1 to
20 cm or longer, but not continuous) by an aerodynamic web formation and
mechanically bonded by needle-punching. The term nonwoven is defined as a
manufactured sheet, web or bat of directionally or randomly oriented fibres,
bonded by friction, and/or cohesion and/or adhesion, excluding paper or products
which are woven, knitted, tufted stitch bonded incorporating binding yarns or
filaments, or felted by wet milling, whether or not additionally needled. The
fibres may be of natural or man-made origin. They may be staple or continuous
or be formed in situ [1, 6].
Materials
For composites composed of natural fibres, blends of natural fibres (e.g. flax and
kenaf or flax and hemp) are particularly interesting. The finer flax fibres impart
55
high stability, but because they impede complete soaking with thermosetting
binders, fractures may develop. Only the mixture with coarser fibres like kenaf
fibres achieves an optimum balance between stability and complete saturation
with the binder. In this project all nonwovens are a mixture of kenaf and flax at
the ratio of 50:50.
2.1 Kenaf fibres
Kenaf is the term for the plant Hibiscus cannabinus L., a warm season annual
closely related to cotton (Gossypium hirsutum L.) and okra (Abelmoschus
esculentus L.). The kenaf fibres are gained from the bark of the stem of the
plants. The staple fibres have a length of 40 60 mm and an average thickness of
25 m. Because of their mechanical properties they are used as reinforcing fibres
in a composite [2].
2.2 Flax fibres
Flax fibres are amongst the oldest fibre crops in the world and the use of flax for
the production of linen goes back 5000 years. Flax fibres are soft, lustrous and
flexible. They are stronger than cotton fibre but less elastic. Because of their
mechanical properties they are used as reinforcing fibres in a composite [3].
2.3 Matrix
The used matrix system in this work is a reactive aqueous acrylic resin, which is
free of phenol and formaldehyde. This one-component resin has good storage
stability and cross-links upon heating to 180 220C. After the impregnation the
fibre mats are dried. The binder begins to cure beyond 100C, and semi-finished
products impregnated with the acrylic resin therefore have excellent storage
stability in comparison to other thermosets (e.g. epoxy resins or polyurethane).
During the hot press step the impregnated mats are formed and the resin is crosslinked. Composite parts bound with the acrylic resin show high stiffness and
strength and meet the ecological requirements of the automotive industry. To
achieve the best mechanical properties and moisture resistance, press moulding
is carried out with tool temperatures of about 200 to 220C and cycle times of
one minute or less [4, 5].
2.4 Additives
In this project two types of additives are used to improve the properties of the
composite. The first is a borate, which contributes to avoid the creation of
mildew, because of the moisture content in nonwoven composed of natural
fibres. The second additive is a paraffin wax, which improves the water
resistance (water uptake and moisture expansion, influence of climatic extremes
on material properties) of the moulded part. In particular the type of paraffin wax
has to be stable against acid, because of the acidity of the used acrylic resin [5].
Technologies of binding
The first task of this work was to find a technology to apply the binder to the
natural fibres. In the following second task the found technology was adjusted to
find the optimised settings for the application. The application process of the
binder can be done in two ways. Either the naturals fibres are impregnated before
they are laid to fibre mats or the application is done subsequent to the
nonwovens process. All 3 tested technologies in this project follow the second
way of impregnation with the following pre-adjustments:
Used binders are liquid, especially aqueous binders
Additives are mixed together with the binder before impregnating
3.1 Foam impregnation technology
The binder is mixed to a foam by a mixing machine and applied to the rotating
rolls. The nonwoven, composed of natural fibres, is impregnated by pressure
between the roller clearance. After that the impregnated nonwoven is dried in an
oven and winded to a roll or cut into formatted pieces.
Figure 1:
Scheme of impregnation by foam.
3.1.1 Evaluation of the foam impregnation technology

+ Energy-saving concerning the drying process of the impregnated nonwoven,
because less water is applied to the nonwoven by the foam compared to other
impregnation technologies.
- Not all binders can be mixed to foam, sometimes a foaming agent is necessary.
- Thick materials (area weight>1000 g/m) require for a complete impregnation
high power of the impregnation plant. Experiments demonstrate that the power
consumption for thick nonwovens with a width greater than 0.5 m exceeds the
engine power; sometimes a thick nonwoven is not continuously impregnated.
- Additionally a complete impregnation of thick nonwovens causes a very high
compression of the natural fibres in the mat, which can lead to damaged fibres.
3.2 Impregnation technology of glass prepregs
The nonwoven is pulled through a basin, which is filled with the liquid binder.
After the impregnation the excess of binder and water is squeezed out between
57
two rolls. After that the impregnated nonwoven is dried in a vertical drying
tower and winded to a roll or cut into formatted pieces.
Figure 2:
Scheme of impregnation technology of glass pregpregs.
3.2.1 Evaluation of the impregnation technology of glass prepregs

+ All liquid binder systems can be used
+ Continuous impregnation for nonwovens in any thickness and width without
damaging of fibres or exceeding of engine power limit.
- To guarantee a continuous impregnation of the nonwoven the aqueous binder
has to be diluted with further water, which cannot be dried by the drying tower.
The vertical drying tower is insufficient to dry the nonwoven quickly enough.
- During storage of the winded impregnated nonwoven the residual moisture and
binder migrate and cause inhomogeneity of the impregnation.
3.3 Foulard dyeing technology
The apparatus comprises a foulard in which the nonwoven is provided with the
binder in a treatment bath and squeezed off to the binder and moisture content.
The impregnated nonwoven is winded to a roll, which can be transported to the
dryer.
Figure 3:
Scheme of a foulard dyeing system.

3.3.1 Evaluation of the foulard dyeing technology
+ All liquid binder systems can be used
+ Continuous impregnation for nonwovens in any thickness and width without
damaging of fibres or exceeding of engine power limit.
+ To guarantee a continuous impregnation of the nonwoven the aqueous binder
has to be diluted with further water. In fact the horizontal dryer is limited to
avoid the curing of the binder but sufficient enough to dry the impregnated
nonwoven to a residual moisture of 10%.
- Dimensioning of the dryer and the drying process is more expensive than the
foam impregnation technology
3.4 Choice of the impregnation technology
Due to the evaluation of the three technologies of binding the decision was made
in favour of the foulard dyeing technology. This technology is the easiest way to
impregnate the nonwoven continuously without risk to exceed the engine power
limit and without risk to damage the fibres because of over-compression. The
external horizontal dryer allows drying of the nonwoven to the residual moisture
of 10%.
Applications and properties
The impregnated mats are moulded to composite parts and the following
properties complying with DIN standards are tested
Table 1:
Measured properties.
Property
Flexural strength [N/mm]
Flexural modulus of elasticity (MOE) [N/mm]
Water uptake [%]
Increase of thickness of samples due to moisture
expansion [%]
Impact strength [mJ/mm]
Standard
DIN EN 310
DIN EN 310
DIN 52351
DIN 52351
DIN EN ISO 179
Many types of applications were made with the impregnated material.

Besides door panels, hat racks and trunk liners a roofliner as application should
be developed.
4.1 Mechanical Properties
Because of the processing of fibre mats (laying and needling) the nonwoven as
well as the moulded parts have different mechanical properties subject to the
direction a sample is taken. Therefore these properties are proved for two
directions, one parallel to the direction of the mat production and one transverse
to this direction. Samples in direction of production have higher mechanical
strength than samples transverse to this direction.
Flexural strength [N/mm]
90
80
70
60
50
40
30
20
10
0
25 % resin
25 % resin;
transverse
15 % resin
15 % resin;
transverse
15 % resin
0.1
0.3
0.5
0.7
0.9
1.1
15 % resin;
transverse
Density [g/cm]
Flexural MOE [N/mm]
Figure 4:
Flexural strength of moulded composites with 2 different contents

of binder as function of density, tested in 2 orthogonal directions.
9000
8000
7000
6000
5000
4000
3000
2000
1000
0
25 % resin
25 % resin;
transverse
15 % resin
15 % resin;
transverse
15 % resin
0.1
0.3
0.5
0.7
0.9
Density [g/cm]
Figure 5:
59
1.1
15 % resin;
transverse
Flexural MOE of moulded composites with 2 different contents of

binder as function of density, tested in 2 orthogonal directions.
Both flexural properties, the flexural strength and the flexural modulus of
elasticity, primarily depend on the density of the moulded composite. The higher
the density of the moulded part is the higher are the flexural properties. The
binder content has only a small effect on the flexural properties; for samples in
production direction and with a higher density the properties are higher, if the
binder content is 15% instead of 25%. In comparison to other composite
materials based on natural fibres the attention should be paid to the high values
of flexural properties for samples with a density of 0.9 g/cm and above: samples
in production direction have a flexural MOE between 7000 and 8000 N/mm and

a flexural strength between 70 and 85 N/mm, samples transverse to this
direction have a flexural MOE about 4500 N/mm and a flexural strength of
about 50 N/mm. The impact strength of all samples is low (10 mJ/mm)
regardless of binder content and of density.
400
25 % resin; Density
= 0.6 - 0.9 g/cm
Water uptake [%]
350
300
250
15 % resin; Density
= 0.3 - 0.9 g/cm
200
15 % resin
150
25 % resin
100
50
0
0.1
Figure 6:
0.3
0.5
0.7
Density [g/cm]
0.9
1.1
Water uptake of moulded composites with 2 different contents of

binder but without hydrophobizing agent as function of density.
4.2 Water uptake and moisture expansion

The water uptake primarily depends on the density of the moulded composite.
The higher the density of the moulded part is the lower is the water uptake. By
higher binder content the ability of water uptake can be decreased. This effect is
much higher if the density of the moulded composite is low.
Moisture expansion [%]
40
38
35
30
25
25 % resin; Density =
0.6 - 0.9 g/cm
25
20
15 % resin; Density =
0.3 - 0.9 g/cm
14
12
15
10
5
0
10
15
20
25
30
Resin content [%]
Figure 7:
Moisture expansion of moulded composites with 2 different

contents of binder but without a hydrophobizing agent.
61
The increase of thickness of moulded composites due to moisture depends

only on the resin content but not on the density of the moulded part. Higher
binder content means that the composite is more rigid against the moisture
expansion and the fibres are better protected against water. A composite with
low density has besides to the hydrophilic behaviour of natural fibres a high
porosity. This explains why the water uptake can be high but at the same time
the thickness doesnt increase to the same degree.
4.3 Improvement of water resistance
The measured properties like water uptake and moisture expansion are subject of
an improvement, mainly for applications with a low density like a roofliner.
Hence a hydrophobizing additive can be mixed into the binder.
To improve the water resistance of the moulded composites 1.5% of additive
is sufficient for moulded parts with a density of 0.91 g/cm.
Table 2:
Binder
[%]
15
15
25
25
Water resistance of composites with density of 0.91 g/cm.
content
Content
additive [%]
0
1.5
0
1.5
of
Water
[%]
58
22
31
16
uptake
Moisture
expansion [%]
30
14
13
6
roofliner - sound absorption
coefficient of absorption
1.20
1.00
0.80
0.60
0.40
0.20
0.00
400
500
630
800
1000
1250
1600
2000
2500
3150
4000
5000
6300
frequency [Hz]
Default material
Sample 02; thickness = 5.5 mm
Sample 04; thickness = 3 mm
Figure 8:
Sample 01; thickness = 5.5 mm

Sample 03; thickness = 3 mm
Sound absorption of roofliners.
4.4 Acoustic properties

Nonwovens with an area weight of 750 g/m and a binder content of 25% are
produced by the foulard dyeing technology. The impregnated mats are moulded
to roofliners and proved in a reverberant chamber. The measured absorbance of
the samples is compared with a default material.

The acoustic properties are an interesting aspect for applications like
roofliners. A roofliner is usually a sandwich construction composed of a foamed
core layer between two reinforcing laminates as top layers (=default material).
The foamed core acts as sound-absorber. The composite material composed of
natural fibres and acrylic resin has a good sound-absorbance. If the frequency is
higher than 1250 Hz the composite is better than the compared default material.
Summary
A new composite material based on natural fibres as reinforcement and an

acrylic thermoset as binder was developed with the following results:
The foulard dyeing technology is the best choice to impregnate both thin and
thick nonwovens composed of natural fibres continuously without risk to
exceed the engine power limit or to damage the fibres.
Advantages for the fabricator:
o The fabricator doesnt need to impregnate the mats immediately
before moulding
o The impregnated mats have a good storage stability compared to
other thermosets
o The acrylic resin has a very good environmental impact (no phenol
and formaldehyde)
A reduction of weight of moulded parts can be realised (-20%) without
losing of mechanical properties.
A low binder content of 1520% is sufficient for binding.
To improve the water uptake and moisture expansion of the moulded
composites a hydrophobizing agent is necessary (1.5% is sufficient).
In comparison to composites with thermoplastic binder the dimensional
stability under heat is excellent.
References
[1]
[2]
[3]
[4]
[5]
[6]
Albrecht W., Fuchs H., Kittelmann W. NonWoven Fabrics Raw

Materials, Manufacture, Applications, Characteristics, Testing Processes.
Verlag Wiley-Vch. 2003.
Clark T. K., R.L. Cunningham, and Wolff I. A. A search for new fiber
crops. 1971. TAPPI 54; (1) p. 63-65.
Dambroth. Seehuber. Flachs: Zchtung, Anbau und Verarbeitung. 1988.
Reck B., Trk J. Thermally curable aqueous acrylic resins a new class of
duroplastic binders for wood and natural fibers. 2nd International Wood
and Natural Fiber Composites Symposium. Kassel, Germany. June 28-29,
1999.
Oberbach K., Baur E., Brinkmann S., Schmachtenberg E. Saechtling
Kunststoff Taschenbuch. 29. Ausgabe. Verlag Carl Hanser. 2004.
Bledzki A. K., Gassan J. Natural Fiber Reinforced Plastics in Handbook
of Engineering Plastics. N.P. Cheremisinoff (editor). M. Deccer Inc. New
York. 1997. p. 787-809.
63
Experimental test of threaded steel rods

glued-in hardwood with epoxy
D. Otero, J. Estvez, E. Martn & J. A. Vzquez
Department of Construction Technology,
Advanced Technical School of Architecture, University of A Corua,
Spain
Abstract
This paper summarizes test results of an experimental study on threaded steel
rods epoxied into structural hardwood members.
Two test campaigns were carried out. In the first one 36 test specimens of
Tali sawn timber (Erythropleum ivorense) with threaded steel rods were tested.
Tests arose from the need to evaluate the behaviour of this type of joint because
of its use in the construction of hollow bars spatial mesh of laminated timber.
Three anchorage lengths and two bar diameters were tested. Moreover, two
different thicknesses of the glue line, 1mm and 3 mm, were investigated.
In order to complement the first tests a new experimental study with 180 test
specimens was carried out. In the second tests, double-sided pull-out tests of
Spanish chestnut timber saw specimens were performed. Small diameter
threaded steel rods (8, 10 y 12 mm) quality 8.8 (yield stress 640 MPa) were used.
In the paper the most significant results of both test campaigns are
summarized, as well as the comparison with the existing design formulae for of
this type of union.
Keywords: joint design, glued-in rods, adhesives for wood, hollow bars, spatial
mesh, destructive testing.
Introduction
Our team has been researching for a long time on space frames. Timber space
frames are a particular aspect of this research. Special attention has been paid to
the use of hollow sections in the making of this type of structure. Several patents
are the outcome of our pioneering work in this field.
doi:10.2495/HPSM06007

A Corua University commissioned the design of the roof structure for a
sports hall on the Zapateira Campus. The project became the first built
application of the research. The main aim was to span over this hall with a
stackable modular space frame made up with hollow timber sections (Figure 1).
Achieving an efficient transmission of loads in the bar node is one of the main
problems of this type of structures. The use of hollow sections increases even
more this problem.
The design solution for the bar end connection with the steel node involves
gluing solid timber sections of Elondo or Tali (Erythrophleum ivorense).
Threaded steel rods (either 24 or 27 mm diameter) are then embedded into the
sawn timber and glued with a two component Hilti HIT-RE 500 epoxy resin
(epoxidic bisphenol A/F and poliamid alifatic bases, both with inorganic filling)
(Figure 2).
Given the influence of the efficiency of this connector on the behaviour of the
structure, we decided to supplement the theoretical calculations with
experimental analysis. This is the main aim of this article.
Figure 1: Stacked space frame modules.
Figure 2:
Cut of the end of the

bar to show how the
Elondo block is glued
into the hollow bars of
glulam.
Theoretical research of the joint strength
The only rules references for the calculation of this type of unions appears
collected as informative annex in the Eurocdigo 5, according to:
65
R ax,k = dequ Lg f vk
(1)
-0,2
f vk = 0,8 10-3 1,5
k d
(2)
For rods glued in grain direction, with:

Rax,k
dequ
Lg
fvk
k
d
Characteristic axial load [N]

Lowest of: hole diameter and 1,25 times the diameter of the bar [mm]
Glued-in anchorage length [mm]
Withdrawal strength [N/mm2]
Timber average density [kg/m3]
Nominal diameter of the threaded rod [mm]
In the results section we will show that experimental values result

significantly lower than estimated by Eurocode formulae values. A number of
other design equations were used to compare results. Several references that
provide equations of design have been consulted, showing so much dispersion in
criteria as in results. For the purposes of comparison the formulae provided by
pioneering studies of H. Riberholt [1, 2] was used in this case, as following:
R ax,k = f ws k d Lg ; to Lg 200 mm
(3)
R ax,k = f wl k d Lg ; to Lg < 200 mm
(4)
with:
Rax,k
fws
fwl
k
d
Lg
Characteristic axial load [N]

Strength parameter [N/mm1,5]
Epoxi type adhesives take the value of 0,520
Strength parameter [N/mm1,5]
Epoxi type adhesives take the value of 0,037
Timber average density [kg/m3]
Nominal diameter of the threaded rod [mm]
Glued-in anchorage length [mm]
Materials characteristics and test devices.
In the first experimental campaign one-sided-push compression tests were

carried out over specimens made of sawn timber of Elondo o Tali
(Erythrophelum ivorense) with glued threaded steel rods. Rods were glued only
in one end of the timber specimen. This solution was directly related to the endelements development for hollow glulam bars. It conditioned, in some ways, the
geometric configuration of the specimens and the diameter of the threaded rods
utilized.

Sawn timber blocks, with threaded rods glued-in, were glued into the hollow
bars subsequently. Cross-section of the blocks was determined from the
characteristics of hollow bars. In accordance with this dimensions, cross-section
of test specimens were 75x75 and 90x90 mm.
For each one of the geometric configurations three anchorage lengths, 200,
250 and 300 mm, were studied. These lengths were inside the compatible levels
with the execution of the hollow bars of the mesh.
Timber of test specimens achieved an average density of 911 kg/m3 and a
characteristic density of 856.36 kg/m3, which corresponds to the 5th percentile.
Galvanized threaded steel bars grade 8.8 (with a yield strength of fy=640 N/mm2
and minimum tensile strength of fu=800 N/mm2) have been used. Rod surface
was not treated before bonding, because this is not a usual practice in normal
construction.
Rods with nominal diameters of 24 and 27 mm were used. It address with rod
diameters used in the construction of the space frame hollow sections. Nominal
rod diameters were dependent on the constructive conditioning of the steel nodes
of the mesh too. For that reason they did not agree with values suggested in
Eurocode 5.
The adhesive was the same as the one used in the construction of the space
frame: Hilti HIT-RE 500. Joints glued with two adhesive thicknesses (1 and 3
mm.) have been studied.
The combination of these geometric variables gave as a result 12 types of
different test specimens, whose characteristics are summarized in the table 1.
Adhesive
thickness
Anchorage
length
Specimen
dimensions
axaxL
[mm] [mm] [mm]
Lb
Lm
10 mm
d e
D
a
b
a
1
2
3
4
5
6
7
8
9
10
11
12
24
24
24
24
24
24
27
27
27
27
27
27
26
30
26
30
26
30
29
33
29
33
29
33
1
3
1
3
1
3
1
3
1
3
1
3
[mm]
200
200
250
250
300
300
200
200
250
250
300
300
[mm]
Edge distance
Hole diameter
Embedded rod
slenderness
Rod diameter
Geometric characteristics of the specimens tested in the first

experimental campaign.
Series
Table 1:
L
d
a/2
d
75x75x210
8,33
75x75x260
10,42
75x75x310
12,50
90x90x210
7,41
90x90x260
9,26
90x90x310
11,11
1,56 d
1,67 d
67
The test sample was longitudinally drilled with an auger drill to the centre of
the section. The drill hole went across all the specimen length. The lower end of
the sample was then closed with a 10 mm thick piece of elastic material (elastic
stopper). The adhesive was blown to fill the drilled cavity with a mixing gun.
Afterwards, the threaded bars were inserted manually by continuous pressing and
rotating. The particular design of the test sample has been chosen in order to
enable the test to be carried out in an INSTRON mod. 8805, 1000kN one-sided
pull-compression universal dynamic press (Figure 3). The elastic stopper at the
lower end of the test sample allows the threaded bar to glide down when it
reaches its load limit. For test reliability purposes, the rod exposed end was
fastened to the load cell by means of a nut. Also, the test sample was placed on
the base plate with a 3 mm thick neoprene sheet in between, in order to take the
imperfections of the support.
A picture of the test device is shown in figure 3.
Figure 3:
Test device used in the

first
experimental
campaign.
Figure 4: Test device used in the

second
experimental
campaign.
Like it is shown in results section, experimental results were lower than

theoretical values that were calculated with design formulae. In order to
complement first tests a new experimental study with 135 test specimens were
carried out. In the second tests double-sided pull-out test of Spanish chestnut
(Castanea sativa) timber saw specimens were performed. Test specimens were

made of Spanish chestnut untreated timber. The cross-section dimensions of the
different test specimens were six times the width of the bars diameter.
Specimens lengths were three times the rod anchorage length.
Glulam is the timber that has been utilized generally for this kind of joints,
and the greatest number of the existing formulations are based on it. Because
chestnut timber was a kind of hardwood with an average density very similar to
glulam it was chosen for the second experimental campaign.
Timber of test specimens achieved an average density of 567.28 kg/m3 and a
characteristic density 474.02 kg/m3 (corresponding 5th percentile). Calculating in
design formulae a value of 475 kg/m3 was used.
Small diameter threaded steel rods (8, 10 and 12 mm) quality 8.8 (yield stress
640 MPa) were used. Five different anchorage lengths were tested, from 60 mm.
until 180 mm. three epoxy systems (Hilti, Sika and Locktite) in glueline
thickness of 1mm were investigated. Attending to the geometric parameters,
there were fifteen different types of test specimens.
Geometric characteristics of the specimens tested in the second experimental
campaign are summarized in the table 2.
As is shown by figure 4, the specimens were tested on double-side-tensile
device in the second experimental campaign. To this end, blind drills were
carried out in both ends of the specimen and the threaded steel rods were glued
inside them. For that, as in the previous case, the adhesive was blown to fill the
drilled cavity with a mixing gun and threaded bars were inserted manually by
continuous pressing and rotating.
Both of the experimental studies were carried out increasing the load with a
constant displacement. The failure value was reached in 52 minutes.
Results
The INSTRON test machine has an electronic device that controls the
displacement of the load cell. It also interprets the results with a computer
system that gives the digitalized relationship between load, displacement and test
time-length. The graphic representation of results showed correlation between
stiffness of the three systems adhesives which were used in the second
More repeated failure modes were shear failure in timber, and timber
splitting. For specimens with the lowest diameter the main sample failure was
tensile failure in steel threaded rods.
Using Eurocode 5 and Riberholt formulae the theoretical values for each
series were calculated. Both experimental and theoretical values were plotted to
assess correlation between them.
One representative example of each experimental campaign is shown in the
next figures. Solid lines in the first one (Figure 6) represent theoretical values for
Elondo specimens, with a characteristic density of 865 kg/m3, a rod diameter of
24 mm and a glueline thickness of 1mm, with anchorage lengths from 150 to 350
mm. Crosses show experimental results, corresponding series 1, 3 and 5.
Table 2:
69
Geometric characteristics of the specimens tested in the second

Le
La
La
Li
Le
da
da
Lb
B1b1
L
Total rod
length
Cross-section
side
Outer rod
length
Lb
Timber
specimen
length
Le
Inner gap
length
Anchorage
length
Li
Adhesive
thickness
La
Hole diameter
Rod diameter
da
Series
[mm]
[mm]
[mm]
[mm]
[mm]
[mm]
[mm]
[mm]
[mm]
1a
1b
1c
1d
1e
2a
2b
2c
2d
2e
3a
3b
3c
3d
3e
8
8
8
8
8
10
10
10
10
10
12
12
12
12
12
10
10
10
10
10
12
12
12
12
12
14
14
14
14
14
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
60
90
120
150
180
60
90
120
150
180
60
90
120
150
180
60
90
120
150
140
60
90
120
150
140
60
90
120
150
140
180
270
360
450
500
180
270
360
450
500
180
270
360
450
500
120
120
120
120
120
120
120
120
120
120
120
120
120
120
120
180
210
240
270
300
180
210
240
270
300
180
210
240
270
300
42
42
42
42
42
60
60
60
60
60
72
72
72
72
72
In view of that theoretical values were referred on characteristic values,

experimental results were significantly lower than theoretical ones.
These results could be due to the high density of the Elondo timber compared
with the glulam density (traditionally used in this kind of joints), and to the small
dimensions of the cross-section of the specimens. This cross-section was
dependent on the hollow bars dimensions, which lead to an edge distance lower
than literature suggested ones.
Figure 5:
Main failure modes: shear failure in timber, timber splitting, tensile

failure in steel threaded rods and shear failure in the timber close
interface timber/adhesive.
LOAD [KN]
400
300
EURO
OLT
RIBERH
200
100
150
CODE
200
250
300
350
ANCHORAGE LEGTH [MM]
Figure 6:
Experimental results and theoretical values, according to Eurocode

and Riberholt formulae, to Elondo specimens with threaded steel
rods of diameter 24 mm and a glueline thickness of 1 mm.
71
To chestnut specimens theoretical values to a characteristic density of 475

kg/m3, a rod diameter of 12 mm and a glueline thickness of 1mm, with
anchorage lengths from 40 to 220 mm, are plotted in Figure 7. Crosses show
experimental results, corresponding series 1, 3 and 5. It squares with test series
3a, 3b, 3c, 3d, and 3e, which failure loads are represented with crosses.
These results are more akin to theoretical values, taking into account that
formulae are referred to characteristic values (5th percentile).
60
LOAD [KN]
50
40
30
20
10
T
OL
ERH
RIB
DE
OCO
EUR
0
40
60
90
120
150
180
200
ANCHORAGE LENGTH [MM]
Figure 7:
Experimental results and theoretical values, according to Eurocode

and Riberholt formulae, to chestnut specimens with threaded steel
rods of diameter 12 mm and a glueline thickness of 1 mm.
Results show that more works needs to be done in joints made with rods
glued in high density hardwood timber. This works will allow one to adapt
formulae to this material characteristic and will allow one to study the influence
of edge distance. Our research team is working about it.
Conclusions
Several series of samples have been tested. The test samples were made with
threaded steel rods glued into two different kinds of hardwood timber. The rods
were glued using three different systems of two-component epoxy-base
adhesive.
The test samples were made varying the following parameters: glued
anchorage length, rod diameter and glue thickness
The obtained results have been compared to existing theoretical formulations.
The comparison showed a significant difference between the theoretical
predictions and the test results when high density hardwood is used.
Acknowledgment
This research is sponsored by the Ministry of Science and Technology through
research project titled UNIONES METLICAS ENCOLADAS CON
ADHESIVOS EN BARRAS DE MADERA (Glued Anchored Timber Joints).
The financial support is gratefully acknowledged.
References
[1]
[2]
[3]
[4]
[5]
[6]
Riberholt H., Glue Bolts in Glulam, Department of Structural

Engineering. Technical University of Denmark. Serie R, No 210, 1986.
Riberholt H., Glued bolts in glulam-Proposal for CIB Code. CIB-W18
Meeting. Parksville, Vancouver Island, Canada. Paper 21-7-2, 1988.
Broughton J.G., Hutchinson A.R, Pull-out behaviour of steel rods bonded
into timber, Materials and Structures. ASCE. AUG. 2001. VOL. 127 No.
8, 2001.
Estvez Cimadevila, F.J.; Vzquez Rodrguez, J.A. & Otero Chans, M.D,
Diseo y dimensionado del nudo extremo de una barra hueca de madera
laminada, CIMAD04. 1 Congreso Ibrico A Madeira na Construao.
Universidade do Minho. Guimaraes. Portugal. Pag 689-698, 2004.
Estvez Cimadevila, J. & Vzquez Rodrguez, J.A, Spatial truss of hollow
bars made of laminated timber supported by walls of reinforced masonry,
Journal of the International Association for Shell and Spatial Structures.
Vol. 45 (2004) n.1. April n.144. IASS. ISSN: 0304-3622, 2004.
Estvez Cimadevila, J.& Vzquez Rodrguez, J.A, Edificio sportivo a La
Corua, Spagna, Construire in Laterizio, n.102. Ao XVII. NoviembreDiciembre 2004.Grupo Editoriales Faenza Edotrice S.p.A. Faenza (Ra).
73
Predicting the mechanical behaviour of large

composite rocket motor cases
N. Couroneau
DGA/CAEPE, St Mdard en Jalles, France
Abstract
A method to develop finite element models of the rocket motor cases of a
strategic missile is presented. Based on the use of multivariate analyses, this
approach is made necessary given the impossibility to control all input data:
difficult characterization of filament wound materials, influence of
manufacturing and qualification processes, unknown fabrication parameters, etc.
An initial reference model is built up using preliminary data and theoretical
winding laws. Having compared the models predictions and available structural
test results, a sensitivity analysis is carried out to discuss the individual influence
of the input data on the accuracy of the predictions. The multivariate analysis
finally enables a global assessment of the parametric analysis results.
Keywords: rocket motor case, composite filament winding, finite element model,
variable scattering, multivariate analysis, test prediction.
Introduction
In addition to providing ground test facilities for the development of the future
generation of French strategic missiles, the Centre dAchvement et dEssais des
Propulseurs et Engins (CAEPE) is responsible for evaluating the degree of
performance, durability and security of the solid propellant rocket motors
constituting the missiles. The mechanical expertise work is carried out using
finite element models able to predict the stresses and strains of the mechanical
parts throughout the fabrication and service life.
Many difficulties arise when attempting to describe the mechanical behaviour
of the rocket motor cases. The filament-wound composite structures have
complex geometry and properties especially in the dome area. As the wound
layers are added on a cylindrical mandrel, the curvilinear path leads to a
doi:10.2495/HPSM06008

continuous change in the winding angle and thickness. In addition, the fiber
angle varies in the thickness direction because the fiber path depends on the
surface on which fibers are wound. Other concerns are due to the difficult
characterization of the different materials and the variability of their properties.
The method chosen here is to take advantage of various complex tests
performed on the full scale specimen in order to identify uncertain material
properties or geometrical parameters. This approach, known as the experimental
/ numerical dialog or inverse identification [1], is presented here for the two
largest composite structures of a strategic missile.
Development of the initial model
2.1 Architecture
The solid propellant rocket motors corresponding to the two first propulsion
stages of the French strategic missiles share a common outer diameter (> 2 m).
Their structure is based on the same architecture and the same materials:
- The carbon-epoxy filament-wound pressure vessel, acting both as the
propellant tank and the combustion chamber,
- The two metal polar mountings located at the aft and front openings to
provide connection with the igniter and the nozzle,
- The two cylindrical skirts, made of both carbon-epoxy tissues and
circumferential windings, ensuring the connection with the rest of the
carrier,
- Rubber connections between the vessel and the skirts on the one hand, and
between the vessel and the polar mountings on the other hand.
2.2 Winding law
The composite envelope is constituted of a succession of circumferential
windings on the cylindrical part and satellite windings running between the two
openings. The laminate lay-up in the cylindrical area is rather simple to describe
with an assumed constant value of winding angle and direction of all plies.
Conversely, the complex dome geometry involves a rapid change in angle and
thickness along a meridian. Different netting theories [2] based on geometrical or
mechanical approaches describe this distribution. The planar theory used here
assumes that the fiber patterns lie in a plane which is tangent to the polar opening
at one end and tangent to the opposite side of the polar opening at the other end.
The winding angle calculated (Fig. 3a) with the planar theory at a given
location is applied as a constant value to all plies. The thickness distribution
(Fig. 3a) shows good agreement with the measured values (Fig. 3b). A
preliminary analysis was performed to investigate the effect of an evolution of
the winding angle in the thickness direction, as suggested by Park et al [3]. The
thickness of the first ply was calculated from the mandrel shape whereas the
subsequent plies were calculated with the updated shape. The difference in angle
between plies was particularly noticeable near the polar bosses where the
75
thickness is more important. However this induced no significant change in the

overall calculated displacements.
Figure 1:
Typical components of a rocket motor case (half-view in section).
Figure 2:
Winding path.
a) Winding angle
Figure 3:
b) Thickness
Planar winding law applied to the first stage.
2.3 Finite element models

The finite element analyses considering the geometrical non-linearities are
performed using ANSYS code. The initial axi-symmetric models are developed
over a small angle in order to use the composite dedicated 8-node finite
elements, allowing for a direct input of the materials lay-up and fibers
orientation. The data are transmitted to Ansys by means of a Fortran routine
incorporating geometric parameters, materials properties and winding laws. A
larger density of elements is used near the openings where the winding angle and
thickness vary abruptly.
a) First stage
Figure 4:
b) Second stage
3D mesh using solid elements.
The model contains general boundary conditions for axisymmetry and full
displacement constraints on the frames to take account of the inter-stage skirts
which are not included in the models.
2.4 Evaluation of the models
Two load cases are selected to evaluate the finite element model performance :
internal pressure (Fig. 5a) and axial compression (Fig. 5b). The internal pressure
load case is corresponding to the pressure proof test carried out to accept the
77
structure prior to its filling up with propellant. It is conducted with an internal

pressure 15% higher than the operating pressure and an axial load applied on the
aft polar mounting to represent the thrust loads. The axial compression load case
is applied on the rear frame with the front frame being blocked. It is
representative of an aggression of the missile when stored in the submarine and it
also represents the load transmitted to the second stage during the first stage
flight. These two tests are carried out and analysed by the contractor in charge of
the rocket motor cases design and fabrication.
a) Internal pressure
Figure 5:
b) Axial compression
Load cases for evaluating the models.
a) First stage
b) Second stage
c) First stage
d) Second stage
Figure 6:
Model evaluation for the internal pressure load case.

The results obtained for the internal pressure load case are shown in Figure 6.
The computed and measured displacements are plotted all along a meridian of
the composite vessel. This comparison shows a good agreement except in the
regions near the polar openings at the rear and front. The average difference is
about 1.5 mm for the 1st stage and 1.1 mm for the 2nd stage, with maximum
values of 5 to 6 mm.
For the axial compression load case, only the axial displacements of the skirts
and the cylindrical part of the structure are used, the other values being too small
for comparison. No experimental results are available here so the contractor's
calculations are used instead. The calculations are all in good agreement with an
average difference of about 0.2 - 0.3 mm for the two models.
a) First stage
Figure 7:
b) Second stage
Model evaluation for the axial compression load case.
Optimization of the initial model
3.1 Parameters influence analysis

An analysis is conducted to identify the influence of each parameter considered
individually on the response of the model and the difference between
calculations and experiments. Only the internal pressure proof test is considered
here as there are no experimental data for this load case and the difference with
the contractor calculations is very satisfactory.
The main reasons for discussing the accuracy of the input data of the model
are listed below:
1) Mechanical properties of composite materials : lack of representativity of
the characterization tests performed on uni-directionnal laminate plates with
respect to the actual fabrication process of the structures, variability of materials
properties and possible damage of the matrix and fibers after the pressure proof
test [4],
2) Winding law : Unknown fabrication parameters, theoretical law assumed
without possibility of a direct validation, expected slippage of the fiber during
the winding process,
3) Initial geometry : Possible evolution of the initial geometry during
subsequent fabrication stages [5] or after the pressure proof test.
79
The different parameters above are modified one by one in the reference
model (Tab. 1). For each of them, a possible variation is assumed and each case
results in a new computation being performed. The matrix degradation is
modelled by a decrease of the transverse modulus of the hoop layers, the
winding laws are transformed by an offset or a multiplying coefficient, and the
initial geometry is modified by means of a prior pressurization of the model.
Table 1:
Description of the sensitivity analysis.
The results of the sensitivity analysis are summarized in Table 2. The

modifications leading to an improvement of the results are indicated in bold
characters whereas those leading to a deterioration are printed in italics. The
major factors of influence given the chosen variations of the different parameters
are the structure moduli, the changes applied to the winding laws and the prior
pressure deformation.
The calculations related to a change in the materials properties tend to
indicate an overvaluation of the reference values or a possible matrix
degradation. Concerning the winding laws, the results show a significant latitude
for improving the initial theoretical laws. The 10 bar prior pressure deformation
of the 2nd stage model also results in an important decrease of the difference
between calculations and measurements.
3.2 Multivariate analysis
The multivariate analysis performed using Matlab allows for a global assessment
of the results of the parameters influence analysis. The data processing consists
of a linear system to be solved using the least square method as per eqn (1). The
output data is the best combination of modified parameters to minimize the
difference between experiments and computations :
[A].{x} = {b}
where
(1)
[A] is a matrix containing the results of parameters influence analysis,

{x} is the requested solution,
{b} contains for each of the measurement points the difference between
experimental and numerical values of displacements.

Table 2:
Results of the parameters influence analysis.
The results of the multivariate analysis (Tab. 3) show, on the opposite of the
analysis performed on individual parameters, that no significant modification has
to be applied to the reference materials properties. The 1st stage model can be
optimized by a slight modification of the winding law which was initially
approached by a theoretical law. For the 2nd stage model, the best solution is
essentially obtained through the use of a prior deformation of the model with a
18 bar internal pressure.
Table 3:
Results of the multivariate analysis.
The verification carried out with this new set of input data confirms a
significant improvement of the predictions for the internal pressure load case
(Tab. 4), for the 2nd stage model (Fig. 8b, d). For the 1st stage model (Fig. 8a, c),
the axial behaviour of the rear dome is much improved at the cost of a slight
degradation of radial displacements prediction. The initial calculations for the
compression load case (Fig. 7) are not affected by the modifications.
Table 4:
Figure 8:
81
Comparative predictions for the internal pressure load case.
a) First stage
b) Second stage
c) First stage
d) Second stage
Evaluation of the models for the internal pressure load case.
Conclusion
This study shows the potential interest of multivariate analyses for the
development of mechanical models for structures with uncertainties concerning
the input data (difficult characterization of filament wound materials, influence
of the fabrication and qualification processes, unknown winding parameters,
etc.). The method presented here for two structures and two load cases results in
a considerable improvement of the predictions with a coherent modification of
the input data. This procedure can be broadened to a multiple load case analysis
with non-linear fits for each parameter.
83
Micromechanical modeling of random or

imperfect composites
M. Sejnoha
& J. Zeman
Czech Technical University in Prague, Faculty of Civil Engineering,
Department of Structural Mechanics, Prague, Czech Republic
Abstract
A class of heterogeneous material systems often regarded as random or imperfect
composites is addressed in this paper. The literature now offers a number of contributions that open the way to the analysis of large material systems with complex microstructures while taking advantage of popular micromechanics based
approaches building on periodicity and first order homogenization techniques.
Until recently the attention has been mostly limited to rather classical material
systems such as unidirectional fibrous composites and textiles with emphasis on
various types of imperfections developed during fabrication process on both microscale (level of fiber bundles) and meso-scale (level of textile geometry). From the
basic mechanics point of view, however, it appears logical to exploit the essential
principles of the proposed procedures in bridging the gap between mechanical and
civil engineering applications. In this regard, historical masonry structures classified as systems composed of more than one material component serve as a typical
example of civil engineering applications, which may benefit from standard first
order homogenization schemes extended to account for possibly irregular arrangement of individual stone blocks. In this contribution, both groups of material systems will be treated on the same footing demonstrating the applicability of basic
homogenization techniques as well as similarities between various heterogeneous
material systems when referred to as random or imperfect.
Keywords: random and imperfect composites, textiles, masonry, periodic unit cell,
two-point probability function.
doi:10.2495/HPSM06009
1 Introduction
Techniques of numerical first-order homogenization have acquired a considerable
attention particularly in applications where detailed numerical analysis of complex heterogeneous material systems proved to be prohibitively expensive. Natural
assumption of the existence of periodic microstructure, e.g., periodic distribution
of fibers in the metal matrix or ceramic matrix composites [1], often serve as the
point of departure. The resulting homogenized or effective material parameters are
then used in large scale structural analysis. Unfortunately, in the vast majority of
real material systems the assumption of periodicity generally fails owing to the
presence of various types of geometrical imperfections usually developed during
fabrication. Giving up the benefit of periodic fields in such cases seems, therefore,
reasonable but definitely not very practical. Instead, the recently introduced concept of statistically equivalent periodic unit cell (SEPUC) [2, 3, 4], appears as a
suitable method of attack.
The leading idea of this approach is to replace a complex non-periodic microstructure by a certain periodic unit cell (PUC), which still optimally resembles the
original microstructure in a proper sense. Here, this objective is formalized as a
difference between appropriate statistical descriptors related to the original media
and the periodic unit cell, respectively. If the original microstructure and the periodic unit cell is described by an identical set of parameters, this problem coincides with reconstruction of random materials. If the periodic unit cell is described
by a substantially smaller number of parameters in order to reduce the problem
complexity, it leads to a problem of the best approximation within the selected
statistical descriptors.
It will be shown in subsequent paragraphs that the proposed technique is applicable not only to more or less classical types of composites represented here by
textiles, but also to more conventional class of material systems such as historical
masonry structures.
The principle idea is evident from Figs. 1-3 showing images of real material
or structural systems with corresponding simplified representative volume elements (RVE) presented in terms of certain statistically equivalent periodic unit
cells. When referring to textiles, Figs. 1 and 2, the crucial sources of imperfections
are attributed to generally random arrangement of fibers within the bundle crosssection (microlevel) and the waviness, misalignment and/or non-uniform crosssectional aspect ratio of individual bundles in the longitudinal direction (mesolevel). Qualitatively similar types of imperfections associated with irregular
arrangement of stone blocks, both in terms of block sizes and their location,
together with a variable thickness of the mortar phase are often encountered when
dealing with historical masonry structures. A typical example is plotted in Fig. 3(a)
showing a parapet wall of the Charles Bridge in Prague.
The stepping stone in the analysis of all systems is the possibility to replace
the original color images, Figs. 1(a)-3(a), by their binary counterparts, Figs. 1(b)3(b). The latter representatives of the true micro or meso-structures are further
exploited in the next section when deriving various statistical descriptors. MatchWIT Transactions on The Built Environment, Vol 85, 2006 WIT Press
85
Figure 1: a) Micrograph of random fibrous composite, b) Binary image, c) 5-fiber

and 10-fiber SEPUC.
Figure 2: a) Plane view of textile composite, b) Binary image of transverse section,

c) Mesoscopic SEPUC.
ing the material statistics of the real microstructure with those corresponding to
simplified periodic unit cell in the framework of a certain optimization problem
then allows for deriving the desired geometrical parameters needed in the construction of individual SEPUCs. Knowing the periodic unit cell then opens the way for
the derivation of effective elastic properties using the well known elements of first
order homogenization procedure briefly outlined in Section 3.
Figure 3: a) Parapet wall, b) Binary image, c) Mesoscopic SEPUC.
2 Definition of a periodic unit cell

The crucial step in the definition of optimal periodic unit cell relies on the choice of
proper description of heterogeneous materials with random or imperfect structure.
In the present work, we restrict our attention to two specific descriptors: one- and
two-point probability functions Sr and Srs ; see, e.g., [5] for more details.
To that end, consider a binary heterogenous material formed by phases r and s
and denote the characteristic function of the domain occupied by the r-th phase r .
(When referring to textiles the symbols r and s may represent the fiber (fiber tow)
and matrix phases, while for masonry the two symbols are essentially reserved for
brick (stone) and mortar phases.) Then, the one-point probability function gives
the probability that a point x will be found in a given phase r and the two-point
probability function Srs stands for the probability that the points x and y will be
located simultaneously in phases r and s, respectively:
Sr (x) = P (r (x) = 1),Srs (x, y) = P (r (x)s (y) = 1).
(1)
For the case of statistically homogeneous and ergodic media, information contained in the one-point probability function reduces to the volume fraction of
87
a given phase (Sr = cr ). In addition, the two-point probability function then

depends on (x y) only and can be obtained from the relation

s ,
r
(2)
||Srs = F 1
where || is the area of the analyzed domain, and F 1 () stand for the direct and
inverse Fourier transform and denotes the complex conjugate. For discretized
microstructures, Eq. (2) can be rapidly evaluated by the fast Fourier transform
even for high-resolution bitmaps. In the present study, the bitmaps of Figs. 1(b)3(b) were employed.
Once the original structure has been characterized by an appropriate statistical
descriptor, we can proceed with the definition of the idealized unit cell. A particular parameterization considered in this work appears in Figs. 1(c)-3(c). In case of
masonry structures (for textile composites we refer the interested reader to [2, 4],
Fig. 3(c), the unit cell is fully determined once the width of the unit cell, heights
of each layer of bricks and thicknesses of individual joints are specified. In particular, the geometry of the chosen unit cell is determined by twelve parameters. The
statistically optimal values of these parameters then follow from minimization of
the least square error
E=

2
0
(i, j) Srs (i, j) ,
Srs
(3)
0
where Srs
is the two-point probability function related to the original microstructure while Srs stands for the two-point probability function of the idealized
unit cell. It can be shown that the objective function E is non-convex, multi-modal
and discontinuous due to the effect of limited bitmap resolution. Based on our
previous works, a stochastic global optimization algorithm based on combination
of real-valued genetic algorithms and the simulated annealing method, see [6],
is employed to solve this optimization problem. This approach was successful in
delivering the desired periodic unit cells for all material systems considered herein.
3 First order homogenization

Consider a heterogenous periodic unit cell Y subjected to a uniform macroscopic
(mesoscopic) strain E. In view of the periodicity of the unit cell, the strain and
displacement fields in the PUC admit the following decomposition

(x) = E + u (x) .
(4)
u(x) = E x + u (x),
The first term on the right hand side of Eq. (4) corresponds to a displacement field
in an effective homogeneous medium which has the same overall response as the
composite aggregate. The fluctuating Y -periodic displacement u and corresponding strain enter Eqs. (4) as a consequence of the presence of heterogeneities;
see, e.g., [7] and references therein. Note that the periodicity of u further implies
Figure 4: (a) Localized equivalent strain, (b) Mesoscopic response for 5-fiber and
10-fiber PUCs.
that the average of in the unit cell vanishes. The local stress fields in the PUC
are constrained by constitutive equations and equilibrium conditions
(x) = L(x) : (x),
div (x) = 0,
(5)
respectively. Note that symbol L stands for the (position-dependent) fourth-order

symmetric stiffness tensor. Combining Eqs. (5) and (4) allows us to determine the
distribution of fluctuating displacement u within the unit cell as a function of E
and, subsequently, to evaluate the average stress in the PUC as

1
(u (E)) dY, in particular, = Lhom : E,
(6)
= =
|Y | Y
where Lhom is the homogenized stiffness tensor characterizing the equivalent elastic homogeneous medium.
3.1 Mesoscopic response of polymer matrix based fibrous composites
Application of Eq. (6) to the derivation of mesoscopic response of graphite fiber
tow impregnated by the polymer matrix is shown in Fig. 4. In this particular
example an influence of the number of fibers within the PUC on the mesoscopic
response was explored. While the elastic behavior is essentially independent of the
number of fibers assumed for the PUC (note that even the model with hexagonal
arrangement of fibers gives the same elastic response) the nonlinear behavior suggests possible dependency. This may be attributed to significant non-homogeneous
distribution of local fields manifested, e.g., by highly localized zones of equivalent
local strain, Fig. 4(a), when loading this system beyond the elastic limit. It should
be mentioned that in this particular example the response of the matrix phase was
assumed to be well described by the generalized nonlinear viscoelastic Leonov
model [8].
89
Figure 5: (a) Experimental setup, (b) Simulations for various material parameters of interface elements, (c) Experimental vs. numerical response, (d)
Mesoscopic response.
3.2 Evaluation of effective fracture energy of masonry

The second example demonstrates the use of homogenization theory to the derivation of macroscopic fracture energy as a material property needed in the large scale
analysis of historical structures such as bridges, where detailed modeling of individual phases (stones and mortar) is essentially impossible. This property is found
as the area under the macroscopic stress-strain curve, Fig. 5(d), multiplied by the
PUC area and divided by the total crack length as suggested in [9]. In analogy
with the smeared crack model assumed for individual phases, see [10] for more
details, it can be shown that for the unit cell, Fig. 3(c), loaded by macroscopically
uniform stress or strain in the direction of one of the coordinate axis, say width h,
(all cracks are then assumed to be perpendicular to the loading direction with the
length approaching the other unit cell dimension h1 + h2 + t1 + t2 ) the macroscopic fracture energy receives the value equal to the area under the macroscopic
stress-strain curve multiplied by the crack width h. In such a particular case the

two definitions are not only identical, but yet confirm applicability of the homogenization theory even for quasi-brittle materials.
To arrive at reliable macroscopic response, however, may prove to be rather
complicated. General assumption of perfect and strong bond between individual
phases is usually not acceptable due to weakening effect of air bubbles in the
stone phase in the vicinity of stone-mortar interface. In our particular case, this
zone was represented by interface elements with the behavior governed by the
Mohr-Coulomb failure criterion. The model parameters such as cohesion and angel
of internal friction were derived by matching numerical simulations, Figs. 5(b)(c), with the results derived experimentally, Fig. 5(a). Although more sophisticated techniques based on inverse approach are available in the literature, a simple trial and error method was exercised here to fit individual parameters of the
assumed constitutive model. In particular, a series of possible solutions displayed
in Fig. 5(b) was derived based on randomly generated values of the cohesion and
angel of internal friction. Material parameters corresponding to the best solution, see Fig. 5(c), were then applied in the unit cell analysis to derive the required
macroscopic response to a sufficient degree of accuracy.
4 Conclusion
A rather general approach to the analysis of heterogeneous materials with either
random or imperfect microstructures was reviewed. The basic scheme assumes
formulation of a certain periodic unit cell statistically equivalent (up to two-point
probability function) to real material systems. It is expected that the periodic unit
cell being statistically similar (from the geometrical point of view) to real systems
will also provide similar (at best the same) mechanical response. The robustness of
this approach has been demonstrated through applications to rather different material systems varying from plane weave textiles over to natural stone masonry. A
special attention was further devoted to the derivation of effective fracture energy
of masonry systems to support applicability of homogenization techniques also to
quasi-brittle materials.
Acknowledgment
Grant No. 103/04/1321 and CEZ MSM
The financial support provided by GACR
6840770003 is gratefully acknowledged.
References
[1] Dvorak, G.J., Bahei-El-Din, Y.A. & Wafa, A., Implementation of the transformation field analysis for inelastic composite-materials. Computational
Mechanics, 14(3), pp. 201228, 1994.
[2] Zeman, J. & Sejnoha,

M., Numerical evaluation of effective properties of
graphite fiber tow impregnated by polymer matrix. Journal of the Mechanics
and Physics of Solids, 49(1), pp. 6990, 2001.
91
[3] Sejnoha,
M. & Zeman, J., Overall viscoelastic response of random fibrous
composites with statistically quasi uniform distribution of reinforcements.
Computer Methods in Applied Mechanics and Engineering, 191(44),
pp. 50275044, 2002.
[4] Zeman, J. & Sejnoha,

M., Homogenization of balanced plain weave composites with imperfect microstructure: Part i theoretical formulation. International Journal for Solids and Structures, 41(22-23), pp. 65496571, 2004.
[5] Torquato, S., Random heterogeneous materials: Microstructure and macroscopic properties. Springer-Verlag, 2002.
[6] Matous, K., Leps, M., Zeman, J. & Sejnoha,

M., Applying genetic algorithms
to selected topics commonly encountered in engineering practice. Computer
Methods in Applied Mechanics and Engineering, 190(1314), pp. 1629
1650, 2000.
[7] Michel, J.C., Moulinec, H. & Suquet, P., Effective properties of composite
materials with periodic microstructure: A computational approach. Computer
Methods in Applied Mechanics and Engineering, 172, pp. 109143, 1999.
[8] Leonov, A.I., Non-equilibrium thermodynamics and rheology of viscoelastic
polymer media. Rheol Acta, 15, pp. 8598, 1976.
[9] RILEM TC50 FMC, Determination of the fracture energy of mortar and concrete by means of three-point bend tests on notched beams. Materials and
Structures, 18, pp. 287290, 1985. Endorsed May 1993.
[10] Cervenka,
V., Jendele, L. & Cervenka,
J., ATENA Program Documentation
Part I : Theory. Cervenka

Consulting Company, 2002.
93
Flexural behaviour of ferrocement roof panels

A. S. Alnuaimi, A. Hago & K. S. Al-Jabri
Department of Civil and Architectural Engineering,
Sultan Qaboos University, Sultanate of Oman
Abstract
This paper presents the experimental results of nine roof panels made of
Ferrocement. Two types of channel sections and one type of box section were
tested. All panels were 2m long, 470mm wide and 20mm thick. Channel type A
had side edge beams 95mm deep and channel type B had side edge beams 50mm
deep. The depth of the box section was 95mm. Thin hexagonal wire mesh was
used as reinforcement. The number of wire mesh layers was varied between two
to six. The wires were impregnated midway through the thickness of the panels.
The panels were tested for bending moment with simple supports. The main
variables studied were the number of wire mesh layers, the cross sectional shape
of the panel and the depth of edge beam. Tests revealed that all panels showed
acceptable strength for roofing systems. The increase in the number of wire mesh
layers leads to an increase in the flexural strength. The box section showed
strength similar to that of the channel section with 95mm edge beam. The
channels with 50mm deep edge beams showed strength much less than the ones
with 95mm edge beam and box section.
Keywords: Ferrocement, fibre reinforcement, slab panels, bending, box section
panels, channel panels.
Introduction
Ferrocement is a type of thin wall reinforced concrete commonly constructed of

hydraulic cement mortar reinforced with completely infiltrated, closely spaced
layers of continuous and relatively small size wire mesh. In its role as a thin
reinforced concrete product and as laminated cementbased composite,
Ferrocement has found itself in several applications both in new structures and
repair and rehabilitation of existing structures. This includes all building
members like walls, roofs, columns, beams etc. Compared with the conventional
doi:10.2495/HPSM06010

reinforced concrete, Ferrocement is reinforced in orthogonal directions;
therefore, it has homogenous properties in two directions. Ferrocements low
cost, durability and serviceability were recognized by engineers and builders
throughout the world. It is frequently used for the construction of housing and
buildings, landscape structures, agricultural facilities, public health facilities and
transportation.
In developing countries, the most critical components in dwelling
construction are appropriate roofing, walls and floors. The use of local materials
has not been very successful in producing durable and resistant to fire, insects,
and flood or earthquake roofing materials. As a result, many developing
countries import expensive galvanised iron sheets or use hazardous asbestos
cement sheets as roofing material. Ferrocement appears to be an economical
alternative material for roofing [1, 2].
The use of Ferrocement as roofing and slab elements has been a subject of
investigation by many researchers. For roofing, Ferrocement has been used for
channel type section, folded plates, ribbed slabs, cylindrical shells, circular
domes, funicular shells etc [3]. The use of hollow box section as a roofing
element has been investigated by Mathews et al. [4]. A total of 21 Ferrocement
box sections have been tested under symmetrical line loads applied at one third
span points. The test results confirm that the Ferrocement box hollow sections
have adequate strength, stiffness and other serviceability requirements for
residual applications. Also the theoretical values of cracking load, ultimate load,
deflection and crack width at working load showed good agreement with
experimental values. Kenai and Brooks [5] carried out extensive testing on direct
tensile, four point flexural and drop impact tests on specimens reinforced with
steel wire meshes (13 and 25mm thick) with varied amounts. They used a simple
model based on plastic analysis which was originally proposed by Mansour and
Paramasivam [6]. The model employed a rectangular stress block in the
compression zone and the neutral axis depth was calculated by considering the
equilibrium of tension and compression forces. The ultimate moment was
calculated by multiplying any one of the two forces by the lever arm. Such
models cannot be used in cases where the reinforcing mesh is dispersed in the
middle of the slab. This is because of the small thickness of the Ferrocement slab
panels (about 20mm) which makes it practically difficult to control the uniform
dispersion of the wire mesh through the depth. Ahmed et al. [7] studied the
shear behaviour of Ferrocement channel beams. Their results indicated that
cracking load and ultimate shear strength increase with the increase in the
volume of wire mesh and mortar strength and decrease with the increase of shear
span/depth ratio. Al-Kubaisy and Jummat [8] investigated the use of
Ferrocement in improving the behaviour of reinforced concrete slabs. The
tension zone of each slab was covered with Ferrocement layer. They studied the
effects of the percentage of wire mesh reinforcement in the Ferrocement cover
layer, thickness of Ferrocement layer and the type of connection between the
Ferrocement layer and the reinforced concrete slab on the ultimate flexural load,
first crack load, crack width and spacing, and load-deflection relationship. They
concluded that the use of Ferrocement cover slightly increases the ultimate
95
flexural load and increases the first crack load. Considerable reduction in crack
widths and spacing was observed with specimen with Ferrocement layers. The
performance of Ferrocement panels under normal, moderate and hostile
environments was investigated by Masood et al. [9]. They concluded that the
flexural capacity of the panel increases with the addition of fly ash. Considerable
deterioration of wire meshes fabric was observed due to sustained exposure in
saline casting and curing condition. Recently, Hago et al. [10] conducted 6
experimental tests to study the ultimate and service behaviour of Ferrocement
roof slab panels. The parameters studied include: the effect the effect of the
percentage of wire mesh reinforcement by volume and the structural shape of the
panels on the ultimate flexural strength, first crack load, crack spacing and loaddeformation behaviour. The results demonstrated that the monolithic shallow
edge Ferrocement beams with the panels considerably improves the service and
ultimate behaviour, irrespective of the steel layers used. Also, slabs with channel
sections supported larger ultimate loads and behaved better under service loads
than their flat slabs counterparts. Due to large deflections experienced by the
thin panels, large deflection theory was adopted in the analysis. Good agreement
was obtained between the theoretical and experimental ultimate loads using the
proposed mathematical model.
In this research, nine simply supported slab panels were tested for flexure.
The specimens were arranged in three categories based on the cross-section:
channel section type A, channel section type B and box section. The aim was to
study the effects of the shape of cross-section and the number of wire mesh
layers on the behaviour and ultimate capacity of the tested panels. The panels
were constructed manually in a simple manner, so similar panels can be
constructed and used as roofing system with almost no equipment needed.
Test program
The experimental investigation consisted of fabricating and testing, for flexure,

nine Ferrocement roof panels. All panels were 20mm thick and were reinforced
with thin steel wire meshes sandwiched midway through the thickness. The
panels were divided into three groups according to their shape and number of
wire mesh layers (Table 1). The first group, Channel A, consisted of three
channel-shaped panels. The dimensions were 470mm outer widths and 2100mm
length with two edge beams 95mm deep (Figure 1). The second group, Channel
B, consisted of three channel-shaped panels similar to the first group except that
the edge beam was 50mm deep (Figure 2). The third group consisted of three
box section panels with 470x 95mm outer cross section and 2100mm total
length. The hollow core was 430x55mm as shown in Figure 3. In all tested
panels, the test span was the middle 600mmm of the span.
Material used
Ordinary Portland cement and natural sand were used in making the Ferrocement
concrete in the ratio of 1:2 respectively with a water to cement ratio of 0.55. The

average mortar cube compressive strength was 47.1N/mm2 and the average
prism flexural strength was 5.23 N/mm2 (Table 1). For reinforcement, a
hexagonal wire mesh with closely spaced wires was used in the tested panels.
The wire mesh had a diameter of 0.579mm and a spacing of 12.13mm in both
directions. The number of wire mesh layers varied from two layers to six layers.
The wire mesh was stretched on a frame of 6mm steel bars having yield strength
of 250N/mm2 (Figures 1-3). The cement, sand and water were mixed using a
power driven drum mixer for about five minutes. The mortar was designed to
give 28day strength of about 40N/mm2. Wooden moulds were used to cast the
slabs. A layer of mortar of about 10mm thick was first placed in the mould
followed by the reinforcement cage and then a second layer of mortar was placed
to make the required thickness. Due to the small thickness of the panel, the wire
mesh was placed almost at mid thickness. With each panel, six 100mm cubes
and two prisms were cast to determine the mortar compressive strength and
modulus of rupture. After one day of casting, the panels and cubes were removed
from the moulds and were kept under wet Hessian cloth until the day of testing
which was about 28 days from the date of casting.
Table 1:
Models tested and their material properties.
Model
No.
Dimensions
(mm)
Depth of
edge beam
(mm)
No. of
steel
layers
% Volume
of steel
Ch2-A
Ch4-A
Ch6-A
Ch2-B
Ch4-B
Ch6-B
Box1
Box2
Box3
2100x470x20
2100x470x20
2100x470x20
2100x470x20
2100x470x20
2100x470x20
2100x470x20
2100x470x20
2100x470x20
95
95
95
50
50
50
95
95
95
2
4
6
2
4
6
2
4
6
1.36
1.60
1.76
1.36
1.57
1.77
1.24
1.43
1.62
Compressive Flexural
strength
strength
(N/mm2)
(N/mm2)
54.9
46.0
47.5
42.6
40.6
42.0
44.1
31.0
54.6
5.0
4.8
4.6
6.6
6.8
6.6
5.3
5.5
7.2
Test procedure
All slabs were tested for flexure. They were simply supported with a clear span
of 2000mm and test span of 600mm in mid-span. The load was applied as two
symmetrically arranged concentrated loads, using a spreader steel beam and a 5ton hydraulic jack. The load was measured using an electric load cell of 50kN
capacity and was applied in increments of 0.5kN. The slabs were painted using
white emulsion to assist in detecting the cracks. Deflection under the centre of
the slab was measured using Linear Variable Displacement Transducers
(LVDT). The load cell and LVDT were connected to a data acquisition system.
Surface concrete strains were measured using a digital DEMEC gauges. At each
load increment, careful search was made for cracks on all sides of the slab with
the aid of a magnifying glass and a powerful electric lamp. The crack spacing,
97
the number of cracks, the extent of the cracked zone over the length of the slab
and the ultimate load were all noted. The failure load considered in this
investigation was the load value after which the panel ceases to resist additional
load or the load measured just before sudden collapse. Figure 4 shows typical
tested panel.
1R6
20m
1R6
95m
Wire mesh layers

1R6
20m
1R6
470mm
Wire mesh layers
1R6
1R6
1R6
50mm
1R6
20m
Channel type A section, Ch A.
20mm
Figure 1:
470mm
Figure 2:
Channel type B section, Ch B.
470mm
Figure 3:
95m
Wire mesh layers
20m
1R6
55m
20mm
Box B section, Box.
Along with each panel, six 100x100x100mm cubes were tested for
compressive strength and two 100x100x500mm prisms were tested for modulus
of rupture.
Figure 4:
Test rig and typical panel tested.
3.5
Cracking load (kN)
3.0
2.5
2.0
1.5
Channel A
1.0
Channel B
Box
0.5
0.0
2
4
No. of Mesh Layers
Figure 5:
Cracking load.
5 Experimental observation
5.1 Cracking load
Figure 5 shows that all sections cracked at loads close to each other when the
number of layers was 2. The box section cracking load was less than both
channel sections when the number of layers was increased to 4 or 6. In general,
as the number of wire mesh increases the cracking load increases in all shapes.
99
5.2 Failure load

The box section and channel A section had almost similar failure loads which
were higher than the failure loads of channel B section for all number of layers
as shown Figure 6. The ultimate load was increased with the increase of the wire
mesh layers in all shapes. The increment was less pronounced in the case of
channel B.
9.0
Ultimate load (kN)
8.0
7.0
6.0
5.0
4.0
3.0
Channel A
2.0
Channel B
1.0
Box
0.0
2
4
No. of Mesh Layers
Figure 6:
Failure load.
Max. deflection (mm)
60.0
50.0
40.0
30.0
Channel A
20.0
Channel B
10.0
Box
0.0
2
No. of Mesh Layers
Figure 7:
Maximum deflection.
5.3 Deflection
Figure 7 shows maximum vertical deflection at mid-span. It is clear that when
the number of wire mesh layers was 2, all panels had close to each other
deflection values. With exception to channel B, the maximum deflection was
decreased with the increase of the wire mesh layers from 2 to 4 and 6. Figure 8
shows that, in all panels, as the number of wire mesh increases the deflection
reduces for the same load. Figure 9 shows that there are no major differences in
the behaviour of channel A and box section but channel B behaved in a softer
manner, more deflection for same load.
Load(kN)
Channel A (95mm edge beam)
8
7
6
5
4
3
2
1
0
Ch-2A
Ch-4A
Ch-6A
10
20
30
Deflection(mm)
40
50
(a)
Channel B (50mm edge beam)
Load(kN)
5
4
3
Ch-2B
Ch-4B
Ch-6B
1
0
0
10
20
30
40
Deflection(mm)
50
60
(b)
Load(kN)
9
8
Box-2
Box-4
Box-6
5
4
3
2
1
0
0
10
20
30
40
50
Deflection(mm)
(c)
Figure 8:
(a): Effect of the number of wire mesh layers on channel A

deflection, (b): Effect of the number of wire mesh layers on
channel B deflection, (c): Effect of the number of wire mesh layers
on Box deflection.
101
Load(kN)
2 wire mesh layers

5
4.5
4
3.5
3
2.5
2
1.5
1
0.5
0
Ch-2A
Ch-2B
Box-2
10
20
30
Deflection(mm)
40
50
(a)
4 w ire mesh layers
Load(kN)
6
5
4
Ch-4A
Ch-4B
Box-4
1
0
0
10
20
30
40
50
60
Deflection(mm)
Load(kN)
(b)
6 w ire mesh layers
9
8
7
6
5
4
3
2
1
0
Ch-6A
Ch-6B
Box-6
10
20
30
40
50
60
Deflection(mm)
(c)
Figure 9:
(a): Effect of shape on deflection with 2 wire mesh layers,

(b): Effect of shape on deflection of panels with 4 wire mesh layers,
(c): Effect of shape on deflection of panels with 6 wire mesh layers.
Conclusion
Nine roof panels were tested for pure bending. Results from two types of channel
sections, channel A and channel B, differing in the depth of the edge beam
(95mm and 50mm) and one type of box section were compared. The number of
wire mesh layers was varied from 2 to 6. Results show that channel type A
behaved in a similar way of the box section with close to each other failure loads
and deflection while channel type B was softer regardless of the number of wire
mesh layers. All panels showed acceptable cracking and failure load for roofing
systems.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
Nedwell P.J and Swamy R.N., Ferrocement, St Edmundsbury Press, Bury

St, Edmunds, Suffolk, 1994.
Swamy R.N, Concrete Technology and design, Bell and Bain (Glasgow)
Ltd, 1984.
Al-Sulaimani, G.J., and Ahmad, S.F., Deflection and flexural rigidity of
ferrocement I and Box beam. Journal of Ferrocement 18(1):1-12, 1988.
Mathews, M.S., Sudhakumar, J. Sheela, S. and Seetharaman, P.R.,
Analytical and experimental investigations of hollow ferrocement roofing
units, Journal of Ferrocement 21(1): 1-14, 1991.
Kanai, S, and Brooks, J.J, Tensile, Flexual and Impact Behaviour of
Ferrocement with chicken wire mesh reinforcement, Proceedings of the
Fifth International Symposium on Ferrocement, UMIST, Manchester,
pp.342-355, 1994.
Mansour, M.A. and Paramasivan, P, Cracking behaviour and Ultimate
strength of Ferrocement in flexure, Proceedings of the Second
International Symposium on Ferrocement, Bangkok, Thailand, pp.47-59,
1985.
Ahmed S.F., Saroash H. Lodi, and Juneid Qureshi, Shear behaviour of
Ferrocement thin webbed sections, Cement and Concrete Research,
Vol. 25, No. 5, pp.969-979, 1995.
Al-Kubaisy, M.A. and Jummat, M.Z, Flexural behaviour of reinforced
concrete slabs with Ferrocement tension zone cover, Journal of
Construction and Building Materials, Vol 14, pp 245-252, 2000.
Masood A., Arif M, Akhtar S. and Haquie M, Performance of
Ferrocement panels in different environments, Cement and Concrete
Research, Vol. 33, pp. 555-562, 2003.
Hago, A.W., Al-Jabri, K.S., Al-Nuaimi, A.S., Al-Moqbali, H., and AlKubaisy, M.A, Ultimate and service behaviour of ferrocement roof slab
panels, Construction and Building Materials 19:31-37, 2005.
103
Finite element modeling of actuated

fibre composites
M. Martinez1, A. Artemev2, F. Nitzsche2 & B. Geddes2
1
National Research Council of Canada (NRC),

Institute for Aerospace Research, Ottawa, Ontario, Canada
2
Department of Mechanical and Aerospace Engineering,
Carleton University, Ottawa, Ontario, Canada
Abstract
A finite element method (FEM) study of the actuation and sensing performance
of actuated fibre composites (AFC) is presented. The effect of non-continuous
fibres on the AFC performance was analyzed for actuation and sensing
applications. The results of the FEM analysis of AFC with non-continuous fibres
are compared to experimental results obtained in specimens with fibres damaged
by large deformation. A study of the change in the polarization state in the fibre,
resulting from the formation of a gap, was performed and its results were
incorporated into the FEM. The correlation between the available experimental
data and simulation results is discussed.
Keywords: PZTs, actuated fibre composites, AFCs, piezoelectric fibres, sensors,
smart rotor.
Introduction
Modern military and civilian aircraft structures require the use of monitoring
systems to identify and predict the health of critical components within the
aircraft. Health Prognostic Monitoring (HPM) is in the process of being
introduced to modern aircraft structures. This concept will allow the aircraft to
determine if the life of a critical component is coming to an end, and have the
overall system schedule the time and place for the component to be
repaired/replaced. Sensors are required to acquire the current state of strain,
temperature or other parameters of concern in specific areas of the aircraft,
which is a critical part of HPM. Several sensors show promise for these
doi:10.2495/HPSM06011

applications, such as, strain gauges, fibre optic strain sensors and piezoelectric
sensors. The use of piezoelectric sensors and actuators embedded in laminated
composites is also being studied as a means of reducing noise and vibration
generated by helicopter rotors. A Smart Blade structure would be capable of
reacting to an external electric stimulus measured by AFC sensors and allow for
the deformation of the rotor blade via AFC actuation. This paper concentrates
on the use of piezoelectric fibres as strain sensors. The Smart Blade tailors the
aerodynamic load distribution to attenuate both noise and vibration at flight
conditions spanning from vertical take-off, manoeuvring flight, high-speed
cruise and vertical landing.
The use of piezoelectric materials for both actuation and sensing functions
has made them attractive for use in micro electro mechanical systems [1] as well
as for large-scale applications. For example, piezoelectric materials are being
analyzed for active control of airfoils [24]. However, bulk piezoelectric
materials have several serious drawbacks impeding their applications. In
particular, these ceramic materials have inherently low toughness. A number of
brittle materials have been used, in the form of fibres, in fibre-reinforced
composites. Such materials allow for the use of high strength fibres together with
a high toughness developed in a composite structure. Recently, several attempts
have been made to develop composite materials with piezoelectric fibres [57].
Such actuated fibre composites (AFC) can be used as actuators and/or sensors in
a number of different applications.
The use of AFC for practical applications is still restricted by the lack of
detailed knowledge about the relationship between the structure, manufacturing
process, and performance of the material. Currently, for several AFC systems,
research is being conducted on the relationships between structures, properties of
the constituent components, and the behavior of the system as a whole [810].
One particularly challenging problem is the analysis of long-term behavior and
degradation of AFC materials. Studies into the effect of fibre damage on the
performance of AFC are an integral part of this research [1112]. The presence
of gaps within the fibre has significant consequences for the polarization field of
the fibre. This change in the proximity of the gap size has a major impact on the
electric potential distribution and, thus, on the overall response obtained from the
AFC. This study has been carried out using a custom developed FEM
solver [13].
Domain patterns in broken fibres
A phase field model of ferroelectric materials developed in [14] was used to

study the effect on the polarization of gaps in fibres. In this model the
distribution of the order parameter vector describes the state of the system.
The polarization at any point is determined as P = P0 . The free energy is:
3
F = p
2 p =1
)2 + f () d 3r + 2P02 3
p =1 q =1
d 3k k p k q ~
p (k )~q (k )*
3
2
(2 ) k
105
where is the gradient energy coefficient, is Kronecker delta symbol, f ()

is the chemical energy, P0 is the equilibrium polarization, ~ is the Fourier
transform of , and * indicates a complex conjugate. We used:
2
f () = 0.136 i2 7.076 i4 + 4.352 i2 + 1.86 i2

i
i
i
and other model parameters characterized by the dimensionless combinations
= P02 / f = 10 and / f l 2 = 0.5 , where f is the difference between the

chemical energies of paraelectric and equilibrium ferroelectric phases. The unit
length is equal to the mesh cell length. A 512x64x64 equilateral rectangular
mesh with a fibre (diameter of 48 and length of 512 cells) was used. Fibres with
different gap lengths were simulated. Initially the fibres had homogeneous
longitudinal polarization (Fig.1(a)). The equilibrium domain patterns were found
using the fast Fourier transform for the solution of the stochastic time-dependent
Ginzburg-Landau equation [14] for the minimization of free energy.
Figure 1:
(a) Domain patterns in the broken fibre in the initial state and (b)
after relaxation in fibres with a 6 cell long gap and (c) with a 9 cell
long gap. The polarization direction is shown by arrows.
Figure 1 illustrates a longitudinal section of the system in the initial state

(Fig. 1(a)) and after the completion of the simulation (Fig. 1(b) and (c)). The
analysis shows that the presence of the gap results in the appearance of 90o
closing domains at the ends of the fibre and 180o domains with reverse
polarization extending along the fibre from the fibre end. The formation of this
domain structure leads to a partial decrease in the overall polarization near the
fibre ends. The simulation shows that increasing the gap length results in an
increase in the degree of reduction of the average polarization and the increase in
the length of the affected zone. Computational constraints do not allow
simulation of fibres with realistic dimensions. However, the simulation results
clearly indicate that the formation of a polydomain zone with a reduced poling
effect can be expected near the ends of broken fibres. The depth of such a zone
for real fibres can be estimated as approximately equal to the fibre diameter.
Finite element models of AFC
Two Finite Element (FE) models were constructed. The first model, for the
simulation of sensing applications, consisted of an AFC subjected to a strain of
0.9 micro strains. The second model for the simulation of actuation applications
represented an AFC subjected to 900V. In both models the effect of gaps with
and without a depolarized field was studied. The models were constructed using
the material properties for PZT-5A fibres, an epoxy matrix and two Kapton
layers. The piezoelectric fibres in the AFC model have the following
piezoelectric parameters: d31 of 44010-12 C/N, d33 of 18510-12 C/N, and d24 and
d15 of 56010-12 C/N. Mechanical properties were represented by C11=120 GPa,
C33=110 GPa C44=21 GPa, C66=23 GPa, C12=75.2 GPa and C13=75.1 GPa [13].
The epoxy had a Youngs modulus of 2.58 GPa and Poissons ratio of 0.38. The
Kapton had a Youngs modulus of 2.5 GPa and Poissons ratio of 0.34.
3.1 Sensing AFC Model
Figure 2:
(a) The AFC model used for simulation of the sensing application
shown with an epoxy matrix, (b) inner components of the AFC
model.
The sensing AFC system was set to 2100m in length, 1200m in width and
260m in thickness. It consisted of four piezoelectric fibres of 254 m diameter
and an epoxy matrix contained between two Kapton layers, as shown in
Fig. 2(a). Figure 2(b) shows the same model with the Kapton removed and a
transparent epoxy matrix. An assumption was made that due to the presence of a
30 m gap, the regions of the fibre next to the gap with the length equal to the
fibre diameter, had a zero poling effect. This means that piezoelectric constants
in these regions were equal to zero; however, elasticity constants were the same
as in the rest of the fibre. A defined strain of 0.9 micro strains was applied in an
extension mode. The electric potential difference was measured between the top
and bottom of the fibres.
3.2 Actuation AFC model
The unconstrained actuation AFC system was set to 4600m in length, 1200m
in width, and 260m in thickness. It consisted of four piezoelectric fibres, an
107
epoxy matrix, and copper inter-digitized electrodes (IDE) that were covered by a
thin layer of epoxy. Figure 3(a) shows a view of the AFC model including the
epoxy matrix. Figure 3(b) shows the same AFC with the epoxy matrix made
transparent to show the fibres and the copper electrodes. PZT fibres with 250m
diameters were spaced 300m center to center. The AFC was cantilevered at one
end. The fibres had a varying polarization field along their length. The fibre
under the electrodes was modeled to have a zero polarization.
Figure 3:
(a) AFC with epoxy matrix, (b) inner components of AFC,

(c) cross-sectional view of AFC at the electrodes showing a 60
contact angle between the electrode and the PZT fibre.
The fibres between the electrodes have a polarization along either positive or
negative x-direction, as shown in Fig. 3(b). Switching of the polarization
direction permits the AFC to extend or contract, since the direction of the electric
field also changes from one electrode pair to another. This type of AFC structure
and actuation method were presented and studied in [56]. The four copper
electrodes are spaced 900m apart and are 200m wide and 10m thick. The
electrodes in Fig. 3(b) are labeled E1, E2, E3 and E4. The electric potentials for
E2 and E4 are set to zero, while E1 and E3 are set to electric potentials of 900,
1350, and 1800 volts for different simulation runs. The copper electrodes make
direct contact with the fibres through an angle of 60, as shown in Fig. 3(c). In
order to study the effect of broken fibres on the actuation performance of the
AFC, we created models with different numbers of gaps (0, 1, 2, 5, 10 and 20)
introduced into fibres. We used models in which such gaps were introduced
without changing the polarization state in the rest of the fibre volume, as well as
models in which gaps were accompanied by depolarized regions at the ends of
the fibres with a length equal to the fibre diameter.
Results of FEM analysis
4.1 Results of sensing application analysis

Figure 4 demonstrates the electric potential distribution in the AFC under the
applied mechanical strain. The results obtained by the simulation of the sensing
application of the AFC indicate that the change in the electric potential between

the upper and lower surface of the system in the proximity of the gap drastically
differs from the voltages developed in other regions of the AFC.
Figure 4:
(a) AFC under an applied strain, (b) cross-section of AFC through

the gap in the fibre.
As seen in Fig. 4, the electric potential difference through the thickness of the
fibre is close to zero in the gap and in the depolarized region near it. Such loss of
voltage can be a problem since discontinuities in the AFC tend to appear in the
proximity of the electrodes used to apply voltage to fibres [15]. If the same
electrodes were used for sensing, then they would detect a voltage corresponding
to a strain that is significantly lower than the actual strain. In such a case a dual
electrode systems can improve sensing performance. The first electrode group
may be used for poling of the piezoelectric fibres, while the second electrode
system would be used to monitor voltages produced by the piezoelectric fibres. It
should be noted that in the AFC model where gaps in the fibres were not
accompanied by depolarized regions, the width of the dead zone in the voltage
distribution was much smaller (approximately equal to the gap width) and the
loss of voltage less severe.
4.2 Results of actuation application analysis
The FEM analysis of the AFC under applied voltage (Fig. 5) showed that gaps in
fibres can reduce the actuation performance. If a fibre between two IDEs is
broken, then the electrical field is concentrated in the gap, and two pieces of fibre
between electrodes have a very low electrical field applied, producing no
contribution to the actuation. However, any significant reduction in the overall
109
Percentage of Displacement
actuation performance of the AFC is obtained only when a large number of gaps
is introduced effectively creating a short fibre composite. It is unlikely that such
a density of damage can be developed before overall failure. The FEM results
also indicate that there is very little difference between systems with
depolarization regions added to the gaps in fibres and systems with gaps only
(without depolarization regions).
100.00
80.00
60.00
40.00
20.00
0.00
0
10
Number of Gaps
15
20
Fully polarized
Depolarized fibers
(b)
Figure 5:
(a) Electric potential in the AFC under voltage applied to IDE,

(b) the effect of the number of fibre gaps on the actuation
performance of the AFC for systems with and without depolarized
regions near gaps.
Discussion and conclusions
An experimental study of the degradation of AFC sensing properties caused by

the damage produced by large strains [12] has shown that a significant loss of the
sensing capacity can be produced. However, the sensing capability is almost
completely restored when the applied load is reduced. Simulation results
demonstrate a possible mechanism for such a reversible sensing degradation.
Under a large imposed strain, broken fibres can contain wide gaps producing
large depolarized regions, which result in a significant loss of the sensing
function. When the load is reduced, the gaps in the fibres are closed and a fully
poled state can be restored, resulting in a better sensing performance even though
fibres remain mechanically broken. This simulation also demonstrates that
damage in the fibres can be detrimental to both actuation and sensing
performance of the AFC; however, any significant loss of the actuation

capability can be expected only when a large density of gaps in fibres is
produced.
Acknowledgement
The authors gratefully acknowledge the financial support of Natural Sciences
and Engineering Research Council of Canada (NSERC).
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
Banks H.T., Smith R.C., Wang Y., Smart Material Structures - Modeling,
Estimation and Control. John Wiley & Sons, 1996.
Bent A. A., Active Fibre Composites for Structural Actuation, Ph.D.
Thesis, Massachusetts Institute of Technology, January 1997.
Hall S.R., Prechtl E.F. Smart Materials and Struct. 5 (1996) 26.
Flinn, E.D. Aerospace America 37 (1999) 40.
Odegard G.M., Modeling of Piezoelectric Polymer Composites National
Institute of Aerospace, Hampton, Virginia. NASA/CR-2003-212681NIA
Report No. 2003-10
Wilkie W.K., et al., NASA Langley Research Center Macro-Figer
Composite Actuator (LaRC-MFC): Technical Overview,
Brei D., Cannon B.J., Composites Science and Tech. 64 (2004) 245.
Sporn D., Schoenecker A., Mat. Res. Innovat. 2 (1999) 303.
Pettermann H.E., Suresh S., Int. J. of Solids and Struct. 37 (2000) 5447.
Wilkie W., High J., Bockman J., Reliability Testing of NASA
Piezocomposite Actuators, U.S. Army Research Laboratory and NASA
Langley Research Center, Hampton, Virginia, USA.
Wickramasinghe V.K, Hagood N.W., Smart Mater. Struct. 13 (2004)
11551165 PII: S0964-1726 (2004) 83334-2.
Melnykowycz M., et al., Performance of Integrated Active Fibre
Composites in Fibre Reinforced Epoxy Laminates. to appear in Smart
Materials Structures.
M. Martinez, et al., Finite Element Analysis of Actuated Fibre
Composites. Proceedings of CANSMART 2005, Toronto, Canada (2005)
231.
Semenovskaya S., Khachaturyan A.G., J. Appl. Phys., 83 (1998) 5125.
Wickramasinghe V.K, Hagood N.W., J. of Aircraft, 41 (2004) 931.
111
CFRP strengthening of prefabricated

timber panel walls
M. Premrov & P. Dobrila
University of Maribor, Faculty of Civil Engineering, Slovenia
Abstract
This paper provides an experimental analysis of timber-framed walls, coated
with carbon fibre-reinforced polymers (CFRP) strengthened fibre-plaster boards,
usually used as main bearing capacity elements in the construction of
prefabricated timber structures. The tensile strength of the boards is lower than
the compressive one and essentially lower than the strength of the timber frame,
therefore it is convenient to strengthen boards in their tensile diagonal direction
with high-strength materials in order to gain a higher capacity. It has been shown
that the inclusion of CFRP diagonal strip reinforcement on the load-carrying
capacity can be quite high and that it is maximized when the carbon strips are
connected to the timber frame. On the other hand, the ductility itself was not
significantly improved. The test samples proved an important distinction in
behaviour in timber frame-fibreboard connecting area, dependant on the
boundary conditions between inserted CFRP strips and timber frame.
Keywords: timber, frame walls, CFRP, fibre-plaster boards, experiments.
Introduction
There is an increasing tendency worldwide toward building multi-level

prefabricated timber structures with timber-framed walls as the main bearing
capacity elements. Their load-carrying capacity becomes critical, especially
when taller structures are subjected to heavy horizontal forces, particulary with
structures located in seismic and windy areas. In this case it is sometimes
necessary to reinforce the walls. The treated wall is a composite element
consisting of framed panels made from sheets of board-material fixed by
mechanical fasteners to one or both sides of the timber frame (Figure 1). There
are many types of panel products available which may have some structural
doi:10.2495/HPSM06012

capacity such as wood-based materials (plywood, oriented strand board,
hardboard, particleboard, etc.) or plaster boards and, more recently fibre-plaster
boards. In the following analysis we limited our attention to the fibre-plaster
boards (FPB), recently the most frequently used in Central Europe. One of the
most important reasons for an increased application of these types of gypsum
products is their relatively good fire protection. Additionally, gypsum is a
healthy natural material and is consequently particularly desired for residential
buildings.
In structural analysis panel walls for design purposes can be regarded
separately as vertical cantilever beams with the horizontal force (FH=FH,tot /n)
acting at the top (Figure 1). Considered supports approximate an influence of
neighbouring panel walls and assure an elastic-clamped boundary condition for
the treated wall (Faherty and Williamson [1] or Eurocode 5 [2]).
FH,tot
FH
b
nb
FH =
Figure 1:
b
timber frame
(the studs)
FH ,tot
n
coating boards
Static design and cross section of the treated panel wall.
Design models
2.1 Shear model

Many design models have been proposed in order to analyse and predict the
behaviour of wood-based shear walls and diaphragms subjected to lateral loads.
Kllsner [3] and kerlund [4] proposed an agreeable approach to determine the
load-carrying capacity of the wall unit, based on the following key assumptions:
-
behaviour of the joints between the sheet and the frame members is
assumed to be linear-elastic until failure,
113
the frame members and the sheets are assumed to be rigid and
hinged to each other.
The influence of shear deformations in the fibreboard can be additionally
estimated by introducing the shear angle. Additionally, two models are presented
based on the assumption that the load-displacement relation of fasteners is
completely plastic. Kllsner and Lam [5] presented the walls load-carrying
capacity as a function of fasteners spacing along the upper horizontal timber
member assuming constant fastener spacing along all timber members.
Two simplified computational methods are given in the final draft of
Eurocode 5 [2] in order to determine the load-carrying capacity of the wall
diaphragm. The first simplified analysis Method A, is identical to the Lower
bound plastic method, presented by Kllsner and Lam [5]. This method defines
the walls shear resistance (Fv,d) as a sum of all the fasteners shear resistances
along the loaded edges in the form of:
-
bi
ci
s
Ff,Rd . lateral design capacity per fastener,
bi .. wall panel width,
s .. fastener spacing,
Fv ,d = F f ,Rd
1 for bi b0
ci = bi
b for bi b0
0
where b0 = h/2
(1)
(2)
This is only an approximated and simplified definition, which can be

applicable for wood-based panels where the strength is relatively high and the
elements tend to fail because of fastener yielding.
The second simplified analysis Method B is applicable to walls made from
sheets of wood-based panel products only, fastened to a timber frame. The
fastening of the sheets to the timber frame should either be by nails or screws,
and the fasteners should be equally spaced around the perimeter of the sheet.
According to Method A the sheathing material factor (kn), the fastener spacing
factor (ks), the vertical load factor (ki,q) and the dimension factors for the panel
(kd) are included in the design procedure in the form of:
bi
9700 d
ci k d k i ,q k s k n ;
s0 =
s0
k
k . characteristic density of the timber frame
Fv ,d = F f ,Rd
d fastener diameter,
(3)
2.2 Composite model

All the above mentioned methods are usually unsuitable for treated walls
sheathed with fibre-plaster boards (FPB). The main assumptions do not exactly
coincide with the real state of FPB, in which the tensile strength is evidently

lower than the compressive one. Consequently, cracks in a tensile zone usually
appear under heavy horizontal loads before stresses on the fasteners reach their
yielding point, and the fibreboards do not behave usually as rigid elements
(Dobrila and Premrov [6]). However, by employing FPB as a coating material, a
horizontal load shifts a part of the force over the mechanical fasteners to the
fibreboard and the wall acts like a deep beam. Distribution of the horizontal force
by composite treatment of the element depends on the proportion of stiffness.
The effective bending stiffness (EIy)eff of mechanically jointed beams taken from
Eurocode 5 [2] can be written in the form of:
n
( EI y )eff = E i I yi + yi Ai a i2 =
i =1
ntimber .
i =1
(E
I yi + E i
2
yi Ai a i timber
nboard
(E i I yi )board
(4)
j =1
where n is the total number of elements in the considered cross-section and ai is a

distance between global y-axis of the whole cross-section and local yi-axis of the
i-th element with a cross-section Ai (see Figure 2). It is evident that the force
distribution in this case strongly depends on the stiffness coefficient of the
connecting area (y), which mostly depends on the fasteners slip modulus (Kser)
and fasteners disposition, as well as on the type of the connection. An
experimental analysis on the influence of fasteners spacing on behaviour of the
treated walls can be found in [6].
Strengthening of fibre-plaster boards
As described, the FPB are usually a weaker part of the presented composite
system, because their tensile strength is evidently smaller than the wood strength
of all members in the timber frame. Thus, especially in multi-level buildings
located in seismic or windy areas, cracks in FPB usually appear. In these cases
the FPB lose their stiffness and therefore their resistance should not be
considered at all. Stresses in the timber frame under a horizontal loads are
usually not critical. There are several possibilities to reinforce panel walls in
order to avoid cracks in FPB:
- by using additional boards. The boards are usually doubled:
- symmetrically (on both sides of a timber frame),
- non-symmetrically (on one side of a timber frame),
- by reinforcing boards with steel diagonals,
- by reinforcing boards with carbon or high-strength synthetic fibres.
3.1 Strengthening with additional FPB
In [6] we presented the first possibility experimentally using additional FPB,
which gave higher elasticity of elements, whilst bearing capacity and especially
ductility were not improved in the desired range.
115
3.2 Strengthening with diagonal steel strips

With the intention to improve the resistance and especially the ductility of the
walls it is more convenient to insert diagonal steel strips, which have to be fixed
to the timber frame. In this case only a part of the horizontal force is shifted from
boards over the tensile steel diagonal to the timber frame after the appearance of
the first crack in the tensile zone of FPB (see [6]). An enlarged effective crosssection of FPB (A*1b) can approximately be computed considering the
compatibility conditions between the actual reinforced and fictitious unreinforced element. The computational procedure is described in details in
Premrov and Dobrila [7] and will not be presented here. With regard to the
fictive enlarged cross-section of FPB we proposed two approximate analytical
models using either fictitious thickness (t*) or fictitious width (b*) of fibreboards:
t* =
A1*b
1 E
1
= t + s sin 2 cos A1s ,0
b
Gb
b
(5)
b* =
A1*b
1 E
1
= b + s sin 2 cos A1s ,0
t
Gb
t
(6)
In the above equations represents the angle of inserted steel diagonals with
the net area (A1s,0). A non-dimensional coefficient is shear cross-section
coefficient defined as a proportion between the shear and actual cross-sectional
area of the FPB with the shear modulus (Gb). Alongside the steel diagonals
influence these models enable simultaneous consideration of the fasteners
flexibility between the board and the timber frame and any appearing cracks in
the tensile area of the FPB. Un-reinforced panels can be computed using actual
dimensions of the FPB. Numerical results presented in [7] on diagonally steel
reinforced elements show good agreement with measurements performed on the
test samples.
3.3 Strengthening with diagonal CFRP strips
As the tensile strength of FPB is obviously lower than the compressive strength
and corresponding capacity of timber frame, the treated elements tend to fail
because the cracks are forming in the tensile area of the FPB, therefore this
tensile area could be reinforced with high-strength materials. This strengthening
concept is such that the composites would contribute to tensile capacity when the
tensile strength of FPB is exceeded. No FRP applications on the treated fibreplaster boards were found in the literature.
Test experiment: strengthening with CFRP strips
4.1 Test configuration

Three sample groups from total of nine test samples were tested in order to carry
out appropriate experimental research on the influence of CFRP strengthened

walls. All test groups consisted of three walls of actual dimensions h=263.5cm
and b=125cm. The cross-section presented in Figure 2 was composed of timber
studs (2x9x9cm and 1x4.4x9cm), timber girders (2x8x9cm) and Knauf fibreplaster boards (Knauf [8]) of thickness t=15mm. They were fixed to the timber
frame using staples of 1.53mm at an average spacing of s=75mm.
yi
At, Et
yi
Ab , E b
t =1.5
Vz
9.0
4.4
9.0
9.0
ai = 58
b =125 cm
Figure 2:
Cross-section of test samples.
The static model according to Figure 1 was used for all groups of test
samples. The samples were actually rotated by 900 according to Figure 1 and
they were therefore subjected to vertical force acting at the end of the elements
(Figure 3a).
The FPB were reinforced in the tensile diagonal area using
SikaWrap-230C strips (Sika [9]) made from carbon high-strength fibre
reinforced polymers of thickness 1.2 mm. Strips with different widths (300 or
600 mm) and of different boundary conditions were glued to the FPB.
The first group (G1) of three test samples was additionally reinforced with two
CFRP diagonal strips (one in each FPB) of width 300 mm which were glued on
the FPB using Sikadur-330 LVP. The strips were additionally glued to the timber
frame (Figure 3a,b) to ensure the transmission of the force from FPB to the
timber frame.
The second group (G2) of three test samples was additionally reinforced with
two CFRP diagonal strips of width 600 mm. The strips were glued on FPB and
to the timber frame as in G1 (Figure 3b) to ensure the transmission of the force
from FPB to the timber frame.
The third group (G3) of three test samples was additionally reinforced with two
CFRP diagonal strips of width 300 mm as in G1 but they were not glued to the
timber frame (Figure 3c).
Material properties for the test samples for all groups were the same (Table
1). Values for timber of quality C22 are taken from EN338 [10], the
characteristics of fibre-plaster boards from Knauf [8] and for carbon strips Sika
[9] data were used.
a.)
117
b.)
c.)
Figure 3:
a.) G1; the static system, b.) G2; the CFRP strip is glued on the
FPB and additionally to the timber frame, c.) G3; the CFRP strip
is not glued to the timber frame.
Table 1:
E0,m
Timber
FPB
SikaWrap
Properties of used materials.

Gm
fm,k
ft,0,k
fc,0,k
[N/mm2]
[N/mm2]
[N/mm2]
[N/mm2]
[N/mm2]
10000
3000
231000
630
1200
/
22
4.0
/
13
2.5
4100
20
20
/
m
[kg/m3]
410
1050
1920
4.2 Test results and analysis

The force forming the first crack (Fcr) in the FPB, the crushing force (Fu), the
maximal cantilever bending deflection (w) under the acting force (F) and the slip
() in the tensile area between the FPB and the timber frame were all measured.
The measured values for the un-strengthened (UNS) test samples were taken
from [6] and included for information and comparison only.
Average force forming the first crack (Fcr):
G1: Fcr,1 = 24.28 kN
G2: Fcr,2 = 32.13 kN
G3: Fcr,3 = 35.90 kN

UNS: Fcr,uns = 17.67 kN

Average crushing force (Fu):
G1: Fu,1 = 40.33 kN
G2: Fu,2 = 46.27 kN
G3: Fu,3 = 36.26 kN

UNS: Fu,uns = 26.02 kN
It is evident that the elastic resistance (force forming the first crack) essentially
increased for all kinds of CFRP strengthened test samples, but mostly for
samples G3, where the CFRP strips were not fixed to the timber frame. The
CFRP influence was not so obvious at samples G1, where carbon strips of the
same dimensions were additionally glued to the timber frame.
On the other hand, when comparing the measured results of the crushing
force, a greater improvement can be noticed in the groups where the CFRP
diagonals were glued to the timber frame. Compared to the un-strengthened test
sample, the crushing force in samples G2 was increased by 78%. In samples G3
the crushing force practically coincided with a force forming the first crack, so
cracks hardly appeared at all, which is not a good solution to ensure better
ductility, necessary for seismic design.
For further analysis it is important to present measured maximal cantilever
deflections (w) (Figure 4) and slips () in the connecting area (Figure 5).
F[kN]
Fcr,3
Fcr,2
UNS
G1
G2
G3
Fcr,1
w[mm]
Figure 4:
Measured average bending deflections (w).
Closer look at the graph in Figure 4 at F>17 kN reveals an obvious difference

in the behaviour of the test samples when the CFRP strips were glued to the
timber frame (samples G1 and G2) or if they were not (samples G3). Beside the
fact that samples G1 and especially G2 demonstrated higher load-carrying
capacity than samples G3, it is also important to mention that samples G1 and
G2 produced substantially smaller slip than samples G3, which never exceeded
1mm at the first crack forming (Figure 5). Therefore it can be assumed that the
yield point of the fasteners was not achieved before cracks appeared at all.
Consequently, the walls tend to fail because of the crack forming in FPB. In this
case of strengthening the ductility of the whole wall element (see Figure 5 for
samples G1 and G2) practically coincides with the ductility of FPB, as
proposed with d1 and d2 coefficients.
119
F[kN]
Fcr,3
Fcr,2
G1
G2
G3
Fcr,1
[mm]
Figure 5:
Measured average slips () in the connecting area.
In contrast, in G3 model, where the CFRP strips were unconnected to the

timber frame, the slip () between the FPB and the timber frame was evidently
higher than in samples G1 and G2, and exceeded 3mm when the first crack in
FPB appeared (Figure 5). The load-displacement relation (F-) of the fasteners
was in this case at the force which produced first cracks almost completely
plastic. Since the tensile strength of FPB is essentially improved, the walls tend
to fail because of fastener yielding and therefore the Lower bound plastic
method (EC5 Simplified Method A) can be used to determine the wall`s load
carrying capacity, (eqn 1).
Conclusions
As shown, there is practically no influence on the element stiffness of any

reinforcement before cracks appeared in the un-strengthened FPB. However,
after the first cracks in un-strengthened FPB appeared, the test samples
demonstrated an important difference in behaviour dependant on the boundary
conditions between the inserted CFRP strips and the timber frame.
If strips are glued to the timber frame the fasteners produced substantially
smaller slip, which never exceeded 1mm when the first cracks appeared.
Therefore it can be assumed that the yield point of the fasteners is not achieved
before cracks appeared at all and the elements tend to fail because cracks appear
in the tensile area of FPB. Therefore, it is not recommended to use Eurocode 5
[2] simplified methods to predict the element resistance. Simple mathematical
models with a fictive enlarged cross-section of FPB are proposed in [7].
In the case where the CFRP diagonals are unconnected to the timber frame,
the slip between the FPB and the timber frame is evidently higher and the loaddisplacement relation of the fasteners is, after the cracks appeared, almost
perfectly plastic. Since the tensile strength of FPB is with CFRP highly
improved, the walls tend to fail because of fastener yielding, similar as at woodbased sheathing boards. Therefore the Lower bound plastic method can be
used to determine the walls load carrying capacity.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
Faherty, K.F. & Williamson, Wood Engineering and Construction

Handbook, McGraw-Hill Publishing Company: 1989.
CEN/TC 250/SC5 N173, Eurocode 5: Design of Timber Structures, Part
1-1 General rules and rules for buildings, Final draft prEN 1995-1-1,
Brussels, 2003.
Kllsner, B., Panels as wind-bracing elements in timber-framed walls.
Swedish Institute for Wood Technology Research, Report 56, Stockholm,
1984.
kerlund, S., Simple calculation model for sheets on a timber frame, Bygg
& Teknik, No.1, 1984.
Kllsner, B. & Lam, F., Diaphragms and shear walls. Holzbauwerke:
Grundlagen, Entwicklungen, Ergnzungen nach Eurocode 5, Step 3,
Fachverlag Holz: Dsseldorf, pp. 15/1-17, 1995.
Dobrila, P. & Premrov, M., Reinforcing methods for composite timber
frame-fiberboard wall panels. Engineering Structures 25(11), pp. 13691376, 2003.
Premrov, M. & Dobrila, P., Approximate analytical solutions for diagonal
reinforced timber-framed walls with fibre-plaster coating material.
Construction and Building Materials, 18 (10), pp. 727-735.
Knauf Gipsfaserplatten Vidivall/Vidifloor, 2002.
Sika, Sicher bauen mit System. Technische Merkbltter. Ausgabe 5, 2003.
European Committee for Standardization, EN 338:2003 E: Structural
timber Strength classes, Brussels, 2003.
121
Evaluation of the structural integrity of

a sandwich composite train roof structure
K. B. Shin1, B. J. Ryu1, J. Y. Lee1 & S. J. Lee2
1
Department of Mechanical Design Engineering,

HANBAT National University, South Korea
2
Division of Rolling Stock, Hankuk Fiber Glass Co. Ltd., South Korea
Abstract
We have evaluated the structural integrity of a sandwich composite train roof
which can find a lightweight, cost saving solution to large structural components
for rail vehicles in design stages. The sandwich composite train roof was 11.45
meters long and 1.76 meters wide. The FE analysis was used to calculate the
stresses, deflections and natural frequencies of the sandwich composite train roof
against the weight of air-condition system. The 3D sandwich FE model was
introduced to consider the hollow aluminum frames jointed to both sides of the
sandwich train roof. The results shown that the structural performances of a
sandwich composite train roof under the loading conditions specified were
proven and the use of aluminum reinforced frame was beneficial with regard to
weight saving in comparison to steel reinforced frame. Also, we have
manufactured the prototype of sandwich composite train roof on the basis of
analysis results.
Keywords: aluminum honeycomb structure, train roof structure.
Introduction
The use of composite materials on land transportation is steadily increasing.

Railroad cars, mass transit vehicles and a wide range of ground transportation
systems offer expanding opportunities for composite materials. Driving forces
for the use of composites in ground transportation applications include low
manufacturing investment cost, cost reduction through parts consolidation,
weight saving, good mechanical properties, excellent durability and dimensional
stability [1].
doi:10.2495/HPSM06013

The reduction of structural weight of one large component usually triggers
positive synergy effects for other parts of the vehicle. For example, a reduction
of the mass of a railway carbody could lead to weight savings in the traction
system, suspension, brakes and other subsystems. The manufacturing
technologies for steel or aluminum carbodies of railway rolling stocks are
currently highly developed. Additionally, cost and weight savings and the
increase of availability of the rolling stock are difficult to achieve using the
conventional metal materials. Therefore, the use of sandwich composite
materials for the railway carriage structure has been proposed and recommended
in transportation applications because they offer not only high specific stiffness
and strength but also low manufacturing investment cost as compared to other
conventional materials [2]. Recently, composites have been applied for the
lightweight design of the train roof structure, side structure and carbody shell
structure. A prototype composite train roof was developed to find a lightweight,
cost saving solution to large structural components for rail vehicles.
Additionally, a composite roof structure improves on torsion stiffness of a
railway carbody [3]. Other advantages included: (1) reduced production costs,
(2) reduced part count, (3) elimination of an additional thermal insulation, (4)
improved passenger comfort. Turbostar in the UK, PUMA Train Express in
Germany and RARe-520 in Switzerland are good examples. These examples
applied advanced composite material to train roof structure.
The objectives of this paper are as following: (1) Evaluation of the structural
integrity of developing sandwich composite train roof structure, (2)
Manufacturing the prototype of sandwich composite train roof on the basis of
analysis results.
Design of the lightweight composite train roof structure
2.1 Structure design

The sandwich composite train roof structure was 11.45 meters long and 1.76
meters wide. The composite train roof structure was manufactured using a
sandwich composite, an inner stiffener, an edge stiffener and a hollow aluminum
extrusion as shown in figure 1. The sandwich structure consists of aluminum
skin and aluminum honeycomb core. The inner stiffener was used to increase the
bending stiffness of transverse direction and longitudinal direction. The edge
stiffener improved the twisting stiffness of the train roof structure. Figure 2
shows the section of the lightweight composite train roof. The air-condition unit
will be located on the roof structure. The hollow aluminium extrusion frame will
be jointed to the side wall structure of carbody as shown in figure 2.
2.2 Construction materials
The sandwich structure was composed of an aluminum skin and an aluminum
honeycomb core. The same material should be used for the skin and carbody in
order to minimise the stress variation at joints and prevent from galvanic
123
corrosion with aluminum honeycomb core. Table 1 shows the mechanical

properties of material used in the structural analysis.
The metal stiffener increased the bending stiffness of the sandwich roof
structure. The inner stiffener will be manufactured by aluminum or steel
material after evaluation of the weight, stress, and deflection of the sandwich
composite train roof using structural analysis.
Figure 1:
Figure 2:
3D CAD model of sandwich composite train roof structure.
The drawing of frame section of sandwich train roof.
Evaluation of the structural integrity
3.1 3D Sandwich FE model

In order to verify the structural integrity of sandwich train roof, ANSYS V9.0
was used at modal and structural analysis. An 8-node sandwich shell element
(shell 91) is recommended to analyze the sandwich composite structure in

ANSYS. However, the structural behaviours of the hollow aluminum extrusion
frame that is bonded at both sides of the sandwich train roof structure are
important in this project. Therefore, the sandwich train roof structure should be
modelled using the proper 3D FE modelling technique to consider the joint
between the sandwich roof structure and hollow aluminium extrusion frames.
Table 1:
Type
The mechanical properties of materials used in the structural

analysis.
Materials
Skin
Al. 5052
Core
Aluminum
honeycomb
core
Dimension
t=1.2mm
Cell size = 3/8 inch

Cell wall thickness = 70m
Cell depth = 32mm
Properties
E=69GPa
=0.33
=2700(kg/m3)
E11=8.27MPa
E22=1.31MPa
E33=1276MPa
G12=0.0001GPa
G23=117GPa
G13=296GPa
12=0.75
23=0.0001
13=0.0001
=100(kg/m3)
E : Youngs modulus, G : Shear modulus, : Poissons ratio, t : Thickness, : Density
Table 2:
Cases
1
2
3
4
5
The analysis cases to select the proper 3D sandwich FE model.

Elements
Sandwich shell 91(8-node) + sandwich option (Reference)
Layered solid 46 (8-node)
Layered solid 191 (20-node)
Shell 63(skin)/Solid 45(core)
Shell 93(skin)/Solid 186(core)
The dimension and material properties of sandwich composite are shown in

table 1. Five cases were used to select the proper 3D FE models of sandwich
train roof as shown in table 2 and they were conducted using modal analysis
which can verify its mass and stiffness. The block Lanczos method was used in
modal analysis and it was compared and checked by natural frequency. Table 3
shows the shell element for the skin part and the solid element for the core part
should be used to replace the sandwich shell element. 3D layered solid element
of case 2 and case 3 could not substitute the sandwich shell element because it do
not simulate the bending and shear behaviour of sandwich structure.
Accordingly, when the sandwich composite roof structure is jointed with
hollow aluminium extrusion frame, 3D FE sandwich modelling is necessary to
evaluate the structural integrity of the sandwich train roof.
Table 3:
125
The results of natural frequencies for the selected elements at

table 2.
Mode
1
2
3
4
5
6
7
8
9
10
Case 1*
1.87(B)
5.16(B)
8.65(T)
10.14(T)
16.8(B)
17.48(T)
25.13(B)
26.68(T)
35.14(B)
36.42(T)
Case 2
1.64(B)
2.22(T)
3.22(B)
4.52(T)
5.33(B)
6.98(T)
7.96(B)
9.64(T)
11.13(B)
12.58(T)
Case 3
1.25(B)
3.45(B)
5.68(T)
6.78(B)
11.25(B)
11.51(T)
16.87(B)
17.61(T)
23.64(B)
24.11(T)
Case 4
1.86(B)
5.12(B)
8.46(T)
10.05(B)
16.62(B)
17.10(T)
24.82(B)
26.11(T)
34.61(B)
35.66(T)
Case 5
1.86(B)
5.12(B)
8.44(T)
10.04(B)
16.60(B)
17.06(T)
24.77(B)
26.05(T)
34.5(B)
35.0(T)
B = Bending mode, T = Twisting mode, * = Reference
3.2 Analysis cases and FE model

Structural integrity of sandwich composite roof structure has been evaluated by
modal analysis and structural analysis as shown in table 4. We have considered
the aluminium and steel stiffener to examine the structural behaviour of
sandwich train roof caused by the difference in stiffener material. The constraints
of joint parts between the sandwich train roof and the hollow aluminium
extrusion frame have been considered for two cases in analysis: (1) contact
condition, (2) coupled condition. The contact condition is nonlinear analysis and
can simulate the real jointing condition using bonding option in ANSYS. The
coupled condition is linear analysis and can save the calculating time. Figure 3
shows the finite element models for the sandwich composite roof structure. The
stiffener and hollow aluminum extrusion were modelled using shell element. The
sandwich composite was modelled using the shell element (skin) and solid
element (core).
Table 4:
The analysis cases of sandwich train roof considered in design

stages.
Modal Analysis
Structural Analysis
Cases
1-1
1-2
1-3
1-4
1-5
1-6
1-7
1-8
Element
Shell63
/Solid45
Shell63
/Solid45
Jointing condition
Bonded contact
Coupled
set
Bonded contact
Coupled
set
Stiffener materials
Aluminum
Steel
Aluminum
Steel
Aluminum
Steel
Aluminum
Steel
Figure 3:
Finite element model for sandwich composite train roof.
Figure 4:
Table 5:
Mode(Hz)
1
2
3
4
5
6
7
8
9
10
Weight(kg)
The boundary condition for modal analysis.
Summary of modal analysis using simply supported BC.

Case 1-1
64.76
65.46
66.64
69.52
74.15
79.30
87.66
95.90
106.19
117.67
393.58
Case 1-2
50.62
51.71
52.70
55.25
59.36
63.95
71.88
79.81
89.59
101.41
596.26
Case 1-3
64.33
65.07
66.25
69.12
73.76
78.93
87.29
95.55
105.85
117.34
393.58
Case 1-4
50.24
51.38
52.38
54.93
59.04
63.65
71.56
79.52
89.30
101.13
596.26
127
3.3 Modal analysis

Modal analysis of the sandwich composite roof structure was conducted to
evaluate its bending mode and four different cases were considered as shown in
table 4. The cant side of the hollow aluminum extrusion was chosen as a
boundary condition as shown in figure 4. Table 5 shows the results of natural
frequencies. Bending mode only was occurred for these cases. It shows that
natural frequencies of a simply supported sandwich train roof could be changed
by the material of the stiffener. The changes in natural frequencies due to the
material used for the stiffener can affect the whole carbody structure. If the
natural frequencies of the whole carbody structure do not meet the design
requirements they can be altered by changing the material of the stiffener.
3.4 Structural analysis
The objective of this chapter is to investigate the structural integrity of the
sandwich composite train roof structure with the weight of air-conditioning unit
(ACU). There are four cases as given in table 4.
Figure 5:
The loading and boundary conditions for structural analysis.
Table 6: Summary of structural analysis using bonded contact condition.
Max. deflection(mm)
Max. principal
stress(MPa)
Total weight(kg)
Case 2-1
Case 2-2
(Bonded contact/Aluminum stiffener)
(Bonded contact/Steel stiffener)
0.42
0.39
8.69
8.93
393.58
596.26

3.4.1 Contact condition (case 2-1 and 2-2)
A contact element was used at the joint between the sandwich composite
structure and the hollow aluminum extrusion frame. An aluminum and a steel
stiffener was considered in each case. Figure 5 shows the boundary and loading
condition for the structural analysis. The results of structural analysis using the
contact condition are given in table 6. The deflection and stress have similar
results using either aluminum or steel stiffeners. The reason is that the bending
stiffness of aluminium stiffener is lower than that of steel, but the weight of the
aluminum stiffener is comparatively lighter than that of the steel stiffener.
By using the aluminum stiffener, we can obtain a weight saving of 34% in
comparison with the steel stiffener. Thus in consideration of weight reduction,
the use of the aluminum stiffener can be more effective as both stiffeners are safe
with regards to stiffness and strength.
3.4.2 Couple condition (case 2-3 and 2-4)
The coupled condition was used at the joint, between the composite structure and
the hollow aluminum extrusion frame. An aluminium and a steel stiffener was
considered in each case. Table 7 shows the results of the structural analysis for
the coupled condition. The results of deflection and stress for the couple
condition are rather high in comparison with that of the contact condition. The
reason is that the analysis of the coupled condition is linear. As with the contact
condition, however, there are few differences in the results of deflection and
stress depending on whether an aluminum stiffener or steel stiffener is used.
When considering the effect of weight reduction, it is preferable to use aluminum
stiffener.
Table 7: Summary of structural analysis using coupled condition.
Max. Deflection(mm)
Max. principal
stress(MPa)
Total weight(kg)
Case 2-1
Case 2-2
(Coupled set/Aluminum stiffener)
(Coupled set/Steel stiffener)
0.73
0.67
15.89
15.83
393.58
596.26
3.4.3 Comparison of analysis results of contact and couple conditions

The values of the deflection and stress of the sandwich train roof using the
contact condition at the joint tend to be lower than that of the sandwich train roof
using the coupled condition. We found that the gradient of displacement
demonstrated a different tendency to each joint marked at figure 6 and 7.
When the contact condition was applied in structural analysis, it could
simulate the real joining condition for joints between the sandwich composite
train roof and the hollow aluminum extrusion frames. However, the coupled
condition only shared the nodal degree of freedom (DOF) at the joints, and it
could not simulate the real joining condition. So, the contact condition is
recommended to analyse the structural behaviour of the real joining parts
although it comparatively takes a long time to calculate the results.
(a) Contact condition
129
(b) Coupled condition
Figure 6:
Contours of deflections at contact and coupled conditions.
Figure 7:
Comparison of deflections at contact and coupled conditions.
Figure 8:
The prototype sandwich train roof.

3.4.4 Manufacturing of prototype sandwich composite train roof
We found that the sandwich composite train roof with the aluminum skin,
aluminum honeycomb core and the aluminum stiffener was best choice on the
basis of the structural analysis. Figure 8 shows the prototype sandwich
composite train roof, which was manufactured using the autoclave moulding
process.
Conclusions
In this paper, we have obtained the following conclusions.

(1) The sandwich composite train roof structure with an aluminum stiffener is
beneficial with regard to weight saving and structural integrity.
(2) We observed that the quality of the reinforcing material could be altered to
achieve the design parameters for natural frequencies of the whole carbody
structure.
(3) As a result of structural analysis of the sandwich composite train roof
structure, there was no striking difference in deflection and stress by changing
the quality of the reinforcing material. We found that the advantage of the light
weight of the aluminum and the bending stiffness of steel counterbalance each
other. Therefore, if we select weight as the priority order, aluminum reinforcing
material would be recommended.
(4) As a result of analysis of joints classified by contact and coupled
conditions, there is no striking difference in natural frequency. However,
structural analysis demonstrated a difference. This was caused by the linear
assumption of the contact condition in natural frequency analysis. On the other
hand, using structure analysis, conditions were calculated by using a nonlinear
term at the contact condition and the linear term at the coupled condition. So,
structure analysis demonstrated a difference between deflection and stress. We
recommended the use of the contact condition for more detailed structural
analysis.
References
[1] K B. Shin, C G. Kim, C S. Hong., Correlation of accelerated aging test to
natural aging test on graphite-epoxy composite materials, Journal of
Reinforced Plastics and Composites 22(2003) 849-866.
[2] K B. Shin, S H. Hahn., Evaluation of the structural integrity of hybrid
railway carriage structures including the aging effects of composite
materials, Composite Structure 68 (2005) 129-137.
[3] J. Cantrill, R. Mableson. Development and prototyping of a lightweight
composite train roof, 14th COMPOSITE Workshop (2003).
131
Measurement of the fiber stress distribution

during pull-out test by means of
micro-Raman spectroscopy and FEM analysis
K. Tanaka1, K. Minoshima2 & H. Yamada3
1
Department of Mechanical Engineering, Doshisha University, Japan

Department of Mechanical Engineering, Osaka University, Japan
3
Department of Mechanical Engineering, Kyoto University, Japan
2
Abstract
A single-fiber pull-out model composite for an aramid/epoxy system was
specially designed to measure the stress distribution of the aramid fiber
embedded in the matrix using micro-Raman spectroscopy. The stress transfer
length of the fiber obtained was about 400-500m, which was equal to the result
of FEM analysis. Just after the initiation and propagation of the fiber/matrix
interfacial debonding, the fiber was broken, and the fiber in the matrix had the
axial tensile residual stress. The tensile residual fiber axial stress showed the
maximum at around the tip of the interfacial debonding. The stress was reduced
and became almost equal to zero after being immersed in deionized water at
80 C for 44h. This behavior agreed with the result of FEM analysis, in which
the friction coefficient was introduced in the fiber/matrix interface.
Keywords: interfacial properties, micro-Raman spectroscopy, pull-out test,
aramid fiber, friction.
Introduction
Investigation of the fracture strength and fracture mechanism of the fiber/matrix

interface is extremely important, because the mechanical properties of the fiber
reinforced composites depend strongly not only on the properties of the fibers
and the matrix but also on the fiber/matrix interfacial ones. We carried out the
single fiber pull-out tests to evaluate the influence of water absorption on the
interfacial properties of aramid/epoxy composite (Tanaka et al. [1]). In this
doi:10.2495/HPSM06014

study, however, the pull-out load, at which the unstable crack propagated through
the total embedded fiber length, was used for the measure of the interfacial
strength and the stress distribution along the fiber in the matrix was not clarified.
Moreover for the study of interfacial crack propagation under fatigue loading, it
is extremely important to understand the stress distribution along the fiber close
to the crack tip (Minoshima et al. [2]).
Raman spectroscopy is a new technique to directly measure the strain or
stress distribution along the fiber embedded in a matrix (e.g. Galiotis et al. [3],
Patrikis et al. [4], Cervenka et al. [5]). In this work, a single-fiber pull-out model
composite for an aramid/epoxy system was specially designed and Raman
spectroscopy was used to clarify the difference of stress distribution along the
fiber before and after the interfacial debonding.
Experimental procedure and FEM analysis
2.1 Specimen preparation and experimental procedure

The technique to measure the stress of the fiber using laser Raman spectroscopy
is based on the fact that the Raman frequencies are strain (stress) dependent.
Therefore, using the calibration curves of the peak wavenumber of Raman
spectrum vs. stress of a fiber, the measured Raman peak wavenumber can be
converted to an axial stress. In this investigation, the Raman spectra were
obtained by means of a laser Raman micro spectrometer (Japan Spectroscopic
Co. Ltd, NRS-2000). The 514.5nm line of an argon-ion laser was used and the
laser beam was focused to a 1.5m spot on the fiber by an optical microscope.
The aramid fiber used for single fiber tensile tests was Kevlar 49
manufactured by Du Pont, USA. The fiber had an average diameter of 12 m.
Specimens were prepared by following the recommended testing procedure as
described in ASTM D3379/JIS R7601. Polyester thin film (Thickness: 100m)
was used for a tab and a single fiber was glued to it, giving a gauge length of 12
mm, as illustrated in fig.1. Quasistatic tensile tests were carried out using a
tensile testing machine with a load cell of 1 N Capacity, which can be installed in
the laser Raman micro spectrometer. After chucking a specimen, the tab was cut
and the load was applied to the fiber.
Fiber
2-6
12
Cut off
12
Figure 1:
Single fiber tensile test specimen. All dimensions are in mm.
133
The specimen construction for a single fiber pull-out specimen is shown in

fig.2. An epoxy resin, Epikote 828 and curing agent Epicure Z (Japan Epoxy
Resins Co. Ltd.) were used in 5:1 weight ratio. The resin was put on the center of
the tab with a fiber, as illustrated in fig.2(a). Specimens were placed in an oven
for two hours at 80C and then for two hours at 150C. The epoxy resin was
remaining on the tab after curing and the fiber was embedded in the matrix
between A and B (fig.2(b)). Pull-out tests were performed using a tensile testing
machine in the laser Raman micro spectrometer. Raman spectrum was measured
along the fiber for both the embedded fiber and the free fiber at several stress
levels, as shown in fig.3. In this study the stress of the fiber adopted was the
nominal stress, which was calculated from the value of the load cell.
(a) Specimen construction.
(b) Shape and dimensions of pull-out specimen. All dimensions are in mm.
Figure 2:
Single fiber pull-out specimen.
2.2 FEM analysis

FEM analysis was conducted using the commercial code ANSYS to estimate the
stress distribution during the pull-out test and after fiber/matrix interfacial
debonding. Specimen was simplified to the model shown in fig.4(a) and analysis

was conducted using 2D model without and with the interfacial debonding of
200 m in length (fig.4(b)). For FEM analysis, thermal expansion during
specimen preparation, friction at the fiber/matrix interface and swelling
expansion of resin was taken into account. Mechanical properties used in this
analysis are shown in table 1 (Hull and Clyne [6], [7], Kawabata et al. [8]).
Friction coefficient () and maximum friction stress (max) at the interface were
calculated from the result of the previous study [1]. =0.6 and max =11.9 MPa
were obtained and used in the analysis. Swelling of the resin was set at 0.6% [1]
and the swelling of the fiber was neglected.
Figure 3:
Schematic drawing of the pull-out test procedure (C-C cross

section of fig.2).
(a) Model of the specimen.

Figure 4:
(b) 2D model with boundary conditions.
FEM model with interfacial debonding.
Table 1:
135
Mechanical properties of the fiber and resin [6-8]: (a) Epoxy resin
(Epikote828 + Epicure Z), (b) Fiber (Kevlar 49).
(a)
Youngs modulus
2.48 GPa
Poissons ratio
0.4
6010-6 C-1
Coefficient of thermal expansion
(b)
Youngs modulus
Poissons ratio
Coefficient of thermal expansion
EL
129.6 GPa
ET
2.49 GPa
LT
0.62
TT
0.31
-210-6 C-1
5910-6 C-1
L: longitudinal direction, T: Radius direction.
Shift
Intensity a.u.
10
Raman peak of neon lamp
Figure 5:
0 GPa
2 GPa
1600
1400
1200 1 1000
Raman shift cm
Examples of the Raman spectrum of Kevlar 49.
Results and discussion
Raman spectra of a Kevlar 49 single fiber without loading and with loading of 2
GPa tensile stress are shown in fig.5. The increase in tensile stress resulted in a
clear shift of the spectrum to a lower wavenumber. In this study, a strong band

around 1615cm-1, which corresponds mainly to the phenyl ring/C-C stretching
(Parthenios et al. [9]), was used to evaluate the stress.
The influence of the applied load on the Raman peak position for a Kevlar 49
single fiber is shown in fig.6. Four specimens were tested and each result was
shown by the same symbol. A linear decrease in the Raman peak position with
stress was found, irrespective of the specimens. Therefore, the liner regression
line for the measured results shown in the figure was used to estimate the tensile
stress of the embedded fiber. The maximum difference between the liner
regression line and the tensile stress of each specimen was 0.1 GPa. Therefore
our system achieves the stress measurement accuracy of 0.1 GPa.
1616
Peak wavenumber cm
1614
1612
1610
1608
1606
1604
1602
Figure 6:
1
2
Tensile stress GPa
Relationship between tensile stress of Kevlar 49 single fiber and

peak wavenumber.
The stress distribution of the axial tensile stress of the fiber during the pullout test was shown in fig.7 with the result of FEM analysis. The applied nominal
stress to the fiber is also plotted in the figure. The origin point of the distance
along the fiber is the meniscus point from which the fiber was embedded in the
epoxy matrix. The fiber at the positive value of the distance along fiber was
embedded in the resin matrix, whereas the fiber at the negative value
corresponded to the unembedded free fiber. The stress distribution was measured
when the applied stress was 0, 0.47, 0.94, 1.41, 1.89 and 2.36 GPa. The stress of
the embedded fiber decreased gradually along the fiber from the meniscus point,
which means that the stress was transferred from the fiber to the matrix.
Considering from the fact that the fiber axial stress at the distance of 400 m and
that of 500 m were almost equal, the stress transfer length was about 400-500
m for the Aramid/epoxy system adopted in our study. This behavior is agreed
with the result of FEM analysis.
Before debonding
2.36GPa
1.89GPa
1.41GPa
0.94GPa
0.47GPa
Applied stress
0GPa
Fiber axial stress GPa
Mesured
with Raman FEM
spectroscope
Embedded fiber
Figure 7:
137
200
0
200
400
Distance along fiber m
Fiber axial stress along the fiber before debonding with the results
of FEM analysis.
When the applied stress was increased to 2.65 GPa, the fiber/matrix
interfacial debonding was initiated and propagated, and at the same time the fiber
was fractured. After this interfacial debonding, the stress distribution in the resin
was again measured. The stress distribution after the fiber/matrix interfacial
debonding differed from that before the debonding, which is shown in fig.8.
Although external stress was not applied to the fiber because of fiber fracture, the
fiber axial stress in the debonded region had the residual tensile stress and had
the maximum at the distance of around 200 m. The position of the interfacial
crack tip was 180 m, which was confirmed after Raman measurement by using
optical microscope. This crack tip was located almost at the position of
maximum residual stress. This residual stress was considered to be caused by the
holding of the previous stress by the fiber/matrix interfacial friction acting on the
debonded interface. This speculation was supported by the fact that the friction
stress plays an important role in the propagation of the fiber/matrix interfacial
debonding and the propagation rate of interfacial debonding slowed down and
retarded when the a constant fatigue load was applied to the specimen [2].
After debonding, this specimen was kept in dry air (relative humidity: 25%)
for 144 h and after that it was immersed in deionized water at 80 C for 44 h.

