Development of a Progressive Failure Finite Element Analysis

For a Braided Composite Fuselage Frame

Daniel C. Hart

Thesis submitted to the Faculty of the

Virginia Polytechnic Institute and State University
in partial fulfillment of the requirements for the degree of

Master of Science
in
Aerospace Engineering

Dr. Eric Johnson, Chair

Dr. Richard Boitnott
Dr. Rakesh K. Kapania

July 2002
Blacksburg, Virginia

Keywords: Triaxial braid, J-section frame, postbuckling, delamination, progressive failure,

crashworthiness, fuselage frame

Development of a Progressive Failure Finite Element Analysis

For a Braided Composite Fuselage Frame
Daniel C. Hart
(ABSTRACT)
Short, J-section columns fabricated from a textile composite are tested in axial compression to study
the modes of failure with and without local buckling occuring.The textile preform architecture is a
2x2, 2-D triaxial braid with a yarn layup of [ 0 18k 64 6k ] 39.7% axial. The preform was resin
transfer molded with 3M PR500 epoxy resin. Finite element analyses (FEA) of the test specimens
are conducted to assess intra- and inter- laminar progressive failure models. These progressive
failure models are then implemented in a FEA of a circular fuselage frame of the same cross section
and material for which test data was available. This circular frame test article had a nominal radius
of 120 inches, a forty-eight degree included angle, and was subjected to a quasi-static, radially
inward load, which represented a crash type loading of the frame. The short column test specimens
were cut from some of the fuselage frames. The branched shell finite element model of the frame
included geometric nonlinearity and contact of the load platen of the testing machine with the frame.
Intralaminar progressive failure is based on a maximum in-plane stress failure criterion followed by
a moduli degradation scheme. Interlaminar progressive failure was implemented using an interface
finite element to model delamination initiation and the progression of delamination cracks. Inclusion
of both the intra- and inter- laminar progressive failure models in the FEA of the frame correlated
reasonably well with the load-displacement response from the test through several major failure
events.

Acknowledgments

First and foremost I would like to thank my parents, Dennis and Anita, my brother Justin, and
most importantly my wife Vanessa for their continued support in pursuit of my degrees. I
would also like to thank Dr. Eric Johnson, the committee chair, for his support throughout my
academic career and committee members Dr. Richard Boitnott and Dr. Rakesh K. Kapania.
While completing my graduate research I was aided by the Impact Dynamics Branch
at the NASA Langley Research Center in Hampton Virginia. I would like to thank Dr. Richard
Boitnott, who was not only on my committee but worked with me and helped orchestrate the
testing done at the facility. Also helping me complete the testing were the technicians of the
building. Operating the testing machine were George Palko, who likes to poke everyones
ribs, Nelson Seabolt, the West Virginia boy who keeps everyone in line, and Ricky Martin,
who puts up with those two on a daily basis. Helping with the data acquisition set up were Jim
Richardson, Ed Knode, and Sotiris Kellas all of whom I would like to thank for their
contributions. I would also like to thank Lisa Jones who was our NASA contact at the Impact
Dynamics Branch.

iii

Helping with the analysis I need to thank Vinay Goyal, Carlos Dvila, Navin Jaunky, and
Nicolas Chretien.
Thank you to everyone who helped me along the way, especially my close friends. Their
friendship throughout the years has allowed me to benefit from and enjoy my collegiate career. So
thank you, in no particular order, Todd Norell, Tony Reichel, Mark Nelson, Mike Henry, Kevin
Waclawicz, Trevor Wallace, Matt Short, Mike Elander, Henrik Pettersson, and Tony Zerante.
I must acknowledge the NASA grant numbers NAG-1-2309 & NAG-1-01123 for the
funding of this research effort, and both the Virginia Tech College of Engineering and NASA
LaRC for the use of their computing facilities.

iv

Table of Contents

Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
Chapter 1: Advancing The Analysis of Textile Composites For Crashworthy Structures
.......................................................................1
1.1 Braided Composites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Composite Structure and Crashworthy Design . . . . . . . . . . . . . . . . . . . . . 2
1.3 Previous Fuselage Frame Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Objective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Chapter 2: Material Properties and Cross Section Anomalies . . . . . . . . . . . . . . . . . . . 9
2.1 Triaxial Braid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1.1 Fiber . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.1.2 Epoxy Resin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2.2 Material Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2.3 Failure Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.4 Manufacturing Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.4.3 RTM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.4.4 Inner Flange . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.4.5 Outer Flange . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
Chapter 3: Short Column Compression Tests, Design and Results . . . . . . . . . . . . . . . 17
3.1 Nomenclature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.2 Preliminary Compression Testing and FEA . . . . . . . . . . . . . . . . . . . . . . . 18
3.3 Short Column Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.4 Specimen Numbering and Origin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.5 Instrumentation and Test Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.6 Testing Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.6.1 1.5-inch Specimen Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.6.2 Four-inch Specimen Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.6.3 Six-inch Specimen Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
Chapter 4: Finite Element Analysis of Short Column Compression Tests . . . . . . . . . 38
4.1 ABAQUS Shell Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.2 Finite Element Model Geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.3 Element Size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.4 Displacement Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.5 FEA Results and Test Data Comparison . . . . . . . . . . . . . . . . . . . . . . . . . . 43

vi

4.5.1 1.5-inch Column . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

4.5.2 Four-inch Column . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.5.3 Six-inch Column . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
Chapter 5: Frame Segment Finite Element Analysis and Comparison . . . . . . . . . . . . 56
5.1 Finite Element Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
5.2 Frame FEA Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
Chapter 6: Interface Element and Progressive Failure Analysis . . . . . . . . . . . . . . . . . 61
6.1 Failure Mechanisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
6.2 Intralaminar Failure and Degradation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
6.3 Delamination Initiation and Progression . . . . . . . . . . . . . . . . . . . . . . . . . . 64
6.4 FEA Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
6.4.1 Short Column Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
6.4.2 Frame Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
Chapter 7: Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
7.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
7.2 Finite Element Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
7.3 Short Column Test and FEA Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
7.4 Fuselage Frame FEA Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
7.5 Final Thoughts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Appendix A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
A.1 Strain Gage Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
vii

A.2 Specimen Strain Gage Patterns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88

a.2.1 1.5-inch Column . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
a.2.2 Four-inch Column . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
a.2.3 Six-inch Column . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
Appendix B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
B.1 Sample ABAQUS Input File For Six-inch Column Model . . . . . . . . . . . 91
b.1.1 hdmesh6.inp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
b.1.2 Material Reduction Scheme For Progressive Failure . . . . . . . . . . . 98
Vita . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

viii

List of Figures

Fig. 1.1

Sketch of the frame test apparatus [3]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

Fig. 1.2

J-section fuselage frame. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

Fig. 2.1

Geometric discontinuity in the web to outer flange junction created by the

manufacturing process. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

Fig. 3.1

General specimen nomenclature along with strain gage locations for the six-inch
specimens. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

Fig. 3.2

Two-inch short column specimen after failure. . . . . . . . . . . . . . . . . . . . . . . . . 19

Fig. 3.3

Short column specimens mounted in end fixtures with strain gages attached.
Column gage lengths are 6, 4, and 1.5 inches. . . . . . . . . . . . . . . . . . . . . . . . . . . 21

Fig. 3.4

Universal testing machine with six-inch specimen mounted between the cross head
and loading platen. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

Fig. 3.5

Specimen #1, 1.5-inch column after failure. . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

Fig. 3.6

Specimen #2 under load with local buckling of the front side outer flange visible.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

ix

Fig. 3.7

Specimen #2 after failure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

Fig. 3.8

Load versus displacement data for specimen #5. . . . . . . . . . . . . . . . . . . . . . . . 30

Fig. 3.9

LVDT data for specimen #5. Numbers indicate LVDT locations at the mid height
of the cross section shown on the right. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

Fig. 3.10 Strain response data for the outer flange of specimen #5. . . . . . . . . . . . . . . . . . 32
Fig. 3.11 Strain response of the axial filler material for specimen #5. . . . . . . . . . . . . . . . 33
Fig. 3.12 Strain response for the web and inner flange of specimen #5. . . . . . . . . . . . . . 34
Fig. 3.13 Specimen #3 showing local buckling of the outer flange. Noted are the approximate
wave crest and node locations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
Fig. 3.14 Six-inch specimen after failure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
Fig. 4.1

Comparison of the end shortening results using S4R and S4R5 shell elements. 40

Fig. 4.2

Geometry used to create short column models with the correct curvature. . . . . 41

Fig. 4.3

End shortening response for specimen #1, 1.5-inch column. . . . . . . . . . . . . . . 44

Fig. 4.4

Compressive strain response in the front side outer flange for specimen #1, 1.5-inch
column. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

Fig. 4.5

Compressive strain response in the inner flange of specimen #1, 1.5-inch column.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

Fig. 4.6

ABAQUS linear buckling results for a four-inch column. On the right is a front on
view of the outer flange, showing the symmetric buckling waves on the front and
back side outer flange. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

Fig. 4.7

Finite element model of six-inch column. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

Fig. 4.8

Load-shortening data for specimen #3 and FEA results. . . . . . . . . . . . . . . . . . . 50

Fig. 4.9

Axial strain response for in the front side outer flange of specimen #3 and FEA
strain response. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

Fig. 4.10 Strain response of the filler material at the web to outer flange junction and FEA
results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

Fig. 4.11 Axial strain response in the web and FEA results for the six-inch column. . . . 54
Fig. 4.12 Inner flange axial strain response and FEA results for the six-inch column. . . 55
Fig. 5.1

Refined shell element model of a 48 degree frame segment. . . . . . . . . . . . . . . 57

Fig. 5.2

Midspan displacement and force response plots from linear and nonlinear analysis
with and without contact compared to test data of frame B. . . . . . . . . . . . . . . . 59

Fig. 6.1

Upper (S+) and lower (S-) surfaces of the interface element. . . . . . . . . . . . . . . 66

Fig. 6.2

Representative traction-stretching curve for the springs . . . . . . . . . . . . . . . . . . 67

Fig. 6.3

Load versus displacement response for the six-inch column from the test and FEA
with progressive failure models. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

Fig. 6.4

Load displacement responses from analyses including intra- or inter- laminar

progressive failure models compared to test data. . . . . . . . . . . . . . . . . . . . . . . . 72

Fig. 6.5

Load versus displacement responses for the test and FEA including both the intraand inter- laminar progressive failure models. . . . . . . . . . . . . . . . . . . . . . . . . . 74

Fig. 6.6

Load displacement responses from the analysis including both the intra- and interlaminar progressive failure models and the test. The domain of the intralaminar
failure extended to web and inner flange. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

Fig. 6.7

Predicted element failures using the progressive failure model. . . . . . . . . . . . . 77

Fig. 6.8

Comparison of the element failures predicted by the progressive failure model and
failures that occurred on frame B. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

Fig A1.

Axial strain gage locations for the 1.5-inch specimens. All dimensions in inches.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88

Fig A2.

Axial strain gage locations for the four-inch specimens. All dimensions in inches.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

Fig A3.

Axial strain gage locations for the six-inch specimens. All dimensions in inches.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

xi

List of Tables

Table 1.1 Frame B Dimensions And Cross-Sectional Data . . . . . . . . . . . . . . . . . . . . . 6

Table 2.1 Tri-Axial Braid Properties For Vf = 55.26% . . . . . . . . . . . . . . . . . . . . . . . . 12
Table 3.1 Local Buckling Results For Short J-Section Columns From ABAQUS . . . 22
Table 3.2 Test Set Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Table 5.1 Responses From Several Analyses Compared to The Test of Frame B . . . 60
Table 6.1 Reduction Factors For Intralaminar Material Degradation . . . . . . . . . . . . . 65
Table 6.2 Interfacial Strengths [15] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

xii

CHAPTER 1

Advancing The Analysis of

Textile Composites For
Crashworthy Structures

Composite materials are presently used in multiple industries including aerospace, marine,
automotive, and trucking. Structural components such as frame rails, stiffeners, skins, and
bulkheads fabricated from advanced composites can be found in lightweight flight, space, and
land vehicles. The architecture of these composites range from unidirectional tapes to weaves
and braids. The most recent advancements have been in two-dimensional (2D) and threedimensional (3D) weaves and braids, which are termed textile composites due to the
fabrication process. Textile composite materials are attractive because of their strength to
weight ratio, energy absorbing characteristics, and the possibility of mass producing
specialized parts. The present research is a continuation of a previous study, which will be
reviewed in this chapter along with an overview of composites in structural design.

1.1 Braided Composites

Advancements in lightweight structures follows the development of new materials and new
fabrication processes. The subject material in this research is a textile composite consisting of
a preform of yarns, or tows, of graphite fibers interlaced in three in-plane directions, which is
subsequently impregnated with epoxy resin by a resin transfer molding (RTM) process. The

preform architecture for this research is a 2x2 2D triaxial braid, details of which will be discussed
in a subsequent chapter.
Textile composite materials have good energy absorbing characteristics because of their
multiaxial strengths, which is attractive for aircraft and automotive industries, to name a few.
Braided composites sacrifice some in-plane stiffness and strength to improve out-of-plane
strength and resistance to delamination with respect to tape laminates. This interlaminar strength
is derived from undulation of the yarns in the thickness direction and subsequent nesting of the
braided layers that occurs during the RTM manufacturing process. Nesting occurs during the
RTM process, pressure from the influx of resin forces layers to move around slightly adjusting the
orientation of some yarns. Yarns of coincident layers end up side by side in the matrix, thus
making a connection between the layers through the cured resin of the matrix.
Delamination in tape laminates is one reason they have poor energy absorbing
characteristics under lateral impact loading. Once delamination initiates there is nothing but the
brittle matrix to carry out-of-plane stresses, thus leading to separation of two layers and to
potential catastrophic failure. The braided architecture and fabrication leads to thickness direction
reinforcement and nesting as previously mentioned. Nesting offers an increased amount of
interlaminar interaction compared to unidirectional composites, which increases the strength in
the thickness direction. Energy absorbing characteristics of textile composites stems from the
material architecture and through-the-thickness strength [1].

