This book addresses the issue of designing the microstructure of

fiber composite materials in order to obtain optimum perfor-

mance. Besides the systematic treatment of conventional con-
tinuous and discontinuous fiber composites, the book also presents
the state-of-the-art of the development of textile structural com-
posites as well as the nonlinear elastic finite deformation theory of
flexible composites.
The author's experience during twenty years of research and
teaching on composite materials is reflected in the broad spectrum
of topics covered, including laminated composites, statistical
strength theories of continuous fiber composites, short fiber
composites, hybrid composites, two- and three-dimensional textile
structural composites and flexible composites. This book provides
the first comprehensive analysis and modeling of the thermo-
mechanical behavior of fiber composites with these distinct micro-
structures. Overall, the inter-relationships among the processing,
microstructures and properties of these materials are emphasized
throughout the book.
The book is intended as a text for graduate or advanced
undergraduate students, but will also be an excellent reference for
all materials scientists and engineers who are researching or
working with these materials.
Microstructural design of fiber composites

Cambridge Solid State Science Series

EDITORS:
Professor R. W. Cahn
Department of Materials Science and Metallurgy,
University of Cambridge
Professor E. A. Davis
Department of Physics, University of Leicester
Professor I. M. Ward
Department of Physics, University of Leeds
Titles in print in this series
B. W. Cherry Polymer surfaces
D. Hull An introduction to composite materials
S. W. S. McKeever Thermoluminescence of solids
D. P. Woodruff and T. A. Delchar Modern techniques of surf ace
science
C. N. R. Rao and J. Gopalakrishnan New directions in solid state
chemistry
P. L. Rossiter The electrical resistivity of metals and alloys
D. I. Bower and W. F. Maddams The vibrational spectroscopy of
polymers
S. Suresh Fatigue of materials
J. Zarzycki Glasses and the vitreous state
R. A. Street Hydrogenated amorphous silicon
A. M. Donald and A. H. Windle Liquid crystalline polymers
T.-W. Chou Microstructural design of fiber composites
TSU-WEI CHOU
Jerzy L. Nowinski Professor of Mechanical Engineering
University of Delaware

Microstructural
design of fiber
composites

The right of the

University of Cambridge
to print and sell

»m granted by
Henry VIII in 1534.

and published continuousl

since 1584.

CAMBRIDGE UNIVERSITY PRESS

Cambridge
New York Port Chester
Melbourne Sydney
CAMBRIDGE UNIVERSITY PRESS
Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, Sao Paulo

Cambridge University Press

The Edinburgh Building, Cambridge CB2 2RU, UK

Published in the United States of America by Cambridge University Press, New York

www.cambridge.org
Information on this title: www.cambridge.org/9780521354820
© Cambridge University Press 1992

This publication is in copyright. Subject to statutory exception

and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without
the written permission of Cambridge University Press.

First published 1992

This digitally printed first paperback version 2005

A catalogue recordfor this publication is available from the British Library

Library of Congress Cataloguing in Publication data

Chou, Tsu-Wei
Microstructural design of fiber composites/Tsu-Wei Chou.
p. cm.—(Cambridge solid state science series)
ISBN 0-521-3 5482-X
1. Fibrous composites. 2. Microstructure. I. Title.
I . Fibrouscomposites.
TA481.5.C48 1992
620.1'18—dc20 90-43347 CIP
ISBN-13 978-0-521-35482-0 hardback
ISBN-10 0-521-35482-X hardback

ISBN-13 978-0-521-01965-1 paperback

ISBN-10 0-521-01965-6 paperback
To Mei-Sheng, Helen, Vivian and Evan
 Contents

Preface xvii

1 Introduction 1
1.1 Evolution of engineering materials 1
1.2 Fiber composite materials 1
1.3 Why composites? 10
1.3.1 Economic aspect 10
1.3.2 Technological aspect 12
1.4 Trends and opportunities 17
1.5 Microstructure-performance relationships 19
1.5.1 Versatility in performance 20
1.5.2 Tailoring of performance 22
1.5.3 Intelligent composites 24
1.6 Concluding remarks 26

2 Thermoelastic behavior of laminated composites 29

2.1 Introduction 29
2.2 Elastic behavior of a composite lamina 30
2.2.1 Elastic constants 30
2.2.2 Constitutive relations 33
2.3 Elastic behavior of a composite laminate 39
2.3.1 Classical composite lamination theory 39
2.3.2 Geometrical arrangements of laminae 44
2.4 Thick laminates 46
2.4.1 Three-dimensional constitutive relations of a
composite lamina 46
2.4.2 Constitutive relations of thick laminated
composites 48
2.5 Thermal and hygroscopic behavior 54
2.5.1 Basic equations 55
2.5.1.1 Constitutive relations 55
2.5.1.2 Thermal and moisture diffusion
equations 61
2.5.2 Hygroscopic behavior 62
2.5.2.1 Moisture concentration functions 62
2.5.2.2 Hygroscopic stress field 65
 Contents

2.5.3 Transient interlaminar thermal stresses 67

2.5.3.1 Transient temperature field 67
2.5.3.2 Thermal stress field 68
2.5.4 Transient in-plane thermal stresses 73
2.5.4.1 Transient temperature field 74
2.5.4.2 Thermal stress field 77

3 Strength of continuous-fiber composites 80

3.1 Introduction 80
3.2 Rule-of-mixtures 81
3.3 Stress concentrations due to fiber breakages 85
3.3.1 Static case 85
3.3.1.1 Single filament failure 85
3.3.1.2 Multi-filament failure 88
3.3.2 Dynamic case 94
3.4 Statistical tensile strength theories 98
3.4.1 Preliminary 98
3.4.2 Strength of individual fibers 102
3.4.3 Strength of fiber bundles 104
3.4.4 Correlations between single fiber and fiber
bundle strengths 106
3.4.4.1 Analysis 106
3.4.4.2 Single fiber strength distribution 108
3.4.5 Experimental measurements of Weibull shape
parameter 109
3.4.5.1 Single fiber tests 110
3.4.5.2 Loose bundle tests 113
3.4.6 Strength of unidirectional fiber composites 115
3.4.6.1 Equal load sharing 115
3.4.6.2 Idealized local load sharing 118
3.4.6.3 Monte-Carlo simulation 132
3.4.7 Strength of cross-ply composites 134
3.4.7.1 Energy absorption during multiple
fracture 134
3.4.7.2 Transverse cracking of cross-ply
laminates 136
3.4.7.3 Statistical analysis 145
3.4.7.4 Transverse cracking and Monte-Carlo
simulation 150
3.4.8 Delamination in laminates of multi-directional
plies 157
 Contents X1

3.4.8.1 Free-edge delamination 160

3.4.8.2 General delamination problems 165
3.4.9 Enhancement of composite strength through
fiber prestressing 166

4 Short-fiber composites 169

4.1 Introduction 169
4.2 Load transfer 169
4.2.1 A single short fiber 170
4.2.2 Fiber-fiber interactions 171
4.3 Elastic properties 176
4.3.1 Unidirectionally aligned short-fiber composites 177
4.3.1.1 Shear-lag analysis 177
4.3.1.2 Self-consistent method 177
4.3.1.3 Bound approach 178
4.3.1.4 Halpin-Tsai equation 181
4.3.2 Partially aligned short-fiber composites 182
4.3.3 Random short-fiber composites 187
4.4 Physical properties 194
4.4.1 Thermal conductivity 194
4.4.2 Thermoelastic constants 198
4.5 Viscoelastic properties 200
4.6 Strength 201
4.6.1 Unidirectionally aligned short-fiber composites 202
4.6.1.1 Fiber length considerations 203
4.6.1.2 Probabilistic strength theory 205
4.6.2 Partially oriented short-fiber composites 216
4.6.3 Random short-fiber composites 224
4.7 Fracture behavior 227

5 Hybrid composites 231

5.1 Introduction 231
5.2 Stress concentrations 233
5.2.1 Static case 233
5.2.2 Dynamic case 243
5.3 Tensile stress-strain behavior 247
5.3.1 Elastic behavior 249
5.3.2 First cracking strain 251
5.3.3 Differential Poisson's effect 254
5.3.4 Differential thermal expansion 256
xii Contents

5.4 Strength theories 256

5.4.1 Rule-of-mixtures 257
5.4.2 Probabilistic initial failure strength 257
5.4.3 Probabilistic ultimate failure strength 262
5.5 Softening strips 273
5.6 Mechanical properties 275
5.7 Property optimization analysis 279
5.7.1 Constitutive relations 279
5.7.2 Graphical illustration of performance
optimization 282

6 Two-dimensional textile structural composites 285

6.1 Introduction 285
6.2 Textile preforms 287
6.2.1 Wovens 288
6.2.2 Knits 292
6.2.3 Braids 294
6.3 Methodology of analysis 300
6.4 Mosaic model 302
6.5 Crimp (fiber undulation) model 308
6.6 Bridging model and experimental confirmation 314
6.7 Analysis of the knee behavior and summary of
stiffness and strength modeling 319
6.8 In-plane thermal expansion and thermal bending
coefficients 327
6.9 Hybrid fabric composites: mosaic model 335
6.9.1 Definitions and idealizations 336
6.9.2 Bounds of stiffness and compliance constants 340
6.9.2.1 Iso-strain 341
6.9.2.2 Iso-stress 343
6.9.3 One-dimensional approximation 344
6.9.4 Numerical results 345
6.10 Hybrid fabric composites: crimp and bridging models 348
6.10.1 Crimp model 349
6.10.2 Bridging model 352
6.10.3 Numerical results and summary of thermoelas-
tic properties 354
6.11 Triaxial woven fabric composites 356
6.11.1 Geometrical characteristics 356
6.11.2 Analysis of thermoelastic behavior 358
6.11.3 Biaxial non-orthogonal woven fabric
composites 365
 Contents xiii

6.12 Nonlinear stress-strain behavior 366

6.13 Mechanical properties 368
6.13.1 Friction and wear behavior 368
6.13.2 Notched strength 371

7 Three-dimensional textile structural composites 374

7.1 Introduction 374
7.2 Processing of textile preforms 376
7.2.1 Braiding 377
7.2.1.1 2-step braiding 377
7.2.1.2 4-step braiding 379
7.2.1.3 Solid braiding 382
7.2.2 Weaving 382
7.2.2.1 Angle-interlock multi-layer weaving 383
7.2.2.2 Orthogonal weaving 387
7.2.3 Stitching 387
7.2.4 Knitting 389
7.3 Processing windows for 2-step braids 389
7.3.1 Packing of fibers and yarn cross-sections 390
7.3.2 Unit cell of the preform 395
7.3.3 Criterion for yarn jamming 398
7.4 Yarn packing in 4-step braids 402
7.4.1 Unit cell of the preform 402
7.4.2 Criterion for yarn jamming 403
7.5 Analysis of thermoelastic behavior of composites 405
7.5.1 Elastic strain-energy approach 406
7.5.2 Fiber inclination model 407
7.5.3 Macro-cell approach 414
7.5.3.1 Geometric relations 414
7.5.3.2 Elastic constants 416
7.6 Structure-performance maps of composites 419
7.7 Mechanical properties of composites 428
7.7.1 Tensile and compressive behavior 428
7.7.2 Shear behavior 431
7.7.3 Fracture behavior 432
7.7.3.1 In-plane fracture 432
7.7.3.2 Interlaminar fracture 435
7.7.4 Impact 440

8 Flexible composites 443

8.1 Introduction 443
8.2 Cord/rubber composites 445
xiv Contents

8.2.1 Rubber and cord properties 446

8.2.2 Unidirectional composites 447
8.2.3 Laminated composites 449
8.2.4 Cord loads in tires 453
8.3 Coated fabrics 456
8.4 Nonlinear elastic behavior - incremental analysis 459
8.4.1 Geometry of wavy fibers 460
8.4.2 Axial tensile behavior 462
8.4.2.1 Iso-phase model 462
8.4.2.2 Random-phase model 465
8.4.2.3 Nonlinear tensile stress-strain
behavior 467
8.4.3 Transverse tensile behavior 471
8.4.3.1 Iso-phase model 471
8.4.3.2 Random-phase model 472

9 Nonlinear elastic finite deformation of flexible

composites 474
9.1 Introduction 474
9.2 Background 478
9.2.1 Tensor notation 478
9.2.2 Lagrangian and Eulerian descriptions 480
9.3 Constitutive relations based on the Lagrangian
description 485
9.3.1 Finite deformation of a composite lamina 485
9.3.2 Constitutive equations for a composite lamina 487
9.3.2.1 Strain-energy function 487
9.3.2.2 General constitutive equations for a
unidirectional lamina 488
9.3.2.3 Pure homogeneous deformation 490
9.3.2.4 Simple shear 492
9.3.2.5 Simple shear superposed on simple
extension 495
9.3.3 Constitutive equations of flexible composite
laminates 499
9.3.3.1 Constitutive equations 499
9.3.3.2 Homogeneous deformation 500
9.3.3.3 Simple extension of a symmetric
composite laminate 502
9.3.4 Determination of elastic constants 505
9.3.4.1 Tensile properties 505
9.3.4.2 Shear properties 506
 Contents xv

9.4 Constitutive relations based on the Eulerian

description 508
9.4.1 Stress-energy function 509
9.4.2 General constitutive equations 511
9.4.3 Pure homogeneous deformation 514
9.4.4 Simple shear superposed on simple extension 515
9.4.5 Determination of elastic compliance constants 517
9.5 Elastic behavior of flexible composites reinforced with
wavy fibers 519
9.5.1 Introduction 519
9.5.2 Longitudinal elastic behavior based on the
Lagrangian approach 520
9.5.3 Longitudinal elastic behavior based on the
Eulerian approach 522

References 526

Author index 556

Subject index 563
 Preface

The science and technology of composite materials are based on a

design concept which is fundamentally different from that of
conventional structural materials. Metallic alloys, for instance,
generally exhibit a uniform field of material properties; hence, they
can be treated as homogeneous and isotropic. Fiber composites, on
the other hand, show a high degree of spacial variation in their
microstructures, resulting in non-uniform and anisotropic pro-
perties. Furthermore, metallic materials can be shaped into desired
geometries through secondary work (e.g. rolling, extrusion, etc.);
the macroscopic configuration and the microscopic structure of a
metallic component are related through the processing route it
undergoes. With fiber composites, the co-relationship between
microstructure and macroscopic configuration and their dependence
on processing technique are even stronger. As a result, composites
technology offers tremendous potential to design materials for
specific end uses at various levels of scale.
First, at the microscopic level, the internal structure of a
component can be controlled through processing. A classical
example is the molding of short-fiber composites, where fiber
orientation, fiber length and fiber distribution may be controlled to
yield the desired local properties. Other examples can be found in
the filament winding of continuous fibers, hybridization of fibers,
and textile structural forms based upon weaving, braiding, knitting,
etc. In all these cases, the desired local stiffness, strength, toughness
and other prespecified properties may be achieved by controlling
the fiber type, orientation, and volume fraction throughout the
structural component.
Second, the external geometrical shape of a structural component
can also be designed. Advances in the technology of filament
winding enable the automated production of components with
complex contours. It is now also feasible to fabricate three-
dimensional fiber preforms using advanced textile technology. As
the ability to fabricate larger and more integrated structural
components of net shape is further enhanced, the need to handle
and join a large number of small parts, as is currently done with
metallic materials, diminishes.
xviii Preface

The integrated and system approach, ranging from microstructure

to component net shape, offers almost unlimited opportunity in
composites processing and manufacturing. The figure below depicts
the interdependence of processing, microstructure, properties,
responses to external fields (physical, chemical and mechanical),
and performance of composites.

Processing Materials Analysis and

science science modeling Durability
Proc essing Microstructure Property Responses to
external fields
J
1\ Responses to
Microstructure crostructu external fields Performance

Design science

Optimization

The purpose of this book is to address the issue of designing the

microstructure of composites for optimum performance. This is
achieved through the selection of fiber and matrix materials as well
as the placement of both continuous and discontinuous fibers in
matrix materials. Continuous fibers can assume straight or wavy
shapes; they can also be hybridized or woven into textile preforms.
The wide range of microstructures available offers tremendous
versatility in the performance of composites; the ability to design
microstructures enables performance to be optimized.
The book is intended as an intermediate-level textbook for
students and a reference for research scientists and engineers.
Readers need some background and preparation in materials
science and applied mechanics. The first chapter examines the
driving forces for advances in fiber composites, as well as the trends
and opportunities of this rapidly developing field. Besides providing
a concise summary of the linear elastic laminate theory, Chapter 2
examines some of the recent developments in the mechanics of
laminated composites. Particular emphasis is given to thick lamin-
ates, hygrothermal effects and thermal transient effects. Chapter 3
treats the strength of continuous-fiber composites. Analyses of the
local load redistribution due to fiber breakages are presented first.
They are followed by a fairly comprehensive treatment of the
statistical tensile strength theories which encompasses the behavior
of individual fibers, fiber bundles, unidirectional fiber composites,
 Preface xix

cross-ply composites and laminates of multi-directional plies.

Various modes of failure of laminated composites are examined.
Section 3.4.6.2 is contributed by S. L. Phoenix, and Sections 3.4.7.4
and 3.4.8 are contributed by A. S. D. Wang. Chapter 4 deals with
the elastic, physical and viscoelastic properties as well as the
strength and fracture behavior of short-fiber composites. The effects
of variations in fiber length and orientation are examined using a
probabilistic approach. In Chapter 5, fiber hybridization serves as a
vivid example of how the performance of composites can be
controlled through the selection of material systems and their
geometric distributions. The synergistic effects between the com-
ponent phases with low elongation and high elongation fibers are of
particular interest. Chapter 6 is devoted to two-dimensional textile
structural composites based on woven, knitted and braided pre-
forms. A comprehensive treatment of the techniques for analyzing
and modeling the thermomechanical behavior of two-dimensional
textile composites is presented. Chaper 7 introduces recent de-
velopments in the processing of three-dimensional textile preforms
based on braiding, weaving, stitching and knitting. The processing-
microstructure relationship is demonstrated by the establishment of
'processing windows' for a specific forming technique. Then the
microstructure-property relationship is exemplified through the
construction of 'performance maps'. Mechanical properties of
polymer- and metal-based composites using three-dimensional tex-
tile preforms are reviewed. Chapters 8 and 9, in contrast to the
earlier chapters, treat the topic of finite elastic deformation of
flexible composites. The fundamental characteristics of flexible
composites and the technique for analyzing them are presented in
Chapter 8. A rigorous treatment of the constitutive relations of
flexible composites is developed in Chapter 9 based upon both the
Lagrangian and Eulerian descriptions of finite elastic deformation.
Overall, the inter-relationship among processing, microstructure,
property, responses to external fields, and performance of compos-
ites is emphasized throughout this text.
The contents of this book have evolved from my experience
during two decades of teaching and research of composite materials
at the University of Delaware. Stimulation from students and
co-workers was indispensable to the preparation of this book. The
contributions of the individuals with whom I had the privilege and
pleasure to interact are too numerous to cite here. However, this
book serves as a tribute to the intellectual achievements of them all.
The generous support provided by the National Science Founda-
xx Preface

tion, Department of Energy, Department of Transportation, Army

Research Office, Office of Naval Research, Naval Research Labo-
ratory, Air Force Office of Scientific Research, NASA, industrial
companies and the Center for Composite Materials of the Univers-
ity of Delaware for conducting the research reported in this book is
greatly appreciated. Ding-Guey Hwang, Shen-Yi Luo, Joon-Hyung
Byun and Wen-Shyong Kuo read the manuscript and gave critical
comments. Te-Pei Niu, Yih-Cherng Chiang, Mark Deshon and
Alison Gier provided valuable assistance in the preparation of the
manuscript.
Lastly, I should like to express my deep appreciation to the
following persons. The late Prof. Alan S. Tetelman of Stanford
University first pointed out to me the technological potential of
fiber composites. As a colleague of mine at Delaware, Prof. R.
Byron Pipes has greatly enriched my perspective on the subject
matter. The scholarship and guidance of Prof. Anthony Kelly have
always been a source of inspiration to me. Prof. Jerzy L. Nowinski
encouraged me throughout the course of this endeavor.
 Introduction

1.1 Evolution of engineering materials

Compared to the evolution of metals, polymers and ceramics,
the advancement of fiber composite materials is relatively recent.
Ashby (1987) presented a perspective on advanced materials and
described the evolution of materials for mechanical and civil
engineering. The relative importance of four classes of materials
(metal, polymer, ceramic and composite) is shown in Fig. 1.1 as a
function of time. Before 2000 BC, metals played almost no role as
engineering materials; engineering (housing, boats, weapons, uten-
sils) was dominated by polymers (wood, straw, skins), composites
(like straw bricks) and ceramics (stone, flint, pottery and, later,
glass). Around 1500 BC, the consumption of bronze might reflect the
dominance in world power and, still later, iron. Steel gained its
prominence around 1850, and metals have dominated engineering
design ever since. However, in the past two decades, other classes
of materials, including high strength polymers, ceramics, and
structural composites, have been gaining increasing technological
importance. The growth rate of carbon-fiber composites is at about
30% per year - the sort of growth rate enjoyed by steel at the peak
of the Industrial Revolution. According to Ashby the new materials
offer new and exciting possibilities for the designer and the potential
for new products.

1.2 Fiber composite materials

Fiber composites are hybrid materials of which the com-
position and internal architecture are varied in a controlled manner
in order to match their performance to the most demanding
structural or non-structural roles. The fundamental characteristics
of fiber composites have been summarized by Vinson and Chou
(1975), Chou and Kelly (1976), Chou, Kelly and Okura (1985),
Kelly (1985), and more recently by Chou, McCullough and Pipes
(1986), from which the following is excerpted.*
On the face of it a composite might seem a case of needless
complexity. The makings of ideal structural materials would appear
* From 'Composites', Chou, McCullough and Pipes. Copyright © (1986) by
Scientific American, Inc. All rights reserved.
 Fig. 1.1. The evolution of materials for mechanical and civil engineering. (After Ashby 1987.)
10 000BC 2020

Glassy metals
Al—Li alloys Development slow
Dual phase steels mostly quality
Microalloyed steels control and
New super-alloys processing
Wood
Skins
Fibers
Light alloys

Composites | Super-alloys Conducting

polymers
Straw bricks Paper High-temperature
Titanium "j
Zirconium > Alloys polymers
etc. J High-modulus

Ceramic composites
Polyesters ./'Metal matrix
Epoxies ^ r composites
PMMA Acrylics ^ K e v l a r - F R P
PC PS PP ^XCFRP Jf I Ceramics
FRP_
pyro_ Tough engineering
Cermets£)£(> ceramics (A12O3, Si 3 N 4 , PSZ etc.
I I
10 000BC 5000 BC 100011500 1800 2010 i 2020
Year
 Fiber composite materials 3

to be at hand, in the midsection of the periodic table. Those

elements, among them carbon, aluminum, silicon, nitrogen and
oxygen, form compounds in which the atoms are joined by strong
and stable bonds. As a result, such compounds, typified by the
ceramics, for instance, aluminum oxide, silicon carbide and silicon
dioxide, are strong, stiff and resistant to heat and chemical attack.
Their density is low and furthermore their constituent elements are
abundant.
Yet because of a serious handicap these substances have rarely
served as structural materials. They are brittle and susceptible to
cracks. In bulk form the substance is unlikely to be free of small
flaws, or to remain free of them for long in actual use. When such a
material is produced in the form of fine fibers, its useful strength is
greatly increased. The remarkable increase in strength at small
scales is in part a statistical phenomenon. If one fiber in an
assemblage does fail, moreover, the crack cannot propagate further
and the other fibers remain intact. In a similar amount of the bulk
material, in contrast, the initial crack might have led to complete
fracture.
Tiny needlelike structures called whiskers, made of substances
such as silicon carbide and aluminum oxide, also contain fewer flaws
and show greater strength than the material in bulk form. Whiskers
are less likely to contain defects than the bulk material, not only for
statistical reasons but also because they are produced as single
crystals that have a theoretically perfect geometry. The notion that
many materials perform best as fibers also holds for certain organic
polymers. Composites are a strategy for producing advanced
materials that take advantage of the enhanced properties of fibers.
A bundle of fibers has little structural value. To harness their
strength in a practical material the designer of a composite embeds
them in a matrix of another material. The matrix acts as an
adhesive, binding the fibers and lending solidity to the material. It
also protects the fibers from environmental stress and physical
damage that could initiate cracks.
The strength and stiffness of the composite remain very much a
function of the reinforcing material, but the matrix makes its own
contribution to properties. The ability of the composite material to
conduct heat and current, for example, is heavily influenced by the
conductivity of the matrix. The mechanical behavior of the compos-
ite is also governed not by the fibers alone but by a synergy between
the fibers and the matrix.
The ultimate tensile strength of a composite is a product of the
4 Introduction

synergy. When a bundle of fibers without a surrounding matrix is

stressed, the failure of a single fiber eliminates it as a load carrier.
The stress it had borne shifts to the remaining intact fibers, moving
them closer to failure. If the fibers are embedded in a matrix, on the
other hand, fracture does not end the mechanical function of a
fiber. The reason is that as the broken ends of the fiber pull apart,
elastic deformation or plastic flow of the matrix exerts shear forces,
gradually building stress back into the fragments. Because of such
load transfer the fiber continues to contribute some reinforcement
to the composite. The stress on the surrounding intact fibers
increases less than it would in the absence of the matrix, and the
composite is able to bear more stress without fracturing. The
synergy of the fibers and the matrix can thus strengthen the
composite and also toughen it, by increasing the amount of work
needed to fracture it.
Although the general requirement that the matrix be ductile
provides some guidance for choosing a matrix material, the most
common determinant of the choice is the range of temperatures the
composite will face in its intended use. Composites exposed to
temperatures of no more than between 100 and 200°C usually have
a matrix of polymer. Most composites belong to this group.
Polymer matrices are often thermosets, that is polymers in which
bonds between the polymer chains lock the molecular structure
into a rigid three-dimensional network that cannot be melted.
Thermosets resist heat better than most thermoplastics, the other
class of polymeric materials, which melt when they are heated
because no bonds cross-link the polymer chains. Epoxies are the
most common thermosetting matrix for high-performance compos-
ites, but a class of resins called polyimides, which can survive
continuous exposure to temperatures of more than 300°C, have
attracted considerable interest. If the resin is a thermoset, the
structure must then be cured, subjected to conditions that enable the
polymer chains to cross-link. Often the composite must be held at
high temperature and pressure for many hours.
In part to shorten the processing time, thermoplastic matrix
materials are attracting growing interest; one promising example is a
polymer called PEEK (polyetheretherketone). Consolidating a
composite that has a thermoplastic matrix requires only relatively
short exposure to a temperature that is sufficient to soften the
plastic. The melting temperature of some thermoplastic matrices is
so high that they rival thermosets in heat resistance: PEEK, for
example, melts at 334°C. Thermoplastics have the additional
advantage of being tougher than most of the thermosets.
 Fiber composite materials 5

Temperatures high enough to melt or degrade a polymer matrix

call for another kind of matrix material, often a metal. Along with
temperature resistance a metal matrix offers other benefits. Its
higher strength supplements that of the reinforcing fibers, and its
ductility lends toughness to the composite. A metal matrix exacts
two prices: density that is high in comparison with polymers, even
though the light metals such as aluminum, magnesium and titanium
are the most common matrices, and complexity of processing.
Indeed, whereas the production of many advanced polymer matrix
composites has become routine, the development of metal matrix
composites has progressed more slowly, in part because of the
extreme processing conditions needed to surround high strength
fibers with a matrix of metal.
Metal matrix composites might assume a place in the cooler parts
of the skin of a hypersonic aircraft, but at the nose, on leading
edges of the wings and in the engines temperatures could exceed the
melting point of a metal matrix. For those environments, there is
growing interest in a class of composites that have matrices as
resistant to heat as the fibers themselves, and also as lightweight
and potentially as strong and stiff, namely, ceramics. Because they
are brittle, ceramics behave differently from other matrices. In
metal and polymer matrix composites the fibers supply most of the
strength, and the ductile matrix acts to toughen the system. A
ceramic matrix, in contrast, is already abundantly stiff and strong,
but to realize its full potential it needs toughening. The fibers in a
ceramic matrix composite fill that need by blocking the growth of
cracks. A growing crack that encounters a fiber may be deflected or
may pull the fiber from the matrix. Both processes absorb energy.
The ceramic matrix gives such composites great temperature
resistance. Borosilicate glass reinforced with carbon fibers retains its
strength at 600°C. Such matrices as silicon carbide, silicon nitride,
aluminum oxide or mullite (a complex compound of aluminum,
silicon and oxygen) yield composites that remain serviceable at
temperatures well above 1000°C. The heat resistance of a ceramic
matrix composite, however, complicates its fabrication.
The characteristics of these three classes of composites can be
exemplified by the relation of stress and strain for the unreinforced
polymer, metal and ceramic as compared with curves for the
corresponding composites. Whereas unreinforced epoxy stretches
easily, an epoxy matrix composite containing 50% by volume of
silicon carbide fibers is far stiffer (Fig. 1.2a). In an aluminum matrix
the same volume of reinforcement, in this case aluminum oxide
fibers, also improves stiffness dramatically (Fig. 1.2b). Because the
 Introduction

Fig. 1.2. Stress-strain curves for (a) SiC/epoxy, (b) A12O3/aluminum, and
(c) SiC/borosilicate glass composites. (From 'Composites,' Chou,
McCullough and Pipes). Copyright © (1986) by Scientific American, Inc.
All rights reserved.
1000

800 -

600 -
a.
2

0.2 0.4 0.6 0.8 1.0

Strain (%)
(a)

800

600 1
1
Stress

400 - /^ AliOj/aluminum

/
200 Aluminum ^ ^ ^ _ _

n 1 1 1 1
0.0 0.2 0.4 0.6 0.8 1.0
Strain (%)
(b)
 Fiber composite materials 7

fibers are brittle, the composite fails at a much lower strain than
unreinforced aluminum does. A similar fraction of silicon carbide
fibers stiffens a matrix of borosilicate glass only slightly but
toughens it considerably, increasing the percentage by which it can
be strained without breaking (Fig. 1.2c). The fibers do so by
restraining the growth of matrix cracks that might otherwise lead to
fracture.
Related to ceramic matrix composites in character but distinctive
in manufacture is a composite in which both the matrix and the
reinforcing fibers consist of elemental carbon. Carbon-carbon
composite is reinforced by the element in a semicrystalline form,
graphite; in the matrix the carbon is mostly amorphous. A carbon-
carbon composite retains much of its strength at 2500°C and is used
in the nose cones and heat shields of re-entry vehicles. Unlike most
ceramic matrix composites, it is vulnerable to oxidation at high
temperatures. A thin layer of ceramic is often applied to the surface
of a carbon-carbon composite to protect it.
The combination of fiber and matrix gives rise to an additional
constituent in composites: an interface (or interphase) region. Chemi-
cal compatibility between the fibers and the matrix is most crucial at
this region. In polymer and metal matrix composites a bond must
develop between the reinforcement and the matrix if they are to act

Fig. 1.2. (cont.)

1000

800

SiC/borosilicate glass

0.2 0.4 0.6 0.8 1.0

Strain (%)
(c)
8 Introduction

in concert. A prerequisite for adhesion is that the matrix, in its fluid

form, be capable of wetting the fibers. Fibers that would otherwise
not be wetted by their matrix can be given a coating that fosters
contact by interacting with both the fibers and the matrix. In some
cases varying the matrix composition can also promote the process.
Once the matrix has wetted the fibers thoroughly, intermolecular
forces or chemical reactions can establish a bond.
The properties of an advanced composite are shaped not only by
the kind of matrix and reinforcing materials it contains but also by a
factor that is distinct from composition: the geometry of the
reinforcement. Reinforcing geometries of composites can be
grouped roughly by the shape of the reinforcing elements: particles,
continuous fibers or short fibers (Fig. 1.3). Sets of parallel con-
tinuous fibers are often embedded in thin composite layers, which
are assembled into a laminate. Alternatively, each ply in a laminate
can be reinforced with continuous fibers woven or knitted into a
textile 'preform'. Recently developed geometries dispense with
lamination: the fibers are woven or braided in three dimensions
(Fig. 1.4), a strategy that in some cases enables the final shape of
the composite to be formed directly.
Progress toward managing the many variables of composite
design has encouraged investigators to contemplate new com-
plexities. An ordinary composite reinforced with stiff, straight fibers
usually displays a nearly constant value of stiffness. New composites
designed to display specific non-linear relations of strain and stress
are now attracting interest. One such example, a flexible composite
consisting of undulating fibers in an elastomeric matrix, can

Fig. 1.3. Particle- and fiber-reinforced composites. (From 'Composites'

Chou, McCullough and Pipes.) Copyright © (1986) by Scientific
American, Inc. All rights reserved.

Particles Short fibers

Continuous fibers
 Fig. 1.4. Preforms of textile structural composites. (From 'Composites' Chou, McCullough and Pipes )
Copyright © (1986) by Scientific American, Inc. All rights reserved.

Bi-axial weave Tri-axial weave Knit Multi-axial multi-layer warp knit

Three-dimensional Three-dimensional braiding Three-dimensional Angle-interlock construction

cylindrical construction orthogonal fabric
10 Introduction

elongate readily at low stresses but stiffens when the fibers become
fully extended. A hybrid composite strengthened with two kinds of
fibers, some of them brittle and inextensible and the others ductile
and tough, can display the opposite behavior. The stiff fibers cause
stress to increase very sharply at low strains, but when the strain is
sufficient to break the stiff, brittle fibers, the curve of stress over
strain flattens. The ductile fibers come into play, and as a result the
composite becomes more extensible. The hybrid design can yield a
material that combines much of the stiffness of an ordinary
composite containing only stiff fibers with increased toughness.
Overall, the opportunity in the engineering of fiber composites is
the potential to control the composition as well as internal geometry
of the materials for optimized performance.

1.3 Why composites?

The question of 'Why composites?' was raised in the 1975
text by Vinson and Chou (1975). The rationale provided then
focussed on
(a) the limitations in strength and ductility for metallic alloys
from the viewpoints of theoretical cohesive strength of
solids and the arrangement of crystalline defects,
(b) the need of a balanced pursuit in strength and ductility and
the potential of achieving both in fiber composites, and
(c) the strength limitation of metallic alloys at elevated tem-
peratures and the potential of carbon-carbon composites
and refractory metal wire reinforced super-alloys.
The field of fiber composites has witnessed drastic changes and
advancement since the mid-1970s because of the availability of
several ceramic fibers, high-temperature thermoplastics, glass-
ceramic matrices, and intermetallic solids for composites. Although
the fundamental physical principles governing the synergism of the
component phases in composites should not change, the advance-
ment in materials technology coupled with that in processing,
surface science and instrumentation has greatly changed the per-
spective of composite technology. In the following, the answer to
the question of 'why composites?' is re-examined from both
economic and technological points of view.

1.3.1 Economic aspect

For the discussion of the economic aspect of advanced
materials in general and fiber composites in particular, it is
 Why composites? 11

worthwhile referring to a recent survey entitled Problems and

Opportunities in Metals and Materials: An Integrated Perspective by
the U.S. Department of the Interior (Sousa 1988). The report
asserts that the future growth prospects seem best not in tonnage
commodities but rather in materials that are more technology-
intensive and more high-value-added. As the economy grows and
matures, the rate of growth in consumption of tonnage metals first
exceeds, eventually parallels, and finally trails that of the economy
as a whole.
Figure 1.5 shows the estimated current relative market maturity
of the major metals and other materials. The vertical dimension
indicates intensity-of-use (amount/GNP). The potential of polymer,
metal and ceramic based composites is obvious. This figure also
demonstrates a hard fact of life that eventually catches up with
virtually any product-that of market saturation and, as the
inexorable evolution of technology proceeds, eventual displacement
and decline.
By incorporating different materials into composites, the synthe-
tic class of materials can thus draw on the essential characteristics of
diverse materials: the high strength, ductility, thermal-electrical
conductivity and formability of metals, the low cost fabrication,
light weight and corrosion resistance of polymers, and the strength,
corrosion resistance and high-temperature performance of ceramics.

Fig. 1.5. Relative market maturity of materials. (After Sousa (1988).)

Commodity plastics - Aluminum

Stainless steel Copper
Super-alloy - Carbon steel
Specialty metals
Traditional engineering plastics -
High-performance engineering plastics -
Engineering plastics, alloys and blends —
Fiber optics -
Advanced polymer .
matrix composites
Advanced metal
matrix composites
Structural ceramics -

Growth Growth
maturing <GNP
Growth
= GNP
12 Introduction

The survey of the U.S. Department of the Interior forecasts the

total demand for advanced materials in the U.S. in the year 2000 to
be approximately $55 billion annually, roughly the same magnitude
as the current U.S. steel market. By comparison, a Japanese
Ministry of International Trade and Industry report showed that the
Japanese annual demand for advanced materials is expected to be
about $34 billion. The breakdown of the market in terms of
material categories is (1) advanced polymer composites: 22%
(U.S.), 7.6% (Japan); (2) advanced metal alloys and composites:
35% (U.S.), 28.3% (Japan); (3) advanced ceramics: 30% (U.S.),
35.9% (Japan); (4) engineering plastics: 13% (U.S.), 28.3%
(Japan). Although the rudimentary nature of such forecasts cannot
be overemphasized, the transition from a metals economy to a
materials economy, and the importance of composite materials to
the economy of advanced materials, is unmistakable.

1.3.2 Technological aspect

From the technological viewpoint, advanced composite
materials can offer a competitive edge in many products, including
aircraft, automobile, industrial machinery and sporting goods,
provided their overall production costs can be reduced and their
performance improved. According to the study New Structural
Materials Technologies made by the Congress of the United States,
Office of Technology Assessment (1988), the broader use of
advanced structural materials requires not only solutions to techni-
cal problems but also changes in attitudes among researchers and
end-users. The traditional approach based upon discrete design and
manufacturing steps for conventional structural materials needs to
be replaced by an integrated design and manufacturing process
which necessitates a closer relationship among researchers, design-
ers, and production personnel as well as a new approach to the
concept of material costs. A fully integrated design process capable
of balancing all of the relevant design and manufacturing variables
requires an extensive database on matrix and fiber properties,
the ability to model fabrication processes, and three-dimensional
analysis of the properties and behavior of the resulting structure.
Knowledge of the relationships among the constituent properties,
microstructure and macroscopic behavior of the composite is basic
to the development of an integrated design methodology.
To further understand the impacts of advanced structural mate-
rials on manufacturing, this report examines the following two
possibilities: substitution by direct replacement of metal com-
 Why composites'? 13

ponents in existing products and the use in new products that are
made possible by the new materials. Direct substitution of a ceramic
or composite part for a metal part is not likely to take full
advantage of the superior properties and design flexibility of
advanced materials. Substitution of conventional structural metals
such as steel and aluminum alloy by composites is highly unlikely.
Because of their low cost and manufacturability, these metals are
ideally suited for applications in which they are now used. On the
other hand, the metal industry has responded to the potential of
direct substitution by developing new alloys with improved pro-
perties, such as high-strength, low-alloy steel and aluminum-
lithium. According to this assessment, significant displacement of
metals could occur in four potential markets: aircraft, automobiles,
containers and constructions.
In the choice of material substitution, a variety of factors need to
be taken into account. Compton and Gjostein (1986) analyzed the
weight saving and cost for material substitution for ground trans-
portation. Weight reduction that can be achieved in designing a part
by substituting a light-weight material for a conventional one
depends critically on the part's function. A unit volume of cast
aluminum weighs 63% less than an equal volume of cast iron. Cast
iron, however, is stiffer than cast aluminum. Therefore if a
hypothetical cast-aluminum part is to be as stiff as a cast-iron one,
more aluminum would have to be used and the weight saving would
be reduced to 11%. If equal loading carrying capacity is required in
the hypothetical aluminum part, the weight saving would be 56%.
(In actual design situations the weight saving offered by the
substitution of aluminum for cast iron ranges from 35 to 60%.)
Similarly, aluminum and fiber-reinforced plastics are much lighter
than mild (ordinary) steel by volume. The weight savings, however,
are much smaller if equal stiffness or equal collapse load and
bending stiffness (a measure of structural strength) is needed.
High-strength steel is no lighter by volume than mild steel, nor is it
stiffer. Where structural strength is the main concern, however,
high-strength steel does offer a weight saving: 18% in the example
discussed by Compton and Gjostein.
In terms of innovative designs and new products based upon
advanced composites, the automotive industry undoubtedly pro-
vides an excellent paradigm. The use of polymer matrix composites
for primary body structures and chassis/suspension systems is under
evaluation by the major automobile manufacturers. The potential
advantages of using composites are: weight reduction and resulting
14 Introduction

fuel economy; improved overall quality and consistency in manu-

facturing; lower assembly costs due to parts consolidation; lower
investment costs for plant, facilities, and tooling; improved corro-
sion resistance; and lower operating costs. The major barriers to the
large-scale applications of composites are the lack of high-speed,
high-quality, low-cost manufacturing processes; uncertainties re-
garding crash integrity and long-term durability; and lack of
adequate technologies for repair and recycling of polymer compos-
ite structures. According to Compton and Gjostein, glass fiber
reinforced composites are capable of meeting the functional re-
quirements of the most highly loaded automotive structures. Candi-
date fabrication methods include resin transfer molding, compres-
sion molding, and filament winding. Among these methods, resin
transfer molding seems the most promising, although none of these
methods can satisfy all of the production requirements at this time.
There is no doubt that the large-scale adoption of polymer matrix
composites for automotive structures would have a major tech-
nological impact on the fabrication and assembly of automobiles.

Fig. 1.6. Temperature capabilities of polymer, metal and ceramic matrix

materials. (After Mody and Majidi 1987, with permission from the Society
of Manufacturing Engineers.)

1
Polymer matrix Metal matrix Ceramic matrix
composites composites composites

>2000-

u
1500 I
1200
1000

600

300

£ £ E ^
X
3 u
Ra E•= S .a .a .a
T3
8. "55 D 3 3 O
smal eim
Pol)/imi

-
3OI1
cd
"7-1 E <
the rmoi

-5 ^ H g. ^3 ™ —
S c^ =

Z
 Why composites? 15

Another technological aspect that motivates the use of fiber

composites pertains to the demand of an elevated temperature en-
vironment (Steinberg 1986). Temperature capabilities of polymer,
metal and ceramic matrix materials are shown in Fig. 1.6 (Mody
and Majidi 1987). The demand for high-temperature applications of
composites is best exemplified by the need for aerospace materials.
The U.S. goals for subsonic, supersonic and hypersonic flight and
for space explorations require alloys and composites with superior
strength, light weight and resistance to heat. According to Stein-
berg, the evolution of aircraft has required continual improvements
in materials because increased speed raises the heating of the skin
from friction with the air and increased power raises the tempera-
ture of the engine. Figure 1.7 shows the changes in skin tempera-
tures from aircraft of the 1930s to the proposed Orient Express
which is a transatmospheric craft capable of cruising at great speed
in space. The skin materials have progressed from wood and fabric
to advanced alloys of aluminum, nickel and titanium and graphite
fiber reinforced polymer composites.
Figure 1.8 shows the changes in engine temperature from engines
cooled by water to those of scramjets. The need for composites in
engine components can be understood from the evolution in engine

Fig. 1.7. Evolution of aircraft skin temperatures. (From 'Materials for

Aerospace', Steinberg). Copyright © (1986) by Scientific American, Inc.
All rights reserved.

Skin temperatures

1930s trainer
50°C

r—
Second World
War fighter 90°C

980s interceptor 1430°C

Space Shuttle
1 090°C

Orient Express
1650°C
1 1 1 1
500 1000 1500 2000 2500
Temperature (°C)
16 Introduction

performance. According to Steinberg, the thrust delivered by a big

jet engine for transport and cargo aircraft has increased about six
fold over the past 30 years, approaching 294 000 newtons
(66 000 pounds) now. During the same period the weight of the
engine has increased by a factor of only two or three. The
thrust-to-weight ratio of the military aircraft may approach 15:1 by
the year 2000. The performance of jet engines has been made
possible partially with improvements in turbine blades. It is
predicted that with the further improvements in blades and other
aspects of aircraft propulsion, a typical propulsion system in the
year 2000 will be likely to contain about 20% each of composites,
steel, nickel and aluminum, 15% titanium, 2% ordered alloys
(aluminides, e.g. titanium-aluminum or nickel-aluminum) and 1%
ceramics (Steinberg 1986).
Clark and Flemings (1986) have also examined the present and
future material systems for meeting the engine operating tempera-
ture requirements. In Fig. 1.9 the lowest band on the graph
indicates the temperature increase that has been achieved so far
through improvements in nickel-based super-alloys, the standard
turbine material. It is believed that in the coming decades alloy
turbine blades made of metal strengthened by directional crystal
structures, and blades protected by a coating of ceramics or special

Fig. 1.8. Evolution of aircraft engine temperatures. (From 'Materials for

Aerospace' Steinberg). Copyright © (1986) by Scientific American, Inc.
All rights reserved.

Engine temperatures
First World War
water-cooled
150°C

1930s air-cooled
<00°C

Fan-jet
1
Turbojet I 1090°C
Scramjet
••j;^j" -~ ;<m
11930°C
1 1 1 1
500 1000 1500 2000 2500
Temperature (°C)
 Trends and opportunities 17

alloys, will allow an increase in turbine-inlet temperatures. How-

ever, ultimately, the demand for very-high-temperature material
can only be met by ceramic matrix composites and carbon-carbon
composites.

1.4 Trends and opportunities

Kelly (1987a&b), in a recent outline of the trends in
materials science and processing, examined the status of fiber
composites. It was concluded that the development of this field has
been mainly driven by the aerospace industry. This development
has contributed to the growth of a relatively small body of new
science which related the colligative properties of fiber composites
to the properties of the individual components. There have been
interesting combinations of properties not hitherto available in
single phase materials, for example, a negative thermal expansion
and a negative Poisson's ratio. However, there have not been large
non-linear synergistic effects. There is perhaps not much new
science of the colligative properties of composites. However, in
Kelly's view, the studies of design of fiber composites are critical for

Fig. 1.9. Rise in the operating temperature of jet engines with time.
(From 'Advanced Materials and Economy' Clark and Flemings). Copy-
right © (1986) by Scientific American, Inc. All rights reserved.
1600

1500 Ceramic matrix

composites

Directional
metal structures
2L
o

Conventionally cast
super-alloys

1970 1980 1990 2000 2010

Year
18 Introduction

their applications. Furthermore, there may be much new science on

how to produce composites.
The significant trends in structural composites point to the
direction of low-temperature metal matrix, resin matrix, metal-
resin matrix, rubber matrix, cement-ceramic matrix and elevated-
temperature composites. Non-structural composites are increasingly
being recognized for their unique opportunities in electric, mag-
netic, superconducting and biomedical applications. A brief sum-
mary of those trends follows (see Kelly 1987a).
A major motivation behind the development of low-temperature
metal matrix composites in the U.S. has been for the utilization of
high-stiffness continuous fibers in a matrix material without the
disadvantage of thermosetting resins of low thermal conductivity,
high thermal expansion, dimensional instability, hygrothermal de-
gradation, material loss in high vacuum, susceptibility to radiation
damage, and lower temperature brittleness. The lighter metals do
not possess these disadvantages; their low atomic number (Z) is
important in a neutron-rich environment. It is useful to bear in
mind that five out of the 13 lowest-Z solids are metals. Some of
these metals, together with their atomic number and density, are
listed below: lithium (Z = 3, density = 0.53 Mgm ~3 ), sodium
(11,0.97), potassium (19,0.86), calcium (20,1.55), magnesium
(12,1.741), beryllium (4,1.85), and aluminum (13, 2.7).
Reinforcement of a light metal, e.g. aluminum and magnesium, is
attractive in the automobile industry in reducing creep at moderate
temperatures and improving wear resistance. Coating for carbon fibers
is necessary for incorporation into aluminum and magnesium matrices.
Thermoplastic resins have certain advantages over thermosets in
their infinite shelf life, good resistance to water and solvents, and
ductility. Thermoplastics are attractive particularly from the view-
point of composites manufacture because they are rapidly proces-
sable, and are better adapted to automated manufacturing. Also,
they can be recycled and joined by welding.
Laminates formed by bonding metal sheets to fiber-resin com-
posites take advantage of the synergistic effects of hybrid compos-
ites. For instance, the combination of aluminum foil with
Kevlar/epoxy composite results in enhanced fatigue resistance and
compressive strength.
Rubber (elastomeric) matrix can be reinforced with short and
continuous fibers and can provide the capability of large non-linear
elastic deformation. Automobile tires and coated fabrics are ex-
amples in this category.
 Microstructure -performance relationships 19

Contrary to the large deformation of rubber type flexible com-

posites, ceramic based composites offer the other extreme on the
scale of deformation. The brittle nature of ceramic solids requires a
new way of thinking in 'reinforcement'. Fibers are added for the
purpose of improving toughness against fracture and ductility in
terms of energy absorption and deformation range.
Ceramic matrix composites, directionally solidified eutectics,
intermetallic solids, certain types of metal based composites, and
carbon-carbon composites are the candidate materials for elevated-
temperature applications. Among these, carbon-carbon composites
present the ultimate in high-temperature materials under reducing
conditions. They have many tribological applications. Protection
against oxidation and densification of the matrix are major chal-
lenges to carbon-carbon composites.
Finally, the potential of non-structural composites has not been
fully explored. Kelly (1987a) cited the examples in making special
devices. For example, a magnetoresistive device obtained by
coupling a metal rod with a semiconductor matrix provides a
contactless potentiometer or a fluxmeter, or coupling a piezoelectric
and magnetostrictive material gives a magnetoelectric material. The
potential for biomedical applications of flexible composites also
exists (see Chou 1989).

1.5 Microstructure-performance relationships

Chapters 2-9 examine the stiffness, strength and failure
behavior of several types of composites: laminated composites
composed of continuous fibers; composites reinforced with short
fibers in biassed or random orientations; composites with two types
of fibers in intermingled, interlaminated or interwoven forms;
composites reinforced with textile preforms; and flexible composites
exhibiting large deformations. The mathematical tools for analyzing
their thermomechanical properties have been presented. Most
significantly, an effort has been made to delineate the relationship
between the behavior and these composites.
In the following, a comparison is first made among the stress-
strain behaviors of three composite systems. The purpose is to
demonstrate the versatility in composite performance through the
design of microstructure. This is followed by specific examples
of tailoring the material performance through microstructural
design. Lastly, the emerging field of 'intelligent composites' is
introduced.
20 Introduction

1.5.1 Versatility in performance

For the purpose of demonstrating the versatility of the
performance of composites, the stress-strain relationships of three
types of composites are examined. Figure 1.10 shows the stress-
strain curves of a unidirectional carbon fiber reinforced glass matrix
composite (Nardone and Prewo 1988). The behavior is typical for
brittle matrix composites based upon polymer and glass/ceramic
matrices. The knee phenomenon of the stress-strain curve re-
sembles the yield behavior of metallic alloys.
Figure 1.11 gives the stress-strain curves of interlaminated
carbon/glass hybrid composites. The ability of the low elongation
phase (carbon) in developing multiple fractures enables the hybrid
composites to sustain deformations at a level much higher than that
of the all-low elongation fiber composite. The energy absorption
capability as indicated by the area under the stress-strain curve is
also much higher than that of the all-carbon fiber composite. The
shape of this stress-strain curve resembles those of ductile metals
with strain-hardening behavior.
The stress-strain data of a flexible composite (Fig. 1.12) show
rapid increase in stress and stiffness at large deformation (Chou
1989). It resembles the behavior of certain biological materials such
as soft animal tissues (Humphrey and Yin 1987; Gordon 1988).
Fig. 1.10. Tensile stress—strain curves of a carbon/borosilicate glass
composite. (After Nardone and Prewo 1988).

130

- Transverse
strain yf 110
700 -
- 90

500 - 70
\ ^^Longitudinal
- \ -^ «*^ strain
50
300 -
30

/
100 - 10
\ E
1 1 I
-0.10 0.10 -0.10 0.50
Strain (%)
 Load (kg)

Stress (MPa)

• 1 1 1 1 1 1
5
o» 5?

Exper iments
>
C/J
o
3
s
a
o
8
o • •

3
•o
o

Acoustic emission rate

(counts)
22 Introduction

It is interesting to note that through the selection of fiber and

matrix materials, as well as their geometric arrangements, a broad
spectrum of material performance can be accomplished. It is
feasible to design the physical and mechanical properties of
composites which not only duplicate the performance of some
existing materials but also fulfil the most demanding structural roles
not envisioned before.

1.5.2 Tailoring of performance

The structure-performance relationships of the various
types of fiber composites are further demonstrated in this section.
First, for continuous fiber composites, the problem of edge de-
lamination is used as an example. Next the variation of composite
electric properties with the configuration of reinforcements is
demonstrated.
Consider the [±45°/0?/902]s laminate. The effect of fiber orienta-
tion on the deformation of each individual lamina is highly
anisotropic (Fig. 1.13). The compatibility of displacements among
the laminae induces interlaminar stresses through the thickness
direction of the laminate. Sun (1989) has demonstrated that the
opening mode of delamination can be minimized through fiber
hybridization, stitching, the use of adhesive layers, ply termination,
and modification of edge geometry.
Figure 1.14 shows the free-edge interlaminar normal stresses in
the all-carbon composite and the hybrid composite formed by
replacing 90° plies with a glass/epoxy composite. A significant

Fig. 1.13. Effect of fiber orientation on the deformation of composite

laminae. (After Sun 1989.)

\
45°
I
90°
 Microstructure-performance relationships 23

reduction in interlaminar normal stress is achieved with hybrid

laminates. The experimentally measured delamination initiation
stress and failure stress are 324.3 MPa and 800.4 MPa, respectively.
The corresponding stresses for the hybrid laminate are 800.4 MPa
and 883.2 MPa, respectively. Thus, the addition of the glass/epoxy
plies significantly improves the delamination stress. The gain in
failure stress is not as significant since the 0° plies in both laminates
dominate the ultimate strength.
Reinforcements in the thickness direction can suppress inter-
laminar failure. Figure 1.15 shows the X-ray radiographs of
[±45°/05/90°]s laminates under uniaxial tension. The specimen with
through-the-thickness stitches along the free edges experiences
much less delamination than the specimen without stitches.
Besides relying on textile performing techniques such as stitching,
weaving and braiding, delamination in brittle resin matrix compos-
ites can be remedied by adding a ductile matrix in the form of thin
adhesive layers. The resulting composite has a hybridized matrix. It
has been demonstrated in [07907457—45°]s carbon/epoxy lamin-
ates that by reducing the free-edge effect the laminate strength can
be greatly improved. Furthermore, the laminate strength becomes
an isotropic property which can be predicted by the classical failure
theory. The use of adhesive layers in laminates subject to low-
velocity impact also proves to be effective in suppressing the
development of matrix cracking and delamination.

Fig. 1.14. Free-edge interlaminar normal stresses on the mid-surface in

carbon/epoxy and carbon/glass/epoxy laminates. (After Sun 1989).
0.25
Carbon
Hybrid

iS 0.125

J_ J_ J_ J_
0 4 8 12 16 20 24
Distance from free edge/ply thickness
24 Introduction

The transport properties, e.g. electrical conductivity, thermal

conductivity, dielectric constants, magnetic permeability and
diffusion coefficients of composites, are also sensitive to the
microstructure of the reinforcements. McCullough (1985) has
demonstrated the importance of structural features that promote
transport along the preferred path, i.e. percolative mechanisms.
Consider, for instance, the electrical behavior of metal-filled poly-
mers. The effective resistivity changes sharply from non-conducting
to conducting behavior upon crossing a 'percolation threshold'.
Figure 1.16 illustrates such a transition for a composite containing
conductive fillers (pf = 10~6 Q cm) in an insulating polymer matrix
(pm = 1016 Qcm). The decrease in resistivity with the increase in
filler volume fraction is attributed to the enhancement in probability
of particle-particle contact. McCullough has concluded that these
contacts promote the formation of continuous conduction paths that
mimic the behavior of conducting fibers.

1.5.3 Intelligent composites

Traditionally, fiber composites have been designed and
manufactured with the purpose of serving very specific functional
Fig. 1.15. X-ray radiographs showing delamination in unstitched (left) and
stitched (right) [±45°/02/90°]s laminates under uniaxial tension. (After
Mignery, Tan and Sun 1985.)

551 MPa(80ksi) 641 MPa(93ksi)

 Microstructure-performance relationships 25

goals. Such goals and considerations may include stiffness, fracture

toughness, fatigue life, impact resistance, electromagnetic shielding,
corrosion resistance, and biocompatibility, just naming a few. With
the expansion in available material systems for composites, advance-
ments in fabrication technologies, and improvements in analysis
and design techniques, it becomes increasingly feasible for develop-
ing multi-functional fiber composites for which a number of
functional goals are satisfied simultaneously, and the performance
can be optimized.
A new breed of multi-functional composites is dubbed 'smart
composites' or 'intelligent composites'. Takagi (1989) has defined
intelligent materials as 'those which can manifest their own func-
tions intelligently depending on environmental changes'. Thus,
intelligent composites can react to the thermal, electrical, magnetic,
chemical or mechanical environment and adjust their performance
accordingly. It should be borne in mind that intelligent composites
are made possible only through the design of their microstructures.
There are two basic requirements for intelligent composites to
'think' for themselves. First, the ability to detect the change in the
environment, such as pressure, strain, temperature, and electro-
Fig. 1.16. Illustration of chain formation in a participate filled composite.
Open circles and closed circles indicate, respectively, isolated particles and
contacting particles participating in chain formation, p and V{ denote
resistivity and filler volume fraction, respectively. (After McCullough
1985.)

1016

o
C3

Filler volume fraction, Kf

26 Introduction

magnetic radiation is necessary. Next, the ability in feedback and

control is also needed so corrective actions can be taken.
An example of intelligent composites under consideration by
researchers is the skin of an aircraft wing (see Port, King and
Hawkins 1988). The resin-based composite skin in this case has
built-in optic-fiber sensors which through the pulses of laser light
can detect internal defects and damages, the weight of ice or
incoming electromagnetic radar waves. Signals from the sensors
would be analyzed by patches of chips mounted on a flexible printed
circuit board bonded over the skin.
It has been suggested that implanting monolithic microwave
integrated circuit chips around an airplane's surface would produce
a huge, omnidirectional antenna that would be far more effective
than the small forward looking units now mounted on its nose.
Other applications of intelligent composites have been envisioned
for the purpose of in-flight damage assessment capability on
airplanes and orbiting spacecrafts, prelaunch checks for leaks and
structural integrity of the casing around rockets, altering the
stiffness of sporting equipment such as golf club and fishing rods in
response to the changing operating conditions, and monitoring the
sway of high-rise buildings induced by hurricane winds or
earthquakes so measures to compensate such deformations can be
activated (Port, King and Hawkins, 1988). Some of the issues of
intelligent structures have been discussed by Rogers (1988).
In summary, the challenges of intelligent composites are mani-
fested by the following factors: (a) development of sensing,
feedback and control systems as well as the technologies for
fabricating composites imbedded with such devices, (b) implemen-
tation of the required changes in the shapes of the structural
components, for example the change of the angle and shape of an
airplane's wing, and (c) perhaps the most challenging task, the
ability of a material to change its performance, for example the
stiffness or transport properties.
The combination of the structural and non-structural roles of a
composite in an integrated manner will undoubtedly change the
performance of fiber composites in a way not envisioned in the past.

1.6 Concluding remarks

Having examined the evolution of engineering materials,
and the role of fiber composites in materials technology, it is
perhaps useful to put in perspective the research and economic
opportunities of advanced composites.
 Concluding remarks 27

First, from the viewpoint of materials research, it is important to

recognize that the distinction between the three classes of materials,
i.e. metals, ceramics and plastics, is disappearing. As observed by
Kelly (1987a), there are now plastics as strong as metals which show
some electrical conductivity. Metals are being made which are
super-plastic and can be subjected to deformations in processing
like conventional polymeric materials. Also the three classes of
materials are beginning to show the same limits of strength and
stiffness; fibers made from all three can attain stiffness and strength
close to the theoretically predicted values. Furthermore, the pro-
perties of all three classes of materials can be modified and
improved by the use of surface coatings.
As the distinction between the three classes of materials disap-
pears, new possibilities and opportunities arise. One of these,
according to Kelly, is the possibility of designing materials not so
much for final properties but equally in terms of processability.
These thoughts have profound implications for the future technol-
ogy of fiber composites:

(1) The commonality in processing shared by the three classes

of materials, e.g. super-plastic forming of metal and poly-
mers, injection molding of polymers and ceramic powders,
will enable more extensive and effective transfer of know-
how among the three basic disciplines and effect efficient
processing technology for fiber composites.
(2) The commonality in performance shared by the three
classes of materials, e.g. stiffness, strength, thermal expan-
sion, enables the material scientist to engineer composites
with a broad spectrum of component materials. Conse-
quently, hybridizations of materials, e.g. glass and low-
melting-point metal, ceramics and thermoplastics, and
polymer and metal in laminates or other interdispersed
composite forms can be achieved and the properties op-
timized (e.g. composites composed of metal and polymer
components of nearly the same stiffness but different
fatigue resistance, or thermal expansion coefficient).
(3) The similarity in material property and behavior implies
that analytical and design methodologies originally
developed for a specific class of composites may be
transferable to others. A notable example is the
fracture and failure behavior of ceramic and polymer
based composites.
28 Introduction

(4) The complex task inherent in conceiving components and

their materials and developing the proper design methodol-
ogy will grow increasingly dependent on computers and
multi-disciplinary teams. Such an approach will harness the
full potential of composites for the technologies of the
future.
2 Thermoelastic behavior of laminated
composites

2.1 Introduction
Laminated composites are made by bonding unidirectional
laminae together in predetermined orientations. The basis for
analysis of thin laminated composites is the classical plate theory.
When the thickness direction properties significantly contribute to
the response of the laminate to an externally applied elastic field,
the classical plate theory breaks down.
Fundamental to the treatment of thin laminates is the knowledge
of the thermoelastic properties of a unidirectional lamina. These
properties are predictable from the corresponding properties of
constituent fiber and matrix materials as well as the fiber volume
fraction. Having established the elastic response of a unidirectional
lamina, the behavior of laminated composites is then analyzed from
the strain and curvature of the mid-plane of the laminate as well as
the force and moment resultants acting on its boundary edges.
Because of the complexity of the constitutive equations for a
general anisotropic laminated plate, simplifications of the stress-
strain relations are accomplished through the manipulation of the
geometric arrangement of the laminae. The lamination theory is a
relatively mature subject; its treatment can be found in text books
of, for instance, Ashton, Halpin and Petit (1969), Jones (1975),
Vinson and Chou (1975), Christensen (1979), Tsai and Hahn
(1980), Carlsson and Pipes (1987), and Chawla (1987), and in the
review articles of Chou (1989a and b). A modification of the classical
plate theory is in the inclusion of higher order terms in the
displacement field expansion to account for the transverse shear
deformation. An outline of such modifications adopted by various
researchers is presented.
The classical thin laminated theory has been extended to take
into consideration the effects of thermal and moisture diffusions,
with particular emphasis on the transient behavior. Because of the
large differences in the magnitudes of the thermal conductivity and
moisture diffusion coefficients, the thermal and hygroscopic prob-
lems can be solved separately and their linear elastic fields can be
superposed. Stress concentrations due to transient thermal effects
30 Thermoelastic behavior of laminated composites

are of particular interest in the study of laminate thermal shock

resistance.
The mechanics of the thermoelastic behavior of laminated com-
posites is fundamental to the understanding of the strength, fracture
and fatigue behavior of all continuous-fiber composites including
those reinforced with textile preforms.

2.2 Elastic behavior of a composite lamina

2.2.1 Elastic constants

It is well known that for a homogeneous isotropic material
(i.e. the material properties are independent of the location and
direction), two independent material elastic constants are sufficient
to specify the constitutive relations. These could be any two of the
five constants commonly used: E (Young's modulus), v (Poisson's
ratio), G (shear modulus), K (bulk modulus), and k (plane strain
bulk modulus). The relations among these constants are
G = E/2(l + v)
K = E/3(l-2v) (2.1)
2
k = £72(1 - v - 2v )
Twenty-one independent constants are necessary to describe the
elastic stress-strain relation of a generally anisotropic material (i.e.
the material properties are different in different directions). How-
ever, due to the material symmetries, the number of the independ-
ent constants can be greatly reduced. Consider a lamina (Fig. 2.1)
composed of unidirectional straight fibers in a matrix. Assume that

Fig. 2.1. A unidirectional fiber composite lamina.

JC3 (thickness direction)

2 (transverse direction)

x i (fiber direction)
 Elastic behavior of a composite lamina 31

it is homogeneous on a scale much larger than that of the inter-fiber

spacing. Then, the unidirectional lamina can be treated as a
homogeneous orthotropic continuum (i.e. having three mutually
perpendicular planes of symmetry). The coordinates x1—x2—x3
shown in Fig. 2.1 are known as the material principal coordinates,
where xx is parallel to the fibers and x2 lies in the plane of the lamina.
For circular cross-section fibers randomly distributed in a unidirec-
tional lamina, the lamina can be further assumed macroscopically as
transversely isotropic, namely the material properties in planes
transverse to the fiber direction are isotropic. Then, there are only
five independent constants. The commonly used engineering elastic
constants for the transversely isotropic lamina, referring to the fiber
(xx) and in-plane transverse (x2) directions, are denoted by En
(longitudinal Young's modulus), E22 (transverse Young's modulus),
v12 (Poisson's ratio due to loading in the xl direction and contrac-
tion in the x2 direction), and G12 (in-plane shear modulus). These
four independent elastic constants can be determined experimen-
tally by three simple tensile tests of composite specimens with fiber
orientations of 0°, 90° and [±45°]2s; the relevant testing standards
are ASTM D3039-76 and ASTM D3518-76. The fifth independent
constant, representing the transverse isotropic properties, could be
either v23 (transverse Poisson's ratio) or G23 (transverse shear
modulus); the two are related by

£22
G23 (2.2)
2(1 + v 23 )

;r enjg ineering c

v2i =
EuVl2
£33 = £22

G 13 = G12 (2.3)
V 32 = v23
y = V12

V 3 1 = V2i

Various micromechanical models are available for predicting the

elastic properties of unidirectional laminae from their constituent
properties. Most of the matrices and some of the fibers used in
composites can be considered as isotropic. Let the elastic constants
32 Thermoelastic behavior of laminated composites

of Eq. (2.1) for the isotropic fiber and matrix materials be denoted
by the subscripts f and m, respectively. Also, the fiber volume
fraction of the composite is indicated by Vf. Assuming no void in the
composite, the volume fraction of matrix is
Vm = l-Vt (2.4)
The following relations due to Hashin and Rosen (see Rosen 1973)
are quoted for their concise forms and, hence, ease in application.

Gm

(2.5)
— +—
kt km

where
*f = £ f / 2 ( l - v f - v ? )
km = £m/2(l-vm-v2m)
+ (Vfkf+Vmkm)Gm
t
k* —-
Vmkf+Vfk
 Elastic behavior of a composite lamina 33

Fibers such as carbon and Kevlar exhibit anisotropic behavior;

their thermoelastic properties along and transverse to the fiber axis
are significantly different. These fibers are considered to be trans-
versely isotropic, and thus five independent constants are needed to
describe their elastic properties, namely, Eu, E2f, G12f, v12f and
G23f. The following expressions, due to Chamis (1983), describe the
elastic properties of a unidirectional lamina composed of anisotropic
fibers in an isotropic matrix:

£^22 - ^33 -
1 - V£l - EmIE2()
Gm
G,2 = G,3 = (2.7)
1 - V t (l - Gm,/G12f)
Gm
G23 =
- Vf(l - GJG23t)
X2 = Vi, = v1,fVf + vV
E22 t
= 1

Halpin and Tsai (1967) have developed some semi-empirical

relations for the laminar elastic properties. These expressions
contain certain parameters which are influenced by the geometry of
the reinforcing phases, their packing in the composite, and the
loading conditions. Estimates of the values of these parameters can
be obtained by comparing the Halpin-Tsai equation predictions
with the numerical solutions employing formal elasticity theory
(Halpin 1984). The effect of interfacial debonding on elastic
properties has been discussed by Takahashi and Chou (1988).

2.2.2 Constitutive relations

Consider a unidirectional lamina exhibiting orthotropic
symmetry. The constitutive relations, referring to the material
principal coordinates x1-x2-x3, assume the general form (Vinson
and Chou 1975):
/en\ /Su 512 5,3 0 0 0\ M,\
x v
£22 sl2 5 22 5 23 0 0 0
£33 513 523 5 33 0 0 0 033
(2.8)
2f23 0 0 0 S44 0 0
2£13 0 0 0 0 s5S 0
\o 0 0 0 0
34 Thermoelastic behavior of laminated composites

Here oijy the stress tensors, are defined in Fig. 2.2. efy are the strain
tensors defined in a manner analogous to the stress components; it
should be noted that the engineering shear strain yl7 = 2e;/ (i =£/). Stj
denote the components of the compliance matrix. For the case of a
transversely isotropic lamina with the x2-x3 plane being isotropic,
the compliance constants are related to the engineering elastic
constants as:
1

•^22 — £33 —
£,22

"J12 ~ "13 ~ "Z ~ ^ (2.9)

S 44 = —

55
** G 12

Fig. 2.2. Stress tensor components.

JC3
 Elastic behavior of a composite lamina 35

Equation (2.8) can be inverted to obtain the following stress-

strain relations

(° u
\ ICn Cu Cn 0 0 o\ /£11\
o22
o33
Cx2
Cn
c
C22 C33
c13 0
0
0
0
0
0
£22

£33
O23 0 0
23
0 C44 0 0 2£ 2 3
(2.10)
o13 0 0 0 0 c055 0 2£L3

\oJ 30 0 0 0 '66/ \2el2

where Ciy are the components of the stiffness matrix. Again, for the
case of transverse isotropy in the x2-x3 plane, the following
relations hold:
Cn = Eu(l-v223)/A
C22 = C 3 3 = £ 2 2 ( 1 - v 1 2 v 2 1)/A
C44 = G
23
(2.11)
Cn = Cn = (v21 = (v1
C23 = (v23 + v12v21)£22/A
A = 1 - 2v 1 2 v 2 1 - v i a - 2v12v21v23
For a unidirectional composite lamina where the thickness is
much smaller than the in-plane (xx-x2) dimensions, it is sufficient to
consider the two-dimensional constitutive relations. Following the
convention used in the composites literature, the following con-
tracted notations,OTand eif are introduced for the stress and strain
components, respectively. Their relations to the tensorial stress and
strain components are:
o3= o33, o3 = o33,

and o 6 =cr 12 (=r 12 )

£ = £ = £
2 22> 3
and
Under plane stress condition (i.e. A33 = o13 = o23 = 0), and using
the contracted notations, Eq. (2.8) can be reduced to

(2.12)
36 Thermoelastic behavior of laminated composites

where the compliance constants Stj are given in Eq. (2.9). Also
£3 = Sl3ox + S23o2 and s 4 = s 5 = 0. By inverting Eq. (2.12), the
following two-dimensional stress-strain relations are obtained:
(Qn Qn 0 \ /£,\
= G12 Q22 0 £2 (2.13)
\ 0 0 QJ\EJ
H e r e , the lamina exhibits orthotropic symmetry. T h e Qkj in E q .
(2.13) are known as the reduced stiffness constants, a n d are related
to the engineering constants as follows:
rr
Qu=-
v12v21

^ (2.14)
1 -v12v21

g _
1 - v12v21
066 = G 1 2
It should be noted that the Qtj s o obtained by assuming the plane
stress condition of the unidirectional lamina are not identical to the
Ctj given in Eq. (2.11). In fact, the difference between Ctj and Qtj
increases as the lamina becomes more isotropic. The inter-relations

Table 2.1. Inter-relations among the different forms of elastic constants.

After Chou (1989b)

Engineering E u E22 v12 v21 G12

Compliance 1/S U 1/X,, S^/Su -S,,/S,, l/5 66

Reduced (G,,Q 2 2-e" 2 )/Q 2 2 ( O i " ^ " 0? 2 )/Gii C2/G22 Gi2/Qn QM,
stiffness

Compliance su S22 5n S^
Reduced G22/G11G22 " Q\2) QUKQUQH- On) Gu/(Qi,G 2 2 - Qi2) l / 0 « ,
stiffness
Engineering l/Eu l/£ 2 2 -v12/Eu UG l2
constant

«ed«c<;rf G,, G22 Gl2 Q66

Engineering # n / ( l - v 12 v 21 ) £ 2 2 (1 - V12V2l) v 1 2 £ 2 2 /(l - v 12 v 2I ) G12

constant
Compliance S22/(SnS22 - S2I2) 5 n /(5,,5 2 2 - 5,,) -5 12 /(S,,S 22 - S2l2) 1/S«,
 Elastic behavior of a composite lamina 37

among the engineering constants, compliance constants and reduced

stiffness constants are summarized in Table 2.1.
For a unidirectional lamina oriented at an angle 6 with respect to
the reference axes x-y (Fig. 2.3), the stress-strain relations in the
x-y coordinates are
«\ / G n G12 QisN/fixxX
"yy = Gl2 Q22 G2 6 (2.15)
J \Gl6 G26 G 6 6 ,
where Qijy the transformed reduced stiffness, are given by
Gn = G n cos4 6 + 2(<212 + 2<266) sin2 6 cos2 6 + Q22 sin4 6
G12 = (G11 + G22 - 4G 66 ) sin2 0 cos2 6

Q22 = Gn sin4 6 + 2(G i 2 + 2G 66 ) sin2 0 cos2 6 + Q 22 cos4 0

Gi 6 = (Gn - G12 - 2<266) sin 9 cos3 6
(2.16)

G26 = (Gn1 - G12 - 2Q 66 ) sin3 6 cos 8

+ (G12 - G 2 2 + 20 6 6 ) sin0 cos3 0
G 66 = (Gn + G22 - 2Gi2 - 2G 66 ) sin2 6 cos2 6
+ G66(sin4 6 + cos4 0)

Fig. 2.3. Fiber axis at an angle 6 from the lamina reference axis x.

z, X) (thickness direction)

X2 (transverse direction)

x\ (fiber direction)
38 Thermoelastic behavior of laminated composites

Note that in the x-y coordinate system the notations of xxy and yxy
are introduced for the shear stress and strain, respectively. The
unidirectional lamina referred to the x-y axes is termed generally
orthotropic.
Equation (2.15) can be inverted to obtain the strain-stress
relations in the following general form:

(2.17)

in which the S/y- are the transformed compliance constants and their
relations to 5,-,- and 8 are
S n = Su cos4 8 + (2512 + S66) sin2 8 cos2 8 + S22 sin4 6
S l2 = 512(sin4 0 + cos4 0) + (5n + S22 - S66) sin2 8 cos2 0
5 22 = Su, sin4 0 + (25 i2 + 566) sin2 0 cos2 0 + S22 cos4 6
S16 = (2S n - 2S l2 - S66) sin d cos3 6
- (2S22 - 2S12 - S66) sin3 6 cos 8 (2.18)
3
5 26 = (2Sn, - 25 12 - 566) sin 0 cos 0
— (2S22 — 2S12 — S66) sin 8 cos3 0
S66 = 2(25n, + 2,S22 - 4S12 - 566) sin2 8 cos2 0
+ 566(sin4 8 + cos4 0)
The engineering constants of the unidirectional lamina referring
to the x-y axes, which are not aligned with the material principal
directions, can be expressed as functions of the off-axis angle, 8, by
using Eqs. (2.9) and (2.18)

^ ^ ) sin2 8 cos2 8 + ^ s i n 4 8

1 1 1 , ,
— + — sin2 8 cos2 8
(2.19)
1 1 / 1 2v \ 2 1 4
— = sin4 8 + 1— sin 8 cos2 8 + cos 8
Eyy £7,1 \G, 2 £,, / £2 2

l £-22

+ (sin4 0 + cos4
G12
 Elastic behavior of a composite laminate 39

The variations of Exx, Gxyy and vxy, with fiber orientation angle,
6, for a Kevlar-49/epoxy composite are shown in Fig. 2.4.
Jones (1975) discussed the extremum (largest or smallest) values
of composite elastic properties, which do not necessarily occur in
the principal material directions. It can be shown that Exx is greater
than both En and E22 for some values of 6 if

G12>, (2.20)
2(l + v12)
and that Exx is less than both En and E22 for some values of 6 if

G12> (2.21)
2{EjE22 + vl2)

2.3 Elastic behavior of a composite laminate

2.3.1 Classical composite lamination theory

Based upon the constitutive relations for a lamina com-
posed of a generally orthotropic material, Eq. (2.15), the constitu-
tive relations for a laminate formed by bonding several laminae

Fig. 2.4. Variations of engineering elastic constants with fiber orientation

angle, 6, for a Kevlar-49/epoxy composite with Vf = 0.6, £ H = 76GPa,
E22 = 5.5 GPa, G12 = 2.3 GPa and v 12 = 0.34.

0 30 60 90
Fiber orientation, 6 (degrees)
40 Thermoelastic behavior of laminated composites

together is presented in this section. The orientation and material

system of each lamina are general. Figure 2.5 depicts the geometry
of an Az-layered laminate of thickness h; the x—y plane coincides
with the laminate geometric middle plane. Following the approach
of the classical, linear, thin plate theory, the following assumptions
are made (see Vinson and Chou 1975).
(1) A lineal element of the plate extending through the plate
thickness, normal to the middle surface {x-y plane) in the un-
stressed state, upon the application of load: (a) undergoes at most a
translation and a rotation with respect to the original coordinate
system, and (b) remains normal to the deformed middle surface.
This assumption implies that the lineal element does not elongate
or contract, and remains straight upon load applications.
(2) The plate resists lateral and in-plane loads by bending,
transverse shear stress, and in-plane action, not through block-like
compression or tension in the plate in the thickness direction.
Based upon the foregoing assumptions, also known as the
Kirchhoff hypothesis for plates, the strain components can be
derived

(2.22)

Fig. 2.5. An n-layered laminate.