Raman spectrum were measured during and after these conditioning. Figure 8
shows the influence of conditioning in dry air and successive hot water on the
fiber axial stress along the fiber after debonding. Analytical results are also
plotted in the figure. The conditioning in dry air did not affect so much on the
stress distribution, but after immersed in hot water the residual stress decreased
to almost 0 GPa. The absolute values obtained by the FEM analysis were not
equal to the experimental results, but the behavior that the residual stress showed
the maximum stress at around the crack tip of interfacial debonding and it
decreased by conditioning in hot water agreed with the result of experiment.
Experimental results
After debonding (without applied stress)
Just after debonding
44 h later
48 h later
Dry air (RH: 25 %)
96 h later
144 h later
In deionized water at 80 for 44 h
Analytical results
Just after debonding
Applied stress
In deionized water at 80 for 44h
Fiber axial stress GPa
Figure 8:
Embedded fiber
200
0
200
400
Distance along fiber m
Influence of conditioning in dry air and successive hot water on the

fiber axial stress along the fiber after debonding.
Conclusions
Stress distribution along the fiber of an Aramid fiber/epoxy model composite

during the pull-out test was measured using micro-Raman spectroscopy. The
stress transfer length of the fiber was about 400-500m. Although the external
stress was not applied to the fiber, after the fiber/matrix interfacial debonding,
the fiber in the debonded region had the residual stress caused by the holding of
139
the previous stress by the fiber/matrix interfacial friction on the debonded

interface. This residual stress showed the maximum at the crack tip of the
interfacial debonding and the residual stress decreased to almost 0 GPa after
immersed in hot water.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
Tanaka, K., Minoshima, K. Grela W. & Komai, K., Characterization of

the Aramid/Epoxy Interfacial Properties by means of Pull-out Test and
Influence of Water Absorption, Composites Science and Technology,
62(16), pp.2167-2174, 2002.
Minoshima, K., Tanaka, K., Araki, Y. & Komai, K., Characterization of
the propagation of aramid/epoxy interfacial debonding under fatigue
loading, JSME Mechanical Engineering Congress, Japan, No.03-1,
pp.363-364, 2003.
Galiotis, C. Young, R J. Yeung, P H J. & Batchelder, D N., The Study of
Model Polydiacetylene/epoxy Composites Part I The Axial Strain in the
Fiber, Journal of Materials Science, 19, pp 3640-3648, 1984.
Patrikis, A K. Andrews, M C. & Young, R J., Analysis of the Single-Fiber
Pull-out Test by means of Raman Spectroscopy: Part I. Pull-out of Aramid
Fibers from an Epoxy Resin, Composites Science and Technology, 52,
pp 387-3965, 1995.
Cervenka, A J. Bannister, D J. & Young, R J., Moisture absorption and
interfacial failure in aramid/epoxy composite, Composites Part A, 29A,
pp 1137-1144, 1998.
Hull D. & Clyne T.W., An Introduction to Composite Materials,
Cambridge University Press, 1990.
Technical paper of curing agent for Epikote resin, Japan Epoxy Resins
Co. Ltd.
Kawabata, S., Sera, M., Kotani, T., Katsuma, K., Niwa, M. & Xiaoxin, C.,
Anisotropic Mechanical Properties of Advanced High Performance Fibers
Obtained by Single Fiber Testing System, Proceedings of the Ninth
International Conference on Composite Materials (ICCM/9), Madrid, 6,
pp.671-677, 1993.
Parthenios, J., Katerelos, D. G., Psarras, G. C. & Galiotis, C., Arramid
Fibres; a Multifunctional Sensor for Monitoring Stress/Strain Fields and
Damage Development in Composite Materials, Engineering Fracture
Mechanics, 69, pp.1067-1087, 2002.
Section 3
Natural fibre composites
(Special session organised by
T. Katayama and H. Takagi)
143
Effect of surface treatment to tensile static

and creep properties for
jute fiber reinforced composite
K. Takemura
Department of Mechanical Engineering, Kanagawa University, Japan
Abstract
Nowadays, there is a serious environmental problem with waste disposal etc. In
fiber reinforced composite, glass fiber reinforced composite is difficult to reuse
and recycle. Instead of glass fibers, natural fibers such as bamboo kenaf, jute,
and hemp fibers are focused on for environmental friendliness. The mechanical
properties for some natural fiber reinforced composites have been studied. The
number of papers for green composite is greatly increased.
In this study, jute fibers are used as reinforcement because of the huge
production and low cost. Polypropylene is used as a matrix because of recycling
properties and cost. The strength and creep properties for Jute Fiber Reinforced
Plastics (JFRP) are examined.
In Japan, natural fibers have been used as fishing nets for a long time.
Astringency of a persimmon is used for hemp fiber as a reinforcement and
binder. So, the surface treatment is conducted for the fiber using the astringency
of a persimmon. The mechanical and creep properties for jute fiber reinforced
composite and the effect of surface treatment are investigated.
From the results, the following conclusions are obtained.
The treatment using the solution of astringency of persimmon is effective to
increase mechanical properties. In the case of using the treatment, some densities
are effective and 25% solution is most effective. For the creep properties, there is
an effect of surface treatment to creep strain. The effect is large when the load is
small and the effect emerges in the initial stage.
Keywords: jute fiber, polypropylene, surface treatment, strength, creep.
doi:10.2495/HPSM06015
Introduction
Fiber Reinforced Plastics (FRP) is widely used for airplane, automobile parts,
fishing pole and any other engineering products. But, recently there are some
environmental problems, especially the disposal of glass fibers could not be
burned out. So, natural fibers such as jute, bamboo, hemp and flax are focused as
reinforcement [1-5] In Japan, third international workshop for green composite
was held in 2005. Many researchers are interested in this issue.
Jute is focused from the viewpoint of cost. And it is easy to take jute from
general market. So Jute Fiber Reinforced Plastics (JFRP) is used in this study.
The static tensile tests for this composite are conducted in order to examine
strength and stiffness etc. Natural fiber has been used for fishing net for a long
time especially in Japan. Many chemical treatment methods have been used to
the natural fibers. Among them, the astringency of a persimmon has been used
for the improvement the face of natural fibers. So, the surface treatment method
is conducted for this natural fiber using an astringency of a persimmon. So,
mechanical and creep properties for the natural fiber reinforced composite and
the effect for astringency of a persimmon are investigated.
Specimen and experimental procedure
2.1 Specimen
Plain woven jute fiber is used as reinforcement. The directions of jute fibers are
parallel and perpendicular to a load direction. Polypropylene resin (Shin Kobe
Electric Inc. PP-N-AN) is used as matrix. The geometry of specimens is
referenced to JIS (Japanese Industrial Standard) 7054. The length of the
specimen is about 200mm. The breadth and thickness are about 15mm and
2.5mm respectively. For surface treatment specimen, the plain woven fiber is
laid in the solution of an astringency of a persimmon.
2.2 Static tensile test for jute fiber
Shimadzu autograph tensile test machine (AG-IS) is used for static tensile test.
The test condition is referenced to JIS7054. The surfaces of the composite are
observed using Scanning Electron Microscope (SEM) to confirm the effect of
surface treatment.
2.3 Creep test for jute fiber
Creep Tester 100LER (Toyo Seiki. Co.) is used for this creep test. The test is
continued by a failure of specimen. When a specimen does not fail, the creep test
stops at 100hours. Three tests can be conducted simultaneously. Creep extension
and strain are measured. This test is referenced to JIS7087. The creep load is
used for the consideration of static strength.
145
3.1 Static test

Static load-strain curve for jute fiber is shown in Fig.1. The 50% solution is used
for surface treatment. In this figure, the stiffness of treated specimen is lower
than that of untreated one. But the maximum stress increases for the treatment. In
the case of that using 100% concentration, the stiffness is not improved. So it is
found that 50% diluted solution is effective for the strength. Because the fiber is
not tensioned at the initial stage, the initial stiffness seems lower.
250
Untreated
Load(N)
200
Treated(50%)
150
100
50
0
0
Strain(%)
Figure 1:
Static tensile results for jute fiber by surface treatment.
Figure 2 shows the stress-strain curves for jute fiber composite. The solution
densities for surface treatment are 0, 25, 50, 75, 100%. For the viewpoint of
stiffness, 25% solution is most effective. For untreated specimen, the stiffness is
lower than that of 25% specimen. But the strength is not low. For 100% solution,
the strength and stiffness is worst. So it is understood that the treatment using an
astringency of a persimmon is effective for the composite, but the densities are
important. For 100% solution, it seems that the fiber has some damage on its
surface. The knee point stress is important for its use as application. The knee
point stresses for 25% and 50% are almost same. The knee point stresses for 0%
and 75% are the same. The knee point stress for 100% solution is obviously
lowest. From this figure, 50% solution may seem to be as effective as 25%
solution.
The photograph of jute fiber is shown in Fig.3. From the surface of treatment,
there is much resin on the surface. It is understood that the adhesive strength on
the surface is improved. And, the fiber is not straight. It seems that the fiber has
ductile properties. So, the stress-strain curve has knee point. And high stiffness is
obtained from the surface.
Figure 2:
Figure 3:
Static Stress-Strain curves of JFRP composite.
Fracture surface of jute fiber composite.
3.2 Creep test

Figure 4 shows the relationships between creep strain and time for untreated
composite specimen. For smaller load, there seems to be little transit creep
region. On the other hand, for bigger load, there is obviously transit creep region.
The slopes of constant creep region seem to be same. So, the difference of the
curve is emerged in transit region. The transit region continues by 15 hours.
147
0.04
200N
300N
Strain
0.03
0.02
0.01
0
0
Figure 4:
20
40
60
Time
80
100
Creep curves for untreated JFRP.
Table 1 shows the creep strains for 25% treated and untreated composite
specimen. From this table, it is understood that the effect of surface treatment is
emerged at lower load. In the case of 300N, the creep strains seem to be same.
From that result, there is obviously an effect of surface treatment. But when the
load is bigger, the effect is hidden. In the case of 300N, the shapes of the curves
are the same. On the other hand, in the case of 200N, the difference is emerging
in the transit region. So, the effect of surface treatment is bigger at the initial
stage for lower load. At the initial stage, there is the friction on the surface of
jute fibre. So, it is thought that the effect of this region is greater.
Table 1:
Comparison of creep strains for untreated and 25% treated jute

fiber composite at the end of tests. (Unit %).
200N
300N
Untreated
1.4
2.4
Treated
1.2
2.4
Because the effect of surface treatment is obtained in some conditions, the

other treatment using katch is conducted. Katch is the obtained from the skin of
tree. The result is shown in Fig.5. From this figure, it is understood that any
effects are not obtained for 200N case. For 300N case, the strain is bigger. It is
not understood that the katch is effective for these conditions. But in any other
conditions, it may have any effects. There are any possible other treatment
method, so some other methods have to be tried.
As the next stage, the surface treatment using water is conducted. Figure 6
shows the relationships between creep strain and time. For 200N, the creep strain
is bigger than that of untreated specimen. So, it is not understood that the surface
treatment using water is effective for creep properties. For the reason, an
astringency of a persimmon has a role to smooth the surface of jute fiber. For
only water, the effect of smoothing is not emerging.

0.04
200N
Strain
0.03
300N
0.02
0.01
0
0
Figure 5:
20
40
60
Time(h)
80
100
Creep curves of katch treated JFRP.
0.04
Strain
0.03
0.02
200N
300N
0.01
0
0
Figure 6:
20
40
60
Time(h)
80
100
Creep curves of water treated JFRP.
The effect of an astringency of a persimmon seems to be tannin.

effect of tannin will have to be examined in detail.
So, the
Conclusions
In this study, the static and creep properties of jute fiber reinforced composite are
examined and the effects of an astringency of a persimmon are also examined.
As a result, following conclusions are obtained.
(1) There is an effect of astringency of a persimmon to static tensile properties.
(2) The effect is dependent on the solutions densities. When the density is 25%,
it is more effective.
(3) For creep properties, there is an effect of surface treatment to creep strain.
The effect is big when the load is small. The effect emerges in the initial
stage.
(4) The effect of surface treatment using only water is not obtained. On the
other hand, in this case, the creep strain becomes big.
149
References
[1]
[2]
[3]
[4]
[5]
D. Nabi Saheb and J.P. Jog, Natural Fiber Polymer Composite: A Review,
Advances in Polymer Technology, Vol.18, No.4, 351 (1999).
H. Takagi and Y. Ichihara, JSME International Journal, Series A, 47,551
(2004).
K. Goda et al., Proc. of the 2nd Int. Workshop on Green Composites,
96(2004).
B. Singh et al., Composite Science and Technology, 60,571-589(2000).
J. Gassan, at. al., Composite Science and Technology, 60,2857 (2000).
151
Effects of forming conditions on mechanical

properties of resinless bamboo composites
H. Takagi1 & H. Mori2
1
Department of Mechanical Engineering,

The University of Tokushima, Japan
2
Graduate School of Engineering,
The University of Tokushima, Japan
Abstract
This paper deals with the development of resinless, unidirectional bamboo
fibre-reinforced eco-composites. Resinless bamboo composites were fabricated
by the hot-pressing of unidirectional bamboo fibres extracted by a steam
explosion method. The effects of forming conditions on their mechanical
properties were investigated. The forming temperature was changed every 20C
from room temperature to 180C, and the forming pressure used was 10 MPa
and 50 MPa. Except for the conditions below 60C, their tensile strength and
fracture elongation decreased with increasing forming temperature. Average
tensile strength and Youngs modulus of the resinless bamboo composites
fabricated at 120C and 50 MPa were 322 MPa and 37 GPa respectively.
Keywords:
natural fibre, bamboo, resinless, binderless, unidirectional
composites, tensile strength, modulus.
Introduction
Biodegradable composites that are also known as green composites are mostly
made from natural fibers and a biodegradable resin. Therefore the composites are
one of the most environment-friendly materials, because they are finally
biodegradable, and thus can be resolved into carbon dioxide and water through
the ideal decomposition stage. Hence, many researchers have been conducting
the research on biodegradable composites, and especially on their mechanical
properties [1-7]. However, most of the biodegradable resins are more expensive
than conventional plastics, such as PP and PE. In addition, extra energy is needed
doi:10.2495/HPSM06016

when the biodegradable resin is made from natural plants, such as corn and
potato.
Recently much attention has been paid to the characteristics of engineered
wood powder products, which are consolidated under a plastic condition derived
from a hot-pressing at high-temperature and high pressure [8-11]. This treatment
is called plasticizing. The wood-plastics made from waste wood powders
have also been developed [11], and they have a good prospect of new ecomaterials. However, their strength is low compared with that of conventional
engineering plastics. Thus, they need to be strengthened in order to extend their
application field.
In recent years, because of a domestic demand of bamboo decreases in Japan,
many bamboo forests have been going to ruin. The ruin of bamboo forests causes
several environmental problems such as landslides and encroachments against
other plants; therefore, much autonomy wants to cut down the bamboo and to
utilize them as a raw material of commercial products.
To overcome these problems, in this paper, we have tried to fabricate resinless,
unidirectional bamboo fiber composites by hot-pressing of steam-exploded
bamboo fibers (abbreviated as bamboo fibers hereafter) without using any
adhesives and binders. Particular focus is placed on the effect of hot-pressing
conditions on their tensile properties.
Experimental method
2.1 Bamboo fibers

Bamboo fibers were extracted from raw bamboo (phyllostachys pubescens) stem
by using a steam explosion method. The condition of the steam explosion was
180C and 40 min. After the steam explosion, bamboo stem became fragile, and
long bamboo fibers were easily taken out from the bamboo stem by hand. Soft
cells attached on the surface of extracted bamboo fibers were removed by wiping
with a wet cloth.
2.2 Preparation of unidirectional resinless bamboo composites
First, the bamboo fibers were cut into about 100 mm in length. Then, the bamboo
fibers were set into a metallic mould and finally hot-pressed at various
temperatures and pressures for 10 min. The dimensions of specimens are about
10 mm 100 mm 1 mm.
2.3 Tensile tests
Tensile tests were carried out using an Instron universal test machine (Model
5567). The tensile tests were performed at a crosshead speed of 1.0 mm/min,
with a gage length of 25 mm. Aluminium tabs in 2 mm thickness were glued at
the both ends of the tensile specimen to prevent damages caused by gripping.
The shape and dimensions of the tensile specimen are shown in fig. 1.
153
100
Figure 1:
Sample
10
Aluminium tab
30
Shape and dimensions of tensile specimens (unit: mm).
Under the moulding condition of R.T. and 10 MPa, bamboo fibres did not adhere
to each other; therefore, it was impossible for this sample to carry out the tensile
test. However resinless bamboo composites can be fabricated by hot-pressing at
temperatures over 80C. The colour of surface changed from dark brown to
black with increasing the moulding temperature. This colour change might be
attributed to the chemical reaction of lignin comprised in bamboo fibres.
Figure 2 shows typical stress-strain curves of resinless bamboo composites
moulded at 50 MPa as a function of moulding temperature. The initial slope of
the curve (corresponding to Youngs modulus) becomes steeper with increasing
the moulding temperature, and simultaneously tensile strength also increases.
The stress of composites moulded below 180C gradually decrease in the latter
part of the deformation. This phenomenon should be responsible for the
cumulative fibre fracture. However, the stress-strain curve for the composites
moulded at 180C becomes a straight line with the same steep slope as 120C. At
the same time, fracture behaviour is brittle, and fracture elongation becomes
considerably small value less than 0.005. In the case of moulding pressure of 10
MPa, almost the similar stress-strain curves are obtained as show in fig. 3.
The relationship between tensile strength of bamboo composites and moulding
temperature is shown in fig. 4. The tensile strength reaches a maximum value of
about 300 MPa at 120C at both of the moulding pressures. It can be seen from
this graph that the bamboo composites become brittle at temperatures over
120C, and bamboo fibres do not adhere to each other at temperatures below
120C.
The relationship between Youngs modulus and moulding temperature is
presented in fig. 5. Youngs modulus initially increases with the moulding
temperature, and then levels off at temperatures above 120C. This rise in
Youngs modulus is consistent with the increase in the density of samples. This
is because modulus as well as stress also increases with the moulding
temperature.
Figure 6 shows the relationship between fracture elongation and moulding
temperature. The fracture elongation is almost constant below 120C, and

decreases with increasing the moulding temperature thereafter. This drop in
fracture elongation might be attributed to the increased brittleness of bamboo
fibres derived from high temperature exposure [12].
500
R.T.
80C
120C
180C
Stress (MPa)
400
300
200
100
0
0
Figure 2:
0.01
0.02 0.03
Strain
0.04
0.05
Stress-strain curves of resinless bamboo composites moulded at 50

MPa.
500
80C
120C
180C
Stress (MPa)
400
300
200
100
0
0
Figure 3:
0.01
0.02 0.03
Strain
0.04
0.05
Stress-strain curves of resinless bamboo composites moulded at 10

MPa.
155
Tensile strength (MPa)
600
10MPa
50MPa
500
400
300
200
100
0
Figure 4:
150
200
50
100
o
Moulding temperature ( C)
Variation of tensile strength of resinless bamboo composites with

moulding temperature.
Young's modulus (GPa)
80
60
40
20
0
Figure 5:
10MPa
50MPa
50
100
150
200
o
Mouding temperature ( C)
Variation of Youngs modulus of resinless bamboo composites

with moulding temperature.
Fracture elongation (%)
2.5
1.5
1.0
0.5
0
Figure 6:
10MPa
50MPa
2.0
50
100
150
200
o
Moulding temperature ( C)
Variation of fracture elongation of resinless bamboo composites

with moulding temperature.
Conclusions
The tensile properties of resinless bamboo composites were investigated. The

major results obtained are summarized as follows:
1) Resinless bamboo composites can be made by hot-pressing at temperatures
over 80C. The optimum moulding temperature is 120C among the
conditions investigated.
2) Youngs modulus of resinless bamboo composites initially increases with
the moulding temperature, and then levels off at temperatures above 120C.
This rise in Youngs modulus is consistent with the increase in the density of
samples.
3) The fracture elongation of resinless bamboo composites is almost constant
below 120C, and decreases with increasing the moulding temperature
thereafter. This drop in fracture elongation might be attributed to the increased
brittleness of bamboo fibres.
Acknowledgement
The authors greatly acknowledge the partial financial support provided by the
president of The University of Tokushima in 2005.
157
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
Wollerdorfer, M. & Bader, H., Influence of natural fibres on the

mechanical properties of biodegradable polymers, Industrial Crops and
Products, 8(2), pp. 105112, 1998.
Luo, S. & Netravali, A. N., Interfacial and mechanical properties of
environment-friendly green composites made from pineapple fibers and
poly(hydroxybutyrate-co-valerate) resin, Journal of Materials Science, 34,
pp. 37093719, 1999.
Mohanty, A. K., Misra, M. & Hinrichsen, G., Biofibers, biodegradable
polymers and biocomposites: An overview, Macromolecular Materials
and Engineering, 276/277, pp. 124, 2000.
Shibata, M., Takachiyo, K., Yosomiya, R. & Takeishi, H., Biodegradable
polyester composites reinforced with short abaca fiber, Journal Applied
Polymer Science, 85(1), pp. 129138, 2002.
Takagi, H. & Ochi, S., Characterization of high-strength green
composites using Manila hemp fibers and starch-based resin, Proceedings
of the Third Japan-Canada Joint Conference on New Applications of
Advanced Composites (JCJC-III), pp. 1927, 2003.
Nishino, T., Hirao, K., Kotera, M., Nakamae K. & H. Inagaki, H., Kenaf
reinforced biodegradable composite, Composites Science and
Technology, 62, pp. 12811286, 2003.
Takagi, H. & Ichihara, Y., Effect of fiber length on mechanical properties
of green composites using a starch-based resin and short bamboo fibers,
JSME International Journal, Series A, 47(4), pp. 551555, 2004.
Hillis, W. E. & Rozsa, A. N., High temperature and chemical effects on
wood stability, Wood Science and Technology, 19, pp. 5766, 1985.
Angls, M.N., Reguant, J., Montan, D., Ferrando, F., Farriol, X. &
Salvad, J., Binderless composites from pretreated residual softwood,
Journal Applied Polymer Science, 73(12), pp. 24852491, 1999.
Salvad1, J., Velsquez1, J. A. & Ferrando, F., Binderless fiberboard
from steam exploded Miscanthus Sinensis: optimization of pressing and
pretreatment conditions, Wood Science and Technology, 37(3-4), pp.
279286, 2003.
Miki, T., Takakura, N., Iizuka, T., Yamaguchi K. & Kanayama, K.,
Production of long rods by sequential extrusion of wood powders, Journal
of Materials Processing Technology, 140, pp. 407412, 2003.
Ochi, S., Takagi, H. & Niki, R., Mechanical properties of heat-treated
natural fibers, Proc. of 1st High Performance Structures and Composites,
pp. 117125, 2002.
159
Compression moulding of jute fabric

reinforced thermoplastic composites
based on PLA non-woven fabric
T. Katayama1, K. Tanaka1, T. Murakami1 & K. Uno2
1
2

Marubeni Intex Co., Ltd., Japan
Abstract
To solve the problem of the large environmental burden in the disposal of FRP
(Fibre Reinforced Plastics), natural fibre and biodegradable resin have received a
lot of attention as easily degradable materials in the natural environment. To
enlarge the usage of the composites based on natural fibre and biodegradable
resin, cost reduction and enhancing its strength are essential subjects. In this
study, in order to develop the environment friendly biodegradable composites
that have high performance of strength, rigidity and productivity, the non-woven
stacking method was proposed and the influence of moulding conditions on
impregnating property and mechanical property of Jute Fabric Reinforced
Thermoplastic composites (JFRTP) based on PLA non-woven fabric was
discussed. It takes several minutes for PLA to impregnate into fibre bundles after
the material reaches preset moulding temperature due to its high melt viscosity
and twist of fibre bundles. The properly moulded JFRTP specimen reached about
115 MPa in bending strength, which is comparable to GFRTP with 20 wt % long
fiber made by injection moulding.
Keywords: green-composite, natural fibre, jute fibre, PLA, non-woven fabric,
compression moulding, bending property.
Introduction
While growing interest in environmental issues, environmental burden of FRP

disposal has been a big problem in recent years. The conventional materials for
FRP, such as glass fibre or carbon fibre as reinforcement, and epoxy or
polyamide as matrix are hardly degraded in natural environment and hardly
doi:10.2495/HPSM06017

incinerated due to their flame resistance. To settle this problem, the composites
based on natural fibre and biodegradable resin, called Green-composite, have
got a lot of attention as easily degradable materials in natural environment. To
enlarge its usage, reducing the manufacturing cost and enhancing the strength are
essential subjects.
For the reinforcement of the green composite, natural fibres, such as bamboo
fibre, ramie, flax, kenaf jute, are used. Among these natural fibres, jute is one of
superior materials due to its low cost, large amount of production and high
specific strength and has got a lot of attention (Acha et al. [1]).
Polylactic acid (PLA) is usually used for the matrix of the green composite,
because of its high strength, rigidity, melting point and productivity compared
with other biodegradable resins. In the case of using PLA as matrix, film type
PLA were usually used and they were laminated and moulded under
compression with sheet type of reinforcing fibres, which is called Film
Stacking, one of the well-applied impregnating methods of FRTP (Fibre
Reinforced Thermoplastics) (Plackett et al. [2], Khondker et al. [3]). However, it
is difficult to change the thickness of the film and handling of the film has also
some problem, therefore a new impregnating method is expected to be
developed. Non-woven fabric thermoplastic, which is used well for clothing,
filter, diaper and so on, is applied as matrix resin in this study. Using non-woven
fabric as matrix, mass per unit area or thickness can be easily changed. In this
study, for the purpose of improving the productivity of not only Greencomposite but also conventional FRTP, we propose a impregnating Non-woven
Stacking method replaced by Film Stacking, in which reinforcing fibre sheets
and matrix resin in the state of non-woven fabrics are laminated and moulded
under compressive stress. Jute Fabric Reinforced Thermoplastics (JFRTP) using
laminated jute plain woven fabrics and PLA non-woven fabrics were moulded
under compressive stress and the influence of moulding time and moulding
pressure on density and bending property were discussed.
Experimental procedure
2.1 Materials and method

Plain woven jute fabric (fig.1) of 430g/m2 is used for reinforcement and PLA
non-woven fabric sheet (fig.2) made by melt-brown method is used as matrix in
this study. The density, melting point and fibre diameter of PLA are 1.25g/cm3,
140 C, 2 - 3m, respectively.
10mm
Figure 1:
Plain woven Jute fabric.
50m
10mm
(a)
Figure 2:
161
(b)
PLA non-woven fabric (a) Macroscopic photography and (b)

Microscopic photography.
PLA non-woven fabric sheet is manufactured by melt-blown process. Figure

3 shows the schematic view of the melt-blown process. Melted resin with hot air
is blown from the nozzles and non-woven sheet will be formed on the conveyer.
Nozzle
Melting resin
Non-woven fabric
Conveyer
Figure 3:
Meltblown process.
Compression moulding is carried out by vacuum press moulding machine.

Jute fabrics in 450 mm 450 mm are stacked between four PLA non-woven
fabrics in 450 mm 450 mm shown in fig.4. All the composites are processed at
48 % fibre volume fraction. After setting the materials on the die, moulding is
carried out under vacuum condition of 0.1MPa. All the moulding conditions are
shown in table 1. After moulding, the composites are air-cooled at laboratory
temperature.
To evaluate the impregnation, density of the moulded product is measured
and compared with theoretical density that is calculated under the assumption of
no void in the composites.
2.2 Mechanical testing
Bending tests were conducted under a constant displacement rate of 1mm/min
using an Instron 5566 universal testing machine. The length and width of the
specimen are 60 mm and 15 mm, respectively and three point bending tests were
conducted at span length of 40 mm. The thickness of the specimen depended on
the moulding condition and varied from 1.8 mm to 3.1 mm.

Non-woven fabric
of matrix resin
Figure 4:
Table 1:
Laminated condition.
Molding pressures and times.
Pressure (MPa)
Time (s)
Reinforcing
fabric
1.0
20
40
60
2.0
3.0
120 180
4.0
300
600
3.1 Impregnation
The relationship between density and moulding time is shown in fig. 5. Before
60 s, the density increases drastically irrespective of the moulding pressure. After
300 s, however, it become stable and the density under moulding pressure of 3.0
MPa almost equal to that under 4.0 MPa.
Figure 6 shows the cross section of the moulding product under the moulding
pressure of 1.0MPa, whose results shows more clearly the difference of
moulding time on the impregnating state than other moulding pressure. Before
180 s, many visible voids are observed. After 300 s, however, matrix is
impregnated into fibre bundles. This observation corresponds with the results of
measured density change shown in fig. 4. Considering the fact that the higher
moulding pressure than 3.0 MPa and longer moulding time than 300 s dont
163
affect so much on the density and impregnation, moulding pressure of 3.0MPa

and moulding time of 300 s are optimum moulding value for the material used in
our study.
2.0
1.8
Density (g/cm )
1.6
1.4
1.2
1.0
0.8
Theoretical Density
1.0 MPa
2.0 MPa
3.0 MPa
4.0 MPa
0.6
0.4
0.2
0.0
Figure 5:
100 200 300 400 500 600 700

Molding Time (s)
Relationship between density and moulding time.
Figure 7 shows the temperature change at the centre of the material during
moulding under the moulding pressure of 1.0MPa. After 10 s from the start of
moulding, the material reaches the melting point of PLA (140 C) and after 60 s
it becomes stable. In spite of the higher temperature than the melting point of
PLA after 10 s, it takes almost 300 s for PLA to impregnate into the fibre
bundles. The high melt viscosity of PLA and the twist of jute fibre bundles are
considered to be the reason for this.
3.2 Bending property
Figures 8 and 9 show the influence of moulding condition on the bending
strength and the bending modulus, respectively. Before 300 s, the bending
strength and modulus drastically increase but after that they become stable. This
behaviour is similar to the result of density shown in fig.5. From the viewpoint
of the mechanical properties, moulding pressure of 3.0MPa and moulding time
of 300 s are optimum moulding value.
Figure 10 shows the stressdisplacement curve of bending test of the
specimen moulded under pressure of 3.0 MPa. In the case of specimen moulded
for 300 s and 600 s, the bending stress is decreased rapidly after reaching the
maximum value and they fractured in brittle manner. This behaviour is one of
the evidence that the resin was impregnated into the fibre bundles.
Figures 11 and 12 show the comparison of the bending strength and modulus
between our JFRTP and other materials. Injection moulded composite of glass
fibre reinforced polypropylene with 20wt% long fibre and jute fibre reinforced
polypropylene with 51wt% long fibre (Tanaka et al. [4]) are shown in these
figures. The PLA/JF product in our study has better bending properties than
GFRTP with 20 wt % long fiber.
500m
500m
(a) 20s
(b) 40s
500m
500m
(c) 1min
(d) 2min
500m
500m
(e) 3min
(f) 5min
500m
(g) 10min
Figure 6:
Cross section photos of the moulded product (Moulding pressure:

1.0MPa).
Molding Temperature ( )

240
220
200
180
160
140
120
100
80
60
40
20
0
Figure 7:
165
100
Temperature
1.0MPa).
profile
200
300
400
Molding Time (s)
during
500
moulding
600
(moulding
pressure:
Bending Strength (MPa)
140
120
100
80
60
1.0 MPa
2.0 MPa
3.0 MPa
4.0 MPa
40
20
0
Figure 8:
100
200 300 400 500

Molding Time (s)
600
Relationship between bending strength and moulding time.
Conclusions
To develop the environment friendly biodegradable composites that have high

strength and high productivity, non-woven stacking method was proposed and
Jute Fabric Reinforced Thermoplastic composites (JFRTP) based on PLA nonwoven fabric were moulded and the influence of moulding conditions on
impregnation and mechanical property was discussed. It takes several minutes
for PLA to impregnate into fibre bundles after the material reaches preset
moulding temperature, due to its high melt viscosity and twist of fibre bundles.
The properly moulded JFRTP specimen reached about 115 MPa in bending
strength, which is comparable to GFRTP with 20 wt % long fiber made by
injection moulding.

7.0
Bending Modulus (GPa)
6.0
5.0
4.0
2.0
1.0
0.0
Figure 9:
1.0 MPa
2.0 MPa
3.0 MPa
4.0 MPa
3.0
100
200 300 400 500

Molding Time (s)
600
Relationship between bending modulus and moulding time.
Bending Stress (MPa)
140
120
20s
40s
60s
120s
180s
300s
600s
100
80
60
40
20
0
Bending Strength (MPa)
Figure 10:
3 4 5 6 7 8
Displacement (mm)
9 10
Relationship between bending stress and displacement.
140
120
100
80
60
40
20
0
PP/GF 20wt% PP/JF 51wt% PLA/JF 53wt%
Long Fibre [4] Long Fibre [4] Plain Weave
Figure 11:
Bending strength of the 3 types of FRTP [4].
Bending Modulus (GPa)
167
8
7
6
5
4
3
2
1
0
PP/GF 20wt% PP/JF 51wt% PLA/JF 53wt%
Long Fibre [4] Long Fibre [4] Plain Weave
Figure 12:
Bending modulus of the 3 types of FRTP [4].
Acknowledgement
This study was partially supported by the Academic Frontier Research Project on
NewFrontier of Biomedical Engineering Research of Ministry of Education,
Culture, Sports, Science and Technology.
References
[1]
[2]
[3]
[4]
B.A. Acha, N.E. Marcovich, M.M. Reboredo, Journal of Applied Polymer

Science, vol.98, No.2, (2005), pp.639-650.
D. Plackett, T.L. Andersen, W.B. Pedersen, L. Nielsen, Composites
Science And Technology, vol.63, No.9, (2003), pp.1287-1296.
O.A. Khondker, U.S. Ishiaku, A. Nakai, H. Hamada, Journal of Polymers
and the Environment, vol.13, No.2, (2005), pp.115-126.
T. Tanaka, N. Tashiro, Japan Plastics, vol.56, No.7 (additional volume),
(2005), pp.95-104.
169
Quality control of fibers end-milled from

bamboo pipe using spiral tool path
K. Ogawa1, E. Aoyama2, T. Hirogaki2, Y. Tomioka2
& H. Nakagawa1
1
Department of Mechanical Systems Engineering,

The University of Shiga Prefecture, Japan
2
Abstract
The utilization of unused forest products is beneficial, and therefore the
application of natural fibers to FRP products has been proposed. Bamboo fibers
have attracted particular attention because bamboo has the fastest growth rate
among various types of renewable natural fibers. Moreover, bamboo fibers have
high specific strength and stiffness appropriate for structural materials. In the
present study, we explain a new method designed to obtain high-quality bamboo
fibers by effectively end-milling them with a machining center (MC) using
numerical control (NC) and an automatic tool changer (ATC), which enables the
manufacture of a wide variety of products. NC programs encoding a spiral tool
path were used for the bamboo pipe in order to obtain bamboo fibers effectively.
The fiber length and diameter were evaluated at various feed rates, cutting
speeds, and depths of cut in the fiber direction. Microscopic observation showed
that the length of the fibers obtained can be controlled by selecting the depth of
cut along the bamboo fiber direction and the diameter can be controlled by
adjusting the feed of the end-mill center, which is determined by spindle speed
and feed speed of end-mill under a constant end-mill diameter. Thus the desired
fiber shape can be controlled with consistently high accuracy using end-milling
by adjusting the cutting conditions. Moreover, using spiral tool path end-milling
with a straight cutting edge tool can efficiently acquire high quality straight
bamboo fibers with no thermal damage.
Keywords: bamboo fiber, end-mill, machining center, spiral tool path.
doi:10.2495/HPSM06018
Introduction
Recently, industrial products made of fiber reinforced plastic (FRP) such as glass
fiber-reinforced plastic (GFRP), are being utilized extensively because they have
high specific tensile strength (that of GFRP is as high as steel) but are
significantly lighter. The use of such materials may have a negative
environmental impact because of problems associated with their disposal or
destruction. Conversely, the utilization of unused forest products is beneficial to
the environment, and therefore a method of acquiring and applying natural fibers
to FRP products has been proposed [1-7]. Bamboo fibers, in particular, have
attracted attention because bamboo has the fastest growth rate among the various
types of renewable natural fibers, and is grown throughout Japan. Moreover,
bamboo fibers have high specific strength and stiffness as appropriate for
structural materials [8, 9]. Bamboo fibers have been obtained from the wood-like
plant through various methods, such as crushing or heat steaming the bamboo
stems. Using these methods, however, it is difficult to control the diameter,
length and other dimensions, as well as to obtain large quantities of a consistent,
uniform shape as necessary for high-quality fiber components in industrial
materials [10]. This paper attempts to obtain high quality bamboo fibers by endmilling with a machining center (MC) as an alternative to previous conventional
methods. The MC makes it possible to manufacture many kinds of products and
the proposed method boasts some advantages over conventional methods. The
shape of the bamboo fibers can be controlled because all the machining center
processes are controlled by digital program. Moreover, the spiral tool path
should produce uniform bamboo fibers and allow accurate control of the shape
of fibers by variation of the cutting conditions.
Experimental method
2.1 Material
Naturally growing Mousouchiku bamboo was used as the base material. The
bamboo was cut using a metal saw into 100 mm-long bamboo pipes in order to
remove the bamboo joints. The bamboo pipes were then set on the table in an
MC and end-milled. The waste bamboo obtained was used as the bamboo fibers
used in the experiments.
2.2 Machining equipment and conditions
A ROBODRILL (Type: T14iDs, FANUC Co., Ltd.) was used as the MC for
milling. The end-mill tool used in the experiments was a square type with two
straight cutting edges, and made of high-speed steel without coating films on the
tool surface. The end-mill diameter was 6 mm. Table 1 shows the machining
conditions. The tool followed a spiral path from the perimeter to the center with
a cutting depth of 50 m per one cycle in the radius direction of bamboo, as
shown in Fig. 1.
Table 1:
171
Machining conditions.
Spindle speed S (rpm)

Feed speed F (mm/min)
Depth of cut in radius direction Rv (mm)
Depth of cut in bamboo fiber direction Ad (mm)
End-mill radius r (mm)
Number of end-mill cutting edge Z
2500, 5000, 10000,

15000, 20000
250, 500, 750,
1000, 2000
0.05
2.5, 5, 8, 10
3
2
Spiral tool path
End-mill
Bamboo pipe
Figure 1:
Spiral tool path.

End-mill
Bamboo pipe
Figure 2:
Convex contour cutting.
2.3 Theoretical cutting configuration

Figure 2 shows a diagram of the geometrical configuration. Here, fr is the feed of
the cutting edge (mm/tooth), fen is the feed of the end-mill center (mm/tooth), r is
the end-mill radius (mm), Rv is the depth of cut in the radius direction per cycle

(mm), Rb is the bamboo radius after cutting (mm), is the angle between the fr
and fen directions (rad), and en is the cutting engagement angle (rad). The tool
path approaches to a circumference path because the depth of cut in the bamboo
pipe radius direction is very small comparing with the end-mill radius and the
bamboo pipe radius.
The cutting arc length L and the maximum thickness h are obtained
respectively based on the geometrical configuration shown in Fig. 2.
(1)
L = r sin en
h = f r sin( en + )
(2)
Here, the depth of cut in the radius direction per cycle Rv is very small
compared with the end-mill radius r and the bamboo pipe radius Rb. As a result,
the cutting engagement angle en is extremely small, so as to be nearly zero.
Therefore, eqn (3) can be obtained from eqn (1) using the approximation
sinenen.
(3)
L = r en
Eqns (4) and (5) can be obtained by the geometrical relation of the triangle
that consists of the bamboo pipe center point, the end-mill center point, and the
sharp end of the end-mill cutting edge, as shown in Fig. 2.
(4)
r sin en = (Rb + Rv ) sin
r cos en + (Rb + Rv ) cos = Rb + r
(5)
Eqn (6) is obtained from the relation that the ratio of the feed of the cutting
edge fr and the feed of the end-mill center fen is equivalent to the ratio of the
distance from the spiral path center to the end-mill center point and to the sharp
end of the end-mill cutting edge.
f en
R +r
= b
fr
Rb + Rv
(6)
Here, the eqn (7) is conducted by eqns (2), (4), (5), and (6).
h = f en sin en
(7)
Therefore, eqns (8) and (9) can be obtained by replacing the parameters
utilized in this study shown in Table 1 into eqns (3) and (7).
r Rv
r
(8)
Rv
R
2 v
r
r
(9)
L = r arccos
h=
F
S Z
2.4 Evaluation method of bamboo fiber quality

The length and diameter of one hundred randomly selected bamboo fibers were
measured using an optical microscope. Then aspect ratio, calculated by dividing
the fiber length by the diameter, was then calculated to give an indication of the
length and narrowness of the obtained fibers. Here, a theoretical fiber diameter
Dth seems (L+h)/2 in this method.
15
Theoretical
200
100
Fiber length (mm)
Fiber diameter ( m)
300
10
5
0
100
200
300
400
Cutting speed (m/min)
100
200
300
400
(a) Fiber diameter
(b) Fiber length

Coefficient of variation
100
Aspect ratio
80
60
40
20
0
0
0.5
Fiber diameter
Fiber length
Aspect ratio
0.4
0.3
0.2
0.1
0.0
100
200
300
400
(c) Aspect ratio

Figure 3:
173
100
200
300
400
(d) Coefficient of variation
Change of bamboo fiber shape

(fen=0.05 mm/tooth, Ad =10 mm).
with
cutting
speed
3.1 Influence of cutting speed on acquired bamboo fiber quality

Figures 3(a), 3(b) and 3(c) show the changes in fiber shape with cutting speed
under a constant feed of the end-mill center of 0.05 mm/tooth and a cut of depth
in the fiber direction of 10 mm. Figure 3(d) shows the coefficient of variation of
the acquired bamboo fiber shapes at various cutting speeds. The coefficient of
variation is calculated by dividing the standard deviation by the mean value.
Here, the cutting speed V (m/min) and the feed of end-mill center fen can be
calculated by eqns (10) and (11), respectively.
V = 2r S
F
f en =
S Z
(10)
(11)

Therefore, the feed speed F was adjusted for various spindle speed in order to
match the constant feed of the end-mill center. Under these conditions, the endmilled sections of the fibers are theoretically uniform. That is, L and h are 548.5
m and 9.1 m, respectively, from eqns (8) and (9). As a result, the theoretical
diameter is about 276.5 m.
End-mill
Bamboo pipe
(a) Optical view

Figure 4:
37 C
40 C
(b) V=94 m/min

(S=5000 rpm)
(c) V=188 m/min (d) V=377 m/min

(S=10000 rpm)
(S=20000 rpm)
Cutting
temperature
as
monitored
(fen=0.05 mm/tooth, Ad =10 mm).
67 C
by
thermography
The results verify that the diameter of the end-milled bamboo fiber decreases
with cutting speed until about 200 m/min though obtained fiber diameter has
good agreement with theoretical one under low cutting speed. However, above
this speed, the fiber diameter remains almost constant. On the other hand, the
fiber length is almost constant, at 10 mm, for any cutting speed. As a result, the
aspect ratio increases with cutting speed until about 200 m/min and then plateaus
at a fiber diameter of about 150 m at the cutting speeds above 200 m/min.
Higher cutting speed with greater feed speed improves the machining
throughput. However, the coefficient of variation for the fiber diameter, fiber
length and aspect ratio show a similar tendency of increasing with cutting speed.
Therefore, a cutting speed of about 200 m/min is suitable in order to obtain high
efficiency machining and maintain low variation. It seems that the bamboo fiber
diameter is controlled by cutting speed under about 200 m/min.
On the other hand, the increased cutting heat generated may damage the
bamboo fiber in end-milling; the cutting temperature increases with cutting
speed. The cutting temperature in steel rises to 800 C or higher. The melting
points of bamboo pipe components of cellulose, hemi cellulose, and lignin are
about 240, 180, and 420 C, respectively. The temperature in bamboo cutting was
monitored by thermography as a non-contact method. It can be seen that the
temperature in bamboo cutting also increases with cutting speed as shown in
Fig. 4. However, up to a cutting speed of 200 m/min the temperature grows to no
more than about 70 C. It can be concluded that under these conditions bamboo
fibers end-milled from bamboo will sustain no thermal damage.
3.2 Influence of cut of axial depth along the fiber direction on acquired
bamboo fiber quality
Figures 5 shows the microscopic images of end-milled bamboo fibers under
various axial cut depths. It can be seen that uniform, straight fibers with no
175
thermal damage are acquired. However, fibers end-milled appear as powders

with smaller axial cut depth.
5mm
(a) Ad =2.5 mm
Figure 5:
(b) Ad =5 mm
(c) Ad =10 mm
Microscopic images of fibers

(fen=0.05 mm/tooth, V=188 m/min).
200
bamboo
100
10
5
0
5
10
15
Axial depth of cut (mm)
5
10
(a) Fiber diameter

80
60
40
20
0
0
5
10
15
(c) Aspect ratio

Figure 6:
15
(b) Fiber length

Coefficient of variation
100
Aspect ratio
from
15
Theoretical
Fiber length (mm)
Fiber diameter ( m)
300
end-milled
0.5
Fiber diameter
Fiber length
Aspect ratio
0.4
0.3
0.2
0.1
0.0
0
10
15
(d) Coefficient of variation
Change of bamboo fiber shape with depth of cut in fiber direction.

(fen=0.05 mm/tooth, V=188 m/min).
Figures 6 shows the changes in fiber shape with axial depth of cut under a
constant feed of the end-mill center of 0.05 mm/tooth and a cutting speed of
188 m/min, the conditions determined to be optimum for acquiring high quality
bamboo fibers efficiently as described in section 3.1. Moreover, L and h are
15
500
400
300
200
Ad=2.5mm
Ad=5mm
Ad=10mm
Theoretical
100
0
Fiber length (mm)
Fiber diameter ( m)
respectively 548.5 m and 9.1 m, respectively, under the same condition of

constant fiber section for various cutting conditions as given in section 3.1.
Ad=2.5mm
Ad=5mm
Ad=10mm
10
5
0
0.0
0.1
0.2
0.3
Feed of end-mill center
(mm/tooth)
0.0
(a) Fiber diameter
Aspect ratio
100
0.1
0.2
0.3
(mm/tooth)
(b) Fiber length

Ad=2.5mm
Ad=5mm
Ad=10mm
80
60
40
20
0
0.0
0.1
0.2
0.3
(mm/tooth)
(c) Aspect ratio

Figure 7:
Change of bamboo fiber shape with feed of the end-mill center and
the depth of cut in fiber direction (fen=0.05 mm/tooth,
V=188 m/min).
It is clear that the fiber diameter measured shows almost constant value,
which is almost half of theoretical one, with the axial depth of cut as shown in
Fig. 6(a). On the other hand, the fiber length increases with the axial depth of cut
as shown in Fig. 6(b). The result shows that the fiber length agrees very well
with the axial depth of cut, as the data points lie almost exactly on a line of slope
1 (Fig.6(b)). As a result, aspect ratio increases with axial depth of cut as shown
in Fig. 6(c). Fibers with a large aspect ratio are desirable to confer superior
material properties to the FRP. It may be effective to increase the axial depth of
cut in order to obtain high quality bamboo fibers. However, the coefficient of
variation of fiber diameter, length and aspect ratio seem to have minimum values
at an axial cut depth from 5 to 8 mm. Therefore, it seems effective to set the axial
cut depth from 5 to 8 mm in order to acquire high quality bamboo fibers.
177
3.3 Influence of feed of end-mill center on acquired bamboo fiber quality

In this section, we investigate the change in shape of the fiber section with
changes in the feed of the end-mill center adjusting spindle speed and the feed
speed based on eqn (11). In this study, cutting arc length L is constant because L
is a function of the end-mill radius and the depth of cut in the radius direction per
one cycle as described by eqn (8). However, end-mill radius is a constant 3 mm
and the depth of cut in the radius direction per one cycle is a constant 0.05 mm.
On the other hand, the maximum thickness of h changes with spindle speed and
feed speed under a constant end-mill radius, depth of cut in the radius direction
per one cycle, and number of the cutting edge of the end-mill as shown in eqn
(9). Therefore, the shape of bamboo fibers end-milled under various values for
the feed of the end-mill center, determined by dividing the feed of the end-mill
center by the spindle speed, and the axial cut depth in the bamboo fiber direction
are investigated in this section. Larger values for the feed of the end-mill center
give thinner end-milled bamboo fibers. The maximum thickness of h is 3.0, 9.1,
and 36.4 m for values of the feed of the end-mill center of 0.02, 0.05, and 0.20
mm/tooth, that is, the theoretical fiber diameter is 276, 279, and 292 m,
respectively, in this study.
Figure 7 shows the experimental results. It is clear that the smaller diameter
bamboo fibers can be obtained with smaller values of the feed of the end-mill
center, independent of the axial depth of cut in the fiber direction as shown in
Fig. 7(a). In the case of a feed of the end-mill center of 0.20 mm/min, the fiber
diameter measured agrees with theoretical one of 292 m. However, the fiber
diameter becomes smaller in case of smaller feed of the end-mill center of 0.05
mm/tooth. On the other hand, the fiber length is controlled very well by the axial
depth of cut at a feed of the end-mill center of 0.05 mm/tooth or greater.
However, the fiber length becomes smaller in the case of a small feed of the endmill center of 0.02 mm/tooth for each axial depth of cut. As a result, the aspect
ratio attains a maximum at a feed of the end-mill center of 0.05 mm/tooth when
h is 9.1 m. Therefore, a feed of the end-mill center of 0.05 mm/tooth seems
suitable for obtaining superior bamboo fibers in end-milling of bamboo pipe.
Conclusions
End-milling using a machining center that follows a spiral tool path was tried for
the production of high quality bamboo fibers. The shape of the fibers end-milled
from bamboo pipe was evaluated under various cutting conditions, giving the
following results.
(1) End-milling with a machining center that follows a spiral tool path is
effective for obtaining high quality, straight bamboo fibers with no thermal
damage.
(2) It is possible to accurately control the bamboo fiber length by adjusting the
axial depth of cut.
(3) A higher cutting speed can provide bamboo fibers with smaller diameter.
However, the decrease in bamboo fibers diameter with higher cutting speed
saturates at a certain cutting speed.

(4) Suitable conditions were determined for acquiring bamboo fibers of the
desired shape.
Acknowledgement
This research is supported by the Research and Development Center for
Engineering Application of Bamboo Resources at Doshisha University.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
Ogawa K., Hirogaki T., Aoyama E. and Katayama T., Data mining of
optimum condition to acquire bamboo micro-fiber with mechanical
method, High Performance Structures and Materials II, WIT PRESS,
pp.441-450, 2004.
Jana S. C. and Prieto A., On the development of natural fiber composites
of high-temperature thermoplastic polymers, Journal of Applied Polymer
Science, Vol. 9, No. 28, pp.2159-2167, 2002.
Li H. and Sain M. M., High stiffness natural fiber-reinforced
polypropylene
composites,
Polymer-Plastics
Technology
and
Engineering, Vol.42, No.5, pp.853-862, 2003.
Luo S. and Netravani A. N., Interfacial and mechanical properties of
environment-friendly green composites made from pineapples fibers and
poly (hydroxybutyrate-co-valerate) resin, Journal of Materials Science,
Vol. 34, No. 15, pp.3709-3719, 1999.
Ogawa K., Hirogaki T., Aoyama E., Tomioka Y., and Shiomi T., Shape
Control of Fibers End-milled from Bamboo with a Machining Center,
Proc. of The 3rd International Conference on Leading Edge
Manufacturing in 21st Century, pp.565-570, 2005.
Thwe M. M. and Liao K., Environmental effects on bambooglass/polypropylene hybrid composites, Journal of Materials Science,
Vol. 38, No. 2, pp.363-376, 2003.
Martikka H. and Katajisto J., Study of Natural Fiber Reinforced
Biodegradable Composite Materials for Designing Optimally Sustainable
Products, Proc. of the 23rd Riso International Symposium on Materials
Science, pp. 251-258, 2002.
Takagi H. and Ichihara Y., Effect of fiber length on mechanical properties
of green composites using a starch-based resin and short bamboo fibers,
JSME International Journal, Series A: Solid Mechanics and Material
Engineering, Vol. 47, No. 4, pp.551-555, 2004.
A. Varada RaJulu, S. Allah Baksh, G. Ramachandra Reddy and K.
Narasimha Chary, Chemical Resistance and Tensile Properties of Short
Bamboo Fiber Reinforced Epoxy Composites, Journal of Reinforced
Plastics and Composites, Vol. 17, No. 17, pp. 1507-1511, 1998.
Seema Jain, U. C. Jindal and Rakesh Kumar, Development and fracture
mechanism of the bamboo/polyester resin composite, Journal of Materials
Science Letters, 12, pp. 558-560, 1993.
179
Characteristic behaviors of CFRP and GFRP

at cryogenic temperature under static and
cyclic loadings
S. Kubo, K. Okubo & T. Fujii
Department of Engineering, Doshisha University, Japan
Abstract
Characteristic behaviors of plain-woven Carbon Fiber Reinforced Plastics
(CFRP) at cryogenic temperature were investigated under static and cyclic
loading. The test results were compared with those of Glass Fiber Reinforced
Plastics (GFRP). Tensile behaviors of monofilaments were also evaluated. The
test results showed that two knee-points appeared in the stress-strain curve of
GFRP under static load, while that of CFRP was almost linear. Both strength and
failure strain of the CFRP at cryogenic temperature were lower than those of
CFRP at room temperature, because the epoxy resin was brittle at cryogenic
temperature. At cryogenic temperature, the knee-point was also shown in the SN curve of GFRP of the fatigue test but that was not shown in the S-N curve of
CFRP. The elastic modulus of CFRP at cryogenic temperature suddenly
decreased in the final stage of fatigue, while such change was not observed at
room temperature. The thermal fatigue test where no cyclic loads were applied to
the specimen was also conducted in order to investigate the damage progression
due to temperature change. After the thermal fatigue test, the residual strength of
the CFRP and GFRP were measured at room temperature. The change in residual
strength was explained by the difference of the coefficient of thermal expansion.
This paper also mentioned that, at cryogenic temperature, the stiffness reduction
under cyclic loading was related to the local thermal stress by thermal cyclic
fatigue. It should be said that the carbon fiber was failed with accompanying
critical crack propagation of the matrix due to the brittle of resin at cryogenic
temperature.
Keywords: fiber reinforced plastics, cryogenic temperature, coefficient of
thermal expansion, residual strength, stiffness reduction.
doi:10.2495/HPSM06019
Introduction
Fiber reinforced plastics (FRP) are widely used for the structural component due
to their low coefficient of thermal expansion as well as high specific strength and
stiffness [13]. The FRPs are also expected in use at harsh environment, for
example outer space and liquid fuel tank for disposable rocket. The GFRP (Glass
Fiber Reinforced Plastics) has been partly used at low temperature of a ship, and
Sakurai [4] discussed the strength of GFRP for them. However their use is
limited but aluminium alloy is still used for many applications. To apply the
FRP, for example, for the fuel tank instead of the aluminium alloy, behaviors of
FRP such as CFRP (Carbon Fiber Reinforced Plastics) and GFRP under the
cryogenic temperature should be cleared.
The purpose of this study is to investigate the mechanical properties of CFRP
at room and cryogenic temperature using liquid nitrogen, in comparison with
those of GFRP. The mechanical properties of carbon and glass monofilaments at
room and cryogenic temperature were also shown. The coefficients of thermal
expansions were measured for FRP as well as the epoxy resin for polymer
matrix. The tensile tests of CFRP and GFRP were conducted at room and
cryogenic temperature. The tension-tension cyclic loading was applied to the
CFRP and GFRP with hydraulic servo testing machine having nitrogen cryostat.
Thermal fatigue tests (without loading) were also conducted in which
environmental temperature was alternated between room and cryogenic
temperature. The residual strengths of the damaged specimens were measured at
room temperature after the thermal fatigue test.
Aluminium
R70
Cut off
20
8
80
150
Adhesion
t = 2 mm
Figure 1:
Monofilament
Grip area
25
Diameter 6.2-8.3m
Dimensions of specimens of composite and monofilament.
2.1 Materials and specimens

Plain-woven carbon (PYROFIL TR3110M: Mitsubishi Rayon Co., Ltd.), epoxy
resin (bisphenol A type) (E-828: Japan Epoxy Resins Co., Ltd.) and
Cyclopolyamide (E-113: Japan Epoxy Resins Co., Ltd.) were used as
reinforcement, matrix and hardener for CFRP (Vcf=65.3%). Glass cloth
(MS253C: Asahi Fiber-Glass Co., Ltd.), polyester resin (5595APT-S: DH
Material Inc.), methyl ethyl ketone per oxide (nacalai tesque) and cobalt
181
naphthenate (6% solution) (Nacalai Tesque Co., Ltd.) were used as

reinforcement, matrix, hardene and catalyzer for GFRP (Vgf=35.0%),
respectively. The specimen was fabricated by hand lay-up method. The
dimensions of the dog-bone shaped specimens of the composites were 150 mm
long, 20 mm wide and 2 mm thick as shown in Fig. 1. Aluminium tabs with 35
mm long, 20 mm wide and 2 mm thick were glued on the end of specimens. The
dimensions of the specimens of carbon and glass monofilaments are also shown
in Fig. 1.
2.2 Experiments
2.2.1 Static tensile tests
The carbon and glass monofilaments were extracted from their fabrics. In the
tensile tests for monofilaments, at first, the diameter of monofilaments was
measured using optical microscope. The tensile testing machine for
monofilaments was shown in Fig. 2(a). Tensile load was applied to the
monofilament at room and liquid nitrogen temperature. The test conditions of
monofilaments specimens were referred to Japanese Industrial Standards (JIS
R7606). The strain was calculated from the crosshead displacement.
The tensile testing machine for the composites and resin were also shown in
Fig. 2(b). Extensometer was attached to the specimen to measure the strain.
(a)
Figure 2:
(b)
Tensile tests at cryogenic temperature. (a): for monofilament,

(b): for composite.
2.2.2 Thermal mechanical analysis (TMA)

The coefficient of thermal expansion m of the pure epoxy resin, c of the CFRP
were measured from 60oC up to room temperature (RT) by the thermal
mechanical analysis (SHIMADZU TMA-60) referred to JIS (K7197) [5]. The
heating rate was 5oC/min.
2.2.3 Tension-tension fatigue tests
At room and cryogenic temperature, the tension-tension fatigue test was
conducted with a Shimadzu Co. EHF-UB50kN at constant amplitude of load
with sinusoidal wave at a frequency of 5 Hz and stress ratio R (R=min/max)=0.1
for CFRP and R=0 for GFRP.

2.2.4 Thermal fatigue tests
The conditions are shown in Table 1. The temperature of the specimens was
alternated from cryogenic temperature to room temperature. The specimens were
alternatively transferred with an aluminum frame into LN2 from ambient air. At
first, specimens were cooled down to 196 oC and kept for 2 min. Then they
were held in ambient air for 5 min. To reduce the time required for specimens to
return to the ambient temperature, a fan was used during the periods when the
specimens were held in ambient temperature [6].
The residual stress was measured after thermal fatigue test.
Table 1:
Materials
Tmax (oC)
CFRP and GFRP
23
(oC)
Thermal fatigue test condition.

Tmin (oC)
-196
(oC)
Maximum number of cycles

1, 3, 5, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100
3 Results and discussions

3.1 Strength and failure strain of monofilaments
The mechanical properties of glass and carbon monofilament at room and
cryogenic temperature are shown in Table 2. The strength and the failure strain
of glass monofilament were increased about 99% and 82% respectively at
cryogenic temperature, compared with those at room temperature. The strength
of carbon monofilament was also increased about 11%, while the failure strain of
carbon monofilament was decreased about 10% at cryogenic temperature. Fig. 3
and 4 show the fracture surfaces of the glass and carbon monofilament at room
and cryogenic temperature, respectively. At room temperature, the typical
cleavage fracture surfaces were observed in the surfaces of glass and carbon
monofilament. On the other hand, at cryogenic temperature, there was rough area
in the center of the fracture surface of glass and carbon monofilament.
Table 2:
Mechanical properties of carbon and glass monofilaments.
Materials Diameter
[m]
Temperature
[oC]
23+2 (Room Temp.)
CF
240
3503
1.34
269
3168
1.22
68.5
1330
1.70
67.7
2650
3.10
6.08.5
-196 (Cryogenic Temp.)
23+2 (Room Temp.)
GF
Elastic modulus Tensile strength Failure str

[%]
[GPa]
[MPa]
7.814.1
-196 (Cryogenic Temp.)
The reduction in failure strain and strength of carbon monofilament is

considered as the brittle behaviors of the carbon fiber at cryogenic
183
temperature [7]. However the glass monofilament had rather different

characteristics from those of carbon monofilament, in which the failure strain of the
glass monofilament was increased at cryogenic temperature even though it has
rough fracture surface.
5m
x7.0 k
x7.0 k
5m
Figure 3: Fracture surface of carbon

monofilament (left: at CT,
right: at RT).
5 m
9.0 k
7.0 k
5m
Figure 4: Fracture surface of glass

monofilament (left: at CT,
right: at RT).
3.2 Static mechanical properties of CFRP and GFRP

Static strengths were listed in Table. 3. The stress strain (S-S) curves of CFRP
and GFRP at room and cryogenic temperature were shown in Fig. 5 and 6,
respectively. The S-S curves of CFRP were linear at both of room and cryogenic
temperature. The S-S curve of GFRP at room temperature was also linear.
However, that of GFRP at cryogenic temperature showed two knee-points. The
elastic modulus of the resin is usually increased while its failure strain is
decreased at cryogenic temperature. Actually, the elastic modulus of CFRP and
GFRP at cryogenic temperature increased in comparison with that at room
temperature in this test. At cryogenic temperature, the strength and failure strain
of GFRP should be contributed by the increase of the strength and failure strain
of glass monofilament at cryogenic temperature. Little changes of the
mechanical properties of CFRP were found in the tensile test at room and
cryogenic temperature.
1000
500
Stress[MPa]
800
Stress[MPa]
600
RT
CT
600
400
200
400
300
200
RT
CT
100
0.0
0.5
1.0
Strain[%]
1.5
2
3
Strain[%]
Figure 5: S-S curves of CFRP at RT Figure 6: S-S curves of GFRP at RT

and CT.
and CT.

Table 3:
Static strength of CFRP and GFRP at RT and CT.

Static strength : st [MPa]
Stress level
=max/st [%]
CFRP
RT
639.1
CT
630.0
GFRP
RT
294.6
CT
564.5
3.3 Failure strain of resin

The stress strain curves of the monolithic epoxy resin are shown in Fig. 7. At
room temperature, the failure strain was significantly higher than that at
cryogenic temperature. It is generally known that the failure strain decreases at
cryogenic temperature in comparison with that at room temperature due to the
brittle behavior of polymer resin [8].
The observed failure strain of pure epoxy resin used in this study was about
0.21% at cryogenic temperature. According to the coefficient of thermal
expansion m, the thermal strain of monolithic epoxy resin was calculated about
0.41% at cryogenic temperature. It was half in failure strain at cryogenic
temperature, compared with that at room temperature.
Stress[MPa]
60
45
30
RT
CT
15
0
0.0
0.5
1.0
1.5
2.0
Strain[%]
Figure 7:
S-S curves of pure epoxy resin at RT and CT.
3.4 Effect of thermal expansion coefficient on residual strength

Carbon monofilament has small minus number of the coefficient of thermal
expansion because carbon monofilament behaves like ceramic [9]. In this study,
the coefficients of thermal expansions of CFRP and its pure resin were measured
by TMA to estimate the residual stress at 60oC up to room temperature
(Table 4). The coefficient of thermal expansion of the epoxy resin was lower
than that of the CFRP.
The thermal fatigue stress was applied to the CFRP and GFRP specimens
[1012]. The temperature change of the specimens was repeated from cryogenic
185
temperature to room temperature. Figure 8 and 9 show the S-S curves of CFRP
and GFRP, respectively after thermal fatigue test, in which the thermal stress
was applied to the specimens before the static test.
Residual strengths [13] of CFRP and GFRP were almost equal to their each
static strength even after 100 cycles of thermal fatigue loading. However the
residual elastic modulus of CFRP was decreased by the cyclic thermal stress.
There was the large difference in the coefficient of thermal expansion of carbon
fiber and that of pure epoxy resin. Therefore the epoxy resin in CFRP should be
damaged during the thermal fatigue test [14], because the magnitude of thermal
free strain was larger than the failure strain of pure epoxy resin at cryogenic
temperature as shown in section 3.3.
Table 4:
The coefficient of thermal expansion of CFRP and the pure epoxy

resin.
Materials
CFRP
Tmin (oC)
23
-60
The coefficient of thermal expansion (-oC)

6.8e-5
4.9e-5
700
300
600
250
500
Stress[MPa]
Stress[MPa]
Epoxy resin
Tmax (oC)
400
300
200
0 cycle
100 cycles
100
0
200
150
100
0 cycle
100 cycles
50
0
0.5
1
1.5
Strain[%]
Figure 8: S-S curves of CFRP after

thermal fatigue test at RT.
0.5
1.5
Strain[%]
Figure 9: S-S curves of GFRP after

thermal fatigue test at RT.
On the other hand, in Fig. 9, the residual elastic modulus of GFRP was equal
to the static elastic modulus. This is explained by that the resin in GFRP was not
damaged at cryogenic temperature, because the coefficient of thermal expansion
of glass fiber was almost equal to that of pure vinyl ester resin.
3.5 S-N curves of CFRP and GFRP
The S-N curves of CFRP and GFRP at room and cryogenic temperature are
shown in Fig. 10 and 11. CFRP had long durability at cryogenic temperature in
high cycle fatigue compared with that at room temperature. The fatigue lives of
GFRP at cryogenic temperature were also longer than those at room temperature.
700
650
600
550
500
450
CT
RT
400
100 101 102 103 104 105 106 107
Number of cycles to failure Nf
Figure 10:
S-N diagram of CFRP at

RT and CT.
Maximum stress max [MPa]
Maximum stress max [MPa]
The knee-point appeared in the S-N curve of the GFRP at cryogenic temperature.
The maximum stress at the knee point in S-N curve was almost coincident to the
stress at knee point under static load at cryogenic temperature. It should be said
that the stress at static knee-point is a threshold in cyclic fatigue loading which
determines the damage progression mode in fatigue of GFRP.
The elastic modulus decay diagrams of CFRP was shown in Fig. 12, where
=90% at room and cryogenic temperature. The elastic modulus of CFRP at
cryogenic temperature suddenly decreased in the final stage of fatigue, while
such change was not observed at room temperature. Same result was observed in
the paper [15]. At the cyclic number to failure in the condition of room
temperature, the sharp decreasing in elastic modulus occurred at cryogenic
temperature.
600
500
400
300
200
100
CT
RT
0
100 101 102 103 104 105 106 107
Number of cycles to failure Nf
Figure 11:
S-N diagram of GFRP at

RT and CT.
Figure 13 shows the elastic modulus decay diagrams (=90%) of GFRP at

room and cryogenic temperature. At cryogenic temperature, the elastic modulus
significantly decreased at first cycle of fatigue, while that was not decreased at
room temperature. This should be explained by that, at cryogenic temperature,
fatigue damage was accelerated in early stage of fatigue because the specimen
was subjected to the load over the stress at knee point, even while the damage
was not produced by local thermal stress of thermal fatigue test as shown in
section 3.4.
The fracture surface of CFRP after fatigue loading test (=90%) at room and
cryogenic temperature is shown in Fig. 14. At room temperature, fracture of the
current CFRP was caused with the meta delamination appeared before the failure
of fatigue at room temperature in CFRP. On the fracture surface, at cryogenic
temperature, the epoxy resin was remained on the carbon fibers, while debonding
was found around carbon fiber on the fracture surface of CFRP at room
temperature. It should be said that, at cryogenic temperature, the carbon fiber
was failed with accompanying critical crack propagation of the matrix due to the
brittle of resin.
120
Normalized modulus EN/E0 [%]
Normalized modulus EN/E0 [%]
100
80
60
40
RT
CT
20
0
100
101
102
103
104
105
Number of cycles N
Figure 12: Elastic modulus decay of

CFRP at RT and CT.
187
120
100
80
60
40
RT
CT
20
0
100
101
102
Number of cycles N
103
Figure 13: Elastic modulus decay of

GFRP at RT and CT
RT
CT
Figure 14: Fracture surface of CFRP after fatigue test (=90%).
Conclusion
(1) The fatigue lives of CFRP and GFRP at cryogenic temperature were higher
than those at room temperature.
(2) At cryogenic temperature, the magnitude of thermal strain of pure epoxy
resin was higher than the failure strain of the pure epoxy resin.
(3) At cryogenic temperature, the epoxy resin adhered carbon fiber remained on
the fracture surface of CFRP applied cyclic loading (=90%), while debonding
was found around carbon fiber on the fracture surface of CFRP at room
temperature.
(4) At cryogenic temperature, the elastic modulus of CFRP decreased in the final
stage of fatigue, while that of GFRP decreases after the first cycle of fatigue.
(5) At cryogenic temperature, the stress at static knee-point of GFRP was the
threshold in cyclic loading, which determined the damage progression mode in
fatigue.
Acknowledgement
This study was supported by the Ministry of Education, Culture, Sports, Science
and Technology for RCAST (Doshisha University).
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
H. Fukuda and G. Ben, Introduction to Mechanics of Composite, KOKIN

Publishers, Tokyo, 1999
T. Ishikawa, H. Kumazawa, Y. Morino, Y. Hayashi, ACCM-1, 1998, 4371-4
F. Tomioka, H. Wada, JSME-A, 1992, 58-550
A. Sakurai, Reinforced Plastic, Vol. 1, 10-17
B. Fiedler, M. Hojo, S. Ochiai, K. Schulte, M. Ochi, Finite-element
modelling of initial matrix failure in CFRP under static transverse tensile
load, Composites Science and Technology, 61, 2001, 95-105
Satoshi Kobayashi, Kazuhiro Terada, Nobuo Takeda, Evaluation of longterm durability in high temperature resistant CFRP laminates under thermal
fatigue loading, Composites: Part B, 34, 2003, 753-759
K. Yamada, K. Okubo, T. Fujii, 9th FRC2002, 2002, 201-206
C. Henaff-Gardin, M.C. Lafarie-Frenot, Specificity of matrix cracking
development in CFRP laminates under mechanical or thermal loadings,
International Journal of Fatigue, 24, 2002, 171-177
D. Hull and T. W. Clyne, An Introduction to Composite Materials, Second
Edition, 1996
N. K. Naik, V. K. Ganesh, Thermo-mechanical behaviour of plain weave
fabric composites: Experimental investigations, Journal of Materials
Science, 32, 1997, 267-277
K. F. Rogers, D. M. Kingston-Lee, L. N. Phillips, B. Yates, M. Chandra,
S. F. H. Parker, The thermal expansion of carbon-fibre reinforced plastics,
Journal of Materials Science, 16, 1981, 2803-2818
J. M. Gaitonde, M. V. Lowson, Low-temperature Thermal Expansion of
PEEK, HTA and Some of Their Composites Reinforced with Carbon
Fibres, Composites Science and Technology, 40, 1991, 69-85
M. de Freitas, R. de Carvalho, Residual strength of a damaged laminated
CFRP under compressive fatigue stresses, Composites Science and
Technology, 66, 2006, 373-378
M.C. Lafarie-Frenot, N.Q. Ho, Influence of free edge intralaminar stresses
on damage process in CFRP laminates under thermal cycling conditions,
Composites Science and Technology, 2005
Surya D. Pandita, Ignaas Verpoest, Tension-tension fatigue behaviour of
knitted fabric composites, Composite Structures, 64, 2004, 199-209
189
Mechanical properties of loosing

natural fiber reinforced polypropylene
K. Mizuta1, Y. Ichihara1, T. Matsuoka2, T. Hirayama2 & H. Fujita3
1
Graduate Student, Doshisha University, Japan

3
Hyogo Prefectural Institute of Technology, Japan
2
Abstract
As interest in studying environmental issues has grown, the mechanical
properties of eco-friendly materials have been studied. In our research, a card
machine that enables raw materials to loose and to mix uniformly was used.
Fibrous polypropylene and cotton or ramie fiber were used as raw materials
since many textiles and apparel are made of such fibers. From a viewpoint of
reuse, the loosing technique of the card machine can be effective. We designed
anisotropic composites by using the card machine followed by the hot press
process. The tensile, flexural and compression properties were investigated.
Ramie fiber reinforced polypropylene (RP) was stronger than cotton fiber
reinforced polypropylene (CP); however, breaking elongation of composites of
CP is superior to that of RP because of the character of natural fiber. The density
of composites applied to the loosing technique also affected the strength of
composites. It was clear that density is an important parameter to describe the
performance of a green composite. Moreover, it was found that the initial
fracture in flexural behavior is affected by the compression strength of
composites using the loosing technique.
Keywords: cotton fiber, Ramie fiber, loosing technique, tensile property, flexural
property, compression property, density.
Introduction
The increasing concerns for environmental issues and the growing eco-friendly
society direct universities and makers to development of Green composites
[1-4]. Green composites should be friendly for both environment and human
doi:10.2495/HPSM06020

body. Materials that can cause pollution and environmental load are not real
green composites. Properties of various green composites have been studied;
however, there are few reports about relationship between density and
mechanical properties of composites. In our research, composites were made by
using the loosing technique that enables raw materials to loose, to disperse and to
mix uniformly. Natural fibers vary in their geometrical and mechanical
properties. Therefore, it is important to grasp the interrelation between density
and mechanical properties of composites. Tensile, flexural, and compression
tests on cotton and ramie fiber reinforced polypropylene were performed to
investigate these correlations.
Experiment
2.1 Material
Polypropylene (PP), a matrix resin, is a fibrous resin with a melting point of
165. Cotton and ramie fibers were used as reinforcement. Length of cotton
fiber was 30 mm, and ramie fiber was 80 mm.
2.2 Production of composite
2.2.1 Loosing method
Loosing of fibrous PP resin and natural fiber was performed by a card machine.
In mixing of the fibrous resin and the natural fiber, it is important to distribute
the fiber uniformly in resin. Therefore, a loosing fibrous resin sheet and a loosing
fiber sheet were separately made by the card machine before mixing fibrous rein
and fiber.
2.2.2 Molding method
Specimens were manufactured by a vacuum compressed molding method as
shown in Fig.1. In this research, molding conditions were selected as shown in
Table 1.
Pressure
Heater
Puressre plate
Vacuum bag
Spacer
Sealant
Composites
Heater
Pressure
Figure 1:
Vacuum pump
Vacuum compressed molding method.
Table 1:
Molding condition.
Weight fraction of fiber [%]
30
50
Molding pressure [MPa]
10
15
Molding temperature []
Molding time [min]
191
210
10
2.3 Experimental method

2.3.1 Static tensile test
The test pieces were cut to a rectangle (201302 mm) from a plate, and gage
length was 80 mm. The static tensile tests were carried out at the test speed 1.0
mm/min under a room temperature by an INSTRON testing machine (Type
4467).
2.3.2 Flexural test
Dimensions of specimen were 15502 mm. A three points flexural test was
carried out at the test speed 1.0 mm/min under a room temperature by an
INSTRON testing machine (Type 4467). Span length was 30 mm.
2.3.3 Compression test
Dimension of specimens were 3255 mm, and gage length was 17 mm. The
compression tests were carried out at the test speed 1.0 mm/min under a room
temperature by a testing machine (Ez Test-500, SHIMADZU. Corp.). Specimen
holding assembly was used to avoid a buckling.
3.1 Tensile test

3.1.1 Effect of loosing technique
Fig.2 shows typical S-S curves of cotton fiber reinforced PP (CP) and ramie fiber
reinforced PP (RP) in loosing direction. The number following CP or RP means
fiber contents. RP is stronger than CP in any fiber contents, while deformation
ability of CP is superior to that of RP by more than 50%. It was caused by
properties of natural fiber; cotton fiber has many crimps and twists complexly
each other, while ramie fiber is relatively high strength and modulus. Fig.3
shows the effect of the loosing technique, where represents the angle between
loosing direction and tensile direction. The loosing technique could order fiber
orientation in both CP and RP, and anisotropic composites could be obtained.
3.1.2 Effect of fiber length on loosing process
Strong orientation couldnt be seen in short ramie (SRP: fiber length=30 mm) in
Fig.3. This result indicates the difference of the effect of fiber length on loosing

degree. It was clear that to order orientation of short ramie fibers was more
difficult than long fibers by measuring fiber orientation angle from flatwise
direction of specimen. It was one of the reasons that made SRP weak compared
to RP.
80
RP50
SR50
RP30
SRP30
70
Stress [MPa]
60
50
40
CP50
CP30
30
PP
20
10
0
0.0
1.0
Figure 2:
2.0
Strain [%]
3.0
4.0
S-S curves of composites.
80
70
SRP( = 0 )
Stress [MPa]
60
RP( = 0 )
50
SRP( = 90)
CP ( = 0 )
40
RP( = 90) CP( = 90)
30
PP
PP
20
10
0
0.0
1.0
2.0
3.0
4.0
Strain [%]
Figure 3:
Anisotropy of loosing natural fiber reinforced polypropylene (Fiber

contents: 30 wt.%).
3.2 Flexural test

Fig.4 shows flexural behavior of composites. Table 2 shows flexural strength and
flexural modulus of composites. Flexural modulus was improved in all materials,
while, flexural strength was not increased well on the whole. However, density is
an important parameter to estimate the loosing natural fiber reinforced
193
polypropylene. The relationship between density and strength properties is

discussed in the next section.
Flexural strength [MPa]
80
R50
70
R30 PP
60
C50
50
40
30
C30
20
10
0
0
Figure 4:
Table 2:
Displace me nt [mm]
Flexural behavior of NFRP.
Flexural strength and modulus of NFRP.
Flexural strength [MPa]

Flexural modulus [GPa]
PP
C30
C50
R30
R50
65
1.9
55
2.2
64
2.7
69
4.4
66
3.8
3.3 Effect of density on mechanical properties

3.3.1 Correlation between density and flexural strength
Fig.5 shows the effect of density on flexural strength of CP and RP. Composites
using loosing technique have low density because the loosing sheet that consists
of fibrous PP and natural fiber contains a lot of voids. However, there is a
specific positive correlation between them. Strength of composites, which had
densities from 0.75 to 0.85 g/cm3, was almost same to that of PP.
That means specific strength increased by around 20% since the density of PP
is 0.91 g/cm3. It was also suggested that high density composites yield high
strength. Therefore, high density composites were manufactured applying higher
pressure. In this paper, we classified into low density materials (LD) and high
density materials (HD). As a result, composites with around 1.0 g/cm3 of density
were obtained. Flexural strength of high density composites HD C30 and HD
R50 could be improved except for HD R30 as shown in Fig.5.
3.3.2 Correlation between density and tensile strength
Static tensile tests were also carried out to investigate tensile behavior of high
density composites. As a result, tensile strength of high density composites
decreased (Fig.6). In this decrease effect, the higher the natural fiber content was,
the lower the tensile strength of the composites was. This indicates the effect of
molding pressure on natural fiber. In the high pressure molding process, massive

damage of natural fibers could be the cause of this result. Observation of natural
fibers that were exposed to fracture surfaces is shown in the next section.
The relationship between some parameters such as strength, crystallization
degree and density of PP resin is nearly linear [5]. It is thought that the
difference of crystallization degree between low density composites and high
density composites is also one of the reasons for these results.
Flexure strength [MPa]
100
90
80
70
LD C30
LD C50
LD R30
LD R50
HD C30
HD C50
HD R30
HD R50
PP
60
50
40
30
0.60
0.70
0.80
0.90
1.00
1.10
Density [g/cm ]
Figure 5:
Correlation between density and flexural strength.
Tensile streng th [M Pa ]
70
60
50
LD C30
40
LD R30
LD C50
30
LD R50
HD C30
20
HD C50
10
HD R30
HD R50
0
0.60 0.70 0.80 0.90 1.00 1.10 1.20
3
Density [g/cm ]
Figure 6:
Correlation between density and tensile strength.
195
3.4 Difference of fracture surface between low & high density material
Fig.7 shows SEM photograph of tensile fracture surface of low density and high
density CP, respectively. There are many pull out single fibers in both, however
these forms are different. A notable point is the difference among their lengths.
In the high density composite, fibers were evenly broken shortly. Therefore, it
showed relatively flat fracture surface. It was also observed in the fracture
surface of CP.
The form of the tensile fracture surface is determined by the relative degree
and friction force between fiber and resin. If the ratio of tensile strength to
interfacial shear strength is relatively small, then the fracture surface is flat [6].
In this research, high pressure was applied to manufacture high density
composites. Fatal damage caused by high molding pressure decreased the
potential of reinforcement by the natural fiber. Therefore, it is thought the
massive damage on the crossing point of fibers made high density composites
low strength.
(a) Low density CP

Figure 7:
142m
(b) High density CP
142m
Fracture surface of CP (Fiber contents: 50wt.%).
3.5 Compression test

Strengthening was not found in flexural strength in spite of the improvement of
tensile strength. Therefore it is thought that compression property weakens the
flexural strength. To investigate the cause of this degradation, compression test
was carried out. Fig.8 shows compression strength of composites. As a result,
compression strength decreased by 30-50% compared to polypropylene,
however, high density materials were superior to low density materials only in
compression strength. It was identified compression failure causes in flexural
behavior. More detail of relationship between compression strength and flexural
behavior is discussed in the next section.
3.6 Effect of compression strength on flexural behavior
According to Bazhenov [7], flexural stress-displacement curve can be divided
three parts as Fig.9. The transition point from region I to region II is called Knee
Compression strength [MPa]
point. At Knee point, composite reaches compression yielding point, and then
relationship between flexural stress and displacement starts non liner behavior.
Fig.9 shows flexural stress at Knee point in flexural behavior.
50
45
40
35
30
25
20
15
10
5
0
Low de nsity
High density
PP
C30
C50
R30
R50
Material
Flexural stress
Figure 8:
Effect of density on compression strength.
. Elastic
. Compression yielding
. Tensile fracture
Knee point
X-head displacement
Figure 9:
Schematic illustrating three stages in flexural deformation.
Compared Fig.8 with Fig.10, we confirmed correlationship between flexural

stress at Knee point and compression strength. Compression stress was nearly to
the flexural stress at Knee point about both CP and RP. Furthermore, high
density materials tend to be superior to low density materials in Fig.8. Reflecting
such a tendency, Knee point stress of high density materials is superior to that of
low density materials (Fig.10). Therefore, it is necessary to keep compression
strength as much as possible to improve flexural strength of loosing natural fiber
reinforced polypropylene.
Stress at Knee point [MPa]
50
197
Low density
45
High density
40
35
30
25
20
15
10
5
0
PP
C30
C50
R30
R50
Material
Figure 10:
Stress at knee point in flexural test.
Conclusion
Cotton fiber and ramie fiber reinforced polypropylene were manufactured using
loosing technique. The experimental data showed that:
(1) Loosing technique enables to uniform fiber orientation in both cotton fiber
reinforced polypropylene and ramie fiber reinforced polypropylene.
(2) Long fibers (fiber length=80 mm) were easier than short fiber (fiber
length=30 mm) to orient natural fiber uniformly in ramie fiber reinforced
polypropylene.
(3) There was a positive correlationship between mechanical properties such as
tensile strength and flexural strength and density in low density materials.
(4) Damage of natural fiber by high pressure caused weakness of tensile
strength of composites.
(5) Initial fracture in flexural behavior of loosing natural fiber reinforced
polypropylene was affected by compression strength. Therefore,
improvement of compression strength is necessary to obtain higher
performance of loosing natural fiber reinforced polypropylene.
Acknowledgements
Authors would like to express special thanks to CHISSO Ltd. for providing
fibrous PP resin in this study. This study was supported by the Academic
Frontier Research Project on New Frontier of Biomedical Engineering
Research of ministry of education, culture, sports, science, and technology.
References
[1]
Kenji Ohnishi, Kazunori Umeoka, Hideyuki Ando, and Wenhai Liu.

Matsushita Electric Works, Ltd., Technical Report, pp58-63, 2003.

[2]
[3]
[4]
[5]
[6]
[7]
Teruo Kimura, J. Soc. Mat. Sci, Japan, Vol.50, No10, pp.1158-1163,

2001.
Hiroshi Yamamoto, Takashi Matsuoka, Kazuhiko Sakaguchi, and
Hiroyuki Fujita, Proceedings of Third International Workshop on Green
Composites, pp.95-100, 2005.
Anil N Netravali, Proceedings of Third International Workshop on Green
Composites, pp.11-15, 2005.
Kaneyuki Takagi, Heizo Sasaki, Polypropylene resin, Nikkan Kogyo,
Tokyo (1970).
D. Hall, An Introduction to Composite Materials, Baifukan Ltd., Tokyo
(1984).
S.L. Bazhenov, Composites 26, pp.757-765, 1995.
Section 4
Material and mechanical
characterisation
201
Identification of strain-rate sensitivity

parameters of steel sheet by
genetic algorithm optimisation
G. Belingardi, G. Chiandussi & A. Ibba
Dipartimento di Meccanica, Politecnico di Torino, Italy
Abstract
The influence of the loading speed on mechanical response of structural
materials can be accounted by means of strain-rate sensitivity parameters. The
aim of the present work is to show a numerical technique based on an inverse
approach to determine strain-rate sensitivity parameters of steels for car body
constructions. This technique is based on the numerical simulation of a simple
test according to the ASTM D5420/96 standard by means of a finite element
explicit code. The test consists of a falling tup with a spherical head impacting
on a thin sheet. Some experimental tests are conducted at different speeds, from
quasi-static to impact loading conditions, on a specimen made of XE280P steel.
A series of simulations are performed, changing the strain-rate sensitivity
parameters in each run according to a genetic algorithm strategy. The strain-rate
parameters that lead to the best fit of the experimental load-displacement curve
with the numerical result are the assumed material characteristic parameters. The
Cowper-Symonds and Johnson-Cook strain rate models have been taken into
consideration.
Keywords: strain-rate sensitivity, optimisation, genetic algorithm, finite element
analysis.
Introduction
Since the second half of the past century, research in the automotive industry
focused on safety improvement. Increasing customer interest about passive
safety and the ever stricter regulations, both in the US and EU, pushed towards
more reliable vehicle structure design and a deeper insight into material
behaviour. The finite element method applied to crash simulations with explicit
doi:10.2495/HPSM06021

codes and the increasing computing capability showed the way to new
possibilities for prediction of passive safety performance of vehicles. But these
opportunities have to be accompanied by adequate information on material
mechanical behaviour and by proper material modelling: both are fundamental
for correct numerical simulation. As a consequence material characterisation and
model identification through experimental tests have gained a primary role.
One of the most critical uncertainties in crash simulations relates to the
influence of the loading speed on the material mechanical characteristics. This
influence can be accounted by means of the strain-rate sensitivity parameters that
are substantial for most materials. Large strain-rate sensitivity of low-carbon
steels, such as the deep-drawing steels of most automotive body, is very well
known.
Strain rate sensitivity characterisation is quite complex because of the variety
of tests that could be performed and the difficulty in data acquisition. Uniaxial
tensile tests can be performed only at low strain rate values with traditional
hydraulic testing machines, moreover it is not possible to perform tests at strictly
constant strain rates and often the strain rate is not uniform on the specimen.
Other type of tests and testing machines, as for example the split Hopkinson
pressure bar, have been developed to characterise strain rate sensitivity of
materials. The instrumented drop dart test has proven to give interesting
opportunities in pointing out the material sensitivity to the strain rate and has
been successfully used for this type of characterisation.
In the present work an optimisation technique, based on a genetic algorithm,
is applied in order to determine parameters for strain-rate sensitivity models for a
high strength steel to be adopted in car body constructions. It is based on the
numerical simulation, by means of a finite element explicit code model, of the
simple drop dart test performed at different strain rates and the interactive
changing of material parameters until an optimum fit of experimental data is
reached. The Johnson-Cook and the Cowper-Symonds strain rate sensitivity
models are analysed in the present work. The considered material is the steel
XE280P laminated in sheets of 1.5 mm thickness.
Experimental tests and numerical models
Quasi-static uniaxial tensile tests and a series of simple bending tests at different
strain rates were performed to characterise the XE280P steel sheets and to
identify parameters of models. Both types of test are necessary because they
involve different mechanisms of deformation and help the stability and
completeness of the identification.
The uniaxial tensile tests were performed by means of a general purpose
hydraulic testing machine (DARTEC HA100). The usual dog bone shaped
specimens were manufactured from a laminated sheet of 1.5 mm thickness. The
obtained true stress vs. true strain curve is fundamental for the implementation of
the material behaviour in FE codes and was used for a first direct
characterisation (fig. 1).
203
600
True stress [MPa]
500
400
300
Elastic modulus:
Yield stress:
Maximum stress:
Failure stress:
200
100
0
0,00
Figure 1:
0,05
0,10
180 GPa
280 MPa
517 MPa
396 MPa
0,15
0,20
True strain [-]
0,25
0,30
Experimental stress-strain curve from quasi-static uniaxial tensile

test.
35
v = 76,5 mm/s
30
Force [kN]
25
v = 6,26 m/s
20
15
v = 0,1 mm/s
10
5
0
0
Figure 2:
10
15
Stroke [mm]
20
25
Experimental force-displacement curves of bending tests performed

at different loading speeds.
Bending tests based on the ASTM D5420/96 standard were performed in

order to characterise the strain rate sensitivity parameters of the steel. The test
consists on a falling tup with a spherical head impacting a steel sheet. The
specimens are square shaped but are completely constrained on a circumference
of diameter 76 mm by means of a blank-holder plate. The tests are conducted at
different speeds, from quasi-static loading conditions by means of the hydraulic
testing machine, to impact loading conditions, by means of a drop-dart testing
machine characterised by settable impact speed up to 6.26 m/s. The falling mass
(including the dart) is 20 kg. Each test gives a load-stroke curve. Force values
are measured with a load cell placed at the dart head while displacement is
obtained by LVDT direct measurement for the quasi-static testing conditions and
by double integration of the acceleration in time for the impact testing
conditions. Direct speed measurements are used in impact test for confirmation.
Figure 2 shows some experimental curves. Modifications induced by the load
application speed are evident. Numerical simulations of the bending tests were

carried out by means of the code RADIOSS. Different meshes and element
formulations were tested, and both demonstrated to have little influence on the
final results. The default Belytschko formulation with global plasticity algorithm
of Radioss was chosen for shell elements. Two models were used: a simplified
model with a complete constrain on the clamping circumference and a more
complete model with rigid plates holding the square specimen (fig. 3). Both
models take advantage of the symmetry of the system.
Figure 3:
On the left: simplified model of the bending test; the external

circumference is completely constrained. On the right: exploded
view of the complete model with the specimen holding plates.
Beyond the yield stress, the static plastic behaviour of the material is
modelled, as usual, by the following Hollomon formulation:
s = A + B pl n
(1)
where pl is the plastic component of the total strain, while A, B and n are
material constants. For the dynamic behaviour two strain rate sensitivity models
have been considered. Both models modify the static stress-strain curve by
adding a multiplying factor depending on the strain rate and both models are
characterised by the presence of two parameters that have to be identified. The
Cowper-Symonds model has the following formulation:
1p
= s ( ) 1 +

D

(2)
where s() is the static stress and D and p are material constants. The JohnsonCook model includes a further factor to account for the temperature dependence
that is not considered in this work:

= s ( ) 1 + C ln
0

T
1
Tm
where C and 0 are the material constants.

(3)
205
Identification by genetic algorithm optimisation
The identification of the static and strain rate sensitivity parameters was
performed by means of an optimisation technique. A set of simulations were
executed, changing the values assigned to the material characterising parameters
in each run according to a genetic algorithm strategy. For each run a complete
generation of a defined number of models was analysed. The parameters that
lead to the best fit of the numerical result with respect to the experimental result
have been selected as the optimal material model parameters. In particular the
following objective function to be minimised was used:
N
F par1 ,..., parq =
(y
exp,i
y mod,i
)2
i =1
(4)
where parq are the parameter values to be identified, yexp,i and ymod,i are the
experimental and numerical responses respectively, and N is the number of curve
points. The objective function was evaluated on the stress-strain curve for the
uniaxial tensile test and on the force-displacement curve for the bending tests. In
both cases curve data up to the maximum of, respectively, the stress or the force
were considered.
The genetic algorithm is defined by the mechanisms of selection,
recombination and mutation. The selection is used to choose the parent models
and is performed randomly by means of probability functions derived from the
model fitness: the higher is the model fitness the higher is the probability to be
chosen as parent of the new generation. The model fitness is defined as the
difference between the value of the objective function for the worst model of the
generation and the value of the objective function for the model itself. The
selection is limited to the models that constitute the 50% of the cumulated
fitness, decreasingly ordered. The recombination is the main feature of the
genetic algorithm. It defines the son models by combining different parent
models:
x a = (1 ) x + x ; x b = x + 1 x
(5)
Two son models (xa, xb) are generated from two parent models (x, x). and
are random numbers generated by a normal probability function with null
mean value. Finally a mutation, limited to 20% of the admissibility domain, is

imposed and the generated models are computed by FE simulations. The best
fitness model selected among the previous generation models and the new
generated model constitute the present generation.
Parameters identification was also performed by a response surface and a
gradient method in order to evaluate, by comparison, the performance of the
genetic algorithm.

3.1 Static model parameters identification
Parameter identification of the static model of eqn. (1) by means of uniaxial
tensile test does not need FE simulations. The good fitting of the model obtained
by genetic algorithm is shown in fig. 4, together with the objective function level
curves on the A-n parameter plane. These curves highlight a large region of A
and n values with a nearly constant value of the objective function that means
possible instability in parameter identification. Multiple optimum couples of
values of these parameters are possible and for this reason a fixed A parameter
defined by a direct identification have been used.
600
0,44
500
B=491,678 MPa
Experimental curve
300
7-9
9-11
11-13
13-15
15-17
17-19
19-21
21-23
23-25
25-27
0,36
0,32
Optimised model curve
0,28
200
n [-]
Stress [MPa]
0,40
400
0,24
0,20
100
0,05
Figure 4:
0,10
True plastic strain [-]
0,15
130
145
160
175
190
205
220
235
250
265
280
295
0,16
0
0,00
0,20
0,12
A [MPa]
Identified model by uniaxial tensile test and level curves of the

objective function on the A-n parameters plane.
The bending test performed at low speed can also be used for the static
parameters identification. In this case different and not uniform mechanisms of
deformation are involved and FE simulations are needed to have the numerical
force-displacement curve to be compared with the experimental one.
600
Experimental curve
Model gradient method
Model genetic algorith
25
20
Stress [MPa]
Force [kN]
30
15
10
5
0
0
Figure 5:
10
Stroke [mm]
15
20
500
400
300
Experimental curve
200
100
0
0,00
0,05
0,10
0,15
Plastic strain [-]
0,20
Simulated static bending test and stress-strain models obtained by

optimisation with bending test and simplified FE model.
The parameters identified by means of the genetic algorithm and of the

gradient method based on the simplified structural model of the bending test
(fig. 3(a)) bring to the force-displacement and stress-strain curves shown in fig.
5. Both optimisation method work well to fit the experimental curve of the
bending test, but the identified parameters are rather different from each other
and the corresponding stress-strain characteristics are rather far from the
experimental one. This confirms the possible criticality of the identification
207
method that could lead to the same well simulated curves with different
parameter values. Moreover it shows that the FE model of the bending test could
be not sufficiently accurate, both for what concern the mesh refinement and for
what concern the through thickness integration.
600
Experimental data
25
20
Stress [MPa]
Force [kN]
30
15
10
5
0
0
Figure 6:
10
Stroke [mm]
15
500
400
300
Experimental data
200
100
0
0,00
20
0,05
0,10
0,15
Plastic strain [-]
0,20
Simulated static bending test and stress-strain models obtained by

optimisation with bending test and complete FE model.
The use of the complete structural model (fig. 3(b)) for the identification
leads to the results shown in fig. 6. A more accurate solution is obtained: the
material model fits better the experimental stress-strain curve of the tensile test,
while keeping the same quality of fitness for the force-displacement curve of
bending test. All the identified parameters are summarised in table 1.
Table 1:
A [MPa]
B [MPa]
N [-]
F [MPa, kN]
Optimal values for the static model.
Uniaxial tensile
test
Genetic Gradient
241.7
213.5
482.5
491.7
0.3298
0.2857
8.328
8.186
Bending test
Simplified model
Genetic Gradient
238.7
305.2
499.0
521.1
0.5913
0.9116
0.361
0.179
Bending test
Complete model
Genetic Gradient
282.4
273.1
532.1
528.4
0.5323
0.4967
0.173
0.185
3.2 Strain rate sensitivity model identification

For the identification of the strain rate sensitivity parameters drop tower tests
were performed with two different falling heights of the dart: 1 m and 2 m. In
order to achieve a more suitable solution the static parameters identified by
means of the uniaxial tensile test were used and only the strain rate parameters
were searched with the described optimisation procedure. The objective function
to be minimised take into account the force-displacement curves of both test
conditions:
N1
F(p1 , p 2 ) =
1
2
(F
exp1,i
Fmod1,i
i =1
N1
N2
)2
+
1
2
(F
exp 2,i
Fmod 2,i
i =1
N2
)2
(6)
6,0
5,5
5,0
4,5
4,0
3,5
3,0
2,5
2,0
1,5
1,0
0,5
0,0
Objective F [kN]
Objective F [kN]
3.2.1 Cowper-Symonds model

The identification of Cowper-Symonds parameters by genetic algorithm was
performed by means of 16 models for each generation. For the complete model,
the evolution of objective function values for each generation in the best and
worst cases are shown in fig. 7(a). In figure 7(b) the values obtained by the
gradient method for each iteration are shown for effectiveness comparison. The
genetic algorithm optimisation shows a more rapid convergence, even if a greater
amount of simulations were used. The identified parameters and the final
objective function values are summarised in table 2. The force-displacement
curves derived from the simulations with the optimised complete model are
shown and compared to the experimental ones in fig. 8.
DYNAMIC
Genetic
algorithm
Model: Cowper-Symonds
Minimum of generation
Maximum of generation
1
Generation
10
11
12
13
10
9
8
7
6
5
4
3
2
1
0
14
DYNAMIC
Gradient
method
Iterations
Figure 7:
The objective function behaviour with the genetic algorithm and

the gradient method.
Table 2:
Optimal values of the Cowper-Symonds strain rate sensitivity

model.
D [s-1]
p [-]
F [kN]
Complete model
Genetic
Gradient
4987
3013
1.619
1.503
1.817
1.822
Simplified model
Genetic
Gradient
2399
3006
1.329
1.416
1.187
1.209
3.2.2 Johnson-Cook model

The identified parameters of the Johnson-Cook model by means of the genetic
algorithm and the gradient method, with both the simplified and complete
models are summarised in table 3. Five models were computed for each
generation of the genetic algorithm optimisation. Force-displacement curves
obtained by simulations with the complete model optimised by the genetic
algorithm are shown and compared to the experimental ones in fig. 9.
Conclusions
A genetic algorithm optimisation procedure was applied to the identification of

strain rate sensitivity models and proved to be effective. Compared to an
optimisation procedure based on the gradient method it reached a better quality
result in less iterations, even if the number of simulations to be run could be
high.
209
35
DYNAMIC
30
Genetic
algorithm
Force [kN]
25
20
15
Experimental curve - h=1 m
Simulation curve - h=1 m
10
5
0
0
Time [ms]
Figure 8:
Comparison between experimental force-time curves of dynamic

bending tests and simulations with Cowper-Symonds model.
Table 3:
Optimal values for the Johnson-Cook strain rate sensitivity model.
0
Complete model
Genetic
Gradient
Simplified model
Genetic
Gradient
1.00
1.00
[s-1]
0,01149
1,375
C [-]
F [kN]
0,01223
1,372
0,01028
1,329
0,01024
1,330
35
DYNAMIC
30
Genetic
algorithm
Force [kN]
25
Model: Johnson-Cook
20
15
10
5
0
0
Time [ms]
Figure 9:
Comparison between experimental force-time curves of dynamic

bending tests and simulations with Johnson-Cook model.

Simple drop tower bending tests were performed at different loading speeds
on specimen made of XE280P steel. Explicit simulations by means of FE models
were run in order to identify the material strain rate characteristic parameters
according to the Cowper-Symonds and the Johnson-Cook laws. The
identification of static model parameters was also performed with the same
algorithm and the same bending test at quasi-static loading conditions, but large
identification uncertainty and drawbacks were encountered. A classical uniaxial
tensile test was chosen for a more reliable and proper identification of static
parameters and for the validation of the bending test modelling.
Acknowledgement
The authors kindly acknowledge the European Commission that, by funding the
ADVANCE project through contract GRD1-200-25914, made this research
activity possible.
References
[1]
[2]
[3]
[4]
[5]
[6]
Jones, N., Structural Impact, Cambridge University Press, 1989.

Symonds, P.S., Survey of methods of analysis for plastic deformation of
structures under dynamic loading, Report No. BU/NSRDC, Brown
University, 1967.
Jones, N., Some comments on the modelling of material properties for
dynamic structural plasticity, Proc. of Int. Conf. Mechanical Properties of
Materials at High Rates of Strain, Oxford, pp. 435-445, 1989.
Johnson, G.R. & Cook, W.H., Fracture Characteristics of three metals
subjected to various strains, strain rates, temperatures and pressures, Eng.
Frac. Mech., Vol. 21, No. 1, pp. 31-48, 1985.
Avalle, M., Belingardi, G., Vadori, R. & Masciocco, G., Characterization
of the strain rate sensitivity in the dynamic bending behavior of mild steel
plates, Proc. EUROMAT 2000, Tours, pp. 505-510, 2000.
Avalle, M., Belingardi, G. & Gamarino, M., An inverse approach for the
identification of strain-rate sensitivity parameters of sheet steels, Proc.
SUSI 2004 (Structure under Shock and Impact), Creta (Grecia), pp.13-22,
27-29 March 2004.
211
Mechanical characterisation of a
viscous-elastic plastic material, sensitive to
hydrostatic pressure and temperature
V. D. Le1, M. Caliez1, M. Gratton1, A. Frachon1 & D. Picart2
1
Laboratoire de Mcanique et Rhologie, EA 2640,

Ecole dIngnieurs du Val de Loire,
Universit Franois Rabelais de Tours, Blois cedex, France
2
C.E.A. Le Ripault, Monts, France
Abstract
This paper deals with the characterization of the static mechanical behaviour of
an energetic material. Due to the constituents (crystals and a polymeric binder),
the behaviour is influenced by the pressure, the temperature and the strain rate.
The temperature, considered varying slowly, is a parameter and the
computational problems are uncoupled. Therefore, a complete experimental
protocol and a model have been developed. Inspired from the Visco-Scram
model, the behaviour is described using a general Maxwell model in which all
the branches are affected by an isotropic damage. The first branch takes into
account elastic-plastic behaviour. The yield stress is given by a parabolic
criterion, characterized using compressive, tensile and tri-axial tests. The
hardening is isotropic and the plastic flow rule is nonassociated. The other
branches are viscoelastic. A genetic algorithm is used to optimise the viscoelastic
parameters, previously obtained using DMA measurements. Comparisons
between the model and experiments are proposed for different temperatures,
strain rates and pressures. At last, a user material subroutine has been developed
in Abaqus Standard and finite element computations of the Brazilian test are
compared to the experimental response.
Keywords: energetic material, parabolic plastic criterion, genetic algorithm,
DMA, viscoelasticity, Isotropic damage, Brazilian tests.
doi:10.2495/HPSM06022
Introduction
The material is made of organic and energetic crystals mixed with a few
percentage of a polymeric binder. After an isostatic compaction forming process,
the material has a small porosity of a few percent. Samples can be machined in
several geometric shapes, which are more than ten times the length of the
material heterogeneity.
In order to survey the possible aging of this material, an accurate
determination of the mechanical properties has to be done. Unfortunately, this
material being available in small amounts, the characterization must be made
using a reduced number of standard tests. An unusual experimental procedure is
proposed in this paper for this kind of material. When monotonic loading paths
were used to determine for example the influence of the strain rate, each sample
is submitted to complex loading paths including relaxation, recovery and cyclic
conditions.
The observation of loading-unloading diagram on figure 1 shows some of
the main features of the material at room temperature, and entails specific
arrangements for the mechanical tests. 1) Hydrostatic pressure sensitivity: to
consider it, an initial hydrostatic loading path (0 MPa, 5 MPa and 10 MPa) is
made before the run of a uniaxial compression load. 2) Viscosity: different strain
rates (5.10-6 s-1 to 10-3 s-1) have been used to observe such effect. The parameters
of the viscoelastic part of the behaviour have been determined using a DMA
apparatus (Dynamic Mechanical Analysis). 3) The plastic strains are determined
using relaxation and recovery delays. 4) Initial elastic behaviour: standard tests
made in various material directions show an initial isotropic behaviour.
5) Damaging: systematic cyclic loading-unloading programs have been
performed. 6) Dispersion: to ensure a minimum statistical validity, each loading
program is repeated five times.
The temperature is considered as a parameter in the model. The material
been temperature dependent, compressive and tensile tests are perform at four
different temperatures: 5C, 20C, 35C and 50C. The figure 2 shows the effect
of the temperature on monotonic compressive experiments.
The main models available in the literature ([13]) for this kind of material
have been developed for transient dynamic behaviour and are not adapted for a
quasi-static study. For example, the influence of the pressure is omitted, even as
the difference of behaviour observed in tension and compression loading paths.
The Maxwell model we have chosen (fig. 3) is close to the constitutive law
proposed in [2]. Several damageable viscoelastic branches and one damageable
elastoplastic branch are used. The main difference with the Bennett and coworkers model is the presence of the elastoplastic branch.
The determination of the plastic (resp. viscoelastic) behaviour is described
in the second (resp. third) part of this paper. In particular, a genetic algorithm has
been used to optimize the determination of the viscoelastic parameters. The
fourth part deals with the damage rule. It can be noted that the characterization of
the plasticity, the damage and the viscosity are uncoupled. The damage rule is
assumed to affect also the viscoelastic branches.
213
Lastly, some comparisons are presented between the model response and
the available experimental data.
Figure 1:
Cyclic tests for a given Figure 2:

strain rate: pressure effect
on the global responses of
the material (20C).
Compressive tests for a

given
strain
rate:
temperature effects.
Damageable
Ep
E1
Ej
En
Figure 3:
1
j
n
Rheologic diagram of the viscoelastic plastic damageable model.
The test procedure is described in [4]. The first stage (when this one exist)
of the loading program is a hydrostatic loading phase. Then, five or six uniaxial
loading-relaxation-unloading-recovery cycles are done. The test is driven by one
of the two longitudinal gages. Relaxation times have been defined to guarantee
an almost complete relaxation of viscous stresses. All the tensile tests and the
compressive tests have been done for several temperatures (5C, 20C, 35C and
50C).
Some compressive DMA experiments have been realized using samples of
50 mm long, and a cross section of 4x4 mm2. A small initial preload of 10 m
and a strain amplitude of 5 m are used to stay in the viscoelastic domain.
The range of frequencies going from 0.004 to 40 Hz, the strain rate ranges from
2.10-6 s-1 to 2.10-2 s-1.
Elastoplasticity
The macroscopic stress tensor is defined as the sum of the stress of each
branch:
ep
j =1
(1)
where the subscript ep (resp. j) denoted the elastoplastic branch (resp. the jth
viscoelastic branch). An additive decomposition is assumed for the strain
mechanical part (superscript m) and the strain thermal part (superscript th):
= m + th .
(2)
An additive decomposition of the mechanical strain of each branch is also
assumed between: an elastic part (superscript e) and a plastic (superscript p)
part (resp. viscous) (superscript p or v):
p
elasto-plastic branch,
(3)
m = eep + ep
= ej + vj visco-elastic branches.
(4)
2.1 Thermo-elasticity
The elastic part is supposed linear damageable and given by the following
equation:
E
ep = (1 d ) ( ) : eep ,
(5)
where ( ) is the elastic tensor of the virgin material and is defined as a

function of the temperature, and d a damage parameter (see below). To identify
the elastic mechanism, we need to isolate the elastoplastic behaviour (in
particular to be sure that the viscous stresses are relaxed). To this end, the ends
of the relaxation and recovery phases are used to determine the elastic modulus.
The initial Youngs modulus Eep is a function of the temperature and is defined
in the table 1.
A value of 0,3 is obtained for the Poissons ratio for all the temperatures,
using longitudinal and transversal strain measurements. This ratio is used for all
the viscoelastic branches. The thermal expansion coefficient is c = 50.10-6 K-1.
Table 1:
Temperature dependency of the Young modulus.
Temperatures
Initial Young modulus Eep (MPa)
5C
3400
20C
2900
35C
2700
50C
2450
2.2 Yield criterion

A review of the main criteria used to describe isotropic plasticity is presented in
[4]. A criterion has been developed at Cambridge University in view of soil
modeling and is famous today as the Cam-clay model [5]. Numerous
215
adaptations of this model were then developed for various applications [68]. As
the forming process of the material is an isostatic compaction up to a pressure of
200 MPa (which amplitude will never be reached in quasi-static applications),
the criteria is supposed open on the hydrostatic negative axis. Open threshold are
usually derived from Mohr-Coulomb, Mises-Schleicher [9], Drucker-Prager [10]
and more recently Hoek-Brown formulations [11]. Raghava et al. [12] applied
the Mises-Schlechers threshold to polymers. The evolution of this criterion is
described by two hardening variables, associated to tensile and compressive
response. Lastly, a unified model is proposed by Aubertin and Li [5] in order to
reproduce all the kinds of criterion (elliptic, parabolic, hyperbolic).
Due to a lack of data about the nature of the hardening mechanisms, an
isotropic hardening parameter, denoted k, is introduced in the model. Then, a
saturation of the hardening mechanism at the maximum stress is taken into
account in the model. A nonlinear plasticity criterion reproduces the evolution of
the yield stress (fig. 4). The following relation is used:
f ( Q , P , k ) = eq k = 0
where Q = (1 3 )
with
eq = Q 2 +
is the octahedric stress,
k2
X (k )
P , (6)
the deviatoric stress
and P the pressure. The set of yield curves is completely defined as soon as the
function X(k) and the hardening law are defined.
The following guidelines help for the determination of the function X(k).
First, it is assumed that the yield curves do not cross themselves in the P-Q
plane, each one being embedded in those of higher levels, all of them being
embedded in the extreme curve. This is a necessary - but not sufficient condition for the phenomenon to be governed by a unique state variable which is
the isotropic strain-hardening parameter. Elementary algebra shows that the
following relation satisfies the previous assumption:
X (k ) = X 0 + ( X m X 0 )
k k0
km k0
(7)
where the parameter X0 = 1,5 MPa, positioning the summits of the initial
parabola, is supposed not being temperature depending, and the three parameters
Xm, k0 and km are temperature depending. The Xm and km parameters are
determined using the ultimate yield stress curve relating the maximum stress
states in the P-Q plane. The values are given table 2.
Table 2:
Temperatures
Xm (MPa)
k0 (MPa)
km(MPa)
c1
c2
Temperature dependency of the hardening parameters.

5C
1,635
0,80
3,88
700
150000
20C
1,620
0,55
3,5
500
100000
35C
1,610
0,45
3,26
500
100000
50C
1,592
0,35
2,82
500
90000

The hardening parameter k has to be related to an effective plastic strain
variable, denoted p. In order to obtain a unique curve k(p) for all the available
experimental data, p is defined as the cumulated deviatoric plastic strain (fig. 5).
Figure 4:
Figure 5:
Initial and saturated yield criterions for the different temperatures.
Hardening parameter k versus the effective plastic strain p for the

different temperatures.
For the hardening law, the following hyperbolic relation is used to interpolate the
data:
1
k ( p ) = k0 + ( km k0 ) 1
2
1 + c1 p + c2 p
where c1 and c2 are two parameters temperature depending (table 2).
(8)
217
2.3 Flow rule

The flow direction is determined using the ratio between volumic and deviatoric
effective plastic strain rates. This ratio, usually called dilatancy and denoted ,
is given by the following relation:
eppV
eppD
p
with eppV = Tr ( ep
)
and
pD
pD
: ep
eppD = 3 ep
, (9)
pD
ep
being the deviatoric plastic strain rates.
The flow rule expression is then:

p
ep
ep
=
+
3
3Q
3
I .
,
(1 + 2 )
(10)
being the plastic multiplier and the plastic flow direction being normalized.
The dilatancy is characterized from the experimental results as a constant
( = 0,15). As a result, a nonassociated plastic law is justified.
Viscoelasticity
3.1 Dynamic Mechanical Analysis

DMA experiments are used to make a first identification of the linear
viscoelastic parameters. The stress response to a unit sinusoidal strain solicitation
for this kind of model can be break up in an in-phase part (related to the storage
modulus Estor) and an out of phase part (related to the loss modulus Eloss). The
analytical response of a generalized Maxwell model is known and depends of the
distributions of the Young modulus (Ej) values and of the viscosity parameters
(j) values. In order to limit the number of parameters (here n = 10 branches) for
a more accurate determination, the following relations are proposed from the
DMA results [4]:
j = 10
( A1
j 1
n 1
+ A2
((
and E j = A3 + A4 e
A5
j 1
n 1
1 .
(11)
Then, the number of unknowns decreases to five (A1 to A5). These relations allow
reproducing at the same time the storage and the loss modulus.
3.2 Genetic algorithm
The previous set of parameters is used here to determine the bounds of each
parameter. We are in the case of a combinatory optimisation problem where a
large number of solutions could be suitable. We have chosen to perform an
inverse identification of the viscoelasticity parameters directly from the
experimental tests. Classical optimisation methods, like conjugated gradient,

have been dismissed because of the possible large number of solutions. Then, a
genetic algorithm [13] has been used in this study.
Genetic algorithms are based on the Darwinian principle of survival to the
fittest. An initial population of a given size is created from a random selection
of parameters values. Each parameter set represents individual chromosomes.
Each individual is assigned of a fitness based on how well each individual
chromosome allows it to perform in its environment. The algorithm produces
new generations by applying three evolution operators: selection, crossover and
mutation. For each generation, the fit individuals survive and the weak die.
Evolution operators create new individuals (children) from two selected parents,
and these children replace the weak individuals for the next generation.
Successive generations are created until very fit individuals are obtained. This
algorithm offers the advantage of exploring all the solutions space to find a
global optimum of an optimization problem. A sensitivity analysis of the
parameters is not required.
3.3 Objective function
The objective function is a direct measure of the quality of a solution. The goal is
here to minimize the gap between the experimental strain-stress curves and the
corresponding simulated curves. Due to different times of recording during the
experiments, we propose a fitness function that represents the spatial gap,
weighted by the segment length Li between two consecutive data, in the stressstrain space, i.e.:
nb exp pts
Obj =
i =1
( ie ( t ) is ( t ) ) . Li ,
(12)
nb exp pts
L
i =1
where superscripts s and e respectively denote simulations and experiments.

3.4 Results
The simulation uses the data corresponding at the end of load, the end of
relaxation, the end of discharge and the end of recovery. From the values of the
longitudinal strain and time in these points, the strain rate is rebuilt, constant by
piece. The strain increment is then given from the strain rate and the step of time
of the program.
At 20C, all the tests (compressive tests with different strain rates, tensile
tests and triaxial tests) are used to optimise the five parameters Ai. 60 generations
constituted of 500 individuals have been performed and leads to the optimized
values of the five Ai parameters at 20C. The restricted quantity of material does
not allow making all those tests for all the tested temperatures. So, for the 5, 35
and 50C temperature values, only the tensile and compressive tests, for only one
strain rate, are done. Therefore, we supposed that the temperature dependency of
the Young modulus of the viscoplastic branches is the same that the elastoplastic
branch one: E j ( ) = E ep ( ). E j (20 C ) / E ep (20 C ) .
219
Therefore, for the tests at 5C, 35C and 50C, only the parameters A1 and
A2 have to be identified with the genetic algorithm. The results are presented in
the table 3.
Table 3:
Temperature dependency of Ai parameters defining the j viscosity

parameters.
Temperature
5C
Tests com- Compressive,
puted from: one strain rate
A1
2,78
A2
2,63
A3
A4
A5
20C
All
3,98
1,64
160,31
16,15
1,21
35C
50C
Tensile/Compressive, Tensile/Compressive,
one strain rate
one strain rate
2,15
2,62
2,60
2,32
Isotropic damage
Assuming an isotropic damage, experimental data shows that the phenomenon

regularly increases with the highest positive principal strain (fig. 6). This
observation indicates that the most probable damaging mechanism is the result of
the development of internal micro-defects (cavities, cracks) with tension [14
18]. A damaging factor d is classically defined as:
d=
E0 E
E0
(13)
where Eo and E are the initial and current Youngs modulus. A constant
Poissons ratio is assumed here.
Figure 6:
Damage versus the maximum positive strain.
Experimental values of d immediately result from the measurements of E.