1.2 Composite Structure and Crashworthy Design

In the event of a crash, and without specifically incorporating energy absorbing concepts into
vehicles structure, most of the impact energy is transmitted through the structure to the occupants

and payload. It follows that the need for understanding the deformation and progressive failure
process of these new materials when subjected to crash loads is crucial in designing safe and
useful structures. Weight sensitive structural designs with laminated composites incorporate thinwall construction, which under compressive load can buckle. Assuming some postbuckling load
carrying capacity, these thin-walled, laminated components may begin failing by material failures
within a lamina (matrix cracks, fiber rupture, fiber-matrix debonding, etc.), or by delamination of
adjacent lamina. Structural components having postbuckling strength, albeit at reduced stiffness,
may be viewed as beneficial with respect to energy absorption, since finite loads can be carried
over large displacements; crumple zones in an automotive structure is one example. If the design
is to carry service loads without buckling then preventative measures can be used, for example the
addition of material layers. The issue becomes how to design a composite structure to carry
service loads without failure and yet under crash loads to fail in such a manner so as to absorb as
much energy as possible. To accomplish the latter, we must first understand the failure
mechanisms and predict the progressive failure sequence.
Consider a survivable crash scenario of a transport aircraft on an airport runway. The force
acting on the passenger due to deceleration of the aircraft must be limited while the kinetic energy
due to impact is absorbed. This kinetic energy is absorbed by stroking of landing gear, crushing of
the fuselage sub-structure below the passenger deck, and proper seat design. For crushing of the
structure below the passenger deck, it has been shown from drop tower tests of fuselage sections
that the fuselage frames play a very important role in the response and energy absorbing
characteristics [2]. Also it was found, for the limited vertical drop speeds of a survivable crash,
that the dynamic and static failure sequences are not that much different. Hence, it is instructive to
understand the fundamental mechanics of the progressive failure sequence of a quasi-statically

loaded fuselage frame under a crash-type load. In particular, the focus of this study is the response
and progressive failure analysis of fuselage frames fabricated from a triaxial braided composite
material. Tests of these frames were conducted in Ref. [3] and these tests are discussed in the next
section.

1.3 Previous Fuselage Frame Tests

In a previous study[3], circular braided composite fuselage frames segments with J cross sections
were subjected to quasi-static, radial inward loading until several major failure events were
recorded. A sketch of the frame mounted in a universal testing machine is shown Figure 1.1 .

fixed cross head

platen
end block

braided frame
end block

I-beam

moveable table of the testing machine

Fig. 1.1 Sketch of the frame test apparatus [3].

Frame geometry is shown in Figure 1.2 , with dimensions of the frame denoted by B in Ref. 3
listed in Table 1.1. Response data from the tests were compared to corresponding results obtained
from a finite element analysis (FEA) of the frame using the ABAQUS/Standard [4] software
package. The open section curved beam element labeled B32OS in the ABAQUS library of

elements was used in the analysis. FEA results yielded an initial frame stiffness within 1% of test
measurements.

wo
to

A
ri

ro

ro

tw

ti

ri
wi

Section A A
Fig. 1.2 J-section fuselage frame.

Table 1.1 Frame B Dimensions And Cross-Sectional Dataa

Inner Flange Width ( w i )

1.2500

Web Thickness ( t w )

0.1590

Inner Flange Thickness ( t i )

0.2040

Opening angle ( )b

42

Outer Flange Width (w o)

2.7700

Cross Section Area in2

1.2168

Outer Flange Thickness ( t o ) 0.0885

Second Area Moment About 1-axis ( I 11 ) 3.9230 in4

Cross Section Height ( h )

4.8000

Second Area Moment About 2-axis ( I 22 ) 0.1914 in4

Radius of Inner Flange ( r i )

117.85

Torsion Constant ( J )

0.0104 in4

Height of Centroid From r i

2.4270

Radius of Outer Flange ( r o ) 122.65

a. Dimensions in inches unless otherwise noted.

b. For the frame mounted in the end blocks shown in Figure 1.1 .

The largest circumferential strains measured in the test were compressive and were
located at the center of the frame where the load was applied through contact with the platen of
the testing machine. However, compressive strains measured by electrical resistance strain gages
near the site of the first major failure event were about 40% lower than the theoretical
compressive failure strain of the material predicted by computer program Textile Composite
Analysis for Design, or TEXCAD [5].The deformation of the frame in the vicinity of the load
platen consisted of a local buckling and postbuckling response of the radially outboard flanges
and distortion of the cross section. Local postbuckling of the flanges reduces the direct
compressive load carried by the flanges at the expense of increased direct compression carried by
the junction between the flanges and web. The first major failure event in this local deformation
state can be due to a material compressive strength failure in the junction, or by delamination
initiating in the flange, or by a combination of both mechanisms. Delamination has been shown to

be the dominant failure mechanism in thin composite compression panels when the width to
thickness ratio, b/t, is between 10 and 20 [6]. The effective b/t ratio for the outer flange in the
cross section of frame B is 15.6.
Cross-sectional distortions and local buckling events cannot be modeled on the basis of
beam theory alone, since beam theory is based on the assumption that cross sections do not distort
in their own plane. Although the linear beam FEA can predict the initial structural stiffness, it
cannot be expected to account for cross-sectional distortions that precede the first major failure
event. To predict local buckling and postbuckling events in the vicinity of the load platen, we at
least need a branched shell FEA of frame B.

1.4 Objective
The overall objective of this research is to improve the analysis for the response of the forty-eight
degree frame segments tested in Ref. [3], and to predict the failure sequence. To capture the
distortion of the cross section and local buckling events observed in the tests, a geometrically
nonlinear, branched shell finite element model is implemented in ABAQUS/Standard, which
includes contact modeling of the platen with the frame. In order to accomplish the prediction of
the failure sequence, it was decided to conduct short column compression tests of the J-section
textile composite frame, since failure in the forty-eight degree frame tests initiated under
compression in the outer flange at the location of the applied load. These short column tests
provide further information on the failure mechanisms that occur in the as fabricated J-section
under direct compression and under local postbuckling. Data obtained from short column tests is
then used to improve the finite element analysis of failure of the forty-eight-degree textile
composite fuselage frame.

An overview of the textile material is given in Chapter 2. Here the material architecture
will be discussed along with strength properties, and the cross-sectional characteristics of the Jsection resulting from the manufacturing process is presented. Following this is the short column
test design and test results are presented in Chapter 3, with a discussion of the finite element
models and FEA correlation with test data in Chapter 4.
The FEA of the forty-eight degree frame segment is presented in Chapter 5. Here the
model is discussed along with the results and how they correlated with test data.
Discussion of the progressive failure analysis (PFA) is presented in Chapter 6. PFA
includes intra- and inter- laminar failure. Interlaminar failure is accomplished via delamination
and intralaminar failure is based on maximum values of the in-plane stresses followed by a
stiffness reduction scheme. A special interface finite element is used to model delamination.
Chapter 7 contains with a brief summary and concluding remarks.

CHAPTER 2

Material Properties and

Cross Section Anomalies

Architecture of the textile composite and geometry of the cross section are discussed in this
chapter.

2.1 Triaxial Braid

The preform of the 2x2 2D triaxial braid is produced on a braiding machine which consists of
yarn carriers that rotate around a cylindrical mandrel in circular paths, and this mandrel can
move in the axial direction. The rotational speed of the yarn carriers relative to the
translational speed of the mandrel controls orientation of these braider yarns. Fixed, straight
axial yarns are introduced at the center of the orbit of the yarns carriers, and the braider yarns
lock the axial yarns in the center of the fabric. A flat braided sheet is obtained by cutting the
cylindrical sheet from the mandrel and stretching it out flat. Triaxial braids offer strength in
three planer directions with braider yarns in symmetrical bias angles measured from the axial
yarn. The triaxial braid is called a quasi-laminar, or a 2D, composite since it has distinct

layers, which can be separated without breaking fibers prior to the resin transfer molding (RTM)
process. Modest volume fractions of through-the-thickness fibers provide delamination
resistance, but the majority of the fibers provide high in-plane stiffness and strength. The 2x2
pattern has two yarns following the same pattern, so in the braid there are two braider yarns
between cross over points [1]. Thickness is controlled by making multiple rounds about the
mandrel, essentially laying down nearly identical layers that have no mechanical connection.
Consolidation of the layers is achieved in the RTM process, which also causes some nesting of the
layers. During the RTM process pressure from the influx of resin can push the 2D braided layers
together causing yarns to intermingle creating a mechanical interaction between layers by locking
yarns from two coincident layers in the cured matrix.
Architecture definition and nomenclature for 2D triaxial braids lists bias braid angles, bias
and axial yarn size in thousands of fibers, and overall percentage of axial yarns in a unit cell. A
unit cell is defined as the smallest grouping of yarns that can be copied to make up the periodic
pattern. A 2x2 unit cell has a set of 2 + and 2 - bias yarns intersecting with an axial yarn on
either side of the intersection. For example, the architecture for the specimens of this project is
[018K/ 646K] 39% axial, which states that nominal bias yarns are at an angle of 64 measured
from the axial yarns and they contain 6,000 fibers per yarn and there are 18,000 fibers per axial
yarn and 39% of the unit cell volume is made up of axial yarns.
2.1.1 Fiber

The yarns consist of AS4 Gr/Ep fibers combined into tows of 18,000 and 6,000 fibers. Dry tows
are used, which means that there is no surface treatment of the fibers during the manufacturing

10

process. Individual tows of 6000 fibers have a modulus of elasticity equal to 3.31 10 psi and an
5

ultimate tensile strength of 5.9 10 psi.[7]

2.1.2 Epoxy Resin

During the RTM process 3Ms PR500 epoxy resin is introduced into the mold using a vacuum.
5

The resin has the following properties, modulus of elasticity of 5.07 10 psi, ultimate tensile
strength of 8,300 psi, and an ultimate flexure strength of 18,450 psi.[8]

2.2 Material Properties

The computer program Textile Composite Analysis for Design [5], or TEXCAD, was used to
predict material properties, which were partially verified with tensile coupon tests [9]. Other
properties include a yarn packing density of 0.75 and axial yarn spacing of 0.2087 in. (5.3mm)
TEXCAD input includes yarn and resin information, such as axial and bias yarn angles, number
of filaments in the yarns, and the dimensions of unit cell and yarns. Using micro-mechanical
models TEXCAD predicts the Young moduli, shear moduli, Poissons ratios, and failure
strengths. The TEXCAD predictions and tension test results are listed in Table 2.1 for a fiber
volume fraction V f = 55.26 %. This fiber volume fraction was determined by the density method
from coupons cut from the web of one the frames[9].

11

Table 2.1 Tri-Axial Braid Properties For Vf = 55.26%

Properties
Axial Modulus (E11,psi)

TEXCAD[5]

Transverse Modulus (E22,psi)

6.59 10

Through Thickness Modulus (E33,psi)

Poissons Ratio ( 12 )

1.53 10
0.231

0.26

Poissons Ratio ( 13 )

0.216

N/A

Poissons Ratio ( 23 )

0.298

N/A

In Plane Shear Modulus (G12,psi)

1.91 10

Transverse Shear Modulus (G13,psi)

0.601 10

Transverse Shear Modulus (G23, psi)

Maximum Compressive Axial Stress (XC, psi)

0.645 10
71,000

N/A

Maximum Tensile Axial Stress (XT, psi)

91,370

76,880

Maximum Compressive Transverse Stress (YC, psi)

56,890

N/A

Maximum Tensile Transverse Stress (YT, psi)

73,140

N/A

Maximum In-Plane Shear Stress (SC, psi)

30,460

N/A

Tensile Failure Strain( 1t T, )

14071

10588

Compression Failure Strain ( 1c , )

10108

N/A

7.06 10

Tension Tests[3]
6

7.09 10
N/A

N/A

N/A

N/A

N/A

2.3 Failure Criteria

Failure criteria specifically developed for triaxial braided materials are less common than criteria
developed for tape laminates due to the complex architecture of the triaxial braid. With undulating

12

braider yarns, and to some extent undulating axial yarns due the compaction of the RTM process,
the micromechanical analysis required to model the stresses and predict failure modes is difficult.
Approaches to predict failure in textile composites with periodic geometry are based on
unit cells. Spatially translated copies of a unit cell construct the entire textile composite. These
approaches for failure prediction can be based on failure loci for individual yarns and matrix in
terms of the local stress state in a unit cell, or be based on in-plane stresses averaged over the
volume of the unit cell. The micromechanical approach is used in computer program TEXCAD to
predict failure of a unit cell based on constituent stresses obtained from the isostrain assumption.
Several features in TEXCAD aid in estimating failure [1]. These features include: bending of the
undulation yarns modeled as the response of a curved beam on an elastic foundation, which
includes the effect of yarn splitting; the nonlinear shear response of impregnated yarns
represented by a power law relation; first order effects of geometric nonlinearity due to yarn
straightening or wrinkling; and, there is a stiffness reduction algorithm that is applied when local
damage is detected. As a result of these features in TEXCAD, fiber dominated failure of the yarns
is predicted using a maximum stress criterion for both tension and compression axial yarn stresses
(11). Matrix dominated failure within the yarns is predicted using a maximum stress criteria for
each fiber dominated failure mode: transverse tension (22, 33), transverse shear (23), and
longitudinal shear (12, 13). Interstitial matrix material failure is predicted by using two failure
criteria: a principal stress criterion (used in the absence of applied shear stresses), and a maximum
octahedral shear stress criterion (used in the presence of shear stresses). Composite failure is
predicted when (i) axial yarn failure is detected anywhere in the RUC, or (ii) all yarns fail in the
same failure mode and failure is detected in the interstitial matrix material [10]

13

Techniques designed for laminated composites involve a combination of failure criterion.