 Elastic behavior of a composite laminate 41

Here, e°x, e°y and yxy are the laminate mid-plane strain, which are
expressed in terms of the mid-plane displacements u° and v° in the
x and y directions, respectively:
3u° du° dv°
dy
^T (2-23)
dy dx
The mid-plane curvatures are related to the z direction mid-plane
displacement w°
2
32w° 32w°
w 32w°
Krr = — (2.24)
dx dy
Note that Kxy represents the twist curvature of the mid-plane.
Figure 2.6 depicts the deformation associated with a typical
cross-sectional element in a thin plate.
Also, following the approach of the classical plate theory, the
resultant forces and moments, instead of the stresses, are utilized in
the constitutive relations. Referring to Figs. 2.7 (a) and (b), the
force and moment resultants of the laminate are obtained by
integrating the stresses of each lamina, through the laminate
thickness, h\
rh/2
rHU
(Nx, Ny, Nxy) = (oxx, oyyy rxy) dz (2.25)
•>-h/2

rh/2
, Oyy, (2.26)
(Mx, My, Mxy) = -h/2
J-h/
Fig. 2.6. Deformation of a typical cross-sectional element in a thin
laminated plate.
A

B *

C
D

Undeformed cross-section Deformed cross-section

42 Thermoelastic behavior of laminated composites

Substitution of Eqs. (2.15) and (2.16) into Eqs. (2.25) and (2.26)
results in the following:

Nx An A12

A16 A2

BX2 B22 B26 II Ayy (2.27)

16 #26 #66

Fig. 2.7. (a) In-plane force resultants, (b) In-plane moment resultants.

(a)

M,

(b)
 Elastic behavior of a composite laminate 43

(2.28)

where

k=l

^J{Qli)k{hl-h\.l) (2.29)
k=l

In Eqs. (2.27)-(2.29), Aijf Bijy and Dtj are called extensional

stiffness, extension-bending coupling stiffness, and bending stiffness,
respectively. The summation in Eqs. (2.29) is carried out over all
the laminae; (QIJ)K refers to the reduced stiffness of the A:th layer.
Eqs. (2.27) and (2.28) are often expressed in the condensed form as

(2.30)

where [K] is composed of KXX, Kyy and 2Kxy.

The constitutive relations of Eqs. (2.27) and (2.28) can be
rearranged into other useful forms by partially or totally inverting
them. The totally inverted forms of Eqs. (2.27) and (2.28) are given
in the following condensed matrix expressions:

[K] = [B'][N] + [D'][M]

where

[A'] = [A*]-[B*][D*-1][C*]

[fi'] = [fl*][D—] = -[D*-'][C*]

[D>] = [D* " ' ]
44 Thermoelastic behavior of laminated composites

and (2.32)

An application of Eqs. (2.31) is found, for instance, when the

stress and moment resultants acting on a laminated plate are
specified. Then, with the knowledge of the elastic constants, the
mid-plane strain and curvature of the laminate can be determined.
The strain components of a specific lamina in terms of the plate
reference axes can be derived from Eq. (2.22) and the correspond-
ing stresses from Eq. (2.15). The existing criteria for laminar
failure, due to combined in-plane stresses or strains, require the
knowledge of stresses and strains along the fiber as well as the
transverse directions. This information can be readily obtained by
transformation of the stress and strain components to the principal
material directions. Thus, the correlation between external loading
on the laminated plate and the failure of an individual lamina can
be established.

2.3.2 Geometrical arrangements of laminae

It has been established in Eqs. (2.29) that the elastic
behavior of a composite laminate composed of unidirectional
laminae is determined by the constituent material properties as well
as the orientation and location of the individual laminae. These
geometric aspects of the laminae are indicated by following the
convention of the composites literature. For example, [0°/45y
-45°,/455/0°] indicates the stacking sequence of a laminate with one
layer at 0°, two layers at 45°, four layers at —45°, two layers at 45°,
and one layer at 0°. Because of the mid-plane symmetry, this
stacking sequence can also be expressed as [0°/45y— 455]s. Follow-
ing this convention, the basic arrangements of laminae can be
expressed as [0°] for unidirectional, [0°/90°] for cross-ply, and
[+6/ — 6] for angle-ply. The implications of the laminar geometrical
arrangements on the laminar elastic behavior, namely, the [A], [B],
and [D] matrices, are discussed below.
The [A] matrix relates the stress resultants with the mid-plane
strains. The couplings between normal stress resultants and mid-
plane shear strains, as well as shear stress resultants and mid-plane
normal strains, are due to the components Al6 and A26- There is
 Elastic behavior of a composite laminate 45

also the coupling between mid-plane stress resultants and the

bending and twisting of the laminate through the [B] matrix.
In particular, the components Bl6 and B26 relate normal
stress resultants with the twisting of the laminate. The [B]
matrix also plays a role in the coupling between the moment
resultants and in-plane strains. Finally, the Dl6 and D26 terms are
responsible for the interaction between the bending moment and
twisting.
The various coupling effects in laminated composites can be
minimized or eliminated through suitable choices of the laminae
stacking sequence. As can be seen from Eqs. (2.29), the Btj terms
involve the squares of the z coordinates of the top and bottom faces
of each lamina. Each term of Btj vanishes if for every lamina above
the mid-plane there is a lamina, identical in properties and
orientation, located at the same distance below the mid-plane. Such
mid-plane symmetry arrangements eliminate the bending-stretching
coupling. The terms A16 and A26 both vanish under either of the
following conditions: (a) all of the laminae assume 0°, 90° or
cross-ply [0°/90°] configuration; (b) for every lamina of +0
orientation there is another lamina of the same property and
thickness with a —6 orientation. The terms D16 and D26 are zero for
the cases: (a) all of the lamina assume 0°, 90° or cross-ply
configuration; and (b) for every lamina oriented at + 0 at a given
distance above the mid-plane there is an identical layer at the same
distance below the mid-plane oriented at - 6. It is obvious that the
D16 and D26 terms are not zero for any mid-plane symmetric
laminate, except for the cases of all 0°, all 90° and cross-ply.
However, the magnitude of these terms can be made small by
increasing the number of layers in the angle-ply configuration.
Table 2.2 shows the effect of stacking sequence on the [A], [B] and

Table 2.2. Effect of stacking sequence on [A], [B] and [D] matrices. After
Chou (1989b)

6 = o°, 90° 0°/90° + 02/-0,/ + 02/-0, Same

+ 0,/-0,. . . -0,/+0,. . . number of
(anti-symmetry) (symmetry) + 6 and - 6
layers

Al6, A26 zero zero zero - zero

Bn> B22, Bu, £66 zero zero zero -
BX6> B26 zero zero zero -
D i 6 , D26 zero zero zero - -
46 Thermoelastic behavior of laminated composites

[D] matrices. The optimization of laminate design for strength has

been discussed by Fukunaga and Chou (1988a and b).

2.4 Thick laminates

The term 'thick laminates' here is used to describe compos-
ite plates of which the thickness direction properties significantly
contribute to the response of the material. Exact elasticity solutions
of thick plates have demonstrated that the classical lamination
theory of Section 2.3 is not applicable to the thick laminates.
Experimental results (for example, Whitney 1972, and Stein and
Jegley 1987) have shown significant departure from lamination
theory predictions, for such properties as maximum deflections and
natural frequencies, when (a) the plate thickness-to-width ratio and
(b) the in-plane Young's modulus to interlaminar shear modulus
ratio become high.
One reason for the departure of thick plate behavior from
classical thin plate theory prediction is the presence of transverse
shear deformation. The effect of transverse shear deformation is
pronounced in anisotropic materials with high ratios of in-plane
Young's moduli to interlaminar shear moduli; this is typical in
laminated composites. Other assumptions of the classical plate
theory (see Section 2.3) such as negligible transverse normal strains
(ez = 0), and the linear in-plane strain variation with the z
coordinate all contribute to the limitations of the theory. Further-
more, the strong interlaminar shear existing in thick laminates is
responsible for delamination, particularly near the free edges. Thus,
it is imperative to determine the magnitude and distribution of
interlaminar shear in thick laminates.
In the following, the three-dimensional constitutive relations
of a thick composite lamina are introduced first. Then, the classical
and higher order theory for thick laminated composites is
discussed.

2.4.1 Three-dimensional constitutive relations of a composite

lamina
The three-dimensional constitutive equations of a compos-
ite lamina referring to the principal material coordinate system
xl-x2-x3 (Fig. 2.1) have been introduced in Eqs. (2.8) and (2.10),
for the case of orthotropic symmetry. The relations between the
 Thick laminates 47

stiffness constants and engineering elastic constants are:

Cn = En(l- v 23 v 32 )/A
C22 = £ 2 2 ( 1 - v 1 3 v 3 1 ) / A
C 33 = E 3 3 (l - v 12 v 21 )/A
C44 = G23
(2.33)
C55 = G 1 3

C12 = (v 2i + VaaVgOfin/A = (v 12 + v 13 v 32 )E 22 /A
C13 = (v31 + v21v32)En/A = (v 13 + v 12 v 23 )£ 33 /A
C23 = (v 32 + v 1 2 v 3 1 )£n/A = (v 23 + v 13 v 21 )£ 33 /A
A = 1 - v 12 v 21 - v 23 v 32 - v 13 v 31 - 2v 13 v 21 v 32
The general three-dimensional constitutive relation of a compos-
ite lamina referring to the reference coordinate x—y—z (Fig. 2.3) can
be obtained from Eq. (2.10) by tensor transformation:

oj Cn c12 c13 0 0 c^
Oyy
c12 c 1 2 C23 0 0 c26
C13 C23 C 3 3 0 0 c36 (2.34)
OyZ 0 0 0 C44 C45 0
0 0 0 C45 cS5 0
c 1 6 26 C36 0 0
C
Here, the x-y plane coincides with the xx-x2 plane and the angle
between the xx and x axes is 6. The stress and strain tensors in these
two coordinate systems are related by
[oxx jon\ /£xx\ /£ll\
Oyy
1
O22 yy £22
Ozz
- rri" o33 Zzz £33
(2.35)
Oyz O23 £yz £23
On

V 12 /
Oxz £xz £13
\Oxy,
\ol2j [£xvly
\ £x /
The transformation matrix is
cos 2 e sin2 0 0 0 0 2 cos 6 sin 9 \
2
sin cos2 0 0 0 0 - 2 cos 0 sin 9
0 0 1 0 0 0
m= 0 0 0 cos —sin 8 0
0 0 0 sin6 cos 8 0
\—cos 6 sin 6 cos 6 sin 6 0 0 0 cos2 6- sin2 0 /
(2.36)
48 Thermoelastic behavior of laminated composites

[T]"1 is obtained by changing 6 to - 0 in [T]. The stiffness matrix is

derived from
[C] = [T]-l[C][T]~l (2.37)
with t indicating the matrix transpose and the explicit expressions of
[C] are
C n = C n cos4 6 + 2(C 12 + 2C66) sin2 0 cos2 0 + C 22 sin4 0
C12 = ( C u + C 22 - 4C66) sin2 0 cos2 0 + C12(sin4 0 + cos4 0)
=
C13 C13 cos2 0 + C23 sin2 0
C 16 = ( C n - C 12 - 2C66) sin 0 cos3 0
+ (C 12 - C22 + 2C66) sin3 0 cos 0
C22 = C n sin4 6 + 2(C 12 + IC^) sin2 0 cos2 0 + C22 cos4 6
C23 = C 13 sin2 6 + C 23 cos2 0
C26 = ( C n - C12 - 2C66) sin 3 0 cos 6
(2.38)
+ (C 12 - C 22 + 2C66) sin 6 cos3 0
c33 = c33

C 3 6 = (C 1 3 - C 23 ) sin 6 cos 6
C 4 4 = C 4 4 cos 2 6 + C 5 5 sin 2 0
C 4 5 = ( C 5 5 — C 44 ) sin 6 cos 0
C 5 5 = C 5 5 cos 2 0 + C44 sin 2 6
C 6 6 = ( C n + C 2 2 — 2 C 1 2 - 2C 6 6 ) sin 2 6 cos 2 6
+ C 6 6 (sin 4 0 + cos 4 0)

2.4.2 Constitutive relations of thick laminated composites

The classical laminated plate theory does not take into
account the effect of transverse shear stress and strain. The
inclusion of transverse shear deformation in the classical thin plate
theory is achieved by allowing the transverse shear strains, exz and
eyz, to be non-zero. This gives rise to definitions of the shear force
resultants:

(Gx, Qy) = \ K z , Oyz) dz (2.39)

These shear force resultants can be related to the transverse shear

strains through the appropriate constitutive relations, Eq. (2.34)
(see Vinson and Chou 1975).
 Thick laminates 49

Several higher order plate theories have been proposed to

account for the transverse shear deformation. This is achieved by
retaining higher order terms in the displacement field expansions,
which are assumed in the form of power series of the z coordinate.
The accuracy of these theories is generally greater for a greater
number of terms retained in the series, but the complexity of the
governing equations places severe limits on the number of terms for
which solutions are realistically attainable.
Among the various proposed displacement field expansions, the
simplest one includes the linear term in z; it has been adopted by
many workers (for example, Reissner 1945, Whitney and Pagano
1970),
u(x, yy z) = u°(x, y) + ztyx{x> y)

w(x, y, z) = w°(x, y)
where u, v and w are the displacement components in the x, y and z
coordinates (Fig. 2.2), respectively; w°, v° and vv° denote the
mid-plane displacements of a point (JC, y); and xpx and ipy are the
rotations of the normal to the mid-plane about the y and x axes,
respectively. It is noted that, unlike the classical plate theory, due
to the existence of transverse shear deformation,

dw°
dx
(2.41)
dw°

The new curvatures expressions, which are different from Eq. (2.24)
are given by

dx dx

d% d2w°
(2-42)
dy dy2

xy +
2\dy dxI dxdy
50 Thermoelastic behavior of laminated composites

Then, the strain-displacement relations of linear elasticity are

_ du _ du° d\px

dv_ dv° d%
eyy = — = — + z ——
dy dy dy

(2.43)
(dv dw\
e -1! l - V v + dw°\1
dw\ dw°\
Bxz - 1
2
{— H"dxJ
'! 1
2\

dv\ 1 [du° du° /ay

txy = ^ 1
2 L dv

By substituting Eqs. (2.34) and (2.43) into Eqs. (2.25), (2.26)

and (2.39), the constitutive relations of the laminated plate in
terms of stress resultants and displacement variables can be
obtained as

Au A12 A16 Bn Bl2

Ny A12 A22 A26 B12 B22
Nxy A16 A26 Aee B16 B26
Mx Bu Bl2 B16 Ai Dl2
My B12 B22 B26 D12 D22
Mx D

\dy dxj
(2.44)
 Thick laminates 51

and

(
3w° \
dy+%\
a o (2-45)

dx *x)
where
r
a dz (1, 7 = 1, 2, 4, 5, 6)

(fl,y, Ay) = I Qj(z, z2) dz (i, j = 1, 2, 6)

and ck in Eq. (2.45) is a correction factor for the fcth lamina which,
according to Lo, Christensen and Wu (1977a), is determined by
matching the approximated solution with the exact elasticity sol-
ution in order to satisfy appropriately the requirements of vanishing
transverse shear stress on the top and bottom surfaces of the thick
plate.
Having obtained the constitutive relations, the problem of thick
laminated plates can be solved by substituting Eqs. (2.44) and
(2.45) into the plate equation of motion. Then, a set of partial
differential equations in terms of the displacement variables w°, v°,
n>°, tyx and xpy can be derived. These unknowns can be solved with
the appropriate initial and boundary conditions, which are deter-
mined from the total energy of the system (Whitney and Pagano
1970).
The approach outlined above demonstrates an example of the
high order laminated plate theories, where only the in-plane
displacement terms linear in z are included in Eqs (2.40); and it
differs from the classical plate theory only by the terms \px and \py as
shown in Eqs (2.41). As pointed out by Lo, Christensen and Wu
(1977a&b), despite the increased generality of the shear deforma-
tion theory, the related flexural stress distributions show little
improvement over the classical laminated plate theory. Thus, it is
apparent that higher order terms are needed in the power series
expansion of the assumed displacement field to properly model the
behavior of thick laminates.
Among the various higher order displacement fields proposed,
Lo, Christensen and Wu (1977a&b) suggested the following dis-
52 Thermoelastic behavior of laminated composites

placement forms:
u = u° + z*l>x + z2%x + z3c/)x
v = v° + z\py + z% + z3<py (2.46)
w = w° + zipz + z2%z
where the cubic terms in z for the in-plane displacement field and
the square terms in z for the out-of-plane deformations are used; a
total of 11 displacement functions (M°, V°, VV°, ipx, \pyy ipz, %xy %y,
!-z, 4>x and (t>y) are involved. Much improvement over the classical
theory predictions is observed; however, the complexity of the
analysis has increased tremendously.
The format of solution to higher order systems generally involves
the application of the principle of potential energy to derive the
pertinent governing equations of equilibrium. Using the strain-
displacement relations and the assumed displacement field, in
conjunction with the equations of equilibrium, a set of partial
differential equations in terms of the displacements used is derived.
The number of equations is determined by the number of terms
retained in the assumed displacement form. With the appropriate
initial and boundary conditions, the solution of these equations
describes the elastic behavior of the plate. The details of such
approaches can be found, for example, in the work of Whitney and
Pagano (1970), Whitney and Sun (1973), Lo, Christensen and Wu
(1977a&b), and Reddy (1984).
Although accounting for the higher order plate deformation in
thick laminates involves a great deal more complexity than the
classical thin plate approach, it is evident that the extra effort to
accurately describe their fundamentally different elastic behavior is
required. The numerical results of the flexural stress distribution in
an infinite [+30, — 30]s laminate of carbon/epoxy composite, sub-
jected to a pressure q, on the top surface (z = h/2) of the form

JTX
q=qosm— (2.47)

are shown in Fig. 2.8 (a) and (b) (see Lo, Christensen and Wu
1977b). Here the length L characterizes the load distribution. The
in-plane stress oxx is normalized as oxx = oxx/q0S2, S = L/h. The
results indicate that the higher order theory is necessary for
determining the deformation of plates with small L/h ratio.
Sun and Li (1988) and Luo and Sun (1989) have adopted a
global-local method for the analysis of thermoelastic fields of thick
 Thick laminates 53

Fig. 2.8. (a) Flexural stress distributions for a [+30, - 3 0 ] s angle-ply

laminate for L/h = 10. (b) Flexural stress distributions for a [+30, - 3 0 ] s
angle-ply laminate for L/h = 4. exact elastic solution; . . . . higher
order laminated plate theory; classical laminated plate theory.
(After Lo, Christensen and Wu 1977b.)
z/h

(a)

0.5
0.4
0.3
0.2
0.1
L
-1.0 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1.0
--0.1
--0.2

--0.3

--0.4

--0.5
(b)
54 Thermoelastic behavior of laminated composites

laminated composites consisting of a repeating sublaminate (the

typical cell). The effective moduli and thermal expansion
coefficients are obtained from the sublaminate and used to obtain
the global (average) stress and strain solutions. A refining proce-
dure is then introduced in which the global solution is used directly
to recover the stresses in the lamina or used as boundary conditions
in a sublaminate to perform the exact thermoelastic analysis.

2.5 Thermal and hygroscopic behavior

Besides externally applied load, deformations in laminated
composites can also occur due to changes in temperature and
absorption of moisture. This is known as the hygrothermal effect.
As polymers undergo both dimensional and property changes in a
hygrothermal environment, so do composites utilizing polymers as
the matrix. Since fibers are fairly insensitive to environmental
changes, the environmental susceptibility of composites is mainly
through the matrix. Consequently, in a unidirectional composite the
temperature-moisture environment has a much greater effect on
the transverse and shear properties than the longitudinal properties.
The thermal diffusivity and moisture diffusion coefficient are used
as measures of the rates at which the temperature and moisture
concentrations change within the material. In general, these para-
meters depend on the temperature and moisture concentration.
However, over the range of temperature and moisture concentra-
tion that prevails in typical applications of composites, the thermal
diffusivity is usually several orders of magnitude greater than the
moisture diffusion coefficient. Consequently, thermal diffusion takes
place at a rate much faster than moisture diffusion, and the
temperature will reach equilibrium long before the moisture con-
centration does. This allows one to solve the heat-conduction and
moisture-diffusion problems and the resulting elastic fields
separately.
The knowledge of anisotropic heat conduction is basic to the
solution of thermal stresses in laminated composites. Investigations
of such problems have been performed by Poon and Chang (1978),
and Chu, Weng and Chen (1983) using transformation theory, by
Chang (1977), Huang and Chang (1980), and Nomura and Chou
(1986) using Green's function method, by Tauchert and Akoz
(1974) using a complex variable method, and by Katayama, Saito
and Kobayashi (1974) using a finite difference technique. The
solution of the steady-state thermoelastic problem of anisotropic
material appears to be initiated by Mossakowska and Nowacki
 Thermal and hygroscopic behavior 55

(1958), Sharma (1958), and Singh (1960). Then Takeuti and Noda
(1978), Sugano (1979), and Noda (1983) have examined the
transient temperature and thermal stress fields of transversely
isotropic elastic medium.
In the category of thermally and elastically orthotropic media, the
steady-state temperature and thermal stress field have been investi-
gated for problems of semi-infinite domain (Akoz and Tauchert
1972), a slab bounded by two parallel infinite planes (Tauchert and
Akoz 1974) and a rectangular slab (Akoz and Tauchert 1978). The
transient thermal stress analysis of thermally and elastically or-
thotropic laminae has been performed by H. Wang and Chou (1985,
1986), Wang, Pipes and Chou (1986), and Y. Wang and Chou
(1988, 1989); their approaches are recapitulated in the following.
In Section 2.5.1, the thermoelastic constitutive equations for a
three-dimensional orthotropic material are introduced. These equa-
tions are then simplified to the two-dimensional case of unidirec-
tional laminae, and the classical lamination theory is generalized to
take into account the thermal and hygroscopic effects. Then, three
transient thermal and hygroscopic problems are discussed to illus-
trate the formulation of the boundary value problems and the
solution techniques. The first problem is for the diffusion of
moisture through the thickness of a laminated composite (Section
2.5.2). It is assumed that the diffusion equation is one-dimensional
(z direction), while the elastic field is two-dimensional (x—y plane).
The second problem focuses on the effect of heat conduction on
interlaminar thermal stresses (Section 2.5.3). It is assumed, in this
case, that heat flows across the width of a laminated plate
(one-dimensional heat conduction) and the resulting thermal stress
field is three dimensional. Finally, a two-dimensional heat conduc-
tion problem is formulated for a rectangular-shaped unidirectional
lamina subjected to thermal boundary conditions at its four edges.
The two-dimensional thermal elastic field is obtained. In all three
problems, the thermal transient effects on stress distribution are
demonstrated.

2.5.1 Basic equations

2.5.1.1 Constitutive relations

Deformations of a unidirectional lamina resulting from
hygrothermal effects can be described by a modified set of linear
constitutive equations: i.e., the total strain minus the non-
mechanical strain is linearly related to the stress.
56 Thermoelastic behavior of laminated composites

The non-mechanical strain is measured from a stress-free re-

ference state, and the elastic moduli used in the calculation are
taken at the final environmental conditions. For example, in the
fabrication of polymer matrix composite laminates, the curing of an
individual ply results in different deformations along the fiber and
transverse directions. The constraint of deformation of a single ply
due to the presence of other plies in a multi-directional laminate
gives rise to residual stresses. Since most of the cross-linking in the
polymer occurs at the highest curing temperature, the polymer
matrix can be considered as still viscous enough to allow complete
relaxation of the residual stress. Thus, the highest curing tempera-
ture can be regarded as the stress-free temperature.
By taking into account the non-mechanical strain in Eq. (2.10) for
hygrothermally induced deformation, the laminated plate analysis
developed in Section 2.3 can be modified to determine the overall
elastic response. The stresses due to moisture absorption and
temperature change are identically analogous, in that they are
dilatational and self-equilibrating when the whole laminate is
considered. In general, the longitudinal properties of polymer
matrix composites are far less sensitive to temperature and moisture
than the transverse and shear properties of unidirectional compos-
ites, because of the excellent retention of mechanical properties by
the fibers. The greatest reduction in properties occurs when
temperature and moisture are combined, such as in hot and humid
environments. However, the combination of temperature and
moisture could render a laminate free of residual stresses. This can be
understood by considering, for example, a [0°/90°] cross-ply based
upon a resin matrix. The thermal stress induced from fabrication
is tensile in the transverse direction of a ply, while the residual
stresses induced by moisture absorption are compressive. Some
details of analysis of such phenomena are developed in the following.
Referring to the principal material coordinate axes of a unidirec-
tional lamina, the three-dimensional orthotropic stress-strain rela-
tions of Eq. (2.10) can be written as

M c12 c 1 3
0
0
0
0
0l
0 a22T- a22T-
<7 33
c12c22C23 c 2 3 C33 0 0 0 £33- #33 T — /3 3 3 /n
^23 0 0 0 C44 0 0 2e 2 3
0 0 0 0 C55 0 2e 1 3
W 0 0 0 0 0 ^66/ \ 2el 2
(2.48)
 Thermal and hygroscopic behavior 57

where au are the coefficients of thermal expansion and /?l7 are the
coefficients of hygroscopic expansion; the subscripts of these
coefficients indicate the principal material axes xt (i = 1-3). Also, T
denotes a small uniform temperature change from the 'stress-free'
temperature; m is the change in moisture concentration referring to
a 'moisture-free' environment. Both ocuT and pum indicate non-
mechanical strains.
Referring to the reference axes x-y and following Eq. (2.34), Eq.
(2.48) can be rewritten as

/c, [ c12 c 1 3 0 0
Oyy
c12 C22 C
23
0 0 C26 1

c013 C23 C33 0 0 C36

Oyz 0 0 C 44 c4, 0
Oxz \ 0 0 0 C 45 c55 0
\oxyj \C16 C16 C 16 0 0

I - P x , cm \
Syy — ar yy r -0» ; m
2eyz -ft.
X (2.49)

\
2 e x y - axyT -ft yml
where
a^x = O"n cos 2 0 + ar22 sin 2 8 pxx - f3u cos 2 9 + /3 22 sin 2 9

&yy = a22 C°S2 0 + C£ \ \ 6 y = /3 2 2 cos 2 6 + p n sin 2 0

— a-22) sin 6 cos 6 = (j8n-j322) sin6 cos

(2.50)

and 0 is defined in Fig. 2.3.

The relations given in Eq. (2.49) require that the thermoelastic
deformations of the medium are accurately described by linear
coefficients of thermal expansion over the range of temperatures of
interest, an often used assumption. Similarly, the deformations
induced by the hygroscopic nature of the medium are characterized
by linear coefficients of hygroscopic expansion, an assumption
which follows from existing experimental data.
58 Thermoelastic behavior of laminated composites

The elastic constitutive relations for a laminate subjected to both

thermal and hygroscopic environments have been formulated by
Pipes, Vinson and Chou (1976). For the purpose of laminar
analysis, Eq. (2.49) is reduced to

' oxx\ Qn Qii 2I6\ / £xx -axxT- /3xxm \

= (Gi2 G22 2 2 6 11 8yy - ayyT ~ /3yym I (2.51)
WGi6 Q26 Q e 6 / \ 2 £ X y - axyT -pxy

Substituting Eq. (2.51) into Eq. (2.25) and following the notation
of Eq. (2.30), the constitutive equation is expressed in the following
condensed form:

[N] = [A][e°] + [B][K] - [N]T - [N]m (2.52)

In Eq. (2.52), the effective thermal force resultants, [N]T, and

effective hygroscopic force resultants [N]m are introduced with the
following definitions:

rh/2
Nj= (Quaxx + Ql2«yy + Ql(,axy)T(z, t) dz
J-h/2

A? J
J
-A/2

-h/2
J-h/2
(Qi2axx + Q22ayy + Q26axy)T(z, t) dz (2.53a)

rh/2
-h/2
NTxy = (Ql6<Xxx + Q26ttyy + Q<&ocxy)T{z, t) dz
J-A/2
J-h/2

/-A/2
= (Gllflc. + Ql2pyy + Q\6PXy)m(z> 0 dz
J-A/2

J -A/2

-A/2
(QuPxx + Q220yy + Qxfixy)m(z, t) dz (2.53b)

rh/2
(Q ,6PXX + Qlrfyy + Q6(,l3xy)m(z, t) dz
J-A/2

where t denotes time. Consider the A:th layer of the laminate; and
define J2 r(£ 0d£ = R(z, t) and J" z m(|, 0 d | = H(z, t). Then,
 Thermal and hygroscopic behavior 59

Eqs. (2.53) are written as summations

k=\

+ Q2(,<Xyy + Q
k=\
x[R(hk,t)-R(hk_ut)]
(2.54a)

i k , t) ~ H{hk.x, t)]

y.[{H{hkli)-H{hk.1,t)\
(2.54b)
Parallel to the treatment of in-plane response, the flexural
response of the laminate is obtained by substituting Eq. (2.51) into
Eq. (2.26)
[M] = [B][e°] + [D][K] - [MY - [M]- (2.55)
T
Here, the effective thermal moment resultant, [M] , and effective
hygroscopic moment resultant, [M]m, are defined as
rfl/2

Mj = (Qu«xx + Ql2«yy + Qx6<xxy)T{z, t)z dz

J-hli
rh/2
M,'J = I(Qi2<xxx + Q22<xyy + Q26«xy)T (z, t)z dz (2.56a)
J-h/2
rh/2
rh/Z

= ( 6 l 6 * « + Q26<*yy + Q66axy)T(z, i)z dz

J-h/2

J -h/2

-ft/2
(Gllfc* + Ql2pyy + Ql6PXy)m(z, t)Z dz

^•ft/2
M™ = (G12&* + Q22fiyy + Q26Pxy)m(z, t)z dz (2.56b)
J-h/2
J-h/2
rh/2
= (Gl6j8« + G26/3y, + Q66pxy)m(z, t)Z
JJ-h/2
60 Thermoelastic behavior of laminated composites

By introducing the integrals of R(zy t) (i.e., S(zy t) = Jz R(%, t) d§),

and H(z,t) (i.e., /(z,i) = f//(§,0dg), Eqs. (2.56) are also
expressed as summations:
n
MTX = 2 ( G n * « + Qxioiyy + Qi6«xy)k[hkR(hk, t)
k=l

k^, t) - S(hk, t) + S(hk_u t)]

k, t)

^, t) - S(hk, t) _u t)]

+ QlfiOCyy + 066 KR Qlk, t)

k.l, t) - S(hk, t)+S(hk_ut)

_u t)]

AC = S (Gn/5« + Qufiyy + QufiMWhk, t)

- hk.xH{hk_u t) - J(hk, t) + J(hk^, t)]

AC = E (Gi2j8« + e pyy + Q26pxy)k[hkH(hk, t)

k=\

- hk_lH(hk_l) t) - J(hk, t) + J(hk_u /)] (2.57b)

AC y = {Qlffixx + + Q(*PXy)k[hkH{hk, t)

t_!,t)-J(hk,t)+J(hk_ltt)]

Finally, Eqs. (2.52) and (2.55) are combined as

/N,\ / Nx\ lN?\
Ny Ny
NXy NXy
Mr Ml M™
My Mj Af™
Mxyj \Ml/Mj \M?yj
IAn Ai2 AX6 BX] #12 #,6\/^\
Al2 A22 A R #22 #26 e°
tyy
AX6 A26 A R #26 #66 y°
^^66 ^16 Yxy (2.58)
#11 #12 fil6 D n A2 A 6 Kxx
fi.2 B22 B26 Dl 2
D22 D26 Kyy
l#16 fi26 B66 Dl 6 D26 Djj 2KXJ
 Thermal and hygroscopic behavior 61

The response of a laminate subjected to known mechanical force

and moment resultants, and both thermal and hygroscopic effects,
can be determined by calculating the effective thermal and hygro-
scopic resultants and inverting Eq. (2.58). The inversion would
yield laminate mid-plane strains, £°, and curvatures, K. The strains
of the laminae could then be calculated by Eq. (2.22). Given the
strains, the stresses within each lamina could be determined
according to Eq. (2.51).

2.5.1.2 Thermal and moisture diffusion equations

The three-dimensional heat conduction equation for a
general anisotropic solid of constant conductivity coefficients is
given by (Ozisik 1980)

Krr— :T-I-/C yy~ ^+/C,,— r + 2/Cru— y - \-2,K

3xz dy1 3z2
VV
3x dy 3x 3z

x B2T BT
+ 2Kjz ——- =pCp — (2.59)

Here, Kjj denote the coefficients of heat conduction, p is mass

density, and Cp is the specific heat. The temperature of the elastic
medium, T, is a function of location (x, y, z) and the time, t. It is
understood that there is no internal heat generation of the elastic
body.
For a thermally orthotropic material, with respect to the re-
ference axes x—y—zy Eq. (2.59) is simplified as

KT^l+KT^l+KT^I=pC 3T

Here, the thermal conductivities Kjxy Kjy and KTZZ are related to the
conductivities along the material principal direction, i.e. Kjx, Kj2
and Kj3 using transformation equations identical in form to those
given in Eqs. (2.50).
An equation identical in form to Eq. (2.59) can be written for
moisture diffusion. Consider, for instance, the diffusion of moisture
along the laminate thickness (z) direction, the governing equation is
reduced to
2
,„ m dm
62 Thermoelastic behavior of laminated composites

where K™z is the moisture diffusion coefficient and m=m(z,t)

denotes the moisture concentration distribution. Equation (2.61) is
further discussed in Section 2.5.2.
In Section 2.5.3, the transient interlaminar stress induced by heat
conduction through the laminate width (y) direction is discussed.
Then, T = T(yy t), and Eq. (2.59) for each lamina is reduced to

K = pc £.62)
^ »Tt
In Section 2.5.4, heat conduction in the plane of a unidirectional
composite is considered. The governing equation for heat conduc-
tion becomes

K + K C (Z63)
^

2.5.2 Hygroscopic behavior

2.5.2.1 Moisture concentration functions

Pipes, Vinson and Chou (1976) assume that the classical
diffusion equation (see Jost 1960) governs the absorption and
desorption of moisture by a hygroscopic material as given in Eq.
(2.61). Consider first the case of moisture absorption. If the
laminate is assumed to be initially moisture free, while its surfaces
z = ±h/2 are exposed to a moisture concentration Mo, then
moisture concentration in the laminate at position z and time t is

m(z, t) = Mo\l - 2) mn cos(anz) (2.64)

where

_ (2n + 1)JT

and

4 f (-1)"

From Eq. (2.64), the effective hygroscopic force resultant can be

 Thermal and hygroscopic behavior 63

readily determined by combining Eqs. (2.54b) and (2.64):

Nf =E (Gll/8« + QllPyy +

\hk- hk_x - X —- (sin(anhk) - sin(anftJt_1))

= 2 65
' „ n( - )
T Am, . ]
L n=0 an " J

x\hk- hk_l - 2 —2

The effective hygroscopic moment resultant is then determined

from Eqs. (2.57b)

U(hl - Al-,) - £ — (A* sin(aBAft

L «=o ««

f (anAfc_0)
-I

a
(2.66)

m
V " z' \ / 1
64 Thermoelastic behavior of laminated composites

hl - h U ) - i — (hk sin(aA)
n=0 an
- hk_x %m(anhk^)

~ S -f (cos(anhk) - cos{anhk_x))

Next, consider the desorption of moisture. The laminate contain-

ing a uniformly distributed moisture concentration, Mo, is exposed
to a moisture-free environment on its surfaces z = ±h/2. The
solution of the diffusion Eq. (2.61) corresponding to these boundary
conditions is

m
m{z, t) = M o E n cos(anz) (2.67)

The corresponding effective hygroscopic force and moment resul-

tants are

(2.68)

x (sin(anhk) -

^Z —-(A* sin(anhk)
n=0 «n

S "T (cos(a«/iit) - c
 Thermal and hygroscopic behavior 65

k=\
x
E — (hk sin(anhk) - hk_x s i n ( (2.69)
Ln=O fl«

-
+ Z £ (cos(anhk) - cos(anhk_x))J
«=o«« -I

x 2—
(anAft_!))
n=0
2.5.2.2 Hygroscopic stress field
Pipes, Vinson and Chou (1976) have illustrated the hygro-
scopic effects on a carbon-epoxy system (T300/5208) comprising a
six-ply laminate of [0°/+45°/—45]s, where each lamina is of the
thickness h. It is assumed that the diffusion coefficient, K™z) and the
coefficients of expansion, a and /?, are constant over the ranges of

Fig. 2.9. Moisture distribution profiles during absorption. (After Pipes,

Vinson and Chou 1976.)

z/fc 0 m/Mo

-3L
66 Thermoelastic behavior of laminated composites

temperature and moisture concentration of interest. The material

properties are E n = 143GPa, E22 = 10.1 GPa, v12 = 0.31, G12 =
4.14 GPa, pn = 0, /322 = 6.67 x 10~3/wt% and h=0.1397 mm.
Figure 2.9 illustrates the moisture profiles across the laminate
which is moisture free at time t = 0, and then exposed to an
environment on both surfaces of moisture concentration Mo. The
range of K™zt values is between 1 x 10~5 and 5 x 10~4. It is seen
that by K™zt = 5 x 10~5 the moisture concentration at the mid-
surface is 20% of that at the surface.
Figure 2.10 shows the profiles of oxx, which is compressive in the
outer, 0°, laminae, because of the expansion caused by the moisture
gradients of Fig. 2.9, and the inner four laminae at ±45° are all in
tension. Stress values are maximum at the outer surfaces. The
profiles of oyy follow the same trend as oxxy and oyy > oxx at each
time. In both cases the steady state is achieved at K™zt > 5 x 10~4.
Figure 2.11 shows that txy = 0 in the outer two layers because
they are at the orientation of 6 = 0°; the same would occur for any

Fig. 2.10. Distribution of stress, oxx during moisture absorption. (After

Pipes, Vinson and Chou 1976.)

-3
 Thermal and hygroscopic behavior 67

layers at 6 = 90° in balanced laminates. The in-plane shear stresses

increase with time, because of the increasing strains caused by
increased moisture content; by the time of steady state (Kztlzt>
5 x 10~4) the shear stresses are much larger than either the oxx or
oyy stress. These large shear stresses imply large interlaminar shear
stresses, ozx and ozy, near laminate discontinuities.

2.5.3 Transient interlaminar thermal stresses

2.5.3.1 Transient temperature field

Consider an x direction infinite laminated plate subjected to
a temperature field T =To on two edges (y = ±b) at time t = 0+
(Fig. 2.12). By assuming that the temperature field in each layer is
independent of the thickness direction, i.e. T = T(y, t), the heat
conduction equation for each lamina follows Eq. (2.62).

Fig. 2.11. Distribution of stress, rxy, during moisture absorption. (After

Pipes, Vinson and Chou 1976.)

z/h

-10 -6 -2 10
68 Thermoelastic behavior of laminated composites

The boundary and initial conditions are

T(±b,t)=To T(y, 0) = 0 (2.70)
The solution of the governing equation Eq. (2.62) by the method of
separation of variables is

(2.71)

where

(2n

K
y

2.5.3.2 Thermal stress field

Y. Wang and Chou (1989) have considered the transient
thermal stresses in an orthotropic composite laminate. Since the

Fig. 2.12. Geometry of an angle-ply laminate for analytical modeling.

 Thermal and hygroscopic behavior 69

thermal boundary conditions are uniform along the surfaces y =

±b, the displacements are independent of the x axis and expressed
as:
u= u{y,z,t)
v= v(y,z,t) (2.72)
w = w(y, z, t)i
The stress-strain relations for such an orthotropicllotrc
follow Eq. (2.49):
n —C F 4 " Cl2eyy 4 C 16 e xy -z42C16exy-
oyy = Cl2exx 4' (--22£yy + C26exy -h 2C 2 6 e x y - *2r
=
°zz Cl3exx +' C23£yy + C36sxy -z+ 2C36sxy -
(2.73)
Oy z Z-L^44CyZ

°xz = 2C55£XZ
• C26£yy + C36£zz +h 2C 6 6 £ x y ~ &6T
where
a- = ^ J . C 1 1 + <XyyCl2 + - a^Ci 6

a- 2 = ar^C 12 + 0tyyC22 4- a z z C 2 3 ; 4 axyC26

(2.74)
6c3 = a
xxCn + +ayyC23 + azzC33 +^c 3 6
a6=aC[C16 + Xyy C 2 6 4 +azzC3t,44 ar, y C 66
The equilibrium equations can be written in terms of the
displacements:
d2u cfu a2u d2w _ _ dT
Cs6
5 /+ Cs5
az2 + Cz6
a/ + C36
a y 3 z " a 6 dy

(
-26 n 2+ (
"22 - 2
+
^44 _ 2
+
I*-'23 + *--44; _ - — *2 _
3y 3>> dz dy dz dy
C (Q4 + C) C
C36
aidz + (C
4 C23)
did~z + c" I f +
^a ? °
^ =
(2.75)
The equilibrium equations can be solved by a singular perturbation
technique (Van Dyke 1975). It is assumed by Y. Wang and Chou
(1988, 1989) that, for h/b sufficiently small, i.e. <10% (see Fig.
2.12), the linear and higher order terms of h/b can be neglected and
70 Thermoelastic behavior of laminated composites

a zeroth order perturbation approach (Hsu and Herakovich 1976,

1977) is applied. The solution of Eqs. (2.75) for the A:th layer in the
interior region (Y = y/b < 1) of a laminate is

D(Y,t) (2.76)

2w

where
2z

q2Qi(Y,t)-qiQ2(Y,t)

Qi(X, t) = i ( ^ a 3 - &i) Tk(Y, t)h^ (2.77)

, 0 = l Tk(Y, t) dY
k= \

h{k) = A:th layer thickness

The subscript k indicates the A:th lamina of the n-layer laminate.
Following Hsu and Herakovich (1976, 1977), a stretching trans-
formation parameter is introduced to obtain the solution for the
boundary layer region (Y ~ 1):
f? = (1 - Y)/(h/b) (2.78)
 Thermal and hygroscopic behavior 71

Then the equilibrium equations (2.75) become

- c?V- 3 2 W _/h\(
/ A W && 6 \ 3 T
66 2 55 2 26
3rj 3Z 3ri2 3n
rj dZ ~ \b)\C mJ
\b)\C 3Y

32U - &V 32V - 32W

C + C C L
26 ^TT 22 T~2 + 44 T ^ ~ ^ 23 + L44J
(2.79)
_ / A \ / &2 \ 3T - 32U - - 32V
36 23 44
\b/\Cmi,J 3Y dridZ 3r)3Z
m
c
where Cti — CijICmaiX, and Cmax is the largest among all the Cfy
values. The following expressions of the displacement field are
assumed for matching the solutions in both interior and boundary
layer regions, based upon Prandtl's matching principle:
= B(Y,t) + PeXri cos(dZ)
V(k) = D(Y, t) + Relr> cos(dZ) (2.80)

Here, 5(F, t)> D{Yy t) and £(y, r) are the interior region solutions
(Eqs. 2.77); P, R and 5 are coefficients to be determined for the
correction terms; 6 is an undetermined positive constant; A is the
negative characteristic of Eqs. (2.79). It is seen from Eqs. (2.80) that
away from the boundary layer region (77 » 1), the correction terms
have no influence on the displacement field; their effects become
significant in and near the boundary layer region.
Substituting the U(k\ V(k) and W(k) expressions into the
equilibrium equations (2.79), the six roots of A for non-trivial
solutions of P, R and S are obtained:

A3,4=±M (2.81)
A5>6 = ±ck8
where ak, bk and ck are three positive constants. The positive roots
of A are dropped to avoid divergence in the displacement field.
Thus, the displacements for both the interior and boundary layer
72 Thermoelastic behavior of laminated composites

regions can be written as follows:

U(k) = B(Y, t) + (P1e-ak6ri + P2fbk6ri + P3e"c*6f|) cos(<5Z)
= D(Y, t) + (Rtf-""6" + R2e-bk6r> + R3e-c*Sri) cos(<5Z)
= E(Y, t)Z + (S^-""6" + S2e-btSri + S3e-Cta") sin(<5Z)
(2.82)
There are ten unknowns for the displacement solution of the A:th
layer (Pl9 P2, P3, Ru R2, R3, Su S2> S3 and d).
The available equations for the solution of these constants are: (i)
three stress boundary conditions, oyy(b, z) = oxy(by z) = oyz(b, z) =
0; (ii) six equilibrium equations (2.79); and (iii) the integrated
equilibrium condition

Jf1/2
= j ox
1\
Y,-)bdY (2.83)

A four-layer angle-ply composite is taken as a numerical ex-

ample. Each layer is 5 mm in thickness h(k\ 200 mm in width (b).
The SiC/borosilicate glass laminate is used as a baseline composite
system for demonstration of the results. The transient interlaminar
normal stress distribution of a [—45°/45°]s SiC/borosilicate glass
laminate, which is subjected to a sudden edge heating of the
magnitude To = 1°C at t = 0 + , is demonstrated in Fig. 2.13. No stress
singularity is found as a consequence of the assumed displacement

Fig. 2.13. Transient interlaminar thermal stress of a SiC/borosilicate glass

[-45°/45°]s laminate for Vf = 30% and To = 1°C. (After Y. Wang and Chou
1989.)

2000

1500 --

20 min >
X 1000 - 2 min \
10 s

500

i \ 1
0.90 0.925 0.95 0.975 1.00
Y = y/b
 Thermal and hygroscopic behavior 73

field, but it is apparent that the interlaminar normal stress con-

centration increases very significantly as approaching to the free
edge of the plate (Y = 1). The stress at Y = 1.0 is about three to
twenty times higher than that at Y = 0.90 for t = 10 s to ». As the
heating proceeds, the overall interlaminar normal stress increases
smoothly, while the stress which is very close to the boundary
remains almost constant. Also, the interlaminar normal stress tends
to zero away from the free edge of the laminate due to the adoption
of the classical lamination theory in the interior region.
Figure 2.14 shows the results of a parametric study of the stress
solution sensitivity to the composite elastic and thermal properties.
Here the [—45°/45°]s SiC/borosilicate glass laminate is taken as the
baseline system, and A indicates an increment. The Young's
modulus (£33) and thermal expansion coefficient (a33) along the plate
thickness direction have a more significant effect on the stress ozz than
the thermal conductivity (K33) and specific heat (Cp). The transient
thermal stress analysis can be applied for the characterization of
thermal shock resistance capability of composite materials. (See, for
example, Cheng 1951; Kingery 1955; Y.Wang and Chou 1991.)
2.5.4 Transient in -plane thermal stress
Having discussed the thermoelastic field due to one-
dimensional heat and moisture diffusion in Sections 2.5.2 and 2.5.3,

Fig. 2.14. Parametric studies of stress solution sensitivity to composite

elastic and thermal properties. The base material is a SiC/borosilicate glass
[-45°/45°]s laminate. Calculations of \Aozz\/azz are based upon £ = 2min
and Y = 0.99. (After Wang and Chou 1989.)

|AAT33

C
P

1 -
Aa 3 3

y 033
* - •

1 1 1 1 1
2 3 4 5
|Aocc
74 Thermoelastic behavior of laminated composites

a two-dimensional transient heat conduction problem is examined in

the following. The model material considered is a unidirectional
lamina. The interaction of thermal stresses among the layers of a
laminate is thus not included in order to clearly demonstrate the
effect of transient heat conduction.

2.5.4.1 Transient temperature field

Consider the two-dimensional problem of an orthotropic
slab with a rectangular region (0 < x < /X, 0 < y < /2) as shown in
Fig. 2.15. The slab is initially held at a uniform temperature and
then the edge y = l2 is suddenly subjected to an arbitrary tempera-
ture distribution or heat flux f(x). The two-dimensional tempera-
ture distribution, T{x> y\t) in the rectangular region is assumed to
satisfy the heat conduction equation (2.63).
The initial condition is

T(x,y;0) = 0 for f = 0 (2.84)

The boundary conditions of the rectangle assume the following

Fig. 2.15. Thermal stress variations with time for K = 0.1 at the cross-
section x = 0. (After H. Wang and Chou 1985.)

1.0 r 1 1 T- 1 T = 0.01
* 5 ^
Pll

y
1-

0.8 -- Oxy = 0 —
1

0.6 -
/ = 0.5
i
i

0
K x
/ v / T — °°

0.4 --

1
i
0.2 -
-
00 1 1 1 1
-1.3 -1.1 -0.9 -0.7 -0.5 -0.3 -0.1 0.1 0.3
Dimensionless stress, d/y
 Thermal and hygroscopic behavior 75

general forms:

dT
-a1- — +b1T = 0 forx=0 (2.85a)

dT
a 2- + b2T = 0 iorx = h (2.85b)

dT
-a3— +b3T = 0 fory=0 (2.85c)
dy
dT
a4 — + b4T=f(x) for y=l2 (2.85d)
dy

Here, at (i = 1, 2, 3, 4) are conductivities for the respective direc-

tions, and bi are the coefficients of surface heat transfer. The
various types of boundary conditions can be obtained through the
proper selections of the constant ratio bja i (see, for example,
Carslaw and Jaeger 1959). Equations (2.85a)-(2.85c) correspond to
zero surface temperature or heat flow, whereas the non-
homogeneous boundary condition of Eq. (2.85d) is for an arbitrary
variation of surface thermal condition.
Equation (2.85d) suggests the use of the principle of superposi-
tion. The problem has a steady-state solution as f-» °°. it is assumed
that

T{x,y\t) = 4>(x,y)+V(*>y\0 (2.86)

such that <j)(x, y) and ip{x, y\ t) satisfy

U
xx - 2 ^ W T^- (2.87)

a 1 - z +fe1rf)= O forx=0 (2.88a)

d(b
— + b2<p = 0 forx = l1 (2.88b)
' ox

dd>
a3—t — + b3(t) = 0 fory=0 (2.88c)
3
3y

3d)
^ + b4<f)=f(x) for y = l2 (2.88d)
dy
76 Thermoelastic behavior of laminated composites

and

d~t (2.89)
for f = 0 (2.90)
dtp
1 for x = 0 (2.91a)
3x
for x = /, (2.91b)
rx
dtp
3 for y = 0 (2.91c)
° dy
dtp
ttA for y —12 (2.91d)
dy
H. Wang and Chou (1986) have obtained the general solution of <p
and ip with the unknown constants in the infinite series expressions
to be determined by the boundary conditions of Eqs. (2.88) and
(2.91) and initial condition of Eq. (2.90).
An example of this solution technique is given by H. Wang and
Chou (1985) for a slab initially held at a constant temperature and
suddenly subjected to an arbitrary temperature variation along one
of its edges. The constants in Eqs (2.85) are ax = —1, a2 = a3 = b{ =
0, and b2 = b3= l. The temperature field solution is

T(x, y;t)= 2 1n cos dnx sinh -=y

+ E 4m cos dnx sin -^y exp[-d(d2n + n2m)t]\

(2.92)
where
U Axx/pi_p, A — P^yyl r^xX

2 K 6
Lm(on, Und = j2 4( 5 n) 77-71—;—-jsinh -^ l2
h
Kl \K,
2 11 r/,
f". (2-93)
«) -j T\
i cos dnx Ax
o , ol
n .
n 2 oj 2
a4 — cosh — - + b4 sinh ——
A A A

—-—JT n = 1,2, 3, . . . ,
 Thermal and hygroscopic behavior 11

Also, fim/K are the positive roots of

If a 4 = 0 and b4 = 1,

^ = ^jt = om m = ly2,3y...,<*> (2.94)

A /2
H. Wang and Chou (1986) have tabulated the solution of tempera-
ture field from the various combinations of aly a2y a3y bXy b2 and b3
values of Eqs. (2.85).

2.5.4.2 Thermal stress field

Consider a unidirectional fiber composite; let the principal
material directions xx and x2 coincide with the reference axes x and
y, respectively. The stress-strain relations follow Eq. (2.49) with
the Cij replaced by Ci}. Depending upon the thickness of the elastic
medium in the z direction, the thermoelastic problem is in the state
of either plane strain or plane stress. In the case of plane strain, the
stress components in the x—y plane are related to the in-plane
displacements, u{xy y\i) and v(xyy\t)y and the temperature,
T(xy y; t)y by substituting the strain-displacement relations into the
stress-strain relations of Eqs. (2.73). The results are

oxx = Cu — + Cn — -alT(x,y;t)

du dv
Oyy = C12 — + C22 — - a2 T(x, y; t)
(2.95)
du dv ,
tfzz = C 13 — + C 23 — - a3 T(x, y; t)

where
drj = Cn<xn +
&2 = Cl2an + C22a22 + C23a33 (2.96)
6c3 = CX3axx + C23a22 + C33a33
The relations corresponding to Eqs. (2.95) for plane stress condition
are obtained by replacing Ciy and at by Ctj — C3J/C33 and at —
6c3C3iIC33y respectively.
78 Thermoelastic behavior of laminated composites

The displacement equations of equilibrium governing the plane

strain conditions are

dtu ^a,fr Mr ,J^L_- SJL

(2.97)
2 3 T
3u 3^ ^_^L_ -

The equilibrium equations are solved by introducing the displace-

ment potentials iply xp2 and 0 defined by

(2.98)
^ + V2p + X^
dy dy dy
where vx, v2 and A are unknown constants. Also, 0 is the
homogeneous solution and \pl and ip2 are particular solutions of
Eqs. (2.97).
An example of the transient thermal stress solution is given by H.
Wang and Chou (1985) for a rectangular slab (-/ x <JC</ x and
0 < y < / 2 ) with fibers oriented in the x direction. The initial
temperature of the slab is T = 0. Then the following form of
temperature rise at the upper edge (y = l2) is adopted:

T=f(x) = T0cos^-x fovt>0 (2.99)

while the temperature over the remainder of the boundary is

maintained at the initial value. All edges of the rectangle are
assumed to be traction free:
oxx = oxy = 0 for x = ±lx
(2.100)
Oyy = OXy = 0 f°r y = ®> h

The thermal and elastic properties as given by Akoz and Tauchert

(1978) simulating a boron/epoxy composite are adopted for the
numerical calculations. Owing to the symmetry of the assumed
temperature rise, only one half of the rectangle ( 0 < x < /x) needs to
be considered. Thus, the boundary condition Eq. (2.85d) is reduced
to a4 = 0 and bA = 1. For the convenience of presenting the
 Thermal and hygroscopic behavior 79

numerical results, the following dimensionless quantities are intro-

duced for temperature stress, time, and lamina dimension, respec-
tively: T = T(x,y;t)/To; otj = otj{xyy;t)/&2TO9 R = DT/L2LY and 1 =
klh-
Figure 2.15 shows the y direction variation of thermal stresses at
the cross-section x = 0 for the various dimensionless time intervals.
It is clear that large longitudinal stresses oxx occur in the vicinity of
the heated boundary, where the relatively large temperature
gradient, dT/dy, exists. On the other hand, the transverse stresses,
oyyy and the shear stresses oxy are fairly small. Also, for axxy the
maximum transient tensile stress is 25% higher than that in the
steady state; the maximum transient compressive stress near the
upper edge (y = l2) is 78% higher than the corresponding steady-
state stress. An examination of the plots of oxx and oxy at a given
time interval indicates that each stress is in self-equilibrium when
the slab is free to deform, i.e. no boundary constraints. This is
consistent with the nature of thermal residual stresses.
 Strength of continuous-fiber composites

3.1 Introduction
Fiber-reinforced composites are a valuable class of en-
gineering materials because they can exhibit both high stiffness and
strength simultaneously, in contrast to more homogeneous materials
which are generally brittle and defect sensitive. In fiber composites,
the inherent lack of toughness of the reinforcing fiber, or its
sensitivity to microstructural defects, is overcome by the local
redundancy of the composite structure, so that its strength may be
utilized effectively. Individual fibers are relatively weakly coupled
by the matrix so that failure of one fiber does not generally
precipitate immediate failure of the composite as a whole, allowing
high strength and stiffness to be achieved in the fiber direction.
The tensile failure of a fiber-reinforced material is a complex
process which involves an accumulation of microstructural damage.
Unlike homogeneous brittle materials, fiber composites do not
contain a population of observable pre-existing defects, one of
which ultimately precipitates failure. Instead, an accumulation of
fiber or matrix fractures develops as the material is loaded and this
constitutes a 'critical defect' in a macroscopic view of the fracture.
Fracture mechanics may successfully account for the strength of
single fibers, but it is inadequate to extend its application to
unidirectional fiber composites when the overall behavior is domin-
ated by the probability of defects in fibers propagating under the
stress concentrations surrounding previous fiber fractures as well as
the probability of defects in the matrix which are responsible for the
multiplication of transverse cracks. Consequently, the statistical
process of damage development in composites needs to be em-
phasized (Manders, Bader and Chou 1982).
The development of a rigorous analysis of fracture, considering
all the sequences of fiber and matrix fractures which result in
fracture of the composite, is a formidable task, and for this reason
the strength of composites with realistic dimensions is much less
well understood than their elastic properties.
This chapter treats the strength of continuous fiber composites
with a combination of statistical and fracture mechanics approaches.
The statistical analysis of unidirectional composites is better de-
 Rule-of-mixtures 81

veloped than that for cross-ply laminates. No comprehensive

statistical methodology is available at this time for treating the
strength and failure of composites from the fiber and bundle level
up to composite laminates. Thus, the fracture mechanics approach
to laminate failure is necessary.
In this chapter, the classical approximation of the rule-of-
mixtures is adopted as a starting point for composite axial strength.
This approximation is substantially altered due to stress concentra-
tions induced at fiber breakages. The statistical variations of fiber
and bundle strengths are then discussed. The knowledge of the
stress redistribution at fiber breaks is then incorporated into the
statistical strength analysis of unidirectional fiber composites. Next,
the strength analysis is extended to the case of cross-ply laminates
which serve as model systems for laminate composites. Finally, an
attempt is made to shed some light on the failure of laminated
composites in general where both inter- and intralaminar failures
play key roles in the failure modes. A method of analysis based
upon the fracture mechanics approach is introduced. Section 3.4.6.2
is contributed by S. L. Phoenix, and Sections 3.4.7.4 and 3.4.8 are
contributed by A. S. D. Wang.
Another approach to the strength and damage of fiber composites
is based upon the overall properties degradation. The strength
behavior can be modeled by regarding the composite with damage
as a continuum with changing microstructure. A phenomenological
theory of constitutive behavior then provides relationships between
the severity of damage and the overall stiffness properties of a
composite (Reifsnider, Henneke, Stinchcomb and Duke 1983;
Talreja, 1985, 1986, 1987, 1989).
Strength theories dealing with short-fiber and hybrid composites
are discussed in Chapters 4 and 5, respectively.

3.2 Rule-of-mixtures
The classical approximation of unidirectional continuous-
fiber composite strength takes the form of the rule-of-mixtures. By
assuming equal strain in the fiber and matrix phases, the stress in
the composite under uniaxial loading can be expressed as (see Kelly
and Nicholson 1971 and Vinson and Chou 1975)

o c =o f V f +o m (l-V f ) (3.1a)
where a and V{ denote, respectively, stress and fiber volume
fraction. The subscripts c, f and m are for composite, fiber and
82 Strength of continuous-fiber composites

matrix, respectively. Then, the ultimate composite strength is

<jcu = Ofavf+amu(l-Vf) (3.1b)
Here, the subscript u denotes ultimate strength. Equation (3.1b) is
valid provided that both the fiber and matrix have the same ultimate
strain.
Equation (3.1b) is not sufficient in determining the strength of
continuous-fiber composites. Aveston, Cooper and Kelly (1971)
have discussed the strength of composites based upon the transfer
of load at the fiber/matrix interface and the mode of failure. For the
case of brittle fiber-reinforced ductile matrix, the matrix ultimate
strain is often higher than that of the fiber (Fig. 3.1a); then single
fractures of the composite occur when
ofu V( + a L ( l - Vf) > a m» (l - Vt) (3. 1c)

Fig. 3.1. (a) Stress-strain relation of a brittle fiber/ductile matrix compos-

ite, (b) Composite strength vs. fiber volume fraction for brittle
fiber/ductile matrix composites, (c) Stress-strain relation of a ductile
fiber/brittle matrix composite, (d) Composite strength vs. fiber volume
fraction for ductile fiber/brittle matrix composites.

Multiple
fracture
Single fracture
of fiber

(b)
 Rule-of-mixtures 83

where afmu is the stress in the matrix when the fibers fail. The matrix
is unable to withstand the additional load transferred to it due to
the fiber fracture, and thus single fracture prevails at sufficiently
high fiber volume fractions. At low fiber fractions,

- Vf) < a mu (l - Vf) (3- ld)

and the load is essentially born by the matrix material. The failure
of the composite is characterized by multiple fractures of the fibers
into shorter and shorter segments as the strain on the matrix
increases (Fig. 3.1b). Experimental data on the ultimate strength of
unidirectional fiber composites usually fall within the triangular
region of Fig. 3.1b specified by the solid line segments.
Provided the failure strain of the matrix is sufficiently large, the
fibers are fractured into lengths between x and 2x. Assuming a
constant fiber-matrix interfacial shear stress T, the fiber fracture

Fig. 3.1. (cont.).

Single fracture Multiple fracture

of matrix g
1
1
1

_ 0f u

to
1
1
1

(d)
84 Strength of continuous-fiber composites

spacing is determined from a simple force balance

where r denotes fiber radius.

The above analysis does not fully account for the fact that the
strength of a fiber is a statistical quantity which results from flaws
being randomly distributed along the length, as is discussed later.
One result is that the strength depends on the fiber length, and thus
is not really a fixed quantity afu. Using the accepted
Poisson/Weibull model, Henstenburg and Phoenix (1989) have
developed a modified version of Eq. (3.2) which includes a factor
connected to the variability in fiber strength. The revised formula
typically produces values which are 15 to 20% larger. Also, these
authors have delved further into the nature of the statistical
distribution for fragment length, and experimental examples can be
found in Netravali, Henstenburg, Phoenix and Schwartz (1989).
For the case of ductile fiber-reinforced brittle matrix composites,
multiple fracture of the matrix occurs when the fiber ultimate strain,
£fu, is higher than that of the matrix, £mu (Fig. 3.1c). The condition
of multiple fracture, according to Aveston, Cooper and Kelly (1971), is
o'fxyi+omu{\-Vf)<ofxxVf (3.3)
Here, o'fn is the stress in the fiber at the failure strain of the matrix.
A single fracture of the composite occurs if the fibers cannot
withstand the increase in loading due to the matrix failure (Fig.
3.1d).
The spacing between two adjacent matrix cracks can again be
determined from a simple force balance, and the separation
distance is between x' and 2xf

In deriving Eq. (3.4), it is understood that the number of fibers per

unit area transverse to the fiber direction is given by Vf/jrr2.
Composites containing ductile fibers in a ductile matrix have
shown work-hardening behavior. Mileiko (1969) has theorized that
the instability or necking of the matrix can be suppressed due to
the constraint of the matrix, and the ultimate strain of the composite,
in this case, is shown to lie in between the ultimate strains of the fiber
and matrix materials.
 Stress concentrations due to fiber breakages 85

3.3 Stress concentrations due to fiber breakages

Fiber breakages in a continuous-fiber composite can occur
at fabrication or during the early stage of loading. Stress redistribu-
tion takes place in the vicinity of a fiber breakage because load can
no longer be transferred along the fiber in a continuous manner.
The resulting stress concentrations in the neighboring fibers are
detrimental to the strength of continuous-fiber composites. In the
following, the shear-lag analysis is introduced to examine both the
static and dynamic stress concentrations in unidirectional
continuous-fiber composites.

3.3.1 Static case

The problem of static stress concentration in composites has
been treated by the shear-lag method (see Hedgepeth 1961;
Hedgepeth and Van Dyke 1967; Fichter 1969, 1970; Van Dyke and
Hedgepeth 1969; Zweben 1974; Fukuda and Kawata 1976a, 1980;
Goree and Gross 1979, 1980; Hikami and Chou 1990), elasticity
theory (see Burgel, Perry and Scheider 1970; Takao, Taya and
Chou 1981), and numerical methods (see Carrara and McGarry
1968; Chen 1971).
Among these approaches, the shear-lag method, which is based
upon simplified assumptions, often provides good physical insights
of rather complex problems. The shear-lag method was first
adopted by Hedgepeth (1961) to treat multi-filament failure prob-
lems of unidirectional composites. The technique also has been
extended to include the effects of plasticity of the matrix (Hedge-
peth and Van Dyke 1967; Goree and Gross 1979; Hikami and Chou
1984a), and the condition of interfacial debonding (Van Dyke and
Hedgepeth 1969). The major assumptions of this method are that:
(1) the fibers sustain only the axial loads, and (2) matrix between
fibers transmits only the shear force.
In the following the single filament failure model of Fukuda and
Kawata (1976a) is reproduced first to demonstrate the fundamentals
of this method, and the nature of stress redistribution in unidirec-
tional composites. Next, the work of Hikami and Chou (1990) is
introduced for the explicit solutions of multi-filament failure
problems.

3.3.1.1 Singlefilamentfailure
Figure 3.2 shows the model of analysis by Fukuda and
Kawata (1976a) which contains three parallel fibers with the middle
one being broken. This model can also be considered as the
86 Strength of continuous-fiber composites

two-dimensional representation of a laminate with a broken middle

layer. Because of symmetry, only half of the model needs to be
considered and the fibers are denoted as n = 1 and 2. The
equilibrium of forces in the fibers in the free-body diagram of Fig.
3.3 gives

(3.5)

-Tl=0 (3.6)
db

Fig. 3.2. A three-fiber composite model for shear-lag analysis.

y

Unit cell
n =2
pa

Fiber /
n = 1
\
-x \

Matrix

Fig. 3.3. Free-body diagrams for the 'unit cell' of the composite shown in
Fig. 3.2.

n =1
 Stress concentrations due to fiber breakages 87

where Px and P2 denote fiber axial force per unit thickness, and XX is
the matrix shear stress. Let the displacement of the nth fiber be
denoted as un. Then,

Pn{x) = E d ^ ^ n = l,2 (3.7)

rl(x) = -(u2(x)-u1(x)) (3.8)

where E is the Young's modulus of the fibers; G is the effective shear

modulus of the matrix; h is the effective fiber spacing; d is the fiber
width; and the lamina is of unit thickness.
Using Eqs. (3.7) and (3.8), and the following non-dimensional
parameters

(3.9)
a = Eh/Gd (3.10)
Equations (3.5) and (3.6) become

£ a 4 § + «2-«i=0 (3.11)

du2

From Eq. (3.11), u2 can be expressed by ux and its derivatives as

follows:

u2 = u1-^a—j (3.13)

Substitution of Eq. (3.13) into Eq. (3.12) yields

dV 3A2ux
0 (314)

The general solution of Eq. (3.14) is

ux=A + B1E, + CeA§ + De-^ (3.15)

where A = V(3/ar) and A, B, C and D are integration constants.
Substituting Eq. (3.15) into Eq. (3.13), the general solution of u2 is
obtained as

^ (3.16)
88 Strength of continuous-fiber composites

Ay Bf C and D in Eqs. (3.15) and (3.16) can be determined from

Eq. (3.7) and the following boundary conditions:
5 - 0 = 0, (P,)e=o = 0, ( P i ) 5 - = Po (3.17)
Finally, the fiber displacements and axial loads are obtained

(3.18)

(319)
P2 = Po(l + i e - « )
Values of Px and P2 in Eq. (3.19) are shown in Fig. 3.4. The stress
concentration factor of this model, (P2/P0)%=o> is 1.5. According to
Eq. (3.19), the distributions of fiber displacements and axial loads
are functions of the material constant A. However, the stress
concentration factor is independent of A. The above treatment has
been extended to composites containing a finite number of fibers
with any number of adjacent fiber breakages on the same transverse
plane.

3.3.1.2 Multi-filament failure

Hikami and Chou (1984b, 1990) have examined the two-
dimensional multi-filament failure problem of unidirectional fiber

Fig. 3.4. Variations of fiber axial forces.

__
 Stress concentrations due to fiber breakages 89

composites, focussing specifically on the stress concentration factors

of fibers adjacent to the cracks. The physical problems are analyzed
by the two-dimensional shear-lag method under two loading condi-
tions: (A) uniform tensile force applied to all fibers at infinity (Fig.
3.5), and (B) concentrated force dipole applied at a particular fiber,
n = b — af on the crack plane (Fig. 3.7).
These analyses are unique in that the general solution of the
governing equations of the elastic field has been obtained in explicit
forms in terms of the Legendre polynomials for the loading
condition (A). Based upon this solution, closed form expressions of
stress concentration factors in all fibers have been derived. These
analyses also provide rigorous proofs of both Hedgepeth and Van
Dyke's inspection (1967) on the general form of the tensile stress
concentration factor at the tip of a crack and Fichter's inspection
(1969) on the general form of the shear stress concentration factor
for the loading condition (A). Since there exists a reciprocal
relation between the influence function matrices for the loading
conditions (A) and (B), the solution for the condition (B) can be
readily derived from the solution for the condition (A).
The analyis considers a two-dimensional unidirectional con-
tinuous-fiber composite containing a slit notch in the transverse
direction, as shown in Fig. 3.5. The fiber direction is taken along
the x axis. The broken fibers are denoted as n = 1, 2, 3, . . . , b,
starting from the left tip of the notch with b being the total number
of fibers in the notch.
Under the assumption of shear-lag analysis, the matrix material
transfers only shear force, in(x), per unit fiber length between two
adjacent fibers. Thus fn(x) is related to the difference of displace-
ments un(x) in the fiber direction as

fn(x) = j{un+1(x)-un(x)} (3.20)

where G is the effective shear modulus of the matrix, and h is the

effective fiber spacing. The tensile force Pn(x) per unit thickness in
the nth fiber is related to the displacement by

Pn(X) = E d ^ (3.21)
djc
where d is the width of the fiber. The equilibrium of forces in the x
direction gives

^+tn-fn^ =0 (3.22)
90 Strength of continuous-fiber composites

The non-dimensionalized axial force, displacement and coordin-

ate are given, respectively, by:

(3.23)

I = \/(G/Edh)x
Then, the equilibrium equation (3.22) can be written as

(3.24)

The boundary conditions are:

Fn(0) = 0 (lsn
£/„(<)) = 0 (n<0, (3.25)

Fn(±=o) = 1 (all n)

Fig. 3.5. Model of a multi-filament crack in a unidirectional composite

under uniform force at infinity (After Hikami and Chou 1990.)

n = -1 0 1 2 3
t b b+1
i i ; \ 1

a=b- 1
t f t \\T
2 1 0 5 = 1 2 3

I
 Stress concentrations due to fiber breakages 91

for loading condition (A), and

Fn(0) = 0 (n = 1, 2,..., b - a - 1, b - a + 1 , . . . , b)
Un(0) = 0 (n<0, n>b + l)
Frt(±oo) = 0 (all/i) ('

for loading condition (B).

The general solutions of the multi-filament failure problem have
been obtained explicitly by Hikami and Chou (1990) using the
Legendre polynomials and Fourier transformation. The stress
concentration factors in all fibers on the crack plane are given in
closed forms. First, for the loading condition (A), the stress
concentration factor of the sih fiber ahead of the tip of a crack
containing b broken fibers is given by

2s • (2s + 2) - (2s + 4) • • • (2s + 2b - 2)

X
(2s - 1) • (2s + 1) • (2s + 3) • • • (25 + 2b - 3) • (2s + 2b-l)
(3.27)
As a special case of Eq. (3.27), the stress concentration factor in
the first intact fiber (s = 1) adjacent to b broken fibers is
4-6-8---(2fe + 2)
Afc- (3.Z8)

Hedgepeth (1961) deduced Eq. (3.28) by inspecting the numerical

results of the cases b = 1,2, ... ,6. This inspection on the general
form of the stress concentration factor has been rigorously proven
by Hikami and Chou. Figure 3.6 depicts the numerical results for

Furthermore, the maximum shear stress takes place in the matrix

at the tip of the crack. Thus, the dimensionless displacement at the
crack tip £4(0) is termed the maximum shear stress concentration
factor, 5max. Hikami and Chou (1990) have obtained
o _ "(26-1)!
max
~ 22b[(b-l)\f
Fichter (1969) deduced the above result by calculating the cases of
b = 1,2, ... ,6. The axial stress in fibers away from the crack plane
has also been obtained.
In the case of loading condition (B), Fig. 3.7, Hikami and Chou
92 Strength of continuous-fiber composites

Fig. 3.6. Stresses concentration factor K% in the (b + s)th fiber, b denotes

the number of broken fibers; 5 = 1 corresponds to the special case of
Hedgepeth (1961). (After Hikami and Chou 1990.)

Ki

1 2 10 12

Fig. 3.7. Model of a multi-filament crack in a unidirectional composite

under concentrated force dipole in the (b — a)th fiber on the crack plane.
(After Hikami and Chou 1990.)

n = -1 0 1 •b-cf b\b+1

I I I I. I

a = b - 1'
t t t \ \
2 1 0 5=1213
 Stress concentrations due to fiber breakages 93

(1990) assume that a unit force dipole is applied on the (b - a)th

fiber. Then the more general cases with multiple dipoles can be
obtained by the linear combination of the solutions of the simple
problem.
The closed form solution of the stress concentration factor at the
sth fiber in front of the tip of a crack containing b fibers and a unit
force dipole at the n (=b — a)th fiber is given as
1(2a + 1)!!(2b-2a-1)!!(2s-3)!!(2s + 2 b - 2 ) ! ! 1
6
~2 (2a)!! (26 -2a -2)!! (2s - 2 ) ! ! (2s + - l ) H (s + a)
(3.30)
where !! denotes double factorial (i.e. nV. = (n\)l). The highest fiber
stress concentration takes place at the edge of the crack (s = 1)
(2a+ 1)!! ( 2 6 - 2 a - 1 ) ! ! (26)!!
(3.31)
(2a + 2)!! (2b - 2a - 2)!! (2b + 1)!!
Figure 3.8 depicts the numerical results for K\'a.
For a semi-infinite crack the stress concentration factor at the 5th
fiber from the crack tip due to the unit applied force dipole at the

Fig. 3.8. Stress concentration K\'a in the (b + l)th fiber when the unit load
is applied at the (b — a)th fiber, b denotes the number of broken fibers.
(After Hikami and Chou 1990.)

0.5 -
94 Strength of continuous-fiber composites

ath fiber has the following value:

The axial fiber stress distributions away from the fracture plane for
both loading conditions (A) and (B) also have been obtained by
Hikami and Chou (1990). Also Fukuda and Kawata (1980) have
shown in their analysis of a finite number of fibers that the stress
concentration factor tends to that of Hedgepeth as the total fiber
number increases.
The static stress concentration factors in a layer of unidirectional
composites containing dacron fibers imbedded in a polyure-
thane elastomer have been measured by Zender and Deaton
(1963). The number of fiber breakages in this experiment
is controlled by partially slitting the specimens in the transverse
direction. The slit length determines the number of broken
fibers. The results of the experiments show reasonably close
agreement with the theoretical analysis. It should be noted that
although the broken fibers induce the adjacent fibers to fail in the
vicinity of the cut, the chances are that such a location is not
the weakest location of the fiber. This has to do with the statistical
nature of fiber strength distribution and will be discussed in
Section 3.4.
The problem of static stress concentration factors in a three-
dimensional fiber array has been examined by Van Dyke and
Hedgepeth (1969). They consider square and hexagonal arrays
where a specified number of fibers are broken. Other stress
concentration problems including the effects of finite length of fibers
(Fichter 1970), relative locations of fiber breaks (Chen 1973), holes
(Kulkarni, Rosen and Zweben 1973) and notches (Zweben 1974)
also have been treated.

3.3.2 Dynamic case

When fibers are suddenly broken in a composite under
stress, the load in the broken fibers must be transferred through the
matrix to the adjacent fibers in order to restore equilibrium. Of
interest is not only the resulting static stress, but also the dynamic
overshoot which occurs during the transient phase. Hedgepeth
(1961) examined the dynamic aspect of stress concentration for the
two-dimensional fiber array as shown in Fig. 3.5. The analytical
model is also based upon the assumptions of the shear-lag analysis;
 Stress concentrations due to fiber breakages 95

that is, it is composed of tension-carrying elements connected by

purely shear-carrying material.
The formulation of boundary value problem for the evaluation of
dynamic stress concentration is outlined below. The fibers are
separated by a constant distance and are numbered from n = —°° to
n = oo (Fig. 3.5). The coordinate along the fiber is denoted by x and
the displacement of the nth fiber at the location x and time t is given
by un(x, i). Similarly, the force per unit thickness in the nth fiber is
denoted by Pn(x, t) and is given in terms of un by

Pn = Ed— (3.33)

where E and d are, respectively, the fiber Young's modulus and

width. The equilibrium of an element of the nth filament then
requires
32un G d2un
Ed ^ + —(u
+ { n + l—2u +u
2 + )
n n _ 1 ) = m (3.34)

Here, G and h denote matrix shear modulus and width, respec-

tively; m is the mass per unit area of the nth filament.
In general, for b broken filaments, let 1 < n < b denote the
broken filaments. The boundary conditions are:

P B (0,0 0 ( l n b )
( }
un(O,t) = O (n<0 orn>Hl) '
For large x, of course, the force in each filament approaches the
uniform applied force per unit thickness, Po. Thus
Pn(±co, t) = P0 (3.36)
For the time-dependent problem, the following initial conditions are
required:
Pn(x, 0) = Po

Using a Laplace transform of the time-dependent differential

equation and boundary conditions, the resulting equations are
similar in form to those of the static problem discussed in Section
3.3.1. The variation of stress concentration factor with time is shown
in Fig. 3.9 for one, two and three broken fibers. As can be seen
96 Strength of continuous-fiber composites

from Fig. 3.9, the stress concentration factor, Klb, varies with the
dimensionless time t (=t/}/(md/G)), and approaches the steady-
state value. In all cases, the first peak is the largest one and the
value of the stress at this peak determines the dynamic
overshoot.
Hedgepeth (1961) defines the dynamic-response factor as the
ratio between the maximum stress and the static stress. Values for
one, two and three broken fibers are, respectively, 1.15, 1.19 and
1.20. It can be shown that the dynamic-response factor approaches
1.27 as the number of broken fibers tends to infinity. Further
discussions of dynamic stress concentration factors are given in
Section 3.4.9.
Following the approach of Hedgepeth (1961), Ji, Liu and Chou
(1985) have investigated the variation of dynamic stress concentra-
tion along the length of a fiber next to a broken fiber. Define the
dimensionless parameter in fiber axial location as

(3.38)

The asymptotic expressions of the stress concentration factor

Fig. 3.9. The variation of dynamic stress concentration factor Klb with
dimensionless time / for b = 1, 2 and 3 (After Hedgepeth 1961.)