An hyperbolic relation, eqn. (14), is used to reproduce an average evolution of

the damaging factor, providing that its value is bounded to 1; d1, d2 and d3 are
three parameters. The subscript + means the positive part.
1
,
( I ) > + + d 2 .1d = d 1 . sup < max
N
N
1 + d 3 sup < max
I >+
N
=
1
3
I
to
N
time
time
(14)
I =1 to 3
The damage rule is reported in the fig. 6. One can see that the model
response is identified using the compression measurements. The hydrostatic data
are not taken into account because the pressure stops an eventually growth of the
microcavities. The introduction of a second damage mechanism, as proposed in
[19], to model the tensile damage is in study. The identified values are the same
for all the temperatures: d1 = 0,5, d2 = 0,75 and d3 = 75.
Model versus experiments
The constitutive law has been implemented in the finite element code
Abaqus/standard. The model is compared to experimental data with unloading
cycles to access to the plasticity level and the damage level. Those seem quite
well reproduced even if the transversal model response does not present enough
damage in compression (fig. 78). This observation can be associated to an
anisotropic damage (which has been neglected here).
Figure 7: Compressive test (3.10-6 s-1;

20C);
model
versus
experiment.
Figure 8: Compressive test (1.5 103 -1

s ; 20C); model versus
experiment.
The plasticity branch reproduces very well the pressure effect that causes
the difference between tensile and compressive responses (fig. 7, 9, 10). The rate
effect is also quite well reproduced on the two compressive tests even if the
unloading curves do not present the same nonlinearity (fig. 78). In the same
manner, the viscous effects in the triaxial test with 10 MPa of confinement
pressure (fig. 10) is underestimated. Certainly, these phenomenons is associated
to an internal friction in the material or a viscosity pressure dependency. The
pressure dependency is certainly the reason of the discontinuity of the Ai
221
parameters at 20C (table 3), temperature for which all the tests are computed
and in particular the triaxial ones.
Figure 9:
Figure 11:
Tensile test (3.10-5 s-1);

model versus experiment.
Figure 10: Triaxial test (3.10-5 s-1;

20C);
model
versus
experiment.
Global response of the Brazilian test; model versus experiment.
The temperature dependency of the model parameters gives a quite good

adequacy of the model versus the experimental results (fig. 9).
The implementation of the constitutive law in the finite element code
Abaqus has allowed one to compare simulations to more complex experimental
configurations as three-point bend tests and Brazilian experiments. The global
response of the compression diametrical test is given in fig. 11. One can see that
too much nonlinearity appears in the model response. This point is in study, the
three-point bend test giving providing first answers.
Conclusion
An experimental procedure has been carried out to characterize a complex

material behaviour. A multibranch viscoelastic plastic and damageable model
and the corresponding identification procedure have been developed. A genetic
algorithm optimisation has one allowed to find accurately some of the viscous
parameters. This model has been implemented in the finite elements software

Abaqus using a user subroutine UMAT. The comparisons between simulations
and experiments show a good agreement. Our future works are now devoted to
the improvement of the damage rule and of the failure threshold. Lastly, the
anisotropy observed during the experiments has to be introduced in the model.
The pressure dependency of the viscosity is in study.
Acknowledgement
The authors address a special thanks to J.L. Brigolle for its contribution to this
study, especially for the realization of the experiments.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
F.L. Addessio, J.N. Johnson, A constitutive model for the dynamic

response of brittle materials, J. Appl. Phys., Vol. 67, 1990, pp 3275-3286.
J.G. Bennett, K.S. Haberman, J.N. Johnson, B.W. Asay, B.F. Henson, A
constitutive model for the non-shock ignition and mechanical response on
high explosives, J. Mech. Phys. Solids, Vol. 46-12, 1998, pp. 2303-2322.
R.M. Hackett, J.G. Bennett, An implicit finite element material model for
energetic particulate composite materials, Int. J. Numer. Meth. Engng,
Vol. 49, 2000, pp. 1191-1209.
M. Gratton, V.D. Le, A. Frachon, M. Caliez, D. Picart, Mechanical
behaviour of a viscoelastic plastic granular material: Experimental
procedure and modelling, WSEAS trans. on comp., Is.1, Vol.5, Jan.2006,
pp.149-156.
M. Aubertin, L. Li, A porosity-dependant inelastic criterion for
engineering materials, Int. Jour. of Plast., Vol. 20, 2004, pp. 2179-2208.
J.H. Prevost, R. Popescu, Constitutive Relations for Soil Materials,
Electronic Journal of Geotechnical Engineering, First Issue, 1996.
O. Coussy, Mcanique des Milieux Poreux. Editions Technip, 1991.
P.Y. Hicher, J-F. Shao, Elastoplasticit des sols et des roches - Modles
de comportement des sols et des roches - 1, Herms Science Publications,
2002.
F. Schleicher, Z. Angew, Math. Mech., 6:199, 1926.
D.C. Drucker, A.M. Asce, R.E. Gibson, & D.J. Henkel, Soil Mechanics
and Work-Harding theories of plasticity, Transactions American Society
of Civil Engineers, Vol. 122, 1957.
X.D. Pan, J.A. Hudson, A simplified three dimensional Hoek-Brown yield
criterion, Rock Mechanics and Power Plants, M. Romana (ed.),
Rotterdam: Balkema, 1988, pp. 95-103.
R. Raghava, R.M. Caddekk, G.S.Y. Yeh, The macroscopic yield
behaviour of polymers, J. Material Science, Vol.8, 1973, pp. 225-232.
D.E. Goldberg, Genetic algorithms in search, Optimization and Machine
Learning, Addison Wesley Publishing Company, 1989.
J. Mazars, Mechanical damage and fracture of concrete structure, Proc.
I.C.F. 5, Cannes, France, 1981, pp. 1499-1506.
[15]
[16]
[17]
[18]
[19]
223
A. Dragon, D. Halm, Th. Dsoyer, Anisotropic damage in quasi-brittle

solids: modelling, computational issues and applications, Comput.
Methods Appl. Mech. Engrg., 183, 2000, pp. 331-352.
D. Halm, A. Dragon, An anisotropic model of damage and frictional
sliding for brittle materials, Eur. J. Mech., A/Solids, Vol.17, no.3,1998,
pp.439-460.
J. Mazars, A description of micro and macroscale damage of concrete
structures, Engineering Fracture Mechanics, Vol.25, no.5/6, 1986,
pp.729-737.
J. Mazars, F. Ragueneau, G. Pijaudier-Cabot, Continuum damage
modelling for concrete structures in dynamic situations, Continuum
Damage Mechanics of Materials and Structures, O. Allix and F. Hild
(Editors), Elsevier Science Ltd., 2002.
J. Lee, G.L. Fenves, Plastic-damage model for cyclic loading of concrete
structures, Journal of Engineering Mechanics, vol.124, no.8, pp.892-900,
1998.
225
Identification of the material properties of

composite beams: inverse method approach
E. Euler1, H. Sol1 & E. Dascotte2
1
Department of Mechanics of Materials and Construction,

Vrije Universiteit Brussel, Brussels, Belgium
2
Dynamic Design Solutions NV, Leuven, Belgium
Abstract
In load carrying applications one is mainly interested in the stiffness properties
of the structural component. Elastic and shear moduli supplemented with section
properties render the stiffness properties of the component. Moduli of isotropic
materials are well known and well documented. However, this is not the case for
composite materials.
The developed procedure belongs to the group of mixed numerical
experimental methods. The method makes use of modal data to determine the
effective orthotropic material properties of composite beams. Modal reference
data is experimentally obtained from the beam at hand. The other modal data set
is obtained from a finite element model of the same beam. The orthotropic
material properties, also called parameters, in the finite element model are then
modified in such a way that both sets of modal data match. If those two sets
match, the virtual model has in a global sense the same mass and stiffness
properties as the real model.
A program written in FEMtools is applied to five test cases. Results are
discussed.
Keywords: inverse methods, FEA, orthotropic material properties, composite
materials, modal data.
Introduction
Most engineers have considerable experience in the design of simple structural

components using isotropic materials. However, today more and more composite
materials are used for structural elements.
doi:10.2495/HPSM06023

In load carrying applications one is mainly interested in the stiffness
properties of the structural component. Elastic and shear moduli supplemented
with section properties render the stiffness properties of the component. Moduli
of isotropic materials are well known and well documented. This is not the case
for composite materials. The effective laminate properties depend on fiber
material, matrix material, ply orientation, laminate thickness, stacking sequence,
etc. Infinite combinations are possible, resulting in a huge amount of different
moduli. It is clear that not all possible scenarios are studied in literature.
If all the details of the composite material are known, the engineer can
calculate effective laminate properties and use them in theories like for example
the first-order shear deformation theory for thin-walled laminated beams. If the
ply lay-up results in coupling phenomena, the engineer has to use advanced
software to simulate the behaviour of the structure. All this requires substantial
effort.
In some cases, not all the details of the composite material are known. In this
case, elastic properties can be determined by experiment. Drawback, these
experiments are destructive in nature. Another possibility is to obtain stiffness
properties of the whole structure by conducting an experiment. Drawback here,
influence of boundary conditions. All this requires substantial effort.
This paper presents a method to determine the global effective orthotropic
material properties of composite beams by measuring a certain amount of natural
frequencies of the structure.
2 Outline: method and program

First, experimental modal analysis (EMA) [1] is used to extract the natural
frequencies of the physical structure. This modal data is used as reference
response data during the procedure.
Next, a mathematical model of the structure is created. In the physical model
all mass and stiffness related properties are known except for the anisotropic
material properties. All known properties are implemented as such into the
mathematical model. The real composite material is modelled as a global
homogeneous orthotropic material in the mathematical model. This mathematical
model is solved for modal data using finite element analysis.
Finally, two sets of non matching modal data are available. One set composed
of experimentally obtained reference data. The other set contains calculated data
from the mathematical model. The orthotropic material properties are then
modified in such a way that both sets of modal data match. If those two sets
match, the virtual model has in a global sense the same mass and stiffness
properties as the real model. This principle is called model updating.
The procedure to identify orthotropic material properties by natural frequency
measurement is automated in the form of a user-friendly FEMtools program.
Program flow is visualized in the next flowchart.
Input Geometry
227
Check Geometry not ok.

New Input Geometry
Input initial material Properties

Obtain converged mesh with respect to first ten frequencies {freq}analyt.
The frequencies {freq}analyt are based on initial material properties.
The ten {freq}analyt and {}analyt are shown.
The sensitivity matrix material properties versus frequencies {freq}analyt is
shown.
Determine how many N response frequencies {freq}exp are taken into
account.
Determine which material properties are identified.
Fem database contains:

First N {freq}analyt
First N {}analyt
Initial Material Properties
Test database contains:

N experimental frequencies {freq}exp
First N {}analyt
Corresponding with physical model
Model Update
Modified fem model is shown.
Reference physical test model is shown.
MAC-matrix fem versus test is shown.
Correlation tracking curve is shown.
Modified Material Properties listed.
No Successful
Identification
Successful Identification
Model updating: mathematics
From a mathematical point of view, the difficulty with model updating is that the
relation between output vector and the parameter vector is nearly always nonlinear. This means that updating the parameter values from an initial value to a
final value has to be done iteratively. The value of the output for some new
parameter values can be estimated with a Taylor expansion. The Taylor series
can be cut off after the linear term or can be cut after some higher order terms.
Figure 1 illustrates the mathematics involved in model updating.
The matrix [S ] that appears in the linear Taylor term is called the sensitivity
matrix. It contains the partial derivatives of the output components for the
different parameter values. The success of model updating is highly dependant

from the numerical condition of the sensitivity matrix because [S ] must be
inverted in every iteration step to obtain the parameter correction {p} .
Figure 1:
Mathematics of model updating (left) and minimizing of cost

function (right).
Figure 2:
Simple cost function (left) and (right) a more elaborated cost

function.
Convergence from an initial parameter value
{p}0
to the final value is
obtained by minimization of a cost function in every iteration step. Graphically,

this means that the cost function evolutes iteratively from an initial point in the
(m+1) dimensional parameter space towards a global minimum. The parameter
values in the global minimum are the optimal parameter values. Figure 1
illustrates the principle.
One of the most simple cost functions leads to a weighted least squares
estimator. In this cost function, a weighting matrix [ W ] is pre- and post
multiplied with the difference between the measured and the computed response
229
in a point in the parameter space. The weighting matrix [ W ] allows to express a

different confidence in the different measured data points. This is illustrated in
Figure 2.
A more elaborated cost function also takes the initial parameter values into
account. A second term is added to the cost function in which a weighting matrix
is pre- and post multiplied with the difference between the initial and the current
parameter value. Again, a weighting matrix allows to express a different
confidence in the different initial parameter values.
The matrix [ W ]p represents the weighting matrix expressing the confidence
in the model parameters, while
[W]y
is a weighting matrix expressing the
confidence in the reference response test data.
Model updating: important considerations
The success of model updating strongly depends on the following considerations.

4.1 Accuracy numerical model
A first aspect is the quality of the mathematical model. All known mass and
stiffness properties must be correctly represented in the mathematical model.
Secondly, this model is solved using the finite element method. Error due to
discretization of the mathematical model is introduced. The discretization error
must be kept to a minimum. A coarse mesh density is not as flexible as the
mathematical model with infinite degrees of freedom. Hence, the calculated
natural frequencies will be overestimated. The model updating will result in
physically incorrect material properties. Discretization error is estimated by
comparing successive solutions with refined mesh density. More elements will
result in a more exact solution of the mathematical model. When the difference
between successive solutions is minimal, the mesh is convergence. The
accuracy of the analysis is related to the mesh density.
4.2 Experimental error
Incorrect input cannot result in physical correct material properties. Experimental
error can be divided into two categories. Random errors can be treated with
statistical procedures. Modal analysis software is capable in minimizing random
errors. Systematic errors are a lot more difficult to detect and to solve. A
damaged accelerometer will produce systematically an error on his output.
Experimental tests on an analytical known problem can indicate a systematic
error.
4.3 Controllability
A numerical model is controllable if it is possible to tune the model output from
an arbitrary point {y}0 in the parameter space to a measured point {y}exp with the

selected parameter set. Non-controllability can be turned into controllability by
selecting more or more appropriate parameter. A parameter is appropriate if the
sensitivity with respect to the response is sufficiently high.
4.4 Observability
A numerical model is observable if measurement of the output contains
sufficient information for the identification of the selected parameters. A natural
frequency does not contain information concerning the colour of the structure. It
is impossible to identify a colour by measuring the natural frequency of the
structure. Of course, one wants to identify all four orthotropic material properties
by measuring a certain amount of frequencies. Unfortunately, depending on the
structure, not every parameter will be identifiable.
To investigate observability, FEMtools offers sensitivity sum curves. Such a
curve sums all sensitivity values for all responses as a function of parameter
number. A low sensitivity sum value shows that none of the responses contains
sufficient information for the identification of that parameter. If this occurs, one
can add more experimental measurements or conclude that the parameter in
question can not be identified by frequency measurement.
4.5 Initial parameter values
Model updating requires initial parameter values {p}0 . The quality of the initial
parameter values can affect both the speed of convergence and whether or not
convergence to the true parameters is achieved. The parameter estimation
problem is often posed as a problem of constraint minimization and in the case
of non-linearity in parameters; a particular minimum is sought on a surface
which contains many minima and maxima. Usually either a local minimum is
sought, when there is confidence in the initial model, or else the problem is to
determine the unique global minimum.
To obtain initial values for the longitudinal moduli Ex and Ey, the first bending
frequency of a beam specimen is determined. To identify Ex a beam model in the
span direction of the structure is used. To identify Ey a beam model in the width
direction of the structure is used. Knowledge of the bending frequency and
dimensions of the specimen renders the elastic modulus of the material. The
formula used to estimate the elasticity modulus [2]:
E = 7.89e-2 * (frequency)2 * (length)4 * density * transverse section *
(Moment of inertia) 1
231
Model updating: FEMtools
5.1 Sensitivity calculation

The matrix [S ] that appears in the linear Taylor term is called the sensitivity
matrix. It contains the partial derivatives of the output components for the
different parameter values. There are two basic approaches to compute
sensitivities: (i) using differential sensitivities and (ii) using a finite difference
approximation. Which one to use depends on the parameter type. For nonproportional parameters, such as the Youngs modulus for orthotropic materials,
FEMtools uses finite difference sensitivities. In this method derivatives are
approximated with a forward finite difference approach. This is done using the
results of two finite element analysis for two states of the parameter pj. The
element ij of the matrix [S ] becomes
y i y i (p j p j ) y i (p j )
=
p j
p j
The sensitivities discussed so far are absolute sensitivities. This means that they
use the units of the response and parameter value. The absolute sensitivities can
be made independent of the units used for the response and parameter values.
They are then referred to as normalized sensitivities. A normalized sensitivity
shows the percentage change of the response value for one percent change of the
parameter value. The element ij of the matrix [S n ] can be written as
Sij(n) =
y i p j
p j y i
5.2 Mode shape pairing

During model updating, the algorithm will try to drive the predicted analytical
response to the experimental reference data in the test database. This implies that
the algorithm knows which analytical response has to match with which
experimental response. This can be defined using sequential mode shape pairing.
Sequential mode pairing means that analytical mode 1 will be paired with
experimental mode 1, analytical mode 2 with experimental mode 2, etc. If during
model updating a switch of mode shapes occurs, this method fails. The resulting
frequencies will probably be close but the mode shapes are different. There is no
possibility to connect a mode shape to a particular experimental reference
response.
To deal with the above problem, the program will copy the analytical mode
shapes, predicted by the initial values for parameters Ex, Ey, Gxy and xy, to the
test database. The user of the program knows exactly the sequence of these
modes [the program will previously show them]. It is now up to the user to
connect the correct experimental reference frequency with the corresponding
mode shape. In other words, the test database must reflect the correct physical

response [natural frequency with corresponding mode shape]. Automatic mode
shape pairing can now be used to drive the analytical response to the
experimental reference response in the test database.
During model updating, automatic mode shape pairing makes a relation
between those frequencies which have the highest Modal Assurance Criterion
(MAC). The MAC is a measure of the squared cosine of the angle between two
mode shapes. To compute the MAC between an analytical and experimental
mode shape, the following equation is used:
({ } { })
)=
({ } { })({ } { })
2
MAC(analyt , exp
After model updating, a MAC value can be calculated between the updated fem
model and the physical test modal. If no mode switch occurred during model
updating this MAC matrix is a diagonal matrix.
5.3 Aspects of convergence
CCABSOLUTE: Average value of weighted absolute relative differences
between predicted and reference resonance frequencies.
CCMEAN: Average value of weighted relative differences between predicted
and reference resonance frequencies.
In model updating, the above correlation coefficients are interpreted as an
objective function that needs to be minimized. With each iteration loop the
values of the correlation coefficients will be verified to check if a convergence
criterion is satisfied. The following criteria are used:
(1) The value of the reference correlation coefficient is less than an
imposed margin:
CCt < 1
(2) Two consecutive values of the reference correlation coefficient are
within a given margin.
CC t +1 CC t < 2
The iteration loop in model updating will be stopped as soon as one of these tests
is satisfied. Default values for 1 and 2 are used namely 0.08.
Example: symmetric U-profile
6.1 Section properties

Dimensions according to figure 3:
Moment of inertia Iy: 439635 mm4
Moment of inertia Iz: 251905 mm4
Torsional stiffness factor J: 1990 mm4
233
6.2 Volume properties

Volume: 867793 mm3
Mass: 1599 gram
Density: 1.8426e-9 Mg/mm3
Length: 1470 mm
65 mm
26.5 mm
3.18 mm
19 mm
3.18 mm
3.18 mm
63.5 mm
Figure 3:
Symmetric U-profile.
6.3 Estimation initial values

An initial value for E x is obtained by measuring the first bending mode of a
beam specimen in the span direction of the structure. An initial value for E y is
obtained by measuring the first bending mode of a beam specimen in the width
direction of the structure. Typical engineering properties of a glass-polymer
composite are used for G xy and xy [3].
Ex
Initial
Method used
value
to obtain value
(N/mm)
22000
Measured
Ey
11800
Measured
Gxy
5000
Literature
vxy
0,3
Literature

6.4 Calculated natural frequencies and mode shapes (based on initial
values)
Mode 1: 32
Mode 2: 61
Mode 3: 116
Mode 4: 140
Mode 5: 197
Mode 6: 199
Mode 7: 207
Mode 8: 217
Mode 9: 221
235
Mode10: 235
6.5 Sensitivity matrix

The sensitivity matrix shows the sensitivities of the four material parameters
versus the first ten resonant frequencies. Parameter 1, 2, 3 and 4 equals
respectively Ex, Ey, Gxy and xy. Figure 4 shows the sensitivity matrix.
The sensitivity matrix indicates that the first torsion mode contains
information about the shear modulus Gxy. The second mode - a complex bending
mode around the Y-axis is sensitive to a change in value of Ex and Gxy. The
parameter Ey can be identified by using resonant frequency five and six. None of
the responses are sensitive to a change in value of xy. No attempt should be
made to identify xy by measuring natural frequencies.
Figure 4:
Sensitivity matrix.
6.6 Model updating results

Natural frequencies and corresponding mode shapes are measured using a laser
vibrometer as measuring device and a shaker as excitation device. There was no
indication, what so ever, of modes five and six in the experimental measured
data. Probably due to the fact that a single shaker is not optimal to excite mode
five. Additional, both modes are extremely close to each other. Consequently,
further discussion will focus on the determination of Ex and Gxy. The first four
modes are used to identify Ex and Gxy.
Conclusions
This paper presents a method to determine the global effective orthotropic

material properties by measuring a certain amount of natural frequencies of a
composite beam structure. The program - developed in FEMtools - is used to
determine the properties of composite beams with closed cross-sections and open
cross-sections. Some general trends are clearly observed and stated hereafter.
Closed box beams behave relatively straightforward. In general, the first
modes are bending modes around the principal axes of the cross-section. The
torsion mode is found in a higher region since the torsional stiffness factor of a
closed cross-section is rather high. The bending modes can be used to identify Ex
and the torsion mode can be used to identify Gxy. The natural frequencies of this
kind of beams are not sensitive to a change in value of Ey and v xy . It is not
possible to determine these parameters by measuring natural frequencies.
Beams with open cross-section are much more complex in behaviour and
general conclusions can not be drawn. For this kind of beams, certain complex
mode shapes are sensitive to a change in value of multiple parameters. Moreover,
frequencies exists which are particularly sensitive to a modification of Ey and can
be used to identify this parameter. It is not possible to identify values for v xy .
Initial values
[N/mm]
Ex
Parameter
selection
Ey Gxy Vxy Ex Ey Gxy Vxy
22000 11800 5000 0,3 X
Mode
Fem data based on
shape
initial values
switch
versus Exp. data [Hz]
occur
Fem Exp. Description
Y N
value value
mode
32
32.5
1 torsion
1 complex
bending Y
116 137.5 1 bending Z
2 complex
140 146.9
bending Y
Final parameter
values [N/mm]
Ex
Ey Gxy Vxy
28944 11800 5014 0,3
61 66.25
The program needs initial values for orthotropic material properties, before
identification can start. A deviation of 25 % given on estimated initial values,
results in the same final updated values for the parameters. Hence, the final
updated results are almost not sensitive to a deviation of initial values.
The program can also be used to study the influence of the length of the
structure on the possibility to identify certain parameters. Consequently, an
optimal length can be determined for which the first (two) frequencies are very
sensitive to a change in value of a preferred parameter. Making it possible to
identify orthotropic material properties in a more easy and structured manner.
237
References
[1] Modal Testing: Theory and Practice, DJ Ewins, Research studies press
[2] Engineering Vibration, DJ Inman, Prentice Hall
[3] Stress analysis of fiber-reinforced composite materials, MW Hyer, McGrawHill
239
Full-field optical measurement for material

parameter identification with inverse methods
J. Gu1, S. Cooreman1, A. Smits1, S. Bossuyt1, H. Sol1,
D. Lecompte2 & J. Vantomme2
1
Department of Mechanics of Material and Construction,

Vrije Universiteit Brussel, Brussels, Belgium
2
Department of Materials and Construction, Royal Military Academy,
Brussels, Belgium
Abstract
The application of FE simulation in manufacturing processes and virtual
prototyping increases every day. In order to allow accurate simulations, correct
constitutive models are needed as input to the FE software. A modern and
promising way to identify the material parameters in those constitutive models is
inverse modeling. Full-field measurement is a suitable way to get the
necessary experimental data. The technique has many advantages such as large
information contents, non-contacting measurement, and versatile size of
observation region, among others. However, there is no standardization yet for
this kind of measurements. Therefore, there are many disagreements among
researchers about how to design DICT experiments and how to correctly collect
the data from DICT experiments. This paper will concentrate on discussing the
key points of those problems as well as presenting some work experience with
the DICT.
Keywords: inverse method, FEA, full field measurement, digital image
correlation.
Introduction
Identification of cracks and defects, and estimation of distributions of material

properties from experimental data are inverse problems, which are not well
recognized till the middle of 1980s. Thanks to the development of information
technology by leaps and bounds, increasing efforts have been devoted to advance
doi:10.2495/HPSM06024

in inverse problem research, especially at beginning of 1990s [1]. The main
reason is that the production cycles in industry, such as automobile and aircraft
industry, become shorter since 1990s. To face this challenge, many industries
take advantage of computer simulations to provide powerful insight into
structural behavior of mechanical systems, manufacturing processes, and many
other engineering problems, which can reduce the dependency of the
manufacturers on expensive and time consuming hardware prototyping [2]. A
common requirement for the success of the numerical models used in computer
simulation software is the input of the correct material properties. The properties
of many materials however can not be found in literature. A modern alternative
to find the material properties is inverse modeling.
The principle of inverse modeling is shown in Figure 1. The input quantities
are assumed to be perfectly known and are the same for the experiment as for the
numerical model. The unknown parameters in the numerical model are tuned in
such a way that the computed output matches the experimentally measured
output, e.g. displacements or strains, as closely as possible.
Figure 1:
The principle of an inverse method for material identification.
In general, the elasto-plastic material parameters are determined by means of

standard tests, such as tensile tests, bending tests and torsion tests. However, the
stress and strain fields occurring during these traditional tests do not represent
the complex stress and strain fields which are generated during real production
processes. With those techniques, the specimen can be tested in such a way that
the material circumstances are comparable with the circumstances of the material
241
during the service life of the construction or during the manufacturing process,
e.g. Single Point Incremental Forming process (SPIF).
The main problem of adopting more complicated tests in the past was always
hindered by the fact that complex displacement fields simply could not be
measured. In recent years however, an increasing number of important
developments in the field of full-field displacement measuring have been
presented. Moreover, modern measurement equipment, as there are Flow
Induced Birefringence, Electronic Speckle Pattern Interferometry, Digital Image
Correlation techniques, have become commercially available and more
accessible than they used to be in the past. Furthermore, more information about
material parameters can be caught from one experiment, so called full field
measurement.
In the current article, an example of full-field measurement by means of
digital image correlation technique (DICT) will be presented. This technique
exploits a good prospect for inverse modeling since it has some important
advantages, for instance non-contacting measurement, large observation region,
etc. However, there is no standardization for such kind of measurements.
Therefore, there are many disputations about how to design DIC experiment and
how to correctly collect the data from DIC experiment. In the following the
influence of subset size, step size and strain window size on the final strain
computation will be discussed in detail. Some experiences for DIC experiment
will be introduced for sharing the knowledge.
Full field digital image correlation experiment
In general DICT is just a method to measure displacements. However, some

extra calculations on the displacement field allow us to extract strain data from
the displacement measurement. In this case, this technique will be applied for the
measurement of the heterogeneous deformation fields during a forming process
2.1 Principle of DIC
Basically, the displacement measurement goes as follows: a random speckle
pattern is applied on the surface of the object of interest. During the forming
process, a number of pictures of the object of interest are taken with one or more
CCD (Charge Coupled Device) cameras (if two cameras are used, it is possible
to measure the displacements in three orthogonal directions). Each picture
corresponds to a different state of deformation (usually the first picture is taken
at zero loading and is called the reference image). Finally the speckle pattern
allows us to correlate the different images to each other and as a result the
displacement field (relative to the reference image) in the different images can be
measured. As was already mentioned, some extra calculations make it possible to
extract strain data from this deformation measurement. Figure 2 summarizes the
above described technique.
Figure 2:
Diagram of digital image correlation experiment.
The CCD cameras use a small, rectangular piece of silicon, which has been
segmented into arrays of individual light-sensitive cells, also known as photo
sites. Each cell constitutes one element of the whole picture and is called a
pixel. Every pixel stores a certain grey scale value varying from 0 to 255, in
agreement with the intensity of the light, reflected by the surface of the tested
specimen. Thus, an image can be looked at as a matrix in which every element
represents the grey value of the corresponding pixel.
In order to run the correlation algorithm, the image is divided in a number of
subsets. A subset represents a part of the whole image. The size of a subset can
be varied by the user and the choice of a good subset size depends on the
deformation. As was already mentioned, this technique uses a random speckle
pattern that is simply sprayed onto the surface of the object or that is offered by
the texture of the specimens material. The objective is to obtain an image with a
varied and distinctive pattern, which enables the correlation algorithm to trace
the subsets of the reference image in the deformed images [3, 4]. The concept
behind the DICT-software 2D matching algorithm is that the distribution of grey
values in a rectangular area (subset) in the picture taken of the specimen in the
undeformed state, corresponds to the distribution of grey values of the same area
in the picture taken at the deformed specimen. In this way the motion and the
deformation of the subsets during the deformation of the object are determined.
This will finally lead to a displacement field for the whole area of interest. Some
additional calculations allow to extract strain data from this displacement field.
Figure 3 shows such a subset in the undeformed (left) and deformed (right)
configuration.
Figure 3:
243
A subset in the undeformed (left) and deformed (right)

configuration.
Measurements in three dimensional space require a measuring arrangement

with two cameras, which should be placed with an angle of approx. 20 between
them. A 3D measuring arrangement has to be calibrated prior to a measurement
in order to be able to determine the image-forming qualities of the cameras and
the lens distortions. With the aid of the determined image-forming qualities the
subsequent calculations can be carried out.
10
140
50
280
15
Figure 4:
The geometry of the specimen and the generated speckle pattern

zone.

2.2 Description of DIC experiment set-up
A tension test on an aluminium beam with a circular hole in the center is selected
to study the identification of metallic plastic parameters, for instance, the
anisotropic elasto-plastic law based on Hill48 yield locus with kinematic
hardening. The geometry of the specimen is shown in Figure 4. The aluminium
specimen was loaded in tension up to 80 kN. A speckle pattern was applied to
measure the displacement field in the shaded zone by means of two CCD
cameras. The images are regularly taken throughout the tension test. The
software offered by Limess Company was used to calculate the strains in the
shaded regions (see Figure 4).
Post processing of DIC results
During the post processing of the DIC experiment, the area-of-interest (AOI)
should be selected. Following that, subset size (pixel) and step size (pixel) are
chosen for calculating the displacement field. The strain window size is chosen
for the strain calculation by Vic 3D software. One image, taken when force
applied on the specimen reached a value of 40kN, is selected to study the
influence of Subset, Step and Strain window size on the strain value.
Figure 5:
Upper left picture, the contour of strain y calculated with subset

11 pixels; upper right picture, the contour of strain y calculated
with subset 17 pixels; lower picture, the contour of strain y
calculated with subset 23 pixels.
245
3.1 Different subset

The subset size controls the area of the image that is used to track the
displacement between images. The subset size has to be large enough to ensure
that there is a sufficiently distinctive pattern contained in the area used for
correlation.
To show the influence of the subset size on the strain computation, a subset
size of 11, 17 and 23 pixels was applied. The step size and the strain window
size were 5 pixels and 15 respectively. It is found that some region with speckles
bigger than subset lost the correlation, for instance, the upper left picture in
Figure 5.
3.2 Different step
The step size controls the spacing of the points that are analyzed during
correlation. If a step size of 1 pixel is chosen, a correlation analysis is performed
at every pixel inside the area of interest.
To show the influence of the step size on the strain computation, a step size of
1, 5 and 11 pixels was applied. The subset size and the strain window size were
23 pixels and 15 respectively. It is found that too small step sizes result in a lot
of noise on the calculated strain field. This causes some difficulties to collect
strain data. To big step sizes result in too much averaging. Hence the actual
strains are underestimated, especially in regions of high strain gradients, e.g.
around the hole, as shown in Figure 6.
Figure 6:
Upper left picture, the contour of strain y calculated with step

1 pixel; upper right picture, the contour of strain y calculated with
step 5 pixels; lower picture, the contour of strain y calculated with
step 11 pixels.

3.3 Different strain window size
The strain window size can be used to adjust the size of the local neighborhood
in which the derivatives of the displacement field are calculated. The value
indicates the number of neighboring points instead of pixel.
To show the influence of the strain window size on the strain computation, a
strain window size of 5, 15 and 25 was applied. The subset size and the step size
were 23 pixels and 5 pixels respectively. It is found that too small strain window
sizes result in a lot of noise on the calculated strain field (Figure 7), whereas too
large strain window sizes result in too much averaging, especially in the region
of high strain gradients like around the hole.
Figure 7:
Upper left picture, the contour of strain y calculated with strain

window size 5; upper right picture, the contour of strain y
calculated with strain window size 15; lower picture, the contour of
strain y calculated with and strain window size 25
Comparison of DIC results and FE simulation results
A FE simulation of this tensile test was performed by means of ABAQUS

software, and the obtained strain contours in x and y direction are compared with
those from the DIC experiment, respectively (see Figure 8 and Figure 9). Since
the specimen is symmetric, only one eighth of specimen was modeled. The
results of the FE simulation approach those of the DIC experiment quite well.
The FE strains in x direction vary between 0.00011 and -0.00082; the strains in x
direction, computed by the DIC software vary from 0.00033 to -0.00092. The FE
strains in y direction vary from 0.0024 to 0.000097; the strains in y direction,
computed by means of the DIC software vary between 0.0026 and 0.
247
Figure 8:
Left picture is the contour of strain in x direction obtained from FE

simulation. The scale of strain is from 1.11e-4 to -8.19e-4; right
picture is the contour of strain in x direction obtained from DIC
experiment. The scale of strain is from 3.31e-4 to -9.19e-4.
Figure 9:
Up picture is the contour of strain in y direction obtained from FE

simulation. The scale of strain is from 0.0024 to 9.7e-5; the picture
below is the contour of strain in y direction obtained from DIC
experiment. The scale of strain is from 0.0026 to 0.
Conclusion
In the present paper, the feasibility and reliability of data collection from fullfield optical measurement is studied. It has been shown that the size of subset,
step and strain window have a large influence on the accuracy of the measured
displacements and calculated strains. Unsuitable subset, step size and strain
window size will cause experiment data either underestimated or scattered.
In the post processing of DIC experiment, generally the larger the subsets are,
the better the results are. However, large is a relative concept. It is important that
the subset size is chosen in accordance with the expected deformations. It is clear

that for steep gradients in the displacement or strain field, a large subset will
smooth the real behaviour and thus yield erroneous results.
References
[1] Shiro Kubo, Inverse problems. Atlanta Technology Publications, ISBN #1883793-01-07
[2] Sol H., Oomens C.W.J., Material identification using mixed numerical
experimental methods. 1997 Kluwer Academic Publishers. pp 1-9
[3] Synnergren P., Sjdagk M., A stereoscopic digital speckle photography
system for 3-D displacement field measurements. Optics and Lasers in
Engineering, 31, pp 425-433, 1999
[4] Fricke-Begemann T., Burke J., Speckle interferometry: three-dimensional
field measurement with a single interferogram. M., A stereoscopic digital
speckle photography system for 3-D displacement field measurements.
Applied Optics 40(28), pp 5011-5022, 1 October 2001
249
Multiaxial characterization of the mechanical

behaviour of aluminium foam
L. Peroni, M. Avalle & P. Martella
Department of Mechanics,
Politecnico di Torino (Technical University of Turin), Italy
Abstract
In the past ten years many new processes for making foamed metals, mostly
aluminium or aluminium alloys, have been developed. As a matter of fact,
closed-cell aluminium foam offers a unique combination of properties such as
low density, high stiffness, strength, and energy absorption capability.
One of the main differences in the mechanical behaviour of cellular materials
with respect to classical homogeneous materials such as solid metals is that foam
failure is not independent from a hydrostatic state of stress. Therefore, it is not
possible to describe the failure surface from a single, usually uniaxial, test but it
is necessary to perform tests with different combination of deviatoric and
hydrostatic stress components.
Within the European Project APROSYS, whose main objective is increasing
the safety of all road-users, the mechanical behaviour of aluminium foam, under
multiaxial loading, was investigated by the authors. In this paper the results of
the hydrostatic and hydro-compression experimental tests are reported. From the
results of these tests, it has been possible to obtain the yield locus of the
aluminium foam in the deviatoric-hydrostatic stress components space.
Keywords: aluminium foam, hydrostatic, hydrocompression.
Introduction
It is well know that the mechanical behaviour of a cellular material, like foam, is
not independent from a hydrostatic state of stress as, on the contrary, it happens
for classical homogeneous materials such as metals. Therefore, it is not possible
to describe the failure surface from a single simple uniaxial test (in tension or
compression if there is a different behaviour) but it is necessary to perform tests
doi:10.2495/HPSM06025

with different combinations of deviatoric and hydrostatic stress components [1].
A test that gives pure deviatoric stress is, for example, a shear test. A test that
produces pure hydrostatic stress in the material is the hydrostatic or triaxial test.
A tensile test gives a combination of deviatoric and hydrostatic stress
components.
In a material insensitive to the hydrostatic component of stress (as usually are
considered metals) the failure limit in a deviatoric-hydrostatic space is a
horizontal line. When the deviatoric stress reaches the yield line yielding occurs,
therefore for metals it is possible to indefinitely increase the hydrostatic stress
without yielding. If this hypothesis can be accepted a single experimental test
gives the complete yield locus for this material.
The yield locus is different for foams (and other plastic materials) because
there is an influence of the hydrostatic stress component [23]. A single
experimental test is not enough to characterize the material behaviour
completely. As a consequence a series of different tests are necessary in order to
cover the widest possible range of combinations of deviatoric and hydrostatic
stress components.
Experimental tests
Within the 6th Framework Programme European project APROSYS [4], a

complete characterisation of aluminium foam [58] was carried out at
Politecnico di Torino. The experimental tests were performed in the Laboratory
of the 2nd Faculty of Engineering, in Vercelli. In this paper the hydrostatic
compression and the hydro-compression tests are reported.
dev
Hydro-compression tests area

Uniaxial compression
Hydro-compression
Hydrostatic
compression
= atan(r)
1
Figure 1:
R
R
A
p = A / B
dev = A R
hyd =
hyd
A + 2 R
3
r = dev / hyd
Hydrostatic and hydro-compression tests area in the deviatorichydrostatic stress components space.
Hydrostatic compression tests were performed first to assess the behaviour of

the tested aluminium foam in a pure hydrostatic stress condition. Then hydroWIT Transactions on The Built Environment, Vol 85, 2006 WIT Press
251
compression tests were carried out changing the ratio p in order to obtain
different hyd - dev combinations. As shown in fig. 1, combining in different
ways the axial stress A and the radial stress R it is possible to investigate the
highlighted area of the hyd - dev plane. The limits of this area are the pure
hydrostatic test (A = R) and the pure compression test (R = 0).
2.1 Testing device
Performing hydrostatic and hydro-compression tests on a metal foam is a
challenging task. The foam specimen have to be completely separated from the
fluid used to load it; for this reason it was covered with a latex sheath in order to
avoid seepage of the fluid in the specimen. The cover have to be enough strong
to not break during the test and, at the same time, it have to be very light and thin
in order to not change the test results. Furthermore in the hydro-compression
tests it is necessary to control the axial and the radial pressure separately. For all
this reasons a testing device (fig. 2) was designed and built expressly to perform
these tests. The test chamber, containing the foam specimen, was filled with a
fluid (mixture of water and glycol) and mounted directly on the fixtures of a
general purpose hydraulic testing machine DARTEC HA100 controlled with a
DARTEC 9600 electronic unit.
Figure 2:
Schematic of the loading system.
In the hydrostatic tests the specimen was put on the vertical rod connected to
the testing machine. This rod could move inside the test chamber in the axial
direction changing the volume of the chamber and, consequently, the pressure of
the fluid inside the chamber, i.e. the pressure on the specimen.

In the hydro-compression tests the moving rod connected to the hydraulic
testing machine pressed the specimen against a fixed rod inside the chamber
generating the axial stress A on the specimen. Meantime, a double-effect
pneumatic piston, was connected to the lateral surface of the test chamber. The
rod of this piston could move inside the test chamber in the radial direction
changing the volume of the chamber and, consequently, the pressure inside the
chamber, generating the radial stress R on the specimen. In both tests the
electronic unit performed the test control and the data acquisition, while a PC
equipped with a National Instruments acquisition board was used to acquire the
load, stroke and pressure. In the hydro-compression tests the PC carried out the
control of the radial pressure R (pressure inside the test chamber) by moving the
piston rod, whose displacement was measured through a potentiometer. The load
applied though the rod connected to the testing machine was measured with a
100 kN, class A, strain-gage load cell, while the stroke of the rod was measured
by means of a LVDT transducer connected to the hydraulic actuator. The
pressure inside the chamber was measured with a 350 bar strain-gauge pressure
transducer applied to the chamber.
2.2 Samples
The samples were cylinders cut from blocks, made of aluminium foam, produced
by the Fraunhofer-Institute for Applied Materials Research (IFAM) in Bremen
(D). These blocks were obtained through a powder metallurgical process for
preparing foamed metals. According to this process, commercial powders are
mixed with small quantities of a powdered foaming agent by means of
conventional techniques. The mixture is compacted to a semi-finished product of
low porosity by applying compaction techniques such as extrusion or coextrusion. The result of the compaction process is a foamable semi-finished
product that can be worked into sheets, profiles, etc. by applying conventional
deformation techniques. During a final heat treatment at temperatures above the
melting point of the corresponding alloy, the material expands and develops its
highly porous, closed-cell structure.
Three blocks (308 mm 125 mm 45 mm), identified as B2, B3 and B4,
were cut in seven sub-blocks each to obtain the samples. These prismatic square
section sub-blocks were turned with a CNC lathe so as to get circular cylindrical
specimens with a nominal diameter and height of 41 mm. As in the quasi-static
compression tests (not reported in this paper) an anisotropic behaviour of the
specimens was found, the specimens were machined in two different ways. In
the ones obtained from block B2 the axial direction was parallel to the topbottom direction of the block (fig. 3). On the contrary, the specimens from
blocks B3 and B4 the axial direction was parallel to the transversal direction of
the block (fig. 3). In the foaming process the top-bottom direction is the first
foaming direction (direction for which the boundary planes of the block were
reached first), while the transverse and the longitudinal directions were the
second and the third foaming direction respectively.
Top-bottom
253
Specimens from block B2

Specimens from blocks B3 and B4
Longitudinal
Transversal
Figure 3:
Position of the axis of the cylindrical specimens with respect to the

block directions.
The properties of the specimens used for hydrostatic compression and hydrocompression tests are reported in the following tables.
Table 1:
Properties of the specimens for hydrostatic compression tests.

DIMENSIONS (mm)
SPECIMEN
CODE
Diameter
Height
DENSITY
(kg/dm)
AXIAL
DIRECTION
H-B2-1
41.2
41.8
0.66
Top-Bottom
H-B2-2
40.8
39.7
0.66
Top-Bottom
H-B3-1
40.8
45.5
0.64
Transversal
H-B3-2
40.8
45.3
0.37
Transversal
H-B4-1
40.8
41.2
0.27
Transversal
H-B4-2
41.0
41.0
0.36
Transversal
H-B4-3
41.0
40.7
0.40
Transversal
Table 2:
Properties of the specimens for hydro-compression tests.

DIMENSIONS (mm)
SPECIMEN
CODE
Diameter
Height
DENSITY
(kg/dm)
SPECIMEN AXIAL
DIRECTION
HC-B2-1
40.6
40.5
0.40
Top-Bottom
HC-B2-2
40.8
39.2
0.34
Top-Bottom
HC-B2-3
40.6
40.2
0.57
Top-Bottom
HC-B3-1
40.8
45.3
0.54
Transversal
HC-B3-2
40.8
45.3
0.47
Transversal
HC-B3-3
40.8
45.0
0.38
Transversal
HC-B4-1
40.8
45.4
0.60
Transversal
3.1 Hydrostatic compression tests

The results of the hydrostatic compression tests are reported in fig. 4. As it is
possible to notice, the results are strongly dependent from the density which is
the apparent density, the average value within the volume of the sample. Since
there is a large scatter in size of the cells, cell thickness, and cell distribution the
apparent density is affected by a very large scatter; as a consequence the test
results are rather scattered. However a general behaviour can be detected. The
elastic modulus, the yield stress and the slope of the stress-strain curve after
yielding, all increase with increasing density.
Hydrostatic stress (MPa)
30
H-B2-1 0.66
H-B2-2 0.66
H-B3-1 0.64
H-B3-2 0.37
H-B4-1 0.27
H-B4-2 0.36
H-B4-3 0.4
25
20
15
10
5
0
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Volumetric strain (-)
Figure 4:
Hydrostatic compression tests: stress-strain curves
Furthermore, the tested foam showed an anisotropic behaviour, which can be

detected by examining the deformed shapes of some specimens after the tests
(fig 5). In specimen H-B3-1 the directions of the cross section are the
longitudinal and the top-bottom directions of the base block. Because the topbottom direction is weaker than the longitudinal one, the cross section of the
deformed specimen is elliptical (fig 5.(a)). This didnt occur in the H-B2-2
specimen for which the directions of the cross section are the longitudinal and
the transversal directions of the base block, which have approximately the same
strength. As a consequence, despite the slightly lower density, the yield stress of
specimen H-B3-1 is higher than the one of the H-B2-1 and H-B2-2 specimens.
The hydrostatic yield stress for the specimens with a transversal axial
direction was evaluated as intersection of the two lines interpolating the first part
(elastic) and the second part (plateau) of the experimental stress-strain curve. An
equation, which relates the hydrostatic yield stress to the density, was found
through an interpolation of these experimental data.
yield-hyd = 13.381.773
(1)
255
Density in the previous equation is expressed in kg/m and stress in MPa.
(a)
Figure 5:
(b)
Cross section of H-B3-1 (a) and H-B2-2 (b) specimen after the
tests.
In fig. 6 hydrostatic tests are compared with uniaxial quasi-static uniaxial

tests (not reported in this paper) for two different densities. As it can be noticed
the yield stress is higher for the uniaxial test because in this loading condition the
specimen material is not constrained in the radial direction and therefore it can
move laterally before the axial yield. Furthermore, the densification rate is
greater for the hydrostatic loading condition.
Hydrostatic stress (MPa)
80
S-B1-1 0.36 Uniaxial
70
60
H-B4-2 0.36 Hydrostatic
50
S-B1-2 0.64 Uniaxial
40
H-B3-1 0.64 Hydrostatic
30
20
10
0
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
Volumetric strain
Figure 6:
Comparison of uniaxial quasi-static tests and hydrostatic tests.
3.2 Hydro-compression tests

As explained above, one of the main aims of the experimental tests was
evaluating the failure locus of the foam as shown in fig. 1. For this reason each
hydro-compression test was performed applying a proportional load, i.e. a fixed
ratio R / A, as shown in the following table. In addition two hydrostatic tests,

on specimen H-B3-1 and H-B3-2 (reported in the previous section) were carried
out using the testing device for the hydro-compression tests with a ratio p = 1.
Table 3:
Hydro-compression tests settings.
SPECIMEN
Density
(kg/m3)
p = R / A
r = dev / hyd
HC-B2-1
0.40
p = 0.658
r = 0.445
HC-B2-2
0.34
p = 0.366
r = 1.111
HC-B2-3
0.57
p = 0.658
r = 0.445
HC-B3-1
0.54
p = 0.658
r = 0.445
HC-B3-2
0.47
p = 0.366
r = 1.111
HC-B3-3
0.38
p = 0.658
r = 0.445
HC-B4-1
0.60
p = 0.366
r = 1.111
H-B3-1
0.64
p = 1 (hydrostatic)
r = 0
H-B3-2
0.37
p = 1 (hydrostatic)
r = 0
Examining the experimental data (axial and radial stress-strain curves), it was
possible to notice a different behaviour in the axial and radial directions. This
behaviour is shown clearly in the fig. 7, where the results of the hydrostatic test
performed on specimen H-B3-2 are reported. This test was performed with the
experimental device for the hydro-compression tests so it was possible to
evaluate the axial and the radial characteristics separately.
Radial strain (-)
0
0.05
0.1
0.15
0.2
8
Axial stress
Radial stress
7
6
7
6
Radial stress (MPa)
Axial stress (MPa)
0.05
0.1
0.15
0.2
Axial strain (-)
Figure 7:
Hydrostatic test (p = 1) on specimen H-B3-2.
In fig. 7, it is possible to notice that, even if the load ratio p = 1 was kept at a
constant level during the test, the axial and the radial yield stress were quite
257
different. This means that the ratio between the axial yield stress and the radial
yield stress was different from the ratio p at which the experimental test was
carried out. This event is rather clear in the hydrostatic tests, while in the hydrocompression tests it occurred only some times and not in a so evident way.
However, due to this phenomenon, it was not possible to use the experimental
data directly to obtain the failure locus because it was not clear which one of the
yield stresses should have been chosen. In order to overcome this problem the
following procedure was used. First of all, the hydrostatic stress-volumetric
strain curve was calculated from the axial and radial stress-strain curves through
the following equations:
hyd = (A + 2 R) / 3, vol = A + 2 R
(2)
Then the yield stress for each test was evaluated considering the hydrostatic
stress-volumetric strain curve. In order to remove the effect of the density on the
results, each yield stress, calculated as described above, was divided by the
theoretical yield stress evaluated through eqn (1). Therefore a normalised yield
yield-adim stress was obtained
Finally, the normalised deviatoric and hydrostatic stress components were
obtained combining the following equation:
R = p A, hyd = (A + 2 R) / 3, dev = | A R |
(3)
These components are plotted in the dev hyd (fig. 8) in order to obtain the
failure locus.
1.8
Uni-axial compression test
dev normalised
1.6
1.4
1.2
HC-B4-1
1
HC-B3-2
0.8
0.6
HC-B3-1
HC-B3-3
H-B3-2
H-B3-1
0.4
0.2
0
0
0.25
0.5
0.75
hyd normalised
Figure 8:
Failure locus.
1.25
4 Summary and conclusions

Performing hydrostatic and hydro-compression tests on a metal foam proved to
be a challenging and rather complicated task. The insulation of the specimen
from the fluid used to load it was particularly difficult; indeed in many tests the
cover used to protect the specimen from seepage of the fluid broke and the fluid
penetrated into the specimen causing an early end of the test. For this reason
many tests were carried out in order to set up the testing device and procedure; in
this paper only the most important results are reported.
To sum up, the mechanical behaviour of aluminium foam was investigated
through a series of experimental tests performed within the European Project
APROSYS. As for foams the yielding is not independent from the hydrostatic
stress component, a single experimental test is not enough to characterize the
material behaviour completely. Therefore pure hydrostatic compression and
hydro-compression tests were carried out through a testing device expressly
developed. From the results of these tests it has been possible to obtain the yield
locus of the aluminium foam in the deviatoric-hydrostatic stress components
space.
Acknowledgements
The financial support of the European Commission by means of the APROSYS
project and Mr. Dirk Lehmhus of Fraunhofer-Institute for Applied Materials
Research (IFAM) are gratefully acknowledged.
References
[1] Gibson, L.J, Ashby, M.F., Cellular solids: structure and properties (Second
Edition), Cambridge University Press, 1997.
[2] Collins, J.A., Failure of materials in mechanical design, John Wiley & Sons,
1980.
[3] Khan, A.S., Huang, S., Continuum theory of plasticity, John Wiley & Sons,
1995.
[4] Integrated Project on Advanced PROtection SYStems (APROSYS). On line:
www.aprosys.com.
[5] Ashby, M.F., Evans, A., Fleck, N.A., Gibson, L.J., Hutchinson, J.W.,
Wadley, H.N.G., Metal foams A Design Guide, Butterworth Heinemann,
2000.
[6] Ehlers, W., Mullerschon, H., Klar, O., On the behaviour of Aluminium
Foams under Uniaxial and Multiaxial Loading, Verlag Mit, 1999.
[7] Deshpande, V.S., Fleck, N.A., Isotropic constitutive models for metallic
foams, Journal of the Mechanics and Physics of Solids, Pergamon, 1999.
[8] Deshpande, V.S., Fleck, N.A., Multiaxial yield behaviour of polymer foams,
Acta Materialia, Pergamon, 2001.
259
Characterisation of the high strain rate

properties of Advanced High Strength Steels
J. Van Slycken1, P. Verleysen1, J. Degrieck1 & J. Bouquerel2
1
Department of Mechanical Construction and Production,

Faculty of Engineering, Ghent University, Belgium
2
Department of Metallurgy and Materials Science,
Faculty of Engineering, Ghent University, Belgium
Abstract
In the automotive industry a lot of energy is put into the development of
lightweight auto body structures that are able to outperform the classic
structures. For these purposes tremendous advances have been made in the field
of multi-phase steels. Complex steel grades have been developed with
exceptional mechanical properties: they combine high strength values (yield
strength, tensile strength, etc.) with an excellent ductility. TRIP steels
(TRansformation Induced Plasticity steels) show these properties pre-eminently.
To guarantee a controlled dissipation of the energy released during a crash,
knowledge and understanding of the impact-dynamic material properties is
essential. In this paper the results are presented of an extensive experimental
program to investigate the strain rate dependent mechanical properties of
different TRIP steels. The influence of different alloying types (Al, Si, SiAl, etc.)
on the static and dynamic stress-strain behaviour is investigated. A split
Hopkinson tensile bar set-up was used for the experiments. Microstructural
observation techniques such as different optical methods, SEM and XRD were
used to reveal the mechanisms governing the observed high strain rate
behaviour. From the results it is clear that the excellent mechanical properties are
not only preserved at higher strain rates, but still improve. The influence of the
alloying elements is comparable in the static and dynamic case: aluminium tends
to increase the elongation level of the material, whereas silicon improves the
stress that is achieved.
Keywords: high strain rate behaviour, split Hopkinson tensile bar, TRIP steel,
alloying elements.
doi:10.2495/HPSM06026
Introduction
Engineered steels provide automotive designers and manufacturers with the

unique option to combine lightweight design with the traditional advantages of
steel: low cost and eco-efficiency. Under the impulse of several steel-auto
partnerships, including the Ultra Light Steel Auto Body (ULSAB) program, new
types of high-strength steel, called Advanced High-Strength Steels (AHSS)
(Bleck [1]), are engineered to complete the traditional steel grades. In the
ULSAB-Advanced Vehicle Concept (AVC) program, the need to reduce the
added mass which is required to satisfy future safety mandates presents the
opportunity to apply these newer types of high-strength steels in the design of an
efficient lightweight body structure. Members of the AHSS family include Dual
Phase (DP), Transformation Induced Plasticity (TRIP), Complex Phase (CP) and
Martensite Steels (MS).
The low alloy TRIP steels show high strength values in combination with an
excellent deformability, making them the material of choice for impactabsorbing structural parts of auto-bodies (Bleck [1]). TRIP steel has a
microstructure of soft ferrite () grains with bainite () and retained austenite
() (figure 1). The retained austenite transforms into martensite () during
deformation. The hard martensite delays the onset of necking leading to high
total elongation values and high crash energy absorption. TRIP steels can
therefore be engineered or tailored to provide excellent formability for
manufacturing complex parts. In addition, these steels can be designed into the
automotive body structure to offer excellent crash energy absorption.
Figure 1:
SEM-micrograph showing the multiphase microstructure of a nondeformed CMnAl-TRIP steel (magnification: x 5000).
The need for more optimized crashworthiness analysis in the automotive

industry makes high strain rate tensile testing of sheet steels very important. It is
well known that steels display positive strain rate performance, i.e. at the higher
rates of strain which are typically associated with crash events, steels have higher
261
strengths and consequently higher crash energy absorption. Different types of

testing techniques have been used to generate data under these dynamic
conditions, each serving a specific range of strain rates. One of the most
commonly used setups is a split Hopkinson bar setup. Mostly, a pressure
apparatus is used to obtain the dynamic parameters (Davies and Hunter [2]), a
tensile setup is however preferred for the testing of sheet metals. Stress-strain
curves at strain rates varying from 500 to 2000 s-1 can be obtained with the
tensile setup developed at Ghent University.
The behaviour of steels during high strain rate loading is the result of the
interaction between two opposing processes: strain rate hardening and thermal
softening, which is due to adiabatic heating during deformation. In the case of
TRIP steels, thermal softening also affects the - transformation: the increased
temperature reduces the transformation rate. This paper presents the results of an
extensive experimental program, which is set up to assess the dynamic
mechanical behaviour of TRIP steels. Special attention is paid to the influence of
the different alloying elements in TRIP steels on the stability of the austenite
phase and thus on the transformation in the material.
Materials and methods
2.1 Materials
A key parameter for the TRIP effect is the stability of the meta-stable austenite
which is mainly determined by the austenite particle size and the composition,
especially the carbon content is an important parameter (Itami et al. [3]).
Alloying elements such as silicon, aluminium and phosphor in TRIP steels principally added to inhibit carbide precipitation during the second stage
isothermal holding temperature in the production process of TRIP steels - also
significantly influence the thermodynamic stability of the austenite phase. The
influence of the alloying elements is assessed by the study of four different TRIP
steel grades: CMnAl-, CMnSi-, CMnSiAl- and CMnSiAlP-TRIP. Specific care
was taken to keep the same carbon content for each steel grade. Table 1 lists the
composition of the low alloy TRIP steels.
Table 1:
Chemical composition of the investigated TRIP steel grades (in

weight percent).
Steel grade
CMnAl-TRIP
CMnSi-TRIP
CMnSiAl-TRIP
CMnSiAlP-TRIP
C
0.24
0.25
0.25
0.20
Mn
1.61
1.67
1.70
1.56
Al
1.54
0.78
0.69
0.29
Si
0.091
1.28
0.55
0.38
P
0.006
0.012
0.011
0.012
Silicon strengthens the ferrite phase considerably, but high Si-contents results
in an adherent FeO.SiO2 oxide layer on the sheet surface, which generates
surface defects on the hot rolled sheet and which is difficult to remove by

pickling. Moreover, due to this oxide layer CMnSi-TRIP steels are complicated
to galvanize and thus to process in continuous galvanizing lines. Therefore, Al
has been used to replace the silicon in TRIP steels. Al increases the ductility
significantly because it has a lower solid solution strengthening potential for the
ferrite phase (Girault et al. [4]).
2.2 Experimental testing procedure
A Split Hopkinson Tensile Bar (SHTB) apparatus is used to characterize the
dynamic properties of the investigated materials. It mainly consists of two bars:
an input and an output bar between which a specimen of the test material is
attached (figure 2). A tensile wave is produced by an impactor that is accelerated
towards the anvil of the input bar. After the impact this incident wave travels
along the input bar towards the specimen where it interacts with the sample and
is partly reflected back into the input bar. The other part, the transmitted wave,
travels along the output bar. The strain histories of the different waves (incident,
reflected and transmitted wave, respectively denoted as i(t), r(t), t(t)), are
recorded by means of strain gauges mounted on both bars. By adjusting the
impact speed of the impactor, the strain rate can be varied.
Figure 2:
Schematic representation of the Split Hopkinson Tensile Bar

(SHTB) set-up.
According to the one-dimensional wave theory and the assumption of a uniaxial and homogeneous stress and strain in the specimen, the stress, strain and
strain rate in the specimen can be written as follows (Kolsky [5]):
(t)=
EbAb
U -U
2C
t ( t ) , ( t ) = ob ib =- b
As
Ls
Ls
( ) d ,
r
Vob -Vib 2Cb

(1)
=r ( t )
Ls
Ls
where, As and Ls are the cross-sectional area and the length of the testing region
of the specimen, respectively. Cb is the one-dimensional elastic stress wave
velocity in the input/output bar material, Ab is the cross-section area and Eb is
Youngs modulus of the input/output bar. Uib and Uob are the displacements of
the specimen/input bar interface and the specimen/output bar interface,
respectively. Vib and Vob are the corresponding interface velocities.
( t ) =
263
Results
3.1 Influence of the alloying elements

Figure 3 shows a comparison of the engineering stress-strain curves of the
different investigated steel grades at a strain rate of ~1600 s-1. The four steel
grades exhibit a remarkable uniform elongation despite their high strength levels.
This is due to the occurrence of the strain-induced transformation and thus, to the
TRIP effect. The difference in their respective strength and elongation levels can
be attributed to their chemical composition. The differences in dynamic flow
stress are fully consistent with the ones in the static case: silicon addition
increases the strength considerably, whereas aluminium has little effect on the
strength level (Girault et al. [4]) and the CMnSiAl-TRIP steel shows an
intermediate behaviour. The addition of phosphor on the other hand, has limited
influence on the behaviour at the early stages of deformation. When the strain
increases, the strain hardening of the CMnSiAl-TRIP steel is more important
than of the CMnSiAlP-TRIP, which reaches maximum stress and uniform
elongation at a lower level of deformation.
Figure 3:
Comparison of the engineering stress-strain curves of the

investigated steel grades at high strain rate (~1600 s-1).
3.2 Strain hardening

To investigate the hardening behaviour, the strain hardening coefficient or nvalue, appearing in the Hollomon stress-strain relation of eqn (2)
(Stout and Follansbee [6]) is calculated as a function of strain for the different
stress-strain curves.
= K n
(2)

where is the true stress, p the true plastic strain and K is a material constant. A
strain window of 0.05 in which the strain hardening coefficient is averaged out,
is used to minimize the effect of oscillations in the stress-strain curves. In figure
4 the evolution of the strain hardening coefficient as a function of the true strain
is represented for the TRIP steels. In the early stages of deformation, the strain
hardening is highest for the CMnSiAl (P)-TRIP steels, followed by the CMnSiTRIP. The TRIP steel with high aluminium content shows the lowest strain
hardening in the beginning. The hardening of CMnAlSiP-TRIP begins to
stabilize and in the region between 7 and 14% of true strain the hardening
behaviour of CMnAl, CMnSi and CMnAlSiP is similar. After approximately
14% of true strain, the strain hardening coefficient of the CMnSi-TRIP steel
begins to decrease. The decrease of the strain hardening of the CMnAl- and the
CMnSiAl (P)-TRIP steels begins later during the deformation. The CMnAl-TRIP
steel shows a slighter decrease in strain hardening at the end of the deformation.
Figure 4:
Comparison of the strain hardening coefficients or n-value of the

investigated TRIP steel grades in function of the true strain at a
strain rate of ~1200 s-1.
When looking at the uniform elongation, CMnAl-TRIP steel shows the

highest n-value, whereas CMnSi-TRIP steel shows the lowest. Figure 5 displays
the strain hardening rate (d/d) for the investigated steel grades during dynamic
deformation. The strain hardening rate of ferrite is known to increase with silicon
additions and to be insignificantly affected by the presence of aluminium.
Accordingly, this behaviour is seen in dynamic conditions during the early stages
of deformation. CMnSi-TRIP shows the highest strain hardening rate, whereas
CMnAl-TRIP has the lowest values. CMnSiAl shows an intermediate behaviour.
After approximately 7% of true strain, the strain hardening rate of the CMnSiTRIP and the CMnSiAl (P)-TRIP steels become similar. The behaviour of the
CMnAl-TRIP steel shows a more constant evolution during deformation,
whereas the other steel grades have a steeper decrease in strain hardening rate.
Figure 5:
265
Comparison of the strain hardening rate d/d of the investigated

TRIP steel grades in function of the true strain at a strain rate of
~1200 s-1.
3.3 Influence of the strain rate

In figure 6 the tensile strength in function of the strain rate is presented for the
CMnAl-, CMnSi- and the CMnSiAl-TRIP steel grades. The influence of the
strain rate is material dependent, but positive for both the CMnSi-TRIP steel and
the CMnAl-TRIP steels: stress levels rise as the strain rate increases whereas for
the CMnSiAl-TRIP this influence is limited. Quite some scatter can be noticed
for the tensile strength values for the CMnSi-TRIP steel.
Figure 6:
Tensile strength as a function of strain rate for the CMnAl-, CMnSi

and the CMnSiAl-TRIP steel grades.
Different parameters can be used to evaluate the crash-resistance performance

of steels. Since, for most applications, the material rarely deforms up to fracture,

the yield stress and energy absorbed by the material at certain levels of
deformation gives valuable information. In figure 7 the energy dissipated by the
CMnAl-, CMnSi- and the CMnSiAl-TRIP steel at 10% strain is given as a
function of the strain rate. From this figure it is clear that all steel grades exhibit
a positive strain rate dependency. The CMnSi-TRIP steel shows the highest
energy absorption values, the other two steel grades have similar energy
absorption levels.
Figure 7:
Energy absorption until 10% deformation as a function of strain

rate for the CMnAl-, CMnSi and the CMnSiAl-TRIP steel grades.
In dynamic conditions the strain rate has limited influence on the material
properties. If these dynamic properties are compared to properties after static
deformation on the other hand, an important difference can be observed. Table 2
shows the mechanical properties of the CMnAl-TRIP steel after static (strained
at a constant strain rate of 10-4 s-1) and after dynamic deformation. An important
increase in both yield and tensile strength can be noticed when comparing static
to dynamic conditions. The uniform elongation however shows limited changes,
as well as the energy dissipation until 10% deformation.
Table 2:
Static
704 s-1
1246 s-1
1898 s-1
Mechanical properties of CMnAl-TRIP steel in static and dynamic

conditions.
Upper yield
strength,
MPa
Tensile
strength,
MPa
Uniform
elongation, -
506
622
646
672
689
786
800
814
0,265
0,264
0,301
0,252
Energy
dissipation at
10%
deformation,
106 J/m3
55,91
59,16
61,39
64,99
267
This can be explained when looking at the X-ray diffraction (XRD) patterns
in order to reveal the mechanisms governing the complex mechanical behaviour.
In figure 8 XRD patterns of non-deformed specimens are compared with patterns
of specimens after static and dynamic tests. The measurements are performed on
a Siemens D5000 diffractometer. Only the ferrite (b.c.c.) and austenite (f.c.c.)
peaks are considered in the measurements. Diffractograms were obtained in the
19-40 2-range using a filtered molybdenum K radiation.
Figure 8:
XRD pattern for the CMnAl-TRIP steel, 2-range between 19 and

40.
The diffractogram of the non-strained sample presents an initial amount of

12% of retained austenite. After both static and dynamic tensile tests, no (220)
austenite peaks, except the (200) one, are present. In the dynamic tests, no
austenite was expected because of the importance of the adiabatic heating during
deformation. The temperature in the specimen is expected to reach, at the end of
the test, about 90-100C. The temperature increase inhibits the austenite to
martensite transformation kinetic (Samek et al. [7]).
Conclusions
Results are presented of an extensive study of the strain rate dependent

behaviour of TRIP steels. These materials combine high strength with high
ductility and offer therefore an excellent crash energy absorption potential. Split
Hopkinson tensile bar tests are performed to obtain the stress-strain curves at
higher strain rates. The influence of the strain rate on the behaviour is especially
important when comparing static with dynamic testing conditions. The amount
of austenite that is transformed to martensite and the adiabatic heating during
dynamic deformation plays an important role. Special attention is paid to the
influence of alloying elements such as Al, Si and P on the dynamic behaviour of
TRIP steels. As in the static case, silicon contributes to a significant solid
solution strengthening of the ferrite matrix. TRIP steels with high silicon content

show therefore high strength levels. TRIP steels mainly alloyed with aluminium
on the other hand exhibit lower strength values but higher elongation levels. The
work hardening of the latter steel grade is more constant during deformation,
whereas TRIP steels with high silicon content show higher work hardening.
These properties can further be used to develop several material models. The
strain rate dependent behaviour cannot be described in a general way and various
types of constitutive relations have been proposed. The validation of these
models can be used for crashworthiness analysis in the automotive industry.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
Bleck, W., Using the TRIP effect - the dawn of a promising group of cold
formable steels. Proceedings of the International Conference on TRIPAided High Strength Ferrous Alloys, ed. B. C. De Cooman, pp. 13-24,
2002.
Davies, E. D. H. & Hunter, S. C., The dynamic compression testing of
solids by the method of the split Hopkinson pressure bar. J. Mech. Phys.
Solids, 11, pp. 155-179, 1963.
Itami, A., Takahashi, M. & Ushioda, K., Plastic Stability of Retained
Austenite in the Cold-Rolled 0.14-Percent-C-1.9-Percent-Si-1.7-PercentMn Sheet Steel. Isij International, 35(9), pp. 1121-1127, 1995.
Girault, E., Mertens, A., Jacques, P., Houbaert, Y., Verlinden, B. & Van
Humbeeck, J., Comparison of the effects of silicon and aluminium on the
tensile behaviour of multiphase TRIP-assisted steels. Scripta Materialia,
44, pp. 885-892, 2001.
Kolsky, H., An investigation of mechanical properties of materials at very
high strain rates of loading. Proc Phys Soc Lond Sec B, 62, pp. 676-700,
1949.
Stout, M. G. & Follansbee, P. S., Strain rate sensitivity, strain hardening,
and yield behavior of 304L stainless steel. Journal of Engineering
Materials and Technology - Transactions of the ASME, 108(4), pp. 344353, 1986.
Samek, L., De Cooman, B. C., Van Slycken, J., Verleysen, P. & Degrieck,
J., Physical Metallurgy of Multi-Phase Steel for Improved Passenger Car
Crash-Worthiness. Steel Research International, 75(11), pp. 716-723,
2004.
269
Evaluation of bond strength in

Roller Compacted Concrete under
various normal pressures
M. Madhkhan & A. Arasteh
Department of Civil Engineering, Isfahan University of Technology, Iran
Abstract
The construction of dams using Roller Compacted Concrete (RCC) is a relatively
new technology, which has rapidly developed in recent years. To increase the
safety factor in designing the various horizontal construction joints in RCC
dams, an experimental research using interlayer cement grout has been carried
out to improve the bond strength. A device with the ability of exerting the
normal pressure in the direct shear method has been designed and used. 144 tests
were carried out to assess the effect of interlayer cement grout along with
exerting the normal pressure. The tests include different proportions of water to
cement of the interlayer grout (0.5, 0.75, 1) and different values of normal
pressure (0, 5, 10, 15 kg/cm 2 ) in the ages of 7, 28 and 90 days. The results
indicate that the use of interlayer cement grout in the range of the normal
pressure exerted gives higher bond strength compared to the state of no grout. By
reducing the ratio of water to cement of the grout, the bond strength increases.
Keywords: Roller Compacted Concrete, horizontal construction joint, interlayer
cement grout, bond strength, normal pressure.
Introduction
The RCC dams are built in relatively thin layers, with a thickness of
approximately 30 centimeters placed, spread and compacted over each other;
therefore a large number of horizontal joints are formed in the dam. The quality
of the concrete at the joint surface is of great importance to ensure the integration
of layers. But in practice, this surface has a lower quality compared to the
concrete mass; and a lower tensile and shear strength and a higher permeability
is expected [1].
doi:10.2495/HPSM06027

The shear strength is a function of cohesion and angle of internal friction. The
minimum shear properties occur at the horizontal construction joints between the
lifts of RCC. Generally, the shear strength of the joint is dependent upon the
amount of cement used in the RCC mixture, grading and quality of aggregates,
amount of compaction of each layer, existing conditions of each layer surface
(including weather conditions, construction conditions, and etc), layer age,
improvement of layer surface by adding bedding mixture and index of maturity
or curing. A large range of the bond strength depends on the RCC mixture and
above named factors. As a general range, the cohesion can be taken between 0.1
and 2.5 MPa and the angle of friction between 30 and 65 degrees [2].
Although the normal pressure varies from one point to another in dam height,
but the experimental researches and the bond (shear) strength determination are
mostly carried out in the absence of normal pressure. The shear strength
parameters stated in the technical data, are based on coring studies and joint tests
on specimens taken from the dam built or being built. In this research, the
influence of different normal pressures on the bond (shear) strength of RCC
construction joints is studied using common laboratory specimens and by using
interlayer cement grout.
RCC materials
2.1 Cement
The heat generation control is of great importance in selecting the cement type.
Heat generation is typically controlled by using pozzolans and slags, therefore
Portland cements including Type , and V can be used [3]. In the present
research, exports cement Type of the Isfahan cement factory was used with the
chemical analysis given in Table 1. The quantities of C 3 S, C 2 S, C 3 A, C 4 AF
and insoluble residue in this cement are respectively 46.13, 27.38, 9.61, 9.98 and
0.44%.
Table 1:
Chemical analysis of the cement used in the RCC mix design (%).
SiO 2
Al 2 O 3
Fe 2 O 3
CaO
MgO
SO 3
Na 2 O
K2 O
21.68
5.72
3.28
63.53
1.75
1.63
0.2
0.52
2.2 Pozzolans
Ninety percent of the dams built until 1998 contain some kind of additional
material such as slag or pozzolan, and in only 10 percent of the cases, the cement
has been used individually to make RCC [3]. To replace a portion of the cement
by pozzolan (25%) a natural pozzolan powder was used in the RCC mix design.
The chemical analysis is given in Table 2.
Table 2:
271
Chemical analysis of natural pozzolan (%).
SiO 2
Al 2 O 3
Fe 2 O 3
CaO
MgO
SO 3
Na 2 O
K2 O
57.35
18.2
6.23
1.45
1.93
3.5
1.72
2.3 Aggregates
Considering that the aggregates fill 75 to 85% of the RCC volume [4], Their
properties are an effective factor on fresh and hardened RCC. The gravel is
crushed river aggregates with a maximum size of 25.4 millimeters in accordance
with ASTM C33, and the sand is natural sand with 4.2% of the material passing
No. 200 sieve (non-plastic) in accordance with ASTM C33. The gravel/sand
ratio in the RCC mix design is 55 to 45. The grading of the coarse and fine
aggregates is given in Table 3.
Table 3:
Grading of the aggregates (Gravel and sand) in the RCC mix

design.
Coarse
Agg.
Fine
Agg.
Agg.
size (mm)
137
12.5
9.5
4.75
2.36
1.18
Passing
Percent
92.3
52.1
28.2
1.3
0.0
0.0
Agg.
size (mm)
4.75
2.36
1.18
0.6
0.3
0.15
Passing
Percent
96.5
87.1
66.9
42.8
27.7
12.2
Table 4:
Water
19.0
Mix design specifications (in one cubic meter).

Weight of mix material (kg/m3)
Aggregates
Cementitious material
Gravel
Sand
Cement
Pozzolan
1161
950
139
35
Tests program
The above materials, cubic RCC specimens with 15x15x15 cm dimensions were
made using the optimum mix design [5] given in Table 4. The specimens were
made in two layers and each layer was compacted for 15 seconds using the
electric vibrating hammer (ASTM C1435).
The specimens are cured and protected by common laboratory methods to
carry out shear strength tests at the ages of 7, 28 and 90 days.

The tests carried out include different proportions of water to cement of
interlayer grout (0.5, 0.75, 1) and different values of normal pressure (0, 5, 10,
2
and 15 kg/cm ).
3.1 Test method for determining the bond strength
Considering that the construction joint can be referred to as an imposed crack
(failure), for determination of the RCC bond strength according to previous
researches [6], a device with a similar mechanism to the soil direct shear
apparatus was designed and used. The RCC specimen was placed in the frame
(figure1). The frame consists of a fixed plate and a moving plate. The normal
pressure is exerted according to the mechanical theory of high strength bolts and
by tightening the bolts on the moving plate. The device was then placed under
the pressure jack (with the pressure being exerted in the same direction as the
contraction joint). Finally, two sliding plates were placed tangent to the joint and
on opposite sides of it. By exerting the pressure on the jack, shear failure would
occur at the joint (figure 2).
Figure 1:
Device for exerting normal pressure to the RCC specimen.
Analysis of results
4.1 Bond strength at different ages

In Table 5, the average results of the bond strength determination at the ages of
7, 28 and 90 days (each test done at least 3 times) are given with the
specifications of the specimen (interlayer grout) and also test along (4 values of
normal pressure).
The variations of bond strength versus normal pressure for different values of
proportion of the interlayer grout are shown in figures 3, 4 and 5 (for different
273
ages of specimens). Any test where no grout is used has the least bond strength,
and where the w/c ratio is equal to 0.5 the most bond strength is obtained. By
decreasing the w/c ratio, the bond strength increases. This matter can be
observed more significantly in the diagrams of figure 4 (at the age of 28).
Figure 2:
Table 5:
Exertion of normal and shear forces.
Average results of RCC bond strength at the ages of 7, 28 and 90

days.
2
Age
(days)
28
90
w/c
ratio
No grout
1
0.75
0.5
No grout
1
0.75
0.5
No grout
1
0.75
0.5
Normal
pressure
= 0
2
kg/cm
16
17.2
17.6
19.4
21.8
23.6
23.3
24
25.4
25.3
27.4
28.5
Bond strength [] (kg/cm )

Normal
Normal
Normal
pressure
pressure
pressure
= 5
= 10
= 15
2
2
2
kg/cm
kg/cm
kg/cm
20.9
22.2
26.7
19
24.45
25.9
23
24.15
28.3
20.3
25.2
28.6
23.1
30.4
31.1
25.8
31.9
33.2
29.8
30.1
35.4
30.8
32.1
35.4
27.1
34.6
35.1
30.9
35.6
36.1
28.4
35.0
37.9
28.8
33.5
39.2
Bond Strength (kg /cm 2)

30
28
26
24
22
20
18
16
14
12
10
no grout
R2 = 0.9419
w /c of grout=1
R2 = 0.945
w /c of grout=0.75
w /c of grout=0.5
R = 0.9496
no grout
R2 = 0.9595
w /c of grout=1
w /c of grout=0.75
5
10
15
Norm al Pressure(kg/cm 2)
Figure 3:
20
w /c of grout=0.5
Bond strength versus normal pressure diagram at the age of 7 days,

comparing different proportions of w/c.
Bond Strength (kg /cm 2)
38
36
no grout
R2 = 0.9099
34
32
w /c of grout=1
R = 0.9147
30
w /c of grout=0.75
w /c of grout=0.5
28
26
no grout
R2 = 0.8853
24
w /c of grout=1
R2 = 0.9386
22
w /c of grout=0.75
20
0
Figure 4:
5
10
15
Norm al Pressure(kg/cm 2)
20
w /c of grout=0.5
Bond strength versus normal pressure diagram at the age of 28

days, comparing different proportions of w/c.
The RCC shear strength parameters which include the interlayer cohesion (C)
and the angle of interlayer friction () have been derived for each age and each
proportion of w/c. The results are given in table 6.
4.2 Effect of specimen age on the bond strength
For all normal pressure values and proportions of w/c, on average, the 7-day
bond strength is approximately 77.7% of the 28-day bond strength. Also, the 90WIT Transactions on The Built Environment, Vol 85, 2006 WIT Press
275
day bond strength has an increase of 10.1% compared to the 28-day bond
strength (figure 6).
40
R2 = 0.9328
B o n d Stren g th (kg /cm 2)
38
no grout
36
w /c of grout=1
34
w /c of grout=0.75
R2 = 0.8959
32
w /c of grout=0.5
R2 = 0.8868
30
28
no grout
R2 = 0.9007
26
w /c of grout=1
24
w /c of grout=0.75
22
0
Figure 5:
5
10
15
Normal Pressure(kg/cm2)
Bond strength versus normal pressure diagram at the age of 90

days, comparing different proportions of w/c.
Table 6:
Specimen
Age (days)
7
28
90
w /c of grout=0.5
20
RCC shear strength parameters (C, ).
w/c ratio
No grout
0.5
0.75
1
No grout
0.5
0.75
1
No grout
0.5
0.75
1
Interlayer
Cohesion, kg/cm
16.44
18.51
18.28
16.87
21.32
25.25
24.16
23.39
25.06
26.98
26.46
26.41
Interlayer
friction angle
33.7
32.96
33.58
32.41
35.14
35.4
36.2
34.9
36.2
36.4
37.3
36.6
4.3 Effect of w/c ratio of interlayer grout and specimen age on shear
strength parameters
The interlayer cohesion is intact with the bond strength, resulting that the
maximum and minimum cohesion occur at w/c=0.5 and state of no grout used,
respectively (figure 7). The angle of interlayer friction has no specific trend, and
Bond strength (kg/cm2)
its maximum and minimum values occur at w/c=0.75 and w/c=1.0, respectively
(figure 8).
34
32
30
28
w /c=-
26
w /c=1
24
w /c=0.75
22
w /c=0.5
20
0
10
20
30
40
50
60
70
80
90
100
Age(days)
Figure 6:
Bond strength versus age curves for different proportions of w/c.
Cohession (kg/cm2)
30
28
26
24
22
w/c=-
20
w/c=1
18
w/c=0.75
16
w/c=0.5
14
0
10
20
30
40
50
60
70
80
90
100
Age(days)
Figure 7:
Cohesion versus age curves for different proportions of w/c.
Conclusions
In this research, a device with the ability of exerting the normal pressure was
designed to assess the RCC bond strength. The analysis of the overall results
obtained from the optimum mix design, different w/c ratios (0.5, 0.75, and 1) and
2
different values of normal pressure (0, 5, 10 and 15 kg/cm ) at the ages of 7, 28

and 90 days gives the following conclusions:
The Mohr coulomb equation [ = C + tan ] has regressed the bond strength
versus normal pressure for each proportion of w/c and also for the state of no
grout used, along with the high correlation coefficients derived from above.
277
Friction(angle)
38
36
34
w /c=w /c=1
32
w /c=0.75
w /c=0.5
30
0
Figure 8:
10
20
30
40
50
60
Age (days )
70
80
90
100
Angle of friction versus age curves for different proportions of w/c.
The minimum bond strength was obtained in the tests where no grout was
used, and where the interlayer grout is used, the bond strength would decrease
whilst the w/c ratio is increased.
The variation of the interlayer cohesion including all proportions of w/c and
also the state of no grout used is intact with the variation of the bond strength,
whereas the angle of interlayer friction does not follow any special pattern.
References
[1]
[2]
[3]
[4]
[5]
[6]
Roller Compacted Mass Concrete, American Concrete Institute, ACI

207.5R, 1999.
Schrader, E.k. Shear Strength and Lift Joint quality of RCC International
Journal on Hydropower & Dams, v6, n1, 2002, p 46-55.
Dunstan, M.R.H Latest Development in RCC Dams, Proceedings of
International Symposium on Roller Compacted Concrete Dams, Chengdu,
China, April 21-25, 1999.
Guide for Selecting Proportions for No-Slump Concrete, American
Concrete Institute, ACI 211.3R, 1997.
Ramazanian pour, A. A & Hassankhani, A., Optimum Mix Design of
Roller Compacted Concrete, Proceeding of 6th International Conference
of Civil Engineering, Isfahan, IRAN, 2003.
Bogat, H. & RAHGOZAR, R., Effect of Polymer Grouts on RCC dams
bond strength, Proceeding of 6th International Conference of Civil
Engineering, Isfahan, IRAN, 2003.
279
High performance fibres and the mechanical

attributes of cut resistant structures made
therewith
S. Rebouillat & B. Steffenino
DuPont International S.A., Geneva, P.O. Box 50, Geneva, Switzerland
Abstract
High performance materials are rarely used as 100% compound, therefore most
of the time when reference is made to these categories of materials it is as a
matter of fact a reference to a system comprising a well-combined ingredients
exhibiting together high performances in use. Fibres, micro-fibres and now nanofibres are definitely part of the high performance reinforcing ingredients, which
in a system largely contributes to the performance in use of the designed system.
High performance fibres, and more specifically aramids for their outstanding
heat and mechanical properties, although most often used as reinforcing
compound, are also used more uniquely and directly in situations where
mechanical and thermal protections are required. It is therefore useful as a
preliminary grounding to review these types of materials in terms of their
properties. Then their attributes and potential contribution alone or in a system
become more specific and can be extrapolated more easily. The mechanical
performance and more specifically the cut resistance of articles made of high
performance materials are generally measured against several norm
specifications. Those are not designed to predict the real in use performance of
the considered protective articles. This gap is even larger when the cut resistance
of the compounding fibre increases. Although relationships exist between the
various cut performance levels obtained with the ASTM, ISO and EN norms
(respectively American, International and European standards), there is a need
for a more fundamental understanding and interpretation of the data. The ASTM
and ISO approaches remain the most suitable methods for that purpose. Various
parameters, such as the high performance material and fibre composition, the
effect of kinetic such as the cutting speed and the speed profile, the force
distributions, the artefacts generated by various surface coatings are presented
and discussed aiming at improving the interpretations of the data generated
through the norm procedures. Therewith, significant progress in the
understanding of the cut fundamentals and the norm harmonization are made
available to the users of high performance fibres, such as KEVLAR.
Keywords: high performance materials and fibres, textile structures, coated
textile materials, mechanical protection, cut mechanism, cut fundamentals,
KEVLAR.
doi:10.2495/HPSM06028
Introduction
Aromatic polyamides came as breakthrough materials in commercial

applications as early as 1961, with the market launch of the meta-aramid fibre
Nomex, which opened new horizons in the field of thermal and electrical
insulation.
A much higher tenacity and modulus fibre was developed and
commercialised, also by DuPont, under the trade name KEVLAR in 1971.
Scientists, in the fields of liquid crystals, polymers, rheology and fibre
processing, as well as, process and system engineers spent several years prior
and during the early stage of its market introduction to establish the basics and
fundamental understanding necessary to take full advantage of this new class of
high performance materials. The outstanding potential of these materials derived
mostly from the anisotropy of their superimposed substructures presenting
pleated, crystalline, fibrillar and skin-core characteristics.
In this work we will focus on the main high performance fibres and the
applications of the commercial aramids. Basic papers outlining the engineering
of such products, especially their surface modifications, are available and
provide a good basis for a scientific background [17].
We will study in more detail the protective solutions obtained from high
performance fibres. A gentle force of 0.01 N applied on a razor blade, such as the
one used by the hairdresser, represents more than 1000 bar or 104 (N/cm2), there
is no doubt that the performance of an effective cut protective barrier has to be a
high performance.
In the professional activities the awareness and the prevention regarding hand
injuries have substantially contributed to contain the injury level, nonetheless the
hand protection remains an area where the room for improvement is substantial.
Thanks to the personal protective equipment, such as gloves, made of high
performance fibres, the situation is improving on a worldwide basis. General
guidance about hand protection is provided in the ANSI/ISEA 105-2000 [8].
Norms, developed in 1987 in Europe and in 1997 in the USA, have played an
important role in classifying cut resistant materials. Glove performance can now
reach the upper range of the cut scale level defined by certain norms with a
substantial dexterity enhancement.
Norm levels are nowadays considered more as entry criteria in the proposition
of a given hand protection than as a truly predictive tool to anticipate the real cut
risk where experience and expert judgment prevail.
The fundamental understanding of the cut mechanism also plays an essential
role to complement this approach. The test equipment, which are described in the
various norms relative to cut protection, such as ASTM, EN and ISO
(respectively American, European and International standards), are good tools to
develop a basis for the cut fundamental understanding. The work presented in
this paper provides and illustrates a number of situations where cut fundamentals
can be link with the properties of high performance fibres.
There is no attempt to extrapolate the results to a real risk encountered in the
field, therefore this study cannot constitute a basis for risk analysis but can serve
281
as guidance towards an improved understanding of factors which may influence

the cut performance of protective equipment.
High performance fibres
2.1 Main high performance fibres

The following designation was adopted in 1974 by the United States Federal
Trade Commission to describe the aromatic polyamide based fibres under the
generic term aramid: a manufactured fibre in which the fibre forming substance
is a long chain synthetic polyamide in which at least 85% of the amide
(- CO - NH -) linkages are attached directly to two aromatic rings.
The earliest representative of this class is poly-m-phenyleneisophthalamide,
which was commercialised by DuPont in 1967 as Nomex aramid fibre. Its
chemical formula is:
Nomex
The discovery in 1965 of high tenacity, high modulus fibres from liquid
crystalline solutions of synthetic para-aromatic polyamides led to the commercial
production of KEVLAR aramid fibre by DuPont Co in 1971[9]; the
corresponding chemical formula is given below:
KEVLAR
KEVLAR fibres are poly (p-phenylene terephthalamide) (PPTA) [10].

Another para-aramid, Twaron, similar to KEVLAR, and an aromatic
copolyamide appeared on the market towards the end of the 80s. Teijin
commercialised the copolyamide Technora fibre.
For the sake of completeness, Monsanto in the 1970s developed, based on an
aromatic polyamide-hydrazine composition, an aromatic copolyamide fibre
under the code X500 which almost reached the market.
PBO (commercialised under the trade name Zylon) and PIPD, often referred
as to M5, are other high performance fibres sufficiently close to the aramids to
be mentioned here [1].
2.2 Spinning fibres
Production of fibres initially involves heating the spinning polymeric solution up
to a suitable processing temperature, which is of the order of 80oC for the highly

concentrated solutions in 100% (water-free) [11] sulphuric acid. The solution
state corresponds to a nematic liquid crystalline phase. The concentration limit
for the polymer in spinning solution is 20 wt %.
Polymer spinning solutions are extruded through spinning holes and subjected
to elongational stretch across a small air gap, illustrated in fig. 1. The spinning
holes fulfill an important function. Under shear, the crystal domains become
elongated and orientated in the direction of the deformation [1214]. Once in the
air gap, elongational stretching takes place. This is effected by making the
velocity of the fibre as it leaves the coagulating bath higher than the velocity of
the polymer as it emerges from the spinning holes. This ratio is often referred to
as the draw ratio, which can be fine-tuned to obtain higher tenacities and moduli
with lower elongations and denier. The resulting stretch in the air gap further
perfects the respective alignment of the liquid crystal domains.
Figure 1: Schematical representation of the extrusion of the liquid crystalline

solution in the dry-jet wet spinning process.
Overall a higher polymer orientation in the coagulation medium corresponds
to higher mechanical properties of the fibre. Because of the slower relaxation
time of these liquid crystal systems, the high as-spun fibre orientation can be
attained and retained via coagulation with cold water [15]. Essentially, the
crystallinity and orientation of the solution are translated to the fibre. These
factors allow the production of high strength, high modulus, as spun fibres.
2.3 Structures and properties
Aramid fibres have unique properties that set them apart from other fibres.
Aramid fibre tensile strength and modulus are significantly higher than earlier
283
organic fibres and fibre elongation is lower. Aramid fibres can be woven on
fabric looms more easily than brittle fibres such as glass, carbon or ceramic.
They also exhibit inherent resistance to organic solvents, fuels, lubricants and
exposure to flame.
The superimposed structures, such as the crystallites, the fibrils and the skincore boundaries, are definitely unique attributes, which can be partially tailored
through the fibre process engineering.
2.4 Mechanical properties
Typical stress-strain curves of different KEVLAR fibres are provided in
fig. 2(a), which clearly outlines the modulus increase from KEVLAR 29 to
KEVLAR 149. There are differences between KEVLAR 49 and KEVLAR
29. The respective moduli, brought by various spinning conditions and posttreatments performed on para-aramid precursors are generally considered as
intermediate between those of graphite, boron and glass fibres. The linear stressstrain behaviour of para-aramids is special compared to most man-made fibres
which tensile behaviours are depicted in fig. 2(b).
Figure 2:
Typical stress/strain curves of (a) KEVLAR fibres and (b) other

commercially representative industrial yarns.
Creep is measured either by the length variation under tension or by the stress
decrease at constant gage length. Para-aramids, which exhibit little creep, differ
significantly from other highly oriented polymeric fibres, such as HMPE fibres,
which can break after several days under intermediate load due to their high
creep properties associated with a stress slip of molecules compared to a
structure tightening in the case of para-aramids. The temperature, the load
relative to the fibre ultimate strength, the water content and other parameters
affect creep.

2.5 Applications
The present section takes into consideration a selected range of applications of
the aramids with an attempt to underline some of the system engineering aspects.
This approach may provide not only a better understanding of the major reasons
for these fibres to contribute to the best balanced performance of the system, but
also may stimulate new ideas in the way these fibres are used in the considered
system or how they could be used in a new one.
As an example, the understanding of how a para-aramid fibre can be shaped
into an optimised fabric pattern to resist to a fragment impact, can provoke new
ideas regarding the engineering of these fibres to offer cut and heat protection in
other sectors where the kinetics may be much closer to static conditions.
This system approach, led to the development of some of the aramid fibre
applications to be outlined in this section illustrated in table 1 [11].
Table 1:
Aramid market segments and key attributes.
The associated fig. 3 provides a concrete illustration of the dynamic network

established since the inception of the aramids. This network outlines the dynamic
285
integration of aramids in advanced and modern technologies with a constant

adaptation to the new challenges associated with various generations of product.
There are end-use market segments, which have been quite rethought because
of the aramids, such as the asbestos replacement by para-aramid pulps. There are
other areas, which will continue to evolve because of the ever-growing stringent
requirements for energy saving in transportation for example.
Figure 3:
Applications of para-aramid fibres.
The aramid contribution in this sector is outstanding. Clearly,

communications, including transportation as well as transmission, leisure and
sports, life protection and, health and safety in general, have been tremendously
improved and adapted to modern technologies in part because of the aramids.
High performance fibres for protective solutions
The cut performance of articles made of high performance materials is generally

measured against several norm specifications [16, 17]. Those are not designed to
predict the real in use performance of the considered protective articles. This
gap is even larger when the cut resistance of the compounding fibre increases.
Although relationships exist between the various cut performance levels
obtained with the ASTM, ISO and EN norms, there is a need for a more
fundamental understanding and interpretation of the data. The ASTM and ISO
approaches remain the most suitable methods for cut level exceeding 3.
Figure 4(a): Test equipments (top) with schematic representation (bottom) of

the force application, (EN388).
3.1 Cut testing methods EN, ISO, ASTM
The first standardized cut test method was developed in France in 1987
(prEN388 [18], hereafter called EN) followed by a method developed in 1997 by
the American Society for Testing Material, (ASTM- F1790-97 [19], hereafter
called ASTM) and its most recent international adaptation via the International
287
Standardisation Organization (ISO-13997 [20], hereafter called ISO). General

guidance about hand protection is summarized in the ANSI/ISEA 105-2000 [8],
which relates to theses various norms.
Each testing procedure is described in the norm documents, [1820],
mentioned above. The three related test equipments are depicted in figs. 4 (a),
4(b) with a schematic representation of the blade orientation, the load force
application and the blade displacement.
There are similarities between the ASTM and ISO procedures and
equipments, but there are fundamentals differences between those two and the
EN equipment and test method.
Figure 4(b): Test equipments (top) with schematic representation (bottom) of

the force application, blade displacement and blade orientation,
(ASTM/ ISO).
The EN test method is based on the measurement of the substrate cut
resistance when submitted to the cutting effect generated by the rotation of the
circular blade on the sample tested, which is under a relatively moderate pressure

provided by a 5 N load force applied on the blade. The number of revolutions of
the blade to cause the cut-through of the specimen will be used to calculate the
cut index and therewith the cut performance. A cotton fabric is used as the
reference material.
The ASTM and ISO methods on the opposite are based on the measurement
of the load necessary to provoke the cut-through of the sample after a sliding
distance of the rectangular blade of 20 mm (most recently adopted harmonized
distance). The corresponding load can reach, for knitted samples made of high
performance fibres, the equivalent of 1000 to 3500 g, i.e. 10 to 35 N, which is 2
to 7 times higher than the load force applied on the circular blade in the case of
the EN method. This is a difference worth highlighting.
3.2 Experimental part
The design of the experiments was selected in order to clarify and better
understand some of the aspects, which were partly introduced in the materials
and methods section.
For the EN test method, the blade cutting edge degradation during the course
of the testing of high cut performance materials, hereafter HCP materials, has
been illustrated and studied by conducting tests on a stop and go basis, i.e.
changing the blade during the test every 1, 2, 3, and 5 cycles as illustrated in
table 2. Both 100% KEVLAR and 5.5 % steel reinforced para-aramid knitted
structures were used.
Table 2: EN test method: the blade cutting edge degradation profile during the
course of the testing of high cut performance materials.
The analysis of the influence of PVC dots, melted and cured onto the knitted
surface to enhance the grip properties of the fabric or the gloves, was performed
using the EN and the ASTM procedures. This further illustrates some of the
specific features of the two methods. Fig. 5 provides the resulting data outlining
NL1, which is the load, in g, provoking the cut through of the specimen for a
blade motion of 25.4 mm.
The motion of the blade on the sample tested has been studied in terms of
speed using the ASTM and ISO blade calibration material, i.e. a 1.6 mm
neoprene layer, and 100% para-aramid felt, which characteristics are indicated in
fig. 6(a) and 6(b).
As depicted in fig. 7 a load cell, linked to a computer, was placed in the
balanced arm holding the blade in order to estimate the force opposing the blade
motion during the ASTM testing. Therewith the ratio of this resisting force to the
applied load force has been calculated for various materials, such as neoprene,
289
100% KEVLAR knitted gloves and another HCP material. Results are
summarized in fig. 8
Figure 5:
PVC dots influence on the EN (top table) and the ASTM results
(top table and curves).
Figure 6(a): Blade speed influence on the cut performance of the calibration
material.
Figure 6(b): Blade speed influence on the cut performance of a 100% paraaramid textile.
Figure 7:
CPPT (ASTM) equipped with a compressive load cell.
Finally, a series of experiments was run to estimate the similitude between

cutting and sliding process. In fig. 12 a sliding experimental assembly is depicted
and was made in order to estimate the sliding force. A large part of the force
applied onto a cutting device can be dissipated via the sliding friction, associated
sometime with the compression and the relaxation of the materials under cut.
Therefore it is important to relate the cut forces measured during cut with the
291
more easily accessible sliding and friction forces. It is not our intention to
pretend that a direct liaison can be established between these two physical
phenomena but, in view of the lack of published data on the subject matter, it is
valuable to bring another piece of comprehension to this scientific matter of
relevant value to personal protection. During an accidental cut the sliding
attributes of the personal protective equipment are determinant.
Figure 8:
F2 / F1 ratio for various materials submitted to ASTM testing.
Figure 9: Various knife cutting edge profiles and recommended uses.
Figure 10:
A proposed cut mechanism sequence: Densification-CompressionCut-Relaxation-Densification....
3.3 Results and discussion

The cut mechanism and related fundamentals are rarely described in the
literature. The large diversity of situations where cuts are involved and the rather
large number of physical phenomena superimposed in a cutting operation, render
the scientific approach further more difficult.
There are some qualitative reflection points, which may guide the novice and
help the reader through this section.
Materials, which are generally considered as cut resistant materials, are also
exhibiting:
a good tear resistance
a reasonable puncture resistance towards sharp but not too thin perforators
a certain resistance to abrasion or ability to fibrillate

In the cutting process itself one may consider the following as useful
considerations:
the amount of material per unit volume, which the cutting threat has to go
through
the deformability of this material and the tendency for this material to
maintain a high density under pressure still keeping its flexibility
the mobility of the fibres on which the blade is sliding; a revolving

material being more difficult to cut.
the blade dulling effect of the material being cut. Glass is a good example.
293
the eventual surface or bulk phase changes occurring during the cut either
due to local heating under friction or heating.
.
the water or liquid retention and absorbency of the protective barrier.
Fibres with skin-core structures are generally more protective. The thickness
of the skin is a positive attribute. Para-aramid fibres are cut resistant.
Rather brittle materials tend to be less effective. On the other hand ductility can
be a good attribute. Steel and glass tend to support these two statements.
As exhibited in fig. 9 the design of blade profiles, leaving apart the steel
selection and processing, remains an art. For example what are the criteria taken
into account to make an effective knife cutting edge to cut tomatoes versus bread
versus meat versus cheese, versus wood, versus textile, etc.
When one observes a few cutting edges, it seems, as depicted in fig. 10, that
the sequence of -1- material blockage-densification and compression followed by
-2- the effective sliding and cut of the previously compressed materials and -3the relaxation and separation of the already cut material is a sequence which
may be taken into account to better understand the selection of a large proportion
of cutting edge profiles. That densification-cut-relaxation sequence repeats itself
along the blade as the cutting operation proceeds. The threshold compression
ratio at which cut can progressively take place depends on the fibre mobility, the
deformation/cut speed, the local humidity as well.
3.3.1 Blade degradation during EN testing
Using the EN norm, the degradation effect of the materials on the blade cutting
edge was already demonstrated by Lara [21] with a series of test performed on
fibreglass reinforced glove samples. According to this study, a fraction of cycle
a cycle being represented by one blade full rotation counter-clockwise and its
reverse rotation clockwise - was initially sufficient to cut the cotton calibration
fabric. The same blade, after running for 50 revolution cycles on the HCP glasscontaining material, then cut the reference cotton fabric after 25 cycles compared
to 0.8 cycles initially. This shows that less than a cycle was necessary to dull the
blade.
Glass fibres are known for their dulling effect. A major concern shall be
outlined regarding products, which have a dulling effect. Shall they be
considered as effective high cut performance materials?
A partial answer is established by the fact that the ASTM/ISO norms are
more and more used for these types of materials. These norms being less
sensitive to the dulling effect since one spot of the blade is in contact with the
sample to cut only once. The EN norm recent revisions suggest the use of the
ISO testing in such cases. This is a significant step and improvement.
3.3.2 Speed effect on cut resistance
In which proportion the blade speed can affect one material more than another.
For example figs. 6(a), 6(b) compares the speed effect, on a neoprene rubber
layer and a KEVLAR sample stamped out of a knitted glove, for a speed
variation ranging from 14 to 28 mm/s.
Such a variation affects much more the neoprene load-to-blade displacement
curve than the load-to-blade displacement relationship of the para-aramid

samples. This may raise some concern about the 2.5 mm/s versus the 110 mm/s
blade displacement speed respectively selected for the ISO and the EN
procedures. The speed selected for the ASTM approach, i.e. 14 mm/s, and the
highest speed of 28 mm/s selected in the present study lies in between the ISO
and the EN selected values. Should those values be reconsidered and
harmonized? This question goes beyond the scope of this study.
Nonetheless, some additional data will be discussed later on and may
converge with this observation providing some elements of explanation. One can
already underline that in a textile structure, the fibres are relatively free to move,
especially when the textile is knitted which adds bulk to the structure. This
mobility can compensate for the blade displacement speed variation since the
fibre elements provoke an interface slippage.
Since these standards have been put together mostly for the testing of hand
protective equipment mostly made of HCP fibres such as para-aramids, the
results shown in fig. 6(b) tend to support the rather limited effect of speed within
the range tested. Although the emergence of dipped supported gloves may
provoke some more questioning of the speed effect since a dipped or coated
glove is no more than a textile structure covered by a rubber layer.
3.3.3 Coating/printing effect on the cut resistance of high performance
textile materials
The testing of gloves with PVC dots is reported in fig. 5. For the dotted side, the
level 3 given by the EN cut testing is not consistent with the 770 g load
necessary to provoke the cut through after a 25 mm blade displacement using
ASTM method. One would expect at least a load of 1100 g, i.e. 1.5 folds, to
reach an EN level 3. Furthermore, the non-coated side of the glove, i.e. without
PVC dots, reach an inferior level 1versus level 3 for the dotted side. This does
not seem to make sense in terms of cut performance.
For the ASTM analysis, one observes that, for the non-dotted side, the
necessary load to cut through after a 25 mm blade displacement is 820 g. This is
more in line with the expected results since the coating is not a cut resistant
material and is a spot by spot treatment. Overall in this specific case, the EN
testing largely over-estimates the real cut performance.
The EN testing does not appear to reflect the real cut performance of the
material. Since it tends to over-estimate it, precautions shall be taken regarding
the interpretation of the results. In terms of ASTM or ISO testing, if the
polymeric layer deposited on the textile substrate is continuous and interlocked
into this substrate, the cut performance is likely to be less than the substrate
alone since the fibres mobility is restricted. This is not the observation made
from fig. 5 since the PVC dotting is discontinuous.
The blade dulling, the speed effect and the interference of non-cut performing
polymeric coating have been discussed and may find some additional
explanations in the study, which follows.
3.3.4 Forces involved during cutting/sliding process
The apparatus designed to measure the cut performance as per the ASTM
procedure is shown in fig. 4 along with a schematic diagram, which provides the
295
blade orientation and displacement, and force application point. This force is
generated by weights placed on a plate mounted on a lever arm assembly in a
way that doubles the resulting force applied to the specimen. The test apparatus
also consists of a motor-driven balanced arm, which holds the blade.
In order to measure the co-linear force component resisting to the
displacement of the blade, a 50 N (or 200N) load cell, type S2 from HBM
Germany, was installed in the arm holding the blade as per the diagram of fig. 7.
This load cell was connected to a computer interface, module MP55 from HBM
Germany. The resisting force, F2 , was recorded during the cutting process
using the CATMAN software from HBM Germany. The normal force, F1 ,
applied onto the blade edge is assumed to be directly proportional to the double
weight force value, which is an approximation since during the displacement,
and due to the thickness of the sample, the blade cutting edge does not remain
strictly perpendicular to the sample contacting point. This is depicted in fig. 7.
Fig. 8 provides various F2 / F1 ratio.
For the neoprene calibration material the resisting force, F2 , is about 2.2
times superior to the normal applied force, F1 .

The resisting force found in this study is almost equal to the one, identified as
a frictional force, reported by Mass et al. [22] in his study of the basic principles
used in the development of a new cut-test machine now specified in the ISO test
method.
Therefore, our findings, although conducted independently on a different
testing apparatus, ASTM versus ISO, are fully in line with the work conducted
by Mass et al. [22] at the IRSST, Qubec Canada. This is a good demonstration
of the interchangeableness of the two equipments now both recommended in the
ASTM norm.
Fig. 8 also provides the F2 / F1 ratio for a KEVLAR knitted structure

and a ceramic coated para-aramid woven fabric.
The two ratios are more or less of the neoprene, measured ratio which may
bring a question regarding the suitability this rubber material in the ASTM and
ISO norms as a blade sharpness normalizing factor.
Furthermore, the amount of friction generated to cut trough neoprene may
explain why a dotted knitted structure exhibits an artificially high cut EN
performance. The necessity to use the blade cutting edge only once during a test
is essential in a situation where abrasion and resulting friction of the surface of
the tested material are high.
The apparatus designed to measure the sliding forces is shown in fig. 11 along
with a schematic diagram, very close to the fig. 7.
In this case the resisting force is generated in part by the weight placed onto
the specimen but also in part by the specimen surface characteristics inclusive of

its deformability under pressure. The resisting force is measured during the
linear displacement at a constant controlled speed equal to 4.2 mm/s.
Figure 11:
Figure 12:
Sliding test: (side view) with schematic representation (bottom) of

the force application, specimen displacement on calibrated
engineered surface.
F2 / F1 ratio for various in case of cut testing and sliding tests.
Figure 12 outlines the similitude of the forces involved in the case of cutting
versus the ones involved in the case of sliding. An amazing and revealing level
of concordance can be observed, which means that one could definitely relates,
to a certain extend, the surface wear and friction with the cut mechanism, which
is by far more complex. This goes beyond the scope of this study, although
297
during an accidental cut the sliding attributes of the personal protective

equipment are determinant.
High molecular weight polyethylene textile structures are known to change
their sliding and friction properties with moderate temperature as well as under
significant abrasion and wear while other materials such as para-aramids tend to
fibrillate bringing somehow and to a certain extend additional grip sometime to
the detriment of appearance.
Conclusion
High performance materials based on high performance fibres significantly

contribute to the emergence of new technologies under development today by
fibre producers as well as converters and end-users relying on the unique
attributes of para-aramids such as KEVLAR in diversifying applications.
Thermal and mechanical attributes reviewed, measured and partly discussed in
the previous part of this work are contributing to better protection. Their
scientific contain remains to be further explored.
Several tests methods, based on their equivalent standard procedures, were
reviewed beyond their normal application frames.
In term of harmonization, the considered test conditions reinforce the
appropriateness of the convergence towards a single norm, still allowing for
multiple choices at the equipment and procedure levels to best represent the
suitability of the material versus the expected cut protection level.
From this study one can appreciate as an improvement of the EN norm the
possibility to test as per ASTM or ISO norm procedure especially when abrasive
materials beyond level 3 are tested.
Similarly, the interchangeability of the ASTM and the ISO apparatus appears
more and more useful as the fundamental understanding of test condition limits
increases. Several situations studied in the present work support this aspect.
In terms of evolution, the emergence of high cut performance materials
reinforced with glass or steel fibres, the market progression of coated protective
equipments, the improvement of the test variability, and the substitution of the
calibration material seems to be good focus points for reflection in the near term.
Several conclusions drawn in the course of this study may serve as a basis.
The dulling effect of glass fibres towards the blade cutting edge can be
misleading in terms of cut performance analysis. Similarly the artificial cut
contribution of polymeric dots deposited on the textile substrate can also cause
difficulties in terms of the cut performance interpretation. Those two specific
situations are avoided when the blade cutting edge is in contact with the sample
only once during the testing. ASTM or ISO are therefore more favourable in
such cases.
A large amount of work remains to be done to fully integrate all the
components of the identified improvements in a global understanding of how the
cut mechanism and testing can evolve to be better predictive tools. Reference
[16] provides more insight into this study.

The present study has brought, hopefully, a refreshing scientific description of
the state of the art knowledge of high performance fibres and related protective
solutions with an effort to stimulate the desire to go beyond the knowledge and
applications of today. There are properties of the aramids, a high performance
family of fibres, which can still be optimised via the system approach. These
materials have the potential to participate to the solutions of tomorrow.
Acknowledgements
The authors wish to express their gratitude to Andr Miret-Casas for his
meticulous technical assistance and enthusiasm. Portia Yarborough, chairman of
the F23 ASTM committee brought valuable comments in her review. DuPont
fibre production centre in UK was instrumental to the independent evaluation
and analysis of the materials tested.
Special thanks are also due to Andr Courgey from the European Automotive
and Transportation Institute in France; his contribution to the wear and friction
study is especially recognised.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
J. W. S. Hearle, High Performance Fibres, Woodhead Publishing Limited

2001, ISBN 1 85573 539 3.
S. Rebouillat, M. Escoubes, R. Gauthier, J. Adhesion Sci. Technol., 10,
1996, pp. 635-650.
S. Rebouillat, M. Escoubes, R. Gauthier, A. Vigier, Polymer, 36, 1995,
pp. 4521-4523.
S. Rebouillat, M. Escoubes, R. Gauthier, A. Vigier, J. App. Poly. Sci., 58,
1995, pp. 1305-1315.
S. Rebouillat, J.B. Donnet, H. Guo, T.K. Wang, J. App. Poly. Sci., 67,
1998, pp. 487-500.
S. Rebouillat, J.B. Donnet, T.K. Wang, Polymer, 38, 1997, pp. 22452249.
S. Rebouillat, J.C.M. Peng, J.B. Donnet, Polymer, 40, 1999, pp. 73417350.
American National Standard for Hand Protection Selection Criteria
ANSI/ISEA 105-2000
S.L. Kwolek, Am. Inst. Chem., The Chemist, 57(11), 1980, 9
L. Penn, H.A. Newey, and T.T. Chiao, J. Mat. Sci, 11, 1976, 190.
H.M. Caesar, Twaron Products BV, Chemical Fibre International, 50,
2000, 161-164
S.L. Kwolek, P.W. Morgan, J.R. Schaefgen and L.W. Gulrich,
Macromolecules, 10, 1977, 1390
P.W. Morgan, Macromolecules, 10 (6), 1977, 1381.
T.I. Bair, P.W. Morgan and F.I. Killian, Macromolecules, 10, 1977, 1396.
S.L. Kwolek, US Patent Office, Pat No 3 671 542, 1972, Pat No 3 819
587, 1974.
[16]
[17]
[18]
[19]
[20]
[21]
[22]
299
P. Yarborough, C. N. Nelson, Performance of Protective Clothing, Global

Needs and Emerging Markets, ASTM International, STP1462, 2005.
Tejani, N., Blocker, R, Schiffelbein, P., and Rivet, E., Cut Protection
Performance Test for Measuring Cut Resistance of Materials Used in
Protective Clothing, Performance of Protective Clothing: Sixth Volume,
ASTM STP 1273, Jeffrey O. Stull and Arthur D. Schwope, Eds.,
American Society for Testing and Materials, 1997.
BS EN388, 1994, Protective gloves against mechanical risks
ASTM F1790-97, Standard test method for measuring cut resistance of
materials used in protective clothing
ISO13997, ISO / TC 94/SC 13/WG 5 N 173, Protective clothingMechanical properties- Determination of resistance to cutting by sharp
objects
Lara, J., Turcot, D., Daigle, R, and Payot, F., Comparison of Two
Methods to Evaluate the Resistance of Protective Gloves to Cutting by
Sharp Blades, Performance of Protective Clothing: Fifth Volume, ASTM
STP 1237, James S. Johnson and S. Z. Mansdorf, Eds., American Society
for Testing and Materials, 1996.
Mass, S., Lara, J., Sirard, C., and Daigle, R., Basic Principles Used in
the Development of a New Cut-Test Apparatus for Standardization,
Performance of Protective Clothing: Sixth Volume, ASTM STP 1273,
Jeffrey O. Stull and Arthur D. Schwope, Eds., American Society for
Testing and Materials, 1997.
301
The study of surface oxidation of tin(II)

fluoride and chloride fluoride materials
by Mssbauer spectroscopy:
to oxidize or not to oxidize, that is the question
G. Dns, E. Laou, M. C. Madamba & A. Muntasar
Laboratory of Solid State Chemistry and Mssbauer Spectroscopy,
Department of Chemistry and Biochemistry, Concordia University,
Montral, Canada
Abstract
Experimental methods designed to study the bulk of materials do not necessarily
detect the changes taking place at the surface of the crystallites. For example,
divalent tin-containing materials appear to be stable at ambient conditions in air,
provided they are not hygroscopic, X-ray powder diffraction shows only the
peaks of the expected tin(II) phase. However, we have observed that the
Mssbauer spectrum of polycrystalline samples contain, in addition to the
expected tin(II) peak(s), a small peak at 0 mm s relative to CaSnO3 at ambient
conditions, that can be attributed only to tin(IV) coordinated by oxygen. A
detailed study of this phenomenon has shown that Mssbauer spectroscopy is
quite sensitive for detecting thin layers of oxide at the surface of crystallites of
tin(II). This phenomenon has been exploited for the study of spontaneous
oxidation of various tin(II) fluoride and chloride-containing materials, some of
these fluorides being the highest performance fluoride-ion conductors known to
date. It was observed that passivation is quite efficient in the fluorides, and in
the chloride fluorides that have all their tin(II) covalently bonded. On the other
hand, the materials containing a mixture of covalently bonded tin(II) and the Sn2+
stannous ion namely, the Ba1-xSnxCl1+yF1-y solid solution, show a higher rate of
oxidation, which is highly dependent on the method of preparation and the
composition parameters, x and y.
Keywords: passivation, oxidation, disordered phases, fluorite-type structure,
Mssbauer spectroscopy, X-ray diffraction, ionic conductivity, BaClF.
doi:10.2495/HPSM06029
Introduction
The MF2 fluorides of large divalent metals (M = Ca, Sr, Ba & Pb) have the well
known fluorite type structure, with a cubic unit-cell, space group Fm3m, and a
cubic coordination of the metal ion (fig. 1) [1]. This contrasts with SnF2 tin(II)
fluoride (stannous fluoride) that has a molecular structure; it is made of Sn4F8
tetramers and the Sn-F bonds are strongly covalent [2]. When SnF2 and MF2 are
combined together, some new materials are produced, some of which are
ordered, others disordered. The MSnF4 compounds have a M M Sn Sn
order parallel to the c axis of the unit-cell (fig. 1) [3]. Tin and lead belong to
group 14, and therefore they have four valence electrons, and in the +2
suboxidation state, the 5s2 (Sn) or 6s2 (Pb) electrons are unused and form a nonbonding electron pair, also called a lone pair. When bonding is ionic, the
orbitals are not hybridized and the lone pair is located on the ns native orbital
that has spherical symmetry, unless it is distorted by polarization, and a quite
regular coordination is observed, and the lone pair is said to be non-stereoactive
since it does not modify the coordination. In fluorite-type PbF2, lead has a
cubic coordination, therefore its lone pair is located on the native 6s orbital and
bonding is ionic, like in BaF2 (fig. 1). In tetragonal MSnF4, the coordination of
M is a distortion of the MF8 cube of the fluorite structure, with in addition, four
longer interactions, to form an overall MF4F4F4 where the bond length
increases as follows: M-F (from M-F-M bridges) < M-F (from M-Feq-Sn
bridges) < M-F (from M-Fax-Sn bridges), where Feq and Fax form equatorial (4)
and axial (1) bonds with tin, respectively (fig. 1). For the sake of clarity, the BaF bonds are not shown on figure 1. The lead coordination in PbF2, is still
ionic since Sr2+ and Ba2+ take the same coordination in SrSnF4 and in BaSnF4,
respectively. In contrast, tin forms covalent bonding in MSnF4, with a very short
axial Sn-F bond, and the lone pair is located on a hybrid orbital, in transposition to F (fig. 1). The tin lone pair is said to be stereoactive since it changes
drastically the tin stereochemistry (lower coordination number, highly distorted
coordination). A consequence of the M/Sn order and of the stereoactivity of the
tin lone pair is that the lone pairs cluster in sheets that are cleavage planes, which
make the structure very highly two-dimensional. Examples of disordered
structures are PbSn4F10, the M1-xSnxF2 solid solution (M = Ca and Pb), and MSnF4 (M = Ba and Pb) obtained by ball-milling. The disordered phases have
an Fm3m cubic unit-cell like the fluorite-type MF2, and their diffraction pattern
shows no peak splitting or additional low angle peak that would indicate the
presence of a lattice distortion or a superstructure. Therefore, Sn and the other
metal M are fully disordered on the metal ion site of the MF2 structure (fig. 2).
The M bonding is ionic and the coordination is cubic with some disordered local
distortions in the neighborhood of tin, while tin bind to only one of the faces of
the F8 cubes, with its stereoactive lone pair pointing towards the center of the
cube, and thus its bonding is covalent. BaClF crystallizes in the PbClF structure
(space group P4/nmm). Its structure is also a tetragonal distortion of the fluoritetype BaF2, with layers of F and Cl ordered parallel to the c axis of the unit-cell,
making it a layered structure, in contrast with the three-dimensional cubic
303
structure of BaF2 (fig. 1). We have recently found two methods for substituting
substantial amounts of Ba by Sn in BaClF, and also some F by Cl (y>0), or some
Cl by F (y<0), making it a doubly disordered Ba1xSnxCl1+yF1-y solid solution. Xray diffraction shows that Ba and Sn are fully disordered on the Ba site of the
BaClF structure. In addition, while the two anionic sites remain ordered, for y>0
the y Cl in excess are disordered with the remaining (1y) F on the F site,
whereas for y<0, the (y) excess F are disordered with the remaining (1+y) Cl on
the Cl site. However, the electronic situation at tin cannot be assessed by X-ray
diffraction. All the tin(II)-containing fluorides studied here are high performance
fluoride ion conductors (about three orders of magnitude higher than the
conductivity of the corresponding MF2).
Figure 1:
Projection of the crystal structure of BaSnF4, BaF2 and BaClF.
All these materials appear stable in air at ambient conditions, and X-ray
diffraction shows no sign of oxidation to tin(IV). The present study shows that
tin-119 Mssbauer spectroscopy is an excellent method for detecting minute
amounts of tin(IV) oxide in tin(II) compounds, and that all materials show at the
minimum some surface oxidation, and some passivate very well while other do
so less efficiently.
Experimental
Fluoride materials synthesis was carried out either in aqueous medium or at high
temperature under dry conditions. Tetragonal -PbSnF4 was obtained by
precipitation on addition of an aqueous solution of lead(II) nitrate to a solution of
SnF2. Ca1-xSnxF2 precipitated when a solution of calcium nitrate was added to a
solution of SnF2 at high Ca/Sn ratio, or by leaching, on stirring, CaSn2F6 in a
large amount of water. The synthesis of Pb1-xSnxF2 and BaSnF4 was carried out
under nitrogen in sealed copper tubes heated at 500 oC and quenched. Barium

tin(II) chloride fluorides were prepared by reaction of barium nitrate and tin(II)
fluoride in aqueous solutions versus the X = BaCl2/(SnF2 + BaCl2) molar ratio.
Three new materials were obtained: BaSn2Cl2F4 (0.25<X<0.35),
BaSnClF3.0.8H2O (0.50<X<0.80), and a material that has the diffraction pattern
of BaClF (X>0.87). Chemical analysis showed that it is a Ba1-xSnxCl1+yF1-y solid
solution (x = 0.15, y = 0.11). A very similar BaClF phase was obtained by
synthesis in dry conditions when mixtures calculated according to eqn (1) were
heated three days under nitrogen in sealed copper tubes at temperatures varying
from 350 to 800 oC.
1 2x y
2
Figure 2:
BaF2 + xSnF2 +
1+ y
2
BaCl2 Ba1 x Sn xCl1 + y F1 y
(1)
Model of a disordered fluorite type M1-xSnxF2 structure.
X-ray powder diffraction was carried out on a PW-1050/25 Philips

diffractometer automated with the SIE RAY 122 system from Diffraction
Technology, using the K radiation of copper (K1 = 1.54051 ) and a
monochromator. Mssbauer spectra were recorded in a TN7200 multichannel
analyzer from Tracor Northern, using an Elscint driving system, a Harshaw
(Tl)NaI detector and a 10 mCi CaSnO3 -ray source. All chemical isomer shifts
were referenced relative to a standard CaSnO3 absorber. All spectra were
collected at ambient temperature and fitted using the GMFP software. Samples
were studied when young (freshly prepared) and old (after several years storage
in air at ambient temperature).
305
Transmission
1
0 .9 6
0 .9 2
(a )
0 .8 8
Transmission
0 .9 9 5
0 .9 7 5
0 .9 5 5
(b )
0 .9 3 5
-8
-6
-4
-2
Ve l o c i t y ( m m /s )
Figure 3:
Ambient temperature 119Sn Mssbauer of BaSnF4 prepared by the

dry method at 500 oC for 2 hours: (a) young sample, (b) sample
aged for 69 months at ambient conditions.
3.1 Oxidation of the fluorides

Various ordered and disordered tin(II)-containing fluorides were checked for
oxidation, using X-ray diffraction and Mssbauer spectroscopy. Figure 3 shows
the spectrum of BaSnF4 when freshly prepared and after 69 months of age. Both
spectra have a large quadrupole doublet in the 2-4 mm/s velocity range, which is
characteristic of the presence of a stereoactive lone pair on tin, and confirms that
bonding is covalent (fig. 1). In addition, a weak peak is observed at ca. 0 mm/s.
The peaks positions are given relative to a tin(IV) oxide standard, CaSnO3, and
therefore, the peak at 0 mm/s in BaSnF4 shows some tin(IV) oxide is contained
in the ample, even when freshly prepared. This can be assigned to Sn(II) to
Sn(IV) oxidation of the surface of the solid particles by atmospheric oxygen.
After aging 69 months in air at ambient conditions, the tin(IV) signal at ca. 0
mm/s has not increased significantly, therefore no much further oxidation
occurred during the nearly 6 years of aging. It is therefore established that there
is an initial very rapid surface oxidation of the particles that creates a passivating
layer, which protects the bulk of the particles against further oxidation. X-ray
diffraction does not show any peak foreign to BaSnF4, and more particularly no

SnO2 peak. This suggests that either the amount of SnO2 is too small to be
detected by diffraction, or the oxide is amorphous, or the particles are too small
to diffract. This establishes the superiority of Mssbauer spectroscopy to detect
small amounts of oxide in tin(II) halides. The relative intensity of Mssbauer
peaks is a function of both the amount of each kind of tin and their recoil-free
fraction fa. The recoil-free fraction is the fraction of 119Sn nuclei that absorb
rays without dissipating the recoil energy through phonons (lattice vibrations).
The stronger the lattice, the higher the recoil-free fraction, and the stronger the
Mssbauer peaks at equal amount of tin. Since Sn(IV) forms very strong bonds
to oxygen in the highly tight rutile structure of SnO2, it has a high recoil-free
fraction and gives a disproportionably high signal, in comparison with small
recoil-free fraction tin(II) halides, that have much weaker bonds. It results that
the fraction of SnO2 in a tin(II) halide is more than ten times smaller than its
relative signal. Therefore, the amount of SnO2 in BaSnF4 is much smaller than
the intensity of it signal suggests. Disordered systems, even very poor in tin, like
the M1-xSnxF2 at small tin content (x = 0.27 for M = Ca, x = 0.10 for M = Pb)
also show no SnO2 diffraction peak (fig. 4Ab), and the tin(IV) Mssbauer peak is
barely detectable (fig. 4Cb), showing excellent passivation.
(A)
Figure 4:
(B)
(C)
(A) X-ray powder diffraction pattern of CaF2 and Ca1-xSnxF2 (x =

0.27), (B) X-ray powder diffraction pattern of BaClF and
Ba1xSnxCl1+yF1-y (x = 0.15, y = 0.094), (C) 119Sn ambient
temperature Mssbauer spectrum of Ba1-xSnxCl1+yF1-y (x = 0.15, y
= 0.094) and Ca1-xSnxF2 (x = 0.27).
307
3.2 Oxidation of the chloride fluorides

The stoichiometric chloride fluorides, BaSn2Cl2F4 and BaSnClF3.0.8H2O have a
behavior similar to that of the fluorides. Freshly prepared samples have a tin(IV)
oxide peak, and therefore there is a rapid initial surface oxidation (fig. 5), and it
also does not show any significant increase after aging, therefore efficient
passivation occurs. X-ray diffraction does not detect the oxide. The stronger
tin(IV) Mssbauer signal of BaSn2Cl2F4, in comparison to BaSnClF3.0.8H2O,
does not mean that the former compound contains more SnO2. It could just
mean that tin(II) in BaSn2Cl2F4 has a lower recoil-free fraction than in
BaSnClF3.0.8H2O. Very low temperature spectra would be required to determine
the reason for their unequal tin(IV) peak intensity.
Transmission
(a)
0.93
0.86
Transmission
0.97
(b)
0.94
0.91
0.88
0.85
-8
-6
-4
-2
Ve locity (mm/s)
Figure 5:
Ambient temperature 119Sn Mssbauer spectrum of: (a) BaSn2Cl2F4

and (b) BaSnClF3.0.8H2O.
The behavior of the non-stoichiometric Ba1-xSnxCl1+yF1-y is much more

complex. The Mssbauer spectrum of the precipitated solid solution is made of
a single tin(II) line at ca. 4 mm/s, characteristic of ionic bonding, and of a tin(IV)
oxide peak at ca. 0 mm/s (figs. 4Cb & 6). Like in the materials above, no SnO2
could be detected by X-ray diffraction (fig. 4Bb). Contrary to the fluorides and
the stoichiometric chloride fluorides, a significant increase of the intensity of the
tin(IV) oxide Mssbauer peak takes place on aging for 4 years. Therefore,

although the material is still quite well protected against oxidation, it is not fully
passivated.
Transmission
0.99
0.98
0.97
0.96
0.95
(a)
(d)
(AM-234, X = 0.860, aged for 2.5 m)
0.94
0.93
-8
-6
-4
-2
Velocity (mm/s)
Transmission
0 .9 8 5
0 .9 6 5
( A M - 2 3 4 , X = 0 .8 6 0 , a g e d f o r 4 8 .0 m )
0 .9 4 5
(b)
0 .9 2 5
-8
-6
-4
-2
V e l o c i t y i n ( m m /s )
Figure 6:
Ambient temperature 119Sn Mssbauer spectrum of the precipitated

Ba1xSnxCl1+yF1-y versus age: (a) 2.5 month old, (b) 48.0 months
old.
The rate of oxidation of the Ba1-xSnxCl1+yF1-y solid solution prepared by

high temperature direct reactions in dry conditions is highly dependent on the x
and y composition parameters (fig. 7). At low tin content (low x), freshly
prepared samples have a tin(IV) oxide Mssbauer peak quite stronger than the
tin(II) peak (fig. 7a1). This shows that initial resistance to oxidation is much
weaker than in the low tin content fluoride solid solutions. For example, the
tin(IV) peak of Pb1-xSnxF2 at low tin content (x = 0.10) is barely detectable.
After less than three years aging, the tin(IV) peak of figure 7a1 has much
increased and the tin(II) peak is now a minority, therefore passivation is much
poorer. When the tin content increases to x = 0.121 at y = 0 (Cl/F = 1), the
309
tin(IV) peak is weaker than the tin(II) peak, however, it is stronger than that of
stoichiometric chloride fluorides.
(AM-1180, x = 0.05, y = 0.10)
(AM-1180, x = 0.05, y = 0.10)
0.999
0.994
Transmission
Transmission
0.995
0.99
0.985
0.984
(a1)
(a2)
0.98
0.979
(AM-315, x = 0.121, y = 0.000)
AM-315, x = 0.121, y = 0.00
0.99
0.99
0.98
0.98
Transmission
Transmission
0.989
0.97
0.96
(b1)
0.95
0.97
0.96
(b2)
0.95
0.94
0.94
(AM-1155, x = 0.225, y = 0.2, RT aged)
(AM-1155, x = 0.225, y = 0.200)

0.998
Transmission
Transmission
0.995
0.988
(c 1)
0.99
(c 2)
0.985
0.98
0.978
-8
-6
-4
-2
-8
-6
-4
Velocity (mm/s)
Figure 7:
119
-2
Velocity (mm/s)
Sn Mssbauer spectra of Ba1-xSnxCl1+yF1-y at RT, prepared by

direct reactions under dry conditions (freshly prepared and aged):
(a1 & a2) AM-1180, x = 0.050, y = 0.10 [freshly prepared (a1) and
aged for 32 months (a2)], (b1 & b2): AM-315, x = 0.121, y = 0.000
[freshly prepared (b1) and aged for 36.7 months (b2)], (c1 & c2):
AM-1155, x = 0.225, y = 0.200 [freshly prepared (c1) and aged
31.0 months (c2)].