One of the most commonly used criterion accounts for four typical compressive failure modes in
unidirectional composites, matrix failure in tension or compression, fiber matrix shear failure,
fiber buckling, and delamination [11]. In comparison to textile composite failure, specifically
triaxially braided composites, the major failures are shear micro cracking between fiber bundles,
local delamination between plies under compression, fiber kink bands forming under
compression, and micro cracking around the bias rovings under axial tension [1]. Complex
geometry and fabrication anomalies make determination of failure types difficult.
The outputs from TEXCAD include the material properties of the given composite along
with the failure strengths. These strengths can be incorporated in a simple in-plane maximum
stress failure criterion which will be discussed subsequently.

2.4 Manufacturing Process

Complex material architecture makes material property prediction difficult. Along with the
variables associated with material architecture, damage associated with the manufacturing
process occurs, damage happens during braiding where yarns may become over tensioned or tows
may be frayed. During the curing process often times yarns become crimped or misaligned
decreasing the strength at random locations in the part. Complexity of the part and the technique
used to manufacture the shape also plays a roll in the strength properties of the structure.

14

2.4.3 RTM

Resin transfer molding is a common manufacturing technique that uses a female mold to create
the hard form of the finished product and a vacuum process to pull the resin into the molded
preform and pull all gases out of the mold, creating a part with very few defects.
2.4.4 Inner Flange

The number of braided layers in the web and inner flange is the same, but, as can be seen from the
thicknesses listed in Table 1.1, the thickness of the radially inboard flange is greater than the
thickness of the web. Hence, because of the manufacturing process there is a reduced fiber
volume fraction in the radially inner flange. Reduction of the fiber volume fraction is due to an
increase in the volume of resin in the inner flange. The fiber volume fraction in the inner flange,
( V f ) i , was approximated by ( V f ) i = ( t w V f w ) t i = 0.4307 , here t w is the thickness of the web,
t i is the inner flange thickness, and V fw is the fiber volume fraction in the web. According to
Naik[12], the relationship between the fiber volume fraction and the moduli are approximately
linear. These data from Ref. 6 are for a material with an architecture of [018k/ 67.46k], yarn
packing density of 0.75, and axial yarn spacing of 0.2094 in. (5.32 mm). Other than the 3.4
difference in the braider angle this architecture matches that of the frames. Using slopes measured
6

from the plots in [12], the reduced moduli for the inner flange are E 11 = 5.83 10 psi and
6

E 22 = 5.09 10 psi. The variation in the shear moduli with fiber volume fraction is negligible,
so it and other property data for the inner flange are as listed for the web and outer flange in
Table 2.1.

15

2.4.5 Outer Flange

Production of the outer flange is done by dividing the layers of the web into two and folding them
over [13]. Eight layers of the web are split into a set of four layers, when folded over a void is
created due to the curvature of the folding process. At the point of splitting of the fabric, a Vshaped void is created at the web-flange junction. This void was filled with a braided strand of
axial fibers run circumferentially and consolidated with the frame in the resin transfer molding
process as shown in Figure 2.1.
Braided filler strand

Matrix

Braided
fabric

Fig. 2.1 Geometric discontinuity in the web to outer flange

junction created by the manufacturing process.

16

CHAPTER 3

Short Column
Compression Tests,
Design and Results

Quasi-static compression tests of short columns allows for the study of failure mechanisms
associated with the cross sectional shape of the fuselage frames. The goals of these tests are to
study the mechanisms of failure and to measure the compressive strength of the section with
and without local buckling occurring during the response. Fabrication details can affect the
compression strength. For example, one of the failure events observed in the full frame tests
was the separation from the frame of a ligament of the axial filler strand contained in the
junction of the outer flange and web [9]. In this chapter the design of the specimens and the
results from the testing will be discussed.

3.1 Nomenclature
While discussing the results of testing, some nomenclature specific to the testing
configuration and cross section is helpful. In Figure 3.1 we define branches of the cross
section and give nominal locations of axial strain gages bonded to the six-inch columns.

17

Because of the frame curvature, the terms inner and outer refer to radial distances from the center
of curvature.
Back Side Outer Flange

Inner
Flange

Back Side

10 9
4
3

1 2
Front Side

6 5 11

11
8 6 5 13

1 2
z

10 9 12

8 13

Front Side Outer Flange

x

Top View

View From Front Side

Fig. 3.1 General specimen nomenclature along with strain gage

locations for the six-inch specimens.

3.2 Preliminary Compression Testing and FEA

Flat-end compression tests of unsupported two-inch columns, cut from the undamaged portions of
frame B, were loaded in compression until failure. Loads and end displacements were recorded in
a series of four tests with the fourth test specimen fitted with two sets of back-to-back axial strain
gages on the web and inner flange. Brooming of the column end initiated failure in the inner
flange, which progressed through the web to the outer flange as shown in Figure 3.2.

18

Progression of failure followed green arrow.

Brooming began here, inner flange.

Fig. 3.2 Two-inch short column specimen after failure.

A finite element model was developed using S4R5 shell elements that included the frame
curvature, 3-D orthotropic material properties, and reduced moduli for the inner flange due to a
decrease in the fiber volume fraction. The buckling load determined from the finite element
analysis was within 2% of the maximum compressive load in the test of the two-inch-long
column. The correlation of the buckling load and the failure load may indicate that local bucking
occurred nearly simultaneously with end brooming failure.

3.3 Short Column Design

Because a mixture of failure mechanisms were exhibited by the unsupported two-inch columns a
second series of tests with better end supports were designed. Buckling analyses of a linear
prebuckling equilibrium state under specified end shortening were undertaken on shell element
models of various column heights. The axial displacements at buckling were then specified in a

19

separate equilibrium analysis to determine the load at buckling by summing up the nodal reaction
forces predicted by the equilibrium analysis.
Approximate failure displacements for the column heights were determine by the
following simple column strain-displacement relationship.
failure l = l

(3.1)

In Equation 3.1 failure is the compressive failure strain predicted by TEXCAD as listed
in Table 2.1, l is the column height, and l is the end-shortening displacement for the
compressive strength failure mode of the material.
When the displacement associated with material failure is less than the displacement at
buckling, a compressive material failure should be observed during testing. By comparing the
displacement buckling results and the displacement for material failure determine by Equation 3.1
a specimen height was chosen for compressive material failure. Analysis indicated that a longer
specimen was likely to be dominated by local buckling, so a reasonable column height was
chosen. To observe a mix of failure mechanisms a height between the material failure and
buckling failure was decided upon. Column heights chosen were 1.5, 4, and 6 inches, these
correspond respectively to compressive material failure, mixed failure modes, and a local
buckling failure. Specimen ends were supported in an epoxy potting mixture to prevent end
brooming and back-to-back strain gages were bonded to the flanges of the specimens in a pattern
to capture local bucking deformations. Axial gages were bonded near the junction of the web and
outer flange to capture the increase in compressive strains at the junction during postbuckling.

20

Three of the short column test specimens, one of each length, mounted in the end fixtures with
strain gages attached, are shown in the photograph of Figure 3.3.

Fig. 3.3 Short column specimens mounted in end fixtures with

strain gages attached. Column gage lengths are 6, 4, and
1.5 inches.

Buckling displacements and loads for the chosen column heights are listed in Table 3.1.
For the material failure mode, the end shortening for the1.5-inch gage length specimen was 0.015
inches, which is less than the FEA predicted local buckling displacement of 0.019 inches. There
were two distinct results calculated by ABAQUS for the six-inch specimen, the results are listed
in the last two rows of Table 3.1. When the column was modeled with only the six inch gage
length, the buckling prediction is a sine wave in the radially outer flange. If the column is
modeled with the extra height to include the potting, the buckling shape becomes three half waves
with the central wave having the largest amplitude bordered by two smaller half waves of

21

opposite amplitude. The buckling shapes are negatives of each other for the front and back side of
the outer flange.

Table 3.1 Local Buckling Results For Short J-Section Columns From ABAQUS
Gage Length (in)

At Buckling

1.5
4

End-shortening (in)
0.0191
0.0168

Compressive Force (lbs)

109142
39074

0.0235

34084

6a

0.0241

33491

a. Model includes extra length of the column potted in the end fixture.

3.4 Specimen Numbering and Origin

There were three replicates for each height chosen for a total of nine specimens. When these
specimens were cut from the fuselage frames an extra two and a half inches of length was added
to the gage length to account for the potting and machining the specimens flat and parallel. Short
column specimens were cut from undamaged portions of previously tested fuselage frames A and
C [9]. There were three testing sets representing the three heights. Short column data for each
testing set represented two frames; for example, one 8.5 inch section was cut from frame A and
two specimens were cut from frame C to make up the six-inch gage length test set. Complete test
set information is listed in Table 3.2. Frame A was used for material coupons in the previous
frame tests and therefore was not subjected to damage, while frame C was previously loaded
through multiple failures. Specimens were extracted from portions of frame C as far from visible

22

damage zones as possible. It was assumed that damage was detectable by visible inspection of the
external surfaces, and that no internal damage was present in the sections deemed to be
undamaged. Specimen #3 cut from frame C was about eight inches away from a damage zone,
and #4 was about 4.5 inches away from a damage zone.
Table 3.2 Test Set Information
Test Set
1.5 gage length
4 gage length
6 gage length

Specimen Origin and Numbers

Frame A
7
8
9

Frame C
1&6
2&5
3&4

The production quality of frame A is suspect, with matrix rich and deprived regions,
surface voids, and misaligned braids and fiber tows. Since these fabrication imperfections are
likely to be present in specimens 7, 8, and 9 cut from frame A, and since the measured failure
loads were consistently around 5,000 lbs lower than failure loads of specimens cut from frame C,
it was determined that the test data for these specimens was not representative of the frame B
material. Hence, the data from tests of specimens 7, 8, and 9 will not be studied further.

3.5 Instrumentation and Test Procedure

Specimens were fitted with up to thirteen electrical resistance strain gages oriented in the axial
direction. Back-to-back gages were bonded to the flanges of the specimens in a pattern to capture
local bucking deformations. Axial gages were bonded near the junction of the web and outer
flange to capture the increase in compressive strains at the junction during postbuckling. The six-

23

inch specimens have thirteen axial gages and the 1.5- and four-inch specimens have nine gages.
Detailed figures of strain gage locations are located in the Appendix A.
The short columns were loaded quasi-statically in compression using a 120 kip universal
testing machine at the NASA Langley Research Center. A photograph of a specimen mounted in
the test frame is shown in Figure 3.4. Specimen response was measured from strain gages and
linear variable displacement transducers (LVDT), and also recorded by digital video.
To insure uniform load distribution and alignment, the specimens were initially loaded
slowly to a level of 10 kips and the strain data was monitored. If non-uniform axial strains
occurred, then shims were inserted under the specimen to achieve a uniform initial compressive
strain. When the initial strain data indicated uniformity, the specimen was loaded until failure.
The testing machine was operated under load control using hydraulic pressure at an approximate
loading rate of 500-1000 lbs. per second. Loading was controlled manually, using fine and coarse
control valves in the hydraulic system. The cross head of the machine is positioned by two large
threaded columns, and these columns are fixed in part to the machine through two large springs.
For the large loads encountered at failure in these tests, the sudden increase in compliance of the
test specimen can cause a momentary dynamic loading condition possibly related to the release of
spring energy. This dynamic response of the load frame was noticeable in the responses
monitored by the instrumentation, and consisted of large measured displacement changes with
negligible amounts of load change.

24

Fig. 3.4 Universal testing machine with six-inch specimen

mounted between the cross head and loading platen.

25

3.6 Testing Results

3.6.1 1.5-inch Specimen Results

Failure of the short columns occurred rapidly with little visual and only minimal audible
evidence. Prior to failure popping noises were audible, and small localized surface deformations
were seen in the web. Catastrophic failure of the specimen was confined to a small length along
the load axis in the gage section near one of the potted ends, and extended over the entire cross
section. In this failure zone, matrix fractures followed the bias yarns, plies delaminated, yarns
fractured, and in the web fabric appeared to broom out of plane. Interior layers of the inner flange
crushed and pushed outward causing the surface layers to delaminate.
The 1.5-inch column failed between 63 and 64 kips. Load-displacement response
remained approximately linear throughout the test. No buckling was evident, and the mode of
failure appears to be dominated by material compressive strength.In the video, the inner flange
fails just prior to specimen failure. Failure was dominated by localized crushing and delamination
as seen in Figure 3.5. Localized matrix failures were seen on the surface in 3-4 locations when
loading exceeded 45 kips

26

Fig. 3.5 Specimen #1, 1.5-inch column after failure.