2.5 n

2.0

1.5

b=1
1.0

0.5

10 15
 Stress concentrations due to fiber breakages 97

K\{%, t) for the fiber s = 1 at x = 0 due to the fracture of the fiber

n = b = 1 (see Fig. 3.5) has been obtained. The results are depicted
in Fig. 3.10, and the following observations can be made: (a) the
fiber axial stress is always tensile at § = 0. For § # 0, the initial
stress induced by fiber fracture is compressive, and the magnitude
of this initial compressive stress increases with £; (b) the dynamic
stress concentration factor, which is defined by the maximum initial
tensile stress, decreases as § increases, i.e. away from the plane of
fiber fracture; (c) the dynamic stress concentration factor is appreci-
able (say, K\ (§, ?)> 1.1) within the range of 0 < £ < l . When
t > 10, the dynamic stress concentration factor results for § < 1
approach the static stress concentration values. The change of stress
concentration factor with the location on a fiber needs to be taken
into account when there is a scattering in fiber strength and
variation of fiber strength with fiber length. The results of Ji, Liu
and Chou indicate that the variation of stress concentration is
significant for § < 1, namely x is of the order of fiber diameter times

Fig. 3.10. Dynamic stress concentration factor K\ with dimensionless time

t for 0 < § < 1. (After Ji, Liu and Chou 1985.)
1.6 i-

= 0.25

10
98 Strength of continuous-fiber composites

V(£/G)- For § > 1.0-2.0, the dynamic response diminishes with

increasing t value, and the static stress concentration factor ap-
proaches 1.0; there is virtually no static stress concentration. On the
other hand, dynamic response in fiber stress concentration exists at
small i even for £ > 2; this factor needs to be taken into account in
the statistical composite strength models.
The variation of stress concentration along the length of a fiber
has implications on the dynamic failure characteristics of fiber
composites. For instance, in the experimental observation of Ji
(1982), carbon composite specimens often fracture at locations near
specimen end-tabs. The reflection and hence magnification of the
stress waves at specimen ends could cause fiber fractures at
locations away from the plane of the existing fiber breakages.

3.4 Statistical tensile strength theories

3.4.1 Preliminary
Statistics is concerned with scientific methods for collecting
and analyzing data, as well as drawing valid conclusions and making
reasonable decisions on the basis of such analysis. Spiegel (1961)
and Kirkpatrick (1974) provide introductions to the basics of
statistics. Statistical treatment of composite strength has emerged as
an important analytical tool for the obvious reason that the
strengths of brittle fibers and yarns are statistical in nature, and not
deterministic such as in metals. A concise outline of the fundamen-
tals in statistics based upon Spiegel (1961) is given below.
In collecting data concerning characteristics of a group of objects,
it is often impractical to observe the entire group or population if it
is large. A small part of the group examined is known as a sample.
Valid conclusions can often be inferred from analysis of the sample.
Because such inference cannot be absolutely certain, the language
of probability is often used in stating conclusions.
When summarizing large masses of raw data, it is often useful to
distribute the data into classes or categories. The number of
individuals belonging to each class is called the class frequency.
Figure 3.11 gives a graphical representation of the frequency
distribution of the measured strength of carbon fibers (M. G. Bader
and B. Gul-Mohammed, private communication, 1990; see also
Dhingra 1980). The relative frequency of a class is the frequency of
the class divided by the total frequency of all classes and is generally
expressed as a percentage. A histogram can be approximated by a
continuous frequency distribution curve as shown schematically in
 Statistical tensile strength theories 99

Fig. 3.12. Also shown in Fig. 3.12 is the cumulative frequency,

which, for a particular class or strength level, is the total frequency
of all classes observed at equal to and less than this particular class.
Cumulative frequency can also be presented on a relative or
percentage basis.
Several types of averages can be defined for a given frequency
distribution. The most commonly used ones may include the
arithmetic mean, geometric mean, quadratic mean (root mean
square), median and mode. The degree to which numerical data
tend to spread about an average value is called the variation or
dispersion of the data. The standard deviation is often used to
measure dispersion, and is defined as the root mean square of the

Fig. 3.11. Distributions of carbon fiber tensile strength in air at gauge-

lengths of 5, 12, 30 and 75 mm. (After Bader and Gul-Mohammed 1990.)

I I I
1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 1.5 2.0 2.5 3.0 3.5
Strength (GPa) Strength (GPa)

75 mm
12 mm

I I I
1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 1.5 2.0 2.5 3.0 3.5
Strength (GPa) Strength (GPa)
100 Strength of continuous-fiber composites

deviations from the mean. Furthermore, the variance is defined as

the square of the standard deviation, and the coefficient of variation
is the ratio of the standard deviation to the mean. The coefficient of
variation is independent of units used and it fails to be useful when
the mean is close to zero.
The probability of occurrence of an event e is denoted by
Pr = P{e} (3.39)
The probability of non-occurrence of the event is denoted by
1_pr = P{note} = l-P{e} (3.40)
Some basic relations of probabilities of events are summarized
below. Consider two events ex and e2. The probability that e2 occurs
given that ex has occurred is the conditional probability of e2
relative to ex\ it is denoted by P{e2\ex}. If ex and e2 are
independent events and hence the occurrence or non-occurrence of
e2 is not affected by el f then
P{e2 | c,} = P{e2} (3.41)
Otherwise, they are dependent events. The probability that both el
and e2 occur is denoted by
P{ele2} = P{e1}P{e2\el} (3.42)

Fig. 3.12. Relative frequency and cumulative frequency vs. fiber tensile
strength.

Fiber tensile strength

 Statistical tensile strength theories 101

For independent events, the above equation is simplified to

= P{e1}P{e2} (3.43)
In the case of three events ex, e2 and e3, Eq. (3.42) is modified to
become
P{eie2e3} = P{e1}P{e2 | Cl}P{e3 | exe2} (3.44)
If ex and e2 are mutually exclusive events, namely the occurrence of
one excludes the occurrence of the other, Eq. (3.42) becomes
P{e,e2}=0 (3.45)
Finally, the event that either e t or e2 or both occur is given by
P{ex + e2) = Pie,} + P{e2} - P{exe2} (3.46)
For the special case of n mutually exclusive events e lf e2, . . . , en,
the probability of occurrence of either ex or e2 or • • • en is then
P{e1 +e 2 + - - - + e n } =P{ex) + P{e2} +• • • + P{en} (3.47)
The applications of these relations to the probabilities of various
events in composite failure are given in this chapter as well as in
Chapters 4 and 5.
The function representing the frequency distribution in Fig. 3.12
is also known as the probability density function. The knowledge of
the probability density function is fundamental to any analysis based
upon a statistical approach. One of the well-known probability
density functions is the normal distribution given by

where x and s are the mean and standard deviation, respectively. It

can be shown, for normal distribution, that 68.27% of the cases are
included between (x —s) and (x +s), and 99.73% of the cases are
between (x — 3s) and (x + 3s). Given a continuous probability
density function p(x) the cumulative distribution function is defined
by

P(x)= f p(x)dx (3.49)

Other commonly used distribution functions may include the

binomial distribution, Bernoulli distribution, and Poisson distribu-
tion. However, the Weibull distribution (Weibull 1939a&b, 1951) is
102 Strength of continuous-fiber composites

probably best known in composite strength theories. Weibull

proposed a cumulative distribution function in the general form of
0 x<xu
^ (3-50)
exp^-^ J x>xu
where m is a shape parameter and xo is a scale parameter. The
function (x — xu)m/xo has the characteristics of being positive,
non-decreasing and vanishing at constant value of xu, which is not
necessarily equal to zero.

3.4.2 Strength of individual fibers

Coleman (1958) examined the strength of long fibers from a
common source (say, from the same spool) for the case that their
tensile strengths are independent of the rate of loading. To obtain a
form for the cumulative strength distribution function P(crf),
Coleman observed that (a) when a fiber is tested it breaks at its
weakest cross-section, (b) the strength of a fiber must be positive
regardless of the fiber length, and (c) P(o{) must be a monotonically
increasing function of af. Coleman postulated that a fiber may be
regarded as composed of a set of N non-interacting unit lengths (or
links). It is further assumed that all the links in a fiber have the
same cumulative strength distribution function P(of).
The probability that a link has a strength greater than af is
1 — P(Of), and the probability that all links do not fail at of is
[1 - P(of)]N (Eq. (3.43)). It follows then the probability that at least
one link breaks at a{ is
Pf(af) = l - [1 - P(af)f (3.51)
Pf(of) can be regarded as the cumulative distribution function of the
strength of fibers.
Coleman has shown that P(of) has the form of a Weibull
distribution. For long fibers (N—><x>), Eq. (3.51) gives the cumula-
tive probability of failure

(3.52)

Pf(of) is the probability of failure of a fiber at a stress level equal

to or less than af. Here, L is the length ratio with respect to a
reference length, ao is the scale parameter for unit fiber length ratio
(i.e. L = 1), and p is the shape parameter. Equation (3.52) implies
 Statistical tensile strength theories 103

a probability density function of

r (° (3.53)
Pt\Of) — Loo pof exp ~L\ —
L \o,
Following Coleman, the mean fiber strength, ot, and standard
deviation, s, are given by

(3.54a)

s= (3.54b)

where F denotes the gamma function. An important feature of Eq.

(3.54a) is that the fiber strength depends upon the fiber length. The
coefficient of variation, which is a function of jS only, is

• - 1 (3.55)
Of

Over the range of practical interest, /J is approximately equal to

1.2/(coefficient of variation). Thus, /3 is an inverse measure of the
dispersion of material strength. For values of /? between 20 and 2,
the coefficient of variation can be expressed approximately as
/3-0.92 .V a l ues of p between 2 and 4 correspond to brittle fibers,
whereas a value of 20 is appropriate for a ductile metal, ft is about 4
for carbon fibers, between 2.7 and 5.8 for boron fibers and about
11 for glass fibers. The factor ooL~v^ in Eqs. (3.54) is often
referred to as a characteristic strength level of the fibers (Kelly
1973, Rosen 1964).
Manders and Chou (1983a) have shown that the scale and shape
parameters of the Weibull distribution function for fiber strength
can be estimated from experimental measurements in a number of
ways. First, by taking logarithms of Eq. (3.54a), it is seen that a
graph of ln(af) against ln(L) is linear and has gradient — 1//3. The
shape parameter can be obtained in this way by testing single fibers
of a range of gauge-lengths. The second procedure is to plot the
cumulative distribution on appropriate logarithmic axes as follows.
The cumulative probability of survival is simply
P. = l - Jf (3.56)
104 Strength of continuous-fiber composites

and Eq. (3.52) can be rewritten after taking logarithms as

(3.57)

Taking logarithms a second time with a change of sign

ln(-ln(P s )) = ln(L) + p ln(af) - j8 ln(a o ) (3.58)

shows that a graph of ln(—ln(Ps)) against ln(af) is linear with

gradient jS (at fixed gauge-length).
The procedures outlined above rely on testing many separate
fibers. If a single fiber could be uniformly stressed along its length it
would fracture into a series of unequal fragments of which the
average length would decrease with higher applied stress. The
distribution of lengths between fractures at any given stress should
be exponential following Eq. (3.57), and plotting ln(Ps) against L
should give a straight line passing through the origin with gradient
—(Of/Oo)13. Taking logarithms with a change of sign gives

ln(-gradient) = p ln(af) - j8 ln(a o ) (3.59)

so that a graph of ln(—gradient) against ln(cxf) is linear with gradient

3.4.3 Strength of fiber bundles

Having examined the strength of single fibers, the strength
theory of fiber bundles can be developed (see Daniels 1945, Epstein
1948, Coleman 1958, Kelly 1973, Phoenix 1974). Following the
treatment of Coleman (1958), a bundle composed of a very large
number, M, of fibers of equal length is considered. The fibers are
further assumed to have the same cross-sectional area and the same
shape of stress-strain curves, but differ in their values of the
elongation at break. It can be shown that the probability density
function of bundle tensile strength ab (breaking load for the
bundle/total fiber cross-sectional area) tends for large M toward a
normal distribution (Eq. 3.48)

(3.60)
sb\{2n)

with a mean bundle strength

oh = otm[l-Pt(°fm)] (3.61)
 Statistical tensile strength theories 105

and standard deviation

sb = dfmV{Pf(<Tfm)[l " P f (o fm )]}M -1/2 (3.62)

Here, Pf(crfm) is the cumulative fiber strength distribution function

and ofm is the value of fiber stress a{ which gives af[l — Pf(crf)] its
maximum value, namely

— { a i l - Pf(af)]}O{=Ofm =0 (3.63)

Equation (3.63) implies that the maximum fiber stress afm is found
from the condition that at failure the load borne by the bundle is a
maximum.
Assuming Pf(Of) follows the Weibull distribution of Eq. (3.52) for
fiber length L, Eqs. (3.61) and (3.63) give, respectively,

afm = oo{LP)~yft (3.64)

and

ab = oo(LI3e)-l/fi (3.65)

where e = 2.71828 • • •. Equation (3.65) implies that the proportion

of surviving fibers is exp(-1//3). The strength of loose bundles is
lower than the mean strength of single fibers of the same length by
the ratio of Eq. (3.65) to Eq. (3.54a), which is termed the 'Coleman
factor'

(3.66)

It is noticed that when there is no dispersion in the strength of the

component fibers of a bundle ab = af. As the coefficient of variation
of the fibers increases above zero, however, the bundle strength
efficiency decreases monotonically and approaches zero in the limit
of infinite dispersion. ab/af~70% for the coefficient of variation
about 17%.
The ratio given in Eq. (3.66) is independent of the length of the
fibers so that the strength of loose bundles decreases with length in
the same way as the mean strength of single fibers. The Weibull
106 Strength of continuous-fiber composites

parameters can therefore be obtained by plotting ln(strength of

loose bundle) against ln(length) as described above for single fibers.
The above analysis is concerned with single bundles, whereas
some situations are better modeled as a chain-of-bundles, such as a
moderately twisted yarn where the link length is a frictional load
transfer length among fibers. A review of this problem is given by
Smith and Phoenix (1981).

3.4.4 Correlations between single fiber and fiber bundle strengths

Equation (3.54a) indicates that the Weibull shape para-
meter of single fiber strength can be determined from the measure-
ment of strength at several fiber gauge-lengths. There are short-
comings in such measurements. First, it is rather tedious to extract
individual fibers from a bundle and to perform numerous tests on
fibers with very small diameters. Second, the extraction of fibers
from a bundle inevitably has 'selected' the stronger ones, since the
weaker fibers are prone to damage and fracture in the process.
Third, experiments based upon laser diffraction fringes have shown
that the measured fiber diameters vary along the fiber length due to
fiber twist and the non-circular fiber cross-section.
In this section, following the approach of Chi, Chou and Shen
(1984), a theoretical expression of the load-strain relationship for a
bundle of fibers under tension is derived first. Then, two methods
for determining the two parameters of Weibull distribution for
single fiber strength are developed. This is done by analyzing the
characteristics of the load-strain curves. The open circles in Fig.
3.13 show the experimental results of a displacement-controlled
test for a loose bundle of carbon fibers.

3.4.4.1 Analysis
The correlation between single fiber and fiber bundle
strengths is established based upon the following assumptions: (1)
the single fiber strength under tension obeys the cumulative Weibull
distribution function, Pf(crf), of Eq. (3.52); (2) the relationship
between stress, af, and strain e{ for a single fiber obeys Hooke's law
up to fracture:

of = Efef (3.67)

where Ef is the fiber Young's modulus; (3) the applied load is

distributed uniformly among the surviving fibers at any instant
during a bundle tensile test.
 Statistical tensile strength theories 107

To establish the tensile load-strain (F—e) relation, Eq. (3.52) is

rewritten in terms of fiber strain:

(3.68)

Here, so is the scale parameter for unit fiber length ratio (i.e. L = 1)
and is given by
£o = Oo/Ef (3.69)
Assume iso-strain conditions for the fibers in a bundle. At an
applied strain, £f, the number of surviving fibers in a bundle, which
consists of No fibers, is
N = No[l - Pf(£f)] = No exp[-L(£ f / eo )P] (3.70)
N can be related to the applied tensile force, F> on the bundle by
F = OfAN = AEfefNo exp[-L(e { /£ o f] (3.71)

Fig. 3.13. Comparison of a theoretical F-ef curve (solid line) with

experimental data (open circles) for carbon fiber, E{ = 225 GPa, df=
No = 1000, P = 4.5 and so = 0.026. (After Chi, Chou and Shen 1984.)

60

50

40

30

20

10

^ l
0.0 0.005 0.010 0.015
108 Strength of continuous-fiber composites

Equation (3.71) is the relationship of F-ef for a bundle of fibers

under tension, where A is the cross-sectional area of a single fiber.
If A, No, L, Ef, e0 and jS are known, the F - s f curve for a bundle of
fibers can be drawn. The solid line in Fig. 3.13 shows the result of
the theoretical prediction.
According to Eq. (3.71), the F—ef curve is continuous and
smooth. After reaching the point of maximum load, F max , the
tensile force on the bundle decreases gradually to zero. The slope of
the curve, 50, at e{ = 0 is
So = AEfNo (3.72)
and the tensile load defined by the tangent line of the F-e f curve at
£f ~ 0 is
F*=AEiNoef (3.73)
Based upon the F—ef relation, the survivability of single fibers in the
bundle can be determined from Eqs. (3.71) and (3.73)

^=l-jPf(£f) = Ps (3.74)

Next, the strain corresponding to the maximum load on the F-e{

curve, £m, is obtained from dF/ds f = 0

(3 75)
.
Thus, the maximum load is

(3.76)

From Eqs. (3.72), (3.75) and (3.76), the slope of the straight line
connecting the origin and the point (F max , sm) in Fig. 3.13 is

K) (3.77)

As a result,

(^) (3.78)

3.4.4.2 Single fiber strength distribution

Based upon the analysis of the fiber and bundle strength
relations, Chi, Chou and Shen (1984) proposed the following two
 Statistical tensile strength theories 109

methods for determining single fiber strength distribution (shape

parameter (5 and scale parameter eo) from measurements on fiber
bundles, and constructing the theoretical F-sf curve.

Method (A)
The method is based upon Eqs. (3.68) and (3.74) and the
experimental F-E{ curve. The procedure is outlined below:
(1) Calculate So from Eq. (3.72) and the data of A, Ef and No
of the fiber bundle.
(2) Calculate F* from Eq. (3.73), F* = SoEf. Measure F from
the F—sf curve. Then determine from Eq. (3.74) the fiber
survivability as a function of strain, Ps(£f) = FIF*.
(3) The shape parameter, /?, can be obtained from the gradient
of the graph of ln(-ln(Ps)) vs. ln(£f), based upon the
relation
ln(-ln(Ps)) = ln(L) + j8 ln(ef) - j8 ln(eo) (3.79a)
(4) The scale parameter, eo, is determined either from Eq.
(3.75) using the measured em value, or from the value of
ln(L) — /3 \n(eo) measured from the graph of ln(-ln(Ps)) vs.

Method (B)
In this method, Fmax and sm are known from experiments.
The calculation steps are:
(1) Determine 5O, 0 and eo from Eqs. (3.72), (3.78) and (3.75),
respectively.
(2) From Eqs. (3.71) and (3.72), the F-e{ relation can be
written as

F = 5 o £ f exp(-L(—) ) (3.79b)
\ \£cr '

3.4.5 Experimental measurements of Weibull shape parameter

It is understood that the shape parameter /3 gives a
measurement of the scattering of the strength data. On a Pf(of) vs.
of plot, the range of strength distribution is narrower for higher p
values. The discussions of Sections 3.4.2-3.4.4 for the estimation of
the Weibull shape parameter are summarized below (see Manders
and Chou 1983a; Chi, Chou and Shen 1984).
110 Strength of continuous-fiber composites

Single fibers
(i) Variation of mean strength with length: The method
requires tests at different gauge-lengths. Fiber diameters
are measured to obtain true stress.
(ii) Distribution of strength at fixed gauge-length: Diameters
are measured to obtain true stress. The method measures
both inherent variability, and also artificial scatter intro-
duced by experimental techniques.
(iii) Distribution of lengths between multiple fractures of a
single fiber: Estimate is based on strain, not stress. The
method requires correction for non-uniformity of strain
near fractures.

Loose bundles
(iv) Variation of mean strength with length: The method
assumes identical fiber diameters and stiffness,
(v) Proportion of surviving fibers is obtained from the
load-strain curve. Estimate is based on strain not stress.
The method assumes fibers are identical,
(vi) Determination of the initial slope of the load/strain curve
and the strain corresponding to the maximum load on the
bundle.
Manders and Chou (1983a) have established the Weibull shape
parameter based upon the methods (i)-(v) by performing tests on
a single batch of PAN-based carbon fiber (Hercules AS-4, 12 000
filament unsized tow) while the loose bundle tests (iv) and (v) are
carried out with the E-glass fiber (St. Gobain, vetrotex type DCN56
filament, unsized tow). Chi and Chou (1983), and Chi, Chou, and
Shen (1984) have examined methods (i), (ii), (v) and (vi) using
Thornel-300 carbon fibers and bundles containing 1000 fibers.

3.4.5.1 Single fiber tests

In order to obtain the strengths of single filaments and their
distributions, it is necessary to measure the diameters and ultimate
tensile load of the filaments. For the measurement of filament
diameters, Chi and Chou (1983) used a helium-neon laser, and the
diameters were determined from the laser diffraction fringes (see
Lipson and Lipson 1981) The results indicate an average filament
diameter of 7.12 //m with the standard deviation of 0.2 /im.
 Statistical tensile strength theories 111

Fiber strength measurements are performed for fiber gauge-

lengths of 10, 30 and 60 mm, and the number of measurements are
80, 81 and 64, respectively. The results are presented on the
Weibull probability paper as shown in Fig. 3.14. Here o{ denotes
fiber ultimate strength; Pf(af) is the fiber cumulative probability of
failure at stresses equal to or less than of and ln{—ln[l — Pf(af)]} is
a representation of failure probability. The variations of fiber failure
probability with strength can be approximated as linear with the
exceptions of the low strength range for 60 mm length fibers, and
the high strength range for 10 mm and 30 mm length fibers.
The Weibull shape parameter, /3, can be obtained by following
method (i), by plotting the mean fiber strength ln(af) vs. fiber
length ln(L) (see Eq. (3.54a)) as shown in Fig. 3.15. A measure-
ment of the slope of the straight line gives the value of /3 = 6.2. It is
worth noting that because of the high scatter in strengths a large
number of tests needs to be performed to determine with high
accuracy whether the Weibull distribution is an accurate description
of strength, and this is where the loose bundle approach is

Fig. 3.14. Strength distributions of single filaments on Weibull probability

paper. (After Chi and Chou 1983.)

of (MPa)
800 1000 1200 15001700 2000 2500 3000 3500 4500 5500
1 I 1 I I
0.9
o L = 60 mm 0.8
• L = 30 mm
* L = 10 mm 0.6

0.4
-1
0.2

-2
0.1

0.05
-3
0.03

0.02
-4
1 * 1 1 1 1 1 1 1
6.7 6.8 7.0 7.2 7.4 7.6 7.8 8.0 8.2 8.4 8.6
In Of
112 Strength of continuous-fiber composites

advantageous. If method (ii) is followed, then the slope is measured

from the linear portion of the data of Fig. 3.14 for ln{ln[l - Ff(af)]}
vs. ln(crf) (see Eq. (3.58)). The distributions of the three sets of data
are reasonably linear and parallel, and an average of the approxim-
ate gradients is taken to obtain the shape parameter of 5.3.
Method (iii) requires multiple fracture tests of a single fiber. In
the experiments of Manders and Chou (1983a), single carbon fibers
are bonded to the surface of a 2 mm thick filled PVC carrier sheet
using a film of polystyrene adhesive approximately 50 fim thick. The
fiber is strained to successively higher levels by bending the carrier
strip around mandrels of decreasing radii. The strain in the fiber is
virtually uniform because the ratio of the carrier thickness to fiber
diameter is —200. The combination of adhesive and carrier is found
to be quite resistant to repeated straining, and facilitates visual
location of the fiber fractures. At each strain level the lengths
between fiber fractures are measured by travelling microscope and
are ranked and plotted as the cumulative distribution on the
logarithmic axes. According to Eq. (3.57), the distributions should
be linear and pass through the origin, but, while they are relatively
straight, they intersect the fracture spacing axis at some positive
intercept. The minimum crack spacing given by the intercept
represents the effective 'unstressed' length of fiber over which the

Fig. 3.15. Relationship between filament average strength and gauge-

length. (After Chi and Chou 1983.)

5000

4000

3000 -
—'—

2000 -

1000 1 1 1 1 1 1
10 20 40 70 100
In (/(mm))
 Statistical tensile strength theories 113

load builds up. The logarithm of the gradient has been plotted
against the logarithm of strain and at low strains, the curve is
relatively linear with a gradient corresponding to /3 = 6.4. At high
strain the curve becomes horizontal because the fiber debonds from
the adhesive film and no new fractures occur. Despite this short-
coming the technique is able to measure the shape parameter for
shorter fibers than the other techniques.
Henstenburg and Phoenix (1989) have considered the problem of
measuring the Weibull parameters for fiber strength using data from
a multiple fracture test of a single fiber. Using a Monte-Carlo
approach they arrived at a simple method which applies to fibers of
length equal to the mean fragmentation length.

3.4.5.2 Loose bundle tests

In the loose bundle tests of Manders and Chou (1983a),
based on method (iv), tows of different lengths are cemented into
grooved end-tabs while particular care is taken to ensure that none
of the fibers are slack. Manders and Chou obtained between five
and ten results for each gauge-length, and they are plotted in the
same way as for the single fiber tests in Fig. 3.16. The cross-sectional
area of the tow is calculated from the manufacturer's value of its
density and weight per length. Because each failure of a loose
bundle involves the independent fracture of many fibers, there is
much less scatter than for the single fibers. According to Manders
and Chou, the mean strength ratios of loose bundles and single
carbon fibers of the same length range from 0.67 to 0.85 for fibers
with lengths between 10 and 200 mm, and this compares quite well
with the theoretical Coleman factor which ranges from 0.65 to 0.76
for fibers with )3 equal to 5 and 10, respectively (Coleman 1958). The
discrepancy may be due to the fact that the strengths are not
perfectly Weibull distributed, and that the optical technique for
measuring fiber diameter overestimates the cross-sectional area of
non-circular crenelated fibers. Also, fiber breaks may be pre-
existing in the bundle, becoming more noticeable at longer bundle
lengths.
It has been noticed that both single fibers and loose bundles show
an increase in strength variability at longer gauge-lengths. This
could be interpreted as the influence of a relatively small population
of severe and broadly distributed flaws. The majority of short
gauge-lengths would not contain one of these severe flaws and the
population would have little influence on the mean strength, but
longer fibers would be more likely to contain one or more such
114 Strength of continuous-fiber composites

flaws and their mean strength would be significantly lowered. The

observation of similar behavior in glass fiber suggests that a 'double'
Weibull distribution with two shape and scale parameters may be
more appropriate (Metcalfe and Schmitz 1964; Harlow and Phoenix
1981a & b). It is also noticed, in the case of loose bundles, that the
recoil and entanglement of failed fiber causes neighboring fibers to
fail, thereby weakening the bundle.
In the loose bundle tests of Chi, Chou and Shen (1984), the shape
parameter and scale parameter were determined based upon
methods (v) and (vi), which correspond to methods (A) and (B) of
Section 3.4.4.2. The relevant data are: No = 1000, fiber diameter =
7 jum, £ f = 255GPa and gauge-length = 60 mm. The shape para-
meters obtained from methods (v) and (vi) are 4.6 and 4.5,
respectively. The scale parameter, £o, corresponding to a fiber of
unit length (l mm in this case), is 0.026 for both methods. The
experimental data points indicating the load-strain (F-£f) relation-
ship are shown in Fig. 3.13. The consistency between the theory and
experiment is rather satisfactory in the range of bundle strain not
much greater than sm.

Fig. 3.16. Variation of mean strength with length for loose bundles of
carbon and E-glass fibers. (After Manders and Chou 1983a.)
Length (mm)
10 20 50 100 200 500 1000
22.0

3.0

21.5
2.0

a.
.5 O

1.0
20.5

20.0 0.5
-5 -4
In (length (m))
 Statistical tensile strength theories 115

3.4.6 Strength of unidirectional fiber composites

This section deals with statistical strength theories of
unidirectional fiber composites. Upon the fracture of a fiber, the
load originally carried by the fiber needs to be transferred to its
neighboring fibers. A simple approximation of the load redistribu-
tion is to assume that the load is shared equally by all the unbroken
fibers. A more precise treatment takes into account the local
concentration of load on neighboring fibers. A Monte-Carlo simula-
tion is also presented to illustrate the statistical nature of composite
failure.

3.4.6.1 Equal load sharing

In general, the high-strength high-stiffness fibers used in
composites are brittle and their tensile strength should be charac-
terized statistically. Parratt (1960) notes that the tensile failure of
composites reinforced with brittle fibers occurs when the fibers have
been broken up into lengths so short that any increase in applied
load cannot be transmitted to the fibers because the limit of
interface or matrix shear has been reached. Rosen (1964), following
Gucer and Gurland (1962), considers fibers as having a statistical
distribution of flaws or imperfections that result in individual fiber
breaks at various stress levels. The fracture initiated in a fiber is
contained by the matrix material. Composite failure occurs when
the remaining unbroken fibers, at the weakest cross-section, are
unable to resist the applied load. Then composite failure results
from tensile fracture of the fibers. In Rosen's failure model, the
composite is assumed to be strained uniformly and the load in a
broken fiber is distributed equally among the remaining unbroken
fibers in a cross-section. Harlow and Phoenix (1978a) have labelled
such a model as equal load sharing. Scop and Argon (1967) also
have dealt with the problem of equal load sharing in their treatment
of the strength of laminated composites.
Figure 3.17 depicts Rosen's failure model. In the vicinity of an
internal fiber end in such a composite, the axial load carried by the
fiber is transmitted by shear through the matrix to adjacent fibers
(see Section 3.3.1). A portion of the fiber at each end is, therefore,
not fully effective in resisting the applied stress. At some distance
from an internal break, the fiber stress will reach a given fraction of
the undisturbed fiber stress. Rosen considers that the fiber length 6,
measured from the fiber end, over which the stress is less than a
given fraction (i.e. 90%) of the uniform stress that would exist in
infinite fibers, as ineffective. 8 is thus known as the ineffective
116 Strength of continuous-fiber composites

length. The model composite in Fig. 3.17 is assumed to be

composed of a series of layers of height 5. The segment of a fiber
within a layer may be considered as a link in the chain that
constitutes the fiber. Each layer is then a bundle of such links and
the composite is a series of such bundles.
The treatment of a fiber as a chain of links is appropriate to the
hypothesis that fracture is a result of local imperfections in the
fibers. The links may be considered to have a statistical strength
distribution that is equivalent to the statistical flaw distribution
along the fibers. Rosen defines the link dimension by a shear-lag
analysis of the stress distribution in the vicinity of a fiber end (see
Section 3.3.1). The length of the composite specimen is designated
by L and the number of links is given by N = Li d.
The relationship between fiber strength and the strength of links
has been briefly discussed in the formulation of Eq. (3.51).
Obviously, the probability density function P\(ox) for fiber links can
be characterized if the experimental data on fiber strength distribu-
tion Pf(of) are known. Suppose that the fibers are characterized by a
strength distribution of the Weibull type (Eq. (3.53)), the link
strength density function can be readily written as

(3.80)

Fig. 3.17. Chain-of-links model for a unidirectional fiber composite.

i i i i
II _ ' 6 J I A
r - - r - " r ~r _ ~r -1

r - - r _ " r ~r - ~r 1-1- 1
r - ~ r - - r - - r - ~r II \N

r - ~ r - ~r - ~ r - " r -1
i—

r _ - r _ - r - - r - " r II - 1
r~ 4 "' J
L • FJKpT
Fiber L Matrix
 Statistical tensile strength theories 111

For a bundle of links and a large number, M, of fibers, the

distribution of bundle strength ph(ob) and the mean bundle strength
ah are given by Eqs. (3.60) and (3.61), respectively.
The bundles may be treated as links in a chain, which now
represents the whole composite of Fig. 3.17. The weakest link
theorem can again be applied to define the failure of the composite.
For N bundles forming a chain (composite) the probability density
function pc(oc) for the average fiber stress at composite failure, ac,
is given by
pc(°c) = NPh(ac)[l - P h {a c )} N - 1 (3.81)
where

ph(o)do (3.82)

The notations of pc(oc), pb(oh)y pf(o{) and PI(CR,) have been used to
denote the strength density functions of the fibers at the level of
composite, bundle, fiber and link, respectively. Thus, it is under-
stood that ac, ab, a{ and ox all refer to stresses in the reinforce-
ments; the contribution of matrix to composite strength is not
considered.
The most probable composite failure stress a* is obtained by setting

-£-[Pc(oc)]Oc=Oc*=0 (3.83)

Following Rosen (1964), the substitution of Eq. (3.81) into Eq.

(3.83) yields

It can be seen from Eq. (3.62) that, for composite dimensions large
relative to fiber cross-section ( M » l ) , sb~ *0 and Eq. (3.84) is
reduced to the mean bundle strength expression of Eq. (3.65)

ac* = ao(6pe)~up (3-85)

When the fiber volume content is considered, the tensile strength of
the composite is given by Vf a*. In Eq. (3.85), the ineffective length
S can be determined from the stress analyses discussed in Section
3.3.1. It is obvious that the composite strength is enhanced due to a
reduction in fiber ineffective length and fiber strength dispersion.
The statistical nature of fiber fracture and the resulting weakest link
mode of failure have been demonstrated experimentally in a
118 Strength of continuous-fiber composites

glass/epoxy system by Rosen (1964). This experiment also points

out the very significant phenomenon in brittle fiber composites:
fiber breakages may exist in a composite of continuous fibers at
stress levels well below the maximum load.
If the composite strength (Eq. (3.85)) is compared with the mean
strength of the tested fibers of length L (Eq. (3.54)), some
interesting conclusions can be drawn (Rosen 1970). Figure 3.18
shows that for reference fibers of ineffective length 6, the strength
of the composite is less than the mean fiber strength. When the fiber
length is greater than d, the composite strength is larger than the
mean fiber strength of a fiber bundle of length L>76. Also for a
fiber strength coefficient of variation (s/o) less than 15% (or the
shape parameter )3 > 8), the composite strength is close to the mean
fiber strength, as shown in Fig. 3.18.

3.4.6.2 Idealized local load sharing

When a fiber breaks in a composite there is inevitably a
redistribution of load in the vicinity of the fiber breakage. Thus,
local load sharing takes place (see Zweben 1968; Scop and Argon
1969; Zweben and Rosen 1970; Fukuda and Kawata 1976b; Harlow
and Phoenix, 1978a&b; Harlow 1979; Phoenix 1979). The localized
nature of stress redistribution around a random fiber break has been
discussed in Section 3.3. Zweben (1968) first considered the

Fig. 3.18. Composite strength/mean fiber strength vs. /3 at various L/d

values.
 Statistical tensile strength theories 119

micromechanical stress transfer process and the probabilistic aspects

of the generation of clusters of breaks to form catastrophic breaks.
Fukuda and Kawata (1976b) generalized the original concept of
Zweben and derived the cumulative strength distribution for the
composite.
In the following, an analysis is presented under the Weibull
distribution for fiber strength, and somewhat simplified assumptions
on local fiber load sharing but with the advantage that various
quantities can be worked out either exactly or asymptotically. The
result is that insight can be gained on the approximate Weibull
behavior for composite strength where the Weibull parameters for
the composite will be connected to various fiber and matrix
properties, and in particular to the composite volume. The size
effect law for the composite will also be discussed. Most of the
features have been experimentally observed but have been difficult
to explain. The ideas for this section are taken from Harlow and
Phoenix (1978a&b, 1979, 1981a&b); Smith (1980, 1982); Phoenix
and Smith (1983); Smith et al. (1983); and Phoenix, Schwartz and
Robinson (1988).
The model considered is the planar, chain-of-bundles model
of Fig. 3.17 where M is the number of fibers and N is the number of
bundles each with fiber elements of length 8, which might better be
termed 'the effective load transfer length'. Following the notation of
Phoenix, the cumulative distribution function for the failure of a
single fiber element of length 8 is taken as the Weibull distribution
and expressed as
F(o) = 1-exp{-(o/o6)p} A>0 (3.86)
where o is the fiber stress, and (3 and a6 are the Weibull shape and
scale parameters, respectively. (At this point it should be men-
tioned that 8 should take into account certain statistical aspects of
fiber strength which modify its magnitude somewhat as described by
Harlow and Phoenix (1979), and Phoenix, Schwartz and Robinson
(1988). Roughly, 8 varies inversely as the shape parameter /3.)
According to principles discussed earlier, the strength of a fiber
element of length 8 can be expressed in terms of those for a longer
reference length L (used, say, for tension tests) according to
/8\
a6 = oL(jj (3.87)

where A L is the Weibull scale parameter for fiber strength at the

reference length. Often o8 will be about double A L in magnitude.
120 Strength of continuous-fiber composites

The local load-sharing rule is 'idealized' as follows: In a bundle, if

the stress is nominally a (ignoring the matrix), a surviving fiber
element carries load Kro, where
Kr = l + r/2, r = 0, 1,2,3, . . . (3.88)
and r is the number of consecutive failed elements immediately
adjacent to the surviving element (counting on both sides). At the
same time a failed fiber element carries no stress over length 6.
Essentially the load of a failed fiber is shifted equally onto its two
nearest surviving neighbors, one on each side. This rule is more
severe than the true situation where the stress redistribution is
somewhat more diffuse, as described say by Hedgepeth (1961), but
it captures the essential features and has the advantages of
simplicity and being fully described for all configurations.
Before proceeding with an approximate analysis of this model, it
is useful to review an extensive numerical analysis performed by
Harlow and Phoenix (1978a&b), where the basic insight into its
behavior was uncovered. To eliminate boundary effects, they
considered circular bundles (composite tubes), and studied the
behavior of the cumulative strength distribution as the bundle size
M increases. They defined GM{a) as the cumulative distribution
function for failure of a bundle with M fibers under the stress oy and
worked out exact formulas for GM(a) for M up to 5 by considering
all configurations of failed and surviving fibers and all ways that
failure could proceed through these configurations and then sum-
ming all probabilities for these ways. For example, for M = 2,
G2(a) = F{of + 2F(a)[F(2a) - F(a)]
= 2F(a)F(2o) - F{of (3.89)
where in the intermediate step the first term represents direct
failure under the applied stress of both fiber elements, and the
second term represents the two ways one element can fail under the
direct stress and the other under the overstress, which is naturally
taken as 2a in this situation (rather than 3a/2). For M = 4, they
obtained by a tedious calculation
G 4 (A) = 1 6 F ( 4 A ) F ( 2 A ) F ( 3 O / 2 ) F ( A ) - 4F(4a)F(2a)F(a)2
- 4F(4o)F(3o/2)2F(o) + 4F(4a)F(a)3
2
- 8F(2a) F(3a/2)F(a)
+ 2F(2a)2F(af - 8F(4o)F(3o/2)F(o)2
+ 4F(3o/2)2F(o)2-F(a)4 (3.90)
 Statistical tensile strength theories 121

Generally no simple pattern emerged except that each term

involved a product of M quantities in F. The evaluation procedure
was automated on a computer, but results were only obtained at
that time for M up to 9 because of the tremendous increase in
computational complexity resulting from the increasing number of
ways the bundle can fail as the bundle size increases. (Even with
present supercomputer capability the limit is still about M = 14.) At
the same time we desire results for M orders of magnitude larger.
Suspecting an eventual weakest-link type relationship, Harlow
and Phoenix (1978b) considered plotting the 'renormalization'

WM(a) = l - [l - GM(o)]VM (3.91)

since in reverse this yields the weakest-link relation

GM(a) = l - [l - WM(a)]M (3.92)

They discovered an extremely rapid numerical convergence
WM(o)-*W(o) asM^oo (3.93)

where W(a) was called the characteristic distribution function for

failure. This convergence is shown in Fig. 3.19 for the Weibull
shape parameter /? = 5, which is typical of brittle fibers. The
coordinates are Weibull coordinates (ln{—ln(l - W)} vs. \n(a/od))
wherein a Weibull distribution always plots as a straight line. For
each value of a the convergence is abrupt at some value of M,
which increases slowly with decreasing values of a. Also the
convergence becomes complete far into the lower tail of W(a)
(probabilities below 10~10) for M = 9. In an extremely complex
calculation, Harlow and Phoenix (1981a&b) uncovered the analyti-
cal character of W(a) in terms of the largest eigenvalue of a
Markov recursion matrix. It suffices to say here that W(a) has
no simple analytical form, though shortly we will develop an
approximation which will give us considerable insight.
The importance of W(o) is that, from Eq. (3.92), the distribution
function for bundle failure can be given extremely accurately by the
approximation
GM(o) ~ l - [l - W(o)]M (3.94)
and this works for M many orders of magnitude larger than the
values used in the calculation of W(a) on the computer. Perhaps
one should note that any boundary effects, which may come into
play for small bundles, are being ignored.
122 Strength of continuous-fiber composites

Because the composite is seen as a weakest-link arrangement of

its N bundles (Fig. 3.17), and the bundles are treated as statistically
independent, the cumulative distribution function for the failure of
the composite, denoted as HM N{o), is given as
HM,N(o) = 1 - [1 - GM(o)Y (3.95)
Combining Eqs. (3.94) and (3.95) and writing V = MN yields the
accurate approximation
HM>N(o) ~ l - [l - W(o)]v (3.96)
which surprisingly, perhaps, is a result which is symmetric in M and

Fig. 3.19. Convergence of the renormalized distribution functions WM(o)

to the characteristic distribution function W(o) as M increases. (After
Harlow and Phoenix 1978b.)

-1.6

0.2 0.3 0.4 0.5 0.6 0.7 0.8 1.0

o/a<j
 Statistical tensile strength theories 123

N. Note that by the binomial expansion HMN(a) ~ VW(a). Thus if

V is large, say 106 elements, it is necessary to know W(a) where its
value is much less than 10~6. As mentioned, this is provided for in
Fig. 3.19. Note that despite the fact that Eq. (3.96) is a 'weakest-
link' relation, in terms of V = MN elements, there is no identifiable
and independent material element to which one can attach W(o).
At best, W(a) characterizes the effects of local failure events which
are actually statistically dependent.
Figure 3.20 displays W(a) for values of /3 from 3 to 50. Now Fig.
3.20 can be used to construct a figure for HMN(o) upon noting that
l n { - l n ( l - / / ) } = l n { - l n ( l - W ) } + \nV, which on Weibull prob-

Fig. 3.20. Characteristic distribution function W(o) for various values of

the Weibull shape parameter p for fiber strength. (After Harlow and
Phoenix 1978b.)

-1.6

0.2 0.4 0.5 0.6 0.7 0.8 1.0

o/o6
124 Strength of continuous-fiber composites

ability paper amounts to a simple translation of each curve upward

(or the left-hand scale downward) the amount \n(V) on the
right-hand scale provided for this purpose. Figure 3.21 shows the
result of such a translation for V = 106 elements, which amounts to
a display of the original region on Fig. 3.20 below 10~5. This yields
plots of the cumulative distribution function of composite failure,
HM N{o), for various /3 for a relatively small composite specimen.
Several features of Fig. 3.21 warrant discussion. First, all the lines
are approximately straight over a very wide probability range,
which suggests that the strength of a composite approximately (but
not exactly) follows a Weibull distribution. In fact, an empirical plot
to cover the probability range shown would require testing about

Fig. 3.21 Cumulative distribution function HM N(o) for composite strength

for volume MN = 106 and various values of the fiber shape parameter /3.
(After Harlow and Phoenix 1978b.)

-1.4 -1.0 -0.6 -0.2

10

0.3 0.4 0.5 0.6 0.7 0.8 1.0

o/o6
 Statistical tensile strength theories 125

20000 specimens, and using standard statistical techniques it is

probable that a Monte-Carlo simulation would not lead to rejection
of the hypothesis that the Weibull distribution is actually the correct
distribution! Second, the lines show only a modest change in slope,
by a factor of less than three, as the original Weibull shape
parameter for the fiber /? decreases from 50 to 3, which is a factor of
more than ten. Since the slope is directly proportional to the
Weibull shape parameter, this indicates that the effective Weibull
shape parameter for the composite decreases modestly, from about
50 to 20 as that for the fiber decreases drastically, from about 50 to
3. On the other hand, the horizontal location of the plots is quite
strongly influenced by the value of /?, which suggests that an
increase in variability in fiber strength substantially decreases
composite strength. It is seen, for example, that the median
strength drops from about 0.75od, to about 0.2o6 as /3 drops from
50 to 3. Note also that the median strength of the composite is much
less than that for a fiber element of length d, being only about 5 for
the typical case )8 = 7. On the other hand, standard tension tests on
fibers are performed at gauge-lengths L about two orders of
magnitude larger than 8, and by Eq. (3.87) their strengths are
about one-half of o6. Fortuitously then, the strength of the
composite will be little different from the strength of the fiber from
typical laboratory tension tests as is often observed. Finally, the
method of constructing Fig. 3.21 indicates that there is a mild size
effect in composite strength and a mild shift in the effective Weibull
shape parameter for the composite. Had a larger volume V = 109
been chosen rather than 106, the curved nature of the graphs on Fig.
3.20, from which Fig. 3.21 was derived, would produce a slightly
lower strength and a slightly higher effective shape parameter for
the composite depending on /3.
Attention is now turned toward a simple but approximate
theoretical explanation based on some key ideas motivated by the
above numerical analysis and results. First, the range for the
composite failure stress lies« a6, as we saw from Fig. 3.21. (Note
that both the median and the stress at 0.99 probability of failure lie
well below o6 for typical values of /3 below 15.) Second, the initial'
failures, that is fiber elements which fail directly under the applied
stress a, are viewed as 'seeds' for the growth of failure clusters,
which are lateral strings of adjacent fiber breaks contained within
bundles. Third, the number of such seeds is easily seen to follow the
binomial distribution with parameters MN and F(a) (the number
depends, of course, on o) with the mean number being MNF(o).
126 Strength of continuous-fiber composites

Fourth, cluster growth from a seed is viewed for calculation

purposes in terms of the sequential failure of adjacent fibers in a
bundle, with growth in either direction to form a string. Fifth,
instability occurs when a string of k breaks occurs such that
F(Kk_xo)<\, say, but F{Kko)~l\ thus, subsequent fiber failures
become almost certain leading to catastrophic growth of a trans-
verse 'crack' and failure of the composite. This value of k> which
depends on the stress level o> is called the critical crack size, and in
view of Eq. (3.88) is better defined as the k value for which
K^o^o^K.o (3.97)
Sixth, the following analysis is based on the Weibull shape
parameter p for fiber strength being 'large', but fortunately the
results work quite well for j8 down to about 4.
Proceeding with the analysis, it is first important to realize that
the initial breaks or 'seeds' are actually quite far apart. For
example, from Fig. 3.21 we recall that the median composite
strength was about 0.27o6 for /? = 5, and F(0.27o6) = 0.0014. This
means that the average spacing of seeds along a fiber is the inverse
of this value times 6, or about 7006, and laterally in a bundle is
about 700 fiber diameters. Moreover this spacing grows larger as the
composite volume increases due to the size effect. To see why, we
note that the size effect means that the median strength will
decrease as the volume increases. As an example, repeat the
process used to develop Fig. 3.21 from Fig. 3.20 but for a volume
MN = 109 instead of 106. One can see that the median strength will
now be only 0.22ad instead of 0.27a6 and since F(0.22ad) =
0.00052, the average spacing is almost 20006. Note that although
the seeds are now farther apart (fewer per unit volume), there are
more of them in the composite because the volume grew by a factor
of 103. (It may come as a surprise to the reader that a small
composite will show lots of single breaks per unit volume just
before failure, but a large composite will show relatively few!)
Thus, as a first approximation we can ignore the possible interac-
tions of two clusters growing near each other since the critical k will
turn out to be quite small.
The probability of a given fiber element becoming a seed and its
immediate neighbors developing further into a failure string of size
k is approximately
F{seed and string} « F(a)2F(Klo)2F(K2a) • • • 2F(Kk_ l o) (3.98)
where the factors '2' appear because, at each step of the growth
 Statistical tensile strength theories 127

beyond the seed, there are two choices for the next failure (one on
each side) which approximately doubles the probability for that
step. Thus, such a string can stretch out variously to the left, or to
the right, or be centered relative to the original break. Clearly Eq.
(3.98) ignores considerable detail about the events of cluster
growth, as discussed more fully in Phoenix and Smith (1983), but it
works mainly because F(Kjo)»F(Kj_xo) when /3 is large. (The
nature of the simplification can be appreciated upon studying Eqs.
(3.89) and (3.90) for small bundles where in each case the first term
will dominate all the others when /? is large.) Using a Taylor series
expansion in {o/o^ it can be seen that

F(o)~{olo6y (3.99)
This is especially true when a « o8, but it turns out that for present
purposes we can take this as a good approximation for 0 < a < o6,
particularly in Eq. (3.98). Substituting Eq. (3.99) in Eq. (3.98), we
have
F{seed and string} ~2k-\o/of>)fS(K1o/od)13 ••• {Kk.xoIorf
k p kp
= 2 ~\KxK2 • • • Kk^) (o/a6) (3.100)
This factorization and collapse of terms, to yield an exponent of kfi
instead of /?, is an important feature which follows from the use of
the Weibull distribution. It is the point at which the effect of
micromechanical 'redundancy' in the composite emerges as a
reduction in variability.
In the composite there are MN potential seed fibers, each of
which may produce a string, and the composite will fail if at least
one such event occurs. Treating the MN seed and string events as
statistically independent (which works because of the wide spacing
mentioned above), we actually have a weakest-link situation so that
the probability of composite failure is

HMA°) « 1 - [1 - P{seed and string}]M7V

« 1 _ [i _ 2k~\KxK2 • • • Kk_y(o/od)kTN (3.101)
b n b
From the calculus, (1 — ao ) —> exp{—naa } as n->™ so that

(3.102)
which is of the Weibull form, though k depends on the stress o
following Eq. (3.97).
128 Strength of continuous-fiber composites

Before discussing several important features of Eq. (3.102), it is

useful to develop a connection to the characteristic distribution
function W(a). For k = 1, 2, 3, . . . , let
fW((7) = 1 - exp{-dk(o/o6)kfi} (3.103)
where
dk = 2k~\KlK2---Kk^Y (3.104)
Equation (3.103) gives us a family of Weibull distributions with
increasing shape parameter kfi in k. Furthermore, following Eq.
(3.97) we can partition the important stress range 0<a<a6 into
the segments
Os/K^o^ojK^ £= l,2,3,... (3.105)
and for each k restrict the corresponding distribution to its
appropriate stress range. Then Eq. (3.102) becomes
]MN (3.106)

where k and o are chosen t o follow E q . (3.105). A n approximation

to W(a) then follows from a comparison of E q s . (3.96) and (3.106)
yielding

W(o) ~ F[k](o) (3.107)

where again k and a satisfy Eq. (3.105).
Figure 3.22 shows a plot of W(a) for p = 5 together with the
family of Weibull distributions ¥[k\a) for k =1,2,3, . . . , where
each is extended over the whole stress range 0 < a < o6. For each
stress level o one of these Weibull distributions comes very close to
W(o), and indeed it is normally the one whose k value satisfies Eq.
(3.105). Unfortunately, Eq. (3.107) has a jagged appearance when
plotted because of small 'jumps' occurring as k changes at the
transition stresses of the boundaries of Eq. (3.105). A graphically
pleasant 'repair' with a smooth appearance is to work with the inner
'envelope' of the family of Weibull distributions, that is

W(a) « min{F[11(a), F [ 2 \ o), F[3](a)> • • } (3.108)

Figure 3.22 indicates that this approximation works extremely well.
In principle we could develop similar graphs t o Fig. 3.22 for t h e
other cases /? = 3, 7, 10, . . . , 50 in Fig. 3.20. In developing Fig.
3.21 from Fig. 3.20 for a given volume V, it is quickly seen that o n e
of t h e Weibull cases, that is o n e value of k> would 'dominate' for
each value of j8, which is why each line in Fig. 3.21 is approximately
straight. F o r each plot, t h e appropriate k a n d Weibull shape
 Statistical tensile strength theories 129

parameter k/3 would be determined through Eq. (3.105) from the

relevant stress range in Fig. 3.21, especially near the median. For
example, for /? = 10, the case k = 3 is appropriate in developing Fig.
3.21, as the effective WeibuU shape parameter for composite strength
is about 3 x 10 = 30 (as determined from the slope of the p = 10 line
in Fig. 3.21).
It is now possible to determine the appropriate WeibuU distribu-
tion for each plot in Fig. 3.21. Substituting the appropriate WeibuU
distribution F[k\a) into Eq. (3.106) (which actually returns us to
Eq. (3.102)) yields the following WeibuU approximation for
composite strength:

« 1 - exp{ - (alakMN)kf>) (3.109)

where
ok,MN= (3.110)

For each value of jS, this WeibuU approximation closely fits the plot
on Fig. 3.21, provided k is chosen by the above graphical scheme.

Fig. 3.22. Envelope construction from WeibuU family ¥[k](o) to ap-

proximate the characteristic distribution function W(o) for composite
strength. Reprinted with permission from International Journal of Solids
and Structures, 19, Phoenix and Smith, Copyright © (1983), Pergamon
Press, pic.

0.90 * = 7/5/5/4/3/2/1
0.50
0.10
- F'"(o) 1 //^^^^
lO-2

lO-3

3. io-4

io-5
- Jw_W{o)
io-6
A//
io-7
'jfl p=5
io-8

io-9
/I// , . , . .
0.2 0.3 0.4 0.5 0.6 0.7 1.0
o/o6
130 Strength of continuous-fiber composites

Of course k will change if the volume V = MN is changed

significantly.
At this stage it is important to recall the interpretation of k as the
'critical crack size'. It is now appreciated that given the composite
volume MN and the Weibull shape parameter p for the fiber
strength, a special value of k emerges which is the size of the
longest crack or string of fiber breaks when such a composite
fractures. This value of k also determines the effective Weibull
shape parameter for composite strength, k(i. Thus far, the calcula-
tion of the appropriate k value has been performed graphically, but
it is possible to estimate k explicitly. The method is given in
Phoenix and Smith (1983), and begins by the study of
(3.111)
For large MN, this leads to the appropriate k being the value which
satisfies
y{k) > ]n(MN)/p > y(k - 1) (3.112)
where
y ( r ) = r \n(Kr) - { ] n ( K 0 + \n(K2) + ••• + l n ( K r _ 0 } (3.113)
for r = l, 2, 3, . . . and y(0) = 0. For Kj = l +j/2, we obtain the
values given in Table 3.1. According to Eq. (3.112) the critical
value of k depends on the ratio \n(MN)/ p, and thus it increases
slowly as the composite volume is increased but decreases more
rapidly as the variability in fiber strength is decreased (ft is
increased).
As an example, for the case /3 = 5 on Fig. 3.21, the graphical
procedure puts the stress range near 0.27ad which by Fig. 3.22 or
Eq. (3.105) puts k = 5. On the other hand, ln(106)/5 = 2.76, and by
Eqs. (3.112) and (3.113) and Table 3.1 one also obtains k = 5. Thus
the effective Weibull shape parameter for the composite being
represented is kf5 = 25.

Table 3.1.

r Y(r) r Y(r)
0 0 5 3.15
1 0.405 6 3.95
2 0.981 7 4.78
3 1.65 8 5.62
4 2.38 9 6.48
 Statistical tensile strength theories 131

Finally, it is interesting to consider the ultimate size effect for the

composite. In the case of a Weibull distribution, we recall that the
strength decreases as the volume V in proportion to V~1/(3. On the
other hand, the curvatures of the lines on Figs. 3.20 and 3.21,
together with our finding that k slowly increases as the volume
V = MN increases suggest that the strength of the composite will
not ultimately have a Weibull size effect, but one which is
increasingly milder as V increases. Smith (1980, 1982) considered
this question and concluded that
composite strength « p2l~ypodl\n.(V) (3.114)
which indicates that the strength decreases as the inverse of the log
of the volume. It turns out that Eq. (3.114) tends to be an
overestimate and a composite must be astronomically huge (V >
1020) for this result to be accurate.
In conclusion, a few extensions and limitations of the above
analysis should be mentioned. As stated earlier, the results given
are based on fi being 'large'. This allowed us to write the
approximation Eq. (3.98), which led us to Eq. (3.100) and then to
the definition of dk in Eq. (3.104). As mentioned earlier, the
calculation of the event implied in Eq. (3.98) is more complex if
'double counting' of certain failure possibilities is to be avoided. For
example, for k = 2, a more accurate rendition is
P{seed and string} - 2F(o)[F(K1o) - F(a)] + F(o)2
= 2F(o)F(Klo)-F(of
« PC/CO"-l](a/a a ) 2 " (3.115)
Y
so d2 should be [2{KX - 1] rather than just 2{Kxy. The same sort
of analysis shows that d3 should actually be 4(K1K2)P — (K})2^ ~
(K2)f3 - 2(Kl)l3 + l and so on for higher k. But it turns out that
these refinements make very little difference, especially when
calculating the scale parameter values okMN in Eq. (3.110) where
the error is typically one or two per cent.
The above results were developed for the idealized case of local
load sharing defined by Eq. (3.88), but appear also to work for
more realistic cases provided one chooses Kr to be the largest load
sharing constant at the edge of a failure configuration. Generally
such values of Kr tend to be smaller than 1 + r/2 (see, for example,
Hedgepeth 1961). Following through the above analysis, the main
effects are not only to increase the scale parameters for strength,
thus increasing the composite strength itself, but also to increase the
132 Strength of continuous-fiber composites

critical k values thus reducing the composite variability. Second, an

analysis has been carried out by Smith et al. (1983), for three-
dimensional composites, with the parallel fibers forming a two-
dimensional hexagonal array. Here the clusters of broken fibers can
take on many different geometric configurations other than a linear
string, but for large p one still comes up with a form for dk that is
similar in structure to Eq. (3.104) except that 2k~l is replaced by a
much more complex configurational constant. Many of the ideas
carry through except that one no longer finds quite the same simple
relationship between the critical cluster size k and the effective
Weibull shape parameter for composite strength. The strength of
such a three-dimensional composite is typically larger than in the
two-dimensional planar case described above. The reason is that
while there are many more failure configurations, the load sharing
occurs over many more fibers at the boundary of a failure cluster so
that the reduction in the Kr values more than compensates for the
increased number of failure possibilities, especially for larger j3.
Finally, experimental data to illustrate the above features have
been presented by Phoenix, Schwartz and Robinson (1988), who
also extend the ideas, through viscoelasticity of the matrix, to
explain creep rupture phenomena under constant stress.

3.4.6.3 Monte-Carlo simulation

The Monte-Carlo method is a numerical technique suitable
for simulating complicated stochastic processes, and it has been
employed to analyze a wide range of physical processes of a
statistical nature (Oh 1979). The Monte-Carlo simulation of com-
posite strength can be regarded as testing the composite materials
'analytically' in an automated fashion. In each Monte-Carlo experi-
ment, random numbers are generated and assigned to the underly-
ing random variables and the outcome of the process of interest can
be observed. When the number of such independent experiments
is sufficiently high, the observations will yield a good assessment of
the statistical characteristics of the process. In dealing with the
strength of fibers as well as composites, the Monte-Carlo experi-
ment involves the partitioning of fiber or a composite into elements,
then random numbers are assigned to the strength of the elements.
For a given applied load, the stress in the elements of a fiber or a
composite can be determined as described in Section 3.3. From the
assigned strength value and the arrangement of breaks of elements
the failure load is then obtained. In the following, fractures of fibers
as well as composites based upon the Monte-Carlo simulation
 Statistical tensile strength theories 133

(Fukuda and Kawata 1977; Oh 1979; Manders, Bader and Chou

1982) are considered. Several common procedures for generating
the normal random numbers are available.
Fukuda and Kawata studied the fracture of a two-dimensional
fiber composite based upon the Monte-Carlo method by choosing a
mean strength of 100 and a standard deviation of 10. A simulation
of the fracture process is shown in Fig. 3.23 for Ef/Em = 20, and
M = N = 20, where M and N are defined in Fig. 3.17. The elements
or links in the partitioned composite specimen are specified by the
position (/, y). Here, 0 indicates that the link is not broken and the
other numerals indicate the sequence of link breakage. As the
initial condition, each link (/, y) is assumed to have a stochastic
strength, STR(/, y), which is the normal random number with a
specific value of mean and standard deviation. Both the Weibull
distribution and normal distribution have been used for expressing
the link strength distributions. Stress concentration factors of all
links, SCF(/,y), are initially assigned as 1. A link with the least
value of STR(i,y)/(SCF(/,y) is sought, and let this link be (io,yo).
The link breaks first at the tensile stress of STR(/O,yo). When this
link breaks, stress concentration occurs in the two adjacent links
( i o , ; o ± l ) . The values of STR(i, y)/SCF(i, y) are again calculated
for the remaining M x N -1 links. A link which has the least of this
value breaks second. This procedure is repeated until all the links in
a plane transverse to the loading direction (y = 1, 2, . . . , M) are
broken.

Fig. 3.23. Monte-Carlo simulation of fiber link fractures. (After Fukuda

and Kawata 1977.)
0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 33 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 39 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 38 0 0
0 0 0 40 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 7 0 0 0 0 0 0 0 0 0 0 0 43 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 41 0 0 0 0 0 0
18 44 45 3 1 4 6 47 0 0 0 0 011 0 0 0 0 0 0 0 0
42 0 0 0 0 0 0 0 10 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 30 0 0 0 0 0
0 0 0 0 0 0 5 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 9 0 0 0 0 0 3 0 0 0 0 0 0 0 0
62 59 58 57 56 55 54 53 52 51 50 49 48 60 61 37 16 36 35 4
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 19 0 0 0 0 0 0 20 0 0 0 0 32 12 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 021 0 0 1 3 0 0 0
0 0 0 0 0 0 0 0 34 29 28 14 27 26 25 6 15 24 22 8
0 17 0 0 0 0 0 0 0 2 3 0 0 0 0 0 0 0 0 0 0
134 Strength of continuous-fiber composites

The result given in Fig. 3.17 resembles the sequence of fiber

failure observed in the experimental work of Rosen (1964). The
predictions of composite strength are shown in Fig. 3.24. It should
be noted that the Monte-Carlo approaches are generally limited to
MN < 50 000 under current supercomputer power which may not be
enough for a realistic composite. Also the Monte-Carlo approach is
inherently poor at handling the lower tails of the distributions.

3.4.7 Strength of cross-ply composites

Cross-ply construction is the simplest form of lamination of
unidirectional laminae. This simple geometric configuration facilit-
ates the understanding of the fundamental problems concerning
laminate strength. It provides a model system for investigating the
matrix cracking of laminates under tensile loading. This section
analyzes the problem from both deterministic and statistical view-
points. The treatment of Aveston, Cooper and Kelly (1971) of
multiple fracture, although it deals with unidirectional composites, is
basic to matrix cracking of laminated composites in general. Hence,
it is outlined first.

3.4.7.1 Energy absorption during multiple fracture

Section 3.2 discusses the mode of fracture of unidirectional
composites as affected by the ultimate failure strains of the fiber and

Fig. 3.24. Numerical results of Monte-Carlo simulation. (After Fukuda

and Kawata 1976b.)

1.0
i i i i i I I i i i •» I

Crack initiation

Fracture

0.5

0.0 i I U m i
0.6 0.7 0.8 0.9
Stress/mean strength of links
 Statistical tensile strength theories 135

matrix materials as well as the fiber volume fraction. The energy

absorption of composites during the failure process was first
investigated by Aveston, Cooper and Kelly (1971). Contributions
to the fracture surface energy during single fracture may be
derived from deformation of the fiber or matrix, the work done
in fracturing the fiber-matrix interfacial bond, and work
done in pulling the fibers out of the matrix against frictional
forces. It is found that the work of fracture increases with
increasing fiber diameter and decreasing fiber-matrix interfacial
strength.
Multiple fracture of fibers occurs in ductile matrix composites at
low fiber volume fraction. Multiple fracture of matrix, on the other
hand, takes place in brittle matrix composites at high fiber volume
fraction, as a result of applied tensile loads or thermal stresses
induced by cooling from the stress-free temperature. The energy
consideration for the development of multiple matrix cracking in a
unidirectional lamina subject to axial tensile loading is introduced
below (see Aveston, Copper and Kelly 1971; Aveston and Kelly
1973, 1980; Kelly 1976).
Consider the formation of a single matrix crack normal to the
fiber direction, at the strain emu under conditions of fixed load. It is
assumed that the stress in the matrix is equal to the matrix fracture
stress and there is a decrease in the combined energy of the
specimen and the loading system. The energy changes due to the
formation of a crack at a fixed load include AW = the work done by
the applied load per unit area of the composite, ydh = energy
absorbed per unit area of debonded fiber, Us = the work done per
unit area of the composite against the frictional force between the
fiber and matrix, A£/m = the elastic strain energy lost due to the
relaxation of the strain in the matrix, and A£/ f =the increase in
strain-energy of the fibers per unit area of the composite. If the
surface energy in forming a matrix crack is ym, a crack will occur
provided

2y m (l - Vf) + y db + £/s + AUt < AW + AUm (3.116)

The terms in Eq. (3.116) have been evaluated by Aveston, Cooper

and Kelly under the assumption that the changes in stress (strain) in
the matrix and fiber due to the formation of the crack vary linearly
with distance from the crack surface. By further assuming purely
frictional bond between the fiber and matrix, Eq. (3.116) yields the
136 Strength of continuous-fiber composites

following expression for the failure strain of the matrix:

where R = fiber-matrix interfacial shear strength (See Eq. (3.2)),

r = fiber radius, and E = Young's modulus with the subscripts f, m
and c indicating fiber, matrix and composite, respectively. Equation
(3.117) indicates that the composite strain at the formation of the
transverse matrix crack can be enhanced by suitable control of the
elastic moduli of the fiber and matrix, fiber volume fraction and
diameter, matrix surface energy, and the fiber-matrix interfacial
strength.
Budiansky, Hutchinson and Evans (1986) have generalized the
results of Aveston, Cooper and Kelly for unbonded, frictionally
constrained slipping fibers initially held in the matrix by thermal or
other strain mismatches. The other case considered by Budiansky et
al. for the onset of matrix cracking involves fibers that initially are
weakly bonded to the matrix, but may be debonded by the stresses
near the tip of an advancing matrix crack. McCartney (1987) has
used an energy-balance calculation for a continuum model of brittle
matrix cracking in a uniaxially fiber-reinforced composite and
confirmed that the Griffith fracture criterion is valid for matrix
cracking.

3.4.7.2 Transverse cracking of cross-ply laminates

Multiple transverse cracks in the matrix of unidirectional
fiber composites have been observed in a number of systems, for
example, glass-reinforced cement, and gypsum reinforced with
polyvinyl chloride or glass, where the failure strains of the fibers are
greater than those of the matrices. Transverse cracking also occurs
in the 90° plies of cross-ply laminates. Experimental observations
and analytical modeling of this behavior have been made by Bailey,
Garrett, Parvizi, Bader and Curtis (see Garrett and Bailey
1977a&b; Parvizi and Bailey 1978; Parvizi, Garrett and Bailey 1978;
Bader, Bailey, Curtis and Parvizi 1979; Bailey, Curtis and Parvizi
1979; Parvizi 1979; Bailey and Parvizi 1981 who followed
Aveston and Kelly's shear-lag approach and interpreted this pheno-
menon by the concept of constrained cracking). Manders, Chou,
Jones and Rock (1983) proposed a statistical treatment of multiple
cracks. Wang, Crossman, Warren and Law (see Wang and Crossman
1980; Crossman, Warren, Wang and Law 1980; Crossman and
Wang 1982; Wang 1984), on the other hand, theorized it based
 Statistical tensile strength theories 137

based upon the strain-energy release rate of crack extension. The

theory of Bailey et al. is introduced in this section. The work of
Manders et al. is discussed in Section 3.4.7.3 and that of Wang et al.
is introduced in Section 3.4.7.4.

(A) Cross-ply laminate

The cross-ply construction of [079070°] is shown in Fig.
3.25. For the cases of glass/epoxy and carbon/epoxy systems, the
mechanical properties of unidirectional laminates are shown in
Table 3.2. The glass/epoxy 0° test curves are essentially linearly
elastic to fracture but the 90° specimens show a pronounced knee at
a strain of about 0.3%, after which a whitening effect can be
observed. The 0° carbon/epoxy test curves are elastic to failure but
they are not linear, there being an increase in the modulus with
increasing strain. The 90° carbon/epoxy is linear to failure with no
knee or acoustic emission prior to failure. The failure strains of the
90° specimens in both systems are characteristically low due to
strain concentrations in the matrix (see Kies 1962).

Fig. 3.25. Illustration of a [0°/9070°] specimen.

y
138 Strength of continuous-fiber composites

When extended in tension, initial failure of the cross-ply laminate

is usually in the central 90° ply, which cracks in a direction normal
to the applied tension and parallel to the fibers in that layer (Fig.
3.26). The failure sequence in both laminates follows a similar
pattern. Two knees appear on the stress-strain curve of
glass/epoxy, first at 0.3% strain, associated with the visual whiten-
ing effect and at 0.5% strain due to transverse cracking, but this is
not apparent in the carbon/epoxy laminate. On further extension,
more cracks are formed until the whole gauge portion of the
test-piece is filled with a regular array of cracks. The strain at which
the first crack occurs increases as the thickness (2h) of the 90° layer
is reduced and at the same time the crack spacing tends to become
smaller. In the case of the thinnest transverse layers, transverse
cracking is not observed at all before the final catastrophic failure of
the test-piece. Microscopy has shown that the earliest indications of
failure are debonds at or near the fiber/matrix interface. These
occur at strains even lower than those at which the whitening is
observed in the glass/epoxy systems. The next stage is a coalescence
of a number of debonds to form a microcrack, which grows rapidly
when it reaches a critical size, about three to four fiber diameters.
Longitudinal splitting is observed to occur in the 0° plies of the
cross-ply laminate at strains intermediate between the transverse

Table 3.2. Mechanical properties of unidirectional laminates (after

Bader et al. 1979), Reprinted with permission from Mechanical
Behaviour of Materials-Copyright © 1979, Pergamon Press, pic).

0° 0° 90° 90°
Property CFRP* GRP** CFRP GRP Units

Low-strain
Young's
modulus 127 42 8.3 14 GPA
Fracture
stress 1.7 0.92 0.039 0.056 GPa
Fracture
strain 1.2 2.2 0.48 0.50 %
Poisson's
ratio 0.29 0.27 0.02 0.09 -

* CFRP: carbon fiber-reinforced plastic

**GRP: glass fiber-reinforced plastic
 Statistical tensile strength theories 139

cracking strain for the 90° plies and final failure (Fig. 3.27).
Longitudinal splitting is due to mismatches in the Poisson's ratios
and the coefficients of thermal expansion of the 0° and 90° plies.
The strain to initiate splitting increases as the thickness of the
longitudinal plies is reduced. Splitting has not been observed in the
carbon/epoxy cross-ply laminates.

(B) Transverse crack spacing

The low strain failure behavior was first explained by Kies
(1962), who predicts the magnification of strain in the matrix
when a unidirectional composite is stressed in the transverse
direction. In the limit when the fibers are almost touching one
another, the strain magnification factor approaches the value
Ef/Em. It should be noted that even at comparatively low fiber
volume fractions there are invariably regions in the lamina where
fibers almost touch one another. The glass fibers are nearly

Fig. 3.26. Transverse-ply crack in a [0°/90°/0°] carbon fiber-reinforced

cross-ply laminate with an inner-ply thickness of 2h — 0.125 mm. (After
Bailey, Curtis and Parvizi 1979.)
140 Strength of continuous-fiber composites

isotropic, but the transverse Young's modulus of carbon is much

lower than its longitudinal modulus and it is this modulus which
should be used for calculating the strain magnification factor. The
first matrix crack usually forms between fibers which are touching or
nearly touching along a direction perpendicular to the loading axis.
The crack density, and hence the crack spacing, is related to the
geometry of the laminate. These can be explained by the cross-ply
laminate shown in Fig. 3.25. When the strain has reached the
fracture strain, £tu, of the 90° ply, the first crack occurs in the
transverse ply, and an additional stress AA is placed on the
longitudinal plies. From a shear-lag analysis similar to that given in
Section 3.3.1,
ACT = Aa o exp( - (3.118)
where
± ECG12 (b + h
EUE22 \ bh

Fig. 3.27. Longitudinal-ply splitting in a [0°/90°/0°] glass fiber-reinforced

cross-ply specimen. (After Bailey, Curtis and Parvizi 1979.)
 Statistical tensile strength theories 141

Ec is the laminate Young's modulus in the y direction, En and £22

are the Young's moduli of a unidirectional ply in the fiber and
transverse directions, respectively, and G12 is the shear modulus of
a unidirectional ply. This additional stress has its maximum value
AA O in the plane of the crack (y = 0) and decays with distance y
from the crack plane as some load is transferred back into the
transverse ply through interlaminar shear stress
dAcr
r^-b — (3.119)

The tensile load in the transverse ply is zero at the crack plane
but builds up by shear transfer from the longitudinal plies. At a
given distance y from the crack, the load F in the inner ply is given
by

F= [y2cTYdy (3.120a)

where c is defined in Fig. 3.25. The first crack in the transverse ply
occurs when the load carried by it is equal to 2chatu where atu
denotes the ultimate tensile strength of the 90° ply in the cross-ply
laminate, which may be different from the transverse tensile
strength of a unidirectional ply. This load is then transferred onto
the longitudinal plies. Another crack can only occur when the
transverse ply is again loaded to 2chatu. The transverse ply will not
be loaded to this value except at infinity and A a o = otuh/b, if the
applied stress on the laminate is maintained at cra = Ecetu after the
first cracking. For another crack to occur, a a and hence Aao must
be increased to such a value that F = 2chotu.
If the first crack is assumed to take place in the middle of the
specimen (y = 0) of length a, the following cracking sequence will
occur:
(1) Initial crack at aa = Ecein, and
F = 2bc AAO[L - exp( - V(0)}>)] (3.120b)
(2) Second and third cracks occur simultaneously at the ends of
the specimen when the applied load increases to such a
value that

h
Aao = atu £ [1 — exp(— V(0)«/2)]~1 (3.121)
b
The crack spacing is a12.
142 Strength of continuous-fiber composites

(3) The next series of cracks will occur midway between the
present cracks. The total shear stress between two existing
cracks is

-exp[V(0)O'-fl/2)]} (3.122)
and from Eq. (3.120a)
F = 2bc Aa o [l + exp( - V(0)«/2)
- 2 exp( - V(0W4)] (3.123)
The value of ACRO when the cracks occur now at intervals of
a/4 is

= atu - [1 + exp( -
1
(3.124)
(4) For crack spacing of a/S

Aa o = atu - [1 + exp( -
o
(3.125)
This crack sequence will continue until the strength of the lon-
gitudinal plies is exceeded or the spacing between neighboring
cracks is so small that the normal stress in the 90° ply cannot be
built up to atu.

(C) Transverse cracking constraint

The strain required to initiate transverse cracks is greater
when the transverse lamina is thinner, and in some cases cracking is
constrained completely up to the strain at which the longitudinal
laminae fail catastrophically. This phenomenon of constrained
cracking is attributed to the fact that in order for a crack to form it
must be both mechanistically possible and energetically favorable.
The former requirement is satisfied for cross-ply laminates from the
viewpoint of strain magnification as discussed in (B). The effect of
lamina thickness on the transverse failure strain can be understood
from the viewpoint of energetics.
For a specimen under constant load, a crack initiates if the
following condition is satisfied:
AW>AU+t/D + 2yA (3.126)
 Statistical tensile strength theories 143

where AW is the work done by the applied stress per unit area of the
specimen, AU is the increase in stored energy per unit area of the
specimen, £/D is the energy loss per unit area due to any dissipative
processes present (e.g. sliding friction between debonded fiber and
matrix), y is the fracture surface energy per unit fracture surface
area, and A denotes the fracture surface area. It has been found
that for practical ply thicknesses the interface between the lon-
gitudinal and transverse plies remains bonded during the cracking of
the transverse ply and the laminate behaves in a fully elastic
manner, thus Eq. (3.126) becomes

AW>AU +2 -y, (3.127)

h+b
Here, yt is the fracture surface energy of the transverse ply in a
direction parallel to the fibers. Since half of the work done by the
applied stress is stored as elastic energy of the specimen, it follows
that

^ - b (3.128)

When the first crack occurs in the transverse ply at a strain of etu
an additional stress ACT, Eq. (3.118), is thrown onto the outer plies
and the laminate increases in length by da, given by
rfl/2ACT
da = 2 — dy (3.129)
Jo En
For a/h»l, Eq. (3.129) becomes
2hE22etu
5a= (3 130)
-——7—— .
The work done by the applied stress cra at the strain of first
transverse failure is
= daaa (3.131)
Hence

The substitution of Eq. (3.132) into Eq. (3.128) yields the minimum
value of the transverse failure strain
144 Strength of continuous-fiber composites

The theoretical values of the minimum cracking strain have been

calculated from Eq. (3.133) as a function of h and are compared
with the experimental results in Fig. 3.28 for glass/epoxy laminates.
Close agreement is observed between theory and experiment in the
region where h< 0.25 mm, indicating an energy controlled crack
propagation. For the thicker laminates, however, this theory does
not apply and cracking occurs at a constant strain of 0.5% which is
close to the cracking strain of the unidirectional 90° lamina.
According to Bader et al. (1979), microscopic cracks usually
develop in glass- and carbon-reinforced plastic laminates in regions
where fibers lie normal to the principal tension axis, at strains which
are, at the most, only 30% of the final failure strain. Thus designers
are faced with a dilemma: whether to base the design on strains
below the cracking threshold (typically 0.5% for glass-reinforced
plastics) or the ultimate failure strain, which might be 1.5% or
more. Microcracks which do not appear to be detrimental to the
short-term mechanical properties of laminates may act as nuclei for
further local damage leading to ultimate failure under cyclic loading
and a hostile environment. Experimental evidence suggests that the
formation of transverse cracks and longitudinal splitting can be con-
strained or inhibited by constructing the laminate from thinner
individual plies.