Upon aging to three years, the tin(IV) peak dominates slightly. This shows
that both the oxidation is slower at higher tin content, and passivation is
somewhat better, although not great. Near the upper limit of the solid solution in
tin (x = 0.225) and at high chlorine content (y = 0.200), the tin(II) signal
predominates by far, however, it is also very weak, which shows that its recoilfree fraction is extremely small. After aging two and a half years, the tin(IV)
signal is twice as strong as the tin(II) signal, therefore initial oxidation was not
very large, however, passivation was not very good.
These results can be understood, at least in part, from the bonding model of
tin in the solid solution. At low x, Sn is diluted in the solid and there is little
chance of a Sn being neighbor to another Sn. The coordination of an isolated Sn
is mostly determined by the size of the Ba site it occupies. Since Ba2+ is quite
larger than Sn2+, the latter is loose in its site, i.e. its bonds are weak, resulting in a
weak recoil-free fraction and fast oxidation (figs. 7a1 & 7a2). At higher tin
content, the rest of the solid network has to accept adjustments to satisfy the tin
bonding requirements, which allows for stronger bonds, hence a higher recoilfree fraction and some improvement of passivation (figs. 7b1 & 7b2). However,
at high x and high y, the excess chlorine, probably clusters around tin than
around barium because there is more room around the smaller tin. However,
since Sn-Cl bonds are weaker than Sn-F bonds, it results in a decrease of recoilfree fraction and passivating power (figs. 7c1 & 7c2).
Conclusion
Mssbauer spectroscopy was an essential method for understanding the complex

bonding of tin(II) and its oxidation to tin(IV) in the fluorides and chloride
fluorides.
Acknowledgements
This work was made possible by the support of Concordia University and the
Natural Science and Engineering Research Council of Canada. Grateful thanks
are also due to the Procter and Gamble Co. (Mason, Ohio) for supporting our
Mssbauer laboratory.
References
[1]
[2]
[3]
Rayner-Canham, G. & Overton, T., Descriptive Inorganic Chemistry, 3rd

ed., Freeman, New York, pp. 78-80, 2002.
Dns, G., Pannetier, J., Lucas, J. & Le Marouille, J.Y., About SnF2
stannous fluoride. I. Crystallochemistry of -SnF2. J. Solid State Chem.,
30(3), pp. 335-343, 1979.
Birchall, T., Dns, G., Ruebenbauer, K. & Pannetier G., A neutron
diffraction and 119Sn Mssbauer study of PbSnF4 and BaSnF4. Hyperf. 29,
pp. 1331-1334, 1986.
311
Study of double ionic disorder (cationic and

anionic) and disorder of two kinds of tin(II)
(ionic and covalent) within the same material:
the incredibly complex Ba1-xSnxCl1+yF1-y
solid solution and its study by
Mssbauer spectroscopy
S. Boufas2 G. Dns1, J. Kochuparampil1, H. Merazig2 &
A. Muntasar1
1
Laboratory of Solid State Chemistry and Mssbauer Spectroscopy,

Department of Chemistry and Biochemistry, Concordia University,
Montral, Canada
2
Laboratoire de Chimie Molculaire, du Contrle de lEnvironnement et
des Mesures Physico-Chimiques, Dpartement de Chimie, Universit
Mentouri, Constantine, Algeria
Abstract
A doubly disordered Ba1-xSnxCl1+yF1-y solid solution has been prepared. X-ray
diffraction shows that Ba and Sn are fully disordered on the Ba site of the BaClF
structure. In addition, while the two anionic sites remain ordered, for y>0 the y
Cl in excess are disordered with the remaining (1-y) F on the F site, whereas for
y<0, the (-y) excess F are disordered with the remaining (1+y) Cl on the Cl site.
119
Sn Mssbauer spectroscopy showed the Ba1-xSnxCl1+yF1-y solid solution is a
truly unique material. Indeed, in addition to the rarity of being a doubly
disordered solid solution over a wide range of chemical composition, it also
exhibits a variable mixture of ionic tin(II) (Sn2+ stannous ions, with a
non-stereoactive lone pair) and covalently bonded tin(II) (with a stereoactive
lone pair). A new material, Ba2SnCl6, has been prepared and its crystal
structure solved. It contains octahedrally coordinated Sn2+, with a perfectly
non-stereoactive lone pair, for the first time in a chloride. This makes the lone
pair a potential charge carrier.
Keywords: disordered phases, BaClF structure, Mssbauer spectroscopy, X-ray
diffraction, mixed conductivity.
doi:10.2495/HPSM06030
Introduction
BaClF crystallizes in the PbClF structure. Its structure is a tetragonal distortion

of the fluorite-type BaF2, with alternating layers of F and Cl ordered parallel to
the c axis of the unit-cell (fig. 1), making it an anisotropic structure, in contrast
with the isotropic cubic structure of BaF2 (fig. 1) [1]. The Ba2+ ions are bonded
to four fluoride ions at the corners of a square, like half of its cubic coordination
in BaF2. An apical Ba-Cl bond is located in trans-position relative to the fluorine
square, with in addition four chloride ions at the corners of a square parallel to
the F4 square and rotated by 45o parallel to (a,b). It results that the barium
coordination in BaClF is a BaF4ClCl4 monocapped square antiprism, in contrast
with the BaF8 cube found in BaF2. The BaClF structure is the result of F Cl
order of the fluorite structure parallel to the c axis of the unit-cell. An alternate
tetragonal distortion of the fluorite structure, also with order parallel to the c axis
is the BaSnF4 structure, with the metals being ordered (Ba Ba Sn Sn). The
structure of BaSnF4 is very strongly two-dimensional (fig. 1).
In the present work, the reaction of barium chloride and tin(II) fluoride in
aqueous solutions, at a high BaCl2/SnF2 ratio, gives a precipitate that appears at
first glance, from its X-ray diffraction pattern, to be BaClF. A similar phase,
although not completely identical, was obtained by high temperature reaction of
a mixture of BaF2, BaCl2 and SnF2 in dry conditions. The detail investigation of
the material obtained, particularly the study of tin by Mssbauer spectroscopy, is
presented here. In addition, the first perfectly ionic tin(II) in a chloride fluoride
has been discovered in a new compound, Ba2SnCl6.
Figure 1:
Projection of the crystal structure of BaSnF4, BaF2 and BaClF.
Experimental
The BaClF phase precipitate was obtained when aqueous solutions of barium
chloride and tin(II) fluoride were mixed, at a X = BaCl2/(SnF2 + BaCl2) molar
313
ratio above 0.85, when barium chloride was added to tin(II) fluoride (BaSn),
or the other way around (SnBa). Synthesis under dry conditions was
performed when mixtures calculated according to eqn (1) were heated for three
days under nitrogen in sealed copper tubes at temperatures varying from 350 to
800 oC.
1 2x y
2
BaF2 + xSnF2 +
1+ y
2
BaCl2 Ba1 x Sn xCl1 + y F1 y
(1)
Crystals of Ba2SnCl6 were collected from a batch obtained by slow evaporation

of an aqueous solution of BaCl2.2H2O and SnCl2.2H2O. Chemical analysis was
carried out by use of a Perkin Elmer 503 atomic absorption spectrometer (Ba and
Sn), an Orion Research Laboratories 94-09 fluoride ion selective electrode (F),
and by argentometry according to the Fajans titration method (Cl). The analytical
methods required substantial adjustments in order to eliminate interferences
created by the presence of tin and fluorine [2]. X-ray powder diffraction was
carried out on a PW-1050/25 Philips diffractometer automated with the SIE
RAY 122 system from Diffraction Technology, using the K radiation of copper
(K1 = 1.54051 ) and a monochromator. Identification to known phases was
done by Search-Match using the PDSM software from Feinn-Marquat and the
JCPDS PDF1 database from the ICDD. Mssbauer spectra were recorded in a
TN7200 multichannel analyzer from Tracor Northern, using an Elscint driving
system, a Harshaw (Tl)NaI detector and a 10 mCi CaSnO3 -ray source from
Ritverc. All chemical isomer shifts were referenced relative to a standard
CaSnO3 absorber. The spectra were fitted using the GMFP5 software.
The X-ray powder diffraction pattern of the precipitate (fig. 2b) has the same set
of peaks as BaClF (fig. 2c), at the same positions, and therefore, it seems that the
product obtained is simply BaClF. This seemed confirmed by Search-Match
that resulted in BaClF being the only reasonable match. Although the match was
reasonably good (Similarity Index SI>150 and all 15 peaks matched), a close
examination of the peak position revealed a small, but significant shift, shown by
arrows on figure 3a, relative to unsubstituted BaClF (fig. 3b). The shift of all
peaks is consistent with a 0.91% decrease of the a parameter and a 0.97%
increase of c, while the unit-cell volume decreases by 0.86% and the c/a
tetragonal distortion increases by 1.89%. The change of unit-cell parameters
suggests a change of chemical composition. A logical change was the
substitution of a fraction of Ba by Sn, to give non-stoichiometric Ba1-xSnxClF.
Chemical analysis was performed to determine the tin molar fraction x, and it
showed that the compound is also non-stoichiometric for the anions, such that
the formula is best written as Ba1-xSnxCl1+yF1-y (x = 0.15, y = 0.11), to show it is
a substituted BaClF phase, as shown by X-ray diffraction.
(a)
220
212
104
211
202
103
003
112
200
102
110
101
Intensity
001
002
Intensity
(b)
(c)
5
10
15
20
25
30
35
40
45
50
55
60
Deg. 2 theta
Figure 2:
X-ray diffraction of (a) Ba1-xSnxCl1+yF1-y prepared by the dry

method (x = 0.150, y = 0.050, 43 hours at 500 oC), (b) precipitated
Ba1-xSnxCl1+yF1-y (X = 0.923), (c) BaClF precipitated by addition
of HF(aq) to a solution of BaCl2.2H2O.
Preparations carried out by the dry method yielded a material that also gives
the BaClF diffraction pattern (fig. 2a). A wide double non-stoichiometric solid
solution was obtained: 0x0.25, -0.15y0.25. The inversion of the relative
intensities of the (002) and (110) peaks could be in part due to the change of
chemical composition. It is likely that it is mostly due to preferred orientation
favoring the (00l) peaks in the latter, i.e. the material tends to cleave parallel to
(a,b). This is consistent with the crystal structure (fig. 1): the long Ba-Cl bonds
(Cl atoms at the corners of a square) are the weak links. In addition, figure 3
shows that the peak shift in the solid solution prepared by the dry method is
negligible, and it was found it changes very little with the composition
parameters x and y. The different unit-cell parameters of the solid solution
prepared by the two methods shows that something differentiates them from one
another, although both are doubly disordered BaClF structures. For these
315
110
110
101
(c)
(a)
20
002
101
Intensity
(b)
110
002
101
Intensity
002
disordered phases, only average Ba/Sn values can be obtained by diffraction

alone. Could the difference be due to the way tin binds to the BaClF network?
25
30
Deg. 2 theta
Figure 3:
Enlargement of a portion of the X-ray powder diffraction pattern

of: (a) Ba1-xSnxCl1+yF1-y prepared by the dry method (x = 0.150, y =
0.050, 43 hours at 500 oC), (b) BaClF precipitated by addition of
HF(aq) to a solution of BaCl2.2H2O, (c) precipitated Ba1-xSnxCl1+yF1y (X = 0.923).
(b)
Figure 4:
Electrical monopolar and electrical quadrupolar interactions

generated by different kinds of divalent tin, and their effect on the
Mssbauer spectrum: (a) unhybridized lone pair (e.g.
semiconducting CsSnBr3), (b) hybridized lone pair (e.g. BaSnF4),
(c) unhybridized lone pair in the case of Ba1xSnxCl1+yF1-y
discovered in the present work.
A selective probing of the tin sites, that ignores barium, is required.

Mssbauer spectroscopy of the 119Sn nuclide probes the tin sites only, regardless
of crystallinity or order/disorder. Figure 4 shows the expected Mssbauer
spectrum for the modes of bonding of divalent tin. Tin belongs to group 14 and
period 5 of the periodic table of the elements, and therefore it has 4 valence
electrons. It results that it forms two stable oxidation states, +2 and +4. In the
suboxidation state +2, tin has a non-bonded electron pair (lone pair). Two types
of bonding can be found in halides: ionic bonding (most iodides and some
bromides) or covalent bonding (all fluorides and chlorides, and most bromides).
When bonding is ionic, the tin lone pair is located on the 5s native orbital (nonstereoactive lone pair). Figure 4a shows that in the case of ionic bonding in an
undistorted coordination, the 5s orbital is perfectly spherical and there is no
317
electric field gradient (efg) at the Sn2+ site. It results no splitting of the nuclear
levels, and therefore only one nuclear transition takes place (1/2 3/2), and a
single Mssbauer absorption line is observed. This contrasts with the case of
covalent bonding, where the tin lone pair is located on one of the hybrid orbitals
(stereoactive lone pair). This lone pair is highly axial and therefore generates a
large efg, (Vzz)val. In addition, the stereoactive lone pair distorts very strongly
the tin polyhedron of coordination, which generates another term of the efg,
(Vzz)latt. The overall effect is the splitting of the first excited state which results
in two nuclear transitions, and the Mssbauer spectrum is a doublet, with a large
quadrupole splitting (fig. 4b). In addition, the higher the valence s electron
density acting at tin, the higher the isomer shift , i.e. the line shift relative to a
CaSnO3 standard. As can be seen from figure 4, this makes the isomer shift of
Sn2+, 4.0-4.1 mm/s (5s electron density concentrated around the tin atom) higher
than that for Sn(II)cov, 2.5-3.5 mm/s (5s electron density pulled away in hybrid
orbitals).
The ambient temperature spectrum of precipitated Ba1-xSnxCl1+yF1-y is a
single line at 4.07 mm/s (fig. 5a1) which shows that tin is present in the form of
Sn2+ stannous ions. For the Ba1-xSnxCl1+yF1-y solid solution prepared in dry
conditions, the Mssbauer spectrum is a function of the x and y composition
parameters. For a high tin content (x = 0.225), the spectrum is also a single line
at ca. 4 mm/s if Cl/F>1 (y = 0.25), and therefore bonding is ionic (fig. 5b1). For
the same x value and Cl/F<1 (y = -0.15), the tin(II) region of the spectrum is a
very highly asymmetric doublet, which is actually due to the sum of a doublet
and a single line. Since for fluorides, the high velocity line of a doublet is at
about 4 mm/s, it adds to the single line of ionic tin to makes a spectrum that
looks like a strongly asymmetric doublet. Therefore, at high tin content, and
when there is more fluorine than chlorine, the Ba1-xSnxCl1+yF1-y solid solution
contains a mixture of Sn2+ ions and covalently bonded tin(II). The line at ca. 0
mm/s in the spectra of figures 5c1 and 5c2 is due to a tin(IV) impurity generated
by surface oxidation of the particles. This is common in tin(II) compounds. The
ambient temperature spectra have a highly unequal intensity, and strangely
enough, the poorest in tin is the strongest (fig. 5a1). In addition, the spectrum of
figure 5b1 is much weaker than that of figure 5c1, even though they contain
exactly the same amount of tin. The line intensity is a function of the recoilless
fraction fa, which is itself a function of bond strength.
All the Mssbauer observations make it possible to build a model of tin
bonding in Ba1-xSnxCl1+yF1-y solid solution: (i) at low tin concentration (small x),
like in the precipitated solid solution, due to dilution tin is isolated from other tin
atoms, the solid network is determined by the size of the Ba2+, Cl- and F- ions,
and since tin occupies a barium site, it takes the same mode of bonding, i.e. ionic
(figs. 6a); (ii) at high tin content (x = 0.25), the tin concentration is high enough
to have some tin clustering, and in an excess of fluorine (y = -0.15)each tin atom
is connected by bridging fluorine, and this result in covalently bonded tin with a
stereoactive lone pair (fig. 6b). The sample of figure 5c1 contains a mixture of
the two models of figures 6a and 6b. However, at the same tin concentration (x
= 0.25), in the presence of an excess of chlorine (y = 0.25), the excess Cl is likely

to coordinate tin where there is more room than around barium, due to the
smallest size of tin, and this probably creates a situation similar to anhydrous
SnCl2. Mssbauer spectroscopy has shown that ionic bond is present in SnCl2.
This is probably the reason for the single Sn2+ line of figure 5b1. The weaker SnCl bonds, compared to Sn-F bonds, are the reason for the very weak lattice
(extremely weak line intensity at ambient temperature).
1
0.93
0.98
Transmission
Transmission
0.98
(X = 0.896, Ba ---> Sn, T = RT)

0.96
X= 0.896,
Ba->Sn, T = RT
0.88
X= 0.896,
Ba->Sn, T = 17K
(X = 0.896, Ba ---> Sn, T = 17 K)
(a)
(a1)
(a
)
0.94
(a(b)
2)
(a
0.83
0.78
-8
-6
-4
-2
Velocity (mm/s)
Transmission
0.96
T r a n s m is s io n
0.995
0.99
(b1)
(x = 0.225,
0.225, y = 0.25, T
x=
y = 0.250, T = RT
(b1)
0.92
x= 0.225,
(x
y == 0.225,
0.250,y = 0.25, T
= 17 K)
T = 17K
0.88
0.84
0.8
-8
0.985
(b(b22))
-6
-4
-2
Velocity (mm/s)
0.995
T r a n sm issio n
0 .9 5
0.99
0.985
0 .9
x =x=
0.225,
y = - 0.15, T = RT
0.225,
y = -0.150,
T = RT
0.98
0.975
(c1)
(x =x=
0.225,
y = - 0.15, T= 17
0.225,
K)
y = -0.150,
T = RT
0 .8 5
(c1)
(c2)
(c2)
0 .8
0.97
-8
-6
-4
-2
-8
-6
-4
Velocity (mm/s)
Figure 5:
119
-2
V e lo c it y ( mm/ s )
Sn Mssbauer spectrum at 293 K (a1, b1, c1) and 17 K (a2, b2, c2)
of samples: (a1 & a2) AM-1273 (precipitated at ambient conditions,
X = 0.896, BaSn), (b1 & b2) AM-1270 (dry method, x = 0.225, y
= 0.250), (c1 & c2) AM-1148 (dry method, x = 0.225, y = -0.150).
319
(a)
(b)
Figure 6:
Bonding model in Ba1-xSnxCl1+yF1-y: (a) ionic bonding for isolated

Sn2+ ions at low x, (b) covalent bonding to Sn(II) in Sn clusters at
higher x.
On cooling to 17 K, thermal motion is frozen and therefore the Mssbauer

spectra are stronger. Since weak and strong lattices have their thermal motion
frozen, the difference between them is very small at low temperature, and weak
lattices benefit more from cooling (fig. 5b1). The reason for some of the
difference between the precipitated solid solution and that prepared in dry
conditions is unclear at this point. However, our current research under progress
has shown that there is a phase transition on heating the precipitates, to give a
high temperature phase metastable at ambient temperature, that is similar to the
samples prepared in dry conditions at ambient conditions.
The theoretical spectrum for purely ionic tin(II), i.e. the stannous Sn2+ ion, is
a single line at high velocity (ca. 4.1 mm/s), shown on figure 4a. However,
SnCl2 has an orthorhombic structure, and therefore the tin site is distorted, which
results in a small efg due to possible polarization of the lone pair and site
distortion, therefore one can expect a moderate broadening of the line. The same

applies to ionic tin in Ba1-xSnxCl1+yF1-y, where the presence of a mixture of
fluorine and chlorine around tin also contributes to the efg (fig. 4c). Therefore,
no strictly purely ionic tin(II) had been obtained before in a chloride or in a
fluoride. Now, this has been done: crystals of Ba2SnCl6 were obtained and its
crystal structure has been solved. The unit-cell is cubic, and tin is located in an
undistorted octahedral site (fig. 7). This is the first case of a perfectly spherical,
i.e. completely non-stereoactive, lone pair, in a tin(II) chloride. Efforts are now
focusing on producing a sufficient quantity in order to collect its Mssbauer
spectrum. It should have the narrowest possible line for Sn2+, and give, for the
first time, the true isomer shift of purely ionic tin in a chloride matrix. The full
crystal structure of Ba2SnCl6 will be published elsewhere.
Figure 7:
Crystal structure of Ba2SnCl6.
Conclusion
The combination of Mssbauer spectroscopy with X-ray powder diffraction has

made possible a deep understanding of tin bonding in the Ba1-xSnxCl1+yF1-y solid
solution. For the first time, a doubly disordered solid solution, with Ba2+ ions,
Sn2+ ions and covalently bonded tin(II) disordered on the same site has been
obtained. In addition, this is the first time that ionic tin(II) and covalently
bonded tin(II) have been reported to be present in the same sample, and also to
be varied by changing the stoichiometry.
Acknowledgements
This work was made possible by the support of Concordia University, the
Natural Science and Engineering Research Council of Canada and the universit
Mentouri. Grateful thanks are also due to the Procter and Gamble Co. (Mason,
Ohio) for supporting our Mssbauer laboratory.
321
References
[1]
[2]
Flahaut, J., Les structures de type PbFCl (EOI) et type anti-Fe2As (C38)
des composs ternaires deux anions MXY. J. Solid State Chem., 9, pp.
124-131, 1974.
Muntasar, A., Preparation, characterization and properties of novel
materials in the BaCl2/SnF2 system, PhD thesis, Concodia University,
Montreal, pp. 39-89, 2002.
323
Study on fatigue and energy-dissipation

properties of nanolayered Cu/Nb thin films
Y.-C. Wang, T. Hoechbauer, J. G. Swadener, T. Darling, A. Misra,
R. Hoagland & M. Nastasi
Centre for Integrated Nanotechnologies Group, Materials Science and
Technology Division, Los Alamos National Laboratory, Los Alamos, USA
Abstract
Energy dissipation and fatigue properties of nano-layered thin films are less well
studied than bulk properties. Existing experimental methods for studying energy
dissipation properties, typically using magnetic interaction as a driving force at
different frequencies and a laser-based deformation measurement system, are
difficult to apply to two-dimensional materials. We propose a novel experimental
method to perform dynamic testing on thin-film materials by driving a cantilever
specimen at its fixed end with a bimorph piezoelectric actuator and monitoring
the displacements of the specimen and the actuator with a fibre-optic system.
Upon vibration, the specimen is greatly affected by its inertia, and behaves as a
cantilever beam under base excitation in translation. At resonance, this method
resembles the vibrating reed method conventionally used in the viscoelasticity
community. The loss tangent is obtained from both the width of a resonance
peak and a free-decay process. As for fatigue measurement, we implement a
control algorithm into LabView to maintain maximum displacement of the
specimen during the course of the experiment. The fatigue S-N curves are
obtained.
Keywords: nanolayered thin films, fatigue, loss tangent.
Introduction
Materials consisting of nano-scale microstructures are of scientific and industrial

interest due to their unusual mechanical properties. It has been shown that nanolayered multilayers exhibit ultra-high yield strength [1]. Transmission electron
microscopy (TEM) studies on cold-rolled Cu/Nb multi-layers showed that the
doi:10.2495/HPSM06031

films exhibit large plastic deformation without the formation of dislocation cell
structures [2]. Based on the high tensile strength and lack of dislocation cell
structure formation during monotonic loading, we hypothesize that the fatigue
strength and failure mechanisms in nanolayered metals may differ significantly
from bulk metals. However, the fatigue and energy-dissipation properties of selfsupported nano-layered thin films have not been well studied.
Experimental methods for measuring mechanical fatigue properties of
materials have been well documented in many textbooks [3, 4]. In recent years,
new methods are continuously under development, such as the methods for the
fatigue investigations in gigacycle regime [5]. Similarly, there is a need to
develop methods to evaluate fatigue properties of novel nanoscale thin film
materials. We develop a method to perform fatigue tests on self-supported films
via oscillating cantilever beams at their resonant frequencies. This method is
superior to traditional methods, such as the tension-tension and rotating beam
fatigue methods, in that high stress and high cycle fatigue can be accomplished
simultaneously with fewer alignment problems. Our method applies fully
reversed stress to specimens, i.e. the load ratio min /max = -1. Furthermore, our
resonant frequency device also facilitates measurement of linear viscoelastic
damping.
Figure 1:
Schematic of the resonant frequency device for the measurement

of fatigue and energy-dissipation properties of self-supported thin
films.
Experimental
The schematic of our resonant frequency setup [6] is shown in Figure 1. A

bimorph piezoelectric plate-like actuator (Moran Electro Ceramics, Fairfield, NJ,
U.S.A.) with the dimensions of 15 mm in length, 6 mm in width and 0.64 mm in
thickness was used to generate excitation at the support of a cantilever beam
specimen. The bimorph actuator was clamped into the foundation with proper
electrical connections, and driven by a function generator (Agilent 33250A,
325
Agilent Technologies, Inc., Loveland, Colorado, U.S.A.). The free length of the
piezo actuator was 10mm. The multilayer specimens were mounted onto the top
of the piezo actuator with a set of aluminium clamp. The weight of the
aluminium clamp was 235.4 mg. The deflection of the specimen and aluminium
clamp was monitored by a fibre-optic measurement system (MTI-2000
FotonicTM sensor, MTI Instruments, Inc., Albany, NY, U.S.A.) with the probes
labelled Fibre 1 and Fibre 2 in the figure. The location of Fibre 1 was chosen to
be close to the fixed end of the specimen to avoid signal overload due to large
deflection. Lock-in amplifiers (Model SR830, Stanford Research Systems, Inc.,
Sunnyvale, CA, U.S.A.) were connected with the fibre-optic probes to reduce
noise in the displacement signals. Experiments were performed under moderate
vacuum conditions typically about 50 mtorr to reduce air damping and achieve a
reasonably high Q factor. Since our primary interest is to observe the initiation of
fatigue cracks, not crack propagation, we define fatigue failure when a 10 Hertz
reduction in specimens resonant frequency is observed.
Cu/Nb multilayers, with individual layer thickness of 40 nm and consisting
of 500 bilayers of Cu and Nb with a total thickness of 40 m, were prepared by
magnetron sputtering deposition on silicon wafers that were cut about a half way
through thickness by a wafer saw into a pattern of 2 mm long and 1 mm (or 0.5
mm) wide rectangular strips on the wafer surface. A typical cross-sectional TEM
image of the Cu/Nb multi-layers is shown in Figure 2 along with selected area
diffraction pattern. Note the polycrystalline structure of the layer with in-plane
grain size on the order of the layer thickness. The multi-layer exhibited a strong
Kurdjumov-Sachs orientation relationship: {111}Cu // {110} Nb, and <110>Cu
// <111>Nb. The parallel {111}Cu and{110} Nb planes formed the interface.
Figure 2:
Cross-sectional TEM micrograph (left) and the corresponding

selected area diffraction pattern (right) of a sputter-deposited
Cu/Nb multilayer with 40 nm individual layer thickness.
It is known that the Cu and Nb films manufactured by the magnetron sputtering

technique respectively have <111> and <110> textures parallel to their growth
direction (i.e. normal to layer surfaces). However, their in-plane properties are

composed of poly-crystals in random directions. The in-plane Youngs modulus
(E) of a material along a direction can be calculated as follows [7].
1 l 4 m 4 n 4 2m 2 n 2 2l 2 n 2 2m 2 l 2
=
+
+
+
+
+
,
F1
F2
F3
E E1 E2 E3
(1)
where
2
1 C22C33 C23
1 C11C 33 C132 1 C11C22 C122
=
,
=
,
=
,
E1
E2
E3
2 2 (C12C13 C11C23 ) 1
=
+
C44 ,
F1
2 2 (C13C23 C33C12 ) 1
=
+
F2
C55 ,
2 2 (C12 C23 C22C13 ) 1

=
+
C66 ,
F3
C11 C12 C13

= C12 C22 C23
C13 C23 C33 .
Here Cij represents elastic constants and ( , m, n) is the direction cosine.
For single crystal Cu, C11 = 169.8 GPa, C12 = 122.6 GPa and C44 = 75.3 GPa
[8]. The calculated principal Youngs moduli are E1 = E2 = E3 = 66.99 GPa. The
Young modulus along <111> is E<111> = 191.2 GPa. As for single crystal Nb, C11
= 245.6 GPa, C12 = 138.7 GPa, C44 = 29.3 GPa [8]. It is found that E1 = E2 = E3
= 145.48 GPa and E<110> = 93.2 GPa and E<111> = 83.2 GPa.
It can be seen from Figure 3 (a), the theoretical results of the in-plane
Youngs modulus of single crystalline Cu and Nb, that Cu in (111) plane is
transversely isotropic but Nb in (110) is direction-dependent. A simple uniform
homogenization for Nb along different directions results in an averaged in-plane
Youngs modulus about 100 GPa. Since our multilayers have equal volume
fraction for Cu and Nb throughout thickness, we adopt the Voigt homogenization
procedure, and obtain the homogenised in-plane Youngs modulus to be 115
GPa. Figure 3 (b) shows comparison between experimental and calculated inplane Youngs modulus. Good agreement is achieved.
A typical SEM picture of the fatigue crack observed on the Cu/Nb
multilayers before complete fracture is shown in Figure 4 (a). The length of the
sample was 7.1 mm, and the crack caused an about 60 Hz reduction in resonant
frequency. This sample was cycled at the stress amplitude 800 MPa for 8500
cycles. The position of the crack is very near the aluminium clamp. The bright
spots are the cyanoacrylate glue used to pre-mount the specimen on the piezo
327
actuator before mounting the aluminium clamp and fatigue testing. Fatigue
damage such as extrusions and intrusions are not observed around the crack on
the surface where maximum stress was exerted. A SEM image of annealed Cu
fatigued by the resonant frequency device is shown in Figure 4 (b). This sample
was fatigue tested at a stress after several thousand cycles. It can be seen that two
cracks were initiated at the edges of specimen, and propagated inwards. The
bright regions along the cracks are slip-related intrusions and extrusions.
160
4800
Ave. E 111 = 131.4 GPa for Cu

Ave. E 110 = 100 GPa for Nb
140
Resonant frequency, Hz
Directional E, GPa
Cu in (111)
120
Nb in (110)
100
Theoretical calculations based on

E = 115 GPa
4000
80
3200
2400
1600
800
<100>
60
-100
Figure 3:
-50
0
, deg
50
100
(a)
10
(b)
(a) Theoretical calculation of directional in-plane Youngs

modulus of the Cu and Nb with the texture of <111> and <110>
directions normal to the layer surface, respectively, and (b)
experimental results of resonant frequencies versus sample lengths.
(a)
Figure 4:
4
6
Length, mm
(b)
SEM images of (a) fatigued 40 nm Cu/Nb multilayers with a total

thickness of 40 m and (b) fatigued annealed 40 m-thick Cu,
clamped at the top of the image.
The S-N curves of several materials, quenched and tempered (550F) 4340
steel, the 40 nm Cu/Nb multilayers, mild steel, 6061-T6 and annealed Cu, are
shown in Figure 5. The nanolayered Cu/Nb exhibits a fatigue limit at about 450
MPa. It can be seen that the high fatigue endurance of the 40 nm Cu/Nb
multilayers correlates well with its high tensile strength of around 1400 MPa. We
obtain the ratio of fatigue endurance limit to ultimate tensile strength to be about
0.35 for the 40 nm Cu/Nb multilayers [9].
4340 Steel
Quenched and Tempered (550F)
Maximum Stress, MPa
1000
40 nm CuNb
multilayers
Al 6061-T6
100
Annealed Cu
100
10
10
10
10
10
10
Figure 5:
S-N curves (maximum stress vs. number of cycles to failure

curves). Symbols are defined as follows. Circles: 4340 steel.
Diamonds: 40nm Cu/nb multilayers. Triangles: mild steel. Up-side
down triangles: annealed Cu.
The deformation mechanism of the fatigue of the 40 nm Cu/Nb multilayers

is not completely understood. Since the formation of dislocation cell structures is
unlikely due to the nanoscale thickness of an individual layer between two
adjacent interfaces, extrusion or intrusion is not observed on the surface
subjected to maximum stress. This is consistent with our experimental findings.
However, concerning detailed deformation mechanisms, specifically how the
nanolayered thin films accumulate damage during cycling loading that
eventually leads to fatigue crack initiation require further TEM study, which will
be presented in a future article.
As for linear viscoelastic damping measurement in terms of loss tangent (i.e.
tan ), two approaches are adopted. From a resonant peak, one can use the
FWHM method to calculate tan , as follows [10].
tan =
0 3
(2)
Here 0 is the resonant frequency of the specimen and the frequency span at
half maximum intensity. The other approach is to extract loss tangent from a free
decay process, as follows.
tan =
1
A
ln 0 .
k Ak
(3)
329
Here k represents number of peaks in a time interval; A0 and Ak are amplitudes

respectively corresponding to the beginning and end of the time interval.
Figure 6 (a) and (b) show the results of our damping measurements on the
40 nm Cu/Nb multilayers. From resonant peaks, we obtain the loss tangent of the
nanolayered thin films, which is approximately 3.8 x 10-3 (thick solid line). After
20 million cycles of excitation under a small driving force (dotted line), tan is
5.3x10-3. The shift in resonant frequency, about 3 Hz, is unlikely due to fatigue
of the specimen. This resonant frequency shift may be system related, and it is
the reason why the 10-Hertz rule is adopted for determination of fatigue failure.
We remark that the measured loss tangent for the nanoscale films appears higher
than expected. This may be due to defects inherently from the magnetron sputter
deposition. Further experiments are underway to determine possible parasitic
damping from the support of the cantilever or air damping.
6
Cycled
-3
tan = 3.8x10 (virgin)
15
tan = 5.3x10 -3 (cycled)

10
5
0
-2
-4
L=11.3 mm
-6
-5
380
Figure 6:
Virgin
20
Amplitude, volts
Specimen displacement amplitude, m
25
-8
385
390
Frequency, Hz
395
400
tan = 3.3 x 10 -3
50
100
150
(a)
200
Time, ms
250
300
(b)
(a) Measured tan delta from resonant frequency peaks of a 6.5-mm

long specimen, and (b) loss tangent from the free decay method of
an 11.3 mm long specimen. Both results were obtained under a
pressure of 20 mtorr.
Conclusions
The resonant frequency method was constructed and shown to be efficient for
mechanical fatigue and energy-dissipation investigations of self-supported thinfilm materials. Cu/Nb multilayers with 40 nm individual layer thickness exhibit
over an order of magnitude greater fatigue strength compared to bulk Cu. No slip
band intrusions and extrusions were observed near the fracture surface. The ratio
of the fatigue endurance limit to ultimate tensile strength was around 0.35,
consistent with similar empirical scaling observed in other materials. Loss
tangent of the nanoscale films was about 4 x 10-3. However, potential parasitic
damping from the cantilever supports requires further study.
Acknowledgements
This research was supported, in part, by the Department of Energy, Office of
Science, Office of Basic Energy Sciences. YCW acknowledges support from the
LANL Directors postdoctoral fellowship.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
Clemens, B.M., Kung, H., & Barnett, S.A. (1999). Structure and strength
of multilayers. MRS Bulletin 24(2), 20-26.
Misra, A., Zhang, X., Hammon, D., & Hoagland, R.G. (2005b). Work
hardening in rolled nanolayered metallic composites. Acta Materialia 53,
221-226.
Frost, N.E., Marsh, K.I., & Pook, L.P. (1974). Metal Fatigue. Clarendon
Press, Oxford, UK.
Suresh, S. (1998). Fatigue of Materials. Cambridge University Press,
Cambridge, UK.
Bathias, C., & Paris, P.C. (2005). Gigacycle fatigue in mechanical
practice. Marcel Dekker, New York.
Wang, Y. C., Hoechbauer, T., Swadener, J. G., Misra, A., Hoagland, R.
G., Nastasi, M., Mechanical fatigue measurement via a vibrating
cantilever beam for self-supported thin solid films, submitted (2005).
Love, A.E.H. (1944). A Treatise on the Mathematical Theory of
Elasticity. Dover Publications, New York.
Hirth, J. P. and Lothe, J. (1982). Theory of Dislocations, Wiley, New
York.
Wang, Y. C., Misra, A. and Hoagland, R. G., Fatigue properties of
nanoscale Cu/Nb multilayers, submitted (2005).
Lakes, R.S. (2004). Viscoelastic measurement techniques. Review of
Scientific Instruments 75(4), 797-810.
331
Advances in computational modeling

through the use of higher-level
microstructure characterization
M. Groeber1, M. Uchic2, D. Dimiduk2, Y. Bhandari3 & S. Ghosh3
1
Department of Materials Engineering, The Ohio State University, USA

Air Force Research Laboratory, MLLMD, Wright-Patterson AFB, USA
3
Department of Mechanical Engineering, The Ohio State University, USA
2
Abstract
In this paper, orientation maps of consecutive serial sections are collected in an
automated manner by a Focused Ion BeamScanning Electron Microscope
(FIB-SEM) outfitted with an EBSD system. Micro-Imager, a program developed
in this work, uses the 2D EBSD maps to define microstructural features such as
grains and grain boundaries. Parameters used to characterize microstructure are
also calculated by Micro-Imager for every section. The statistical measurements
of each section are compared to assess variability in the microstructure. The 2D
sections are reconstructed into a volume by Micro-Imager3D, another program
developed in this work. Statistics analogous to those measured in 2D are
calculated and compared to the expected distributions predicted by the 2D
measurements coupled with stereology. As a result, quantitative descriptions of
microstructure are made and improvements over conventional methods are
yielded. Information about individual constituents allows correlations between
distributions to be derived. The correlations drawn allow models to account for
aspects of microstructure that have classically been overlooked. The resulting
3D grain structure serves as a realistic model microstructure.
Keywords: quantitative characterization, equivalent representation, modelling.
Introduction
The ability to characterize microstructure is an important tool for materials

scientists and computational modelers, because it allows one to predict the
capability of a material for a given application. For example, it is well known
doi:10.2495/HPSM06032

that the grain size of a material has a strong effect on mechanical properties;
therefore an accurate measure of the grain size distribution is desirable to predict
material performance. Classic methods for characterizing microstructure usually
involve viewing an image from a sectioned surface, where the area of interest
would be mechanically polished [1]. Stereology or other methods can be used to
interpolate three-dimensional (3D) statistics from the 2D microstructural images.
However, there are some microstructural parameters that cannot be inferred
from 2D sections [2]. In addition, many stereological parameters yield only
average values to describe microstructural features. Recognizing the fact that
many properties (especially those associated with failure) require extreme values
of the microstructure [3], it is evident that characterizing the full distribution of
these features may be more appropriate for some predictive models [4,5]. This
need to more completely characterize microstructure has led to more direct
methods, such as serial sectioning, that allow one to obtain true 3D
microstructural data [6,7]. For this study, a Focused Ion Beam (FIB) is used to
serial-section a specimen and an EBSD system is used to obtain an orientation
map of each section. This method shows great potential because the dual beam
FIB-SEM microscope can be automated to perform this analysis without user
interaction.
Post-processing of the orientation data is performed using a program
developed in this study called Micro-Imager. Micro-Imager automatically
defines grains and grain boundary segments using the 2D EBSD maps, and
calculates statistics of the microstructure based on this information. MicroImager characterizes the microstructure more completely; because it calculates
full distributions of parameters in both the 2D sections and the 3D reconstructed
volume. This study focuses on the task of developing refined data collection and
also improving statistical analysis techniques to increase predictive capabilities.
This methodology may be particularly useful for generating a host of statistical
microstructural correlations that enable the construction of truly representative
material microstructures as input for modeling and simulation programs.
Brief overview of the FIB-EBSD serial sectioning process
The serial sectioning experiment in the Dual Beam FIB-SEM is comprised of

moving the sample repeatedly between two microscope stage positions the
sectioning position and the ion imaging/EBSD analysis position. Image
recognition is critical for precise alignment at the two stage positions, as the
microscope stage is only accurate to within few microns when moving between
the two stage positions. The image recognition of the DB235 ensures a consistent
slice thickness, which is important factor for reconstruction of the serialsectioning data into a 3D volume and is one of the distinct advantages of this
automated process. The key advantage of acquiring the orientation information is
it allows unsupervised segmentation of grains, which is difficult using only
secondary electron or even ion images. A more detailed explanation of the serial
sectioning process can be found in [8].
333
3D orientation data reconstruction and grain segmentation

and measurement by Micro-Imager
3.1 Grain segmentation with Micro-Imager

Micro-Imager is a newly developed software program that automatically locates
and approximates the complex grain boundary positions using a series of line
segments and an error-per-unit-length algorithm. This process corrects artifacts
that are produced when collecting EBSD data as well as simplifies the structure
for subsequent mesh generation for computational modeling.
Grain boundary detection consists of comparing orientations of neighboring
data points. The midpoint between two data points is tagged as a boundary point
if there is a misorientation of more than 4 degrees. This process is completed for
all data points, and then special boundary points are located. Special points are
the points that three or more grains share. The initial approximation of the grain
boundaries is simply the connection of these special points.
Micro-Imager allows the user to adjust the value of the acceptable error-perunit-length. As the user lowers the acceptable value, the approximated
boundaries will have an error-per-unit-length that is intolerable. The specialpoint connection will be changed into two equal-length segments that share an
endpoint on a defined boundary point and have the same bounding special points
as the initial segments endpoints. A schematic of the evolving grain boundary
approximation is shown in Fig. 1. Low tolerances closer represent the actual
data, but do not simplify the structure. Higher tolerances simplify the structure,
but it is important not to oversimplify and make erroneous microstructures. A
more detailed description of the grain segmentation process is discussed in [8].
Figure 1:
Schematic showing the evolution of the grain boundary

approximation in Micro-Imager. Here t(x) represents the true grain
boundary shape and the line segment(s) labeled a(x) represent the
approximated boundary.
3.2 Measurement of microstructural parameters by Micro-Imager

Once Micro-Imager has created a representative microstructure, it has the ability
to extract statistical information about the grain morphology. In the initial phase

of this work, five parameters were selected for measurement. These parameters
are characterized in 2D, and corresponding 3D measurements have also been
calculated for some of these parameters. The five parameters measured in 2D
are: misorientation, number of neighboring grains, number of grain boundary
edges, grain boundary perimeter, and grain area. Grain volume and number of
neighboring grains were also measured in 3D.
Statistical analysis
4.1 Micro-Imager analysis versus stereology

In this work, measurements of two microstructural parameters by Micro-Imager
are compared to the same parameters measured by conventional stereology to
validate the accuracy of Micro-Imager calculations. The two parameters are
grain area and grain boundary surface-area-per-volume.
(A)
(B)
Figure 2:
Comparison of parameter measurements from Micro-Imager and

Fovea Pro, where thickness means the sectioning depth. (A)
Average grain area and (B) Average grain boundary-surface-areaper-volume.
335
The measurement of grain area in Micro-Imager is straightforward. The

area of each grain is simply the number of data points originally assigned to that
grain multiplied by the area associated with each data point. The calculation also
includes the error calculated during grain boundary approximation discussed
previously. The stereological measurement of grain area was performed using
Adobe Photoshop with the image analysis toolkit Fovea Pro. The grain area was
determined using the lineal mean intercept measurement in Fovea Pro.
Figure 2a shows a comparison of average grain area determined by MicroImager and Fovea Pro for section. To calculate the grain area using the lineal
mean intercept method, grains were considered to be spherical, i.e. the grain area
was assumed to be that of a circle with a diameter equal to the lineal mean
intercept. There are two curves for the grain area as calculated by Fovea Pro; the
higher curve includes a multiplicative factor that accounts for the error in the
lineal intercept method for circular grains in a 2D image [10].
Figure 2b shows a comparison of surface area-per-volume measurements.
Micro-Imager calculates this value by adding the length of all the grain boundary
segments for a grain and dividing by the total area of the grain. A stereological
factor is used to translate the value of boundary-length-per-area to boundarysurface-area-per-grain-volume [9]. The calculation using Fovea Pro counts the
total number of intersections of a grid of cycloids with the grain boundaries. The
number of intersections is divided by the length of the cycloids. This value is
multiplied by a stereological factor to convert the intersections-per-length to
boundary-surface-area-per-volume [9]. The radius of the cycloids will affect the
results, because as the radius approaches zero the intersections approach exactly
the grain boundary lines and the cycloids fill the full area, yielding the same
value as the length-per-area determined by Micro-Imager. It can be seen in Fig.
2b that measurements in Fovea Pro converge to the measurement by MicroImager. This is one advantage of Micro-Imager; that some microstructural
parameters can be calculated in a manner that is not sensitive to the measurement
method, in contrast to traditional stereological techniques.
(A)
Figure 3:
(B)
Probability Density Functions generated by Micro-Imager from a

single 2D section. (A) grain diameter (B) misorientation angle.
The value of grain diameter is calculated as that of a circle with the
same area as the grain.

4.2 Probability density functions
In addition to calculating the mean value of the parameters, the full distributions
have been plotted and fitted to determine probability density functions (PDFs),
which provide a much more complete characterization. Figure 3 shows the PDFs
of grain size and misorientation angle. Probability density functions are useful in
creating statistically equivalent microstructures, because they clearly describe the
likelihood of a value of a parameter to occur.
4.3 Parameter variation between slices
Figure 4 is a 3D surface plot constructed from the series of individual probability
density functions. This plot does not substitute for a PDF from a 3D
reconstruction, but it can easily display the variability of a parameter through the
thickness of the serial sectioning data set. That is, if the PDFs vary markedly,
then one 2D section may not be sufficient to describe a particular microstructural
feature.
Figure 4:
PDFs of grain area shown collectively as a surface plot in 3D to

highlight variation from section-to-section.
Two additional applications of this methodology to examine parameter

variation are shown in Figs. 5 and 6. In Fig. 5, a histogram of the grain area is
plotted, which was constructed by taking the average value of the bin across all
the 2D sections. The error bars denote the maximum and minimum values for
each bin across the entire set. The figure illustrates that there can be significant
deviations in the distribution of grain size across the 2D imageseven for sizes
which are at or near the mean.
337
Figure 5:
Histogram of grain area that is constructed from the averaged bin

values for each size range for all of the sections. The error bars
show the maximum and minimum values for each size throughout
the entire data set.
Figure 6:
Plot showing the values for average grain area and maximum grain
area for each slice.
Figure 6 tracks the average grain area and the maximum grain area on every
slice. One can observe that the average grain area appears essentially constant
for every slice, but there is noticeable variation in the maximum. The data
shown in this section further illustrates the potential need for more detailed
microstructural analysis than simply the average of a parameter. Often, the
average value for a particular parameter appears constant from slice-to-slice, but
can vary markedly in the extreme values. If only the average value of a feature

is needed for property prediction, then a single section of sufficient size may be
adequate. However, if knowledge of the extreme values of a micro-constituent is
needed, then the information provided by only one or two sections from a
material cannot confidently describe the rest of the microstructure. Furthermore,
there are some parameters that can only be calculated from a true 3D
microstructural environment [2].
Figure 7:
Plot of correlation between grain volume and number of

neighboring grains in 3D. Plot shows that small can grains cluster,
but large ones tend not to.
Figure 8:
Plot showing the change in the distribution of neighboring grains as

grain size increases.
339
4.4 Correlations between microstructural parameters

In addition to representing distributions of individual microstructural parameters,
it may be important for many properties to consider the correlation between
multiple parameter distributions. A grouping of similarly oriented grains or a
clustering of extremely large or small grains can greatly affect local responses.
Figure 7 displays the correlation between grain size and the number of neighbors
of the grain. It can be seen in Fig. 7 that small grains can have many neighbors,
implying clustering; however, larger grains do not appear to cluster. Accounting
for relationships between parameters yields a more accurate description of
microstructure. Figure 8 shows distributions of the number of neighbors for
various grain sizes. It is clear that as the grain size increases the distributions
shift to larger numbers of neighbors as well as broaden. Parameter correlations
can yield microstructure models that may produce more realistic results for
locally driven properties.
Figure 9:
Plot comparing the unbiased distribution of grain volumes from the

reconstructed microstructure and the average extrapolated
distribution of grain volumes from all the 2D sections.
4.5 Comparison of 2D and 3D measurements

The two parameters correlated in Fig. 7, grain volume and number of neighbors,
are computed in an unbiased manner from the reconstructed volume. These
unbiased measurements can be compared to extrapolated 2D measurements.
Comparisons of the two measurements will allow for the evaluation of the
extrapolation technique. Figure 9 shows a plot of the unbiased distribution of
grain volumes from the reconstructed microstructure along with the average
extrapolated distribution of grain volumes from all the 2D sections. Figure 9

shows that the extrapolated distribution differs noticeably from the actual 3D
distribution. The extrapolation technique used in this comparison, which
assumed the grains to be shaped as Kelvin polyhedra, over-predicts the number
of very small grains and the size of the largest grains. As a result, the average
grain volume for the two distributions is nearly the same, but the extremes of the
distributions are quite different.
Parameter measurements on the reconstructed volume provide unbiased
descriptions of the grain morphology. Comparisons such as that in Fig. 9 will
show where classic stereological relations breakdown. In addition, there are
anomalies in the reconstructed volume that maybe deviate significantly from
classic assumptions of microstructure. For example, the average number of
neighbor grains in the reconstructed microstructure may agree with theoretical
calculations, but theoretical calculations greatly under-predict the maximum
number of neighbor grains.
5 Conclusion
Material characterization is a key in the prediction of material function. The
ability to understand the microstructure is the first step in being able to optimize
parameters that inevitably lead to superior materials. This study has introduced
both experimental and computational procedures that enable new techniques to
the forefront of microstructural analysis. This paper presents a snapshot of the
representation and analysis methodology as it exists currently. The focus of this
work is to continue to improve the representation of microstructures to create
realistic computational models for prediction of material capabilities.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
Samuels LE. Metallographic Polishing by Mechanical Methods. Metals

Park, OH: American Society for Metals; 1982.
DeHoff R T. Quantitative serial sectioning analysis: preview. J.
Microscopy 1983; 131: 259-263.
Ghosh S and Moorthy S. Particle Fracture Simulation in Non-Uniform
Microstructures of Metal-Matrix Composites. Acta Materialia 1998; 46:
965-982.
Kurzydlowski K J, Ralph B, Bucki J J, and Garbacz A. The grain
boundary character distribution effect on the flow stress of polycrystals.
Materials Science and Engineering A 1995; 205: 127-132.
Kurzydlowski K J. On the dependence of the flow stress on the grain size
distribution in polycrystals. Scripta Metallurgica 1990; 24: 879-884.
Li M, Ghosh S, Richmond O, Weiland H, and Rouns T N. Three
dimensional characterization and modeling of particle reinforced MMCs,
Part I: Quantitative description of microstructural morphology. Material
Science and Engineering A 1999; A265: 153-173.
Li M, Ghosh S, Richmond O, Weiland H, and Rouns T N. Three
dimensional characterization and modeling of particle reinforced MMCs,
[8]
[9]
[10]
341
Part II: Damage characterization. Material Science and Engineering A

1999; A266: 221-240.
Groeber M, Haley B K, Uchic M D, Dimiduk D M, and Ghosh S.
Towards 3D Reconstruction and Characterization of Polycrystalline
Microstructures Using a FIB-SEM System. Submitted to Materials
Characterization 2005.
Russ J C and Dehoff R T. Practical Stereology. New York: Kluwer
Academic/Plenum; 2000.
Thompson A W. Calculation of True Volume Grain Diameter.
Metallography 5 1972; 366-369.
343
Deformation of aluminum alloys AD-1,

AMg-6 and D-16 at dynamic compression
and temperatures of 25250oC
V. A. Pushkov, S. A. Novikov, V. A. Sinitsyn,
I. N. Govorunov & O. N. Ignatova
Russian Federal Nuclear Center VNIIEF, Russia
Abstract
Using the Kolsky method, the authors studied dynamic diagrams of uniaxial
compression of aluminum alloys AD-1, AMg-6, and D-16. These alloys are
applied in aviation, space technologies, shipbuilding and many other industries.
Alloys similar to the above mentioned alloys have been studied abroad as well
(for example, aluminum alloys 1100-0, 5182, 6061-T6); however in general data
on dynamic deformation of the mentioned aluminum alloys have been obtained
only at normal temperature. In this work, we performed systematized research of
the aluminum alloys at strain rates of 2001400s-1 and temperatures of 25
250o.
Keywords: Kolsky method, compression, temperature, strain rates, aluminum
alloys, temperature and velocity dependencies V-0.2, model of strength.
Introduction
Study of dynamic mechanical properties of structural materials is an urgent

problem for many areas of science and engineering, and VNIIEF experts are
actively involved into efforts for solving this problem. The Kolsky method [1] is
one of the reliable methods for study of dynamic mechanical properties of
materials at strain rates of 102104s-1. In its classical form, it is intended to study
dynamic diagrams of uniaxial compression V-H. The method falls into the
category of tests with the parameter =dH/dt=const, i.e. with constant strain rate.
The method essence is quasistatic loading of a sample placed between two steel
bars by passing and reflected waves of stresses. The bars are in elastic state,
while the sample undergoes elastic-plastic deformation. Various techniques of
doi:10.2495/HPSM06033

dynamic tests are presently based on the Kolsky method principles. These
techniques are used in tests with extension, torsion, torsion simultaneously with
compression or extension, crack-resistance tests, tests for study of the
Baushingers effect [2, 3].
Research results
Using the Kolsky method, the authors studied dynamic diagrams of uniaxial
compression of aluminum alloys AD-1, AMg-6, and D-16. These alloys are
applied in aviation, space technologies, shipbuilding, and many other industries.
Available data on these alloys are mostly data obtained at static velocities of
loading [4]. There are some data obtained under dynamic loading, but only for
normal temperature [58]. Alloys similar to the above mentioned alloys have
been studied abroad as well (for example, aluminum alloys 1100-0, 5182, 6061T6), strain diagrams have been obtained for them as well, but they are also for
normal temperature [911]. In this work, we performed systematized researches
of the aluminum alloys at strain rates of 2001400s-1 and temperatures of
25250o. Samples in as-delivered condition having sizes 88mm were
studied. In each experiment, diagrams of dynamic compression were obtained,
and yield strengths -0.2 were determined. The diagrams were plotted in the
stress intensity strain intensity coordinates. Recalculation was performed
using the dependencies taking account for change of the effective Poissons ratio
in the area of elastic-plastic transition, eqns (1) and (2):
i=/[1-ln(1+)]2
(1)
i=2(1+)ln(1+)/3
(2)
In eqns (1) and (2) =0.5-0.5(1-2)/(). Fig 1 shows averaged diagrams ii of the alloys. To make the figure simpler, the diagrams are given for each
material at temperatures of 25 and 250o (the diagrams at 150o are located in
the middle). Experiments revealed that strain hardenings of alloys D-16 and
AMg-6 were approximately the same, and it was higher than that of alloy AD-1.
D-16 is the hardest, and AMg-6 and AD-1 are located in the decreasing order
(see Fig 1).
The yield strengths temperature dependencies -0.2=f(T) and velocity
dependencies -0.2=f() are shown in Fig 2 and Fig 3. The same as in Fig 1, the
dependencies in Fig 3 are given at temperatures of 25 and 250o. These
dependencies have generally the linear character. It should be noted that the
velocity dependence -0.2 does not evidently result from curves 7, 8 (for AMg-6)
and curves 11, 12 (for AD-1) from Fig 1. However, in Fig 3, where results of
each experiment are pointed, one can see, nevertheless, the weak dependence of
-0.2 on for the mentioned materials. It follows from Fig 2 and Fig 3 that the
strongest change of -0.2 occurs in D-16 with growth of and , and the weakest
change occurs in AD-1.
i, MPa
345
1
5
2
400
3
7
6
8
9
200
10
11
12
0
0
0.1
0.2
Figure 1: Averaged diagrams of dynamic compression of aluminum alloys.

D-16: 1=25o, =8001100s-1; 2-=25o, =310-500s-1;
3-=250o, =9001050s-1; 4-=250o, =280550s-1; AMg-6: 5
=25o, =11001300s-1; 6-=25o, =250550s-1; 7-=250o,
=12001400s-1; 8-=250o, =190530s-1; AD-1: 9=25o,
=8001200s-1; 10-=25o, =200370s-1; 11-=250o, =920
1100s-1; 12-=250o, =330600s-1.
So, for alloys D-16, AMg-6, AD-1 in the considered temperature-velocity
conditions of loading, we observe in this or that extent the drop of -0.2 with
temperature growth, and the increase of -0.2 with strain rate growth. It is
characteristic of many other materials and alloys [11]. The obtained data are in
agreement with data of the other authors. For instance, it is mentioned for D-16
in [7] that the yield strength of the alloy is +0.2=290300 MPa at =20o and
500s-1 (extension). In [11], for alloy 6061-6 (that is an analog of D-16), there
is -0.2=300310 MPa at =20o and =900s-1 (compression) from the diagrams
-. It is noted in [5] for AMg-6 that at =20o and =270s-1 and 2000s-1, values
of the yield strengths are 205 MPa and 230 MPa, respectively. It is characteristic
that these data confirm the velocity dependence of -0.2 for AMg-6 in [5]. In
[7, 8], the authors mention that +0.2=170175 MPa at =20o and 500s-1
(extension). At the same time, it is noted in [7, 8] that the yield strength +0.2 is

actually unchanged, when is growing from static values to dynamic values (up
to 103s-1). This difference from our data can be caused by different histories of
AMg-6 deformation. Time of loading growth in sample in our experiments was
70s, and it was 5s in [7, 8]. However, this problem requires more
comprehensive study.
-0.2, MPa
300
200
2
3
4
100
0
5
6
0
100
, 0
200
Figure 2: Temperature dependencies of -0.2 of aluminum alloys. D-16: 1

=8001100s-1; 2 =280550s-1; AMg-6: 3 =1020-1400s-1; 4
=190620s-1; AD-1: 5 =8001200s-1; 6 =200600s-1.
0,2, MPa
1
300
2
3
200
100
0
5
6
0
400
800
1200
, s-1
Figure 3: Velocity dependencies of -0.2 of aluminum alloys. D-16: 1=25o;

2=250o; AMg-6: 3=25o; 4=250o; AD-1: 5=25o; 6
=250o.
No similar data have been revealed for AD-1 in Russian literature. In foreign
literature, Lindholm and Yeakley [10], for alloy 1100-0 (that is an analog of
347
AD-1) tested by the Kolsky method, increase of strength properties is noted

with strain rate growth.
Modeling
To create a more reliable model of behavior of aluminum alloys, additional tests

were performed with dynamic extension and normal temperature. Basing on
experimental results, we developed the phenomenological model of strength [13]
for description of our obtained dynamic diagrams of compression and extension
of aluminum alloys AMg-6, AD-1, D16 at strain rates of 102103 s-1 and
temperatures of 25250 (298523 K). The model content is the following. If
medium is elastic-plastic, stress intensity i (yield strength at uniaxial
p
compression and extension) can be presented as function of three variables i ,
ip , characterizing its stress-strain state, eqn (3):

i = i ( ip , ip , T )
(3)
p
p
Here i - intensity of plastic strains, i - intensity of rate of plastic strains, -
current temperature. Stress intensity i can be presented as product of simple

functions, where each function depends only one variable:
i = A f 1 ( iP ) f 2 (iP ) f 3 (T )
(4)
In eqn (4), f1 (ip ) describes strain hardening, f 2 (ip ) influence of rate of

plastic strain, and f 3 ( ) thermal softening. In the expanded form, the
constitutive equation of dynamic deformation is the following eqn (5):
m
ip
i = A 1+ a ( ) 1+b ln p 1T k
i0
p n
i
(5)
Here , , b, m, n, k are constant values, which can be determined using

experimental data; iop =1s-1 normalizing value of ip ,
r
, m melting
m
temperature; Tr current temperature.

For aluminum alloys, steel and copper tested by the SHPB method, Table 1
presents values of constant coefficients and exponents selected basing on
available experimental data. The area of application of the suggested model is
limited by the area of experimentally obtained characteristics. It is equal to:
ip 0.20 ; ip 1,4103 1/s ; m

Table 1:
Material
AMg-6
AD-1
D-16
Values of constant coefficients and exponents, eqn (5).
process
compression
extension
compression
extension
compression
extension
, MPa
200
130
43.5
30
240
220
a
2.45
2.40
2.8
2.8
2.28
2.28
n
0.4
0.4
0.335
0.330
0.28
0.28
b
510-4
510-4
210-4
210-4
1.210-4
1.210-4
m
2.55
2.50
3.6
3.6
3.3
3.3
k
2
2
2
2
2
2
Table 2: Experimental and calculated values of conventional yield strength at

compression of aluminum alloys AD-1, AMg-6 and D-16 at various
strain rates and initial temperatures of samples.
Material
,oK
298
AMg-6
423
523
298
AD-1
423
523
298
D-16
423
523
,s-1
-0.2 aver, GPa
250550
0.1870.12
11001300
360620
10201300
190530
12001400
200370
0.2100.08
0.1690.09
0.1900.14
0.1590.12
0.175
0.0550.04
0.201
0.155
0.169
0.170
0.177
0.058
8001200
230290
9001200
330600
9201100
0.0670.03
0.0490.03
0.053 0.03
0.0450.02
0.0490.02
0.065
0.051
0.056
0.043
0.049
310500
0.3100.09
0.320
8001100
280520
8001100
280550
9001050
0.3380.1
0.2630.14
0.2880.045
0.2450.09
0.2650.11
0.340
0.280
0.300
0.245
0.263
test
-0.2 aver, GPa

calculation
0.189
Tables 2 and 3 present values of conventional yield strength at compression

(-0.2) and extension (+0.2), which are experimental and calculated by (5) for
aluminum alloys AD-1, AMg-6 and D-16 at various strain rates and initial
temperature of samples. And for example, Fig 4 shows - diagrams of AMg-6
compression and extension, which are experimental and calculated by eqn (5).
One can see in Table 2, Table 3, and Fig 4 that the suggested elastic-plastic
model is in satisfactory agreement with the experimental data within test error.
Table 3:
349
Experimental and calculated values of conventional yield strength at

extension of aluminum alloys AD-1, AMg-6 and D-16 at various
strain rates and normal temperature of samples.
Material
,oK
AMg-6
298
AD-1
298
D-16
298
, s-1
+0.2 aver, GPa

test
+0.2 aver, GPa

calculation
640800
12001450
210370
9101100
230370
6501130
0.1410.05
0.1560.08
0.0400.08
0.0430.08
0.2850.05
0.3030.014
0.147
0.152
0.040
0.045
0.288
0.300
Conclusion
Studies of dynamic diagrams of uniaxial compression of alloys AMg-6, AD-1,

D-16 were performed at strain rates =2001400 s-1 and temperatures =25
250o. Temperature and velocity dependencies of yield strengths -0.2 were
obtained. Among the tested alloys, D-16 has the highest strength. The strongest
change of -0.2 with growth of and is revealed for D-16, and the weakest for
AD-1. Additional tests were performed with dynamic extension and normal
temperature.
Figure 4: Experimental (different form points) and calculated (solid and dotted
lines) - diagram of compression and extension of aluminum alloy
AMg-6 at strain rate of 103 s-1. 10=298o (compression), 2
0=423o (compression), 30=523o (compression), 40=298o
(extension).

Basing on experimental results, we developed the phenomenological model of
strength for description of our obtained dynamic diagrams of compression and
extension of aluminum alloys. The presented results can be helpful for
conduction of various strength calculations, as well as for prediction of material
behavior under dynamic loading [13].
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
Kolsky H., An Investigation of the Mechanical Properties of Materials at

Very High Rates of Loading. Proceedings of the Physical Society, Section
B, 62, pp. 676-700, 1949.
Muzychenko V.P., Kashenko S.I. and Guskov V.A., On application of
split Hopkinson bar method. Plant laboratory (Rus.) 1, pp. 58-66, 1986.
Novikov S.A., Pushkov V.A., Sinitsyn V.A. and Tsoi P.A., Study of the
Baushingers effect at dynamic loading. Appl. Mech. and Tech. Phys.
(Rus.), 4, pp. 163-169, 1995.
Shalin R.E. Aviation materials. Reference book, v.4, part 1, ONTI-VIAM,
Moscow, pp. 180-185, 1982.
Bolshakov A.P., Novikov S.A. and Sinitsyn V.A., Study of the tension
and compression uniaxial dynamic diagrams of copper and alloy AMg6.
Problems of strength (Rus.), 10, pp. 87-88, 1979.
Glouschenkov V., Novobratsky R. and Bourmistrov A., Influence of the
spread in values of aluminum alloy dynamic properties upon the final
results of magnetic-pulse strain, Proc. of the DYMAT-91, Physics Eds.,
Association DYMAT: Strasbourg, pp. C3/331-C3/334, 1991.
Stepanov G.V., Astanin V.V. and Romanenko V.I., Study of properties of
the aluminum alloys at dynamic loading. Problems of strength (Rus.), 2,
pp. 59-63, 1983.
Popov N.N., Ivanov A.G., Strekin V.P. and Barinov V.M., Study of AMg6 dynamical behaviour. Problems of strength (Rus.), 12, pp. 50-54, 1981.
Higashi K., Mukai T. and Kaizu K., The microstructural evolution during
deformation under several strain rates in a commercial 5182 aluminum
alloy, Proc. of the DYMAT-91, Physics Eds., Association DYMAT:
Strasbourg, pp. C3/347-C3/352, 1991.
Lindholm U.S., Yeakley L.M., High strain rate testing: tension and
compression. Experimental Mechanics, 8(1), pp. 1-9, 1968.
Mayden S., Green S., Compressive strain-rate tests on six selected
materials at strain rate from 10-3 to104 in/in/sec. Applied Mechanics, Ser.
, v. 33(3), pp. 20-30, 1966.
Novikov S.A., Pushkov V.A., Dynamic crack-resistance of 4-based
materials, Proc. of the Conf. on Urgent Problems of Protection and
Safety, Publ. House of SIC Spetsmaterials: St. Petersburg, pp. 4-8, 1999.
Glushak B.L., Ignatova O.N., On modeling of dynamic deformation
diagrams. Questions of Atomic Science and Technique (Rus.), 2,
pp. 45-49, 1998.
351
X-ray residual stress measurements on

plasma sprayed molybdenum coatings
K. Hirukawa1, K. Akita2, S. Tobe3, T. A. Stolarski4 & S. Ohya2
1
Research Division in Engineering,

Graduate School of Musashi Institute of Technology, Japan
2
Musashi Institute of Technology, Japan
3
Ashikaga Institute of Technology, Japan
4
School of Engineering and Design, Brunel University, UK
Abstract
Plasma sprayed molybdenum coatings have been applied to friction parts in the
automotive field because of their high wear resistance. Clarification concerning
the significance of mechanical properties and the residual stress state of the
coating is important to improve the performance of the coating. However there
are some difficulties in the measurement of mechanical properties of the coating,
especially Youngs modulus, because it has complex structures including
lamella, pores, micro cracks and so on. Youngs modulus is required to
determine the x-ray stress constant for x-ray residual stress measurements. Strain
gauges are often used to measure the strains of a component. If glue is applied to
thermal spray coatings to attach a strain gauge, it might be that it penetrates into
the pores and becomes solidified, hence incorrect strains will be measured.
In this research, firstly, Youngs modulus and x-ray stress constant, K, for
molybdenum thermal spray coating were determined experimentally by
four-point bending tests. The effects of quick-drying glue on strain measurement
using a strain gauge were investigated. Secondly, the residual stresses of the
coatings were measured using x-ray diffraction. Six types of substrates with a
different thermal expansion coefficient (TEC) were prepared. The materials
used, in ascending order of TEC, were Super Invar, molybdenum, SCM440,
SK5, SUS304 and A6063. Molybdenum powder was sprayed on the substrates in
air with various thicknesses.
The following results were obtained. (1) The effect of quick-drying glue on
the mechanical strain measurement on the sprayed coating was negligible. (2)
The Youngs modulus of the coating was lower than that of the commercial
molybdenum sheet. (3) Linear strain response against applied mechanical loads
was observed in the case of the polished coating surface. (4) The x-ray stress
constant K of the coating and the bulk were almost the same. (5) Residual
stresses on the coating surfaces were of tensile type on all substrates used in this
study. (6) Tensile residual stresses of the coatings increased with the decreased
thermal expansion coefficient of substrate.
Keywords: x-ray stress measurement, residual stress, plasma spraying,
molybdenum coating, thermal expansion coefficient, quick-drying glue, strain
response.
doi:10.2495/HPSM06034
Introduction
Thermal sprayed molybdenum coating is one of the surface modification

applications improving the wear resistance. Mo coating also has high heat
resistance so that Mo coatings are applied to the engines and sliding parts of
automobiles [1].
The effect of controlled residual stress on wear resistance has been reported
[2]. Evaluation of the residual stress is essential to improve the enabled
functions. In this study, firstly, the effects of quick-drying glue on mechanical
Youngs modulus measurement using strain gauge were considered. Secondly,
the effects of thermal expansion coefficient (TEC) of substrates on residual stress
were investigated.
Experiment procedure
2.1 Specimen preparation

Materials used as substrates, in ascending order of TEC in air, were Super Invar,
molybdenum, SCM440, SK5, SUS304 and A6063. Molybdenum powder was
used as a spraying material. TEC of all materials used are shown in table 1. A
condition substrate TEC < coating TEC can be provided when Super Invar is
used as substrate in the case of molybdenum coating despite it has low TEC.
Two types of substrate were used for mechanical Youngs modulus measurement
and residual stress measurement. The former was 100 15 3mm, the later was
15 15 3mm.
Table 1:
Thermal expansion coefficient of materials at the room temperature

.
, 10-6.
Super
Molybdenum SCM440
SK5
SUS304
A6063
Invar
ASTM
AISI
Molybdenum AISI 4140 SK50H AISI 304 ISO AlMg0.5Si
F1684
JIS
0.5
4.9
11.8
11.8
17.3
24.3
2.2 Spraying condition

For Youngs modulus measurement, molybdenum was sprayed with the
thickness of 300m by the Atmospheric Plasma Spraying system. The spraying
condition is shown in table 2. There were two surface conditions. One is assprayed surface and another is polished with waterproof abrasive paper.
For residual stress measurement, six types of substrate were sprayed with
various thicknesses. Spraying was applied with the different number of passes.
Table 2:
Plasma spraying conditions.
Spraying material
Mo powder
Method
APS
Current
800 A
Voltage
26 V
Spraying distance
Maximum thickness
353
100 mm
300 m
3.1 Measurements of mechanical Youngs modulus E

Loads were applied to coating-substrate composite beam by four-point bending.
Tensile or compressive stress was generated at coating surface. The range of
applying loads was determined in the elastic region (within -1600 10-6 strain)
determined by Acoustic Emission test [3]. In this paper, the elastic region of the
coatings was defined below AE increasing point.
In measuring mechanical strain using the strain gauge, it is said that quickdrying glue penetrates into the pores and solidifies thus making strain
measurement ambiguous. In our study, strain gauges were attached using quickdrying glue; Youngs modulus of the glue after solidified was 3.2 GPa. With
reference to Youngs modulus, the coating was measured with various conditions
of coating surfaces and substrate thicknesses. All the strain gauges were held for
24 hours after bonding.
Both strains of substrate and coating were measured at the same time.
Youngs modulus of coating, EC, was calculated by the following equation (1)
tC AS ES C +1 EC Z S
S
EC =
C
+1 Z S tC AC
S
(1)
from the theory of composite beam. Where t is the thickness, A is the crosssection area and Z is the geometrical moment of area, respectively. Two suffixes
C and S mean coating and substrate. Coating surfaces are postulated as uniaxial
stress.
Furthermore the coating was broken by bending just after cooling with liquid
nitrogen. The fractured surface was observed by a Scanning Electron Microscope
to clarify the penetration depth of quick-drying glue.

3.2 Measurements of x-ray stress constant K
Tensile and compressive stress was applied to coating surfaces within the elastic
region described previously. Loads were applied gradually and slope of 2-sin2
diagram [4] M was determined in each case under the condition shown in table 3.
2-sin2 diagrams were obtained by averaged diffraction angle in three times
continuous measurements. X-ray stress constant K was calculated with M-app
diagram. The equations for K and M are as follows.
K=
Table 3:
app
M
M=
(2 )
(sin2)
(2)
Conditions of x-ray diffraction.
Characteristic x-ray
X-ray wavelength
Diffraction plane
Diffraction angle 2
K filter
Peak determination
Stress determination
sin2 range, step
Detecter
Tube voltage, current
Peak count
Diameter of incident x-ray
V-K
0.250483 nm
Mo211
154.267 deg
Ti
Center of FWHM
2 -sin2 method
0.0 - 0.6 , 0.1step
PSPC
30 kV, 8 mA
2048
3 mm
3.3 Residual stress measurements by x-ray diffraction

A flattened particle had a thickness of about 4m and x-ray penetration depth
was about 0.9m. This means that the lattice strain of extreme upper part of the
coating surface could be measured in this study. Residual stresses were measured
on the principle of the -goniometer method
4.1 Penetration depth of quick-drying glue, and Youngs modulus

Fig.1 shows the cross-sectional secondary electron images of fractured surface of
the free-standing coating. The chemical composition of quick-drying glue
provides white image when exposed to the electron beam irradiation. The right
image indicates that the glue penetration depth was about 40m.
355
Youngs modulus of the coating was measured using three specimens with
different conditions. Tensile and compressive stresses were applied to as-sprayed
or polished surface, and thicknesses of substrates were 3mm and 5mm. As a
result measured EC was about 235 GPa on an average. EC seems to be lower than
that of bulk molybdenum (323 GPa [5]) because of the porosity of the coating.
Then Youngs modulus of quick-drying glue was 3.2 GPa. The penetration
depth of quick-drying glue was about 10% of the coating thickness, and the
Youngs modulus of the glue was about 1% of the coating. It means that the
effect of quick-drying glue on mechanical strain measurement on the sprayed
coating was negligible.
Coating surface
Coating surface
50m
As-sprayed
Increment of slope of 2T-sin2\

diagram M , deg
Figure 1:
50m
Glue applied
Cross-sectional secondary electron images of fractured surface of

the free-standing coating.
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
-400
-300
-200
-100 -0.1 0
-0.2
Bulk : Bulk molybdenum
-0.3
Sub3 : Substrate t=3mm
-0.4
Sub5 : Substrate t=5mm
Tens. : Tension
-0.5
Comp. : Compression
-0.6
-0.7
Bulk Tens.
-607 MPa/deg
Bulk Comp.
-547 MPa/deg
Sub3 Tens.
-536 MPa/deg
Sub3 Comp.
-630 MPa/deg
Sub5 Tens.
-523 MPa/deg
Sub5 Comp.
-530 MPa/deg
100
200
300
400
Mechanical applied stress Vapp , MPa

Figure 2:
The relationship between slope of 2-sin2 diagram M and

mechanical applied stress app.
4.2 X-ray stress constant K

In the case of as-sprayed surface slope M was not changed regularly. This is the
same result with a previous study [6] so that details will be skipped. Fig.2 shows

the relationship between M and app on polished coating specimens. These
specimens had different thicknesses of substrates as t=3mm (Sub3) and t=5mm
(Sub5). Bulk molybdenum (Bulk, t=3mm) was also tested. Increment of slope M
was evaluated differently depending on applied stress, tension or compression.
Applied stresses were calculated based on Youngs modulus of coating, EC.
Calculated K values are written in explanatory note in Fig.2.
Considering that the results of bulk molybdenum are criterion, K of thermal
sprayed molybdenum coating were almost the same with that, and independent
to experimental conditions. This means that the strain response was improved
because of polishing the coating surface.
To conclude the averaged K value -563 MPa/deg would be used for residual
stress determination.
Surface residual stress Vr , MPa
200
9 passes
4 passes
2 passes
1 pass
0.5 pass
150
B
100
50
0
0
10
15
20
25
Thermal expansion coefficient of substrate D, 10-6

Figure 3:
The relationship between TEC of substrates and surface residual

stress r.
4.3 The relationship between TEC and residual stress

Fig.3 shows the relationships between TEC of substrates and surface residual
stress of coatings. This result includes the changes of scanning passes of the
spraying gun. Residual stresses of all specimens were tension. Here, the melting
point of molybdenum is about 2893 K [7]. Therefore the temperature difference
was about 2273 K when the molten particles impacted the pre-heated substrate.
As a result the temperature difference caused significant shrinkage when the
coating began to cool. Therefore the tensile residual stress was generated.
Except for two plots signed A and B in fig.3, residual stresses tend to increase
with decreasing the TEC of substrate. It suggested that the shrinkage of substrate
contributed to that of lamella in cooling process so-called the bimetal.
357
Conclusion
(1) The effect of quick-drying glue on mechanical strain measurement on the

sprayed coating was negligible.
(2) The Youngs modulus of the coating was lower than that of the bulk
molybdenum.
(3) Linear strain response against applied mechanical loads was observed in the
case of polished coating surface.
(4) X-ray stress constant K of the coating and the bulk were almost the same.
(5) Residual stresses on the coating surfaces were of tensile type on all substrates
used in this study.
(6) Tensile residual stresses of the coatings increased with decreasing the thermal
expansion coefficient of substrate.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
Sulzer Metco, Automotive Solutions Kit, p.8, 2002.

XIE C X and OUYANG M-L, Wear, 137, No.2, p.159, 1990.
K. Akita, S. Tobe, Journal of the Society of Materials Science, Japan,
p.741, 2004.
Standard method of X-ray stress measurement for steel, The Society of
Materials Science Japan, p.71, 2002.
Technical Data Modulus of Elasticity of Metals, The Japan Society of
Mechanical Engineers, p.209, 2001.
S. Takahashi, I. Uchibayashi, H. Misawa and H. Suzuki, 34th Symposium
on X-ray Studies on Mechanical Behavior of Materials, The Society of
Materials Science Japan, p.61, 1998.
E. Yajima, R. Ichikawa, K. Furusawa, T. Miyazaki, T. Kozakai, Y.
Nishino, Mechanical and Metallic Materials for Young Engineers,
Maruzen CO., LTD, p.343, 2002.
359
Compressive residual stress generation process

by laser peening without pre-coating
H. Tanaka1, K. Akita2, Y. Sano3 & S. Ohya2
1
Graduate Student, Musashi Institute of Technology, Tokyo, Japan

Musashi Institute of Technology, Tokyo, Japan
3
Power and Industrial Systems Research and Development Center,
Toshiba Corporation, Yokohama, Japan
2
Abstract
Laser peening without pre-coating has been applied to water-immersed
specimens of high tensile strength steel. In order to understand the generation
process of compressive residual stress, specimens with various laser irradiation
patterns, i.e., single spot, line scanning and area scanning, were prepared.
Detailed distributions of residual stress on the specimens were measured using
synchrotron radiation. Large tensile residual stresses, which might be caused by
the thermal effect of the laser pulse, were observed in the center region on the
single spot. It decreased towards the edge of the spot, and changed to
compression around the spot. The compression became larger with the increasing
pulse numbers irradiated on the same spot. In the line scanning, a tensile residual
stress was observed in the final spot of the line, which decreased and changed to
compression as the distance from the final spot increased. The residual stress on
the area scanning was compression as a whole. It was estimated that the
compressive residual stress in the area scanning would be generated from the
overlapping effect of the compressive field made around each laser spot. The
residual stress generated by laser peening without pre-coating is considered to be
the superimposition of the tensile and the compressive components due to
thermal effect and plastic deformation, respectively.
Keywords: laser peening, surface residual stress, single pulse, line irradiation,
laser pulse density, x-ray stress measurement, synchrotron diffraction.
doi:10.2495/HPSM06035
Introduction
Laser peening is one of surface treatment technologies to improve fatigue

strength and resistance to stress corrosion cracking (SCC) [1, 2]. Laser peening
utilizes a mechanical interaction between the surface of material and plasma
caused by the irradiation of a nanoseconds-order laser pulse. Laser peening
introduces compressive residual stress in surface layer (more than 1 mm in
depth) and, therefore, has been applied to reactor core shrouds of nuclear power
plants to prevent SCC [3].
Our laser peening method is able to generate the compressive residual stress
on the material surface without any ablative coating [4], and the laser power is
relatively low. Therefore the treatment process is simple and the laser pulses can
be led easily through optical fibre. However, the residual stress generation
mechanism is not simple because the thin surface layer of material is exposed to
the high temperature plasma and it can cause a tensile residual stress on the
component.
Mukai et al. reported that the surface residual stress changes to the
compression side when the number of laser pulses irradiated per unit area was
increased. Recently, microscopic distributions of residual stress on a single laser
spot were measured using synchrotron radiation source [5, 6, 7]. As a result, the
tensile residual stress was observed on the surface of the center of the single laser
spot. The phenomenon that the residual stress changes from tension at the first
pulse to compression in the area irradiation is considered to be a key factor to
clarify the residual stress generation mechanism.
In this study, laser pulses were irradiated on the same spot to investigate the
effect of overlapping of laser pulses. A line scanning specimen was also prepared
to investigate the effect of offset over lapping of laser pulses. Microscopic
distributions of residual stress were measured on the surface of these specimens
to discuss the residual stress generation mechanism.
Fundamental mechanism of residual stress generation by

laser peening
Fig. 1 shows the fundamental process of laser peening. Laser beam focused by a
lens is irradiated on a specimen surface in water. The material surface layer
becomes plasma if the power density of the laser pulse at the material surface
exceeds the threshold of the ablation of the material. The inertia of water
prevents the expansion of the plasma, and the energy of the plasma concentrates
in a narrow space. The resulting plasma pressure becomes 10100 times higher
than that in air and it reaches 1 to 10 GPa [8]. The shock wave is generated by
this pressure, and it propagates in the material. Plastic deformation occurs near
the surface of the material due to the dynamic stress of the shock wave, and the
material is hardened through the process. The compressive residual stress is
generated by the elastic restraint from the un-deformed region. The generation of
the compressive residual stress has been explained basically by this mechanism,
361
since the depth profile of the residual stress by laser peening is well reproduced
by elastic-plastic dynamic simulations.
Laser pulse
Lens
Plasma
Water
Compression
Test sample
During laser irradiation
Figure 1:
After laser irradiation
Fundamental process of laser peening.
3.1 Material and specimens

Material used in this study is high tensile strength steel, HT1000. Table 1 and
Table 2 show the chemical compositions and the mechanical properties,
respectively. Fig.2 shows the micrograph of the HT1000 surface. The grain size
is around 10 m or less, so it is suitable for the specimen of microscopic X-ray
stress measurements. The dimensions of the specimens are 20 mm 80 mm with
the thickness of 15 mm and 28 mm 30 mm with the thickness of 17 mm. To
minimize the effect of residual stress by machining, the electrolytic polishing
was applied on the surface of the specimens prior to laser irradiation.
Table 1:
Si
Mn
0.13
0.25
0.92
Table 2:
Chemical compositions of HT1000.
0.011 0.001
Mo
Nb
Cr
0.39
0.02
0.04
0.87
Mechanical properties of HT1000.
Yield strength [MPa] Tensile strength [MPa]

965
1055
Elongation [%]
23
3.2 Laser peening conditions

The schematic of experimental setup for laser peening is shown in Fig.3. The
fundamental wave of a Q-switched Nd:YAG laser ( = 1.06 m) was frequencydoubled to a water-penetrable wave ( = 532 nm) by a second harmonic
generator with a nonlinear optical crystal. The laser beam was focused by a lens

and irradiated on a specimen in a water jacket through an optical window. The
diameter of the laser spot was about 1 mm. The specimen was fixed on a
specimen holder and driven to x- and y-directions in the water jacket during laser
irradiation.
30 m
Figure 2:
Micrograph of HT1000 specimen surface.

Mirror
Frequency-doubled
Nd:YAG laser
X-Y table
Test sample
Lens
Mirror
Window
Water jacket
Figure 3:
1 pulse 4 pulses 10 pulses 40 pulses
15
20
Experimental setup of laser peening.
27
18
18
80
Figure 4:
Scheme of pulse laser irradiation on the same spot on HT1000

specimen.
Fig.4 illustrates the single spot specimen where laser pulses irradiated at the
same point 1, 4, 10, and 40 times. The line and area scanning were performed as
363
shown in Fig.5 (a) and (b). The lines with pulse densities of 1, 5, 10 and 100
/mm were prepared in the line scanning, changing the scanning speed. In the area
scanning, an area of 10 10 mm2 is irradiated by scanning laser pulses twodimensionally, as shown in Fig.5 (b). The coverage was 800% which was
defined in the same way in shot peening. The laser irradiation conditions are
summarized in Table 3.
17
17
(D)100 /mm
(C)10 /mm
28
(B) 5 /mm
28
Laser irradiated area

(10mm10mm)
Y
X
(A) 1 /mm
Figure 5:
30
30
(a) Front surface.
(b) Back surface.
Laser pulses irradiation scheme (Numerals in (a) mean the number

of laser pulses per unit length. Arrows mean the laser scanning
direction).
Table 3:
Laser peening conditions on HT1000 specimens.
Material
Pulse energy
Spot diameter
Pulse duration
Coverage
Pulses irradiated per
unit length
Single pulse
HT1000
215 mJ
1.0 mm
8 ns
800% ( Fig,5 (b) )
1 /mm (A)
5 /mm (B)
10 /mm (C)
100 /mm (D)
1, 4, 10 and 40 pulses
3.3 Residual stress measurement

The microscopic residual stress distributions near single laser pulse spot, laser
irradiated line and laser irradiated area were measured on the HT1000 specimens
using the synchrotron radiation source at the beam line 3A of Photon Factory,
KEK, Tsukuba, Japan. Synchrotron radiation is suitable for the microscopic

residual stress measurement required in this research because the intensity of the
radiation is much higher than that of the typical laboratory X-ray. The
wavelength used was 0.228 nm. -Fe211 diffraction was measured. The X-ray
irradiated area was 0.2 mm or 0.5 mm in diameter. Residual stress was derived
by the sin2 method. X-ray stress constant, K, is assumed to be the mean value
of the calculated ones using Reuss and Voigt models and the elastic compliance
of a single crystal of -Fe. The X-ray stress measurement conditions are
summarized in Table 4.
Table 4:
Conditions of X-ray stress measurement.
X-ray source
Material
Diffraction plane
Wavelength
Diffraction angle
Detector
Irradiated area
X-ray stress constant, K
Synchrotron radiation
KEK, PF BL3A
HT1000
- Fe211
0.228 [nm]
154 [deg]
PSPC
0.2 [mm] or 0.5 [mm]
-353 [MPa/deg]
4.1 Surface residual stress distributions on the overlapped laser spot

Fig. 6 shows the schematic illustration of the measuring positions of surface
residual stress in and around the overlapped laser spot. The X-ray irradiated area
was 0.2 mm in diameter.
Y
X
Laser spot 1.0mm

: Stress measured positions, 0.2 mm
Figure 6:
Schematic illustration of residual stress measurement positions in

and around a single pulse laser irradiated spot (see Fig. 7).
Fig. 7 shows the surface residual stress distributions along X-axis of Fig. 6.
Large tensile residual stresses of about 600-800 MPa were observed in all the
365
laser spots with 1 to 40 pulses. The tensile residual stresses decreased toward the
edge of laser spot, and changed to compression in the outside of the spot. The
compression was about -200 MPa in the single pulse spot. When the number of
laser pulses increased, the compression increased at 0.3mm outside from the
edge of laser spot. The maximum compressive residual stress reached to -600
MPa for 40 pulsed spot.
Residual stress, X [MPa]
Laser spot
1000
800
600
400
200
0
-200
-400
-600
-800
1 pulse
4 pulses
10 pulses
40 pulses
Figure 7:
0.5
Distance from spot center, X [mm]
1.5
Surface residual stress distribution in and around the laser spots

irradiated 1, 4, 10 and 40 pulses at the same position.
Laser scanning direction
1 pulse/mm (A)
Final spot
5 pulses/mm (B)
10 pulses/mm (C)
100 pulses/mm (D)
X
: Stress measured positions, 0.5
mm
Figure 8:
Schematic illustration of residual stress measurement positions on

the center line of laser irradiated line.
4.2 Surface residual stress distributions in laser irradiated line and area
The interaction of adjoined laser spots was examined using the line-scanned
specimen. Fig.8 shows the positions of residual stress measurement on the laser

irradiated line. Fig.9 shows the surface residual stress distributions along the
lines. The horizontal axis of the figure shows the distance from the center of the
final laser spot center of each line. The X-ray irradiated area was 0.5 mm in
diameter. The tensile residual stresses were observed near the final spot for the
cases A, B and C. The tensions decreased with increasing the distance from the
final spot. The tensile residual stress was decreased with increasing the number
of laser pulses per unit length, and it changed to the compression side in case of
D with the 100 pulses/mm irradiation density. On the area-scanned specimen
with the coverage of 800%, the surface residual stress was compression with
about -150 MPa.
Final spot
Residual stress, X [MPa]
800
1 pulse/mm (A)
5 pulses/mm (B)
10 pulses/mm (C)
100 pulses/mm (D)
600
400
200
0
-200
-400
0
0.5
1.5
2.5
Distance from edge of irradiated line, X [mm]

Figure 9:
Surface residual stress distributions along the center of laser

irradiated line.
Discussion
The residual stress at the center of the single spot was tensile even if laser pulses
were irradiated 40 times at the same position (Fig. 7). Meanwhile when the laser
pulses were scanned in a line as shown in Fig. 9, tensile residual stresses were
rapidly decreased with increasing distance from the final spot. These facts mean
the overlapping of laser pulses at the same position dose not generate
compression at the center of laser spot, on the other hand the off-center
overlapping of laser pulses decreases the tension as shown for the lines A, B and
C in Fig. 9. The reason of this phenomenon is considered that the compressive
residual stress region around laser spot overlaps with tension region at the spot
center of the adjoining spot by off-center overlapping.
When the irradiation density was increased to 100 pulses/mm (Line D), the
residual stresses changed to compression. However, although the irradiation
density per unit length of the area irradiation was 3.2 pulses/mm, the residual
stress was compression. These facts mean that an off-center overlapping of laser
scanning lines is also important to generate higher compressive residual stresses.
367
Our laser peening process does not use any pre-coating. Therefore, the material
surface is exposed to high temperature plasma, and the resulting surface residual
stress depends mainly on the balance of the compressive stress component
generated by cold working due to high-pressure laser plasma and the tensile
stress component generated by rapid cooling due to thermal effect. Thermally
affected depth is only several micrometers [9] and the effect seems to be reset
pulse by pulse. However, the plastic deformation caused by cold working can be
accumulated [10]. Therefore, the tensile component might be constant and the
compressive component increases when the laser pulse density increases. Thus,
the whole stress level is changed to the compression side, when the number of
laser pulses increases, as shown in Fig. 10.
Residual stress
th
Tension
1
Laser pulse density
r = th + sw
Compression
sw
r: Surface residual stress on laser peened area
th: Tensile stress component generated by thermal effect
sw: Compressive stress component generated by
compressive residual stress around laser spot
Figure 10:
accumulation
of
Schematic of the residual stress generation mechanism on laser

peening without pre-coating.
Conclusion
Overlapped single laser pulse, line irradiation and area irradiation were
performed on high tensile strength steel, HT1000 and the residual stress
distributions near the irradiated spot, line and area were measured using a
synchrotron radiation source. The results obtained through the experiments are
summarized as follows:
(1) Residual stress was tension in the single spot even if 40 laser pulses were
irradiated at the same position. Compressive residual stress was observed in the
outside of the spot.
(2) Off-center overlapping of laser pulses generates the compressive residual
stress on the surface of HT1000 because the compressive stress region at the

outside of laser spot would change the tensile residual stress at the spot center to
compression.
Acknowledgement
This work has been performed under the approval of the Photon Factory
Program Advisory Committee of the High-Energy Accelerator Research
Organization, Japan (Proposal No. 2003G032).
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
Y. Wakabayashi, K. Masaki, Y. Ochi, T. Matsumura, Y. Sano, T. Kubo,

Japan Society Of Mechanical Engineers M&M2003 pp.283-284 (2003).
P. Peyre, C. Braham, J. Ldion, L. Berthe and R. Fabbro, Journal of
Materials Engineering and Performance, 9, pp.656-662 (2000).
Y. Sano, et al, Proc. of the 7th Int. Symp., JWS, pp.439-444 (2001).
Y. Sano, M. Kimura, M. Obata, N. Mukai, A. Sudo and S. Shima, 6th
International Conference on Nuclear Engineering (ICONE-6236), 1998.
Y. Yoshioka, Proc. 38th Symposium on X-ray Studies on Mechanical
Behaviour of Materials, 83 (2002).
K. Akita, Y. Sano, T. Kubo, Y. Yoshioka and H. Suzuki, Int. Conf. On
Advanced Technology in Experimental Mechanics 2003 (ATEM03),
(2003).
K. Akita, H. Tanaka, Y. Sano and S. Ohya, Material Science Forum, 490491, pp.370-375 (2005).
Y. Sano, N. Mukai, K. Okazaki and M. Obata, Nuclear Instruments and
Methods in Physics Research B pp.432-436 (1997).
Y. Sano, M. Kimura, K. Sato, M. Obata, A. Sudo, Y. Hamamoto, S.
Shima, Y. Ichikawa, H. Yamazaki, M. Naruse, S. Hida, T. Watanabe, Y.
Oono, 8th International Conference on Nuclear Engineering (ICONE8441), 2000.
Y. Sano, N. Mukai, M. Yoda, K. Ogawa and N. Suezono, Materials
Science Research International, Special Technical Publication-2, pp.453458 (2001).
369
The scale effect of roughness

in contact problems
S. Mezghani1, A. Jourani1 & H. Zahouani2,3
1
Laboratoire de Tribologie et Dynamique des Systmes,

UMR CNRS, 5513, France
2
Ecole Nationale dIngnieurs de St Etienne, France
3
Ecole Centrale de Lyon, France
Abstract
In this paper we have used the Hlder exponent to characterise the scale of
roughness and to study the scale effect of high spatial frequencies on elastic
contact between solids. The mathematical approach shows that the Hlder
exponent of roughness is a sophisticated tool for modelling realistic surface
roughness at different scales of observation. The incidence of Hlder exponent
on the prediction of pressure and displacement is studied in elastic contact
between a smooth plane and rough surface.
Keywords: Holder exponent, scale effect, elastic contact.
Introduction
Surface topography plays an important role in a multitude of physical and

tribological phenomena such as contact mechanics, friction, adhesion, wear,
wettability, lubrication, etc. Surface topography causes discrete contact points,
when two rough nominally flat surfaces are brought together, the real area of
contact is the accumulation of the area of the individual contact points. For most
metals at normal loads, this will be only a small percentage of the apparent
contact area. Typical models of surface deformation are either elastic, plastic or
mixed elastic-plastic, and can be represented schematically as a function of
surface topography and material constants (h, s, Rc , E, H) with h the root
mean square of height, s the root mean square of summits, Rc the mean radius of
summit curvature, E and H are respectively the Youngs modulus and the
hardness of the solid which are the intrinsic parameters of the solid. On the other
doi:10.2495/HPSM06036

hand the principal statistical parameters of roughness, such as variance of height,
slope and curvature are unfortunately not always independent of length scales.
The definition of slope or summit of surface roughness systematically involves a
high wave vector cut off due to the measurement process itself and not to the
surface characteristics. Consequently, instruments with different resolutions and
scan lengths yield different values of these statistical parameters for the same
surface. Therefore it is very important to characterize rough surfaces by intrinsic
parameters which are independent of the sampling length or area.
Hlder exponent which indicates the scale of perturbation of surface
topography has been widely used in recent years [1-5], this is due in part to the
observations that fractal geometry can reflect an intrinsic property of random
phenomena and can be applied to characterize surface topography and contact
mechanics. Majundar and Bhushan [5,7] explored these applications using the
theory developed by Mandelbrot [6] concerning the cumulative distribution of
islands on the earths surface relief, which follows the power law N a 2 ,
where N is the total number of islands whose area larger than a, and D is the
fractal dimension of its coast line ( 0 < D < 1 ).
In this work we have used the Hlder exponent to characterize the scale of
roughness and to model the effect of roughness scale in contact mechanics.
In the second part we have studied, through experiments, the mechanism of
asperity deformation using two numerical approaches based on geomorphologic
characterization and fractal geometry using the concept developed by
Mandelbrot.
Elastic model of rough contact
The majority of the methods used to determine the relationship between normal
load and displacements, except finite elements methods, are based on the
assumption of elastic solid, homogeneous and semi - infinite The basic
assumptions necessary to use the continuous equations of elasticity:
The slopes of roughness are weak;
The contact area is small in front of other dimensions.
With zero loads, surfaces of the two bodies in the immediate neighbourhoods
of the point of contact can be comparable with paraboloids of equations:
z1 = A1x2 + A2xy + A3y2
z2 = B1x2 + B2xy + B3y2
(1)
Each surface is characterized at the point of contact by its principal radii of

curvature R1 and R'1, R2 and R'2 is the angle which forms between them.
Constant A and B depend on the size of the principal curves of surfaces and
1
1
1
1
.
+
+
+
R1 R '1 R2 R '2
the angle with A + B =
By choosing an orientation of axes X and Y, X1 and Y1, the terms in xy

become negligible. The expressions of the equation (1) become:
z1 =
371
1
1 2
1 2
1 2
y
x +
y , z2 = ' x22 +
' 1
'' 1
'' 2
R
R
2 R1
2 R1
2
2
2
(2)
The geometry of the contact between two unspecified solids is thus brought back
to the problem of the contact between two paraboloids:
h( x, y ) = Ax 2 + By 2 =
1 2
1 2
x +
y
2 R1
2 R2
(3)
The equations of linear elasticity make it possible to connect displacement wi(x,

y) of a M(x point, y) of the surface of a semi elastic body infinite, with the
pressure p(x', y') in Me (x', y') pertaining to the surface of contact, by means of
the following integral equation of Boussinesq:
1 i 2
wi ( x , y ) =
Eequ
Ac
p( x ', y ')dx ' dy '
( x x ')
+ ( y y ')
(4)
with Ac the surface of contact
S1
Z1
S2
Figure 1:
w2(x,y)
w(x,y)
Z2
Contact
region
Rgion
du
Rgion
ducontact
contact
w1(x,y)
General diagram of the areas of contact between two solids.
So if one calls w(x, y), the difference between the M1(x point, y, z1) of the
surface of the solid 1 (S1) and the M2(x point, y, z2) of the surface of solid 2
(S2), the geometry enables us to write the following relation:
w( x, y ) = w1 ( x, y ) + w2 ( x, y ) =
x2
y2
2 R1 2 R2
(5)
where w1(x, y) and w2(x, y) are relative displacements on the two solids, R1et R2
the principal radii of curvature of the two solids.
The determination of displacement in a point of surface due to the field of
pressure is carried out by the superposition of the influence of all the efforts.
Displacement at the point ij can then be written in the form:
wij = Cijkl pkl
(6)
with C ijkl the matrix of the coefficients of influence.
P(x,y)
P(x,y)>0
P(x,y)=0
Figure 2:
Pressures field at the contacts points between the two bodies.
The problem is thus to find pressures, such as pij > 0 , such as
C ijkl p kl + h( x, y ) = 0
(7)
If one consider the pressure as an input signal and displacement like a response
of exit by the following relation to a dimension (Johnson 1985):
w( x) = h( x) p( x)
(8)
This equation is well defined in the space field in the form of a product of
convolution: If the pressure applied is form
p( x) = B0 sin( 2x / )
(9)
The obtained displacement has the following expression (Johnson 1985)
w( x) =
(1 2 )
B0 sin( 2x / ) + cst
E
(10)
what gives like transfer function of the system in the field of the space
frequencies:
H ( ) =
2(1 2 )
E
(11)
and in the field of the space frequencies one has like response
W ( ) = H ( ) P( ) =
2(1 2 ) P( )
(12)
In addition, by using the properties of the Fourier Transform one can write that
the derivative of displacement can be written in Fourier domain as
W ( ) = iW ( )
by replacing in the equation of the pressure one can write:
(13)
P( ) =
F ( ) =
373
E
W ( )
i 2(1 2 )
(14)
E
W ( )
i 2(1 2 )
(15)
These last relations are important characteristics of contact problems. They

imply that the contact pressure and slope of displacement as well as the load and
displacement are two pairs of parameters which have similar frequency
composition.
The passage in the spectral field makes it possible to solve this last equation,
while subjecting the functions a two-dimensional transform of Fourier. What
makes it possible to write a relation between the spectrum of the pressures and
the spectrum of displacements as [1, 2 ,3] :
W ( x , y ) =
2
2(1 ) P ( x , y )
(16)
2
2
where = x + y represent a space frequency.
The determination of the field of pressure p(x, y) is determined by opposite
transform of Fourier for each level of bringing together :
-1
p ( x, y ) = TF
2
2(1 )
W ( x , y )
(17)
The spectrum of pressure is thus spread out between the low and high
frequencies:
vx =
1
1
1
1
, vy =
mm-1 , hx =
mm-1
, hy =
Lx
Ly
2x
2y
with Lx, Ly , the size of the surface, x , y the sampling steps in the x and y
directions, which initially define rough surface before contact.
For normal macroscopic imposed load F, one numerically brings closer two
surfaces and to solve the equation for each level of bringing together . For each
normal position of the rigid plan, one determines contact pressures p(x, y) by
using the direct and inverse Fourier transform of the contact equations.
Convergence is reached when the imposed effort is equal to the sum of the local
efforts calculated for each iteration:
F = pij ( x, y ) xy
i, j
(18)
Hlder exponent of rough surfaces
As the details of roughness: Z(x) depends on the length scale, we assume each
realization of Z(x) to be a continuous, but non-differentiable function. It means
that the presence of any small roughness elements may prevent us from reaching
a satisfactory limit of
(Z ( x + ) Z ( x) ) as 0
(19)
with the sampling step of roughness measurement

A simple way to obtain this behaviour for a function Z(x) is to assume that
the increment of Z(x) is related to by the self-affinity relation:
Z ( x + ) Z ( x) H , 0 < H < 1 0
(20)
lim 0 H 1 , only exists if H=1.

The derivative of Z, which is proportional to the limit, for 0<H<1, this derivative
is infinite, although the function remains continuous, and for H=0 the function
itself becomes discontinuous. So by varying it from 0 to 1, parameter H
characterizes the transition from a non continuous to a differentiable function,
the range 0<H<1 corresponding to non differentiable functions which become
smoother as H increases. Thus H can be considered as an indicator of roughness.
Since we are concerned with functions which can be considered as representing
random processes, we state the self-affinity relation in terms of variance of the
increments of Z:
(Z ( x + ) Z ( x) )2
2 H 0<H<1, 0
(21)
This relation can be generalized to dimensions E > 1 in the following way
(Z ( x + ) Z ( x) )2
2H
, 0 < H < 1, 0
where x represents a point in E-dimensional Euclidean space RE and
(22)
stands
for the usual norm in this space. This is not a new condition; it is proven for all
values by the fractional Brownian motions.
The spectral exponent
From the previous relation it appears that the functions

Z(x+)-Z(x) and
( Z ( x + s ) Z ( x))
sH
(23)
Considered as functions of , have the same law of probability. If we assume

that Z(0) = 0, then function Z(x) verifies the statistical self- affinity relation
Z ( sx ) = s H ( Z ( x )
(24)
It means that the two functions Z(x) and Y ( x ) =
375
Z ( sx )
are statistically
sH
indistinguishable.
Voss (1985) showed that the fractal dimension is related to parameter H by
the relation D = E+1-H where E is the dimension of the Euclidean space RE.
These functions, which correspond to a non-stationary random process with
stationary increments, present many other interesting properties. For instance
their power spectrum S() presents a well-defined decrease following the power
law
S ( ) =
= 2H + E
2H +E
(25)
The power spectrum of roughness verifies another self-affinity relation
S (k ) =
1
k
2H +E
S ( )
(26)
If one makes the analogy with the relation between the spectrum of pressure and
the spectrum of the slopes of displacements which vary in the same space
frequency band, one can notice that the law of power spectrum of pressure
depends on the Hlder exponent as following
P ( ) =
E
W ( )
i 2(1 2 )
P (k ) =
1
k
2H +E
P( )
(27)
(28)
This relation, shows well the effect of the exhibitor of Hlder on the spectrum of
pressure, and consequently on the local singularities of pressures and
displacements which will induce local plasticity.
The effect of the Hlder exponent of roughness on elastic

contact
In order to study the influence of the Hlder exponent of roughness on the

parameters of elastic contact: pressure, displacement and real contact area, we
propose to generate three rough surfaces parameterised in roughness by their
Hlder exponent. Each of those surfaces is the sum of a periodic surface
(sinusoidal) and a random rough surface generated by the random displacement
model [14-15] with different roughness Hlder exponent H=0.1, H=0.5, H=0.8
(figure 3). Those surfaces obtained (figure 4 {a1, a2, a3}) were put in contact
under a pressure of 400 MPa with a rigid plan.
Pressure fields are calculated for each of these three rough surfaces as using
the model of the elastic contact (figure 4). We notice that the pressure field
gotten is as much more discontinuous than the surface is rougher.
(a)
Figure 3:
S1
S2
S3
Figure 4:
(b)
(c)
Three examples of simulation by the algorithm of the midpoint

method with random successive additions (a) H=0.1 (b) H=0.5 (c)
H=0.8.
(a1)
(b1)
(c1)
(a2)
(b2)
(c2)
(a3)
(b3)
(c3)
Roughness surfaces: (a1) H=0.1, (a2) H=0.5, (a3) H=0.8; Pressure

field: (b1) H=0.1, (b2) H=0.5, (b3) H=0.8; deformed surfaces: (c1)
H=0.1, (c2) H=0.5, (c3) H=0.8; {Applied pressure 400 MPa}.
The following table (Table 1) regroups the whole of contact parameters below
(Bearing surface area and mean contact pressure) of the three different rough
surfaces.
Table 1:
377
Contact parameters for various roughness surfaces.
Surface roughness
S1 (H=0.1)
S2 (H=0.5)
S3 (H=0.8)
Bearing area (%)

11.66
27.78
28.60
Mean Pressure (GPa)

3.40
1.41
1.38
These results (Table 1) show that the roughest surface (S1) which has the
lower Holder exponent is the more resistant in contact. Indeed, it presents the
lowest bearing area surface (11.66%) and the most important mean pressure
(3.40 GPa). This proves the narrow relation between punctual pressure fields
singularity and the Holder exponent indicator of roughness.
Conclusion
The incidence of the Hlder exponent of roughness is important in contact

mechanics, wettability and lubrication of rough material, where the knowledge
of the surface bearing area, the developed area or void volume is directly related
to the scale of observation. In contact mechanics there are two incidences. The
first is related to the size of the spot area and surface bearing area, the second is
connected to the evolution of the void and material volume during contact as a
function of observation scale. The mathematical modelling study shows that the
Hlder exponent of roughness is a sophisticated tool for modelling realistic
surface roughness at different scale of observation. It can be used as the input
data in contact mechanic modelling. This scale parameter can be used as an
indicator of the real length, area and volume of rough material, independently of
the scale of observation. This scale parameter is very promising to simulate the
behaviour of rough morphology in static contact.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
Johnson K.L., N J Mech. Phys. Solids 1970.

Ju Y., Farris T.N., Spectral analysis of two dimensional contact problems,
Transactions of the ASME, Vol 118, pp 320-329, 1996.
Liu Sh., Wang Q., Liu G., A versatile method of discrete convolution and
FFT for contact analyses. Wear 243 pp. 101-111, 2000.
Gagnepin J.J and Roques-Carmes C., Fractal approach to twodimensional and three-dimensional surface roughness, Wear, Vol 109,
pp.119-114, 1986.
Majunder A. and Bhushan B., Role of fractal geometry in roughness
characterisation and contact mechanics of surfaces, ASME J. Tribology,
Vol. 112, pp.205-216, 1990.
Majunder A. and Tien C.L, Fractal characterisation and simulation of
rough surfaces, Wear, Vol 136, pp. pp. 313-327, 1990.
Majunder A. and Bhushan B., Fractal model of elastic-plastic contact
between rough surfaces, J. Tribology, Vol 13, pp.1-11. 1991.

[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
Lopez J., Hansali G., Zahouani H., Leboss J.C., Mathia T., 3D Fractalbased
characterisation
for
engineered
surface
topography,
Int.J.Mach.Tools Manufact, Vol 35, pp.211-217, 1995.
Mandelbrot B., Stochastic models for the Earths relief, the shape and the
fractal dimension of coastlines, and the number-area rule for islands,
Proceedings of the National Academy of Science, Vol 72, pp.3825-3838,
1975.
Richardson L.F., The problem of contiguity: An appendix of statistics of
deadly quarrels, General Systems Yearbook 6, pp.687-689, 1961.
Mandelbrot B., How long is the coast of Britain - Statistical self-similarity
and fractional dimension, Science, Vol 155, pp.636-638, 1967.
Brown S.R. and Scholz C.H., Broad bandwidth study of the topography of
natural rock surfaces, J. Geophys. Res., 90, pp. 12575-12582. , 1985.
Felder J., Fractals, Plenum Press, New York, 1988.
Mandelbrot B. and Van Ness J.W, Fractional Brownian motions,
fractional noises, and applications, SIAM Review 10, 422-437, 1968.
Russ John C., Fractal surfaces, Plenum Press, New York, 1944.
Saupe D. and Peitgen H.O., The science of fractal images. SpringerVerlag pp.82-91, 1988.
Zahouani H., Vargiolu R., Fractal models of surface topography and
contact mechanics, Vol28, N 4-8, pp 517 534, 1998.
Section 5
High performance
concretes
381
Copper slag as fine aggregate for

high performance concrete
K. S. Al Jabri
Department of Civil and Architectural Engineering,
Sultan Qaboos University, Sultanate of Oman
Abstract
This paper presents results from an experimental investigation carried out to
study the potential use of copper slag as fine aggregate on the strength of both
normal and high strength concrete. Concrete mixtures were prepared using
different proportions of copper slag as partial and full replacement of fine
aggregate. The percentage of copper slag added by weight ranged between 10100% of sand used in concrete.
For each concrete mixture, six
150mmx150mmx150mm cubes, three 300mmx150mm dia. cylinders and three
100mmx100mmx500mm prisms were cast. Density, compressive, tensile and
flexural strengths were determined at 28-day of curing. Cube compressive
strength was also determined at 7-day of curing. Results demonstrated that there
is general an increase in the density and workability of both normal and high
strength concretes as copper slag quantity increases. Also results showed that
the compressive strength of concrete is generally improved, compared with the
control mix, with the increase of copper slag up to a certain copper slag content
beyond which the strength generally reduces. Mixes with large copper slag
percentage showed signs of bleeding and segregation due to the significant
increase of workability.
Keywords: copper slag, fine aggregate, concrete, strength, workability, density.
Introduction
Aggregates are considered one of the main constituents of concrete since they
occupy more than 70% of the concrete matrix. In many countries there is
scarcity of natural aggregates that are suitable for construction whereas in other
countries the consumption of aggregates has been increased, in recent years, due
doi:10.2495/HPSM06037

to the increase demand by the construction industry. In order to reduce
dependence on natural aggregates as the main source of aggregates in concrete,
artificially manufactured aggregates and artificial aggregates from industrial
wastes provide an alternative for the construction industry. Therefore utilisation
of aggregates from industrial wastes can be alternative to the natural and
artificial aggregates.
Copper slag is an industrial by-product material produced from the process of
manufacturing copper. It has been estimated that approximately 24.6 million ton
of slag is generated from world copper industry [1,2]. In the Sultanate of Oman,
approximately 60,000 tons of copper slag is produced every year by Oman
Mining Company. Although the majority of the produced copper slag is used in
the sand blasting industry and in the manufacturing of abrasive tools, the
remainder is disposed without any further reuse or reclamation. There are many
mechanical and chemical characteristics that qualify copper slag to be used in the
concrete as partial replacement of Portland cement or a substitute of aggregates.
For example, copper slag has number of favourable mechanical properties to be
used as an aggregate such as excellent soundness characteristics, good abrasion
resistance and good stability. Also, copper slag exhibits pozzolanic properties
since it contains a low content of CaO, and it can possesses cementitious
properties as CaO content increases or under activation of NaOH. Use of copper
slag in concrete industry as replacement of cement or/and fine aggregates can
have the benefit of reducing the costs of disposal and minimising environmental
pollution [1,2].
Little research work has been conducted to investigate the potential use of
copper slag as fine aggregate on the properties of normal concrete. Akihiko and
Takashi [3] investigated the use of slag from copper smelting as a fine aggregate
in concrete. From mortar strength tests with a cement/slag/water ratio of
1/2/0.55, the ball milled slag gave a higher strength. The effects of using several
types of slag on mortar and concrete reaction, reinforcing steel corrosion,
abrasion, workability and slump, shrinkage, and freezing and thawing
characteristics were examined. The effect copper slag as replacement of fine
aggregates on the strength, setting time and durability of concrete mixtures was
also studied by Toshiki et al. [4]. Recently, Demirboa and Gl [5] used blast
furnace slag aggregates to produce high strength concrete.
Different
water/cement ratios were used to determine the 7- and 28-day compressive
strength and other properties. Results showed that compressive strength of blast
furnace slag aggregate concretes were approximately 60-80% higher than control
concretes containing no slag for different water/cement ratios. It was concluded
that blast furnace slag aggregate, in combination with other supplementary
cementitious materials, could be utilised in making high strength concretes.
Research objectives
The main objective of this study was to investigate the use of copper slag as
partial and/or full replacement for fine aggregate in normal and high strength
concrete mixtures. The following were specific tasks:
383
(1) Evaluate the effect of copper slag addition on the workability

and density of normal and high strength concretes
(2) Conduct compressive, tensile and flexural strength testing on
concrete mixtures
Materials
3.1 Cement
The cement used in this study was ordinary portland cement (OPC) purchased
from Oman Cement Company. This cement is the most widely used one in the
construction industry in Oman.
3.2 Coarse and fine aggregates
Coarse aggregates (i.e. 20 mm and 10 mm) and fine sand were purchased from a
nearby crusher in Al-Khoudh area, which are typically the same materials used
in normal concrete mixtures. The gradation test conducted on aggregates
showed that they met specifications requirements.
3.3 Copper slag
Copper slag is a by-product material produced from the process of
manufacturing copper. As the copper settles down in the smelter, it has a higher
density, impurities stay in the top layer and then are transported to a water basin
with a low temperature for solidification. The end product is a solid, hard
material that goes to the crusher for further processing. Copper slag used in this
work was brought from Oman Mining Company, which produces an annual
average of 60,000 tons.
3.4 Silica fume
The silica fume used in the production of high strength concrete was supplied
and added to the mix in a powder form (Elkem Emsac 500s).
3.5 Superplasticizer
In order to improve the workability of high strength concrete, superplasticizer in
the form of a polynaphthalene sulphonate-based admixture (conplast SP430) was
used. This had 40% active solids in solution.
Laboratory testing program
4.1 Mix design and sample preparation

The mix proportions chosen for this study are given in Table 1 for both normal
strength concrete (NSC) and high strength concrete (HSC). The constituents
were weighed in separate buckets. The materials were mixed in a rotating pan in
accordance with ASTM C192-98. The overall mixing time was about 4 minutes.

The mixes were compacted using vibrating table. The slump of the fresh
concrete was determined to ensure that it would be within the design value and
to study the effect of copper slag replacement on the workability of concrete.
The specimens were demoulded after 24 hours, cured in water and then tested at
room temperature at the required age.
Table 1:
Mix proportions and water-to-cement (w/c) ratios for normal and

high strength concretes.
Type
Cement
NSC
HSC
416
400
Mix proportions (kg/m3)

Silica Sand
10mm
20mm
fume
Agg.
Agg.
721
338
790
44
710
1190
-
w/c
Wat
er
207
140
ratio
SP
l/m3
0.5
0.35
7.9
To determine the unconfined compressive strength, six cubes

(150mmx150mmx150mm) were cast for each mix and water-to-binder ratio, and
three samples were tested after 7 and 28 days of curing. Three 150 mm diameter
x 300 mm long cylinders were prepared for each mix in order to determine the
28-day tensile strength of concrete. Also, to determine the flexural strength
(modulus of rupture) for each mix, three 100mmx100mmx500mm prisms were
cast and tested after 28 days of curing.
4.2 Testing procedure
After curing, the following tests were carried out on the concrete specimens:
7- and 28-day cube compressive strength test was conducted in

accordance with BS 1881: Part 116 using a loading rate of 2.5 kN/s;
28-day cylinder tensile (splitting) strength test was done in accordance
with ASTM C496-96 using a loading rate of 2 kN/s; and
28-day flexural strength test was conducted in accordance with ASTM
C78-94 using a simple beam with third point loading at a loading rate of
0.2 kN/s.
All tests were conducted using a DARTEC compression machine.
Test results and discussion
The properties of normal and high strength concretes in terms of density, slump,
compressive strength, tensile strength and modulus of rupture are shown in
Tables 2 and 3, respectively.
The effect of copper slag addition as replacement of fine aggregate on the
density of normal and high strength concretes is shown in Fig. 1. It can be seen
from Fig. 1 that there is general increase in the density of both normal and high
strength concretes with the increase of copper slag replacement in the concrete.
The density of concrete was increased by almost 5% for both concrete types.
This is mainly due to the higher specific gravity of copper slag which was 3.4
compared to fine aggregate which has a specific gravity of 2.8.
Table 2:
Properties of normal concrete at 7- and 28-days of curing.
Mix
No.
Mix
Type
Control
(100% S)
10%CS +
90S
20%CS +
80%S
30%CS +
70%S
40%CS +
60%S
50%CS +
50%S
70%CS +
30%S
90%CS +
10%S
100%CS
+ 0%S
2
3
4
5
6
7
8
9
385
Strength (MPa)
(Fcu)+
(Ft)+
Density
(kg/m3)
Slump
(mm)
(Fcu)*
2524
50
33.2
44
7.7
2515
90
37.1
44.9
3.5
7.2
2540
90
39.6
48.5
3.7
7.2
2558
75
37
48.9
3.2
6.9
2550
145
36.2
48.1
3.8
6.5
2560
180
39.5
53.1
4.1
7.3
2601
185
37.2
46.6
3.6
6.3
2597
200
37.9
50.1
3.6
7.2
2653
210
38
45
3.4
5.9
(Fcr)+
S= Sand
CS= Copper Slag, *= Cured at 7 days
Table 3:
Properties of high strength concrete at 7- and 28-days of curing.
Mix
No.
Mix
proportions
Control
(100% S)
10%CS
90S
20%CS
80%S
40%CS
60%S
50%CS
50%S
60%CS
40%S
20%CS
80%S
100%CS
0%S
2
3
4
5
6
7
8
S= Sand
+
+
+
+
+
+
+
+= Cured at 28 days.
Density
(kg/m3)
Slump
(mm)
(Fcu)*
Strength (MPa)
(Fcu)+
(Ft)+
(Fcr)+
2568
28
76.9
93.9
5.4
14.6
2530
28
79.6
99.8
5.2
13.6
2588
50
74.5
95.3
6.2
12.4
2586
125
64.8
79.6
4.6
10.8
2625
115
77.8
96.8
6.1
12.9
2658
128
69.0
83.0
4.8
11.1
2673
102
63.8
79.0
4.7
10.3
2700
150
63.4
82.0
4.4
10.1
CS= Copper Slag, *= Cured at 7-days += Cured at 28-days.