.Strain gage locations on the 1.5-inch specimen are similar to the six-inch specimen, gage
pairs 1and 2 and 3 and 4 are positioned back to back on the inner flange and web respectively. On
the outer flange, pairs 5 and 6 are on front side with 7 and 8 on the back side. Gages 6 and 8
mounted on the web side of the outer flange. Measuring the strain on the outer flange surface over
the junction is gage 9.
3.6.2 Four-inch Specimen Results

Failure responses for the four-inch columns varied between the specimens. They exhibited
combined failure modes consisting of local buckling of the outer flange and some compressive
material failure of the inner flange. Snap-through occurred in the outer flange of specimen #5
during its local postbuckling response, but did not occur in specimen #2. However, in specimen
#2 the front and back side outer flanges exhibited unsymmetrical bending modes.

27

Local buckling of the front side outer flange occurred at a load level of around 38 kips for
both specimens. The buckling mode shape is slightly unsymmetrical and is shown in Figure 3.6.
The crest of the buckling wave appears to be just above the mid height of the gage length. From
video of the test, the back side outer flange buckles in the same shape but with its wave crest just
below the mid height. Localized material failures near the web to outer flange junction coincide
with changes in the flange displacement and shape. As observed from video of the tests, popping
is associated with increased bending of the outer flange. Failure in the outer flange consisted of
cracking and delamination. Prior to cross section failure there is failure and separation of the axial
strand of filler material at the outer flange junction. This leads to crack propagation across the
outer flange. The failure location coincides with a node in buckling wave for specimen #5, and at
a wave crest for specimen #2, as is shown in Figure 3.7.

28

Mid Height
Wave Crest

Fig. 3.6 Specimen #2 under load with local buckling of the front
side outer flange visible.

Fig. 3.7 Specimen #2 after failure.

29

Failure loads were specimen dependent, and occurred at 57 kips and 53 kips for specimens
2 and 5, respectively. Except for some initial take up, and within a few thousand pounds of failure,
loading was very linear with respect to the end shortening displacement throughout the test, as is
shown in Figure 3.8. Near the initiation of local buckling, localized material failures can be seen
and heard as small distortions in the material surface and audible pops. As the load is increased
from initial buckling, the bending deformations in the outer flanges change shape from
unsymmetrical waves to a symmetric half wave. The unsymmetrical buckling shape may be
associated with the slight curvature of the specimen or with uneven support at the gage length
edges. Because the buckling shape was consistent between specimens, material imperfections in
the specimen do not appear to govern the buckling mode shapes.
60000

50000

Load (lbs)

40000

30000

20000

10000

0
0.00

0.01

0.02

0.03

0.04

0.05

0.06

Displacement (in)

Fig. 3.8 Load versus displacement data for specimen #5.

30

The snap through occurring in specimen #5 is evident in both the LVDT displacement data
shown in Figure 3.9, and in the strain gage data shown in Figure 3.10. Data from LVDT # 2 and
0.1

3
LVDT 1

0.08

Flange Displacement (in)

LVDT 2
0.06
LVDT 3

0.04

0.02
0
-0.02

1
-0.04
-0.06
-0.08
-0.1
0.00

1
0.01

0.02 0.03 0.04 0.05

End Displacement (in)

0.06

Fig. 3.9 LVDT data for specimen #5. Numbers indicate LVDT
locations at the mid height of the cross section shown on
the right.
LVDT # 3 in Figure 3.9, indicate that both sides of the outer flange are displacing radially outward
until snap through, and then the front side displaces toward the inner flange. On the outer flange
back-to-back strain gage pairs 5 and 6 are on front side with back-to-back gages 7 and 8 on the
back side. The axial strain reversal of gages 5 and 6 near the applied displacement of 0.050 inches
shown in Figure 3.10 is evidence of the snap through event of the front side flange. Snap through
was not observed with specimen #2

31

.
1.4
Gage 5

1.2

Gage 6

Compressive Strain %

Gage 7

Gage 8
8

0.8
0.6
0.4
0.2

0
-0.2
6

-0.4
0.00

0.01

0.02

0.03

0.04

0.05

0.06

Displacement (in)

Fig. 3.10 Strain response data for the outer flange of specimen
#5.
Axial strain measured on the surface of the filler material at the junction is linear with
respect to the end-shortening until snap through occurs at an approximate displacement of 0.05
inches. See Figure 3.11. When the bending response of the front outer flange changes direction
due to snap through, there is an abrupt increase in compressive strain in the filler material, which
precipitates its separation from the junction.

32

0.9

Compressive Strain %

0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0.00

0.01

0.02

0.03

0.04

0.05

0.06

Displacement (in)

Fig. 3.11 Strain response of the axial filler material for specimen
#5.
Inner flange and web responses are essentially purely compressive until snap through of
the front side outer flange occurs, as indicated by the responses of strain gages 1-4 in Figure 3.12.
Back-to-back gages 1 and 2 are bonded to the surfaces of the inner flange, and back-to-back gages
3 and 4 are bonded to the surfaces at center of the web. When the bending response of the front
outer flange changes direction due to snap through, the amount of bending strain carried by the
web increases drastically as is shown by the responses from gages 3 and 4. The loss of rotational
restraint provided by outer flanges to the web due to snap through, results in the dramatic increase
in web bending. Gage 4 is on the back side of the web and is clearly going into tension in Figure
3.12.

33

1.2
1

Gage 1
Gage 2

Compressive Strain %

Gage 3
0.8

Gage 4

0.6
0.4
2
0.2
0
-0.2
-0.4
0.00

0.01

0.02

0.03

0.04

0.05

0.06

Displacement (in)

Fig. 3.12 Strain response for the web and inner flange of
specimen #5.

3.6.3 Six-inch Specimen Results

The failure response of the six-inch column specimens were fairly uniform. Buckling initiated in
the six-inch columns around 33-35 kips and the columns failed around 50 kips. The loaddisplacement response became slightly non linear near the buckling load predicted by FEA.
Buckling was predicted to occur around 34 kips, and during testing buckling of the outer flange
began between 33 and 35 kips. See Figure 3.1 for the nomenclature used to describe the crosssectional configuration. The out-of-plane deformation of the front side outer flange changed from

34

a single half wave to three half waves along the load axis. The larger amplitude wave in the
middle of the flange deflected toward the inner flange, and the two adjacent smaller amplitude
waves deflected in the opposite sense as shown by the photograph in Figure 3.13. The out-ofplane displacement of the outer flange on the back side was a single half wave displacing away
from the inner flange, with the amplitude of this back side flange smaller than the amplitude of the
front side flange. The out-of-plane displacement of the web was toward the back side of the
section in a half wave along the load axis and a half wave perpendicular to this axis.
In the postbuckling response and prior to ultimate failure, both sides of the matrix,
surrounding the filler material at the junction of the web and outer flange, fracture along the load
axis. Subsequently the axial filler material failed and separated from the outer flange. Failure
progressed across the outer flange to the edges and through the web to the inner flange. The outer
flange of specimen #3 failed near the node line location of the local buckling mode as shown in
the photograph in both Figure 3.13 and Figure 3.14. However, specimen #4 failed near the
location of the large amplitude buckle crest in the outer flange.

35

Wave Crest

Node

Fig. 3.13 Specimen #3 showing local buckling of the outer

flange. Noted are the approximate wave crest and
node locations.

36

Fig. 3.14 Six-inch specimen after failure.

Failure response data for the short columns will be shown in comparison to FEA results in
a subsequent chapter.

37

CHAPTER 4

Finite Element Analysis

of Short Column
Compression Tests

Finite element modeling for the compression response of the short column test specimens is
discussed. The finite element analyses include geometric nonlinearity, but the material law is
assumed to remain linear elastic.

4.1 ABAQUS Shell Elements

Two shell elements from the ABAQUS library were considered; these are elements S4R5 and
S4R [14]. The S4R5 element is a quadrilateral, small strain, thin shell element. Strain
measures are suitable for large rotations but small strains. There are five degrees of freedom at
each of the four nodes; the three displacements of the reference surface, and the two
projections of the change in the unit normal along orthogonal directions tangent to the
reference surface that measure transverse shear strains. To model classical shell theory,
Kirchhoffs constraints are imposed numerically at a set of reduced integration points on the
reference surface; i.e., a discrete Kirchhoff constraint. The transverse shear stiffness acts as a

38

penalty that enforces the constraint of vanishing transverse shear strains at the set of reduced
integration points. The element formulation accepts the drilling degree of freedom, so that there
can be six degrees of freedom at a node, if six degrees of freedom are specified by the user at that
node, or if the node is in contact with an element containing six degrees of freedom per node.
When six degrees of freedom is indicated at a node, the extra rotational degree of freedom is
constrained locally by applying a small stiffness penalty to a strain measure representative of the
same rotation of the shells reference surface.
Element S4R is a quadrilateral, finite-membrane-strain, shell element for use with both
thin and thick shells. It has six degrees of freedom at each node of the four nodes. The shell theory
used depends on the thickness of the element. Thickness change as a function of the in-plane
deformation is allowed. For thick shells a first-order shear deformation theory is used. For thin
shells a discrete Kirchhoff constraint is imposed to reduce the transverse shear deformation.
A comparison of the S4R5 and S4R shell elements was made using the refined mesh
model of the six-inch short column, which will be discussed in more detail in Section 4.5.3 on
page 48. The load end-shortening response plot from the FEA of this specimen is shown in
Figure 4.1. The S4R element predicts a stiffer response in the nonlinear region of the response
plot than the S4R5 element.
Initial analyses of the short column specimens employed the S4R5 element. To improve
the correlation with the six-inch short column tests, element S4R was used as the mesh was
refined. As is depicted in Figure 4.1, the analysis with S4R resulted in a stiffer postbuckling
response, which correlated better with the test data. Also, the six degrees of freedom per node for
element S4R facilitated the use of the decohesion element to model delamination. The decohesion
39

element was implemented in the user-written subroutine UEL of ABAQUS to model progressive
failure, which is discussed in Chapter 6. The S4R element was used in the final analyses where
the element sizes were on the order of a sixteenth of an inch.
50000

S4R

Load (lbs)

S4R5

35000

20000

5000
0.00

0.02

0.04

0.06

Displacement (in)

Fig. 4.1 Comparison of the end shortening results using S4R

and S4R5 shell elements.

4.2 Finite Element Model Geometry

Cross-sectional dimensions used for finite element modeling are those listed in Table 1.1 for
frame B, which are taken from Ref. [9]. Using the radius of the inner flange, r i , and the radius of
the outer flange, r 0 from this table combined with the desired short column height H , the half

40

angle of the circular arc of the inner flange, a i , and the half angle of the arc of the outside flange,
a 0 , are determined. The basic geometry is specified in the finite element model using r i, a i, r 0, a 0
as shown in Figure 4.2. Two cylindrical surfaces representing the flanges were created with radii

ri

ai

H2

Mid Plane
Origin

r0

a0

H2

Fig. 4.2 Geometry used to create short column models with the
correct curvature.
r i and r 0 . The area bounded by radii r i , r 0 , and the vertical height H represented the flat surface
of the web. The three surfaces were meshed with quadrilateral elements, and then material
properties were assigned to the elements sets constituting these surfaces.

4.3 Element Size

The element mesh was refined to accurately represent the local buckling witnessed in the front
side outer flange during testing. To capture the buckling deformation, it is usually required to
have about five nodes per half wave. The original element size was on the order of a half an inch,
which was then refined to a quarter of an inch to get improved results. Further refinement was
done to the elements using biased meshing techniques. With the bias meshing tool the initial

41

element size and the final element size can be selected within a subdomain, then the meshing tool
automatically determines the number and varying element sizes to go from the initial to final
element size. Element sizes ranged from a sixteenth of an inch to a quarter of an inch in the
meshes used for the analyses.

4.4 Displacement Control

Loading was specified by controlled shortening. The node set making up the top edge of the
model was displaced in the negative 3-direction, corresponding to axial compression, and
restrained in all other degrees of freedom. The nodes making up the bottom of the column were
constrained in all degrees of freedom; i.e., clamped.
Both geometrically linear and nonlinear elastic analyses were performed. Linear analyses
were used to see if model fidelity was sufficient to match the initial column stiffness measured in
the tests, but linear analysis is not capable of predicting the shortening at buckling. To determine
the postbuckling load-shortening response, nonlinear analyses use an iterative solution procedure
within each incremental load step. Assuming equilibrium was established at the last load step,
Newtons method is used to converge to a solution for the next load increment. Load step
increments are automatically determined by ABAQUS. An example of an ABAQUS input deck
for a nonlinear analysis, called hdmsh6.inp, is listed in Appendix B. The initial displacement load
increment is specified as 0.01 in this input deck and the maximum displacement is specified as
0.06. This value of the maximum displacement is specified for the 3-direction of the DISP node
set, which is the set of nodes at the top end of the model.