Fig. 3.28. Plot of the theoretical and experimental transverse cracking

strain, (ec)min, as a function of the inner-ply thickness, 2h, for glass-
reinforced sandwich laminates. The outer ply thickness is 0.5 mm. — Eq.
(3.133); — cracking strain of the unidirectional 90° lamina; $ experiment.
Reprinted with permission from Bader et al. in Mechanical Behaviour of
Materials, Copyright © (1979), Pergamon Press pic.

1.0 2.0 3.0 4.0 5.0

1h (mm)
 Statistical tensile strength theories 145

3.4.7.3 Statistical analysis

The deterministic multiple cracking theory of Garrett,
Bailey and Parvizi attempts to account for the measured distribution
of crack spacing in [079070°] glass fiber/resin matrix laminates.
Manders et al. (1983) have proposed a statistical model which fits
the experimental data and predicts a dependence of strength on
size. The origins and implications of this variability of strength are
discussed below after descriptions of the experimental observations.
The three-ply [079070°] laminates of Manders et aL are composed
of Silenka E-glass fibers in an Epikote epoxy resin. The central 90° ply
is 1.1mm thick and is sandwiched between two 0.55 mm plies. A
close match between the refractive indices of the fiber and matrix
makes the laminate virtually transparent so that cracking and
microscopic damage in the 90° can be closely observed (Fig. 3.29).

Fig. 3.29. Photographs of specimens at the indicated strain levels (%)

under bright-field ((a) to (i)) and dark-field ((j) to (r)) illumination,
showing multiple transverse cracks in the 90° ply, stress 'whitening' and
longitudinal splitting in the 0° plies. (After Manders et al. 1983.)

(a)0,56 (b)0.72 (c) 0.90 (d) I 10 (e) 1.30 (f) 1.70 {g)1.90 (h)2.40 (i) 2.80
20 mm

00.0 (k) 0.34 (1)0.36 (m)0.54 (n) 0.72 (o) 0.90 (p) 1.08 (q) 1.50 (r)1.90
I 20 mm |
146 Strength of continuous-fiber composites

The pattern of cracks is photographed at regular intervals of applied

load using either bright- or dark-field illumination. The dark-field
illumination showsfiber—matrixdebonding ('stress whitening' which
scatters light) with good contrast, whereas bright-field illumination
gives better definition of the cracks, although in this case the
fiber-matrix debonding appears dark with relatively poor contrast.
The thermal residual tensile strain of the 90° ply is estimated to be
about 0.22% due to cooling from the postcure temperature of 150°
to ambient.
As the specimens are loaded the initial whitening progressively
increases, most noticeably at about 0.34% strain (Fig. 3.29k). A
knee is visible in the stress-strain curve of Fig. 3.30 at about 0.1%
which is attributed to the onset of fiber-matrix debonding. Cracks
appear instantaneously at about 0.4% strain, often in the bands
of more pronounced whitening (Fig. 3.291 and m). It is concluded
from this observation that a crack forms by the joining up of the
fiber-matrix debonds. The beginning of multiple cracking is re-
flected on the stress-strain curve by a second knee. The rate of
crack formation with applied strain decreases throughout the
loading. At higher strains the crack spacing becomes more uniform.
At a strain of about 0.7% stress whitening appears in the lon-
gitudinal 0° ply (Fig. 3.29n-r); this is seen as darkening in Figs.

Fig. 3.30. Low-strain portion of a stress-strain curve. Changes of gradient

are associated with a rapid increase in stress whitening and with the
beginning of multiple cracking. (After Manders et al. 1983.)
1ZJ

100 Ultimate failure

at 2.4%
xess MPa)

75 —

^ ^
50 Stress whitening ^
|-^- Multiple cracking
\
25

0 S^ | 1 1 1 1 1 1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.1
Strain (%)
 Statistical tensile strength theories 147

3.29(b)-(i), and it develops into longitudinal cracks at about 1.8%

strain.
Manders et al. have measured the positions of every crack in a
photograph by traveling microscope and calculated the spacings
between cracks and their cumulative distribution functions for each
load. These distributions illustrate the overall trend towards closer
spacing at higher strains. In their study of the variation of crack
spacing with stress, Manders et al. assume that the 90° ply is an ideal
homogeneous brittle material with an inherent distribution of
strength which is described by a cumulative distribution function
termed So for failure of a unit volume. It is also expected that the
strength of the 90° ply will be statistically the same throughout its
volume; i.e. the constituent volumes which are substantially larger
than the microstructure should have strengths which are independ-
ent of each other and which are identically distributed.
Thus, the cumulative distribution function of strength Sv for a
volume V can be written as
1-SV = (1-SO)V (3.134)
Then the 'risk of rupture', Rv, proposed by Weibull (1939a and b) is
given by
- SV) = V ln(l - So) = -Rv (3.135)
Let
So) = - 0 ( a ) (3.136)
then the risk of rupture dR for a volume element dV is
dR = - l n ( l - So) dV = 0(a) dV (3.137)
For a non-uniform state of stress

Rv= I 0(a)<\dV (3.138)

Jv
and
Sv = l- exp(-Ry) = 1 - expf - J 0(a) dv] (3.139)

Assuming that the stress is uniform in the cross-sectional area, A,

the volume integral may be replaced by an integration over the
length L. Then Eq. (3.135) becomes
ln(l-Sv) = -A(f)(o)L (3.140)
The quantity A(j) is found from the gradient when ln(l - Sv) is
plotted against L.
148 Strength of continuous-fiber composites

Manders et al. adopted a two-parameter Weibull distribution for

the strength of the 90° ply in which

A<p=A (3.141)
£o

The constants oo and eo are the scale parameters in terms of stress

and strain, respectively, and /? is the shape parameter. Taking
logarithms of Eq. (3.141) gives
= p In e - p In eo + In A (3.142)
It is seen from Eq. (3.142) that a graph of the gradients obtained
from ln(l - Sv) vs. L and applied strain is linear with gradient P if
the Weibull distribution is valid. This is demonstrated in Fig. 3.31,
which shows two linear regions intersecting at a strain of about
0.4% (corrected for thermal residual strain), or 0.6% of applied
strain. The values of P are about 8.5 and 1.0, respectively, for low

Fig. 3.31. Variation of gradients with 90° ply strain, corrected for residual
thermal strain. Solid and open circles correspond to two nominally
identical specimens. (After Manders et al. 1983.)

2
ao

5.0 -

4.0 -

-6.0 -5.0 -4.0 -3.0

In (strain)
 Statistical tensile strength theories 149

strain and high strain. The two intercepts (In A —f}In so) for the
two linear segments are 47 and 11.
Finally, Eq. (3.140) can be evaluated after substitution of Eq.
(3.142) using the fitted values of /3 = 8.5 and intercept = 47 to
obtain median crack spacings (Sv =0.5) as a function of strain. The
results of the theoretical correlations are shown by the solid curve in
Fig. 3.32.
It is suggested by Manders et al. that the deterministic model of
Garrett, Bailey and Parvizi and the probabilistic models are
complementary. At low strains, the crack spacing is large and the
length necessary to build up stress in the 90° ply on either side of a
crack is relatively small. Therefore, most of the region between
cracks is fairly uniformly stressed and the positions of new cracks
are determined by the distribution of flaws in the matrix; a new
crack rarely forms exactly midway between two existing cracks.
Consequently, the distribution of crack spacings covers a wider
range than the factor of two predicted by Garrett, Bailey and
Parvizi. At high strains the opposite is true. The region between
cracks is non-uniformly stressed. Since the highest stress is found
midway between two existing cracks, this is where the new crack
forms as described by the deterministic model. When the crack

Fig. 3.32. Crack spacing vs. strain. Solid curve is based upon the statistical
model predictions. (After Manders et al. 1983.)

S.

0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2
Strain (%)
150 Strength of continuous-fiber composites

spacing is significantly higher than the 'unstressed length' (approxim-

ately equal to the 90° ply thickness) the probabilistic model is
appropriate, and when it is of similar magnitude the deterministic
model is more appropriate.
Further analytical treatments of the statistical strength of cross-
ply laminates can be found in the work of Fukunaga, Peters,
Schulte and Chou (1984) and Peters and Chou (1987).

3.4.7.4 Transverse cracking and Monte-Carlo simulation

The occurrence of transverse cracks in cross-plied laminates
under ascending tension can be regarded as a kind of stochastic
process due to the presence of randomly distributed microflaws. As
discussed in Section 3.4.6.3, a stochastic process can be simulated
by the Monte-Carlo procedure. In this case, it is postulated that
'intralaminar flaws' exist randomly in the unidirectional ply, which
lie in the ply thickness direction and align with the fibers, Fig.
3.33(a). When the transverse ply in the cross-plied laminate is
subjected to tension, these flaws effect the observed transverse
cracking. For purpose of simulation, the identity of the intralaminar
flaws is represented by randomly generated 'effective flaws'. The
effective flaws are not, of course, the real flaws. However, if chosen
properly, they represent an inherent property of the ply system and
effect the essential characteristics of the transverse cracking process
in the simulation model.
Wang and Crossman (1980) first conducted an energy analysis to
predict the onset of a single transverse crack based on the classical
fracture mechanics concept, in conjunction with the effective flaw
postulation. Their analysis was validated by a series of experiments
(see Crossman, Warren, Wang and Law 1980; Crossman and Wang
1982). Later, Wang, Chou and Lei (1984) and Wang (1984, 1987)
incorporated the energy method into a Monte-Carlo procedure to
simulate the stochastic nature of multiple cracking. In this section,
the work of Wang et al. is discussed in some detail.

(A) Ply-elasticity and three-dimensional stress states

At the outset, it is useful to describe briefly the basis of the
energy method. The method is simply derived within the confines of
ply elasticity and the classical theory of fracture mechanics. The
theory of ply elasticity regards each unidirectional ply as a three-
dimensional, elastic, homogeneous and anisotropic solid; and the
laminate is modeled as a three-dimensional layered medium con-
taining flaws. An individual effective flaw is handled as a small
 Statistical tensile strength theories 151

crack; hence the elastic stress field surrounding the flaw is almost
always three-dimensional. Under certain simplifying assumptions,
however, some three-dimensional fields may be reduced to general-
ized plane-strain fields. Even then, numerical techniques are usually
required for solutions (see Pipes and Pagano 1970; Wang and
Crossman 1977).

(B) Effective flaw distribution

The exact mechanism of transverse cracking is rather
complicated when viewed at the fiber-matrix scale. It is usually
postulated that the crack is caused initially by the coalescence of
material microflaws which lie aligned with the fibers in the
transverse ply. When viewed at the ply scale, however, a transverse

Fig. 3.33. Schematic view of (a) effective intralaminar flaws, and (b)
effective interlaminar flaws. (After Wang 1987.)

Intralaminar
flaws

Interlaminar
flaws

Interlaminar
edge flaw

(b)
152 Strength of continuous-fiber composites

crack represents a separation of the transverse ply along the

fiber-matrix interface (see Fig. 3.29). To facilitate a mathematical
description of the event at the ply level, the concept of effective
flaws is now introduced. Assume that in each unidirectional ply
there exists a characteristic probability density distribution of
effective flaw sizes as shown in Fig. 3.34. The linear size of the flaws
is denoted by 2a and the location by x. Then, the discrete random
variables {ai9 i = 1, 2, . . . , M} and {xh i = 1,2, . . . , M} char-
acterize the size and the location distributions of the flaws.
When two or more plies are grouped together, such as in the
[0°/90°n/0°] laminate (with n > 1), the flaw size distribution in the
grouped 90° plies is represented by the volumetric rule (see Lei
1986):
\2/A
Hn = «,(") (3.143)
where i = 1, 2, . . . , M and A is a constant related to the distribu-
tional characteristics of {«,-}.
For simplicity, the flaw location distribution in the grouped 90°
plies is assumed to be independent of n
Xi,n = i = 1, 2, . . . , M (3.144)

Fig. 3.34. (a) The size {2at) and location (xt) of an intralaminar flaw, (b)
The probability density distribution of effective intralaminar flaw size in
transverse plies. (After Wang 1987.)

(a)

(b) f(a)
 Statistical tensile strength theories 153

(C) Onset of the first transverse crack

The [0790°n/0°] laminate shown in Fig. 3.34a is now used to
illustrate the energy method. Consider that the laminate is under
both the applied tensile strain exx and the temperature change AT
(A T is positive for a temperature drop). Let the distribution of the
flaws be characterized by Eqs. (3.143) and (3.144), Fig. 3.34b. With
the size and the location of a particular flaw known, an elastic stress
analysis can be performed; and by treating the flaw as a small crack,
one can also calculate the crack-tip strain-energy release rate
G(ainf EXX, AT) (see Wang 1987). The condition governing the
propagation of the small crack into a full transverse crack is then
given by

G(ai>n,exx,AT) = Glc (3.145)

where Glc is the material fracture toughness for mode I matrix crack
propagation.
Now, for the first crack to form, it is assumed that the crack is
caused by the largest of {ain}, denoted by «max. The critical
laminate strain (exx)CT for the onset of the first crack is then
determined from Eq. (3.145) by setting a in = amax. Now, this first
crack is physically detectable.

(D) Shear-lag effect

When the first transverse crack is formed, the local tensile
stress oxx formerly existing in the unbroken 90° plies is now zero. If
the 0°/90° interface bonding is strong, a localized interlaminar shear
stress rxz is then developed in the vicinity of the transverse crack, as
shown in Fig. 3.35. This interlaminar shear stress decays exponen-
tially a small distance away from the transverse crack; while within
the same distance, the tensile stress oxx in the 90° plies regains its
original magnitude. This local stress-transfer zone, or the shear-lag
zone, is proportional to the thickness of the grouped 90° plies, 2nt.
When there is an effective flaw located near a transverse crack,
Fig. 3.36, the flaw may be under the shear-lag zone of the
transverse crack. The degree of the shielding effect depends on the
relative spacing, s/nt. Specifically, if the size of this flaw is 2a and
the associated strain-energy release rate at the flaw tip is
G(ay sxx, AT, s)y then the shear-lag effect on the strain-energy
release rate can be expressed by the factor, R(s), defined by

R(s) = G(a, exxy AT, s)/G(a, exxy AT) (3.146)

154 Strength of continuous-fiber composites

where G(a, exx, AT) is calculated without the influence of shear-

lag. It may be noted that the range of the retention factor R(s) is
between zero and unity over the range of the shear-lag zone, as
shown in Fig. 3.36, for a carbon/epoxy composite.
When a flaw is situated between two consecutive transverse
cracks, then it is under the shear-lag effect from both cracks. The
associated strain-energy release rate, G*, is given by
G*(a, ex exx, AT)R(sK) (3.147)
where sL and sK are the distances from the flaw to the left crack and
to the right crack, respectively.

(E) Multiple cracks as a function of loading

After the formation of the first crack from the largest flaw
in {ai>n}, subsequent cracks can form from the remaining flaws at
laminate strains appropriately higher than (SXX)CR. A search is then

Fig. 3.35. (a) A transverse crack in a cross-ply laminate, (b) Local stress
transfer caused by transverse cracking and the shear-lag zone. (After
Wang 1987.)

(a)

(b)

•>• x/nt
 Statistical tensile strength theories 155

commenced to determine the next flaw that yields the highest

strain-energy release rate G* (with due regard to the shear-lag
effect cast by the existing cracks). The applied laminate strain
corresponding to the next crack, which should be higher than
(£**)CR> is determined by using G* in Eq. (3.145).
Successive searches for the next most energetic flaw follow, and
the entire load sequence of transverse cracks is simulated until it is
no longer energetically possible to produce any more transverse
cracks, or until some other failure modes (e.g. delamination, fiber
break, etc.) set in during the loading process.

(F) Determining the effective flaw distribution

One difficulty in the above simulation procedure lies in the
fact that the effective flaws are hypothetical quantities, and that
they must be chosen properly to yield the essential features of
transverse cracking. Appropriate experiments are required to de-
termine the effective flaw distribution.
In the work of Lei (1986), the effective intralaminar flaw
distribution in the AS4-3501-06 carbon-epoxy unidirectional ply
was determined by testing [0°2/90O2]s tensile coupons. In the test,
transverse cracks were detected by X-radiography and were re-
corded as a function of the laminate tensile stress. The shaded band

Fig. 3.36. The energy retention factor, R(s), vs. s/nt due to the shear-lag
effect (after Wang 1987.)

10 12
s/nt
156 Strength of continuous-fiber composites

in Fig. 3.37 is formed by plotting data obtained from four

specimens, in terms of crack density (cracks per unit length of
specimen) versus the applied laminate stress. This band, repre-
senting a cumulative formation of the transverse cracks during
loading, resembles a form of the output from a certain stochastic
process.
It is noted that the experimental band possesses a certain position
on the stress scale, a certain characteristic curvature in the
coordinate plane and an asymptotic value on the crack density scale.
These features will now be used to determine the effective flaw
distribution in the [90°4] layer. To do so, a random number
generator is used to form a set of M random values in the interval of
(0,1). These M values are assigned to be {JCJ, the locations of M
flaws along the unit length of the [90°4] layer. The sizes of the M
flaws {aiA} are assumed to fit a Weibull cumulative function,
F(a) = 1 - e x p [ - (a/a)p] (3.148)

Fig. 3.37. Cumulative crack density (number of cracks per millimeter

specimen length) vs. applied laminate stress for [0o2/90o2]s laminates. The
shaded data band indicates experimental range of four specimens. The
dots represent results of Monte-Carlo simulations. (After Wang 1987.)
2.0
Experimental range
(four specimens)

1.5

1.0

u
0.5

0.0
250 500 750 1000
Applied stress (MPa)
 Statistical tensile strength theories 157

At this point, the parameters M, a and /J are assumed known.

And a new set of M random values is again generated in the interval
(0,1). These values are assigned to {^}, corresponding to the
values of F(a) at a = ai4. The flaw size {aiA} is then determined
using Eq. (3.148).
With the assumed values of a, j3 and M, a simulation of the
transverse cracking process as described earlier can now be per-
formed. An appropriate choice of a, /? and M is one that simulates
closely the experimental data band shown in Fig. 3.37.
Generally, a affects primarily the curvature of the band, /3 shifts the
band along the stress scale, and M determines the asymptotic value
of the band on the crack density scale (see Lei 1986). Figure 3.37
shows also the simulated crack density vs. laminate stress data from
five simulation specimens. Properly selected values of a, fi and M
can fit the experimental data band very well.
Once the values a, /? and M are chosen, the effective flaw size
distribution in any number of grouped 90° plies can be found using
Eq. (3.143); and then the transverse cracking in the grouped 90°
plies in laminates can be simulated. Figure 3.38 shows the simulated
results for four [0°2/90°]s coupons along with the experimental data
band from four test specimens. Figure 3.39 shows a similar
comparison between experiment and simulation for four [0o2/90o4]s
coupons. In both Figs. 3.38 and 3.39, the simulated data were based
on the flaw distribution found from the [0°2/90o2]s coupons in
conjunction with Eq. (3.143).
As was mentioned in Section 3.4.6.3, the Monte-Carlo method
depends on the nature of the input random variables; and in this
case, the input is the distribution of the assumed effective flaws. In
the examples discussed above, the values of a, (3 and M determined
by fitting the experiment could not be proved unique. Nevertheless,
the simulation, which is performed in conjunction with fracture
mechanics analysis, provides not only a quantitative description of the
mechanisms but also an assessment of the statistical characteristics
of the transverse cracking process.

3.4.8 Delamination in laminates of multi-directional plies

Delamination is another mode of failure in multi-directional
laminated plates and shells. At the ply level, delamination may be
viewed as a plane crack propagating in the interface between two
adjacent plies, Fig. 3.40. Cracking of this kind is peculiar because
the crack plane is parallel rather than perpendicular to the applied
tension; the driving force stems from the interlaminar stresses. As most
158 Strength of continuous-fiber composites

laminates are designed to carry in-plane loading, interlaminar

stresses are generally absent throughout the laminate except near
free edges, cut-outs, large defects and other such locations where
local interactions from mismatched ply properties cause stress
concentrations. Again, these local stress fields are almost always
three-dimensional in character.
The three-dimensional stress analysis model and the energy
method discussed in Section 3.4.7.4 can be applied to describe the
initiation and propagation of delamination. Crossman et al. (1980)
and Wang and Crossman (1980) followed this approach and
investigated free-edge delamination in laminates loaded in uniaxial
tension; Wang, Slomiana and Bucinell (1985) considered free-edge
delamination in compressively loaded laminates; and Wang,
Kishore and Li (1985) examined delamination near interacting
laminate defects. In all cases, experimental correlation was per-
formed to validate the analysis.

Fig. 3.38. Cumulative crack density (number of cracks per millimeter

specimen length) vs. applied laminate stress for [0°2/90°2]s laminates. The
shaded data band indicates experimental range of four specimens. The
dots represent results of Monte-Carlo simulations. (After Wang 1987.)

2.0

Experimental range
(four specimens)
1.5

I
1.0

2
u
0.5

0.0
250 500 750 1000
Applied stress (MPa)
 Statistical tensile strength theories 159

Fig. 3.39. Cumulative crack density (number of cracks per millimeter

specimen length) vs. applied laminate stress for [0°2/90°4]s laminates. The
shaded data band indicates experimental range of four specimens. The
dots represent results of Monte-Carlo simulations. (After Wang 1987.)
1.5

Experimental range
(four specimens)

1.0

0.5

0.0
250 500 750
Applied stress (MPa)

Fig. 3.40. Inter-ply cracking (edge delamination) in a multi-ply laminate.

(After Wang 1987.)
160 Strength of continuous-fiber composites

For conciseness, the problem of free-edge delamination in

laminates loaded in uniaxial tension is discussed in this section, and
only the logic underlying the formulation of the analytical method is
presented.

3.4.8.1 Free-edge delamination

The free-edge delamination problem has attracted con-
siderable interest for both its scientific challenge and engineering
importance. Early laboratory tests have shown that laminate tensile
strength can be greatly reduced if free-edge delamination occurs
during the course of loading (Pagano and Pipes 1971; Bjeletich,
Crossman and Warren 1979). A similar effect on laminate compres-
sive strength has also been confirmed (Wang, Slomiana and
Bucinell 1985). Further analyses of the delamination mechanisms
have established that the physical behavior of delamination is
profoundly influenced by ply stacking sequence, ply fiber orienta-
tion, individual ply thickness and laminate width to thickness ratio
(Crossman and Wang 1982).
While there have been many predictive models describing de-
lamination growth, the energy method developed by Wang and
Crossman (1980) accounts for all these intrinsic and extrinsic factors
operating in a severely concentrated three-dimensional stress field
near the free edges.
To illustrate this method, the symmetric laminate having straight
edges shown in Fig. 3.40 is considered as an example. Assume that
the laminate under the applied laminate tensile strain exx is such
that free-edge delamination is induced in one of its ply interfaces.
The problem is then to determine which interface is most likely to
delaminate and at what load.
For long, symmetrically stacked and finite-width laminates, it may
be assumed that the laminate stress field is independent of the
loading axis, x. Hence, it can be described by ply elasticity
formulation under the generalized plane strain condition (Pipes and
Pagano 1970). The induced free-edge delamination would then
extend uniformly along the length of the laminate and advance from
the free edges toward the center of the laminate piece, as shown in
Fig. 3.40; and the delamination crack can be considered as a
self-similar line crack with a linear size, a> propagating in the
preferred ply interface.

(A) Effective interlaminar flaws and conditions for propagation

To render a prediction for delamination initiation, the
 Statistical tensile strength theories 161

assumption of effective flaws (Section 3.4.7.4) will again be invoked

here. In this case, random interfacial flaws are assumed to exist in
each ply interface of the laminate as illustrated in Fig. 3.33(b). In
particular, along the laminate free edges there is a dominant
interlaminar edge flaw. It is further assumed that this flaw is located
in a known interface and has a linear size, ao, in the sense depicted
in Fig. 3.40. This flaw is treated as a starter delamination crack,
with its size ao still a random variable. Thus, one can proceed to
calculate the crack-tip strain-energy release rate G(ao, exx) if the
elastic constants of the unidirectional plies, the ply stacking
sequence, the ply fiber orientations and the ply interface in which ao
is residing are known.
The general character of G(ao, exx) as a function of ao is shown in
Fig. 3.41 (for a unit of the applied laminate strain exx). G rises
sharply from zero at ao = 0, and reaches an asymptotic value, Gasy,
as ao becomes greater than am. It should be noted that in Fig. 3.41,
Gasy can be expressed in terms of e2xx. The physical meaning of am is
that, at this size, the delamination no longer interacts with the
free-edge boundary. Generally, this boundary effect extends
roughly to a distance of about one-half the thickness of the laminate.
Beyond this distance, the delamination problem merely involves the
extension of cracks between two anisotropic elastic media and the
free-edge effect vanishes.
The calculated strain-energy release rate G may be expressed

Fig. 3.41. Variation of the strain-energy release rate G with delamination

crack size ao for a given exx value. (After Wang 1984). Copyright ASTM,
reprinted with permission.)

G(a0,

a0
162 Strength of continuous-fiber composites

explicitly in terms of the applied laminate strain exx:

G(ao, exx) = Ce(ao)2te2xx (3.149)

with Ce(ao) an exclusive function of delamination size ao. In the
above, 2t is this thickness of the ply.
Effects of thermal residual stresses due to cooling in fabrication
can be readily included in the calculation of G. If the laminate
stress-free temperature is To and the ambient temperature is T, then
the laminate is exposed to a temperature drop of AT = To — T. The
calculated strain-energy release rate G can be expressed in explicit
terms of exx and A T as

G(ao, exxy AT) = [\l{Ce)exx + V(C r ) AT]22t (3.150)

where CT is also an exclusive function of ao.
From fracture mechanics, the condition governing the onset of
delamination is given by:
G(aO9exx,AT) = Gc (3.151)
where Gc is the fracture toughness of the laminate under
delamination.
Equation (3.151) provides a prediction for the critical laminate
strain exx at the onset of delamination when the delaminating
interface, the values of ao and Gc are given. These values, however,
are not readily available; a further analysis of the problem is still
needed.

(B) The effective edge flaw size

Given the functional character of G(ao, exx) shown in Fig.
3.41, a one-to-one relationship between the critical exx and ao can
be obtained from Eq. (3.151) assuming the delaminating interface
and the associated Gc are known. If ao is represented by some
probability density function, f(ao), then there is a corresponding
range of exx for which Eq. (3.151) is satisfied (see Fig. 3.42). The
limiting value of exx as ao becomes equal to or greater than am is
determined by setting Gasy/Gc= 1. This serves as the lower-bound
of the critical strain, EXX. Since am is about one-half the thickness of
the laminate, it is small compared to the observable delamination
size in relatively thin laminates. In effect, the lower-bound value for
sxx is usually regarded as the critical strain at the onset of
delamination.
 Statistical tensile strength theories 163

(C) The critical delaminating ply interface

Given a specific laminate, the most probable delaminating
interface cannot be presupposed from experience. It requires an
analysis in which the values of Gasy/Gc on all possible interfaces
can be compared. According to Eq. (3.151), delamination shall
occur on the interface which yields the largest value of Gasy/Gc (for
the same exx).
While Gasy at each interface can be calculated readily, the Gc
associated with each interface may differ from one interface to
another. To elucidate this fact, consider a specific example: the
[±25°/90°]s laminate made of the AS4-3501-06 carbon-epoxy
system. Based on the generalized plane strain model mentioned
earlier, the entire laminate stress field is calculated first. Of interest
are the interlaminar stresses near the free edges before delamina-
tion. Figure 3.43 shows near the free edge, the through thickness
distribution of the interlaminar normal stress ozz. Note that ozz is
tensile and unbounded approaching the -25790° interface; and is
tensile but bounded on the laminate mid-plane (90790° interface).
Figure 3.44 shows the interlaminar shear stress rxz near the free
edge. Here, an unbounded RXZ exists on both the 25°/—25° and the
-25°/90° interfaces. These results suggest only qualitatively that
free edge delamination may occur either in the 90790° interface as
a mode I crack, or in the — 25790° interface as a mixed-mode
(mode I and mode III) crack.
Further energy analysis provides (Gy)asy for mode I cracks in the

Fig. 3.42. Relation between applied strain EXX and flaw size a o . Flaw size
distribution f(a o ) is shown schematically. (After Wang 1987.)

G(a o ,£ < v ) = G c
I \
Onset strain
range
i
I 1
1
i
1
/(flo) 1
/ \
164 Strength of continuous-fiber composites

mid-plane, and (Gx + G m ) asy for the mixed-mode crack in the

-25790° interface. In the latter, the mixed-mode ratio for Gm/G, is
also obtained.
Fracture toughness for mode I delamination may actually be
different from that for mixed-mode delamination. Indeed, interfa-
cial fracture of various mixed modes often manifest themselves

Fig. 3.43. The distribution of normal stress ozz through the laminate
thickness. (After Wang 1987.)

25° y

-25° \ ^

90°

1 1 1 1 1 \l 1 fc
-10 -5 10 15 20 25

Fig. 3.44. The distribution of shear stress xxz through the laminate
thickness. (After Wang 1987.)

-15 -10 35
 Statistical tensile strength theories 165

through differences in the fractured surface morphology, which in

turn implies differences in the Gc value measured on the ply
scale (Bradley and Cohen 1985). Laboratory tests using carbon-
epoxy specimens have shown that Gc for mode I delamination is
generally lower than Gc for mixed-mode delamination. And, the
latter often increases with the amount of the shearing mode. The
cause for variable Gc in mixed-mode delamination is complex;
several recent studies cited local crack-tip matrix yielding and fiber
bridging across the crack surfaces possibly due to shear deformation
(see Russell and Street 1985). To use Eq. (3.151) for mixed-mode
delamination, Gc must be first obtained as a function of mixed-
mode ratio.
For the example problem, as it turned out, mid-plane delamina-
tion was predicted because it yielded a larger Gasy/Gc than the
—25790° interface. The prediction agreed with the experiment (see
Wang, Slomiana and Bucinell 1985). It should be noted that besides
Eq. (3.151) many other fracture criteria for mixed-mode cracks
have been suggested in the literature.

3.4.8.2 General delamination problems

The free-edge delamination problem discussed above serves
to illustrate the basic rationale in the formulation of the energy
method. The assumption of effective interfacial flaws allows a
fracture analysis from which the onset of delamination could be
determined. The assumption may seem awkward at first glance;
but it is no more inconvenient than to assume the existence of a
stress-based interlaminar strength that is used to determine de-
lamination onset in the highly concentrated free-edge stress fields.
It should also be remarked that delamination problems encoun-
tered in practice are very complicated. Frequently, the delamination
plane has a two-dimensional contour. To describe the growth of a
contoured delamination may require a criterion which is direction-
ally dependent, due to different material characteristics along the
contoured crack front. In addition, delamination growth in practical
laminates is almost always accompanied and/or preceded by other
types of damages such as transverse cracks. Interactions amongst
the various local cracks with delamination can be both deterministic
and probabilistic in nature. The energy method discussed in this
section appears to have sufficient generality for application to the
more complex delamination problems. Generic extension of the
method could conceivably be developed which can provide quan-
166 Strength of continuous-fiber composites

titiative, if approximate, predictions for a wide class of delamination

problems.

3.4.9 Enhancement of composite strength through fiber

prestressing
The scattering in fiber strength has been attributed to the
existence of surface and bulk defects (See Section 3.4.2). Owing to
the statistical strength distribution of fibers, it is necessary to design
fiber composite structural components based upon a high level of
survivability. The enhancement of composite strength can be
achieved by eliminating some of the weak spots or defects in the
fibers. One way of attaining this goal is to stress the fibers and to
induce fracture at the defect sites before they are incorporated into
the matrix.
Mills and Dauksys (1973) were the first to adopt the concept of
fiber prestressing. In their work, carbon fiber prepregs are pre-
stressed at temperatures as low as — 18°C. The prestress of prepregs
by bending induces non-uniform tensile stress which reaches maxi-
mum values at the outer surfaces with fibers near the center of the
prepreg stressed the least.
Manders and Chou (1983b) provide a theoretical analysis of
enhancement of strength in composites reinforced with previously
stressed fibers. The basis of their reasoning is as follows. The failure
of a fiber in an aligned composite causes a stress wave to propagate
outwards placing a dynamic overstress on the neighboring fibers
(see Section 3.3.2). The resulting dynamic stress concentration is
generally greater than the static stress concentration which prevails
after the system has settled, and increases the probability that
adjacent fibers also fail, weakening the composite. This analysis
shows how weak fibers may be prefractured to eliminate the
dynamic overstress, thereby increasing the strength of the compos-
ite. Manders and Chou discussed this strength enhancement with
reference to the level of prestress, fiber variability, stress concentra-
tions, and size of the composite.
Chi and Chou (1983) have measured in a systematic fashion the
effect of fiber prestressing on the mean strength of composites as
well as the dispersion of composite strength. Thornel-300 carbon
fibers are used as the reinforcement materials for composites. A
loose bundle contains 1000 fibers with a fiber diameter of 7 fim. In
order to obtain consistent results in composite strength enhance-
ment, it is essential that all the defect sites of the fibers with
strength less than a certain value should be broken when they are
 Statistical tensile strength theories 167

subject to prestressing at a given level. It would be most ideal if a

uniform tensile stress could be applied uniformly to each small
segment of a fiber with length comparable to the ineffective length
of the fiber. However, this is impractical in real experiments, where
the gauge-length for fiber testing is much larger than the ineffective
length. Thus, a fiber already broken at its weakest site can no
longer be stressed under tensile loading.
The prestressing of carbon fibers is achieved by pulling the bundle
through a pair of circular bars of the same diameter at a tensile
force of 30 g. The relationship among the maximum prestress in
fibers, (jp, the bar diameter, D, and the fiber diameter, d, is
op = Efd/D (3.152)
where Ef denotes the fiber axial Young's modulus. The stress in the
fiber caused by the applied tensile force is much smaller than op and
hence it is neglected. Composite specimens are fabricated by
impregnating prestressed and non-prestressed fiber bundles in

Fig. 3.45. Negative strength enhancement in composites reinforced with

prestressed loose carbon fiber bundle. (After Chi and Chou 1983.)
o c (MPa)
2200 2600 3000 3400
i i r \
/ = 76.2 mm
o Non-prestressed 0.99
• Prestressed
0.9

0.7

0.5
I
0.3

0.1

0.05

0.03

1 1
7.7 7.9 8.1
In o c
168 Strength of continuous-fiber composites

epoxy resin. The strength data obtained for prestressed fiber

composites with gauge-length of 76.2 mm are shown in Figs. 3.45
and 3.46, using Weibull probability paper. Here, oc denotes
composite strength, P(oc) is the cumulative strength distribution
and ln{-ln[l — P(oc)]} indicates the failure probability. The D
values for specimens presented in Figs. 3.45 and 3.46 are 0.711 mm
and 1.168 mm, respectively; the resulting ap values are 2.21 GPa
and 1.35 GPa. The mean strength of the composites with non-
prestressed fiber bundles is 3.01 GPa. The strength data of Fig. 3.45
show negative enhancement while significant strength enhancement
can be seen in Fig. 3.46. It is noted that the strength data of
prestressed composites can be fitted approximately by straight lines.
Chi and Chou (1983) have concluded that the composite strength
for high survivability (low failure probability) is low. These low
strength tails can be eliminated by stressing the loose fiber bundles.
Enhancement in strength as high as 25% for survivability of
99.9% has been achieved.

Fig. 3.46. Positive strength enhancement in composites reinforced with

prestressed loose carbon fiber bundle. (After Chi and Chou 1983.)
o c (MPa)

2200 2600 3000 3400

1 I 1 1 1 1 1
/ = 76.2 mm
o Non-prestressed 0.99
o
• Prestressed o —
1 - o°*- 0.9

0 -

•
}'-- 0.7

I -
0.5

0.3
o
o° •
o
- 2 ~~- o
o - 0.1
o •

- 3 _- 0 - 0.05

0.03

1 1 1 1 1
7.7 7.9

Ino c
 Short-fiber composites

4.1 Introduction
Composites reinforced with discontinuous fibers are catego-
rized here as short-fiber composites. The fiber aspect ratio
(length/diameter = I Id) is often used as a measurement of fiber
relative length. Depending upon the dispersion of fibers in the
matrix, the relevant d values may include those of the filaments,
strands, rovings, as well as other forms of fiber bundles. Although
discontinuous fibers such as whiskers have been used to reinforce
metals and ceramics, the majority of short-fiber composites are
based upon polymeric matrices. Discontinuous fiber-reinforced
plastics are attractive in their versatility in properties and relatively
low fabrication costs. The concern of the rapid depletion of world
resources in metals and the search for energy-efficient materials has
contributed to the increasing interest in composite materials.
Discontinuous fiber-reinforced plastics will constitute a major por-
tion of the demand of composites in automotive, marine and
aeronautic applications.
A discontinuous fiber composite usually consists of relatively
short, variable length, and imperfectly aligned fibers distributed in a
continuous-phase matrix. In polymeric composites the fibers are
mostly glass, although carbon and aramid are also used; non-fibrous
fillers are often added. The orientation of the fibers depends upon
the processing conditions employed and may vary from random
in-plane and partially aligned to approximately uniaxial.
The understanding of the behavior of short-fiber composites is
complicated by the non-uniformity in fiber length and orientation as
well as the interaction between the fiber and matrix at fiber ends
(Chou and Kelly 1976, 1980). These factors are examined in the
following discussions on the physical and mechanical properties of
short-fiber composites.

4.2 Load transfer

Various attempts have been made to evaluate the stress
transfer from the matrix to the fiber in a short-fiber composite.
Analyses based upon the shear-lag theory, elasticity theory, and
finite element method have been performed. Considerations re-
170 Short-fiber composites

garding fiber aspect ratio (Fukuda and Kawata 1974), the effects of
bonded ends and loose ends as well as the geometric shapes of fiber
ends (Burgel, Perry and Schneider 1970), and the distribution of
radial and circumferential stresses near the interface at fiber ends
(Haener and Ashbaugh 1967; Carrara and McGarry 1968) have
been made. Experimental measurements of interfacial strength have
been made using a single fiber pull out test (Favre and Perrin 1972),
a fiber fragmentation test (Wadsworth and Spilling 1968), a
microtension test (Miller, Muri and Rebenfeld 1987) and a micro-
compression test (Mandell, Grande, Tsiang and McGarry 1986).
(Also see Piggott 1987 and Piggott and Dai 1988.)
Although the shear-lag approach is not as rigorous as the other
methods, it does provide a simplistic analysis for gaining some
insights into a complex problem and it will be employed in the
following. The fiber axial and interfacial stresses are discussed with
or without the consideration of interactions among neighboring
short fibers.

4.2.1 A single short fiber

Cox (1952) first dealt with the problem of a single short
fiber embedded in an infinite matrix material. In this essentially
one-dimensional approach the load on the fiber is considered to be
built up entirely due to the generation of shear stress in the matrix.
Under the assumptions of shear-lag analysis no tensile stress is
permitted to transmit across a fiber end.
Consider a long cylindrical composite of radius R which contains
a fiber of radius ro and length / along the cylinder axis. The
composite as a whole is subjected to a normal strain e in the
direction of the fiber. The assumption of the shear-lag analysis leads
to the following relation:

^ = H(u-v) (4.1)

where u(x) is the displacement of the fiber at the point x\ v(x) is

the matrix displacement; H is a constant; and P is the fiber axial
force. The force-displacement relation of the fiber is

P = EtAt^ (4.2)

where E{ and Af denote the fiber axial Young's modulus and cross-
sectional area, respectively. Substituting Eq. (4.2) and du/djc =
constant = e into Eq. (4.1), and applying the boundary conditions
 Load transfer 111

P = 0 at x = 0 and /, the fiber axial stress, crf, and interfacial shear

stress, r, are obtained:
cosh/3(//2-jt)-
cosh (pi/2) J
(4.3)
sinhp(l/2-x)
T =

where
/5 =

Here, Gm and Ef are the matrix shear modulus and fiber Young's
modulus, respectively. Figure 4.1 shows schematically the variation
of Of and r along the length of the fiber. The largest axial stress in
the fiber occurs at the center and it reaches Efe for a very long fiber.
The magnitude of T reaches its maximum at the fiber ends, i.e., at
x = 0 and /, and it vanishes at the middle point of the fiber.

4.2.2 Fiber-fiber interactions

The interactions among fibers in a short-fiber composite are
more complex than those in a continuous fiber composite. This is
because the axial load carried by a short fiber has to be transferred
to the neighboring fibers at locations near its ends. To illustrate the
load transfer in a short-fiber composite, the work of Fukuda and
Chou (1981a) is recapitulated in the following. This approach,
based upon the shear-lag model, introduces axial load into the

Fig. 4.1. The variation of of and r along a short fiber.

X= I
172 Short-fiber composites

matrix, and the fiber ends are assumed to be bonded to the matrix.
These assumptions of the modified shear-lag analysis are valid if the
bonding between the fiber and matrix at the fiber end is perfect such
as the cases often observed in metal matrix composites and in
polymeric matrix composites under compression.
The two-dimensional model for analysis is given in Fig. 4.2,
where the hatched parts of the matrix sustain axial load and behave
as if they are fibers with a Young's modulus different from the actual
fibers. The fiber diameter and matrix layer width are denoted by d
and h, respectively. A representative region in Fig. 4.2 containing
fiber ends is divided into n parts along the fiber direction x. Fibers
in this region are numbered from / = 1 to / = ra. Figure 4.3 shows a
free body diagram of a fiber and the adjacent matrix. The
equilibrium of forces in the x direction gives

dx

dPi,
i = 2, . . . ,m-\) (4.4)

dPm)
dx-

Fig. 4.2. The general model of analyses, = x/d. (After Fukuda and
Chou 1981a).
i v

rd
i
't'K'"t ' '

h
infn

i =1 1 1 1
42 (4)
X] Xi X

j= 1
 Load transfer 173

where Pti and R,Y are, respectively, the axial force of the ith fiber and
the interfacial shear stress in the yth region. The condition of linear
elastic deformation leads to the following stress-strain relations:

dx
(4.5)

where E, G and u denote the Young's modulus of the fiber, shear

modulus of the matrix, and displacement of the fiber, respectively.
The subscripts / and y indicate the ith fiber and yth region as shown
in Fig. 4.2. Thus, Etj is either E{ (Young's modulus of the fiber) or
Em (Young's modulus of the matrix).
The above general formulation is now applied to a model
composite shown in Fig. 4.4. This model is composed of a row of
short fibers of equal length and two surrounding long fibers. This
simple model is adopted for demonstrating the load transfer of short

Fig. 4.3. Free body diagram of the ith fiber. (After Fukuda and Chou
1981a.)

i i

h
1 r

i i

d
1 f
174 Short-fiber composites

fibers. Given i (=1 and 2) and; (=1 and 2) as shown in Fig. 4.4, the
following general solutions of u^ and Ptj are obtained from Eqs.
(4.4) and (4.5):
un =

P2l = £ f { S i - U i
(4.6)

M22 = A2

where
£ = */<*, atj = EvhlGd
k = EjEf

aya2)
l1( Bly Cly Dlf A2, B2, C2 and D2 are unknown constants.
The axial force and boundary conditions of the model of Fig. 4.4
are
(i) symmetry conditions
=e, (4-7)

Fig. 4.4. An example of a three-rowfibermodel. (After Fukuda and Chou

1981a.)
j = i 2

2
i
h
r
1 1
1
1= 1
0 1
Ui
 Load transfer 175

(ii) continuity conditions

(Mll)§ = §0 = ( M l 2 ) i = § 0 , ( W 2 l ) § = §o = ( M 22)i = §o.

(Ai)g=t o = (^12)5=?,,, (^2i)g=go = (^22)5=5,,

(ii) equilibrium of force
Pn + 2P21 = 3PO, P12 + 2P22 = 3FO (4.9)
where 3FO denotes the total applied load, and § o and ^ are given in
Fig. 4.4. The above conditions provide nine equations, of which
eight are independent, to determine the eight integral constants of
Eqs. (4.6). Finally, the axial load distribution becomes:

hA
Pn/Po 1 ~ k) 2(^ - go) cosh At§
it

P2JP0 = 1 + Xi{\~k) sinh A 2 (|, - | o ) cosh A,§

(4.10)
1+ 2 A z ( )sinhAl?ocosh
I V^
3 f A;A n k}
P22/P0 = YTk\l — I F sinh k
^°cosh Az(
^" §

where
F = kk2 cosh 2(gi g o) s m n i ? o
2 + A:
+ ^ ~ Ai sinh A2(gx - go) cosh A ^

The displacement field can also be obtained with the given

boundary conditions.
Limiting cases such as a single short fiber, a semi-infinitefiberand
two semi-infinite fibers separated by a gap can be deduced from Fig.
4.4. Furthermore, the solution for the case where no load is
transferred at fiber ends can be obtained by setting k = 0 in Eqs.
(4.10). Figure 4.5 shows the axial load distributions, for several
Ef/Em values, in the continuous and discontinuous fibers. Fukuda
and Chou (1981a) also concluded that the axial load distributions
near the fiber ends are essentially the same for fibers of different
length for given hid and Ef/Em values. This is demonstrated in Fig.
4.6 for the fiber configuration of Fig. 4.4. This finding is consistent
with Rosen's (1964) definition of ineffective length (Section 3.4.6.1)
176 Short-fiber composites

which is independent of the actual fiber length (Chen 1971; Fukuda

and Kawata 1977). Fukuda and Chou (1981b) have also considered
the effects of load transfer at fiber ends and plastic deformation in
the matrix.

4.3 Elastic properties

The elastic behavior of short-fiber composites has been
extensively studied. It is convenient to subdivide short-fiber com-
posites into three categories, according to their fiber orientations,

Fig. 4.5. Effect of Ef/Em on fiber axial load distribution, for h/d = l,
%Jd = 100, %Jd = 120. go and | t are defined in Fig. 4.4. (a) Continuous
fiber, and (b) discontinuous fiber. (After Fukuda and Chou 1981a.)

1.5r

P/Po

0 20 120

Fig. 4.6. Axial load distribution for fiber 1, and definition of fiber
ineffective length. h/d = l, Ef/Em = 20, and 6 = ineffective length. (After
Fukuda and Chou 1981a.)
-.1.0
l—Ud = 200

Fiber 1'

'h

200 60 50 40 30 20

= ( / i - x)/d )/d
H
 Elastic properties 111

for the purpose of stiffness discussions: (1) unidirectionally aligned

short fibers, (2) partially aligned short fibers, and (3) random short
fibers.

4.3.1 Unidirectionally aligned short-fiber composites

For unidirectionally aligned short-fiber composites, the
focus is on the effect of fiber length. Two major approaches are
presented for the prediction of elastic moduli of aligned short-fiber
composites. The first one is based upon a self-consistent model and
the second one gives the upper and lower bounds of elastic moduli.
The validity of some semi-empirical and numerical solutions is also
examined.

4.3.1.1 Shear-lag analysis

Using Cox's fiber stress expression of Eqs. (4.3), the
average fiber stress is

Based upon Eq. (4.11), the effective axial Young's modulus of the
short-fiber composite is approximated by
-Vf) (4.12)
where

and it represents a reduction of the composite elastic modulus due

to the finite length of the fiber.

4.3.1.2 Self-consistent method

The self-consistent method is a rigorous approach based
upon the assumptions that the fiber and matrix materials are
isotropic, homogeneous and linearly elastic, the fiber-matrix inter-
facial bonding is perfect, and the composite with aligned fibers is
macroscopically homogeneous and transversely isotropic. As re-
viewed by Chamis and Sendeckyj (1968), there exist two basic
variants of the self-consistent approach, namely the method by Hill
(1965a&b) and that used by Kilchinskii (1965, 1966) and Hermans
178 Short-fiber composites

(1967). Hill followed the method proposed by Kroner (1958) for

aggregates of crystals and modeled the composite as a single long
fiber embedded in an unbounded homogeneous medium which is
macroscopically indistinguishable from the composite. The model of
Kilchinskii and Hermans, on the other hand, consists of three
concentric cylinders: the innermost cylinder has the elastic pro-
perties of the fiber, the middle one simulates the pure matrix
material, and the outer one is unbounded and has the properties of
the composite. Hill has shown that the prediction of the self-
consistent method is more reliable at low and intermediate fiber
contents.
The approach of Hill has been adopted by Chou, Nomura and
Taya (1980) to treat the stiffness of short-fiber composites. In their
work, a single inclusion is assumed to be embedded in a continuous
and homogeneous medium (see Hill 1952; Eshelby 1957; Hashin
and Rosen 1964; Mura 1982). The inclusion has the elastic
properties of a short fiber while the surrounding material possesses
the properties of the composite. It is the unknown elastic property
of the composite that needs to be found. The work of Chou et al.
does not restrict the number of component phases in the composite
and is hence applicable to hybrid composites (Chapter 5). Numeri-
cal examples of this self-consistent approach are given for the
special case of a binary system of one kind of fiber in a matrix.
Figure 4.7 shows the variation of longitudinal modulus En of a
glass/epoxy system with inclusion volume fraction V{ at three
different inclusion aspect ratios {lid). For lid =100 the self-
consistent theory predicts that the inclusions behave like continuous
fibers and the rule-of-mixtures is valid. Also shown in Fig. 4.7 are
the predictions of the semi-empirical relation of Halpin and Tsai
(see Halpin 1984). The discrepancy between the self-consistent
theory and the Halpin-Tsai equation is most pronounced at
intermediate values of the aspect ratio. Comparisons of the self-
consistent approach with experiments are given in Section 4.3.2,
where the effect of fiber misorientation is taken into account.
The predictions of elastic stiffness for particulate-filled composites
have been performed by a number of investigators, including
Kerner (1956), van der Poel (1958), Hashin and Shtrikman (1963)
and Budiansky (1965). The self-consistent theory reduces to
Budiansky's solution for the special case of l/d = 1.

4.3.1.3 Bound approach

Nomura and Chou (1984) also adopted an alternate ap-
proach to short-fiber composite effective moduli by deriving their
 Elastic properties 179

upper and lower bounds. This approach was motivated by the work
of Eshekby (1961), Hashin (1965a), Kroner (1967, 1972, 1977),
Dederichs and Zeller (1973), Zeller and Dederichs (1973), Wu and
McCullough (1977), and Christensen (1979). Nomura and Chou
adopted a perturbation expansion of the composite local strain
based upon the elastic Green's function. The effective elastic
constants can be expressed in infinite series form. When the series
are written in terms of the stiffness constants, the first term is the
well known Voigt average (1889). The first term of the series
represents the Reuss average (1929) when the expression is written
in terms of the compliance constants. Based upon the assumptions
that the short fibers are modeled as aligned ellipsoidal inclusions
and distributed in the matrix material in a statistically homogeneous
manner, Nomura and Chou have evaluated the series expressions of
the elastic constants up to the third-order term. A variational
treatment has been utilized to derive the bounds of the effective
elastic moduli of the unidirectional short-fiber composite.

Fig. 4.7. The variation of Ell/Em with V{ at various l/d values for
EjEm = 20, vf = 0.3 and vm = 0.35. self-consistent approach;
Halpin-Tsai equation. (After Chou, Nomura and Taya 1980).

11.0

10.0

9.0
lid = 100
8.0

7.0

6.0

5.0

4.0

3.0

2.0

1.0
0.1 0.2 0.3 0.4 0.5
180 Short-fiber composites

Figure 4.8 illustrates the variation of the axial Young's modulus En

(normalized by the matrix modulus Em) with fiber volume fraction
Vf at fiber aspect ratios IId = 1,5 and °°, for glass/epoxy compos-
ites. The solid lines indicate the upper and lower bound predictions;
the predictions of the self-consistent model of Chou, Nomura and
Taya (1980) are indicated by broken lines. The self-consistent
model prediction is close to the lower bound at low fiber volume
fraction and approaches the upper bound at high fiber volume
fraction. The gap between the bounds at a fixed fiber volume
fraction narrows as the fiber aspect ratio increases. For long
continuous fibers, the bound approach and the self-consistent model
all predict the rule-of-mixtures relation. Although fiber volume
fraction in the full range of 0 to 1 is used in Fig. 4.8, it is understood
that the maximum attainable fiber volume fraction in a composite is
determined by the fiber geometric packing and fiber cross-sectional
shape.
Figure 4.9 shows the comparison of the bound approach with
Hashin's (1965a) results for the effective axial shear modulus Gl2 of
continuous fiber composites. The theory of Nomura and Chou
(1984) predicts tighter bounds than those of Hashin. This is due to

Fig. 4.8. The variation of En/Em with Vf for £f/£m = 20, vm = 0.4 and
vf = 0.3 bound approach; self-consistent model. (After
Nomura and Chou 1984).

20 r-

15

10

I I _L
0.0 0.2 0.4 0.6 0.8 1.0
 Elastic properties 181

the fact that Hashin's result is equivalent to the evaluation of the

series expression of the elastic constants up to the second-order
term. The bounds of effective elastic moduli of multi-phase systems
such as hybrid composites can also be examined by this approach.

4.3.1.4 Halpin—Tsai equation

The Halpin-Tsai equation (see Halpin 1984) was obtained
by reducing Hermans' solution (1967) to a simpler analytical form
while the filament geometries are taken into account through the
use of some empirical factors. The pertinent relations are

+
(4.14)

Fig. 4.9. The variation of G12/Gm with Vf for Ef/Em = 20, v m = 0.4,
vf = 0.3 and l/d—>*>. bound approach; self-consistent model;
bounds of Hashin and Shtrikman. (After Nomura and Chou
1984).

25 r

0.0 0.2 0.4 0.6 0.8

182 Short-fiber composites

where

P = En, E22, G12 or G 23

Pf = E{ (for E n and E22) or Gf (for G12 and G23)
Pm = E m (for E u and £22) or Gm (for G12 and G23)
Other solutions of effective elastic constants can be found from,
for instance, the numerical work of Conway and Chang (1971), and
Chang, Conway and Weaver (1972). Experimental data on short-
fiber composite elastic properties have been reported by Lees
(1968), and Blumentritt, Vu and Cooper (1974).

4.3.2 Partially aligned short-fiber composites

It is usually desirable to orient the fibers for enhanced
stiffness and strength properties. However, perfect alignment of
short fibers in a composite is normally very difficult to achieve.
Partial fiber alignment is typical in, for example, injection molded
composites. Several different approaches have been adopted by
researchers to predict the stiffness of short-fiber composites with
biassed fiber orientation. The following discussions of these ap-
proaches begin with a brief summary of the original treatments on
misaligned continuous fibers.
The first attempt in examining the effect of fiber orientation is
attributed to the work of Cox (1952), who studied the elastic
properties of paper and other fibrous materials. Cox's model is
concerned with continuous fibers of negligible thickness with
orientations either random or defined by some distribution rules.
The contribution of matrix to stiffness is ignored. It is also assumed
in this model that under load the fibers do not slide across each
other at the points of intersection (see Cook 1968).
Cook (1968) provides the elastic properties of continuous fiber
composites in three dimensions. The systems of misorientation
examined by Cook include the axially symmetric type, a fan shaped
 Elastic properties 183

array and systems of crossed fibers. Fiber orientation distribution

functions are generated analytically to describe the characteristics of
these systems. A case of most practical significance is the axially
symmetric fiber distribution, which is also termed the witch's broom
by Cook. The degree of fiber scatter from perfect alignment is
described by the root-mean-square deviation of orientation from the
symmetry axis
run

•-r Jo
r](6)e2sin Ode (4.15)

where r\{6) is the fiber orientation distribution function. According

to Cook, for a composite such as glass fibers in a polymer resin
(VfEf/VmEm~20) the orientation effect can be minimized if the
fibers are sufficiently long and, hence, a high degree of orientation
can be achieved. On the other hand, for whisker reinforced
composites the reduction in stiffness may be significant if the fibers
are short and alignment becomes a difficult technical problem. Cook
reported that for a silicon nitride whisker reinforced epoxy resin
composite examined, stiffness reduction of 4-19% could occur for
the root-mean-square scatter between 4.5° and 10°.
Fukuda and Kawata (1974) considered the Young's modulus of
short-fiber composites and took into consideration variations in
both fiber length and orientation. The analysis is based upon the
plane stress elasticity solution of load transfer between the fiber and
matrix in a single short-fiber model, and the assumption of
negligible interactions between neighboring fibers. The prediction
of the composite Young's modulus is given in the general form
Ec G CeEM + EM- Vf) (4.16)
The factors Cz and Ce reflect the effects of fiber length and
orientation distributions, respectively. Both Q and Ce are unity in
the case of aligned continuous fibers.
Figure 4.10 shows the variations of Q with the factor
(l/d)(Ef/Em) where IId denotes the fiber aspect ratio. The open and
solid circles in Fig. 4.10 are experimental values of Anderson and
Lavengood (1968) for glass/epoxy and boron/epoxy, respectively.
The solid line is obtained from a two-dimensional analysis and the
broken line is the modified result when the fiber circular cross-
sectional shape is taken into account. Figure 4.11 shows the
variations of Ce with r](6), which is the probability density of fibers
at the orientation angle 6. Fukuda and Kawata assume that
Jo/2 Tj(O) dd = 1. Three forms of rj(6) are assumed: rectangular,
184 Short-fiber composites

sinusoidal and triangular. 6O defines the range of fiber angular

distribution. Comparisons of the predictions of Eq. (4.16) with the
measured modulus of an ar-SiC whisker/aluminum composite
(Schierding and Deex 1969) are favorable.
The above discussions have centered upon either continuous
fibers or short fibers in planar arrangement. A treatment of the
three-dimensional fiber orientation effect has been developed by
Chou and Nomura (1981). They considered an axially symmetrical
fiber orientation distribution. Referring to Fig. 4.12, the general
orientation of a short fiber can be considered as derived from the

Fig. 4.10. Relation between C, and (l/d)/(Ef/Em). (After Fukuda and

Kawata 1974).

0.5

o Glass/epoxy
• Boron/epoxy

0
0.1 0.2 0.5 1 2 5 10 20 50 100
(l/d)(Ef/Em)

Fig. 4.11. Relation between C e and 0O. (After Fukuda and Kawata 1974).
1.0

0.5

nl 1 1 1
0 10 20 30 40 50 60 70 80 90

d0 (degrees)
 Elastic properties 185

original position along the z axis by two rotations. The cor-

responding rotational angles are cp and Q, as indicated in Fig. 4.12.
The transformation matrix, from the original coordinate system to
the current system, is defined as
sin 6 cos cp sin 6 sin cp cos 6
cos 6 cos (p cos 6 sin q? —sin 6 (4.17)
—sin (p cos (p 0
Let the bold-faced letter indicate a tensor. The transformation of an
elastic stiffness tensor C (or compliance tensor S) of a unidirection-
ally aligned short-fiber composite can be performed through the
application of the T matrix and the resulting tensor is denoted by C
(or S')« The effective elastic tensor of a misaligned short-fiber
composite is then given by

Z" = jci(d,cp)rl(e,cp)dA

= [ dcp j C'(0, cp)r)(0, cp) sm dd6 (4.18)

Jo Jo

Here, rj(6y cp) in the above equation is the probability density

function of fiber orientation determined from experiments. The
integration is carried out over the surface area of a unit sphere to
include all the fibers in the composite.

Fig. 4.12. Reference coordinate axes.

dA
186 Short-fiber composites

Two cases of fiber orientation distribution are of practical

importance. In the case of injection molded objects, fiber orienta-
tion distribution is independent of the angle cp if the direction of
flow is along the z axis, and rj = rj(6). The composite in this case is
isotropic in the plane transverse to the z axis, and C" is independ-
ent of cp. In sheet molding compounds, it is reasonable to assume
that the short fibers all lie on the xz plane and the problem is
two-dimensional. The transformation matrix is
sin 0 0 cos 6
T= cos 6 0 —sin 6 (4.19)
0 1 0
Equation (4.18) is then reduced to

C" = 2n \ C'(d)r)(6)sin0d6 (4.20)

It has been pointed out in the variational treatment of Section

4.3.1 that the first term in the series expression of composite
stiffness constant or compliance constant gives the well known
Voigt's upper bound or Reuss' lower bound. The averaging
principles of Voigt and Reuss were first used to predict the elastic
properties of a polycrystalline aggregate in terms of the basic
properties of a single crystal and its orientation in the aggregate.
The Voigt and Reuss averages are equivalent to assuming that the
single crystals are arrayed in parallel and in series, respectively.
These concepts of Voigt and Reuss averages are also useful in
dealing with misaligned composites. They can be expressed in the
general forms for the stiffness constant C and compliance constant S
as

À= f C(r, 0, <p)dv/\ dV
Jv I Jv
(4.21)
(S)= I S(r, 6, <p)dv/\ dV
Jv I Jv

where, in general, C and S are functions of position (r, 6, cp) as

shown in Fig. 4.12. Furthermore, it can be shown that in the Voigt
and Reuss averaging processes for small fiber misalignment there is
negligible difference between the model involving a distribution of
fiber orientations and the model in which all the fibers are aligned
along the direction of the root-mean-square average angle (see
 Elastic properties 187

Knibbs and Morris 1974). The treatment of effective elastic moduli

for partially aligned short-fiber composites can also be achieved
through the laminated plate analogue, which is discussed in Section
4.3.3.

4.3.3 Random short-fiber composites

The treatment of Cox (1952) discussed in Section 4.3.2
deals with the stiffness of continuous fibers distributed in a plane.
For completely random fiber distribution, Cox's results are reduced
to the simple forms
Ec = EfV{/3
Gc = EtVf/8 (4.22)

where Ec, Gc and vc are, respectively, the Young's modulus, shear

modulus and Poisson's ratio of the composite. The random distribu-
tion of fibers imparts isotropic properties of the composite at the
macroscopic scale. Hence, Ec, Gc and vc satisfy the relationship for
isotropic materials:
G c = £c/2(l + v c ) (4.23)
The contribution of matrix material is neglected in this treatment
but has been taken into account in the works of Arridge (1963), and
Pakdemirli and Williams (1969), who also derived approximate
expressions for Ec and Gc.
Nielsen and Chen (1968) proposed that the in-plane Young's
modulus of a random composite with continuous fibers can be
approximated by an averaging process. Basic to this process is the
knowledge of the elastic moduli of a unidirectional fiber composite
measured at an angle 6 from the fiber direction (see Eqs. 2.19). The
effective in-plane Young's modulus of a random composite, for
example, is then given by

E(6)dd (4.24)
Jl Jo

In applying Eq. (4.24), the fiber volume fraction of the composite

used for calculating E(6) should be the same as that in the random
composite. It should also be noted that E(6) is not a component of
a tensor. Hence, the averaging process defined in Eq. (4.24) bears
no relation to the Voigt and Reuss averages discussed in Section
4.3.2.
188 Short-fiber composites

The elastic moduli of a composite where the short fibers exhibit

in-plane random orientation can also be examined by the method of
a laminate analogue (Halpin 1969; Halpin and Pagano 1969;
Halpin, Jerine and Whitney 1971). The following discussions are
based upon reviews by Kardos (1973) and Nicolais (1975). In
the laminate analogue the mechanical response of the composite is
simulated by that of a laminate composed of unidirectional short
fibers (Kardos 1973). The laminate is symmetric about the mid-
Fig. 4.13. (a) Laminate analogue of a composite with random in-plane
orientation of short fibers. The quasi-isotropic laminate has the [+45°/ —
45°/90°/0°]s configuration, (b) Dependence of tensile modulus on volume
fraction of 3.2 mm E-glass/polycarbonate composites for random in-plane
(dashed curve) and biassed (solid curves) fiber orientations. is
quasi-isotropic calculation; weighted distribution calculations; O, •
experimental data. Fiber aspect ratio is about 313. (After Halpin, Jerine
and Whitney 1971).
Quasi-isotropic laminate

Random in-plane orientation

 Elastic properties 189

plane and has the same number of +6 and —0 orientation plies

(Fig. 4.13a).
The concept of the laminate analogue is outlined in the following.
First, the four independent elastic moduli Eu, E22y v12 and G12 of a
unidirectional short-fiber lamina can be derived from the fiber and
matrix properties based upon the self-consistent model, the varia-
tional method, or the Halpin-Tsai equation. The stiffness matrix
components Qtj are given by Eqs. (2.14). The effective engineering
stiffness constants En E22, v21 and G12 for the aligned short-fiber
lamina can be expressed in terms of the g,7's as given in Table 2.1.
The stiffness matrix components Qtj for a unidirectional lamina
oriented at an angle 0 with respect to the x axis are given in Eqs.
(2.16). They can also be written in the following alternate forms:
Qn = UX + U2 cos 20 + U3 cos 40
Q22 =Ul-U2 cos 20 + f/3 cos 40
<212=£/4-£/3cos40
(4.25)
Q 66 =£/ 5 -£/ 3 cos40
Gi6=2C/isin20+ {/3sin40
Q26 = \u2 sin 26 - U3 sin 46

Fig. 4.13. (cont.).

40

30

20
a*
o
15

10

Random
Non-random

_L 1
0 0.1 0.2 0.3 0.4 0.5 0.6

(b)
190 Short-fiber composites

where the Ut are defined as

When the plies are stacked together to form a laminate, the

in-plane stretching stiffness Ai} is given by Eqs. (2.29). For the case
of a balanced angle-ply (±6) composite with mid-plane symmetry,
the bending stiffness Btj (Eqs. 2.29) and the coupling terms Al6 and
A26 vanish, and the Atj components can be represented by
An = [Ul + U2 cos 26 + U3 cos 40]h
A22 = [Ux - U2 cos 26 + U3 cos 46]h
(4.26)
A l2 = [U4-U3 cos 48]h
A66 = [U5-U3 cos 46]h
Here, h denotes the total laminate thickness. Following the same
reasoning for the derivation of Eq. (2.15), the effective engineering
constants of the laminate are given by
AA—A2 ~ A2
A22- h
AnA22 A22
£22
A,r h
(4.27)
V,2 =
—Al2
_A66
G12
h
If a random short-fiber composite assumes the form of a thin
sheet while the sheet thickness is less than the average fiber length,
the composite can be modeled as a 'quasi-isotropic laminate'. In
principle, the laminate can be constructed by stacking up unidirec-
tional laminae in all orientations to achieve a balanced and
symmetric arrangement. Because the fiber orientation covers all the
values between 0° and 180°, the angular dependent terms in the Atj
 Elastic properties 191

components of Eqs. (2.16) cancel one another. Consequently, the

effective engineering constants can be simplified as

f/i

f/i - 2f/5
(4.28)

It is obvious that the above elastic constants satisfy the necessary

relation for in-plane isotropy. Expressions for random fiber com-
posite elastic constants equivalent to Eqs. (4.28) also have been
obtained by Akasaka (1974). Halpin, Jerine and Whitney (1971)
have demonstrated the validity of the laminate analogue by
comparing the analytical predictions with the measurement of
effective tensile modulus of E-glass/polycarbonate with random
fiber orientation as shown in Fig. 4.13(b).
The laminate analogue can also be applied to quasi-isotropic
short-fiber composites using lay-ups such as 0°/±60° and 0°/
±45790° (Warren and Norris 1953). Other works dealing with the
elastic stiffness of random fiber composites can be found from Tsai
and Pagano (1968); Manera (1971); Christensen and Waals (1972);
Wilczynki (1978); and Hahn (1978). As pointed out by Bert (1979),
the accuracy of these approximations is affected by the fiber volume
fraction and the ratio Ef/Em. The laminate analogue can also
be used for examining the elastic properties of short-fiber com-
posites with layered microstructures. Figure 4.14 shows the scanning
electron micrograph of the cross-section of an injection molded poly-
ethylene terephthalate with short glass fibers. This type of layered
structure has been found in many types of short-fiber reinforced
thermoplastics.
Attempts also have been made to predict the stiffness of
composites with random fibers in three-dimensional distribution.
Rosen and Shu (1971) and Christensen and Waals (1972) examined
the case of continuous fibers. Halpin, Jerine and Whitney (1971)
treated the case of layers of plain woven fabric in which the unit
weave cell is pierced by a straight yarn perpendicular to the fabric
plane. The problem of random short fiber orientation in three
dimensions has been treated by Chou and Nomura (1981). By
taking rj — 1/2N in Eq. (4.18), elastic moduli for completely random
orientation can be obtained. Figure 4.15 illustrates the theoretical
192 Short-fiber composites

variations of Ec/Em with V{ for a random glass/epoxy system and

the experimental data of Manera (1971).
The laminated plate analogue developed above can also be
applied to consider in-plane partially aligned short fibers (Halpin,
Jerine and Whitney 1971; Kardos 1973) discussed in Section 4.3.2.
In this case the angular fiber distribution function rj(6) needs to be
measured from the composite specimen. The laminate simulating
the composite is treated as composed of weighted groups of angle
plies (±0) with fixed fiber volume fraction. The percentage of
materials oriented at the angles ±6 is obtained from rj(6). The
contributions to the overall response of laminate stiffness from
different layers are proportioned to their fractional thickness in the
laminate.
Table 4.1 gives an example of the orientation distributions of
discontinuous glass fibers in a polymeric matrix. The composite is

Fig. 4.14. SEM micrograph of short glass fiber/polyethylene terephthalate

showing layered structure of fiber orientations. (After Friedrich and
Karger-Kocsis 1989.)

200 /xm
 Elastic properties 193

compression molded from extrudate. It can be seen that most of the

fibers are oriented quite close to the extrusion direction. Whereas
previously each ±6 ply was weighted equally in summing up the
stiffness contributions to the laminate, one must now account for
the fact that more of the laminate thickness may be made up of one
angle than the other. Define a(6)/h as the percentage of the
material oriented at the angles ±0, and it is obtained from the
experimental angular distribution rj(6) where jo r](0) d6 = 1. The
stiffness moduli Atj of the laminate is related to the stiffness of the
plies Aij(Ok), oriented at the angles ±0ky by

1«=i^W (4-29)
where n is the total number of plies.
In summary, the calculation of the effective engineering stiffness
of short-fiber composites with biassed fiber orientations should first
follow the procedure outlined in Section 2.3 to obtain the Atj
components for each fiber angle. These are then summed according
to their fiber angular distributions such as that given in Table 4.1
and Eq. (4.29) to obtain the Atj terms. The engineering constants
are then obtained from Eqs. (4.27). The solid lines in Fig. 4.13(b)

Fig. 4.15. The comparison of EJEm ( bound approach; self-

consistent model) with experimental data for Ef/Em = 32.4, vm = 0.4,
vf = 0.25 and //</->«>. (After Chow and Nomura 1981.)

10

6
E

0.0 0.2 0.4 0.6

194 Short-fiber composites

are theoretical predictions of the tensile moduli based upon this

procedure. The bumps in the curves are attributed to the fact that
the angular distribution functions are not smooth functions of fiber
volume fraction.

4.4 Physical properties

The physical properties described below include thermal
conductivity and thermal expansion coefficients. These properties
are essential to the study of the thermomechanical behavior of
short-fiber composites.

4.4.1 Thermal conductivity

The important transport properties of composites include
dielectric constant, heat conduction, electrical conduction, magnetic

Table 4.1. Fiber orientation distributions in composites

compression molded from rod extrudate. Short glass fiber aspect
ratio « 313. After Halpin et al (1971)

Orientation 6 (degrees) Percent fibers having 6 orientation

2.5 23.4 25.4 25.0 36.5

7.5 17.9 18.1 23.8 23.9
12.5 12.0 12.3 16.4 14.2
17.5 16.0 7.7 10.0 5.7
22.5 6.2 6.4 6.8 3.0
27.5 5.9 5.6 4.8 2.7
32.5 4.4 4.6 3.1 1.8
37.5 4.6 3.1 2.4 2.0
42.5 2.6 3.4 1.6 1.0
47.5 1.7 1.9 1.3 0.4
52.5 0.4 1.3 0.8 0.7
57.5 0.7 0.7 1.1 0.8
62.5 1.0 1.4 0.9 0.5
67.5 0.7 1.1 0.7 0.7
72.5 0.1 2.1 0.4 0.5
77.5 0.9 0.9 0.6 0.8
82.5 0.5 2.3 0.3 0.9
87.5 1.0 1.4 0.1 0.8
Fiber volume fraction 20 30 40 50
 Physical properties 195

permeability and diffusion coefficients. Since all these properties are

second rank tensors, only the bounds of thermal conductivity are
demonstrated.
The linear relation between the heat flux q and gradient of
temperature T is given by

q = k(-Vr) (4.30)

where k denotes thermal conductivity and is assumed to be a

function of position only. It is understood that k is a symmetric
tensor quantity. The governing equation for a steady-state heat
conduction is

V-q = 0 (4.31)

Several approaches to this subject have been employed by

researchers. These include the statistical method by Beran (1965),
Beran and Molyneux (1966), and Hori and Yonezawa (1975) as well
as the self-consistent and variational approaches of Hashin and
Shtrikman (1962), Hashin (1968) and Willis (1977).
Nomura and Chou (1980), following their development of bounds
of elastic moduli (1984), derived bounds of effective thermal
conductivity of unidirectional short-fiber composites. The short
fibers are again modeled as ellipsoidal inclusions of the same length
and are distributed in a statistically homogeneous manner in the
matrix material. The composite exhibits transverse isotropy. This
approach is also valid for composites containing more than one type
of fiber. Consider the case of a binary system and denote the
thermal conductivity and volume fraction of the fiber and matrix
phases by kt, V{ and km, Vm, respectively. The bounds of the
effective composite conductivity kn along the fiber directions are

kf km
kmj

Vm(kf-kmf(l-h
ff m m K i}
(vm-vf)(kt-km)(i-h(t)) +vtkt+vmkm '
The bounds of the conductivity k22 and k33 in the transverse
196 Short-fiber composites

direction are

= £ *22
,,=
- *33
+
f k

VtVm(kt-km)2h(t) (4.33)
(Vm - Vf)(kf - km)h{t) + 2{Vtkt + Vmkm)
where

t2-i

(4.34)

and f denotes the aspect ratio l/d of the short fiber.

For the special case of spherical inclusions (h(t) = l), the com-
posite is isotropic and Eqs. (4.32) and (4.33) are simplified as

VfVm(kf-km)2
(4.35)
t ~ Vm)(kf - km) + 3(Vtk{ + Vmkm)

In the case of continuous fibers, h{t) = 1 and Eqs. (4.32) and (4.33)
become
kn = Vfkf+Vmkm (4.36)
(km + kf)kmkf s (Vmkm + V(ktf + kmkf
k22( = k33) <
(Vfkm + Vmkf)2 + kmkf
(4.37)
Figure 4.16 illustrates the variations of ku/km with fiber volume
fraction of an E-glass/epoxy system for the limiting cases of //<i—»<»
and l/d = 1. The bounds of kn converge to a single line for
continuous fibers as indicated by Eq. (4.36).
For axially symmetrical fiber arrangement at an angle 0 with
respect to the XX axis, the fiber orientation effect can be investigated
as in Section 4.3.2. By transforming the effective thermal conduc-
tivity tensor ktj based upon the [T] matrix of Eq. (4.17) and
 Physical properties 197

subsequently integrating the tensor components over the In range

of cp, the resulting components are transversely isotropic with
respect to the x2-x3 plane:

(4.38)
+ cos2 e

By substituting the bounds of ktj (Eqs. (4.32) and (4.33)) into the
above expressions, the bounds of thermal conductivity can be
expressed as functions of fiber orientation 6. Again, Eq. (4.18) can
be used to find the effective thermal conductivity of a composite
with a given rj(6).
For completely random fiber orientation, the result can be
simplified to

2*22
(4.39)
3

Fig. 4.16. The variation of the upper and lower bounds of kn/km with Vf
for an E-glass/epoxy system. (After Nomura and Chou 1980.)
198 Short-fiber composites

4.4.2 Thermoelastic constants

Knowledge of the thermoelastic constants, including ther-
mal expansion coefficients and thermal stress coefficients, is basic to
the understanding of the hygrothermal effects in composites. So far
as it is assumed that these quantities obey the linear constitutive
equation, their solutions can be obtained in a manner similar to the
determination of effective elastic moduli or thermal conductivities.
The problem of effective thermoelastic constants for non-
homogeneous materials has been investigated by several research-
ers. The works of Kerner (1956), Levin (1967), Schapery (1968) and
Budiansky (1970) are mainly concerned with composites reinforced
with spherical inclusions. Rosen and Hashin (1970) extended
Levin's model of a binary composite to general anisotropic compos-
ites by adopting a variational approach. Laws (1973) studied the
thermoelastic behavior of anisotropic composites based upon Hill's
self-consistent approximation.
By focussing attention on thermostatics and considering
composites at uniform temperature, heat conduction can be ex-
cluded and the problem is uncoupled with that given in Section
4.4.1. Consider a composite subjected to a stress field, a, and a
uniform temperature rise, AT. The total strain of the elastic
medium is given as

e = So+aAT (4.40)

where S denotes the elastic compliance tensor, and a is the thermal

expansion coefficient. The constitutive relation of the thermal
elastic field can also be expressed in the following general form:

o = C(s-aAT) (4.41)

where C is the elastic stiffness tensor.

Nomura and Chou (1981) have shown that for composites
reinforced with ellipsoidal inclusions and exhibiting statistical homo-
geneity, the effective thermoelastic constants can be evaluated
following the technique for deriving elastic moduli. Figure 4.17
shows the variation of atj (normalized by the fiber thermal
expansion coefficient af) with V{ and fiber aspect ratio l/d for a
glass/epoxy system, assuming E{ = 72.3GPa, E m = 2.76GPa, vm =
0.35, vf = 0.2, am = 36 x KT 6 / ° C and af = 5.04 x 10- 6 /°C At a
given fiber volume fraction, the thermal expansion coefficient along
the fiber direction (an) is smaller than that transverse to the fiber
direction (ar22)- Figure 4.18 shows a comparison of the theoretical
Physical properties 199

Fig. 4.17. The variation of aJa{ with V{ and l/d for an E-glass/epoxy
system. l/d=l; l/d = 5; //d = °°. (After Nomura and
Chou 1981.)
12 i
11
\
FT" *^ \
N \
10 N
l» \
\ \
\ \\
9 * N
\ \
\
> . \
•• \ \ \\
8 -i\ \ \
\
\'\ \ \ \

_ 7 V
\ \
~ \ \
\ \ \ \
a" 6- \
\ \ \ \ \
\ \
5 - \ \
\ a s
\ \ \
\
4 \ \
\
\
s
-
\ \
i

3 —
\
\
\
an X W ^ /
X •
>
2 -
— — """••—•

1 1 i
0.0 0.2 0.4 0.S 1.0
vf

Fig. 4.18. Comparison of the predicted a^/a, with experimental data of

l. (1978).