Density (kg/m3)
2700
Normal concrete
High strength concrete
2650
2600
2550
2500
2450
2400
1
M ix No.
Figure 1:
Effect of copper slag addition on the density of normal and high

strength concrete
Slump (mm)
250
Normal concrete
High strength concrete
200
150
100
50
0
1
M ix No.
Figure 2:
Effect of copper slag addition on the workability of concrete.
The workability of concrete was assessed based on the measured slump of

fresh concrete as shown in Fig. 2 for different proportions of copper slag. It is
clear from Fig. 2 that the workability of concrete increases significantly with the
increase of copper slag in concrete mixes. For the control mix (i.e. mix 1), the
387
measured slump was 50mm and 28mm for normal and high strength concretes
respectively whereas for mixes 8 and 9, with 100% replacement of copper slag
for both NSC and HSC respectively, the measured slump was 210mm and
150mm for both types of concrete. This suggests that copper slag has much less
absorption characteristics than sand which explains the considerable increase in
workability as copper slag quantity increases. This increase in workability can
have beneficial effect on concrete in the sense that concrete mixes with low
water-to-cement ratios can be produced which can have good workability,
greater strength and improved durability than conventional concrete. However,
it should be noted that mixes with high contents of copper slag showed signs of
bleeding and segregation of which can have detrimental effects on concrete
performance.
Compressive Strength (MPa)

60
50
40
30
7-days
20
28-days
10
0
1
Mix No.
Figure 3:
Average cube compressive strength of normal concrete at 7 and 28days of curing.
The average unconfined 7- and 28-days compressive strengths are shown in

Figs. 3 and 4 for NSC and HSC, respectively. Results from both Figures show
that the compressive strength of concrete is generally increases as copper slag
quantity increases up to 50% addition of copper slag beyond which the
compressive strength is slightly decreased due to the significant increase in the
workability causing reduction in the strength. High water content in the mix
causes the particles of the constituents to separate leaving pores in the hardened
concrete which consequently reducing the strength. For normal concrete, Mix 6
(50% sand and 50% copper slag) yielded the highest compressive strength value
of 53 MPa whereas the control mix gave a compressive strength value of 44
MPa. Mix 9 (with 100% copper slag replacement) yielded a compressive
strength of 45 MPa which is comparable with the control mix (Table 2). For
high strength concrete, the highest compressive strength was achieved by Mix 2
with 10% replacement of copper slag. The highest compressive strength was

99.8 MPa compared with 93.9 MPa for the control mix. However, the mix with
100% replacement of copper slag (Mix 8) gave the lowest compressive strength
value of approximately 82 MPa which is almost 12.5% lower that the control
mix (Table 3). Tables 2 and 3 indicate that the tensile strength of concrete and
the modulus of rupture showed a similar behaviour to the compressive strength
results.
Compressive Strength (MPa)
100
90
80
70
60
50
40
30
20
10
0
7-days
28-days
Mix No.
Figure 4:
Average cube compressive strength of high strength concrete at 7

and 28-days of curing.
Conclusions
This paper presented the results of a research study on the effect of using copper
slag as a substitute for fine aggregate in normal and high strength concrete
mixes. Normal and high strength concrete mixes were prepared with different
proportions of copper slag. Concrete mixes were tested for workability, density,
compressive strength, tensile strength and flexural strength. Results showed that
the increase in the copper slag content causes an increase in both workability and
density of normal and high strength concretes. Results also showed that the
compressive strength of concrete is generally improved, compared with the
control mix, with the increase of copper slag up to a certain percentage beyond
which the strength was slightly reduced. Mixes with large copper slag
substitution suffered from bleeding and segregation due to the increase in
workability. Generally it is concluded that copper slag can be used as fine
aggregate to produce concrete with good strength.
References
[1]
Shi, C., and Qian, J.; High Performance Cementing Materials from
Industrial Slags A Review. Resources, Conservation and Recycling,
29:195-207, 2000.
[2]
[3]
[4]
[5]
389
Gorai, P., Jana, R. K., and Premchand. Characteristics and utilisation of

copper slag a review. Resources, Conservation and Recycling; 39:299313; 2003.
Akihiko, Y., and Takashi, Y. Study of Utilisation of Copper Slag As Fine
Aggregate for Concrete (in Japanese). Ashikaya Kogyo Daigaku Kenkyu
Shuroku; 23:79-85, 1996.
Toshiki A, Osamu K, Kenji S. Concrete with copper slag as fine
aggregate (in Japanese). J. Soc Mater Sci Jpn; 49:1097-102, 2000.
Demirboa, R., and Gl, R., Production of High Strength Concrete by Use
of Industrial by-products, Building and Environment, in press, 2006.
391
Application of FRC in tunnel reinforcement

P. Prochzka
Czech Technical University in Prague, Czech Republic
Abstract
Fibers in concrete should primarily restrain the material from cracks during the
curing process and dehydration. For standard evaluation of the influence of the
fibers, the three or four point test is studied, both experimentally and numerically
on a straight beam. In tunnel lining the situation is slightly different. While the
three or four point test leads to the assessment of the structure being only
bended, in tunnel reinforcement not only this but also bending with normal force
has to be observed. Moreover, the curing is affected by the non-uniform
distribution of moistening, as from the side of the surrounding rock the measure
of wet is principally higher than on the face where ventilation speeds up the
dehydration. These two assumptions are considered and treated experimentally
and numerically. Because a real tunnel lining is too large, a scale model is used
and, consequently, more samples can be studied. Similarity rules and an
appropriate calibration have to be obeyed. This paper is principally focused on
the theory of possible coupled modeling of behavior of the one-sided moistened
tunnel lining.
Introduction
Recently, the author published several papers on behavior of steel FRC with
Dramix type fibers in underground structures, especially in tunnel linings, [1].
One-sided moistened concrete beams have been studied from the mechanical
point of view. Scale models were prepared and studied.
The process, which we deal with in this paper, starts with coupled modeling
targeted to description of one-sided moistened beams. Then arches are modeled
and the mechanical properties are taken from the straight beam behavior. From
the numerical point of view a linear system of equations for unconstrained
problem has to be solved and when restrictions are imposed on stresses a
nonlinear problem is to be observed. The solution of such a problem can be
doi:10.2495/HPSM06038

formulated in terms of series including eigenparameters, which simulate measure
of different moistening at different layers in the structure. The application of the
proposed problem is suitable for structures from composite materials, for
laminated or layered structures, in tunnel construction, especially in NATM, etc.
In order to approximate the real structure of tunnel lining, in our study a
laminated hollow cylinder will be investigated from theoretical standpoint.
Uniform distribution of eigenparameters is considered in each layer.
Experiments with a beam
The behavior of one-sided moistened FRC was studied on straight beam, since
curvilinear beam could be modeled in a cumbersome way. The samples were
prepared in an enameled basin, the tap water in which was pored, which was
permanently added to hold the water level at a proper elevation. The wetted
medium surrounding the sample simulated real situation on site. Consequently,
the sample was prepared from two kinds of layers: the first layer was built up
from high porously liquid-absorbing mass. This mass was separated from the
concrete (the sample itself) by a filter paper to ensure convenient removal of the
concrete slender element from the equipment for preparing the samples. As a
consequence of this structure the lower side of the samples was permanently
highly moistened. This was also the aim of the experimental study
The samples were built up by auxiliary timber frame; its thickness along three
sides had the value of the right layer of the samples. The fourth side was free in
the beginning of the layers creation. Then the first layer of the fiber reinforced
concrete was constructed and the rectangular frame was completed by gluing the
last part (the shape of it corresponded to the height of the layer). This stage of
building the experimental equipment is depicted in Fig. 1, where also wire
holders for sticking next molding are seen to create the following layers of the
concrete. In this figure still no water is pored, and the three remaining moldings
inside the form are still not yet put inside the form to create the first concrete
layer.
In order to create the thicknesses of the layers as accurately as possible, the
moldings were put into the form along all sides of the equipment, so that the
room for the first and the subsequent layers was exactly defined. In other words,
at each stage of layers creation, one planar frame was built up with the same
thickness of moldings (as desired).
A plateau with the set up of fibers was drawn in Corel Draw on hardened
paper of 200 g/m2 In order to hold the cold temperature, which could cause
temporary sticking of fibers to the paper, the temperature of the paper with the
lay-up of fibers was dropped by putting the system in a freezer. Frozen at 32
degrees of Celsius every paper was three times sprinkled to assure the adhesion
of fibers to the paper. The number of fibers was calculated in such a way that the
fiber volume ratio was 1 percent. In each layer are 270 fibers in 11 interfaces of
concrete layers (6 and 5 fibers are repeated in rows). The thicknesses of the
layers are 6 + 4 +...+ 4 + 2 mm.
393
The concrete mixture was made of Knauf BP-8, the stones of which larger
than 0.8 mm were removed by filtering.
For 28 days in room temperature the slender beam was dried in such a way
that from the upper side a fan two times a day for 2 minutes was used from a
distance of 2 m. From below steady delivery of water was assured.
The final stage is seen in Fig. 2. The water was poured in the basin
immediately after the first layer of concrete was put in the frame. The sample
was not fully in a directly contact with the water, but along the interface of the
sample with the plateau the moistening of the lower layer was ensured. The
dimensions of the whole sample element were: 150 x 250 x 48 mm3. This sample
was cut into beams of a cross section 2.8 x 4.8 mm2, which were loaded by a
concentrated load in the middle of span of the beams. The ratio of thickness to span was about 1/3. From that point of view, the mechanical behavior was closer
to a plate.
Figure 1:
Experimental equipment with the basin and frame at the

commencing stage.
Then a comparative study provided a time-dependent distribution of strains in

the beam during the bending test. The one-sided moistened beam during the
curing process and standard fiber reinforced beam are compared in both
experimental studies, static and time dependent. Because a non-reinforced beam
has very poor bearing capacity, it is not considered in these comparative studies.
A standard cement mixture has modulus of elasticity E = 20 GPa and
Poissons ratio v = 0.16. Steel fibers possess modulus of elasticity E = 197 GPa
and Poissons ratio v = 0.3. Regularly distributed Dramix-type fibers are used. A
typical bending test with one concentrated load in the middle of 128 mm span of
beams was applied. The beams were loaded on the moistened side to worsen the
situation. It was expected that the pore pressure would expand the layers on the

moistened side, while in the dry side the pore pressure slowly disappears. This
caused a sort of prestress at the moistened side.
Figure 2:
Final stage of the sample element after construction.
The aim of this study was to discover how the situation would worsen during
the bending test comparing these samples with that without loading by the
moistening.
In Fig. 3, a relation applied force - deflection is depicted. Two cases of the
beams are compared: moistened and external moistening free plates are
considered. Both types of beams were cured 28 days and then the experiments
were accomplished.
From Fig.3 one can see a sudden drop at the value of loading about 1.8 kN in
the case of moistened beam and about 2.6 kN in the case of the wet-free beam.
These drops differ also in deflections, which were in the first case 0.18 mm and
0.22 mm in the second case. Sudden opening of a crack below the concentrated
load causes the drops of loading. After increase of loading peaks are reached.
While the externally moistened beam attains 3.8 kN, the moistening free beam
4.5 kN. After this, the beams still bear, but their bearing capacity lowers and
softening of the aggregate occurs. This is caused by a pulling of fibers out of the
concrete matrix. When the destruction of the beams is attained, it can be seen
that no steel fiber were disconnected, fibers were only pulled out of the concrete.
The different behaviors of both beams are caused by pore pressure, which is
stored in cured beams. In mathematical models this property is very easy to
describe. Pore pressure is a special case of eigenstrain tensor, , or eigenstress
tensor, , which both are expressed in the generalized Hookes law as:
395
= L( ) or = L + ,
where is the stress tensor and is the strain tensor. L is the elastic material
stiffness matrix.
Another type of experiment was carried out on these beams. A time
dependent behavior of both types of beams is observed by bending with the same
disposition as in the previous experiments, see Fig. 4.
Figure 3:
Figure 4:
Deflection - concentrated load relation.
Time dependent behavior of the beams (creep).

From the picture it is seen that the behavior of both types of beams is not
different that much. Only in the neighborhood of the peak values the differences
are larger than for the lesser values of the concentrated load.
From both examples we have a conclusion that both bearing capacity and
time dependent deflections are larger in moistened beams. Nevertheless, the
differences are not that much decisive that the steel fiber reinforced concrete has
to be removed from the production for the type of structures, such as
underground works (tunnels, retaining walls), foundations, floor decking, etc.
Numerical description of a cylindrical segment
In this section we concentrate our attention on a special type of one-sided

moistened FRC structure, which can be used in design of underground works,
foundation engineering, floor decking of industry halls, sport halls, etc. The
mechanical behavior of the moistened arches was compared with the fiberreinforced straight beams. The reinforcement of both types of beams was put into
structures regularly (uniformly) in a very careful way, as described in the above
text.
Figure 5:
Geometry and disposition of the beam under study.
It was shown that the bearing capacity of the moistened structure is less than
that of beams created in the standard way. Also the creep seems to be more
important and larger in the case of moistened structure.
Relations on a single layer
The clamped arch can be solved by Fourier method: the unknowns and the
loading are split into series in one direction and in the remaining direction the
ordinary differential equation is solved. The problem of clamped arch is divided
into solution of simply supported arch and the rotation at clamped edges is
annihilated by applying external bending moment. The solution of the latter case
can be found in [2]. The first problem is a generalization of the problem from
[3]. Results from similar cases can be seen in [4].
397
Introducing cylindrical coordinate system 0rz, components of the

displacement vector are considered: u = u r , v = u , w = u z , where u r is the
displacement in the radial direction and u is the displacement in the
circumferential (hoop) direction and w = u z is the displacement in the axial
direction. The cylindrically orthotropic structure is a particular example of a
cylindrically anisotropic solid with the constitutive equations written as (in this
section the indices describing number of the layer are omitted):
r
z
r
rz
Lrr

L r
L
= rz
0

0
0

L r
L
Lz
0
0
0
Lrz
Lz
L zz
0
0
0
0
0
0
G r
0
0
0
0
0
0
Gz
0
r r
r r
zr zr
2( r r )
2( z z )
G rz 2( rz rz )
0
0
0
0
0
(1)
where the stress tensor was related with the strain tensor and the eigenstrain
tensor through the material stiffness matrix L, and Lij , G r , Gz , G rz are the
stiffness coefficients. Nine stiffness coefficients and six eigenstrains describe
this kind of anisotropy. Cylindrical orthotropy is characterized by the fact that
properties in the tangential, radial and axial directions are distinct; in other
words, the material is orthotropic in a Cartesian system which is located at any
point within the structure, with the three axes pointing in the axial, tangential and
radial directions respectively. If Lrr > L then the material is called radially
orthotropic, and if L > L rr then it is called circumferentially orthotropic.
In the special case of transversally isotropic solid which may represent the
average structure of fiber reinforced material, nine independent material
constants in (1) are related as:
Lrr = L ,
Lrz = Lz ,
G rz = Gz ,
Lrr - Lr = 2G r
In what follows we concentrate our attention on 2D case, assuming a long

enough cylinder. Then the number of displacement components is reduced to
u = u r , v = u , w = u z , and three stresses remain non-zero: r , , r . The
cylindrical coordinate system reduces to polar one.
Recall basic geometrical relations:
r =
u
,
r
1 v u
+ ,
r r
r =
1 u v v
+
r r r
(2)
If we adopt the assumption z = 0 and isotropic material is considered, Hookes

law takes the form:

r r =
1
( r ),
E
1
( r ),
E
r r =
1
r (3)
G
where is Poissons ratio. Substituting the last equations in the geometrical

equations yields the inverse relations of Hookes law for expressing stresses in
the terms of displacements:
1 v
u
u
r + Lr
+
r = Lrr ( r r ) + Lr ( ) = Lrr
r
r
1 v u
u
r
+ + L
= Lr
r
r
1 u
v v
r = G r r = G
+
r
r
r
r
(4)
(5)
Two equations of equilibrium provide relations among three components of

stress tensor:
r 1 r
+ r = 0,
+
r
r
1
+ 2 r = 0
+
r
r
(6)
Expanding two components of displacements into sine and cosine series and
denoting the k-th term by U r( k ) and U ( k ) yields
u r( k ) (r , ) = U r( k ) (r ) cos 2k ,
u( k ) (r , ) = U ( k ) (r ) sin 2k ,
(7)
u z( k ) = 0
where U r( k ) U ( k ) and U ( k ) V ( k ) are unknown functions of r which need to
be determined from the equations of equilibrium. In particular, the substitution of
(1) in the stress-strain relations gives stresses which, when substituted into the
equations of equilibrium in cylindrical coordinates, provide the following
equations for evaluation of U ( k ) and V ( k ) .
Similarly the eigenparameters will be expanded into series with the k-th
components:
r( k ) (r , ) = R ( k ) (r ) cos 2k , ( k ) (r , ) = F ( k ) (r ) cos 2k ,
r(k ) (r , ) = M ( k ) (r ) sin 2k
(8)
399
Components of strains, (2), and stresses, (6), are then expressed as

(identification (k) of the number in the series is dropped out and the prime
denotes derivative by r):
r( k ) = U ' cos 2k , ( k ) =
1
(2kV + U ) cos 2k , r(k ) = 2k U + V ' 1 V sin 2k
r
r
r
(9)
r( k ) = Lrr (U ' R ) + Lr (2kV + U F ) cos 2 ,

r
(k )
= Lr (2kV + U F ) + L (U ' R) cos 2 ,

r
2k
r(k ) = G U + V ' V M sin2

r
r
(10)
(11)
Substituting (10), (11) to (6) and using the last four relations, two equations for
unknown amplitudes of displacements read as:
(2kV '+U ' F ') +

r
2k (1 ) 2k
V
1
+
U + V ' M [2kV + U F ] = 0
r
r
r
U ' ' R' '+(1 )(U ' R' ) +
M
F
3
8k 2 + 1
4k +
4k ( 3)
V ' ' V '+
V
U '+
U M '+
+ R '+ 2 = 0
2
r
r
r
R
1
r
(1 ) r
The last two equations lead to the solution, which defines the displacements
on each lamina. Four integration constants are determined from interfacial
conditions between adjacent layers. The angle determined from the above
approach is eliminated by introducing moments at the end points to get
conditions for clamped segment. The influence of such a moment to the segment
is taken from [1], where the approach is described in details.
Conclusions
In this paper the procedure for calculating a segment of one-sided moistened

tunnel lining from FRC. Information from straight one-sided moistened beams
has been transferred to the segment and then the numerical procedure is
suggested. Since it allows us to apply semianalytical (Fourier) solution, the
procedure is relatively easy. First simply supported segment is considered, then
using the approach from [2] for involvement of given moment at the end points,
the clamped edge can be simulated.
Acknowledgment
This paper was prepared under financial support of GA AV R, project No. IAA
2119402.
References
[1]
[2]
[3]
[4]
Prochzka, P., Starikov, N.: Optimal Slope of Dramix Type Fibers in

Reinforced Concrete, 6th World Congresses of Structural and
Multidisciplinary Optimization, Rio de Janeiro, 30 May - 03 June 2005,
Brazil.
Lekhnitski, S.G. Anisotropic plates. Gordon and Breach Sci. Publ. NYC London Paris, 1968.
Chen, T., Dvorak, G.J. and Benveniste, Y. Stress fields in composites
reinforced by coated cylindrically orthotropic fibers. Mechanics of
Materials 9 (1990) 17-32.
Sokolnikoff, I.S. Mathematical theory of elasticity. 2nd edition, McGrawHill, New York, 1996.
401
Innovative procedure to produce high

performance pretensioned concrete girders
combining high strength concrete and normal
or special concrete types
C. Vzquez-Herrero & F. Martnez-Abella
Department of Construction Engineering, School of Civil Engineering,
University of La Corua, Spain
Abstract
As is well known, prefabrication procedures allow significant reductions in
construction time. In many countries, precast concrete bridges are customary and
made up of pretensioned prestressed girders. Both the compressive and tensile
strength of the concrete mix must be enough at transfer (which takes place
typically two or three days after concrete casting) in order to avoid excessive
prestress losses, and undesirable cracking. Many attempts have been made to
increase the spectrum of the usual high performance concretes to other concrete
types (i.e. normal strength concrete, lightweight concrete and recycled aggregate
concrete). Why are these special concrete types not used in the current
production of large pretensioned concrete girders? Technical problems have been
observed at the critical zones: development length, geometrical discontinuities,
supports, etc. A new technique has been developed by the authors to produce
hybrid pretensioned concrete elements, combining the usual high strength
concretes in the critical zones with other concrete types in the rest of the
elements.
Keywords:
prestressed concrete, precast concrete, pre-tensioning, pretensioned, pretensioned, high performance, high strength, prestressed, girders,
beams.
Introduction
As is well known, the prefabrication procedures allow significant reductions in

the construction time. In many countries precast concrete bridges are customary,
made up of pretensioned prestressed girders deck.
doi:10.2495/HPSM06039

In post-tensioned structures, the transfer of prestress is done through
mechanical anchorage. However, in pretensioned structures strand force is
transmitted to concrete by bond, which is essential to:
Transfer the prestressing force to the beam during the service life.
Transfer length is defined (after Weerasekera [1]) as the distance over
which the strand should be bonded to concrete in order to develop the
effective prestress stress (fpe) in the strand (see Figure 1) .
Guarantee the anchorage of the prestressing force. The flexural bond
length is defined as the distance beyond the transfer length required to
achieve the prestress design stress (fps) during the service life (see
Figure 1). The development length is the sum of the transfer length and
the flexural bond length.
Accordingly two different zones are determined along these elements: the
external zones, where the bearing capacity is lower than the design capacity
(development length) and the inner zone, where prestress is fully effective.
Both the compressive and tensile strength of the concrete mix must be high
enough at transfer (which takes place typically two or three days after concrete
casting) in order to avoid excessive prestress losses, and undesirable cracking.
The stresses created by bond in concrete are so high that the use of high strength
concrete is usually required to prevent cracking along the transfer length.
Figure 1:
Transfer length and development length.
Many attempts were made to increase the spectrum of the conventional high
performance concretes to other concrete types (i.e. moderate strength concrete,
lightweight concrete and recycled aggregate concrete).
The use of lightweight concrete having similar strength compared to the
current high strength concrete, would allow either the reduction of the self
weight of these elements while maintaining their span length, or the increase of
the span length while maintaining the self weight.
On the other hand, a reduction of the requirements of the resistant properties
of concrete would imply a significant cost reduction. The use of normal strength
403
concrete is desirable, instead of the current high strength concrete. However this
possibility must be carefully studied, in order not to affect the performance of the
structure during the service life.
Environmental regulations are hardening in developed countries. The
extraction of aggregates is being penalized. In the future, a considerable increase
of the cost of natural aggregates will be experienced. Consequently, the
applicability of recycled aggregate concrete to the prefabrication processes must
be studied.
Why are these special concrete types not used in the current production of
large pretensioned concrete girders? In our opinion it is due to their high cost and
eventually to the technical problems at the critical zones: development length,
geometrical discontinuities, supports, etc.
Nowadays, the bearing capacity of the pretensioned concrete elements is
limited by the feasibility of introducing the maximum prestress in the cross
section due to the restrictions of minimum center-to-center strand distance as
well as on minimum cover.
Several attempts to use high strength lightweight concrete instead of the
current normal weight concrete, with the same design criteria have failed due to
longitudinal cracking along the ends of the girders (Vzquez-Herrero and
Martnez-Abella [2]). These cracks are generated by the high stresses in the
boundary between strands and surrounding concrete:
Normal stresses between strand and the surrounding concrete, caused by the
recovery of the strand diameter after transfer. As a result transverse tensile
stresses appear in the surrounding concrete (bursting stresses) along the
whole length of the pretensioned concrete elements.
Bond stresses along the transfer length. The resulting transverse stresses
which appear in the surrounding concrete are named splitting stresses.
Longitudinal cracking along the pretensioned concrete elements is

unacceptable, because it nullifies the protection against corrosion of the strand
provided by concrete cover. So it is necessary to prevent this kind of damage.
The dispositions that have been usually considered in order to reduce the risk
of longitudinal cracking are the following:
Methods to enhance the confinement of the strands. With this aim transverse
reinforcement is disposed, either through stirrups or spiral reinforcement.
This procedure does not prevent cracking, but it certainly reduces crack
width, and prevents losing the confining effect. These methods imply a
considerable increase of operational costs.
Methods to reduce bond between active reinforcement and concrete by
different means, i.e.: reduction of the strand surface roughness, use of
lubricating substances, use of partially debonded tendons, etc. An adverse
effect is collateral to these measures: the increase of the development length
means a reduction of the bearing capacity of the girders at this zone.
Besides, it is difficult to exert and control the rate of bond reduction desired.

In this paper an alternative procedure to produce pretensiond concrete
elements, which allows the use of a normal strength concrete is presented. This is
explained in the following section.
A procedure to produce hybrid pretensioned concrete

elements
The scope of the procedure is to produce pretensioned concrete elements with

what will be named objective concrete, which can be either a normal strength
concrete or a special type of concrete (i.e. lightweight concrete, recycled
aggregate concrete, fiber reinforced concrete, etc.).
The critical zones of the structure are determined by :
The bending moments along the structure determine the situation of the
zones where the prestressing force is not constant, which means the
existence of bond stresses.
The stress concentration zones, such as supports, holes, zones where the
geometry changes abruptly,...
The objective concrete will be used in most of the concrete girders, but in the
critical zones, which will be executed using a high strength concrete mix that has
proved a good performance. The adequate performance of the objective concrete
in the non critical zones must be checked experimentally, in order to determine
the adequacy of the material for this use.
The production procedure is similar to current girders, but for the different
phases, which will be determined by the different critical zones. The boundary
surface between the different concrete types can be accomplished by the use of
any kind of formwork, or by the natural slope adopted by concrete.
Figure 2: Hybrid pretensioned concrete girder.

Figure 2 presents an example of a simply supported beam, which will take
part of an isostatic bridge with precast girders deck. The depth of the critical
zone (see flag 1 in Figure 2) is determined by the upper strand level, adding the
convenient cover. The critical zones are delimited by the development length of
the lower tendons (see flag 2 in Figure 2). The adopted angle (see flag 3 in
Figure 2) of the boundary must be fixed to completely define the critical zone.
405
The reinforcement required to transmit the horizontal component of the shear

stresses (see flag 4 in Figure 2) must be conveniently anchored at both sides of
the interphase. The rest of the element is cast with the objective concrete.
A procedure to assess the performance of hybrid

pretensioned concrete elements
In order to study the feasibility of producing hybrid pretensioned concrete

beams, the folowing procedure is proposed:
1.
2.
3.
Initially, several hybrid pretensioned concrete prisms must be produced

and tested (see Figure 3). The critical zones have to be performed using a
high strength concrete mix, and the rest using the objective concrete.
These elements must be monitored for at least several months after
transfer, to detect eventual cracking or unexpected growth of the transfer
length.
If the behavior of the prisms is considered satisfactory, the second stage
consists of producing at plant several hybrid pretensioned girders, these
elements have to be monitored, and some of them tested to failure (see
Figure 4) to estimate their bearing capacity and development length.
As a final stage of the design process, a prototype structure will be
designed, built and monitored for several years. Only if this final stage is
successful, current production of the hybrid pretensioned concrete
elements can be attempted.
Figure 3:
Pretensioning bench and hybrid pretensioned concrete prism.
Figure 4:
Flexural test on a pretensioned concrete girder.
The behavior of hybrid pretensioned elements of

lightweight concrete-high strength concrete
An experimental research was planned to study the feasibility of producing

pretensioned concrete elements with lightweight concrete (LC, see Table 1). The
proposed LC mix contains expanded clay as lightweight aggregate. Previously an
experimental research had been undertaken to study the feasibility of lightweight
concrete pretensioned elements, but longitudinal cracking appeared at the
transfer length regions of these elements some time after transfer (VzquezHerrero and Martnez-Abella [2]).
The pretensioned concrete prisms (square cross section 105 mm wide, 3750
mm long) were produced at the School of Engineering Laboratory, in a 4 m
pretensioning bench designed for this research (see Figures 3 and 5). This part of
the experimental research has been the subject of the undergraduate research
project of Flix Saudo-Herrera [3]. The critical zones (1500 mm long from the
end of the elements) were cast with a high strength concrete (RC, see Table 1),
and the inner part (750 mm long) was cast afterwards with the lightweight
concrete mix.
Table 1:
Prop.
UNITS
Concrete
specific
weight
KN/m3
Properties of the concrete mixes.
Average
Average
Average
Average
Average
Average
concrete
direct
concrete
direct
split tensile modulus of
cylinder
tensile
cylinder
tensile
strength, at elasticity, at
strength, at
strength, at strength, at strength, at
2 days
2 days
2 days
2 days
28 days
28 days
MPa
MPa
MPa
MPa
MPa
MPa
RC
22
50
3.2
30441
3.3
67
>3.3
LC
17
48
2.7
19939
2.7
56
2.9
407
Strand type was Y1860 S7 (1860 MPa), with a diameter of 15.2 mm (0.6).
Strand was tensioned until a stress of 75% of the minimum ultimate stress was
attained. From the stressing process to the prestress force release 48 hours
elapsed. Transfer was performed through gradually by flame heating.
The prisms were continuously monitored during the transfer process and
afterwards. Transfer length was measured periodically after transfer, and its
values kept steady within the high strength concrete critical zones. The prisms
did not show any visible cracking for several months, but after 3-5 months
longitudinal cracks appeared along the lightweight concrete inner part of all the
hybrid specimens. Cracks progressed throughout the whole lightweight concrete
inner part.
Consequently, the tested lightweight concrete mix used is not able to
withstand the bursting stresses resulting from the strand diameter recovery which
takes place after transfer due to the Poisson effect. Although the performance of
the hybrid prisms is adequate for the first several months after transfer, durability
cannot be assessed throughout the whole service life of the structure.
Figure 5:
View of the pretensioning bench and the prism during testing.
Conclusions
After the experimental program carried out on hybrid pretensioned concrete

elements, the following conclusions have been established:
A procedure has been developed to design and produce hybrid pretensioned

concrete elements (girders, beams, hollow core plates, etc.) using a given
objective concrete in most parts of the element. The critical zones will be
performed using a concrete mix that has shown a good performance (i.e.
high strength concrete). Special attention must be paid to the design of the
boundary surface between both concrete mixes.
An objective high strength lightweight concrete was tested, by means of

producing hybrid pretensioned concrete prisms. This lightweight concrete
mix is not adequate for the production of hybrid pretensioned concrete
girders, using 15.2 mm diameter strand, with a cover of 50 mm, measured
from the strand center.
The prism test has shown a good procedure to detect potential problems of
the objective concrete, for a given strand size and cover. In our opinion this
first stage is very convenient to study the feasibility of the proposed
objective concrete mixes.
In our opinion further research should be done using a stronger lightweight
concrete aggregate, strong enough to restrain cracks propagation. Eventually
the use of fiber reinforced concrete, either lightweight or normal weight
concrete, is another way to increase concrete tensile strength.
Further research will be done at the University of La Corua dealing with

different objective concrete types (i.e. normal weight concrete of moderate
strength, fiber reinforced concrete and recycled aggregate concrete) in order to
widen the scope of pretensioned concrete elements.
Acknowledgements
This research project has been sponsored by the cement and concrete firm
Corporacin Noroeste, and by the projects XUGA22801A97 and
CICYTMAT20001-0765 (Ministry of Science and Technology, Spain). The
authors appreciate the contribution of the engineers Mr. Humberto Vzquez, Mr.
Arturo Martnez, Mr. Marcos Fernndez and Mr. Juan Rabual, and of Professor
Dr. Pablo Rodrguez-Vellando for the revision of this text.
References
[1] Weerasekera, I.R., Transfer and Flexural Bond in Pretensioned Prestressed
Concrete, PhD Thesis, UMI: Michigan, 1991.
[2] Vzquez-Herrero, C., Martnez-Abella, F., Bond Properties of High Strength
Pretensioned Lightweight and Normal Weight Concrete, Proceedings of
Bond in Concrete, Budapest, pp. 791-798, 2002.
[3] Saudo-Herrera. F., Experimental study of the behavior of hybrid
pretensioned concrete elements using high strength concrete and lightweight
concrete, Undergraduate Research Project (in Spanish), University of La
Corua, 2003.
409
Properties of high performance concrete:

the effect of cracks
E. Mahonkov, M. Jiikov & R. ern
Department of Structural Mechanics, Faculty of Civil Engineering,
Czech Technical University in Prague, Czech Republic
Abstract
The moisture transport and storage parameters of high performance concrete
belong to the most critical parameters in designing and using complex reliability
based models for service life prediction of concrete structures. However, mostly
it is supposed that the material is compact without any significant cracks, which
is not always true. In this paper, the effect of thermally induced cracks on the
basic moisture transport and storage properties is analyzed. Experimental results
show that the presence of cracks in hardened high performance concrete
mixtures dramatically changes all measured hygric parameters. Also, the effects
of both microsilica and the size and presence of aggregates on moisture transport
and storage parameters are very important. The liquid water transport is found to
be affected by cracks, microsilica and aggregates in a much more significant way
than water vapour transport.
Keywords: high performance concrete, moisture transport and storage
properties, cracks.
Introduction
The progressive building materials development improves their quality and

utility value. In order to achieve better material properties, besides the new
technologies also specific admixtures are employed. The accurate admixtures
dosage and technological conditions observation can improve the material
parameters like strength and durability. Most frequently, the mineral admixtures,
for example pozzolana, fly ash or flue cinder, are used in concrete production.
Fine siliceous materials usually called silica fume belong to relatively newly
applied materials improving concrete quality.
doi:10.2495/HPSM06040

Silica fume can influence the hydration process on the one hand physically,
ultra fine elements fill space between hydrating cement grains, on the other one
chemically; they can react together with calcium hydroxide to form qualitatively
different binding products. It leads to lower porosity, it changes the micro-pores
distribution, the cementing agent is compacting and the strength, insolubility and
corrosion durability of final product are improved.
Despite the gradual shift in the HPC design and application philosophy
towards a generally recognized necessity to measure a wider scale of HPC
parameters the mechanical properties still remain the far most frequent
parameters investigated in the research work being done on HPC. Water and
water vapor transport and storage properties of HPC were not yet in the center of
interest of most researchers until now although they possess a very high
predicative potential concerning the HPC quality.
In this paper, we analyse the influence of cracks on moisture transport and
storage parameters of two different types of concrete, namely the high
performance concrete C90/105 containing microsilica and the same material
C60/75 without microsilica.
Materials
Basic tested material was high performance concrete C90/105 containing silica
fume suspension, microsilica, denoted as BI in subsequent text. For the sake of
comparison, the same concrete mixture C60/75 (denoted as BII) was prepared
without microsilica. In order to analyse the effect of aggregates on moisture
transport and storage parameters, two other mixtures based on BI and BII were
prepared. The first one (denoted as BBI and BBII) was cement mortar without
the 8-16 mm aggregate fraction. The second one was cement paste denoted as PI
and PII. Table 1 presents the composition of the studied cement mixtures.
Table 1:
Composition of studied cement mixtures.
Composition in g
**
SiO2
Aggregates
Woer- ***
CEM I
Lentan
suspen- 0-4
ment
4-8
8-16
52.5 R
VZ 33
sion
mm mm
mm FM 794
BI
480
72
664 207
995
7.74
2.58
BII
470
0
668 209 1001
5.17
2.35
BBI
480
72
664 1202
0
7.74
2.58
BBII
470
0
668 1210
0
5.17
2.35
PI
2346
352
0
0
0
38
12.67
PII
2348
0
0
0
0
26
12
*
SiO2 suspension water suspension consisting of 88-95% of SiO2 and
amounts of calcium oxide, magnesium oxide and nitrogen oxide.
**
FM 794 plasticizer on the basis of polycarboxylateether.
***
Lentan VZ 33 hydration retarder on the saccharose basis.
Type of
concrete
w/c
0.36
0.33
0.38
0.35
0.31
0.34
small
411
Each mixture was cast into the forms, after one day the samples were
removed from the forms and then varnished using curing solution. After the
hardening period of 28 days, the samples were cut to required dimensions. These
samples were then heated up to 600C in an oven to reach cracks, see Figure 1.
Figure 1:
The HPC BI sample after thermal loading.
3 Experimental methods
3.1 Basic material parameters
As fundamental physical material characteristics, bulk density b [kgm-3],
vacuum saturation moisture content wsat [kgm-3], porosity [Vol.-%] and matrix
density m [kg m-3] were determined. They were obtained using the gravimetric
method and the vacuum saturation method. The vacuum saturation moisture
content was calculated according to the equation
wsat = w
msat m0
= w ,
msat ma
(1)
where w is the water density [kg m-3], m0, msat and ma are the mass of dry
sample, water- saturated sample and mass of the immersed water - saturated
sample [kg], respectively, and is the open porosity, which is defined as the
ratio of the volume of open pores in material to its total volume. Matrix density
was calculated as
mat =
m0 ,
V (1 )
(2)
where V is sample volume [m3].

The measurement of basic parameters took place in a conditioned laboratory
at the temperature of 221 C and 25-30% relative humidity. Each result
represents the average value from three to five measured values.

3.2 Water vapour and water transport parameters
Two versions of the common cup method were employed in the measurements
of the water vapor diffusion coefficient [1]. In the first one the sealed cup
containing silica gel (5% relative humidity) was placed in a controlled climatic
chamber with 97% relative humidity and weighed periodically. In the second one
the cup containing the saturated solution of K2SO4 (97% relative humidity) was
placed in the 25% relative humidity environment. The measurements were done
at 20C in a period of two weeks. The steady state values of mass gain or loss
were determined by linear regression for the last five readings.
The water vapor diffusion coefficient D [m2s-1] was calculated from the
measured data according to the equation
D=
m d R T
S M p p
(3)
where m the amount of water vapor diffused through the sample [kg], d the
sample thickness [m], S the specimen surface [m2], the period of time
corresponding to the transport of mass of water vapor m [s], pp the difference
between partial water vapor pressure in the air under and above specific
specimen surface [Pa], R the universal gas constant, M the molar mass of water,
T the absolute temperature [K].
On the basis of the diffusion coefficient D, the water vapor diffusion
resistance factor was determined:
Da
D
(4)
where Da is the diffusion coefficient of water vapor in the air.

The water sorptivity was measured using a standard experimental setup. The
specimen was water and vapor-proof insulated on four edges and the face side
was immersed 1-2 mm in the water, constant water level in tank was achieved by
a Mariott bottle with two capillary tubes. One of them, inside diameter 2 mm,
was ducked under the water level, second one, inside diameter 5 mm, was above
water level. The automatic balance allowed recording the increase of mass. The
water absorption coefficient A [kgm-2s-1/2] was then calculated using the formula
i = A t ,
(5)
where i is the cumulative water absorption [kg/m2], t is the time from the
beginning of the suction experiment. The water absorption coefficient was then
employed for the calculation of the apparent moisture diffusivity in the form [2]
413
A ,
app
wc w0
(6)
where wc is the saturated moisture content [kgm-3] and w0 the initial moisture
content [kgm-3].
In the experimental work, the following samples were used: water vapor
diffusion coefficient 10 cylinders with the diameter 105 mm and thickness 20
mm, water sorptivity - 5 specimens 50 x 50 x 20 mm.
3.3 Sorption isotherms
The water adsorption and desorption in a porous material are based on van der
Waals forces between the surface of the porous matrix and water molecules. The
dry material mass increases after a contact with moist air because of gradual
bonding of water molecules from the air to the pore walls, in the case of
adsorption. Desorption is reversed physical phenomenon, the initial state is
capillary saturated sample. At the moment of achieving the equilibrium state
between the water vapour pressure in the moist material and in the surrounding
air this process is stopped.
The mass of samples was measured in specified periods of time until steady
state value of mass was achieved. Then, the moisture content by mass was
calculated according to the equation
um =
mw m0
m0
(7)
where mw is the mass of wet sample and m0 is the mass of dry sample [kg].
The samples were placed into the desiccators with different salt solutions to
simulate different values of relative humidity, see Table 2.
Table 2:
Relative humidity over saturated salt solutions.
Temperature/Relative humidity [-]

20C
23C
25C
LiCl
0.111
0.111
MgCl2.6H2O
0.330
0.329
0.328
NaNO2
0.654
0.643
NaCl
0.755
0.754
0.753
NH4Cl
0.792
0.788
0.786
KCl
0.851
0.845
0.843
K2SO4
0.979
0.976
Salt
In the measurements of adsorption and desorption isotherms, for each relative

humidity 3-10 test specimens of each material with overall dimensions of 40 x
40 x 10 mm were prepared. The measurements were done at 231C. The initial

states were dry samples in the case of adsorption, and capillary saturated samples
in the case of desorption.
4 Experimental results and discussion

The results of basic parameter measurements are presented in Table 3.
Table 3:
Basic material parameters.
Without cracks
Type of
mixture
mat
wsat
With cracks
mat
[kgm-3]
[%]
[kgm-3]
wsat
[%]
BI
2423
2760
122.0 12.20 2391
2856
162.2 16.30
BII
2388
2647
95.1
9.80 2350
2713
133.2 13.40
BBI
2240
2652
150.6 15.60 2147
2627
182.0 18.30
BBII
2210
2578
140.5 14.30 2205
2669
174.8 17.40
PI
1988
2809
248.4 29.20 1942
3068
375.8 36.70
PII
1991
2779
285.2 28.30 2019
3088
355.7 34.60
The highest bulk density exhibited the hardened HPC mixture BI. The second
highest had the BII mixture without microsilica. Nearly the same values had
mixtures denoted BBI and BBII, cement pastes had the lowest values. The bulk
density of cement mixtures with cracks was lower then of the same without
cracks. The matrix densities differed only in the range of about 5% and
increased about 10% in the case of mixtures with cracks compared to mixtures
without cracks. The highest vacuum saturated moisture content had the cement
pastes PI and PII, where the difference was within the error range, and increased
about 30 % in the case of mixtures with cracks compared to mixtures without
cracks. The most porous materials were cement pastes PI and PII, the presence of
microsilica and admixtures did not have a significant effect. The porosity
increased from 3 to 6% in the case of mixtures with cracks compared to mixtures
without cracks.
The results of measurements of water and water vapour transport parameters
of the studied hardened cement mixtures are presented in Tables 4 and 5.
The water vapour diffusion coefficient was affected by the addition of
microsilica only for cement paste PI in a remarkable way where it decreased
three to four times. However, for the mixture BII its effect was within the error
range of the measuring method. The cracks caused dramatic increase of the water
vapour diffusion coefficient in all cases.
The effect of microsilica on the liquid moisture transport parameters was
more pronounced than on water vapour transport. The water sorptivity for the
415
basic mixture BI decreased five times due to the microsilica addition, and for the
cement paste PI by about 30%. The moisture diffusivity of mixtures with
microsilica decreased in a corresponding way. In the case of cement mixtures
with cracks the moisture diffusivity increased one order of magnitude compared
to mixtures without cracks.
Table 4:
Water vapour transport properties of cement mixtures.

Without cracks
Type
of
mixture
BI
BII
BBI
BBII
PI
PII
-6
D [10 m s ]
9725%
RH
1.06
1.10
0.287
0.288
0.136
0.357
Table 5:
Type of
mixture
BI
BII
BBI
BBII
PI
PII
2 -1
597%
RH
0.397
0.375
0.132
0.151
0.079
0.291
[-]
9725%
RH
22.5
20.9
80.3
81.4
170.7
65.4
597%
RH
59.3
61.4
175.4
152.1
295.9
81.7
With cracks
D [10-6 m2s [-]
1
]
97597525% 97% 25% 97%
RH
RH
RH
RH
1.48 1.55 15.8 15.4
2.22 1.76 10.4 13.2
1.73 1.96 13.8 12.3
2.11 2.23 11.0 10.6
2.29 3.20 10.2
7.2
2.22 1.76 10.4 13.2
Water transport properties of cement mixtures.
Without cracks
A [kg m-2s-1/2] app [m2s-1]
3.63E-03
8.9E-10
1.80E-02
3.3E-09
1.76E-02
1.5E-08
4.04E-02
8.3E-08
2.36E-02
7.0E-09
3.52E-02
1.8E-08
With cracks
A [kg m-2s-1/2] app [m2s-1]
4.84E-02
9.2E-08
4.23E-02
1.1E-07
8.08E-02
2.5E-07
6.21E-02
2.0E-07
2.77E-01
5.5E-07
1.59E-01
2.0E-07
The effect of aggregate presence was found to be quite different for liquid
water transport than for water vapour transport. The water sorptivity was for the
basic hardened HPC mixture BI several times lower than for the cement pastes.
On the other hand, the water vapour diffusion coefficient of cement pastes was
always lower than for the basic hardened HPC mixtures. The 5-97% RH values
of the water vapour diffusion coefficient were always lower than the 97-25% RH
values, which is in accordance with the previous measurements on many other
materials including concretes and cement pastes.
Figure 2a-f shows the measured adsorption (lower curves) and desorption
(upper curves) isotherms. We can see that cement pastes had higher water
adsorption capacity than the basic hardened HPC mixtures. This is an expected
behaviour due to absence of the inert aggregates in cement pastes. The presence
of microsilica slightly increased the water adsorption capacity of basic hardened

HPC mixture but for cement pastes an opposite behaviour was observed.
Therefore, any clear effect of microsilica on water adsorption capacity was not
evidenced. The sorption curves are almost the same for the mixtures without and
with cracks.
0.04
0.06
BII without cracks

BII with cracks
moisture content [kg kg-1]
BI without cracks
BI with cracks
0.04
0.02
0.02
0
0
0.2
0.4
0.6
0.8
0.2
relative humidity [-]
0.4
(a)
0.8
BBII without cracks

(b)
BBI without cracks

BBI with cracks
0.04
0.02
BBII with cracks

0.04
0.02
0
0
0.2
0.4
0.6
0.8
0.2
0.4
0.6
(c)
(d)
0.14
0.12
0.12
0.1
0.8
0.06
0.06
0.08
0.06
0.04
PI without cracks
0.02
PI with cracks
0.1
0.08
0.06
0.04
PII without cracks
0.02
PII with cracks
0
0
0.2
0.4
0.6
0.8
0.2
(e)
Figure 2:
0.6
0.4
0.6
0.8
(f)
Sorption isotherm of mixture (a): BI, (b): BII, (c): BBI, (d): BBII,
(e): PI and (f) PII.
Conclusions
The experimental work performed in this paper has shown that the presence of
cracks in hardened HPC mixtures dramatically changed all measured parameters.
Also the effect of both microsilica and aggregates on the moisture transport and
storage parameters was very important. The liquid water transport was found to
be affected by cracks, microsilica and aggregates in a much more significant way
than water vapour transport.
417
Acknowledgement
This research has been supported by the Ministry of Education of Czech
Republic, under grant No. MSM: 6840770003.
References
[1]
[2]
S. Roels, J. Carmeliet, H. Hens, O. Adan, H. Brocken, R. ern, Z.

Pavlk, C. Hall, K. Kumaran, L. Pel, R. Plagge, Interlaboratory
Comparison of Hygric Properties of Porous Building Materials. Journal of
Thermal Envelope and Building Science 27(2004), 307-325.
R. ern, J. Podbradsk, J. Drchalov, Water and water vapor
penetration through coatings. Journal of Thermal Envelope and Building
Science 26(2002), 165-177.
419
Seismic upgrading of square and rectangular

RC columns using FRP wrapping
M. A. N. Abdel-Mooty1, M. E. Issa2, H. M. Farag3 & M. A. Bitar4
1
Department of Construction Engineering,

American University in Cairo, Egypt
2
Department of Structural Engineering, Cairo University, Egypt
3
Department of Structural Engineering,
Military Technical College, Egypt
4
Department of Structural Engineering,
Military Technical College, Egypt
Abstract
This paper presents the results of an experimental study to evaluate the
effectiveness of using FRP wrapping for repair and seismic upgrading of square
and rectangular RC columns in buildings. The factors affecting the performance
of FRP wrapping in rectangular columns under the action of both axial and
combined axial and lateral loads are considered in this paper. These factors
include the rectangularity ratio of the column cross section, the thickness of the
FRP jacket, and the use of carbon verses glass FRP for column jacketing.
Techniques to improve the performance of strengthening rectangular columns
were also proposed and evaluated in the paper. Such techniques include rounding
the sharp edges of the columns, and transferring square into circular columns
using mortar. A total of fourteen half-scale reinforced concrete columns, divided
into two groups, are tested in this research. The first group of columns consists
of three square columns and three rectangular columns which are tested under
axial loads. The second group consists of four square columns and four
rectangular columns tested under combined axial and lateral loads.
Keywords: seismic upgrading, column strengthening, CFRP and GFRP.
Introduction
The use of FRP wrapping for repair, strengthening, and seismic upgrading of
columns has gained increasing attention in recent years [13]. FRP wrapping
was proposed for increasing the ductility of column under axial and axialWIT Transactions on The Built Environment, Vol 85, 2006 WIT Press
doi:10.2495/HPSM06041

flexural loading through confinement [48], improving insufficient shear
strength [9], and insufficient lap splice length [10]. Due to uniform confinement,
FRP wrapping has been proven effective for strengthening circular columns [4,
10]. Other studies showed the improvement in seismic capacity of square
columns wrapped with FRP jacket [7, 8]. It was recommended that the sharp
corners of the column must be rounded to avoid premature failure of the FRP
jacket due to stress concentration at the corners. Studies conducted on wrapping
rectangular columns showed inferior performance to that of circular and square
columns [5, 11].
The work presented in this paper describes the first phase of a two-phase
research program addressing the performance of RC rectangular columns
strengthened using FRP wrapping jacket. The first phase is an experimental
study to evaluate the effectiveness of using FRP wrapping for repair and seismic
upgrading of square and rectangular RC columns in buildings. The second phase
is a theoretical investigation to develop a numerical model capable to accurately
predict the actual behaviour of the repair techniques. The factors affecting the
performance of FRP wrapping in rectangular columns under the action of both
axial and combined axial and lateral loads are considered in this paper. These
factors include the rectangularity ratio of the column cross section, the thickness
of the FRP jacket, and the use of carbon verses glass FRP for column jacketing.
Techniques to improve the performance of strengthening rectangular columns
are also proposed and evaluated in the paper. Such techniques include rounding
the sharp edges of the columns and transferring square into circular columns
using mortar. A total of fourteen half-scale reinforced concrete columns, divided
into two groups, are tested in this research. The first group of columns consists
of three square columns and three rectangular columns which are tested under
axial loads. The second group consists of four square columns and four
rectangular columns tested under combined axial and lateral loads.
Experimental program and setup
The test program in this research aims at studying the various parameters
affecting the behaviour of reinforced concrete rectangular columns strengthened
using FRP wrapping under both axial and seismic forces. These parameters
include the rectangularity ratio of the column cross section, the thickness of the
FRP jacket, and the use of carbon verses glass FRP for column jacketing. A total
of fourteen half-scale reinforced concrete columns, divided into two groups, are
tested in this research. The first group of columns, Group I, consists of three
square columns; CS1, CS2 and CS3; and three rectangular columns; CR1, CR2
and CR3; which are tested under increasing axial loads to failure. The second
group, Group II, consists of four square columns; CS4, CS5, CS6 and CS7; and
four rectangular columns; CR4, CR5, CR6 and CR7; tested under combined
axial and lateral loads. Constant axial loads, 350 kN for square columns and 700
kN for rectangular columns, were applied to Group II columns during increasing
cyclic lateral loading up to failure. This load level is approximately 30% of the
axial load carrying capacity of the tested columns. The details of the test
specimens and program are summarized in Table 1.
Table 1:
Testing load and set-up
Group I:
Axial load
Summary of test program and setup.

Specimen crosssection and
reinforcement
Specimen
No.
Square Columns
CS1
Control
200200 mm
CS2
CFRP
CS3
CFRP
curving
sides
CR1
Control
CR2
CFRP
CR3
CFRP
Square Columns
CS4
Control
200200 mm
CS5
GFRP
CS6
CFRP
CS7
2-layers
CFRP
CR4
Control
CR5
GFRP
CR6
CFRP
CR7
2-layers
CFRP
6/150 mm
FRP
1500
mm
412
Rectangular Col.
Designation
400200 mm
6/150 mm
612
Group II:
Lateral cyclic
loading with
constant axial
load
6/150 mm
412
1600
mm
FRP
Rectangular Col.
750
mm
400200 mm
6/150 mm
400
mm
Table 2:
612
Physical and mechanical properties of GFRP and CFRP.
Property
Tensile strength in fibre direction (MPa)
Elongation at breaking (%)
Tensile modulus (GPa)
Nominal laminate thickness (mm)
Fabric width (mm)
GFRP
575
2.2
26.1
0.17
300
CFRP
965
1.33
73
0.13
350
421

The average concrete strength for the test specimens after 28 days from
casting is 32.5 MPa. Mild steel of diameter 6 mm was used for stirrups while
high grade steel of diameter 12 mm was used for longitudinal reinforcement.
Two types of fibre reinforced polymers were used for jacket wrapping of the
column specimens namely Glass FRP and Carbon FRP laminates. The physical
and mechanical properties for both GFRP and CFRP are displayed in Table 2.
Rigid frame
Jack and
load cell
Steel cap
LVDT
LVDT &
load cell
Steel cap
a) axial load
b) axial and lateral cyclic load
Figure 1:
Schematic diagram for test set-up.
a) Control Specimen CR1 during testing

Figure 2:
Axial load testing and failure of control specimen CR1.
a) CR6 during testing

Figure 3:
b) Failure of CR1
b) Failure of CR6
Lateral cyclic load testing and failure of specimen CS7.
Group II columns were cast integrally with a 1000400400 mm footing for

square columns and a 1200400400 mm footing for rectangular columns. The
footing had six bars of diameter 12 mm as top and bottom reinforcement with
423
stirrups 10 mm diameter spaced at 10 cm. The footing dimensions and

reinforcement were chosen such that its deformations would not affect the
measurements in the test zone.
All columns of Group II were tested under constant axial load and increasing
cyclic load to failure with columns CS4 and CR4 used as reference. FRP
wrapping of columns was applied within the potential plastic hinge zone of the
column for a length of 750 mm from the footing top face. Columns CS5 and
CR5 were wrapped by one layer of GFRP, Columns CS6 and CR6 were
strengthened by one layer of CFRP, and Columns CS7 and CR7 were wrapped
by two layers of CFRP.
The installation of the FRP jacket necessitates surface treatment. Sandpaper
and sand blasting were used for concrete surface preparation before applying the
epoxy resin used for installing the FRP wrapping. The sharp corners of the
rectangular columns were rounded to avoid premature failure of the jacket due to
stress concentration at sharp corners. Epoxy was applied uniformly on the entire
face of the concrete column as well as on the FRP laminate. The fabrics were
then compressed to the concrete surface with a roller. The thickness of the epoxy
layer was controlled to be about 1.2 mm. The specimens were treated at room
temperature for at least 24 hours before testing.
Group I specimens were instrumented with 2 LVDTs at the mid-height to
measure out of plane deformation. Strain gauges were attached to longitudinal
reinforcements and the stirrups at the column mid-height to measure longitudinal
and transverse strain during loading. Surface strain gauges were also mounted to
the concrete surface and to the FRP jacket. Those measurements will be used for
verification of the detailed numerical models developed in the second phase of
this research. As for Group II specimens, LVDTs are used to measure the lateral
displacement at the top and mid-height of the columns during testing.
Furthermore, strain gauges are used to measure vertical and lateral strain on
reinforcement and concrete and FRP surfaces near the column base. The
measured load, displacements, and strains at the various locations were fed into
the data acquisition system and recorded for further processing and analysis.
Figure 1 shows schematic diagrams of the test set-up for Group I and Group II
specimens. Figures 2 and 3 show sample specimens during testing and at failure.
Avreage stress at failure (N/mm2)
60
50
40
CR2
CS1
CS2
CS3
30
20
10
0
0
0.5
1.5
2.5
Strain measured on longitudinal steel RFT(%)
Figure 4:
Variation of axial strain with axial stress for square columns.
60
50
40
CR2
CS1
CS2
CS3
30
20
10
0
0
0.5
1.5
2.5
Transverse strain measured on stirrups (%)
Figure 5:
Variation of transverse strain with axial stress for square columns.

60
CR1
CR2
CR3
(CR1)"
(CR2)"
50
(CR3)"
40
30
20
10
0
0
Figure 6:
0.5
1.5
2.5
Strain measured on longitudinal steel RFT (%)
Variation of axial strain with axial stress for rectangular columns.
60
50
CR1
(CR1)
40
(CR2)"
CR2
CR3
(CR3)"
30
20
10
0
0
0.5
1.5
2.5
Transverse strain measured on stirrups (%)
Figure 7:
Variation of transverse strain with axial stress for rectangular

columns.
425
Analysis of test results
Group I specimens are tested under increasing axial load to failure. The square
column control specimen formed inclined cracks at an average stress of
30.2 N/mm2 followed by fast progressive failure in the form of falling off the
concrete cover and buckling of longitudinal reinforcement. As for specimen
CS2, and CS3, the confinement provided by the CFRP jacket delayed the failure
to higher stresses values, 48.2 N/mm2 for CS2 and 48.9 N/mm2 for CS3. At
failure the CFRP jacket was ruptured due to hoop tensile stresses after which
buckling of steel longitudinal bars occurred. The rectangular specimens showed
similar behaviour up to failure except that the effect of confinement due to the
CFRP jacket was less pronounced. Figures 47 show the variation of axial strain
and lateral strain with axial stress for square and rectangular column specimens
respectively. Figure 8 compares the average axial stress at failure for axially
loaded specimens. FRP confinement creases the ultimate capacity of axially
loaded square columns by approximately 50%. The confinement effect was less
pronounced for rectangular columns where the increase in the ultimate capacity
was about 21%. Transferring square column to circular one slightly increase the
average failure stress by 1.5%, however the overall load capacity of the column
is increased by 5.2% from 193 kN to 203 kN due to increasing the cross section.
Average stress at failure (N/mm2)
60
48.25 N/mm2
50
48.88 N/mm2
38.875 N/mm2
40
35.625 N/mm2
30.75 N/mm2
30.2 N/mm2
30
20
10
CS1
Figure 8:
CS2
CS3
CR1
CR2
CR3
Average axial stress at failure for axially loaded columns.
Group II specimens were tested under constant axial load (30% of the axial
load capacity) and increasing cyclic load to failure. Figure 9 shows the loaddisplacement hysteresis loops for specimen CR6 during testing. All specimens in
this group showed similar behaviour. As load increased cracking started first at
the base, and then more cracks started to develop within approximately 450 mm
from the base. With further increasing load, crack opened and concrete in
compression crushed forming plastic hinge. Figure 10 shows the envelope for
lateral load variation with lateral displacement for square specimens. The
variation of lateral load with strain measured on steel stirrups for square
specimens is displayed in Figure 11. The load-displacement envelope for
rectangular columns is shown in Figure 12. The ultimate lateral load increased
by 11%, 22%, and 44% for square column wrapped by GFRP, CFRP and two

CFRP layers, respectively. Similar improvements were observed for rectangular
columns where lateral load increased by 12.5%, 25%, and 33.3% for rectangular
column wrapped by GFRP, CFRP and two CFRP layers, respectively. Figure 13
compares the lateral load capacity for cyclically loaded columns.
400
300
Lateral load (kN)
200
100
0
-40
-30
-20
-10
10
20
30
40
-100
-200
-300
-400
Displacement (mm)
Figure 9:
Load-displacement hysteresis loops for specimen CR6.

150
Lateral load (kN)
100
CS4
CS5
CS6
CS7
50
0
-50
-40
-30
-20
-10
10
20
30
40
50
-50
-100
-150
Lateral displacement (mm)
Figure 10:
Lateral load vs. lateral displacement envelop for square columns.

150
Lateral load (kN)
100
CS4
CS5
CS6
CS7
50
0
0
0.5
1.5
2.5
3.5
-50
-100
-150
Transverse strain on stirrups (%)
Figure 11: Lateral load vs. lateral strain on steel stirrups for square columns.
427
400
300
CR4
CR5
CR6
CR7
Lateral load (kN)
200
100
0
-40
-30
-20
-10
10
20
30
40
-100
-200
-300
-400
Lateral displacement (mm)
Figure 12: Lateral load vs. lateral displacement envelope for rectangular
columns.
350
300 kN
Maximum lateral load (kN)
300
320 kN
270 kN
240 kN
250
200
130 kN
150
100
90 kN
100 kN
110 kN
50
CS4
CS5
CS6
CS7
CR4
CR5
CR6
CR7
Figure 13: Ultimate lateral load for cyclic loaded specimens.
4 Summary and conclusions

The performance of fourteen RC square and rectangular columns strengthened
using FRP wrapping jacket under axial loading and increasing lateral cyclic
loading was experimentally evaluated in this paper. The loads, displacements
and strains during testing were recorded for further numerical study. It was found
that strengthening RC columns using FRP wrapping is much more efficient in
square columns than in rectangular columns. Failure stresses increased due to
FRP wrapping by 50% in square columns and 21% in rectangular columns.
Transferring square columns to circular columns to improve FRP confinement
increases the ultimate stress by only 1.5%.
Group II columns subjected to combined axial and lateral loads were
wrapped with FRP in the plastic hinge zone. FRP wrapping delayed concrete
cracking, and increase ductility. The rupture of the FRP wrap occurs gradually
with increasing load giving more warning before failure. The ultimate lateral
load is increased by 11%, 22%, and 44% for square column wrapped by GFRP,

CFRP and two layers of CFRP respectively. Similar improvements in the lateral
load capacity were observed for rectangular columns (12.5%, 25%, and 33.3%).
References
[1] Lehman, D.E., Gookin, S.E., Nacamuli, A.M. & Moehle, J.P, Repair of
earthquake-damaged bridge columns. ACI Structural journal, 98(2), pp.
233-242, 2001.
[2] Mosallam, A.S. (ed.). Innovative System for Seismic Repair and
Rehabilitation of Structures Design and Application, Proc. (SRRS2),
Technomic Publishing co., Inc., 2000.
[3] Tan, K.H. (ed.). Fibre-Reinforced Polymer Reinforcement for Concrete
Structures, Proc. of the Sixth International Symposium on FRP
Reinforcement for Concrete Structures, Singapore, 8-10 July, 2003.
[4] Fam, A.Z. & Rizkalla, S.H., Confinement model for axially loaded concrete
confined by circular fiber-reinforced polymer tubes. ACI Structural journal,
98(4), pp. 451-461, 2001.
[5] Wang, Y.C. & Restrepo, J.I., Investigation of concentrically loaded
reinforced concrete columns confined with glass fibre-reinforced polymer
jackets. ACI Structural journal, 98(3), pp. 377-385, 2001.
[6] Chaallal, O. & Shahawy, M., Performance of fiber-reinforced polymerwrapped reinforcement concrete column under combined axial-flexural
loading. ACI Structural journal, 97(4), pp. 659-668, 2000.
[7] Mahmoud, Kh., Fouad, E., Ramadan, M.O. & Abd-Elalim, A., Behaviour
of axially loaded square RC column confined with sandwich FRP wraps.
Proc. International Conference: Future Vision and Challenges for Urban
Development, Paper SG96F, Cairo, 20-22 December 2004.
[8] Allam, H.M., Strengthening of square columns by a New Technique. Proc.
International Conference: Future Vision and Challenges for Urban
Development, paper SG167F, Cairo, 20-22 December 2004.
[9] Haroun, M. A. & Elsanadedy, H.M., Seismic retrofit of shear-deficient
reinforced concrete bridge columns by advanced composite jackets. Proc.
Structural Composite for Infrastructure Applications, Aswan, Egypt, 2002.
[10] Haroun, M.A., Mosallam, A.S., Feng, M.Q. & Elsanadedy, H.M.,
Experimental Investigation of Seismic Repair and Retrofit of bridge
columns by composite jackets. Journal of Reinforced Plastics and
Composites, 22(14), pp. 1243-1268, 2003.
[11] Hosny A., Shahin, H. Abdelrahman, A., & El-Afandy, T., Uniaxial tests on
rectangular columns strengthened with CFRP. Proc. Structural Composite
for Infrastructure Applications, Aswan, Egypt, 17-20 Dec., 2002.
429
Arch bridge made of reactive powder concrete

D. Cizmar 1, D. Mestrovic2 & J. Radic2
1
2
Ekonerg Institute, Zagreb, Croatia

Faculty of Civil Engineering, University of Zagreb, Zagreb, Croatia
Abstract
Concrete is the most commonly used building material. It is used for buildings,
industrial structures, bridges and dams. Every day the concrete is being
improved, to achieve better characteristics, reduce price and to be environmental
acceptable. In the introduction an historical overview of concrete is given
beginning in the 1950s when the compression strength was 40 N/mm2. Today
this is the Industry standard. Concrete with compression strength greater than 40
N/mm2 is called high performance concrete (HPC). The first HPS concretes
were made in 1960. Around 1990, Reactive Powder Concrete (RPC) started to
appear. Their strengths go up to 800 N/mm2. Today, HPC is limited to bridges,
and RPC is still very rarely used. In this work we present a method for making a
RPC composite material with a compression strength of up to 170 N/mm2.
Preparation and testing of material is performed in the laboratory of the Faculty
of Civil Engineering in Zagreb. In addition to the mechanical properties, the
durability parameters are also tested. Detailed concrete mix proportions are given
in the article. The possibility that RPC could be used to construct an arch bridge
over Bakar strait (Croatia) is analyzed. The bridge would have arch span of 432
m. Due to very high compression strength of RPC the span length may be
increased, therefore reducing the total construction cost. RPC is also used to
increase the resistance to freeze-thaw cycles, increase abrasion resistance, and
reduce chloride permeability. Many large Adriatic bridges have experienced
durability problems that are primarily the result of using standard concrete. Each
of these durability enhancements provided by RPC decrease maintenance costs
and lengthens the service life of a structure, which is vital for bridges in the
Adriatic region.
Keywords: arch bridge, reactive powder concrete (RPC), construction
technology.
doi:10.2495/HPSM06042
Introduction
Cement is around 12 millions years old. Reactions between limestone and oil
shale during spontaneous combustion occurred in Israel to form natural deposits
of cement compounds. The deposits were first described by Israeli geologists
between 1960 and 1970. In 3000 years BC, the Egyptians used mud mixed with
straw to bind dried bricks. They also used gypsum mortars and mortars of lime in
the construction of the pyramids. Chinese used cementitious materials to hold
bamboo together in their boats and in the Great Wall of China. The Romans used
pozzolana cement from Pozzuoli, Italy near Mt. Vesuvius to build the Appian
Way, Roman baths, the Coliseum and Pantheon in Rome, and the Pont du Gard
aqueduct in southern France. They used lime as a cementitious material. Pliny
reported a mortar mixture of 1 part lime to 4 parts sand. Vitruvius reported a 2
parts pozzolana to 1 part lime mixture. Animal fat, milk, and blood were used as
admixtures (substances added to cement to improve properties).
Figure 1:
Maxentius basilica.
Fig. 1 shows the Maxentius basilica built in the 4th Century. These materials
were not used again until the beginning of the 19th Century. In 1824, Joseph
Aspdin of England invented Portland cement by burning finely ground chalk
with finely divided clay in a lime kiln until carbon dioxide was driven off. The
sintered product was then ground and he called it Portland cement named after
the high-quality building stones quarried at Portland, England. Portland cement
started the era of modern composition cements. In 1828, I. K. Brunel made the
431
first significant engineering application of Portland cement, using it to fill a

breach in the Thames Tunnel. In 1867, Joseph Monier of France reinforced
William Wands (USA) flower pots with wire ushering in the idea of iron
reinforcing bars (re-bar). In 1889, the first concrete reinforced bridge was built.
Around 1950, concrete with a compression strength of 40 N/mm2 was made. In
the 1960s, High Performance Concrete (HPC) was made. HPC is the name
given to a class of materials that exhibits properties superior to those of
conventional concrete. The superiority may lie in one or more of several
attributes, such as strength, stiffness, freeze-thaw durability, or resistance to
chemical attack. The properties are selected on the basis of the requirements of
the particular application. Its compression strength ranges from the 50 N/mm2 to
100 N/mm2. In the 1970s, fiber reinforcement in concrete was introduced.
During the 1980s, superplasticizers were introduced as admixtures. Around
1990, Reactive Powder Concrete (RPC) appeared. Their compression strength
goes up to 800 N/mm2. While HPC is increasingly used for bridges, RPC is still
very rarely used.
Components for RPC mixture
2.1 Introduction
This article presents the possibility of making RPC concrete with compression
strengths up to 200 N/mm2. Four different mixtures are analyzed. The first is
mixture of hybrid micro-fiber concrete; the others are composed of only one type
of fibers.
2.2 Cement
As the class of cement increases the compression strength increases. For this
mixture we have selected the Portland cement PC 55 with no mineral
ingredients.
2.3 Fine aggregate
Aggregate that is used for the making of this mixture is quartz aggregate. Two
fractions of this material are used: one with soil size of 0.1250.25 mm, other
with 0.250.5 mm, This effectively means that maximal soil size is 0.5 mm.
2.4 Steel fibers
Two different types of steel fibers are used (shorter and longer fibers). Shorter
fibers are 13 2 mm long, diameter is 0.2 0.02 mm. Minimal tensile strength is
2600 MPa. Longer fibers have curved ends; their length is 40 3 mm and
diameter 0.5 0.02 mm. Minimal tensile strength is 2600 Mpa. In both cases, the
high tensile fibers are used to achieve necessary ductility.
2.5 Silica fume
Silica fume is a pozzolanic additive, with specific area of 20 m2/g.

Other parameters are:
- density
- sieve residue
45 m
- pH value
8.44
2.23 g/cm3
5.6%
2.6 Superplasticizer
Superplasticizer is primary used to decrease the participation ratio of water in the
concrete mixture. Chosen superplasticizer is based on policarbon-silate. It is a
brown fluid, dissolves in water and doesnt contain chlorides.
Other parameters are:
- density
1.064 kg/dm3
- pH value
7
- alkalinity
0.31 %
- viscosity on 20C
134 mPas
Table 1:
Ingredients per m3 of RPC.
Mixture number
M1
M2
M3
M4
Steel fibers
(kg/m3)
SF1 (40/0.5)
SF2 (13/0.2)
76
190
228
228
234
720
955
720
980
230
239
230
303
123
1112
105
945
123
1111
105
965
Superplasticizer
(kg/m3)
30
35
31
40
Water (l/m3)
190
215
190
209
24C
5%
2.41 kg/m3
140 mm
25C
5%
2.35 kg/m3
250 mm
24.5C
5%
2.36 kg/m3
190 mm
26C
5%
2.306kg/m3
220 mm
Cement (kg/m3)
Fine aggregate
(kg/m3)
Quartz sand
0.1250.25
0.250.5
(kg/m3)
Concrete
properties
temperature
pores
density
consistency
433
2.7 Compatibility of cement and superplasticizer

The compatibility of PC 55 cement and superplasticizer Glenium ACE 30 is one
of the important requirements for achieving a good RPC mixture. Compatibility
is measured by a change in the consistency of the concrete. To achieve proper
treatment of concrete it must retain its characteristics for approximately 1.5 h for
site application and 0.5 h for prefabrication.
Compression and tensile strength measurement were conducted on four

specimens with a prismatic shape (40 x 40 x 160 mm). The specimens were 28
days old. First, the tensile strength was measured, then the compression strength.
Table 2 presents the experimental results for each mixture. Fig. 2 shows the
specimens after the tests.
Table 2:
Flexure strength (mean)

(MPa)
Compression strength (mean)

(MPa)
M1
M2
46.9
42.8
132.0
155.6
M3
M4
42.8
48.8
153.3
174.8
Figure 2:
Experimental results.
Specimens after testing.
Gas permeability test
Gas permeability is tested according to Croatian regulations EN 993-4.

Specimens were cylindrical, the diameter and length were 50 mm. They were
taken from a 10 x 10 x 50 cm prism (first mixture). The specimens were put in

dry chamber until constant mass was achieved. The specimens were than cooled
to room temperature, polished and coated with epoxy. A pressure difference of 3
bars wasnt detected, which means that gas permeability should have been very
low (none was detected).
Capillary water test
A capillary water test is conducted according to Croatian regulations

HRN.U.M8.300:1985 in the laboratory of the Civil engineering faculty in Zagreb
on specimens 28 days old. The diameter of the specimen was 15 cm and hadd a
height of 10 cm. Before test the specimens are dried on 105C and then left for 2
days in laboratory. Sealing putty is applied on one side. Specimens are then
weighed.
The capillary water test was performed in intervals of 1, 3, 5, 10, 15, and 30
minutes after immersing in water, and then after 1, 2, 3, 4, 5, 6, and 24 hours.
Results show the linear proportion between the capillarity water and square root
of time. The height of capillarity water could not be determined because it didnt
cross the area sealed with sealing putty. The starting absorption capacity for
ordinary concrete is 0.25 ml/m2/s after 10 min, 0.17 ml/m2/s after 30 min, and
0.10 ml/m2/s after 1 hour. These results show (table 3) that RPC has very little
water permeability. Its absorption after 10 minutes is much lesser than that of
ordinary concrete after an hour.
Table 3:
Capillary water test.
Time
0 min
1 min
3 min
5 min
10 min
15 min
30 min
1h
2h
3h
4h
5h
6h
24 h
A sample
(kg/m2/s)
0
0.000613
0.00029
0.00022
0.000122
0.000089
0.0000526
0.0000314
0.0000182
0.0000133
0.0000105
0.00000888
0.00000451
0.00000242
B sample
(kg/m2/s)
0
0.000754
0.000306
0.000212
0.000122
0.0000896
0.0000518
0.0000310
0.0000181
0.0000133
0.0000107
0.00000888
0.00000451
0.00000251
C sample
(kg/m2/s)
0
0.00066
0.000291
0.000198
0.000117
0.0000864
0.0000511
0.0000302
0.0000176
0.0000131
0.0000103
0.00000864
0.00000439
0.00000241
6 Bridge over Bakar strait

We will now explore the possibility that a future arch bridge over Bakar strait
could be made of RPC. By building this bridge the highway would be 7 km
435
shorter. The bridge arch span would be 423 m, vertical clearance is 72 m. Bridge
has 22 piers (spans are 30+20x38+30 m). Fig. 3 represents the longitudinal
section of the bridge. A cross section of the bridge is shown on fig. 4. The cross
section is constant and aerodynamic. With the exception of the foundations,
abutments, and highest piers, the whole bridge will be build from pre-cast
segments 3.8 m long. To achieve symmetric load bearing structure will be
supported from both sides. Segments that weight 718 kN are being composed
and pre-stressed in pressurising station behind the abutments and then pushed.
Computer simulation of the Bakar bridge is presented in fig 5.
Figure 3:
Figure 4:
Figure 5:
Longitudinal section of the bridge.
Cross section of the bridge.
Computer simulation of the bridge.
Basic design
Calculations are made for stresses during usage, strains during construction,
stresses and strains of bearing structure during construction and usage. With the
exception of dead and traffic load these calculations cover temperature load,
creep, lateral wind, and earthquake load. Table 4 shows maximal stresses for
each part of the bridge. Road slab and transverse beams are the only parts of the
structure where tensile stresses are about 15 MPa. Due to very high stresses,
passive reinforcement with ropes will be used. In transverse beams internal prestressing with addition of passive reinforcement will be used. This conception is
not very often with RPC and in future phases of the project it should be carefully
considered.
Table 4:
Part of structure
Road slab
Transverse beam
Highest pier
Arch
Maximal stresses.
Maximal compression (Mpa)

15.30
10.24
15.36
67.35
Maximal tension (Mpa)

15.30
15.50
6.92
10.61
Conclusion
In this article we have presented the possibility that a future arch bridge over the
Bakar strait (Croatia) could be made from RPC. Due to the very high
compression strength of RPC it can be used for large spans. RPC also has superb
durability parameters as well as a high abrasion resistance and reduced chloride
permeability. This makes RPC an ideal material for bridges on the Adriatic cost
because durability problems are primary related to the fact that
the
reinforcements protective layer is rapidly being destroyed. Winds that reach up
to 250 km/h drift large amounts of chlorides that destroy the bridge structure,
primary the arch and the columns. Each of these durability enhancements
provided by RPC decrease maintenance costs and lengthen the service life of a
structure, which is vital for bridges in the Adriatic region.
References
[1] Candrlic, V., Concrete arch bridge over Bakar straits. Proceedings for
Conference of Croatian builders, eds. V. Simovic, Cavtat, pp. 358-364,
2001.
[2] Jagar, A., High performance concrete, Faculty of civil engineering: Zagreb,
2003.
437
[3] Mrakovcic, S., Precast arch bridges made of reactive of powder concrete,
Faculty of civil engineering: Zagreb, 2001.
[4] Edward Nawy, G., Fundamentals of high-performance concrete, John Wiley
& Sons: New York, 2001.
439
Elastic and inelastic seismic response

comparison of reinforced concrete buildings
with normal resistance concrete and
with high resistance concrete
J. A. Avila1, 2 & D. Rivera2
1
Institute of Engineering, National University of Mexico (UNAM),

Mexico
2
Faculty of Engineering, National University of Mexico (UNAM), Mexico
Abstract
The elastic and inelastic seismic response of 3, 9, 17 and 25 levels of reinforced
concrete buildings (offices), located in the soft soil of Mexico City, with normal
resistance concrete (fc= 250 kg/cm2) and with high resistance concrete (fc= 700
kg/cm2) are compared. The design with normal and high resistance concrete is
made with the RDF-93 and RDF-04 codes respectively. The design results
(transversal sections dimensions, vibration periods, lateral displacements,
reinforcement steel, etc) are compared, after making a spectral modal dynamic
analysis, as well as the non-linear responses (lateral displacements, global and
local ductility demands, global distribution of plastic hinges, etc) from the
step-by-step dynamic analysis with the SCT-EW record of the 1985 earthquakes.
Introduction
The elastic and inelastic behavior of 3, 9, 17 and 25 levels reinforced concrete

buildings (offices) is compared; it is designed with the RDF-93 Code [1] for
normal resistance concretes and with the RDF-2004 Code [2] for high resistance
concretes. The service (the ratios of relative lateral displacement do not exceed
of 0.012 times the story high) and failure (given resistances to satisfy the seismic
behavior factor Q= 3 requirements) limit states are satisfied according to the
Complementary Techniques Norms in both Codes. The structures are designed
doi:10.2495/HPSM06043

for the soft soil conditions (zone III for the RDF-93 and zone IIIb for the RDF2004) in Mexico City. For the design, the spectral modal dynamic seismic
analysis is used, considering the elastic-lineal three-dimensional behavior; the
vibration periods, maximum lateral displacements, ratios of relative lateral
displacements to story high, shear forces, mechanics design elements (internal
actions), and longitudinal and transversal reinforcements are compared. Based on
these designs, the non-linear response is later determined making step-by-step
dynamic analysis in time history, using the SCT-EW record representative soft
soil and the bigger damages zone during the 1985 September earthquake in
Mexico City. The maximum local ductility in beams and columns are calculated,
as well as the global ductility and the tendencies that the failure mechanics
develop. Finally, conclusions and recommendations to follow in the practical
design of the kind of structures are presented, regarding both concrete resistance
type.
1.1 High resistance concretes
Some of the high resistance concrete mechanical properties are different from the
conventional concrete. Due to this some doubts about if the actual structural
design procedures for conventional concretes are also used in high resistances
ones. In the Techniques Norms of RDF-2004 it is allowed to use high resistance
concretes with values of fc until 700 kg/cm2, except the structures designed with
a seismic behavior factor Q= 4 and members facing flexion-compression that
take part of frames resisting more than 50% of the seismic actions and which
design axial load (Pu) is bigger than 0.2 PRO, where PRO is the design resistant
axial load; in this cases only concretes with fc until 550 kg/cm2 (55 MPa) can be
used. This Norms propose to use the following equation in order to obtain the
elasticity modulus in high resistance concretes:
Ec= 7700 (fc)1/2 + 163000 (kg/cm2) or Ec= 2400 (fc)1/2 + 16300 (MPa)
Fig. 1 shows some curves stress-deformation for different values of fc. Some
of the main parameters to consider in high resistance concretes are the following:
a) workability, b) permeability, c) volumetric changes, d) durability in useful
life. For design consideration purposes, flexion and axial load, many of the
design procedures used for the conventional concretes have been applied and in
most of the cases good results have been obtained.
The maximum unitary deformation value in the concrete cu= 0.003 adopted
in conventional concrete design, has also adequate, but less conservative for high
resistance concretes. The equivalent rectangular block of compression stress is
proposed, but its application in high resistance concrete members facing to
flexion and flexure-compression has been questioned because in the Norms a
superior limit of fc from which such block is invalid has not been specified. The
concrete ductility is going to diminish as long as the fc value increases. The
transversal reinforcement presence increases the resistance and ductility of a high
resistance concrete column, but in less magnitude than the normal concrete due
441
to the tendency of lateral deformations to be considerably smaller, which

produces a less effective confinement. If the axial load is bigger than 0.40 PRO
the use of high resistance steel for the transversal reinforcement is recommended,
trying to avoid the reinforcement congestion in the structure nodes. The
confinement level is directly proportional to the concrete resistance and does not
consider the axial load level in the structural element. Therefore in members
where the fc resistance is high, we have a high transversal steel percentage and
in consequence construction problems.
Despite of the augmentation in concrete costs as its resistance increases, the
total cost of the structure will be less, because the concrete resistance
augmentation causes an important reduction in the structural elements
dimensions as well as in the reinforcement areas.
Figure 1:
Curves stress-deformation for several concretes.
Structures description
Two kinds of structures are considered: A case (normal concrete with fc= 250
kg/cm2 and B case (high resistance concrete with fc= 700 kg/cm2). The elasticity
modulus for both cases is Ec= 14000 (fc)1/2 (221,359 kg/cm2) and Ec= 7700
(fc)1/2 + 163000 (366,723 kg/cm2), respectively. The concrete is class 1 with
volumetric weight c= 2400 kg/cm2 and = 0.2; the reinforced steel used has a
yield stress fy= 4200 kg/cm2.
Figs. 2 and 3 show the type plant and transversal cuts of 3 and 9 level
buildings; fig. 4 shows the type plant of 17 and 25 level buildings. Also, the

structural characteristics of each building are presented. The floor system is
concrete slab type of 10 cm thickness.
2.1 Design criteria
800
800
800
800
800
250
800
250
The gravitational loads effects and those of second order (P-) are included in
the analysis. For A case the seismic design spectra of soft soil is considered
(Zone III) of the RDF-93 and the Zone IIIb spectra of the RDF-2004 for the B
case, considering the seismic behavior factor Q= 3 (see fig. 5).
The reinforcement steel areas of the structural elements design (beams and
columns) is made with the last mechanical elements obtained from the structural
analysis for the critical load combination; it is according to the general
requirements and ductile frames of the Concrete Norms. These designs include
the load factor and the resistance reduction factor.
Figure 2:
SECONDARY
BEAMS
VIGAS
SECUNDARIAS
(CENTIMETERS)
Type plant and three-dimensional 3 level model.
150

1
2
800
3
800
4
800
800
800
800
800
E
VIGAS
SECUNDARIAS
SECONDARY
BEAMS
(CENTIMETERS)
astic
responses
ROOF
3.50
al8 sections and reinforcement areas

3.50
s7 of the3.50
structural elements of the transversal sections in B cas
is6 reduction is more important in columns, so that the resistanc
mpression3.50in the concrete has a bigger influence in the behavior o
h5igh axial loads. In the majority of B case elements, the require
areas w3.50
ere smaller; only the some external axes columns, th
4
esult similar
in both cases.
3.50
3
3.50 s of vibration
ntal period
2
3.50
res the fund
amental periods of vibration of analyzed A and B cas
1
A case structures tend to be less rigid.
7.50
STREET LEVEL
daLSmental periods of vibration comparison, 3, 9, 17 and 25 leve

3.75
dels (A and
B cases)
BAS
(METERS)
3.00
CIM
Figure 3:
Type plant and transversal cut, 9 level model.
443

2
800
800
800
800
800
800
D
VIGAS SECUNDARIAS
SECONDARY BEAMS
(CENTIMETERS)
Figure 4:
Type plant, 17 and 25 level models.
Design spectra, zones III and III-b

0.50
Sa/ g
0.45
0.40
0.35
0.30
0.25
0.20
0.15
0.10
0.05
0.00
0
Ti [seconds]
RDF-93,Q= 1
Figure 5:
RDF-04, Q= 1
RDF-93, Q=3
RDF-04, Q=3
Design spectra RDF-93 and RDF-2004, Q= 1 and Q= 3.
Design elastic responses
3.1 Transversal sections and reinforcement areas

The dimensions of the structural elements of the transversal sections in B case
are smaller. This reduction is more important in columns, so that the resistance
increase to compression in the concrete has a bigger influence in the behaviour
of elements with high axial loads. In the majority of B case elements, the
required reinforcement areas were smaller; only the some external axes columns,
the reinforcement result similar in both cases.
445
3.2 Fundamental periods of vibration

Table 1 compares the fundamental periods of vibration of analysed A and B case
buildings. The A case structures tend to be less rigid.
Table 1:
Fundamental periods of vibration comparison 3, 9, 17 and 25 level

models (A and B cases).
MODEL
3
9
17
25
Fundamental period of vibration [seconds]

X
Y
0.80
0.80
0.69
0.70
0.74
0.63
1.49
1.45
1.07
1.44
1.41
1.09
1.91
1.91
1.38
1.84
1.84
1.42
2.10
2.10
1.39
1.88
1.88
1.38
CASE
A
B
A
B
A
B
A
B
Fig. 6 has the September 19th 1985 earthquake SCT-EW accelerogram, used
later in the inelastic analysis. Fig. 7 shows the elastic and inelastic response
spectra of this record ( = 5%) and of RDF-2004, as well as the fundamental
periods of vibration location (X direction) of all the structures, A and B cases.
When the level number increases, the difference between both cases tends to be
bigger.
0.20
0.15
0.10
0.05
Sa/g
a/ g
0.00
-0.05
-0.10
-0.15
-0.20
0
10
15
20
25
30
35
40
45
50
tiempo [s]
Time (seconds)
Figure 6:
th
September 19 1985 earthquake SCT-EW accelerogram.
3.3 Total maximum lateral displacements

In both design cases very similar responses are presented, so that, if the A case
structures result less rigid than those of B case is because they were designed
with the RDF-93 with a maximum spectral level of 0.40, in comparison with the
0.45 value for the RDF-2004.
3.4 Ratios of relative lateral displacement to story high (drifts)
The 3 and 9 level buildings present less drifts because in these the resistance
conditions ruled, therefore it was necessary to increase the dimensions in some

of the structural elements regarding to the required in the service limit state,
checking not to exceed the 0.012 permissible limit.
1.20
1.20
1.10
1.10
1.00
1.00
A case
B case
0.90
0.80
0.80
0.70
0.70
Sa/g
Sa/g
0.90
0.60
0.60
0.50
0.50
0.40
0.40
0.30
0.30
0.20
0.20
0.10
0.10
0.00
0.00
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
0.0
0.5
1.0
1.5
Period (seconds)
duct=1
duct=2
Figure 7:
duct=3
duct=4
RDF-04 Q=1
2.0
2.5
3.0
3.5
4.0
Periodo [s]
Period
(seconds)
Periodo [s]
RDF-04 Q=3
duct=1
duct=2
duct=3
duct=4
RDF-04 Q=1
RDF-04 Q=3
Fundamental periods of vibration location regarding the design

spectra of the RDF-2004 and of response spectra SCT-EW, A and
B cases.
Inelastic step-by-step seismic analysis responses
4.1 Maximum lateral displacements and global ductility demands

Generally the inelastic response was less than the elastic, especially in high
buildings (17 and 25 levels), because in these there was bigger energy
dissipation; in the 3 level building both responses were very similar. Table 2
presents the global ductility maximum demands (G) developed in each structure;
excepting the 3 level building, the mG values are bigger in A case buildings with a
bigger energy dissipation than those in B case; it is important to notice that in
any case the G was bigger than the design value of Q= 3.
4.2 Ratios of relative lateral displacement to story high (drifts)
The inelastic response diminishes considerably from the elastic, until reaching
permissible level values; the A case responses tend to be bigger.
Table 2:
MODEL
3
9
17
25
Global ductility demands (mG) 3, 9, 17 and 25 level models (A and

B cases).
AXE
A
3
A
C
A
B
A
B
max [cm]
5.48
5.64
41.12
43.06
51.00
48.98
77.26
76.73
Global ductility demand (G)

A case
B case
y [cm]
max [cm]
y [cm]
G
4.82
1.14
5.27
3.74
5.42
1.04
5.90
3.15
17.40
2.36
32.30
13.80
16.70
2.58
29.10
14.90
22.40
2.28
39.88
19.85
18.50
2.65
38.36
18.84
29.30
2.64
45.18
20.61
30.70
2.50
44.10
28.56
G
1.41
1.91
2.34
1.96
2.01
2.04
2.19
1.54
447
4.3 Ratios of basal shear force-roof lateral displacement

Table 3 compares the maximum values of the roof lateral displacement and the
basal shear force, elastic and inelastic behavior, of the analyzed structures (3, 9,
17 and 25 levels), A and B cases. Excepting the 3 level building, the ratios of
basal shear force-roof lateral displacement result bigger in B case.
Table 3:
MODEL
Ratios of basal shear force-roof lateral displacement.
AXE
Roof lateral displacements and

maximum basal shear forces
A case
B case
V [t]
[cm]
[cm]
5.47
161
5.68
5.45
156
5.27
5.61
60
5.89
5.64
57
5.90
49.30
859
27.90
41.12
382
30.93
44.80
706
28.11
43.05
406
29.11
149.00
2710
93.14
51.00
582
44.81
148.00
2610
92.00
48.97
566
46.58
143.00
3920
107.84
77.26
1261
55.56
142.00
3830
107.57
76.73
1269
55.39
CASE
Elas.
Inel.
Elas.
Inel.
Elas.
Inel.
Elas.
Inel.
Elas.
Inel.
Elas.
Inel.
Elas.
Inel.
Els.
Inel.
A
3
3
A
9
C
A
17
B
A
25
B
Level
V [t]
203
164
56
52
525
331
518
329
1792
424
1757
407
3277
921
3229
910
Level
18
17
16
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
26
25
24
23
22
21
20
19
18
17
16
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
-8
-6
-4
-2
Ductility maximum demand

A case
Figure 8:
-6
-5
-4
-3
-2
-1
Ductility maximum demand
B case
A case
B case
Local ductility maximum demands (L) in beams, 17 and 25 level

buildings, A and B cases.
4.4 Local ductility maximum demands (L) in beams and columns and
global distribution of plastic hinges
The 3 level building does not present yield in columns. The beams in all
buildings have very similar responses, independently of the case type; the beams
maximum demands are rarely bigger in A case. The L maximum values in
beams as in columns are obtained in the 9 level building, but inside the
admissible limits from the practical point of view. Fig. 8 compares the local

ductility maximum demands developed in the beams of the 17 and 25 level
buildings central axes, A and B cases, for illustrative purposes.
In all buildings, A and B cases, global distribution of plastic hinges presents a
general tendency towards the failure mechanism known as strong column-weak
beam type. This means that plastic hinges are in the most part of the beams and
only in some columns, which is according to the actual design philosophy of
RDF-93 and RDF-2004 codes.
Conclusions
Use of high resistance concretes (fc > 400 kg/cm2) in reinforced concrete
columns improves in a great deal its axial load capacity; therefore in front of
actions where compression high loads predominate its behavior is very adequate;
nevertheless, in front of less magnitude loads or even tensions its use results
almost no efficient because the resistance is given mainly by the reinforcement
steel. The most part of the energy dissipation is in the beams, no matter what
type of structure it is. The A case structures present a bigger amplitude in the
hysteretic cycles of the ratios of basal shear force-roof lateral displacement, with
bigger dissipated energy amount for inelastic deformations. The maximum
values of local ductility demands are found inside the permissible by the RDF-93
and RDF-2004. The inelastic response is considerably reduced regarding to the
elastic one, except in the 3 level building, where the elastic and inelastic
responses tend to be practically the same. In every analyzed case, reviewing the
shear forces history, the corresponding resistance is never reached; this is, there
is always a resistance reservation, which is bigger in B case buildings, so that for
this case a bigger quantity of transversal reinforcement by confinement was
required because the design conditions for transversal reinforcement must be
more strict as the fc value increases.
References
[1]
[2]
Diario Oficial de la Federacin, Reglamento de Construcciones del

Distrito Federal, 1993.
Gaceta Oficial del Gobierno de la ciudad de Mxico, Reglamento de
Construcciones del Distrito Federal, 2004.
449
The strength effects of synthetic zeolites on

properties of high performance concrete
P. Frontera1, S. Marchese1, F. Crea1, R. Aiello1 & J. B. Nagy2
1
Department of Chemical Engineering and Materials,

University of Calabria
2
Laboratoire de RMN, Facults Universitaires Notre-Dame de la Paix
Abstract
The influence of several kinds of synthetic zeolite as mineral admixtures on the
workability and performance strength of mortars cement and concrete has been
investigated. In the first series of experiments zeolites have been used to replace
10% of cement in the preparation of mortars. The second series of experiments
have regarded the preparation of high performance concrete. In order to evaluate
the influence of alkali cations, two different forms of zeolite A have been used
with one sodium (NaA or 4A) and the other with calcium (CaA or 5A) and also
to evaluate the filling effects of particles zeolites 4A and 5A with different
crystals sizes have been used. Furthermore we have used three types of cement
with a different content of clinker: CEM I 42.5R, CEM II/A-S 42.5 R and CEM
III/A-S 42.5 N. Every prepared sample has been water-cured at 20C for 2, 7 and
28 days. The results of samples (mortars and concrete) made with zeolites were
compared with those obtained with silica fume, usual fine material used in the
preparation of high performance strength concrete.
Keywords: admixtures, zeolites, mortars, concrete, compressive strength.
Introduction
High strength concrete exhibits the rheological, mechanical and durability

properties better than those of conventional concrete. The use of high concrete
results in many advantages, such as reduction in beam and column sizes and
increase in the building height. Moreover high strength concrete can perform
much better in extreme and adverse climatic condition, reducing maintenance
and repair costs. Silica fume is among one of the most recent pozzolanic
doi:10.2495/HPSM06044

materials currently used in high performance concrete production, this material is
a by-product of the manufacture of silicon of various silicon alloys. The use of
silica fume is very expensive, because it is light and has a low bulk density,
which may cause difficult in transporting and handling.
The goal of this research was to study production and utilization of alternative
materials that is suitable for use in concrete industries as a substitute for the
expensive silica fume. The material examined has been the synthetic zeolites
(LTA, FAU), with different content of alkali cation (Na+, Ca++) and micronic
particles size. The structure of the natural and synthetic zeolitic materials is
characterized by a large number of channels and cavities which exhibit a high
surface area. The natural zeolites are hydrated alumino-silicate minerals
containing alkaline and alkaline-earth metals, formed by the alteration of
volcanic ash which is mainly an amorphous, siliceous material. They are
considered as materials with high pozzolanic activity for their open structure and
chemistry [1-5]. The pozzolanic activity of zeolites depends on the presence of
reactive SiO2 and Al2O3 which react with the Ca(OH)2 liberated during the
hydratation of cement and converting it into C-S-H gels and aluminates. As a
result, the microstructure of hardened cement concrete is improved and the
concrete becomes more impervious. In fact, a work of Chan et al., found that the
use of natural zeolites as admixtures in concrete concurs to make high
performance concrete (HPC concrete with high strength over 80 MPa), that
usually are obtained with admixtures such as silica and fly ash [5].
In literature numerous reports exist on the use of natural zeolites in the
concrete [7-11], little work instead consider the use of synthetic zeolites in
cement industry [12-14]. This is probably linked with the wide availability of
natural zeolites that economically concur being more favourable for large scale
applications. However, the limit in the use of natural zeolites is in the fact that
they have variable chemical compositions depending on their origin, and in
addition, variable amounts of impurities can be present. Another limitation is that
the natural zeolitic material before being used as additives in concretes must be
finely crushed. For this reason we use the synthetic zeolites in form of powders,
with fixed chemical compositions, which have also a low cost.
In this work the effect of synthetic zeolite A has been examined, in particular
the zeolite 4A and 5A, which differ for the presence of Ca++ cations. Moreover,
crystals with different sizes have been used in order to evaluate the filling effect
of the particles.
Experimental
2.1 Materials
- Portland cement CEM I 32.5 R with content of clinker >95% (UNI EN 197/1).
- Portland- slag cement CEM II/A-S 42.5 R with content of clinker between 80
and 94% and content of slag ranging between 6-20%( (UNI EN 197/1).
- Blastfurnace cement CEM III /A-S 42.5 N with content of clinker between 35
and 94% and slag ranging between 6 and 65 % (UNI EN 197/1).
451
Four different synthetic zeolites were prepared: two of which having small
particle size:
- zeolite 4A with size of crystals ranging about 1-5 m (NaAL) (Fig.1).
- zeolite 4A with size of crystals ranging about 0.2-1 m (NaAS) (Fig.2).
- zeolite 5A with size of crystals ranging about 1-5 m (CaAL) (Fig.1).
- zeolite 5A with size of crystals ranging about 0.2-1 m (CaAS) (Fig.2).
Silica Fume (SF) (trade name MAPEPLAST SF) produced and supplied by
MAPEI (Italy).
Coarse Aggregate rounded with 25 mm maximum diameter in accordance with
UNI 8520-2.
Fine Aggregate is standard sand in accordance with UNI EN 933-3.
The Superfludifiant used was acrylic polymers without formaldehyde with 30.5
solid content (trade name Dynamon SP1) produced and supplied by MAPEI
(Italy).
2.2 Preparation of zeolites
The small crystals of zeolites 4A were prepared starting from the following
initial reaction mixture: 2.37Na2O 1Al2O3 1.90 SiO2 62 H2O
The synthesis procedure was the following: sodium aluminate was added to a
solution of sodium hydroxide solution and after the homogenisation the silica
source was added. The solution was stirred for 15 min at room temperature and
then submitted to hydrothermal treatment, in static conditions, at 25C for about
7 days.
The large crystals of zeolites 4A were prepared starting from system having
the composition: 3.2 Na2O 1Al2O3 1.90 SiO2 95 H2O. The reagents admixture
was made as described above. Successively, the solution obtained has been
submitted to hydrothermal treatment at 95 C for 2 h, in stirred conditions.
After the hydrothermal treatment every samples are recovered, washed with
distilled water, dried at 110C for 12 hours and successively equilibrated at room
temperature before X-ray and microscope analysis.
The X Ray analysis confirms the presence of only zeolite A in all samples
obtained.
Through microscope analysis it is possible to evaluate the different
dimensions of crystals obtained (Fig.1, Fig.2).
The zeolites 5A were prepared by ionic exchange of zeolite 4A, 250g of 4A
have been contacted with 1liter of a 1M aqueous solution of calcium nitrate at
room temperature. This exchange procedure was applied three times. The
efficiency of exchange has been checked by EDAX (Energy Dispersive X-Ray
Analysis), the zeolite 5A obtained has a ratio of ionic content (Ca++/ (Ca+++
Na+)) of about 0.9.
2.3 Preparation of mortars
Cement mortars were prepared for determination of compressive strength of the
mixtures. Zeolites were used as direct replacement of cement on a weight to
weight basis at the levels of 10%. The sand/cement ratio was 3 and the

water/solid ratio was 0.5. To obtain the cement mortars all solid and water was
mixed for 5 minutes and then cast in prismatic moulds.
The specimens were demolded after 24 hours of moist-curing at 23C.
Companion specimens were placed in water at 20C after demolding.
Compressive strength was determined at designated ages.
Figure 1:
A) SEM micrographs of the small crystal of zeolite A (NaAS,

CaAS) B) SEM micrographs of the large crystal of zeolite A (NaAL,
CaAL).
2.4 Preparation of concrete

The concrete mixtures were prepared at water to cementitious material ratio 0.35.
Zeolite was used in addition of cement on a weight to weight basis at the levels
of 10%. Also the concrete mixtures containing silica fume were prepared at a
water-to cementitious ratio of 0.35.
Cube specimens of concrete were cast in 100mmx100mmx100mm standard
moulds, respectively. The specimens were removed from the moulds after 1 day
and were then cured in water at 20C until the age of strength testing. The mix
proportions of concrete prepared are shown in Table 1.
Table 1:
Mixture proportions of concrete.
Mix Proportion (kg/m3)

W/C Zeolite water Cement
0,35
50
175
500
Fine
aggregate
693
Course
Aggregate
1040
Superfluidifiant
5
2.5 Characterization and testing

The compressive strength of the concrete and mortars were tested at the ages of
2, 7 and 28 days using a machine (MATEST in accordance with UNI 6686-1,2)
the loading rate of 1000 N/mm2sec.
Moreover, in the case of preparation of concrete the workability has been
measured, following the prescription of UNI 9418, we have executed the slump
test on fresh concrete.
453

3.1 Properties of mortars
The results of the compressive strength test of the mortars are shown in Table 2.
Table 2:
Compressive strength of mortars.

Compressive strength (MPa)
Cement
Zeolites
2 days
7 days
28 days
CEM I
-NaAS
NaAL
CaAS
CaAL
SF
-NaAS
NaAL
CaAS
CaAL
SF
-NaAS
NaAL
CaAS
CaAL
SF
18.9
17.7
21.9
26.1
25.9
20.3
22.5
20.6
28.0
26.8
30.0
21.1
15.7
15.7
20.4
20.1
22.3
13.5
29.2
21.2
30.1
34.5
33.4
29.5
32.6
28.5
34.3
35.7
37.2
28.6
23.5
20.9
26.7
30.4
32.6
22.9
34.0
25.3
31.3
36.2
39.1
36.4
42.8
39.9
37.7
42.7
42.8
40.5
42.5
23.9
30.5
38.0
40.8
39.0
CEM II/A-S
CEM III/A
The replacement of 10% of CEM I, with most of zeolites, generally improves

the compressive strength of the mortars cement, both at early ages and longer.
Only the zeolites NaAS and NaAL make worse the compressive strength of the
mortars made with CEM I. The mortars prepared with CEM I and Zeolite CaAS
and CaAL exhibit compressive strength alike and higher, respectively, than the
mortars with CEM I and silica fume.
In the case of CEM II/A-S, the replacement of 10% of this type of cement in
the preparation of mortars with all zeolites and also with silica fume does not
improve significantly the compressive strength at later ages as 28 days, but at
early ages it is possible to remark an increase of compressive strength.
For the CEM III/A-S, that is a cement with a high content of fly ash with a
low rate of hardening, the replacement of 10% of zeolites NaAL, CaAL, CaAS
improves the compressive strength at early ages, concurring to transform a
cement of slow hardening in to a rapid hardening.
The increase in the strength of mortars due to the replacement of zeolites
CaAS, CaAL and silica fume can be attributed to the improved aggregate-matrix

bond associated with formation of a less porous transition zone and a better
interlock between the paste and sand.
Furthermore, the zeolitic materials with calcium ions play an important role in
improving the fine aggregate-paste bond through the formation of more calcium
hydrates as showed by microscope analysis (Fig. 2).
Figure 2:
SEM micrograph of fibrous C-S-H (Calcium Silicate Hydrate) in

the mortars made with zeolite CaA.
3.2 Properties of fresh concrete

The concrete mixes were tested in the fresh state for the workability. The
standard test measured the slump of fresh concrete and the results are
summarized in figure 3.
300
slum p [m m ]
250
200
CEM I
150
CEM II
100
CEM III
50
0
--
Figure 3:
NaAL CaAS CaAL
SF
Workability of fresh concrete prepared with cements with and

without 10% of various admixture.
According to these results, most of the concrete made with zeolites CaAs and
CaAL, had a high slump and were workable. In particular, the fresh concrete
made with zeolites are more workable than that made with silica fume. This is
455
probably caused by the presence of amount of water in zeolitic materials (20

weight % of zeolites).
The concretes made did not show tendencies to segregate or bleed much,
because of the high total cementitious contents, including zeolites or silica fume
and low water to cementitious ratio.
3.3 Properties of hardened concrete
Compressive strength [MPa]
For investigating the strength development of different concrete mixtures, the

results of compressive tests are plotted in Figures 4-6.
Figure 5:
60
50
40
30
20
10
0
10
20
30
Days
Compressive strength of concretes prepared with CEM I with and

without additions: ()CEM I, (y) NaAL, (S) CaAS, (T) CaAL, ()
SF.
Figure 4:
70
70
60
50
40
30
20
10
0
10
20
30
Days
Compressive strength of concretes prepared with CEM II/A-S with

and without additions: ()CEM II/A-S, (y) NaAL, (S) CaAS, (T)
CaAL, () SF.
Figure 6:
70
60
50
40
30
20
10
0
10
20
30
Days
Compressive strength of concrete prepared with CEM III/A-S with

and without additions ()CEM III/A-S, (y) NaAL, (S) CaAS, (T)
CaAL, () SF.
The concrete made with zeolites CEM I and both CAS and CAL exhibits
comparable values of compressive strength at 2, 7 and 28 days. They do not have
relevant differences caused by the size of the particles, probably due to the fact
that at a ratio a/c= 0.35 the pozzolanic effects of zeolites are more important than
the effects of particles packing. Also it is possible to notice at early ages a more
increment of compressive strength for samples prepared with zeolites instead of
silica fume, caused by accelerating effects of zeolites. For the concrete made
with CEM II/A-S the addition of zeolites does not improve significantly the
compressive strength at 28 days, the greater increment is observed by addition of
silica fume. It is probably due to more finesse of CEM II/A-S, that involves a
greater mechanical resistance of the samples without admixtures, in fact at 28
days we have obtained compressive strength a little advanced with the fine
additions as silica fume.
The concretes made with CEM III/A-S and zeolites CaA, exhibit an
increment of resistance with respect to concrete without admixture of about 9 %.
The greater increments are obtained at the early ages, caused by the alkali
content of zeolites that accelerate the process of activation of the fly ash.
As expected, in agreement with the results of mortars, the strength of the
concretes made with zeolite NaA, was less than all the other mixes at all ages
thus reflecting the fact that the sodium content influences negatively the
development of hardening properties of cement concrete. The increase in
strength in systems of concrete containing zeolite CaAS and CaAL can be
explained in a way very similar to the strength increase in mortars mixes. The
aggregate-matrix bond improvement induced by zeolites materials is probably
the result of a combined filler and pozzolanic effect. The filler effect leads to
reduction in porosity of transition zone and provides a dense microstructure and
457
thus increases the strength of the system. The pozzolanic effects help the
formation of bonds between the densely packed particles in the transition zone
through the pozzolanic reaction with the calcium hydroxide liberated during the
hydration of Portland cement to form extra binding calcium silicate hydrates
which lead to further increase in the strength.

Based on the results of this study it can be concluded that some artificially made
zeolites can improve the strength of mortars in a way comparable to the usually
addition of silica fume. The increase in the strength can be attributed to the
improved aggregate-matrix bond resulting from the formation of a less porous
transition zone in the zeolites concrete as silica fume concrete.
The zeolites with calcium ions can be used to produce high strength concrete
in the range 60-80 MPa 28 days compressive strength, with high workability
(slump flow more than 100 mm), using total cementious content of about 400
kg/m3. Therefore, high strength mortars and concrete with 10% calcium zeolite
by weight of cement can be produced and marketed to provide technical and
economical advantages in specific local uses.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
Ortega E.A., Cheeseman C., Knight J., Loizidu M., Properties of alkaliactivated clinoptilolite, Cement Concrete Research 30 (10), pp. 16411646, 2000.
Poon C.S., Lam L., Kou S.C., Lin Z., A study on the hydratation rate of
material zeolite blended cement pastes, Construction and building
materials, 13 (8), pp. 427-432, 1999.
Fragoulis D., Chaniotakis E., Stamatakis M.G., Zeolitic tuffs of Kimolos
Island, Aegean Sea, Greece and their industrial potential, Cement
Concrete Research, 27 (6), 889-905, 1997.
Shi C., Day R.L., Pozzolanic reaction in the presence of chemical
activators: Part II Reaction products and mechanism, Cement Concrete
Research, 30 (4), pp.607-613, 2000.
Chan S. Y.N., Ji X., Comparative study of the initial surface absorption
and chloride diffusion of high performance zeolite, silica fume and PFA
concretes, Cement Concrete Composites, 21 (4), pp. 293-300, (1999).
Taylor H.F.W., Cement Chemistry, Academic Press. London, 1992.
Naiqian F., Properties of zeolitic mineral admixtures concrete, in: S.L.
Sarkar & S. N. Gosh Eds, Mineral Admixtures in cement and concrete.
ABI Books. India pp. 393-447, 1993.
Sersale R., Frigione G., Portland-zeolite-cement for minimizing alkaliaggregate expansion, Cement Concrete Research, 17, pp.404-410, 1987.
Atkins M., Glasser F. P., Jack J.J., Zeolite P in cements: its potential for
immobilizing toxic and radioactive waste species, Waste Management, 15
(2), 127-135, 1995.

[10]
[11]
[12]
[13]
[14]
Fu Y., Ding J., Beaudoin J. J., Zeolite-based additives for high alumina
cement products, Advanced Cement Basic Materials, 3 (1), pp. 3742,1996.
Gervais C., Ouki S. K., Performance study of cementitious systems
containing zeolite and silica fume: effects of four metal nitrates on the
setting time, strength and leaching characteristics, Journal of Hazardous
Materials 93 (2), pp.187-200, 2002.
Mazzasa F., Pozzolana and Pozzolanic Cements in P.C: Hewlett. Editors.
Leas Chemistry of Cement and concrete. Ed. Arnold, pp.241, 1995.
Su N., Fang H.-F., Chen Z H, Liu F.S, Reuse of waste catalysts from
petrochemical industries for cement substitution, Cement Concrete
Research, 30 (11), pp. 1773-1783, 2000.
Ujike I., Properties of concrete containing artificial zeolite made by
chemical conversion of fly ash in R. K. Dhir. M. J. Mccarthy Editors.
Concrete Durability and Repair Technology. Thomas Telford pp.123-131,
1999.
459
Behaviour model and experimental study for

the torsion of reinforced concrete members
C. E. Chalioris
Department of Civil Engineering,
Democritus University of Thrace, Greece
Abstract
The typical behavioural curve of a reinforced concrete element with longitudinal
bars and stirrups under torsion comprises two distinct regions; the elastic part
until the first cracking and the part after cracking. The different character of the
response in these regions reveals the different nature of the load resisting
mechanism in each case. The present work addresses an approach that combines
two different analytical models in order to predict the entire torsional behaviour.
The prediction of the elastic part until the first cracking is achieved using a
smeared crack analysis for plain concrete in torsion, whereas for the description
of the post-cracking response the softened truss model is used. Further, the
results of an experimental investigation on the behaviour of 15 reinforced
concrete beams subjected to pure torsion are also presented in this paper. The
reported results include the behavioural curves and the values of the initial
torsional stiffness, the cracking torque moment and the ultimate torque capacity
of the beams. Analyses for the torsional behaviour of the tested beams using the
proposed approach were performed and analytical curves are produced and
compared with the experimental ones. A good agreement between predicted and
experimental results is observed.
Keywords: beams, reinforced concrete, smeared crack model, softened truss
model, torsion, tests.
Introduction
Several theoretical and experimental researches have emphasized the differences

between the pre-cracking and post-cracking behaviour of reinforced concrete
members in pure torsion [1]. The first, pre-cracking part is characterized by the
doi:10.2495/HPSM06045

high value of torsional stiffness. The element behaves as a homogeneous one and
the influence of reinforcement is of minor importance. The second, after
cracking part is characterized by a further increase of the torque moment at a
lower rate, depending on the volume of the transversal reinforcement. The
consistently decreasing torsional stiffness reveals the different nature of the
bearing mechanism. This justifies the fact that theories for the prediction of the
ultimate torsional strength of reinforced concrete elements generally fail to
describe the elastic response and to predict the torque moment at cracking.
Application of the two well-known space truss models to reinforced concrete
elements under torsion exhibits very reliable results for the prediction of the
post-cracking behaviour and the ultimate torque. The space truss model with
spalling of concrete cover [2] and the space truss model with softening of
concrete [3, 4] are experimentally verified and they can predict the ultimate
torsional strength, the angle of twist, the steel and concrete strains throughout the
post-cracking response of reinforced concrete beams, assuming that the section is
cracked from the beginning. However, deficiency of the space truss models
consists the fact that predictions of the elastic stiffness of analytical torque - twist
curves lies considerable below the test curves in all the examined cases in the
literature [3, 59].
The elastic torsional behaviour till the development of the first cracks of
concrete of a reinforced concrete element is quite similar to a plain concrete
member response. Thus, the ultimate torque moment of a plain concrete element
is approximately equal to the torque moment at cracking of the same element
with longitudinal bars and stirrups. The classical Saint Venant theory to the
torsion problem of plain concrete members, although it properly describes the
elastic behaviour, fails to predict the ultimate torsional strength even in the case
of plain concrete [3, 10].
Recently, a new method for the analysis of plain concrete elements in torsion has
been proposed by Karayannis [10]. This approach is using an efficient numerical
scheme for the torsional analysis of concrete that although initially is based on
the elastic theory, it utilizes a special numerical technique properly modified to
include the smeared cracking approach. Extensive comparisons between
analytical results yielded by this smeared crack analysis and experimental data
derived from a broad range of parametrical studies established the validity of this
analytical model [1113].
This paper reports experimental results of tests on 15 beam specimens
subjected to pure torsion and addresses the combination of two analytical models
for the prediction of the entire torsional behaviour of reinforced concrete
elements. The prediction of the elastic behaviour and the estimation of the torque
moment at cracking are achieved using the smeared crack analysis for plain
concrete in torsion [10], whereas for the description of the post-cracking
response and the calculation of the ultimate torque moment the softened truss
model developed by Hsu and Mo [3, 4] is used. Analyses for the prediction of
the torsional behaviour of the tested beams using the proposed approach were
performed. Predicted torsional behaviour curves are produced and compared
with the experimental ones.
461
Experimental program
The experimental program comprises 15 beams of rectangular cross-section

sorted in five groups (R4, R6, RH4, RH6 and RH8) and tested under pure torsion
action. The cross-section dimensions of the beams of the groups R4 and R6 were
10/20 cm (height to width ratio h/b = 2), whereas the beams of the groups RH4,
RH6 and RH8 had cross-section dimensions 10/30 cm (h/b = 3). Each group
comprises three specimens. The transversal reinforcement used in these three
specimens was 8 mm diameter closed stirrups at a uniform spacing of 200 mm,
150 mm and 100 mm, respectively for each beam.
The longitudinal reinforcement used in the beams of groups R4 and RH4 was
four longitudinal bars of diameter 8 mm (48) located at the corners of the
closed stirrups. The longitudinal reinforcement of beams of groups R6 and RH6
comprised 68; four of them located at the corners of the stirrups and the other
two bars located at the midheight of the closed stirrups (total longitudinal
reinforcement 68 uniformly distributed). Finally, the beams of group RH8 had
eight bars of diameter 8 mm (88); four of them located at the corners of the
stirrups and the other four bars were uniformly distributed at the height of
stirrups sides. Steel yield strength was 518 MPa for the high bond longitudinal
steel bars and 365 MPa for the mild plain steel stirrups. Reinforcement
arrangement for the tested beams is also presented in Table 1.
Table 1:
Group
R4
R6
RH4
RH6
RH8
Reinforcement and concrete strength of tested beams.
Beam code
name
R4-20
R4-15
R4-10
R6-20
R6-15
R6-10
RH4-20
RH4-15
RH4-10
RH6-20
RH6-15
RH6-10
RH8-20
RH8-15
RH8-10
Longitudinal
bars
48
68
48
68
88
Stirrups
8/20 cm
8/15 cm
8/10 cm
8/20 cm
8/15 cm
8/10 cm
8/20 cm
8/15 cm
8/10 cm
8/20 cm
8/15 cm
8/10 cm
8/20 cm
8/15 cm
8/10 cm
fc (MPa)
fsp (MPa)
20.96
2.89
24.59
3.33
26.56
3.04
24.90
3.42
27.39
3.09
The cement used in this experimental work was a locally manufactured

general purpose ordinary Portland type cement. Sand with a high fineness

modulus and coarse aggregates with a maximum size of 9.5 mm were used. The
concrete mixture was made using cement, sand and crushed aggregates in a
proportion 1:2.8:1.2, respectively, and water to cement ratio equal to 0.43. Also
included in Table 1 are the concrete compressive and tensile strength values as
deduced from supplementary compression and splitting tests, respectively.
The total length of the beams was common for all specimens equal to 1.60 m.
Experimental setup and instrumentation are shown in Figure 1. All specimens
were tested monotonically under pure torsion [14]. They were supported on two
roller supports that allowed beam to twist and to elongate longitudinally at both
ends. The load was applied on the ends of two steel arms fixed at the end parts of
each tested beam, through a steel spreader. The end parts of the beams were
properly over-reinforced so that they can bear without cracking the imposed
loading. Cracking and, finally, failure localized at the middle part of each beam.
The load was imposed consistently in low rate and was measured by a load
cell. The average angle of twist per unit length of the tested beams was estimated
using two linear variable differential transducers with high accuracy. These two
LVDTs measured the opposite deformations of each specimen as it rotates.
Initial elastic torsional stiffness, torque moment at cracking and ultimate
torque moment, as measured from the tests, are presented in Table 2. Further, the
entire torsional behaviour of the beams of groups R4, R6, RH4, RH6 and RH8 is
also presented in Figures 2, 3, 4, 5 and 6, respectively, in terms of torque
moment versus angle of twist per unit length experimental curves.
P
load cell
steel spreader beam

LVDT
LVDT
160
0
steel arm
tested beam
roller support
Figure 1:
Test setup.
Behavioural model and comparisons with test results
In this study, for the prediction of the entire torsional behaviour of reinforced
concrete elements, the combination of two different theories is adopted. The
elastic till the first cracking part is described by a smeared crack analysis for
463
plain concrete in torsion [10] and the post-cracking part is described by the wellknown softened truss model [3, 4].
It is justified that for the elastic till the first cracking part the percentage of
steel has a minor effect on the torsional response and reinforced concrete
elements behave as plain concrete members [1]. Therefore, the analytical
smeared crack model for plain concrete in torsion proposed by Karayannis [10]
is applicable to reinforced concrete beams for the prediction of the first elastic
part till the developing of concrete cracking. This approach is using an efficient
numerical scheme for the torsional analysis of concrete that is initially based on
the elastic theory and utilizes a special numerical technique properly modified to
include the smeared cracking approach. The model is based on an analytical
technique that employs constitutive relations expressed in terms of normal stress
and crack width, for the behaviour of the crack process zones. Detailed
derivation of the equations and the solution technique of this theory can be found
in reference 10.
Table 2:
Beam
code
name
R4-20
R4-15
R4-10
R6-20
R6-15
R6-10
RH4-20
RH4-15
RH4-10
RH6-20
RH6-15
RH6-10
RH8-20
RH8-15
RH8-10
Test results and analytical predictions of the tested beams.
Initial stiffness
(10-3 kNcm2/rad)
exp .
exp. calc.
calc.
3771
3084
3229
3452
3557
3442
5913
6689
6186
6494
7215
6755
7744
7191
6658
3346
3624
6500
6292
6596
1.13
0.92
0.97
0.95
0.98
0.95
0.91
1.03
0.95
1.03
1.15
1.07
1.17
1.09
1.01
Cracking torque
(kNcm)
exp .
exp. calc.
calc.
217.05
201.25
200.79
240.12
259.63
264.60
322.00
364.54
418.78
378.70
355.10
372.10
345.27
330.37
368.30
204.65
235.89
337.82
379.97
343.39
1.06
0.98
0.98
1.02
1.10
1.12
0.95
1.08
1.24
1.00
0.93
0.98
1.01
0.96
1.07
Ultimate torque
(kNcm)
exp .
exp. calc.
calc.
238.49
264.87
325.40
287.28
318.38
374.22
394.76
501.27
583.42
481.14
586.91
661.63
503.69
611.95
694.97
263.26
277.52
292.30
288.75
322.81
346.65
439.99
500.89
525.01
444.21
509.88
541.61
462.20
551.14
603.72
0.91
0.95
1.11
0.99
0.99
1.08
0.90
1.00
1.11
1.08
1.15
1.22
1.09
1.11
1.15
Among the various theories available in the literature, researchers have a

common consensus that the space truss models (the spalled and the softened
truss model) are the most rational and powerful models for dealing with torsional
problems. In the present study, the theory of the softened truss model is used [3],
[4]. This method relies on solving three equilibrium and three compatibility
equations along with the constitutive laws of an element taken from a member
subjected to pure torsion. The equations of equilibrium and compatibility are
based on the assumption that the material is continuous. Especially for the

concrete in compression, it is considered that concrete struts strength is greatly
reduced by the diagonal cracking caused by tension in the perpendicular
direction (concrete softening). For further details refer to references 3 and 4.
500
Torque moment (kNcm)
2 8
Smeared crack analysis

Softened truss model
8/s
400
Proposed approach
2 8
300
200
100
s = 20 cm
0
0.0
0.5
s = 15 cm
1.0 1.5 0.0
0.5
1.0
s = 10 cm
1.5 0.0
-3
0.5
1.0
1.5
Angle of twist per length (x 10 rad/cm)

Figure 2:
Experimental and calculated curves of the beams of group R4.
600
2 8
500

8/s
2 8
Proposed approach
2 8
400
300
200
100
0
s = 20 cm
0.0
0.5
1.0 1.5 0.0
s = 15 cm
0.5
1.0
s = 10 cm
1.5 0.0
0.5
1.0
1.5
-3

Figure 3:
Experimental and calculated curves of the beams of group R6.
900
2 8
800
465
700
Proposed approach
8/s
600
2 8
500
400
300
200
s = 20 cm
100
0
0.0
0.5
s = 15 cm
1.0 1.5
0.0
0.5
1.0
s = 10 cm
1.5 0.0
0.5
-3
1.0
1.5

Figure 4:
Experimental and calculated curves of the beams of group RH4.
1000
2 8
900

800
Proposed approach
2 8
8/s
700
600
2 8
500
400
300
200
s = 20 cm
100
0
0.0
0.5
1.0
s = 15 cm
1.5 0.0
0.5
1.0
s = 10 cm
1.5 0.0
-3
0.5
1.0
1.5

Figure 5:

Analyses for the prediction of the torsional behaviour of the tested beams
using the proposed combined approach were performed. Smeared crack analysis
was performed in order to determine the elastic response and the torque moment
at cracking, whereas the softened truss model was performed for the prediction
of the post-cracking response and the ultimate torque moment. The predicted
values of the initial torsional stiffness, the torque moment at cracking and the
ultimate torque moment are presented and compared with the measured ones in
Table 2. Furthermore, analytical torque curves for the torsional behaviour of the
beams of groups R4, R6, RH4, RH6 and RH8 are presented and compared with
the experimental ones in Figures 2, 3, 4, 5 and 6, respectively.
1000
2 8
900
800
2 8
700
2 8
600
2 8

8/s
Proposed approach
500
400
300
200
s = 20 cm
100
0
0.0
0.5 1.0 1.5 0.0
s = 15 cm
0.5
1.0
s = 10 cm
1.5 0.0
-3
0.5
1.0
1.5

Figure 6:
Conclusions
The following concluding remarks are drawn from the results reported herein.
Volume of the transversal and the longitudinal reinforcement significantly
affects the torsional behaviour of the tested beams, as it was expected. Two
distinct regions can be observed in a typical experimental torque twist curve of
the tested reinforced concrete beams. The different character of the response in
these regions reveals the different nature of the load resisting mechanism in each
case. In order to describe the entire torsional behaviour of the tested beams and
based on the above observation, the combination of two different theories is
adopted. For the estimation of the elastic till the first cracking response, a
smeared crack analysis for plain concrete in torsion is used, whereas the
467
prediction of the post-cracking behaviour is achieved using the softened truss

model. Comparisons between the predicted torsional behaviour and the measured
one showed that this combined approach yields realistic torsional curves for the
entire response of the element and is capable to predict with satisfactory
accuracy the initial torsional stiffness, the torque moment at cracking and the
ultimate torsional strength.
References
[1]
Hsu, T.C., Unified Theory of Reinforced Concrete, CRC Press, Inc., Boca
Raton, Fla., 1993.
[2]
Mitchell, D. & Collins, M.P., Diagonal Compression Field Theory A
Rational Model for Structural Concrete in Pure Torsion, ACI Proceedings,
71(8), pp. 396-408, 1974
[3]
Hsu, T.C. & Mo, Y.L., Softening of Concrete in Torsional Members
Theory and Tests, ACI Proceedings, 82(3), pp. 290-303, 1985.
[4]
Hsu, T.C., Toward a Unified Nomenclature for Reinforced-Concrete
Theory, Structural Engineering, ASCE, 122(3), pp. 275-283, 1996.
[5]
Mansur, M.A., Nagataki, S., Lee, S.H. & Oosumimoto, Y., Torsional
Response of Reinforced Fibrous Concrete Beams, ACI Structural, 86(1),
pp. 36-44, 1989.
[6]
Wafa, F.F., Shihata, S.A., Ashour, S.A. & Akhtaruzzaman, A.A.,
Prestressed High-Strength Concrete Beams under Torsion, Structural
Engineering, ASCE, 121(9), pp. 1280-1286, 1995.
[7]
Rahal, K.N. & Collins, M.P., Simple Model for Predicting Torsional
Strength of Reinforced and Prestressed Concrete Sections, ACI Structural,
93(6), pp. 658-666, 1996.
[8]
Ashour, S.A., Samman, T.A. & Radain, T.A., Torsional Behavior of
Reinforced High-Strength Concrete Deep Beams, ACI Structural, 96(6),
pp. 1049-1058, 1999.
[9]
Chalioris, C.E., Study of the Behaviour and the Failure Mechanisms of
Plain and Reinforced Concrete Elements in Torsion, PhD dissertation,
Department of Civil Engineering, Democritus University of Thrace,
Xanthi, Greece, 1999.
[10] Karayannis, C.G., Smeared Crack Analysis for Plain Concrete in Torsion,
Structural Engineering, ASCE, 126(6), pp. 638-645, 2000.
[11] Karayannis, C.G., Nonlinear Analysis and Tests of Steel-fiber Concrete
Beams in Torsion, Structural Engineering and Mechanics, 9(4),
pp. 323-338, 2000.
[12] Karayannis, C.G. & Chalioris, C.E., Experimental Validation of Smeared
Analysis for Plain Concrete in Torsion, Structural Engineering, ASCE,
126(6), pp. 646-653, 2000.
[13] Karayannis, C.G. & Chalioris, C.E., Strength of Prestressed Concrete
Beams in Torsion, Structural Engineering and Mechanics, 10(2),
165-180, 2000.