42

4.5 FEA Results and Test Data Comparison

Only the 1.5- and six-inch columns are analyzed in detail. The 1.5-inch short columns failed
primarily in a material compressive strength mode, while the six-inch short columns failed in a
local postbuckling response mode. The failure of the four-inch short columns was more complex,
and appeared to exhibit combinations of both modes. Since the goal of the short column tests was
to study the mechanisms of compression failure with and without local buckling occurring during
the response, it was decided that the four-inch short column test was less important in assessing
criterion to use for each individual mode in the full frame analysis. Hence, the mixed mode failure
of the four-inch columns would not aid in achieving the goal of the current research and therefore
was not studied.
4.5.1 1.5-inch Column

The finite element model of the 1.5-inch column consisted of 304 S4R5 shell elements and 378
nodes, with element size on the order of a quarter of an inch. The load-shortening response plot
from the 1.5-inch specimen #1 is shown Figure 4.3. The response is nearly linear both in the test
6

and in the nonlinear FEA. The slope of the load-shortening test data is 2.1 10 lbs/in and the
6

FEA predicts a slope 2.4 10 lbs/in, which is a 14.3% discrepancy. The maximum shortening in
the test shown in the response plot is 0.0325 inches at a load of 64,700 lbs. At this same
shortening the FEA predicts at load of 78,000 lbs.
The axial strain responses from back-to-back gages bonded to the front side outer flange
of the 1.5-in. specimen are shown in Figure 4.4 along with the predictions from the FEA. The

43

100000
90000

Test Data
FEA

80000

Load (lbs)

70000
60000
50000
40000
30000
20000
10000
0
0.000

0.010

0.020

0.030

0.040

Displacement (in)

Fig. 4.3 End shortening response for specimen #1, 1.5-inch

column.
strains predicted at these locations from the FEA are labeled as section points S.P.1 and S.P. 3 in
the plot. Section points refer to the location through the thickness of the element where the strains
and stresses are computed. Since the FEA model has one shell element comprising all the lamina
through the wall thickness, S.P. 1 and S.P. 3 are located nearest the external surfaces of the wall.
The strains predicted from FEA correlate very well with the strain gage data until an end
shortening of about 0.029 inches. At this shortening gage 6 reads 1.1% compressive strain and
gage 5 reads 0.55% compressive strain. The discontinuity in the gage responses at the shortening

44

of 0.029 inches indicates an initiation of some failure event. The difference in these back-to-back
axial strains, or bending strain, is about 0.58%, and their average, or membrane strain, is about
0.84%. Gage 6 saturates as the shortening increases from 0.029 inches.
s.p. 1
2

Gage 5
Gage 6

Compressive Strain %

FEA Section Point 1

1.5

FEA Section Point 3

6
1
s.p. 3
0.5
5
0
0.00

0.01

0.02

0.03

0.04

Displacement (in)

Fig. 4.4 Compressive strain response in the front side outer

flange for specimen #1, 1.5-inch column.

In contrast to the response of the outer flange, the bending in the inner flange is negligible
prior to the first major failure event, as is shown by the responses of back-to-back axial strain
gages 1 and 2 in Figure 4.5. The discontinuities in gage 1 and 2 responses occurs at shortening of
45

0.0314 inches where the gages read a compressive axial strain of 0.69%. The axial strain
responses in the inner flange and the web are very similar, so for simplicity only the results for the
inner flange are shown. Video of the 1.5-inch column tests shows the inner flange failing prior to
the failure of the entire cross section, which is not the sequence seen for the six-inch specimens.
The compressive axial strain of 0.69% at the discontinuity in the gage responses is less than the
1% compressive failure strain predicted by TEXCAD for this material. The FEA predicts larger
compressive strains in the inner flange and web than what occurs in the test. At an end shortening
of 0.0314 inches FEA predicts a compressive strain of 0.94%. It is not clear why the correlation
between analysis and test is poorer in the inner flange and web with respect to the good
correlation achieved in the outer flange. The stiffer response of the FEA with respect to the test on
the load-shortening plot in Figure 4.3 correlates with the predictions of the axial strains in the
inner flange and web.
4.5.2 Four-inch Column

The only FEA completed on the four-inch column was a linear ABAQUS buckling analysis. This
analysis used 888 S4R5 shell elements with 950 nodes and an element size on the order of a
quarter of an inch. Boundary conditions and extra length representing the potting was included,
along with curvature and reduced inner flange properties. As can be seen in Figure 3.6 the local
buckling shape of the outer flange is a non symmetric half wave. ABAQUS predicts a symmetric
buckling pattern as shown in Figure 4.6. This discrepancy in the buckling wave pattern between
test and analysis maybe due to the support conditions in the test as idealized in the analysis. The
depth of the potting compound in the fixture varies around the specimen, and this variable depth is
not modeled in the FEA.Though a difficult task, perhaps more attention needs to be given in

46

making the potting resin cure evenly so that potting thickness does not vary around the cross
section of the specimen.
1.2

Compressive Strain %

Gage 1
Gage 2
FEA

0.8

0.6

0.4

0.2

0
0.00

0.01
0.02
0.03
Displacement (in)

0.04

Fig. 4.5 Compressive strain response in the inner flange of

specimen #1, 1.5-inch column.

47

Fig. 4.6 ABAQUS linear buckling results for a four-inch

column. On the right is a front on view of the outer
flange, showing the symmetric buckling waves on the
front and back side outer flange.

4.5.3 Six-inch Column

Short column specimens 3 and 4 have a six-inch gage length. The FEA model contains 5698 S4R
shell elements and is shown in Figure 4.7. Element size in the model is on the order of a sixteenth

48

of an inch, and this refined mesh was developed primarily for the failure analysis to be discussed
in Chapter 6.

Fig. 4.7 Finite element model of six-inch column.

The load-shortening data from the test of the six-inch specimen #3 along with the FEA
predictions are shown in Figure 4.8. The analysis and test compare very well in the initial portion
6

of the response. The initial stiffness is 1.1 10 lbs/in from the test data and 1.06 10 lbs/in from
the FEA results, which is 3.6% discrepancy. The analysis begins to soften relative to the test data
as the load increases from 30 kips.

49

60000
Test Data
FEA

50000

Load (lbs)

40000

30000

20000

10000

0
0.00

0.01

0.02

0.03

0.04

0.05

0.06

Displacement (in)

Fig. 4.8 Load-shortening data for specimen #3 and FEA results.

The axial strains responses from back-to-back gages 5 and 6 located on the front side of
the outer flange are shown in Figure 4.9. The strains predicted at these locations from the FEA are
labeled as section points S.P.1 and S.P. 3 in the plot. The FEA and test are in close agreement in
the initial portion of the response but begin to deviate in the postbuckling region. Fair correlation

50

is achieved in the postbuckling response regime. At an end shortening of about 0.045 inches gage
1

Test Data, Gage 5

Test Data, Gage 6

Compressive Strain %

0.8

FEA, Section Point 1

FEA, Section Point 3

S.P. 1
0.6

0.4

0.2

5
S.P. 3

0
0.00

0.01

0.02
0.03
0.04
Displacement (in)

0.05

0.06

Fig. 4.9 Axial strain response for in the front side outer flange
of specimen #3 and FEA strain response.
6 reads 0.89% compressive strain and gage 5 reads 0.11% compressive strain. As the shortening
increases from 0.045 inches, responses from gages 5 and 6 exhibit at small jump and then the
difference in the strains, or bending strain, begins to decrease.

51

The axial strain response from gage #11, which is located on the external surface of the
outer flange just above the web junction, is shown in Figure 4.10. It is along this line that extra
filler material was added in the fabrication process. During fabrication, the eight plies of fabric in
the web of the preform were split into two four ply fabrics to form the outer flanges forming a
void shown in Figure 2.1. Post inspection of the failed specimens revealed fracture surfaces that
progressed in a direction parallel to the load axis in the matrix surrounding the filler material, and
in some instances the filler material separated from the junction. Gages 11 to 13 were located on
the line of filler material to a see if this structural feature influenced failure. As is shown in
Figure 4.10, at an end-shortening displacement about 0.045 in., which corresponds to strain of
0.54%, the gages exhibit a jump suggesting a failure initiating in the junction. Recall that gages 5
and 6 on the outer flange in Figure 4.9 also jumped at this end-shortening displacement. The
junction remains relatively straight during the local postbuckling response of the flange. In
postbuckling the flange carries proportionally less of the direct compression and the junction
carries proportionally more direct compression as the end-shortening increases. A large increase
in the bending response of the flanges between end-shortening displacements from 0.030- to
0.045-inches is evident in Figure 4.9.

52

0.8
Test Data, Gage 11

Compressive Strain %

FEA
0.6

0.4

0.2

0
0.00

0.01

0.02

0.03

0.04

0.05

0.06

Displacement (in)

Fig. 4.10 Strain response of the filler material at the web to

outer flange junction and FEA results.

Back-to-back axial strain gage data from the web indicates a small amounts of bending
occurring in the initial portions of the test as shown in Figure 4.11. The predictions of these strain
responses from the FEA are very good in the initial portions of the test, but the correlation
degrades in the postbuckling response. The FEA predicts a large amount of bending in

53

postbuckling relative to the strain gage data. The analysis predicts the back side of the web going
into tension at a lower value of shortening than occurred in the test.
1.4
1.2

Compressive Strain %

Test Data, Gage 3

S.P. 3

Test Data, Gage 4

FEA, Section Point 1

FEA, Section Point 3

0.8
0.6
0.4
0.2
0

-0.2
-0.4
S.P. 1
-0.6
0.00

0.01

0.02
0.03
0.04
Displacement (in)

0.05

0.06

Fig. 4.11 Axial strain response in the web and FEA results for
the six-inch column.

54

The back-to-back axial inner flange strains from the test and FEA of the six-inch column
are shown in Figure 4.12. Correlation between the strain gage data and FEA is very good. through
most of the response.
1
Test Data, Gage 1

S.P. 1

Test Data, Gage 2

Compressive Strain %

0.8

FEA, Section Point 1

FEA, Section Point 3

0.6

S.P. 2

0.4

0.2

0
0.00

0.01

0.02
0.03
0.04
Displacement (in)

0.05

0.06

Fig. 4.12 Inner flange axial strain response and FEA results for
the six-inch column.
Finite element analysis results for the short columns correlates well in the initial linear
region and fairly well once test data goes non linear. The load drops measured during testing were
not predicted. Further improvements to the analysis need to be implemented.
55

CHAPTER 5

Frame Segment Finite

Element Analysis and
Comparison

Finite element modeling for the response of the full 48-degree frame tested in Ref. [3] is
presented. Issues discussed include mesh development, FEA representations of the applied
load, and geometrically linear and nonlinear response analyses. The material law is assumed
to be linear elastic.

5.1 Finite Element Model

Finite element models of frame B [3] were developed in ABAQUS/Standard using the
dimensions listed in Table 1.1, material property data for the web and outer flange listed in
Table 2.1, reduced Youngs moduli as discussed in Section 2.4.4 on page 15, and shell
element S4R. Clamped boundary conditions were prescribed in the analyses to simulated the
frame mounted in the end blocks. One model, the coarse mesh model, had 999 nodes and 870
elements with an average element dimension of 1.2 in. by 1.0 in. A second model with a finer
mesh containing 3757 nodes and 3530 elements with average element dimensions of 0.6 in.

56

by 0.5 in. was also considered. Both meshes gave the same initial stiffness of the frame. However,
to adequately capture the local buckling response and to model contact between the platen of the
testing machine and the frame, the mesh was refined again, and this model had 13,217 elements
and 17,190 nodes. The model with the most refined mesh is shown in Figure 5.1, which shows a
graduated mesh to conserve degrees of freedom. The regions near the supported ends of the frame
retained an element size of a half an inch. A 45-inch-long section centered at the midspan with
mesh size on the order of a quarter inch was bordered by two 2-inch-long transition sections that
connected to the coarse mesh near the supports. The finest mesh was located at the midspan of the
frame; here the element dimensions are 0.23 in. by 0.226 in the outer flange, 0.226 in. by 0.218 in
the web, and 0.23 in. by 0.226 in. the inner flange. The refined mesh scheme was used to
determine the results presented in this chapter, because deformations varied rapidly near the apex
where the load platen contacts the frame.

Fig. 5.1 Refined shell element model of a 48 degree frame

segment.

57

Two mid-span loading schemes were used to simulate the frame loading condition as
depicted in the sketch of the frame test apparatus shown in Figure 1.1. The first scheme was a line
load specified at midspan. Along the line of nodes across the width of the top flange at midspan,
the displacements in the direction out of the plane of the frame were specified to be zero, and the
displacements in the radial inward direction were specified to be a non-zero value. The second
scheme was a contact algorithm available in ABAQUS. A rigid plane parallel to the platen was
displaced in the negative 2 direction toward the outer flange of the frame.