0 _L _L
0 0.2 0.4 0.6 0.£ 1.0
V(
200 Short-fiber composites

prediction of Nomura and Chou with the experimental results of

Yates et al. (1978) on a carbon/epoxy system where Ef/Em = 53.4,
vm = Vf = 0.34, am = 5 x 10"5/°C and af = 0.5-1.9 x 10-5/°C.

4.5 Viscoelastic properties

The viscoelastic properties of composite materials were first
examined by Hashin (1965b, 1969, 1972), who dealt with matrices
reinforced with spherical inclusions and continuous fibers. Hashin
showed that viscoelastic problems in composite materials can be
solved by considering the corresponding problems in elasticity.
Although application of the elastic-viscoelastic correspondence
principle (see, for example, Christensen 1971) is well known, there
are practical difficulties. This is due to the fact that very often the
creep compliances or relaxation moduli of the constituents of a
multi-component system are not known, and, even if they are given,
the inverse transformation process would be formidable. Approxi-
mate methods for inverting the Laplace transform have been
proposed by Schapery (1967, 1974).
The work of Laws and McLaughlin (1978) on viscoelastic
composite materials adopted a self-consistent approximation. They
derived the creep compliance, and numerical calculations were
performed for the limiting cases of composites containing spherical
inclusions and continuous fibers. Eimer (1971) derived formal
effective relaxation moduli expressions of multi-phase media by
considering the many point correlation functions.
Chou and Nomura (1980) and Nomura and Chou (1985) obtained
the effective relaxation moduli of short-fiber composites based upon
their work on effective elastic properties. Explicit expressions of
composite relaxation moduli are given in terms of the elastic and
viscoelastic properties of the constituent phases, fiber volume
fraction, and fiber aspect ratio. Numerical calculations for a typical
glass/epoxy composite system based upon the collocation ap-
proximation method as well as the self-consistent model have
been performed by Nomura and Chou. It is assumed that the fiber
is elastic while the matrix phase is viscoelastic. Figure 4.19 shows
the time dependence of the effective axial Young's modulus of
relaxation (normalized by the initial value of the matrix Young's
modulus) for the fiber volume fraction of Vf = 0.2 and fiber aspect
ratios IId = 5 and oo. The matrix behavior is shown by the
lowermost curve in Fig. 4.19. The effective axial Young's modulus
of relaxation at each fiber aspect ratio is calculated from the
effective relaxation moduli (upper curve), the self-consistent model
 Strength 201

(middle curve), and the effective creep compliances (lower curve).

The self-consistent approximation always lies in between the
predictions of the two other approaches. The results also indicate
that the increase in fiber length or aspect ratio makes the effective
axial Young's modulus of relaxation less sensitive to the time effect.
The fiber length effect also has been examined by Nomura and
Chou for other effective moduli, i.e. the transverse Young's
modulus of relaxation and the shear relaxation modulus, and they
found no such sensitivity for these effective relaxation moduli, as in
the elastic case.

4.6 Strength
Unlike continuous-fiber composites the mechanical be-
havior of short-fiber composites is often dominated by complex
stress distributions due to fiber discontinuities. In particular, the
local stress concentration at fiber ends plays a critical role in
affecting the performance of short-fiber composites, and it often
reduces the strength of a short-fiber composite to a level far less
than that of a continuous-fiber composite with the same fiber
volume content. Several theories (see Vinson and Chou 1975) have
been proposed to predict the strength of discontinuous-fiber corn-
Fig. 4.19. Time dependence of effective axial Young's modulus £ L /£ m for
I/d = 5 and °° and Vt = 0.2. The viscoelastic material properties are
£ m (0 = £m(0) = 3.2GPa, Em(oo) = 0.04 GPa, vm(0) = 0.365, vm(») =
0.485, Es = 71.5 GPa and v, = 0.2. t denotes time. For each l/d value, the
upper, middle and lower curves are obtained from the effective relaxation
moduli, self-consistent model and effective creep compliances, respec-
tively. (After Nomura and Chou 1985.)

6 r

hj

tq

12
log (time (s))
202 Short-fiber composites

posites. One type of theory is based on a modification of the

'rule-of-mixtures', which was originally developed for continuous-
fiber composites. Since the axial stress distribution in a short fiber is
not uniform, the rule-of-mixtures has been modified by researchers
to take into account the effect of fiber length.
Among short-fiber composites, aligned-fiber composites have
many attractive properties (see Edward and Evans 1980; Richter
1980; Manders and Chou 1982). When complicated shapes and
double curvatures are fabricated by matched-die molding tech-
niques, aligned short-fiber composites have an advantage over their
equivalent continuous mats (Kacir and Narkis 1975). The ability of
aligned-fiber composites to elongate both parallel and perpendicular
to the fiber direction without splitting complements the pre-
dominant shear deformation of woven materials. Because of their
useful properties, highly aligned short-fiber composites have been
commercially produced by the centrifuge (Edward and Evans 1980)
and hydrodynamic alignment (Richter 1980) processes.
In the following, the strength of short-fiber composites is dis-
cussed first for the case of aligned fibers. Then, the effect of fiber
orientation is considered for partially aligned and random fiber
arrangements.

4.6.1 Unidirectionally aligned short-fiber composites

To examine the strength of short-fiber composites it is
necessary to recall the original strength predictions developed by
Kelly and co-workers (see Kelly and Davies 1965; Kelly and Tyson
1965a&b; Kelly 1971; Hale and Kelly 1972) for continuous-fiber
composites. The ultimate axial tensile strength expression of Kelly
et al. is (see Section 3.2)

ocu=ofuVf+o'mu(l-Vf) (4.42)

where acu and crfu are the ultimate tensile strengths of the composite
and the fiber, respectively. afu is identical with the fracture strength
of brittle fibers. o'mxx denotes the stress in the matrix at the failure
strain of the composite.
Equation (4.42) was derived based upon the assumptions that the
tensile strain in the composite is uniform along the axial direction
and the applied load is distributed among the fibers and the matrix.
When fibers are discontinuous, the iso-strain condition of Eq. (4.42)
is no longer valid. The difference of the strains in the fiber and
matrix near a fiber end induces shear stresses along the fiber axis.
 Strength 203

The shear forces acting near both ends of a fiber stress the fiber in
tension or compression. It is through this transferring of stress that
applied load can be dispersed among the short fibers.

4.6.1.1 Fiber length considerations

Figure 4.20 shows schematically the variation of fiber axial
tensile stress with fiber length. The profile of linear stress variation
from fiber ends originates from the assumption of constant interfa-
cial shear stress. The fiber critical length /c is defined as the
minimum fiber length necessary to build up the axial stress to afu.
The ultimate strength of a short fiber can be realized if its length
reaches /c.
Kelly and Tyson (1965a) proposed a linear transfer of stress from
the tip of a fiber to a maximum value when the strain in thefiberis
equal to that in the matrix. By assuming constant interfacial stress
r, the fiber critical length can be easily derived by considering the
balance of tensile and shear stresses:

r is the shear strength of either the matrix or the interface,

whichever is smaller. Experimental measurement techniques for /c
have been discussed by Vinson and Chou (1975).
Using the concept of critical fiber length and replacing afu in Eq.
(4.42) by the average fiber stress af, Kelly (1973) derived the
following expression of composite strength for / > lc:

= M l " (1 " 6)Ul] + <Jmu(l - Vf) (4.44)

where 6 is defined as the ratio of the area under the stress
distribution curve over the length IJ2 in Fig. 4.20 to the area of

Fig. 4.20. Variations of fiber tensile stress with fiber length.

Of

Ofu

A\\\\\\l K\\\\\\\\\\\\\1 i\\\\\\\\

K l c l = lc
204 Short-fiber composites

ofJJ2. For constant interfacial shear strength, S = \ and

<TCU = M l - IJ21)V{ + aL(l -Vf) I > /c
(4.45)

Equations (4.45) predict that for short fibers with ///c = 10, a{
reaches 95% of the value for continuous fibers. Equations (4.45)
have been shown to be a good approximation for metallic (Kelly
and Tyson 1965a&b; Kelly 1973) and polymer matrices (Kelly 1973;
Riley and Reddaway 1968; Hancock and Cuthbertson 1970). It
should be noted that Eqs. (4.45) do not consider fiber end stress
concentration which occurs in short-fiber composites. There exist
several variants of Kelly's formulation of short-fiber composite
strength. For example, Outwater (1956) has taken into considera-
tion the effect of interfacial friction load due to resin cure
shrinkage. However, there lies the difficulty of measuring the
friction coefficient and radial shrinkage pressure (Kardos 1973).
For pure elastic deformation of the fiber, afu = Efecu where £cu is
the composite ultimate strain. Equation (4.43) can be rewritten as

(4.46)
a 2T
For composites with variation in fiber length, Bowyer and Bader
(1972) pointed out that at any value of composite strain ec there is a
critical fiber length given by

Fibers shorter than 4 will carry the average stress

of = lj (4.48)
which is always less than |£f£c- Fibers longer than le carry the
average stress

which is always greater than \Efec.

Following Bowyer and Bader, for a composite containing a
spectrum of fibers of different lengths, its strength can be estimated
by dividing the length of fibers into sub-fractions at a given
 Strength 205

composite strain level (Lees 1968). Sub-critical fractions are de-

noted by /, and their respective volume fractions Vt while super-
critical fractions are denoted by /y and VJ. Thus the composite stress
can be expressed as

Oc = f Tl¥i + f Etec(l - ^f)V, + Emsc(l - Vt) (4.50)

Equation (4.47) indicates that at low composite strain le is small and
all fibers will contribute to the reinforcement as given by Eq. (4.49).
As the strain is increased, a progressively smaller proportion of the
fibers will reinforce according to Eq. (4.49) and an increasing
proportion will follow Eq. (4.48). Thus, the load-extension curve
for such a material as indicated by Eq. (4.50) is expected to show
smaller slope as the strain is increased. The work of Bowyer and
Bader on short-fiber-reinforced thermoplastics has further shown
that improvements in the fiber-matrix bond strength have led to
small improvements in strength. Also the fibers which are too short
to be strained coherently with the matrix tend to fail at very low
strains preventing the potential of the longer fibers from being
realized. Thus the very short fibers should be eliminated if full
strengthening potential is to be achieved.

4.6.1.2 Probabilistic strength theory

The following discussions of the probabilistic strength
theory of short-fiber composites begin with a consideration of fiber
length variations and their effect on fiber axial stress distribution.
Then, the influence of local stress concentrations due to fiber-fiber
interaction is introduced. A probabilistic strength theory is de-
veloped to consider the maximum stress concentration induced by
the clustering of ends of short fibers.

(A) Modification of the rule-of-mixtures

Consider a unidirectional short-fiber composite material
with fibers of uniform length and strength. The mechanisms of
failure can be categorized according to fiber length (Fig. 4.21).
When fibers are very short, a crack formed at a fiber end can
circumvent the neighboring fibers without breaking them (Fig.
4.21a). Final failure of the composite is then attributed to fiber
pull-out. On the other hand, if fibers are sufficiently long, fiber end
cracks will cause fracture of the neighboring fibers and, hence,
failure of the composite (Fig. 4.21b). The strength model of Fukada
206 Short-fiber composites

Fig. 4.21. Two failure modes in short-fiber composites. (After Fukuda and
Chou 1981b.)
Failure surface

(b)

Fig. 4.22. Stress distribution in a short fiber. (After Fukuda and Chou
1981b).
 Strength 207

and Chou (1981a&b), and Hikami and Chou (1984a&b) aim at the
latter case.
The composite ultimate strength acu is defined as the stress level
which causes first fiber fracture. Consequently, the maximum stress
in a fiber is of primary importance in predicting composite strength.
Figure 4.22 shows schematically stress distributions in a short fiber.
Here crmax and oo are, respectively, the maximum and plateau stress
of the profile. The average fiber stress at failure is given by

o(x)dx (4.51)

In the case the composite has a distribution of fiber length, Eq.

(4.51) should be replaced by

ot=\ f(l)\)\'a(x)dx}dl (4.52)

Jo '-I JO >
where / ( / ) is a probability density function of fiber length and has

r
the following characteristics:

(4.53)
Jo

f
Jo
f (l)l dl = J (4.54)

/ i n Eq. (4.54) denotes the average fiber length. Then af of Eq.

(4.52) should be used in the rule-of-mixtures expression of Eq.
(4.44). The values of af and oo are not the same. However, the
difference diminishes as the fiber length increases. For relatively
large fiber aspect ratios it is reasonable to assume of ~ oo.
Furthermore, by defining the stress concentration factor K in the
following expression:
tfmax = Ofu = K(JO (4.55)

Eq. (4.44) can be written as

°Cu=Yvf+°™(1-vJ (4-56)
A.

(B) Critical zone model

A systematic experimental study of short-fiber composite
strength has been performed by Curtis, Bader and Bailey (1978)
using polyamide thermoplastic reinforced with short glass and
208 Short-fiber composites

carbon fibers. Their experimental findings led Bader, Chou and

Quigley (1979) to propose a damage model. The basic concepts are
that microcracks are most likely to develop at fiber ends at
microscopic strains well below the fiber failure strain, and that
failure is finally initiated in a critical cross-section that has been
weakened by the accumulation of cracks.
Figure 4.23 depicts a typical volume element in a short-fiber
composite used by Bader, Chou and Quigley. The width of a
'critical zone' in the strength model is denoted by pi where
0 < P < 1 is a constant parameter and / is the average fiber length.
The critical zone width is assumed to be of the same order as the
fiber ineffective length (Sections 3.4.6.1 and 4.2.2).
A discontinuous fiber can end in the zone (ending fiber), in which
case it bears no load, or it can bridge the zone (bridging fiber) and
contribute to the strength of the critical zone. The probabilities of
finding an ending fiber and a bridging fiber are /3 and 1 - /3,
respectively. All fibers are assumed to have uniform strength afu.
Within each transverse section of the composite, ending fibers and
bridging fibers are distributed randomly. A typical fiber configura-
tion on a transverse section in a two-dimensional fiber array is
shown in Fig. 4.24. The ending fibers and bridging fibers are
depicted, respectively, by solid circles and open circles. Under the
applied stress, the stress in the bridging fibers is enhanced by the

Fig. 4.23. A typical critical zone in a short-fiber composite. (After Bader,

Chou and Quigley 1979).

Fiber end in zone

Fiber bridging zone

 Strength 209

stress transferred from the neighboring ending fibers. For example,

the stress in the bridging fiber no. 8 in this figure is enhanced by the
ending fibers nos. 1, 5, 6, 7, 9, 12 and 13. In other words, it is
enhanced by the neighboring fiber-end-gaps A, B, C and D.
The strength of the composite is determined by the relative
numbers of fibers that bridge the zone vs. those with ends within the
zone. These latter will develop matrix cracks when the strain
exceeds a critical value. The critical situation arises when the
bridging fibers are unable to sustain the load transfer due to matrix
cracking and failure occurs. The critical stress and strain values for a
wide range of fiber aspect ratio, fiber critical length, fiber-matrix
interfacial strength and critical zone width have been evaluated by
Bader, Chou and Quigley.

(C) Stress concentration

The stress concentration factor for the unidirectional fiber
arrangement of Fig. 4.25 is difficult to evaluate in a precise manner.
The following assumptions are adopted to facilitate the calculation
of K: (a) fibers are of the same length, /; (b) they are arranged in
rows along the axial direction; (c) the spacing between two
neighboring rows is uniform (Fig. 4.25a); and (d) fibers with ends in
the critical zone of width /?/ are assumed to have the ends aligned
along the cross-section zz' (Fig. 4.25b). This collection of fiber ends
is termed a 'fiber-end-gap' in a two-dimensional array. It is assumed
that the fiber length / is much larger than the critical length lc and,
hence, results for stress concentrations due to the fracture of long
fibers can be used. Also, in Fig. 4.25(a), the number 1 and number
4 fibers are labeled as 'bridging fibers' and the number 2 and
number 3 fibers as 'ending fibers'.
Since the stress concentration factor, K, cannot be readily
calculated by considering the enhancement effect from all the
fiber-end-gaps, assumptions need to be introduced for the load
sharing rule. Hikami and Chou (1984a) have examined the first and

Fig. 4.24. Schematic cross-sectional view of fiber configuration. Solid

circles depict ending fibers and open circles indicate bridging fibers. A
group of adjacent ending fibers is termed afiber-end-gap(i.e. A, B, C and
D). (After Hikami and Chou 1984a.)
A B C D

10 11 12 13
210 Short-fiber composites

simplest approximation for K by only considering the stress

enhancement effects of the first nearest neighboring fiber-end-gap of
a bridging fiber. This is known as the weak local load sharing rule
and the assumption is allowable if the probability of finding the
ending fibers is relatively small. Using the shear-lag method, the
stress concentration factor due to the presence of n, and n r ending
fibers (Fig. 4.26) has been obtained by Hikami and Chou (1984a
and b, 1990).
It can be shown that the failure of the {nt + l)th fiber does not
cause the composite failure since the stress concentration factor for
the (n, + l)th fiber is larger than that for the zeroth bridging fiber
after the failure of the («/ + l)th bridging fiber. Clearly, the failure
of the zeroth bridging fiber causes the total failure of the composite.
Thus neglecting the load bearing capacity of the matrix, the strength
of the composite is given by

am = oJKb (4.57)

The explicit expression of elastic stress concentration factor Kh due

to b broken fibers is given in Section 3.3.1.2.

Fig. 4.25. (a) Critical zone in a two-dimensional fiber array, (b) A

fiber-end-gap. (After Fukuda and Chou 1981b.)
| 1 1
A i i I i
1
1
'
1 ll ll 1

1 1
•> 1 i i i I
1 1 '
1 1 i i 1 •^ 1
i r
i i
1 (3/ h
(a)

I 1

II II l

(b)
 Strength 111

The explicit expression of stress concentration factor for compos-

ites with plastically deformed matrices (Fig. 4.26) has also been
obtained by Hikami and Chou (1984a). For the small-scale plastic
deformation case, the plastic stress concentration factor, Kb, can be
expressed in series expansion form in terms of the dimensionless
plastic deformation zone length a.
In the large-scale plastic deformation case, Kh at the tip of a
fiber-end-gap can be approximated by

(4.58)
n \a.
where
To=rm\/(hEt/GmA() (4.59)

Fig. 4.26. Model of stress concentration calculations for a fiber-end-gap in

short-fiber composites with matrix plastic deformation zone at the tip of
the gap. (After Hikami and Chou 1984.)

n = 0 1 2 • • • ni,|+ 1

t t t t

I
I S, Plastic zone

Elastic deformation
Bridging fiber zone in matrix
Ending fiber
212 Short-fiber composites

Also, a a = applied stress, b — number of fibers in the gap, y' =

Euler's constant (—0.577), r m = matrix shear strength, Gm = matrix
shear modulus, E{ = fiber axial Young's modulus, h = fiber spacing,
and A{ = fiber cross-sectional area. The fibers are of unit thickness.

(D) Probability distribution of maximum fiber-end-gap

The fiber-end-gap size has been analyzed by Hikami and
Chou (1984a) for the case of the two-dimensional array shown in
Fig. 4.25(b). Focussing attention on a single fiber end, the
probability, Pn) that this fiber end is in the gap consisting of n fiber
ends is
Pn = nf}n-\l-l3)2 (4.60)
and

S Pn = 1 (4.61)
n= l

The probability that a given fiber end is not in any one of the gaps
with more than n fiber ends is

G» = l ~ S Pi (4.62)
l' = M + I

When the above probability is independent for each fiber, the

probability that there is no gap larger than size n is
P(n) = (Qn)N (4.63)
where N is the total number of fibers in the composite. However,
actually Qn for a given fiber is not independent of the other fibers.
When N is sufficiently larger than the average gap size, h, it is more
suitable to express P(n) of Eq. (4.63) as
P(n) = (Qn)N/* (4.64)

where

n=2nPn (4.65)

Using Eqs. (4.60) and (4.62), Eq. (4.64) can be rewritten as

P(n) = { l - ) 6 > ( l - / } ) + l]}A"'' (4.66)
and
 Strength 213

P(n) can be used to determine the strength of short-fiber compos-

ites through the relation between gap size, n, and the corresponding
stress concentration. Figure 4.27 demonstrates the variation of P(n)
with N and /}. It can be shown that P(n) behaves like a step
function and P(n) changes from 0 to 1 at n = M, where M is
determined from

(4.68)

M obtained from Eq. (4.68) is termed the 'most probable maximum

gap size'. Figure 4.28 shows M as a function of /3 and N. For actual
composites, the values of M do not vary tremendously with /3 and
N. When N is sufficiently large, using the formula 1 — x =exp(—x),
P(n) can be approximated as

P(n) SB exp[-N/3nn(l - /5)2] (4.69)

(E) Strength predictions

Based upon the considerations of fiber-end-gap size and
stress concentrations, Hikami and Chou (1984a) have proposed a
modification of the rule-of-mixtures for composite strength. The
composite ultimate strength acu is defined as the stress level at
which fracture of the composite occurs. Based upon the approxima-

Fig. 4.27. Cumulative probability distribution functions for the maximum

fiber-end-gap size. O: N = 106, /3 = 0.2; • : N = 106, P = 0.1; A: N = 108,
/S = 0.2. (After Hikami and Chou 1984a.)
214 Short-fiber composites

tions discussed above, ocu is given as

(4.70)

Here, o'mu is the matrix stress at the ultimate tensile strain of the
fiber. aa is the applied fiber stress at the instant when the fiber stress
at the site of stress concentration reaches ofu. Thus, aa satisfies the
following relation:

afu = K[aa-rjomy(l-Vf)/V{] (4.71)

for the weak local load sharing rule, where K = Kbor Kb. omy is the
matrix yield strength. The parameter 77 in Eq. (4.71) reflects the
loading condition of the matrix in the fiber-end-gap. If the matrix is
brittle, a crack can propagate in the matrix along the fiber-end-gap
prior to the failure of the intact bridging fiber. In this case, the
matrix in the fiber-end-gap will bear no load and rj is taken to be
zero. However, in a ductile matrix composite the matrix in the
fiber-end-gap can deform plastically to the yield strength, omy. Then
each fiber in the fiber-end-gap sustains the stress amy(l — Vf)/Vf,
thus reducing the applied stress aa, and 77 = 1. Since the fracture of
a composite initiates at the weakest point, the stress concentration
factor for the most probable maximum gap size M of Eq. (4.68)
should be used.

Fig. 4.28. Most probable maximum gap size, M, vs. critical zone para-
meter, p. N denotes the total number of fibers. (After Hikami and Chou
1984a.)

0.01
 Strength 215

In the case of three-dimensional fiber arrays, the problem is more

complicated and there is no rigorous probabilistic treatment avail-
able. The shape of the fiber-end-gap cannot be uniquely defined for
a given number of fiber ends and it is fairly involved to obtain the
highest stress concentration factor in the intact bridging fibers.
Furthermore, the fiber failure process here is more complex than
that in the two-dimensional case. To circumvent these difficulties,
Fukuda and Chou (1981b) took only compact fiber-end-gaps as the
first approximation. Following this approximation, Hikami and
Chou (1984a) have considered the special type of fiber-end-gap
which consists of square-arrayed ending fibers. A typical example of
such a fiber-end-gap is shown in Fig. 4.29, where ending fibers are
indicated by solid circles and bridging fibers by open circles in the
two-dimensional square lattice. Approximations for the most prob-
able maximum gap size and the resulting composite strength have
been obtained and the details can be found in the reference.
The relation between the fiber volume fraction, Vf, and composite
strength normalized by the matrix stress at failure, ocja'mu, is
shown in Fig. 4.30 for the case of an elastic matrix. The properties
of a glass fiber/thermoplastic matrix composite are used; fiber length
(/) = 1 mm; fiber diameter (d) = 0.01 mm; fiber critical length
(/c) = 0.1 mm; and critical zone parameter (j8) = 0.1. Also
oJo'mu = Ef/Em = 35.2.
In Fig. 4.30, line A shows the simple rule-of-mixtures for
continuous fibers, while line B depicts the rule-of-mixtures modified
for short fibers. Neither case takes the effect of local stress

Fig. 4.29. Schematic cross-sectional view of a three-dimensional fiber

array. Solid circles indicate ending fibers and open circles are for bridging
fibers. (After Hikami and Chou 1984a.)

Square-packed fiber-end-gap

ooo\o*ooooo

OOfOOQOOOO
Ending fiber Bridging fiber
216 Short-fiber composites

concentrations into consideration. Lines C and E indicate the

results of Hikami and Chou (1984a) for a three-dimensional fiber
array and a two-dimensional fiber array, respectively, based on the
local load sharing rule. The composite strength is expected to lie
between these two bounds, which are far less than the values
obtained from the rule-of-mixtures because of local stress
concentrations.

4.6.2 Partially oriented short-fiber composites

Cox (1952) first proposed the idea of orientation factor in
the strength equation for continuous fiber composites to account for
fiber misalignment. Bowyer and Bader (1972) adopted this concept
in their study of short-fiber systems, and Eq. (4.50) was modified by
multiplying the fiber dependent terms on the right-hand side of this
equation by the orientation factor Co, Co = 1 for perfectly aligned
fibers and Co assumes values less than unity for partially oriented
fibers. Bowyer and Bader concluded that the orientation factor is
independent of strain and is the same for all fiber length at least at
small strains. The orientation factor can then be calculated from
Eq. (4.50) based upon the knowledge of fiber length distribution,
interfacial bond strength and composite ultimate tensile strength.
Curtis, Bader and Bailey (1978) investigated the strength of a

Fig. 4.30. Strength of the composite as a function of V{. A: rule-of-

mixtures; B: Kelly and Tyson (1965b); C: three-dimensional fiber array,
weak local load snaring; E: two-dimensional fiber array, weak local load
sharing. (After Hikami and Chou 1984a.)
 Strength 217

poly amide thermoplastic reinforced with glass and carbon fibers,

and calculated the fiber orientation factor from the measured
composite modulus and the knowledge of the fiber and matrix
properties. Their results indicate that fiber alignment increases with
increasing fiber volume fraction, which agrees with the qualitative
assessment of optical micrographs.
In general, when there are variations in both fiber length and
orientation, the rule-of-mixtures (Eq. (4.42)) can be modified as
°cu = oiuVfF(lc/l)Co (4.72)
Here, the factor F{ljl) is a function of fiber average length / and
critical length /c. Equations (4.45), for instance, give the forms of
F(ljl) for aligned short fibers of uniform length. If the necessary
information with respect to fiber orientation is known, Co can be
estimated analytically.
Fukuda and Chou (1982) have used a probabilistic theory to
predict the strength of short-fiber composites with variable fiber
length and orientation. They introduced two kinds of probability
density functions to describe the fiber length and orientation
distributions and neglected the effect of stress concentration in this
particular treatment. The analytical result of composite strength is
given only in the form of an average value. The theory of Fukuda
and Chou is introduced below in three parts.
(A) Geometrical consideration of a single short fiber
First, the geometrical arrangement of a single short fiber is
described. Figure 4.31(a) shows an obliquely positioned short fiber
Fig. 4.31. Several notations on short-fiber arrangement, (a) Obliquely
oriented fiber, (b) Bridging fiber and ending fiber, (c) Critical angle.
(After Fukuda and Chou 1982.)
Bridging
A
fiber

(a) (b) (c)

218 Short-fiber composites

of length /. In accordance with the terminology of Section 4.6.1, a

bridging fiber and an ending fiber are defined in Fig. 4.31(b); that
is, if a fiber crosses a critical zone (Section 4.6.1.2) of width 01, it is
termed a bridging fiber; and if the end of a fiber is within the critical
zone, it is defined as an ending fiber. Here, / denotes average fiber
length. The probability density function of fiber length distribution
h(l) satisfies the following condition:

f A(/)d/ = l

Then, the average fiber length is defined as

(4.73)

7 = f lh(l)dl (4.74)
Jo
From Fig. 4.31(a),
lz = lcos6 (4.75)
and from Fig. 4.31(c) the critical angle 6O within which a fiber of
length / is a bridging fiber becomes
(4.76)
for pi< /. If pi>l, 6O cannot be defined, and a fiber in such a case
is inevitably an ending fiber. If the fibers are distributed randomly
with respect to the z axis, the probability pe that a fiber of length / is
an ending fiber in the critical zone becomes
_ pi _ r pin cos e (o < e < # o and #7 < /)
P
l
and the probability ph for finding a bridging fiber is, by definition,
ph=l-pe (4.78)
The probability density function with respect to fiber orientation
(g(6)) should satisfy the condition

f
Jo
Jl
6=1 (4.79)

(B) Load transfer in a short fiber

First, consider a short fiber situated parallel to the applied
tensile stress, ao, along the z axis. The average fiber stress is

dz (4.80)
/ Jo
 Strength 219

The fiber axial stress of(z) has, in general, the profile shown in Fig.
4.1. Consider the simplest form of of(z) by assuming a constant
interfacial shear stress (Fig. 4.20). Then afo becomes

(4.81)
(KQ

The average force in a fiber of cross-sectional area A{ is oioAf.

Next, consider a single short fiber situated at an angle 6 to
the applied stress ao. The applied stress can be decomposed
into an axial and a shear component, with respect to the fiber axis,
as
o'o = ao cos 2 d (4.82)

TO = ao sin 0 cos 6 (4.83)

If the effect of ro on the fiber stress distribution can be neglected,
the average force of the fiber becomes Afafocos2 6 and the z
direction force component is

Fz=AfGfocos36 (4.84)

(C) Strength of short-fiber composites

Based upon the above preparations, the strength of short-
fiber composites can be derived. In the following discussion, h(l)
and g(6) are assumed to be independent of each other. This means
that g(6) is the same for all the samples with different fiber length
distributions. A rectangular-shaped specimen with the lengths of the
three mutually perpendicular edges denoted by a, b and c is
considered. The c axis is so chosen as to be parallel to the z axis.
The volume of the specimen is

V = abc (4.85)
and from the definition of fiber volume fraction, Vf becomes
Vf = NAf1/V (4.86)
where N and Af denote, respectively, the total number of fibers and
fiber cross-sectional area.
Recall that Eq. (4.76) gives the length of the projection of a fiber
on the z axis. Then the average length of the projection of fibers
220 Short-fiber composites

can be written as

J
rJt/2 /•«>
I cosO h(l)g(O) dl dd
o Jo

J
rJt/2
g(d)cosdd0 (4.87)
o
The value of Nlz gives the total length of projection of all fibers on
the z axis and if this value is divided by the specimen length c, the
average number of fibers which cross an arbitrary section in the
specimen normal to the z axis is obtained. That is,

JVc=— = ? r g(6)cos6dO (4.88)

c Af Jo
Equation (4.77) gives the probability of a specific fiber being an
ending fiber. Therefore, the average probability of finding an
arbitrary fiber being an ending fiber is
rn!2 p
qe= f\pM)g{O) dl dO (4.89)
Jo Jo
Similarly, the average probability of finding an arbitrary fiber being
a bridging fiber is

?b=f f Pbh{i)g(e)didd
Jo Jo
= l ~ qe (4.90)
Substituting Eqs. (4.77) and (4.78) into Eqs. (4.89) and (4.90),

qe= T de(f g(0)h(l)dl + I -^-

Jo \Jb Jpjl cos 0

+f j g(6)h(l)dld6 (4.91)

Then, the total numbers of ending and bridging fibers in the

specimen are
Ne = Neqe (4.93)
Wh = Ncqb (4.94)
 Strength 221

Strictly speaking, the value of Ne is not precise because only one

cross-section, for example AA' in Fig. 4.31(b), has been examined.
The fibers denoted by 2 and 3 in Fig. 4.31(b) are not considered.
However, the objective is to calculate the number of bridging fibers,
which is not affected by Ne in the subsequent discussions.
Based upon Eq. (4.84) for the z direction component of the
axial load of one specific fiber, the average value among the
bridging fibers is

(4.95)
Jo Jfi
Then the total load that all of the bridging fibers can sustain in the
zone fi is
FT = Nb-Fz (4.96)
and the composite strength becomes

tfcu = ^ + < C ( l - V f ) (4.97)

ab
where the matrix is assumed to sustain part of the applied load.
Substituting Eqs. (4.81), (4.84), (4.88) and (4.91)-(4.96) into Eq.
(4.97), the composite ultimate strength is determined as

J
rjr/2 r6a

g(6) cos 6 dB\ g(6) cos3 6 d6

o Jo

f -v<) a U l

Equation (4.98) is a general strength expression of short-fiber

<498)

composites. In order to conduct further analysis, it is necessary to

know the functions g(6) and h(l) together with afu, O'MU, V{ and /c.
Some limiting cases of Eq. (4.98) are discussed in the following.
First, consider a unidirectional short-fiber composite with uniform
fiber length /. Equation (4.98) can be reduced to

ae = oM l - j8)(l - 1 ) + aL(l - Vf) (7> lc)

(4.99)
oc = ofuVf(l - 13) — + a: u (l - Vt) {I < lc)
222 Short-fiber composites

Equations (4.99) coincide with the result of the original failure

model of Bader, Chou and Quigley (1979).
Secondly, consider the effect of fiber length distribution on the
strength of a unidirectional composite. By assuming the limiting
case of /?—»0, namely all fibers are bridging, and the following
probability density function of fiber length distribution

(4.100)

Eq. (4.98) is reduced to

l l

(4.101)

where Si(jc) is the integral sine function defined by

(4.102)

The result of F{ljl) from Eq. (4.101) is shown in Fig. 4.32 by the
solid line. In the case of constant fiber length, the strength can be
obtained from Eqs. (4.45) and the value is also shown in Fig. 4.32
by a broken line. It can be concluded from Fig. 4.32 that the
strength of a composite material is reduced if the fiber length is not

Fig. 4.32. F(ljl) vs. IJl fiber length distribution considered;

fiber length assumed to be constant. (After Fukuda and Chou
1982.)

I
1
IJl
 Strength 223

uniform. However, the difference in composite strength between

the non-uniform fiber length system (Eq. (4.100)) and the uniform
fiber length system is not very significant and, hence, the ordinary
theory based upon an average fiber length may be used as a first
approximation.
As a third example, the case of uniform fiber length and biassed
fiber orientation distribution is considered. The following two types
of fiber orientation are examined.

(a) g(6) = Hoc for 0 < 0 < a and g(6) = 0 for 0 > or;
(b) g(6) = (ji/2a) cos(jr0/2a) for 0 < 0 < a and g(6) = 0 for
6>a.

These functions are taken so as to satisfy Eq. (4.79). The shapes of

these functions are shown schematically in Fig. 4.33 and 6 is defined
in the three-dimensional view of Fig. 4.12. Note that g(6) does not
mean the probability per unit area. The probability per unit area is
proportional to g(6)/sin 6. The limit of )8 —> 0 is again considered.
At this limit, 60 tends to JZ/2 from Eq. (4.76). Considering this
condition, Co is calculated from Eq. (4.98) for the two types of g(6)

Fig. 4.33. Values of Co for two types of fiber orientation distribution.

(After Fukuda and Chou 1982.)

15 30 45 60 75 90
a (degrees)
224 Short-fiber composites

given above:
/ x ,.„ sin
sinaa 11// 11 . . 3 .. \
(a) umC o = sin3a++-sin
— sin3ar -sina)
or
4 /
p^o a a \12 4 / (4.103)
(b) limCo = -^ [ - J - sin ^ (1 + q) + —*— sin ^ (1 -
/3^o 16 L1 + g 2 1~<7 2

where q = 2a/jt. These values are shown in Fig. 4.33. Bowyer and
Bader (1972) estimated the value of Co by their experimental data.
For laboratory glass/nylon injection molded materials, C o was 0.66.
If a retangular distribution for g(6) is used, the value of a
corresponding to Co = 0.66 is approximately 45° from Fig. 4.33.
The orientation factor Co discussed here is slightly different from
the factor Ce discussed in Section 4.3.2. The bridging effect of fibers
is considered in the derivation of C o , while the Poisson's effect of
the composite is taken into account in evaluating Ce. Co and Ce are
essentially the same for the limiting case of /3 —» 0. The effect of /3 is
discussed in Section 4.6.3.

4.6.3 Random short-fiber composites

Both Lees (1968a&b) and Chen (1971) attempted an
averaging technique to treat the strength of random fiber compos-
ites. They adopted the failure mechanisms of Stowell and Liu (1961)
and Jackson and Cratchley (1966), namely fiber failure, matrix
failure in shear and matrix failure in plane strain. The operative
failure mechanism in composites is dictated by the angle between
the fiber direction and the direction of applied stress

{ ax = o'Jcos2 6

o2 = rjsin 6 cos 6
(0<0<dx)

(0X < 0 < 02)

2 ( ]
o3=ojsm 8 (0 2 <0<?r/2) '
where o'c denotes the strength along the fiber direction of the
unidirectional composite given by a rule-of-mixtures type of re-
lationship. r m and am are, respectively, the shear and tensile failure
stresses of the matrix and the interface. Local stress perturbation
due to fiber-fiber interaction can also be included in o^ of Eq.
 Strength 225

(4.104). The strength for random fiber composites can be obtained

by considering the angular strength dependence as a piecewise
continuous function integrated over 90°:

2 r r01 f°2 ra i
<xc = - 01d0+ <72d0+ cr3d0 (4.105)
a Uo Je, Je2 J
The predictions of this approach agree reasonably well with
experimental results on glass-reinforced polyethylene and PMMA
random mat (Lees 1968a) as well as random Al 2 O 3 -aluminum-
silicon and glass/epoxy composites (Chen 1971).
Treatments of the strength of random short-fiber composites can
also be found in the works of Lee (1969), Lavengood (1972),
Kardos (1973), McNally (1977) and Blumentritt, Vu and Cooper
(1975). The method of laminate analogue discussed for stiffness
(Section 4.3.3) can also be applied to prediction of the strength of
two-dimensional random fiber composites; the strength behavior of
an isotropic laminate can be simulated by unidirectionally oriented
plies laid up to approximate random orientation.
The strength prediction method of Fukuda and Chou (1982) can
also be applied to determine the orientation factor Co (Eq. (4.72))
for random fiber composites. By assuming that the fiber length is
uniform and is larger than the critical length / c , Eq. (4.98) becomes

cu = otavf(i -£)j g(d) cos e de J °%(0) cos3 e dd

By comparing Eqs. (4.72) and (4.106), the following expression for

Co is obtained:
rJt/2 rdo

Co = g(6) cos 6 d9 g(0)cos 3 6 dd

Jo Jo

g(8)dd (4.107)
Jo \ cos 8
Now consider both two-dimensional and three-dimensional ran-
dom fiber arrays. In a two-dimensional random array model, g(e)
must be constant in the whole region of 0 < e < nil, and

g(8) = 2/n (4.108)

226 Short-fiber composites

from Eq. (4.79). Substituting Eq. (4.108) into Eq. (4.107), the
following result is obtained:

(4.109)
The solid line of Fig. 4.34 depicts this result. As /3 increases, the
composite contains more ending fibers and fewer bridging fibers, and
hence the reinforcing effect of fibers is reduced. In the limit of
/?—>0, all fibers are bridging fibers, and Co tends to 0.27 for this
two-dimensional case. Bowyer and Bader (1972) used the value of 5
by quoting the result of Cox (1952) for the orientation factor of
Young's modulus of a random composite. Cox's value is also shown
in Fig. 4.34 by the solid circle.
In the case of a three-dimensional random fiber model, referring
to Fig. 4.12, g(6) can be expressed as
g(0)d6 =
where the hemispherical surface area is S. Therefore,
g(6) = sin 6 (4.110)

Fig. 4.34. Fiber orientation factor Co of random fiber array model.

two-dimensional random array; three-dimensional random array.
Solid and open circles indicate Cox's results. (After Fukuda and Chou
1982.)

0.4 1-

0.3

Q,

0.1

I I
0.01 0.1 1.0
p
 Fracture behavior 227

In this case, Eq. (4.107) becomes

C0 = i(l-/3 2 )(l + /S 2 )(l-)8 + /31ogj8) (4.111)
This result is shown in Fig. 4.34 by a broken line. In the limit of
)3-»0, Co becomes | and this value is again less than Cox's
prediction of \ as indicated by the open circle.

4.7 Fracture behavior

Among the various types of short-fiber composites, the
fracture behavior of polymer based composites is relatively well
understood. The failure of short-fiber composites often initiates at
micro voids and microcracks. These defects exist in the reinforce-
ments, the matrix, and the interphase material and are introduced
in the fabrication process. The final failure of a short-fiber composite
is the result of several micromechanical mechanisms. The macro-
scopic appearance of the fracture depends on which of these
mechanisms dominate the overall fracture process.
According to Friedrich (1985, 1989) and Friedrich and Karger-
Kocsis (1989), the major failure mechanisms of short-fiber compos-
ites include (a) matrix deformation and fracture, (b) fiber/matrix
debonding, (c) fiber pull-out, and (d) fiber fracture. A schematic
fracture path through a short-fiber-reinforced polymer is given in
Fig. 4.35; the individual failure mechanisms are also demonstrated.
The extent to which a specific failure mechanism occurs depends
on the properties of the fiber, matrix, and interphase as well as the
geometric form and arrangement of the fibers. As discussed in
Sections 4.2.1 and 4.6.1, the efficiency in load transfer between a
fiber and its surrounding matrix depends on the length of the fiber
relative to its critical length, /c. If the length of the fiber is shorter
than /c, fiber pull-out and matrix fracture are the dominating
mechanisms of energy absorption. On the other hand, when the
fiber length is longer than /c, the fibers will, in some cases, break
and in other cases be pulled out; the fiber location and orientation
with respect to the crack is an important factor in determining
which failure mechanism takes place.
Friedrich (1989) has examined the fracture energy of aligned
short-fiber composites and given the following observations:
(1) The matrix material supplies a certain portion of the
fracture energy of the composite. For a brittle polymer
matrix, this portion is small in comparison to fiber fracture
or interfacial failure. Then the fracture energy of the
228 Short-fiber composites

composite as a result of fiber reinforcement is higher than

that of the unfilled matrix. However, in the case of a ductile
polymer matrix, the energy absorption in the fracture
process is higher than those due to fiber related mechan-
isms. Thus, the fracture energy decreases as fiber volume
fraction increases.
(2) The fiber/matrix interface shear strength, which affects the
fiber critical length (Eq. (4.43)), is strongly influenced by
the temperature of the environment. Higher temperatures
result in higher /c. Furthermore, the temperature also
influences the matrix fracture behavior.
(3) The fracture toughness Kc of a short fiber composite is
related to the fracture energy Gc and elastic modulus E by
Kc = V(GCE). Some qualitative observations can be made
concerning this relationship. First, the addition of fibers to
a brittle polymer matrix enhances Kc due to a simultaneous
increase in Gc and E. Second, the addition of fibers to a

Fig. 4.35. Schematic fracture path through a short-fiber-reinforced poly-

mer, and individual mechanisms of failure: (A) fiber fracture, (B) fiber
pull-out, (C) fiber/matrix separation, and (D) plastic deformation and
fracture of the polymer matrix. (After Friedrich 1989.)

Fiber

Crack

10 |/m
 Fracture behavior 229

very ductile thermoplastic matrix results in an increase in E

but a decrease in Gc. Thus, Kc may decrease or remain
unchanged.

It is also noted that the addition of fibers can result in

constraining effects on the matrix and a change of the stress state.
Consequently, this leads to limited plasticity in the matrix and stress
concentrations at fiber ends. The implications of the stress con-
centration on the fracture of short-fiber composites are discussed
below.
Experimental work for identifying fiber end stress concentration
was first performed by MacLaughlin (1966), who used a photoelas-
tic method to investigate the effect of fiber end shape and gap size
on the shear stress near a single short fiber. MacLaughlin (1968)
extended the photoelastic study to a square-ended short fiber
flanked by continuous fibers. Photoelastic methods were also used
by Chen and Lavengood (1969) to examine the distribution of fiber
stress and interfacial shearing stress around a short square-ended
fiber.
Theoretical analyses of fiber end stress concentrations have been
discussed in Section 4.2. Iremonger and Wood (1967, 1969), Muki
and Sternberg (1969, 1970, 1971), Chen and Lewis (1970), Stern-
berg (1970), Sternberg and Muki (1970), Baker and MacLaughlin
(1971) and Takao, Taya and Chou (1981) have also presented
analytical solutions with particular emphasis on fiber end separation
distance, fiber volume fraction, fiber and matrix modulus ratio, and
fiber end geometry. Several general conclusions can be drawn from
the analyses: (1) the primary parameters affecting the stress
concentrations are gap size, fiber volume fraction and fiber-matrix
modulus ratio; (2) square-ended and tapered-end fibers give higher
stress concentrations than round-ended fibers; (3) stress concentra-
tion increases with decreasing fiber end separation distance; (4)
higher stress concentrations exist at the fiber-matrix interface when
the end gap is a void as compared to a gap filled with matrix. It is
understood that in real composites the fiber ends are usually oblique
and uneven and that the concept of fiber end separation distance is
difficult to apply to a randomly distributed and misaligned fiber
system. A significant finding of the stress analyses surveyed above is
that the concentration of stress in the matrix near the discontinuity
of a fiber is very severe even under moderate load application.
Composite failure initiation, either by fracture of the matrix or by
debonding, is likely to occur at these locations.
230 Short-fiber composites

The experimental work of Curtis, Bader and Bailey (1978) on

glass and carbon fiber reinforced poly amide 6.6 has demonstrated
the embrittlement effect of short-fiber composites. Theoretical
modeling of the fracture of short-fiber composites can be found in
the work of Taya and Chou (1981, 1982), Ishikawa, Chou and Taya
(1982), Takao, Chou and Taya (1982) and Takao, Taya and Chou
(1982). The environmental effect on the fracture of short-fiber
composites has been examined by Friedrich, Schulte, Horstenkamp
and Chou (1985), Hsu, Yau and Chou (1986), and Yau and Chou
(1989).
 Hybrid composites

5.1 Introduction
The term 'hybrid composites' is used to describe composites
containing more than one type of fiber materials. Hybrid compos-
ites are attractive structural materials for the following reasons.
First, they provide designers with the new freedom of tailoring
composites and achieving properties that cannot be realized in
binary systems containing one type of fiber dispersed in a matrix.
Second, a more cost-effective utilization of expensive fibers such as
carbon and boron can be obtained by replacing them partially with
less expensive fibers such as glass and aramid. Third, hybrid
composites provide the potential of achieving a balanced pursuit of
stiffness, strength and ductility, as well as bending and membrane
related mechanical properties. Hybrid composites have also dem-
onstrated weight savings, reduced notch sensitivity, improved frac-
ture toughness, longer fatigue life and excellent impact resistance
(Chou and Kelly 1980a). Some of the pioneering studies on this
topic can be found in the work of Wells and Hancox (1971),
Hayashi (1972), Kalnin (1972), Hancox and Wells (1973), Bunsell
and Harris (1974), Harris and Bunsell (1975), Walton and Majum-
dar (1975), Aveston and Sillwood (1976), Bunsell (1976), Harris
and Bradley (1976), Zweben (1977), Arrington and Harris (1978),
Badar and Manders (1978, 1981a,b), Marom, Fischer, Tuler and
Wagner (1978), Rybicki and Kanninen (1978), Summerscales and
Short (1978), Aveston and Kelly (1980), Wagner and Marom
(1982), Fukuda (1983a-c), and Harlow (1983)4.
Depending upon the arrangements of fibers and pre-preg layers,
hybrids can be categorized into the following types. In the first type
the different fiber materials are intimately mixed together and
infiltrated with a matrix simultaneously. The hybrid in this case is
described as intermingled (Aveston and Kelly 1980) or intraply
(Chamis and Lark 1978) (Fig. 5.1a). The second type of hybrid is
made by bonding together separate laminae each containing just
one type of fiber in a matrix, and is known as interlaminated
(Aveston and Kelly 1980) or interply (Chamis and Lark 1978) (Fig.
5.1b). The third category of hybrids consists of fabric reinforce-
ments where each fabric contains more than one type of fiber and it
232 Hybrid composites

can be termed as interwoven (Chou and Kelly 1980a) (Fig. 5.1c).

Hybrid composites consisting of resin composite plies, metal composite
plies and metal foils also have been explained. When a laminated
hybrid is composed of plies of different matrices it needs to be
fabricated by a consolidation procedure that is compatible with all
matrix materials.
Optimization of composite properties can usually be achieved
through a suitable combination of fiber types. Reinforcements for
hybrids include boron, carbon, glass and aramid fibers. Limited
applications of ceramic and metallic filaments have been explored
(Renton 1978). Intermediate modulus epoxies, thermoplastics and
polyimides are the common polymeric matrices for hybrid compos-
ites. Current applications of hybrid composites can be found in
aircraft fuselage, wing and tail structures, helicopter rotor blades
and automobile parts as well as in an array of sports equipment,
ranging from sailboats and racing cars to bicycle frames and hockey
sticks.
The fundamental questions pertinent to the study of hybrid
composites are (a) how is the load shared among the constituent

Fig. 5.1. Types of hybrid composites: (a) intermingled; (b) interlaminated;

and (c) interwoven.

0 OO OO OO OO OO
oo oo oo oo o c
ooo ooo ooo ooo ooo o
o oo on" o o o o ool
(b)

(c)
 Stress concentrations 233

fibers? (b) are there synergistic effects among the different types of
fibers? and (c) will certain combinations of fiber types and micro-
structure designs produce an overall desirable structural performance?
In order to gain a basic understanding of these problems, the
various aspects of the mechanical behavior of hybrids are examined.
To simplify the consideration of deformation, the following discussions
are primarily restricted to unidirectional composites and their
laminates. Woven hybrid composites are examined in Chapter 6.

5.2 Stress concentrations

The load redistribution in unidirectional continuous fiber
hybrid composite laminae due to fiber breakages is examined in this
section. Stress concentration factors are obtained for both inter-
mingled and interlaminated hybrids. The solution techniques are
demonstrated for both static and dynamic responses. The ter-
minologies of low modulus (LM) and high modulus (HM) are used
to distinguish the two kinds of fibers in the model lamina. For
hybrid composites such as glass/carbon and Kevlar/carbon com-
binations, LM and HM fibers correspond to HE (high elongation)
and LE (low elongation) fibers, respectively. The fiber ductility or
elongation to break does not enter into the present analysis in an
explicit manner. The shear-lag technique demonstrated in Chapters
3 and 4 is again adopted in the following.

5.2.1 Static case

Consider a unidirectional lamina composed of HM and
LM fibers in alternating positions. Each pair of neighboring HM
and LM fibers is designated as the group m. Asterisks (*) are used
to denote quantities related to LM fibers.
Fukuda and Chou (1983) have examined the three types of
combinations of fiber discontinuity depicted in Fig. 5.2. Let nx and
n2 be the number of broken HM and LM fibers, respectively. Thus,
in Fig. 5.2, (a) nx = n> n2 = 0; (b) nx = n, n2 = n — l; and (c)
nl = n2 = n. The counting of broken fibers starts at m = 0 and ends
at m = n — 1.
The axial loads of the mth pair of fibers are denoted by pm{x) and
Pm{*)\ the displacements are um{x) and u^{x), and x = 0 denotes
the plane of fiber fracture. Based upon the assumptions of shear-lag
analysis (Hedgepeth 1961; Ji, Hsiao and Chou 1981), the force
234 Hybrid composites

equilibrium equations of the rath HM and LM fibers are,

respectively.

(5.1)
d2«* G
E*d
dx2 h

Here, E and G denote the fiber extensional modulus and the shear
modulus of the matrix, respectively. The lamina is assumed to be
of unit thickness; d and h denote, respectively, fiber width and
spacing.
Under the assumption of linear elastic deformation, the force-
displacement relations become

—,
(5.2)

Fig. 5.2. Arrays of discontinuous fibers: (a) nl = n, n2 = 0; (b) nl = n,

n2 = n — l; (c) nl = n2 = n, nx and n2 are, respectively, the number of
discontinuous HM and LM fibers. (After Fukuda and Chou 1983.)

i
A B C D

\m = n — \
T

>='

HM fiber

(b) (c)
 Stress concentrations 235

The boundary conditions are

Pm(°°)=P

(5.3)
pm(0) = 0, pm(0) = 0 for broken fibers
" m (0) = 0, um(0) = 0 for unbroken fibers
To simplify Eqs. (5.1)-(5.3), the following dimensionless para-
meters are introduced:

n __Pm p* Pm
rm— rm —

E
-f) «4>/(¥) ™
R=E*d*/Ed

Thus, Eqs. (5.1) and (5.2) become

^ - ^ + £/* + f/* _, -2Um = 0

(5.5)
R^-¥?+Um+, + Um-2U*=0

(5.6)
m — t\ ,.

By adopting the concept of influence functions proposed by

Hedgepeth (1961), the dimensionless displacements are expressed
as
236 Hybrid composites

where V, V*, W and W* are the influence functions. Then, from

Eqs. (5.5), the following two sets of equations in terms of the
influence functions are obtained:
(I)

(5.8)
d2V*
p +y + v7 v n

with the boundary conditions of

Vm(0) = l (m=0) Vm(0) = 0 (m*0)
V*(0) = 0 (5.9)
dK,(°°) = 0 W*mH _ Q
d§ d§
(II)
d2W

(5.10)
d2W*
f+W W
with the boundary condition of
Wm(0) = 0
W*(0) = l (m=0) W*(0) = 0 (»i#0) (5.11)

df d§
Since Eqs. (5.8) and (5.10) are identical in form, only the solution
procedure of Eqs. (5.8) is given below. For solving Eqs. (5.8), the
following Fourier series expressions are introduced

V= 2 K»e"™° V*= S V*e-"" e (5.12)

or, inversely,
If- 1 f -
Vm = —\ Veimedd F* = —\ y*e™ed0 (5.13)
 Stress concentrations 237

Then, multiplying Eqs. (5..8) by e "" and summing over all m gives
d2? ^ _

(5.14)
2
d V* U fi• 1 /
r D

where

Also, from Eqs. (5.9),

0, 6) = \ V*(0, 6) = 0
(5.16)

The solutions of Eqs. (5.14) under the boundary conditions of Eqs.

(5.16) are
K = C,e-Al5 + C2e-A25
(5.17)
V* = C3e~k^ + C 4 e" A ^
where
A! = V[a + V ( a 2 - ^ ) ] A2 = V[a - V(a2 - b)]

(5.18)
l~i2
izA 12
r
C
2 - ,2 12
A^ A2 Aj A2
2
_ (2-A )(2-Al) _

Substituting Eqs. (5.17) into Eqs. (5.13) and considering a, b, k lt

b , Cly C2 and C3A as even functions with respect to 6, the
following results are obtained for the influence functions:
i r
Vm=-\ (C]e"A|? + C2e"A2§) cos(m0) d0
n Jo
(5.19)
r C 3 A ( e e ) ^ i
71 Jo 2(1 + cos 6)
238 Hybrid composites

Differentiating Eqs. (5.19) and substituting the result for £ = 0 into

Eqs. (5.7) and the third condition of Eqs. (5.3), the values of Uk(0)
(0 < k < nx - 1) and Ut(0) (0 < k < n2 - 1) can be obtained.
The dimensionless axial loads, Pm and P£ are calculated by
substituting Eqs. (5.7) into Eqs. (5.6). The stress concentration
factor of the mth group of fibers is defined as Pm(0)/Pm(<») or
P*(0)/P*(»).

(5-20)

For an HM or an LM fiber adjacent to a discontinuous fiber, the

stress concentration factor can be calculated by substituting the
corresponding value of m into Eqs. (5.20). For instance, the stress
concentration factors for fibers B and A of Fig. 5.2 are, respectively,
Pl-m/Pl-iH and Pni(0)/Pni(«>).
Fukuda and Chou (1981, 1983) have evaluated Eqs. (5.20) and
the results are presented for (a) comparisons with the solutions of
Hedgepeth (1961) for non-hybrid composites, and (b) hybrid
composites. First, for a non-hybrid composite, there is only one
type of fiber in the lamina and R = l. Two limiting cases are given
below.

(A) nl = l and n2 = 0
Consider, for instance, Fig. 5.2(a). The fiber immediately
adjacent to the broken fiber is the one designated as an LM fiber in
the m = 0 pair. Therefore, the stress concentration factor is
dV* 4
P:(0)/P*o(o°) = l+-rf Uo(0) = - (5.21)
ds -J

This result coincides with that of Hedgepeth (1961), as expected.

The same conclusion can be reached by considering the case of
«j = 0 and n2 = 1 in Fig. 5.2(a).
 Stress concentrations 239

(B) n1 = n2 = 1
The stress concentration factor is given by

4W6
(5.22)
3A5

Next, for the unidirectional hybrid lamina (R i= 1), Eqs. (5.20)

have been solved by numerical integrations using a trapezoidal rule.
The results are shown in Figs. 5.3-5.5.
The limiting case of the fracture of one HM fiber (n1 = l, n2 = 0)
is demonstrated in Fig. 5.3. The fibers adjacent to the broken HM
fiber of particular interest are the LM fiber of m = 0 and the HM
fiber of m = 1. The stress concentration factors of these two fibers
(K^M, ^ H M ) are plotted in Fig. 5.3 as functions of the stiffness ratio
R = E*d*/Ed. For R = 1 (i.e. non-hybrid case), # L M =1.33 is
obtained from Eq. (5.21). The limit of /£—»0, on the other hand,
indicates that the extensional rigidity of the LM fiber is in-
finitesimal. Then Fig. 5.3 again becomes a model with only one type
of fiber. The fiber nearest to the broken fiber in this case is the HM
fiber of m = 1 and therefore KHM^> § at the limit R —»0.
Figure 5.3 also indicates that the stress concentration factor of the

Fig. 5.3. Stress concentration factor vs. stiffness ratio R = E*d*/Ed =

extensional stiffness of LM fiber/extensional stiffness of HM fiber. (After
Fukuda and Chou 1981.)

.2 2 -

0 ' 0.02 0.05 0.1 0.2 0.5

R(= E*d*/Ed)
240 Hybrid composites

HM fiber in the hybrid lamina is lower than 3 because of the

presence of the LM fiber between the discontinuous and continuous
HM fibers. For example, at R = 0.5, A"LM = 1.67 and KHM = 1.11.
Figure 5.4 depicts the relations between stress concentration
factors and the total number of broken fibers, nx + n2, for the case
of R = l. The letters A—F correspond to fibers A—F in Fig.
5.2. Curves D and E show the stress concentration factors of the
fibers immediately adjacent to the broken fibers. Curves C and F
give the results for the second nearest fibers to the broken fibers.
The cases of A and B in Fig. 5.2 give stress concentration factors
insensitive to the number of broken fibers.
Figure 5.5 shows the stress concentration factors for R = 3 which
approximately corresponds to carbon/glass hybrid composites. The
stress concentration factor of the HM fiber nearest to the discon-
tinuous fibers, i.e. fiber C or E in Figs. 5.2(b) and (c), respectively,
is smaller than that of fiber D (Fig. 5.2b) for a fixed number of
broken fibers. This means that, as far as the HM fibers are
concerned, the stress concentration is reduced in a hybrid compos-
ite. Thus, it is possible for the high modulus fibers in a hybrid

Fig. 5.4. Stress concentration factors vs. total number of broken fibers for
R = l. (After Fukuda and Chou 1983.)

R = 1 >

3 —
1
1 1 1 1 1

C,F
2 - /

1 - A

I I
5 10 15 20
Number of discontinuous fibers
 Stress concentrations 241

composite to sustain higher loads than in the all-high modulus fiber

composite, and a 'hybrid effect1 could be realized.
The stress concentration factors of the LM fibers are shown by
curves D and F in Fig. 5.5. A comparison of curves D of Figs. 5.4
and 5.5 indicates that the stress concentration on the LM fiber
increases as R is reduced. This implies that the LM fibers are more
susceptible to fracture in a hybrid composite than in a non-hybrid
composite.
When the number of fibers in the composite model is high, the
solution procedure of the governing equations becomes very com-
plex. Fukunaga, Chou and Fukuda (1984) have evaluated the stress
concentration factors using an eigenvector expansion method.
Tables 5.1 and 5.2 show the numerical results of their analysis based
upon a glass/carbon intermingled hybrid composite (R = 5). The
solid and open circles represent HM and LM fibers, respectively.

Fig. 5.5. Stress concentration factors vs. total number of broken fibers for
a hybrid composite of R = \. (After Fukuda and Chou 1983.)

5r

R =1

_L
5 10 15 20
Number of discontinuous fibers
242 Hybrid composites

Table 5.1 gives the stress concentration factors for various VHM
values, where KiJtk, for instance, is the stress concentration factor of
the kth fiber due to the breakage of the /th and /th fibers for various
fiber relative volume fractions. The fiber arrangements given in
Table 5.1 are repeated to generate the composite, but the position

Table 5.1. Stress concentration factors (SCF) for various VUM

values. After Fukunaga, Chou and Fukuda (1989)

VHM

1.0 ).75 0.5 0.25

o # o # o otoo
SCF 12 34 1 2 3 4 12 34 12 34

Kl,2 1.333 L.356 1.777 1.141

A-,.3 1.067 1.077 1.121 1.036
Kh4 1.029 [.041 1.060 1.017
K 2 l 1.333 [.347 1.131 1.829
K2,3 1.333 1.347 1.131 1.829
K2A 1.067 1.102 1.030 1.208
*3.. 1.067 1.077 1.121 1.036
K32 1.333 L.356 1.777 1.141
KXi 1.333 L.727 1.777 1.311
K4l 1.029 L.007 1.010 1.017
K4,2 1.067 1.018 1.030 1.036
K4,3 1.333 1.128 1.131 1.318
^12.3 1.600 L.654 1.412 2.144
^12,4 1.143 L.221 1.135 1.304
^13.4 1.419 L.832 1.952 1.340
Kl32 1.802 L.772 2.767 1.291
^14,2 1.412 L.378 1.817 1.180
^14,3 1.412 L.210 1.261 1.359
^23.1 1.600 L.654 1.412 2.144
^23.4 1.600 >.275 2.038 1.913
^24,1 1.419 L.362 1.146 1.886
^24,3 1.802 L.494 1.270 2.252
^34.1 1.143 L109 1.172 1.077
^34,2 1.600 L.478 2.038 1.258
•^123,4 1.829 2.116 2.510 2.175
^234,1 1.829 L.768 1.522 2.382
^341.2 2.022 L.914 3.105 1.418
^412,3 2.022 L.822 1.569 2.612
 Stress concentrations 243

of the fractured fiber (or fibers) is not repeated. In the numerical

calculations 80 fibers are used in each composite model.
Table 5.2 presents the effect of fiber bundle size on stress
concentration for the relative fiber volume fraction of VHM = 50%.
Two (case 2), three (case 3) or four (case 4) fibers of the same type
can be placed adjacent to one another besides the alternating
arrangement of one HM and one LM fiber (case 1). It is evident
that the stress concentration factor is sensitive to the microscopic
fiber arrangements. Knowledge of the stress concentration factors
in various hybrid fiber arrays is essential to the evaluation of hybrid
composite strength.

5.2.2 Dynamic case

The dynamic stress concentration in hybrid composites due
to fiber breakage has been examined by Ji, Hsiao and Chou (1981).
Figure 5.6 shows an interlaminated hybrid composite for the
analytical model; it is composed of a layer of HM fiber and a layer
of LM fiber embedded in a common matrix. The fibers are aligned
along the x axis, and h1 and h2 denote the fiber spacings. A fiber in
each array is numbered by an integer n (-°°<w<°°). The
displacement field of a fiber as a function of location and time is
denoted by un(x, t) for an HM fiber, and by u*(x, t) for an LM
fiber. Similarly, the axial forces in the fibers are denoted by pn(x, t)
and Pn(x, t). Ji and colleagues have analyzed the dynamic stress

Table 5.2. Stress concentration factors for various bundle sizes.

After Fukunaga, Chou and Fukuda (1989)

1 2 3 4
Case 1 O • O •;O • O
• O
Case 2 % #O O •:o
o • O
Case 3
• oo• • • •
o:o o #
Case 4
o oo : • • •
o:o o o •
Fiber location 1 2 3 4

Case 1 1.777 X 1.777 1.121

Case 2 1.376 X 1.762 1.175
Case 3 1.354 X 1.354 1.113
1.082 1.366 X 1.763
Case 4 1.343 X 1.351 1.112
1.079 1.365 X 1.764
244 Hybrid composites

concentration factor of the fiber n = 1 or — 1 in the HM fiber array

when the fiber n = 0 suddenly breaks.
The fundamental equations governing the deformation of the HM
and LM fibers are approximated by
2
rr,i 9 un _ Ghlt Gh2
&r h2
32un
= Y dt2
(5.23)
EM
dx1 + U1-2u*n+u*n.1) + ~(nn-K)

dt2
for all n (= -oo, . . . , - 1 , 0, 1, . . . , <*>). In Eqs. (5.23), m and ra*
are the fiber masses per unit length, £* and E are the fiber Young's
moduli, A* and A denote fiber cross-sectional areas, and G is the
matrix shear modulus.
The boundary conditions are
p Q (o,o = o
Mn(0, t) = 0 (n¥=0) pn(±°°, t)=p (alln)
(5.24)
E*A*
= 0 (all n) K(±°°.0 = p (all n)
EA

Fig. 5.6. A model of interlaminated hybrid composite. (After Ji, Hsiao

and Chou 1981.)

yf HM fiber
y
. LM fiber

7
6
5
4
•~r 3 I *• » 2 *
2
1
0
-1 -1
_2 -2
_ -i
-3
-4 -4
 Stress concentrations 245

The initial conditions are

(5.25)

for all n.
The forces and displacements of the fibers are related by

(5.26)

By introducing the following dimensionless parameters,

pn-- _£? p*—Pn

"n —
P P

un l/EAGh1 \ j EAGhl
Un- !
np V\
"-p Vl A2
L
A2 )
(5.27)
\ /1 *
V \EAh2r V \mh2,
r
E*A* m* hi
R EA M
m

Eqs. (5.23) can be rewritten as

~ 2
(5.28)
R

Also, Eqs. (5.24) become

PO(0,T) = 0

Un(0, R) = 0 (n# 0), PB(±«, T) = 1 (all n) (5.29)

t/;(0, T) = 0 (alln), P : ( ± » , T) = /J (all n)

246 Hybrid composites

and Eqs. (5.25) can be written in terms of displacements only as

(5.30)
0) =

Equations (5.28)-(5.30) can be solved following the approach of

Hedgepeth. Three essential steps are involved: the Laplace trans-
form in time, the use of the technique of influence function, and the
Fourier series representation of the unknown functions. Ji and
colleagues have evaluated the dynamic stress concentration factor of
the HM fibers immediately adjacent to the broken fiber. The most
severe stress concentration factor #i(0, r) occurs at § = 0. It is
interesting to note that the solution of K^O, r) is composed of two
parts which are related to the HM and LM fibers, respectively.
Figures 5.7-5.9 depict the variation of the stress concentration
factor K1(0) R) with the dimensionless time rfor m*/m = 1, 2 and 6,
respectively. The two components of the solution are denoted by K[
and K'[. Also in these figures, E*A*/EA = h2/h1 = 1. Figure 5.7 is for
the case of non-hybrid composites and the solution of Hedgepeth
for a unidirectional lamina is also shown. The summation of K[ and

Fig. 5.7. The variation of stress concentration factor Kx with dimension-

less time r for m*/m = E*A*/EA = h2/hl = 1. K[ and K'[ are the two
components of Kx. (After Ji, Hsiao and Chou 1981.)
2.01-

Hedgepeth?s curve

1.5

1.0

0.5

10 12
 Tensile stress-strain behavior 247

K'[ gives the total stress concentration factor. As the difference in

mass density of the HM and LM fibers increases, the locations of
the peak values of K[ and K'{ are out of phase. As a result, the
maximum value of K[ + K'[ is reduced (Figs. 5.8 and 5.9).
The analysis of Ji and colleagues concludes that the time
variations of the stress concentration factors related to the two
component fiber materials are always out of phase in a hybrid
composite with fibers of different mass densities. The parent HM
fiber composite always provides the upper bound for the stress
concentration of the hybrid, since this is the case where there is no
difference in phase and magnitude of the K[ and K'[ values.
Furthermore, the magnitudes of the K[ and K'[ are determined by
the extensional stiffnesses of the component fibers.

5.3 Tensile stress-strain behavior

An idealized stress-strain curve of a hybrid composite
containing both high elongation (HE) and low elongation (LE)
fibers is depicted in Fig. 5.10 (Aveston and Kelly 1980). For hybrids
with good bonding between the component phases, the stress-strain
curve is given by OABC. The important features of this curve
include the elastic behavior indicated by OA, the first cracking

Fig. 5.8. The variation of stress concentration factor KA with dimension-

less time T for m*/m = 2, and E*A*/EA = h2/hl = i. K\ and K'[ are the
two components of Kx. (After Ji, Hsiao and Chou 1981.)
2.0 r-
248 Hybrid composites

Fig. 5.9. The variation of stress concentration factor Kx with dimension-

less time T for m*/m = 6, and E*A*/EA = h2/hl = 1. K[ and K'[ are the
two components of Kx. (After Ji, Hsiao and Chou 1981.)
2.0 r-

1.5

l.O

- . • * .
0.5

6 10 12
T

Fig. 5.10. A typical stress-strain curve of hybrid composites. (After

Aveston and Kelly 1980.)

£,£!.„

£I.U(1 + a/2) £Hu

Strain
 Tensile stress-strain behavior 249

strain £Lu, the relatively flat portion of the curve AB, the
subsequent rise of the curve at a smaller slope (BC) and the final
failure strain of the hybrid given by the point D. The subscripts H
and L are used to denote parameters related to high and low
elongation fibers, respectively. The various features of the stress-
strain curve are discussed in the following.

5.3.1 Elastic behavior

Theoretical predictions of the elastic moduli of multi-phase
short-fiber composites have been performed by Chou, Nomura and
Taya (1980) using a self-consistent approach, Nomura and Chou
(1984) based upon a bound approach, and Taya and Chou (1984)
employing a combination of Eshelby's (1957) equivalent inclusion
method and Mori and Tanaka's (1973) back stress analysis.
For unidirectional hybrid composites composed of continuous
fibers, Chamis and Sinclair (1979) have examined the elastic
properties based upon composite micromechanics approach, linear
laminate theory, and finite element analysis. It has been concluded
that these methods predict approximately the same elastic pro-
perties. The through-the-thickness properties predicted by the
micromechanics equations are in good agreement with the finite
element results.
For simplicity, the results from the micromechanics approach
are introduced below. The analytical model is an intermingled
hybrid composite composed of two components, termed primary
composite and secondary composite by Chamis and Sinclair (Fig. 5.11).
Since these two components are interchangeable, they can be

Fig. 5.11. An intermingled hybrid composite lamina composed of primary

(HM) and secondary (LM) phases.
HM phase
LM phase
250 Hybrid composites

considered the high modulus (HM) and low modulus (LM) com-
ponents. In the following, the subscripts 1, 2 and 3 denote,
respectively, the directions along the fiber, transverse to the fiber
and through the thickness in an intermingled hybrid lamina.
The effective longitudinal Young's modulus is approximated by
the iso-strain assumption along the 1-direction:

En = E™ + (E^-E^)VLM (5.31)

Here, Ef™ and E\™ denote the longitudinal Young's moduli of the
HM and LM composites, respectively. FLM is the volume fraction of
the LM composite and VLM + VHM = 1. Both E ™ and EnM can be
expressed in terms of the properties of the fiber and matrix as given
in Eqs. (2.7).
The transverse Young's modulus is obtained by assuming that the
HM and LM components are connected in series in the 2-direction

_ rHM
(5 32)
.

Here, is™ and E22™ are the transverse Young's moduli of the HM
and LM composites, respectively, is™ and E 2™ can also be
expressed in terms of fiber and matrix elastic properties as well as
the fiber volume fraction of each component phase (Eqs. (2.7)).
The effective Young's modulus in the through-the-thickness
direction has been modeled by Chamis and Sinclair assuming that
the component phases are in parallel in the 3-direction. Thus,
rHM
{t33 - E33 )K LM (5.33)

Here, £ " M and E\™ are the transverse Young's moduli of the
unidirectional composites composed of HM and LM fibers, respec-
tively. Since the unidirectional all-HM or all-LM fiber composite is
assumed to be transversely isotropic the expression of E33 (=E22) of
Eqs. (2.7) can be used to relate E\f*(E3™) to the constituent
material properties. However, it should be noted that for a
yarn-by-yarn unidirectional intermingled hybrid lamina, E22=£E33
and thus the composite is not transversely isotropic.
The in-plane shear modulus G12 is obtained by an iso-stress
approximation

(5.34)
 Tensile stress-strain behavior 251

Similarly, the interlaminar shear modulus G23 is expressed as

(5.35)

The interlaminar shear modulus G13 is obtained by assuming that

the component phases are connected in parallel in the 3-direction.
Thus analogous to Eq. (5.31) the following expression can be
obtained:
+
13 = " 1 3 ( G \ ? —"13 J^LM (5. 36)
in Eqs. (5.34)-(5.36), the shear moduli of the component phases
can again be related to the constituent fiber and matrix properties
by using Eqs. (2.7).
The Poisson's ratios v12 and v32 are derived by assuming parallel
elements in the 1- and 3-directions.
v12 = v?2M + (v\™ - v™)V L M (5.37)
v32 = v?2M + (v!f2M - v3H2M)VLM (5.38)
For the derivation of v 13 , the iso-strain and iso-stress states
are assumed in the 1- and 3-directions, respectively. The result is
v f V LM _ HM\
vv _ ,.HM
v
. ^LM^Vl3 Vl3 ) ,- „„.
+ pjy;

33

Experimental measurements of the elastic properties of inter-

mingled hybrid composites can be found in the work of Chamis and
Sinclair (1979) and Gruber and Chou (1983). The thermal prop-
erties of unidirectional intermingled hybrid composites have also
been examined by Chamis and Sinclair (1979). It has been
recommended that linear laminate theory be used to predict the
thermal expansion coefficients.

5.3.2 First cracking strain

The linear portion of the stress-strain curve often extends
beyond the failure strain of the pure LE fiber composite (point A'
in Fig. 5.10) and the first cracking of the LE fibers occurs at the
strain £Lu (point A in Fig. 5.10). This phenomenon is known as a
'hybrid effect'. The treatment of fiber first cracking strain of
Aveston and Kelly (1980) is introduced below.
A unidirectional hybrid composite composed of both high and
low elongation components can continue to bear the total load after
252 Hybrid composites

the first cracking of the low elongation component if the following

condition is satisfied:

(5-40)
- O

Here, au denotes the failure stress, V indicates fiber volume

fraction, the subscripts L and H denote the LE and HE com-
ponents, respectively. Also, o'H (=£LuEH) is the stress of the HE
component at the failure strain of the LE component. For both HE
and LE components behaving elastically up to £Lu, the above
condition can also be expressed in terms of the failure strain of the
high elongation component
£HU^£LU(1 + " ) (5.41)
where

«=f ^ (5-42)
and E denotes the Young's modulus. Obviously the effect of stress
redistribution due to fiber breakage is not considered here.
The magnitude of the first cracking strain £Lu and the extension of
the curve between A and B in Fig. 5.10 can be understood using the
concept of multiple cracking and constrained failure (Chapter 3). It
has been established that the first failure strain is size dependent.
The term size means essentially the effective diameters of fiber tows
and their spacings at a fixed fiber volume content as well as the
thickness of lamellae in an interlaminated hybrid.
When an interlaminated hybrid composite is deformed beyond
the first cracking strain, parallel cracks will appear in the low
elongation phase with the crack planes normal to the loading axis.
The spacing of the cracks depends on the bonding (i.e. elastic or
frictional) between the component plies, after the initial cracking. If
they remain elastically bonded, the crack spacing is determined by
the maximum interfacial shear stress at the crack, rmax. On the
other hand, if load transfer occurs through frictional bonds between
the component phases, the crack spacing is determined by the
limiting bond strength R. In both cases, the formation of a crack
through the ply thickness of the low elongation component results
in the relaxation of the material in this ply on both sides of the
crack. The nearly flat portion of the composite stress-strain curve,
i.e. AB in Fig. 5.10, is the consequence of multiple cracking of the
low elongation phase and the associated extension of the specimen.
 Tensile stress-strain behavior 253

The stress originally carried by the low elongation component on

the crack plane has to be shifted to the high elongation phase. The
maximum additional stress thrown onto the high elongation phase
can be estimated by
aa oaEnB
Ao = -' H E - — (5.43)
where aa is the applied stress and Ec is the Young's modulus of the
composite. The first term on the right-hand side of Eq. (5.43) is the
stress on the crack plane while the second term gives the stress away
from the crack plane in the high elongation component. The
additional load carried by the high elongation component induces
an extension 61 of the specimen ends.
The idea of load transfer between the HE and LE components
explains why the strain £Lu, at which first cracking occurs, depends
upon the dispersion of the component phases. When thinner fibers
or lamellae are used, the interfacial area per unit volume between
the two component phases increases. This also means increased
efficiency of load transfer from the HE component bridging the
crack back to the LE layers. As a result, the additionally strained
length of this component and, hence, the displacement of the
specimen ends, 61, are decreased. The product of 61 and £ c £ Lu gives
the upper limit of the work available from the loading system to
form the crack. Assuming a constant surface work of fracture y, the
decrease in 61 will reach such a point that eLu must increase above
the value for the pure LE composite before the required work of
fracture can be extracted from the system. In the case of inter-
laminated hybrids with an elastically bonded interface the cracking
strain of the low elongation phase of thickness d can be estimated
by (Aveston and Kelly 1980)

(5 44)
) -
where a is defined in Eq. (5.42). Equation (5.44) predicts that £Lu
varies with the inverse square root of ply thickness. This prediction
of the hybrid effect is consistent with the experimental observation
of carbon/glass sandwich laminates that the carbon ply failure strain
is greater when its absolute thickness is smaller.
The theory of multiple cracking in fiber composites has very broad
applicability. Multiple cracking occurs in the brittle phase which
could be either the matrix or the fiber phase of an aligned fiber
254 Hybrid composites

composite, the low elongation layers of a non-hybrid laminated

composite, or the layers reinforced with LE fibers in an interlamin-
ated hybrid composite. Multiple cracking of non-hybrid composites
has been examined in Chapter 3. Aveston and Kelly (1980) have
summarized the analytical expressions of the minimum crack
spacings, the first cracking strain and the maximum interfacial stress
for hybrid and non-hybrid composites with both elastic and sliding
friction bonds.
The above discussions have provided the answer to the question
posed in Section 5.1 concerning synergistic effects in hybrid
composites. The answer to the question of load sharing is also
positive. The bond between the fiber and matrix in a hybrid ensures
that the LE fiber continues to carry part of the applied load and to
contribute to the overall stiffness after first cracking (Bunsell and
Harris 1974). The load sharing by the LE fiber is evident from the
observation of multiple fractures in well bonded interply hybrids
and by the bursts of acoustic emission accompanying the repeated
load drops on the stress-strain curve. Finally, the rise of curve
BC in Fig. 5.10 is attributed to the loading of the high elongation
fibers. The failure strain of the hybrid composite (point D) is lower
than that of the high elongation fiber composite (point G). This is
because the multiple fracture of the LE fibers and partial debonding
between the HE and LE fiber reinforced laminae. Consequently,
the HE layers in an interlaminated hybrid composite cannot be
stretched uniformly along their length to the ultimate strain level.
On the other hand, if the debonding is complete at the first fiber
cracking strain, £Lu, the stress-strain curve follows the path OAEF.
Note that EF and BC in Fig. 5.10 have the same slope.