[14]
Chalioris, C.E., Cracking and Ultimate Torque Capacity of Reinforced

Concrete Beams, Proc. of the Int. Symposia Celebrating Concrete: People
and Practice, Vol. Role of Concrete Bridges in Sustainable Development,
University of Dundee, Scotland, UK, pp. 109-118, 2003.
469
A study on use of blended ferrocement:

a high performance material for
repair/strengthening of brick masonry columns
T. Kibriya
National University of Sciences and Technology, Risalpur, Pakistan
Abstract
The use of ferrocement, containing blended cement, to repair and enhance the
strength of brick masonry columns of old buildings and structures showing
distress has been investigated for use as a rehabilitation/strengthening material.
Intensive experimental study has been carried out on plain brick masonry
columns and brick masonry columns with single, double and triple layers of
ferrocement, with/without the use of chemical bonding agents. Use of blended
cements containing agricultural waste i.e. rice husk ash have not as yet been
investigated at all for use in ferrocement. Blended cement containing 75% OPC
and 25% rice husk ash has now been investigated for use in ferrocement.
Ferrocement containing blended cement has been observed to substantially
increase the load carrying capacity of brick masonry columns along with
decreasing the strains by about 50%. Strength improvements of up to 160% in
the load carrying capacity of columns with single layered ferrocement have been
observed whilst about 100% strength improvements have been observed in
columns with 2/3 layers of ferrocement applications. The slope of the
stress-strain curve resembles typical curves characteristic of high strengths. The
results of this study have been highly encouraging and the use of
ferrocement with blended cement is recommended for the repair and
rehabilitation/strengthening of brick masonry columns. The use of rice husk ash
for blending with ordinary Portland cement produces a stronger, impermeable
and durable material in addition to ease of in-situ application, reducing the
disposal problem of this agricultural waste and reducing the costs of repairs.
Keywords: strengthening, rehabilitation, brick masonry columns, blended
cement, old buildings, retrofit material.
doi:10.2495/HPSM06046
Introduction
Many old buildings along with few bearing historical architecture of past
civilisations were constructed by using brick masonry. Most of these old
buildings/structures have shown signs of distress/damage due to aging and action
of various offensive agencies. To maintain the safety requirements, most such
structures now require replacement/ rehabilitation/ strengthening. Strengthening
of existing columns may also be required for expansion works in the existing
buildings/structures. With the trend of conserving the heritage and specific
architecture/ faade along with avoiding environmental/disposal problems of
demolition waste, stress is laid on rehabilitation/ strengthening of the existing
elevations/structures. Resource conservation and paucity of development funds
in developing/under developed countries suggests economical options of
rehabilitation/ strengthening of existing structures. To maintain economy and
environment, use of waste materials in concrete have been investigated widely.
Lot of work has been done on use of waste brick aggregates (widely available
from old dilapidated buildings in the South Asian region) in normal/high strength
concrete by Kibriya [14] and use of agricultural waste in blended cement and
normal/high strength concrete Kibriya [58]. Ferrocement carries certain
advantages in rehabilitation/strengthening works like in-situ application, ease of
construction, use of locally available materials, no formworks, no special
skills/training required, easily moulded in any shape, very low permeability,
easily repairable, resource saving and durability vide Kibriya [9, 10], Paul and
Pama [11], ACI [12] and BOSTID [13]. The repair and strengthening of existing
structures has been a more challenging problem for civil engineers than
construction of new structures.
Use of blended cements containing agricultural waste i.e. rice husk ash have
not at all been investigated as yet for use in ferrocement. Higher content of silica
found in rice husk ashes along with low calcium oxide content is advantageous
as far as its cementing properties are concerned. Blended cements containing rice
husk ash as a partial replacement material have been found to improve strength
and durability characteristics substantially when used as cementitious material
and in high strength concretes for various purposes. Blended cement containing
75% OPC and 25% rice husk ash has now been investigated for use in
ferrocement. Selection of 25% replacement of OPC with rice husk ash has been
on the basis of excellent performance of such blended cements and concrete in
the recent studies by Kibriya [58].
2 Research methodology
This experimental investigation was carried out on nine different groups of
columns. Each group consisted of three specimen 225x225x750mm. The groups
were labeled in alphabetical order and the individual specimens were thus
numbered as 1, 2 and 3 and so on. 1st Group i.e Group A comprised of plain
brick masonry in 1:4 cement mortar and were considered as control specimen.
Group B comprised of brick masonry columns in 1:4 cement mortar with
471
12.5mm thick, 1:4 cement mortar plaster. Group C were brick masonry columns
in 1:4 cement mortar, simply plastered with 1:2 cement mortar. Groups D, E and
F comprised of brick masonry columns in 1:4 cement mortar with one, two and
three layers of ferrocement applied respectively. Each layer of ferrocement
applied was 15mm thick applied in two steps i.e. a layer of 7mm 1:2 cement
mortar, application of wire mesh and another 8mm layer of 1:2 cement mortar.
W/C ratio of mortar was maintained at 0.3. Blended cement containing 25% rice
husk ash and 75% ordinary Portland cement was used. Rice husk ash used for
blending was obtained by burning rice husk in an industrial furnace with
temperatures maintained around 500 to 6000C. Ashes were then cooled to 200C
and subsequently ground in a laboratory ball mill for about 120 minutes and
thereafter sieved through a #325 sieve. Ashes passing 100% through the sieve
were then used for blending with ordinary Portland cement. Sand with medium
grading was used for mortar. Upto 1% high range water reducing admixture was
used for preparing workable mortar mix. Wire mesh of expanded metal having
openings of 15mm and wire thickness of 1mm was applied with yield strength of
227 N/mm2 and ultimate strength of 379 N/mm2. 1mm thick, 2.5cm long nails
were used to bind wire mesh layers to the core. Groups G, H, and I comprised of
specimen similar to groups D, E and F with the exception that a chemical
bonding agent Sika-Latex was used in addition to nails to bind the ferrocement
layer to the core. The specimen were moist cured for 28 days at a temperature
around 200C before testing. The details are shown in Table 1.
Testing procedure
The column specimen were tested for compression using universal compression
testing machine with a 255x255mm, 5mm thick plate on top and bottom. The
observations were made for stress-strain behavior, crack pattern, cracking and
failure loads. The measurements for strain were done with the help of two
compressometers attached to the specimen. The results of the tests are tabulated
in Table 2. Permeability testing was carried out by the help of capillary rise test
according to RILEM CPC13 and Initial Surface Absorption (ISAT) tests
according to BS 1881. In addition, 50mmx50mm cubes were caste, moist cured
for 30 days and immersed in water for 30 days to observe any weight changes.
Analysis of test results
The ultimate failure and cracking load capacities of the specimen are given in
Table 2. The stress-strain curves and trend lines based on the average readings of
the specimen groups are shown in Figures 1 and 2.
4.1 Brick masonry columns (group A)
The failure loads were generally identical. The average failure load for case A
was 142kN. The crack initiation started at about 70% of the failure loads. Major
cracks were vertically oriented, however, a few horizontal cracks were also

observed. The crack widening and propagation was quite fast after their first
appearance. The stress - strain behaviour of this group is shown in Figs 1and 2.
Table 1:
Type of Specimen
Details of specimen.
Group
Designation
A
B
No of
Layers
_
_
Bonding
Agent
_
_
Mortar
Ratio
_
1:4
Masonry Col
Masonry Col With
Plaster
Masonry Col with C
_
_
1:2
Plaster
Masonry Col + D
1
Nails
1:2
Ferro cement
Masonry Col +
E
2
Nails
1:2
Ferro cement
Masonry Col +
F
3
Nails
1:2
Ferro cement
Masonry Col +
G
1
Nails &
1:2
Ferro cement
S. Latex
Masonry Col +
H
2
Nails &
1:2
Ferro cement
S .Latex
Masonry Col +
I
3
Nails &
1:2
Ferro cement
S. Latex
Note: Brick Masonry in 1:4 Cement Mortar.
One layer of ferrocement mortar 15mm thick with 1:2 cement mortar.
Wire mesh with 1mm wire, 15mm mesh.
Cementitious material 75% OPC + 25% Rice Husk Ash
W/C ratio 0.3 of ferrocement.
Moist curing - 28 days
4.2 Brick masonry columns with 1:5 cement mortar plaster (group B)
The results of all specimen of group B clearly suggest that simple mortar
application can increase the failure loads by up to 35%. However, it seems to
have no effect on crack initiation and its growth as the crack growth was rapid
as in case of group A. The cracks generally started at about 70% of failure load.
The average failure load was 195kN as given in Table 2. The strains also reduced
somewhat, as compared to plain brick masonry columns as shown in Figure 1.
4.3 Brick masonry columns with 1:2 cement mortar plaster (group C)
The specimen of group C exhibited fairly good results in terms of failure loads.
The average failure load was calculated as 208kN. This clearly shows an
increase of about 46% over the failure load of the control specimen of Case A
473
and about 10% increase over columns plastered with 1:5 mortar ratio. The stressstrain curve is fairly straight with reduced strains as shown in Figure 2.
Table 2:
Type of Specimen
Group
Masonry Col
Masonry
With Plaster
Col
Masonry Col with

Plaster
Masonry Col +
Ferro cement
(1 Layer)
Masonry Col +
Ferro cement
(2 Layers)
Masonry Col +
Ferro cement
(3 Layers)
Masonry Col +
Ferrocement(1)
Bonding Agent
Masonry Col +
Ferrocement(2)
Bonding Agent
Masonry Col +
Ferro cement(3)
Bonding Agent
D
E
F
G
H
I
Summary of test results.

First
cracking
Load (kN)
86.19
113.00
104.10
146.40
124.60
157.10
142.00
128.00
146.60
187.30
196.30
212.20
157.30
186.00
175.80
202.10
195.80
186.40
204.70
192.10
224.20
162.50
197.60
166.26
142.90
189.70
178.80
Failure
Load
(kN)
138.4
141.2
146.8
192.1
198.3
196.4
211.1
208.4
204.7
340.9
394.2
367.2
299.4
268.6
279.5
310.4
301.5
270.8
354.8
396.2
377.1
280.4
359.7
260.9
281.7
320.1
318.4
Mean
Failure
Load (kN)
%
Increase
142.13
Control
195.6
37.6
208.1
46.5
367.4
158.7
282.5
98.9
294.3
107.3
376
164.8
300.3
111.5
306.7
116
4.4 Brick masonry columns with single layer of ferrocement (group D)

The specimen of this group having one layer of ferrocement showed a marked
difference in results by increasing the failure load to 367kN, an increase of 158%
over the control. The crack propagation was not rapid and widening of cracks
occurred gradually thus showing a ductile failure. The increased failure load and
mode of failure clearly suggests usefulness of ferrocement in strength
enhancement of brick masonry columns. Failure strains were observed to be
around 60% of the strains in the control.

A
1000
900
800
700
STRESS
600
500
400
300
200
100
0
0
0.0005
0.001
0.0015
0.002
0.0025
0.003
STRAIN
Figure 1:
Average stress-strain curves of all groups.
4.5 Brick masonry columns with double/triple layer of ferrocement (groups

E and F)
The specimen of groups E and F treated with ferrocement having 2 and 3 layers
respectively of wire meshes did not give results better than group D having one
layer of ferrocement. The ultimate loads were 282kN and 294kN respectively for
2 and 3 layers of ferrocement, an increase of about 100% in the load carrying
capacity of columns as compared to 158% increase due to a single layer of
ferrocement application. The absence of higher contribution by ferrocement with
more meshes was due to budging and early bond failure between the mesh layers
and smaller size of top and bottom steel plates (255mm) than the over all cross
section of the samples having multi-layers of wire meshes. The stress-strain
curve is shown in Figures 1 and 2.
475
1200
1000
STRESS
800
600
400
200
0
0
0.0005
0.001
0.0015
STRAIN
D
0.002
0.0025
0.003
Linear (A)
Linear (B)
Linear (C)
Linear (D)
Linear (E)
Linear (E)
Linear (F)
Linear (G)
Linear (H)
Linear (I)
Figure 2:
Trend line - stress-strain curves.
4.6 Brick masonry columns with single layer of ferrocement and bonding
chemical applied (group G)
In the course of samples preparation of this group, a bonding agent, Sika-Latex
was used in addition to nailing the mesh layer. The average failure load 376kN
of this group although significantly higher i.e. 164% than the control specimen,
did not show much improvement over its corresponding group D of samples
having one layer of mesh nailed to the core, without any bonding agent.
However, the Sika-Latex bonding was effective in controlling the spalling of
outer mortar layer on failure. The failure strains were almost 10% lower than the
specimen with single layer of ferrocement without any bonding agent. The
propagation of cracks upto failure was gradual.

4.7 Brick masonry columns with double/triple layer of ferrocement and
bonding chemical applied (groups H and I)
The columns with 2/3 layers of ferrocement application failed at 300kN and
306kN respectively showing 111% and 116% strength improvements over the
control specimen. The specimen of these groups did not show much improved
results over similar groups E and F specimen having same number of meshes but
without application of bonding agent. Sika-Latex agent was not effective in
maintaining bond between the mesh layers. The stress-strain diagrams are shown
in Figures 1 and 2.
4.8 Permeability testing on ferrocement specimen
The capillary rise test indicated that ferrocement used was watertight. The w/c
ratio of ferrocement mortar used was 0.3 which theoretically should also be
watertight. Same results were confirmed by ISAT. Due to impermeability of
ferrocement, it is expected to be highly durable in offensive environments.
Weight recordings on 50mm cubes kept in water for 30 days indicated constant
weight before and after the soaking period indicating the ferrocement was
watertight and impermeable.
Discussion of results
From the summary of these results it can be clearly observed that the load
carrying capacity of ferrocement encased brick masonry columns increased
significantly. It is evident that a significant increase in failure load capacity can
be achieved using only one layer of ferrocement. Failure strains were observed to
be reduced by about 50% on application of single layer of ferrocement.
Additional layers (2 or 3) of ferrocement did not contribute much due to bond
failure between them and while bulging outwards these layers pulled the inner
mesh along and thus affecting the results. The contribution of Sika-Latex
bonding agent was not much except for controlling the spalling of mortar at
failure. The permeability of ferrocement layer is negligibly low thereby
increasing the durability of the ferrocement applied columns.
It has been observed that ferrocement containing blended cements has
performed even better than ferrocement with ordinary Portland cement on
comparing results of recent studies. Strength improvement of about 30% along
with complete watertightness in ferrocement with blended cement have been
observed. The results of this study have been highly encouraging and use of
ferrocement, with blended cement containing 25% rice husk ash, for repair and
rehabilitation/strengthening of brick masonry columns has been recommended
for practical use in the field. Use of rice husk ash for blending with ordinary
Portland cement produces a stronger and durable material in addition to ease of
in-situ application, reducing the disposal problem of this agricultural waste and
reducing the costs of repairs.
477
Conclusions
From the above study the following conclusions are drawn:

The ferrocement (using blended cement) treatment enhances the load
carrying capacity of brick masonry columns by as much as 150% to
160%. This increase can be economically achieved with one layer of wire
mesh used in ferrocement nailed to core and with or without application of
any chemical bonding agent. There were almost 50% reduction in strains
at failure and much improved modulus was observed.
Ferro cement having 2 and 3 layers of wire meshes are uneconomical and
did not appreciably improve the load carrying capacity as noted in this
experimental study.
The initial resistance to cracking and thereafter crack growth mechanism

of brick masonry columns improved quite significantly due to ferrocement
treatment.
The permeability of ferrocement layer is negligibly low thereby increasing

the durability of the ferrocement applied columns.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
Kibriya. T., Crushed burnt clay bricks as aggregates for histrength

concrete, Proceedings of McMat 2005, Joint ASCE/ASME/SES
Conference on Mechanics and Materials, Baton Rouge, Louisiana, USA,
2005.
Kibriya, T., Properties of Concrete with crushed brick aggregates,
Dissertation submitted for partial fulfillment of Ph.D requirement, City
University, London, U.K. 1991.
Kibriya, T. and Speare, P.R.S., The use of crushed brick coarse aggregates
in concrete, International Congress on Concrete in the Service of
Mankind, Dundee, Scotland, pp 204 207, 1996.
Kibriya. T., Durability of concrete with crushed bricks coarse aggregates,
8th Islamic Countries Conference on Statistical Sciences, University of
Bahrain, Bahrain, pp 161 164, 2002.
Kibriya. T. and Baig. N., High performance concrete with blended
cements using agrowastes, Proceedings of McMat 2005, Joint
ASCE/ASME/SES Conference on Mechanics and Materials, Baton
Rouge, Louisiana, USA, 2005.
Kibriya. T. and Khan. S., High performance concrete pavements using
agrowaste blended cement, Proceedings of McMat 2005, Joint
ASCE/ASME/SES Conference on Mechanics and Materials, Baton
Rouge, Louisiana, USA, 2005.
Kibriya. T. and Khan. S., Performance of ecologically friendly waste in
high strength concrete, IV Regional Conference on Civil Engineering
Technology, Joint ASCE/ESIE Conference, Cairo, Egypt, 2005.
Kibriya. T. and Baig. N., Agricultural wastes in construction, Blended
cements containing rice straw ash, IV Regional Conference on Civil
[9]
[10]
[11]
[12]
[13]
Engineering Technology, Joint ASCE/ESIE Conference, Cairo, Egypt,

2005.
Kibriya. T., Ferrocement as retrofit material for old structures,
Proceedings of McMat 2005, Joint ASCE/ASME/SES Conference on
Mechanics and Materials, Baton Rouge, Louisiana, USA, 2005.
Kibriya. T., Ferrocement An innovative material for
rehabilitation/strengthening of brick masonry columns, IV Regional
Conference on Civil Engineering Technology, Joint ASCE/ESIE
Conference, Cairo, Egypt, 2005.
Paul. B.K. and Pama. R.P., Ferrocement, 1989.
ACI 549, Ferrocement, American Concrete Institute, 1988.
National Academy of Sciences, Ferrocement - Application in Developing
Countries, A report of an Adhoc Panel of the advisory Committee on
technological innovation, BOSTID, Washington, D.C., 1973.
Section 6
Damage and
fracture mechanics
481
Crush behaviour of open cellular lattice

structures manufactured using
selective laser melting
M. Santorinaios, W. Brooks, C. J. Sutcliffe & R. A. W. Mines
Department of Engineering, University of Liverpool, England, UK
Abstract
Open cellular lattice structures, made out of stainless steel, have been
manufactured using the micro fabrication procedure of selective laser melting. A
single geometry has been considered, namely vertical struts with cross bracing
with a cell geometry of a cubic form. Blocks of this material, with dimensions of
25 mm cube, have been tested under compression and shear loading. Three cell
sizes have been considered, namely 1.25, 2.5 and 5 mm. Uniaxial compression
and shear force versus displacement and failure characteristics of these materials
are discussed, In this way, the micro mechanical characteristics of the material
and manufacturing parameters are related to bulk properties.
Keywords: cellular, micro fabrication, selective laser melting, mechanical
properties.
Introduction
Foams have been used for a number of years as core materials in sandwich
construction for transport and general engineering systems. There is a wide
variety of foam ranging from low density, high end technology applications, e.g.
the Polymethacrylimide foam Rohacell 51WF for aerospace use [1], to high
density/low end technology applications, e.g. the aluminium foam Alporas for
civil engineering use [2]. The usual design parameters for sandwich construction
focus on elastic response (specific stiffness and strength) [3], but more
specialised behaviours include fatigue behaviour, crashworthiness and foreign
object impact performance [3, 4]. The mechanical behaviour of sandwich
structures is dependent on the mechanical properties of the foam core and
composite skin. As far as the foam core is concerned, the core can be subject to
doi:10.2495/HPSM06047

multi axial stresses and progressive failure in applications such as foreign object
impact loading of sandwich panels [4].
The properties of foam cores are dependent on the parent material and the cell
geometry [5]. Hence, for a given parent material, properties are adjusted by
varying the foam density. This means that foams tend to be quasi isotropic, and
that they tend to have a restricted range of properties. It should also be noted
that low end foams have highly variable microstructure, which means that
sophisticated modelling of mechanical behaviour is negated by the variation in
bulk properties [4].
There is currently large interest in creating cellular structures with tailored
cellular properties. These can be closed cell, open cell and framework type
geometries. A number of manufacturing procedures are possible. This includes
investment casting, deformation forming, brazing etc. [6]. Limitations of such
techniques include the fineness of the structure, and the actual cell geometry.
From the theoretical point of view, research is currently underway that addresses
the analysis and optimisation of lattice structures in the context of core materials
in sandwich construction [7].
Rapid prototyping techniques have been developed over the years to create
complex geometries [8]. Selective laser melting is a new technique which allows
the creation of metal structures with fine detail [9]. The technique that has been
developed has the potential to create open cellular lattice structures with features
with a resolution of 50 micro meters [10].
Selective laser melting of open cellular lattice structures
A schematic of the selective laser melting process is shown in Figure 1. The

metallic powder is levelled off using a plate. A pre-programmed laser then
traverses the powder surface, selectively melting areas of the surface. A new
layer of powder is then deposited, and the process is repeated. In this way, a
structure can be created in 50 micrometer layers.
Figure 1:
The Selective Laser Melting manufacturing process [9].
483
There are a large number of possible geometries possible, which gives large
design freedom. The laser spot has a size of 50 micrometers, and any structural
feature can be built to this resolution, as long as the newly solidified material can
be built on top of the previous (solidified) layer. Figure 2 shows two cellular
structures that can be created. The geometry is based on standard RP support
geometries, but has features typical of general open cellular lattice structures.
The structure consists of vertical members with diagonal cross members only, as
currently, the process cannot build horizontal members. The powder used was
Stainless steel 316L from Sandvik [11], and the powder grain size was between
10 and 38 micrometers. The finer the powder, the more expensive and dangerous
(from combustion) it is.
Figure 2:
Builds for cell sizes of 2.5mm and 1.25mm.
Figure 2 shows the cellular structures with 1.25mm and 2.5mm cell sizes. A
5mm cell size was also considered, but this proved to be problematic to make.
The struts tended to sag during build, which meant that the final quality was
unsatisfactory. In a single manufacture, six blocks were built on a steel plate
base. The manufacture time for these six blocks was about 4 hours. The time of
the process is dependent on the size and complexity of the cellular structure.
There are about 500 laser passes to create the height of blocks. The blocks were
cut from the steel bottom plate using a sharp knife, in an effort to reduce damage
to cells. An important issue is the size and quality of the struts. This has been
discussed in detail in Brocks et al. [10]. Two laser exposure times and two laser
powers were used (see Table 1).
Table 1:
Summary of builds (X number of specimen, 1 to 5).
Build ID
Cell size
(mm)
5
2.5
2.5
1.25
1-X-5
2L-X-2.5
2H-X-2.5
3-X-1.25
Rel. Dens.
(%)
0.8
4.1
5.4
12.5
Laser Exp
(ms)
2.5
2.5
3.0
3.0
Las. Pwr
(A)
2.25
2.25
2.60
2.60

Figure 3 shows the detailed quality of the structure, in the form of an actual
picture and in the form of a shadowgraph. The laser power was 2.6A and the
laser exposure time was 3 milliseconds. It can be seen that the struts are by no
means cylindrical, and that there is the possibility of local imperfections. A
feature of selective laser melting is molten material balling, which gives rise to
beads being formed [12]. This shows the importance of careful control of
manufacturing parameters. The strut diameter is on average 100 micrometers
[10].
Figure 3:
Photo and shadow graph of detail of 2.5mm cell sized specimen.
Compression tests and results
25 mm cube specimens were crushed along the cell vertical direction in a 50kN
Instron testing machine at a displacement rate of 5mm/min. The specimens were
crushed between two steel platens and the stress and strain data was derived from
load cell and crosshead data. Engineering stress and strain measures are used.
Figure 4 gives stress v strain data for the three cell sizes, with stress being
plotted on a log scale, for clarity. Figure 5 gives the failure mode for the 2.5mm
cube cell size case. It can be seen that the progressive crush is stable and that the
major failure mode results from elastic-plastic buckling of the vertical struts.
This gives rise to a shear band forming in the complete specimen. Cells also
fail near the base, which shows the need to improve the test to ensure failure in
the bulk material. Zhou et al. [13] give an analysis of plastic buckling of struts,
and show that the critical buckling load, Pcr is given by:
Pcr =
12C
l
(1)
485
where C is a constant relating the plastic hinge couple of the strut, and l is the
length of the strut. Hence the critical failure load should be inversely
proportional to the strut length. Figure 6 checks out this hypothesis. Equation (1)
is calibrated from the 2.5mm cell data. The effect of the change of the number of
vertical struts as a function of cell size is taken into account. The general trend is
correct, showing that compression failure for this block and loading
configuration is governed by strut length, strut cross section and parent material
property.
Figure 7 gives the effect of manufacturing parameters. It compares the
average crush stresses from the 2H tests with two 2L tests. For details of
specimens, see Table 1. It can be seen that manufacturing parameters are very
important, and it should be noted that these need to be optimised before
optimising the cell geometry.
It would be interesting to study the crush of the block in the lateral direction.
It is proposed that, crush strength would be reduced, as there are no struts in that
direction. Therefore this cellular material will be highly anisotropic.
a
b
c
Figure 4:
Stress v strain for three cell sizes (a. 1.25mm, b. 2.5mm, c. 5mm).
Figure 5:
Failure of 2.5mm cell size sample (H-2-2.5).

8.00E+06
Crush Stress (N)
7.00E+06
6.00E+06
5.00E+06
4.00E+06
3.00E+06
Eqn.1
2.00E+06
1.00E+06
0.00E+00
0
10
Cell size (mm)
Figure 6:
Predicted (from Equation (1)) and actual failure loads.
2H
Figure 7:
Comparison of the effect of manufacturing parameters on

compression stress strain behaviour (2.5mm cell size).
Shear tests and results
Foam cores in sandwich structures are subject to multi axial stresses, and so
these have to be studied [4]. A tensile Arcan test [14] has been developed, in
which combinations of tensile direct and shear stress can be applied to a block of
foam. In this paper, shear only will be described. The 25mm cube blocks are
glued onto steel plates, which are then mounted in the Arcan Rig [14]. Figure 8
gives shear stress v strain data, and Figure 9 gives the failure modes for 1.25mm
and 2.5mm cell sizes.
487
3-2
2H
Figure 8:
Figure 9:
Stress strain data from Arcan shear tests.
Comparison in failure modes for shear loading (2.5 and 1.5mm cell
sizes).
In both cases, failure initiates at the boundaries, showing the need to develop
the test further, to ensure the initiation of failure for the bulk material. In the
2.5mm cell size case, failure initiates by the buckling of the diagonal struts,
whereas in the 1.25mm cell size case, the whole block rotates, and failure is by
tensile rupture of the vertical struts. The latter failure mode gives a more abrupt
reduction in stress after failure. It would be interesting to conduct both biaxial
and hydrostatic tests, as these are important for localised loadings in sandwich
structures [4,15]. For the 2.5mm cell size, the shear failure stress is about 6MPa
which should be compared to the compression value of 10MPa. This would
indicate a failure surface typical of foams [15]. However, more unusual cellular
structures may well give more unconventional failure surfaces.
General discussion
Figure 10 compares the bulk performance of the current cellular structure with
other foams [16]. It can be seen that the structure performs well, and it should be

noted that no attempt has been made to optimise the cell geometry. Therefore, it
can be concluded that there is potential to markedly improve mechanical
properties, over and above those of foams.
Figure 10:
Comparison of current cellular structure with envelope for foams

[16] (A 1.25mm cell size, B 2.5mm cell size).
To summarise, there is scope to optimise the current cellular geometry. This

needs to be addressed from a fundamental point of view basically the
manufacturing process can create any cellular structure with a resolution in
structural features of about 50 micrometers.
The topology of the open cellular lattice structure can be optimised for elastic
stiffness in bulk properties, and for non-linear (collapse) behaviour. A promising
approach has been discussed by Bendsoe and Sigmund [17], in which a two
dimensional solid is defined, and is populated by blocks (20 by 40, for example).
Boundary conditions are then applied to the geometry, and a linear finite element
analysis is carried out. Blocks that are least structurally effective are then
deleted, giving rise to the most efficient overall structure. Simple cases of beam
bending have been analysed using MATLAB [17] and the approach can be
extended to three dimensions. The optimisation approach has been applied to
elastic stiffness problems [17], buckling problems [18] and crashworthiness
problems [19].
Concluding remarks
A rapid prototyping manufacturing method is being developed to realise open

cellular lattice structures that can be tailored for specific structural applications.
A wide variety of structural geometries are possible. The synthesis of optimum
structures (for a given application) is theoretically complex, and design tools
need to be developed to maximise the potential of the selective laser melting
manufacturing process. The analysis method of topology optimisation shows
good potential for quantifying optimum performance.
A related issue concerns the simulation of the mechanical behaviour of these
types of materials in the context of the actual structure. Periodic structures with
489
small cell size give rise to a large number of structural features, which leads to
unrealistic numerical models for full micro mechanical modelling. Also, in the
case of foreign object impact loading, material behaviour is dynamic and highly
non linear with damage occurring. There is a need to develop homogenised
models, that characterise the effective behaviour of periodic structures [17].
Acknowledgements
The SLM machine has been funded by the EPSRC Grant number GR/598405/01
with contributions from MCP and Sandvik Osprey.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
Rohacell 51 WF Foam, Roehm Ltd., Bradbourne Drive, Tilbrook, Milton

Keynes, Bucks, MK7 8AU, England, 1999.
Alporas Data Sheet, Shinko Wire Company, Amagasaki, Japan, 2001.
Zenkert, D.A., Introduction to Sandwich Construction, EMAS Publishers,
2000.
McKown, S and Mines, R.A.W., Impact behaviour of metal foam cored
sandwich beams, European Conference on Fracture 16, Alexandropoulos,
Greece, Paper No. 254, July 2006.
Gibson, L.J., and Ashby, M.F., Cellular Solids, Cambridge University
Press, 1999.
Wadley, H.N.G., Fleck N.A., Evans, A.G., Fabrication and structural
performance of periodic cellular metal sandwich structures, Composites
Science and Technology, 2004.
Liu, J.S., Lu, T.J., Multi objective and multi loading optimisation of ultra
lightweight truss materials, International Journal of Solids and Structures,
Vol. 41, pp619-635, 2004.
Chau, C.K., Leong, K.F., and Lim, C.S., Rapid Prototyping (2nd Edition),
World Scientific, 2004.
MCP Realiser II, MCP Tooling Technologies Ltd., Whitebridge Way,
Whitebridge Ind. Park, Stone, Staffordshire, ST15 8LQ, England, 2005.
Brooks, W.K., Tsopanos, S., Stamp, R., Sutcliffe, C.J., Cantwell, W.J.,
Fox. P., Todd, J., 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.
Data sheet for microfine powders, 316L Austenitic stainless steel, Sandvik
Osprey, Neath, Scotland, 2005.
Kruth, J.P., Froyen L., Van Vaerenbergh, J., Mercelis, P., Rombouts, M.,
Lauwers, B., Selective melting of iron based powders, Journal of
Materials Processing Technology, No. 149, pp616-622, 2004.
Zhou, J., Shrotriya, P., Soboyejo, W.O., On the deformation of aluminium
lattice block structures: from struts to structures, Mechanics of Materials,
Vol. 36, pp723-737, 2004.

[14]
[15]
[16]
[17]
[18]
[19]
McKown, S., Mines, R.A.W., Measurement of material properties for

metal foam cored polymer composite sandwich construction, Journal of
Applied Mechanics and Materials, Vol. 1-2, pp211-216, 2004.
Li, Q.M., Mines, R.A.W, and Birch, R.S., The crush behaviour of
Rohacell 51WF structural foams, International Journal of Solids and
Structures, Vol. 37, pp 6321-6341, 2000.
Ashby, M.F., Materials selection in engineering design, Butterworth
Heinemann, Oxford, 2005.
Bendsoe, M.P. and Sigmund, O., Topology Optimization: Theory,
Methods and Applications, Springer, Berlin, 2004.
Neves, M.M., Sigmund, O., Bendsoe, M.P., Topology optimisation of
periodic microstructures with buckling criteria, WCCM 5, Vienna, July
2002.
Pedersen, C.B.W., Topology optimisation of energy absorbing frames,
WCCM 5, Vienna, July 2002.
491
Stress intensity factors for cracked cold-drawn

steel wires under tensile loading
B. Lin1 & G. Lu 2
1
School of Naval Architecture, Ocean and Civil Engineering,

Shanghai Jiaotong University, Peoples Republic of China
2
Department of Bridge Engineering,
Tongji University, Peoples Republic of China
Abstract
Cold-drawn steel wires have the excellent mechanical properties of strength and
toughness and are widely used in prestressed concrete structures. During
cold-drawn operations residual stresses are generated in steel wires. In this paper,
by the residual stress distribution from the finite element analysis (FEA), a
weight function method (WFM) has been used to evaluate the effective stress
intensity factors (SIFs) for cracked cold-drawn steel wires under tensile loading.
The calculation results have been compared with those obtained by the
two-dimensional (2D) FEA, which considers the residual stress redistribution in
the presence of a crack. In the present study, the effective SIFs calculated using
the WFM have shown a good agreement with those derived from the 2D FEA.
Keywords: stress intensity factor, residual stress, weight function, cold-drawn
steel wire, finite element analysis.
Introduction
In the cold-drawn procedure, the tensile strength and toughness of steel wires
increase, but a residual stress field also appears in them. After the treatment of a
further drawing with a very small area reduction or a combination of heating and
stretching the wire, the decreased residual tensile stresses still exist in the steel
wire surface [1]. It is well known that the presence of residual tensile stresses
will reduce the fatigue life of steel wires. With increasing in residual tensile
stresses, the fatigue crack growth rate increases for stress ranges close to the
fatigue limit [2]. The calculation of the SIF is one of key factors in fatigue life
doi:10.2495/HPSM06048

determination of the components. The superposition principle, which assumes
that the stresses applied by external loading superimpose linearly on residual
stresses, can be used to calculate the effective SIFs in a cracked body in which a
residual stress field acts [3]. If the weight function, which depends only on the
geometry of the body, is obtained, the contribution of residual stresses on the
effective SIFs for cracks in mode I can be determined by the WFM [4]. As for
cold-drawn steel wires under tensile loading, the net stresses in the steel wires
under service loading remain tensile. When a crack is produced such that it lies
wholly in a tensile region the faces of the crack will remain fully open and the
applicability of the linear superposition principle is valid [5]. Additionally, the
produced crack will bring the residual stresses ahead of a crack tip to
redistribute, [5,6]. So, the availability of the WFM, which assumes that the
residual stress distribution keeps invariable, must be validated when dealing with
residual stresses and fatigue crack growth.
In this paper, the 2D FEA, which considers the residual stress redistribution
in the presence of a crack, is carried out to validate the WFM for calculations of
effective SIF for cracked cold-drawn steel wires under tensile loading. Section 2
presents the 2D model for the cracked steel wire. Calculations of the effective
SIF by the weight function analysis are described in Section 3 and by the 2D
FEA in Section 4. Section 5 gives the corresponding analysis results and
discussion. The present work is concluded in Section 6.
app
app
A
C
x
app
(a)
Figure 1:
O
C
a
a
2b
D
A -A
app
(b)
(c)
3D geometry and 2D model of a wire used in the analysis.
Two-dimensional model
The surface crack of the cold-drawn steel wires produced by mechanical fatigue
can be assumed to be a semi-elliptical geometry (a/b<1, a is the crack depth or
minor axis of the ellipse, b the major axis of the ellipse), perpendicular to the
tensile loading direction [7], as show in fig. 1(a), 1(b), and the maximum SIF
value was obtained at the crack center C, according to the results given by
Astiz [8]. Following a well known purely linear elastic solution in the vicinity of
493
a 3D crack with an arbitrary curved front [9], the calculation of the SIF can be
based on 2D plane strain hypothesis on the longitudinal section of a cracked
cold-drawn steel wire under tensile loading, fig. 1(c).
Weight function analysis
As for a cracked cold-drawn steel wire under tensile loading, the effective SIF at
the crack front is decided by the combined effect of local residual stresses and
the stresses applied by external loading. According to the linear superposition
principle, the effective SIF (Keff) is the algebraic sum of the external SIF (Kapp)
due to the external applied loads and the residual SIF (Kres) due to the residual
stresses, and can be expressed as:
Keff = Kapp + Kres
(1)
3.1 Residual SIF
It is well known that, once the weight function is determined for a particular
cracked body, the SIFs can be obtained for any stress field by an appropriate
integration. So, in this section, the WFM is used to evaluate the residual SIF
(Kres).
Based on the FEA [1], the distribution of the residual stresses for colddrawn steel wires with a thermomechanical treatment is shown in fig. 2, where r
is the polar coordinate of the wire and R the radius of the wire, and the equation
used for the calculation of the residual stresses in the plane problem is as
follows:
res( r R ) = 235.71 + 411.88 r R 987.84 (r R )2 + 3133.93 (r R )3
4
5
(2)
3257.62 (r R ) + 1029.93 (r R ) .
In the edge-cracked 2D model, fig. 1c, a residual SIF can be calculated by
integrating the weight function m(x,a) and the residual stress distribution res(x)
acting on the crack length a [10], which is expressed as follows:
Kres = res( x )m( x , a )dx
(3)
500
(MPa)
Longitudinal redidual stress
Figure 2:
-500
1.0
0.8
0.6
0.4
Relativ depth, (r/R)
0.2
0.0
Residual stress distribution for a steel wire used in the analysis.

The analytical weight function for an edge crack in a finite width plate, as
shown in fig. 3, was given by Kaya and Erdogan [11]:
m( x ,a ) =
x a
G ,
a D
a
1
D
3/ 2
x
1
a
(4)
where
2
a x
a x
a x
a
x a
G , = g1 + g 2 + g3 + g 4 ,
D a
D a
D a
D
a D
5
a
a
a
a
a
g1 = 0.46 + 3.06 + 0.841 + 0.66 1 ,
D
D D
D
D
2
a
a
g 2 = 3.52 ,
D
D
2
3/ 2
a
a
a a
a
g3 = 6.17 28.22 + 34.54 14.39 1
D
D
D
D D
5
a
a
a
5.881 2.64 1 ,
D D
D
2
a
a
a
a
g 4 = 6.63 + 25.16 31.04 + 14.41
D
D
D
D
a
+ 21
D
3/ 2
a
a
a
+ 5.041 + 1.98 1 .
D D
D
y
x
P=1
Figure 3:
Weight function notation for an edge crack in a finite width plate.
495
3.2 External SIF

An empirical solution of the SIF for a single edge cracked plate applied under
the axial stress, fig. 1(c), given by Tada et al. [12] is used to calculate the
external SIF (Kapp), which is expressed as follows:
Kapp = app a F ( a D )
(5)
where app is the external applied stress, D the diameter of the steel wire and F
(a/D) given by:
F( a / D ) =
a
a
0.752 + 2.02 + 0.371 sin
a
2D
D
2D
tan
.
a
a
2D
cos
2D
app
L=10 mm
a
D/2=2.5 mm D/2=2.5 mm
Figure 4:
2D finite element model.
Finite element analysis
4.1 2D finite element model

2D finite element model of the cracked steel wire was developed for numerical
stress analysis. A detail of a typical mesh was shown in fig. 4. A different mesh
was used for each crack length to present the equivalent level of mesh
refinement. The ABAQUS finite element program [13] was chosen for the
present analysis. 8-node biquadratic solid element for plane strain available in
the ABAQUS code element library was used in the model, and the quarter point
technique, used for numerical evaluation of the SIF, was applied to improve
precision of the results. Due to symmetry, as shown in fig. 1(c), only the top
half-plane was modelled for the specimen with a homogeneous linear elastic

material using a Youngs modulus, E, of 207 GPa, Poissons ratio,, of 0.3 and
Limit strength, fptk, of 1940 MPa .
The SIF for the elastic singular stress field was calculated by means of the
J-integral for 10 different paths, which was obtained using the domain integral
formulation in ABAQUS. Calculations coupled with the residual stress field
were carried out for each crack length. Two levels of applied loading, app= 0.4
and 0.7fptk, were employed and applied to the top edge of the model. From the
average J-integral value, the effective SIF was computed.
4.2 Verification of the finite element model
In order to verify the finite element models accuracy to calculate SIFs for
different crack lengths, comparisons between these solutions and empirical
solutions given by eqn. (5) were made only under tension loading. The
normalised SIF, KI app a , is presented in fig. 5, showing a fairly good
agreement and all differences being within 1%. Based on these results, the
present finite element model was considered suitable for the 2D analysis of
cracked cold-drawn steel wires under tensile loading.
Normalised SIF, I/app(a)
2.4
FEM
Tada et al.
2.1
1.8
1.5
1.2
0.0
Figure 5:
0.1
0.2
0.3
Normalised crack length, a/D
Comparison of KI
calculations.
app
0.4
a between FEM and Tada et al.s
4.3 Residual stresses

The initial stress method illustrated by Lei et al. [14] can be used to introduce a
residual stress distribution into the finite element model as an initial condition,
and a fine mesh has been employed to provide for an accurate description of the
residual stress distribution in the model. But the introduced residual stresses are
497
not self-equilibrating; usually resulting in a stress distribution that differs

significantly from desired one. This problem was overcome by using two load
steps in the linear elastic analysis. In the first step, reactions required to balance
the out-of-balance load vector caused by the residual stresses were derived with
all degrees of freedom constrained and no other applied loadings. In the second
step, the obtained reactions were converted to applied force, and an equilibrium
step analysis was executed for the uncracked model with the standard boundary
conditions. After this procedure, the contour of the residual stresses (unit: MPa)
is presented in fig. 6. It is shown that the residual stress distribution in the
uncracked model accords well with the target distribution shown in fig. 2.
Figure 6:
Contour of residual stresses in the uncracked model.
4.4 Residual stress redistribution

After the residual stresses are introduced into the uncracked model, a crack can
be introduced into the FE model by releasing nodes located on the crack face
until the desired crack length is reached. When a node is released from its
displacement constraint, the reactions previously acting on the node disappear,
and moreover, the equilibrium state in the elements around the node is also
changed. Subsequently, a new equilibrium state is re-established. The shift of the
equilibrium state will bring the residual stresses in these elements to redistribute.
Analysis results and discussion
Both the WFM and the FEM described above were used to evaluate the variation
of the effective SIF as a function of the crack length. The effective SIFs
calculated from the WFM and those derived from the FEA are compared in fig. 7
for seven different crack lengths. From this figure, one can find that, for different
crack lengths under two loading levels, the maximum difference in values
between the two methods is 0.87%, which exhibits a good mutual agreement.
Comparison results between the FEA and the WFM showed that, when
calculating effective SIFs with a linear elastic analysis for cracked cold-drawn

steel wires under tensile loading, the phenomenon of redistribution was not
obvious to residual stress problems, which keeps a good agreement with the
results for cold-worked cracked holes given by Moreira et al. [15].
240
Effective SIF (MPam)
200
WFM, app=0.4fptk
WFM, app=0.7fptk
FEA, app=0.4fptk
FEA, app=0.7fptk
160
120
80
40
0.0
Figure 7:
0.1
0.2
0.3
Normalised crack length, a/D
0.4
Comparison of residual SIF between weight function and finite

element results.
Conclusions
In this paper, we have demonstrated that, according to the obtained residual

stress fields from documents, the effective SIF can be calculated for a cracked
body using either the WFM or the FEA. The following conclusion can be drawn:
the reliability of the WFM results is demonstrated by comparing with FE
calculations, and the complexity of the analysis is remarkably reduced.
References
[1]
[2]
[3]
[4]
Elices, M., Influence of residual stresses in the performance of cold-drawn

pearlitic wires. Journal of Materials Science, 39(12), pp. 3889-3899,
2004.
Llorca, J. & Sanchez-Galvez, V., Fatigue limit and fatigue life prediction
in high strength cold drawn eutectoid steel wires. Fatigue and Fracture of
Engineering Materials & Structures, 12(1), pp. 31-45, 1989.
Parker, A.P., Stress intensity factors, crack profiles and fatigue crack
growth rates in residual stress fields. ASTM STP 776, pp. 13-31, 1981.
Glinka, G. & Shen, G., Universal features of weight functions for cracks
in mode I. Engineering Fracture Mechanics, 40(6), pp. 1135-1146, 1991.
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
499
Wilks, M.D.B., Nowell, D. & Hills, D.A., The evaluation of stress

intensity factors for plane cracks in residual stress fields. Journal of Strain
Analysis for Engineering Design, 28(3), pp. 145-152, 1993.
Lee, Y.B., Chung, C.S., Park, Y.K. & Kim, H.K., Effects of redistributing
residual stress on the fatigue behavior of SS330 weldment. International
Journal of Fatigue, 20(8), pp. 565-573, 1998.
Toribio, J. & Toledano, M., A fracture criterion for prestressing steel
cracked wires. Proc. Of the 2nd Int. Conf. On Advances in Steel
Structures, eds. S.L. Chan & J.G. Teng, Elsevier: New York, pp. 947-954,
1999.
Astiz, M.A., An incompatible singular elastic element for two- and threedimensional crack problems. International Journal of Fracture, 31(2), pp.
105-124, 1986.
Hartranft, R.J. & Sih, G.C., Stress singularity for a crack with an
arbitrarily curved front. Engineering fracture Mechanics, 9(3), pp. 705718, 1977.
Cartwright, D.J., Stress intensity factor determination (Chapter 2).
Developments in Fracture Mechanics, eds. G.G. Chell, Applied Science:
London, pp. 29-66, 1980.
Kaya, A.C. & Erdogan, F., Stress intensity factors and COD in an
orthotropic strip. International Journal of Fracture, 16(2), pp. 171-190,
1980.
Tada, H., Paris, P.C. & Irwin, G.R., (eds). The Stress Analysis of Cracks
Handbook, Del Research Corporation: Hellertown, Pennsylvania, 1973.
Hibbitt, B., Karlsson, B. & Sorensen, P., (eds). ABAQUS/Standard Users
Manual, version 6.1, Hibbitt, Karlsson & Sorensen Inc.: Rode Island, RI,
2001.
Lei, Y., O'Dowd, N.P. & Webster, G.A., Fracture mechanics analysis of a
crack in a residual stress field. International Journal of Fracture, 106(3),
pp. 195-216, 2000.
Moreira, P.M.G.P., De Matos, P.F.P. & Pinho, S.T., The residual stress
intensity factors for cold-worked cracked holes: A technical note. Fatigue
and Fracture of Engineering Materials and Structures, 27(9), pp. 879886, 2004.
501
Void growth and damage ahead of a crack in

pressure-sensitive dilatant polymers
H. B. Chew, T. F. Guo & L. Cheng
National University of Singapore, Singapore
Abstract
The deformation of amorphous polymers is highly sensitive to hydrostatic
pressure and the plastic flow is non-volume preserving. Failure of these materials
is often preceded by the formation of small crack-like defects known as crazes,
bridged by the fibrils of the polymer. In this work, a population of discrete voids
is placed ahead of a crack within a boundary layer configuration to study the
effects of pressure-sensitivity and plastic dilatancy on void growth and
damage in polymeric materials. Results show that high pressure-sensitivity
significantly increases both the intensity and spatial extent of damage, which
could promote crazing. These effects can be compounded as the deviation from
the associated flow rule, | |, increases. This study offers some evidence of
pressure-sensitivity and plastic dilatancy as the major contributing parameters to
the formation and growth of microporous crazing zones in polymers.
Keywords: void growth, coalescence, pressure-sensitive yielding, plastic
dilatancy, polymer.
Introduction
The ductile fracture of many engineering materials involves the micromechanical

process of void growth and coalescence. The fracture surfaces of these materials
contain voids of various size-scales. In polymeric materials, the larger voids can
originate from cavitated rubber blends or from decohesion of filler
particle/polymer matrix interfaces, while shear banding and crazing can induce
the formation of the smaller microvoids. The ductile fracture process can be
described by mechanism-based computational approaches which employ voidcontaining cell elements [1]. Porous material models, such as the Gurson
doi:10.2495/HPSM06049

constitutive model [2], are used to describe void growth for each cell element,
and the macroscopic softening during the fracture process. This approach has
been widely used in the prediction of ductile crack growth in metals. More
recently, Guo and Cheng [3] extended the Gurson model to incorporate vapour
pressure effects. The extended Gurson model has been used to study the vapour
pressure assisted void growth and rupture of hygroscopic polymeric materials,
like the die attach and molding compound in electronic packages [46].
While the micromechanisms of ductile fracture for metals and polymers are
somewhat similar, two important characteristics of polymers that differ from
metals are its pressure-sensitive yielding, and its non-volume preserving plastic
flow. To this end, Chew et al. [7] examined the effects of microvoid interaction
in pressure-sensitive dilatant polymeric materials. They showed that increasing
pressure-sensitivity severely reduces the materials stress carrying capacity,
while multiple void interactions was responsible for the sharp post-peak stress
drop and could act as a catalyst for rapid failure. In addition, plastic dilatancy
was found to have some effect in raising the post-peak stress levels, resulting in a
larger work of separation.
An accurate simulation of the void growth behaviour in pressure-sensitive
dilatant polymeric materials can be achieved by explicitly modelling the voids
using refined finite elements. This discrete void approach was previously
adopted by Tvergaard and Hutchinson [8] and Kim et al. [9] for the study of the
ductile fracture of metals, and by Chew et al. [10] for the modelling of moisture
induced failure of hygroscopic polymeric materials.
In this work, a single row of discrete voids is placed ahead of a crack within
a boundary layer configuration to study void growth and damage in pressuresensitive dilatant polymeric materials (Fig. 1). The pressure-sensitive yielding of
the material is described by a linear combination of the mean stress and effective
stress under a non-associated flow rule. The motivation behind this study is to
understand the contributions of pressure-sensitivity and plastic dilatancy to
the formation and growth of long craze zones widely observed in polymeric
materials [11].
X2
2R0
X1
Figure 1:
Schematic of cylindrical discrete voids ahead of a crack in a

boundary layer configuration.
Problem formulation
2.1 Material model

The plastic behaviour of polymeric materials differs considerably from the von
Mises material. Such behaviour can be explained by assuming a yield criterion
503
based on a linear combination of the mean stress and effective stress. Here, the
pressure-dependent yielding of the material is described by
3J 2 + m tan = 0
(1)
where J2 is the second invariant of the deviatoric part of the Cauchy stress tensor
, m the mean stress, the flow stress of the subsequent yield surface, and
the index for pressure-sensitivity. We assume the flow potential to take the form
= 3J 2 + m tan
(2)
where is the index for plastic dilatancy. The plastic deformation rate is given
by the non-associated flow rule
d P = P
(3)
where P is the equivalent strain rate.

The flow stress is a function of the accumulated plastic strain
P
= P dt. For a power-law plastic hardening solid, one has
1/ N
0

E 0
=P
(4)
where N is the hardening exponent ranging from 0 to 1, and 0 the initial yield
stress under pure shear which is related to the initial tensile and compressive
yield stresses 0t and 0c, by
1
t
1 + 3 tan 0 for tension

0 =
1
1 tan t for compression.

0
3
The pressure-sensitivity index can be determined from 0t and 0c as
0c 0t
.
c
t
0 +0
= tan 1 3
(5)
(6)
In polymeric materials, typical pressure-sensitivity levels range between 0 to

23.
2.2 Small-scale yielding
Xia and Shih [1] simplified the ductile fracture process under small-scale
yielding conditions by placing a single row of void-containing cells ahead of the
crack. In the present study, we model void growth and damage via an array of
unit-cells ahead of the crack (see Fig. 1). Each unit-cell is of dimensions D by D
and contains a discrete cylindrical void of initial radius R0. The initial void
volume fraction is given by f 0 = R02 / D 2 . A total of 23 voids are used in the
computation.

(a)
(b)
Figure 2:
Finite element mesh for small-scale yielding analysis. (a) Entire

mesh showing remote boundary. (b) Close-up view of the crack tip.
Along the remote circular boundary (see Fig. 2a), the elastic asymptotic (inplane) displacement field
u1 ( R, ) = K I
1 +
E
R
(3 4 cos )cos
2
2
(7)
1 + R
(3 4 cos )sin
u 2 ( R, ) = K I
E
2
2
is applied under plane strain conditions, where R2 = X12 + X22 and = tan-1(X2/X1)
for points on the remote boundary. KI is the mode I stress intensity factor related
to the J-integral by
1 2 2
J=
KI .
(8)
E
At various stages of loading, the value of the J-integral is calculated on a number
of contours around the crack using the domain integral method. The domain
integral value was found to be in good agreement with the value given by eqn.
(8) for the prescribed amplitude KI. This consistency check assures that smallscale yielding conditions are satisfied.
505
Figure 2 displays an example finite element mesh for the discrete void
model. By taking advantage of symmetry, only one half of the geometry needs to
be modelled. The mesh contains 4-node bilinear quadrilateral hybrid plane strain
elements. The computations are performed within the finite strain setting using
the general-purpose finite element program ABAQUS Version 6.5.1 [12].
Crack tip plastic zone
Before proceeding to the modeling of void growth, we first examine the effects
of pressure-sensitivity on the extent of plastic deformation ahead of the crack.
Some insights can be obtained from the Dugdale strip yield model, which
assumes a long slender plastic zone at the crack tip in a nonhardening pressureinsensitive material ( = 0). The plastic zone size L under a uniaxial tensile
stress T is given as
L=
KI
8 T
(9)
For the pressure-sensitive material in eqn. (1), the tensile stress T at yielding is
T = 0/(1+1/3 tan ). Denoting L0 as the plastic zone size corresponding to =
0, i.e. T = 0, we obtain
2
1
L / L0 = 1 + tan .
(10)
3
This relation suggests an increase in the plastic zone size L with pressuresensitivity . Observe that an increase in from 0 to 20 increases L by around
25%.
A numerical study is next performed to shed some light on the extent of
plastic dissipation in a void-free pressure-sensitive polymer. The void-free
material is subjected to mode I, K-field loading in eqn. (7). An associated flow is
assumed with = . The elastic modulus of the polymer is taken to be E = 500
MPa, with a yield strength of 0 = 5 MPa, and a Poissons ratio of = 0.4. In
accordance with the Dugdale assumption, the strain hardening exponent N is set
to zero (for ideally-plastic material).
Figure 3 displays the plastic zones in the deformed mesh configuration for
three pressure-sensitive levels: = 0, 10 and 20. The plastic zones are
operationally defined by p 0.001 , where p is the accumulated plastic strain.
Observe that at the same external load level, the plastic zone spreads out as the
pressure-sensitivity level increases. For example, at J = 68 kNm-1, an increase in
from 0 to 20 raises the extent of plastic dissipation ahead of the crack by
nearly two-fold (L/L0 2). This general trend of increasing plastic dissipation
with pressure-sensitivity corroborates with the findings from eqn. (10).
Damage evolution in porous polymers
In this section, an array of discrete voids is placed ahead of the crack. From
dimensional considerations, the spatial distribution of the current void volume

fraction f depends on the following dimensionless geometric-material
parameters:
J 0
;
, N , v, , ; f 0 .
0D E
(11)
The properties of the polymeric material are specified by the parameters 0/E =
0.01, = 0.4, and N = 0.1. We focus on the effects of pressure-sensitivity and
plastic dilatancy on void growth and damage ahead of the crack.
(a)
= 0
L0
(b)
= 10
L
(c)
Figure 3:
= 20
Plastic zones in the deformed mesh configuration for several

pressure-sensitivity levels at J = 68 kNm-1.
4.1 Pressure-sensitivity effects
The effects of pressure-sensitivity under an associated flow rule ( = ) are

shown in Fig. 4 for f0 = 0.005 and f0 = 0.05. Computations focusing on the
ductile fracture of metals by Tverggard and Hutchinson [8] have identified two
distinct mechanisms of fracture: (i) single void growth mechanism, and (ii)
multiple void interaction mechanism. Both these failure mechanisms are also
exhibited in our study for amorphous polymers.
At equivalent load levels, we observe that an increase in significantly
increases both the intensity and extent of damage ahead of the crack. Herein, we
operationally define the spatial extent of damage as the distance from the crack
tip to the region where the void porosity has increased by f f0 > 0.01. For f0 =
0.005 at J/(0D) = 0.36, an increase in from 0 to 15 raises the peak porosity
level by 160% and increases the extent of damage by nearly three-fold. Similar
observations are made for f0 = 0.05.
The plastic zone distribution in the deformed mesh configuration for f0 =
0.005 is displayed in Fig. 5. Observe that the plastic zone spreads over a much
larger region as increases. For = 0, the plastic zone is confined to the first
507
two voids ahead of the crack, resulting in near-tip void growth. For = 15, the
plastic zone spreads across the first four voids, leading to the formation of an
extended damage zone. This extended damage zone closely resembles the long
craze zones in polymers. One can therefore infer that pressure-sensitivity is a
major contributing parameter to the crazing phenomenon. Interestingly, we also
note that there is an increasing tendency for the voids to become oblate as they
grow larger. This suggests the effects of void shape to play an important role in
the coalescence stages of fracture.
(a)
(b)
Figure 4:
Distribution of porosity f ahead of crack (X2 = 0) for several

pressure-sensitivity levels with = , (a) f0 = 0.005; (b) f0 = 0.05.

(a)
=0
(b)
=15
Figure 5:
Plastic zones in the deformed mesh configuration for several

pressure-sensitivity levels at J/(0D) = 0.36. f0 = 0.005, (a) = 0;
(b) = 15.
4.2 Plastic-dilatancy effects
The analysis in the previous section employs an associated flow rule, i.e. = .
Experimental studies, however, have shown that the plastic volume change in
polymers does not commensurate with the predictions of the associated flow
rule, i.e. < .
Figure 6 displays the effects of on the damage distribution ahead of the
crack for = 10 and 20. One can see that an increase in the deviation from the
associated flow, | |, significantly increases both the intensity and the spatial
extent of damage. For f0 = 0.005, = 20, at J/(0D) = 0.60, an increase in | |
from 0 to 5 raises the damage intensity level by nearly 60%, and shifts the
damage process zone from X1 = 5D to X1 = 7D. The effects of increasing | |
are found to be less severe for f0 = 0.05. The non-conservative nature of the
associated flow in predicting damage has serious implications in the design of
engineering materials and structures.
Conclusion
The effects of pressure-sensitivity, , and plastic dilatancy, , on void growth

and damage in amorphous polymeric materials have been studied. Our numerical
results for a void-free polymeric material show that pressure-sensitivity
significantly increases the extent of plastic dissipation ahead of the crack. Void
growth and damage in the pressure-sensitive dilatant polymeric material was
subsequently modelled via a single row of discrete voids placed ahead of the
crack. Both the voiding intensity and the spatial extent of damage were observed
to increase with pressure-sensitivity, which could promote crazing. These effects
were further compounded as the deviation from the associated flow rule, | |,
increases. These findings suggest pressure-sensitivity and plastic dilatancy to be
closely linked with the formation of long craze zones in polymers.
509
(a)
(b)
Figure 6:
Distribution of porosity f ahead of crack (X2 = 0) for several plastic

dilatancy levels with = 10 and 20, (a) f0 = 0.005; (b) f0 = 0.05.
References
[1]
[2]
Xia, L. & Shih, C.F., Ductile crack growth - I. A numerical study using
computational cells with microstructurally-based length scales. Journal of
the Mechanics and Physics of Solids, 43, pp. 233-259, 1995.
Gurson, A.L., Continuum theory of ductile rupture by void nucleation and
growth: Part I - Yield criteria and flow rules for porous ductile media.
Journal of Engineering Materials and Technology, 99, pp. 2-15, 1977.

[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
Guo, T.F. & Cheng, L., Modeling vapor pressure effects on void rupture
and crack growth resistance. Acta Materialia, 50, pp. 3487-3500, 2002.
Chew, H.B., Guo, T.F. & Cheng, L., Vapor pressure and residual stress
effects on the toughness of polymeric adhesive joints. Engineering
Fracture Mechanics, 71, pp. 2435-2448, 2004.
effects on failure of an adhesive film. International Journal of Solids and
Structures, 42, pp. 4795-4810, 2005.
effects on mixed mode toughness of an adhesive film. International
Journal of Fracture, 134, pp. 349-368, 2005.
Chew, H.B., Guo, T.F. & Cheng, L., Effects of pressure-sensitivity and
plastic dilatancy on void growth and interaction. International Journal of
Solids and Structures, in Press, 2006.
Tvergaard, V. & Hutchinson, J.W., Two mechanisms of ductile fracture:
void by void growth versus multiple void interaction. International
Journal of Solids and Structures, 39, pp. 3581-3597, 2002.
Kim, J., Gao, X. & Srivatsan, T.S., Modeling of crack growth in ductile
solids: a three-dimensional analysis. International Journal of Solids and
Structures, 40, pp. 7357-7374, 2003.
Chew, H.B., Guo, T.F. & Cheng, L., Vapor pressure and voiding effects
on thin film damage. Thin Solid Films, in Press, 2006.
Kambour, R.P., A review of crazing and fracture in thermoplastics.
Journal of Polymer Science, 7, pp. 1-154, 1973.
Hibbit, Karlsson & Sorensen Inc., Abaqus/Standard Users Manual,
Version 6.5.1, Vol. 1, 2005.
511
An orthotropic damage model for

crash simulation of composites
W. Wang1, F. H. M. Swartjes1 & M. D. Gan2
1
BU Automotive
Centre of Lightweight Structures TUD-TNO,
TNO Science and Industry, Delft, The Netherlands
Abstract
In this paper, a practical orthotropic damage model is developed and tested for
composite materials during crash. The model uses the Hashins failure criteria in
which the fibre and matrix failures are described explicitly, both in tension and
compression. A linear softening degradation is proposed and a close-form
solution of the corresponding damage parameter is provided.
To reduce the mesh dependency, an embedded discontinuous element is
proposed. It is a virtual embedded element in the sense that an actual element is
divided into two zones, one elastic (undamaged) zone and one localization
(damaged) zone. Their equivalence is preserved by constraining the kinematics
and equilibrium equations. Since the damage zone is introduced into the element,
the corresponding dissipated energy due to damage is fixed, independent of the
element size.
The numerical simulations using the developed damage model show a mesh
objective result and correlate well with the energy dissipation in dynamic 3-point
bending experiments. It is concluded that an accurate material calibration is
crucial for the success of failure simulations. The traditional strain mapping
cannot be applied to failure problems in which the strain distribution is highly
non-uniform. Material parameter calibration should take into account the ratio of
the failure zone and the gauge length. A practical approximation formula is
provided.
Keywords: composite, damage model, crash simulation, embedded element.
Introduction
The crash simulation of composites is particularly difficult due to the complexity

of its physics. First, the failure process involves multiscales and is difficult to
doi:10.2495/HPSM06050

observe experimentally. Second, the softening effect leads to the well known
numerical simulation difficulties, cf. mesh dependency, instability and
bifurcation.
In this paper, a practical orthotropic damage model is developed and tested
for composite materials during crash. This new model is implemented in LSDYNA as a user-defined material model. The related issues such as the mesh
dependency and material calibrations of the corresponding damage parameters
are addressed.
Continuum damage constitutive relationship
For plane stress situations in the continuum damage approach, the effective
is related to the total strain by the elastic stiffness tensor C0
stress
= C 0
(1)
The effective stress is mapped to the true stress by the damage operator M
= M ,
0
1 D1 0
M= 0
1 D2
0
0
0
1 D12
(2)
where D1, D2 and D12 are the damage parameters. Inserting Equation (1) into
Equation (2) yields
= M C0 = C
(3)
in which the damaged stiffness tensor C reads
(1 D1 )E11
1
C = (1 D2 ) 12 E11

0
(1 D1 ) 21 E22
(1 D2 )E22
0
(1 D12 ) G
= 1 12 21
(4)
(5)
with E11, E22, G, 12 and 21 are the elasticity parameters of the undamaged
lamina. Note that when damage occurs (Di>0), the stiffness tensor is not
symmetric. To preserve a symmetric stiffness tensor, the Poissons ratio can be
related to the damage parameters [3].
513
Hashins failure criteria
Hashins criteria [2] refers to an unidirectional fibre reinforced composite with

five failure modes:
1) Fibre tensile and compression modes
2
0 failed

e = 11 1
X
< 0 elastic
2
f
(6)
2) Matrix tensile and compression modes

2
0 failed

em2 = 22 1
Y
< 0 elastic
3)
(7)
Shear mode
2
0 failed

e = 1
S
< 0 elastic
2
s
(8)
where X and Y are fiber and matrix strengths and S is the shear strength. Note
that the compression and tensile modes are treated in the same way. For notation
simplicity, the sub indexes for the tensile and compression strength parameters
are omitted.
Note that Hashins criteria have a complete decoupling of the fibre and matrix
failure modes from shear failure. Therefore, the failure criteria may overestimate
the strength of a lamina when shear is present. However, it should be mentioned
that measuring the softening part of the stress-strain curve is very difficult even
in a uniaxial test condition. A multi-axial loading surface is almost impossible to
be validated. Furthermore, the damage evolution or degradation law is, in most
cases, in an uncoupled fashion. Therefore, it is convenient to treat all failure
criteria to be independent.
It must be emphasised that although the failure modes and the degradation
law (which will be introduced in the next section) are decoupled, the complete
stress-strain relationship still contains a coupling due to the stiffness matrix C
(see Equation 3).
4 Degradation rule
Damage occurs when one of the Hashins criteria is satisfied. For simplicity, we
assume a linear degradation law (see Figure 1) after the initial strength 0 is
exceeded. Consequently, the damage parameter D can be determined by the
following relationship
= E ( D) = E0 (1 D ) = 0 h ( 0 )
(9)

where E0 is the elastic modulus, 0 = 0/E0 the initial failure strain and h is the
softening parameter, respectively.
Figure 1:
A linear stiffness degradation.
It must be emphasised that Equation (9) holds for materials in the localisation
zone. Two issues related to this localisation problem should be addressed:
1) The material calibration of localisation related parameters such as the
softening parameter h and the final failure strain f. Note that the strain field
is not uniform as soon as localisation occurs. Therefore, the standard strain
mapping in uniaxial tensile examples: = u/L where L is the gauge length
and u the deformation, is not valid. The results of the experiments should be
interpreted carefully.
2) The mesh dependent problem. For a tensile bar with a uniform cross section
area A, only one element will follow the failure path. Consequently, the total
dissipation energy of the tensile bar reads
1
Wdiss = 0 f 0 l A
2
(10)
where l is the length of the failure element. Clearly, the energy dissipation is
dependent on the element size and, therefore, the simulation result is mesh
dependent.
These two issues will be discussed in the next section. First, the mesh
dependence problem is solved by introducing a localisation zone into a standard
finite element. Next, the corresponding localisation related parameters are
investigated and their relationships are outlined. Finally, formula will be
provided for practical approximation of the final failure strain and the softening
parameter.
Figure 2:
515
Embedded localisation element (left) and its equivalent element

(right).
Embedded localization element

divided into two zones, one elastic (undamaged) zone and one localization
the corresponding dissipated energy due to damage is fixed, independent of the
element size. Therefore, mesh dependency with respect to dissipated energy can
be removed.
Assume that the length of the damage zone is and the length of the element
is L, see Figure 2. When damage occurs, the strain localised in the damage zone
is h and the strain at the elastic zone is e. If we smear the localisation zone and
make an equivalent element, the following kinematics and equilibrium equations
must be satisfied
h + (L ) e = L = u
(11)
E0 e = Eh h = E L =
(12)
where E0 is the elastic modulus, Eh the stiffness at the damage zone and EL the
equivalent (element) elastic modulus, respectively
E h = (1 d ) E
(13)
E L = (1 D ) E
(14)

Note that d and D are damage parameters of the embedded localisation element
and the equivalent element, respectively.
From Equations (11-14) the localisation strain h can be related to the
equivalent (element) strain
h =
1 + d (r 1)
r=
(15)
and the local damage parameter d can be evaluated and then transferred to the
global (element) damage parameter
D=
rd
1 + d (r 1)
(16)
Note that the localisation strain h is a function of damage parameter d. This local
damage parameter can be determined by inserting Equations (13)(15) into the
linear softening law (9)
d=
(E + h ) ( 0 )
E + (E + h ) 0 (r 1)
(17)
With a known element strain , the local damage parameter can be directly
determined. This is the advantage of a linear softening assumption, which makes
the stress update simple and straightforward.
Note that the softening parameters should be carefully calibrated. The
traditional strain mapping cannot be applied to failure problems in which the
strain distribution is highly non-uniform. Material parameter calibration should
take into account the ratio of the failure zone and the gauge length.
For the final failure, d=D=1, from Equation (15), the local failure strain f and
the element failure strain have relationship
f =
r=
(18)
Suppose one uses one element to simulate a tensile bar with a gauge length L.
The material failure strain f reads
f =
Lu u
=
L
(19)
where is the length of the damage zone. Note that the material failure strain is
obtained by dividing the deformation by the length of the damage zone and not
the gauge length.
517
Numerical simulations
6.1 Mesh sensitivity analysis

A mesh sensitivity analysis is performed by simulating the fracture behaviour of
the juvenile part of a dogbone specimen with a multi-element FE model. Three
different meshes are used with element size of 2 mm, 1 mm and 0.5 mm for
mesh 1, 2 and 3, respectively (see Figure 3).
The same material data are used for the LS-DYNA material model 58 and the
new user-defined material model (with the assumption that the failure zone is
equal to 0.1 mm).
In Figure 4 the results of the mesh sensitivity analysis are plotted. Clearly, the
global force is independent on the mesh size.
Figure 3:
Mesh sensitivity analysis: three meshes with element size 2 mm

(left), 1 mm (middle) and 0.5 mm (right).
Figure 4:
Mesh sensitivity analysis for the LS-DYNA material model 58

(left) and the user-defined model (right).

6.2 Three-point bending tests
Next, numerical simulations on three-point bending experiments were performed
using the new developed composite model. These US bumper standard
experiments (NHTSA 49 CFR Part 581: Bumper Standard) were performed on
U-shaped GMT bumper beams with an aluminum impactor having a total mass
778 kg and initial impact velocity of 2.30 m/s.
Figure 5:
FE model for the three-point bending tests on composite U-beams.

Force-deflection curves (filtered with CFC60 filter)
test data
mat58
new model
Force [KN]
Figure 6:
10
20
30
40
50
deflection [mm]
60
70
80
90
Force-deflection curves from test and simulation for the dynamic

three-point bending experiments on composite U-beams.
519
The FE model of the experimental set-up is plotted in Figure 5. The same

material parameters are used for the LS-DYNA material model 58 and the new
user-defined material model.
In Figure 6 the contact forces for the experiment and numerical simulations
are plotted (all data are filtered with a CFC60 filter). Both models predict the
shape of the force-deflection curve accurate until the onset of failure. It is
observed that the new user-defined material model captures the softening range
better than the original LS-DYNA material model 58.
7 Conclusions
An orthotropic damage model is developed and tested for composite materials.
The model uses the Hashins failure criteria in which the fibre and matrix
failures are described explicitly, both in tension and compression. A linear
softening degradation is proposed and a close-form solution of the corresponding
damage parameter is provided. This new model is implemented in LS-DYNA as
a user-defined material model.
divided into two zones, one elastic (undamaged) zone and one localisation
the corresponding dissipated energy due to damage is fixed, independent on the
element size. Therefore, mesh dependency with respect to the dissipated energy
is removed.
It is concluded that an accurate material calibration is crucial for successful
failure simulations. The traditional strain mapping cannot be applied to failure
problems in which the strain distribution is highly non-uniform. Material
parameter calibration should take into account the ratio of the failure zone and
the gauge length. A practical approximation formula is provided.
Reference
[1]
[2]
[3]
Haan, P.A.J. de, Swartjes, F.H.M and Wang, W. Failure of composite

material under impact loading- WP4: Numerical simulation of three-point
bending experiments on bumper beams, TNO Science and Industry,
Internal report 04.OR.SA.032.1/FSW, 2004.
Hashin Z. Failure criteria for unidirectional fiber composites,
Composite struct. 32, 255-264 (1980).
Schweizerhof, K., Weimar, K., Mnz, Th. and Rottner, Th.
Crashworthiness analysis with enhanced composite material models
in LS-DYNA -Merits and limits, 5th Ls-Dyna Users Conference,
Southfield, Michigan, USA (1998).
521
Ring-shaped crack propagation in a cylinder

under nonsteady cooling
V. A. Zhornik, Yu. A. Prokopenko, A. A. Rybinskaya
& P. A. Savochka
Department of Theoretical Physics, Taganrog Pedagogical Institute,
Russia
Abstract
A quasistatic problem of thermoelasticity for a solid infinite cylinder with a ringshaped crack is considered. The cylinder is enclosed in a rigid thin heat
absorbing shell with a slip joint which reduces thermal impact. The external
surface of the shell is subjected to a linear heat transfer by radiation to the
surroundings. We assumed that the cylinder had initially a temperature changing
along the radius and the surrounding temperature was a function of time. The
main result is an expression obtained for the stress intensity factor depending on
time. Various particular cases are considered obtained both by the authors of this
paper and by other scientists. An analysis of the dependences of the stress
intensity factor (SIF) on time shows that with an assigned critical SIF, there are
minimum and maximum crack sizes below and above which the crack will not
grow under the given cooling conditions. For these dimensions, the maximum of
SIF does not reach the critical value. For intermediate crack sizes, growth of the
crack begins at the moment SIF reaches the critical value. Here, the crack first
grows irregularly to the size for which at the given moment of time SIF is equal
to the critical value. Extension of so called hot cracks in a solid cylinder is also
considered. These cracks arise when a cold cylinder is placed in a fusion of the
same material at the melted temperature. Further on this cylinder is rapidly taken
out of the fusion and on its surface the melting layer linked with the cylinder
surface arise. Under cooling of the cylinder, residual tensile stresses arise in the
layer which may cause growth of ring-shaped surface cracks which are
dangerous for further exploitation of this system.
Keywords: thermoelasticity, crack propagation, SIF, powdery covering.
doi:10.2495/HPSM06051
Introduction
A solid infinite cylinder of radius rc with an unloaded co-axial ring-shaped crack

(rd < r < rc) coming to the surface of the cylinder was chosen for crack
propagation analyses in deformed solids under nonstationary thermal loading.
We assume that the cylinder is surrounded by the shell with a slip joint, i.e. the
surfaces of the cylinder and the shell can slide freely relatively to each other to
the axial direction.
It is assumed that the covered cylinder, with a constant initial temperature, is
cooled over is entire surface through heat exchange with its constant
temperature environment. In this case, the heat flows are radial and the crack has
no effect on heat propagation within the cylinder.
Mathematical modeling
In light of this, the solution of the thermoelastic problem is represented as the

sum of two solutions,
ij = ijT + ijP ,
u i = u iT + u iP .
(1)
The first solution, obtained with the assumption that no cracks are present,
satisfies the equations of thermoelectricity for an infinite cylinder in a
generalized plain-strain state. This solution satisfies all boundary conditions
except for those for the edges of the crack, which are loaded by a normal load
T (1 )
zz* = zz
T E (T0 )
= 2 Bi
yn J 0 ( yn ) ( Bi ky
a root of equation
n =1
yn
exp ( yn2 Fo )
2 2
n
r
2 J 1 ( yn ) yn J 0 yn
rc
+ y n2 (1 + 2k )
(2)
yJ1 ( y) = Bi ky 2 J 0 ( y) .
Here E is the elastic modulus of the cylinder material; is the Poisson ratio;
r c
is the Biot criterion;
c 1 + 0 d 2 T
1
= 0
; =
;
1 + 0d
c 1 + 0 d T
k = d/rc; d coating width.
T is the coefficient of lineral expansion; Bi =
523
Calculations
For numeral analyses a solid cylinder of the radius rc = 5103 m made of steel 45
(its chemical composition: Fe, 0,42-0,49% C, 0,17-0,37% Si, 0,5-0,8% Mn; =
7800 kg/m3, c = 470 J/kgK, = 50 W/mK) is chosen, coating is made of bronze
101, (its chemical composition: Cu, 9-11% Sn, 0,8-1,2% P; 0 = 8700
kg/m3, c0 = 380 J/kgK, 0 = 100 W/mK); its width is d = 0,25103m. The
coefficient of heat transfer 0 = 1,5105 W/m2K. The porosity = 0,54. Using
the formula for the porous medium
= 0(1 1,5) for < 0,6
(3)
and the formula
c = c0(1 )
(4)
(5)
= 0(1 )
we obtain: = 4000 kg/m3, c = 175 J/kgK, = 19 W/mK.
As a result k = d/rc = 0,05 << 1, k = 0,0065, Bi = 5. In the case of covering
absence (d = 0) we obtain k = 0, Bi = 0rc/ = 15.
Figures 1 and 2 show the relations between time Fo = at/ rc2 (a heat
diffusing) and axial strains zz* depending on the radius r/rc for the cases of the
cylinder with the covering (Figure 1) and without it (Figure 2). As we can see
from the graph on the figures powdery coverings essentially reduce
thermoelastic stress in cylinder.
Figure 1: Diagram for zz depending on Fo for different r/rc (without

covering).
Figure 2: Diagram for zz depending on Fo for different r/rc (with covering).

In the case of plane-strain stress axial displacement is zero in the plane of the
crack.
When added to the first solution, the second solution from the isothermal
theory of elasticity satisfies all boundary conditions including those for the
crack edges.
Because of symmetry conserning the plane of the crack, the second solution is
for semi infinite cylinder (z > 0) with mixed boundary conditions on the end
wall. Through boundary condition for tangent stress is given on the end wall.
The condition of slipping joint leads to zero radial displacement and zero tangent
stress component.
Mathematical setting of the problem is:
pzz = Tzz ,
u pz = 0 ,
rd < r < rc,
z = 0,
(6)
0 < r < rd,
z = 0,
(7)
0 < r < rc,
z = 0,
(8)
r = rc,
0 < z < ,
(9)
rz = 0
p
rz
= 0,
p
r
u =0
r = rc,
0<z<
(10)
According to the problem (6) (10) the calculation of stress-strain state of the
cylinder with a ring-shaped crack is fulfilled. This solution allowed to us
calculate stress intensity factor (SIF) as a criterion of crack propagation [1, 2].
525
As is known, the solution of the problem from the isothermal theory of
elasticity ij , u i
) reduces to finding the function (, ) , having the form

rc2
Bn ( n )J 0 ( n )e n
n =1 n
(, ) =
(11)
Here, n are the positive roots of the equation of Jl () = 0, = r / rc. The

boundary conditions for the rigid shell (8),(9),(10) are satisfied identically
through selection of the function (, ) . By substitution
Bn ( n ) = b n ( n ) +
rd
2
t
g(t ) cos n dt
2
rc J 0 ( n ) 0
rc
(12)
the unknown Bn is determined from a pair of the ordinary equations constructed

from the mixed boundary conditions assigned in the plane of the crack
pzz = 2F(r ) ,
0 < r < rd,
z = 0,
(13)
u pz
rd < r < rc,
z = 0,
(14)
= ,
Here, is the shear modulus, bn are the coefficients of expansion of the
function F(r ) = zz 2 in terms of J 0 n

T
r
, is the constant normal
rc
displacement along the crack which is determined from the balance condition of
the cylinder in the plane of the crack.
rd
rc g (t )dt =
0
rc
P
F(x )xdx
4 rd
(15)
P is a preset force, applied to the cross-section of the cylinder at infinity.

The final solution of this problem reduces to finding the solution of a
Fredholm equation of the second kind relative to the unknown function g(t)
independent on (in particular can be equal to zero).
g ( ) =
1 P
211
a
(
)
+
yF(yrc )
dy
yF
yr
arcsin
c
2
4rc
y
2 1 yF(yrc )dy 4 K1 ( ) sh
1
ch
+
d yF(yrc )I0 (y )dy +

y 2 2 2 0 I1 ( )
4
K1 ( )
sh
(
)
g
d
I ( ) ch ch d ,
2
0
0 1
0 < < , = rd rc .
(16)

Integral equation (16) with respect to g( ) was solved by a successive
approximation technique where function F(r) was approximated by polynomial
with even degrees from 0 to 8.In this case the maximum approximation error
which occurs at small Fo is no more than 2 %.Stress intensity factor SIF K I is
determined according to [3]
K I = 2
rd
r c g (r d )
(17)
In Figure 3 the dependence of stress intensity factor SIF
K *I =
(1 )
(T0 )
2K I
1/ 2
rc T E
on time for Fo for different ring-shaped crack sizes = rd/rc in the case of rigid
slip joint where d = 0 is shown. In Figure4 the same for d = 2,5104m. In case
of flexible joint SIF is 12 % larger for the same sizes of the cylinder and the
shell.
Figure 3: Dependence SIF on time Fo fir different crack sizes (without

covering).
527
Figure 4: Dependence SIF on time Fo fir different crack sizes (with covering).
4 Analyses and conclusions

Let us analyze ring-shaped crack propagation in a covered cylinder starting from
relation shown in Figure 4.
*
Let us assume critical SIF K IC
= 0.136 (horizontal dotted line). Then, for
example, a very small ring-shaped crack of the size = 0.95 (the depth is 0.2
mm) at the moment of time Fo 0.02 begins to grow unstable, jump-like (point
a) and passes the sizes from = 0.95 to = 0.8 (vertical arrows). After that the
*
crack propagates stable (horizontal arrow) as K I* reaches its critical value K IC
= 0.136. Finally the crack stops at the moment of time Fo = 0.12 (point b) as the
crack size is = 0,5 (the depth is 2,5 mm).
This fact should be taken into account when covering the cylindrical details
by powder materials.
References
[1] Zhornik A.I. Nonsteady problem of the theory of elasticity for a solid
infinite cylinder with a penny-shaped crack, enclosed in a thin shell,
International Applied Mechanics, USA 1993, July P. 47 52.
[2] Zhornik A.I., Zhornik V.A., Kashitsyn L.P. Nonsteady problem of
thermoelectricity for a solid infinite cylinder will a ring-shaped crack,
enclosed in a thin shell, Advanced Computational Methods in Heat Transfer

IV: Book of Proceedings. Fourth International Conference on Advanced
Computational Methods in Heat Transfer Heat Transfer 96
Southampton, UK, Boston, USA, 1996. P. 505 516.
[3] Das B.R. Thermal stress in a long cylinder containing a penny-shaped crack
on the distribution of stress in a long circular cylinder, Int.J. Engng.Sci.1963,v.1 - P.391-406
529
Investigation of the hygrothermal performance

of wooden beam ends embedded in inside
insulated outside walls
H. Stopp & P. Strangfeld
Department of Building Physics, University of Applied Sciences FH
Lausitz, Germany
Abstract
The inside insulation has been investigated in the past by means of scientific
institutions as a necessary method in connection with the thermal improvement
of worth-preserving facades of architectonic heritage. In the case of fair faced
brickwork it is established on homogenous areas of the wall, for instance as a
variant of capillary active systems or as a classical version with a
humid-dependent vapor barrier. With regard to the insulation of special details,
e.g. embedded wooden beam ends within areas of outside walls, the opinion of
experts differs considerably. Both the total ventilation of the beam ends and
complete air tightness are proposed. In line with the new European directive on
the energy efficiency of buildings, innovative ideas should also be developed in
the field of building physics in order to realize the restrictions without damage to
building envelope parts. The main point of this paper is the investigation of the
hygrothermal situation of embedded beam ends in inside insulated outside walls,
dependent on time and climate boundary conditions. The topics are treated by
means of in situ measurements in two test houses in the eastern part of Germany
under condition of use and by experiments in connection with an erected test
stand. A comprehensive numerical simulation of the coupled heat- and mass
transfer supports the measurements to reduce the experimental work. In principle
there are several options to avoid the effects of the hygrothermal deficiencies of
the structure caused or intensified by inside insulation. Hygrothermal effects are
quantified and solutions are found for workmanship without damage.
Keywords:
wooden beam ends, worth-preserving facades, heating energy
saving, hygrothermal performance, damages to structures.
doi:10.2495/HPSM06052
Introduction
In case of worth preserving facades there are some conflicts: for an example on
the one hand the changing climate conditions challenge the mankind to develop
political strategies and to implement measures that reduce greenhouse gas
emissions. Plans of European Member States call for 20% ...40% decrease by
2020. Another key objective, which is linked to the environmental protection, is
to make European states less dependent on oil and gas. Current forecasts indicate
that without measures this dependency will jump to approximately 70% by 2030.
A directive of the European Parliament and of the Council on the energy
performance of buildings was published and discussed in the past. The updated
regulation came into force in Dec.16th 2002 [1]. The directive should be realized
in the member states of the European Union up to 2006. About 41% of the total
final energy demand of Europe is used in the residential and tertiary sectors.
Space heating is by far the largest energy end-use of households in EU Member
States, namely 57%. Therefore the European directive covers four main
elements:
-Establishment of a general framework of a common methodology for
calculation the integrated energy performance of buildings.
-Application of minimum standards on the energy performance to new buildings
and certain existing buildings when they are renovated.
-Certification schemes for new and existing buildings on the basis of the above
standards and public display of energy performance certificates and
recommended indoor temperatures and other relevant climatic factors in public
buildings and buildings frequented by the public.
-Specific inspection and assessment of boilers and heating and cooling
installations.
About 32% of the current stock of the 150 million residential dwellings in the
15 old EU Member States was built prior to 1945, about 40 % between 1945 and
1974 and 28% since 1974. An upgrade of thermal insulation regulations and
improved efficiency for installed equipment for existing dwellings, bringing
them close to current buildings codes, would help to realize an important saving
potential, making it a very desirable and in most cases a cost-effective option.
So the directive is explicit valid for existing buildings too. In the face of the
low turnover rate of buildings with a lifetime from 50 to more than 100 years it
is clear that the existing stock of buildings contains the largest potential for
improving energy performance in short and medium periods of time.
On the other hand due to an architectonic heritage the use of a thermal
insulation composite system or another outside insulation is not possible. An
inside insulation may be an alternative if the hygrothermal and acoustic
behaviour is considered. But in regard with the inside insulation of wooden beam
ends embedded within areas of outside walls, the opinion of experts differ
considerably. For this reason the investigation is to be done both by means of
numerical simulation of the coupled heat- and mass transfer included the effects
of ventilation and a lot of in situ measurements in test houses under condition of
use and by experiments in connection with an installed test stand in a laboratory.
531
Calculation of the hygrothermal performance
Owing to the requirements of the new European directive and to the demand for
a comfortable room-climate the thermal insulation of outside walls of existing
buildings should be increased. A good solution for homogenous areas in case of
worth- preserving facades is the so-called capillary-active inside insulation or a
classical version with a humid-dependent vapor retarder [2]. The results of the
numerical simulation of the coupled heat and moisture transfer, represented in
this paper are carried out by means of the program Delphin [3], validated also
in international projects.
Fig. 1 represents the moderate moisture distribution by means of a capillary
active calciumsilicate inside insulation without use of a classic vapor barrier.
This method guarantees the drying out of moisture caused for instance by driving
rain at the facades. Owing to the capillary diffusivity function of the material
calciumsilicate with a maximum of 10-5m2/s the condensation at the cold site of
the inside insulation is limited in wintertime and due to the deleted vapor barrier
a sufficient drying out process takes place during summer period.
Figure 1:
Moisture field of an inside insulated brick masonry wall by 50mm

calciumsilicate. Indoor climate: 20C; 50%. Outdoor climate: TRY
Middle Europe, hills, Westside
Embedded wooden beam ends in connection with an inside insulated worth

preserving facade can be a hygrothermal problem. A non-reflected application
of energy saving directives particularly with regard to the interior insulation can
cause defects of restoration. The formulated requirements of the energy
directives are raising many questions to the architect, contractor and client of a
restoration project. Therefore the purpose of a special recommendation of the

WTA [4] is to provide support in interpreting and applying the regulations (e.g.
in case of half-timbered framework the additional thermal resistance of an inside
insulation system should be limited to 0.8 m2K/W).
20+
17.5 to 20
15 to 17.5
12.5 to 15
10 to 12.5
7.5 to 10
5 to 7.5
2.5 to 5
0 to 2.5
100
300
12
10
15
y in mm
500
20
20
700
15
10
900
1,100
1,300
100
Figure 2:
200
300
x in mm
400
500
700
Temperature field Jan. 29th, variant heating channel, heating

mobilized, (air temp within channel: 35C, partial pressure of water
vapor: 1170 Pa), TRY Middle Europe.
20+
18 to 20
16 to 18
14 to 16
12 to 14
10 to 12
8 to 10
6 to 8
4 to 6
2 to 4
0 to 2
100
3
300
8
500
y in mm
600
700
10
15
900
5
10
15
1,100
1,300
100
200
300
400
500
600
700
x in mm
Figure 3:
Temperature field Jan. 29th, variant: complete inside insulation

system, TRY Middle Europe.
533
Figure 4:
Effect of thermal bridge caused by a mobilized heating cannel,

additional heat losses of the wall, related high of wall 0.52 m and
2.40 m, width of wall: 1 m.
Figure 5:
Temperature field of the area of the wooden beam end influenced

through a so called passive heating bar (thermal short-circuit by
means of the inserted metallic conductor aluminium) Climate
boundary conditions: indoor 20C/50%; outdoor -10C/80%.

The selected figures 2 and 3 demonstrate the deformation of isotherm-lines by
the operating flow pipe and return, installed within a channel of the insulation
near the floor (the insulating calciumsilicate panel with a thickness of 50 mm is
replaced at the bottom for a height of 80mm) in comparison with an inside
insulation of an outside wall completely (fig. 3). Fig. 2 shows the positive effect
of an operating heating pipe in a channel. The increased temperature of 3 K at
the front side of the beam should not conceal the fact that the calculated
decreasing of the wooden moisture at the same place amounts to about 1mass%
only. The fig. 4 represents the additional heating energy losses caused by the
effect of thermal bridge with different reference areas of the wall. In the fig. 5
the shifted temperature field is shown caused by a so-called passive heating bar.
Figure 6:
Course of the wooden moisture measured in front of the four

wooden beam ends H1...H4 after retrofitting is finished.
Experimental investigation
3.1 In situ measurements

The in situ measurements are carried out under conditions of use in a climate of
Middle Europe in test houses near Berlin [5]. The courses of the wooden
moisture content of four investigated wooden beam ends prove the considerable
quantities of moisture built in during the retrofitting activities. They show the
long periods of drying out in laps of several years also in case of indoor climate
similar to an office, fig 6. The start of the supply of heating at the wooden beam
end H4 by means of a lengthened flow pipe with return as a bypass is to
recognize clearly. The wooden moisture decreases by 2M% in comparison to a
no operating flow pipe (Comment: beam ends H4, H3 and H2 are influenced
through the same pipeline of flow and return for the heaters. Only H4 is to
535
separate from the heating energy supply completely no heater is installed

parallel to the pipeline). The reduction of the relative humidity of the air gape
surrounding the beam head amounts up to about 10%, fig.7.
Figure 7:
Course of the relative humidity of the air gape in the front side of
the beam end H4.
Figure 8:
Course of the temperature in front of the four wooden beams ends

H1...H4 dependent on the heating supply through the flow pipe and
return.
The effect of the additional heating supply through the heating channel to the
temperature in front of the beam ends demonstrates fig. 8. After the heating flow
to H4 is mobilized (Jan.4th.2002) the differences between the temperatures at the
beam ends H4 and H3 are compensated.

Experiments depending on the users activities deal with the influence of the
indoor air on the humidity of the air gaps surrounding the inside insulated beam
ends. The influence of a natural ventilation by opening of windows or/and doors
is an indifferent result in practice and can bring in warm-humid air into the area
of the air gaps, fig. 9.
Figure 9:
Penetration of indoor air to the front side of wooden beam ends

owing to the activities of inhabitants
3.2 Test stand

The effect of an additional energy supply is small in case of a thermal shortcircuit of the natural temperature gradients between the indoor and the air gape
by means of a good conductor without thermal resistance, e.g. through a metallic
bar inserted into the wooden beam end. In opposite to that passive bar the fig. 10
shows the abrupt increasing of the temperature from March 8th in 2003 clearly in
spite of the many measuring points at the prepared wooden beam head. The
period of heating energy supply of an active heating bar is realized in
combination with an underfloor heating.
Conclusions
With regard to the hygrothermal performance of wooden beam ends embedded

in inside insulated worth-preserving outside walls, some results for practice are
represented and evaluated in the following briefly. They are worked out by
numerical simulation and measurements under conditions of use.
Figure 10:
537
Effect of a so called active heating bar within a wooden beam head

mounted in an outside wall of the Building Physics testing floor of
the FH Lausitz.
- A supply of additional heat to the wooden beam ends by an aimed conductivity

of heating energy including available building components represents the best
solution.
- For instance without additional expenses and few effort the heating systems
flow pipe and return fitted in a heating channel should be used. The effect of
thermal bridges is to neglect.
- The idea of a so-called passive or active heating bar, inserted into the wooden
beam end connected with an underfloor heating, is an interesting variant. It
should be tested detailed in future.
- Concerning the influence of the natural ventilation on the hygrothermal
situation of the air gaps surrounding the beam ends the investigations show that
the wooden beam ends are individualists.

- In connection with a bigger stock of existing buildings of an increasing
European Union the topic provides an option for the preservation of a lot of the
building heritage in the new eastern countries. The thermal insulation of outside
walls with worth-preserving facades and embedded wooden beam ends can be
improved without damage.
References
[1]
[2]
[3]
[4]
[5]
Richtlinie 2002/91/EG des Europischen Parlaments und des Rates vom

16. Dezember 2002 ber die Gesamtenergieeffizienz von Gebuden,
Amtsblatt der Europischen Gemeinschaften Nr. L1 pp. 65.
Stopp, H.; Strangfeld, P.: The hygrothermal performance of external walls
with inside insulation. In proceedings of the conference, Performance of
exterior envelopes of whole buildings VIII: Integration of building
envelopes. Clearwater Beach FL, U.S.A., Dec.2001.
Grunewald, J.: Diffusiver und konvektiver Stoff- und Energietransport in
kapillar-porsen Baustoffen. PhD thesis, TU Dresden, Germany, 1996.
Restoration of historical buildings. EnEV: possibilities and limits.
WTA:Wissenschaftlich-Technische
Arbeitsgemeinschaft
fr
Bauwerkserhaltung und Denkmalpflege e.V., D-80686 Mnchen,
Germany, 2002.
Stopp,. H.; Strangfeld, P. et al.: Heizungstechnisch gesttzte kapillaraktive
Innendmmung bei Holzbalkendecken. Bau- und WohnforschungBericht
F 2431, Fraunhofer IRB Verlag 2004.
539
Bond repair of cracked beams

H. Cruz1, J. Custdio1 & D. Smedley2
1
Structures Department, Wood Structures Division,

Laboratrio Nacional de Engenharia Civil, Portugal
2
Rotafix, Rotafix Ltd., Rotafix House, Abercraf, Swansea, UK
Abstract
The repair of cracked timber beams on site frequently involves crack injection
with adhesives followed by the application of the chosen strengthening system,
involving bonded or bolted metal plates or rods or bonded pultruded GFRP or
CFRP profiles. Although strengthening design will not generally take into
account the contribution of injecting adhesives in the crack volume region it is
felt that this contribution to overall strengthening of the component or structure
may be of some significance. A test programme was therefore developed at
LNEC to assess the strength and stiffness of timber beams which were treated
with epoxy adhesives after they had developed failure in a 4 point bending test. It
is recognised that partially broken beams may have very narrow, almost
invisible, fissures, which are difficult to impregnate. As well as the forgoing
partially separated broken beams may present extremely irregular split ends,
which are difficult to bring back into perfect alignment and contact. These two
factors may reduce the effectiveness of the strength restoration and were
considered in the experimental study. However, the specific beam failure
appearance is not likely to give a clear indication of how efficient the
consolidation turns out to be. Test results show that epoxy adhesive injection of
cracked timber beams may lead to a not insignificant, but somewhat variable,
strength recovery (0.45 to 1.11) and to a significant stiffness recovery (0.74 to
0.99) of broken beams. Having these results in mind, although on site repair of
broken timber beams should not rely on epoxy adhesives injection only, its
contribution to the stiffness of strengthened beams is significant and one should
make use of this advantage.
Keywords: cracked timber beams, structural adhesives, consolidation, strength
and stiffness recovery.
doi:10.2495/HPSM06053
Introduction
Old timber structural members may present failure due to biological damage,
which is responsible for timber strength reduction or cross section loss, or due to
unacceptable stresses resulting from inadequate design or from structural
accidental or intentional modifications. Furthermore older structures intended for
a particular purpose are currently being subjected to types of use for which the
original design parameters were not intended.
Although the replacement of damaged members by new timber or other
materials may seem the easiest solution, on site repair may be a better cost option
in cases where only a small timber length is damaged, or when that member is
part of a complex structural system and its removal may jeopardise decorative
ceilings and/or limit the occupancy of the building, Even the replacement of one
carrier beam may with traditional and well-proven carpentry techniques involve
the whole or partial removal of a roof structure with its corresponding time and
cost delay. In fact removal of sound surrounding components often puts hitherto
undamaged and adequate parts of the structure at risk.
The repair of cracked timber beams on site frequently involves crack injection
with adhesives followed by the application of the chosen strengthening system,
namely involving bonded or bolted metal plates or rods or bonded glass fibre
profiles. It is interesting to note that in the early days of injection technology
with timber (borrowed from Concrete repair practice in the late 1950s early
1960s) attempts were made to restore overstressed classical truss rafter systems
with inappropriate low viscosity 2 part epoxy so-called injection systems (Ref
Aberdeen Music Hall) - the lack of success resulted in a secondary repair method
using bolted U beam sections.
Current strengthening design will not generally take into account the
contribution of injecting adhesives in the cracked area, so-called consolidation. It
is felt that this contribution might be significant. There was also the question as
to whether it would be possible to identify in advance some typical failure
configurations likely to benefit in practice from the so-called consolidation
technique.
A test programme was therefore developed at LNEC with the intention of
assessing the strength and stiffness of timber beams which were consolidated
with epoxy adhesives after they had developed failure in 4 point bending tests.
References to studies involving timber crack injection could not be found in
the literature, although they have some similarity with glued timber delamination
repair. For the specific case of crack consolidation, with complex irregular
surfaces of the adherend it is necessary to find adhesives with suitable rheology.
Thick gap-filling adhesive which usually contain fillers are not deemed to be
suitable, whereas very fluid adhesives will penetrate more easily but will not be
capable of filling small voids left inside the timber member, reducing bonding
efficiency at those points. This was observed in previous trials by the authors [1],
where a poor crack consolidation (and corresponding low strength recovery) was
obtained by the combined use of a thick gap-filling adhesive followed by
541
injection with a very fluid one. To avoid this problem, a thixotropic adhesive was
used in the current experimental programme.
Another peculiarity of this kind of repair lies in the fact that preparation of
clean bonding surfaces is not possible. Wood surface oxidation and any possible
air borne or liquid contamination will have taken place since the time between
fracture and repair intervention could be days or even years depending upon
discovery of the fracture and application of the chosen remedial solution.
Materials and method
Test specimens consisted of sixteen 1500x120x70 mm Maritime pine (Pinus

pinaster, Ait.) timber beams. Timber defects like knots were within the limits of
the EE strength grade defined in NP4305 [2], although no restrictions concerning
the presence of pith and the associated juvenile wood (which is not allowed in
grade EE) were imposed.
Two series of beams were studied: 1 to 10 and A to F. Every beam was first
tested up to failure, in 4-point bending, to evaluate its maximum strength and
modulus of elasticity. Moisture content of timber varied between 12% and 14%
(mean value = 13%).
After a period of time (making approximately nine months for beams 1-10
and fifteen months for beams A-F), each beam was consolidated with the
thixotropic epoxy adhesive standard Slow Set CB10T, from ROTAFIX. No
regularization of the split ends was done and the beams were brought back to the
initial (straight) shape by using clamps and steel bars which were removed after
adhesive cure.
Two months later, the consolidated beams were subjected to a 4-point
bending test again. The same tension edge was chosen in both the bending tests,
before and after consolidation.
Moisture content of timber at the time of the second bending test was checked
with the electrical moisture meter and it was seen to be similar to their initial
value.
Bending tests were carried out in general accordance with the specifications
of EN408 [3].
The 1.50 m long specimens, simply supported over a span of 1.40 m, were
symmetrically loaded in bending at two points 0.50 m apart. Small steel plates,
5 cm large, were inserted between the test piece and the loading heads, and
between the test piece and the supports, to minimise local indentation (figure 1).
The total load F (equal to the sum of the vertical loads at the two load points)
was applied at a constant displacement rate of 1 cm/min. Total deformation
(wtotal) was measured at mid span, at mid depth of the cross section.
Bending strength and elasticity modulus of each beam were to be assessed in
a single test; therefore, the referred displacements were measured till failure.
The global modulus of elasticity in bending was determined by applying the
relevant equations in EN 408 [3].
Figure 1:
Test set up and specimen being tested in bending.
Test results and discussion
Table 1 presents individual test results in the first (initial sound state) and in the
second (consolidated) round of tests, concerning the global modulus of elasticity
and bending strength.
In annex A are presented some examples of the beams after the second round
of bending tests. The fibre separation lines on both faces and bottom edge are
enhanced on the photos (fine lines for old cracks before consolidation, thick
lines for cracks developed on the consolidated beams).
Load-deformation diagrams obtained for each beam, in both rounds of
bending tests (initial state and consolidated beam), are presented in the annex B.
The average recovery of strength due to consolidation with the epoxy
adhesive was 0.72 (ranging from 0.45 to 1.11). In most cases, the initial part of
the load-displacement test diagrams obtained before and after consolidation were
very similar. Therefore, similar global modulus of elasticity was obtained 0.90
efficiency was obtained with individual beams (ranging from 0.74 to 0.99).
This seems to indicate that consolidation was effective till a certain point
under moderate loads although it was not enough to recover the performance of
the solid timber beam for stress levels closer to the initial bending strength.
This is confirmed by the fact that bending failure in the consolidated beams
generally affected the same material, that is to say that it occurred in the same
areas where the initial beam had previously failed.
There is not a clear correspondence between crack pattern and efficiency of
the repair: apparently similar cracks have conducted to distinct results within the
range of strengthening values obtained.
Despite the small number of beams in each group, there is no clear distinction
between groups 1-10 and A-F. In both cases, the application of the epoxy
543
adhesive took place several months after failure, thus simulating the application
of adhesive to old timber surfaces.
Table 1:
Global modulus of elasticity and bending strength for initial beams

and beams after consolidation (ini/cons).
Beam
Modulus of Elasticity (MPa)

Econs
Econs/Eini
Fini
Fcons
Fcons/Fini
11921
9684
0.81
29.88
14.00
0.47
16193
15504
0.96
56.99
63.13
1.11
17371
14680
0.85
45.74
20.77
0.45
12479
12033
0.96
43.93
33.18
0.75
12504
10660
0.85
31.49
20.93
0.66
13378
12843
0.96
41.10
42.48
1.03
11706
(1102)
32.68
6.11
15261
15105
0.99
40.42
33.84
0.84
14431
12162
0.84
44.39
24.26
0.55
10
15683
14989
0.96
50.40
27.49
0.55
14169
10519
0.74
59.05
43.54
0.74
13364
11668
0.87
44.54
36.68
0.82
14183
13913
0.98
51.66
27.49
0.53
Mean (1-10)
Maximum Load (KN)
Eini
0.91
0.71
12636
11400
0.90
48.39
23.42
0.48
13565
12345
0.91
58.67
55.30
0.94
Conclusions
It is recognised that partially broken beams may have very narrow, almost
invisible, fissures, which are difficult to impregnate. Together with the fact that
partially separate broken beams may present extremely irregular split ends,
which are difficult to bring back into perfect contact. These two factors may
reduce the effectiveness of the consolidation and were considered in the
experimental study. However, the specific beam failure appearance is not likely
to give a clear indication of how efficient the consolidation turns out to be.
Nevertheless, test results show that epoxy adhesive injection of cracked
timber beams may lead to a not negligible, but somehow variable, strength
recovery (0.45 to 1.11) and to a significant stiffness recovery (0.74 to 0.99) of
broken beams.
Having these results in mind, although on site repair of broken timber beams
should not rely on epoxy adhesives injection only, its contribution to the stiffness
of strengthened beams is significant and should not be ignored.
Annexes
Annex A
Beam 1
Beam 3
Beam 6
Beam E
545
Annex B
547
Acknowledgement
This work was carried out in the scope of the European Research Project
CRAFT-1999-71216 LICONS Low Intrusion Conservation Systems for
Timber Structures (www.licons.org), which had a much wider scope.

The authors wish to thank STAP (www.stap.pt) for carrying out crack
consolidation and ROTAFIX (www.rotafix.co.uk) for providing the thixotropic
adhesive used in this study.
References
[1] COLORETIM CT 98-9548, COmposites LOcal REinforcement for TIMber
structures, Final Report Consolidating tests: bending tests on timber
beams strengthened with FRP.
[2] Norma Portuguesa NP 4305: Madeira serrada de Pinheiro bravo para
estruturas. Classificao visual Visual strength grading of maritime
pine.
[3] European Standard EN 408: Timber structures Structural timber and
glued laminated timber Determination of some physical and mechanical
properties. Draft revision, November 1999.
549
Microtexture and nanoindentation study of

delamination cracking in Al-Cu-Li-X alloys
R. Crooks1, M. S. Domack2 & J. A. Wagner2
1
2
National Institute of Aerospace, Hampton, Virginia, USA

NASA Langley Research Center, Hampton, Virginia, USA
Abstract
Commercial Al-Li alloys have strength and weight advantages over non-Al-Li
alloys. The fracture behavior of these alloys is unusual and has limited their use.
The fracture mode, described as delamination, is intergranular, along the broad
grain boundaries parallel to the rolling plane of the plate. Microtexture analyses
have shown that delaminations occur along boundaries with greater than 30
misorientation. However, it was observed that relatively few of the high angle
boundaries exhibited this behavior. Some grains of the retained deformation
texture show high internal misorientation, which is a measure of stored strain
energy. Delamination tends to occur between these grains and adjacent,
recrystallized grains. Nanoindentation studies indicate a higher hardness for the
high internal misorientation grains. These results suggest that the delamination
could be reduced by processing the alloys to minimize grain-to-grain property
disparities.
Keywords: microtexture, nanoindentation, Al-Li, delamination.
Introduction
Fracture in aluminum aerospace alloys typically occurs by ductile rupture and/or

transgranular shear. Al-Li alloys saw renewed interest in the 1980s as a low
density, high strength alternative to conventional aerospace Al alloys [1]. In
addition to exhibiting ductile rupture and transgranular shear, many Al-Li plate
alloys in the T8 temper (stretched and age hardened), exhibit delamination
fracture. The intergranular fracture in these alloys is referred to as delamination
because fracture along the characteristically thin, elongated grains resembles the
separation of a laminated material. When the primary crack is perpendicular to
doi:10.2495/HPSM06054

the rolling plane in the LT or TL orientation, the triaxial stress state near the
crack tip can be sufficient to separate grain boundaries susceptible to
delamination.
The propensity of Al-Li alloys to delaminate has limited their use in
aerospace applications and is due in part, to the absence of aerospace design
tools to account for delaminations. Development of mechanistic models that
describe delamination fracture behavior could help guide the modification of
ingot processing to reduce the tendency for this type of fracture, thereby
allowing designers and fabricators to realize the full benefits of Al-Li alloys.
There are several theories to explain this tendency for intergranular fracture.
Based on recent journal publications, there are at least three current and
competing explanations for this behavior. These can be described as planar slip,
grain boundary precipitation, and grain boundary segregation.
The planar slip explanation has been strongly advocated by Sanders and
Starke [2] and Csontos and Starke [3]. The reasoning is that intense planar slip
bands arise due to the presence of shearable precipitates, e.g. the , Al3Li phase.
Shearing of the precipitates reduces the resistance to dislocation glide within the
slip system and promotes planar slip. Planar slip bands cause intense stress
concentrations at grain boundaries and promote intergranular fracture. Planar slip
and precipitate shearing also occur in Al-Li-X alloys where the , Al3Li
precipitates are not the dominant strengthening phase. This is evident in the
shearing of the higher strength ternary T1, Al2CuLi precipitates [3-5] and may be
related to disruption of order within the Li-containing matrix.
An unusual feature of these alloys is the presence of large, grain boundary
precipitates which cover large area fractions of high angle grain boundaries.
Vasudevan and colleagues [6-9] have published several studies which conclude
that the grain boundary particles are responsible for the relative weakness of the
boundaries.
Lynch et al. [10, 11], on the other hand, maintain that intergranular fracture in
Al-Li-X alloys is due to grain boundary segregation of Li. Lithium segregation is
postulated to cause a modification of grain boundary structure and a lowering of
grain boundary cohesive strength.
The publication of several conflicting explanations for low energy
intergranular fracture of these alloys shows a lack of conclusive evidence for a
delamination fracture mechanism. In this study some relatively new analytical
techniques, namely microtexture and nanohardness, have been used to obtain
more precise information on the grains adjacent to the fracture path of
delamination cracks in the Al-Li alloy 2090. The grains in unrecrystallized AlLi-X alloy plate tend to be less than 20 m thick, so a plate of 1 cm thickness
would have 500 or more grains in the thickness direction. Delamination is,
however, limited to a small fraction of boundaries (Figure 1). The properties and
microstructures of the grains adjacent to the delaminations may help provide a
more accurate description of the fracture mechanism.
Microtexture, as a commercially available technology, is a development of
the last decade [12, 13]. Microtexture studies rely on the collection of grain
specific orientation information from polished sections of materials. The
551
information is collected in a scanning electron microscope equipped with an

electron backscattered pattern detector (EBSD) system, and uses a computer
interface to position the electron beam at designated points on a tilted, polished
sample; collect diffraction patterns with a phosphor screen coupled to a low-light
camera and a series of computer algorithms to determine the crystallographic
orientation. Data can be collected with spatial resolutions better than 0.1 m and
at rates better than 50,000 points per hour. Most data is collected in an automatic
mode over a grid of selected size and point separation, and the data can be
displayed as maps showing various aspects of the crystallite orientation, the
boundary angle between grains, and even the extent of deformation of
unrecrystallized grains.
Thickness, 0.95 cm
Delamination
Side
Groove
Figure 1:
Side
Groove
Cross section of a fractured specimen of Al-Li, showing

delamination cracks. Growth direction of primary crack is into the
plane of the page. Microtexture results show that delaminations
occur along relatively few high angle boundaries.
In earlier work [14, 15], microtexture methods were used to examine the
misorientation angles of delaminating boundaries, the possible role of special
boundaries, and similarities between boundary planes and precipitate planes.
These studies showed that delamination occurs at high angle boundaries in
excess of 30 misorientation. No conclusive correlation was found with grain
boundary types, crack path and precipitate distributions [15]. However, Kalu and
Wagner [14] showed that delaminations in alloy 2090 tended to occur adjacent to
grains of the {110}<112>, Brass texture component orientation. The Brass

component is an extremely frequent orientation of the rolling texture in this
alloy.
Nanoindentation is another recently developed technique, with a history
including the last twenty or thirty years [16]. With currently available
instruments it is possible to determine hardness and Youngs modulus from
extremely small regions. Loads as small as 1 nN and displacements as small as
0.1 nm can be measured. Measurements of other properties, e.g. local yield
stress, strain-hardening and creep behavior are also possible. Recent
nanoindentation studies of aluminum alloys have examined the micromechanical

properties of particle reinforced aluminum matrix composites [17, 18], and the
properties of precipitate free zones in Al-Zn-Mg alloys [19].
In this study, nanoindentation was used to explore the grain-to-grain property
variations of the Al-Cu-Li-Mg-Zr alloy 2090. Results were correlated with
microtexture studies which characterized both the location of delaminations and
the variation in recovery and recrystallization in the alloy plate.
Experimental
Fractured 2090 compact tension samples were sectioned normal to the fracture
surface and crack direction in a region containing several delaminations.
Sectioning was performed with a slow speed diamond saw. Fracture surfaces
were protected by painting with an acetone soluble stop-off lacquer (Microstop),
mounted in 1.25 diameter epoxy mounts and metallographically polished with a
specimen load of ~ 4lbs./sample. The polishing procedure involved the following
abrasives and times: 600 grit SiC until flat, 1200 grit SiC for 15 minutes, 3 m
Diamond paste with Struers Blue lubricant for 30 minutes on Buehler Texmet
2000 cloth, 1 m Diamond paste with Struers Blue lubricant for 30 minutes on
Buehler Texmet 2000 cloth. Final polish used a 1:1:2 mixture of Stuers colloidal
silica (OP-S), soap (Liquinox) and water. Final polishing was for 1 hour on
Buehler Mastertex cloth. Samples were then freed from the epoxy and cleaned in
acetone and alcohol with ultrasonic agitation. This method provided a flat, fairly
strain-free surface, which was well suited for collection of microtexture data
along the delamination cracks. The surface quality was comparable to that
obtainable from electropolishing methods, but with a much flatter surface. A flat
surface is important since even the minor undulations common to electropolished
samples interfere with the microtexture data collection at the required 70 of tilt,
especially in vicinity of cracks. Grains on either side of a delamination were
studied via microtexture.
A JEOL 6400 scanning electron microscope (SEM) equipped with a
microtexture system was used to study microtexture and the distribution of
nanoindents.
Microtexture data were collected in large sets of ~ 40,000 solved diffraction
patterns per set, with a spatial separation of 0.5 m. Analyses of the data
included grain boundary misorientation; internal misorientation; discrete texture
from points of the data set; identification of grains by proximity to specific,
characteristic orientations (or texture components); and internal misorientation.
Internal misorientation is a measure of the spread of orientations bound by a high
angle boundary; i.e. it is essentially a quantification of the amount of
substructure. Since the recrystallization process removes substructure, a low IM
is an indication of a recrystallized condition. The data were plotted as colorcoded maps, where appropriate. Data were collected from around delamination
cracks and from undeformed material. A simple method of indicating the grain
orientation will be used in this paper, namely showing the distribution of indexed
points on a discrete pole figure. Discrete pole figures associated with
microtexture maps can be used to depict the grain orientations.
553
Regions studied by microtexture were also evaluated by nanoindentation.

Nanoindentation was performed with an MTS Nano Indenter XP system. Low,
constant maximum load indents (1mN) were applied with a Berkovitch diamond
indenter. Nanoindentation studies required use of an SEM for examination of the
individual nanoindentations. The location of nanoindent rows in the SEM was
facilitated by placement of two indents with a much larger load. Single rows of
ten nanoindents were applied across a delamination, with a spacing of 5 m, and
the hardness values were calculated by the method of Oliver and Pharr [16].
Results from microtexture studies of delamination cracks and undeformed
material were compared with nanoindentation measurements straddling
delamination cracks.
High
IM
Figure 2:
Internal misorientation (IM) map from a wide area microtexture

scan of a region below a delamination crack in a L-T, J-R, 2090,
compact tension sample. The range of grey levels corresponds to
an IM range of 0 to 5. The indicated grain shows a high IM and is
within the retained deformation texture brass orientation.
Microtexture study of undeformed material (1 cm below a delamination crack)

revealed a wide variation in internal misorientation from grain to grain. The
contrast was particularly evident between grains of the retained deformation
texture and neighboring, recrystallized grains with dramatically different levels
of IM. The indicated grain in Figure 2, for instance, shows an internal
misorientation near 5, whereas some grains show IM values near 0.
Examination of several delaminations indicated common features which were,

presumably, related to the fracture process. There was a tendency for fracture to
occur between grains with large differences in IM. Low internal misorientation
grains were expected to be soft due to recrystallization. Such grains were
found adjacent to delaminations either as single grains or small necklace
grains. The other side of the delamination was found to be a large,
unrecrystallized grain with a greater amount of work hardening. This
unrecrystallized grain was typically larger, with a much larger internal
misorientation than the neighboring grains. These grains were presumed to be
harder prior to aging.
a
GOSS
Figure 3:
Cu
Microtexture plots showing internal misorientation of grains

adjacent to a delamination crack (arrows) in 2090-T8. From a T-L,
J-R, compact tension sample. a) Image Quality (IQ) map, b)
Internal Misorientation (IM) map, c) schematic of (111) pole figure
showing ideal Goss and Cu orientations and d) discrete (111) pole
figure from the data of b). Darker grain is near a Cu deformation
component orientation. Region below the crack is in the Goss,
recrystallized orientation.
An example of this effect is shown in Figure 3, which shows microtexture

maps of delamination fracture between a Goss, recrystallized grain, and a grain
near the Cu deformation component orientation. The unrecrystallized (Cu) grain
555
shows an IM of near 5, which is about twice as high as the Goss grain across the
crack. Large internal misorientation was taken as an indication of high stored
strain energy, such as would be expected for a grain of the deformation texture.
Low internal misorientation was an indication of recrystallization subsequent to
rolling, which is likely to have developed during solution heat treatment of the
sheet prior to artificial aging.
For sheet or plate to be given a T8 aging treatment, a stretch of about 6 8%
is applied subsequent to solution heat treatment and prior to aging.
Recrystallized grains (without substructure) are then given a significant
dislocation density which aids in precipitate nucleation. The increased density of
precipitates increases the strength considerably. However, the grains with well
developed substructure should exhibit precipitation on subgrain boundaries, and
should develop an even higher dislocation density due to dislocation
multiplication effects during deformation. These grains may develop greater
hardness during aging. A nanoindentation study was used to investigate this
possibility. Figure 4 shows a series of nanoindents traversing a delamination
crack. The grain on the right side of the crack, which was near the Brass
orientation, showed a hardness increase of about 20% greater than the grain
across the crack. This was despite the indent being at the very edge of the crack,
which might be expected to result in a low hardness value.
Figure 4:
Scanning electron micrograph of a delamination crack and

nanoindents. Longitudinal (LS) plane from a 2090 L-T, J-R test.
Nanohardness indents are shown along with larger marker indents.
A nanohardness increase, of about 20%, was found for one point
within a grain adjacent to the delamination (arrow). The higher
nanohardness grain shown has a locally high IM and is close to the
Brass orientation.

Delamination frequently occurs along grain boundaries which are interfaces
between hard and soft grains. Measurements of IM, texture components and
direct measurements of nanohardness support this observation. This may be due
to the tendency of the softer grains to preferentially deform and suffer from poor
accommodation of plastic distortion at the grain boundary, since the minimum
resolved shear stress would be greater in the harder grain. An additional factor
may be the tendency of recrystallized grains to end their growth at obstacles,
such as incoherent intermetallic particles, which establish weak interfaces at the
new grain boundaries.
Conclusions
The following conclusions can be reached from this study:

Delamination occurs preferentially along interfaces between hard and soft
grains, representing, in the first case, grains of the deformation texture with high
stored strain energy and in the second case, recrystallized grains. This is
supported by 1) texture information, which can be compared to well-known
sequences in the processing history, 2) measurements of the intragranular
orientation variation and 3) direct measurements of the nanohardness adjacent to
delamination fractures.
Higher hardness is attributed to stored strain energy of some components of the
unrecrystallized material vs. the softer condition of the recrystallized material.
This observation of fracture along the hard/soft interface does not eliminate any
one of the mechanisms proposed for delamination fracture, but adds yet another
factor.
The correlation of nanoindentation and microtexture data provide significant and
new insights into the mechanical properties of Al-Cu-Li-X alloys.
References
[1]
[2]
[3]
[4]
Sanders, Jr., T.H., & Starke, Jr., E.A., (eds). Al-Li Alloys I, Proceedings of
the First International Aluminum-Lithium Conference, Stone Mountain,
Georgia, 1980.
Sanders, Jr., T.H. & Starke, Jr., E.A., The effect of slip distribution on the
monotonic and cyclic ductility of Al-Li binary alloy. Acta Metall., 30, p.
927, 1982.
Csontos, A.A. & Starke, Jr., E.A., The effect of inhomogeneous plastic
deformation on the ductility and fracture behavior of age hardenable
aluminum alloys. International Journal of Plasticity, 21, pp. 1097-1118,
2005.
Howe, J.M. & Vasudevan, A.K., Structure and deformation behavior of T
sub 1 precipitate plates in an Al-2Li-1Cu alloy. Metall. Trans. A, 19A, p.
2911, 1988.
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
557
Crooks, R., Wang, Z., Levit, V.I. & Shenoy, R.N., Microtexture,
microstructure and plastic anisotropy of AA2195. Materials Science and
Engineering A, A257(1), pp. 145-152, 1998.
Jata, K.V. & Vasudevan, A.K., Effect of fabrication and microstructure on
the fracture initiation and growth toughness of Al-Li-Cu alloys. Materials
Science and Engineering A, A241, pp. 104-113, 1998.
Suresh, S., Vasudevan, A.K., Tosten, M. & Howell, P.R., Acta Metall.,
35, p. 25, 1987.
Vasudevan, A.K. & Doherty, R.D., Acta Metall., 35, p. 1193, 1987.
Vasudevan, A.K., Ludwiczak, E.A., Baumann, S.F., Howell, P.R.,
Dougherty, R.D. & Kersker, M.M., Mater. Sci. Technol., 2, p. 1205, 1986.
Lynch, S.P., Materials Science and Engineering A, A136, p. 25, 1991.
Lynch, S.P., Muddle, B.C. & Pasang, T., Ductile-Brittle Fracture
Transitions in 8090 All-Li Alloys. Acta mater, 49, pp. 2863-2874, 2001.
Adams, B.L., Orientation Imaging Microscopy: Emerging and Future
Applications. Ultramicroscopy, 67, pp. 11-17, 1997.
Schwartz, A.J., Kumar, M. & Adams, B.L., Electron Backscatter
Diffraction in Materials Science, Kluwer Academic/Plenum, New York,
2000.
Kalu, P.N. & Wagner, J.A., A Microtexture Investigation of the Fracture
Behavior of Al-Li Alloy, 2090-T81. The 3rd Light Weight Alloys for
Aerospace Applications, ed. E.W. Lee, TMS, Warrendale, PA, pp. 157167, 1995.
Hales, S.J. & Crooks, R., Recrystallized and unrecrystallized Al-Li alloys
for elevated temperature service. Proc. 3rd International Conf. on
Recrystallization and Related Phenomena, ed. T.R. McNelley, MIAS, pp.
511-518, 1997.
Oliver, W.C. & Pharr, G.M., An Improved Technique for Determining
Hardness and Elastic Modulus Using Load and Displacement Sensing
Indentation Experiments. J. Mater. Res., 7(6), pp. 1564-1583, 1992.
Torralba, J.M., Velasco, F., Costa, C.E., Vergara, I. & Caceres, D.,
Mechanical behaviour of the interphase between matrix and reinforcement
of Al 2014 matrix composites reinforced with (Ni3Al)p. Composites: Part
A, 33, pp. 491-496, 2002.
Liu, C., Qin, S., Zhang, G. & Naka, M., Micromechanical properties of
high fracture performance SiCp-6061/6061 Al composite. Materials
Science and Engineering A, A332, pp. 203-209, 2002.
Ogura, T., Hirosawa, S. & Sato, T., Quantitative characterization of
precipitate free zones in Al-Zn-Mg(-Ag) alloys by microchemical analysis
and nanoindentation measurement. Science and Technology of Advanced
Materials, 5, pp. 491-496, 2004.
559
A contribution to the rehabilitation of

reinforced concrete structures by
non-destructive electrochemical methods
J. L. Rovira Santa Olaya & P. Pardo Trfach
Department of Construction Engineering, Materials Division,
Universitat Politcnica de Catalunya, Barcelona, Spain
Abstract
The aim of this work has been the study of an electrochemical treatment that
allows the prevention and halting of chloride corrosion processes in passive
reinforced concrete. This treatment is based on a desalination system that
removes free chloride in concrete simultaneously increasing the pH around
reinforcement steel until its repassivation.
With this purpose a concrete has been manufactured with mixing proportions
close to that used in civil engineering. The chloride attack has been carried out at
room temperature, and the chloride removal process has been started at several
concrete ages. Chloride ions have been determined before and after the attack, as
well as the corrosion potential, in order to establish comparison between the
grade of corrosion of the reinforcement steel before and after treatment,
according to exposition time and initial corrosion situation. Relationships
between electrochemical parameters have been determined.
The studied procedure is a non-destructive and non-intrusive method that
does not change either the strength of the concrete or its external appearance.
The procedure uses non-pollutant materials, so no environmental damage is
caused.
Keywords: electrochemical chloride extraction, steel-reinforced concrete,
repassivation.
Introduction
Corrosion in steel reinforcement caused by exposition to chlorides is the main

cause of deterioration in steel-reinforced concrete structures. Electrochemichal
doi:10.2495/HPSM06055

chloride extraction (ECE, also named desalination) is a method to repair steel
reinforced structures that have suffered chloride attack [1, 2], although it can be
also applied as a prevention treatment [3].
The treatment consists of the application of an electric current between the
steel, acting as a cathode, and a temporary external anode placed on the concrete
surface. As reported by Arya et al. [4], chloride removal increases with
increasing applied potential, number of reinforcing bars at a particular depth and
initial chloride content, although chloride remaining seems to be independent of
the initial chloride content. This method requires the chloride ions to be in the
pore water to be transported out of the concrete [2]. Removal of 40 to 70% of the
initial chloride has been observed using this technique [1, 5, 6]. On the other
hand, repassivation of steel due to the increase in the potential of corrosion after
the chloride removal has been studied by several authors [5, 6, 7].
In this work, electrochemical chloride removal has been studied on reinforced
concrete that has been subjected to different degrees of corrosion. The
measurement of the potential of corrosion has allowed the probability of
corrosion to be determined before and after the treatment according to these
values for the potential [8]:
- Higher than -200 mV: low probability (0-5%)
- Between -200 and -350 mV: medium probability (5-50%)
- Lower than -350 mV: high probability (50-100%)
Finally, the time period during which the treatment should be maintained to
ensure the rehabilitation of concrete has been studied.
26 cm
11 cm
Figure 1:
Disposition of the reinforcement steel in the test specimens.
Materials and methods
2.1 Production of tests specimens

40 cylindrical standard test specimens were cast and molded according to UNE
EN 83301:91 (150 mm in diameter and 300 mm in length). 33 of the specimens
were reinforced with eight steel bars (AEH-400) connected by two circular
fences as it is shown in fig. 1. The steel surface was mechanically cleaned before
561
use. In order to accelerate the corrosion process, 2 cm of covering layer was

selected to cover the steel bars. This must not affect the results, since the goal of
this work is a comparative study between several corrosion and regeneration
conditions. The concrete composition, shown in table 1, corresponds to a typical
mixing for building purposes. Steel characteristics are also shown in table 1.
40 test specimens
7 control specimens
(without reinforcement)
33 reinforced specimens
11 YY-specimens
11 XX-specimens
1 control specimen
chloride attack (7d)
11 ZZ-specimens
1 control specimen
1 control specimen
1 control specimen
1 control specimen
1 control specimen
ECE
(7d; 3 specimens)
ECE
(7d; 3 specimens)
ECE
(7d; 3 specimens)
ECE
(21d; 3 specimens)
ECE
(21d; 3 specimens)
ECE
(21d; 3 specimens)
ECE
(42d; 3 specimens)
ECE
(42d; 3 specimens)
ECE
(42d; 3 specimens)
Figure 2:
Schedule followed for the experiment.
Test specimens were compacted in a shaking table and demolded after 24

hours. Then, they were maintained in a wet room for six days. After this period,
they were subjected to chloride attack and electrochemical treatment following
the schedule shown in fig. 2.
2.2 Chloride attack
Chloride attack was carried out in two pools of 350 l, containing a solution with
150 g NaCll-1 (52 g Cl-l-1). Water evaporation was controlled, maintaining
chloride concentration within a 2% variation.
2.3 Electrochemical chloride extraction treatment
In order to save time, the treatment for chloride removal was applied to batches
of nine specimens at a time. A current intensity of 1 Am-2 with a potential of 20
V was supplied and maintained constant during the required period of time. Fig.
3 shows the electrochemical chloride extraction system for a specimen. A first 2
cm layer of wet cellulose paste covered the specimen. This layer acts as an
exchange electrolyte between reinforcement bars and a titanium anode mesh.
Finally, the whole system was coated with a second 1.5 cm wet cellulose layer in
order to maintain the wet conditions.