5.2 Frame FEA Results

Response plots of the prescribed radially inward displacement at midspan and the corresponding
reaction force from the analyses and test of frame B are shown in Figure 5.2. The FEA responses
were plotted for a linear analysis with line loading, a geometrically nonlinear analysis with line
loading, and a nonlinear analysis using the contact algorithm. The iterative procedure in the
nonlinear analyses use Newtons method, and the ABAQUS software issues warnings messages
when the eigenvalues of the tangent stiffness matrix at an equilibrium state are negative. As the
load is increased from zero, which corresponds to the origin in the response plot, the first
equilibrium state encountered along a nonlinear equilibrium path where a negative eigenvalue
occurs is an approximation of the critical state. The critical state divides stable and unstable
equilibrium states along the path, and these critical states on the nonlinear equilibrium paths are
noted in Figure 5.2. Note that the slope at the origin of the response plot, or initial stiffness, of all
three finite element analyses are essentially the same as the initial stiffness determined from the
test data. However, the linear analysis cannot predict the softening response exhibited by the test

58

data. Both geometrically nonlinear analyses soften with increasing load, but the nonlinear
analysis with the contact algorithm more closely correlates to the test data.
FEA, Linear Line Load

8000

Test Data
7000

FEA, Non Linear Line

Load
FEA, Non Linear
Contact Load

Reaction Force at Mid Point (lbs)

6000

Critical State
5000

4000

3000

2000

Critical State

1000

0
0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Displacement of Mid Point (in)

Fig. 5.2 Midspan displacement and force response plots from

linear and nonlinear analysis with and without contact
compared to test data of frame B.
A quantitative comparison between the FEA predictions of the critical equilibrium states
and the test data at the first major failure event is given in Table 5.1. The first major failure event
in the test occurred at a displacement of 0.4782 inches and the corresponding force was 5,791.6

59

lbs. The force from the linear analysis at the displacement of 0.4782 inches is 29.4% larger than
the force measured in the test. For the critical states from the nonlinear analysis with contact
loading, the force is 7.5% greater and the displacement is 1.6% lower than test data. As can be
seen from the data in the table, the critical state from the nonlinear analysis using the contact
algorithm correlates closely with the first major failure event in the test.
Table 5.1 Responses From Several Analyses Compared to The Test of Frame B

Analysis type
Linear; Line Load
Nonlinear; Line Load
Nonlinear; Contact
Load

Displacement
(in.)

Midspan
reaction force
(lbs)

% Discrepancy with respect

to testa

0.4782

7498.5

0.4103b

4345.4c

Disp.
0 (From test)
-14.2

0.4042c

5604.3c

-1.6

Force
29.4
-24.9
7.5

a. The first major failure event occurred at a displacement of 0.4782 inches and at a midspan reaction force of
5791.6 lbs.
b. Critical equilibrium states predicted by the geometrically nonlinear analyses.

Further improvements to the analysis are required to represent the load drop that occurred
in the test. In order to represent the load drop failure mechanisms such as delamination and
material degradation need to be represented in the analysis.

60

CHAPTER 6

Interface Element and

Progressive Failure
Analysis

Failure mechanisms observed during short column and frame tests were similar. In this
chapter these characteristics will be briefly discussed along with implementing user-written
subroutines to model these mechanisms in ABAQUS/Standard.

6.1 Failure Mechanisms

Multiple mechanisms were observed during testing, including matrix fractures surrounding
the bias yarns with theses yarns popping out of the plane of the web, delamination, material
brooming, and the separation of the axial strand of filler material between the web to outer
flange junction. The failure sequence of frame B began at midspan in the junction between the
web and outer flange where the load was applied. Here failure began in the outer flange and
then propagated through the outer half of the web. Failure mechanisms can be classified as
intralaminar material failures occurring by several micro mechanisms and interlaminar
failures, or delamination [9]. Separation of the axial filler material in the junction of the web

61

and outer flange can be considered a separate mode of failure, which is a feature unique to the
fabrication of the J-section.
Braided textile material failure mechanisms are hard to observe even in quasi-static testing
because of how quickly failure occurs. Under compressive loading axial tows can kink causing
matrix crazing, bias tows experience micro cracking, and finally delamination between plies is
observed. The exact sequence of events is not understood due to the randomness of material and
manufacturing flaws. Similar events occur under axial tension with delamination being less
severe. Because of the multi axial nature of a triaxial braided material a failure due to any of the
stresses reduces the effectiveness of the material drastically.
We take the approach of using a criterion based on the mode of failure followed by a
scheme to degrade material properties when a mode of failure is predicted to occur. Since we have
predictions of the in-plane strengths of the triaxial braided composite material from TEXCAD,
these data are used to estimate the onset of intralaminar failure from the in-plane stress state
within an element. Intralaminar failure prediction is followed by a scheme to reduce lamina
moduli. The onset of interlaminar failure prediction, or delamination initiation, is based on a
quadratic criterion in terms of the interlaminar tractions and interlaminar strengths. The
progression of delamination cracks is based on a mixed mode fracture criterion involving the
strain energy release rates for each mode. Initiation and progression of the delamination process is
accomplished via special interface elements, or decohesion elements, as presented in Ref. [15].

62

6.2 Intralaminar Failure and Degradation

A FORTRAN subroutine was written to be used in conjunction with ABAQUS to predict
intralaminar failure initiation followed by a degradation in the moduli. The algorithm follows one
used in Ref. [11] for intralamina progressive failure of a unidirectional composite material. The
in-plane strengths for the triaxial braid determined from TEXCAD are listed in Table 2.1 on
page 12. With this limited material strength data, we assumed a maximum in-plane stress criterion
to predict the initiation of failure in an element. The FORTRAN subroutine used is called from the
ABAQUS input deck under the *USER DEFINED FIELD command. The state of stress in an
element from the equilibrium state at the previous load is passed to the subroutine prior to the next
load increment. The in-plane stresses are compared to their corresponding strengths. If the stress
magnitudes equal or exceed their corresponding strengths, then the value of a field variable is
changed from zero to one. Field variables are a means to indicate whether failure has occurred or
has not occurred; a value of zero means no failure and a value of one means failure. The values of
the field variables are returned to the input deck where they are used to determine which lines of
material properties are used to calculate the stress and strains for the next load step.
The maximum in-plane stress criterion is non-interactive and is based on the values of the
axial normal stress S 11 , the transverse normal stress S 22 , and the in-plane shear stress S 12 .
Consequently three field variables are required: Field variable FV1 is associated with the axial
normal stress limits, FV2 is associated with the transverse normal stress limits, and field variable
FV3 is associated with the shear stress limits. The initial values of the three field variables are
zero, representing an intact layer. If S 11 91.37ksi or S 11 71ksi , then FV1 = 1 and the axial
modulus (E11) is multiplied by a factor of 1/106 and the transverse modulus (E22) and the shear
63

moduli (G12, G13, G23) are multiplied by a factor of 1/1000. If S 22 73.14ksi or S 22 56.89ksi ,
then FV2 = 1 , and the same scheme as used before is followed but the transverse modulus (E22)
is multiplied by a factor of 1/106 and the axial modulus (E11) is multiplied by a factor of 1/1000. If
S 12 30.46ksi , then FV3 = 1 and the in-plane moduli (E11, E22) and the out of plane shear
moduli (G13, G23) are multiplied by a factor of 1/1000, the in-plane shear modulus (G12) is
multiplied by a factor of 1/106. For combinations of failure modes, the previous reduction factors
are combined together decreasing each property by the largest factor unless the factors are equal
and greater than 1/106. In the latter case, there is an additional reduction of 10. For example, if the
values of field variables are (1,0,1), which indicates a combination of axial and shear failure, then
E11 and G12 are decreased by 1/106 and E22, G13, G23 are multiplied by 1/104 because the axial
failure factor is 1/103 and the shear failure factor is 1/103. See Table 6.1 for the property
degradation scheme. Also, a sample of the subroutine that implemented this scheme is given in
Appendix B

6.3 Delamination Initiation and Progression

Interface elements are used to model the delamination process between adjacent layers of the
quasi-laminar textile composite. The interface element used to model interlaminar crack initiation
and growth in tape laminates developed by Goyal, Johnson, Dvila, and Jaunky [15] was
implemented in the FEA of the textile columns and frames. Interface elements were located in the
models at the mid surface of the outer flange, where delamination was observed to initiate in the
the tests. The displacements of the sublayers above and below the mid plane where the interface
elements are located determine interlaminar tractions through a phenomenological constitutive

64

Table 6.1 Reduction Factors For Intralaminar Material Degradation

ABAQUS field
variables
FV1 FV2 FV3

Reduction Factor for Each Material Property

E11

E22

E33

12

13

23

G12

G13

G23

1
0

1
0

1
0

1/103

1/103

1/103

0
1

0
0

0
0

1/106

1/103

1
1

1/103

1/106

1/103

1/103

1/103

1/103

1/103

1/106

1/103

1/103

1/106

1/106

1/104

1/104

1/104

1/106

1/104

1/106

1/104

1/104

1/104

1/106

1/106

1/104

1/104

1/106

1/106

1/103

1/106

1/105

1/104

law. Each interface element consists of 2 surfaces and 8 nodes, an upper and lower surface with 4
nodes each. The 4 nodes on the upper surface of the interface element connect to the 4 nodes on
the lower surface of the element in the upper sublayer of the outer flange. The 4 nodes on the
lower surface of the interface element connect to the 4 nodes on the upper surface of the element
in the lower sublayer of the flange. In the undeformed state the two surfaces of the interface
element coincide. The constitutive law that resists the separation of the upper and lower surfaces
of the interface element can be conceptualized as representing a continuous distribution of
nonlinear springs that do not interact with one another; i.e. similar to Winkler foundation. A
graphical representation of the interface element is shown in Figure 6.1

65

S+
X3, x3, U3

S0

Xi (1,2)

Ui

xi+ (1,2)

xi (1,2)

Ui
S
X2, x2, U2

X1, x1, U1

Undeformed
Configuration

Deformed
Configuration

Fig. 6.1 Upper (S+) and lower (S-) surfaces of the interface
The constitutive law relates the tractions to the displacement jumps across the interface.
An exponential form of the this relationship is used Ref. [15], which is illustrated for Mode I
loading (opening mode) in Figure 6.2. In this figure T is the traction, is the displacement jump
normal to the interface, T c is the specified tensile, or peel, strength, and G c is the specified
critical strain energy release rate in Mode I. Tractions first increase with increasing , reach a
maximum at = c , then decrease as > c . The exponential form of the law and specification
of T c and G c determine the critical separation length = c . The critical separation length is a
characteristic length scale for the initiation of two free surfaces of the incipient crack. The
softening portion of the constitutive law represents the presence of the process zone ahead of the
crack tip, which is a craze zone in polymers or a plastic zone in ductile metals. Under combined
Mode I, II, and III loading, delamination initiation is predicted using multi-axial stress criterion
and the propagation of delamination is based on an mixed-mode fracture criterion [15]. A damage

66

parameter is included in the formulation such that on unloading from a state where > c , the
traction does not follow the loading curve. That is, unloading is associated with a loss of work.

T
Tc

Gc
C
A c

Fig. 6.2 Representative traction-stretching curve for the springs

Because there is little test data for debonding strengths of 2D triaxial braided textile
materials, and testing for this data is beyond the scope of the project, values used for the following
analyses will initially be based on unidirectional graphite epoxy composite. Values used for the
three interfacial strengths T ic , i = 1, 2, 3 , and the three critical strain energy release rates
G Ic, G IIc, G IIIc are listed in Table 6.2.
Table 6.2 Interfacial Strengths [15]
c

T1 , T2

T3

G Ic

G IIc , G IIIc

11.6 ksi

8.7 ksi

2.01 lbs/in

8.28 lbs/in

67

6.4 FEA Results

Analyses implementing various combinations of the intralaminar and interlaminar progressive
failure models were undertaken and compared to tests results. In the following sections the results
from the six-inch short column model will be discussed followed by the results from the full
frame analysis.
6.4.1 Short Column Analysis

To evaluate the effectiveness of the interlaminar progressive failure model, it was implemented in
the FEA of the six-inch column. In order for the interface elements to predict delamination the
element size was reduced. Because of computer limitations the smallest element size used was on
the order of 0.0625 inches (1.6 mm). A biased mesh seeding technique was used to go from the
small elements to the elements to that were approximately a quarter of an inch in size. The model
consists of S4R shell elements and the user-defined interface elements. Interface elements were
inserted at the midsurface of the outer flange and are located in the gage length, they do not
extend into the potting regions on the top and bottom of the column. The outer flange thickness
was split into two element sets making up a radially inner and outer sublayers surrounding the
midsurface. The reference surfaces of the element sets in the sublayers inward and outward of the
midsurface were then offset such that they coincided with midsurface of the flange. The offset
reference surfaces of the sublayer elements make the nodes of the sublayer elements coincide
with the nodes on the interface element and facilitates compatibility between elements. Including
the interface elements, the finite element model of the six-inch gage length column had a total of
8,929 nodes and 11,778 elements.

68

Initially the FEA was run with just the interlaminar progressive failure model and no
intralaminar progressive failure model. Without the intralaminar failure, delamination did not
initiate. Delamination initiated in the FEA only when the combination of the interlaminar and
intralaminar progressive failure models were included in the analysis.
Comparison of the FEA was made with the test data on the load end-shortening plot,
which is shown in Figure 6.3. There is good correlation between test and analysis prior to failure
initiation, and the initial stiffness of the finite element model is within 3.6% of the stiffness in the
test. The FEA begins to diverge from the test data after the predicted and observed local buckling
load of the column. After local buckling, the stiffness in the FEA decreases faster than what
occurs in the test. It is not clear why the column is stiffer in postbuckling than predicted by the
analysis. The first major failure event in the test occurs at a displacement of 0.0496 inches and a
load of 49,500 lbs. The first major drop in the load from the FEA occurs at a displacement 0.058
inches and a load of 48,377 lbs. At the major load drops, the displacement from FEA is 16.9%
greater and the load is 2.3% less than the corresponding data from the test. The energy absorbed
by the column, which is the area under the load-shortening plot, to the displacement of 0.058
inches is 13.3% higher in the FEA than in the test. In spite of these discrepancies, the progressive
failure analysis demonstrates it has the capability to represent the severe discontinuities in the
slope of the load-shortening response observed in the test. If better strengths and strain energy
release rates were available for this material to use in the progressive failure model, then the
predictions of the failure sequence in the frames would have a more rational basis and may
improve.