5.3.3 Differential Poisson 's effect

Another factor that needs to be taken into account in the
deformation of hybrid composites is the interlaminar stress induced
due to differential Poisson's effect (Aveston and Kelly 1980). It is
understood that this effect exists in laminated composites with or
without fiber hybridization. To demonstrate the magnitude of the
Poisson's strain and its effect on longitudinal splitting of laminated
composites, a three-layer non-hybrid cross-ply laminate is con-
sidered. The central LE (90°) layer in this case is sandwiched
between two HE (0°) layers. Thus the conclusions derived from this
example are applicable to interlaminated composites in general. The
strain induced by the differential Poisson's effect depends only on
 Tensile stress-strain behavior 255

the volume fraction of the component phases. However, cracking of

the lamina can be minimized by making the LE layers sufficiently
thin.
The critical strain for causing longitudinal split due to Poisson's
effect can be derived based upon Fig. 3.25. Let the subscripts 1 and
2 denote the longitudinal and transverse fiber directions, respec-
tively, and x-y-z are the reference axes of the cross-ply. Under a
simple extension eyy, the following strains are induced in the 0° and
90° layers, if they are deformed independently:

(5.45)
exx(90°)=-v2,£yy
For the composite laminate subjected to eyy, the strain induced in
the x direction is exx. Referring to Fig. 3.25, the transverse strains
induced in the 0° and 90° layers due to the Poisson effect
are
Aexx{W) = exx - exx(0°) = exx + vl2eyy
(5.46)
Ae«(90°) = exx - exx(90°) = exx + v21eyy
Thus
Aexx(0°) - &exx(90°) = (v 12 - v21)eyy (5.47)
The stress oxx is built up in each layer due to the requirement of
compatibility in normal strain in the x direction. These stresses are
given approximately by
a«(0°) = Aexx(0°)E22
(5.48)
oxx(90°) = Aexx(90°)Eu
The force equilibrium of the laminate in the transverse direction
requires
axx(0°)b + axx(90°)h = Aexx(0°)E22b + Aexx(90°)Euh =0
(5.49)
From Eqs. (5.47) and (5.49), the transverse strain induced in the 0°
layer due to the Poisson's effect is
256 Hybrid composites

where Ec = E22(b/h + b) + En(h/h + b) is the effective Young's

modulus along the x direction.
If the 0° and 90° layers in the laminate of Fig. 3.25 are of different
materials, then it is necessary to distinguish the elastic constants in
Eq. (5.50)

£ll(9O>[v12(O°)-v21(9O°)Ky
xA } ( j
bE22(0°) + hEn(90°){}=En

5.3.4 Differential thermal expansion

Additional strain may be induced in a laminated hybrid
composite due to differential thermal expansion of the component
phases. The carbon/glass hybrid system is a typical case where the
axial thermal expansion coefficient of the glass laminae is much
larger than that of the carbon layers. Upon cooling down from the
stress-free temperature, compressive stress develops in the carbon
layers as a result of the differential thermal expansion. This
thermally induced compression can partially account for the hybrid
effect often observed in carbon/glass hybrids (Bunsell and Harris
1974). The constrained thermal strain in such a hybrid with high and
low elongation fibers can be expressed as (Aveston and Kelly 1980)

T ATEHEVHE
£LE = = ( <*LE - « H E ) (5.52)
E
T ATELEVLB
£HE = = ( »HE ~ ^LE) (5.53)
Ec
where AT = stress-free temperature — service temperature, and a LE
and arHE denote the thermal expansion coefficients. The thermal
strain components are independent of fiber or lamina dimension.
The relation between the cracking strain and the dimension of the
low elongation phase is identical under external load and thermal
load. Hence, the treatment of multiple cracking can still be applied
in this case.

5.4 Strength theories

Discussions on the strength of hybrid composites begin
with an introduction on the rule-of-mixtures type of approach which
delineates the contributions of the high elongation and low elonga-
tion fibers to the load carrying capacity of the hybrid. In this
approach, the fibers are assumed to be of uniform strength and the
local stress redistributions due to fiber breakage are not taken into
 Strength theories 257

account. This is then followed by the probabilistic strength theories.

The synergistic effects between the LE and HE fibers in local stress
concentration and fiber strength distributions are modeled to
predict the first failure strength and ultimate failure strength of
hybrid composites.

5.4.1 Rule-of-mixtures
The ultimate tensile strength of a unidirectional hybrid
composite can be estimated from the contributions of the com-
ponent phases at different volume fractions. Consider a binary
composite with low elongation fibers. The addition of a small
amount of higher elongation fiber decreases the strength of the
composite. The ultimate hybrid composite tensile strength, ACU, is
given by (Aveston and Kelly 1980)

tfcu = CJLU V LE + £ L U £ H E VHE (5.54)

where £Lu is the failure strain of the LE fiber in the hybrid. The
failure of the low elongation fiber leads to the fracture of the hybrid
and there is no multiple fracture.
As the content of the high elongation fiber increases, a transition
in failure mode occurs when there is sufficient volume of these fibers
to carry the load upon the fracture of the low elongation fibers. The
fracture mode is multiple fracture of the brittle fibers. The ultimate
tensile strength is represented by

OCU=OHUVHB (5.55)

The volume fraction of HE fibers should exceed the lower limit,

given by Eq. (5.40), to bear the total load at the first cracking
strain. Aveston and Kelly (1980) applied Eqs. (5.54) and (5.55) to
analyze the experimental data of Kalnin (1972), who studied the
failure stress of interlaminated carbon/glass/epoxy hybrids. Figure
5.12 shows the variation of hybrid composite ultimate tensile
strength with the relative glass fiber content. The stress is calculated
by dividing the load by the cross-sectional area of the fiber and,
thus, the contribution of the epoxy matrix is neglected. Aveston and
Kelly suggested that for this particular experimental system the
fracture of carbon fibers did not produce a large stress concentra-
tion leading to the weakening of the glass fibers.

5.4.2 Probabilistic initial failure strength

It has been shown in Section 5.3.2 that the phenomena of
initial failures of unidirectional hybrid composites and cross-ply
258 Hybrid composites

laminates share the same basic physical principle. In unidirectional

hybrid composites consisting of LE and HE fibers, the failure strain
of the LE fibers under tension is often greater than that in the all
LE fiber composite. On the other hand, in the non-hybrid [±
0°/9O°]s (0° < 6 < 90°) laminates the failure strength (strain) of the
90° (LE) layer under tension is greater for smaller thickness of the
inner 90° layers. It is also known that the failure strength (strain) of
the inner 90° layer depends on the material properties of the outer
±6° layers. Thus, in composites which consist of laminae with two
different types of material properties such as hybrid composites and
[±0°/90°]s non-hybrid laminates, the failure strength (strain) of the
LE layers is not an intrinsic material property and it depends on the
HE material properties and the geometric arrangement of the
layers.
Fukunaga, Chou and Fukuda (1984) and Fukunaga et al.
(1984a&b) have examined the initial failure strength of both hybrid
and non-hybrid composites based upon a statistical approach. The
hybrid composite under consideration is a sandwiched structure
composed of unidirectional glass fiber and carbon fiber laminae
(Fig. 5.13a). The non-hybrid composite is a carbon composite with
the [±6°/90°]s configuration (Fig. 5.13b). Both composites can be
depicted by the HE and LE representations of Fig. 5.13(c) where
M H E and MLE denote, respectively, the number of HE and LE

Fig. 5.12. Tensile strength of aligned carbon/glass/epoxy hybrids vs.

relative fiber content. (After Aveston and Kelly 1980.)

°Hu
1500,

1000 O

Single O j Multiple fracture

fracture • * of carbon
500 s
s
s

' 1 1 1
0 0.2 0.4 0.6 0.8 1.0

+Carbon)
 Strength theories 259

layers. Thus, the total number of layers in the hybrid composite,

M = MHE + M L E .
Fukunaga and colleagues modeled the composite as a chain of
short laminates in series as shown in Fig. 3.17. Each laminate has
the length 6 equivalent to the ineffective length, and the specimen
length / = NS. In order to obtain the first ply failure strength of the
whole laminated composite, the following two-parameter cumula-
tive Weibull distribution functions of failure strain for the LE and
HE short layers are assumed

(5.56)

Here e* and /? denote the scale and shape parameters, respectively.

Then, the cumulative distribution function HC(S) for the first ply
failure strain of the composite can be obtained from the weakest
link model

i/c(e) = l - [ l - Gc(e)]" (5.57)

where Gc(s) is the cumulative distribution function for the first ply
failure strain of the short laminate and it is given by

GC(E) = 1 - [1 - (5.58)

Fig. 5.13. (a) Carbon/glass hybrid laminate, (b) [±0°/9O°]s non-hybrid

laminate, (c) Model for analysis. (After Fukunaga et al. 1984c.)

I© ®<
© il

90°

(a) (b)

HE MHE/2
LE
HE M HE /2

(c)
260 Hybrid composites

Substitution of Eq. (5.58) into Eq. (5.57) yields

(5.59)
It should be noted that Eq. (5.59) is concerned with the first ply
failure strain of the composite laminate not the LE phase of the
composite. This is because, from a probabilistic viewpoint, the
failure of the LE layer does not always precede that of the HE
layers. Similar to Eq. (5.59) the cumulative distribution function
/JLE(£) for the first ply failure strain of the all LE fiber composite
(of the same size as the hybrid composite) is given by
# LE (£) = 1 ~ exp[-A^M(£/eLE)^IE] (5.60)
Fukunaga and colleagues have compared the failure strains for
the HE/LE/HE laminate and the pure LE laminate for the failure
strains at 50% failure probability. From Eqs. (5.59) and (5.60)
;
M / er \^HE In 2
+ - HE
M \£ L E / M \ET M J NM
(5.61)
E
In 2
\eiJ NM
where ec and £LE denote, respectively, the median failure strains of
the hybrid and LE composite. In order to obtain an explicit relation
between the ratio of median failure strains, £C/£LE> and the relative
volume fraction of the LE material, MLB/M, the case of /3LE =
/3HE = p is considered. Then, Eqs. (5.61) become

(5 62)
fHM LE + y ( £ * ) " P -
£LE \ M /
where E* = £ H E / £ L E . The ratio £C/£LE in Eq. (5.62) is independent
of the size of the composite. By assuming linear stress-strain
relations for the HE and LE layers, the first ply failure strength
ratio crc/aLE is readily obtained

where E denotes the Young's modulus in the loading direction,

 Strength theories 261

Numerical calculations of Eqs. (5.62) and (5.63) have been

performed for carbon/glass/epoxy unidirectional hybrid composites.
It is assumed that e* = 3.0 and EHE/E^E = 3. Figure 5.14 shows the
comparison of Eq. (5.62) with the experimental results of Bader
and Manders (1981a&b) for HTS carbon/glass hybrid laminates.
The analytical results for jS = 10 seem to show good agreement with
experimental results. Figure 5.15 indicates the variation of initial
composite failure strength with the relative volume fraction of LE
and HE layers. Points A and D in Fig. 5.15 represent the strengths
of glass and carbon composites, respectively. Line BD(oc/o^E =
EJELB) indicates the stress in the hybrid at which failure of the
carbon layer takes place. Line AE(oJal^E = MHE/(MLE + MHE))
represents the stress in the hybrid assuming that the LE layer
carries no load. As the shape parameter ft decreases, the first ply
failure strain and strength of the hybrid composite increase relative
to those of the all carbon fiber composite. When )3—o°, that is for
composites without scattering in the failure strains of the LE and
HE layers, the first ply failure strain of the hybrid composite is
identical to that of the pure LE composite. This relation at /?-» °° in
Fig. 5.15 is equivalent to the rule-of-mixtures for the initial failure
in hybrids.

Fig. 5.14. Comparisons of the analytical predictions of Fukunaga et al.

(1984c) and the experimental results of Bader and Manders (1981a&b) for
HTS carbon/E-glass hybrid laminates. (After Fukunaga et al. 1984c.)

• Experimental data

«• 1 . 2 -'

0.2 0.4 0.6 0.8 1.0

MLE/(MLE + Mw)
262 Hybrid composites

The trend of variation of £C/£LE with the relative 90° layer volume
fraction for [±#790°]s laminates is similar to that given in Fig. 5.14.
For a given value of shape parameter, EJELB decreases with the
increase in the angle 6.
In summary, the probabilistic strength analysis of Fukunaga,
Chou and Fukuda (1984) and Fukunaga et al. (1984a) for the initial
failure in hybrid and non-hybrid composites has demonstrated a
volumetric relation. It has been shown that the initial failure strain
or strength is greater in composites composed of low elongation and
high elongation materials than in the all low elongation fiber
composite. This is the result of a 'size effect'; that is the failure
probability is lower in the composite with the smaller size of the low
elongation material.

5.4.3 Probabilistic ultimate failure strength

In this section three analytic approaches are presented for
predicting the ultimate tensile strength of hybrid composites. All
these methods consider the variability in fiber strength and the
stress redistribution at fiber fracture. First, the analytical model of
Zweben (1977) assumes that the LE and HE fibers are arranged in
alternating positions in a unidirectional lamina. The fracture of an

Fig. 5.15. Composite initial failure strength ac (normalized by a L E ) in

glass/carbon/glass laminates. )3 L E = PUE = P> £ H E - £ LE = 3.0 and
£ H E / £ L E = 3 - (After Fukunaga et al. 1984c.)
 Strength theories 263

LE fiber induces, in the vicinity of the fiber fracture, a local stress

(strain) concentration. The HE fibers will break at the points of
stress (strain) concentration. Zweben hypothesizes that the strain
level at which the first overstressed HE fiber is expected to break is
a lower bound on the ultimate strain of the hybrid composite.
Next, in order to obtain a more realistic view of the multiple
fracture of the LE fibers in a hybrid composite, Fukuda and Chou
(1982a&b) adopted a Monte-Carlo simulation of the ultimate failure
of a unidirectional hybrid composite. The method demonstrates the
diffused nature of fracture of the LE fibers. The constraint on the
propagation of the LE fiber fractures due to the presence of HE
fibers as well as the stress-strain relation of the hybrid composite as
influenced by multiple fiber fractures has been demonstrated.
Thirdly, Fukunaga, Chou and Fukuda (1989) have adopted a
more rigorous approach by applying the methods of Harlow and
Phoenix (1978a&b). The effects on the ultimate strength of hybrid
laminates due to the scatter of laminar strengths, relative fiber
volume fractions, composite size and laminate stacking sequence
have been identified.
The analytical model of Zweben is composed of a single layer of
fibers of axial length / with the same arrangement as that shown in
Fig. 3.17. High elongation (low modulus) fibers and low elongation
(high modulus) fibers are arranged in alternating positions. The
total number of fibers in the composite is M, of which M/2 are LE,
and M/2 are HE fibers. The term 'fibers', according Zweben,
represents both single fibers and yarns.
It is assumed that the fibers support all of the applied load and
the fibers break under load in a random fashion throughout the
composite. Attention is focussed on what happens in the vicinity
of the breaks in the LE fibers. The strain of the composite is
adopted as the independent variable. When an LE fiber breaks at
the strain level e, the two adjacent HE fibers are subjected to a
strain concentration of Khs, where Kh is the strain concentration
factor associated with a single broken LE fiber. Because of the
linear elastic deformation assumed in the model, Kh is also the
stress concentration factor. Since the model assumes that the fiber
axial stress depends only on the axial strain, the stress in these two
HE fibers increases from EHBe to KhEHB£. The axial distance over
which the fiber stress is perturbed due to a fiber breakage is known
as the ineffective length (Rosen 1964). Zweben denoted the
ineffective length associated with a broken LE fiber in the hybrid by
<5
264 Hybrid composites

Similar to Fig. 3.17, the hybrid composite is composed of a series

of links with axial dimension <5h. The total number of links is
Nh = //<5h. Failure of the composite results from the propagation of
fiber breaks due to local strain concentrations.
The cumulative distribution functions for the failure strains of the
LE and HE fibers of length / are assumed in the form of Weibull
distributions:
FLE(e) = 1 - expi-ple")
(5.64)

where py q, r and s are Weibull parameters. The composite strain at

which the fracture of the first overstressed HE fibers occurs in the
hybrid is given by
eh = [ A f ( p r 5 h ( ^ h - l ) ] - 1 ' ^ > (5.65)
For a composite reinforced with M fibers of the same type, say the
LE fibers, Eq. (5.65) becomes
e = [2Mlp2SBB(KlB - l)]" 1 / 2 « (5.66)
Here, X"LE and <5LE are, respectively, the strain concentration factor
and ineffective length of the LE fiber composite.
From Eqs. (5.65) and (5.66), Zweben has obtained the ratio of
the lower bounds of failure strain of a hybrid to that of a LE fiber
composite of the same length

£h_ [MlPr6b(Kll)Y
e [IMlp^UKll)]-112" l }

Equation (5.67) indicates that the ratio of the failure strains

depends on the Weibull strength parameters p> q, r and s> the
specimen length /, the ineffective lengths <5h and <5LE, and the strain
concentration factors KBB and KUB. Zweben has derived approxi-
mate expressions of these parameters in terms of the fiber elastic
properties and the geometric parameters of fiber arrangements in
the composites.
For the convenience of assessing the failure characteristics of
hybrid composites, Eq. (5.67) can be simplified by assuming that
both LE and HE fibers have the same coefficient of variation in
tensile failure strain. It can be shown that for this case q = s and Eq.
(5.67) becomes
 Strength theories 265

From Eq. (3.54a) the mean strains of the LE and HE fibers for the
gauge-length / are given by

(5.69)

where T is a gamma function. Thus, for q=s, Eqs. (5.68) and

(5.69) yield

Equation (5.70) can be further simplified if the fiber coefficient of

variation is small (i.e. 5% or less) and thus the shape parameter is
large (q > 25). For this case KqHE - 1 « #HE and Kq - 1 « Kqy and
Eq. (5.70) is reduced to

XifrV(£)
Equation (5.71) indicates that the ratio of the lower bounds of
failure strains is sensitive to the mean fiber failure strains and the
strain concentration factors and it is less sensitive to the ineffective
lengths under the assumption of low fiber coefficient of variations.
For the case of an intermingled Kevlar 49/Thornel 300 hybrid
composite, Zweben obtained Kh = 1.462, K^E = 1.293, 6J6LB =
1.573/1.531 = 1.03, and CREOLE = 1.63. The fiber strain coefficients
of variation for both HE and LE fibers are assumed to be about 6%
if the Weibull parameter q = 20 is adopted. Using these values, Eq.
(5.71) gives R = 1.22.
In spite of the simplifying assumption used, Zweben's model, in
essence, predicts that the introduction of HE fibers into a LE fiber
composite enhances the failure strain of the hybrid composite. This
effect is attributed to the ability of HE fibers to redistribute the
local stress concentration and act like crack arrestors at the
micromechanical level.
Fukuda and Chou (1982a) have examined the strength of hybrid
composites using the method of Monte Carlo simulation. Figure
5.16 shows an idealized intermingled hybrid composite sheet of unit
thickness. The LE and HE fibers assume alternating positions. For
the purpose of numerical calculations, a five-fiber region with three
LE fibers and two HE fibers is considered. Following the notations
266 Hybrid composites

of the chain-of-links model given in Fig. 3.17, M = 5 and N = 20 are

adopted for Fig. 5.16. Also the extensional rigidity ratios R(=
E*d*/Ed) are assumed to be unity (for a non-hybrid composite)
and I (for simulating a glass/carbon composite).
Using a shear-lag analysis, the stress concentration factors are
evaluated. The fiber breakage patterns considered by Fukuda and
Chou include all the combinations of one-, two-, three- and
four-fiber fracture in a transverse plane. A shear-lag analysis is
applied to evaluate the stress concentration factors of all the intact
links in the layer where fiber breakage has taken place. Results for
two kinds of fiber composites are presented below.
In the first case, there is only one type of fiber; its strength
follows a normal distribution with an average normalized strength
of unity and a standard deviation of 0.1. A typical sequence of link
failure is shown in Fig. 5.17(a), where 0 indicates that the link is
intact and the other numbers show the sequence of fracture of links.
In this model, the link of i = 14, j = 2 breaks first; the link of / = 5,
j = 4 breaks second; finally the failure occurs at the transverse plane
of i = 14. In a total of 100 links, six links are broken. One hundred
iterations have been performed by Fukuda and Chou and the
number of broken links of each iteration is found to be either five or
six. Generally speaking, multiple fiber fractures are not extensive in
non-hybrid composites.

Fig. 5.16. Model for the Monte Carlo simulation. (After Fukuda and
Chou 1982a.)
LE fiber
/ HE fiber
/

N
 Strength theories 267

Figure 5.17(b) shows the failure sequence of links in the hybrid

composite. It is seen that the LE fibers tend to break at the initial
stage of loading and that a total of 16 links are broken for this
model to fail at the plane of / = 10. A minimum of five and a
maximum of 32 link failures are observed during 100 iterations, and
the average number of broken links in one iteration is 16.3. The
degree of multiple fiber fracture is considerably more extensive in
the hybrid composite than the non-hydrid case.
Figure 5.18 shows the examples of stress-strain relations of both
non-hybrid and hybrid composites. The stress ac and strain ec of
composites are normalized by the average ultimate stress (<XLINK)
and average ultimate strain (eLiNK) of the links of LE fibers. The
initial failure strains of the LE fiber composite and the hybrid
composite are nearly the same. However, in the hybrid composite,
initial multiple failures of the LE fibers are arrested by the HE
fibers. As a result, the hybrid composite can withstand more
deformation and hence a higher ultimate failure strain. According

Fig. 5.17. Examples of the fiber link failure sequence, (a) Non-hybrid
composite, (b) Hybrid composite. (After Fukuda and Chou 1982a.)

0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 2
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0 2 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 10 0 13 0 14
0 0 0 0 0 0 0 11 0 0
0 0 0 0 0 3 0 4 0 0
0 0 0 0 0 7 16 8 15 1
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 9 0 0 0 0
0 0 0 0 0 0 0 0 0 0
3 1 4 5 6 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 12 0 6 0 5
LE HE LE HE LE
(a) (b)
268 Hybrid composites

to this example, the ultimate strain increases approximately 50%

over that of the LE fiber composite.
Although the total number of fibers utilized in this numerical
model is extremely small the progressive nature of failure as
indicated in the stress-strain curve resembles that of the ex-
perimental curves obtained by Bunsell and Harris (1974) of
carbon/glass hybrid composites.
The ultimate strength of the hybrid predicted in Fig. 5.18 is lower
than those of the LE and HE fiber composites and the ultimate
strain of the hybrid is lower than that of the HE fiber composite.
This can be understood from the fact that, with the exception of
those on the crack plane, the links of the HE fibers are constrained
by the surrounding LE links and cannot be deformed to their
ultimate strain. As a result, the strength potential of the HE links
is not fully realized.
Fukunaga, Chou and Fukuda (1989) have extended the treatment
of Harlow and Phoenix (1978a&b) to analyze the ultimate tensile
strength of hybrid composites. In their model, the laminate is also
treated as a chain of N short segments arranged in series. Each
segment has a length 6 and the total length of the composite is
/ = Nd. There are four layers in the laminate. Figure 5.19 shows the
stacking sequence of the laminates consisting of LE and HE fiber
layers. The strain concentration factors for the various configura-
tions of layer breakage have been evaluated by the eigenvector
expansion method (Fukunaga, Chou and Fukuda 1984).

Fig. 5.18. Normalized stress-strain relations of non-hybrid and hybrid

composites. (After Fukuda and Chou 1982a.)

1.0
LE fiber
composite

Hybrid

0.5 HE fiber
composite

£c/Elink
 Strength theories 269

The strength analysis is performed on the basis of the knowledge

of the stress redistribution due to layer breakage. It is assumed that
each layer follows a two-parameter Weibull probability distribution
function and the ultimate failure is defined as the failure of all the
layers. The cumulative distribution functions for the ultimate strains
of the LE and HE layers are denoted by FLE(e) and FHE(£),
respectively. They are expressed as in Eqs. (5.56). The failure
patterns of the four-layered hybrid are shown in Fig. 5.20 where the
circles and crosses denote the intact and broken layers, respectively;
the possibilities of one, two, three, and four fractured layers are
given. According to Fukunaga and colleagues, there are altogether
75 possible failure sequences which can be grouped into eight

Fig. 5.19. Stacking sequences of laminates. (After Fukunaga, Chou and

Fukuda 1989.)

1 2 3 4

\ 1
1.00 A O
O O O

A • o o o
0.75 < HE LE
B o
X o
A • • o o
B • o o •
0.50 •
C • o • o
D o • • o
A • • • o
0.25 <
. B • • o •
0.00 A • • • •

Fig. 5.20. Failure patterns of the four-layered hybrid composite. (After

Fukunaga, Chou and Fukuda 1989.)
b : xoo o /: xoo o / : xxxo
r:oxoo g : x ox o m : xxo x
a : oooo d : oox o h:\oox « : xox x p : xxx x
: 1 234 e : ooo x / : oxx o o : oxx x
j:oxox
t:ooix
270 Hybrid composites

different types. The number of different failure sequences in each

type has been identified. A typical failure sequence in each type and
the corresponding failure probability are given in Table 5.3 where
F(x) stands for FLE(f) for the LE layer and FHB(e) for the HE
layer, and Kijtk denotes the strain concentration factor of the A:th
layer due to the fracture of the ith and yth layers.
The cumulative distribution function, G(e), in failure strain for a
laminate segment of length 3 is given by the summation of the
failure probabilities of failure sequences no. 1 through no. 75.
When G(e) is given, the cumulative distribution function with
respect to the average stress, G(a) can be obtained from the
relation o = Ece, where Ec is the effective axial Young's modulus of
the hybrid laminate. Finally, the cumulative distribution function,
H(a), for the hybrid composite of length / = N6 is given by the
weakest link theory as follows:

H(a) = l - [l - G(a)]N (5.72)

Two limiting cases are considered. In the case of non-hybrid

composites, FLE( £ ) = FUB( S ), the above results can be reduced to
those of Harlow and Phoenix if the local load sharing (equal load
sharing) rule is used for the strain redistribution. Another case
deserving attention is the situation of high failure probability, for
which the first ply failure may trigger the complete failure of the
laminate. The cumulative distribution function for the first ply
failure strength of a laminate segment is then given by Eq. (5.58)

Table 5.3. Failure sequence and failure probability of the

four-layered hybrid composite

Type Nos. Failure Failure probability

sequence
I 1-24 abflp F(x)}{F(Ku3x) -F(x)}{F{Kl2 u3x)-F{KX3x)}
{F(Kl23 4x) -F(Kl24x)}
II 25-36 ablp F(x){F(Kl2x) -F(x)}{F(K, x)-F(x)}{F(K12,Ax)-

III 37-48 abfp F(x){F(K, ,x) -F(x)}{F(Kl 3x) - F(K, 3x)}

{F(Kl2 4x) - -F(K,4x)}
IV 49-60 aflp F(x)2{F(K,2 r x ) - F(x)}{F(Kl 2 3 4 x ) — F(KV J T ) }
4

V 61-4 abp F(x){F(K, 2*) -F(JC)}{F(K,4 x)-F(JC)}{F(K,4x)-

F{x)}
VI 65-70 afp F(x)2{F(K^ ,;x ) - F(x)}{F(Kl 23x)-F(x)}
VII 71-4 alp F(xf{F(Kl234x)4x) - F(x)}
VII 75 ap F(x)4
 Strength theories 271

with e replaced by a. Fukunaga and colleagues have shown that

G(a) can be expressed on the Weibull probability paper in the
following form:

1n + j31n
ao M L E + MUBE
ln(M L E + MHBEP) (5.73)
where M = MLE + MHE, - = EUB/EL and oo is the scale
parameter.
Numerical results of the hybrid ultimate strength have been
obtained for p = /3 LE = /3 H E , £ H E / £ L E = ^ and £HE = £ L E = 3.
Figure 5.21 shows the median strength for F LE = 0.5 and four
different laminate configurations. It can be seen that the strength of
hybrid laminates with LE layers clustering together (cases A and B
in Fig. 5.21) is lower than that for the situation where the LE layers
are dispersed more evenly among the HE layers. This obviously
results from the higher stress concentration factors in cases A and B
than in cases C and D.
Figure 5.22 shows the median strength variation with VLE for the
case of TV =100. The points A (A') and D (£>'), respectively,
represent the strengths of an all HE fiber composite and an all LE
fiber composite. Line BD (B'D') represents the stress in the

Fig. 5.21. Effects of stacking sequences on the median strength. (After

Fukunaga, Chou and Fukuda 1989.)
0.6 r-
• • O O A
• O O • B
• O • O C
0.5 --
^s^ O • • O D

\ \\\
0.4 --

0.3 - p = io
- P= 5 "^V^-X— C
~^A
0.2 1 1 1
10 102 103 104
N
272 Hybrid composites

hybrid at which failure of the LE fiber takes place. Line AC

{A'C) represents the stress in the hybrid assuming that the LE
fibers carry no load. According to the rule-of-mixtures, line
AED {A'E'D') marks the ultimate failure strength of the hybrid.
Increases in strength above the rule-of-mixtures are known as a
hybrid effect. The present results show that the strength of hybrid
laminates is lower than that of the all LE and all HE fiber
composites. This finding is consistent with the experimental meas-
urements of hybrid laminate strength (Ji 1982). Some other effects
on the strength can also be seen from this figure. These are the
scatter of the lamina strengths, the LE fiber relative volume fraction
and the laminate stacking sequence. The median strength for /? = 10
is greater than that for /J = 5, whereas the hybrid effect is greater
for jS = 5 than that for /3 = 10. The strength varies with the laminate
stacking sequence as well as the LE fiber relative volume fraction.
The symbols in Fig. 5.22 correspond to the four types of fiber
arrangements (A, B> C and D) in Fig. 5.21. It can also be seen that
the hybrid effect is greatest for the case of V^E = 0.5.
The probabilistic analysis of Fukunaga, Chou and Fukuda (1989)
has led to the following conclusions: (1) The hybrid composite

Fig. 5.22. The median strength vs. relative fiber volume fraction. (After
Fukunaga, Chou and Fukuda 1989.)
0.6 r

C, C
0.25 0.50 0.75 1.00
 Softening strips 273

ultimate strength is greater for larger values of shape parameters,

whereas the hybrid effect is greater for smaller values of shape
parameters. (2) The hybrid effect on the probabilistic ultimate
failure strength is most pronounced for VLE = 0.5. (3) Both the
magnitude and scatter of strengths vary with the length of hybrid
composites. (4) The strength is also affected by the lamina stacking
sequences because of the difference in stress redistributions result-
ing from laminar fracture. For a given relative fiber volume
fraction, laminates with the LE layers uniformly dispersed among
the HE layers are stronger than the laminates with the LE layers
clustering together.

5.5 Softening strips

The idea of using softening strips in laminated composites
has been developed for the purpose of modifying the local stress
state and thus enhancing the load-carrying capability of composite
structures. The process of introducing a softening strip involves
replacing the low elongation plies with high elongation layers in
selected regions, thus forming an interlaminated hybrid composite.
Sites of stress concentrations in composites are particularly de-
sirable for the use of softening strips. This concept is demonstrated
below for notched laminates.
Sun and Luo (1985) have used three composites for fabrication of
the hybrid specimens, i.e. AS4/3501 carbon/epoxy by Hercules,
S2/CE9000-9 glass/epoxy by the Ferro Corp., and Scotchply 1002
glass/epoxy by the 3M Co. The base-line carbon/epoxy laminates
have the lay-up of [±457070707=F45°] and [±4570°]s. Each
laminate contains a circular hole at the center. To create the
softening strips, two plies of S2/CE9000-9 glass/epoxy are used to
replace the three 0° carbon/epoxy plies in the [±457070707*45°]
laminate, and one ply of Scotchply glass/epoxy is used to replace
the two 0° carbon/epoxy layers in the [±45°/0°]s laminate. Each
strip is placed along the axial direction in the center of the laminate.
The width of the softening strip is about twice of the hole diameter.
Table 5.4 summarizes the average failure load of all-carbon and
hybrid composites. For both laminate systems, there is a significant
increase in failure load (22-8%) in the notched hybrid system over
the corresponding all-carbon system. However, the unnotched
hybrid system shows lower strength than the corresponding all-
carbon system as expected. The enhancement in failure load is
realized only in the notched specimens. This can be understood
from a stress analysis of the hybrid and non-hybrid laminates.
274 Hybrid composites

Figure 5.23 presents the results of finite element analysis of the

normal stress along the transverse cross-section through the center
of the circular hole. The distance in Fig. 5.23 is measured from the
edge of the hole. There is a drastic reduction of the stress
concentration in the hybridized region due to the presence of the
low modulus material. This is believed to be the reason for the
higher ultimate strength of the hybrid composite. The stress level
outside of the softening strip is elevated.

Table 5.4. Average maximum load. After Sun and Luo (1985).

[±457070707 * 45° [±4570°] s

all-carbon hybrid A all-carbon hybrid A

(kN) (kN) (%)* (kN) (kN) (%)*

unnotched 51.6 41.0 -20. 5 30.2 21.4 -29.2

notched 27.3 35.0 28. 3 15.6 19.1 22.0
A (%)t -47.0 -14.5 -48.2 -10.9

* A(%) = hybrid maximum load/carbon maximum load - 1

t A(%) = notched maximum load/unnotched maximum load - 1

Fig. 5.23. Finite element analysis of axial normal stress distribution along
the transverse section in the notched laminate of [±45o/0°/0o/0°/ =F 45°]
through the center of the circular hole. (After Sun and Luo 1985.)
500r

400

300

Hybrid
200

Carbon
100

I
2 3
Distance from hole center/hole radius
 Mechanical properties 275

The combination of low elongation and high elongation materials

in interlaminated hybrid composites also offers improved fatigue
and crack growth properties. A typical example of this kind is the
aramid aluminum laminate (also known as ARALL) originated in
the Netherlands. (See, for instance, Marissen 1984; Marissen,
Trautmann, Foth and Nowack 1984; Mueller, Prohaska and Davis
1985; Vogelesang and Gunnink 1986; Bucci, Mueller, Schultz and
Prohaska 1987; Kenaga, Doyle and Sun 1987; Chen and Sun 1989.)
ARALL can be considered as a hybrid of aramid fiber laminae
sandwiched in between sheets of high strength aluminum alloy. The
hybrid composite is 15% less dense than the monolithic aluminum
alloy and shows a 100-1000 fold improvement in fatigue life. The
tensile properties of the hybrid are reported to be 15-30% better.
Thus, ARALL is desirable for fatigue dominated sheet applications
such as in lower wing skins, fuselages and tail skins of aircraft.
Other examples of combinations of high elongation and low
elongation sheets of materials can be found in the use of adhesive
layers in controlling free edge delamination and impact damage.
(See, for instance, Chan, Rogers and Aker 1976; Sun and Norman
1988; Sun and Rechak 1988.)

5.6 Mechanical properties

This section discusses briefly the fracture, impact and
fatigue characteristics of hybrid composites. In the case of inter-
laminated hybrid composites the failure mode is the fracture of the
low elongation plies transverse to the loading axis. It has been
observed in carbon/glass hybrids that the main transverse fracture is
typically of cruciform shape and debonding occurs between the
carbon and glass plies outwards from the line where the transverse
crack intersects the carbon-glass interface. Ultimately most of the
hybrid becomes debonded and the strength and stiffness approach
those of the glass plies alone as shown in Fig. 5.10 (Bader and
Manders 1978; Pitkethly and Bader 1987).
It is important to note that the initial fracture in the carbon plies
does not propagate across the glass plies and load can be progres-
sively diffused back into the carbon plies away from the fracture
plane (Section 5.3.2). These processes of multiple cracking in the
low elongation phase and the associated debonding have a sig-
nificant effect on the total work of fracture in hybrid composites.
The extension of debonding decreases as the absolute thickness of
the carbon layers and volume content of carbon fiber decrease. It
also has been suggested that debonding could be inhibited if the
thickness of the carbon layers is made sufficiently thin.
276 Hybrid composites

The subject of work of fracture has received some attention

(McColl and Morley 1977; Kirk, Munro and Beaumont 1978). It has
been pointed out that fracture mechanisms at the microscopic level
in binary fiber composites can be combined to give different kinds
of macroscopic fracture behavior in hybrid composites. Therefore
the fracture of the carbon/glass hybrid system, for instance, is
expected to show characteristics of the individual composite
systems.
The impact resistance of composite materials can be modified
through a broad range of methods. Jang et al. (1989) have reviewed
the major techniques including the control of fiber/matrix interfa-
cial adhesion, matrix modifications, lamination design, through-the-
thickness reinforcements, fiber hybridization, and utilization of
high-strain fibers. Among these approaches, hybrids offer the
benefit of improving the impact resistance of composites based upon
high modulus fibers.
The total strain-energy of a composite at its ultimate tensile
strength is inversely proportional to the fiber tensile modulus. Thus,
high modulus fibers are not desirable for impact resistance from the
viewpoint of energy absorption. The same is true when the work of
fracture or toughness is considered. Aveston and Kelly (1980) have
assessed the effectiveness of hybrid composites in energy absorp-
tion. Their analysis is recapitulated in the following. Consider the
idealized stress-strain curve of Fig. 5.10, and compare the total
energy per unit volume absorbed in a hybrid up to its ultimate
tensile strength with the sum of energies of its components. By
assuming a mean crack spacing of 1.5JC (for crack spacing between x
and 2x), the strain at the limit of multiple cracking is £Lu(l + (f)a)
where a is given in Eq. (5.42). The ultimate failure strain of the
hybrid is 8Uu - (|)AELU, where £Lu and eUu are the ultimate strains
of the LE and HE components, respectively.
Based upon these strain values, the area under the stress-strain
curve OABC of the hybrid composite is
U, = laEcetu + l2e2HuEHEVHB (5.74)
When the HE and LE components are completely debonded, the
stress-strain curve follows the path OAEF. Then the energy
absorbed at ultimate failure is
U2 = hEceln + ^ H E VHEEHU " ^ H E V H E E L (5.75)
Aveston and Kelly have concluded that for Ul > U2y oc should be
greater than three. For the combination of glass and carbon fiber
 Mechanical properties 277

composites, this can be achieved with a carbon fiber volume fraction

of over 30%. Furthermore, the condition of multiple cracking (Eq.
(5.41)) requires that a>7.1 for eUu = 2 3 % and eLu = 0.28%. This
implies YLE = 48%. Thus, toughening through fiber hybridization
and multiple cracking can be accomplished in carbon/glass compos-
ites with carbon fiber volume fraction between 30% and 48%.
Experimental studies of the impact behavior of hybrid composites
can be found in the work of Chamis, Hanson and Serafini (1972),
Beaumont, Riewald and Zweben (1974), Adams (1975), Adams
and Miller (1975, 1976), Dorey, Sidey and Hutchings (1978);
Adams and Zimmerman (1986); and Jang et al. (1989). Several test
methods have been employed to measure the impact resistance of
composite materials. The Charpy impact test and Izod impact test
have been performed mainly on unidirectional composites, whereas
the drop-weight impact test is usually used for laminated compos-
ites. Other tests such as longitudinal impact, transverse impact and
pure shear impact tests on unidirectional fiber composites have also
been considered.
Both notched and unnotched specimens have been used in
Charpy tests. Instrumented Charpy impact tests can be used to
determine the maximum load on the specimen, as well as to
differentiate between the energy required to initiate damage and the
energy absorbed during damage propagation. The relative sizes of
the two regions for failure initiation and propagation under the
load-time curve provide a qualitative measurement of the ductility
of a composite under impact loading. Two materials with the same
total Charpy energy may have quite different proportions of the
component energies and, thus, distinct mechanisms of failure
(Beaumont, Riewald and Zweben 1974). Most composite materials,
especially laminated systems, can dissipate a considerable amount of
energy in the fracture propagation phase even though the initial
fracture may be of a brittle cleavage mode (Adams and Miller
1975). Interlaminated hybrid composites have been found to
increase delamination under impact loading. Jang et al. (1989) have
reported that the impact energies of the interlaminated hybrids
generally show a negative hybrid effect, i.e. slightly lower energy
dissipation than that predicted by the rule-of-mixtures.
As to the fatigue behavior of hybrids, Phillips (1976) has reported
significant improvements in fatigue resistance of glass composites by
hybridization with carbon fibers. At the stress level of 300 MPa or
about 50% of ultimate, the fatigue life of the three-to-one volume
fraction of glass/carbon hybrid is improved by about 100 fold over
278 Hybrid composites

the all-glass control. This probably arises from the increased

stiffness and, hence, the decreased strain for a given stress.
Figure 5.24 shows the stress vs. log life curves for unidirectional
carbon/Kevlar hybrids tested in repeated tension (minimum
stress/maximum stress = 0.1) reported by Fernando et al. (1988).
The stresses in Fig. 5.24 are peak stresses. The <xmax/log Ncurves show
a uniform variation from the linear form of the curve for plain
carbon/epoxy towards the pronounced step-function shape of the
curve for the plain Kevlar-49/epoxy composite. Adam et al. (1989)
have examined a series of carbon/Kevlar-49/epoxy unidirectional
hybrid composites; the fatigue behavior has been established as a
function of composition and the ratio of the minimum to maximum
stress in cyclic tension and tension/compression. This enables them
to represent all data in a single two-parameter fatigue curve.
Other mechanical property data can be found in the literature for
the non-linear tensile stress-strain relation (Takahashi and Chou
1987), compressive behavior (Chou, Steward and Bader 1979; Chou
and Kelly 1980b; Gruber, Overbeeke and Chou 1982; Kretsis 1987;
Yau and Chou 1989), flexural behavior (Fischer and Marom 1987;
Marom and Chen 1987) and shear property (Kretsis 1987).
Fig. 5.24. Stress/log life (omax/\og N) curves for the family of unidirec-
tional carbon/Kevlar hybrids tested in repeated tension (minimum
stress/maximum stress = 0.1). (After Fernando et al 1988.) amax is the
peak stress. The percentages indicate relative fiber volume fractions of
carbon and Kevlar.

2.0r-
100% Carbon

1.5

o-
O 1.0

0.5

0.0
-1
logN
 Property optimization analysis 279

5.7 Property optimization analysis

5.7.1 Constitutive relations

A method is presented in this section to determine the
concentration of components which can simultaneously optimize
certain mechanical, thermal and electrical properties of a hybrid
composite. This analysis, developed by McCullough and Peterson
(1977) is essential in the pursuit of balanced material properties of a
multi-component system.
The basis of this optimization analysis is the assumption that the
constitutive relations for several major properties can be cast into
simple linear form. Although the hybrids under consideration are
restricted to unidirectional composites, the general format of this
treatment is applicable to structural elements with more complex
fiber arrangements. The constitutive equations for estimating the
longitudinal properties of ternary composites are summarized below.
Here, the volume fractions of the components are denoted by V, and
the weight fraction by w. The subscripts 1, 2 and 3 refer to the
components of the ternary system.
Mechanical properties
modulus E = VXEX + V2E2 + V3E3
strength a = (V1E1 + V2E2 + V3E3)e* (5.76)
Poisson's ratio v = Vxvv + V2v2 + V3v3
Thermal properties
coefficient of a = [VlEla 1 + V2E2a2 + V3E3a3]/
expansion [VjE! + V 2 E 2 + V 3 E 3 ] (5 77)
thermal conductivity K = V, Kt + V2K2 + V3K3
Electrical property
1 Vi V2 V3
resistivity - =11 h—
P Pl P2 P3
(5.78)
Weight density d = Vldl + V2d2 + V3d3
Cost
cost/weight C = Clw1 + C2w2 + C3w3 (5-79)
In the strength expression it is assumed that £* = min(e 1 , e2, e3)
and failure occurs when the composite strain reaches the lowest
failure strain of the three components. Clearly, this assumption
280 Hybrid composites

underestimates the longitudinal strength. In the cases of coefficient

of expansion and resistivity, E, a and p are treated as property
variables. Also, w, is the weight fraction of the /th component.
It is often useful to examine the performance of a composite in
terms of certain 'specific' properties (i.e. the property per unit
weight). A specific property P can be related to the property P and
its weight density d as

(5.80)

where wt is defined by the term in the square brackets. The

equations for mechanical, thermal and electrical properties, as well
as density, have volume fraction as the composition variable while
the specific property equations and the cost equation have weight
fraction as the composition variable. Volume fraction and weight
fraction are related by
Wi = dtVi/d (5.81)
Both weight fraction and volume fraction can be used as concentra-
tion variables in the analysis of performance.
A typical property map for ternary systems can be represented by
triangular diagrams as shown in Fig. 5.25. Each side of the
equilateral triangle is unity in length and it expresses the volume
fraction of a component. The volume fraction of a ternary system is
represented by a point inside the triangle. For example, the hybrid
composite denoted by point R in the property map contains 35%,
25% and 40% of matrix, fiber 1 and fiber 2, respectively. These
concentration readings are obtained by drawing lines from point R
parallel to the three sides of the triangle. The volume fractions of
binary systems and pure components are represented, respectively,
by points located on the sides and the apices of the triangle.
A straight line within the triangle of Fig. 5.25 specifies arbitrarily
selected property levels available to the ternary system. Thus,
line DE represents the various ways of combining the three
components to achieve a specified property level, P, of the
composite. Point E, for example, indicates that the combination of
55% volume fraction of matrix, 0% volume fraction of fiber 1 and
 Property optimization analysis 281

45% volume fraction of fiber 2 will result in a composite with

property P. This same property level P can be achieved by
concentration levels represented by all the other points on line
DE. Also shown in Fig. 5.25 is the shaded area defined by the
concentration line for 20% of matrix material. This particular
volume fraction represents a square array of fibers of equal size. By
determining the minimum desirable volume fraction of matrix
material, the region in the triangle representing a matrix content
below the minimal level can be excluded from further consideration
in the optimization process.
Figures 5.26(a) and (b) present the various levels of longitudinal
specific modulus and longitudinal specific strength of the hybrid
system. The lines of constant properties are constructed from the
following property data for the carbon/boron/epoxy system:
Young's modulus in GPa (345/410/3.4); tensile strength in GPa
(2.1/3.1/0.035); density in 103 kg/m3 (1.66/2.71/1.1); critical strain
in % (0.6/0.7/10).

Fig. 5.25. Typical format of a ternary property map. (After McCullough

and Peterson 1977.)

I £1 I I I r
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
v •
' resin
282 Hybrid composites

5.7.2 Graphical illustration of performance optimization

Figure 5.27 illustrates the property map for two different
composite properties P and Q. For clarity only one of the constant
property levels of property P is shown. It can be readily dem-
onstrated that a designer can achieve a specified property of the
hybrid by simultaneously adjusting the component properties.

Fig. 5.26. Selected property maps for the system carbon/boron/epoxy. (a)
specific modulus, (b) Specific strength. (After McCullough and Peterson
1977.)
Specific modulus of carbon/boron composites (unit: 106 m)

Carbon Boron

0.2 0.4 0.6 0.8

Resin

(a)

Specific strength of carbon/boron composites (unit: km)

0.2 0.4 0.6 0.2

Resin
(b)
 Property optimization analysis 283

Any combination of fiber and resin that falls on line P will yield a
composite with property P. Similarly, a line denoted by Qj
represents a specified level of property Q achieved by combinations
of fiber and resin properties. Any combination of components that
falls simultaneously on lines P and Qj will yield a hybrid composite
with properties P and Qj. Consequently, the intersection of lines P
and Qj uniquely determines the concentrations of the three com-
ponents to achieve properties P and g r
Suppose that Q1 < • • • < Qt < • • • < Q7. If a hybrid composite is
desired such that for a specified property P (e.g. modulus), the
property Q (e.g. strength) is a maximum, then the intersection of P
and Q7 should give the proper combination of concentrations of the
components.
It often occurs in the optimization of hybrid composite design
that, instead of a specified property, a level of performance is
required. For instance, a property level P or greater (or P or less
for properties such as density and cost) may be required. Figure 5.28
illustrates superimposed maps for properties P and Q. The require-
ments are
P>P*
(5.82)
Q>Q*
and
Vmatrix>V0

Fig. 5.27. Superimposed property maps for properties 'P' and 'Q\ (After
McCullough and Peterson 1977.)

vy
resin "
284 Hybrid composites

Fig. 5.28. Superimposed property maps illustrating bounding ranges on

performance requirements. (After McCullough and Peterson 1977.)

Q<Q
v •
Kresin

The shaded regions in Fig. 5.28 represent the concentration ranges

which fail to meet any one or all of the above requirements.
Naturally, the concentration ranges in the unshaded area of the
figure will meet or exceed the specified requirements. The op-
timization of hybrid composite performance can thus be carried out
based upon the basic information given in the property maps of the
type shown in Fig. 5.26.
This schematic treatment illustrates the basic notions for a
performance optimization of multicomponent systems. When the
number of components exceeds three, the graphical method is no
longer applicable. McCullough and Peterson (1977) have developed
algebraic relationships of the properties of multicomponent systems
and the optimization procedure has been structured in the form of a
classical linear programming problem.
 Two-dimensional textile structural
composites

6.1 Introduction
The term 'textile structural composites' is used to identify a
class of advanced composites utilizing fiber preforms produced by
textile forming techniques, for structural applications. The recent
interest in textile structural composites stems from the need for
improvements in intra- and interlaminar strength and damage
tolerance, especially in thick-section composites. Textile composites
offer the potential of providing adequate structural integrity as well
as shapeability for near-net-shape manufacturing (Chou and Ko
1989).
Textile structural composites provide the unique capability that
the microstructure of fiber preforms can be designed to meet the
needs of the performance of composite structures. Textile structural
composites can be fabricated directly to their final shapes or can be
assembled or readily machined to specified contours and dimen-
sions. A total system approach is necessary to optimize the
composite performance through the consideration of preform avail-
ability, cost, ease of processing, needs for secondary work such as
machining, joinability of parts, and the overall performance of the
composite structure. Chapters 6 and 7 discuss the fundamental
characteristics of two- and three-dimensional textile preforms and
the analysis of composite behavior based upon these preforms. The
following discussions of yarn assembly, as well as textile preforms
and characteristics, are based upon the review of Scardino (1989).
The forming of textile preforms requires knowledge of the
structure of yarns and fibers. Yarns are linear assemblages of fibers
formed into continuous strands having textile characteristics, i.e.
substantial strength and flexibility. Figure 6.1 illustrates the ideal-
ized models of yarn structures; a yarn may consist of (a) single or
(b) multiple continuous fibers, or (c) short (staple) fibers, where a
substantial amount of twist or entanglement is needed to overcome
fiber slippage. Yarns made from staple fibers are referred to as spun
yarns. Figure 6.1(d) and (e) show that two or more single yarns can
be twisted together to form ply or plied yarns. Plied yarns can be
further twisted to form multiples (f). Spun yarns can also be
286 Two-dimensional textile structural composites

combined to form plied yarns. Advanced textile structural compos-

ites are mostly based upon continuous filament yarns.
The relative density of fiber packing in the yarn cross-section is
quantified by the fiber packing fraction, which is the ratio of fiber
specific volume (volume/mass) to yarn specific volume. Fiber
packing fractions are determined by a number of factors, including
the number of fibers in a yarn, fiber cross-sectional shape, yarn
tension, level of yarn twist and yarn manufacturing method. The
yarn structures determine the translation of fiber properties into
yarn properties. Consider, for example, the axial yarn elastic
modulus (Ey). Hearle, Grosberg and Backer (1969) have predicted
that Ey = cos2 6E{, where E{ is the fiber elastic modulus and 6
denotes the helical angle of the fibers in a yarn. The translation
efficiencies reflect the effect of fiber orientation relative to the yarn
axis due to the twist as well as the fiber entanglement in the yarns.
The efficiency of fiber packing in a yarn and the fiber-to-yarn
strength and modulus translation need to be taken into account in
the selection of yarns for textile preforms. Further discussions on
the packing of fibers in a yarn are given in Chapter 7.
The selection of fiber preforms as reinforcements for composites
requires additional considerations to those at the yarn level. The
most basic ones, according to Scardino, are the manipulative
requirements in dimensional stability, subtle conformability and

Fig. 6.1. Idealized models of various yarn structures. (After Scardino

1989.)
 Textile preforms 287

deep draw shapeability. A high degree of dimensional stability is

required in pultruded, flat panel or laminated composites. Some
conformability is desirable in slightly curved structural parts.
Considerable extensibility of the preform is necessary for deep-draw
molded composites. These factors not only are pertinent to the
selection of composites processing techniques, but also dominate
the fiber preform microstructure in the finished product. It should
be noted that the orientations of fibers in a preform before and after
matrix impregnation can be very different, and can thus have
significant implication on composite performance.

6.2 Textile preforms

The major textile forming techniques for composites rein-
forcement are weaving, knitting, braiding and stitching. There is the
lack of a definitive criterion for separating textile preforms into the
two-dimensional and three-dimensional types. In Chapters 6 and 7,
a rather loose criterion is applied to distinguish these two types,
based upon the degree of integration of the yarns as well as the
extent of strengthening in the thickness direction of the preform.
Consider, for instance, the traditional weaving, knitting and braid-
ing processes; the interaction of yarns (i.e. interlacing, interlooping)
in the thickness direction is limited to two or three yarn diameters.
As a result, the strengthening effect due to yarn penetration,
although higher than that for conventional laminated composites, is
fairly small. Therefore, these preforms are considered to be two
dimensional. On the other hand, the more recently developed
preforms, such as angle-interlock wovens and solid braids, are fully
integrated structures, and there is a significant degree of strength-
ening in the thickness direction. Thus, these preforms can be
categorized as three-dimensional. The foregoing definitions are
independent of the actual dimensions of the preform.
The uncertainty in the separation of two- and three-dimensional
preforms arises when the integration of yarns in the thickness
direction is of limited extent and the resulting strengthening is not
very significant. An example can be found in multiaxial warp knits.
The layers of essentially straight fibers in such a construction are
connected by knitting yarns. The degree of strengthening in the
thickness direction depends on the type of knitting yarns used.
Figure 6.2 summarizes the major manufacturing techniques for
two-dimensional textile preforms. It is feasible to insert laid-in yarns
into the basic knitted or braided fabric given in Fig. 6.2, thus
significantly modifying the directional stability of the fabric. The
288 Two-dimensional textile structural composites

great varieties of fabric geometry so induced are not shown in Fig.

6.2 for the reason of simplicity. A brief discussion of wovens, knits
and braids for reinforcing composites is given below.

6.2.1 Wovens
Woven fabrics, formed on a loom by interlacing two or
more sets of yarns, are essentially two-dimensional constructions.
When two sets of yarns are interlaced at right angles, the Ion-
Fig. 6.2. Manufacturing techniques for two-dimensional textile preforms.

Two-dimensional textile preforms

Weaving Knitting Braiding

I I
Biaxial Triaxial Weft knit Warp knit Circular Flat Figured

Fig. 6.3. Examples of woven fabric patterns: (a) plain weave (ng = 2); (b)
twill weave (n g = 3); (c) four-harness satin (n g = 4); (d) eight-harness satin
n(g = 8). (After Ishikawa and Chou 1982a.)

(a) (b)

Filling
(Weft)

trr Warp

Interlaced
region
(d)
 Textile preforms 289

gitudinal yarns are known as the warp, and the widthwise yarns are
known as the filling or weft. The individual yarns in the warp and
filling directions are also called an end and a pick, respectively.
Figure 6.3 shows examples of orthogonal woven fabrics. According
to Lord and Mohamed (1982) and Schwartz, Rhodes and Mohamed
(1982), the manufacture of woven fabrics based upon high speed
power looms requires four operations or primary motions: (1)
shedding, (2) filling insertion, (3) beat-up, and (4) warp and fabric
control. Following Lord and Mohamed, and Schwartz, Rhodes and
Mohamed, a brief introduction is given for these four motions.
Shedding involves the movement of the warp yarns to provide a
path for the insertion of the weft yarn. One of the techniques of
shedding uses heddle wires which are grouped into several frames,
known as harnesses or shafts (Fig. 6.4). Each harness is operated by
a separate cam; the purpose of the cam is to lift or lower the
harness. As a result of the movement of the harness, the shed is
formed. A cam loom is generally limited to designs repeating on six
or fewer picks (Schwartz, Rhodes and Mohamed 1982). Besides the
cam system, traditional fabrics are often woven on a dobby head
loom. A commercially available dobby mechanism uses a maximum
of about 24 harnesses, and thus allows the control of interlacing 24
different groups of warp yarns. The Jacquard head provides control

Fig. 6.4. Shedding in fabric weaving. (After Lord and Mohamed 1982.)

Warp beam
Cloth roll
290 Two-dimensional textile structural composites

of every individual yarn across the width of the fabric, and it does
not have the limitation of the dobby loom. Thus, the yarn
interlacing possibilities are greatly enhanced and they are only
limited by the number of warp yarns used.
The number of harness frames required for the shedding opera-
tion depends on the type of weave. Two harnesses are used for
weaving plain fabrics (Fig. 6.3a), and their relative motion is carried
out in two weaving cycles. In one cycle, the front harness is in the
top position and the back harness is in the bottom position. In the
next cycle, the harnesses change positions, and the sequence is
repeated. Obviously when the two sheets of warp yarns are at the
same level, the shed is closed (Schwartz, Rhodes and Mohamed
1982).
Filling insertion, as the term implies, involves the passing of a
filling yarn through the open shed. Figure 6.5 shows schematically
the conventional way of filling insertion by a shuttle. In order for
the filling insertion motion to take place, the shed has to be
sufficiently open and remain open for an adequate period of time.
Consequently, the speed of the weaving process is dominated by the
speed at which the shuttle travels through the shed. The transit time
of the shuttle involves its acceleration and deceleration. Further-
more, it is desirable to remove the filling supply package from the
filling carrier so the carrier could be made smaller and the yarn
movement in the shedding is reduced. All these considerations
provide the impetus for using shuttless looms.

Fig. 6.5. Filling yarn insertion in fabric weaving. (After Lord and
Mohamed 1982.)
Quill (pirn) contains filling and is
mounted inside shuttle

»<?
 Textile preforms 291

The commonly used shuttless systems include rapiers, gripper

projectiles, air jets, and water jets. A rapier is a device made of
metal or a composite material with an attachment on the end to
carry the filling yarn through the shed. For the case of a single rigid
rapier, its length should be at least equal to the loom width. In
order to improve the loom speed, double rigid rapiers have been
used. These consist of a 'giver', which picks up the filling yarn and
carries it to the center of the shed, where the filling yarn is
transferred to a 'taker' for transporting the yarn to the other end of
the shed. The reduction in carrier traveling time thus doubles the
number of picks that can be inserted per unit time. The gripper
loom uses a small projectile to transport the filling yarn. It is
feasible to use many projectiles which may be initiated from both
ends or one end of the loom. In the case of fluid jet looms, the
filling yarn is carried by a high pressure air or water jet.
Finally, the purpose of beat-up is for incorporating the filling yarn
into the body of the fabric after the filling is inserted. This is
accomplished by the use of a wire grate called a reed, through
which the warp yarns are threaded. The reed is first moved
backward to allow the insertion of the filling yarn. When the
insertion is finished, the reed moves forward and drives the filling
yarn into the fabric. For a continuous operation of the weaving
process, it is also necessary to supply the warp yarns continuously,
and to remove the fabric from the loom. It should be noted that the
repeated actions of shedding and beating induce cyclic tension
variations in the yarns (Schwartz, Rhodes and Mohamed 1982). The
control of warp and filling yarn tension is essential in the weaving
process.
Orthogonal woven fabrics exhibit good dimensional stability in
the warp and weft directions. Woven fabrics offer the highest yarn
packing density in relation to fabric thickness. The pure and hybrid
woven fabrics used in composites are mostly in the forms of plain,
basket, twill and satin weaves. Wovens are available in tubular and
flat forms.
Woven fabrics provide more balanced properties in the fabric
plane than unidirectional laminae; the bidirectional reforcement in
a single layer of a fabric gives rise to enhanced impact resistance.
The ease of handling and low fabrication cost have made fabrics
attractive for structural applications. On the other hand, the limited
conformability, poor in-plane shear resistance, and reduced yarn-to-
fabric tensile translation efficiency due to yarn crimp are some of
the disadvantages of woven fabrics.
292 Two-dimensional textile structural composites

Triaxial woven fabrics, made from three sets of yarns which

interlace at 60° angles, provide higher isotropy and in-plane shear
rigidity than orthogonal wovens. However, no woven fabric con-
struction provides sufficient extensibility for deep-draw molding
(Scardino 1989).

6.2.2 Knits
A knitted structure is characterized by its interlacing loops.
Two basic types of knits can be defined according to the general
direction of travel of a looped thread in the fabric (Thomas 1971).
In weft knitting, the thread runs widthwise, and the loops are
formed by a single weft thread (Fig. 6.6a). The loops in a horizontal

Fig. 6.6. Knitted fabrics: (a) weft knit structure; (b) warp knit structure.
(After Thomas 1971.) (c) Knitted fabric with weft and warp laid-ins.
(After Wray and Vitols 1982.)

(b)

Straight warps

Straight filling
 Textile preforms 293

row are built up one loop at a time. In practice, many weft threads
are used simultaneously in weft knitting. In warp knitting, the
orientation of a looped thread is warpwise, and all the loops making
up a single horizontal row are formed simultaneously (Fig. 6.6b).
The principal mechanical elements used in knitting are needles.
According to Schwartz, Rhodes and Mohamed (1982), there are
three major needle types: the latch needle, the bearded needle and
the compound needle. The latch needle has been used most often
and it contains a latch which can be closed in the knitting process.
The loops in knitted fabrics are formed essentially on a very
similar principle. Following Thomas (1971), the looping process is
demonstrated for a single latch needle by the consecutive steps
shown in Fig. 6.7. Consider the needle which has at its stem a loop
already formed during the course of the knitting process (Fig. 6.7a).
A thread is then placed under the hook of the needle. The loop is
restrained in its position whereas the needle is allowed to move
through it. As the needle moves downward, the existing loop will
push the latch and close the hook (Fig. 6.7c). When the top of the
hook reaches the level of the existing loop (Fig. 6.7d), this loop is
pulled out of the way by the yarn tension. Then as the needle moves
upward again, the thread in the hook opens up the latch, and it
becomes the next 'existing' loop. More loops are generated as the
process repeats. Depending on the type of knitting machine, a
variety of needle configurations and looping cycles is available. A
detailed discussion of the knitting processes and knit fabrics has
been given by Schwartz, Rhodes and Mohamed (1982).
Simple weft and warp knits can provide extensibility in all

Fig. 6.7. The latch needle cycle in fabric knitting. (After Thomas 1971.)
V
<o

(a)
(a) (e)
294 Two-dimensional textile structural composites

directions and are thus suitable for deep-draw molding techniques.

Directional stability can be established by adding laid-in (non-
knitting) yarns in the desired directions. According to Scardino,
weft inserted warp knits offer flexibility in performance, from
complete dimensional stability to engineered directional elongation.
Furthermore, weft inserted warp knits with laid-in warp systems
offer high yarn-to-fabric translation efficiencies and greater in-plane
shear resistance than comparable wovens. An example of a knitted
fabric with weft and warp laid-ins is shown in Fig. 6.6(c).

6.2.3 Braids
Braided fabrics are constructed from intertwined yarns. In
order to understand the characteristics of braided preforms, it is
useful to review the basic mechanisms involved in maypole
braiders. The paths traced by the carriers of a maypole braider are
similar to those of the dancers around the maypole.
Douglass (1964) has explained the operation of some common
types of maypole braiders. A simple slide plate machine consists of a
deck, a driving mechanism and a superstructure with the take up
facility and the braiding guide. The deck has two metal plates.
Serpent-like tracks are cut in the upper plate. Between the base
plate and the track plate is a train of gears. Each horngear has a
circular flanged top which is slotted to engage the bottom driving
lugs of the spool carriers. Furthermore, the horngears are so
arranged (Fig. 6.8) that the slots in the top flanges of two
neighboring horngears will meet at the intersection of the tracks.

Fig. 6.8. Horngears and tracking in a tubular braider. (After Douglass

1964.)
Base plate

Horngears Track plate

 Textile preforms 295

Consider a carrier with its lug engaged in one of the horngear slots
and moving in the track; the contour of the track enables the carrier
lug to be transferred from one horngear to the next at the
intersection. Consequently, this process can be repeated and each
carrier can follow a chain of interconnected figure eights in a
continuous manner.
According to Douglass (1964), some common types of maypole
braiders are the 'Soutache' braider, circular (tubular) braiders and
flat braiders. The machine used for Soutache braiding is the
simplest of all the braiders, consisting of two horngears which are
slotted to take 3, 5, 7, 9, 11, 13 or 17 carriers. Figure 6.9 shows a
three-carrier soutache set-up for demonstrating the mechanism of
braiding. Circular braiding machines, on the other hand, have an
even number of carriers, starting with eight and increasing by steps
of four.
Braiders for flat products are characterized by the tracking
system, which does not completely encircle the braiding center (Fig.
6.10). The two horngears at the ends of the track have an uneven
number of hornslots. Unlike the circular machines, a yarn carrier in
this case reverses its path at the terminal gears and as a result flat
braids can be accomplished.
The geometric configurations of some two-dimensional braids are
given in Fig. 6.11. Figure 6.11(a) shows the braid with a 2/2
intersection repeating pattern, and it is known as a regular, plain or
standard braid. Figure 6.11(b) gives a diamond or basket braid

Fig. 6.9. Horngears for a three-carrier 'soutache' braider. (After Douglass

1964.)
296 Two-dimensional textile structural composites

which is characterized by a 1/1 intersection repeating pattern.

Figure 6.11(c) shows a regular braid with warp in-laids and it can be
made by either a circular or flat braider. There are certain significant
similarities and differences between woven and braided fabrics.
Both fabrics utilize two sets of yarns; these are the warp and weft
yarns in weaving and the yarns moving in clockwise and counter-
clockwise directions in circular braiding. As far as interlacing
patterns are concerned, they are unlimited in weaving and very
limited in braiding (normally 1/1, 2/2 and 3/3). The angle of

Fig. 6.10. Horngears and tracking in a flat braider. (After Douglass 1964.)

Track plate ^ - ~ - j - ^ Horngears

Base plate

Fig. 6.11. Geometric configurations of (a) flat braid, (b) diamond or

basket braid and (c) flat braid with warp in-laids. (After Du, private
communication, 1990.)

(a) (c)
 Textile preforms 297

interlacing between the two sets of yarns is 90° in orthogonal woven

fabrics and less than 90° in braided fabrics.
The braiding technique is highly versatile and a great variety of
geometric patterns can be produced. This is demonstrated by the
figured (fancy) braids which have more complex cross-sections (for
example, I and T sections) than the traditional braids, or variable
cross-section shapes along the axial direction. Figure 6.12 shows a
flat-cord-flat fabric, and the tracking system employed for producing
such a fabric (Yokoyama et al. 1989). The figured braids are
categorized as two-dimensional fabrics for the reasons stated
earlier. These fabrics could be considered as three-dimensional

Fig. 6.12. (a) A figured braid of flat-cord-flat construction; (b) configura-

tion of the tracking system. (After Yokoyama et al. 1989.) Parts A and B
are fabricated by flat braiding, and part C is fabricated by a tubular
braiding mechanism. The spindles indicated by the open circles move on
section C of the track and spindles represented by solid circles move
through all sections of the track in the sequence (A)-(C)-(B)-(C)-(A).

Fiber bundle
(a)

Spindle

(A)
Table 6.1. Directional behavior of two-dimensional preforms in unjammed configuration
After Scardino (1989).

Preform Directional Directional Substantial Substantial

construction stability conformability directional in-plane shea
extensibility resistance

MD CD BD MD CD BD MD CD BD MD CD
Woven
Biaxial x xx xx
Triaxial xX xX xX x x
Knitted
Weft X X X X X X
Warp X X X X X X

Braided
Circular
(tubular)
Flat

MD = machine direction; CD = crosswise direction; BD = bias direction.

 Textile preforms 299

provided sufficient structural integration and reinforcement are

achieved in the thickness direction.
A braided fabric exhibits dimensional stability under tension
along the 0° orientation if there are axial laid-in yarns, and along
the ±6° directions of the braiding yarns. Without axial laid-in
yarns, the dimension of a braid can be changed by applying tension
in the 0° and 90° directions. Yarn jamming in a fabric can occur
essentially in two ways. First, for a fabric without laid-in yarns,
application of a tensile force along the 0° or 90° direction will stretch
the fabric until it is jammed and no further movement of the yarns
is possible. Second, yarn jamming could occur during fabrication.
This condition is characterized by the situation that the yarn
covering factor = 1, i.e. there is no void space in between the yarns
of a fabric. Thus, the fabric shape cannot be deformed under load.
The first concept concerning yarn jamming is useful when the focus
is on conformability and large deformation of a fabric. The second
concept is important when one is concerned about, for instance, the
maximum fiber volume fraction in a composite, and the control of
the preforming process. Chapter 7 provides examples of yarn
jamming in the fabrication of three-dimensionally braided preforms.
The major limitations in machine-made braids at the present time
are restricted width, diameter, thickness and shape selection.
There are also non-woven fabrics which are essentially sheet
materials composed of randomly oriented fiber segments bonded
together. There is a lack of geometrically defined arrangements in
non-wovens as compared to wovens, knits and braids. The key
methods of fiber bonding in non-wovens are: sticking fibers together
as in fiber mats, entangling the fibers to give frictional interaction,
stitching through the non-woven web with a textile yarn, and
adhesive bonding. Hearle (1989) has given in-depth treatment of
the mechanics of non-woven fabrics.
Scardino (1989) has examined the behavior of various fabric
structural forms under uniaxial and shear stresses in the machine
(0°), crosswise (90°) and bias (±45°) directions. Table 6.1 sum-
marizes the directional characteristics of two-dimensional preforms.
Based upon such knowledge it is feasible to design fabric preforms
with a specific directional behavior while accommodating certain
manipulative requirements in composites manufacturing.
The analysis and modeling of two-dimensional textile structural
composites in this chapter focus on woven preforms. The models so
developed can be extended to treat braided composites. Composites
based upon knitted and non-woven preforms are not considered.
300 Two-dimensional textile structural composites

Sections 6.3 to 6.11 are excerpted from the work of Chou and
Ishikawa (1989).

6.3 Methodology of analysis

The objective of the analysis in Section 6.3 is to model the
thermomechanical behavior of two-dimensional orthogonal woven
fabric composites. The fabrics are composed of two sets of mutually
orthogonal yarns of either the same material (non-hybrid fabrics) or
different materials (hybrid fabrics). Here, the term 'yarns' repre-
sents individual filaments, untwisted fiber bundles, twisted fiber
bundles or rovings.
An orthogonal woven fabric consists of two sets of interlaced
yarns. The length direction of the fabric is known as the warp, and
the width direction is referred to as the filling or weft. The various
types of fabric can be identified by the pattern of repeat of the
interlaced regions, as shown in Fig. 6.3. Two basic geometrical
parameters can be defined to characterize a fabric: nfg denotes that
a warp yarn is interlaced with every nfgth filling yarn and nwg denotes
that afillingyarn is interlaced with every nwgth warp yarn. The present
treatment is confined to the case of /twg = nfg = ng for both hybrid
and non-hybrid fabrics. Fabrics with n g > 4 are known as satin
weaves. As defined by their ng values, the fabrics in Fig. 6.3 are
termed plain weave (ng = 2), twill weave (ng = 3), four-harness satin
(ftg = 4), and eight-harness satin (rcg = 8). The regions in Fig. 6.3
enclosed by the dotted lines define the 'unit cells' or the basic
repeating regions for the different weaving patterns. It is also noted
that the top sides of the fabrics in Fig. 6.3 are dominated by the filling
yarns, whereas the reverse sides are dominated by the warp yarns.
The theoretical basis of the present analysis is the classical
laminated plate theory, which is given in Chapter 2. Only the key
equations are recapitulated in the following for ease of reference.
Under the assumptions of the Kirchhoff hypothesis, the constitutive
equations are expressed in the condensed form as

Here, N and M are membrane stress resultants and moment

resultants, respectively; e° and K are the strain and curvature of the
laminate geometric mid-plane, respectively. The components of the
 Methodology of analysis 301

stiffness matrices A, B and D are evaluated as follows:

(Ay, BiJ} Di}) = i f * (1, z, z2)(Qtj)k dz (i, / = 1, 2, 6)

(6.2)
where the reduced stiffness constants Qti corresponding to the
lamina defined by hk and hk_1 in the thickness direction are used in
the calculations. The subscripts 1, 2 and 6 in Eq. (6.2) indicate, in
the xyz coordinate system, the x direction, the y direction, and the
x-y plane, respectively. More explicitly, Eq. (6.2) can be written as

A
a=

(6.3)

The inverted form of Eq. (6.1) is given by

(6.4)
lici B'
When the effect of temperature change is taken into account, the
constitutive relation of Eq. (6.1) should be written as
N
(6.5)

where

Gn G12 Gw
Gl2 G22 226 (6.6)

'On G12 Qi6

G12 G22 G26 z dz (6.7)
LGl6 G26 Qte] *xyj k

AT indicates a small uniform temperature change, and a denotes

the thermal expansion coefficients. After inversion, Eq. (6.5)
302 Two-dimensional textile structural composites

becomes
fe°l [A' I B'l(N) (A1)

where
(A1') XA' I R ' K A^
(6.9)
The constants Ar and B' represent, respectively, the in-plane
thermal expansion and thermal bending coefficients.
Based upon the iso-stress and iso-strain assumptions, the above
constitutive equations can be used to obtain the bounds of the
thermoelastic properties. The upper bounds of compliance con-
stants are obtained from the iso-stress assumption; the lower
bounds of stiffness constants are then obtained by inverting the
compliance constant matrix. Similarly, the upper bounds of stiffness
constants are derived from the iso-strain assumption; the lower
bounds of compliance constants are then obtained by inverting the
stiffness constant matrix. Three techniques for modeling the
stiffness and strength properties of fabric composites are introduced
in Sections 6.4-6.6 based upon the laminated plate analysis. They
are known as the 'mosaic model', 'crimp (fiber undulation) model',
and 'bridging model'. The prediction of thermal expansion
coefficients of fabric composites is given in Section 6.8 based upon
these three models. The analytical techniques so developed are also
applied to hybrid fabric composites (Sections 6.9 and 6.10). Finally,
the thermoelastic behavior of two-dimensional textile structural
composites reinforced with triaxial fabrics is presented in Section
6.11. The following discussions are based on the work of Ishikawa
and Chou (1982a-c, 1983a-d), Ishikawa (1981), Chou (1985, 1986,
1989a&b), Yang and Chou (1986, 1987) and Byun and Chou (1989).

6.4 Mosaic model

The basis of idealization of the 'mosaic model' can be seen
from Fig. 6.13. Figure 6.13(a) is a cross-sectional view of an
eight-harness satin. The consolidation of the fabric with a matrix
material is depicted in Fig. 6.13(b), which can be simplified as the
mosaic model of Fig. 6.13(c). The key simplification of the mosaic
model is the omission of the fiber continuity and crimp (undulation)
that exist in an actual fabric.
In general, a fabric composite idealized by the mosaic model can
be regarded as an assemblage of pieces of asymmetric cross-ply
 Mosaic model 303

laminates. Figure 6.14(a) shows the mosaic model of a unit cell for
an eight-harness satin composite. The elastic stiffness constants of a
cross-ply laminate (Fig. 6.14b) can be derived on the basis of Eqs.
(6.3). Assuming that fibers are aligned along the x direction, the
stiffness constants, Qijy of a unidirectional lamina, which has

Fig. 6.13. Idealization of the mosaic model, (a) Cross-sectional view of a

woven fabric before resin impregnation; (b) woven fabric composite; (c)
idealization of the mosaic model. (After Ishikawa and Chou 1983b.)