2.4 Potential measurements
Corrosion potentials were measured using an in situ measurement system
Colebrand with a Cu/CuSO4 reference electrode.
Table 1:
Concrete composition and steel characteristics.
Concrete
Cement CEM I 42,5R
Calcareous Sand 0/5
Calcareous gravel 5/12
Calcareous gravel 12/18
Water
w/c
Compressive strength 28d
Steel bars AEH-400N
Diameter
Modulus of elasticity
Breaking stress
346 kg/m3
1289 kg/m3
521 kg/m3
739 kg/m3
225 L/m3
0.65
45 MPa
3 mm
4100 kp/cm2
4500 kp/cm2
Steel reinforcement (cathode)

Concrete
First wet cellulose layer (2 cm)
Titanium anode mesh
Second wet cellulose layer (1.5 cm)
Figure 3:
Electrochemical extraction system.
For a better monitoring of the corrosion and electrochemical removal processes,

the corrosion potential was measured in two points of each of the two bars which
constituted the reinforcement structure, at 1/3 and 2/3 height from the specimen
basis. The measurements were taken before and after electrochemical treatment.
Table 2 shows the increase in the potential of corrosion in each set of specimens
produced by the chloride attack.
563
3.1 X-specimen set (7 day chloride attack)

The attack on this set was a low intensity attack, simulating an initial stage of
corrosion or a concrete with an excess of chloride. Fig. 4 shows the evolution
with time of the potential of corrosion for this set. As it can be seen in the table
3, the corrosion potential decreases a 28.3% during the first week of the chloride
removal treatment. From the second week to the end of treatment, the reduction
in the corrosion potential was about 6-7% per week. The total decrease with
respect to the initial corrosion potential was 60.2% (from -283 mV to -112.7
mV).
For this set, the first week of treatment caused a reduction in the potential of
corrosion enough to reach the low corrosion probability zone, as it is shown in
fig. 4. This is due to the fact that the corrosion process was hardly initiated after
7 days to chloride exposition, and chloride ions had not penetrated deeply into
concrete. Consequently, chloride removal was carried out quickly.
From these results, it can be concluded that a four week removal treatment
would be enough for structures with a low corrosion degree.
Table 2:
Increase in the potential of corrosion for each set after attack.
Potential before attack

(mV)
Potential after attack
(mV)
Variation (%)
Table 3:
X-specimen set
(7 day attack)
-110
Y-specimen set
(56 day attack)
-110
Z-specimen set
(90 day attack)
-110
-283
-524
-621
157
376
464
Decrease of the potential of corrosion for each set after treatment.
Decrease 1st week of ECE

Decrease (per week) after 3
weeks of ECE
Decrease (per week) after 6
weeks of ECE
Cumulative decrease 1st
week of ECE
Cumulative decrease after 3
weeks of ECE
Cumulative decrease after 6
weeks of ECE
X-specimen
set
(7 day attack)
28.3
Y-specimen
set (56 day
attack)
13.3
Z-specimen
set (90 day
attack)
17.1
5.84
12.6
10.9
6.73
9.38
8.90
28.3
13.3
17.1
40.0
38.6
34.8
60.2
66.7
61.6

3.2 Y-specimen set (56 day chloride attack)
The attack on this set was a medium-high intensity attack, simulating a concrete
with a corrosion degree such that the structure is damaged, but not in a way to be
at risk for the moment. Fig. 4 shows the evolution with time of the potential of
corrosion for this set. As table 3 shows, the reduction of the potential of
corrosion for this set is maintained between 9 and 13% per week, decreasing
with a regular rate along the period of time in which the experiment was carried
out. The total decrease with respect to the initial corrosion potential was 66.7%
(from -523.8 mV to -174.3 mV).
For this set, the low corrosion probability zone was reached after six weeks
of treatment, although a few more weeks could be necessary to ensure the
concrete to stay in this low risk zone.
From these results, it can be concluded that a ten or eleven week removal
treatment would be needed to ensure a correct rehabilitation of structures with a
medium or high degree of corrosion.
Potential of corrosion (mV * -10)
70
X-specimen set
62,1
60
Y-specimen set
52,38
50
Z-specimen set
51,5
45,4
40,46
40
32,18
30
28,3
23,87
20,26
20
17,43
16,99
11,27
10
0
0
Week
Figure 4:
Evolution of the potential of corrosion during the electrochemical

removal of chloride. Horizontal line: 50% probability of corrosion.
Horizontal plotted line: 5% probability of corrosion.
3.3 Z-specimen set (90 day chloride attack)

The attack on this set was a high intensity attack, simulating a very damaged
concrete, which would imply its demolition if it was a structural concrete. Fig. 4
shows the evolution with time of the potential of corrosion for this set. As it is
shown in table 3, the reduction per week of the potential of corrosion for this set
decreases with time from 17.1% to 8.90%. The total decrease with respect to the
initial potential of corrosion was 61.6% (from -621 mV to -239 mV).
565
0,35
0,32
Cl- concentration (%)
0,30
0,28
0,25
0,22
0,20
0,15
0,12
0,10
0,05
0,00
0,1
0,08
Exterior
0,02
0,02
0,05
0,04
Interior
Week
0,60
0,51
0,50
0,44
0,40
0,33
0,29
0,30
0,20
0,10
0,00
0,19
0,17
Exterior
0,12
0,02
0,02
0,07
Interior
Week
0,60
0,57
0,50
0,6
0,48
0,40
0,35
0,34
0,30
0,21
0,20
0,17
Exterior
0,10
0,00
0,09
0,02
0,02
Interior
Week
Figure 5:
Evolution of chloride concentration inside and near surface.

For this set, after six weeks of electrochemical removal, the zone of medium
risk of corrosion (5-50% probability) is reached. An extrapolation of the data
obtained for this set, showed that 12 weeks of treatment would be needed to
reach the low risk of corrosion zone for the rehabilitation of highly damaged
concrete elements.
3.4 Evolution of chloride concentration
Chloride concentration was monitored both inside the test specimens (near
reinforcement) and near the surface of the specimen. Results for the three sets are
pictured in fig. 5.
Chloride concentration decreased between 65 and 71%, and the results were
in concordance with that obtained from the study of the potential of corrosion,
supporting the conclusions previously drawn.
Conclusions
After electrochemical removal treatment, chloride concentration decreased

between 65% and 71% in the three sets of specimens studied. The potential of
corrosion decreased about 60% in all cases, lowering the probability of corrosion
to less than 5% for specimens subjected to low and medium intensity chloride
attack.
Depending on the degree of the aggression by chloride ions, between 4 to 12
weeks of electrochemical treatment can be needed to ensure a correct
rehabilitation of concrete.
The results obtained have shown that electrochemical chloride removal is a
useful technique for rehabilitation of reinforced concrete structures subjected to
corrosion of steel reinforcement by chloride attack.
References
[1]
[2]
[3]
[4]
[5]
Elsener, B., Molina, M. & Bhni, H., Electrochemical removal of

chlorides from reinforced concrete structures. Corrosion Science, 35(5-8),
pp. 1563-1570, 1993.
Tritthart, J., Petterson, K. & Srensen, B., Electrochemical removal of
chloride from hardened cement paste. Cement and Concrete Research,
23(5), pp. 1095-1104, 1993.
Glass, G. K. & Buenfueld, N. R., The inhibitive effects of electrochemical
treatment applied to steel in concrete. Corrosion Science, 42(6), pp. 923927, 2000.
Arya, C., Said-Shawqi, Q. & Vassie, P. R. W., Factors influencing
electrochemical removal of chloride from concrete. Cement and Concrete
Research, 26(6), pp. 851-860, 1996.
Polder, R. B. Electrochemical chloride removal from concrete prisms
containing chloride penetrated from seawater. Construction and Building
Materials, 10(1), pp. 83-88, 1996.
[6]
[7]
[8]
567
Fajardo, G., Escadeillas, G. & Arliguie, G. Electrochemical chloride

extraction from steel-reinforced concrete specimens contaminated by
artificial seawater. Corrosion Science, in press, available on
www.elsevier.com/locate/corsci.
Elsener, B. Half-cell potential mapping to assess repair work on RC
structures. Construction and Building Materials, 15(2-3), pp. 133-139,
2001.
Andrade, C., Monitoring Techniques (Chapter 6). Corrosion of Steel in
Concrete, ed. P. Schiessl, Chapman & Hal: London and New York, pp.
95-79, 1988.
Section 7
Adhesion and adhesives
571
Timber specimens parametrized design for

numerical analysis
E. Martn Gutirrez, J. Estvez Cimadevila, D. Otero Chans &
S. Muiz Gmez
Department of Technology of Construction, A Corua University, Spain
Abstract
The characterization of materials or the study of resistant response associated
with new constructive solutions often convey progressive adjustment of models
of analysis starting from the results obtained by means of test campaigns. The
reliability of the conclusions will obviously be conditioned by the amplitude of
the experimental development which, as a result, should contain a sufficiently
representative sampling of cases. In the situations where significantly multiple
parameters intervene, specimens that involve possible values and combinations
should be examined, which in practice remarkably increases the number of
specimens to be treated. The case being presented deals with the evaluation of
the effectiveness of certain fixed joints carried out by gluing metallic bars into
timber pieces by means of adhesives of different formulations. The set of
calculus patterns, including the extensive variability of the associated
magnitudes, has been treated in a parametrized form. The geometry of the
studied pieces, as well as the sustenance conditions and the load, are
automatically generated by means of software specially designed for those
purposes. The resulting information is structured in an orderly way to facilitate
its reading and also its manipulation if needed. The files of the process obtained
in this way can optionally incorporate the meshing sequences, as well as the
commands of the calculus, analysis and postprocess. The adopted strategy
considerably reduces the time assigned to the phases of the geometrical
definition and resolution, and enormously simplifies the interpretation of the
results.
Keywords: numerical analysis, parametrized design, glued joints.
doi:10.2495/HPSM06056
Introduction
This work is placed in an investigation project directed to the study of the joints
behaviour in timber pieces by means of steel bars glued-in with an adhesive. The
approach causes the development of two parallel complementary lines of
advance: the first one is of experimental nature and the second one is based on
the use of the numerical analysis technique.
On the one hand it is pretended to use the results obtained in the experiments
for calibrating gradually the theoretical models in such a way that they, in their
turn, could be applied for extrapolating not specifically experimented situations
or facilitating the adoption of design decisions in certain practical cases. On the
other hand the numerical analysis can be useful for the selection of those
mechanisms of connection that prove to be more operative in advance, reducing
the temporary and economic costs associated with the creation of the specimens.
Figure 1:
Reconstruction of different specimens.
Specimens are timber prisms of square transversal section made of timber, at

both ends of which two threaded bars are placed, properly fixed by means of
adhesives of epoxi formulation (fig. 1). The resulting specimen is subjected to a
failure test of growing centered tensile load, registering the limit number and the
form of the dominant failure. Throughout the process the equipment registers the
total relative displacement between the ends of the specimen, associated to each
load value, which later permits to configure the corresponding cause-effect
diagrams. In principle a first experimental campaign consisting of a total of 45
pieces for each of the used adhesives is planned. The given number is distributed
in 3 series of 15 specimens in which the joint with 3 possible diameters (8, 10
and 12 mm) and 5 anchorage lengths are analyzed (table 1).
Table 1:
d
10
10 12
12 14
573
The structure of the first experimental campaign.

L
60
90
120
150
180
60
90
120
150
180
60
90
120
150
180
Figure 2:
Li
Lm
Li=L
3.L 500
60
90
120
150
140
60
90
120
150
140
60
90
120
150
140
180
270
360
450
500
180
270
360
450
500
180
270
360
450
500
Le
150
150
150
Lb
210
240
270
300
330
210
240
270
300
330
210
240
270
300
330
Reference
a=6.d
Units: mm
48
60
72
1a
1b
1c
1d
1e
2a
2b
2c
2d
2e
3a
3b
3c
3d
3e
Initial configuration of the specimens.
The attached diagram shows the proposed combinations for the 9 geometrical
parameters to deal with, as well as the type of the timber and the type of the
applied adhesives. This set appears to be sufficiently full for establishing
relevant conclusions in relation with the incidence of each magnitude and even
facilitating formulations that will permit to define the necessary anchorage
lengths.
The structure of the process file
The numerical models are calculated with the help of the application of Ansys
Multiphysics, Swanson Analysis Systems Inc., initially assuming an elastic and
lineal behaviour, noticeably correct in accordance with the experimentally
obtained stress-deformation diagrams for load values not surpassing the 75-80
per cent of the limit number.
The volume of the pieces, the possible parametrization of its geometrical and
mechanical values lead us to the development of a specific software that

automates the definition of the model, taking advantage of the common
commands of the preprocessor. It deals with the creation of a process file for
each specimen that Ansys could manage in an autonomous way but in which the
information would be structured in an orderly way with the purpose of
simplifying its reading and the possible manipulation. In general the mentioned
files respond to the following configuration: numerical definition of
parametrized data; meshing specifications; assignment of the load value; access
to the preprocessor of the system; description of the conditions of visualization
of the model; definition of elements and materials; an orderly localization of the
keypoints; positions that designate lines and arcs; creation of areas; construction
of volumes by means of extrusion operations; material assignment; meshing
operations; access to the calculus module; definition of coercion and application
of the load; access to the postprocessor of the application; energy estimation of
errors structured by materials; and presentation of results.
It should be paid attention to the fact that the file in question is defined in
ASCII format, therefore it turns out to be operable with any conventional line
editor. This circumstance allows to assess, with remarkable simplicity, diverse
alternatives of modelling, for example: typology of the involved finite elements;
adequacy of the parameters that define the mechanical behaviour of the
materials; variations in the conditions of the meshing and adjustment of
dicretization in concrete areas.
The model is configured by means of tridimensional elements of 8 nodes
(Solid45), that present 3 degrees of liberty in each position (translations x, y, z).
After diverse experiments the use of elements of 20 nodes has been discarded,
taking into consideration the fact that it greatly increases the storage
requirements and the time of the calculus without significant differences neither
in the results nor in the reductions in the energy estimations of the possible
committed errors. The associated materials are: steel 8.8 (fy=640 N/mm, fu=800
N/mm) for the threaded bars of the ends, sawn timber (basically chestnut) and 3
types of adhesives of epoxi formulation (Sika AnchorFix-3, Hilti Hit Re-500,
Loctite Hysol 9464A&B). All of them are integrated in the model by their
parameters of elastic behaviour; basically the modules of transversal and
longitudinal elasticity and the coefficient of Poisson in relation with the axes that
define the Cartesian system of reference (is designated as Z the relative to the
direction of the load application). The mentioned magnitudes are initially
decided on the basis of the recommended values of different references [1,2,6],
and considering the experiments previously made upon the isolated materials.
Later a successive adjustment is realized with a purpose of approximating the
obtained results by means of experimental and numerical technique.
Geometrical organization of the model
Attending to the considerations of symmetry it is decided to define a fraction

corresponding to the eighth part of the specimen. At first, the configuration of
the generic transversal section is established that will give place to the complete
volume of the modelling by repetition of its elements, totally or partially. In
575
principle, ample divisions are managed to tackle the first assessment of the
problem. Later comes the restructuring of the configuration on the basis of a
radial scheme that serves to a better estimation of the stress distribution in
different materials, especially in the areas close to the contact surfaces (fig. 3).
Figure 3:
Figure 4:
Sections and axonometry of the model.
Basic numeration of nodes, lines and arcs.
At the beginning of the process the coordinates (x, y) of the 105 points, that
define the base plane, are stored in matrices. In the area nearest to the centre of
the section the mentioned values are deduced by means of a polar system of
reference, while in the exterior panels a Cartesian type of analysis is proposed. It
should be paid attention to the fact that the parametrization of the process of

discretization permits the successive application of the adjustments that will
improve the behaviour of the model.
The generation program supposes that the specimen is formed by 3 well
differentiated areas (fig. 2): the central zone (zl), corresponding integrally to the
timber piece; the relative to the glueline (z2); and the external fraction of the
steel bar (z3). Each of these sets can be subdivided independently in the desired
number of bands. With these premises the definition of the model keypoints is
realized by means of the command:
K,NPT (reference number for keypoint),X,Y,Z (keypoint location)
The generation sequences of the nodes are structured by means of nested
buckles for each of the described areas: an external one that runs across all the
fringes of the zone that has been fractioned; and another internal one that
analyzes the 105 constituting points of each transversal section (18 in the third
sector). In each case the number of the node is resolved by means of a counter,
the coordinates (x,y) are deduced from the previously formed auxiliary matrices,
and the position of (z) is estimated in the external buckle in relation with the
section that is being configured at every moment. The following phase supposes
the creation of lines and arcs with an arrangement corresponding respectively to
the orders:
L,P1,P2 (keypoints at the end of the line or circular arc line)
LARC,P1,P2,PC (center),RAD (radius of curvature of the arc)
In this case, Ansys does not require the numeration of such entities in the
command line, that is why it is considered as appropriate to specify it as a
comment to its term. This possibility simplifies, in its case, the localization of the
elements in the list. The process is configured equally by means of buckles,
updating in each transversal plane the number of its lines and arcs with a total
arrangement of the previously treated sections.
Once the bases of the wire structure of the model are formed, the definition of
the areas based on the lines and the arcs that configure its perimeter is realized.
The process is very similar to the described one in accordance with the previous
phase and is used in any case of the command:
A,P1,P2,P3,P4,P5,P6,P7,P8,P9 (list of keypoints defining area)
For configuring the volumes an extrusion of the definite areas in each section
is realized in such a way that the numeration of both types of entities turns out to
be coincident (fig. 5). To achieve this operation it is necessary to create in
advance auxiliary lines over the axis of the specimen, linking the central nodes
(1) of each of the consecutive two sections. The order that formalizes the
extrusion responds to the following syntax:
VDRAG,N1,N2,N3,,N6 (areas in the pattern),NLP (line defining the path)
Figure 5:
577
Obtaining volumen by extrusion.
At the end of this phase the volumes that are situated between the zones z1
and z2 (additional band of 2 mm) relative to the bottom of the bar and the
adhesive are eliminated. This elimination is decided after the real specimens are
analyzed with a very reduced efficiency of the adhesive in this area, as a
consequence of the filling process itself. In figure 4 the resulting slot caused by
the elimination commands is estimated (VDELE, NV1, NV2, NINC: delete
volumes from NV1 to NV2 in steps of NINC).
To finish the preprocess of the model, the obtained volumes are related to the
involved materials, a task in which controlling the numeration of the first items
proves to be fundamental. The selection is realized by means of ranks, a reason
for which in each section at first the central steel elements are coded, then the
relatives to the band of the adhesive and finally the constituent elements of the
timber.
VSEL,Type,Item,,VMIN,VMAX (values of the item range),VINC (increment)
VATT,MAT,REAL,TYPE (parameters to be assigned)
The meshing only affects the final volumes (VMESH,ALL). With a previous
discretization, a subdivision of each edge in two halves (ESIZE,2) is considered
correct. The application permits to refine the mesh in determined areas, but
advanced studies indicate that the resultant model is sufficiently approximate
without further modifications of the set.
Analysis and interpretation of the results
In the module of the calculus the transversal displacements coerce the symmetry
planes (used for reducing the real specimen to its eighth part). So the external
load is applied to the extreme surface of the steel bar, and the phase of analysis

begins. After the calculus is finished, the postprocessor is opened studying the
magnitudes that define the solution: basically the energy estimation of the errors,
displacements (fundamental for its comparison with the experimental results),
and stress conditions (fig.6). An important question is the distribution of the
tangential stresses over the surface of contact between the adhesive and the
timber, a parameter that could be determinant in numerous extreme situations
(fig.7). For that reason the nodes of the mentioned area included in the plane (xz)
are isolated and a list of resisting normal and tangential stresses is requested. A
third computer program extracts the desired values (xz in this case) and starts to
represent them in an environment of computer aided design (fig. 8).
Figure 6:
Figure 7:
Displacements (z) and tangencial stresses (xz).
Failures associated with the excessive tangential stresses.
The diagrams show curved lines with maximal numbers focused on the ends
of the anchorage area which are much greater than the average value traditionally
used in the references. Both peaks turn to be similar for reduced longitudes,
579
increasing noticeably the exterior one when the union is realized with more
profundity.
Figure 8:
Distributions of the tangential stress.
Conclusions
The means in parallel to experimental technique and numerical analysis, for the
study of the resistant behaviour of timber-steel joints as the described ones,
supposes a sufficiently complete and illustrative strategy. The comparison of the
results allows to calibrate gradually the virtual models in such away that they
prove to be useful for verifying practical cases that are not specifically subjected
to tests, or for predicting possible situations of risk. Likewise, the mentioned
idealizations are useful for assessing constructive solutions at the very beginning,
thus reducing the relative costs of production and of the load of possible
prototypes. In this generic frame the computerized configuration of the models
remarkably simplifies the generation and postprocess tasks appealing to the
parametrization of its geometry.
Acknowledgements
This research is sponsored by the Ministry of Science and Technology through
research project title uniones metlicas encoladas con adhesivos en barras de

madera (glued anchoraged timer joints). The financial support is gratefully
acknowledged.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
Broughton, J.G. & Hutchinson, A.R. Effect on timber moisture content on

bonded-in rods. Construction and Building Materials, volume 15, issue 1,
pp. 17-25, 2001.
Deng, J.X., Moss, P.J. & Buchanan, A.H. Glued bolts in glulam. An
analysis of stress distribution. 5 World Conference on Timber
Engineering, Montreux, Switzerland, 1998.
Estvez, J., Pablos, J., Muiz, S., Freire, M., Vzquez, R. & lvarez, J.
Double-layer space structures of laminated timber tubular members. Proc.
of the 4 International Conference on Space Structures, University of
Surrey, Guildford, UK, Ed. Thomas Telford Ltd.: London, pp. 563-572,
1993.
Estvez, J. & Vzquez, J.A. Spatial truss of hollow bars made of
laminated timber supported by walls of reinforced masonry. Journal of the
International Association for Shell and Spatial Structures, volume 45, n.1,
April n.144.
Estvez, J., Vzquez, J.A. & Otero, M.D. Diseo y dimensionado del
nudo extremo de una barra hueca de madera laminada. 1 Congreso
Ibrico A Madeira na Construao, Universidade do Minho, Guimaraes,
Portugal, pp. 689-698, 2004.
Senno, M., Plazza, M. & Tomasi, R. Axial glued-in steel timber jointsexperimental and numerical analysis. Holz als Roh und Werkstoff,
volume 62, number 2, 2004.
Swanson Analysis Systems, Inc. Ansys Users Manual. Houston, 2005.
(See also http://www.ansys.com).
581
Theory of the elasticity of the materials

of the second order
V. Shorkin1 & V. Gordon2
1
2
Department of Physics, Oryol technical states university, Russia

Department of Mathematics, Oryol technical states university, Russia
Abstract
It is admitted that a solid body is obtained by means of extraction from an
infinite medium, and at the same time its state changes. To describe this
phenomenon there a linear theory of elasticity of isotropic materials of the
second order was used, generalized for the case of availability of the initial
stressed state described by tensors of various ranks. They depend upon the
physical nature of the body, the form of its boundary and position of points for
their determination. The boundary effect existence of surface tension and
surface energy is described. The analysis of known test data regarding values of
these quantities, wave dispersion in resilient media, calculus results on the basis
of the model offered and one-dimensional theory of the crystalline lattice of
elastic energy density for a number materials allowed values of elastic modulus
and initial stresses to be determined. The created model of medium is used for
the description of the stressed state of bodies adhered together, when their
adhesion is considered to be one of the kinds of contact interaction. The calculus
results of adhesion energy are confronted with the known data.
Keywords: theory of the elasticity, materials of the second order, the initial
stressed state, and surface energy, adhesion energy.
Introduction
It is known that particles of solid body - atoms, ions, interact with the aid of the
potential forces [1]. Therefore a change in the position at least of one particle of
a body affects the position and the mutual influence of all remaining particles.
The formation of the new section of free surface, for example, with the division,
is accompanied by the separation of the sets of particles. On the one side, the
doi:10.2495/HPSM06057

adhesion of two bodies is connected with the association of large quantities of
particles. In these processes the inter-atomic bonds break or are formed only
along the interface or association. But the result affects all particles of the
divided or united bodies. The possibility of the calculation of the aforesaid exists
with the discrete description of the behavior of material [2]. This can be made,
also, with the continuous non-local description [3, 4]. But the first method is
difficult to use for describing the materials with the complex structure. The
second method requires information about the potential of non-local interaction.
For describing the surface phenomena it is possible to use gradient theories (for
example [5, 6]). They use a higher order differential equation of motions.
Surface effects can be considered as the effects of boundary layer. However, in
the existing models is not traced clearly the connection of the predicted effects
with the special features of potential interaction of the particles of solid body. In
the model proposed below the underlining of this connection is the basis of
theory.
Model of continuous elastic medium
2.1 Basic assumptions

1.
Interaction of the atoms of real body is described with the aid of the
potential of pair wise interaction. It is assumed that for the continuous
simulator interaction of infinitely small particles is also potential, paired. The
potential of interaction is proportional to their volumes dV and dV1
( = const - material density), that have at the moment of the time t of
G
G
G G
position R and R 1 . Constant of proportionality = R 1, R (further
2.
potential) - the continuous and differentiated function of its arguments.

The studied processes are isothermal, reversed. The mechanical
properties of medium are completely determined by the form of the function
G G
= R 1, R . The uniformity of properties indicates the dependence
G G
G
G G G G
G
G G G
R 1 , R = R 1 R , R1 R = a + u , where a = r1 r the state vector of
particle dV1 relatively dV in the initial state of the considered body,
G G G
u = u1 u relative particle displacement upon transfer of body into the
G
G
G
current state. Values u and u 1 are considered small. Therefore u
3.
)
) (
also low value.

For the exception of the influence of boundary it is assumed that any
body B is obtained by isolation from infinitely extensive homogeneous
G G
continuous medium . Isolation occurs instantly. Potential = R 1, R
does not change. Initial for the body B is the state, in which it was located
after instantaneous isolation. The motion, caused by disruption of the
equilibrium of internal forces, begins at this moment. It is considered that it
rapidly goes out. Therefore, in further reasoning it is not examined.
583
2.2 Stress-strained state and mechanical properties

The characteristics of the strained and stressed state of medium are introduced as
follows. The potential d of particle dV inside the homogeneous medium is
calculated through the formula
G
G
d = R 1 R dV1 dV .
(1)
Its variation relative to the current state of medium is determined by the

equality
G
(
(
) (G
))
(d ) = w dV = R1 R + ( u ) R1 R dV1 dV =
G G T
G
= R1 R ( u )dV1 dV .
( (
(2)
))
G
G
G
Potential gradient F = R 1 R is written in the form
( )
G G
G
G G
GT
F R1 R = F (a ) + F
G
u =0
G
u .
(3)
G
In this case for the relative displacement u is considered valid the idea in the
G
form Taylor series according to the external degrees of vector a [7]
G
u =
k ! [rG(k )u ]G
k =1
G
... a k .
(4)
a =0
The consequence of these operations is the idea of a variation in the density of a

change in the potential of internal forces w , developing in the environment
G
dV of point r , which is equal to the work of internal forces in this region, in
the form
w=
P (m )T ... ( rG (m )u ) .
m =1
where:
(5)

P (m ) = P 0(m ) +
P
0(m )
C (m , n)T ... ( (n )rG u ) ,
n =1
G G Gm
[F (a )a ]dV1 ,
C (m , n ) =
(6)
(7)
G G
G
1
a n aG F a m dV1 .
n! m!
(8)
G
It follows from idea (5) that the infinite sequence rG (m )u of the gradients of
displacements is the characteristic of the strained state in the infinitely small
G
environment dV of point r . Its elements - these are the generalized
displacements, determined in dV . By the generalized forces, which work on
them, appear the stress tensors, determined by equalities (6) - (8). They compose
{ }
the infinite sequence P (m ) of the characteristics of the stressed state of region

dV . It is characteristic that the calculation of interaction of the particles of the
G G
medium by the introduction of potential = R1 , R led to the need for the
calculation of the initial stressed state in the medium, caused exclusively by its
G G
intrinsic properties. These properties are described by function = R1 , R . It
determines the values of the mechanical characteristics (in the classical
understanding) of medium , computed with the aid of (8).
For the homogeneous infinite medium tensors (7) and (8) do not depend
on the position of the point of their determination. The same is correct for the
body B , when isolated it in mentally only with surface S with the external
G
unit normal n . Picture changes with the real isolation, described in the third
assumption of introduction. The action of the appearance of body B on its
particles disappears at this moment, integration in (7) and (8) must be produced
not on , but on B . In this case the position of boundary S relative to each
G
G
point r , from which is counted off the vector a , proves to be different. In spite
G G
of the safety of function = R1 , R with the real isolation of body B from
it ceases to be uniform in the classical understanding.
2.3 External actions and equation of motion

The system of the differential equation of motions and boundary conditions takes
the form
G
G
2u
P (1) P (2 ) P (3) ... + f =
t2
G
G
n P (1) P (2 ) ... S n P (2 ) ... = (0 )
(
(
))
G
n (P ( ) (P ( ) ...))
2
)))
( (
))
G
(n (P ( ) ...)) = ( )
3
G
r V ,
(9)
G
r S ,
(10)
G
r S .
(11)
585
rG the symbol of volumetric gradient; S the symbol of surface

gradient. It was assumed during the construction of system (9) - (11) that, as in
the classical case, external actions on any body and any part of it are divided into
those volumetrically distributed and those superficially distributed. The first
G
make classical sense. Their bulk density is f . The second, just as in the classical
case, characterize contact interaction of different bodies. However, this
interaction is characterized not by one form of forces, but by infinite sequence,
that possess surface densities (k ) , k = 0 ,1, ... . With k = 0 this is the density of
the classical surface forces, which accomplish work on the true displacements of
material points. With k > 0 - this is of density over forces [8]. System (9) - (11)
must be augmented by the differential form of the law of conservation of
momentum of pulse and by initial conditions for pour on displacements and their
speeds.
G G
2.4 Version of the potential = R1 , R determination
It can be realized for the isotropic medium. In this case it suffices to examine
one-dimensional situation. Equation (9) taking into account (6) - (8) which
G
f = 0 takes the form
1 2 u1
c2 t 2
2 u1
x12
+ b12
4 u1
x14
+ b2 4
6 u1
x16
+ ...
(12)
Coefficients are determined by the formulas

C11(1,1)
, D j = D1b j 12( j 1) ,
j = 1, 2 , ... ,
D1 = C11(1,1) , D 2 = C111(1,3) C111(2 ,2 ) + C111(3 ,1) ,
c=
(13)
................................. ................
The search for the solution in the form u1 = e i ( t kx1 ) , i = 1 , gives the
possibility to obtain dispersion law dependence (k ) [1], in the form of the
power series
2
c
= k 2 b12 k 4 + b2 4 k 6 ... .
(14)
The same law can be built theoretically, with the use of ideas of solid state
physics [1], or on the basis of data of experiments [9]. With the comparison it is
G
possible to determine values b j . Function = (a ), a = a can be

represented, for example, in the form of a number of the diminishing exponential
curves
(a ) =
p e
, p =
p =1
a
bp
(15)
Substitution (15) in (7), (8), and then in (13) reduces to the infinite system of
linear algebraic equations for the unknowns p . Its solution determines
potential (15) of material, for which are obtained the values b j .
The theory application
3.1 Additional assumptions

1.
Instead of the infinite sequences of the generalized displacements and

stresses they are used only on their two first elements. This gives the
possibility to use for its solution classical methods, in particular, the method
of separation of variables - Fourier's method. The analytical method of
solving the ordinary differential equations of final order with the arbitrary
differentiated coefficients, which are appeared in the one-dimensional case,
is represented into [10].
2. Maximum simplification in the task is possible to achieve, if we assume that
in the range of values in question under the condition of coefficients are
constant. 2. It is assumed that the material is isotropic. In its near-boundary
layer the values of the tensors of initial stresses and characteristics of elastic
state are constant and equal to those values, which are characteristic for the
boundary plane of semi-infinite body. In this case the density of a change in
the internal forces potential is determined by the equality
w = g ij g ij +
g kk g ll +
+ (2 + ) b1 Z ijk Z ijk + b1 Z ijk Z kij + 0 ij k Z ijk
(16)
1
ui , j + u j ,i and Z ijk = ui , jk the component of the classical strain
2
tensor and second gradient of displacements; , the Lame coefficients;
g ij =
b1 , b2 , 0 additional constants; k the component of the unit vector,

G G
normal to the surface, parallel to boundary, where = n ; P (1)ij and P (2 )ijk
they are determined by differentiation (16) on g ij and Z ijk respectively.
Relationship (16) corresponds to the model of the linearly elastic medium of

the second order with the initial stressed state.
587
3.2 Quantity checking of theory

Within the framework of the theory proposed is examined the task about the
stress-strained state of semi-infinite body, on which act no external forces. On
the basis of its solution and data of experiments regarding the surface energy
W p of solid bodies [11] in accordance with the formulas
0 =
Wp
k
,b =
3W p
2
4k A1
, k=
2 +
, 2 A1 = 2 + , b 2 = b1 2 + b2 2 . (17)
the estimations of additional elastic constants are made for the number of
materials b and 0 . The results of calculations are represented in table 1.
Table 1:
Estimations of additional elastic constants.
Element
W p Dg / m 2
1
2
3
4
5
6
7
Mg
Al
Si
Ni
Cu
Ag
Au
0,73
1,04
1,24
1,44
1,12
0,95
1,23
b 1011 (m )
0 Dg / m 2
2,67
4,65
7,54
2,49
1,89
1,98
1,29
1,28
1,92
3,21
2,50
1,57
1,28
1,46
The use of a dispersion law [1] for determining the parameter b leads to the
results, commensurate with those given.
3.3 Adhesion of two solid bodies
By adhesion of two solid deformed bodies B (1) and B (2 ) in the model of
medium is understood this form of their contact, when the rectilinear material
fiber, normal to the surface of contact S a and which intersects it, in the process
of the deformation of the united of body B = B (1) B (2) preserves not only its
integrity (as in the case of classical, solid contact), but also smoothness of the
distribution of its deformations. This idea gives the possibility to solve the
problem about the stress-strained state of the body, which are in a state of
adhesion as the combined task about the state of piecewise-uniform body
B = B (1) B (2 ) with the condition of the absence of external actions on it.
Adhesion energy F of bodies B (1) and B (2 ) is determined by loss per the unit
surface area, along which occurred the adhesion, their free energy [12]
F = W p(1) + W p(2 ) W p(1,2 ) .
(18)

Here: W p(1,2 ) the surface energy of the system of bodies, concentrated bi S a ;
W p( j ) the surface energies of the disconnected bodies. The theory, examined in
this work, makes it possible to consider the volumetric distribution of these is

specific internal energy by means of calculated value (16). For two plates with
the thickness, for much the great significance of the parameters b ( j ) , is obtained
the formula
F =
W p( 1 )W p( 2 ) ( k ( 1 ) + k ( 2 ) )2
W p( 1 ) k ( 2 )2 + W p( 2 ) k ( 1 )2
, k( j) =
( j)
,
1 ( j )
(19)
( j ) Poisson ratio. The results of the comparison of the results of theory with
the data of experiment are represented in table 2.
Table 2:
The comparison of the results of theory with the data of

experiment.
Metal
adhesion energy on silicon
(Dg
adhesion energy on the
glass Dg m 2
aluminum (Al)
0,214
0,150
silver (Ag)
0,181
0,120
copper (Cu)
0,210
0,110
Conclusion
At present is discussed a question about the correspondence of the discrete and

continuous description of real material [2]. In the represented work it is shown
that it is possible to attain an increase in the degree of the adequacy of
description by the allotment of continuous medium with the properties,
characteristic for the discrete medium. In this work this property is the ability of
particles to interact at the significant distance with the aid of the potential forces.
The gradient model of the second order for the medium with the initial stresses
became the result this of calculation. It proved to be applicable for the use in the
calculations of surface energy and adhesion energy.
References
[1]
[2]
Kittel, Ch., Introduction to Solid State Physics, Science: Moscow,

pp 111148, 1978.
Krivtsov, A.M. & Myasnikov, V.P., Simulation by the method of the
dynamics of particles of the change in the internal structure and stressed
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
589
state in the material under the strong thermal influence. By notify of

Russia academy of science. Mechanic of solid body, 1, pp. 88 - 103, 2005.
Gordon, V.A. & Shorkin, V.S., The non-local theory of the surface layer
of solid body. Proceedings of Ttua State University. Mathematic.
Mechanic. Informatics. 4(2), pp. 55 57, 1998.
Eringen, A. C., Nonlocal continuum field theories, Springer-Verlag:
Berlin, 2002.
Mikaelyan, K.N., Gutkin, M. Yu & Ayfantis, Ye. S., Edge dislocations at
the phase boundaries in the gradient theory of elasticity. Solid state
physics, 2(9), pp. 1613 1620, 2000.
Vitcovsky, I.V, Konev, A.N., Shorkin, V.S., Kzaev, N.D., Rusanov, A.E.,
Khoroshikh, V.M., & Leonov. S.L. Adhesion energy estimation of some
composite materials. Plasma Devices and Operations, 11(2), pp 81-87.
2003.
Korn, G. & Korn, T., Mathematical handbook for scientists and
engineers, eMGRAW-HILL BOOK COMPANY, INC: New York,
Toronto, London, 1961.
Tupin, R.A., The theory of elasticity, which considers moment stresses.
Voltages the mechanics coll. of transfers, 3, pp. 113 140, 1965.
Woods, A.D.B., Cochran, W. & Brockhouse, B.N. Lattice Dynamics of
Alkali Halide Crystals. Phys. Rev., 119, pp.980-999, 1960.
Gordon, V.A., Shorkin V.S. & Borsenkov, M.I., The method of solution of
problems of mechanics of heterogeneous bodies, Oryol State University:
Oryol, pp. 35 46, 2005.
Samsonov, G.V., (ed). Properties of elements. Physical Property,
Metallurgy: Moskow, 1976.
Vvedensky, B.A., (ed). Physical encyclopedias dictionary, Soviet
encyclopedia: Moskow, 1960.
Section 8
Optimal design
593
Optimal design of fibre reinforced

tubular structures
H. Martikka & E. Taitokari
Lappeenranta University of Technology, Lappeenranta, Finland
Abstract
The main purpose of this study is to present results of the design and damage
analysis of industrial vessel shell microstructures made of fibre reinforced plastic
laminates and subject to mechanical and aggressive chemical loads due to
sulphuric and hydrochloric acids and high temperatures. Methods to obtain
optimal performance microstructure are use of balanced and symmetric laminate
layering with optimal fibre strengthening directions and volume fractions.
Optimal selection of chemically resistant fibres and matrix is essential for
endurance. Use of thick enough walls is cost-effective for vessel bottoms. Two
vessel winding manufacturing technologies are compared for optimal selection.
Keywords: composite materials and structures, optimal design, stress corrosion
cracking.
Introduction
Composites used for structural engineering are commonly formed from five
commonest material groups, metals, ceramics, glasses, elastomers and polymers.
Design rules and codes are based continuum models. Another trend is to use
more microscopic modelling. Useful information of composite design is
presented by Agarwal and Broutman [1] and Barbero [2]. Chemical resistances
are studied by Aveston and Sillwood [3] and, Kawada and Srivastava [4]. In
aggressive enough loadings all material alternatives and combinations suffer
accumulation of damage. Typical damage events are stress corrosion wearing of
outer surface, fracture of fibres and degradation of matrix. The main purpose of
this study is to present results of design and damage study analysis of industrial
vessel walls.
doi:10.2495/HPSM06058
Goals, materials and methods
2.1 General goals of designing thin shelled structures for process industry
In metallurgical chemical industry concentrated sulphuric and hydrochloric acids
are processed at about boiling temperature in thin shelled large vessels. The
focus is now on shells made inorganic composites, Figure 1.
Biopolymers
Natural
fibres
Polymers,
elastomers
Ceramics,
glasses, inorganics
Old
Metals T=10y
A B New
T=0.1y
FRP,inorganic
FRP
biobased
MMC
Rm,new
Composites
Metal -metal-interphases
a)
Rm,old
b)
Figure 1:
a) Composite structure relationships. b) Environmental attack on a

two layered composite causing wear and decrease of strength Rm.
2.2 Materials and loads

The principle of designing composite walls is to use functional property
gradients. Chemical resistance is needed close to inner wall. Load bearing
capacity is obtained by making the walls strong and thick enough.
The fibre materials are typically glass fibres. Now elastic modulus is Ef =
70000MPa, fracture strain fu = 0.012. Options are normal low cost E-glass and
special chemically resistant glass. Matrix modulus is typically Em =3400, tensile
strength 83MPa and shear strength of same magnitude.
Load is 20% sulphuric and hydrochloric acid at 104oC.
Background theory
3.1 Micromechanical properties of plane strain orthotropic lamina layers

In principal LT material directions the relation between principal strains and
principal stresses is given by the principal material direction stiffness matrix Q.
It is the Hookes law
L Q11
= Q
T 12
LT 0
Q12
Q22
0
0 L
0 T
Q66 LT
(1)
3.1.1 Stresses in global and principal strength directions

Each lamina gets its input loads from the global strain distribution. Use of the
direction transformation matrix gives the components in principal directions.
c 2 s2
2sc x
L
= s 2 c 2 2sc
y
T
LT k sc sc c 2 s 2 xy
k
{ L } k = [T]{ x } k
595
(2)
Transformation of stiffness matrix components from principal to global xy coordinates gives

Q11 Q11, k = Q11c 4 + Q22 s 4 + 2(Q12 + 2Q66 ) s 2 c 2
Q22 Q22, k = Q11s 4 + Q22 c 4 + 2(Q12 + 2Q66 )s 2 c 2
Q12 Q12, k = (Q11 + Q22 4Q66 )s 2 c 2 + Q12 c 4 + s 4

Q66 Q66, k = (Q11 + Q22 2Q12 2Q66 )s 2 c 2 + Q66
)
(c
(3)
4
+ s4
Q16 Q16, k = (Q11 Q12 2Q66 )sc 3 (Q22 Q12 2Q66 )s3c
Q26 Q21, k = (Q11 Q12 2Q66 )s3c (Q22 Q12 2Q66 )sc 3
The global midplane strain induces global stresses at layer k with angle to xaxis
x
Q11

y = Q12
xy
k Q16
Q12
Q22
Q26
Q16 x 0

Q26 y 0 ,
Q66 xy 0
{ x } k = [Q ]{ x 0 }
(4)
3.1.2 Force and moment resultants of all layers

Force resultants are needed. They are obtained by integrating over the thickness
Nx
Ny =
N xy
x

hk -1 y dz =
k =1
xy
k
k =n h
k
k=n h
k
k =1
k -1
k dz
(5)
= Q = Q ( 0 + z ) = Q 0 + zQ
In component vector form the force resultant vector is
Nx
N y = Aij
N xy
[ ]
x0
x
0
y + Bij y
0
xy
xy
[ ]
where the A and B matrices are

(6)

k =n
[ Aij ] = Qij
k =1
k=n
[Bij ] = 12 Qij k (hk 2 hk-12 )
( hk hk-1 ) ,
(7)
k =1
Use of these is illustrated in the case study model, shown in Figure 2.

k =1
t1
1=+45 h0 = -6
k=2
x
k=3
t2
2=-45
h1=--3 h2 = 0
t3
3=-45
h3=3
k = 4 = n t4
4=+45
z
T
y
Figure 2:
h4 = 6
L
Principle of a symmetric and balanced composite +45/-45/-45/+45.
The ABD matrix relates the midplane strains and plate curvatures to resultant
force vector N and moment vector M. The ABD matrices are extensional
stiffness matrix A, coupling stiffness matrix B and bending stiffness matrix D.
N A B 0
A B 0
=
M B D
B D

A 0
0
B
(8)
In formulating the design goals and constraints the symmetry condition is the
stiffest one. These rules can be utilised to obtain simple and well functioning
microstructures.
3.1.3 Symmetric laminate structure as one optimisation goal
The useful result of using symmetry goal is that the B matrix is made zero. Using
the case study data
[Q ]k=1,+45deg = [Q ]k=4,+45deg
[Q ]k=2,-45deg = [Q ]k=3,-45deg (9)
Gives for the B matrix
[Bij ] = 12 ([Q ]1 (t 2 4t 2 ) + [Q ] 2 (0 2 t 2 ) + [Q ] 2 (t 2 0 2 ) + [Q ]1 (4t 2 t 2 )) = 0
(10)
The simplified result form is
N A 0 0
M = 0 D

A 0 0
0 D
A 0
(11)
597
3.1.4 Balanced laminate structure as one optimisation goal

The next optimisation goal is to simplify the A matrix as much as possible. This
goal is reached by adjusting the transformed stiffness components.
If for every layer there is another almost identical layer but with opposite
orientation somewhere in the laminate, the sign of sine function is changed. This
means that Q16 and Q26 are zero meaning that A16, A26 are zero. For instance
Q16 + sin = + s = (Q11 Q12 2Q66 )sc 3 (Q22 Q12 2Q66 )s3c
(12)
Q16 ' sin( ) = s = (Q11 Q12 2Q66 )sc 3 + (Q22 Q12 2Q66 )s3c = Q16
Using this result in A matrix and setting laminate layers constant h = t
[ A16 ] =... Q16 t + Q16 't = t (Q16 Q16 ) = 0
(13)
Using this one obtains for the A matrix
[ ] Qij k = t((Qij )1 + (Qij ) 2 + (Qij )3 + (Qij ) 4 )

k =1
Aij = t
N x A11

N y = A12
N xy A16
A12
A22
A26
A16 x 0 A11
A26 y0 A12
A66 xy0 0
A12
A22
0
, T = t
0 x0
0 y0
A66 xy0
(14)
(15)
Force resultant depends on strains caused by global load stresses

N x A11
N =
y A12
A12 x 0
N x
x
1
pr 1
= pr
0 , = t = t
N
A22 y
t 0.5
0.5
y
y
(16)
N xy = A66 xy 0
The globally directed stresses acting on or within lamina k are

x,k Q11
=
y,k Q12
Q12 x 0

Q22 k y 0
(17)
From this equation the midplane strains can be solved

x 0 a11
0 =
y a12
a66 =
1
A66
a12 N x
1 A22
=
a 22 N y det A A12
,
det A = A11 A22 A12
A12 N x

A11 N y
whence the globally directed stresses acting on lamina depend on pressure

(18)

x,k
=
y,k
Q11
Q12
Q12 a11

Q22 k a12
a12 1
pr
a 22 0.5
(19)
3.1.5 Stresses in laminate layers

Using the xy global stress in layer k the stresses in principal directions LT of a
layer k can be calculated. These are needed in stress based failure criteria
L,k c 2 s 2
2
c2
T,k = s
LT,k sc sc
2sc x,k

2sc y,k
c 2 s 2 xy = 0
(20)
3.2 Environmental interaction effects

The degradation of composite materials may result from several factors [1]
1) Loss of strength of reinforcing fibres by stress-corrosion
2) Loss of adhesion and interfacial bond strength degradation
3) Chemical degradation of the matrix material
4) Dependence of the matrix modulus and strength on time and temperature.
One present observation is that rise of temperature from 20C 100C lowers the
modulus to half 3400 1700MPa
5) Accelerated degradation by temperature and chemical environment.
3.2.1 Failure mechanisms in fibre reinforced structures

Some failure mechanisms in fibre reinforced structures are shown in Figure 3.
d
5
b)
2a
hackles
4
c)
a)
Figure 3:
d)
a) Crack growth in fibre reinforced material by various

mechanisms, 1. Fibre pull-out. 2. Fibre bridging, 3. Fibre/matrix
debonding, 4. Fibre failure, 5. Matrix cracking. b) In E/VE
composite the solution gets into contact with E-glass to corrode
fibres. In E-glass there are Al2O3 and Fe2O3 compounds which
corrode forming hydroxides. c) In C/VE glass water diffuses into
matrix swelling it. Acid corrodes interfaces but less fibres. The
result is that fibres are loosened and pull-out takes place. d) Fibre
cross-section fracture models.
599
3.2.2 Condition of chemical endurance

The industrial goal is to ensure that the wall does not leak and burst. Both the
strength and the load bearing effective wall thickness T are decreased by the acid
gas and liquid. Margin of safety goal is that stress should be less than strength.
Dominant stress is the hoop stress. It increases when the wall corrodes. The
strength decreases also and the factor safety N should be large enough
x = p
1
N
Rm =
.
Rm = 10 a t b 100 t 0163
1
N
Rm0
(21)
tp
, Rm [ MPa ], t[ years]
(22)
300
Bending Rm= 250MPa,
new samples
tested in air.
Bending Rm=150MPa,
after 2 years in liquid acid
200
Bending Rm= 100MP,

after 2 years in gas
Ultimate
strength Tensile
(MPa) Rm=
100
Tensile test
Log(Rm)-Log(T)
prediction model for samples
held in gas 5 years
Rm= 69 MPa
k
190MPa
new
Tensile
Rm=100MPa
2 years in liquid
Tensile
Rm= 80MPa
2 years in gas
0
0
T, time in environmental test conditions

t1= 0.1years
Figure 4:
Tensile test Rm-T

linear model
Tensile test Lin(Rm)-Log(T)

linear -log prediction model
Experimental testing of E-glass vinylester 25mm thick laminates at

Sulphuric -hydrochloric 20% acid at 104 oC and in gas above it.
Average stress was zero.
3.3 Model of stress corrosion aided crack growth of fibres

Aveston and Sillwood [3] have studied conditions for crack growth in E-glass
unidirectional polyester in 1 N H2SO4.They made several tests a1) static load, a2)
static load with pre-soak for 1.8106 sec, a3) Fatigue 0.1 Hz square wave. All
points were close each other allowing a single model for crack growth rate to be
used nearly independent of the time dependence of load stress as

V=
da
= AK I n = 9.55 10 11 K I 3
dt
units: V
[ ],
m
s
t[sec], [ MPa]
(23)
1
Y ' = Y 2
KI = aY' a , n 3 ,
a is applied stress, static or dynamic, a is crack length at time t, Y is a
geometrical factor for crack. For a crack of length 2a in a plate, Y = 1 , Y =
Since the life of fibres is decisive for the whole composite, the fibre life can
be regarded as the life time of the composite under full loading. Initial crack
size is ai and final aIC. Stress corrosion life is from initial to final crack size
dt =
da
AK I n
da
A aY' a
tC
dt = t C =
AY ' n a n
0
C
1
1 2 ( 2 n ) (24)
a
1 n
ai
For short time testing, the applied stress is raised up to tensile strength while the
crack starts at initial size. Stress intensities at initial and final crack sizes are
K Ii = a Y ' a i , K IC = a Y ' a C K IC = max Y ' a i , max = Rm (25)
Analytical optimisation
This is based on the previous theory.
4.1 Logic of finding optimal structure

1. For Itt = 1 to 3, T(iItt) : T(1) =.01, T(2)=.022, T(3)= .04,wall thickness
2. For Ivf = 1 to 2, Vf (IVf): Vf (1) = 0.14, Vf (2) = 0.46, volume fraction
3. For Iss = 1 to 2, s(Iss) : s = 5000,20,1, aspect ratio choices
Table 1:
Material and model data.
Ef = 70000
f = 0.22
Em = 3400
m = 0.38
s=
2L
,e
d
L =
e 1
e + 2s
Ef
Em
fu ( Im) = 0.012
T =
e 1
e+2
Ef
2(1 + f )
Em
Gm =
2(1 + m )
Gf =
v=
Gf
Gm
G =
, z =1
v 1
v+z
Goal is to calculate elements for the principal LT matrix Q

1 + 2 s LVf
1 + 2 TVf
E L = uE m , u =
, E T = wE m , w =
1 LVf
1 TVf
E
1 + GV f
G LT = gG m , g =
, LT = f Vf + mVm , TL = LT T
EL
1 GV f
(26)
(27)
601
These models are input to principal LT stiffness matrix of orthotropic lamina

EL
LT E L
0
1
LT TL
LT TL
11
11
(28)
ET
= LT E L
22 , 21 = 12
0
22
1

E2
E1
LT TL 1 LT TL
12
G12 12
0
0
The Q matrix is in Hookes law

0 L
L Q11 Q12
= Q
(29)
0 T
T 12 Q22
LT 0
0 Q66 LT
4. for Ith = 1 to 3, Angle (Ith), is varied

(1) = 20, (2) = 40, (3) = 60
Next the transformed Q k for layer k are calculated. The two optimising
conditions of symmetry and balancing have been applied
For k= 1 to 4, calculation of global stresses in lamina k
(30)
s = sin( k ) , c = cos( k )
{ x } k = [Q ]{ x 0 }
(31)
Next k
5. For k = 1 to 4 loop for calculating lamina k stresses and strengths
Calculate A matrix
4
[ Aij ] = t
(( ) ( ) ( ) + ( Q ) )
(32)
Q12 a11

Q22 k a12
(33)
Qij = t Qij + Qij + Qij

k
1
2
k =1
-1
Calculate inverse A = A = a
Calculate globally directed stresses in lamina layer k
x,k
=
y,k
Q11
Q12
ij 4
a12 1
pr
a22 0.5
Calculate lamina stresses in LT directions

L,k c 2 s 2 x,k
x,k
, LT,k = [ sc sc]
= 2
2
c k y,k
T,k s
x,k
(34)
Strength in layer k
Strength in L direction is related to volume fractions of fibres
EL
V
= p fL VfT = 1p VfL
ET
VfT
, VfL + VfT = Vf 0
(35)
Total average volume fraction of fibres if conserved and divided to L and T

directions

Vf 0 + Vm 0 = 1 ,VfL + VmL = 1, VfT + VmT = 1
(36)
Volume fraction portions for L and T directions are

p
1
VfL =
Vf 0 , VfT = VfL = VmL
1+ p
p
Strength formulas
Strength depends on length of fibres and mechanism of strengthening
If s > sc then the overcritical strength formula applies
(37)
LU = fuVfL + m *VmL , TU = fuVfT + m *VmT , LTU = 83 (38)

f
Else if s < sc, then undercritical then strength formula applies

It is assumed m = 1 meaning full slipping
LU = 2ms i (1 m) VfL + m *VmL
(39)
TU = 2ms i (1 m) VfT + m *VmT s < sc
(40)
End if condition s<>sc

Failure criterion Fk for each layer k
Now failure criteria for each layer k, fibre material Im, aspect ratio s and
environmental loadings are obtained
For k = 1 to 4, get F (k)
2
Fk ( k ) = L L T + T + LT 1
LU k LU k TU k TU k LTU k
(41)
Next k
5. Next k,
for each lamina stresses and strengths and failure criteria
4. Next Ith, Angle (Ith) is varied
3. Next Iss, s = 5000, 20, 1, aspect ratio is decreased if
environmental attack by stress corrosion cuts fibres
2. Next IVf, Vf = 0.14 to 0.46 is lower at inner and higher for support layer
1. Next Itt, wall thickness
Ef
Em
Figure 5:
2L
2r=d
mL
Basic models of laminate and fibre strength behaviour.
4.2 Goals and results

The main goal is minimisation of cost subject to desired values of failure criteria
of layers. The cost is related to thickness times cost of a unit mass. The second
603
goal is to constrain product of failure criterions. The third goal is to obtain stress
corrosion life of fibres sufficiently large by fibre and matrix material selections.
A. At new materials, interfacial matrix strength is i=83MPa, LT strengths are
LUk=320, TUk=125. When long fibres s=5000, failure F=0.073. Strains
x=0.0015, y=0.0028 are close to allowed strain 0.0015.
B. Next matrix degradation is assumed from i =83 to 40 causing F rise to 0.1
C. If cracking decreases aspect ratio to s=20, with lowing of matrix strength to
i=40, then F increases further but not much to 0.11.
D. If cracking of fibres is maximal, then aspect ratio is minimal, s =1. If also
matrix is degraded to i=40, then LUk=40 and TUk=40. Failure criteria are large
predicting full failure F >4. Also strains are large x=0.004, y=0.01.
E. Stress corrosion cracking life at pressure 1 MPa gave hoop stress 45. At 1N
H2SO4 acid load reasonable threshold stress intensity is KIi=5 for typical E-glass
giving 0.02 years. Somewhat more chemically resistant E-glass with KIi=1 may
give 0.1 years. Most optimal is to use such fibres which endure these acids well.
Microscopy characterisation
Scanning electron microscopy was applied to study the microstructure of a

sample of thickness 9 mm which had been 2 years in saturated acid vapour gas,
Figure 6. Then it had been tested to fracture in three point bending test.
During the 2 years of acid attack the stress was low. At the final test it was
applied up to fracture load. Evidently fibres had been weakened substantially as
shown by its lowered strength value. Also little sign of matrix was found.
1
2
3
Figure 6:
SEM characterisation. Location (1) is at neutral surface and fracture

mode is mainly shearing and fibre tensile fractures, at left up (230x)
and left down (500x). Location (2), (at middle top 650x and top
down 230x) shows fibres at transverse direction. c) Right down (at
180 x) at location (3) close to the tensile surface shows axial and
transverse fibre fractures but little matrix left.
6 Optimisation of pipe manufacturing and structure

An industrial pipe made of continuous fibre and mat as sketched in Figure 2.
Diameter is 2m and load is internal pressure of 1 MPa. Volume fraction is 0.46,
moduli and Poissons ratios: Ef =72000MPa, f =0.22, Em =3450, m =0.38,
densities f =2590kg/m3, m =1200. Individual ply lamina LT properties were
EL=35300, ET=6150, GLT=2244, LT=0.3, TL=0.053. Allowable strains of 0.0015
were achieved. Two manufacturing methods are competing options:
A. Manufacturing with 90/0deg winding can be done with a simple machine
and control but requires costly axial mats. This gives pipes with about 30%
fibres in hoop and 70% in axial direction.
B. Manufacturing with 60 deg gives equal strains to the hoop and axial
directions but requires a programmable costlier machine. Advantages are that
low cost continuous roving can be used.
Conclusions
The following conclusions can be drawn.

In this study the results are presented of designing optimally industrial large
vessel shell microstructures made of fibre reinforced plastic laminates and
subject to mechanical and aggressive acids at high temperatures.
Optimal performance microstructure is obtained by use of balanced and
symmetric laminate layering with optimal fibre directions and chemically
resistant fibres and matrix. Thicker wall is cost-effective for vessel bottoms.
Industrial end-users require pipes to have same allowed strains in main
directions. Manufacturing choices determine also microstructures: the
90/0deg winding requires simple machine but costly mats, the 60 deg
winding requires complex machine but with inexpensive continuous roving.
Acknowledgements
The assistance in SEM analyses by Mr. Markku Levomki, MSc, is gratefully
acknowledged. This research is supported by EU Asia Link Project (Contract
Reference no.: ASI/B7-301/98/679-023)
References
[1]
[2]
[3]
[4]
Agarwal, B.D.and Broutman L.J.Analysis and performance of fiber

composites, John Wiley & Sons, Inc.1990.
Barbero, E.J. Introduction to composite materials design, Taylor &
Francis, 1999.
Aveston, J. and Sillwood, J.M. Long term strength of glass-reinforced
plastics in dilute sulphuric acid. Journal of materials science 17 (1982)
3491-3498.
Kawada, H. and Srivastava, V.K., The effect of an acidic stress
environment on the stress-intensity factor for GRP laminates. Composites
science and technology 61 (2001) 1109-1114.
605
MINLP optimization of steel frames

S. Kravanja & U. Klanek
University of Maribor, Faculty of Civil Engineering, Smetanova 17, 2000
Maribor, Slovenia
Abstract
In this paper we deal with the topology and standard optimization of unbraced
steel frames with rigid beam-to-column connections. The optimization has been
performed by the Mixed-Integer Non-linear Programming (MINLP) approach.
The MINLP performs a discrete topology and standard dimension optimization,
while continuous parameters are simultaneously calculated inside the continuous
space. As the discrete/continuous optimization problem of steel frames is
non-convex and highly non-linear, the Modified Outer-Approximation/EqualityRelaxation (OA/ER) algorithm has been used for the optimization. Two practical
examples with the results of the optimization are shown at the end of the paper.
Introduction
The paper presents the topology and standard dimension optimization of

unbraced steel frames with rigid beam-to-column connections. The optimization
of frames is performed by the Mixed-Integer Nonlinear Programming, MINLP.
The MINLP is a combined discrete-continuous optimization technique. In this
way, the MINLP performs the discrete topology (i.e. the number of columns and
beams) and standard dimension optimization (i.e. standard cross-section sizes)
simultaneously with the continuous optimization of parameters (e.g. internal
forces, deflections, mass, costs, etc.).
The MINLP discrete/continuous optimization problems of frames are in most
cases comprehensive, non-convex and highly non-linear. This optimization
approach is proposed to be performed through three steps. The first one includes
the generation of a mechanical superstructure of different topology and standard
dimension alternatives, the second one involves the development of an MINLP
model formulation and the last one consists of a solution for the defined MINLP
optimization problem.
doi:10.2495/HPSM06059

The Modified Outer-Approximation/Equality-Relaxation algorithm [1], [2]
and [3] is used to perform the optimization. The objective of the optimization is
to minimize the mass of the structure. The finite element equations are defined as
the equality constraints for the calculation of internal forces and deflections of
the structure. Constraints for the dimensioning of steel members are determined
in accordance with Eurocode 3 [4]. Beside the theoretical basics, an example of
the optimization of a steel frame is presented at the end of the paper.
Steel frames
Steel frames, see Figure 1, are proposed to be analyzed under the combined
effects of the self-weight of frame members, uniformly distributed variable load,
concentrated variable load on each storey and an initial frame imperfection.
Second-order elastic structural analysis is performed by considering a
geometric nonlinearity due to P- and P- effects. While the P- effect is
associated with the influence of the axial force on the beam-column member
flexure, the P- effect denotes the influence of axial force acting through the
relative sideways displacements of the element ends. In this paper, both effects
are included in the nonlinear stiffness matrix of the individual frame member by
usage of stability function approach. Chen and Lui [5] have comprehensively
described this approach and have presented stability functions sii and sij. These
stability functions are different for compressive and tensile axial forces. Since
they give the indeterminate numerical solution when axial force is zero,
simplified expressions for stability functions S1 and S2, introduced by Kim et al.
[6], are rather used. The shear deformation effect is neglected considering the
fact that only slender structural members are subjected to buckling for which
shear deformation is insignificant.
Design/dimensioning of steel frames is performed in accordance with
Eurocodes 3 for the conditions of both the ultimate limit and serviceability limit
states. When the ultimate limit state of beam-column members is considered, the
elements were checked for bending moment, vertical shear, shear buckling,
interaction between bending, shear and axial force and interaction between
bending and axial compression/buckling. The ultimate moment capacity is
calculated by the elastic method. Since the second-order elastic global analysis is
used, the in-plane buckling lengths of compression members are calculated
considering the non-sway mode.
Considering the serviceability limit state, the vertical deflections of beams in
the individual storey were calculated by the elastic method. The total deflections
subjected to overall load max and the deflections subjected to the variable
imposed load 2 are calculated to be smaller than the limited maximum values:
span/250 and span/300, respectively. The horizontal deflections are also checked
for the individual storey and for the structure as a whole. Both types of
horizontal deflections are checked for the recommended limits: the relative
horizontal deflection of each storey should be smaller than each storey
height/300 and the horizontal deflection of the top of the frame must be smaller
than an overall frame height/500.
607
Lc
q
F
H
Lc
Lb
Lb
Lb
Figure 1:
Unbraced steel frame.
MINLP model formulation for mechanical superstructures
The MINLP optimization approach requires the generation of an MINLP frame

superstructure composed of various topology and standard dimension design
alternatives that are all candidates for a feasible and optimal solution. It is
assumed that a general nonconvex and nonlinear discrete/continuous
optimization problem can be formulated as an MINLP problem (MINLP-G) in
the form:
min z = c T y + f ( x )
h( x ) = 0
g( x ) 0
By + Cx b
s.t.
(MINLP-G)
x X = {x R : xLO x xUP}
m
y Y ={0,1}
where x is a vector of continuous variables specified in the compact set X and y
is a vector of discrete, mostly binary 0-1 variables. Functions f(x), h(x) and g(x)
are nonlinear functions involved in the objective function z, equality and
inequality constraints, respectively. Finally, By+Cx b represents a subset of
mixed linear equality/inequality constraints.
The above general MINLP model formulation has been adapted for the
optimization of mechanical superstructures (MINLP-SMS). The resulted
formulation that is more specific, particularly in variables and constraints. It can
also be used for the modelling of steel frames. It is given in the following form:

min z = c T y + f ( x )
s.t.
h( x ) = 0
g(x ) 0
A( x ) a
Ey e
(MINLP-SMS)
Dy + R( x ) r
e
( )
Py + S (d ) s
Ky e + L d cn k
st
m
y Y ={0,1}
In the model formulation, included are continuous variables x={d, p} and
discrete binary variables y={ye, yst}. Continuous variables are partitioned into
design variables d={dcn, dst} and into performance (nondesign) variables p,
where subvectors dcn and dst stand for continuous and standard dimensions,
respectively. Subvectors of binary variables ye and yst denote the potential
existence of structural elements inside the superstructure (the topology
determination) and the potential selection of standard dimension alternatives,
respectively.
The mass or economical objective function z involves fixed mass or cost
charges in the linear term cTy and dimension dependent mass or costs in the term
f(x).
Parameter nonlinear and linear constraints h(x)=0, g(x) 0 and A(x) a
represent the rigorous system of the design, loading, stress, deflection, stability,
etc. constraints known from the structural analysis.
Integer linear constraints Ey e are proposed to describe relations between
binary variables.
Mixed linear constraints Dye+R(x) r restore interconnection relations
between currently selected or existing structural elements (corresponding ye = 1)
and cancel relations for currently disappearing or nonexistent elements
(corresponding ye = 0).
Mixed linear constraints Kye+L(dcn) k are proposed to define the continuous
design variables for each existing structural element. The space is defined only
when the corresponding structure element exists (ye =1), otherwise it is empty.
Mixed linear constraints Py+S(dst) s define standard design variables dst.
Each standard dimension dst is determined as a scalar product between its vector
of standard dimension constants q and its vector of binary variables yst. Only one
discrete value can be selected for each standard dimension since:
d st = qi yist
iI
st
yi = 1
iI
609
Optimization
The Modified Outer-Approximation/Equality-Relaxation (OA/ER) algorithm is

used to perform the optimization. The OA/ER algorithm consists of solving an
alternative sequence of Non-linear Programming optimization subproblems
(NLP) and Mixed-Integer Linear Programming master problems (MILP). The
former corresponds to the optimization of parameters for a frame structure with
fixed topology and standard dimensions and yields an upper bound to the
objective to be minimized. The latter involves a global approximation to the
superstructure of alternatives in which a new topology and new standard sizes
are identified so that its lower bound does not exceed the current best upper
bound. The search is terminated when the predicted lower bound exceeds the
upper bound.
The optimization is performed sequentially in two different phases in order to
accelerate the convergence of the OA/ER algorithm. The optimization starts with
the topology optimization of the frame, while standard dimensions are relaxed
temporary into continuous parameters. In the case of the standard dimension
optimization only, the optimization begins with the continuous NLP optimization
of the frame. When the optimal topology (continuous parameters) is found,
standard sizes of cross-sections are re-established and the standard dimension
optimization of cross-sections is then continued until the optimal solution is
found.
Examples
5.1 Example 1: three-storey frame

The first example shows the standard dimension optimization of the three-storey,
three-bay plane steel frame (see Figure 2). The frame is subjected to the selfweight, to the uniformly distributed imposed load of 50 kN/m and to the
concentrated imposed load of 10 kN on each storey. The frame is considered as a
laterally supported plane frame. The frame superstructure has been generated in
which all possible structures are embedded by different standard sizes variation.
The superstructure comprises 24 different standard hot rolled European wide
flange HEA sections (from HEA 100 to HEA 1000) for each beam and column
separately. The material used is steel S 355.
The optimization was performed by the MINLP optimization approach. The
MINLP optimization model for steel frames was used. The task of the
optimization was to find the optimal structure mass.
The optimization was carried out by a user-friendly version of the MINLP
computer package MIPSYN, the successor of PROSYN [1] and TOP [7]. As
an interface for mathematical modelling and data inputs/outputs GAMS (General
Algebraic Modelling System), a high level language, was used [8]. The Modified
OA/ER algorithm and the two-phased optimization were applied (a single
MINLP), where GAMS/CONOPT2 (Generalized reduced-gradient method) [9]

was used to solve NLP subproblems and GAMS/Cplex 7.0 (Branch and Bound)
[10] was used to solve MILP master problems.
5.0 m
F = 10.0 kN
q = 50.0 kN/m
q = 50.0 kN/m
5.0 m
F = 10.0 kN
15.0 m
5.0 m
F = 10.0 kN
5.0 m
5.0 m
5.0 m
15.0 m
Figure 2:
Three-storey, three-bay steel frame.
The optimization model of the frame contained 1572 (in)equality constraints,

1552 continuous and 357 binary variables (61 after the prescreening). The final
optimal solution of 6214 kg was obtained in the 61st main MINLP iteration (the
subsequent feasible result was not so good). The optimal standard sizes were also
obtained. Only 152 seconds of CPU time were spent on a 2 GHz PC. The
optimal structure of the frame is shown in Figure 3.
5.2 Example 2: single-storey industrial building
The second example presents the topology and standard dimension optimization
of a single-storey industrial building. The building is 20 meters wide, 40 meters
long and 6.5 meters high (see Figure 4). The structure is consisted from equal
non-sway steel portal frames, which are mutually connected with the purlins.
task of the optimization was to find the optimal structure mass, the optimal
topology (the optimal number of portal frames and purlins) and optimal standard
cross-sectional sizes.
611
Variable imposed loads s = 1.60 kN/m2 (snow) and w = 0.137 kN/m2 (wind)
are defined as the uniformly distributed surface load in the model input data.
Both, the horizontal concentrated load at the top of the columns (wind) and the
vertical uniformly distributed line load on the beams (snow and wind) are
calculated considering the intermediate distance between the portal frames.
Internal forces are calculated by the elastic first-order analysis for non-sway
frame mode. The dimensioning of steel members is performed in accordance
with Eurocode 3 for the conditions of both ultimate limit state (ULS) and
serviceability limit state (SLS).
5.0 m
5.0 m
HEA 180
HEA 180
HEA 340
HEA 180
5.0 m
15.0 m
5.0 m
HEA 260
HEA 340
HEA 260
5.0 m
HEA 260
HEA 180
HEA 260
HEA 260
HEA 180
HEA 260
HEA 260
HEA 180
HEA 260
HEA 260
HEA 180
HEA 260
HEA 180
HEA 260
5.0 m
15.0 m
Figure 3:
Optimum design of the frame.
The portal frame superstructure was generated in which all possible

structures were embedded by 30 portal frame alternatives and 20 purlin
alternatives. The superstructure also comprised 24 different standard hot rolled
European wide flange I beams, i.e. HEA sections (from HEA 100 to HEA 1000)
for each column, beam and purlin separately. The material used was steel S 355.
The optimization model contained 124 (in)equality constraints, 111
continuous and 122 binary variables (55 after the prescreening). The final
optimal solution of 62029 kg was obtained in the 3rd main MINLP iteration. The
optimal result includes the obtained optimal topology of 15 portal frames and 12
purlins (see Figure 4) as well as the calculated optimal standard sizes of
columns, beams and purlins (see Figure 5).
6.5 m
14
6m
.0 m
40
2.8
20.0 m
0.50 m
Optimal design of the single-storey industrial building.
HEA 120
HEA 260
HEA 260
6.5 m
HEA 500
HEA 500
6.0 m
Figure 4:
10 2.0 m
20.0 m
Figure 5:
Optimal design of the portal frame.
Conclusions
This paper presents the topology and standard dimension optimization of steel
frames with rigid beam-to-column connections. The optimization has been
performed by the Mixed-Integer Non-linear Programming (MINLP) approach.
The MINLP was found to be very successful optimization technique for solving
the frame structures.
613
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
Kravanja, Z. and Grossmann, I.E., New Developments and Capabilities in

PROSYN - An Automated Topology and Parameter Process Synthesizer,
Computers chem. Eng., 18(11/12), pp. 1097-1114, 1994.
Kravanja, S., Kravanja, Z. and Bedenik, B.S., The MINLP optimization
approach to structural synthesis. Part I: A general view on simultaneous
topology and parameter optimization, Int. J. Numer. Methods Eng., 43, pp.
263-292, 1998.
approach to structural synthesis. Part II: Simultaneous topology,
parameter and standard dimension optimization by the use of the Linked
two-phase MINLP strategy, Int. J. Numer. Methods Eng., 43, pp. 293-328,
1998.
Eurocode 3, Design of steel structures, European Committee for
Standardization, 1992.
Chen, W.F. and Lui, E.M., Structural Stability: Theory and
Implementation, New York: Elsevier Science Publishing Co., Inc., 1987.
Kim, S.E., Lee, J.S., Choi S.H. and Kim, C.S., Practical second-order
inelastic analysis for steel frames subjected to distributed load, Eng.
Struct., 26, pp. 51-61, 2004.
Kravanja, S., Kravanja, Z., Bedenik, B.S. and Faith, S., Simultaneous
Topology and Parameter Optimization of Mechanical Structures,
Numerical Methods in Engineering '92, Proceedings of the First
European Conference on Numerical Methods in Engineering, ed. Ch.
Hirsch et al., pp. 487-495, Elsevier, Amsterdam, 1992.
Brooke, A., Kendrick, D. and Meeraus, A., GAMS - A User's Guide,
Scientific Press, Redwood City, CA, 1988.
Drudd, A.S., CONOPT A Large-Scale GRG Code, ORSA J. Comput., 6
(2), pp. 207-216, 1994.
CPLEX User Notes, ILOG inc.
615
Optimization of timber trusses

considering joint flexibility
S. ilih, M. Premrov & S. Kravanja
University of Maribor, Faculty of Civil Engineering, Slovenia
Abstract
This paper presents the optimization of metal-plate-connected plane timber
trusses with special emphasis on joint flexibility. The optimization is performed
by the non-linear programming approach. Since various truss design parameters
such as type of truss configuration, span/depth ratio, number and type of
diagonal and vertical members and type of joint connections simultaneously
affect each other, all of these parameters are proposed to be considered
simultaneously in a single mathematical model. The optimization model for cost
optimization of timber trusses was thus developed. The economic objective
function was defined to minimize the structures self-manufacturing costs,
subjected to the design, stress and deflection (in)equality constraints. The finite
element equations were as the equality constraints defined for the calculation of
the internal forces and the deflections of the structure. The stiffness matrix of the
structure was composed by considering the fictiously decreased cross-section
areas of all the flexible connected elements. Constraints for the dimensioning of
the timber members were determined in accordance with Eurocode 5 in order to
satisfy the requirements of both the ultimate and the serviceability limit states.
The cross-section dimensions and the number of fasteners were defined as
independent optimization variables. A numerical example demonstrates the
applicability of the presented approach.
Introduction
Timber construction is an important part of the infrastructure in a number of

areas around the world. Wood has proved to be quite a resilient material,
showing relatively high ductility and low density. In addition, the flexibility of
mechanical fasteners provides a high damping capacity between the connected
doi:10.2495/HPSM06060

timber elements. Well-built timber structures maintain a good performance
particularly under the influence of wind and especially earthquake forces.
In the last decades, the application of timber trusses has frequently been
noticed in all aspects of building construction. Timber trusses have become
known for their pleasing architectural appearance, lightweight design and easy
fabrication. The use of timber trusses to bridge over large open areas with a few
or no intermediate supports is still on the increase. These trusses are essentially
lighter than the analogous beam solutions. Many magnificent space and plane
timber trusses have been constructed all over the world, covering public halls,
stadiums, exhibition centres and many other buildings. In this field, metal-plateconnected timber trusses have been found to be favourable structures for roof
framings for spans greater than 20 meters.
In order to design a satisfactory and optimal timber truss structure with the
given load, span and boundary conditions, some main design parameters need to
be considered, on which timber truss behaviour basically depends:
type of truss configuration,
span/depth ratio,
number and type of intermediate members (diagonals and verticals),
type of joint connections.

When a high number of truss design parameters, designer decisions and
factors are involved in the analysis, the designing of timber trusses can become a
difficult and expensive process. This has forced designers to find simpler and
cheaper alternative design methods, adequate at least for the preliminary design
state. Several approximate methods have been developed in the recent past with
different accuracies of suitability and simplification according to real truss
conditions, see [1]. Approximate designing methods which additionally consider
the flexibility of the joints in timber trusses with respect to different diagonal
members can for example be found in [2-5].
The idea of the present study was to together simultaneously consider all
the mentioned design parameters and factors in a single mathematical truss
model, where structural optimization is performed rather than classical analysis.
For more than four decades, trusses have not only been successfully
optimized but also very frequently used to present, test and improve various
optimization techniques. Numerous research papers on this topic have been
published since the early 1960s, e.g. [6]. While many papers discuss the
topology, shape and discrete sizing optimization particularly of steel trusses, e.g.
[7-10]; also the optimization of composite trusses, [11]; timber trusses have been
quite neglected.
The paper presents the sizing optimization of metal-plate-connected timber
trusses considering the flexibility of the embedded fasteners. The optimization
was performed by the non-linear programming (NLP) approach, where all the
mentioned design parameters were simultaneously considered as (in)equality
constraints. The optimization model for the cost optimization of the timber
trusses was developed. An economic objective function was proposed to
minimize the structures self-manufacturing costs, subjected to the design, stress
617
and deflection (in)equality constraints. The finite element equations were as the
equality constraints defined for the calculation of the internal forces and the
deflections of the structure.
Timber truss design criteria
The design constraints for the timber trusses were determined in accordance with
Eurocode 5 [12] in order to satisfy the requirements of both the ultimate (ULS)
and the serviceability limit state (SLS). Considering the ULS, the truss members
were checked for the tensional as well as the compressive/buckling resistance.
The required number of fasteners was also calculated for each joint. At the SLS
the vertical deflections of the truss girders were checked.
Since the bracing members (diagonals and verticals) are flexibly connected,
their stiffness decreases. In finite element analysis we consider the joint
flexibility in such a way that cross-sectional areas Am of all bracing members are
replaced by a fictiously decreased cross-section area Am* [4]:
A*m =
Am
E m ,mean Am
1
1
1+
+
K k
Lm
K
ser k m ,2
ser m ,1
(1)
where km,1 and km,2 represent the numbers of fasteners at both ends of the
considered m-th bracing element and Kser denotes the fasteners slip modulus,
taken for different types of fasteners from Table 7.1 of Eurocode 5. Em,mean stands
for the mean value of the modulus of elasticity.
Optimization of timber trusses
As the optimization problem of timber trusses is non-linear, e.g. the objective

function and (in)equality constraints are non-linear, the non-linear programming
optimization (NLP) approach is used. The general NLP optimization problem
can be formulated as follows:
Min z = f(x)
subjected to:
h(x) = 0
g(x) 0
x X = { x x Rn, xLO x xUP }
(NLP)
where x is a vector of continuous variables, defined within the compact set X.

The variables x are calculated between their lower and upper bounds xLO and
xUP. Functions f(x), h(x) and g(x) are non-linear functions involved in the
objective function z, equality and inequality constraints, respectively. All
functions f(x), h(x) and g(x) must be continuous and differentiable.

In the context of structural optimization, variables include dimensions,
cross-section characteristics, strains, materials characteristics, stresses, economic
parameters, etc. Equality and inequality constraints and the bounds of the
variables represent a rigorous system of the design, loading, stress, deflections
and stability functions taken from the structural analysis. The optimization of the
structures may include various objectives worthy of consideration. The most
popular criterion used today is the minimization of mass. In this paper, an
economic objective function is proposed to minimize the structure's selfmanufacturing costs. Hence, the trade-offs between different materials can be
appropriately accounted for.
The optimization model TTO (Timber Truss Optimization) for the
optimization of timber trusses was developed according to the above NLP model
formulation. GAMS (General Algebraic Modeling System), [13], was used as the
interface for mathematical modeling and data inputs/outputs.
A general timber truss layout with its characteristic elements (upper and
lower chord, verticals, diagonals) is presented in Figure 1, where xi represents the
local longitudinal axis of element i, while yi and zi represent the principal axes of
the cross-section of the element i; the axes X and Y form the global coordinate
system of the structure. Ai and li stand for the cross-section area and the length of
member i, respectively. The cross-sections are considered to be rectangular,
where bi and hi represent the width and the height of the cross-section of the truss
member i.
xi
upper chord
i,
li
vertical
A
i,
dia
l
na
go
hi
X
lower chord
Figure 1:
bi
A i = bihi
Plain timber truss.
An economic objective function is defined in the model to minimize the

structures self-manufacturing costs, subjected to design, stress and stability
constraints, known from structural analysis. Internal forces are proposed to be
determined by the finite element equations, while the dimensioning is performed
in accordance with Eurocode 5. The objective function is thus defined:
min
cost = ct bi hi li + c fm + c fl 2 k m + c fm + c fl k j
i =1
m =1
j =1
(2)
where cost represents the self-manufacturing (material and labour) costs of the
structure; ct denotes the price of the manufactured and embedded timber material
per m3; the sum of the products between widths bi, heights hi and lengths li of i,
iI, timber members represents the volume of the truss in m3 (see Figure 1); cfm
is the material cost of one fastener together with the adjoining steel plates, while
619
cfl denotes the manual labour costs for handling, assembling, drilling and bolting,
defined per one fastener. Considering that the required number of fasteners is
equal for both ends of each intermediate member, the total number of fasteners in
the m-th member is 2km, where km = km,1 = km,2. The last term of the objective
function represents the sum of all fasteners required in joints of the chord
members. Variable kj stands for the number of fasteners of the j-th joint, jJ. It is
evaluated considering the resultant force on account of the axial forces of all
intermediate elements connected to joint j. Since the dimensions of steel plates
depend directly on the number of calculated fasteners, the costs of steel plates are
included in the values cfm and cfl.
The input data of the optimization model is the truss geometry (coordinates
of joints), the supporting and loading conditions, the diameter of the considered
fasteners, the thickness of the metal plates, as well as the material characteristics
of all the used components (timber, fasteners, plates).
The cross-section dimensions bi and hi of i, iI, truss timber members and
the number of fasteners km and kj are defined as independent optimization
variables.
The finite element equations for the calculation of internal forces and
deflections of the structure are defined as equality constraints. The stiffness
matrix of the structure is composed by considering the fictiously decreased
cross-section areas of all the intermediate timber elements (diagonals and
verticals) in accordance with Eq. (1). The ULS and SLS design conditions,
described in Section 2 are defined as inequality constraints.
Numerical example
As a numerical example, a timber truss girder of 30 m span is presented. The

layout of the structure is shown in Figure 2. The truss elements are composed
from a glued laminated timber GL32h according to the EN 1194 [14]
classification. M14 dowels made of steel S 235 are used as fasteners. Steel plates
of 8 mm thickness made of steel S 235 are additionally placed as the central
members of a double shear connection.
The truss is subjected to a permanent load of 2 kN/m and a variable load of
5 kN/m (snow load). The self-weight of the truss members was automatically
determined through the optimization process with respect to the actual calculated
cross-sectional dimensions. The uniform loads are, in the calculation,
approximated as nodal forces (see Figure 2).
Fi
i
F = qL
h0
L = 30 m
Figure 2:
Layout of the timber truss.

The economic parameters considered for the objective function (2) were
costs ct = 900 EUR/m3 for the GL32h timber, while cfl = 1.5 EUR and cfm = 1.0
EUR for one M14 S 235 dowel (including the corresponding parts of the steel
plates). The lower bound on the cross-section dimensions bilo/ hilo were taken as
12/12 cm.
Three different optimization cases of the considered truss were performed,
namely A, B and C. At truss A the maximal height h amounted to 400 cm
(span/depth ratio = 7.5), while at truss B a lower value of h = 250 cm (span/depth
ratio = 12) was considered. In both examples A and B the flexibility of joints
was taken into account. In addition, the lower truss (h = 250 cm) was optimized
again by ignoring the flexibility of joints (truss C). In all cases the inclination of
the upper chord amounted to 7.5%. The vertical displacements of joints at SLS
were limited to L/300 = 10 cm.
Since the NLP optimization handles continuous variables, the obtained
variables of the final/optimal solution take some real values between their
defined lower and upper bounds. At this stage, the structure is fully exploited
considering either the ultimate or the serviceability limit state design conditions.
As the aim of this research is to obtain a structure of practical applicability, the
final continuous optimal solution was reanalysed with the variables rounded to
their nearest upper integer values (i.e. 1 cm for cross-sectional dimensions and 1
fastener). The NLP optimization was performed by the computer program
GAMS/CONOPT2 [15].
Table 1:
The obtained optimal results.

Truss A
Truss B
Truss C
[EUR]
2984.33
3575.14
3322.59
[kg]
1291.29
1486.19
1356.99
16/13
18/15
17/21
21/17
17/20
21/17
181
248
244
Max. deflection without

considering joint flexibility
44.82
79.85
82.62
Max. deflection by considering

joint flexibility
59.88
95.79a
Total costs
Timber mass
Chord dimensions b/h [cm]

lower chord
upper chord
Total number of dowels
It should be noted, that the max. displacements of truss B is not equal to L/300 = 100 mm due to the
rounding of cross sections before the reanalysis. After the first optimization phase (continuous
optimization), the SLS conditions were active and the displacement was equal to 100 mm.
The obtained results of the three performed optimizations are presented in

Table 1. In order to arrive to appropriate conclusions, it is convenient to present
the final results by comparing the optimal costs and the obtained masses between
621
different truss alternatives, see Table 2. The influence of the height of the truss
on the final results is presented by the ratio B/A, which represents the ratio
between the obtained costs (masses) of the truss B (h = 250 cm) and A (h=400
cm). The possible error, caused by neglecting the flexibility of the joints in the
timber trusses, is exposed through the ratio B/C which represents the ratio of the
obtained costs (masses) between truss B (considering joint flexibility) and C
(neglecting joint flexibility).
Table 2:
Comparison of timber mass and total costs.
costs ratio
timber mass ratio
B/A
B/C
B/A
B/C
1.198
1.076
1.151
1.095
From the results presented in Table 1 it is evident, that truss A represents

the best solution. At the final result the deflection constraints are not active i.e.
the ULS design conditions are decisive. A lower span/depth ratio of the truss is
thus more favourable. Despite lower timber mass and total costs, the vertical
deflections of Truss A are considerably smaller when compared to the
deflections of truss B. However, both the results of trusses A and B show a
significant influence of joint flexibility on the final displacements. When joint
flexibility is considered, the deflections increase by over 30 percent (truss A).
Comparing trusses B and C, it is evident that the influence of fastener
flexibility should not be neglected at all. The total costs increase by 7.6%, and
the timber mass by almost 10%. It should be noted, that a more detailed study,
presented in [16], has shown, that the influence of joint flexibility increases with
a higher number of flexibly connected bracing members (diagonals and
verticals).
Conclusions
The paper presents the optimization of metal-plate-connected plane timber

trusses with respect to joint flexibility. The optimization was performed by the
non-linear programming approach. Since various truss design parameters like the
type of truss configuration, the span/depth ratio, the number and type of diagonal
and vertical members as well as the type of joint connections simultaneously
affect each other, all these parameters were proposed to be simultaneously
considered in a single mathematical model.
The optimization model TTO (Timber Truss Optimization) for cost
optimization of timber trusses was thus developed. The economic objective
function was defined in order to minimize the structures self-manufacturing
costs, subjected to the design, stress and deflection (in)equality constraints. The
finite element equations were as the equality constraints defined for the
calculation of the internal forces and the deflections of the structure. The
undesirable slips in the connections of timber trusses additionally resulted in the

reduction of truss stiffness. A stiffness matrix of the structure was therefore
composed by considering the fictiously decreased cross-section areas of all the
flexibly connected elements. The constraints for dimensioning the timber
members were determined in accordance with Eurocode 5 in order to satisfy the
requirements of both the ultimate and the serviceability limit states. The crosssection dimensions and the number of fasteners are defined as independent
optimization variables. The obtained results are not only optimal, but the
optimization also enables that the conditions of either the ultimate or the
serviceability limit states are fully exploited and there is no reserve in the
resistance of the structures. At this point the comparison between results of
different truss design alternatives was performed.
The presented numerical example shows the applicability of the presented
optimization approach as well as the influence of considering the fasteners
flexibility on the optimal self-manufacturing costs. Based on numerical results, it
is recommended to design higher timber trusses with a lower span/depth ratio,
with a smaller number of diagonal and vertical elements and, consequently, by
using chord elements with smaller cross-sections.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
Hoyle RJ, Woeste FE. Wood Technology in the Design of Structures,

Ames, Iowa: Iowa State University Press, 1989.
El-Sheikh A. Approximate Analysis of Space Trusses. International
Journal of Space Trusses 1996; 11: 321-30.
Stalnaker J, Harris E. Structural Design in Wood, NY: Van Nostrand
Reinchold, 1989.
Steck G. Fachwerbinde aus Brettschictholz un Vollholz, Holzbauwerke:
Bauteile Step 2, Konstruktionen, Details nach Eurocode 5, Dsseldorf:
Fachverlag Holz, 1995.
Brninghoff H et al. Eine Ausfhrliche Erluterung zu DIN 1052, Teil 1
bis Teil 3, Beuth Kommentare, Berlin: Beuth Bauverlag, 1988.
Schmidt LA Structural Design by Systematic Synthesis. Proceedings of 2nd
Conference on Electronic computations; NY: ASCE, 1960, 105-22.
Lipson SL, Agrawai KM. Weight Optimization of Plane Trusses. ASCE
Journal of the Structural Division 1974; 100: 865-79.
Prager W. Optimization of Structural Design. Journal of Optimization
Theory and Applications 1970; 6: 1-21.
ilih S, Kravanja S, Bedenik BS. Shape optimization of plane trusses. In:
Hendriks, MAN, Rots JG editors, Finite Elements in Civil Engineering
Applications, Proceedings of the Third DIANA World Conference,
Tokyo, 2002, 369-73.
Kaveh A, Kalatjari V. Topology optimization of trusses using genetic
algorithm, force method and graph theory. International Journal for
Numerical Methods in Engineering 2003; 58: 771-91.
Kravanja S, ilih S. Optimization based comparison between composite I
beams and composite trusses. Journal of Constructional Steel Research
2003; 59: 609-25.
[12]
[13]
[14]
[15]
[16]
623
CEN/TC 250/SC5 N173, Eurocode 5: Design of Timber Structures, Part

1-1 General rules and rules for buildings, Final draft prEN 1995-1-1,
Brussels, European Committee for Standardization, 2002.
Brooke A, Kendrick D, Meeraus A. GAMS - A User's Guide, Redwood
City, CA: Scientific Press, 1988.
EN 1194, Timber structures Glued laminated timber Strength classes
and determination of characteristic values, Brussels, European
Committee for Standardization, 1999.
Drud AS. CONOPT A Large-Scale GRG Code. ORSA Journal on
Computing 1994; 6:207-16, 1994.
ilih S, Premrov M, Kravanja S. Optimum design of plane timber trusses
considering joint flexibility. Engineering structures 2005; 27: 145-154.
625
The HendrickxVanwalleghem design strategy

W. Debacker1, C. Henrotay2, W. P. De Wilde1 & H. Hendrickx2
1
Department of Mechanics of Materials and Constructions (MeMC),

Vrije Universiteit Brussels, Faculty of Engineering, Belgium
2
Department of Architecture (ARCH), Vrije Universiteit Brussels,
Faculty of Engineering, Belgium
Abstract
The optimisation of structures and materials is a justifiably popular engineering
topic. Contemporary research is concentrated, among others, into cost
minimisation, structural efficiency and intelligence, in compliance with
environmental and social preservation. As a counterpart this paper puts the
accent on the time dependent aspect of constructions, such as the life cycle cost,
the possibility to make (non-) structural changes and recycling or reuse of
building material. In search of an optimisation of this dynamic efficiency of
constructions, a design strategy has been developed at the Vrije Universiteit
Brussels (Departments of MeMC and ARCH). This strategy is presented here. It
considers the temporal character of constructions from the first sketches
onwards.
Keywords: adaptability, reuse, design strategy, construction kits, generating
system, temporal/temporary.
Introduction
In society buildings have been - and still are - designed in terms of end states.
The moment the first sketches are drawn, the constructions finality is planned or
denied. Because of their static nature, which they acquired ab initio, most
buildings are not suited to meet the demands of a quickly changing society. As a
result many building components end up as waste or are brought back in
circulation by means of expensive and consuming industrial processes. Changing
functions, quickly evolving living and building trends, the amount of new
materials and improved techniques are some mutations the built environment
has to go through and has to provide appropriate answers to. Although humans
doi:10.2495/HPSM06061

have to cope with an unpredictable future, full of uncertainties, there is one
universal constant: the environment always changes! Hence, a sustainable built
environment requires a dynamic concept; a step-by-step redesign process of
gradual changes in which no end states or final goals can be defined [2].
The strategy HendrickxVanwalleghem [3] includes this dynamic view on
the built environment. By designing adaptable construction systems, which are
compatible with each other, a dynamic and by this a sustainable answer can
be given to an unexpected and unpredictable future. These construction systems
are made of a minimum number of basic elements and a set of combination rules.
They allow the conversion of each artefact to a different configuration, by means
of adding, removing or transforming the basic elements which it is made of. It
offers a high potential of recycling and (direct) reuse. The outcome can be
compared with the Meccano building set, which, in this view, encloses all
materials and techniques, and is applicable to all scales.
Hendrickx and Vanwalleghem proposed a set of standardisation rules, which
they called a generating form and dimensioning system, The generating system
is a central concept in the design strategy, in the sense that it ensures full
compatibility of form and dimensions between all basic elements. The rules are
translated into a fractal model, based on basic forms, such as the square, the
inner circle and its diagonal, and a dimensional range using the operator
multiply or divide by 2 (Fig. 2).
The HendrickxVanwalleghem strategy
Hendrickx and Vanwalleghem developed a dynamic design strategy (further

called HendrickxVanwalleghem strategy). It allows for the design of flexible
construction systems, based on a minimal number of elements and combination
rules.
The design strategy is explained in Fig 1. It consists of 4 layers. The lower
layer is characterised by a material solution, using 2 design tools, shown in the
upper conceptual layer: the generating form and dimensioning system and the
theoretical design catalogues.
The text hereafter explains the set-up of the design strategy, starting with the
material solutions (layer 4).
2.1 Layer 4: Adaptable material solutions
According to European standards [4], a durable house is designed for a life cycle
of 50 years. It is easy to understand that such constructions will undergo radical
transformations and repairs during this extensive period. The Hendrickx
Vanwalleghem design strategy has consequently been developed for
constructions subjected to these transformation processes.
In Fig.1 the reader can see how a minimal habitation unit can be transformed
into alternative configurations. This minimal habitation unit provides its user(s)
with basic functionalities. It can be expanded from its core, according to
evolving possibilities of user(s) and/or the building site. This principle has been
627
successfully applied in (developing) countries, such as India, as it takes into

account a gradual development of the users and the possibility to perform
reversible changes. Of course, such processes should make use of simple and
cost effective building methods.
2.2 Layer 3: Adaptable construction elements
As a consequence of previous argumentation, construction elements must be
developed to be easily adaptable and reusable. This should be done not only for
the non structural elements but also for the load bearing ones, such as walls,
floors, foundations, columns, beams (Fig.1) and, even more important, the
connections. Nevertheless, all these elements can be deconstructed into very
simple basic elements.
2.3 Layer 2: Compatible construction kits
The basic elements can be compared to the letters of an alphabet: they do not
carry a semantic meaning. The basic elements can be combined in different ways
and form a variety of construction elements, which in the actual comparison can
be considered as words. Three types of basic elements can be discerned:
-
line elements (one-dimensional)

plane elements (two-dimensional)
volume elements (three-dimensional)
Figure 1:
The HendrickxVanwalleghem design strategy.

Point elements (zero-dimensional) can be categorised in the previous
classification: e.g. bolts can be considered as scaled volume elements.
A construction kit is defined as a grouping of a FEW simple basic elements
aiming at the assembly of one or more adaptive construction elements. One must
be aware that the building of a complete construction will very often require
several construction kits. It is thus absolutely necessary that all basic elements
found in the same construction kit or even another one, be compatible with each
other. The establishment of explicit standardisation rules is here for stringent.
2.4 Layer 1: Development of adaptable construction systems
2.4.1 Design tool 1: a generating form and dimensioning system
Using their own developed design tool, called a generating form and
dimensioning system, Hendrickx and Vanwalleghem proposed a set of
standardisation rules. It is a central concept in the design strategy, in the sense
that it ensures full compatibility of form and dimensions between all the simple
basic elements.
Hendrickx and Vanwalleghem presume that any tangible element, in any
construction phase, can be approximated with a minimal diversity of basic forms.
They have chosen the square, its diagonals and the inscribed circle, due to an
important property of the former, i.e. its orthonormality. This makes sense since
right angles are found in many material solutions and certainly in the area of
construction.
Figure 2:
Model of the generating system.
To make effective use of the proposed system, the set of basic forms should
be provided with basic dimensions. In order to achieve optimal flexibility and
combination, the basic elements should have the same dimensions. If differences
are unavoidable HendrickxVanwalleghem proposed to solve the problem
using the rules of either halving or doubling. Both are the result of an easy
mathematical manipulation and create a geometrical series. Halving is easy if
one uses flexible elements: they can be folded. Starting with a square with side
x one finds: x, 2x, 4x, 8x, 16x
The fractal model in Fig. 2 can be projected on all materials and all scales and
thus can define the basic elements for each material type. Grouping ALL possible
629
variations within a chosen set of basic elements is named as a construction

system. The types of basic element (first order elements) are defined by both
their form and their constitutive material, as these define how basic elements
should be joined and combined into basic elements of a higher level. Here the
attention should be drawn to the fact that a construction system is not an object,
but rather a set of entities, i.e. smallest elements of the system, between which
predefined relations exist. Those are the dimensional and formal rules imposed
by the generating system. The concept of construction kit can thus be
redefined as a rational selection of SOME basic elements out of one ore more
construction systems. The objective is to generate one or more flexible
constructions and their constitutive parts.
2.4.2 Design tool 2: theoretical design catalogues
An aid to the development of construction systems is achieved developing
theoretical design catalogues. This development is carried out in the following
way.
In a first step each material solution, or more precisely each of its
construction elements, in whichever phase, is objectively and verbally described
based on characteristics, strengths and weaknesses. Each characteristic has one
or more parameters as a counterpart, all bracketed between predefined limits.
This delimitation, for each parameter, is done at the level of the entities.
Considering that all artefacts are measurable and can be depicted, most of the
parameters can be visualised with simple symbols or pictograms and be
categorized in different series. If a graphic representation is not wanted or
impossible, a short verbal description will be sufficient. Through interpolation
and/or combination of the outer elements in the series all variants can be
achieved.
Figure 3:
Theoretical design catalogue of a load bearing corrugated plate.
This can be illustrated with a simple construction element. The bearing

capacity of a steel corrugated plate, subject to transverse loads e.g. used as a
roof element can be described with three parameters: its thickness, the number

of waves per unit length and the height of the waves (or also the predefined form
of the wave). These parameters can be entered into series by defining the
extreme values. A thin plate is the opposite of a thick one: this means that there
is a limit for the ratio thickness/span. The number of waves per unit length and
the height of the waves are limited by plate thickness and fabrication process. All
variants should lie between the thus defined limits.
With arithmetic and geometric calculation rules changes in each series can be
described; every value of the parameter thus gets its place within the series.
Whether the succession of elements is continuous or discontinuous has no
importance, at least on the theoretical level. But practically seen stepwise
variations are preferred: they reduce the number of values and more easily
achieve the goal of using a minimal number of (standardised) basic elements.
The adopted geometrical rule of the game stems from the fractal model in the
generating system, which can easily be seen in following sketch:
Figure 4:
Compatible basic elements.
If all series of individual parameters are grouped, any historical, actual or

future artefact can be theoretically described by a series of lines or vectors which
intersect in a common point. Variations or new solutions are found by translating
one or more lines through the intersection point.
A theoretical design catalogue is thus established, combining and juxtaposing
elements. The emphasis has been put on theoretical, as in practice not all
combinations are possible or technically sound. This means that they are erased
in the practical catalogue.
This approach can be applied to a whole construction. As the bracketing of an
entire construction is difficult, the preference is given to deconstructing it into
physically and non-physically observable elements, without forgetting to keep in
mind the global context. For example, the maximum height of a building is often
defined through ridge height or roof slope of adjacent constructions.
631
The concluding result is a fan of design catalogues, each one based on

combinations of selected parametric rules. They allow to describe any artefact,
existing or not, through translation of one or more series.
The above presented strategy matches perfectly with the idea of open
industrialisation, wherein a minimum of construction elements, belonging to
several construction systems and distributors, based on the same design rules,
can be combined together to form multiple (adaptable) projects [5].
Temporary character of constructions
The awareness of the limited life time of our heritage is ever increasing; slums
are demolished, old train stations are replaced by prestigious ones, offices are
refurbished and monuments are carefully renovated or the object of restoration
Still professional developers and real estate owners pay little attention to the
temporary character of a construction: even during the study and the drawing
phase this aspect is often forgotten or even simply ignored. The point is that if
you want to face changing uses during the life time of the construction, static
solutions will make transformations extremely difficult if not impossible. It
could happen that some structural elements still perform in a satisfactory way,
but the owner will often prefer to demolish and start over. This causes a lot of
debris. Consequently, the actual society is missing a dynamic design strategy,
allowing transformations and adaptations during the life cycle of a construction.
The Hendrickx Vanwalleghem design strategy takes these characteristics of
temporality and adaptability of constructions into account, from the first sketches
on. It allows every construction part to transform into another configuration by
adding, deleting and transforming basic elements of the same system and
combining it with elements coming from other. One of the consequences is that
so-called dry connections are used: bolts and nuts, screws, click-systems
[6, 7].
Figure 5:
Actual model of the life

cycle of a construction.
Figure 6:
Proposed model of the

life
cycle
of
a
construction.

Damaged basic elements can be reused allowing small transformations, even
in fields where the structural, i.e. load bearing properties, are less important
(furniture, window frames). Hence the total life cycle cost of the elements
will decrease.
Modular versus generating
The main asset of a modular construction system is an economical one. Thanks

to (modular) standardization, simplified and cheaper prefabrication processes are
made possible, which consequently speeds up the construction phase. Modular
construction systems are also known as flexible. However, this is not without
any shortcomings!
Figure 7:
The 4 design levels [8].
Changing a module or unit is excluded, because it is has been technically and

structurally denied. Adaptability and by this the designers freedom is
therefore limited to the addition and reduction of fixed modules. A common
employed module is the foot (in the horizontal plane). This module is approx.
30cm and is rightfully successful as a functional, ergonomic and spatial unit. But
it cannot be used at all levels of the design: for technical dimensions it is often
too large, for structural purposes too small (Fig.7). A multi-modular grid
provides an improvement; i.e. a superposition of modular design grids with a
different module related to the respective design level (structural, spatial,
functional or technical). Design at different levels is thus possible, but not
without possible conflicts. Using an arbitrary or no mathematical relation
between the module sizes dimensional problems occur where different grid lines
intersect (Fig. 8). [8]
The standardisation rules of the generating system are based on a fractal
model (Fig.2). Thanks to a single operator (divide or multiply by 2), switching to
different design levels is always possible, and this without jeopardising
compatibility between each basic element. A generating system thus allows the
633
development of (multi-) modular systems, but with the additional property that
they can be used with different design scales.
Furthermore, it is not the module which is standardised but the
(dimensional) modifiable basic elements from which it is composed. The latter is
the key difference with modular construction systems.
Figure 8:
5M
8M
3M
4M
2M
2M
Multi-modular grid.
Figure 9:
Generated compatible
design grids.
Conclusion
The Hendrickx-Vanwalleghem design strategy offers a sustainable answer to the

optimisation of dynamic characteristics of the construction world. Thanks to its
explicit standardisation rules, it maximises adaptability and reuse possibilities
during the constructions entire life. Full compatibility between basic elements of
the same construction kit and elements belonging to others, makes an
unequivocal generating system even more efficient than modular systems.
References
[1] De Wilde W.P., Conceptual design of lightweight structures: the role of
morphological indicators and the structural index. Proc. of 3rd International
Conf. on High Performance Structures and Materials 2006, ed. Wessex
Institute of Technology, United Kingdom, 2006.

[2] Hendrickx, H., Vanwalleghem, H., Sustainable Urban Development. Local
Agenda 21 in Development Perspective, 48p. http://home.tiscali.be/
momentumcongressen/Paper%20Van%20Walleghem.pdf
[3] Hendrickx H., Solutions derived from natural processes harmonising nature
and material culture. Proc. of the 1st Conf. on Design and Nature 2002.
Comparing Design in Nature with science and engineering, ed. C.A.
Brebbia & L.J. Sucharov, Wessex Institute of Technology, United Kingdom
and P. Pascolo, Universita degli di Udini, Italy, 10p. 2002.
[4] European Committee for Standardization (CEN), Eurocode Basis of
structural design: NBN EN 1990, pp. 23-26, 2002.
[5] Henrotay, C., Research report. A generally applicable building system for
shelter in emergency situations. , Vrije Universiteit Brussels, Brussels,
Belgium, 2005
[6] W. Debacker, Sustainable development of the material culture approached
by system thinking, Masters dissertation, Vrije Universiteit Brussels,
Brussels, 2003.
[7] C. Henrotay, Sustainable Development. A general description and research
of its application towards the material support in emergency situations.
Masters Thesis, Vrije Universiteit Brussels, Brussels, Belgium, 2003
[8] De Troyer, F., Naert, N., Grids: help or obstacle?. Building with prefab
concrete: design guide, ed. FEBELCEM / FEBE, Brussels, Belgium, 12p.
635
The optimization of a truss facade

B. Verbeeck1 , W. P. De Wilde1 & Ph. Samyn2
1 MeMC,

Samyn and Partners, Belgium
Abstract
This paper presents the study of a facade. One of the aims of the designing team is
the minimisation of the weight of the facade. Due to architectural constraints, the
geometry of the facade is fixed to a highly statically indeterminate truss. Therefore,
the minimisation of the weight of the facade is limited to a section optimisation.
The objective function is the indicator of volume. Since the number of members
is of the order of 1000, exhaustive search methods are impractical. We used a
simple iterative process to find optimal sections. In the first iteration all sections
are equal. This allows the forces to flow through the facade as if the facade has a
constant stiffness. In the next iterations sections are adapted to the forces that arise
from the previous iteration. This method quickly converges to an optimal section
layout. The results of this method are corroborated by a genetic algorithm. We find
that the truss facade with optimal sections consumes less material than an arch
with push rods, that transfer the forces to the arch. Furthermore, the influence of
buckling can be ignored.
Keywords: morphological indicators, optimization, genetic algorithm, statically
indeterminate, truss.
1 Introduction
This paper describes a methodology to optimize sections of a (highly) indeterminate truss. It is especially applicable for fixed geometries, where only sections
need to be optimized. The method is iterative and consists of updating the stiffness of the members according to forces in that member in the previous iteration.
The main advantage of this process is that it consumes considerable less time than
an ordinary optimization algorithm (genetic algorithm). Typically it will take the
process on the order of 10 iterations before convergence occurs. This means that
doi:10.2495/HPSM06062
Figure 1: Geometry of the facade, which spans 48.6m.

only on the order of 10 finite element analyses are performed (compared to 104 or
more for genetic algorithms). The application of the method is a facade of a large
building. A basic geometry is calculated to illustrate the principle. The results for
an alternative model is shown and compared to an arch. The update method we
use was first proposed in [2].
1.1 Facade
The facade is composed of 9 by 12 cruciformed modules, which measure respectively 5.4m and 3.5m. Due to the constraint that a train has to pass under the
facade, supports are only possible at both extremes (Figure 1). Forces are applied
on all nodes. On the top nodes the force is 306kN (roof). The force on all other
nodes is 89kN .
The other model (Figure 2) has a larger span, but a middle support.
1.2 Morphological indicators
In order to compare the different solutions of the facade, an objective function is
necessary. The aim of the designers is the minimization of weight (or volume).
We therefore use the Morphological Indicators (MI). These dimensionless numbers enable the designer to compare different structural systems on the basis of
weight. The most important MI, the Indicator of Volume W = FVL allows the
comparison of the volume of material for different structural systems. It is the volWIT Transactions on The Built Environment, Vol 85, 2006 WIT Press
637
Figure 2: Alternative model (forces are shown).
ume of an isomorphic structure with unit span, with at least one section of each
element dimensioned on its unit allowable stress, subjected to a load system with
E
compares the displacement
unit resultant. The Displacement Indicator = L
of different structural systems. It is the maximum displacement of an isomorphic
structure with unit span in a material with unit Youngs Modulus, with at least
one section of each element dimensioned on its unit allowable stress, subjected
to a load system with unit resultant. The analytical expressions of both W and
have been established in [3] and [2] for trusses, beams, arches, cables, cable stayed
structures, masts and frames subjected to a limited number of (simple) load cases
and support conditions. For statically determinate structures these Morphological
L
if instabilities, self
Indicator are only function of the geometrical slenderness H
weight and second order effect are neglected. Efficiency curves with respect to
minimum volume of used material can be established (Figure 3).
2 Method
The first iteration consists of making all sections equal, thus making all modules
equally stiff. The system is then solved, using a simple finite element program.
The solution can be interpreted as follows: it is a set of sections that is needed
to transfer the loads to the supports, if the stiffness is homogeneously distributed
over the framework (Figure 4). In reality the stiffness will not be homogeneous,
Figure 3: Efficiency curves.
since member sizes vary. Therefore an update of the framework stiffness will yield
a more accurate flow of forces. This is done in the second and every successive
iteration, until convergence of the objective function W .
3 Results
The process converges rapidly (Figure 5(a)). We observe that after 1 iteration the
value of W is 1.38. After 10 iterations this value becomes 1.27. The optimum
is reached after 31 iterations (W = 1.26) (Figure 5(b)). The objective function
decrease between the first and final (best) iteration is 10%. This decrease becomes
639
Figure 4: Forces in framework after 1 iteration.
1% between iteration 10 and the final one. This means that early on, the process
gives a very good estimation of the optimum. Of course this method has to be
compared to other optimization techniques. Using the genetic algorithm toolbox
of MatLab, a slightly better optimal solution of W = 1.20 is reached. However, it
takes on the order of 104 more time to reach this optimum. This has consequences
on the design process: because one finite element analysis takes about 2 seconds,
the answer of the iterative process is known after 1 minute, whereas the GA yields
its result after 5 hours. The advantage of having a slightly better solution does
not weigh up against the enormous loss in calculation time, especially during the
conceptual design stage.
The results are obtained by incorporating buckling. Optimal solutions vary
slightly when buckling is ignored. There are two reasons for this: member lengths
2
are small and sections are buckling efficient, with form factor ( H
= 8 with
I
section, H height and I moment of inertia) low (circular tubes).
The alternative geometry (Figure 6) yields a W -value of 1.07.
The previous solution obtained by the update iteration is compared to an arch
with push rods. According to [3] the value of W of this structural system is 1.13.
This means that the (alternative) truss facade (will transfer the vertical loads more
efficiently than an arch with push rods. The inclusion of horizontal (wind) loads
will only widen the gap between the two structural systems, because the arch is
inefficient at transferring this type of loads. It is not surprising that the truss solutions exhibit an arch shape of its most compressed members and a suspension cable
shape of its most tensioned members.
(a) evolution of the objective function
(b) final solution
Figure 5: The result of the iterative process.
Figure 6: Result for alternative model.
4 Conclusion and future work

The method proposed in this paper gives the designer of structures a very fast
optimization tool for (highly) statically indeterminate trusses. The process consists
of starting with the truss that consists of members with identical sections. In every
consecutively step, the sections are adapted to the forces in the members of the
641
truss with the previous sections. Within the limits of the conceptual design phase,
the results are accurate enough to guide the designer in her/his choices.
With regard to the application, this method yields member sections that make
the facade very light. The truss solution is even better than the arch with push rods,
wind loads excluded.
A setback of this method is that some boundary conditions (e.g. maximum displacement) are difficult to incorporate. This is due to the method itself. The process
can be used for two situations:
1. fixed geometry: only the member sections are to be optimized;
2. truss members used as grid: the truss works as base structure and members
that are not useful tend to have sections that converge to zero (Figure 7). This
type of optimization echoes the topology optimization method developed by
Bendsoe [1]. The problem will be to find the optimal grid.
Figure 7: Optimization using update method.
References
[1] M.P. Bendsoe. Optimization of Structural Topology, Shape and Material.
Springer, New York, 1995.
[2] P. Latteur. Optimisation et predimensionnement des treillis, arcs, poutres et
cables sur base dindicateurs morphologiques. VUB, Brussels, 2004.
[3] Ph. Samyn. Etude

de la morphologie des structures a` laide des indicateurs de
volume et de deplacement. Academie Royale de Belgique, Brussels, 2004.
643
MINLP optimization of the single-storey

industrial building steel structure
T. ula, U. Klanek & S. Kravanja
University of Maribor, Faculty of Civil Engineering, Maribor, Slovenia
Abstract
In this paper the optimization of the single-storey industrial building steel
structure is presented. The structure consists of the main portal frames, which are
mutually connected with the purlins. All structural elements are proposed to be
built up of standard hot rolled I sections. The structural optimization is
performed by the Mixed-Integer Non-linear programming approach, MINLP.
The MINLP performs a discrete topology and standard dimension optimization,
while continuous parameters are simultaneously calculated inside the continuous
space. Since the discrete/continuous optimization problem of steel frames is
non-convex and highly non-linear, the Modified Outer-Approximation/EqualityRelaxation (OA/ER) algorithm has been used for the optimization. Alongside the
optimal structure mass, the optimal topology (an optimal number of main portal
frames and purlins) as well as all standard cross-section sizes have been
obtained. The paper includes the theoretical basis and a practical example with
the results of the optimization.
Introduction
Single-storey frame structures are extensively used for industrial, leisure and
commercial buildings. In order to obtain efficient frame designs, researchers
have introduced various optimization techniques, appropriate either for the
continuous or the discrete optimization. OBrien and Dixon [1] have proposed a
linear programming approach for the optimal design of pitched roof frames.
Guerlement et al. [2] have introduced a practical method for single-storey steel
structures, based on a discrete minimum weight design and Eurocode 3 [3]
design constraints. Recently, Saka [4] has considered an optimum design of
pitched roof steel frames with haunched rafters by using a genetic algorithm.
doi:10.2495/HPSM06063

One of the latest researches reported in this field is the work of Hernndez et al.
[5], where authors have considered minimum weight design of steel portal
frames with software developed for structural optimization.
This paper deals with topology and standard dimension optimization of
unbraced single-storey industrial building steel structures. The optimization of
portal frames is performed by the Mixed-Integer Nonlinear Programming,
MINLP. The MINLP is a combined discrete and continuous optimization
technique. In this way, the MINLP performs the discrete topology (i.e. the
number of frames and purlins) and standard dimension (i.e. the standard crosssection sizes of columns, beams and purlins) optimization simultaneously with
the continuous optimization of parameters (e.g. a structure mass, internal forces,
deflections, etc.).
The MINLP discrete/continuous optimization problems of frames are in most
cases comprehensive, non-convex and highly non-linear. The optimization is
proposed to be performed through three steps. The first one includes the
generation of a mechanical superstructure of different topology and standard
dimension alternatives, the second one involves the development of an MINLP
model formulation and the last one consists of a solution for the defined MINLP
optimization problem.
The objective of the optimization is to minimize the mass of the single-storey
industrial building. The mass objective function is subjected to the set of the
equality and inequality constraints known from the structural analysis and
dimensioning. The dimensioning of steel members is performed in accordance
with Eurocode 3.
The Modified Outer-Approximation/Equality-Relaxation algorithm is used to
perform the optimization, see Kravanja and Grossmann [6], Kravanja et al. [7-8].
The two-phase MINLP optimization is proposed. It starts with the topology
optimization, while standard dimensions are relaxed temporary into continuous
parameters. When the optimal topology is found, standard dimensions of crosssections are re-established and standard dimension optimization of beams,
columns and purlins is then continued until the optimal solution is found.
Single-storey industrial building
The paper presents the topology and standard dimension optimization of

unbraced rigid single-storey industrial building steel structures, see Figure 1.
Columns, beams and purlins are proposed to be built up of standard hot rolled
steel I sections.
The considered portal frame structures are analyzed under the combined
effects of the self-weight of frame members, uniformly distributed surface
variable load (snow and wind), concentrated horizontal variable load (wind) and
an initial frame imperfection. The purlins are designed to transfer permanent load
(self-weight of purlins and weight of roof) and variable load (snow and wind) to
frame structure. Internal forces are calculated by the elastic first-order analysis.
The dimensioning of steel members is performed in accordance with Eurocode 3
645
for the conditions of both ultimate limit state (ULS) and serviceability limit state
(SLS).
Lf
Lf
Lf
Lf
Lf
Lf
Figure 1:
LL
Single-storey industrial building.
When the ultimate limit state of structural members is considered, the

elements are checked for axial resistance, shear resistance, bending moment
resistance, interaction between bending moment and axial force, interaction
between axial compression/buckling and buckling resistance moment.
Considering the serviceability limit state, the vertical deflections of beams
and purlins are calculated by the force method. The total deflection max
subjected to the overall load and the deflections 2 subjected to the variable
imposed load are calculated to be smaller than limited maximum values:
span/200 and span/250, respectively. The horizontal deflections are also
checked for the recommended limits: the relative horizontal deflection should be
smaller then the height/150 of the portal frame.
MINLP model formulation for mechanical superstructures
It is assumed that a general nonconvex and nonlinear discrete/continuous

optimization problem can be formulated as an MINLP problem (MINLP-G) in
the form:
min z = c T y + f ( x )
h( x ) = 0
g( x ) 0
By + Cx b
s.t.
(MINLP-G)

n
m
y Y ={0,1}
where x is a vector of continuous variables specified in the compact set X and y
is a vector of discrete, mostly binary 0-1 variables. Functions f(x), h(x) and g(x)
are nonlinear functions involved in the objective function z, equality and
inequality constraints, respectively. Finally, By+Cx b represents a subset of
mixed linear equality/inequality constraints.
The above general MINLP model formulation has been adapted for the
synthesis of mechanical superstructures (MINLP-SMS). The resulted
formulation that is more specific, particularly in variables and constraints. It can
be used also for the modelling the steel frames. It is given in the following form:
min z = c T y + f ( x )
s.t.
h( x ) = 0
g(x ) 0
A( x ) a
Ey e
Dy + R( x ) r
(MINLP-SMS)
( )
Py + S (d ) s
Ky e + L d cn k
st
m
y Y ={0,1}
The MINLP model formulation for mechanical superstructures is proposed to be
described as follows:
Included are continuous variables x={d, p} and discrete binary variables
y={ye, yst}. Continuous variables are partitioned into design variables d={dcn,
dst} and into performance (nondesign) variables p, where subvectors dcn and
dst stand for continuous and standard dimensions, respectively. Subvectors
of binary variables ye and yst denote the potential existence of structural
elements inside the superstructure (the topology determination) and the
potential selection of standard dimension alternatives, respectively.
The mass or economical objective function z involves fixed mass or cost
charges in the linear term cT y and dimension dependant mass or costs in the
term f(x).
Parameter nonlinear and linear constraints h(x)=0, g(x) 0 and A(x) a
represent the rigorous system of the design, loading, stress, deflection,
stability, etc. constraints known from the structural analysis.
Integer linear constraints Ey e are proposed to describe relations between
binary variables.
Mixed linear constraints Dye+R(x) r restore interconnection relations

between currently selected or existing structural elements (corresponding
ye = 1) and cancel relations for currently disappearing or nonexisting
elements (corresponding ye = 0).
Mixed linear constraints Kye+L(dcn) k are proposed to define the
continuous design variables for each existing structural element. The space
is defined only when the corresponding structure element exists (ye = 1),
otherwise it is empty.
Mixed linear constraints Py+S(dst) s define standard design variables dst.
Each standard dimension dst is determined as a scalar product between its
vector of standard dimension constants q and its vector of binary variables
yst. Only one discrete value can be selected for each standard dimension
since:
st
yi = 1
d st = qi yist
iI
iI
647
(1)
The optimization model
The optimization model for a single-storey industrial building steel structure

contains the mass objective function, (in)equality constraints and the variables
(dimensions, internal forces, deflections, mass, etc.). Equality and inequality
constraints represent a rigorous system of the design, loading, resistance, stress,
deflections and stability functions. The dimensioning constraints are determined
according to Eurocodes 3 (ultimate and serviceability limit states).
Resistance to bending moment of beams, columns and purlins:
M Sd M el ,Rd
M el , Rd =
Wel f y
M0
(2)
(3)
where MSd is the design bending moment, Mel,Rd is the elastic resistance moment,
fy is the yield strength of structural steel, Wel is the elastic section modulus and
M0 is the partial safety coefficient.
Resistance to axial force of the beams and columns:
N Sd N pl ,Rd
N pl , Rd =
A fy
M0
(4)
(5)
where NSd is the design axial force, Npl,Rd is the compression resistance, A is the
cross-section area, M0 is the partial safety coefficient.
Compression/buckling resistance of columns:

(6)
N Sd N b, Rd
N b, Rd =
A fy
(7)
M1
where Nb,Rd is the compression/buckling resistances, is the reduction factors for

the relevant buckling mode and M1 is the partial safety coefficient.
Shear resistance of beams, columns and purlins:
(8)
VSd V pl , Rd
V pl ,Rd = Av
fy
3 M0
(9)
where VSd is the design shear force, Vpl,Rd is the design shear resistance and Av is
shear area.
Interaction between axial force and bending moment:
N sd
M sd
+
1 .0
N pl , Rd M el , Rd
(10)
Interaction between axial compression/buckling and bending moment lateraltorsional buckling:

N sd
A f y / M1
k LT M sd
LT Wel f y / M 1
1.0
(11)
where kLT is the coefficient at lateral-torsional buckling, and LT is the reduction

factors for the relevant buckling mode at lateral-torsional buckling.
Inequality constraints of serviceability limit states for the vertical deflection
of beams and purlins:
max,L
L
200
max,Lf
2, L
L
250
2,Lf
Lf
200
Lf
250
(12)
(13)
where max.L is the vertical deflection of beams and max.Lf is the vertical
deflection of purlins for the terminated state, 2,L is the vertical deflection of
beams and 2,Lf is the vertical deflection of purlins for the variable load.
649
Inequality constraints of serviceability limit states for the horizontal

deflections of portal frame:
H
150
(14)
The optimization
The Modified Outer-Approximation/Equality-Relaxation (OA/ER) algorithm

(Kravanja and Grossmann [6]) was used to perform the optimization. The
OA/ER algorithm consists of solving an alternative sequence of Non-linear
Programming optimization subproblems (NLP) and Mixed-Integer Linear
Programming master problems (MILP). The former corresponds to the
optimization of parameters for a frame structure with fixed topology and
standard dimensions and yields an upper bound to the objective to be minimized.
The latter involves a global approximation to the superstructure of alternatives in
which a new topology and standard sizes are identified so that its lower bound
does not exceed the current best upper bound. The search is terminated when the
predicted lower bound exceeds the upper bound.
The optimal solution of complex non-convex and non-linear MINLP problem
with a high number of discrete decisions is in general very difficult to be
obtained. The optimization is thus performed sequentially in two different phases
to accelerate the convergence of the OA/ER algorithm. The optimization is
proposed to start with the topology optimization of the frame, while standard
dimensions are relaxed temporary into continuous parameters. When the optimal
topology is found, standard sizes of cross-sections are re-established and the
standard dimension optimization of beams, columns and purlins is then
continued until the optimal solution is found.
The example
The paper presents an example of the topology and standard dimension

optimization of a single-storey industrial building. The building is 30 meters
wide, 50 meters long and 8 meters high, see Figure 2. The structure is consisted
from equal non-sway steel portal frames which are mutually connected with the
purlins.
The portal frame is subjected to self-weight g, uniformly distributed surface
variable load q (snow s and wind w), concentrated variable load P (wind F and
initial frame imperfection F). Variable imposed load (s = 1.60 kN/m2 and w =
0.137 kN/m2) is defined as the uniformly distributed surface load in the model
input data. Both, the horizontal concentrated load at the top of the columns and
the vertical uniformly distributed line load on the beams are calculated
considering the intermediate distance between the portal frames.
Design/dimensioning was performed in accordance with Eurocode 3. The
design loads were calculated for the conditions of both ultimate limit states and
serviceability limit states: (a) for ultimate limit states: 1.35(g + s + w + F + F),
8.0 m
(b): for serviceability limit states: 1.00g + 0.90(s + w + F + F). While internal
forces were calculated by the elastic first-order analysis the deformation of frame
members were calculated by the force method.
The portal frame superstructure was generated in which all possible
structures were embended by 30 portal alternatives, 20 purlin alternatives and
different standard size variation. The superstructure also comprised 24 different
standard hot rolled European wide flange I beams, i.e. HEA sections (from HEA
100 to HEA 1000) for each column, beam and purlin separately. The material
used was steel S 355.
task of the optimization was to find the optimal structure mass, the optimal
topology (the optimal number of portal frames and purlins) and optimal standard
sizes.
The optimization was carried out by a user-friendly version of the MINLP
computer package MIPSYN, the successor of PROSYN [6] and TOP [9]. As
an interface for mathematical modelling and data inputs/outputs GAMS (General
Algebraic Modelling System), a high level language, was used [10]. The
Modified OA/ER algorithm and the two-phased optimization were applied,
where GAMS/CONOPT2 (Generalized reduced-gradient method) [11] was
used to solve NLP subproblems and GAMS/Cplex 7.0 (Branch and Bound)
[12] was used to solve MILP master problems.
The optimization model contained 130 (in)equality constraints, 183
continuous and 122 binary variables. The final optimal solution of 150,87 tons
was obtained in the 12th main MINLP iteration.
The optimal result represents the mentioned optimal structure mass of 150,87
tons, the obtained optimal topology of 12 portal frames an 20 purlins (see Figure
2) and the calculated optimal standard sizes of columns, beams and purlins (see
Figure 3).
m
.54
u4
m
1
0
.
1
50
30.0 m
Figure 2:
Optimal design of the single-storey industrial building.
651
0.50 m
HEA 140
8.0 m
HEA 900
HEA 900
7.5 m
HEA 800
HEA 800
18 1.67 m
30.0 m
Figure 3:
Optimal design of the portal frame.
Conclusions
The paper presents the topology and standard dimension optimization of the
single-storey industrial building steel structure. The optimization is proposed to
be performed by the Mixed-Integer Non-linear Programming (MINLP)
approach. The MINLP was found to be a successful optimization technique for
solving this type of structures.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
E.J. O'Brien, A.S. Dixon, Optimal plastic design of pitched roof frames
for multiple loading, Comput. Struct. 64, 737-740, 1997.
G. Gurlement, R. Targowski, W. Gutkowski, J. Zawidzka and J.
Zawidzki, Discrete minimum weight design of steel structures using EC3
code, Struct. Multidisc. Optim. 22, 322-327, 2001.
Eurocode 3, Design of steel structures, European Committee for
Standardization, 1992.
M.P. Saka, Optimum design of pitched roof steel frames with haunched
rafters by genetic algorithm, Comput. Struct. 81, 1967-1978, 2003.
S. Hernndez, A.N. Fontn, J.C. Perezzn, P. Loscos, Design
optimization of steel portal frames, Adv. Eng. Software. 36, 626-633,
2005.
Kravanja, Z. and Grossmann, I.E., New Developments and Capabilities in
PROSYN - An Automated Topology and Parameter Process Synthesizer,
Computers chem. Eng., 18, 1097-1114, 1994.
approach to structural synthesis. Part I: A general view on simultaneous
topology and parameter optimization, Int. J. Numer. Methods Eng. 43,
263-292, 1998.