69

60000

Test Data
FEA w/Interface Elements &
Progressive Failure

50000

Load (lbs)

40000

30000

20000

10000

0
0.00

0.01

0.02

0.03

0.04

0.05

0.06

Displacement (in)

Fig. 6.3 Load versus displacement response for the six-inch

column from the test and FEA with progressive failure
models.

The first element failures occurred at the ninth load increment where the end-shortening
was 0.0579 inches and the load was 47,467 lbs. Similar to the test, failures occurred rapidly.
These first element failures were located in the junction of the web and outer flange at the axial
location of the wave crest of the largest buckle in the outer flange. Elements also failed in the
front side of the outer flange at the axial location of the smaller buckle over its entire wave length.

70

In the next load increments the many intralaminar failures lead to the large drop in the load shown
in Figure 6.3.
6.4.2 Frame Analysis

Refinements in the forty-eight degree frame model were required to capture the delamination
process in the outer flange. To keep the model of reasonable size for the computational resources
available, the model size had to be decreased for the smaller elements required to model
decohesion associated with delamination. Using symmetry, only half of the frame was modeled.
Boundary conditions modeling symmetry were used at the midspan location. With the physical
model size decreased, the element size was reduced so that the smallest element in the outer
flange was on the order of 0.06 inches (~1.6 mm). The model consisted of S4R shell elements,
R3D8 rigid elements for the platen in the contact analysis, and the user-defined interface
elements. The final model contained 16,770 elements and 15,081 nodes. However, ABAQUS
then created 2,268 additional elements and 6,195 additional nodes of its own to run the contact
analysis.
Interface elements are only included in the midplane of the outer flange over a 10 degree
arc beginning at the midspan because of modeling difficulties and size limitations. Hence, the
model cannot represent crack propagation from the outer flange into the web at the apex of the
frame. Intralaminar progressive failure prediction is included only for the shell elements making
up the outer and inner sublayers of the outer flange surrounding the interface elements.
Initial analyses were performed using the failure criteria separately; in other words, one
analysis used only the interlaminar progressive failure to predict delamination and the other used

71

only the intralaminar progressive failure criterion to degrade the layer material properties. A
comparison of the two analyses is shown on the load displacement plot of Figure 6.4. The severity
of material degradation is evident with the large load drop and softening of the frame due to
intralaminar failure. Element failure initiates at a displacement of 0.415 inches where an 800 lb.
load drop occurs. Element failure initiates at the free edge of outer flange on the back side of the
section. From this free edge, failure spreads a short distance circumferentially in the flange. Then
failures occur in the front side outer flange approximately five degrees from the apex.
8000

Interlaminar Progressive Failure Model

Intralaminar Progressive Failure Model
Test Data

Reaction Force at Mid Point (lbs)

7000
6000
5000
4000
3000
2000
1000
0
0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Displacement of Mid Point (in)

Fig. 6.4 Load displacement responses from analyses including

intra- or inter- laminar progressive failure models
compared to test data.

72

The location of delamination initiation is more difficult to determine because the

ABAQUS software does not have the normal full post-processing capabilities for user-defined
elements. However, from the analysis with interlaminar progressive failure only, delamination
initiates at a displacement of 0.5763 inches and a load level of 6580 lbs, where a sharp drop in
load occurs and the analysis begins to stall due to convergence difficulties.
Implementing both progressive failure models in the FEA predicts a load-displacement
response of the frame that agrees fairly well with the test as shown in Figure 6.5. Initial stiffness
is, again, nearly the same in both the analysis and test. The first major failure event is predicted to
occur at a displacement of 0.428 inches and a load of 5794 lbs. Compared with the test data, the
displacement is 12% lower and the load is 1.2% higher. Calculating the energy absorbed up to a
displacement of 0.58 inches shows that the FEA is 6.5% greater than what occurs in the test,
which is good correlation.
Intralaminar element failure initiates at increment 26 with the first load drop occurring by
increment 28. Failed elements are in the backside outer flange at the midspan and propagate
circumferentially in both the upper and lower surfaces of the outer flange. Interlaminar failure
initiates at the apex of the frame in the outer flange. The first elements to accumulate damage
straddle the junction of the web and outer flange. Delamination progresses circumferentially for 2
inches, in the high density mesh area of the outer flange.
To capture more of the failure domain observed during testing, the intralaminar
progressive failure model was added to the elements comprising the central span of the web and
the section of the inner flange near the clamped end. This analysis captured the load-shortening
response behavior fairly well with a large load drop followed by reloading as is shown in
73

7000

Test Data
Intra & Inter Laminar Failure Criteria

Reaction Force at Mid Point (lbs)

6000
5000
4000
3000
2000
1000
0
0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Displacement of Mid Point (in)

Fig. 6.5 Load versus displacement responses for the test and
FEA including both the intra- and inter- laminar
progressive failure models.
Figure 6.6. The slope of the reloading portion of the response from FEA is within 10% of the test
data at reloading. The energy absorbed by the FEA is 11% lower than measured by the test.
Along with the good correlation of the load-shortening behavior, the location of element
failures correlate well with the locations of failure observed in the test. A graphical representation
of the element failures is shown in Figure 6.7. Initial element failures are predicted by the
intralaminar criterion an increment prior to the first large load drop. Element failures occur near

74

Reaction Force at Mid Point (lbs)

7000

Test Data

6000

Interface Elements & Progressive Failure In

Outer Flange And Web

5000
4000
3000
2000
1000
0
0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Displacement of Mid Point (in)

Fig. 6.6 Load displacement responses from the analysis

including both the intra- and inter- laminar progressive
failure models and the test. The domain of the
intralaminar failure extended to web and inner flange.
the junction of the web and lower sublayer of the outer flange. During the increment where the
first large load drop occurs interface elements straddling the junction of the web and the outer
flange begin to accumulate damage at the apex of the frame, these elements are shown in red.
Intralaminar failures are predicted in the upper and lower sublayers of the outer flange at the
damaged interface elements and into the back side outer flange at the apex, these elements are
highlighted with green. In the second large load drop intralaminar failures begin to fail in the web
at the apex and propagate circumferentially along the junction in the outer flange, these elements

75

are shown in blue. At the end of the total load drop intralaminar failures end while interlaminar
damage continues. By the end of the analysis the elements shown in pink have delaminated, or
accumulated damage. A comparison of the failed frame and the predicted element failures is
shown in Figure 6.8. Failures observed during the test of Frame B included the separation of the
axial filler material in the outer flange at the apex of the frame, cracking of the front and back
sides of the outer flange, and cracking of the web from the outer to the inner flange. Delamination
and intralaminar element failures were predicted at the apex of the frame over the junction
between the web and outer flange. Intralaminar failures were predicted in the back side outer
flange and the web near the apex. The failure pattern and sequence predicted by the progressive
failure model correlated very well with damage observed during the failure of Frame B [9].
Including both inter- and intra- laminar progressive failure models in the analysis
improves the correlation of the response and energy absorbing characteristics to what was
measured in the frame test.

76

Damaged interface elements at first load

drop
Failed elements during the first load
drop
Damaged interface elements at the end
of the analysis
Failed elements by the end of the load
drop
First load drop
(5794-5620 lbs)
Second load drop
(5620-5266 lbs)

Fig. 6.7 Predicted element failures using the progressive failure

model.

77

Fig. 6.8 Comparison of the element failures predicted by the

progressive failure model and failures that occurred on
frame B.

78

CHAPTER 7

Concluding Remarks

7.1 Background
The goal of this research was to improve the analysis of the fuselage frame test conducted in
Ref.[3] past the first major failure event. This circular frame test article had a nominal radius
of 120 inches, a J cross section, a forty-eight degree included angle, and was subjected to a
quasi-static, radially inward load. The largest circumferential strains measured in the test were
compressive and were located at the center of the frame where the load was applied through
contact with the platen of the testing machine. However, compressive strains measured by
electrical resistance strain gages near the site of the first major failure event were about 40%
lower than the theoretical compressive failure strain of the material predicted by computer
program Textile Composite Analysis for Design, or TEXCAD [5].The deformation of the
frame in the vicinity of the load platen consisted of a local buckling and postbuckling
response of the radially outboard flanges and distortion of the cross section. Local
postbuckling of the flanges reduces the direct compressive load carried by the flanges at the

79

expense of increased direct compression carried by the junction between the flanges and web. The
first major failure event in this local deformation state can be due to a material compressive
strength failure in the junction, or by delamination initiating in the flange, or by a combination of
both mechanisms.

7.2 Finite Element Analyses

Branched shell models were developed using the ABAQUS/Standard software for the structural
analyses conducted in this study. The analyses of the frame included geometric nonlinearity and
contact of the load platen of the testing machine with the frame. Branched shell models consisted
of S4R shell elements, R3D8 rigid elements for the platen in the contact analysis, and the userdefined interface elements to model delamination. Shell elements S4R5 were used in some
analyses, but the S4R element models correlated better with the test data in postbuckling.

7.3 Short Column Test and FEA Results

To further study the failure mechanisms associated with the unsymmetrical J cross section, short
column compression tests of the J section were performed. Quasi-static compression tests of short
columns allows for the study of failure mechanisms associated with the cross sectional shape of
the fuselage frames. The goals of these tests were to study the mechanisms of failure and to
measure the compressive strength of the section with and without local buckling occurring during
the response.Three short column heights were tested and then analyzed.
With a refined branched shell analysis of the 1.5- and six-inch short columns, good
correlation was achieved for the initial loading response with element size on the order of quarter
of an inch. Initial stiffness of the 1.5-inch column was within 14% of the test while the six-inch
80

column was within 3.6% of the test. The discrepancy in the FEA, especially with the 1.5-inch
column, may be attributed to the compliance of testing machine, especially for sudden load
reductions occurring at very large load magnitudes, to friction between the potting material and
the specimen, or to the material properties, such as lower elasticity moduli associated with
compression which is common for textile composites. The six-inch short column response
exhibited local buckling and postbuckling.
Finite element analyses (FEA) of the six-inch short column test specimen were conducted
to assess intra- and inter- laminar progressive failure models. Intralaminar progressive failure is
based on a maximum in-plane stress failure criterion using strength data predicted by TEXCAD,
followed by a moduli degradation scheme. Interlaminar progressive failure was implemented
using an interface finite element to model delamination initiation and the progression of
delamination cracks. Implementing the progressive failure models in the FEA of the six-inch
column required a small element size on the order of 0.0625 inches (1.6 mm). The load level at
the first major failure event was predicted within 2.3% of the test and the displacement was 16.9%
greater. Load response characteristics of the FEA correlated well with the test and exhibited the
sharp drop in load. The FEA prediction of the energy absorbed by the column to this failure load
was 13.3% greater than the test. Due to computer limitations the element size could not be further
reduced, but such a reduction may improve the accuracy of the analysis by better capturing the
large stress gradients associated with the large load levels and brittle fracture of the short
columns. Reasonably good correlation of FEA to test data was achieved, which gave confidence
to use the FEA with the progressive failure analysis to model the response of the fuselage frame.

81

7.4 Fuselage Frame FEA Results

Simple branched shell analysis of the fuselage frame segment yielded good results when the
contact algorithms included in ABAQUS were used to represent the loading platen of the
universal testing machine. Initial stiffness of the FEA was the same as the test while load and
displacement at numerical instability were within 7.5 and 1.6% of the test, respectively. These
results were achieved with element size on the order of 0.2 inches.
Implementing the progressive failure models in the frame FEA was done with the
minimum element size on the order of 0.0625 inches (1.6 mm). Again the initial stiffness of the
FEA was nearly the same as that measured during the test. Displacement at the first major failure
event was 12% lower than the test but the load was predicted within 1.2%, and the energy
absorbed was within 6.5% of the test when the intra- and inter- laminar progressive failure model
covered the outer flange only. Failure patterns were also represented well, with element failures
initiating in the outer flange near the apex as was observed in the test. When the intralaminar
failure model was included in the web and outer flange better correlation was achieve with respect
to the FEA response characteristics. Here failure initiated in the junction and progressed in both
the outer flange and the web, which was similar to the test. Loading characteristics correlated as
well, with a load drop and reloading occurring under increased displacement. See Figure 6.6 on
page 75.

82

7.5 Final Thoughts

More in depth material properties for all branches of the cross section due to varying fiber volume
fractions and unit cell dimensions or running TEXCAD again with updated fiber volume fractions
and unit cell dimensions may improve the load response of the FEA. More computing power and
disk space to run refined models with much smaller elements that include failure criteria over the
entire model may improve FEA correlation.
If time allowed, an improved intralaminar failure model for braided textile composites
using micro mechanical analysis of undulating axial tows and bias rovings to predict unit cell
failure and property degradation may be a large improvement in the FEA.