(a)

(b)

(c)

Fig. 6.14. The mosaic model, (a) Repeating region in an eight-harness

satin composite; (b) a basic cross-ply laminate; (c) parallel model; (d)
series model. (After Ishikawa and Chou 1983b.)

a /~ ////

y ////// /
/ / ///// /(//)
///// A///
± A/// (a)
h , ^ x

::::::::::.W (b)

(d)
304 Two-dimensional textile structural composites

orthotropic symmetry in the x-y plane, are given by

EJDV vl2E22/Dv 0
vv2iEn/Dv E22/Dv 0 (6.10)
0 0
where
Dv = 1-v12v2i (6.11)
Here, En and E22 are the Young's moduli, G12 is the in-plane shear
modulus, and v12 denotes the Poisson's ratio relating the transverse
strain in the x2 direction and the applied strain in the xx direction.
The Qij constants are symmetrical, i.e. Qtj = Qn (see Chapter 2).
From Eqs. (6.3) and (6.10), the elastic stiffness constants of the
cross-ply laminate shown in Fig. 6.14(b) can be derived. The
laminate is composed of two unidirectional laminae of thickness
h/2. The total laminate thickness is h and the x-y coordinate plane
is positioned at the geometrical mid-plane of the laminate. Thus, in
Eqs. (6.3), k = 1 and 2 define, respectively, the laminae with fibers
in the y and x directions. The non-vanishing stiffness constants are
Axl= A22 = (Eu + E22)h/(2DV)
A12=v12E22h/Dv
A66 = G12h
fin = ~B22 = (En - E22)h2/(SDV) (6.12)
3
A i = D22 = (Eu + E22)h /(24Dv)
D1 2 =vl 2 E 22 h 3 /(12Dy)

The extension-bending coupling constants Bn and B 22 do not vanish

because Eu¥ zE22- Also, it is understood that Aijy Bijy and Ay are
symmetrical constants.
Using Eqs. (6.12), Eq. (6.1) can be written in the following
explicit form:
An Al2 0 Bn 0
Ny \ = An An 0 0 -Bn
_0 0 .4 « 0 0
(6.13)
Mx] 'Bn 0 0 l 2 0
My\ = 0 -Bn 0 A, 0
Mxv\ 0 0 0 0 D'66
( J
 Mosaic model 305

Inverting Eqs. (6.13), the following are obtained:

~A'n A[2 0
A[2 A'n 0 Ny
. 0 0 4 J

B'n B[2 0 Mx
+ -B[2 -B'u 0 Mv
0 0 0
(6.14)
B'n -B[2 0 Nr
-B[2 -B'n 0
0 0 0
D'n D[2 0
+ D[2 D'n 0
. 0 0 D'66J Mr
In the bound approach, the two-dimensional extent of the fabric
composite plate is simplified by considering two one-dimensional
models where the pieces of cross-ply laminates are either in parallel
or in series as shown in Figs. 6.14(c) and (d). In the parallel model,
a uniform state of strain, e°, and curvature, KY in the laminate
midplane is assumed as a first approximation. For the one-
dimensional repeating region of length n^a, where a denotes the
yarn width, an average membrane stress, Nx, is defined as

NXdy
nRa JQ

f B Kxx) dyl
-1

= (Ane°x + Al2e°yy) + — [aBTu + (nsa - a)

nea

(6.15)

The factor (1 — 2/ng) appears because the terms Bu for the

interlaced region (#n) and non-interlaced region (Bh) have
opposite signs, namely, Bjx = -B\x. It is noted that B\x is derived
for a cross-ply with the same configuration as in Fig. 6.14(b), where
306 Two-dimensional textile structural composites

the upper surface (z > 0) shows fibers in the x direction. Bjx is for a
cross-ply obtained by exchanging the positions of the two laminae in
Fig. 6.14(b). Other average stress resultants can be written similar
to Eq. (6.15) for uniform mid-plane strain, e°, and curvature, K.
The moment resultant, Mxy for example, is
1 r&a
Mx =— M xdy
nga J0

= DUKXX + D12Kyy + ( l - )B\1S°XX (6.16)

Let Ay, Bijy and Dlj be the stiffness constant matrices relating the
average stress resultant N and moment resultant M with e° and K.
Then

^ (6-17)

Ay = Dtj
These components provide upper bounds for the stiffness constants
of the fabric composite based upon the one-dimensional model. If
these stiffness constants are inverted, lower bounds of the elastic
compliance constants can be obtained. All the elastic stiffness
constants A, B and D are computed using the basic laminate where
the top layer is composed of the filling yarn (Fig. 6.14b).
In the series model, the disturbance of stress and strain near the
interface of the interlaced region is neglected. Let the model be
subjected to a uniform in-plane force, Nx, in the longitudinal
direction. The assumption of constant stress leads to the definition
of an average curvature. For instance, the average curvature, kxx,
along the x direction is
1 r«g«
KXX = KX dx
nga Jo
=— [fB' nNxdx+r B'nNxdx]
ngga Uo Ja J
1

nga

= (1--)B[\NX (6.18)
\ flo
 Mosaic model 307

It is also understood that the terms Bfn for the interlaced region
(B[J) and non-interlaced region (B[\) are equal and opposite in
sign. Other average curvature and mid-plane strain expressions can
be written similar to Eq. (6.18) for uniformly applied N and M. Let
A[j9 Bljy and D\j be the compliance constant matrices relating the
average mid-plane strain, e°, and curvature, R, with the stress
resultant, N, and moment resultant, M. Thus

(6.19)

Equations (6.19) give the upper bounds for the composite com-
pliance constants and, after inversion, the lower bounds for the
stiffness constants.
In summary, both upper and lower bounds for the elastic stiffness
and compliance constants can be obtained from the mosaic model.
Numerical results demonstrating the relationship between these
bounds and l/n g are shown in Fig. 6.15 for Au and A'n and in Fig.
6.16 for B'n. The material properties of a carbon/epoxy composite
given in Table 6.2, with fiber volume fraction in the impregnated
yarn of 60%, are adopted in the calculations. Bidirectional fiber

Fig. 6.15. Variations of A'n and Au with l/n g . (After Ishikawa and Chou
1983b.)
0.15 r - i 30.0
UB
\
\
0.1 =.
\ UB = 20.0
B
<2 a.
•-«. LB
— - — . ^ ^ LB
0.05 - = 10.0
Ah
Au

0.0 1 1 1 1 0.0
0.0 0.1 0.2 0.3 0.4 0.5
Ullr,
308 Two-dimensional textile structural composites

composites are represented by the limiting case of lMg—»0 (« g— »°°)

and the upper and lower bounds of the elastic constants coincide
with each other. Plain weaves are represented by the case of
l/n g = 0.5. The coupling effects for plain weave composites vanish,
as can be seen from Eqs. (6.17) and (6.19), and both the upper and
lower bounds of B-j (fi/y) are identical, i.e. zero. However, the
bounds of AVl do not coincide for plain weave composites.

6.5 Crimp (fiber undulation) model

The crimp model is developed in order to consider the
continuity and undulations of fibers in a fabric composite. Although
the formulation of the problem developed in the following is valid
for all ng values, the crimp model is particularly suited for fabrics
with low ng values. The crimp model also provides the basis of
analysis for the bridging model (Section 6.6).
Figure 6.17 depicts the geometry of the model where the
undulation shape is defined by the parameters hx{x), h2(x), and an.
The parameters ao= (a — au)/2 and a2 = (a + a u)/2 are automati-
cally determined by specifying au, which is geometrically arbitrary
in the range from 0 to a. Because a pure matrix region appears in

Fig. 6.16. Variations of the average coupling compliance with l/n g . (After
Ishikawa and Chou 1983b.)

- 0 . 6 i-

-0.5

„ -0.4
I
UB
z
' -0.3

ICQ
-0.2
LB
-0.1

0.0
0.0 0.1 0.2 0.3 0.4 0.5
!/«„
Material properties of unidirectional laminae

Fiber £n(GPa) £22(GPa) G12(GPa) eb2 Thickness 7°C)

volume (mm)
fraction in
impregnated
yarns

xy1 60% 113 8.82 4.46 0.3 — 0.4

65% 132 9.31 9.31 0.28 — 0.4 -25.0 2
ster2 60% 47.5 15.9 6.23 0.27 0.38% 0.4
mide3 50% 41.2 15.7 15.7 0.3 0.5% 0.244
xy4 65% 85.3 5.5 2.54 0.4 — 0.4 3

a, Koyama and Kobayashi (1977), Ishikawa (1981), Ishikawa and Chou (1983b). (2) Kimpara, Hamamoto an
Ishikawa and Chou (1982b). (4) Chou and Ishikawa (1989).
310 Two-dimensional textile structural composites

the model, the 'overall' fiber volume fraction, V{, can be different
from that in the yarn region.
To simulate the actual configuration, the following form of crimp
is assumed for the filling:
0 (0<x<ao)

hi(x) =
ht/2 (ai 2 <x<n g a/2)
(6.20)

The sectional shape of the warp yarn is expressed by

{ ht/2 (0<x<ao)

h2(x)={
( H ) ^ 4 <°°*
x ai2)
*

(a2<x<nial2)
(6.21)

It is assumed that the laminated plate theory is applicable to each

infinitesimal piece of the model along the x axis. Thus, Ay, B^, and

Fig. 6.17. Fiber crimp model. (After Ishikawa and Chou 1982b.)
 Crimp (fiber undulation) model 311

;> are expressed as functions of x (0 < x < a/2) by

rhx(x)-h,l2

Aij()\ij(x)=\ Qfdz
J-h/2

fh2(x) M2

Q7dz+\ Qfdz

= Qf[h1(x)-h2(x) + h-hJ2]
+ Q%0)hJ2 + Qf[h2(x) - hx(x)] (6.22)
Btj(x) = Mjid^ix) - ht/4]ht + \Q%[h2{x) -hx(x)]ht
Dtj{x) = \Qf{[hx(x) - ht/2f - h\(x) + /i3/4}

+ 3htf(x)/2] + hQ%[h\{x) - h\(x)\

where superscripts F, W and M signify the filling yarn, warp yarn
and matrix, respectively. Similar expressions can be written for
a /2 < x <n g a/2.
The local stiffness of the filling yarn, Qfj(O), in the above
equations is calculated as a function of the local off-axis angle, 0(x),
which is defined as

(6.23)

Consider afillingyarn composed of parallel fibers. Thefiberdirection

is denoted as the 1 direction; the 2 and 3 directions are perpendicu-
lar to the fiber and they define the transversely isotropic plane.
Then, from the Young's moduli (Eu, E22= E33), shear moduli
(G12 = G13, G23) and Poisson's ratio (v12) of the filling yarn, the
elastic constants of the filling yarn with respect to the xyz axes in
Fig. 6.17 can be defined (Lekhnitskii 1963). Here, the angle
between the 1 and x axes is 6:

1 cos /1 2v 21 \ ,2 „ . 2, „ sin4 8
+ b ^ - — cos 6sin 0 i s
XX(O) En \ G U £22' E22

2 -
^ \ - (6.24)
1 cos2 8 sin2 6
= 1
G12 G23
:
v21 cos2 6 + v32 sin2 0
312 Two-dimensional textile structural composites

It is also understood from the assumption of transverse isotropy of

the filling yarn that v12 = Vi3, En/vu = E22/v2i> v23 = v32, and
G23 = £22/2(1 + v23).
Thus, the local stiffness constants of the undulated portion of the
filling yarn, referring to the xyz coordinate axes, are given as
functions of the fiber orientation angle 6
EFX(6)/Dy EFx(d)vFyx(6)/Dv 0
E xx(8)vFx(8)/Dv
F
EFyy(8)lDv 0
0 0 GFy(6

(i,y = l,2, 6) (6.25)

where
Dv = 1- (vFyx(0))2EFx(0)/EFyy(d) (6.26)
By substituting Eq. (6.25) into Eqs. (6.22), the local plate stiffness
constants can be evaluated. The local compliance constants, Afij(x)y
Bfij{x)y and D,y(x) are then obtained by inverting the stiffness
constants Ay(*)> Btj{x)y and Dtj{x).
Define the average in-plane compliance of the model under a
uniformly applied in-plane stress resultant by

Af = — I"" AyWdx (6.27)

nga Jo
where the superscript C signifies the crimp model. Since AyOO is a
constant within the straight yarn portion of Fig. 6.17, Eq. (6.27) can
be rewritten as
/ 2a \ 2 C2
A = A li + A
'f V~nJ ' n~a) '"^x)dx < 6 - 28 )
where Ay m the first term on the right-hand side of Eq. (6.28)
denotes the compliance of the straight portion of the yarns, namely
a cross-ply laminate, and is independent of x. The other average
compliance coefficients B'f and D'f are obtained in a similar
manner.

B\f = (l - - W + — f Blj(x) dx (6.29)

f^) (6.30)
na L
 Crimp (fiber undulation) model 313

In the case of ng = 2, B'f vanishes because B[j(x) is an odd function

with respect to x = a/2, the center of undulation, due to the
assumed form of hx(x). Furthermore, Eqs. (6.28)-(6.30) coincide
with the upper bounds of the compliance of Eqs. (6.19) as an tends
to zero. The integrations in Eqs. (6.28)-(6.30) are conducted
numerically because of the complexity of the integrands. The final
results of the average elastic stiffness, Afj9 Bft and Dfjy for the entire
strip can be reached by the inversion of A'f, B\f and D'f. If this
procedure is applied in the warp direction, balanced properties such
as /in = A22 can be realized.
Numerical results demonstrating the relationship between the
in-plane stiffness, An, and l/n g are given in Fig. 6.18 based upon
the unidirectional lamina properties of a carbon/epoxy system
(Table 6.2). In Fig. 6.18, UB and LB represent, respectively, the
results of the upper and lower bound predictions of the mosaic
model; CM denotes the crimp model; circles indicate finite element
results. Figure 6.18 demonstrates the reduction in An due to fiber
undulation, and the reduction is most severe in plain weave
(l/rtg = 0.5) as compared to cross-ply laminates (l/n g = 0).
The relationship between the coupling compliance B'n and l/n g is

Fig. 6.18. A^ vs. \/ns for carbon/epoxy composites, Vf = 60%. Finite

element results are indicated by (O) for the mosaic model and by (•) for
the crimp model. mosaic model; crimp model. (After
Ishikawa and Chou 1982b.)

0.
s

0.0 0.1 0.2 0.3 0.4 0.5

314 Two-dimensional textile structural composites

demonstrated in Fig. 6.16. The results from the crimp model

coincide exactly with those of the upper bound predictions. This is
due to the fact that the second term on the right-hand side of Eq.
(6.29) vanishes due to the assumed asymmetrical shape of fiber
undulation and, hence, the odd function representation of B'tj with
respect to x — a12,

6.6 Bridging model and experimental confirmation

The crimp model which is based upon a single fiber yarn
has led to the concept of a bridging model for general satin
composites. Such a model is desirable because the interlaced
regions in a satin weave are often separated from one another. The
hexagonal shape of the repeating unit in a satin weave, as shown in
Fig. 6.19, is modified to a square shape (Fig. 6.19b) for simplicity of
calculation. A schematic view of the bridging model is shown in Fig.
6.19(c) for a repeating unit which consists of the interlaced region
and its surrounding areas. This model is valid only for satin weaves
where n g >4. The four regions labeled I, II, IV and V consist of

Fig. 6.19. Concept of the bridging model: (a) shape of the repeating unit
of eight-harness satin; (b) modified shape for the repeating unit; (c)
idealization for the bridging model. (After Ishikawa and Chou 1982b.)

i
i

2a
HI 3a "

/
-2a-
(a) (b)
 Bridging model and experimental confirmation 315

straight filling yarns, and hence can be regarded as pieces of

cross-ply laminates of thickness ht. Region III has an interlaced
structure with an undulatedfillingyarn. Although the undulation and
continuity in the warp yarns are ignored in this model, their effect is
expected to be small because the applied load is assumed to be in
the filling direction.
The in-plane stiffness in region III, where « g = 2, has been
derived in Section 6.5 and has been found to be lower than that of a
cross-ply laminate. Therefore, regions II and IV carry higher loads
than region III; all three of these regions act as bridges for load
transfer between regions I and V. It is also assumed that regions II,
III and IV have the same average mid-plane strain and curvature.
Then, the average stiffness constants for the regions II, III and IV
are

(6.31)

Afj and Dfj for the undulated portion III in Fig. 6.19 are obtained
from Af and Df of Eqs. (6.28) and (6.30), and Bf = Q. Aijy Bijy
and Dtj in Eqs. (6.31) for the cross-ply laminates of regions II and
IV in Fig. 6.19 are given in Eqs. (6.12).
It is also postulated that the total in-plane force carried by regions
II, III and IV is equal to that by region I or V. Then, the
following average compliance constants are derived:

(6.32)

-,s_ 1
iJ
~7n~g
where A'ijy B'ijy and D[j are determined by inverting Eqs. (6.31) and
the superscript S denotes properties of the entire satin plane.
Finally, Afjy Bfj and Dfj can be obtained by inverting Eqs. (6.32).
316 Two-dimensional textile structural composites

The fiber crimp model is effective for plain weave composites

whereas the bridging model is valid for satin weave composites.
This is because there are no straight yarn regions surrounding an
interlaced region in the plain weave. Therefore, no bridging effect is
expected in plain weave composites, and the analysis based on
the fiber undulation model provides a reasonable prediction of the
behavior of plain weave composites.
Numerical results for the relationship between the in-plane elastic
stiffness A\x and l/n g are indicated in Fig. 6.20, also using the
unidirectional laminar properties of Table 6.2. A prediction by the
present theory agrees with experimental results (Ishikawa and Chou
1982b). It should be noted that the overall fiber volume fraction of a
fabric composite is slightly less than that of the impregnated yarns
due to the resin rich region in the vicinity of the undulation. For
instance, for a fiber volume fraction of 65%, the average overall
fiber volume fraction in a repeating unit (Fig. 6.19) for ng = S,
ht = h, and au = a is around 62%.
Ishikawa, Matsushima, Hayashi and Chou (1985) have conducted
experimental verifications of the analytical models for elastic moduli

Fig. 6.20. Asu vs. l/n g for carbon/epoxy composites, Vf = 65%.

Upper and lower bounds; bridging model solution; (A, • ) ex-
perimental results for a cross-ply laminate and eight-harness satin,
respectively. (After Ishikawa and Chou 1982b.)
 Bridging model and experimental confirmation 317

of fabric composites. The experimental materials used include plain

weave and eight-harness satin fabric reinforced composites of
carbon/epoxy. Ply numbers are 1, 4, 8 and 20 for plain weave
fabrics and 2 for eight-harness satin. Yarn orientations are [0°/90°],
[157-75°], [307-60°] and [±45°], as defined by the angles between
the loading axis and the yarn direction.
Experimental and theoretical results are compared in Figs.
6.21-6.23. Figure 6.21 presents results of the in-plane stiffness, An,
non-dimensionalized by the corresponding An of the cross-ply
laminate as a function of l/n g . Experimental results of four-ply plain
weave and two-ply 8 harness weave composites are given. The
symbol $ signifies both the averaged value indicated by the solid
circle, and the scattering indicated by the horizontal bars. Theoreti-
cal predictions of the bridging model (BM) are adopted for n g > 4
and the crimp model (CM) for A > ng > 2 according to the reason
stated earlier. Abbreviations LWC and LWA denote, respectively,
the limiting cases where local warping is completely constrained and
allowed. Also, UB and LB are, respectively, upper bound and
lower bound predictions of the mosaic model.
A good correlation between theory and experiments can be
observed for eight-harness satin composites. The experimental data

Fig. 6.21. Relationships between non-dimensionalized in-plane stiffness

and l/wg; $ experiments. (After Ishikawa et al. 1985.)

UB
1.0
^•xv W"""~"~ *"••"•"

\
\ '^/ BM BM

;
/

L ^^^ LB

LWC
LWA

0.0 1 1 1 1
0.0 0.5
!/«„
318 Two-dimensional textile structural composites

lie in between the LMC and LMA predictions. These results suggest
good predictability of the theory based upon the bridging model for
satin weave composites. There exists a significant discrepancy
between the LWC and LWA curves of plain weave composites even
though slight improvement is achieved over the simple bound
theory.
Constraint of local warping is another factor governing the
in-plane modulus. Neighboring layers in a fabric laminate tend to
suppress the warping of one another. Thus, a dependence of elastic
moduli on ply number appears for plain weave composites. This
effect is demonstrated in Fig. 6.22 where experimental results of
specimens of four different ply numbers are indicated. The in-plane
stiffness Au of the fabric composite is non-dimensionalized by An
of the cross-ply. Small variations in the theoretical predictions are
caused by the scattering of the measured hi a. The in-plane modulus
increases from the value for one-ply, which is slightly higher than
the LWA prediction, and reaches values slightly lower than the
LWC prediction.
The in-plane off-axis elastic moduli results are presented in Fig.
6.23. The off-axis behavior is symmetric with respect to 0 = 45°
because it is assumed that the elastic properties in both the filling
and warp directions are identical.
In summary, the experimental results of eight-harness satin
composites coincide very well with theoretical predictions. There is

Fig. 6.22. Dependence of in-plane stiffness on ply number in plain weave

composites: LWC; LWA; ^ experiments. (After Ishikawa
etal 1985.)

1.0 -

I I
-

I I

0.5 -
—- — —•• — - — '"

1 1 1 l l l 1 1 1 1 I
10 20
N
 Analysis of the knee behavior 319

still a discrepancy in the predictions of elastic moduli of plain weave

composites based upon two limiting cases: local warping completely
prohibited or allowed. All on-axis measured moduli fall in between
the two predictions.

6.7 Analysis of the knee behavior and summary of stiffness

and strength modeling
Both the crimp model and bridging model described above
are now extended to the study of the stress-strain behavior of
woven fabric composites after initial fiber failure, known as the
knee phenomenon. The essential experimental fact for the knee
phenomenon is that the breaking strain in the transverse layer, e\,
is much smaller than that of the longitudinal layer in cross-ply
laminates. Only the failure of the transverse yarns, which occurs in
the warp direction in the present model, is considered. Thus, a
failure criterion based upon maximum strain is adopted.
In the following, the crimp model is utilized and attention is
confined to the one-dimensional behavior of fabric composites

Fig. 6.23. Off-axis moduli of plain weave and eight-harness satin. E =

axial Young's modulus; G = in-plane shear modulus; LWC; :
LWA. (After Ishikawa et al. 1985.)
Plain Eight-harness satin -
Experiments:
x 2-ply _

60 -

OH
O

Off-axis angle, ^ (degrees)

320 Two-dimensional textile structural composites

under an applied stress resultant Nx. Then Eq. (6.4) is reduced to

" " (6.33)

where Mx is the locally induced moment resultant due to the

application of Nx. By assuming first that no bending deflection by
the coupling effect is allowed along the x axis,

This assumption can be realized only if the fabric composite plate is

symmetrical with respect to its mid-plane. However, in practical
multi-layer fabric composites arranged symmetrically to their mid-
planes, this assumption is expected to be approximately true. From
Eqs. (6.33) and (6.34)
e°xx = A'[xNx (6.35)
where Ann=A[l-B[21/D'n.
The quantity A[x may be referred to as a modified in-plane
compliance and it is a function of x. Since Nx is uniform along the x
direction, A'[x(x) represents a strain distribution before the first
transverse matrix cracking. Figure 6.24 depicts two examples of the
mid-plane strain distribution relative to that at the point x = 0 in
Fig. 6.17 and for au = a. It can easily be seen that the fiber
undulation causes local softening and that the maximum strain
appears at the center of undulation (x = a/2). Also, the strain along
the thickness direction at each section is uniform and equal to ex x
owing to the classical plate theory and the absence of bending.
Although the strain distribution calculated from finite element
analysis (Ishikawa and Chou 1983b) deviates slightly from the
assumed uniform distribution, the present idealization provides a
simple method for analyzing the knee phenomenon.
Assume that the region of the highest strain reaches the trans-
verse failure strain e\ first, and the damaged area in the warp yarn
propagates as the load increases. It is further assumed that clas-
sical lamination theory is still valid in this failure process, and that
the effective elastic moduli of such a failed region in the warp yarn
are much lower than those of a sound area and can be expressed as
"gn/100 Qn/100 0
G12/100 Q% 0 (6.36)
0 0 (
 Analysis of the knee behavior 321

Here, Q'™ denotes the reduced stiffness of the warp yarns after
failure, and it is assumed that, with the exception of Q™ 2, the Qtj
components are reduced by a factor of 1/100 to reflect the
weakening effect of transverse cracking. The assumption of the
applicability of the classical lamination theory implies that the
complex stress and strain fields around the failed region are
neglected. Such a successive failure process will continue until the
lowest strain in the region reaches E\. At that time, all the warp
regions have failed completely. Beyond this point, the stress-strain
curve becomes a straight line again until the final failure of the
filling yarns.
Next, consider the case where the restraint on bending is
removed. From the classical lamination theory (Chapter 2)
EXX{Z) = E°XX + ZKXX (6.37)
The strain state under an in-plane stress resultant, Nx, is given by
exx{z) = {A[l + zB'n)Nx (6.38)
Thus, the strain field under the prescribed Nx is determined from
A'n, B'n, and z. Since the strain in a vertical section is distributed
linearly according to Eq. (6.37), it is necessary to determine the
height, h3, where the strain reaches the critical value, E\. If the

Fig. 6.24. Relative strain distribution along the x axis in the fiber crimp
model without bending, au = a; carbon/epoxy; glass/
polyester. (After Ishikawa and Chou 1982b.)

0.1a 0.1a 03a OAa 0.5a

322 Two-dimensional textile structural composites

strain at the outer edge of the warp yarns, £2(^2) according to Eq.
(6.37), is larger than e\, then, for ao<x <a/2,

= h2-(h2-hl) (6.39)
e2(h2) - £ 2 (/i x )

Based upon the h3 value, the plate stiffness in Eqs. (6.22) needs to
be modified after the initial failure. For instance, for a o < x < a / 2 ,
AiJ(x) = Qff[h1(x) - h2(x) + h- hJ2]

(6.40)
Modifications similar to Eq. (6.40) are made for Btj and D,y in Eqs.
(6.22).
Figure 6.25 presents two numerical examples for a glass/polyester
plain weave composite of au = a and overall Vf = 36.8% with and
without bending. The finite element analysis and acoustic emission
results of Kimpara, Hamamoto and Takehana (1977) are also given.
Basic material properties are shown in Table 6.2. The prediction for

Fig. 6.25. Stress-strain curves for plain weave composites of

glass/polyester, Vf=36.8%, and experimental data of acoustic emission;
analytical results for no bending; - - • analytical results for uncon-
strained bending; finite element simulation; ( ) total count in
acoustic emission measurement. The vertical arrow indicates the specified
value of e\. (After Ishikawa and Chou 1982b.)

0.
O

1.0
 Analysis of the knee behavior 323

the case without bending compares very favorably with the finite
element simulation. It is quite reasonable that the case with bending
shows much lower stiffness because it is not subjected to lateral
constraints.
In actual plain weave composites, local bending deformation
caused by the coupling effect in each interlaced region is con-
strained by adjacent regions for which the stiffness constants Btj
have opposite signs. Therefore, as far as plain weave composites are
concerned, one-dimensional analysis for the case without bending
should give a reasonable prediction of the knee behavior under
in-plane loading.
The bridging model and the process of successive warp yarn
failure can be combined to analyze the knee behavior in satin
composites. The approaches for plain weave composites are
adopted here. First, define the stiffness of the fabric composite
under an applied stress resultant Nx without bending to be A*x. It is
noted that A*x = 1/AX1 and the compliance A[x follows the definition
in Eq. (6.35). The average stiffness for regions II, III and IV of
Fig. 6.19 is denoted as A\x and calculated by taking the volume
average:
A\x = (l/V*g)A?iC + (1 - 1/VngMfa (6.41)
Then the compliance of the whole satin composite is calculated
from an average over its length. Define the compliance Axl = 1/A*1.
The assumption of uniformity of Nx along the x direction leads to

where the superscript S indicates satin composites. Finally, the

stiffness of the whole satin composite can be obtained as
At? = l/A"ns (6.42)
It should be noted that the inversion of the compliance and stiffness
constants cannot generally be achieved by merely taking the
reciprocal of the respective components (i.e. Ci} ¥= 1/Sy). However,
under the assumptions made in the derivations of Eqs. (6.33)-
(6.35), Eq. (6.42) is valid for this one-dimensional problem.
Expressions similar to Eqs. (6.41) and (6.42) for the case of
unconstrained bending can also be obtained but are omitted here.
The rest of the procedure for analyzing the knee phenomenon
follows that for the plain weave case. The initial failure of the warp
yarns occurs at the point of highest strain, for example the center of
undulation in the case without bending. Also, since there are
324 Two-dimensional textile structural composites

regions of uniform strain such as the bridging zones in this model,

the entire area of these regions may fail simultaneously, according
to the present assumptions.
Figure 6.26 compares numerical and experimental results for
stress-strain curves of an eight-harness satin glass fabric/polyimide
composite (Table 6.2). Since the test pieces were nearly symmetri-
cal with respect to their mid-planes, the analysis of the case without
bending is selected for comparison; the agreement is quite good,
particularly for strain values up to the knee point. A theoretical
stress-strain curve for a plain weave composite of the same material
is also shown in Fig. 6.26. Here, a knee point is defined by a
deviation of 0.01% in strain from the linear strain. Then, the knee
stress in the eight-harness satin is higher than that of the plain
weave, although knee strains are nearly identical. It can be
concluded that the elastic stiffness and knee stress in satin compos-
ites are higher than those in plain weave composites due to the
presence of the bridging zones.
The following is a summary of the stiffness and strength models
for two-dimensional orthogonal woven fabric composites:
(1) A fabric composite can be idealized as an assemblage of
pieces of asymmetric cross-ply laminates. The upper and

Fig. 6.26. Theoretical and experimental stress-strain curves for glass/

polyimide composites, Vf = 50% in impregnated yarns; bridging
model solution without bending for eight-harness satin (overall Vf =
47.7%); fiber undulation model solution without bending for plain
weave (overall Vf = 40.9%); experimental curve; (•) knee points.
(After Ishikawa and Chou 1982b.)

OH
O

£ 0.1 -

0.0
 Analysis of the knee behavior 325

lower bounds of elastic stiffness and compliance of fabric

composite plates in such a 'mosaic model' are obtained
under the assumption of constant strain and constant stress.
(2) The 'crimp model', which is a one-dimensional approxima-
tion and takes into account fiber continuity and undulation,
is particularly suited for predicting elastic properties of
plain weave composites. The analytical results based upon
the crimp model demonstrate that fiber undulation leads to
a softening in the in-plane stiffness as compared to the
mosaic model. However, fiber undulation has no effect on
the coupling constants. Therefore, the solution of the
coupling compliance based upon the mosaic model is
considered to be reliable.

Both the results of the crimping model and of the mosaic model for
the compliance constants A'n and B'u compare very favorably with
the results of a finite element analysis (Ishikawa and Chou 1983b).

(3) In the case of Dn, Ishikawa and Chou (1983b) have

adopted a transverse shear deformation theory for a modi-
fication of the mosaic model, and examined the response of
a fabric composite plate under both cylindrical bending and
lateral force. Numerical results of Dru based upon the
modified transverse shear deformation theory coincide well
with the finite element results.
(4) The effect of fiber undulation shapes on A[ \ in the crimp
model is shown in Fig. 6.27. The geometrical parameters a
and h are chosen to be 1.0 and 0.4, respectively. The
calculations are performed for the range of ajh values
from 0 to a/h, where the case au—>0 corresponds to the
configuration of a mosaic model. The results show that A[\
is susceptible to the shape of undulation, particularly at
small ng values. The highest A[\ value, i.e. the lowest
in-plane stiffness, is obtained at around ajh = 1. On the
other hand, the A[\ values at ajh = 0 and a/h are not far
apart. Because in actual fabrics ajh — a/h, the mosaic
model (ajh = 0) seems to be effective in evaluating the
in-plane stiffness of a fabric.
(5) The crimp model has been applied to examine the knee
phenomenon of plain weave composites. The predicted
knee behavior of a glass/polyester composite without bend-
326 Two-dimensional textile structural composites

ing shows excellent agreement with the stress-strain

curve obtained by using a finite element analysis.
(6) The bound method based upon the mosaic model is useful
for a rough estimation of fabric composite stiffness prop-
perties. The crimp model offers better predictability than
the mosaic model for the in-plane and bending moduli.
However, the crimp model is inadequate for evaluating the
elastic properties of satin weave composites with large ng.
(7) A bridging model has been developed to examine the
stiffness and strength of general satin composites. The
interlaced regions in a satin fabric are often separated from
one another by the non-interlaced regions. Since the
regions with straight yarns surrounding an interlaced region
have higher in-plane stiffnesses than the latter, they carry
higher loads and play the role of load transferring bridges.
(8) The initial elastic stiffness of satin composites can be
predicted by the bridging model. The analysis of an
eight-harness satin carbon/epoxy composite demonstrates
good agreement with experimental data.

Fig. 6.27. Relationship between average in-plane compliance and undula-

tion length. (After Ishikawa and Chou 1983b.)
0.13

0.12 -

0.11 -

0.10 -

0.09 -

1.0 2.0
ajh
 In-plane thermal expansion and bending coefficients 327

(9) The concept of successive failure of the warp yarns and the
bridging idealization have been combined to study the knee
behavior in satin composites. The theoretical results for an
eight-harness satin glass reinforced polyimide composite
compare favorably with the experimental curve. It can be
concluded that the bridging regions surrounding an inter-
laced region are responsible for the higher stiffness and
knee stress in strain composites than those in plain weave
composites.

6.8 In-plane thermal expansion and thermal bending

coefficients
The constitutive equations of a laminated plate taking into
account the effects of a small uniform temperature change are given
in Eqs. (6.5)-(6.9). In the following, the analytical techniques
developed for the mosaic model, crimp model, and bridging model
are applied to analyze the thermal problem.
First, for applying the mosaic model, a long strip of the fabric
composite (Fig. 6.14a) is again considered. The laminate is free of
externally applied load. The average strains and curvatures of a
one-dimensional strip of width a along the filling or warp direction
due to a uniform temperature change, AT, can be expressed in the
following forms:

8
£°xx = — f ATAfx(x) dx = ATA[
nga Jo
(6.43)
Ky = — f^ &TA' y(y) dy = ATA2
nRa JQ

1 f "e" n —1
kxx= ATB'x(x) dx = A T - 8 B'x
nga Jo ne
(6.44)
1 ff" n — 2
kyy = ATS y) dy = A T § y
n~a I 'y( ^T '
It should be noted that B'x has opposite signs in the regions x = 0-a
and x = a-(ng — \)a\ the same is true for B'y.
Because of the nature of the cross-ply laminates Axy and B'xy
vanish. From Eqs. (6.43) and (6.44), the average thermal expansion
328 Two-dimensional textile structural composites

and thermal bending coefficients for the mosaic model are given by
Ax — s\x> y— y
- I 7\ - I ?\ (6.4o)
B'X=[1--)BX, By=(\--)By
\ nj \ nj
Next, the crimp model is applied; the forms of fiber crimp for the
filling and warp yarns follow the assumed shapes of Eqs. (6.20) and
(6.21), respectively. By assuming no in-plane force and moment and
following the derivations of Eqs. (6.43) and (6.44), the fiber crimp
model gives
= ,c_ / 2au\ -, 2 p -
x
\ noal noa )„
(6.46)

B'xc=(l-l)B'x + ^-\ B'x(x)dx

Here,, the superscript C signifies the crimp model. It is understood

that A'xf and B'x^ vanish for cross-ply constructions. Since B'x and B'y
are odd functions of location with respect to the center of
undulation (Eq. (6.20)) the integration in Eqs. (6.47) vanishes and

(6.48)

The expressions of B' from Eqs. (6.45) and (6.48) are identical.
Thus, fiber crimp has no effect on the thermal bending coefficients.
The same conclusion has been obtained for the extension-bending
coupling constant in Section 6.5.
For the in-plane thermal expansion coefficient, it is necessary to
evaluate the integration in Eqs. (6.46). This is done on the
assumption that the classical laminated plate theory is applicable to
each infinitesimal piece of width dx of the one-dimensional strip
shown in Fig. 6.17. The following steps are taken to obtain A'x(x)
and Ary(y). Consider A'x{x) as an example. First, Ax(x) and Bx(x)
 In-plane thermal expansion and bending coefficients 329

are evaluated from Eqs. (6.6) and (6.7) for 0 < j t < a / 2 , and the
results are
Ax(x) = qf(hx(x) - h2(x) +h- ht/2) +
+ qxv{h2{x)-hl{x)) (6.49)
Bx(x) = iqFx(6){h,{x) -ht/4)ht + \q?(h2{x) -hx{x))h, (6.50)
where the superscripts F, W and M signify the filling yarn, warp
yarn, and matrix region, respectively. Next, from Eq. (6.6),
qx = Qn<Xxx + Q\i<xyy + Q\6<*xy\ q¥ x{Q)y in particular, is determined
from the local stiffness matrix Qfj(d), following the procedures
outlined in Section 6.5. Furthermore, the off-axis thermal expansion
coefficients are given by
a¥ xx(6) = cos2 daFn + sin2 d 8a¥ 22
y ( 6 ) = a¥22 (6.51)

where an and oc22 denote, respectively, thermal expansion

coefficients parallel and transverse to the fiber direction in a
unidirectional fiber composite. Thus, Ax(x) and Bx(x) can be
determined from Eqs. (6.49) and (6.50). Then, the A,'(x) and B[{x)
components are obtained by inverting At{x) and Bt{x) as in Eq.
(6.9).
Numerical integration of Eqs. (6.46) has been conducted and the
results for A'xc and B'xc as functions of l/n g are given in Fig. 6.28.
The balanced thermal property such as Ax = A'y for a fabric
composite can be realized if the above procedure of calculation is
conducted for one-dimensional strips along both the filling and warp
directions.
Lastly, the bridging model is applied to analyze the thermal
properties. It has been noted in Section 6.6 that regions II and
IV of Fig. 6.19 are stiffer than the crimped region III and, hence,
they carry more load when an external force is applied in the x
direction. Regions II, III and IV are termed bridging regions. For
the thermal property analysis, assuming no mechanical loading, the
equilibrium of the bridging regions requires

where the superscript C again denotes the crimped region, and N

and M without superscripts are for the cross-ply laminate. Further-
330 Two-dimensional textile structural composites

more, under the assumption of uniform strain and curvature in the

bridging regions II, III and IV, it is defined that

(6.53)
{KC} = {k}
where the bar denotes the average of the bridging regions.
Substituting Eq. (6.5) into Eq. (6.52), and taking into account
Eqs. (6.53), the results are expressed in the condensed form

n g )-l)|-j) (6.54)
Fig. 6.28. Variation of the thermal deformation coefficients with l//t g for
carbon/epoxy composites, Vf = 60_% and aja = 1.0; A'x\
B'x\ ( • ) experimental results of B' x at 300°K. (After Ishikawa and Chou
1983a.)
14.0 -7.0

12.0 -6.0

-5.0

2.0

ii i i i
0.0 i i

0.0 0.1 0.2 0.3

 In-plane thermal expansion and bending coefficients 331

The quantities on the left-hand side of Eq. (6.54) can be related to

the average elastic stiffness in the bridging regions as

Hence, Eq. (6.54) can be written as

(6 56)
-
Here, A'ijy B-Jy and D-j are obviously obtained by inverting An, Bih
and Dy. In comparison to Eq. (6.8) the quantities in the parentheses
on the right-hand side of Eq. (6.56) can be regarded as the average
v_alues for the bridging regions and hence they are denoted by
A* and B*. Thus, we obtain, in index notation

(6.57)

Finally, the whole satin composite of Fig. 6.19 can be regarded as

a linkage of regions I, II-III-IV and V in series. The average
strain and curvature for the entire model are given in condensed form
as

where the superscript s signifies the properties of the satin compos-

ite, and s° and K denote, respectively, mid-plane strain and
curvature for the cross-plies in regions I and V of Fig. 6.19. From
Eq. (6.58), the components of the thermal expansion and thermal
bending coefficients of the satin composite are expressed as

Figure 6.28 shows the numerical results of the analysis based

upon the elastic properties of Table 6.2. Also, an = 0.0 and
a22 = 3.0 x 10~5/°C. The general characteristics of the variations of
332 Two-dimensional textile structural composites

thermal deformation coefficients with l/n g are very similar to those

of the compliance constants Arn and Bfn as discussed earlier. For
the thermal bending coefficients, there is considerable discrepancy
between the results obtained from the one-dimensional models and
the bridging model.
The geometrical shape of the fiber undulation also affects A'x\ this
is demonstrated in Fig. 6.29 using the carbon/epoxy properties of
Table 6.2. The results indicate that the in-plane thermal expansion
coefficient of satin weave composites is less sensitive to ajh than
that of plain weave composites. Furthermore, the fiber crimp model
predicts a larger effect on A'x due to ajh than the bridging model.
In general, the bridging model predictions are also less sensitive to
the ng values than the crimp model predictions. In both models, the
maximum in A'x occurs at ajh ~ 1.
Experimental data on thermal expansion coefficients of fabric

Fig. 6.29. The effect of fiber undulation on A'x of carbon/epoxy compos-

ites; solid lines: crimp model; broken lines: bridging model. (After
Ishikawa and Chou 1983a.)
13.0

1.0 2.0
ajh
 In-plane thermal expansion and bending coefficients 333

composites are quite limited. Rogers et al. (1977, 1981) and Yates et
al. (1978) have performed measurements of thermal expansion of
carbon fiber reinforced plastics. These experiments, however, are
based upon thick specimens with 15-25 plies. Due to the constraint
of the neighboring layers, an individual ply in the laminate is not
free to bend. As a result, modifications to the analysis developed
above are necessary for making a meaningful comparison with
experiments.
It is assumed that the thermal expansion of a lamina without
bending can be realized if there exist bending moments {M}, under
a temperature change, AT, and no in-plane force is allowed. Thus,
(6.60)
Equations (6.8) and (6.60) give
[Df]{M} + AT{Bf} = 0 (6.61)
Then,
~l{Bf} (6.62)
Substituting Eq. (6.62) into Eq. (6.8), and from the expression of
e°, a modified in-plane thermal expansion coefficient for the case
without in-plane force and external bending can be defined as

{A"} = {A'}-[B'][DTl{B'} (6.63)

Equation (6.63) can be evaluated for the mosaic, crimp and
bridging models provided that the appropriate constants are given
for a particular model. Also, note the presence of elastic com-
pliance constants in Eq. (6.63). Thus, it is necessary to evaluate, for
instance, B'f and D[f for calculating A"s, and B'f and D'f for A"c.
The above modifications are of practical significance because it is
desirable to overcome the anti-symmetrical behavior such as that of
B\ by suitable stacking in laminate constructions.
Figure 6.30 gives the variation of A[ with l/ng. The theoretical
predictions are based upon both the crimp and bridging models
using the thermoelastic properties of the unidirectional carbon/
epoxy composite of Table 6.2. The experimental results of Rogers
et al. (1981) for five-harness satin composites are also shown in
Fig. 6.30. Two estimated values for a/h were used for the
analysis, and a/au is assumed to be unity. The bridging model
prediction coincides fairly well with experimental results. It is also
obvious that the in-plane thermal expansion coefficients are more
334 Two-dimensional textile structural composites

sensitive to ng in the case without bending (Fig. 6.30) than in the

case of unconstrained bending (Fig. 6.28).
In summary, the following can be stated regarding the thermal
property modeling of two-dimensional woven fabric composites:

(1) The mosaic model provides a simple means for estimating

thermal expansion and thermal bending coefficients.
(2) The one-dimensional crimp model predicts slightly higher
in-plane thermal expansion coefficients and the same ther-
mal bending coefficients compared to those obtained from
the mosaic model. The limited experimental data on
thermal bending coefficients coincide rather well with the
predictions of the mosaic and crimp models.

Fig. 6.30. Comparison of theoretical predictions with the experimental

results of Rogers et aL (1981) for five-harness satin carbon/epoxy
composites; a/h = 3.75; a/h =7.5; aja = 1.0; CM and BM
indicate fiber crimp and bridging models, respectively; (•) experimental
results at 300 K. (After Ishikawa and Chou 1983a.)
5.U

7.0 _

^^"^^CM
6.0
/
^ ^ ^ ^ ^ ^ m
/
5.0
/ s
// /s
// ••
4.0 //
- /

/ ^ : > B M

3.0

// ^z~"
V

2.0

1.0 i 1 1 1

0.0 0.1 0.2 0.3 0.4 0.5

!/«„
 Hybrid fabric composites: mosaic model 335

(3) The bridging model is particularly suited for the prediction

of thermal expansion constants for satin composites. The
experimental results on in-plane thermal expansion
coefficients for a five-harness satin composite agree well
with the theory.

6.9 Hybrid fabric composites: mosaic model

Hybrid woven fabrics provide a wide variety of material
selection for designers with a new degree of freedom in tailoring
composites to achieve a better balance of stiffness and strength,
increased elongation to failure, better damage tolerance, and
significant improvement of cost-effectiveness in fabrication. A basic
difference between hybrid and non-hybrid composites is that
material variation as well as geometrical variation come into play
for the former case.
Figure 6.31 shows an example of a hybrid fabric composite for
ng = 8. The front view is dominated by filling yarns and the back
side by warp yarns. There are two kinds of fiber materials, denoted
by a and jS, although there is no restriction regarding the number of
fiber materials in a hybrid fabric. For the case of Fig. 6.31, the
pattern of arrangement of fiber types in the filling direction repeats

Fig. 6.31. A hybrid woven fabric with n g = S, nfm = 2 and n wm = 3. (a)

Front view; (b) back view. (After Ishikawa and Chou 1982a.)

Filling | Warp
1110 9 8 7 6 5 4 3 2 1 3 4 5 6 7 9 10 11

1 1
11
1
1 1

11
\ Jca
\
a thread -H P thread

(a) (b)
336 Two-dimensional textile structural composites

for every two warp yarns; thus it is defined that nfm = 2. In the warp
direction, the pattern of arrangement of fiber types repeats for
every three filling yarns, and rcwm = 3. The subscript 'm' indicates a
material parameter. The following analysis is limited to fabrics
containing only two types of fiber densely woven in both directions,
i.e. no gaps are allowed (Ishikawa and Chou 1982a, 1983d).

6.9.1 Definitions and idealizations

It has been adopted that ng (=nfg = nwg) specifies the fabric
geometrical pattern, and nfm and nwm define the fabric material
arrangements. The notation nm will be used when consideration of
the material parameter is not restricted to any one direction.
In the following, the discussions are first focussed on the pattern
of hybrid fabrics in one dimension, along the filling or warp
direction. Under the assumption of the mosaic model (Section 6.4),
a fabric composite can simply be regarded as an assemblage of
pieces of asymmetrical cross-ply laminates.
If ng and nm are numbers not divisible by each other in a given
direction, warp or filling, the pattern of the hybrid fabric will repeat
in that direction for every ng x nm yarns in the orthogonal fabric.
For instance, for ng = 5 and nfm = 3, Fig. 6.32 shows the pattern of
fabric repeats in the filling direction for every 15 warp yarns. In
general, the pattern of a fabric is repeated in the filling direction
after every nf warp yarns, where nf is the least common multiple
(LCM) of nfg and rcfm, or nf = LCM(nfg, nfm). Similarly, it is defined
that nw = LCM(« ).
Although the size of a basic repeating unit in the filling direction,
for instance, is determined by nf, the detail of fiber arrangement

Fig. 6.32. A fabric where ng and n fm are not divisible by each other in the
filling direction; a and fi denote two types of fibers. (After Ishikawa and
Chou 1982a.)

• Filling

a — y — P — a a a a
-a- a P a a or- a a P a : a : /5 a a P:a :
n = 3

• * — — n g = 5 - »
X flfm
"t — 1.
 Hybrid fabric composites: mosaic model 337

may vary. Figure 6.33 shows two cases of fabric pattern in the filling
direction for nfg = 8. Here, a and /? denote the two types of fiber
material of the hybrid. The notation § is used to indicate that the
filling yarn can be of either a or /? type. It is obvious from Fig. 6.33
that the different repeating patterns are generated by continuously
shifting the positions of the warp yarns in the filling direction. In
general, the number of repeating patterns in the filling direction for
a given nf is equal to the greatest common measure (GCM) of nfg
and nfm and is denoted by n n = GCM(rtfg, nfm). Naturally, nfl = 1 for
the case of Fig. 6.32. Again the notation nx can be used if the
discussion is independent of the direction.
Further comments are necessary for identifying the nature of the
interlaced regions of a hybrid fabric. In Fig. 6.33 the interlaced
region is 'homogeneous' if the yarns are identical or 'heterogeneous'
if the yarns are of different types. The notations HOl a , HES'', etc.
are simply for identification purposes pertaining to later discussions.
The types of interlacing are termed 'mixed' if both homogeneous
and heterogeneous interlacing appear in a repeating pattern. This
can occur when ng and nm are numbers not divisible (Fig. 6.32) or
divisible (for instance ng = 8, nfm = 4) by each other.

Fig. 6.33. Homogeneous and heterogeneous interlacings. (a) n fg = S,

n{m = 2 (n f ^ = n g n = l ) ; (b) /i fg = 8, nim = 4 « m = 3, «f m = l). (After
Ishikawa and Chou 1982a.)

4 = P\a\P\a\p\a\P a\a\p\a\a\eT\j} HO1»

4 —*• a: homogeneous H O l a
4 —*• : heterogeneous
= 4! a\P\a\a\a\P\c, HO2" HE2"

4
4 =a\p\a\p\a\P\a 4! 4: PI a|a|a\P\ oja HO3" HE3P

| — • a: heterogeneous H E l a 4 Ela a: homogeneous

4 — • /3: homogeneous HOI'' | ous p: heterogeneous

(a)
4
ill HEPHOI/3

4 —*• a: heterogeneous
i, . p: homogeneous
(b)
 Fig. 6.34. Two-dimensional basic repeating unit of a hybrid fabric for n g = 8 and n fm = 4. (a) n w
ng and n w m are not divisible by each other, (b) n w m = 4; ng and « w m are divisible by each
(1) mixed interlacing; (2) homogeneous interlacing. (After Ishikawa and Chou, 1982a.)

B
J HO1°
HE1 *
HOP H
HO2"
HE1 * HOI"
HE3/3
HE1"
HO2"
HOI" HO2"

s
•F HOlP I HOI"
HO3"
X (1) Mixed
HO2 a
c
II
HEla H
HO3"
HO3"
HE2I3 HO2 «
*
HOI HOI*
HE1 *
HO3 *
HE3/3
HO2" HO2"

HOl a
HOI*
D H (2) Homogeneous
(a) (b)
 Hybrid fabric composites: mosaic model 339

Next, hybrid fabric patterns in two dimensions are identified on

the basis of the above definitions established for one-dimensional
considerations. Figure 6.34 shows hybrid fabric patterns for ng = 8,
rtfm = 4, and nwm = 3 (Fig. 6.34a) and nwm = 4 (Fig. 6.34b). It should
be noted that all the interlacing patterns of Fig. 6.33(b) appear in
Fig. 6.34(a). The area denoted ABCD in Fig. 6.34(a) is a possible
repeating unit of the fabric. However, there are repetitions in the
geometrical and material patterns within this area. The patterns of
AGIE and IFCH are identical. So are the patterns of GBFI and
EIHD. Consequently, the smallest repeating unit of the fabric in
two dimensions can be represented by either AGHD or EFCD. The
area EFCD, for instance, contains 12 (ngnwm/ (ng/ nfi)) filling yarns.
It can be further concluded that if ng and nm are numbers not
divisible by each other in one direction (filling or warp) there exists
only one kind of basic repeating unit for defining the two-
dimensional fabric. This is true regardless of whether ng and nm are
numbers divisible or not by each other in the other direction.
On the other hand, there exists more than one type of basic
repeating unit in the two-dimensional fabric if ng and nm are
numbers divisible by each other in both directions. This is illus-
trated in Fig. 6.34(b). In Fig. 6.34(bl) both homogeneous and
heterogeneous interlacing in the filling direction occur and the
pattern is considered to be 'mixed' in two dimensions. The pattern
is homogeneous in two dimensions for Fig. 6.34(b2). Two other
mixed patterns exist: [HOI", HE2/S, HO3", HE1"] and [HE1 *
HO2" HO3" HE1"], and no heterogeneous pattern exists for the
geometrical and material parameters given in Fig. 6.34(b).
Let If and /w be the edge lengths of a basic repeating unit in a
two-dimensional fabric. If ng and nwm are numbers not divisible by
each other for the repeating unit AGHD in Fig. 6.34(a).

'w = (nZmCa + nwmCp)ng

(6-64)
lt = n?mCa + nLCp

Ca and Cp denote yarn widths as shown in Fig. 6.31. The area of the
two-dimensional repeating unit is then given by

Ar = ng(nZmCa + nLiCpXnSnC + nLCp) (6.65)

This equation is valid for ng and nm, which are numbers not
divisible by each other in either the filling or the warp direction as
well as in both directions.
340 Two-dimensional textile structural composites

Alternate expressions for Eqs. (6.64) can be given by considering

the repeating unit of EFCD in Fig. 6.34(a):
/w = (nZmCa + nwmCp)nfi
(6.66)
U = (n?mCa + nfmCp)ng/nti
When ng and nm are numbers divisible by each other in both
directions

Ax = — (nZmCa + n^CpXnLCa + nLCp) (6.67)

for nw > %, and

Av = — (nZmCa + nlmCp){n?mCa + n$jCp) (6.68)

for nf > nw.

In summary, the pattern of a regular hybrid satin fabric can be
determined by the parameters Ca, Cp, ng, nm and, hence, nx, n{ and
nw. The 'regularity' of fabrics deserves some comment. The concept
of regularity is based on the geometrical consideration. For in-
stance, in the regular satin weave of Fig. 6.3(d), the geometrical
distribution of the interlaced regions in two dimensions can be
determined uniquely by two vectors, i.e. (3,1) and (1,3). The
vector (3,1) translates to an interlaced region by three yarns in the
filling direction and one yarn in the warp direction. Other combina-
tions of vectors are also possible, for instance (3,1) and (—2, 2) or
( 2 , - 2 ) and (1,3). An example of an irregular satin is shown in Fig.
6.3(c), where ng = 4 and a set of two vectors cannot be found to
generate the locations of all the interlaced regions. The term
'balanced' hybrid fabric is also used in the analysis. In such a fabric,
the total number and arrangements of yarns of each material in the
filling and warp directions are identical. Hence, the relations
An=A22y DU = D22 and Bn = —B 22 hold for a balanced hybrid
fabric.

6.9.2 Bounds of stiffness and compliance constants

On the basis of the idealizations given in Fig. 6.13, the
hybrid fabric composite can be modeled as an assemblage of pieces
of cross-ply laminates. It is further assumed that the shear deforma-
tion in the thickness direction is neglected.
 Hybrid fabric composites: mosaic model 341

There exist four different types of material combinations in a

cross-ply asymmetrical laminate as depicted in Fig. 6.35, where the
upper lamina is assumed to be composed of filling yarns. In the
superscripts used in Fig. 6.35 the first Greek letter identifies the
upper layer material and the second letter is for the lower layer.
The derivations of the components of Aijy Btj and Dtj of Eqs. (6.3)
for the hybrid fabric composites are straightforward. Also, it is
understood that the cross-terms A16, A26y B16, B26, Dl6 and D26 for
these asymmetrical cross-ply laminates vanish; this is also true when
the upper lamina is composed of warp yarns.

6.9.2.1 Iso-strain
The distributions of stress resultant (and moment) and
strain (and curvature) over the laminate mid-plane vary with
location in the hybrid fabric composite. As a first approximation,
the assumption of iso-strain in the mid-plane is adopted. Equations
(6.3) are then applied to a fundamental region in the laminate. This
region, if repeated, should reproduce the geometrical and material
arrangements of the entire idealized fabric. Thus, the behavior of
the fundamental region should reflect that of the whole laminate.
The dimensions of the fundamental region are denoted by lf and /w
in the filling and warp directions, respectively. It is also defined that
r = <VCa (see Fig. 6.31).

Fig. 6.35. Material combinations in a cross-ply asymmetrical laminate.

The elastic constants for the plies are denoted by: (a) A?", B™, D™,
A'f, B'«" and D[°"*; (b) Af, Bf, Df, A\f, B\f and D\f\ (c) A?«,
Bf, Df, A\fa, B\fa and D\fa; (d) Af, Bf, Df, A1™, B\ffi and D',^.
(After Ishikawa and Chou 1982a.)

(a)

(c)
342 Two-dimensional textile structural composites

(A) Aif and Ay

The average stress resultant Nx, for example, is given as an
average over the fundamental region in the x-y plane:
1 f'- ('<

/ w Jo Jo

M w •'0 Jo

(6.69)
Here, £ and rj stand for the a and /? material phases. From Eq.
(6.69) the following expressions for the effective stiffness constants
of a hybrid fabric composite are given:
{Av, Bir Ay) = v f f" f (Ap, B|n, Of) ck dy
*
(6.70)
f * w JO JO

These averages, in their simple forms, provide upper bounds for

the fabric composite stiffness. If these stiffness constants are
inverted, lower bounds for the elastic compliance constants can also
be obtained.
Both Ap and Dp for the upper ply are identical to those for the
lower ply; general expressions of An and Ay can be written
regardless of the relative magnitude of ng and nm. For instance,

A
nfmr)(nZ

+ n?mnimAl«)r + ntnLAfr2] (6.71)

where n?m and nfm denote the number of a and /3 yarns,
respectively, within the repeating length of nfm yarns in the filling
direction. Naturally, n?m + n$m = nfm, and <m + n£m = /twm. The
expression for Ay can be obtained if Ap(%,r] = a,(3) in Eq. (6.71)
are replaced by Dp. Finally, it should be noted that A,y and Ay can
be reduced to the special case of non-hybrid fabric composites
(Ishikawa and Chou 1983b). The upper bounds of Ati and Ay thus
obtained are identical to those of Ati and Dtj of intermingled hybrids
in cross-ply laminate form.

(B) %
The Bij constants can be obtained with the same approach
as for Ay and Ay- However, here it is necessary to distinguish the
 Hybrid fabric composites: mosaic model 343

weaving pattern as indicated by ng and nm. Hence, the algebra is

more complicated. If ng and nm are numbers not divisible by each
other in one direction

(6.72)
In the case where ng and nm are numbers divisible by each other
in both directions, the expression of #,7 depends upon whether the
interlacing is homogeneous, heterogeneous or mixed. For instance,
for the case of homogeneous interlacing where n{ > nw

x [(ngn?m - 2nfl)nZmB?ja + /ig(w{U™«f

+ n?mnlmBla)r + (ngnfm - 2n^mBf r2] (6.73)
Similar expressions can be derived for heterogeneous interlacing. In
the case of mixed interlacing, the expressions depend upon the
details of the material arrangement. However, the differences
among the Bi}s for homogeneous, heterogeneous and mixed inter-
lacings are, in general, not significant within the usual range of r,
around unity. Therefore, Eq. (6.73) can be used as an approxima-
tion of Bij for such r values when ng and nm are numbers divisible by
each other in both directions.

6.9.2.2 Iso -stress

As another method of estimating the bounds of elastic
moduli the assumption of iso-stress is made. Derivations similar to
that of Eq. (6.69) can be performed to obtain the average strain
expression of the hybrid fabric composite. The average elastic
constants are then given by

(A'o, B',, D$ = T ^ f f (>tf, Blf, D&) dx dy (6.74)

By replacing Afy Bf and Dp in Eqs. (6.71)-(6.73) by A^, B'&
and D'^n, explicit expressions of Eq. (6.74) can be obtained. These
are upper bounds for the composite compliance constants; they can
be inverted to obtain the lower bounds for the stiffness constants.
344 Two-dimensional textile structural composites

6.9.3 One-dimensional approximation

The approximate solution presented below is based upon a
combination of the series model of Ishikawa (1981) for non-hybrid
fabric composites and the mechanics of materials approach for
unidirectional composites. The basic assumptions are that the
hybrid fabric composite can be divided into repeating regions in the
form of one-dimensional strips, and the equilibrium and com-
patibility conditions are not exactly satisfied. Figure 6.36 shows that
the hybrid fabric composite is divided into strips along the filling
and warp directions. It is then assumed that the stress resultant (N)
is uniform in each strip.
The division of the strips is made according to the elastic moduli
under consideration; along the filling (x) direction for Au, B n,
Dn, A'n, B'n and D'n and along the warp (y) direction for A22y
B22, D22, A22) B22 and D22. Either the x or the y direction is
admissible for determination of all the other non-zero constants.
Evidently the average strain in an a yarn is different from that of
a /? yarn. The one-dimensional average strain for the a yarn, for
instance, can be written by considering the stress and moment
resultants in the x direction only:

1 f/f
Kax=j\ £°xxdx

~-\N x
['A1^ dx+Mx
n
(6.75)
~ I \
1{L Jo

Fig. 6.36. One-dimensional model of hybrid fabric composites. (After

Ishikawa and Chou 1982a.)

Filling

filling

a filling
 Hybrid fabric composites: mosaic model 345

where £ and rj stand for a and /?. For the case where ng and nfm are
numbers not divisible by each other in the filling direction, the
following expressions for the averaged compliances are obtained for
the ar-filling yarns:

P ) (zAr
ntmr
)
) B\r + nLB'^r) (6.76)
"g \nfm + n
tmr)

f)'a — („<* r\'aa-\- nP n'aPr\

\jifm + n{mr)
Naturally, expressions of Afjy Bfj and D,y for the ar-filling yarns can
be obtained by inverting A\"y &[" and D\j* of Eqs. (6.76). A similar
procedure can be applied to the /3-filling yarns.
Finally, if the average strain and curvature in the ar-yarns ( S A and
a
k ) are not very much different from those in the /3-yarn, it is not
unreasonable to approximate the entire composite plate with a
uniform strain field. Thus, the stiffness constants can be obtained
from a volume average. For example,

1 (6.77)

6.9A Numerical results

Consider the numerical example for the case of a carbon/
Kevlar fabric in an epoxy matrix. The basic elastic properties of
the constituent unidirectional laminae used in this idealized mosaic
model are given in Table 6.2. The fiber volume fraction is chosen to
be 65% in order to match that of the experimental systems.
Figure 6.37 shows the relationship between An/h and relative
fiber volume fraction for 'balanced fabrics' where An=A22- The
carbon and Kevlar yarns are designated as a and jS yarns,
respectively. Fabric parameters are chosen so as to coincide with
those of Zweben and Norman (1976): ng = 8, nfm = 4 and nwm = 4,
while rifm, nfm, nZm, and n^m vary from 0 to 4. The numbers in
parentheses correspond to values of «£,, nf m, nZm, and n^,m. The
ratio of the yarn width, r, varies from zero to infinity as the relative
fiber volume fraction changes. Since ng and nm in this example are
numbers divisible by each other, the lower bound predictions are
affected by the weaving patterns. Only the lower bound for the case
(3, 1; 3, 1) is shown for the full range of relative fiber volume
346 Two-dimensional textile structural composites

fractions. Also, only homogeneous interlacing types are considered

in Fig. 6.37. The upper bounds are identical to one another for
these three weaving patterns and are shown by a straight line similar
to the predictions of the rule-of-mixtures. The circles and triangles
represent the experimental results of Zweben and Norman for
carbon/Kevlar hybrid fabrics and laminates composed of unidirec-
tional laminae of the parent components. The fabrics used in the
experiments are equivalent to, in the present terminology, the
categories of (3, 1; 3, 1) and (2, 2; 2, 2). The relative yarn width is
close to r = 1 and the interlacing pattern is of the homogeneous
type. These results fall in between the bound predictions.
The effect of fabric geometrical patterns of the parent composites
on the bound prediction is worth examining. The bound prediction
of Kevlar/epoxy is represented in Fig. 6.37 by either n"m = nZm = 0
or r = Cp/Ca—* 0°. Similarly, the carbon/epoxy system corresponds

Fig. 6.37. An/h vs. relative fiber volume fraction, h denotes specimen
thickness; UB, upper bound; LB, lower bound. Experimental results of
Zweben and Norman (1976); (•) fabric; (A) laminate. (After Ishikawa
and Chou 1982a.)

UB
6 -

PH
5 -
o

1 1 1 1 1 1 1
1.0
^carbon /(^carbon + ^Kevlar)
 Hybrid fabric composites: mosaic model 347

to the case of either nfm = n£ m = 0 or r = 0. The lower bound

predictions based on different combinations of n g and nm yield
different results. Point A in Fig. 6.37 indicates the combination
n g = 8 and n?m = nZm = 0- Point B is for the limiting case of ng = 8,
(1, 3; 1, 3) and for R—»°°; this case is equivalent to n g = 6 and
"frn = ftwm = : 0. Point C is obtained from the case of ng = 8, (3, 1; 3,
1) and r—»°°. The same weaving pattern can be achieved for n g = 2
and rifm = 0. Discussions similar to the above can be made for the
case of the carbon/epoxy system. Point D is for n g = 8 and
«fm = nim = 0. Point E is for n g = 8, (3, 1; 3, 1) and r - > 0 ; this is
equivalent to n g = 6 and n%m = n1m = 0. The transition of the
geometrical pattern from ng = 8 to either ng = 2 or 6 as r approaches
the limiting values can be understood from Figs. 6.38(a) and (b), as
well as Eqs. (6.71) and (6.73).

Fig. 6.38. The transition of fabric geometrical pattern as affected by r. (a)

n g = S (3, 1; 3, 1) and r = \; this pattern becomes ng = 6 as r—>0. (b)
n g = 8, (3, 1; 3, 1) and r = 8; this pattern becomes n g = 2 as r—»o°. (c)
Homogeneous interlacing for n g = 8, (2, 2; 2, 2) and r = J; this pattern
becomes n g = 4 as r—>0, (d) Heterogeneous interlacing for n g = 8, (2, 2; 2,
2) and r = \ ; this pattern becomes a cross-ply laminate as r—>>0. (After
Ishikawa and Chou 1982a.)

II 1
1 1 1 1

(b)

(d)
348 Two-dimensional textile structural composites

The relationship between Bn/h2 and the relative fiber volume

fraction is demonstrated in Fig. 6.39 for ng = 8 and nfm = rcwm = 4.
Results for both homogeneous and heterogeneous interlacings are
shown. In the case of homogeneous interlacing, i.e. (1, 3; 1, 3), (2,
2; 2, 2) and (3, 1; 3, 1), the basic trends of the lower bounds are
similar to the predictions shown in Fig. 6.37. However, the upper
bound predictions in this case are also affected by the fabric
parameters. In the case of heterogeneous interlacing (2, 2; 2, 2),
both the upper and lower bound predictions tend to be very large
values when r=0 and r—>°°. As a result, extremely large coupling
effects are seen in these limiting cases.

6.10 Hybrid fabric composites: crimp and bridging models

Although the bounds for the elastic properties of hybrid
fabric composites can be conveniently estimated by the mosaic
model, the upper and lower bounds are rather far apart. An
improved analysis based upon the 'crimp model' and 'bridging
model' developed for non-hybrid fabric composites is described.
In the following analysis, we specify ng= 8 and the fiber material
repeating parameters (n"m, nfm; < m , «wm) are of three types: (3,

Fig. 6.39. Bu/h2 vs. relative fiber volume fraction; upper and lower
bound predictions for homogeneous interlacing; upper and
lower bound predictions for heterogeneous interlacing; one-di-
mensional approximate solution. (After Ishikawa and Chou 1982a.)

'« 6

0.0 0.0 0.0

 Hybrid fabric composites: crimp and bridging models 349

1; 3, 1), (1, 1; 1, 1) and (1, 3; 1, 3). Also, homogeneous interlacing

is considered in the analysis. Furthermore, the calculation proce-
dure for the system (1, 3; 1, 3) is the same as that for the system (3,
1; 3, 1) by interchanging the a and j3 materials. Therefore, only the
systems of (3, 1; 3, 1) and (1, 1; 1, 1) need to be considered in the
analysis.

6.10.1 Crimp model

The crimp model takes into account fiber continuity;
sectional shapes of some typical interlacing regions are shown in
Figs. 6.40(a) and (b). The sinusoidal type functions used in Section
6.5 for describing the undulation shapes are adopted here. For the

Fig. 6.40. Typical structures of interlaced regions of hybrid fabric compos-

ites; h denotes plate thickness and ht indicates the total thickness of the
yarns, (a) Filling: a material; warp: a or p material, (b) Filling: /3
material; warp: a or p material. (After Ishikawa and Chou 1983d.)

(a)
350 Two-dimensional textile structural composites

case where the filling yarn is composed of a material (Fig. 6.40a),

the height of the filling yarn is given by

0
(6.78)

When the filling yarn is composed of /3 material, the height of the

filling yarn is given by

(6.79)

where /*t denotes the total thickness of the yarns.

Corresponding to the cases of Eqs. (6.78) and (6.79), the heights
of the warp yarns are given, respectively, by

x
2
(6.80)

hl(x) = •
hl
1-:
(6.81)

It should be noted that Eqs. (6.78)-(6.81) are written for the

portion of the undulated region where the filling yarn is beneath the
warp yarn.
A key assumption made in the fiber crimp model (Ishikawa and
Chou 1982b) is that the classical laminated plate theory is applicable
to each infinitesimal slice of material of width dx. Then the local
plate extensional stiffness coefficients for the portion where the
 Hybrid fabric composites: crimp and bridging models 351

filling yarn is composed of a material, are given by

AA a%/X \ _ /)M u _^1_L.+ n U<*( Y\ — n ha(r\\

ij \ > ~~ Kij'I " y l\x) 2\X)j

where the superscripts F, W and M denote the filling yarn region,

warp yarn region, and pure matrix material, respectively; £ stands
for a or fi material, and h denotes the total laminate thickness,
including the pure matrix layers. Furthermore, the first superscript
of Atj indicates the filling material and the second one the warp
material. This convention is followed for all the stiffness and
compliance constants throughout this analysis.
Likewise for the portion of the laminate in Fig. 6.40(b), where
the filling yarn is composed of /3 material,

+ Q?(x) | + QP(hl(x) - *?(*)) (6.83)

Similarly, expressions for Bf(x), Bf(x)y Df(x) and Df(x) can

also be obtained.
The local thermal deformation coefficients can be obtained by
replacing Qtj in Eq. (6.83) by <2//O} (Eqs. (6.6) and (6.7)). For
instance,

(h - 1 + hftx) - hft

+ qFxa(x) | + q7Kh%(x) - /!?(*)) (6.84)

where qx = Qxxocxx + Q12Qcyy + Qie&xy Explicit expressions of off-

axis properties in the filling yarn region, Qjf(x), are given in
Section 6.5. The local compliance constants A'if'1(x), B'i^r'{x) and
D£\x) are obtained by inverting Ap(x), Bfp(x) and Dfp, where £
and rj indicate a or /3 material. Then, the thermal coefficients A'^11
and B'^n can be obtained from Eq. (6.9).
Finally, consider again the one-dimensional idealized model of a
hybrid laminate. The average extensional compliance for the
352 Two-dimensional textile structural composites

portion containing a yarns is defined as

Af<* = - f/2A^(x)dx = (l-^W^ + - fVMd*

Q JQ \ z / a j ao
(6.85)
For the case of /3fillingyarns,
AfK = (l - U-)Ar + - r A'tix) Ax (6.86)

The superscript ' C in Eq. (6.85) signifies the fiber crimp model. The
other averaged compliance constants B'f"*, B'0CK, D'iJCali and t)fK
can be obtained in a similar manner. Expressions for the averaged
in-plane thermal expansion coefficients can be obtained from Eqs.
(6.85) and (6.86) by replacing A[j by the appropriate A\. Also, the
thermal bending coefficients can be easily obtained. However, it
should be noted that B\f^n and B\c%n do not vanish when the
integrations in Eqs. (6.85) and (6.86) are carried out over the entire
length of a(l + r)/2 (Fig. 6.40), unlike the cases of non-hybrid
fabrics. This fact is caused by the difference in yarn width and
properties of the constituent fibers of the fabric. Finally, the
averaged stiffness constants Af^, B^ and £>ffn can be obtained by
inverting these averaged compliance constants. Then the averaged
thermal constants Af^n and Bf^ are obtained from the inverted
form of Eq. (6.9). It should be noted that the thermoelastic
constants derived here are based upon the definitions of hx{x) and
h2(x) given in Eqs. (6.78)-(6.81), i.e. the filling yarn is situated
beneath the warp yarn. Thus, the coupling stiffness constants for the
right-hand portions of Figs. 6.40(a) and (b) for instance, are
denoted by -B%af> and -B^Pa, respectively.

6.10.2 Bridging model

The case of fabrics with ng = 8, « m, nfm; < m , nim) = (3, 1;
3, 1) and homogeneous interlacing pattern is considered first. A
possible shape of the minimum repeating unit is indicated in Fig.
6.41 as the area ABCD. The three-dimensional view of this
repeating unit showing the interlaced configurations of the a and ft
yarns is given in Fig. 6.42, which consists of five regions Rlf R2,
R3, R4 and R5, arranged in series along the loading direction.
However, other choices for the division in regions are possible. It is
assumed in the following analysis that the resultant force in the
loading direction of every region is identical.
 Hybrid fabric composites: crimp and bridging models 353

To exemplify the analysis of the bridging model, region R2 is

considered. Region R2 consists of four sub-regions labeled R\, R\,
R\ and R\ (see Fig. 6.43). By assuming an iso-strain condition for
the sub-region, the average compliance constant of each region can
be found. Then, the averaged stiffness constants of each sub-region
are obtained by inverting the corresponding compliance constants.
On the basis of the assumption of iso-strain, the average stiffness of
the entire region R2 can be determined. The elastic constants of the
other regions can also be determined following this procedure.
For the fabric composite of Fig. 6.42, it is assumed that each

Fig. 6.41. A hybrid fabric with homogeneous interlacing, for « g = 8,

% m = wwm = 4; a and ft indicate two types of yarn material; ABCD and
EFGD denote two choices of repeating units. (After Ishikawa and Chou
1983d.)

(S material
• a material

3"
Interlaced
region
D

Fig. 6.42. A bridging model for ng = S and the (3, 1; 3,1) case (region
ABCD of Fig. 6.41). (After Ishikawa and Chou 1983d.)
(1 + r)a

/ Nx, Mx

h%(x) h1(x)
354 Two-dimensional textile structural composites

region, Rl9 R2, R3, R4, or R5 carries the same load Nx. Thus, the
compliance constants of the entire composite can be regarded as the
volume average of the compliances of the individual regions. Then,
the inversion of the compliance constants gives the stiffness
coefficients of the entire composite unit cell, Ay, By and Di}. The
basic idea of the analysis briefly outlined above is identical to that of
the 'bridging model' of Section 6.6 in which only non-hybrid
composites are considered. The details of the derivations of elastic
and thermal deformation constants can be found in Ishikawa and
Chou (1983d).
Both regions ABCD and EFGD of Fig. 6.41 can be treated as
repeating regions for the entire fabric composite. A three-
dimensional view of region EFGD can also be found in Ishikawa
and Chou (1983d). As to the case of the (1,3; 1,3) material
combination, the thermoelastic constants can be obtained from the
above procedure by simply interchanging the a and /? materials.
Ishikawa and Chou (1983d) have also examined the case of a fabric
of ng = 8 with homogeneous interlacing and material repeating
parameters (2,2; 2,2). The cases of (3,1; 3,1), (1,3; 1,3) and
(2,2; 2, 2) give all possible fiber material combinations for homoge-
nous interlacing in hybrid fabrics with the given fabric parameters.

6.10.3 Numerical results and summary of thermoelastic properties

Numerical work has been performed to examine the thermo-
elastic properties of a carbon/Kevlar/epoxy hybrid fabric compos-
ite. The basic material properties of unidirectional laminae of
carbon/epoxy and Kevlar/epoxy are given in Table 6.2. The fiber
volume fraction of all the unidirectional laminae is assumed to be

Fig. 6.43. Detailed view of region R2 in Fig. 6.42. (After Ishikawa and
Chou 1983d.)
 Hybrid fabric composites: crimp and bridging models 355

65%, which is slightly higher than the total fiber volume fraction of
the fabric composite due to the presence of pure matrix layers.
Figure 6.44 shows the predictions of the extensional stiffness of
the bridging model as well as those from the bound approach (Fig.
6.37) of Ishikawa and Chou (1982a) for a carbon/Kevlar/epoxy
system of ng = 8. Three different material repeating parameters are
presented and the theoretical curves are obtained by changing r, the
yarn width ratio of a and /3 materials. Because values of r far from
unity are impractical, the curves in Fig. 6.44 are truncated. The
predictions based upon the bridging concept fall in between the
upper and lower bounds and compare very favorably with the
experimental data of Zweben and Norman (1976).
The following is a summary of the analysis of thermoelastic
properties of hybrid woven fabric composites:

(1) The structural characteristics of woven hybrid fabrics have

been identified by the material parameter nm (nfm and nwm)
as well as the geometrical parameter ng (nig and nwg). If the

Fig. 6.44. Axx/h vs. relative fiber volume fraction of carbon/Kevlar/epoxy

composites with « g = 8; bound theory; bridging model; ( •
and A experimental data for fabric and cross-ply laminate composites,
respectively; (h = ht, h/a = 0.4). (After Ishikawa and Chou 1983d.)

7.0

Upper bound ^ ^
6.0 —
• j ^—-—
o ^ ^
i
o 5.0 3, 1)
Bridging "g ^ ^
m O d
1
! ! ^ " l ; l, i) JJ
7*—— 1 *
4.0
e
F (1, 3; 1,3) i — '/ -
ng =
J . — -**
8 ^ j \~~ Lower bound
3.0

0 1 1 1 1 1 1 1 1 1
0.0 0.2 0.4 0.6 0.8 1.0
/ /
I carbon/(I carbon ' ^Kevlar)
356 Two-dimensional textile structural composites

ng and nm of a fabric are numbers not divisible by each

other in one or both directions (filling and warp) there is a
unique interlacing pattern. There is more than one type of
interlacing pattern if ng and nm are numbers divisible by
each other in both directions.
(2) In the analysis of the mosaic model, the fabric composite is
regarded as an assemblage of asymmetrical cross-ply lamin-
ates. Upper and lower bounds for the elastic stiffness
and compliance of hybrid composites have been obtained
assuming iso-strain and iso-stress, respectively. The in-
fluence of fabric parameters on elastic properties can be
assessed using this model.
(3) The magnitude of the coupling terms of Btj and B'tj depends
on whether ng and nm are numbers divisible by each other.
In the case where ng and nm are numbers divisible by each
other in both the warp and the filling directions, the upper
and lower bounds of Btj and the lower bounds of Atj and Dtj
are influenced by the interlacing types.
(4) The transition of ng from one value to another occurs as the
ratio of yarn width of the component fibers approaches zero
or infinity. In such extreme cases, the magnitude of the
coupling terms becomes very large, especially for heteroge-
neous interlacing. The distinct interlacing types for given ng
and nm, however, render nearly identical solutions for the
bounds when the yarn width ratio is around unity.
(5) The one-dimensional fiber undulation or crimp concept has
been modified to treat the interlacing of two different types
of fibers, and it has been incorporated into a general
'bridging model' for predicting thermoelastic properties of
hybrid fabric composites.
(6) The predicted values of the axial elastic stiffness constant
are insensitive to the choice of a repeating unit of the fabric
material.