[8]
[9]
[10]
[11]
[12]

approach to structural synthesis. Part II: Simultaneous topology,
parameter and standard dimension optimization by the use of the Linked
two-phase MINLP strategy, Int. J. Numer. Methods Eng. 43, 293-328,
1998.
Kravanja, S., Kravanja, Z., Bedenik, B.S. and Faith, S., Simultaneous
Topology and Parameter Optimization of Mechanical Structures,
Numerical Methods in Engineering '92, Proceedings of the First
European Conference on Numerical Methods in Engineering, ed. Ch.
Hirsch et al., pp. 487-495, Elsevier, Amsterdam, 1992.
Brooke, A., Kendrick, D. and Meeraus, A., GAMS - A User's Guide,
Scientific Press, Redwood City, CA, 1988.
Drudd, A.S., CONOPT A Large-Scale GRG Code, ORSA J. Comput. 6,
207-216, 1994.
CPLEX User Notes, ILOG inc.
653
Genetically optimised placement of

piezoelectric sensor arrays: linear and
nonlinear transient analysis
J. N. Rao, S. Lentzen & R. Schmidt
Institute of General Mechanics, RWTH Aachen University, Germany
Abstract
In modern structural control the application of discrete modal sensor arrays is a
commonly used technique to obtain the modal state vectors. In this paper, a genetic
algorithm is used to find the optimal positions to place modal sensor arrays on simple structures such as beams, plates and shells. The performance criterion is taken
as the steady state observability Grammian of the system and includes spillover
prevention as well. The performance of optimally placed modal sensors in the linear range is discussed. The variation in the performance of these modal sensors in
the nonlinear range is investigated.
Keywords: geometrical nonlinearity, modal sensors, Lyapunov equation, genetic
algorithm.
1 Introduction
There has been consistent research on smart materials and structures for the last
two decades. Geometrically linear theories and numerical methods have been
developed by many authors, e.g. Crawley and de Luis [1], Tzou and Tseng [2].
Considerably less work can be found in the area of geometrically nonlinear modelling of smart structures. Structural nonlinearity has been taken into account for
interlaminar stress analysis by Icardi and Di Sciuva [3], for large deflection shape
control in Yi et al. [4], Mukherjee and Chaudari [5] and Lentzen and Schmidt [6].
Active buckling control and post-buckling analysis has been done by Krishna and
Mei [7] and Chandrashekhara and Bhatia [8].
Piezothermoelastic analysis including nonlinearity is discussed in Tzou et al. [9]
and Pai et al. [10].
doi:10.2495/HPSM06064

In order to implement structural vibration control in a modern fashion, one
should be able to sense the modal amplitudes. Several possibilities of modal sensing can be found in the literature. Clark and Burke [11], Lee and Moon [12],
Gawronski [13] are some of the earlier researchers, who worked on modal sensors. The optimal locations of sensors in intelligent structures is an important issue
in the research of present day, since an arbitrary decision can degrade their performance. In order to find the optimal positions of these modal sensors, genetic
algorithms (GAs) can be used as a searching method. Some of the earlier successful attempts can be found in the works of Sadri et al. [14], Han and Lee [15], where
GA is used to find optimal places for both sensors as well as actuators.
In the present paper discrete array sensors are used to filter the modal amplitudes. To find the optimal locations of modal sensors the observability of the system should be investigated to obtain the best performance with the least number
of sensors. The performance criterion for the selection of the sensor positions is
taken as the steady state observability Grammian of the system. The steady state
observability Grammian is obtained by solving the Lyapunov equation. The performance of genetically optimised sensor arrays in the linear range is discussed,
and the influence of structural nonlinearity is investigated.
2 Structural model
The FE analysis of the structural response is performed using the theory of composite laminated shells given by Schmidt and Reddy [16]. The geometrically nonlinear strain displacement relations are based on the Reissner-Mindlin hypothesis and are valid for small strains and moderate rotations of the midsurface normals [16]
2
1
0
= +3 + 3 2
1
3 = 3 +3 3
33 = 0
with
= +
1
2
1 0 0
,
2

0
0
0
0
1
1
1
1
v | + v | b b + b v + b v ,

1 1 1
1
1
2
=
b b v v b v| b v | ,
2

1 0
1 1 1
0
1
1 0
1
+ v + v
3 =
and 3 = v v | .
2
2
The following abbreviations were used:

0
1 0
0
0
v | + v | b v 3 ,
=
2
(1)

0
=v | b v 3
655
=v 3, +b v .
and
The internal and external virtual work are evaluated in a total Lagrangian fashion. The second Piola-Kirchhoff stress and Green-Lagrange strain are chosen to
express the mechanical part of the internal virtual work. Consequently, the electric
variables are defined as referring to the initial undeformed configuration [17].
3 Performance criterion
The performance criterion to optimise the modal sensor positions requires modal
voltages. The linear equations of motion of a laminated composite structure with
embedded piezoelectric layers can be written as

q
Fe
M 0
q
Kqq Kq
(2)
=
+
Kq K
Qe
0 0
0
where the above matrices are found in Lentzen et al. [17]. There, M is the mass
matrix, Kqq is the linear elastic stiffness matrix, Kq = Kq are the linear electromechanical coupling matrices and K is the piezoelectric capacitance matrix. The
electromechanical response of the structure is denoted by {q, }, where q and
are the systems generalised displacements and sensor voltages, respectively. The
externally applied forces (Fe ) and charges (Qe ) are denoted by the right subscript e.
After static condensation of the electric variables, equation (2) can be written in
modal form as
R

BjL VL
(3)
pj + j2 pj =
L=1
where pj , p j and p represent the j modal acceleration, velocity and displacement,

respectively. The natural frequency of the j th mode is denoted by j . The piezo
sensor voltages can be represented as
th
L =
CjL pj .
(4)
j=1
Here the induced voltage of the Lth sensor and the control voltage at the Lth actuator are denoted by L and VL , respectively. The sensing constant of the Lth sensor

due to unit modal displacement of the j th mode is denoted by CjL and the j th modal

actuation constant due to unit applied voltage of the Lth actuator is denoted by BjL .
In order to address the observability problem one should convert the standard
FE equations into state space equations as follows [15]:
x = [A]{x} + [B]{u}
(5)
y = [C]{x}
(6)

where
x=
p1
1 p1
. . . pn

Ai =
A = diag(Ai ),
B11
.
B = ..

Bn1
0
0
1
1 C11
..
and C =
.
1
n Cn1
V1
.
u = ..
,
VR
1
.
y = ..
n pn
T

i
0
0
i
(7)
B1R
..
.
0
..
.

BnR
..
.
0
1
1 C1S
..
.

n1 CnS
A linear time invariant system (A, B, C), with s outputs is completely observable,
if any of the following conditions are satisfied [19].
1. The (sn n) observability matrix [O] has rank n, where

T
[O] = [C] [C][A] [C][A]2 [C][A]n1
(8)
2. The observability Grammian [Wo ] is full-rank

T
e[A]t [C][C]T e[A] t dt
[Wo ] =
(9)
In fact [Wo ] is the solution of the Lyapunov equation:

[A][Wo ] + [Wo ][A]T + [C][C]T = 0
(10)
The observability test based on the rank is binary in nature (i.e. it tells us whether
the system is observable or not). Additionally the extent of observability is
required. Therefore, in the present work the objective function proposed by Hac
and Liu [18] is used, given by:

2(no +nr )
2no
2(no +nr )
2no

2n
r
2n
o
J=
j
(j )
j
(j ) , (11)
j=1
j=1
j=2no +1
j=2no +1
where no and nr are the number of observed modes and residual modes, respectively. For sensor optimal places, j is the eigenvalue of the steady state observability Grammian (i.e. [Wo ]). Note that in the objective function (i.e. Eq. 11) a
657
product term is included to account for least controlled modes. If any of the modes
is least controlled then the total objective function value goes to zero.
4 Genetic algorithm
Genetic algorithms (GAs) are random search techniques based on the mechanics
of natural selection and genetics. Genetic algorithms are used to explore the global
extremum of the given linear or nonlinear function. Although randomised, genetic
algorithms can efficiently explore the new generation with better fitness.
The GA is used to maximise J (i.e. Eq. (11)) for a given number of sensors.
4.1 Algorithm
1. Create a random initial population of sensors.
2. Evaluate each member of the current population by computing its fitness
value (i.e. J as given in Eq. (11) ), and select parents based on their fitness
value.
3. Children are produced by mating a randomly selected pair of parents at a
randomly selected site, known as crossover and by making random changes
to a single parent, known as mutation.
4. Replace the current population with the children from the new generation.
5. Repeat the algorithm for a prescribed number of generations.
4.2 Modal sensors
The principle of discrete modal sensor arrays is depicted in Figure 1. By choosing
the gains i in a particular way, the modal sensor will respond only to the mode j.
The gains are obtained by solving the orthogonal system of equations
[G]{j } = {ej },
(12)
where [G] = gki , which is the modal voltage of the k th sensor due to the unit
modal displacement of mode i, {j } are the linear gains of all sensors to sense
mode j and {ej } is the unit vector in the direction j.
Modal sensor signal
Linear combiner
Sensors
Structure
Figure 1: Principle of the discrete modal sensor array.

5 Numerical results
5.1 Clamped plate
A plate, with the dimensions [1101101] mm as shown in Figure 2, is taken as a
numerical example. The plate consists of aluminium and is meshed with a [11x11]
grid. The material properties of the aluminium and PVDF layers are displayed in
Table 1. The position of the four patches has to be optimised in order to observe the
first four modes. The GA parameters used in the analysis are population size (30),
crossover probability (50%), mutation probability (10%) and number of generations (40). After application of GA, the optimal sensor positions for the plate are
obtained as depicted in Figure 2. The first four natural frequencies are calculated
as 728 Hz, 1486 Hz, 1486 Hz and 2193 Hz.
Table 1: Aluminium and piezo material properties.
Aluminium
PVDF
E [Gpa]
[]
70
0.3
2
0.3
[kg/m3 ]
d31 [m/V ]
2700
2800
2.2 1010
d33 [m/V ]
1.062 1011
111
111
121
85
37
41
82
86
38
42
11
110 mm
Figure 2: Optimised configuration.
110 mm
110 mm
81
121
11
110 mm
Figure 3: Arbitrary configuration.
In order to examine the effectiveness of the modal sensors, the transient modal
sensor signals have to be investigated. One of the popular explicit time integration
techniques (i.e. central difference method) is used to integrate the equations (2)
659
in the time domain. As an initial displacement field, the superposition of the first
four modal displacements is prescribed, with the respective modal amplitudes of
0.01, 0.007, 0.007 and 0.003 resulting in a mid-point deflection of 0.3616 mm. Figure 3 shows an arbitrary configuration of sensors that is considered in the present
work for comparison purpose. Figures 4-7 compare the results obtained with the
optimised and the arbitrary configuration. It can be observed in Figures 4-7 that the
modal sensor signals for the optimal sensor configuration are correct and are barely
aliased with those of higher modes. In case of the arbitrary sensor configuration the
modal sensor signals are strongly aliased with those of higher modes. Figures 8-9
show the modal signals for both optimal sensor configuration and arbitrary sensor
configuration in the nonlinear case. It can be concluded that the induced membrane
stresses which are not considered in the linear and modal analysis are the main
cause for the failure of modal sensor arrays in the nonlinear range of deformations.
2
x 10
modal sensor signal, mode2
x 10
optimal configuration
arbitrary configuration
Figure 4: First mode signal.
Figure 5: Second mode signal.
2
x 10
2
x 10
time[sec]
time[sec]
Figure 6: Third mode signal.
time[sec]
time[sec]
Figure 7: Fourth mode signal.
time[sec]
Figure 8: Nonlinear first mode signal.
time[sec]
Figure 9: Nonlinear
signal.
second
mode
6 Conclusions
In the present work, GA is used to find optimal placement for modal sensors. Transient analysis is performed with linear and nonlinear FE, based on first-order shear
deformation moderate rotation theory. By numerical example it is shown that the
principle of modal sensor arrays yields good results for the genetically optimally
placed sensor patches in the range of small displacements. In the geometrically
nonlinear case, it is found that the induced membrane stresses are the prime cause
for the failure of modal sensor arrays.
References
[1] E. F. Crawley and J. de Luis, Use of piezoelectric actuators as elements of
intelligent structures. AIAA Journal, 25, 1373-1385, 1987.
[2] H. S. Tzou and C. I. Tseng, Distributed piezoelectric sensor/actuator design
for dynamic measurement/control of distributed parameter systems: a piezoelectric finite element approach. J. Sound Vib., 138, 17-34, 1990.
[3] U. Icardi and M. Di Sciuva, Large-deflection and stress analysis of multilayered plates with induced-strain actuators. Smart Mater. Struct., 5, 140-164,
1996.
[4] S. Yi, S. F. Ling and M. Ying, Large deformation finite element analyses
of composite structures integrated with piezoelectric sensors and actuators.
Finite Elements in Analysis and Design, 35, 1-15, 2000.
[5] A. Mukherjee and A. S. Chaudari, Piezolaminated beams with large deformations. Int. J. of Solids and Structures , 39, 4567-4582, 2002.
[6] S. Lentzen and R. Schmidt, Nonlinear finite element modelling of composite
structures with integrated piezoelectric layers. High Performance Structures
and Materials II, WIT Press, Southampton-Boston, 67-76, 2004.
661
[7] M. R. M. Krishna, and C. Mei, Finite element buckling and post-buckling

analyses of a plate with piezoelectric actuator. In Proc. of the Conference on
Recent Advances in Adaptive and Sensory Materials and Their Applications,
Blacksburg, USA. 301313, 1992.
[8] K. Chandrashekhara and K. Bhatia, Active buckling control of smart composite plates finite element analysis. Smart Mater. Struct., 2, 31-39, 1993.
[9] H. S. Tzou, Y. Bao and R. Ye, In Smart Structures and Materials 1994: Smart
Structures and Intelligent Systems, Proc. of SPIE, Vol. 2190. 206-214, 1994.
[10] P. F. Pai, A. H. Nayfeh, K. Oh and D. T. Mook, A refined nonlinear model
of composite plates with integrated piezoelectric actuators and sensors. Int.
J. of Solids and Struct., 30, 1603-1630, 1993.
[11] R. L. Clark and S. E. Burke, Practical limitation in achieving shaped modal
sensors with induced strain materials. J. Vib. Acoust., 118, 668-675, 1996.
[12] C. K. Lee and F. C. Moon, Modal sensors/actuators. J. Appl. Mech, 57, 434441, 1996.
[13] W. Gawronski, Modal actuators and sensors. J. Sound Vib., 229, 1013-1022,
2000.
[14] A. M. Sadri, J. R. Wright and R. J. Wynne, Modelling and placement of
piezoelectric actuators in isotropic plates using genetic algorithms. Smart
Mater. Struct., 8, 490-498, 1999.
[15] J. H. Han and I. Lee, Optimal placement of piezoelectric sensors and actuators for vibration control of a composite plate using genetic algorithms. Smart
Mater. Struct., 8, 257-267, 1999.
[16] R. Schmidt and J. N. Reddy, A refined small strain and moderate rotation
theory of elastic anisotropic shells. J. Appl Mech, 55, 611-617, 1988.
[17] S. Lentzen and R. Schmidt, Nonlinear FE-simulation of piezolaminated
plates and shells, Proc. International Congress on Computational Mechanics
& Simulation, ICCMS-04, IIT Kanpur (India), vol. I, 77-85, 2004.
[18] A. Hac and L. Liu, Sensor actuator location in motion control of flexible
structures. J. Sound Vib., 167, 239-261, 1993.
[19] W. Gawronski, 1998, Dynamics and Control of Structures: A Modal
Approach, Springer.
[20] I. Bruant, G. Coffignal and F. Lene, A methodology for determination of
piezoelectric actuator and sensor location on beam structures. J. Sound Vib.,
243, 861-882, 2001.
[21] S. L. Padula and R. K. Kinkaid, Optimisation strategies for sensor and actuator placement. National Aeronautics and Space Administration (NASA),
Langley Research Center, Langley, Virginia, 1999.
[22] D. Halim and S. O. Reza Moheimani, An optimisation approach to optimal
placement of collocated piezoelectric actuators and sensors on a thin plate.
Mechatronics, 13, 27-47, 2001.
663
Numerical analysis of the process

of trapezoidal thread rolling
L. Kukielka & K. Kukielka
The Technical University of Koszalin, Poland
Abstract
In this paper the physical and mathematical models of deformations
(displacements and strains) and stress in the cold process of trapezoidal thread
rolling is presented. The process is considered as a geometrical and physical
non-linear, initial as well as boundary value problem. The phenomena on a
typical incremental step were described using a step-by-step incremental
procedure, with an updated Lagrangian formulation. The state of strains was
described by Green-Lagranges tensor, while the state of stress by the second
symmetrical Pioli-Kirchhoffs tensor. The object was treated as an elastic (in the
reversible zone) and visco-plastic body (in non-reversible zone) with mixed
hardening. The variational equation of motion in three dimensions for this case
was proposed. Then, the finite elements methods (FEM) and dynamic explicit
method (DEM) were used to obtain the solution. The application developed for
the method of finite elements ANSYS 8.1 provides a complex time analysis of
displacement, strains and stresses occurring in the object. The boundary
conditions for a displacement increment determined in model investigations were
used. Examples of calculations of influence of a friction coefficient on the states
deformation and stress were presented.
Introduction
One of the most widespread machine elements is the thread. Above 60% parts of
modern machines, devices and mechanism has threaded holes, whose
performance by screw-tap in high-plastic steels, some non-ferrous metals and
their alloy, pose difficult problems in technological aspects. The difficulty at
threading hard materials characterized by a large ductility and a high elastic
doi:10.2495/HPSM06065

limit, come especially from the tendency to seizure the screw - tap in a threading
hole. However, in spite of peck constructional improvements and many
recommendations in the field of selection conditions for treatment quality
executed, threads do not comply with standard technical requirements, the
required proprieties do not apply to the surface layer, and apart from this, this
tool wears away too fast, while the capacity of the process is low [1].
At the Technical University of Koszalin in Poland, in the Chair of Working
Machines a scientific research is being conducted into methods of realization of
male threads through plastic shaping (e.g. embossing screw threads) with a high
velocity.
Thread rolling is a non-cutting cold-forming process, and this gives the
following further advantages: the actual time for thread rolling lasts only seconds
and, therefore, it is unimportant in respect to the cycle time; the time required for
inspection is very short and the maintenance costs are extremely low; the rolling
attachment operation is of a very high accuracy; changing to another thread size
is possible by using another set of thread rolls and a gauge, provided that the
thread diameter is in the range of the respective rolling attachment size; the short
thread run-out (about pitch) of the thread rolls allows to produce a thread
reaching up to the collar of the component; high resistance to corrosion due to
smooth press polished surfaces; significant in the surface strength; an
uninterrupted grain flow (Fig. 1). The advantageous exploitation proprieties of
screw threads are the result of a plastic strain in surface layer of elements.
Figure 1:
The surface layer after trapezoidal thread rolling.
Company FETTE successfully applies the technology of the rolled screw

thread for embossing on various pieces (Fig. 2). Currently, we are examining
three varieties of threading: tangential, radial and axial. A stand for threads
rolling of the axial method, made by FETTE, is in our Laboratory (Fig. 3).
Figure 2:
Example of work pieces, in which tread rolling was applied.
Figure 3:
665
Machine for axial thread rolling, produced by the FETTE.
The trapezoidal thread rolling is a geometrical and physical non-linear

process, with initial and boundary conditions as a function of time and space.
The boundary conditions in the contact zone are not known. The paper has
focused mostly on the preparation of a physical model of the process and then on
a mathematical model and the solution algorithms obtained for the discrete
systems of equations along with the respective initial and boundary conditions.
The basic problem in this technology is the knowledge of physical phenomena,
especially influence of friction coefficient on states deformation and stress in
the surface layer.
Mathematical model of process
A mathematical model of the process is formulated in increments and contains

the following: a material model, an equation of motion, with initial and boundary
conditions.
2.1 Material model
2.1.1 Incremental model of yield stress
Yield stress y is the most important parameter characterizing the resistance of a
visco-plastic deformation. The incremental model of the yield stress for a typical
step time t=t+t is defined as [2]:
t
t
t y = t F2 ( t y) t y +
tt F1[] ( VP ) tt F1[] t t ( VP ) ( VP )
eq + t
t
t F3 ( t eq ) t eq ,
t st
tt (eqVP )
= 2 t e ij(VP) t e ij(VP) /3 ,
where t (VP)
eq
(VP)
t
eq
2 t e ij(VP) t e ij(VP) /3
(1)
are
the incremental of effective visco-plastic strain and strain rate; st is the

state stress depending on the accumulated effective visco-plastic strain and time,
t
t
t F2 ( t y) t y is the component of change in the initial yield stress Re with a


change of chemical composition; [ tt F1 [ ]/ tt st ] tt F3 ( tt (VP)
) t (VP)
is the
eq
eq
component of change in the temporary yield stress tt y with change of the viscoplastic strain, [ tt F1 [ ]/ tt (VP)
] t (VP)
is the component of change in the
eq
eq
temporary yield stress with change of the visco-plastic strain rate.
2.1.2 Elastic/visco-plastic material model
A new model of mixed hardening for isotropic material which includes the
combined effects of elasticity (a reversible domain), visco-plasticity (a non
reversible domain) (E/VP) is used. The model takes into account the history of
the material.
The constitutive equation of increment components of a total strain tensor
takes form [3]:
t ij
1
[ t D (E) kl tt A]
t ~ ** t ijkl t
1 t S
(2)
and of increment components of the total stress tensor:
t (E)
t ij = t C ijkl t kl
~ ~
(E)
t
tt Sij*[ tt Sij tt Cijkl
t kl t A],
(3)
where:
t ~** t ~* t (TE) t ~
t S = t Sij t Cijmn t Smn
(4)
is a positive scalar variable,

t ~*
t Sij
t~
t Sij
t ~ t (E) t ~
t Sij t C ijkl t Sij
(5)
2t 2 t~ t
t y ( t C+ t E T )
is a component of a stress tensor,

t
tA
is a positive scalar variable,

Piola-Kirchhoff
t (E)
tD
stress
2t
3
t y
t ij
tensor,
y (VP)
eq
t (VP) t
t eq
(6)
is the increment component of the second

t (E)
t D ijkl
are
the
components
of
tensor
=[ tt C (E) ]1 in time t, t ij is the increment component of Green-Lagrange
(E)
strain tensor, t Cijkl
are the components of elastic constitutive tensor.
667
2.2 Incremental model of motion
In this section we develop the equation of a deformation of the object in the

updated Lagrangian formulation. Assuming that numerical solutions are obtained
at discrete time t, the solution for t+t is to be obtained. In such a case,
a functional increment tJ () is formulated for the increment displacement and
reads as follows:
i , t u i )= t J ()
t J ( t u i , t u
(7)
where tu i , tu i , tu i are the ith increment components of the displacement,
velocity and acceleration vectors, respectively. While using the conditions of
stationary of functional t J and a finite element method, we can write an
equation of motion in the form:
[ tt M]{tr} +[ tt CT ]{tr} + ([tt K T ] +[tK T ]){tr} ={tR} +{tF} +{tt F} +{tt R} (8)
where mass matrix [ tt M ] , damping matrix [ tt CT ] , stiffness matrix [ tt K T ] and
force vector {tt FT } are known at time t. However, increment stiffness matrix
[ tK T ] , external incremental load vector {t R} , internal incremental forces
vector { t F} , incremental vectors of displacement {t r} , velocity {t r} and

acceleration {t r} of finite element assembly at a typical step time are not
known. In order to solve this problem we apply the integration methods.
3 DEM solution
Assuming that an increase of temporary step t is very small, it is possible to
execute a linearization of equation (8) and using the incremental decomposition
we obtain:
[M ]{t tt r} + [CT ]{t tt r} + [K T ]{t tt r} = {tt tt FT } +{t tt Q}.
(9)
Then using the central difference method (DEM), in which it is assumed that:
{t r} =
1
1
({t + t r} {t t r}) , {t r} = 2 ({t + t r} 2{t r} + {t t r}) (10)
2 t
t
and substituting the relations in (10) into (9) we obtain:


~
~
[M]{t r} = {t tt Q T } ,
(11)
where:
~
1
1
[CT ],
[M ] = 2 [ M ] +
2t
t
~
{t tt Q T } = {tt tt FT } +{t tt Q} [K T ]{t tt r} +
+
t t
2{ r} {
t 2
r}
[M ] +
(12)
t t
r}
[CT ].
2t
The integration method requires that the time step t is smaller than critical
value tkr, which can be calculated from the mass and stiffness properties of the
complete element assemblage: t t kr = TN / , where TN is the smallest
period of the finite element assemblage with n degrees of freedom.
Model investigation
The model investigation was conducted in order to settle the course deformation
layer top sample executed from the plastic material, as well as with the aim to
qualify boundary conditions for displacements indispensable to numeric analysis
of the trapezoidal tread rolling process. To model investigations were applied to
the samples in a rectangle shape with the following dimensions:
a)
b)
Figure 4:
The stamp (a) and the meshed sample before deformation (b).
Two samples were joined by sides with a plot mesh, and were closed in a metal
form. Then the samples were subjected to the deformation by a perpendicular shift
of rectilinear motion in the model stamp of an outline of trapezoidal thread rolling.
The view of the deformed mesh of complete elements is presented in Fig. 5.
a)
Figure 5:
b)
The mesh after deformation for =0,2 (a) and =0,39 (b).
669
Numerical analysis of state of displacement, strain and

stress of material during trapezoidal thread rolling
The application developed with regard to the method of finite elements in

ANSYS 8.1 programme provides a complex time analysis of states of
displacement, strains and stresses. Digital computing for the process was carried
out with the use of two methods. The first method requires introducing the
boundary conditions for displacements in the contact zone determined by the
model investigation, whereas the second one requires the adequate determination
of the contact zone without an introduction of boundary conditions. The main
aim of the simulation was to define the influence of friction coefficient on the
state of deformation and stress in the surface layer of the object. The numerical
analysis for 2-D states of deformation and 3-D states of stress was applied on the
example of steel C55 (DIN). The example results of deformation, strain and
stress obtained by the numerical analysis for friction coefficient =0,2 and
=0,39 are presented in Fig. 6 and Fig.7.
Figure 6:
Composition of numerical results for friction coefficient =0,2:

1 initial finite element mesh, 2 - finite elements after deformation,
3 intensity of strain, 4 intensity of stress.
The distribution of strain and stress intensity in defined nodes (Fig. 8) is

shown in Figs. 912.

1
2
ELEMENTS
DISPLACEMENT
APR 13 2005
21:59:14
Y
Z
APR 13 2005
21:59:14
STEP=1
SUB =100
TIME=1
DMX =15.6
4
NODAL SOLUTION
NODAL SOLUTION
APR 13 2005
21:59:14
STEP=1
SUB =100
TIME=1
EPTOINT (AVG)
DMX =15.6
SMN =.00995
MX =3.616
SMX
MN X
Z
Z
.00995
.911405
1.813
2.714
3.616
APR 13 2005
21:59:14
STEP=1
SUB =100
TIME=1
SINT
(AVG)
DMX =15.6
SMN =.517E+08
MX
SMX =.178E+10
MN
X
.517E+08
.485E+09
.918E+09
.135E+10
.178E+10
Composition of numerical results for friction coefficient =0,39:

1 initial finite element mesh, 2 - finite elements after deformation,
3 intensity of strain, 4 intensity of stress.
Figure 7:
DISPLACEMENT
APR 19 2005
22:49:06
STEP=1
SUB =100
TIME=1
DMX =16.85
2
DISPLACEMENT
STEP=1
SUB =100
TIME=1
DMX =16.85
node 158
APR 19 2005
22:49:06
node 160
node 127
Figure 8:
Defined nodes in the discretized model.
Figure 9:
The distribution of strain intensity for =0,2.
Figure 10:
The distribution of strain intensity for =0,39.
Figure 11:
The distribution of stress intensity for =0,2.
671
Figure 12:
The distribution of stress intensity for =0,39.
Conclusions
The paper presents a numerical analysis of the process for thread rolling. The
process is considered as a geometrical, physical non-linear initial and boundary
value problem. The phenomena on a typical incremental step were described
using a step-by-step incremental procedure, with an updated Lagrangian
formulation. The variational nonlinear equations of the objects motion for a
typical incremental step were derived from the stationary condition for these
functionals. These equations were unravelled using the spatial digitization by a
finite elements method. The adequate algorithm to solve the equations of the
motion assuming Rayleighs proportional dissipation was developed for this
method. The application developed for the method of finite elements ANSYS 8.1
provides a complex time analysis of displacement conditions, strains and stresses
occurring in the object. The mathematical models of the process, the algorithms
for the solution of discrete equations of motion and the application in the
ANSYS system developed in the present work could be used to improve the
design process for the thread rolling process.
References
[1]
[2]
[3]
yczko K.: Technologia narzdzi i wygniatania gwintw wewntrznych.

Politechnika Czstochowska, 1999.
Kukieka L., Cienkowski W., Dudek P.: Incremental model of yield stress
of metals in the conditions of burnishing rolling operation with electrical
current. Third Internationat Meeting on Computer Methods and
Experimental Measurements for Surface Treatment Effects, eds. M. H.
Aliabadi, C.A. Brebbia, WITPRESS, Boston, pp. 93-102, 1997.
Kukieka L., Krzyyski T.: New thermo-elastic thermo-visco-plastic
material model and its application. Conference in GAMM, Metz, 12-16
April 1999: Ed. WILEY-VCH, pp. 595-596, 2000.
673
Three-dimensional limit analysis of ancient

masonry buildings with rigid block models
A. Orduna
Faculty of Civil Engineering, University of Colima, Mexico
Abstract
Limit analysis with rigid block models is a tool successfully used in recent years
for the assessment of ancient masonry structural elements and small buildings. In
this paper, the interface yield functions for three-dimensional models are defined
at interpolation points, instead of using the generalised stresses approach. This
approach leads to very simple expressions for the yield functions and flow rules
and therefore, renders the mathematical programming problem easier to solve than
the generalised stresses formulation. The solution for the limit analysis problem is
obtained using the load-path following approach. The validation of the present
proposal shows good agreement compared with non-linear finite element results.
Keywords: limit analysis, rigid block assemblages, non-associated flow, numerical
integration.
1 Introduction
A valuable model for the structural assessment of ancient masonry structures is
the limit analysis of rigid block assemblages interacting through no-tension and
frictional interfaces. The reason for this affirmation lies in the fact that masonry
has low tensile strength and quasi-brittle failure; therefore, at collapse, the cracks
render the structure as a set of rigid blocks rocking and sliding between them.
In the rigid blocks modelling strategy, the degrees of freedom are related to
the blocks, and the stress and strain variables are related to the interfaces. Therefore, blocks can be regarded as extended nodes and the interfaces as structural
elements. Previous works in this subject have used mostly a generalised stress
approach for the interfaces [14]. This means that the generalised stresses are,
for instance, the normal and shear forces and the bending and torsion moments.
Besides, Livesley[5, 6], using a different approach, verified the contact at points
located at the interface corners.
doi:10.2495/HPSM06066

In this paper, the use of numerical integration techniques is investigated, as an
alternative to the generalised stresses approach for the yield function description.
Firstly, the formulation for non-associated limit analysis of rigid block assemblages is outlined. Then, the formulation of the yield function and flow rule using
numerical integration is presented. A discussion of the torsion failure on frictional
interfaces follows, where it is emphasised the importance of a correct normal stress
distribution calculation over the interface. A comparison of results for a model
obtained by the proposed approach against non-linear FE method results serves to
validate the proposal. Finally, relevant conclusions are stated.
2 Limit analysis of rigid block assemblages

Eqns. (1-6) are the conditions that a limit analysis solution with non-associated
flow rule must fulfil, see e.g. [7]. Eqn. (1) combines the compatibility and flow
rule conditions. Here, the columns of the matrix N0 contain the flow directions for
each one of the yield functions in the structure; the flow multipliers for each one
is the compatibility matrix and u
of such yield functions form the vector; C
is the vector of block displacement rates. Eqn. (2) is a scaling condition for the
displacement rates that ensures the existence of non-zero but finite values. Here Fv
is the vector of variable loads. Eqn. (3) ensures equilibrium. Here Fc is the vector of
constant loads; is the load factor that measures the amount of the variable loads
is the vector of generalised stresses at the interfaces.
applied to the structure and Q
are the actual stress values at every integration
In this paper the elements of Q
point of all the interfaces in the model. Eqn. (4) guaranties that the yield functions,
vector , are not violated and eqn. (5) ensures that plastic flow implies energy
dissipation. Finally, eqn. (6) guaranties that plastic flow cannot occur unless the
stresses have reached the yield surface. Figure 1 shows a graphical representation
of the static variables at two adjacent blocks, named i and j, and the common
quadrilateral interface, k. Details about these vectors, matrices and functions for
rigid block, three-dimensional models have been given elsewhere [4]. In this paper,
modifications are made with the aim to change the generalised stresses approach
to the integration points approach.
u = 0
N0 C

FvT
(1)
u 1 = 0
(2)
= 0
TQ

Fv C
Fc +
0
(3)
0
T = 0
(4)
(5)
(6)

r
Fi
Fi
675
Fj
s 1k
nk
s 2k
o
Fj
Block j
Block i
Figure 1: Static variables at an interface and adjacent blocks.
3 Numerical integration approach

Consider a quadrilateral shaped interface, k, illustrated in Figure 1. This interface
lies between two infinitely strong blocks, i and j. The cohesion-less Coulombs law
governs the interface failure. The interface is supposed to represent a dry masonry
joint, therefore, it has no tensile strength. The origin of the local coordinate system
is located at the centroid of the interface, o. This coordinate system has axes x1 and
x2 on the interface plane, and xn in the normal direction, forming a right-handed
system. The unitary vectors along the coordinate axes x1 , x2 and xn are s1 , s2 and
n, respectively, Figure 1. The stress vector at a point has components 1 , 2 and
along the axes x1 , x2 and xn , respectively. The normal component, , is positive
in tension.
The resultant stresses over the interface are the shear forces, V1 and V2 , along
the x1 and x2 axes, respectively; the normal force, N; the bending moments, M1
and M2 , along the x1 and x2 axes, respectively and the torsion moment, T . Under a
particular stress distribution, eqns. (712) give the values for the stress resultants.
Here, dA is the differential of area. The second right hand sides of eqns. (712)
are the numerical integration approximations for the integrals in the first right
hand sides. Here, represents the weights of the quadrature, n p is the number
of integration points, assumed equal in both interface in-plane directions, J is
the determinant of the Jacobian of the transformation between the interface coordinate system and the quadrature coordinate system [8]. Since the interfaces are
supposed to be quadrilaterals, this quantity is a constant and can be placed out of
the summation. In the second right hand sides, the variables are evaluated only at
the integration points, according to the indexes l and m.
V1 =
V2 =
np np
1 dA
.
= J
1lm l m
(7)
l=1 m=1
np np
2 dA
.
= J
2lm l m
l=1 m=1
(8)

N=
M1 =
lm l m
(9)
l=1 m=1
x2 dA
.
= J
lm x2m l m
(10)
l=1 m=1
np np
.
= J
np np
M2 =
T=
np np
dA
x1 dA
.
= J
lm x1l l m
(11)
l=1 m=1
np np
.
(1 x2 + 2 x1 )dA= J
(1lm x2m + 2lmx1l )l m
(12)
l=1 m=1
In the approach proposed in this paper, it is not necessary to calculate the stress
resultants over the interfaces, but it can be done by means of eqns. (712). The
yield function characterisation is performed by limiting the normal and shear stresses
at each integration point. The normal stress limits are: from bellow, the effective
compressive stress and from top the zero value due to the no-tension hypothesis.
Therefore, eqns. (13) and (14), represent the yield functions for the normal stress.
Here, fce f is the effective compressive stress, which takes into account the effect of
transverse cracking and the fact that masonry presents quasi-brittle failure, while
the model features perfect plastic behaviour. The shear stresses are limited by the
cohesionless Coulomb criterion, as already mentioned. This criterion conducts to a
quadratic function as illustrated in Figure 2 by the circle of radius , where is
the friction coefficient. Nevertheless, in order to simplify the solution to the mathematical programming problem, a piecewise linear approximation is proposed and
illustrated also in Figure 2 by an inscribed octagon. The resulting yield functions
are expressed by eqns. (15) and (16). It is observed that, due to the absolute value
operands, these two expressions represent the eight linear functions illustrated in
Figure 2.
c
fce f
(13)
s1s4 |1 | + ( 2 1)|2 | + 0
s5s8 ( 2 1)|1 | + |2 | + 0
(14)
(15)
(16)
There are, therefore, ten linear yield functions for each integration point. The
yield functions for all the integration points at every interface in a model can be
cast in matrix form, in such a way that eqn. (4) takes the form of eqn. (17). Here, N
is the matrix resulting from the assemblage of all the yield functions in the model.
0
N Q
(17)
There are three generalised strain or relative displacement rate components at

each integration point: s1 , s2 and n, parallel to the local coordinate directions

2
677
Coulomb criterion
Approximation
Figure 2: Shear yield function.
Table 1: Flow directions for an integration point failure.

Flow multiplier
s1
s2
t
s14
0
1 /|1 |
( 2 1)2 /|2 |
1
0
2 /|2 |
s58
( 2 1)1 /|1 |
x1 , x2 and xn , respectively. The flow directions for the yield modes are in Table 1.
For the compression and tension yield modes the flow consists on a normal relative
displacement rate in the negative and positive directions of the xn axis, respectively.
The flow directions for the sliding failure modes are obtained from the normality
rule but neglecting the relative displacement rate along the xn axis, due to the
hypothesis of zero dilatancy. The rows of Table 1 are used to assemble the columns
of matrix N0 , in eqn. (18), which gives the generalised strains resulting from the
flow rule condition. Here the vector q gathers the generalised strains for all the
integration points in a model.
q = N0
(18)
The displacement rates for a generic block i are the translation displacement
rates at the block centroid uti and the angular displacement rates uri . These vectors are referred to the global coordinates system. Eqn. (19) gives the generalised
strains at the integration point p of the interface k in terms of the displacement
rates at the blocks i and j; see Figure 1. Here, Tkg is the matrix that transforms a
vector from the global coordinates system to the local interface k system; c pk , ci
and c j are the position vectors of the point p, and the centroids of blocks i and j,
respectively. Eqn. (19) can be used to assemble the compatibility set of equations

of a model in the form of eqn. (20). It is evident that eqn. (1) follows directly from
eqns. (18) and (20).

g
q pk = Tk utj uti + urj c pk c j uri c pk ci
u
q = C
(19)
(20)
4 Torsion failure on frictional interfaces

Orduna and Lourenco [4] studied the torsion failure on rectangular frictional interfaces. In that work, the assumptions were made that dilatancy coefficient was zero
and that the normal stresses over the interface have uniform distributions either
over all or part of the interface area. With these assumptions, yield functions and
flow rules were presented in terms of the stress resultants over the interface. Nevertheless, with the proposal presented in this work, where the normal stresses at each
integration point of an interface are independent each other, there are no guarantee
that uniform or even normal stresses distributions are obtained over an interface.
This section presents a discussion about the influence of the normal stresses distribution over the interface on its torsion moment strength.
Consider a rectangular interface with dimensions 2l1 and 2l2 parallel to the local
axes x1 and x2 , respectively, as shown in Figure 3. This interface, placed between
two infinitely rigid bodies, is subjected to a constant normal force in compression, N (negative), at the interface centroid, and fails under the action of a torsion
moment, T . The friction coefficient is and an infinite compressive strength is
assumed by now.
At incipient failure, the sliding displacement rates are zero at the centre of
twisting and vary linearly with the distance from this point. If associated plasticity is assumed (dilatancy coefficient equals friction coefficient), the normal displacement rates would be proportional to the sliding displacement rates. The limit
x2
c
l2
x1
l2
l1
l1
Figure 3: Rectangular joint.

679
analysis solution indicates that the centre of twisting is located at the interface
centroid, therefore, the maximum sliding and normal displacement rates are at the
interface corners. As the blocks are rigid, separation will occur at the inner points,
and contact will be possible only at the interface corners. Under this conditions, it
is straightforward to show that eqn. (21) proportionates the value of the failure torsion moment, where c is the distance from the interface centroid to any corner; see
eqn. (22) and Figure 3. Michalowski and Gawecki [9] arrived to an analogous conclusion for a circular interface where c is replaced by the circle radius in eqn. (21).
It is possible to arrive to the same result always that the dilatancy coefficient is
positive, even if it does not equal the friction coefficient.
T = c(N)

c = l12 + l22
(21)
(22)
If the dilatancy coefficient is negative, the normal displacement rates are again
proportional to the sliding displacement rates. Nevertheless, no separation but penetration exists between blocks, contact is only possible at the very interface centroid and, therefore, the torsion moment strength is zero (contact on a single point).
If the dilatancy coefficient is zero, contact points or areas are not determined
by the failure mechanism and neither, the normal stress distribution. This means
that for every possible normal stress distribution, there exists a torsion moment
strength. It is evident that the values of these strengths lie between zero and that
given by eqn. (21). The most fiscally meaningful case consists on a uniform normal stress distribution over the interface. For this case, eqn. (23) gives the torsion
moment strength, where the torsion constant cT is given by eqn. (24) [4]. It is
observed that eqns. (21) and (23) are the same except for the torsion constant definition.
T = cT (N)

l2
l2
1
l2 + c
l1 + c
+ 2 ln
c + 1 ln
cT =
3
2l2
l1
2l1
l2
(23)
(24)
The cT /c ratio can be calculated for the whole range of interface aspect ratios,
and the observation is made that it has a small variation between 0.50 and 0.54.
Therefore, a first but marginal conclusion is that the torsion constant can be approximated as half the distance from the centroid to the corner. A more important observation is that, for zero dilatancy, the torsion strength of the interface can take values between zero and that given by eqn. (21), but the strength for a uniform normal
stress distribution is about half the way between the former values. Therefore, it
is of fundamental importance, in limit analysis problems with zero dilatancy, to
obtain solutions with even normal stress distributions over the interfaces.
Orduna and Lourenco [4, 10] showed that for this type of problem it is also
important to take into account the loading history. Here, a solution procedure, akin

Table 2: Calculated ultimate load factors for infinite compressive strength.
Procedure
Ultimate load factor
Theoretically minimum
Theoretically maximum
0.427
0.553
FEM
Load-path following
0.479
0.465
to the load-path following one is used, which agrees with the integration points
approach. The main idea behind this procedure is to keep the normal stresses distribution as uniform as possible, while the variable loads are applied by small
increments.
5 Validation
The validation of this type of three-dimensional models is not an easy task due
to the lack of experimental or analytical results to compare with. Therefore, the
example presented here is a comparisons against a finite element (FEM) non-linear
model analysed with the package DIANA [11].
Figure 4(a) shows the model of a masonry hollow pile. The pile is built of
dry masonry blocks with dimensions 0.20.20.4 m. The pile dimensions are
0.60.81.2 m. The material volumetric weight is 20 kN/m3 and the friction
coefficient is 0.7. The permanent loads are the self-weight of the blocks. The variable loads are proportional to the blocks weight, but horizontally applied in the
direction of the larger base side (X direction). The compressive effective stress,
according to the load-path following procedure, is steadily increasing.
Figure 4(b) shows the failure mechanism obtained by the FEM analysis. The
same failure mechanism is obtained by limit analysis and Figures 4(c), (d) show it
from two different viewpoints for a better understanding. A range of ultimate load
factors is possible for this mechanism and for infinite compressive strength. It is
noted that the ultimate load factor is defined as the ratio between the variable loads
causing failure on the structure, and their nominal values, in this case numerically
equal to the self-weight of the blocks. Table 2 presents the ultimate load factor
obtained from different approaches. If the reaction on the overturning blocks set
is concentrated on the interface A only, Figure 4(d), with zero stresses at interface
B, the ultimate load factor would be 0.427, the minimum possible for this mechanism. If there are non-zero normal and shear contact forces on interface B, the
last one opposing to the upper block overturning, the maximum possible ultimate
load factor for the mechanism shown equals 0.553. The ultimate load factor calculated with the load-path following procedure is 3% lower than the FEM value. It
is possible to verify that the load-path following procedure agrees very well with
the FEM results.
681
1.2
0.8
0.6
(a)
(b)
joint A
joint B
(c)
(d)
Figure 4: Masonry pile; (a) model; failure mechanisms for infinite compressive
strength: (b) FEM failure mechanism; (c) and (d) different views of the
limit analysis failure mechanism.
It is interesting to note that the condition of zero stresses at interface B is only

possible for an infinite compressive strength. Under limited compressive stresses,
the hinge on interface A forms slightly inwards and interface B must be in contact
and must transmit normal and shear forces.
6 Conclusions
A limit analysis formulation using numerical integration techniques at the interfaces has been proposed for rigid block assemblages. The importance of even normal stress distributions over the interfaces was demonstrated. Comparisons, not
presented here due to space limitations, show that good agreement is obtained, at
interface level, between the integration points approach and the yield functions
obtained by constant normal stress distributions over regular shaped parts of the

interface. The presented example shows also very good agreement compared with
the non-linear FEM results. It must be recognised that the computing time for the
limit analysis approach and for this large model is not attractive compared with the
FEM process. Nevertheless, it is expected that, optimising the numerical tasks for
the limit analysis approach, advantage can be taken from the simplified nature of
this formulation and the computing times can be significantly reduced.
Acknowledgment
This work was supported by project PROMEP/103.5/04/1322 funded by the Secretary of Public Education (SEP) of Mexico.
References
[1] Begg, D. & Fishwick, R., Numerical analysis of rigid block structures including sliding. Computer Methods in Structural Masonry 3, eds. J. Middleton &
G. Pande, Portugal, pp. 177183, 1995.
[2] Baggio, C. & Trovalusci, P., Limit analysis for no-tension and frictional
three-dimensional discrete systems. Mech Struct Mach, 26(3), pp. 287304,
1998.
[3] Ferris, M. & Tin-Loi, F., Limit analysis of frictional block assemblies as a
mathematical program with complementarity constraints. Int J Mech Sci, 43,
pp. 209224, 2001.
[4] Orduna, A. & Lourenco, P., Three-dimensional limit analysis of rigid blocks
assemblages. Part I: Torsion failure on frictional interfaces and limit analysis
formulation. Int J Solids Structures, 2005. Accepted for publication.
[5] Livesley, R., Limit analysis of structures formed from rigid blocks. Int J Num
Meth Engrg, 12, pp. 18531871, 1978.
[6] Livesley, R.K., A computational model for the limit analysis of threedimensional masonry structures. Meccanica, 27, pp. 161172, 1992.
[7] Orduna, A. & Lourenco, P., Cap model for limit analysis and strengthening
of masonry structures. J Struct Eng, 129(10), pp. 13671375, 2003.
[8] Bathe, K.J., Finite Element Procedures in Engineering Analysis. PrenticeHall, 1982.
[9] Michalowski, R.L. & Gawecki, A., Limit torque for a frictional joint. Mech
Struct Mach, 24(4), pp. 499512, 1996.
[10] Orduna, A. & Lourenco, P.B., Three-dimensional limit analysis of rigid
blocks assemblages. Part II: Load-path following solution procedure and validation. Int J Solids Structures, 2005. Accepted for publication.
[11] TNO Building and Construction Research, Delft, The Netherlands, DIANA
Users Manual Release 7.2, 1999.
Section 9
Reliability of structures
685
Analysis of diffusional stress relaxation in

submicron Cu interconnect structures using
the model with enhanced vacancy diffusivity
in grain boundary region
I. Tsukrov, W. M. Grich & T. S. Gross
Department of Mechanical Engineering, University of New Hampshire,
USA
Abstract
We propose the finite element simulation technique to model the process of
diffusional creep and stress relaxation that occurs in Cu-damascene interconnects
of integrated circuit devices in the processing stage. On the length scale of the
interconnect lines (microns), the stress-induced mass flow constitutes the major
mechanism of inelastic deformation. The mass flow problem is coupled to the
stress analysis through vacancy flux and equilibrium vacancy concentration,
allowing independently for the concentration profile and evolution of the stresses
and strains in an iterative process to be solved. We decompose the total
displacement field into the elastic part and the inelastic mass flow contribution.
Performing the stress analysis in the configuration with accumulated inelastic
displacements, we ensure that the shape of the interconnect line is compatible
with external geometrical constraints throughout the simulation. This approach
has been implemented in the software package that seamlessly integrates the
problem-oriented code with the commercially available finite element program
MSC.Marc. We apply the technique to model the Coble creep phenomenon by
introducing the nanoscale grain boundary region having the thickness of the
order of several layers of atoms. As an illustration, the problem of stress
relaxation in a single grain subjected to prescribed displacements and tractions is
examined.
Keywords: copper interconnects, diffusional creep, grain boundary, nanoscale
deformation, finite elements.
doi:10.2495/HPSM06067
Introduction
The fabrication of Cu-damascene interconnect lines has become an important

field of research since early 1990s, when the first commercial samples were
introduced. Much technological effort has been applied to reduce the feature
size, which currently is on the order of a 100 nm (about 400 atomic radii).
Operating on this length scale requires understanding of the impact of nanoscale
mechanical behavior on reliability of conductors and interfaces. The major
concerns arise due to the fabrication process, which includes reactive ion etching
of a circuit pattern in a dielectric blanket, physical vapor deposition or chemical
vapor deposition of a Ta-based diffusion barrier and deposition of copper.
Thermal cycle from room temperature to 350-400 C is used to anneal out the
device damage from reactive ion etch. It is followed by chemomechanical
polishing to flatten the layer and thus produce a planar layered structure. Figure
1 presents the atomic force microscope image of a real interconnect line
manufactured by IBM.
Typical
Cu grain
Grain
boundary
Ta liner
Low-k
dielectric
m
Figure 1:
Atomic force microscope topographical map of Cu-damascene.
During thermal processing, the conductor is subjected to considerable stresses

from the thermal mismatch between the substrate, the dielectric and the diffusion
barrier. The reliability of the conductors and interfaces depends on the state of
stress. The deformation and stress relaxation in these nanometer scale structures
is highly complex. We may assume that dislocation activity is not relevant at
interconnect size scales and processing temperature ranges, as suggested, for
687
example, by Kobrinsky et al. [1]. The deformation may be attributed to the CuCu grain boundary and Cu-Ta interfacial sliding, as well as to the material build
up due to diffusional creep. The experimental background for this assumption
was provided by Gross et al. [2], who developed an AFM method to measure
out-of-plane deformation resulting from thermal cycling and applied this
technique to observe the Cu-polyimide interconnect structure.
In this paper we develop the numerical method to predict the diffusion-based
deformation that occurs in the process of interconnects fabrication. To describe
this phenomenon adequately, the mathematical formulation is needed that
accounts for the coupled nature of mechanical variables and concentration of
species such as Cu atoms, vacancies and impurities. The decoupling of creep
problem is achieved by its decomposition into the linear elasticity and mass flow
subproblems, which are solved in an iterational process. Rzepka et al. [3] applied
similar technique to a 3-D model of interconnect. We propose a different
approach by utilizing the thermodynamical coupling equations and the concept
of grain boundary region of finite thickness. A general discussion of latticebased thermodynamics with respect to diffusional creep in interconnects was
presented by Garikipati et al. [4]. Although our treatment is considerably
simplified, it nevertheless gives adequate description of the creep/stress
relaxation behaviour, and at the same time can be easily adapted to use with
existing commercial finite element packages. We also devote special attention to
the modelling of grain boundary regions, which are expected to be the major
path of vacancy diffusion at the temperature ranges used in interconnects
processing.
Decomposition of diffusional creep problem into elasticity

and mass flow subproblems
The Arrhenius-type expression for equilibrium concentration of vacancies in the

bulk at a given temperature is obtained from the minimum of Gibbs free energy
Gv written for dilute vacancy concentration (Porter and Easterling [5])
Cveq = exp ( Gv / kT ) .
(1)
In this equation, k is Boltzmann constant and T is the absolute temperature.

The Gibbs free energy can be expressed as Gv = H v T S v , where H v and
S v are the enthalpy and entropy of vacancy formation. The estimate for
entropy of vacancy formation for Cu is typically given as exp ( S v k ) 3 .

The enthalpy is given by H v = Q f + h , where Q f is the activation
energy of vacancy formation, h is the hydrostatic stress, and is the atomic
volume. The energy of vacancy formation at grain boundaries and free surfaces
will be somewhat less due to reduced constraint.
Since there are no hydrostatic stress gradients in a uniaxially loaded solid, the
vacancy concentration gradients that drive Nabarro-Herring and Coble creep

should not exist. Nabarro, Herring, and Coble all assumed that the equilibrium
concentration in the region adjacent to the grain boundary is given by n , the
stress normal to the boundary. This can be rationalized by noting that the elastic
constants of the grain boundary are different from that of the bulk as reported by
Zhou and Huang [6] and other authors. Therefore, we assume that
H v Q f + n which is consistent with the implicit assumptions of Nabarro,
Herring, and Coble.
We propose that vacancy creation or annihilation occurs instantaneously at
grain boundaries and free surfaces, and that these surfaces are infinite sinks and
sources of vacancies. Defining the temperature dependent stress-free vacancy
concentration as C0 = exp ( ( T Sv Q f ) / kT ) , the equilibrium vacancy
concentration in the GB region is given by
Cveq = C0 exp ( n / kT ) .
(2)
We ignore the contribution of dislocations as vacancy sources and sinks since

they are rarely observed in the interior of nanoscale grains (see Kong et al. [7]
and references therein).
The vacancy diffusivity is given by D = D exp ( Q / kT ) where Qm is the
v
v0
activation energy for vacancy motion. We assume that Qm is less in a region

adjacent to the grain boundary due to disorder near the interface and reduced
elastic constraint. We choose the thickness of the enhanced diffusivity region as
3-4 monolayers, assuming the absence of impurities. We define the vacancy
diffusivity at interfaces as Dv , gb = D0 gb exp Qm , gb kT
where Qm , gb = Qm
( < 1) . We are not aware of a precise method to measure the activation energy
for vacancy motion through the grain boundary and will treat it as an adjustable
parameter that may be affected by impurities, grain boundary roughness,
crystallographic orientation and whether the opposing interface has similar
diffusivity.
Neglecting the local vacancy relaxation strain (Hirth and Lothe [8]) the mass
or atomic flux jA is opposite to the vacancy flux jv :
jA = jv .
(3)
The atomic flux field can be treated as the diffusive flow velocity field in a body
that is statically fixed. The gradient of the mass flow velocity jv defines the
rate of creep deformation
cr = 1 2 jv + jv
).
(4)
The creep strains defined above are caused by diffusion mass flow due to stress
field gradients. Assuming that total strain consists of elastic and creep
components, the elastic part of the strain can be related to the total stress by
Hookes law
= C : ( cr )
689
(5)
where C is the elastic stiffness matrix.

To analyze deformation of a grain assembly, we start with the uniform stressfree vacancy concentration C0 . Application of external load or prescribed
displacements may result in stress gradients throughout the structure. This
produces vacancy concentration gradients according to eqn (2). Vacancy
diffusion leads to the accumulation of diffusion creep strains of eqn (4) and the
evolution of stress field. The coupled vacancy diffusion elasticity problem
governed by eqns (2), (4)-(5) is solved by the finite element technique presented
in the next section.
We implement the iterative approach to analyze the transient processes of
diffusion creep or stress relaxation. It is assumed that with time step
appropriately selected, we may adopt the staggered procedure on each iteration,
i.e. solve the elasticity and mass flow subproblems independently, while holding
the variables of another subproblem fixed.
The transient vacancy concentration subproblem is governed by the
conservation law
Cv
+ jv = 0
t
(6)
where vacancy fluxes jv obey the Ficks constitutive equation
jv = Dv Cv .
(7)
On each time step we integrate eqn (6) with fixed boundary vacancy
concentrations given by eqn (2). The diffusive fluxes that occur during the time
step are used to compute the creep strain increment cr by integration of
kinematic relation (4). The increment cr is a component of total strain
increment . We choose the displacement as the independent variable, hence
needs to be expressed in terms of displacement increments u . Assuming
small strain increments, we employ the relation ij = 1 2 ui , j + u j ,i ,

where comma denotes partial differentiation with respect to the corresponding
coordinate. The principle of virtual work may be written in incremental form as
: C : ( cr )dV = P
(8)
where P is the increment of the externally applied load. This equation is the
basis for the finite element formulation of mechanical subproblem.
Stress relaxation in 1D periodic row of square grains
We consider stress relaxation in a 2D (plane strain) grain array shown in Figure

2 subjected to applied strain x = 0.002 . The numerical values used in the
computation are listed in Table 1.

Free surface
symmetry
Grain
boundary
symmetry
Figure 2:
Model of periodic square grain array used for validation of finite

element procedure.
In the analysis of calculated results, we assume that the time dependence of

average stress 11 = 1 V
V 11dV
can be approximated by Maxwell model
relaxation function
(9)
= 0 exp ( t ) .
In this equation, is the Maxwell model relaxation time defined through spring
stiffness and dashpot viscosity as = . We use the concept of
relaxation time to characterize the time period of relaxation of average stress in
the model, and to scale the time t in the graphs throughout the rest of this
section.
Table 1:
Values of parameters used in the stress relaxation modeling of

periodic grain array to compare numerical and theoretical
predictions.
Parameter
Grain width d
Burgers vector b
Grain boundary region thickness
Atomic volume
Melting temperature Tm
Value
100 nm
0.25 nm
1 nm
1.18 10-29 m3
Youngs modulus E
Poissons ratio
128 Gpa
0.33
Initial compressive stress on internal grain boundary 0
286.8 Mpa
Lattice vacancy diffusion pre-exponential D0 L
2 10-5 m2/s
Lattice activation energy for vacancy motion Qm , L
113.3 kJ/mole
Grain boundary vacancy diffusion pre-exponential e D0 gb
5e-15 m3/s
Grain boundary activation energy for vacancy motion Qm , gb
67.98 kJ/mole
Grain boundary activation energy for vacancy formation Q f
83.7 kJ/mole
1356 K

10.000
unloaded
loaded
Displacement y, nm
691
T = 0.3Tm
T = 0.5Tm
T = 0.6Tm
t=0
1.000
T = 0.7Tm
T = 0.8Tm
T = 0.9Tm
t=
loaded at t = 0
0.100
0.010
0.001
-50
-40
-30
-20
-10
10
20
30
40
50
x, nm
Figure 3:
Free surface profile at time t = , with stress relaxation modeled at

different temperatures. Tm is the melting temperature of copper.
Application of fixed strain x results in initial uniform stress 0 . Stress

relaxation occurs as material flows towards the interfaces with the lowest
magnitude of normal stress. For the structure under consideration, the smallest
normal stress is at the free surface. The profiles resulting from the accumulation
of material along the free surface at various temperatures are shown in Figure 3.
The lump at the grain junction region results from material migration along the
grain boundary (Coble mechanism). The height of the lump is approximately the
same for all considered temperatures. It can be explained by the fact that the
increased diffusivity at elevated temperatures is compensated by higher rate of
vacancy flow (eqn 7). We also observe the increased accumulation of material
along the free surface at high temperatures ( T > 0.5Tm ). This effect can be
explained by contribution of vacancy diffusion through the grain interior
(Nabarro-Herring mechanism).
Figure 4 shows the stress 11 distribution along the internal grain boundary at
different time instances. It can be seen that the stress relaxation occurs almost
instantaneously at the junction between grain boundary and free surface, and
gradually advances along the internal grain boundary. The stress distribution is
shown at low and high temperatures, T = 0.3Tm and T = 0.7Tm . The stress
gradients are lower at high temperatures. This effect may be attributed to the
contribution of Nabarro-Herring mechanism to stress relaxation. As seen in
Figure 4, the difference in stress evolution at different temperatures is not
substantial for a one-dimensional array of grains. Hence the relaxation of average

stress as a function of non-dimensional time t can be plotted as a single
curve, see Figure 5. The insets illustrate the evolution of stress field 11 in the
grain. Note that relaxation time is different for different temperatures, for
example 1 s for T = 0.5Tm and 0.001 s for T = 0.7Tm .
1.4
610-4
T=0.3Tm
1.2
T=0.7Tm
610-3
11/0
0.03
0.09
0.8
0.3
0.6
0.6
0.4
y
x
0.2
0
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
y/d
Figure 4:
Evolution of stress distribution along the internal grain boundary (x

= 0) at temperatures T = 0.7Tm (solid lines) and T = 0.3Tm (dashed
lines).
1.0
t=
0.9
0.10
0.8
<11>/0
0.7
0.20
0.30
0.40
0.6
0.5
0.4
0.80
0.90
1.10
0.3
0.35
1.00
0.95
0.2
0.1
t = 0.01
0.0
0.0
Figure 5:
0.1
0.2
0.50
0.65
0.80
t = 0.1
0.3
0.4
0.5
t /
0.6
0.7
0.8
0.9
1.0
Relaxation of average stress 11 . Insets show the distribution of

11 at selected time instances.
693
The numerical predictions are compared to the theoretical estimate based on

the formula given by Frost and Ashby [9]
e Dgb
42
DL 1 +
=
(10)
2
dDL
kTd
where is the shear strain rate, d is the grain width, e is the effective
thickness of grain boundary (on the order of magnitude of Burgers vector b ) and
is the applied stress. Note that the diffusivities DL and Dgb in eqn (10) are
based on the energy of both formation and motion of vacancies. They are
different from the vacancy motion only diffusivities Dv , L and Dv , gb used in
our finite element formulation which explicitly includes the vacancy formation
energy in eqn (2) for stress dependence of vacancy concentration. We obtain the
theoretical estimate of relaxation time by computing the dashpot viscosity from
eqn (10) as = . The shear strain rate in eqn (10) accounts for the
contribution of both Nabarro-Herring and Coble mechanisms to creep rate,
which is consistent with the numerical modeling approach of sections 2 and 3.
Figure 6 shows good agreement of theoretical and numerical predictions for
relaxation time. The largest difference is in the temperature range 600-900 K.
1.0E+09
Maxwell model
Relaxation time , s
1.0E+07
FEA
1.0E+05
1.0E+03
1.0E+01
1.0E-01
1.0E-03
1.0E-05
400
500
600
700
800
900
1000
1100
1200
1300
Temperature T, K
Figure 6:
Comparison of diffusion creep theoretical estimate of relaxation

time to the finite element predictions.
Conclusions
The stress-induced mass flow constitutes the major mechanism of creep in Cudamascene interconnects during thermal processing of integrated circuit devices.

This mechanism can be modelled by the iterative procedure which decouples the
original creep problem into the static stress analysis and transient mass flow
subproblems. Introduction of the grain boundary region with enhanced
diffusivity into the finite element scheme allows direct modelling of effects
associated with Coble creep behaviour.
Acknowledgements
This work was supported by the National Science Foundation, Division of
Manufacturing and Industrial Innovation, under Grant No. DMI-0300216. The
donation of the Cu-damascene sample by IBM is gratefully acknowledged.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
Kobrinsky J., Thompson C. & Gross M., Diffusional creep in damascene

Cu lines. Journal of Applied Physics, 89(1), pp. 91-98, 2001.
Gross T.S., Kamsah N. & Tsukrov I.I., Scanning probe microscopy
generated out-of-plane deformation maps exhibiting heterogeneous
nanoscale deformation resulting from thermal cycling of Cu-polyimide
damascene interconnects. Journal of Materials Research, 16(12), pp.
3560-3566, 2001.
Rzepka S., Meusel E., Korhonen M. & Li C.-Y., 3-D finite element
simulator for migration effects due to various driving forces in
interconnect lines. AIP Conference proceedings, 491, pp.150-161, 1999.
Garikipati K., Bassman L. & Deal M., A lattice-based micromechanical
continuum formulation for stress-driven mass transport in polycrystalline
solids. Journal of the Mechanics and Physics of Solids, 49(6), pp. 12091237, 2001.
Porter D.A. & Easterling K.E., Phase Transformations in Metals and
Alloys, Chapman & Hall: London, 1992.
Zhou L.G. & Huang H., Are surfaces elastically softer or stiffer? Applied
Physics Letters, 84(11), pp. 1940-1942, 2004.
Kong Q.P., Cai B., Lu L. & Lu K., The creep of nanocrystalline metals
and its connection with grain boundary diffusion. Defect and Diffusion
Forum, 188-190, pp. 45-58, 2001.
Hirth J.P. & Lothe J., Theory of dislocations, Wiley-Interscience: New
York, 1982.
Frost H.J. & Ashby M.F., Deformation-Mechanism Maps, The Plasticity
and Creep of Metals and Ceramics, Pergamon Press: 1982.
695
Application of fuzzy sets to structural

reliability of existing structures
I. Mura
Department of Structural Engineering, University of Cagliari,
Cagliari, Italy
Abstract
Assessing the safety of a structure through inspection has always been part of the
practice of the structural engineer. Information obtained through these
inspections can be both objective and subjective. One can update the reliability
of an existing structure based on objective and subjective information through
the fuzzified Bayes theorem.
This paper cites the algorithm for computing the posterior probability of
failure based on visual inspection of existing structures by incorporating fuzzy
set theory into the Bayes theorem.
The failure design probability of a structural steel frame is updated.
Results indicate that the preciseness of the membership function associated
with the input data has little or no influence on posterior failure probability.
Keywords: reliability, Bayes theorem, fuzzy sets, existing structures, steel frame.
Introduction
In engineering practice one often has to deal with qualitative and vague
evaluations, commonly known as fuzzy. Such subjective fuzzy information is
to be found in the results of inspections of structures (such as bridges, frames and
so on) and are accompanied by those of a deterministic and probabilistic kind.
In particular, when we consider the question of evaluation of the safety level
of an existing structure, it appears evident that we cannot avoid taking into
account the quality of workmanship, the state of conservation of the elements
making up the structure and so on. Estimation of these qualities can be expressed
subjectively through variations of a linguistic nature. The values of these
variables, as for example the quality of the workmanship is good, or the state
doi:10.2495/HPSM06068

of conservation of the bolts is poor, being vague and imprecise (which is to say
fuzzy) cannot be defined with any certainty. For this reason, it is impossible to
use them efficiently in the ambit of conventional statistical theories.
Modeling and the taking into consideration of vague and imprecise
information can be performed through use of the fuzzy set theory, which was
formulated by Zadeh in the 1960s and later developed by many other researchers
(see Zadeh [1], Kaufmann and Gupta [2]). The first application of fuzzy set
theory in civil engineering goes back to Blockley [3], who developed a method
for evaluating the influence of the many parameters that may reduce the a priori
safety factor n which determines the probability (pf = 10-n ) of the occurrence of a
given structural failure by means of linguistic variables and a fuzzifier.
Subsequently, the procedure of taking into account subjective information
was developed to allow the reaching of different goals. Blockley [4] extended the
method illustrated in [3] to the study of twenty-three different structural failures.
Brown and Yao [5] estimated the effective strength of cast concrete in a given
structure starting from the results of compression tests on test pieces through the
taking into account of a fuzzy parameter appropriate for defining the quality of
the cast. Itoh and Itagaki [6] and Chou and Yuan [7] considered the problem of
evaluating the reliability of existing structures. The algorithms they described
allow the calculation of posterior probability based on the results of the visual
inspection of structural components by incorporating the fuzzy set theory into the
Bayes theorem. Wu [8] applied the Bayes theorem to evaluation of the reliability
of systems whose identifying parameters are assumed as fuzzy random variables
with an a priori distribution of the fuzzy kind.
Thus by combining the fuzzy set theory (which allows the expression of
linguistic evaluations through specific functions called membership functions)
with the Bayes theorem it is possible to solve the problem of evaluating the
reliability of an existing structure and determine the posterior reliability of
uncertain parameters on the basis of all the results (deterministic, probabilistic
and fuzzy) of inspections.
But the use of membership functions requires the solution of the problem of
their modeling. The analysis and evaluation of damage to a structural element is
in fact a difficult process in which human judgement plays a most important role.
In the literature we can find many different methods proposed for the modeling
of membership functions based on expert judgements. These operations are
generally quite costly and laborious, as well as requiring long periods of time.
Thus there has been a tendency to restrict membership functions to well-known
forms. Triangular, left-shoulder, right-shoulder and trapezoidal or, more
generally, piecewise linear, functions are common. Also used are standard
Gaussian or Sigmoid type curves.
In this work we first illustrate the essential functions of Bayes fuzzified
theorem (Asai and Negoita [10], Kandel [11]). In the illustration, the theorem
will be applied to the updating of design failure probability for the steel frame
structure previously studied in [7], based on inspection results considered as
fuzzy parameters. To evaluate the influence of the form of the membership
functions necessary in defining inspection judgements on the final result, we use
697
those proposed in [3] and [7] (triangular and piecewise linear in form) in our
calculations. The two solutions obtained are then compared.
The fuzzified Bayes theorem
The Bayes theorem provides a method that allows inclusion of new information
in a priori probabilistic evaluations, thus producing a new probability value for
the occurrence of an event (Benjamin and Cornell [9]).
When there is no fuzzification and we take into account the theorem of total
probability, the Bayes theorem is defined by the relation:
P (B J | A) =
P ( A | B J ) P (B J )
m
(1)
P (A|Bk ) P (Bk )
k =1
In Eqn (1) BJ is the Jth unknown random parameter having a known a priori
distribution P(BJ ), whose probability is to be updated; A is the random sample
representing the parameter to be inspected; m is the number of mutually
exclusive and totally exhaustive events. P(BJ | A ) is commonly known as the
posterior probability of BJ after statistical event A has taken place.
Equation (1) is thus valid if event A can be expressed objectively, which is
not the case of a fuzzy event. Therefore the above formulation of the Bayes
theorem cannot be used with information of the type: the crack is small or the
state of bolt conservation is poor, and so on, which are vague and imprecise and
cannot by defined with any certainty. To express such fuzzy information,
specific linguistic variables are introduced. In particular, function (x) is
introduced to express the probability of fuzzy event .
Taking this into account, we find that the conditional probability of fuzzy
event can be expressed as:
~
P A | BJ =
x
~
A
(x ) f x| B J (x ) dx
(2)
where fx|BJ (x) is the conditional probability function that predicts event BJ for
support x. Equation (2) is valid if the support is continuous while if the support is
discrete, the equation is written:
~
P A | BJ = A~ (x ) P x | B J (x )
x
(3)
Briefly stated, the posterior probability of event BJ, which takes into account
the observation of fuzzy event , is obtained by modifying Eqn (1) in accordance
with Eqns (2) and (3).
The fuzzy Bayes theorem ([10], [11]) can therefore be expressed by the
following relations. Where the support is continuous, taking into account Eqn
(2), Eqn (1) is written:
A~ ( x ) f x | B J ( x ) dx P (B J )
P BJ | A = x
m
A~ ( x ) f x | Bk (x ) dx P (B k )
k=1 x
(4)
where the support is discrete, taking into account Eqn (3), Eqn (1) is written:
~
x A ( x ) P x| B J P (B J )
~
P BJ | A =
m
A~ (x ) P x| Bk P (B k )
k=1 x
(5)
3 Examples of application
The example under consideration (see [7]) concerns the updating of the design
failure probability of a steel frame following an inspection. It is assumed that
such a probability is equal to Pf = 10-5, and that this represents a mean reliability
value that could be calculated by considering the construction of a large number
of similar structures. Numerous components (both structural and non-structural)
are inspected. Since the conditions of these components influence the overall
failure probability of the structure, they assume the role of probability
parameters. It is supposed that during the inspection the following parameters are
considered: connections, foundations, alignment, columns, beams, braces, bolts
and paint.
3.1 Statistical parameters
We consider event BJ which may correspond to event B1 (= the structural
component is safe) or to event B2 (= the structural component is near failure).
The conditional probability function of random event BJ, owing to the
support of quality x, supposedly assumes the following quadratic forms:
N 6
g ( X | B1 ) = i
(0.1) 2 (x 10 ) 2 + 1.0
6
(6a)
N 6
g( X | B2 ) = i
(0.1) 2 x 2 + 1.0
6
(6b)
since Ni is the index of the importance of the parameter or structural component i

and x is the support quality. Since the above functions (6) represent probability
functions, they must satisfy the basic axiom of probability. They are therefore
normalized in accordance with the following relation:
g ( x | BJ )
(7)
Pi ( X | BJ ) =
g ( x | BJ )
x
699
0.14
P(X|Safe)
1
0.12
0.10
3
4
5
6
0.08
0.06
0.04
0.02
0
Figure 1:
4
6
Support of Quality, x
10
Conditional probability function for Quality at various importance

indices given as Safe.
0.14
0.12
0.10
P(X|Failure)
6
5
0.08
4
0.06
0.04
0.02
1
0
Figure 2:
4
6
10
Conditional probability function for Quality at various importance

indices given as Failure.
Functions (6) must be defined so as to respect the weight assumed by each

parameter in the evaluation of the overall safety of the structure. The
meaningfulness of each parameter in overall safety is given by the indices of

importance. Parameters are ordered on the basis of their indices of importance.
Indices of unit value correspond to the maximum importance. Numerically
increasing values of the index imply a reduction in the importance of parameters.
In the part that follows it is supposed possible to assume for each index, on
the basis of expert judgement, the values given in Table I. It is to be pointed out
that functions (6) may assume different forms for each parameter. In particular,
the index of importance may in such forms be defined implicitly.
The results of normalization calculations are shown in Figures 1 and 2.
Finally, in order to simplify calculations, we allow the parameters to be
mutually independent, as are their effects on structural reliability.
3.2 Probability of inspection results
It is supposed that the quality of each of the parameters considered above can be
verbally described by fuzzy evaluations of the kind: the state of conservation of
the bolts is poor. Below, in order to evaluate the influence of the modeling on
the results supplied by the fuzzified theorem, two different circumstances will be
considered. The first kind of modeling, which will be referred to as type 1 and
which is taken from [3], is indicated in Figure 3 and corresponds to functions (8).
0.10 x + 1.0
0.40 x + 1.3
P ( x) = 0.30 x + 1.1
0.10 x + 0.5
if
x 1
if 1 x 2
if 2 x 3
if 3 x 5
if
x5
0
0.20 x 0.4
0.40 x 1.0
N ( x) =
0.40 x + 3.0
0.20 x + 1.6
if
if
if
if
if
if
0 x2
2 x3
3 x 5
5 x7
7 x8
8 x 10
(8b)
if
if
if
if
if
x5
5 x7
7 x8
8 x9
9 x 10
(8c)
G ( x) =
0
0.10 x 0.5
0.30 x 1.9
0.40 x 1.7
0.10 x
(8a)
The second kind of function modeling, which will be referred to as type 2, is

taken from [7]; it is expressed by means of relations (9) and is indicated in
Figure 4.
701
1.00
0.80
Normal
Good
Grade of Membership
Poor
0.60
0.40
0.20
0.00
Figure 3:
4
6
10
Membership function for Quality of type 1.
1.00
Grade of Membership
0.80
Normal
Poor
Good
0.60
0.40
0.20
0.00
Figure 4:
4
6
Membership function for Quality of type 2.
10

2.25 x + 1.0
0
P( x ) =
0.20 x
0.20 x + 2.0
N ( x ) =
0.25 x 1.5
0
G ( x ) =
if
x4
if
x>4
if
x5
if
x>5
if
x6
if
x>6
(9a)
(9b)
(9c)
Considering that three quality judgements can be applied to the eight

parameters, 6561 possible combinations of inspection results can be considered.
The three possible cases of inspection results that are considered herein (taken
from [7]) are illustrated in Table I below.
3.3 Updating of failure probability
Having assigned the index of importance to each of the structural parameters
considered, we can calculate the posterior failure probability of the frame both
on the basis of functions (7) and the membership functions reflecting inspection
results (8) and (9). The posterior probability of failure of the frame, taking into
account the occurrence of fuzzy events 1, 2 8, is obtained by extending
Eqn (1) to the case of more than one fuzzy variable. We have:
~ ~
~
P( A1 , A 2 ... A 8 | BJ ) P( BJ )
~ ~
~
P( BJ | A 1, A 2 .... A 8 ) = 2
~ ~
~
P( A 1, A 2 ... A 8 | Bk ) P ( Bk )
(10)
k =1
Having supposed that events 1, 2 8 are independent, we indicate by:

~
P ( A1 |
P( B1J ) = 2
~
BJ ) P ( BJ )
(11)
P ( A 1| Bk ) P ( Bk )
k =1
the posterior probability of event BJ on the basis of observation of the first

parameter. Obviously we also have the following:
P( BJ2 ) =
~
P ( A 2 | BJ ) P ( B1J )
2
~
P( A 2 | Bk ) P( Bk )
(12)
k =1
Having indicated by P(B J) the posterior probability of event BJ following

observation of parameter i, in general we can write:
703
P( BJi ) =
~
P( A i | BJ ) P( BJi 1 )
2
~
i 1
P( A i | Bk ) P( Bk )
(13)
k =1
Equation (13) above is immediately explicable by using Eqn (2) when the
support is continuous and Eqn (3) when it is discrete.
Table 1:
CASE N
Parameter
(1)
(2)
Posterior failure probability for the type 1 (col. 5) and type 2 (col.
6) membership functions adopted.
Connection
Foundation
Alignment
Columns
Beams
Bracing
Bolts
Painting
Connection
Foundation
Alignment
Columns
Beams
Bracing
Bolts
Painting
Connection
Foundation
Alignment
Columns
Beams
Bracing
Bolts
Painting
Index Ni
Quality Ai
Pif ( 1 )
Pif ( 2 )
1
1
1
2
4
4
4
6
1
1
1
2
4
4
4
6
1
1
1
2
4
4
4
6
Normal
Poor
Normal
Poor
Normal
Poor
Poor
Normal
Good
Normal
Poor
Normal
Good
Normal
Poor
Good
Good
Good
Good
Normal
Good
Normal
Good
Normal
1.00E-5
3.16E-5
3.16E-5
6.94E-5
6.94E-5
9.52E-5
1.31E-4
1.31E-4
3.17E-6
3.17E-6
1.00E-5
1.00E-5
7.29E-6
7.29E-6
1.00E-5
1.00E-5
3.17E-6
1.00E-6
3.17E-7
3.17E-7
2.32E-7
2.32E-7
1.69E-7
1.69E-7
1.00E-5
3.11E-5
3.11E-5
6.76E-5
6.76E-5
9.24E-5
1.26E-5
1.26E-5
3.22E-6
3.22E-6
1.00E-5
1.00E-5
7.32E-6
7.32E-6
1.00E-5
1.00E-5
3.22E-6
1.04E-6
3.34E-7
3.34E-7
2.44E-7
2.44E-7
1.79E-6
1.79E-6
(3)
(4)
(5)
(6)
Results of numerical processing and conclusions
Results of calculations performed are shown in Table 1 and correspond to the

three circumstances of the inspection results considered.
In column 5 of Table 1 we find the updated values of failure probability
deduced by adopting functions (8); in column (6) we see those deduced by
adopting functions (9).

Calculations were carried out by adopting the hypothesis of a discrete support
by means of the use of a simple program created specifically for the purpose.
The values of failure probabilities, later updated by means of Eqn (13), on the
basis of inspection results, are found to be practically insensitive to the different
definitions of the membership functions adopted and, to all practical purposes,
substantially coincident.
As can easily be seen, an average failure probability, like that of the design,
is reduced if following inspection it is found that the parameters that globally
define it turn out to be more than average. On the contrary, it increases when it is
found that the parameters that globally define it are less than average.
Generally speaking, the results indicate that the preciseness of the
membership function associated with the input data has little or no influence on
the posterior failure probability. Thus, the fuzzified Bayes theorem has practical
applications in assessing the posterior failure probability of existing structures,
with minimal expert input to establish the necessary membership functions and is
a very powerful tool of analysis.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
Zadeh L.A., Fuzzy sets. Information and Control, Vol. 8, pp. 338-353,
1965.
Kaufmann A., Gupta M.M., Fuzzy mathematical models in engineering
and management science. North-Holland, Amsterdam, 1988.
Blockley D.I., Predicting the likelihood of structural accidents. Proc. Inst
Civ. Eng., 59(2), pp. 659-668, 1975.
Blockley D.I., Analysis of structural failures. Proc. Inst Civ. Eng., 62(1),
pp. 51-74, 1977.
Brown C.B., Yao T.P., Fuzzy sets and structural engineering. Journal of
Structural Engineering, ASCE, 109(5), pp. 1211-1225, 1983.
Itoh S., Itagaki H., Application of Fuzzy-Bayesian Analysis to Structural
Reliability. Proc. of ICOSSAR '89, ASCE, N.Y., pp. 1171-1174, 1989.
Chou K.C., Yuan J., Fuzzy-Bayesian approach to reliability of existing
structures. Journal of Structural Engineering, ASCE, 119(11), pp. 32763290, 1993.
Wu H.C., Bayesian system reliability assessment under fuzzy
environments. Reliability Engineering & System Safety, 83, pp. 227-286,
2003.
Benjamin J.R., Cornell C.A., Probability, statistics and decision for civil
engineers. Mc Graw Hill, New York, N.Y, 1970.
Asai K., Negoita C.V., Introduction to fuzzy set theory. Ohm Press,
Tokio,1978.
Kandel A., Fuzzy mathematical techniques with applications. Addison
Wesley, Reading, Massachusset, 1986.
705
Non-linear response of combined system,

3D wall panels and bending steel frame
subjected to seismic loading
M. Z. Kabir, A. R. Rahai & Y. Nassira,
Department of Civil Engineering, AmirKabir University of Technology,
Tehran, Iran
Abstract
3D wall panels are used in the construction of exterior and interior bearing and
non-load bearing walls and floors of building of all types of construction. This
system consists of a welded wire space frame integrated with a polystyrene
insulation core and two layers of concrete on both sides. In this paper, attention
is focused on the experimental measurements of the seismic response of 3D wall
panels surrounded by a steel bending frame. The approach of quasi-static cyclic
loading is employed using horizontal actuators to the combined system. The
vertical, lateral and horizontal displacements are measured by LVDT equipment.
The failure mechanism of 3D wall panels is described in detail. The evaluation
of strength and stiffness degradation of the whole system is presented based on
the envelope force-displacement curve of actual specimens under cyclic loads.
The results of the current study are shown in the form of ductility factors,
hysteresis loops and load-displacement envelope curves. The comparison
between the ductility of sole steel frames, 3D shear walls and the combined
system as the main theme of the current research is presented. Finally, this work
clarifies the benefits of using 3D wall panels as a strengthening method for
existing steel frame buildings and confirms the feasibility resistance of such
combined systems.
Keywords: 3D-panels, combined system, cyclic loading.
Introduction
3D wall panels are used in construction of exterior and interior bearing and nonload bearing walls and floors of building of all types of construction. This system
doi:10.2495/HPSM06069

consists of a welded wire space frame integrated with a polystyrene insulation
core. The wall panel is placed in position and wythes of concrete are applied to
both sides. Wall panel receives its strength and rigidity by the diagonal cross
wires welded to the welded-wire fabric on each side. This combination produces
a truss behavior, which provides rigidity and shear terms for full composite
behavior. Figure 1 shows schematically the 3D panel.
Figure 1:
3D Sandwich panel.
Salmon and Einea [1] presents the results of full-scale test of prototype
sandwich panel under transverse loading in a vertical position. Nijhawan [2]
measured experimentally the interface shear force and designed the shear
connectors. Eiena et al. [3] used the plastic composite diagonal elements to
implement in sandwich panel as shear connector for increasing the thermal
insulation of this system.
Through this study the behavior of 3D panels in combination to steel moment
frame was investigated, the fracture mechanism of concrete wythes and the
adequacy of steel bars designed based on ACI 318-95 and procedure of PCI
design handbook [4].
Theoretical study
2.1 Ductility capacity

The term ductility in seismic design is used to mean the ability of structure to
undergo large amplitude cyclic deformation in the inelastic range without a
substantial reduction in strength. The ductility is calculated by various types,
which reflects structural behavior. The displacement ductility capacity is
defined as:
max
y
where max max is the maximum displacement and y is the displacement at

yield
Experimental program
Four 3D walls are provided to be combined with portal steel frames. These
specimens are considered to represent the critical and structural elements with a
707
rectangular cross section and are experimentally subjected to low cycle

horizontal loading regimes. The type of failure and behavior of such structural
elements are investigated in details.
Figure 2:
3D panel with steel frame.
3.1 Wall details

The wall specimens have 1200 mm height, 640 mm width, 140 mm constant
thickness including 40mm shotcrete in each side and 60 mm of expanded
polystyrene core. The welded wire fabric is consisted of a cold rolling of steel
bar with final outside diameter of 3.5 mm in accordance with ASTM A82 and
automatic welding process with accordance of ASTM A185. The yield and
ultimate strength of drawn and annealed wires are 4000 and 5200 MPa,
respectively. The shotcrete used for all specimens is used from Portland cement
(II), river sand with maximum 8 mm diameter, drinkable water. The (W/C) is
about 0.5 and the mix is made of 400 kg cement, 1750 kg sand and 180 kg water
for a unit cubic meter shotcrete. Compression tests are carried out on
(150*150*150 mm) standard cubes and provided cores from shotcrete. Tables 1
and 2 are shown results of the compression strength tests.
3.2 Frame details
Two IPE120 as columns and one IPE120 as beam are used for constructing of
flexural steel frame. The beam to column and column to base plate connections
are rigid based on Iran steel structure design code.
Table 1:
Geometry and compressive strength of standard specimens.
Specimen
No.
Wall
Dimensions
(cm)
Specific
Gravity
(kg/m3)
Max. app
lied force
(tone)
035
036
037
038
120*64
120*64
120*64
120*64
2330
2310
2290
2280
87
91
82
85
Cube
compressive
strength
(bars)
387
404
364
378

Table 2:
Geometry and compressive strength of cylindrical cores.
Core No.
Core Dimensions
(Diameter*length)
Specific
gravity
(kg/m3)
Max.
applied
force (kg)
1
2
3
4
5
107.6*54.4
100.4*54.4
99.5*54.4
108.3*54.4
103.3*54.4
2150
2170
2230
2100
2250
5350
5100
5250
5400
5200
Table 3:
Specimen
No.
1
2
Specimen
Dimensions
Tensile strength.
Specific
gravity
(kg/m3)
Max.
applied
force (kg)
Slump
(cm)
2300
2320
3700
3500
8
8
15*30
15*30
(a) side view

Figure 3:
Core
Compressive
strength
(kg/cm2)
230.2
219.4
225.9
232.3
223.7
Cylinder
tensile
strength
(kg/cm2)
52
50
(b) top view

Details of steel frame.
3.3 Loading history and test procedure

To simulate loading sequences that might be expected to occur during
earthquake, simplified types of horizontal cyclic load history are adopted. Since
no standard cyclic test procedures has been introduced for testing of such system,
the horizontal load is applied at a quasi-static rate in displacement controlled
cycles with different patterns which corresponded to three major states, namely
709
cracking state, yielding state and ultimate state. The applied displacement is
started from 0.5 mm to 5 mm in 10 cycles. In the second phase of loading, the
increment of displacement is increased to 1 mm and after 19 cycles in
displacement of 14 mm the displacement increased to 2 mm and after 23 cycle
the increment increased to 4 mm and in 30 cycle system was failed.
Linear transducer of types LVDT is used to measure and monitor the
horizontal displacements at top, mid height and bottom of specimens. The
measured values of applied load and displacement are recorded by a computer
data logger capable of measurement to sensitivity ranges of 0.1 N, 0.001 mm,
respectively.
40
30
20
Displacement (mm)
10
0
0
10
15
20
25
30
-10
-20
-30
-40
-50
Cycle Number
Figure 4:
Figure 5:
Applied successive cyclic displacement.
The crack patterns and failure mechanism in specimen at late stages

of cyclic loading.
3.4 Experimental results

Table 4 shows the results of applied displacement in type 1 specimen during
cyclic loading. Up to 0.4 mm displacement, the panel behaves in elastic zone and
the first crack is occurred at the location of 200 mm below the crest, where the
connector reinforcements are discontinued. In early stages of loading, the cracks
are minor and their direction is mostly horizontal. By increasing applied
displacement level, the cracks propagate in both sides of panel. Figure 5

represents the shear cracks pattern at near final stages. It is seen that, the cracks
are more visualized and many of them are opened.
The direction and propagation of cracks in panels are shown in figure 5. The
first crack is located above the base anchor bars when the applied displacement
reaches to 3 mm. The main cracks in these types of wall panels are at the bottom
area and it continues up to one third of panel height. It is seen that the direction
of crack at the area near the panel edges are mostly horizontal and it obliged to
45 in the middle of panel.
Table 4:
Cycle No.
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
28
Experimental results in cyclic successive loading for combined

system.
Horizontal
Displacement in
both sides (mm)
forth
back
0.270
0.123
0.616
0.347
0.966
0.672
1.323
1.061
1.703
1.470
2.075
1.878
2.450
2.312
2.848
2.733
3.323
3.171
4.015
4.029
4.797
4.907
5.576
5.771
6.274
6.653
6.986
7.415
7.723
8.140
8.455
8.972
9.103
9.787
9.793
10.563
11.087 12.019
12.299 13.404
13.914 15.174
15.787 17.439
17.624 20.469
21.316 24.748
24.728 29.096
28.025 33.483
31.395 38.939
34.315 43.566
Reaction Forces in
Both sides
(kN)
back
forth
Panel Stiffness in
both sides (kN/m)
back
forth
15.144
27.891
39.463
50.134
60.301
69.308
77.705
85.582
92.177
104.830
116.267
126.511
135.015
143.335
153.12
162.463
169.196
176.249
187.332
194.126
194.889
186.294
172.539
150.83
125.381
114.359
104.619
87.384
55884
45237
40847
37871
35397
33387
31713
30042
28468
26106
24235
22688
21520
20516
19825
19214
18586
17996
16895
15783
14006
11800
9790
7075
5070
4080
3332
2546
136762
83804
63561
50860
43414
38511
34540
31671
29432
26847
247.76
23284
21971
21158
20509
19686
18892
18228
17144
15992
14351
12252
8505
6179
4916
3697
2994
2582
16.808
29.113
42.745
53.966
63.828
72.346
79.857
86.559
93.353
108.191
121.580
134.373
146.174
156.906
166.95
176.630
184.905
192.553
206.064
214.369
217.773
213.666
174.111
152.937
143.06
123.793
116.588
112.512
711
Based on ASTM standard method, the applied displacement and their

corresponding reaction forces are plotted in a load deflection hystersis curves.
Their values are listed in table 4.
250
80
200
60
150
50
0
-60
-40
-20
-50
20
40
Reaction Force (KN)
Reaction Force (KN)
40
100
20
0
-80
-60
-40
-20
20
40
60
80
-20
-100
-40
-150
-60
-200
-80
-250
Displacement (mm)
Displacement (mm)
(a) combined system

Figure 6:
(b) steel frame
Load deflection Hystersis energy loops due to successive applied

displacement.
Discussion
Due to the significant stiffness of wall panels, up to 217kN, the majority of

produced energy in cyclic loading is absorbed by the action of combination of
wall and frame. It was shown, [5], the bare shotcrete wall could resist only about
20 kN. The corresponding displacement of total system is about 15.2 mm.
It should be noted that all the combined systems are essentially subjected to
the intended in-plane action of successive displacement where the out-of-plane
movement is prevented, figure 7. Measured maximum value of the vertical
displacement at the top of the wall is found to be insignificant since its value is
about 0.1% of the maximum wall displacement for all specimens.
Figure 7:
Lateral supports for out-of-plane movement.

4.1 Cracking
For all specimens, cracks are initially formed near the bottom part of the tensile
zone of the wall when only about 10% of the final horizontal deformation
capacity is applied. The significant inclined cracks initiated at the tension zone of
the wall during compressive reversal displacements. These cracks continue to
penetrate deeply into the centre of the wall towards the compression zone.
4.1.1 Strength reduction
Table 5 indicates horizontal load-carrying capacity of specimens 035 and 036,
respectively. It is seen from these tables that the cyclic displacement regime
employed in all specimens appeared to have had an insignificant effect on the
ultimate strength of the walls.
4.1.2 Stiffness and deformations
Variation of lateral deflection in successive cyclic loading and also average of
stiffness against horizontal reaction forces are illustrated in table 4. The
displacement relating to the first occurred cracking is about 10%. The horizontal
load versus top displacement is shown in figure 8, indicates a distinctly nonlinear deformation response. In terms of stiffness reduction, it should be stated
that the stiffness is gradually decreased for both, fig. 8.
4.1.3 Ductility
The capacity of the structural elements to deform beyond yield or elastic limit
with minimum loss of strength and stiffness depends upon their ductility. The
load-displacement envelope curve obtained from hystersis loops is sketched in
figure 7. The maximum displacement corresponding to 80% of ultimate load is
introduced as max. The displacement at the first yielding in panel, y, is
determined at the intersection of two lines. The initial tangent of envelope curve
with horizontal line passes at the level of 80% of Pu, [6]. It is described in
figure 8.
250
Reaction Force (kN)
200
150
100
50
0
0
10
20
30
40
Displacement (mm)
Figure 8:
Load deflection envelope curve.
50
Table 5:
The ductility values for wall specimens based on max /y.
Specimen
without gap
Wall panels
713
max
035
21.32
2.44
8.73
036
22.07
2.58
8.55
Conclusions
The current work describes the cooperation between 3D panels as infill wall and
traditional steel frames. The following conclusions remarks are raised based on
previous work by the first author [5]:
- For all specimens, the plastic hinge is formed at the extreme fiber of the wall
section and at the vicinity at the base, above the foundation. The distribution and
propagation of cracks show that 3D sandwich panels with limited height,
behaves in shear performance. The observed horizontal crack at the base prior to
ultimate state may be due to sliding at the vertical reinforcement. This caused
considerable reduction in the strength, stiffness and energy dissipation of the
specimens.
- Externally reinforcing 3D wall panels, which basically behaves in shear
manner, enhances more ductility in performance design approach and increases
substantially load carrying capacity of system.
Finally, this study clarifies benefits of using lightweight prefabricated panels
as a strengthening method for existing of steel frame building and confirms
feasibility resistance of such compound system.
Acknowledgement
The authors gratefully acknowledge the financial support from Pre-fabricated
sandwich Panel firm, SAP company in Tehran for present work
References
[1]
[2]
[3]
[4]
Salmon, D.C. and Einea, A., Partially composite sandwich panel

deflections, Journal of structural engineering, Vol. 121, No.4, pp. 778783, 1995.
Nijhawan, J.C., Insulated wall panels interface shear transfer, PCI journal:
98-101, 1998.
Einea, A., Todros, M.K., Salmon, D.C. and Culp, T.D., Culp, T.D. A new
structurally and thermally efficient sandwich panel system, PCI journal,
39(4):90-101, 1994.
PCI design handbook-precast and prestressed concrete, 3rd edition,
precast/prestressed concrete institute, Chicago, 1985.

[5]
[6]
[7]
[8]
Kabir, M.Z., Jahanpoor, A.R. and Rahbar, M.R., An estimation of

ductility and behavior factor of 3D sandwich shotcreted panels subjected
to monotonic shear loads, ERES Conference, Seville, Spain, 2003.
Uang, C., Establishing R and Cd factors for building seismic provisions,
ASCE, Journal of structural Engineering, vol. 177, 1991.
Duan L. and Reno, M., Performance based seismic design criteria for
bridges, in W.F. Chen (ed), Handbook of structural Engineering, CRC,
Press, 1997.
Newmark, N.M., and Hall, W.J., Earthquake spectra and design,
Earthquake engineering Res. Inst. El Cerrito, Calif., 1982.
715
Author Index
Abdel-Mooty M. A. N. ............ 419
Aiello R.................................... 449
Akita K. ........................... 351, 359
Al-Jabri K. S. ..................... 93, 381
Alnuaimi A. S. ........................... 93
Aoyama E. ............................... 169
Arasteh A................................. 269
Artemev A. .............................. 103
Avalle M. ................................. 249
Avila J. A................................. 439
Belingardi G. ........................... 201
Bhandari Y............................... 331
Bitar M. A................................ 419
Bossuyt S. ................................ 239
Boufas S................................... 311
Bouquerel J. ............................. 259
Brooks W................................. 481
Caliez M. ................................. 211
ern R.................................... 409
Chalioris C. E. ......................... 459
Cheng L. .................................. 501
Chew H. B. .............................. 501
Chiandussi G............................ 201
Cizmar D.................................. 429
Cooreman S. ............................ 239
Couroneau N.............................. 73
Crea F. ..................................... 449
Crooks R. ................................. 549
Cruz H...................................... 539
Custdio J. ............................... 539
Cuypers H. ................................. 21
Darling T.................................. 323
Dascotte E................................ 225
De Bolster E............................... 21
De Temmerman N. .................... 41
De Wilde W. P...............................
................... 3, 13, 21, 31, 625, 635
Debacker W. ............................ 625
Degrieck J. ............................... 259
Dns G. .......................... 301, 311
Dimiduk D. .............................. 331

Dobrila P.................................. 111
Domack M. S........................... 549
Estvez Cimadevila J............... 571
Estvez J. ................................... 63
Euler E. .................................... 225
Farag H. M............................... 419
Frachon A. ............................... 211
Frontera P. ............................... 449
Fujii T. ..................................... 179
Fujita H.................................... 189
Gan M. D. ................................ 511
Geddes B. ................................ 103
Ghosh S. .................................. 331
Gordon V. ................................ 581
Govorunov I. N........................ 343
Gratton M. ............................... 211
Grich W. M.............................. 685
Groeber M. .............................. 331
Gross T. S. ............................... 685
Guo T. F................................... 501
Gu J.......................................... 239
Hago A....................................... 93
Hendrickx H. ........................... 625
Henrotay C............................... 625
Hirayama T. ............................. 189
Hirogaki T. .............................. 169
Hirukawa K. ............................ 351
Hoagland R.............................. 323
Hoechbauer T. ......................... 323
Ibba A. ..................................... 201
Ichihara Y. ............................... 189
Ignatova O. N. ......................... 343
Issa M. E.................................. 419
Jiikov M.............................. 409
Jourani A. ................................ 369
Kabir M. Z. .............................. 705

Katayama T.............................. 159
Kibriya T.................................. 469
Klanek U. ....................... 605, 643
Kochuparampil J...................... 311
Kravanja S. .............. 605, 615, 643
Kubo S. .................................... 179
Kukielka K............................... 663
Kukielka L. .............................. 663
Laou E...................................... 301
Latteur P. ................................... 31
Le V. D. ................................... 211
Lecompte D. ............................ 239
Lee J. Y.................................... 121
Lee S. J. ................................... 121
Lentzen S. ................................ 653
Lin B. ....................................... 491
Lu G......................................... 491
Madamba M. C. ....................... 301
Madhkhan M............................ 269
Marchese S. ............................. 449
Martella P. ............................... 249
Martikka H............................... 593
Martn E..................................... 63
Martn Gutirrez E................... 571
Martinez M. ............................. 103
Martnez-Abella F.................... 401
Matsuoka T. ............................. 189
Merazig H. ............................... 311
Mestrovic D. ............................ 429
Mezghani S. ............................. 369
Mines R. A. W. ........................ 481
Minoshima K. .......................... 131
Misra A. ................................... 323
Mizuta K. ................................. 189
Mahonkov E. .................... 409
Mollaert M................................. 41
Mori H. .................................... 151
Muiz Gmez S. ...................... 571
Muntasar A. ..................... 301, 311
Mura I. ..................................... 695
Murakami T. ............................ 159
Nagy J. B. ................................ 449

Nakagawa H. ........................... 169
Nassira Y. ................................ 705
Nastasi M................................. 323
Nitzsche F................................ 103
Novikov S. A. .......................... 343
Ogawa K.................................. 169
Ohya S. ............................ 351, 359
Okubo K. ................................. 179
Ordua A. ................................ 673
Otero Chans D. ........................ 571
Otero D. ..................................... 63
Pardo Trfach P. ...................... 559
Peroni L. .................................. 249
Picart D.................................... 211
Ponsaert W........................... 13, 31
Premrov M....................... 111, 615
Prochzka P. ............................ 391
Prokopenko Yu. A. .................. 521
Pushkov V. A........................... 343
Radic J. .................................... 429
Rao J. N. .................................. 653
Rahai A. R. .............................. 705
Rebouillat S. ............................ 279
Rivera D................................... 439
Rovira Santa Olaya J. L. .......... 559
Rybinskaya A. A...................... 521
Ryu B. J. .................................. 121
Samyn Ph........................... 13, 635
Sano Y. .................................... 359
Santorinaios M......................... 481
Savochka P. A. ........................ 521
Schmidt R. ............................... 653
ejnoha M.................................. 83
Shin K. B. ................................ 121
Shorkin V................................. 581
ilih S. ..................................... 615
Sinitsyn V. A. .......................... 343
Smedley D. .............................. 539
Smits A. ................................... 239
Sol H. ............................... 225, 239

Steffenino B............................. 279
Stolarski T. A........................... 351
Stopp H. .................................. 529
Strangfeld P. ............................ 529
Sutcliffe C. J. ........................... 481
Swadener J. G. ......................... 323
Swartjes F. H. M...................... 511
Taitokari E. .............................. 593
Takagi H. ................................. 151
Takemura K. ............................ 143
Tanaka H.................................. 359
Tanaka K.......................... 131, 159
Tobe S...................................... 351
Tomioka Y............................... 169
Tsukrov I.................................. 685
Uchic M. .................................. 331
Uno K. ..................................... 159
717
Van Mele T................................ 41

Van Slycken J. ......................... 259
Van Steirteghem J................ 13, 31
Vandenbergh T. ......................... 31
Vantomme J............................. 239
Vzquez J. A.............................. 63
Vzquez-Herrero C.................. 401
Verbeeck B. ....................... 31, 635
Verleysen P.............................. 259
Wagner J. A. ............................ 549
Wang W................................... 511
Wang Y.-C............................... 323
Wastiels J................................... 21
Wuzella G.................................. 53
Yamada H................................ 131
Zahouani H. ............................. 369
Zeman J. .................................... 83
Zhornik V. A. .......................... 521
ula T. ..................................... 643
Structures Under Shock

and Impact IX
Edited by: N. JONES, The University of
Liverpool, UK, C. A. BREBBIA, Wessex
Institute of Technology, UK
The shock and impact behaviour of
structures is a difficult area, not only because
of its obvious time-dependent aspects, but
also because of the difficulties in specifying
the external dynamic loading characteristics
and in obtaining the full dynamic properties
of materials. This book examines the
interaction between blast pressure and
surface or underground structures, whether
the blast is from civilian, military, dust and
natural explosions, or any other source.
Including papers from the Ninth
International Conference on Structures
Under Shock and Impact, the book will be
of significant interest to engineers from
civil,
military,
nuclear,
offshore,
aeronautical, transportation and other
backgrounds. Featured topics include:
Impact and Blast Loading Characteristics;
Protection of Structures from Blast Loads;
Missile Penetration and Explosion; Air Craft
and Missile Crash Against High-rise
Buildings;
Seismic
Engineering
Applications; Energy Absorbing Issues;
Fluid Structure Interaction; Behaviour of
Structural Concrete; Behaviour of Steel
Structures; Structural Behaviour of
Composites; Material Response to High Rate
Loading; Structural Crashworthiness;
Impact
Biomechanics;
Structural
Serviceability under Impact Loading;
Microdynamics; Interaction between
Computational and Experimental Results;
Software for Shock and Impact.
WIT Transactions on The Built
Environment Volume 87
ISBN: 1-84564-175-2 2006 apx 500pp
apx 180.00/US$325.00/270.00
High Performance
Structures and Materials
III
Edited by: W. P. DE WILDE, Vrije
Universiteit Brussel, Belgium,
C. A. BREBBIA, Wessex Institute of
Technology, UK
Most high performance structures require
the development of a generation of new
materials, which can more easily resist a
range of external stimuli or react in a nonconventional
manner.
Featuring
authoritative papers from the Third
International Conference on High
Performance Structures and Materials, this
book addresses issues involving advanced
types of structures, particularly those based
on new concepts or new types of materials.
Particular emphasis is placed on intelligent
smart structures as well as the application
of computational methods to model, control
and manage these structures and materials.
The book covers topics such as: Damage
and Fracture Mechanics; Composite
Materials and Structures; Optimal Design;
Adhesion and Adhesives; Natural Fibre
Composites; Failure Criteria of FRP; Nonlinear Behaviour of FRP Structures; Material
Characterization; High Performance
Materials; High Performance Concretes;
Aerospace Structures; Reliability of
Structures; Ceramics in Engineering.
WIT Transactions on The Built Environment
Volume 85
ISBN: 1-84564-162-0 2006 apx 350pp
apx 129.00/US$233.00/193.50
Find us at
Save 10% when you order from our encrypted
ordering service on the web using your credit
card.
Dynamics in the Practice

of Structural Design
O. SIRCOVICH-SAAR, Labaton, Israel
This book is a practitioner-friendly approach
to dynamics on structural design, oriented
to facilitate understanding of complicated
issues without their elaborate mathematical
formulations.
While the chapters follow logically from one
another, each one deals independently with
a subject in structural dynamics; this
approach allows the engineer to go directly
to the topic of his or her interest at a given
moment.
Throughout each chapter the reader will find
the text set in two different forms, for
different levels of the topic in consideration,
which will enable him to postpone for a
second reading deeper explanations.
Conceived as practical support for engineers
whenever they want to review a subject
related to dynamics in the practice of
structural design, this book can be of great
help for students of engineering.
ISBN: 1-84564-161-2 2005 208pp
apx 66.00/US$118.00/99.00
Computational Mechanics
for Heritage Structures
B. LEF;HERIS, Technical University of
Crete, Greece, M. E. STAVROULAKI,
Technical University of Crete, Greece, A.
C. SAPOUNAKI, , Greece,
G. E. STAVROULAKIS, University of
Ioannina, Greece
For thousands of years people have built
great structures utilizing the mechanics of
beams, arches, columns and their
interactions. It is now possible to use new
computer technology to examine the great
structural achievements of the past and to
apply the lessons learnt from these to current

projects.
Reflecting the authors extensive experience,
and describing the results of restoration
projects they have worked on, this book
deals with applications of advanced
computational mechanics techniques in
structural analysis, strength rehabilitation and
aseismic design of monuments, historical
buildings and related structures. The results
are given with clear explanations so that civil
and structural engineers, architects and
archaeologists, and students of these
disciplines can understand how to evaluate
the structural worthiness of heritage
buildings without the use of difficult
mathematics.
The book is also accessible to specialists with
a less theoretical background such as
historians, sociologists and anthropologists
who wish to develop an awareness of the
significance of past human efforts.
Series: High Performance Structures and
Materials Vol 9
ISBN: 1-84564-034-9 2005
288pp+CD-ROM
apx 130.00/US$234.00/195.00
WIT eLibrary
Home of the Transactions of the Wessex
Institute, the WIT electronic-library provides
the international scientific community with
immediate and permanent access to
individual papers presented at WIT
conferences. Visitors to the WIT eLibrary
can freely browse and search abstracts of
all papers in the collection before
progressing to download their full text.
Visit the WIT eLibrary at
http://library.witpress.com
Computational Methods in
Materials
Characterisation
Editors: A.A. MAMMOLI, University of
New Mexico, USA and C.A. BREBBIA,
Papers from the first international
conference on this subject. Topics covered
include:
Parameter Identification;
Thermomechanical Behaviour; Damage
Mechanisms; Foams; Polymers; and
Interface Phenomena.
Materials, Vol 6
ISBN: 1-85312-988-7 2003 368pp
121.00/US$193.00/181.50
Evolutionary Algorithms
and Intelligent Tools in
Engineering Optimization
Edited by: W. ANNICCHIARICO,
Universidad Central de Venezuela,
Venezuela, J. PERIAUX, Pole Scientifique
Dassault Aviation, France, M.
CERROLAZA, Universidad Central de
Venezuela, Venezuela, G. WINTER,
Universidad de la Palmas de Gran
Canaria, Spain
Evolutionary Algorithms are very powerful
techniques used to find solutions to realworld search and optimization problems.
They are artificial intelligence techniques
which mimic nature according to Darwins
principal of the survival of the fittest in
order to explore and explode large search
space and find near-global optima.
In this book, a large spectrum of innovative
evolutionary and intelligence methods are
presented and used for solving various

application problems, name and among
others: Genetic Algorithms, Evolutions
Strategies or Constrained Optimization,
Genetic Programming, Sampling Methods
in Evolutionary Computation, Tabu
Methods, Metamodelassisted Evolutionary
Algorithms, Multi Objective Robust Design,
Hierarchical
Asynchronous
Parallel
Evolutionary Algorithms, Micro Distributed
Genetic
Algorithms
and
Hybrid
Optimization. The reader will easily access
the comprehensive methods described in the
fourteen chapters that work impressively
well on practical problems representative
of real engineering situations in the areas of
Control, Electronics, Civil Engineering,
Aerospace Engineering, Turbomachinery
and Medical Engineering and Energy.
This collective book, written by
internationally recognized experts in the
fields of evolutionary design optimization,
will therefore be of significant interest and
value to computer scientists, researchers and
post graduate students, and practicing senior
or young engineers involved in complex
design Optimization problems.
ISBN: 1-84564-038-1 2005 360pp
88.00/US$158.00/132.00
All prices correct at time of going to press but

subject to change.
WIT Press books are available through your
bookseller or direct from the publisher.
Impact Loading of
Lightweight Structures
Introduction to Motion
Based Design
Edited by: M. ALVES, University of Sao

Paulo, Brazil, N. JONES, The University
of Liverpool, UK
J.J. CONNOR and B.S.A. KLINK, Massachusetts Institute of Technology, USA
This book features contributions from the

International Conference on Impact Loading
of Lightweight Structures.
The topics covered are all relevant to the
behaviour of Lightweight Structures
subjected to various dynamic loads. Many
leading researchers, from all over the world,
contributed to this conference with articles
on material characterisation, structural
failure and crashworthiness, energy
absorbing
systems,
experimental
techniques, theoretical models and
numerical analysis, each providing
information on the design of modern,
lightweight structures and contributing to a
safer world.
ISBN: 1-84564-159-0 2005 624pp
218.00/US$349.00/327.00
Deployable Structures
Analysis and Design
C.J. GANTES, National Technical
University of Athens, Greece
Accessible to practicing structural engineers
and graduate students with no previous
knowledge of the field, this title formulates
and solves the complex engineering design
problems with which deployable structures
are associated. It also presents the issue of
design of snap-through type deployable
structures in an organized way which will be
of interest to more experienced readers.
Materials, Vol 2
ISBN: 1-85312-660-8 2001 384pp
132.00/US$198.00/198.00
an excellent introductory text to

response control in structural engineering
design.
JOURNAL OF STRUCTURAL ENGINEERING
Bridging the gap between conventional

structural design and the field of motion based
design, this book provides a basis for the
preliminary design of motion sensitive
structures. The authors present a systematic
outline of basic concepts, computational
procedures and behavioural issues for
structural motion control.
ISBN: 1-85312-454-0 1996 440pp
95.00/US$171.00/142.50
WIT Press is a major publisher of engineering

research. The company prides itself on producing
books by leading researchers and scientists at the
cutting edge of their specialities, thus enabling
readers to remain at the forefront of scientific
developments. Our list presently includes
monographs, edited volumes, books on disk, and
software in areas such as: Acoustics, Advanced
Computing, Architecture and Structures,
Biomedicine, Boundary Elements, Earthquake
Engineering, Environmental Engineering, Fluid
Mechanics, Fracture Mechanics, Heat Transfer,
Marine and Offshore Engineering and Transport
Engineering.
WIT Press
Ashurst Lodge, Ashurst, Southampton,
SO40 7AA, UK.
Tel: 44 (0) 238 029 3223
Fax: 44 (0) 238 029 2853
E-Mail: marketing@witpress.com

High Performance Structures and Materials III PDF

Hochgeladen von

Dokumentinformationen

Originalbeschreibung:

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

High Performance Structures and Materials III PDF

Hochgeladen von

Copyright:

Verfügbare Formate

High Performance

WIT Press publishes leading books in Science and Technology.

WIT Transactions on The Built Environment

R van der Heijden

THIRD INTERNATIONAL CONFERENCE ON

INTERNATIONAL SCIENTIFIC ADVISORY COMMITTEE

This page intentionally left blank

This page intentionally left blank

Always prepared to share his time, he headed up and motivated several

Micromechanical modeling of random or imperfect composites

Section 4: Material and mechanical characterisation

Study on fatigue and energy-dissipation properties of

Arch bridge made of reactive powder concrete

Bond repair of cracked beams

Numerical analysis of the process of trapezoidal thread rolling

This page intentionally left blank

This page intentionally left blank

High Performance Structures and Materials III

Conceptual design of lightweight structures:

The so-called morphological indicators (MI) were introduced by P. Samyn [1] in

4 High Performance Structures and Materials III

High Performance Structures and Materials III

the strength of the structural parts is controlled through the indicator of

6 High Performance Structures and Materials III

through the buckling indicator =

, introduced by Latteur [3], in

High Performance Structures and Materials III

Design for stiffness versus design for strength

If one wants to consume a minimum of material it is logical that he follows a

8 High Performance Structures and Materials III

more important: W versus , or strength versus stiffness. This

this parameter could be L / F , which in fact has been introduced by

High Performance Structures and Materials III

As will be shown by Vandenbergh et al. [12] in a subsequent paper, the

10 High Performance Structures and Materials III

High Performance Structures and Materials III

Samyn, P., Etude compare du volume et du dplacement de structures

12 High Performance Structures and Materials III

High Performance Structures and Materials III

Influence of dynamic loads on the optimum

Vrije Universiteit Brussel, Belgium

14 High Performance Structures and Materials III

Introduction to the theory of morphological indicators

Morphological Indicators are design tools allowing the optimisation of structures

to take into account the additional volume of

material due to buckling. is a global measure of the sensitivity of a

is introduced [1, 3, 4].

Indicator of the First Natural Frequency

2.1 Classification of loads

The external permanent loads F p

The self weight F0

High Performance Structures and Materials III

FP = F0 + F p , the sum of all permanent loads

FE = F p + Fl , the sum of all external loads

We can now express the resultant of all loads as follows:

16 High Performance Structures and Materials III

2.2 The first natural frequency of a SDOF system

The parameter indicates the stress level of the structure.

When we calculate the natural frequency we need a larger number of parameters

the ratio of co-vibrating load in SLS and total load in ULS ( z * ).

High Performance Structures and Materials III

2.3 Determination of the correction factor c cor