83

References

[1]Cox, Brian N., Flanagan, Gerry, Handbook of Analytical Methods for Textile
Composites, NASA Contractor Report 4750, March 1997.

[2]Carden, H.D., Boitnott, R.L., and Fasanella, E.L.,Behavior of Composite/Metal Aircraft

Structural Elements and Components Under Crash-Type Loads - What Are They Telling
Us?, Proceedings of the 17th Congress of the International Council of Aeronautical Sciences
Vol. 2, (Stockholm, Sweden, Sept. 9-14, 1990), American Institute of Aeronautics and
Astronautics, Inc., Washington DC, 1990, p. 1195-1208.

[3]Prez, J.G., Johnson, E.R., and Boitnott, R.L., Tests of Braided Composite Fuselage
Frames Under Radial Inward Load, Proceedings of The 41st AIAA/ASME/ASCE/AHS/ASC
Structures, Structural Dynamics and Materials Conference, (Atlanta, GA, April 3-6, 2000),

84

American Institute of Aeronautics and Astronautics, Reston, VA, 2000, (AIAA Paper No. 20001547).

[4]ABAQUS/Standard Version 5.8 Users Manual, Hibbitt, Karlsson, & Sorensen, Inc., 1080
Main St., Pawtucket, Rhode Island, 2001.

[5]Naik, R. A., TEXCAD - Textile Composite Analysis for Design: Version 1.0 Users Manual,
NASA CR-4639, National Aeronautics and Space Administration, Hampton, Virginia, 1994.

[6]Masters, J.E., Foye, R.L., Pastore, C.M., and Gowayed, Y.A., Mechanical Properties of
Triaxial Braided Composites, Experimental and Analytical Results. Journal of Composite
Technology and Research, Vol 15, No. 2, 1993,pp. 112-122.

[7]Hexcel Carbon Fiber AS4 (6000 Filaments), MatWeb [online data base], URL: http://
www.matweb.com/SpecificMaterial.asp?bassnum=EHECF6 [cited March, 2002].

[8]3M ScotchplyTMBrand PR 500 Epoxy Resin Data Sheet [online data base], URL: http://
www.3m.com/us/mfg_industrial/adhesives/pdf/123452.pdf [issue no.2, April 1,1990].

[9]Perez, Jose G., Energy Absorption and Progressive Failure Response of Composite Fuselage
Frames, Master of Science Thesis in Aerospace Engineering, Virginia Tech, Blacksburg, VA
24061-0203, July 1999.

85

[10]Naik, Rajiv A., Failure Analysis of Woven and Braided Fabric Reinforced Composites,
NASA Contractor Report 194981, September 1994.

[11] Dvila, Carlos G., Ambur, Damodar R., McGowan, David M., Analytical Prediction of
Damage Growth in Notched Composite Panels Loaded in Axial Compression, Proceedings of
The 40th AIAA/ASME/ASCE/AHS Structures, Structural Dynamics and Materials Conference, St.
Louis, MI. April 12-15 1999. Paper AIAA-99-1435

[12] Naik, Rajiv A., Analysis of Woven and Braided Fabric Reinforced Composites, NASA
Contractor Report 194930, June 1994.

[13]Barrie, R.E., Skolnik, D.Z.,Evaluation of Textile Composites for Fuselage Frames,

Proceedings 4th NASA/DoD Advanced Composites Technology Conference, NASA CP-3299,
pp. 719-735.

[14]ABAQUS/Standard Version 5.8 Theory Manual, Hibbitt, Karlsson, & Sorensen, Inc., 1080
Main St., Pawtucket, Rhode Island, 1998, Section 3.6.

[15]Goyal, V.K., Johnson, E.R., Dvila, C.G., Jaunky, N., An Irreversible Constitutive Law for
Modeling the Delamination Process Using Interface Elements, Proceedings of The 43rd AIAA/
ASME/ASCE/AHS/ASC Structures, Structural Dynamics and Materials Conference, (Denver, CO,
April 22-25, 2002), American Institute of Aeronautics and Astronautics, Reston, VA, 2002,AIAA
Paper No. 2002-1576.

86

Appendix A

A.1 Strain Gage Information

To measure surface strains electrical resistance strain gages were used. The following
information covers the gages used.
Resistance: 350 0.3 %
Gage Factor at 24 degrees C: 2.080 0.5 %
Nominal dimensions of a strain gage: 0.5 by 0.25 inches

87

A.2 Specimen Strain Gage Patterns

a.2.1 1.5-inch Column

1(2)

3(4)

6(5)

9
8(7)

Fig A1. Axial strain gage locations for the 1.5-inch specimens.
All dimensions in inches.

88

a.2.2 Four-inch Column

1(2)

3(4)

6(5)

8(7)

Fig A2. Axial strain gage locations for the four-inch specimens.
All dimensions in inches.

89

a.2.3 Six-inch Column

13

6(5)

12
1(2)

3(4)

10(9)

8(7)
11

Fig A3. Axial strain gage locations for the six-inch specimens.
All dimensions in inches.

90

Appendix B

B.1 Sample ABAQUS Input File For Six-inch Column Model

b.1.1 hdmesh6.inp
*HEADING, SPARSE
High density mesh for delamination attemp.
8 inch column, 6 inch gage length specimen with potting.
Non Linear Analysis
6 inch gage length with potting
Incorporating 8 node interface elements in the outer flange.
Interface elements designed by Vinay Goyal, et al.
**
*RESTART, WRITE, FREQUENCY=10
**
** 3D space node definitions
**
*NODE
1,

-1.38500,

122.617,

2.85412,

91

2,

-1.38500,

122.617,

2.85412,

.
.
.

(Dots represent a lengthy node or element list.)

**
** Element sets are defined with nodes. From left to right:
**

element number, node 1, node 2, node 3, node 4

** Element normal is defined by node order, i.e. right hand rule

**
*ELEMENT, TYPE=S4R, ELSET=OUTER_FL
1665,

833,

836,

704,

702

1666,

836,

837,

706,

704

.
.
.

*ELEMENT, TYPE=S4R, ELSET=INNER_FL

5519,

2763,

2925,

2924,

2762

5520,

2925,

3071,

3070,

2924

.
.
.

*ELEMENT, TYPE=S4R, ELSET=WEB

2753,

2920,

3066,

3065,

2919

2754,

3066,

3194,

3192,

3065

92

.
.
.

***************************************************************
** INTERFACE

**

***************************************************************
*USER ELEMENT,TYPE=U1,NODES=8,COORDINATES=3, VARIABLES=49,
I PROPERTIES=2,PROPERTIES=8,UNSYMM
1,2,3
*USER SUBROUTINES,INPUT=IE3D8.for
**
**************************************************************
**

DEFINITON OF MATERIAL SETS

**

**************************************************************
**

INTERFACE MATERIAL PROPERTIES

**

**************************************************************
*UEL PROPERTY,ELSET=INTER1
** GIc | GIIc | GIIIc
4.60,

6.330,

6.330,

| SI

| SII

| SIII

| Kp |

NP

7000.0,13000,13000, 2.0e+10, 4.0,

** INTGS | IPOINTS
2,

**
**************************************************************
**
** Element sets are defined with nodes. From left to right:
** element number, node 1, node 2, node 3, node 4, node 5, node 5,
** node 7, node 8
** Element normal is defined by node order, i.e. right hand rule
**

93

*ELEMENT, TYPE=U1, ELSET=INTER1

8739,
951,
8740,
956,

1079,

1080,

954,

949,

945,

952,

1081,

959,

954,

952,

957,

947
1080,
951

.
.
.

*ELEMENT, TYPE=S4R, ELSET=OUTOUTFL

5699,

1172,

1293,

1291,

1170

5700,

1293,

1413,

1411,

1291

.
.
.

*ELEMENT, TYPE=S4R, ELSET=OUTERFLB

1,

2161,

2321,

2319,

2159

2,

2321,

2489,

2487,

2319

.
.
.

**
*SHELL SECTION,MATERIAL=AS4,ELSET=OUTER_FL
0.0885,

94

*SHELL SECTION,MATERIAL=AS4R,ELSET=INNER_FL
0.204,

*SHELL SECTION,MATERIAL=AS4,ELSET=WEB
0.159,

*SHELL SECTION,MATERIAL=AS4,ELSET=OUTOUTFL,OFFSET=-0.5
0.04425,

*SHELL SECTION,MATERIAL=AS4,ELSET=OUTERFLB,OFFSET=0.5
0.04425,

**
*MATERIAL,NAME=AS4
**
*ELASTIC,TYPE=ENGINEERING CONSTANTS
7.06E6,6.59E6,1.53E6,0.231,0.256,0.298,1.91E6,6.01E5,
6.45E5,
**
*MATERIAL,NAME=AS4R
**
*ELASTIC,TYPE=ENGINEERING CONSTANTS
5.8257E6,5.0865E6,1.53E6,0.231,0.256,0.298,1.91E6,6.01E5
6.45E5,
**
*NSET, NSET=FLG_POT, GENERATE
49,

50,

55,

62,

.
.
.

NSET, NSET=WEB_POT, GENERATE

970,

970,

95

972,

977,

.
.
.

*NSET, NSET=EDGE_POT
11,

14,

15,

16,

17,

18,

19,

20,

21,

22,

23,

24,

63,

68,

2920,

2925,

3066,

3071,

3194,

3199,

3320,

3325,

6082,

6083,

6084,

6085,

6209,

6339,

6468,

6594,

8055,

8167,

8183,

8184,

8185,

8186,

8303,

8833,

8848,

8849,

8850,

8851,

8852,

8867

*NSET, NSET=CLAMP
6086,

6214,

6342,

6470,

6596,

6718,

6846,

6972,

7095,

7216,

7341,

7464,

7585,

7704,

7827,

7950,

8069,

8182,

8298,

8411,

8417,

8518,

8523,

8615,

8619,

8702,

8705,

8771,

8773,

8816,

8817,

8860,

8861,

8862,

8863,

8864,

8865,

8866

*NSET, NSET=DISP
130,

136,

211,

216,

300,

304,

417,

420,

556,

558,

697,

698,

833,

840,

841,

842,

978,

980,

1095,

1097,

1212,

1214,

1329,

1446,

1565,

1687,

1822,

1959,

2098,

2251,

2418,

2587,

2758,

2759,

2760,

2761,

2762,

2763

**
*STEP,INC=600,NLGEOM
*STATIC
0.01,0.06,1.0e-10,
**
*BOUNDARY

96

CLAMP,ENCASTRE
DISP, 1,,

0.

DISP, 2,,

0.

DISP, 4,,

0.

DISP, 5,,

0.

DISP, 6,,

0.

WEB_POT, 1,,

0.

EDGE_POT, 1,,

0.

EDGE_POT, 2,,

0.

FLG_POT, 2,,

0.

**
**
*BOUNDARY
DISP, 3,,

-0.06

**
**
*CONTROLS, PARAMETERS=LINE SEARCH
6,,,,
*CONTROLS,PARAMETERS=TIME INCREMENTATION
35,35,,35,35, , , 9
0.40
**
*NODE PRINT,SUMMARY=NO,NSET=DISP
U,
*NODE PRINT,SUMMARY=NO,NSET=DISP,TOTALS=YES
RF,
*NODE FILE,NSET=DISP
U,
*NODE FILE,NSET=DISP
RF,
**
*END STEP

97

b.1.2 Material Reduction Scheme For Progressive Failure

Material definition section taken from an ABAQUS input deck. The three integers after the definition of material properties are the user defined field variables. Field variables are returned from
a user subroutine, the combination of 0s and 1s determines which line of material properties are
used.
**
*MATERIAL,NAME=AS4P
**
*ELASTIC,DEPENDENCIES=3,TYPE=ENGINEERING CONSTANTS
7.06E6,6.59E6,1.53E6,0.231,0.256,0.298,1.91E6,6.01E5,
6.45E5,0,0,0,
7.06E0,6.59E3,1.53E6,0,0,0,1.91E3,6.01E2,
6.45E2,1,0,0,
7.06E3,6.59E0,1.53E6,0,0,0,1.91E3,6.01E2,
6.45E2,0,1,0,
7.06E0,6.59E0,1.53E6,0,0,0,1.91E2,6.01E1,
6.45E1,1,1,0,
7.06E3,6.59E3,1.53E6,0,0,0,1.91E0,6.01E2,
6.45E2,0,0,1,
7.06E0,6.59E2,1.53E6,0,0,0,1.91E0,6.01E1,
6.45E1,1,0,1,
7.06E2,6.59E0,1.53E6,0,0,0,1.91E0,6.01E1,
6.45E1,0,1,1,
7.06E0,6.59E0,1.53E3,0,0,0,1.91E0,6.01E0,
6.45E1,1,1,1,
*DEPVAR
3,
*USER DEFINED FIELD
**

98

Vita

Daniel Constantine Hart was born on January 5, 1977 to Dennis and Anita Hart in New
Haven, CT. After attending high school in McLean, VA, Daniel attended Virginia Tech for
four years and received a B.S. in Ocean and Aerospace Engineering. While an undergraduate
student he worked at Newport News Shipbuilding as a cooperative education student. While
sailing his interest in boats and composite structures increased, one day hoping to engineer
and design sail boats. To further his education in composite structures Daniel obtained a Masters degree in Aerospace Engineering, with a concentration in Structures.

99