6.11 Triaxial woven fabric composites

6.11.1 Geometrical characteristics

Biaxial woven fabrics exhibit relatively low elastic moduli
or low resistance to extension when deformed along the bias
direction (45° to warp and filling) as compared with deformation in
the warp and filling directions. A triaxial woven fabric (Doweave
fabric), is composed of three sets of yarns (two sets of warp yarns
 Tri-axial woven fabric composites 357

and one set of filling yarn), which intersect and interlace with one
another at 60° angles as shown in Fig. 6.45.
For the purpose of identification, the warp yarns are called 'one
o'clock' and 'eleven o'clock' warps. The filling yarn is horizontal
and is interwoven with the warp yarns in different sequences
depending on the fabric style. The geometry of the fabric can vary
from a very open but stable construction, such as basic weave, and
stuffed basic weave (which has additional yarns in the filling
direction), to a tightly packed construction, such as the bi-plane
weave. Figure 6.46(a) shows a schematic diagram of the stuffed
basic weave. The bi-plane weave is quite similar to the basket
weave of biaxial woven fabrics. As shown in Fig. 6.46(b), the filling
yarns in a bi-plane weave are woven both over and under two sets
of warp yarns to form a closed construction.
With the load bearing yarns arranged in three instead of two
directions, the triaxial woven fabrics yield more isotropic responses
to both tensile and shear deformations, offering an alternative to
the inherent structural weakness of conventional biaxial fabrics. The
ability of the triaxial woven fabrics to maintain structural integrity

Fig. 6.45. Triaxial woven fabric. (After Yang and Chou 1989.)

Fig. 6.46. The geometries of (a) stuffed basic triaxial woven fabric and (b)
bi-plane weave triaxial woven fabric. (After Yang and Chou 1989.)

(a)
358 Two-dimensional textile structural composites

even with a very open construction is quite unique among textile

structures.
The filling yarn of a triaxial woven fabric may be composed of
bundles of different size and material from the warp yarns.
Therefore, hybrid fiber constructions are available for triaxial
woven fabrics as for biaxial woven fabrics. Also, by proper
selection of material combinations, yarn sizes and fabric weaving
patterns, a wide range of geometrical and mechanical properties can
be engineered in triaxial woven fabrics.
Although considerable effort has been made to investigate the
mechanical behavior of triaxial woven fabrics (see Dow 1969; Dow
and Tranfield 1970; Skelton 1971; Scardino and Ko 1981; Schwartz,
Fornes and Mohamed 1982; Schwartz 1984), the properties of
composites reinforced with triaxial woven fabrics have not been
adequately evaluated. Dow (1982) developed an analytical method;
the geometrical model used for the calculation of the fiber volume
fraction and elastic properties of the triaxial woven fabric composite
resembles the crimp model of Fig. 6.17. The undulated yarns are
divided into segments and each of these segments is treated as an
off-axis short-fiber composite lamina. The elastic properties of the
triaxial fabric composite unit cell are calculated by averaging the
contribution from each of the short-fiber composites.
In the following, a more refined analytical model is developed to
predict the thermoelastic properties of triaxial fabric composites.
An outline of the methodology of analysis is given first. It is then
extended to biaxial, non-orthogonal woven fabric composites.
Numerical results of the thermoelastic properties are presented as a
function of the fabric construction parameters. The contents of
Sections 6.11.2 and 6.11.3 are excerpted from Yang and Chou
(1989).

6.11.2 Analysis of thermoelastic behavior

For the purpose of analyzing the thermoelastic constitutive
relations of triaxial woven fabric composites, a unit cell of basic
triaxial weave is identified, as shown in Fig. 6.47, which contains
three impregnated yarn bundles oriented in space and interstitial
matrix regions. Repeating the unit cell in the fabric plane obviously
reproduces the complete triaxial woven structure. This methodol-
ogy can easily be extended to treat other types of weaving patterns.
The concept of the 'crimp model' (Ishikawa and Chou 1982b) is
extended to the following analysis. In this model, each impregnated
yarn bundle is further idealized as an undulated unidirectional
 Tri-axial woven fabric composites 359

lamina as shown in Fig. 6.48. The geometrical configuration of each

undulated lamina can be simulated as follows. First, consider the
lamina of filling yarns. The upper boundary for the undulated
configuration is given by (Fig. 6.48a)

(6.87)

Here, xx coincides with the x axis and Ht is the thickness of the

undulated lamina.
Next, the form of fiber undulation in the one o'clock warp lamina
as shown in Fig. 6.48(b) is
l
2- jjfj^ (0<x2<2/2) (6.88)

Here, x2 is in the direction of 60° from the x axis. Similarly, in the

eleven o'clock warp lamina (Fig. 6.48c), the form of fiber undula-
tion is

(6.89)

where x3 is in the direction of —60° from the x axis.

The crimp in the undulated laminae reduces the composite
stiffness as compared with that of straight reinforcements. The
local off-axis angle of each undulated lamina along the xly x2 and x3

Fig. 6.47. Unit cell structure of the basic triaxial woven fabric composite.
(After Yang and Chou 1989.)
360 Two-dimensional textile structural composites

directions can be obtained by

6 = tan" (i = 1, 2 or 3) (6.90)
dx,
The effective thermoelastic properties of each undulated lamina
can be derived through the following procedures. First, the undu-
lated lamina can be regarded as an assemblage of many small pieces

Fig. 6.48. Geometrical configurations of undulated filling and warp

laminae. (After Yang and Chou 1989.)

(a)

H,

(b)

(c)
 Tri-axial woven fabric composites 361

of unidirectional lamina. Each of these segments is uniquely

characterized by an off-axis angle as defined in Eq. (6.90). The
reduced effective thermoelastic properties in the x direction for the
filling lamina are the same as those given in Eqs. (6.24) and (6.51).
By assuming that each of these short composite laminar segments
is subjected to the same stress, the strain in each segment is

(6.91)

The normal strains averaged over the length 2/x along the x
direction are
1 f2/'
IIX JQ
(6.92)
2/l

The average longitudinal Young's modulus, transverse Young's

modulus and Poisson's ratio can be obtained as

Exx = J * Eyy = E22 vxy= - y (6-93)

The average in-plane shear modulus can be obtained by assuming

that each of these segments is subjected to the same shear strain.
Thus,

r 2 / G
i xy(6)dx (6 94)
.
Thus, the averaged stiffness constants of the undulated filling lamina
can be obtained by using Eq. (6.10).
The average thermal expansion coefficients along the x and y
directions are defined as
1 f2/l
&xx = —• (#n cos2 6 + a22 sin2 6) d6

a
yy=~r \ awC0) d* = a22 (6.95)

xy -J_f2'' xy
2/, Jo
362 Two-dimensional textile structural composites

The procedures outlined above can be applied to obtain the

effective thermoelastic properties of both one o'clock and eleven
o'clock warp laminae along the x2 and x3 directions, respectively.
However, the x2 and x3 directions are, respectively, at 60° and —60°
off-axis orientations with respect to the x axis. The effective
properties of these two warp laminae in the x—y plane can be
obtained by the following coordinate transformation (Jones 1975):

Gn = Gii cos4 0 + 2(Q12 + 2g 6 6 ) sin2 0 cos2 0

+ Q22 sin4 0
G12 = ( 2 n + Q22 ~ 4e 6 6 ) sin2 0 cos2 0
+ Gi2(sin4 <f> + cos4 4>)
Q22 = Qu sin4 0 + 2(<312 + 2Q 66 ) sin2 0 cos 2 0 + Q22 cos4 0
Gi6 = (Gn - G12 " 2G66) sin 0 cos 3 0
+ (G12 - G22 + 2G 66 ) sin3 0 cos 0
-
G26 = (Gn G12 - 2G 66 ) sin3 0 cos 0
(6.96)
+ (G12 - G22 + 2G66) sin 0 cos3 0
G66 = (Gn " G22 " 2Gi2 - 2G 66 ) sin2 0 cos2 0
+ G66(sin4 0 + cos4 0)
axx = ocxx cos 2 0 + ayy sin 2 0

ocyy = axx sin 2 0 + ayy cos 2 0

=
«xy (&xx - ocyy) sin 0 cos 0

?y = Q\2«xx + QllOCyy + G ^ ^ y

Qxy = G l 6 ^ x + G26<*yy + G e e * ^
Here, 0 represents +60° and —60°, respectively, for one o'clock
and eleven o'clock warp yarns.
Upon knowing the effective thermoelastic properties of each
undulated lamina in the x-y plane, the composite properties can be
derived under the assumption that each of these undulated compos-
ite laminae is subjected to the same strain along the x direction.
Thus, the effective in-plane thermoelastic properties of the triaxial
fabric composite unit cell are given as (Rosen, Chatterjee and
 Tri-axial woven fabric composites 363

Kibler 1977)

Q* = t V<">Gj?>

n= \

n= \

(6.97)

where V is volume fraction and (n) denotes the yarns in the xlt x2
and x3 directions. The thermal expansion coefficients of the triaxial
woven fabric composite are found from

= S*nq*x
= St2q*x + SWy + SU% (6.98)

Here, S*} is the inversion of Q- of Eqs. (6.97).

By assuming that the yarns have a circular cross-section with
diameter d, and lx = 12 = 13 = I for the unit cell of Fig. 6.47, the
highest fiber volume fraction that can be obtained for a basic
triaxial weave is about 43%. For the yarn spacing/diameter ratios
(l/d) of 2, 3, 4, 5 and 6, the fiber volume fraction (Vf) values are,
respectively, 42.5, 24, 17.5, 14.2 and 11. Higher volume fractions
can be obtained by changing the weave pattern to stuffed basic
weave or bi-plane weave. As the l/d ratio increases, the fiber
volume fraction decreases and the crimp can be minimized. Thus,
the unit cell structure approaches a [0°/±60°] laminate composite
with straight reinforcements.
Figures 6.49(a)-(c) demonstrate the variation of longitudinal
Young's modulus, in-plane shear modulus and longitudinal thermal
expansion coefficient of triaxial woven carbon fabric reinforced
epoxy composites with yarn spacing/diameter ratios. The results of
[07 ±60°] laminate composites as functions of fiber volume fraction
can also be found in these figures. These results all indicate that as
l/d increases, the difference in thermoelastic constants between
woven structures and straight laminae is reduced as expected.
364 Two-dimensional textile structural composites

Even though the stiffness reduction of triaxial woven fabric

composites as compared with [0°/±60°] laminatesis quite severe
when l/d is small, it is feasible to place additional laid-in yarns
(non-crimp yarns) in the filling direction to enhance the axial
properties as shown in Fig. 6.46(a). Furthermore, fiber hybridiza-

Fig. 6.49. Comparisons of the predicted thermoelastic properties of

triaxial woven fabric composite (carbon/epoxy) and [0°/±60°] angle-ply
laminate composite (carbon/epoxy) as functions of yarn spacing ratio
(l/d). (a) Longitudinal Young's modulus, (b) In-plane shear modulus, (c)
Longitudinal coefficient of thermal expansion (CTE). (After Yang and
Chou, 1989.)

40

30 -
°/±60° laminate

20 -
Triaxial
fabric composite
10
1 1 1
4
l/d
(a)

20 —

u
15
"8
10
^ ^ Triaxial fabric composite
1
——
j
[0°/±60° laminate

n 1 1 1
4
l/d
(b)
 Tri-axial woven fabric composites 365

tion allows considerable designflexibilityin meeting the requirements

of high performance composites. Thus, by the proper selection of
material combinations and fabric structural geometry, a wide range
of mechanical properties can be engineered.
6.11.3 Biaxial non -orthogonal woven fabric composites
Biaxial non-orthogonal woven fabric composites can be
produced by flat braiding, or they could occur in the fabrication of
bi-axial orthogonal woven fabric composites. The flow of matrix
material and the curvature of the mold surface could distort
orthogonal yarns into non-orthogonal positions. The geometry of a
non-orthogonal woven fabric is depicted in Fig. 6.50. It can be
treated simply as a triaxial woven fabric without the filling yarn.
Consequently, the methodology developed for the triaxial woven
fabric composites can be readily applied. The composite unit cell is
composed of two undulated laminae interlaced together at any
angle, the magnitude of which depends upon the braiding pattern or
the distortion of the fabric.
Figure 6.51 illustrates the variation of Young's modulus with the
braiding angle or the angle of a biaxial non-orthogonal woven
fabric composite. Yang and Chou (1989) have also shown that as
26 decreases below the right angle, the thermal expansion
coefficient increases along the y direction and decreases along the x
direction; the in-plane shear modulus decreases with the decrease in
the bias angle.

Fig. 6.49 (cont.).

4
Triaxial fabric composite

u [0°/±60° laminate

00
c
o

_J
4

lid
(c)
366 Two-dimensional textile structural composites

6.12 Nonlinear stress-strain behavior

The nonlinear stress-strain behavior of fabric composites
due to transverse crackings initiated in warp yarns has been
discussed in Section 6.7. Although transverse cracking can account
for the nonlinearity at small strains, both the filling yarns and the
matrix rich regions contribute to the overall nonlinear behavior of
fabric composites.
Ishikawa and Chou (1983c) first adopted a one-dimensional
(crimp) model to examine the material nonlinearities in the filling
yarn and the pure matrix region. This approach is then extended to

Fig. 6.50. A non-orthogonal woven fabric composite and its unit cell for
analysis. (After Yang and Chou 1989.)

Fig. 6.51. The predicted longitudinal Young's modulus of non-orthogonal

carbon fabric/epoxy composites as a function of bias angle (Vf = 60%).
(After Yang and Chou 1989.)

200 -
Young s modu lus (GPa)

150

100 — ^s,. x direction

50

v direction
n 1 i
30 60 90
Bias angle, 20 (degrees)
 Nonlinear stress-strain behavior 367

the bridging model for satin weave composites, and combined with
the analysis of transverse matrix cracking to provide a more
comprehensive description of the nonlinear elastic stress-strain
behavior of fabric composites.
The essence of the treatment of Ishikawa and Chou can be
understood by considering the filling yarn depicted in Fig. 6.17.
Segments of this yarn are subjected to off-axis loading in the x—z
plane due to fiber undulation. Thus, nonlinear shear deformation is
induced in the filling yarn due to the axial load. Following Hahn
and Tsai (1973), the nonlinear shear strain-stress relation is
assumed to be
ezx = S55azx + S5555{ozxf (6.99)
Here, S55 and S5555 represent, respectively, the linear and nonlinear
compliance constants. As to the nonlinear shear behavior of the
matrix material under tensile loading, the constitutive relation is
assumed to follow the same form as Eq. (6.99)
e£ = S M + S E n ( O 3 (6.100)
Ishikawa and Chou have performed a numerical analysis of the
stress-strain relation for glass/polyimide. The basic properties of a
unidirectional lamina are given in Table 6.2 and S5555 = 37.0 (1/GPa 3).
The major ambiguity of the analysis lies in the value of the
nonlinear shear compliance, S5555. Because of the lack of ex-
perimental data, an estimated value based upon the stress-strain
curve for a glass/epoxy composite (Jones 1975) is used. The elastic
properties of polyimide are £ = 4.31GPa and v = 0.36. The non-
linear extensional compliance S™ u = 9.88 (1/GPa3) is also assumed
to be the same as that of epoxy. Other assumptions are that
a = au = 0.4 mm, h = ht = 0.244 mm, and the bending-free state of
deformation is valid.
Figure 6.52 indicates the numerical results of this nonlinear
analysis (solid line) and the result from the consideration of
transverse cracking only (dashed line) for the glass/polyimide
composite. Both eight-harness satin (ng = S) and plain weave
(ng = 2) composites are indicated. The experimental stress-strain
data of an eight-harness satin as indicated by the dots are included.
The nonlinear analysis compares better with the experiment in the
range of large strain than the results given by Ishikawa and Chou
(1982b) for matrix cracking only. It is also observed that the
contribution from shear nonlinearity increases at higher stress
levels and for lower ng values. Furthermore, the effect of non-
368 Two-dimensional textile structural composites

linearity on the composite behavior from the filling yarn far exceeds
that from the pure matrix region.

6.13 Mechanical properties

The microstructure of two-dimensional woven fabric com-
posites is responsible for some unique mechanical properties
which are not found in their equivalent cross-ply laminates. The
tension-tension fatigue behavior of woven fabric composites has
been examined by Schulte, Reese and Chou (1987). In the
following, the structure-property relationships are demonstrated in
terms of the friction and wear behavior, and the notched strength of
woven fabric composites.

6.13.1 Friction and wear behavior

When two surfaces interact, contact is made at their
asperities. With the application of a normal load and relative
motion, plastic deformation at the asperity contact zones occurs. As
a result, adhesive junctions are formed which, under the influence
of motion, tend to get fractured. Fracture occurs not at the original
point of contact, but at some point within the softer material.

Fig. 6.52. Non-linear stress-strain relations of glass/polyimide fabric

composites with a=au = 0.4 mm and h=ht = 0.244 mm. (After Ishikawa
and Chou 1983c.)
0.5

1.0
Strain (%)
 Mechanical properties 369

Hence material is transferred from one surface to the other.

Subsequently, these transferred particles come loose due to the
repeated contact.
In sliding wear, material loss is dominated primarily by adhesive
mechanisms and secondarily by surface fatigue and abrasion; the
abrasive component increases with increasing surface roughness. As
compared with the abrasive wear conditions, the sliding wear
process is much milder and is, consequently, extremely sensitive to
the microstructure of the surface being worn. This is especially true
for composite systems (Mody, Chou and Friedrich 1988).
In sliding wear, the sliding velocity (v) effects are manifested in
frictional heating generated at the sliding interface. At some critical
velocity, steady-state wear will no longer prevail, and the coefficient
of friction and/or the wear rate will increase sharply. Reinforcing
fibers usually increase the critical velocity of polymeric matrices.
The influences of contact pressure (p) on sliding wear and of
temperature on limiting pv values are also of major concern. Other
factors include humid environments, counterface properties (i.e.
surface roughness, density and height of the asperities), and the
state of sliding interface (i.e. lubricants, films). The issue of fiber
reinforcement raises additional important parameters, such as the
type of fiber preforms, volume fraction and fiber orientation.
Woven forms of fiber reinforcement have demonstrated superior
wear characteristics for self-lubricating bearings. Mody, Chou and
Friedrich (1988) have investigated the sliding friction and wear of a
neat thermoplastic matrix (PEEK), and examined the changes
achieved by the incorporation of unidirectional continuous and
two-dimensional woven carbon fibers. In their experiments, a
pin-on-disc type wear testing machine is used; the specimen
temperature, the torque generated at the sliding interface, the
sliding velocity (in terms of revolutions per minute), the sliding
distance (in terms of the number of revolutions made), and the
sliding time are monitored. The sliding counterface is a polished
steel surface.
The dimensionless wear rate (w), in the units of um/m (depth
worn per unit distance slid), is computed by using the measured
mass loss (Am) and density (p), along with the apparent contact
area (A) and the sliding distance (L) in the following form:
w = Am/(ALp) (6.101)
The wear resistance of a material is the reciprocal of the wear rate
(w"1). The experiments show that, initially, wear progresses in a
370 Two-dimensional textile structural composites

non-linear fashion. Later, as a definite sliding interface is estab-

lished, the steady-state condition prevails, and the mass loss
increases linearly with increases in sliding time.
Because of the anisotropic nature of fiber composites, it is
important to identify the sliding directions relative to the fiber
orientations. Three principal directions for the unidirectional con-
tinuous fiber composite have been identified, as shown in Fig.
6.53(a). Fibers in the plane of sliding and parallel to the direction of
sliding are termed parallel (P). In-plane fibers oriented transverse to
the direction of sliding are termed anti-parallel (AP), and fibers that
stand normal to the plane of sliding are designated as normal (N).
Following Mody, Chou and Friedrich (1988), six sliding directions
are defined for a five-harness satin composite (Fig. 6.53b). The
warp direction of the fabric, which has 80% of the fibers oriented in
the direction of sliding, is referred to as the parallel direction (P).
On the other hand, the filling direction of the fabric, which has 20%
parallel to the sliding direction, is referred to as the anti-parallel
direction (AP). Having thus defined the P and AP directions for the
woven fabric system, consider a face perpendicular to the warp
direction. This face will have a combination of fibers that stand
normal to it, and parallel or transverse to it, depending on the
direction of sliding on this face. Similarly, for the face orthogonal to
the filling orientation of the fabric, the same reasoning prevails.
From Fig. 6.53(b) it can also be concluded that the pair N P(N P) and
NP(N,AP) is the same as the pair N AP(NP) and NAP(N AP) , if the warp
and filling fiber yarns are the same fiber type. (The notations within
the parentheses represent fibers of those orientations which are
being slid.)
Sliding wear experiments have been conducted by Mody and

Fig. 6.53. Sliding directions with respect to the fiber orientation for (a) a
unidirectional continuous fiber composite, and (b) a two-dimensional
woven fabric composite. (After Mody, Chou and Friedrich, 1988.)
(Filling)/AP (Warp)/P
\ /
1
N

(a)
 Mechanical properties 371

colleagues for unreinforced PEEK matrix, unidirectional continuous

fiber composites, and two-dimensional woven fabric composites at
three temperatures (50, 150 and 240°C) and three pv values (0.3,
0.6 and 0.9MPam/s). Here p denotes contact pressure and v is
sliding velocity. The variations of wear rate for these three material
systems at 50°C and pv = 0.3 MPa m/s are summarized in Fig. 6.54.
The wear rate of unreinforced PEEK is relatively high. In the case
of unidirectional carbon/PEEK composites, the wear rates are
highly anisotropic with the AP direction showing nearly twice the
wear rate of the P and N orientations. For two-dimensional woven
fabric composites, owing to the equivalence of the sliding directions
NP(N,P) to N Ap(NP) , and of NP(N,AP) to N AP(N , AP) , four unique sliding
directions are identified. These include the P-oriented surface, the
AP-oriented surface, the surface containing a combination of N-
and P-oriented fibers (N, P), and the fourth, which has a combina-
tion of N- and AP-oriented fibers (N, AP). Wear rates of these four
surfaces at 50°C turn out to be quite uniform, and thus only their
average value is indicated in Fig. 6.54. Models for the wear
mechanisms of composites as functions of fiber orientation have
been presented by Mody, Chou and Friedrich (1988, 1989).

6.13.2 Notched strength

Curtis and Bishop (1984) and Bishop (1989) have assessed
the strength behavior of woven fabric composites. It has been

Fig. 6.54. Comparisons of the sliding wear rates of unreinforced PEEK, as

well as unidirectional and two-dimensional fabric carbon/PEEK compos-
ites, at 50°C and pv = 0.3 MPa m/s. (After Mody, Chou and Friedrich
1988.)

S
3.

Unreinforced Unidirectional Two-dimensional

PEEK composites fabric composites
372 Two-dimensional textile structural composites

concluded that the woven fabrics are effective in limiting the growth
of damage in laminated composites. It is suggested that woven
fabrics be utilized in the 45° layers of a [0°/±45°] laminate with the
unidirectional non-woven layers providing the needed stiffness and
strength in the loading direction. Bishop has devised a scheme for
laying up balanced fabric laminates without warping and unneces-
sary residual stresses; the line of crimped fibers in the fabric is an
important parameter in the design of the lay-ups. The mechanical
performance of plain and notched laminates under tensile, com-
pressive and fatigue loadings has been reported by Bishop (1989).
To further demonstrate the damage tolerance of woven fabric
composites, the example of molded-in holes is discussed below. The
process of molding holes into the fabric at the laminate fabrication
stage, instead of drilling the holes in the finished laminate, takes
advantage of the microstructure of the woven preform. As a result,
in the vicinity of the hole the fiber volume fraction is increased at
regions where the stress concentrations are high, and the continuity
of fiber is maintained.
Chang, Yau and Chou (1987) and Yau and Chou (1988) have
examined the notched strength in tension and compression for
Kevlar/epoxy and carbon/Kevlar/epoxy hybrid laminates. Speci-
mens with molded-in holes exhibit tensile failure strengths which
are up to nearly 40% higher than those of drilled specimens. Figure

Fig. 6.55. Molded-in holes in a carbon-Kevlar/epoxy [0°]4s laminate.

(After Chang, Yau and Chou 1987.)
 Mechanical properties 373

6.55 shows the fiber geometry around molded-in holes in a

carbon-Kevlar/epoxy [0°]4s hybrid laminate. The compression be-
havior of woven carbon fiber reinforced epoxy composites with
molded-in holes can be found in the work of Ghasemi Nejhad and
Chou (1990a&b).
7 Three-dimensional textile structural
composites

7.1 Introduction
Three-dimensional textile preforms are fully integrated
continuous-fiber assemblies with multi-axial in-plane and out-of-
plane fiber orientations (Chou, McCullough and Pipes 1986; Ko
1989a). Composites reinforced with three-dimensional preforms
exhibit several distinct advantages which are not realized in
traditional laminates. First, because of the out-of-plane orientation
of some fibers, three-dimensional preforms provide enhanced
stiffness and strength in the thickness direction. Second, the fully
integrated nature of fiber arrangement in three-dimensional pre-
forms eliminates the inter-laminar surfaces characteristic of lamin-
ated composites. The superior damage tolerance of three-
dimensional textile composites based upon polymer, metal and
ceramic matrices has been demonstrated in impact and fracture
resistance. Third, the technology of textile preforming provides the
unique opportunity of near-net-shape design and manufacturing of
composite components and, hence, minimizes the need for cutting
and joining the parts. The potential of reducing manufacturing costs
for special applications is high. The overall challenges and oppor-
tunities in three-dimensional textile structural composites are very
fascinating.
Three-dimensional textile preforms can be categorized according
to their manufacturing techniques. These include braiding, weaving,
knitting and stitching, as shown in Fig. 7.1.
There are three basic braiding techniques for forming three-
dimensional preforms, namely 2-step, 4-step and solid braidings. In
the case of 2-step braiding, the axial yarns are stationary and the
braider yarns move among the axials. Thus, the axial yarns are
responsible for the high stiffness and strength in the longitudinal
direction and relatively low Poisson contraction. A high degree of
flexibility in manufacturing can be achieved in 2-step braiding by
varying the material and geometric parameters of the axial and
braider yarns.
Flexibility in the manufacturing of 4-step braids is somewhat
less than that of 2-step braids. All yarn carriers change their
positions in the braiding process and do not maintain a straight
 Introduction 375

configuration. As a result, the preforms exhibit relatively high

Poisson contractions. In order to enhance the longitudinal stiffness
and strength, straight laid-in yarns are often employed.
It can be demonstrated that 4-step and 2-step braidings are
merely variations of a general braiding scheme. By inserting some
axial yarns and placing braiding yarns at proper locations on the
braiding machine, a 4-step braiding process can be converted to a
2-step braiding process.
Besides the more recently developed 2-step and 4-step braidings,
which involve the sequential and discrete movement of yarn
carriers, the maypole type braiding technique is also capable of
producing three-dimensional solid braids. Both square and circular
shapes are feasible. The technology of solid braiding has been well
developed, and commercial machines are available with the maxi-
mum number of carriers currently limited to 24. The application of
solid braids to composite materials is limited to simple shapes.
In woven preforms, there are two major categories. The angle-
interlock multi-layer weaving technique requires interlacing the
yarns in three dimensions. The warp yarn in this three-dimensional
construction penetrates several weft layers in the thickness direc-
tion, and therefore the preform structure is highly integrated. In
orthogonal wovens, the yarns assume three mutually perpendicular
orientations in either a Cartesian coordinate system or a cylindrical
coordinate system. The yarns in the Cartesian weave are not wavy,
and as a result matrix rich regions often appear in the composites.
The process of stitching is mainly based upon an existing
technology for converting two-dimensional preforms to three-
dimensional ones. Because of the simplicity of the stitching opera-
tion, it is feasible to join composite parts continuously in a
cost-effective manner. Both lock stitch and chain stitch have been

Fig. 7.1. Three-dimensional textile preforms.

Three-dimensional textile preforms

, I , ,
I I , ,
Braiding Weaving Stitching Knitting
I I I I
4-step 2-step Solid Angle- Orthogonal Lock Chain Multi-axial
t interlock stitching stitching warp knit
Square Circular
Cartesian Cylindrical
376 Three-dimensional textile structural composites

utilized. Major concerns of the stitching operation include depth of

penetration of the stitching yarns and, hence, the thickness of
two-dimensional preforms that can be stitch-bonded, as well as the
degree of sacrifice of the in-plane properties due to the damage to
in-plane yarns.
The technique of knitting is particularly desirable for producing
preforms with complex shapes because the variability of the
geometric forms is almost unlimited. The large extensibility and
conformability of the preforms enable them to be designed and
manufactured for reinforcing composites subject to complex loading
conditions. The versatility of knitted preforms offers a new dimen-
sion in textile structural composites technology.
In this chapter the discussion of knitting is focussed on the
conversion of two-dimensional structures (for example, unidirec-
tional laminae) to three-dimensional ones through knit-loop-
bonding. In this process, the two-dimensional layers or structures
are formed at the same time when they are bonded. The technology
of multi-directional multi-layer warp knit, for instance, is attractive
because it enables the bonding of the unidirectional lamina by
knitting yarns whereas the yarns in an individual lamina remain
straight. In other words, unlike the stitch-bond of woven fabrics,
the yarns in the two-dimensional structure are not wavy and hence
do not sacrifice their stiffness and strength in the principal material
directions. The manufacturing process is highly integrated, and the
properties in the through-the-thickness direction depend upon the
density and material of the knitting yarn. The potential of knitting
in producing cost-effective thick laminates is attractive.

7.2 Processing of textile preforms

This section outlines the processing techniques of braiding,
weaving, stitching and knitting for making three-dimensional textile
preforms, with particular emphasis on braiding and weaving.
According to Du, Popper and Chou (1991), braiding can form
shapes either by overbraiding mandrels in conventional circular
machines or by using new braiding patterns to form solid shapes
directly. Weaving can be done by using either conventional looms
with multi-layer constructions or entirely new equipment. Knitting
can be used to interconnect fiber arrays that have been arranged by
other techniques. Stitching has been used to interconnect layers of
two-dimensional fabrics for achieving desired thickness and inter-
laminar strength.
 Processing of textile preforms 377

7.2.1 Braiding
Three-dimensional braids have been produced on tradi-
tional horn-gear machines. At the present time, horn-gear based
braiding machines use a small number of yarn carriers (<24) and
cannot form complex shapes. Their applicability is therefore lim-
ited. A number of new machines have been developed to create
complex shapes. These newer braiding processes include 2-step
(Popper and McConnell 1988), AYPEX (Weller 1985), interlock
twiner (Cole 1988), and row and column (Florentine 1982), which
is also referred as Omniweave, Magnaweave, or 4-step in the
literature.
A schematic view of a set-up for the three-dimensional braiding
process is shown in Fig. 7.2. Axial yarns, if present in a particular
braid, are fed directly into the structure from packages located
below the track plate. Braiding yarns are fed from bobbins mounted
on carriers that move on the track plate. The pattern of motion of
the braiders and the presence/absence of axial yarns determine the
type of braids, as well as the microstructure. The processes of 2-step
and 4-step braiding are introduced below.

7.2.1.1 2-step braiding

The preform structure of a 2-step braid includes a large
number of parallel (axial) yarns aligned for efficient reinforcement
and a smaller number of braiding yarns (braiders) that interconnect

Fig. 7.2. A set-up for three-dimensional braiding. (After Du, Popper and
Chou 1991.)
Take-up
mechanism
r" ^
Braid
Forming plate

Braiding yarn
\
Braiding yarn
carrier

Axial yarn
-—i L L L L L L
378 Three-dimensional textile structural composites

the axial yarns and form the fabric shape. The axial array can be
arranged in essentially any shape, including I-beams, box beams,
circular tubes, etc., whereas the braiders are arranged around the
perimeter of the axial array as shown in Fig. 7.3. In the braiding
process, the braiders move through the axial array in two sequential
steps. In the first, the braiders all move in one diagonal line but in
alternating directions (Fig. 7.3a). In the second, they move along
the other diagonal line (Fig. 7.3b).
Although the machine action consists of only two steps, each
braider moves through a larger portion of the structure. This can be
seen by tracing the path of a single braider subjected to the
repeated 2-step machine action. The paths followed by all braiders
will completely intercinch the axial yarns and lock them in the
desired shape.
Compared with other three-dimensional braiding processes, 2-
step braiding has several distinct advantages. A relatively simple
sequence of braider motions can form a wide range of shapes.
During each step of the process, all the braiders are simultaneously
outside of the axial array, and thus it is possible to add various
inserts to the structure or even rearrange the axial array geometry
to change the preform cross-section. Furthermore, this structure can
be made with a high level of fiber packing and a large number of
axially oriented fibers as needed in many applications (Du, Popper
and Chou 1989, 1991).
The 2-step process has motivated a number of researchers. Li and
El Shiekh (1988) modeled the microgeometry using idealized

Fig. 7.3. 2-Step braiding pattern showing the relative motion of yarns.
(After Du, Popper and Chou 1991.)

(a) Step 1 (b) Step 2

Axial yarn

Braider yarn
 Processing of textile preforms 379

circular yarns. Ko, Soebroto and Lei (1988) and Whitney (1988)
have evaluated the mechanical properties of consolidated 2-step
composites.

7.2.1.2 4-step braiding

The 4-step braiding process, so named by Li, Kang and El
Shiekh (1988), requires four distinct Cartesian motions of the yarns
in the fabric cross-sectional plane in each machine cycle. Following
G. W. Du (private communication, 1990), the 4-step braiding
process is depicted in Fig. 7.4 for a 1 x 1 set-up in which the yarn
carriers are arranged in a rectangular plane with eight columns
(m = 8) and four layers (n = 4). Here, the yarn carriers are
indicated by the circles, and can move along the y and z direction
tracks. The process is termed 1 x 1 if the distance traveled by a
carrier in each machine step is equal to the inter-yarn spacing in the
y ox z direction. Other braiding patterns (i.e. 1 x 2 , etc.) are
feasible, which require machine set-ups different from that of Fig.
7.4. It is noted that the carriers occupy alternating positions on the
perimeter of the set-up. The total number of carriers in the m x n
rectangular slab for the l x l braiding pattern is (Li, Kang and El
Shiekh 1988):

N = mn+m+n = (m + l)(n + 1) - 1 (7.1)

Thus, for the 8 x 4 array, there are 44 carriers.
Consider the starting carrier positions as shown in Fig. 7.4(a). In
step-1 of the machine cycle, all the rows of carriers move in the
y direction; adjacent rows move in opposite directions as indicated
by the arrows. In step-2 of the machine cycle (Fig. 7.4b), all the
columns of carriers move vertically; adjacent columns move in
opposite directions as indicated by the arrows. Note that in step-1
and step-2 movements, the carriers on the perimeter of the set-up
remain stationary. The displacement of an individual carrier can be
identified (for example, carriers marked A and B in Fig. 7.4a). Step
3 (Fig. 7.4c) is similar to step-1 except that the directions of
movement of the same row are opposite to each other. The same
comparison can be made between step-2 and step-4 (Fig. 7.4d).
These four steps comprise a machine cycle, because at the end of the
cycle (Fig. 7.4e) the carrier arrangement is the same as that at the
beginning of the machine cycle, although the individual carriers
have changed their locations.
It is interesting to note that the 44 carriers in the slab of Fig.
7.5(a) can be divided into four groups. These are denoted as groups
380 Three-dimensional textile structural composites

Fig. 7.4. Yarn carrier configurations and movements in a 4-step braiding

set-up. (After G. W. Du, private communication, 1990.)

(e)
Processing of textile preforms 381

Fig. 7.5. The four yarn carrier groups in an 8 x 4 slab. Each group defines
a unique yarn path, (a) Yarn carrier location; (b) carrier path for group 1;
(c) carrier path for group 2; (d) carrier path for group 3; and (e) carrier
path for group 4. (After G. W. Du, private communication, 1990.)

(C)

(d)

(e)
382 Three-dimensional textile structural composites

1, 2, 3 and 4. Within each group the carriers are labeled in

alphabetical order from a to k. The characteristic of each group is
that all the carriers within the group share the same path of motion.
For example, carrier la in group 1, moves along the path of
\a^> l&-> lc-> lrf-> le-> 1/-* lg-» l/i-> li-> l/-> lfc-» la (Fig.
7.5b). All the other carriers in this group follow the same path. The
paths of groups 2, 3 and 4 are indicated in Figs. 7.5(c), (d) and (e),
respectively. The path for group 1 carriers in Fig. 7.5(b) is not
marked directly on the carriers to avoid overlapping and confusion;
the same is true for Fig. 7.5(c). The movement of a carrier, for
instance, from position la to lb, or lb to lc, etc. requires one
machine cycle which comprises the steps as shown in Fig. 7.4. The
complete cycle of movement of a carrier, i.e. la—>lfe—>- • -—»la
(returning to the original position) is termed a repeat.
Li, Kang and El Shiekh (1988) have shown that the number of
yarn groups in an m x n slab is given by
G = mn/LCM(m, n) (7.2)
where LCM(m, n) denotes the least common multiple of m and n.
Furthermore, each group has the same number of carriers, which is
N/G. The number of machine cycles required for all the carriers to
return to their original positions is thus also equal to N/G. It should
be noted that the above discussions are valid only for the l x l
braiding pattern.

7.2.1.3. Solid braiding

The term solid braiding is used here to describe the
category of three-dimensional preforms produced by the continuous
intertwining of yarns in the maypole fashion. Figure 7.6(a) shows
the horn-gear set-up for square braiding. The longitudinal and
cross-sectional views of some square braids are given in Fig. 7.6(b).
Solid braids with circular cross-sections are also available. However,
it is not feasible to produce three-dimensional preforms with
complex shapes using solid braiding.

7.2.2 Weaving
Advances in textile manufacturing technology are rapidly
expanding the number, type and complexity of preforms which offer
reinforcements in the through-the-thickness direction. The tradi-
tional weaving technique for producing two-dimensional fabrics has
been modified to achieve a much higher degree of integration in
fiber geometry in the thickness direction. Angle-interlock weaving
 Processing of textile preforms 383

and orthogonal weaving are the two distinct techniques by which

the fibers are incorporated at an angle and parallel to the thickness
direction, respectively.

7.2.2.1 Angle-interlock multi-layer weaving

Angle-interlock multi-layer woven fabrics for thick section
composite applications can be produced on either a dobby loom or
a Jacquard loom. The cam-system is limited to fabricating double-
or triple-layer cloth. Yarns or fibers in angle-interlock multi-layer
wovens are interlaced in a manner similar to two-dimensional
woven structures, except that warp fibers may penetrate more than

Fig. 7.6. (a) Horngear set-up for square braiding. (After Ko 1989a.) (b)
Examples of square braids. (After Steeger 1989.)

(a)
384 Three-dimensional textile structural composites

one layer of weft yarns. The warp direction again coincides with the
machine direction, just as in two-dimensional wovens, whereas the
filling yarn insertion takes place in the transverse direction. Many
other preform configurations are possible, such as those with laid-in
non-crimp yarns (to reduce Poisson's effect), or a combination of
different fiber materials within the same preform (Whitney and
Chou 1988, 1989).
Many variations in the basic geometry of angle-interlock preforms
are feasible, depending on the number of layers interlaced, the
pattern of repeat, and the presence of laid-in yarns. Whitney (1988)
has discussed the fiber architectures in which all warp yarns
interlace the same number of weft yarns. In order to demonstrate
the geometric variability of angle-interlock fabrics, a highly ideal-
ized example is given in the following. Discussions are based on the
fabric structure of the l x l pattern, i.e. the warp yarn orientation
can be represented by one inter-yarn spacing in the horizontal
direction and one inter-yarn spacing in the vertical direction, as
shown in Fig. 7.7.
Following Byun, Leach, Stroud and Chou (1990a), the key
geometric parameters for identifying the preform microstructure
include the number of weft yarns in the thickness direction (n f), as

Fig. 7.7. Three-dimensional angle-interlock woven preforms as identified

by [nf, nft]: (a) [5, 2], (b) [5, 4], and (c) [6, 6]. (After Byun et al. 1990.)
Warp yarn
«fi . / , Weft yarn

(a) (b)

(c)
 Processing of textile preforms 385

well as the number of weft yarns interlocked by a warp yarn in the

thickness direction (n it ) and in the length direction (nn). Parametric
relations are obtained based on the preform structures which have
the following restrictions: (1) The fabric structure is symmetric in
the thickness direction with respect to the mid-plane. (2) The weft
yarns have the same degree of interlocking by warp yarns. (3) The
number of weft yarns in the thickness direction is the same along the
warp direction. Employing the notation of [nu nft\, the woven
preforms of Figs. 7.7(a), (b) and (c) can be identified as [5, 2], [5, 4]
and [6,6].
The following relationship needs to be satisfied to ensure the
interlocking of weft yarns by warp yarns for the l x l pattern:
nft = nn (7.3)
With the above condition, a maximum number of warp yarns can be
achieved in the preform. For a [nfy nn] weave, the total number of
warp yarns (n w ) is 2nf. However, not every warp yarn interlocks
with nft weft yarns. This can be seen from Figs. 7.7(a) and (b)
where the warp yarns at the top and bottom faces only interlace
with the weft yarns in the surface layers. The degree of reinforce-
ment in the thickness direction is related to the number of warp
yarns (nwi) that interlock with the nn weft yarns. When nf = nft, all
the warp yarns interlock with all nft weft yarns, i.e. n wi = 2nft. When
ft, n wi is given as follows:

wi = 2knit for k < nf/nft < k + 1 (7.4)

nwi = (2k - l)ntt for n f /n ft = k (k>2) (7.5)
where k is an integer. For the fabrics of Figs. 7.7(a), (b) and (c), the
n wi values are, respectively, 8, 8 and 12.
Thus, the number of warp yarns (/?wn) which do not interlace n{t
weft yarns is
nwn = 2nf-nwi (7.6)
In Figs. 7.7(a) and (b), n wn = 2 and in these cases each warp yarn at
the surface only interlace with one layer of weft yarns. However,
the nwn warp yarns can also interlace with the multi-layer of weft
yarns. Consider a fabric preform with the [5,3] weave pattern and
nwn = 4. Figures 7.8(a) and (b) show the two possible configurations
of the warp yarns near the free surfaces.
When the condition of Eq. (7.3) is not satisfied, the resulting
fabric is not highly integrated and there are non-interlaced yarns. In
386 Three-dimensional textile structural composites

the following, the case of three-dimensional weaves with non-

interlaced weft yarns is discussed. An additional geometric para-
meter, nws, is identified; it denotes the number of weft rows shifted
by adjacent warp planes. Using the notation of [n{, nfu n ws ], the
fabrics of Figs. 7.9(a), (b) and (c) can be identified as [6,2,1],
[5,3,0] and [5,3,1], respectively.
Finally, the total number of warp yarns can be obtained as
for nft = 2
(7.7)
nf = 2ntt for nft > 3
Thus, the number of warp yarns interlaced through the nft weft
yarns is
n w i = n , - 2/i w s (7.8)
It should be noted that the parametric relations for a three-
dimensional weave in which every weft yarn is interlocked with
warp yarns can also be obtained in the case that Eq. (7.3) is not
satisfied.

Fig. 7.8. Two variations of the [5,3] weave with « wn = 4. the two warp
yarns near the surface interlace with (a) one weft yarn layer or (b) two
weft yarn layers. (After Byun et al. 1990.)

(a) (b)

Fig. 7.9. Three-dimensional angle-interlock woven preforms as identified

by [n f ,n ft ,n ws ]: (a) [6, 1, 1], (b) [5,3,0], and (c) [5,3, 1]. (After Byun et
al 1990.)
"fl

(a) (b) (c)

 Processing of textile preforms 387

7.2.2.2 Orthogonal weaving

Figure 7.10 shows an orthogonal woven fabric where the
yarns are placed in three mutually orthogonal directions. Because of
the nature of fiber placement, matrix rich regions are created in
composites reinforced with orthogonal woven preforms. Since the
thickness direction yarns are incorporated into the preform in the
weaving process, they do not cause damage to the in-plane fibers.
This is different from the case of stitching bonding of two-
dimensional fabrics. Orthogonal woven fabrics can be fabricated by
maintaining one stationary axis either by predeposition of the yarn
system or a space rod which is subsequently retracted and replaced
by axial yarns. The two sets of yarns in the plane perpendicular to
the axial yarns are then inserted in an alternating manner (Ko
1989a). Both Cartesian and cylindrical woven fabrics are available.

7.2.3 Stitching
The process of stitching for making three-dimensional
preforms is relatively simple. The basic needs include a sewing
machine, needle and stitching thread. The processing variables are
stitch density (stitch/unit length), the size of the stitch thread, and types
of stitch. Both lock stitch and chain stitch are available (Fig. 7.11).
A lock stitch becomes unbalanced if the tension in either the bobbin
thread or the needle thread is higher than that in the other thread.
The necessary clearance between the feed and dog as well as the
length of the needle stroke in the case of lock stitching, for instance, are

Fig. 7.10. An orthogonal woven fabric. (After Chou, McCullough and

Pipes 1986.)
388 Three-dimensional textile structural composites

Fig. 7.11. (a) Lock stitch and (b) chain stitch seams. (After Ogo 1987.)

Balanced lock stitch Unbalanced lock stitch

111=113
Needle
thread

Bobbin Bobbin
thread thread
(a)
Chain stitch

(b)

Fig. 7.12. Schematic of the lock stitch process. (After Ogo 1987.)

Needle and needle thread

Bobbin and bobbin thread

 Processing windows for 2-step braids 389

determined by the thickness of the two-dimensional preform to be

stitch-bonded. Figure 7.12 shows the schematic of a lock-stitching
process for bonding fabric layers. The needle thread needs to be
abrasion resistant and can be bent to small curvature in the needle
hole.

7.2.4 Knitting
Three-dimensional knitted fabrics can be produced by
either a weft knitting or warp knitting process. For additional
strengthening in the 0° and 90° directions, laid-in yarns can be
placed inside the knitting loops. Figure 7.13 shows a weft knit fabric
with laid-in weft and warp yarns.
The most promising knitted preform which provides a high
degree of structural integration in the thickness direction is perhaps
the multi-axial warp knit. It consists of warp (0°), weft (90°) and bias
(±6) yarns held together by a chain of tricot stitch through the
thickness of the assembly (Fig. 7.14). Different kinds of multi-axial
warp knits have been developed. The main attraction of the knitted
construction is that it possesses the advantage of unidirectional
laminates while also providing enhanced stiffness and strength in the
thickness direction (Ko, Pastore, Yang and Chou 1986).

7.3 Processing windows for 2-step braids

The purpose of the following discussions is to demonstrate
that knowledge of the microgeometry and structure of textile
preforms provides the basis for understanding flexibility in
processing. The work of Du, Popper and Chou (1991) in 2-step
braiding is recapitulated as an example of such an approach. The

Fig. 7.13. Weft knit with laid-in weft and warp yarns. (After Ko 1989a.)
390 Three-dimensional textile structural composites

key inputs of the analysis are (1) the size, type and shape of
braiders and axial yarns, (2) the braid pattern (size of axial yarn
array), and (3) the advance rate during braiding. The key outputs
are braid dimensions, fiber orientation, inter-yarn void content,
fiber volume fraction, and geometric limits imposed by yarns
jamming against each other. The modeling work is for preforms of
rectangular cross-section. However, the methodology regarding
yarn cross-sections, unit cells, and yarn jamming can be used to
analyze more complex shapes, as well as other types of three-
dimensional fabrics.
The major assumptions are: (1) Multi-filament yarns are used for
both braiders and axial yarns. These yarns are composed of a large
number of fibers, and their cross-sections can be readily deformed
to prismatic shapes. (2) Fiber cross-section is round. (3) Fibers are
parallel along the yarn length, i.e. zero twist. (4) Yarn tension is
high enough to ensure a straight yarn path, except for the braider
yarns, which are bent around the braid surface. (5) Filaments are
inextensible.

7.3.1 Packing of fibers and yarn cross-sections

The fiber volume fraction of a three-dimensional preform
depends on the level that fibers pack against one another in a yarn
and the level to which yarns pack against one another in the
structure. Two methods for estimating inter-fiber packing are
described in this section. Section 7.3.3 discusses yarn packing in
preforms.
The geometry of inter-fiber packing in yarns has been studied
primarily for textile applications (see Hearle, Grosberg and Backer

Fig. 7.14. Mult-axial warp knit fabric. (After Chou, McCullough and
Pipes 1986.)
 Processing windows for 2-step braids 391

1969). Two basic idealized packing forms can be identified: open

packing (Fig. 7.15a) and close packing (Fig. 7.15b), in which
the fibers are arranged in concentric and hexagonal patterns,
respectively.
In open-packed yarns the packing fraction, defined as the
fiber-to-yarn area ratio, has been computed as a function of the
number of fibers. If the outer ring is completely filled and the fibers
are circular, the yarn packing fraction is

(7.9)
° (2Nr-l)2
where NT is the number of rings, and its relationship to the number
of fibers, N{, is given by

For a large number of fibers, KO approaches 0.75.

In close-packed yarns, for any number of circular fibers if the
outer layer is completely filled, the yarn packing fraction equals the

Fig. 7.15. Fiber packing in yarns, (a) Open packing in a circular yarn, (b)
Close packing in a hexagonal yarn, (c) Open packing in a diamond-shaped
yarn, (d) Close packing in a diamond-shaped yarn. (After Du, Popper and
Chou 1991.)

(c) (d)
392 Three-dimensional textile structural composites

area ratio of a circle to the hexagon in which the circle is inscribed:

K
K = = 0.91 (7.11)
:2\/3
The yarn packing fractions predicted by the two models assume
circular and hexagonal yarn cross-sections. However, as shown in
Figs. 7.15(c) and (d), they apply equally well to other shapes if the
number of fibers is sufficiently large. Factors that affect both the
packing of fibers in a yarn and the packing of yarns in a preform
include yarn tension, inter-yarn contact, yarn twist, fiber cross-
section, fiber straightness, manufacturing method, and preform
geometry.
In addition to the level of yarn packing fraction, the yarn
cross-sectional shape plays a significant role in determining how
many fibers can be packed into a fabric. In the textile literature, the
yarns are often assumed to have a circular cross-section (see Peirce
1937; Brunnschweiler 1954). However, it has been shown that the
cross-section of even highly twisted yarns deviates significantly from
a circular shape and the yarn cross-section varies considerably for
different types of preforms. Many attempts have been made to
develop more realistic geometric models for yarns in fabrics by
assuming elliptical and race-track cross-sections (Hearle, Grosberg
and Backer 1969).

Fig. 7.16. Cross-sections of axial yarns in a rectangular braided preform

before consolidation. (After Du, Popper and Chou 1991.)

mm
i - Center axial

\ - Side axial

- Side axial

- Corner axial
 Processing windows for 2-step braids 393

In the following, a model for the yarn cross-section in 2-step

braids is developed. It is observed from composite specimens that
the yarns in the preform have cross-sections as shown in Fig. 7.16.
After matrix addition and consolidation in a mold, the fabric is
observed to be flattened, as shown in Fig. 7.17. The axial yarns
have different cross-sections depending on their locations in the
preform. Central yarns, which form the bulk of the structure, are
diamond-shaped. Axial yarns on the side and corners of the
preform are pentagonal. Braiding yarns, which occupy the space in
between the axial yarns, are rectangular.
The aspect ratio of axial yarns, / a , is related to the inclination
angle of the braiders (Figs. 7.16 and 7.18) and is given by

/ a = ^ = tan0 (7.12)

where / a is influenced by braider yarn tension or external lateral

compression applied at the forming point during the process. It can
also be changed by compacting the entire braided preform during
matrix consolidation. These aspect ratios affect the shape of the
final braid as well as the braider yarn orientation angle (a) and fiber
volume fraction (Vf). With unit axial aspect ratio (6 = JI/4), the
cross-section of the center axial yarns becomes square. In this
special case, the fiber volume fraction is at a maximum.
The axial yarn dimensions can be calculated from the cross-
section of the consolidated braid in Fig. 7.17. These relations are

Fig. 7.17. Cross-sections of axial yarns in a rectangular braided composite.

(After Du, Popper and Chou 1991.)

WM
)WO 00\J [>
AOAOAOAOAOJ 1A
T
394 Three-dimensional textile structural composites

given in terms of yarn area and inclination angle. The yarn area is in
turn evaluated from its linear density and the fiber packing fraction
and fiber density:

p a *- a sin(20)
0.5Aa
(7.13)
, sin 6
0.5Aa
Sn=
aKa cos 6

The parameters Aa and p a are the linear density and the fiber
density of axial yarns, respectively. The packing fraction, tc a, is
assumed to be constant for all axial yarns.

Fig. 7.18. Path of one braiding yarn in the fabric: (a) braid pattern ; (b)
top view; and (c) front view. (After Du, Popper and Chou 1991.)
1 2

(a) .

(b) t

(c)
 Processing windows for 2-step braids 395

For, only one braider yarn is shown in Fig. 7.18 for two steps of
the braiding process. These yarns are assumed to be rectangular
with aspect ratio fh. The aspect ratio is usually much less than unity
because of compression by the axial yarns. The dimension of a
braider yarn can also be calculated from its packing fraction (jcb),
yarn linear density (Ab), and fiber density (p b ):

<714
>

7.3.2 Unit cell of the preform

In order to understand the microscopic arrangements of
yarns, it is necessary to identify the 'unit cell' of the fabric preform.
By definition, a unit cell constitutes the smallest repeating entity in
the structure. The complexity of three-dimensional preform struc-
tures often makes the identification of unit cells a difficult task.
The unit cell of the 2-step braid is composed of four sub-cells,
labelled A, B, C and D in Fig. 7.19. The repeat of these four
sub-cells will generate the whole braided structure. Because of
geometric similarity, any one of these four can be utilized to derive
the basic structural characteristics. The length of the unit cell in Fig.
7.19 is the length of braid formed in one machine step. This length
is actually half of the fabric pitch length (P), as shown in Figs.
7.18-7.20. Five layers are shown in Fig. 7.19. The number of
columns has been assumed to be very large so that the rather
complicated edge configuration of the preform can be avoided.
Figure 7.20 shows the difference between finite and infinite columns
and their effects on braider paths. Figure 7.20(a) shows the yarn
path on the surface of a specimen seven columns wide. Note that
the trace of the braider yarns lies on an inclined line. In an infinitely
wide specimen (Fig. 7.20b) the trace of the braider yarns is
perpendicular to the braiding direction.
The width and thickness of the braided preform can be computed
from Figs 7.17 and 7.18 in terms of m (number of axial columns)
and n (number of axial layers) as well as yarn geometric and
material parameters:

w = (m- l)(25 a cos 0 + J j ^ ) + 2(5m + /bSb) (7.15)

t = {n- l)(s a sin 6 + ^ ^ ) + 2(Sn +/b5b) (7.16)

396 Three-dimensional textile structural composites

From Eqs. (7.15) and (7.16), the aspect ratio of the braided
preform can be obtained as

f = z. ( 7 - 17 )
The braider yarn orientation can be determined by computing the
projected length (i.e. segment P1P2 in Fig. 7.18b) of one braider
over one half of the pitch length. Note that the angle between a
braider and the axial yarns (a) appears to vary on the front view in
Fig. 7.18(c). This apparent variation occurs because a segment of a
braider yarn in the interior of the preform has a different projected
angle compared to a segment on the preform surface. The projected

Fig. 7.19. Unit cell model of a 2-step braided preform showing four
sub-cells. Each sub-cell includes a braider yarn and a number of axial
yarns. (After Du, Popper and Chou 1991.)
 Processing windows for 2-step braids 397

length L p of the segment PXP2 in the axial direction is

Lp = t esc 6 + 2Sn(l - esc 6) + 2Sa cos 6 (7.18)
The braider yarn angle is then given by

(7.19)

The total length of one braider yarn in a unit cell is

(7.20)
sin a 2 cos a
Then the volume of the braider yarn (vh) and the volume of axial
yarns (va) in a unit cell and the total volume of a unit cell (vt) can
be determined:
vb = LbfbSl (7.21)
P
va = - Sa[(n - l)5 a sin(20) + 4Sn cos 0] (7.22)

-fbSb[(n - l)5 a + L p ] (7.23)

The fiber volume fraction Vf (total fiber volume/unit cell volume),

the braider fiber volume fraction Vb (braider fiber volume/unit cell
volume), and the volume fraction of the void Vy (volume of

Fig. 7.20. Effect of fabric width on braid geometry: (a) finite-width

preform; (b) infinite-width preform. (After Du, Popper and Chou 1991.)
m=1 m -*- °°

Sb

Axial yarn

^ - Braider yarn
(a) (b)
398 Three-dimensional textile structural composites

inter-yarn voids/unit cell volume), can then be obtained:

(7.24)

(7.25)

V=l- (7.26)

7.3.3 Criterion for yarn jamming

The allowable microstructural states of a fabric preform are
limited by the condition at which the yarns jam against one another.
Knowledge of yarn jamming is essential in identifying the
processing windows of fabric preforms. Although jamming is
discussed frequently in the textile literature, it is often neglected in
the analysis of composites.
In 2-step braids, the braider angle becomes very small as the
braider yarns become parallel to the axial yarns. However, as the
pitch length is reduced, the braider angle increases, and a limiting
state is reached in which the yarns jam against one another. If all
other parameters remain constant, the pitch length cannot be
reduced further. The state of jamming is illustrated in Fig. 7.21
where the yarn-to-yarn contact is shown. This rather complex

Fig. 7.21. Surface geometry of braid at jamming. (After Du, Popper and
Chou 1991.)
m -»OO
 Processing windows for 2-step braids 399

limiting state, however, can be described simply by

1 (7.27)
sin a
where Pj denotes the pitch length at jamming. Equation (7.27) is
applicable to specimens of finite width.
Due to the edge effect of finite-width structures, the orientation
angle, a, of all braiders are not equal. In a braiding step, the
braiders on the side surface of the preform will have shorter length
than those in the interior. However, since all yarns advance at the
same pitch length, the 'edge' braiders will lie at a somewhat lower
angle than those passing through the center of the structure.
Du, Popper and Chou (1991) have conducted experiments to
measure the geometric and material parameters of 2-step braids. A
comparison of measured and predicted values of two samples are
given in Table 7.1. Braid I is rectangular in cross-section, consisting
of 12-column by five-layer axial yarns (Kevlar-49 with a linear

Table 7.1 Material and geometric parameters of 2-step braided

preforms. After Du, Popper and Chou (1991)

Braid I Braid II

Parameters Measured Computed Measured Computed

K (g/m) 9.57 3.33

K (g/m) 2.39 0.25
m 12 7
n 5 5
P(mm) 17.8 4.98
/a 0.78 1.32
/b 0.12 0.05
Ka 0.70 0.78
Kb 0.70 0.78
t (mm) 9.0 8.9 7.0 7.3
w (mm) 63.0 62.9 15.0 14.6
65.0 65.9 75.0 76.7
«o
V (%)
f 56.0 56.8 73.0 73.6

Braid I: bare fiber preform.

Braid II: infiltrated with matrix.
400 Three-dimensional textile structural composites

density of 9.57 x 10"3 kg/m) and 15 braider yarns (Kevlar-49 with a

linear density of 2.39 x 10~3kg/m). Braid II is also rectangular in
cross-section, with seven-column by five-layer axial arrays and ten
braiders. In Braid II, all axial yarns are made of Kevlar-29 with a
linear density of 3.33 x 10~3 kg/m: Kevlar-49 is used for the braider
yarns which are much finer than the axial yarns with a linear density
of 2.53 X 10 4 kg/m. Braid II was impregnated with an epoxy by
resin transfer molding.
Based upon the relations between process variables and fabric
geometry, it has been shown that the range of allowable fabric
structures is dictated by effects such as yarn jamming and fiber
packing. Figure 7.22 demonstrates the processing window for 2-step
braids when the braider yarn orientation angle and pitch length,
total fiber volume fraction, and yarn linear densities are considered.
The processing window is bounded by two limiting states: yarn
jamming and zero braider angle. Preform constructions correspond-
ing to the curved 'jamming' line are at their tightest possible state,
and constructions corresponding to the a = 0 curve have infinite
pitch length. As the dimensionless pitch length P/Sa increases, the

Fig. 7.22. Fiber volume fraction vs. braider-to-axial linear density ratio.
The allowable process window is shown (*:a = 0.8, ieb = 0.8, 6 = 38°,
fh = 0.2, n = 5). (After Du, Popper and Chou 1991.)

80
^^^Tx
Jamming s

/ xf
70

60

£
50

40
= 1000 (o - 0°)

30 l i n i n g i t iimtl i i limit i i until i iiiitn| | , ,,,,J , , J , , 1 , , „ . ,

10 ~ 10~4 10- 3 10 -2 10~l 10° 10 1
102 103
 Processing windows for 2-step braids 401

fiber volume fraction decreases. Increasing the ratio of braider-to-

axial yarn linear density causes the maximum allowable fiber
volume fraction to go through a minimum. At fixed levels of pitch
length, an increase in Ab/Aa first reduces the fiber volume fraction
because the inclusion of larger braider yarns creates more void
space. However, at large Ab/Aa ratios, a higher fiber volume fraction
is realized. At a ratio of about 200, the fabric reaches a limiting
state in which the braider yarn angle approaches zero due to the
infinite pitch length. The fiber packing in the yarns, taken as 0.8,
limits the maximum fiber volume fraction in the fabric.
A 'microstructure map' of 2-step braids, which gives the relation-
ship among fiber volume fraction, pitch length, braider yarn
orientation angle and braider yarn volume fraction, is shown in Fig.
7.23. The minimum allowable fiber volume fraction increases with a
reduction in braider yarn pitch length. For a fixed pitch length and
above the minimum allowable fiber volume fraction, both Vf and a
increase with an increase in braider fiber volume fraction. This map
demonstrates that a wide range of orientation angle and fiber
volume fraction can be achieved by varying the pitch length and the
amount of braider yarns relative to the axials. Maps of microstruc-
tures provide guidance in designing preforms for a specific
application.

Fig. 7.23. Property volume fraction (Vf) vs. fiber orientation angle (a) for
various pitch length and volume fraction of braider yarns (ic a = 0.8,
jfb = 0.8, 9 = 38°, b/ = 0.2, n = 5). (After Du, Popper and Chou 1991.)
Limit due to fiber packing in yarn

40 50 60 70 80 90
a (degrees)
402 Three-dimensional textile structural composites

7.4 Yarn packing in 4-step braids

Knowledge of yarn packing in three-dimensional structures
is essential for determining the unit cell configuration of a fiber
preform as well as the condition for yarn jamming. In a 4-step
braiding process, the braiding yarn carriers move in a two-
dimensional grid with two sets of perpendicular tracks (Fig. 7.4).
For the sake of simplicity, the following discussions are restricted to
4-step braids without laid-in yarns.

7.4.1 Unit cell of the preform

When the specimen cross-sectional area is large, the domi-
nant unit cell configuration can be represented by a parallelepiped
(Fig. 7.24) with the size of PaxPbxPc. The braiding axis is
assumed to coincide with the x axis. Obviously, for a 1 x 1 braid,
Pb = Pc. A unit cell contains four yarns situated along the diagonals.
It is understood that in Fig. 7.24 the yarns are idealized as
geometric lines and, thus, they intersect at the center of the unit
cell.
The details of the yarn arrangement in a 1 x 1 braid can be
visualized by taking the 123'4' cross-section of the unit cell. This is

Fig. 7.24. Unit cell of a 4-step braided preform. (After Yang, Ma and
Chou 1986).

1i

(b)
 Yarn packing in A-step braids 403

shown in Fig. 7.25. Referring to Fig. 7.24, the yarns along the
diagonals 13' and 4'2 are contained in the cross-section, and are at
an angle a to the braiding axis. Yarns of type 4'2 are shown in
Fig. 7.25 by the inclined sections. Yarns of types 1'3 and 2'4
show elliptical sections. Yarns of type 13' are also parallel to the
cross-section and they are blocked by the other three types of yarns
in the cross-sectional view of Fig. 7.25.
It should be noted that Fig. 7.25 is valid for a < 58°, which is the
critical angle for yarn jamming (Section 7.4.2). The pitch length, P,
which is the preform take-up length for one machine cycle (four
steps), is defined in Fig. 7.25, along with the braider yarn
orientation angle a.

7.4.2 Criterion for yarn jamming

The condition for yarn jamming in a 4-step braided preform
can be understood from the yarn geometric arrangements. The
following assumptions are made in the derivation of the yarn
jamming criterion: (1) the braiding yarns are circular in cross-
section, with diameter d, (2) the yarns are in a stable configuration,
namely, each yarn in Fig. 7.24 is in contact with the other three,
and (3) the braid is of the 1 x 1 type.

Fig. 7.25. Fiber configuration in the cross-section 4'3'21 of Fig. 7.24.

(After G. W. Du, private communication, 1990.)

Yarn 24'
Yarn 2'4
_\
Yarn 24'

4'
404 Three-dimensional textile structural composites

Figure 7.25 shows the yarn configuration of the cross-sectional

plane 123'4' before yarn jamming. The pitch length and fiber
diameter are denoted by P and d, respectively. From the relation of
tangency between an ellipsoid (yarn 1'3) and a line (yarn 24'), the
distance OA can be obtained:

OA = - V(l + sec2 a) (7.28)

Another geometric relation for this yarn configuration is

P tan a = Ad = d{\ + sec a + V(l + sec2 a)) (7.29)
which yields the braider yarn angle and aspect pitch length (P/d):
or = 41.4°; P/d = 4.54 (7.30)
As compaction of preform continues, the yarn configuration
finally reaches a limiting state where both the yarns 2'4 and 1'3
touch the yarn 24'. Figure 7.26 shows the yarn configuration at
jamming. The geometric relation for this case is:
P t<in a = 4d = d(sec a + V(l + sec2 a)) (7.31)
Thus, the criteria for yarn jamming are
a = 57.8°; P/d = 2.52 (7.32)
and the conditions a>57.8° and P/d < 2.5 are physically not
feasible. It is interesting to note that jamming in 4-step braids
occurs at a unique yarn orientation angle, which is independent of
the yarn material and processing parameters.

Fig. 7.26. Yarn configuration at jamming (After G. W. Du, private

communication, 1990.)

Yarn 24'
Yarn 2'4 \ Yarn 1'3
 Analysis of thermoelastic behavior of composites 405

The consequence of such a characteristic of 4-step braiding is the

absence of a processing window for providing the flexibility of
manufacturing. However, such a window can be created by using
two or more types of braiding yarns and inserting laid-in yarns, thus
expanding the ranges of fiber geometric and material parameters.

7.5 Analysis of thermoelastic behavior of composites

The analysis of thermoelastic behavior of three-dimensional
fabric composites can be made based upon the knowledge of the
microstructure of the preforms. For the preforms reviewed in
Section 7.2, their unit cell structures are sufficiently well
established.
For braided composites, the unit cells of both 2-step and 4-step
braids are well known, whereas the unit cells for solid braids
depend on the specific preform designs. In the case of weaving, the
unit cell of an angle-interlock woven may occupy the entire preform
thickness. This is true for the preforms shown in Fig. 7.7 where
there are no repeating units in the thickness direction. The unit cell
structures of orthogonal wovens and stitch-bonded preforms are
similar. Because the thickness direction yarns in both cases are
normal to the free surfaces, they can be considered as limiting cases
of the angle-interlock configuration.
The knitting yarns in a multi-axial wrap knit are severely curved.
Because the knitting yarns usually have fine dimension and low
stiffness, their contributions to the composite thermoelastic pro-
perties are perhaps negligible. When high stiffness knitting yarns
are used, their contributions to the thickness direction properties
need to be taken into account. Because of the low volume fraction
of knitting yarns, relative to that of the in-plane fibers, it is not
unreasonable to neglect the in-plane behavior of the knitting yarns
in the composite.
Unlike the case of unidirectional laminates, the thermoelastic
behavior of three-dimensional composites is complicated by the
fiber configuration in the thickness direction. In the following, three
different analytical approaches are outlined. Among them, the
energy approach considers the elastic strain energies due to the
interaction of yarns at an interlock. The fiber inclination model is
based upon the lamination analogy, whereas the macro-cell ap-
proach utilizes stiffness tensor transformation and an averaging
technique.
406 Three-dimensional textile structural composites

Besides these three modeling techniques, a micro-cell approach

has been adopted by Whitney (1988) and Whitney and Chou (1989)
for analyzing angle-interlock woven composites. This technique is
also based upon the lamination analogy. In view of the large
geometric variability of angle-interlock wovens, their elastic be-
havior perhaps can be more efficiently analyzed by the macro-cell
approach.

7.5.1 Elastic strain-energy approach

An elastic strain-energy approach has been adopted by Ma,
Yang and Chou (1986) to derive the elastic stiffness of three-
dimensional textile structural composites. Although their analysis is
for a 4-step braided composite, the methodology has general
applicability.
In the general case, the unit cell structure of Fig. 7.24 can be
considered as composed of three sets of mutually orthogonal yarns
as well as yarns assuming the diagonal positions. The unit cell is
centered on an 'interlock' of these yarns. The analytical model then
then focuses on the interaction of the yarns at the center of the unit cell.
The following assumptions are made in the analysis: (1) The
baseline and diagonal yarns are regarded as 'composite rods' after
being impregnated with matrix materials. The stiffness and strength
of the composite are mainly derived from the three-dimensional
composite rod structure. (2) The composite rods are homogeneous
and linearly elastic, and have uniform circular cross-sections that do
not flatten under external loading. (3) The composite rods possess
tensile, compressive, and bending rigidities. (4) A jamming
force exists at the region of contact between two interlocking
composite rods. The rods can be treated as either compressible or
incompressible under the action of jamming forces.
Because of the complexity of the yarn configurations at their
interlocking positions, the model does not simulate each individual
'lock' separately. The interactions among the yarns are dealt with in
approximate fashion by projecting the yarn positions onto a set of
mutually orthogonal planes. Within each two-dimensional projec-
tion, the interactions between two yarns are taken into account.
Consider, for instance, the interaction of two baseline composite
rods (Fig. 7.27). Three types of elastic strain energies in the
composite rods are taken into account. These include the strain
energies due to bending, extension and compression over the region
of fiber contact. Based upon the knowledge of the elastic strain
energy of the baseline and diagonal composite rods, the elastic
 Analysis of thermoelastic behavior of composites 407

properties of the composite can be obtained through the application

of an energy principle.

7.5.2 Fiber inclination model

The fiber inclination model developed by Yang, Ma and
Chou (1986) can be understood also by considering the yarn
arrangements in a 4-step braided preform. Consider again the unit
cell structure based upon the yarns oriented along the four body
diagonals in a 4-step braided fabric (Fig. 7.24). The three-
dimensional composite can thus be regarded as an assemblage of
unit cells as shown in Fig. 7.28(a), where the emphasis is placed on
the yarn orientation rather than the interaction among yarns. Here
only one set of diagonal yarns in the composite is shown for clarity.
The zig-zagging yarn segments are not confined to one layer only.
Each yarn in the composite extends through the whole length of the
material and changes its orientation at the interlocks. Furthermore,
straight laid-in yarns along the edges of the unit cell can be added in
the present formulation.
The methodology for the analysis of the fiber inclination model
is based upon a modification of the classical laminated plate theory.
The following geometrical characteristics are assumed by Yang, Ma
and Chou (1986): (1) All the yarn segments parallel to a diagonal
direction in the layer ABCD (Fig. 7.28a), for instance, are treated
as forming an inclined lamina (Fig. 7.28b) after matrix impregna-
tion. (2) Fibers within a lamina are considered to be straight and

Fig. 7.27. Yarn interaction at the point of interlock.

\
408 Three-dimensional textile structural composites

unidirectional. Fiber interlocking and bending due to the change of

orientation from one diagonal direction to another at the corners
of the unit cell are not taken into account. (3) A unit cell in Fig.
7.24(a) can be further considered as an assemblage of four inclined
unidirectional laminae. The intersections among the four inclined
laminae are ignored. Each unidirectional lamina is characterized by
a unique fiber orientation and all the laminae have the same
thickness. Furthermore, the fiber volume fraction of each lamina is
assumed to be the same as that of the composite.
The laminate approximation of the unit cell structure is shown
schematically in Fig. 7.29. The geometrical configuration and
stacking sequence of the inclined laminae composed of yarns in four
diagonal directions in the unit cell are given below. First, the
?i — £i — Vi and %2~ b— ?l2 coordinate systems are assigned to

Fig. 7.28. (a) The idealized zig-zagging yarn arrangement in the braided
preform, (b) Schematic view of the inclined laminae representing the
diagonal yarns of the 'fiber inclination model'. (After Yang, Ma and Chou
1986.)

/ / /
/ /
/ / ,
/^ ^ y
(b)
 Analysis of thermoelastic behavior of composites 409

lamina 4'2'24 and lamina 1'3'31, respectively, as shown in Fig.

7.24(b). Referring to Figs. 7.24 and 7.29, the equations describing
the height of each lower surface of laminae 1 and 3 and the height
of each upper surface of laminae 2 and 4 measured from the base
plane (z = 0) of the unit cell are:

lamina 1 (yarn 4'2): //i(£i) = -4— (0 < ^ t < L)

(7.33)

lamina 2 (yarn 1'3): H2(%2) = -j1 ( 0 < | 2 < L)

(7.34)

lamina 3 (yarn 42'): // 3 (§0 = Pc( l - j -

(7.35)

lamina 4 (yarn 13'): // 4(§2) = pjl -—) (0< lj 2< L)

(7.36)
where L = V(Pa + Pb)-
The yarn orientation angles a, /3 and y in Fig. 7.24 are denned as

a = tan

(7.37)

With the above geometrical relations and assumptions, the

three-dimensional braided composite of Fig. 7.24 can be modeled
based upon the classical lamination theory. The approach is
essentially an extension of the fiber crimp model of Section 6.5.
Thus, the constitutive equations of a laminated plate follow Eq.
(6.1). The stiffness constants, Qijy are given by Eq. (6.10).
Since the undirectional yarns in each of the four laminae of Fig.
7.29 are at an angle, y, with respect to the ^-direction (see Fig.
7.24b), the effective elastic properties of lamina 1 in the §-£ plane,
410 Three-dimensional textile structural composites

for example, are given by Eq. (6.74) as

cos4 y
12 £.22

cos2y (7.38)

= 21 c o s 2 y + sin2 y
The transverse isotropy in the plane perpendicular to the yarn
direction has been taken into account. Then the stiffness matrix,
Qij(y), similar to Eq. (6.25) can be written in terms of E^, E^,
Ggj, v & , and Dy = l — v2-j=(y)£5?(y)/£c?.
For lamina 1, the yarn segments in an inclined lamina also form
an off-axis angle, /?, with respect to the braiding direction (x axis).
Thus, the effective laminar elastic properties in the x direction are

Fig. 7.29. Four unidirectional laminae representing the inclined yarns.

(After Yang, Ma and Chou 1986.)
 Analysis of thermoelastic behavior of composites 411

further reduced, and the stiffness constants of the laminae become

GV
P, r) = On Ql2 Qj6 (7.39)
Q26 G « d
where
cos4 2G c (y)]
x cos2 £ sin2 /3 + — sin 4
/3

COS3 ^ s i n

(7.40)

- cos ft sin3J
G26 =

x cos2 p sin2 p + G§c(y)[cos4 p + sin4 P]

Knowing the effective laminae properties with respect to the
x—y coordinate system, the local plate stiffness matrices Atj{x),
Bij(x) and Dtj{x) can be calculated from the lamination theory:

[(A0(x), Bijix), Dv(x)] = 2 fj G,y(i8, y)[l, z, z2] dz

(7.41)
412 Three-dimensional textile structural composites

The integration is performed through the thickness of the unit cell

of Fig. 7.29. By neglecting the contribution of the pure matrix
region, the extensional stiffness matrix Atj{x)y for instance, can be
evaluated as follows:

Aqix) = Q$\p, y) dz + f(/i, y) dz

J J
«l(ll) H2(.£2)-h'
« 4(?2)
Qf(p, Qf(fi,Y)Az
JH3(5I)
(7.42)
where the superscripts (1), (2), (3) and (4) correspond to the
laminae in Fig. 7.29. Also, h' = h/cos y, where h is the thickness of
a lamina. It should be noted that the signs of the angles /3 and y of
the laminae 2, 3 and 4 depend on the fiber orientations. In order to
avoid over-estimation of the composite properties, the portions of
the laminae which lie outside of the unit cell (such as the region
above 0a'32 in Fig. 7.29) have been excluded from the integration
in Eq. (7.41). The lamina thickness is so determined that the total
cross-sectional area of laminae (1), (2), (3) and (4) in the x-z
plane is equal to that of the unit cell.
The inversion of the local stiffness matrices Atj{x), Btj{x) and
Dij(x) of Eq. (7.41) yields the local laminate compliance matrices
Afa), B'ij(x) and D[j{x). The average in-plane compliances of the
unit cell under a uniformly applied in-plane stress resultant are

(7.43)

Then, the averaged stiffness matrices Aijy Bijy and D/; for the unit
cell can be obtained by the inversion of A\j, 5,y, and D[j. Finally,
effective laminate engineering constants Exx, Eyyy vxy and Gxy can
be expressed as functions of the stiffness constants Atj and the unit
cell thickness.
Figures 7.30 and 7.31 show the comparisons of theoretical
calculations for the axial Young's modulus and Poisson's ratio with
experimental data for three-dimensional braided carbon/epoxy
composites (Yang, Ma and Chou 1986). The basic material
 Analysis of thermoelastic behavior of composites 413

properties are E f = 234A GPa and vf = 0.22 for Celion 12K carbon
fiber, and £m = 3.4 GPa and vm = 0.34 for epoxy matrix. The
average yarn angle in the braided preform is denoted by a.
As the braiding angle becomes smaller, the performance of the
inclined laminae approaches that of the unidirectional laminae. The
interchange of the stacking sequence of the four inclined laminae in
the unit cell does not affect the effective in-plane properties. The

Fig. 7.30. Predicted axial elastic moduli of three-dimensional braided

composites as functions of fiber volume fraction, V{, and fiber orientation
angle, a. • , A, • and X: experimental data. (After Yang, Ma and Chou,
1986.)
150 r- = 15°

O
100

25°

I 50

J_ I I I
10 20 30 40 50 60 70 80
Fiber volume fraction, Vf (%)

Fig. 7.31. Predicted Poisson's ratios of three-dimensional braided compos-

ites as functions of fiber volume fraction, Vf, and fiber orientation angle, a.
# , • , and X: experimental data. (After Yang, Ma and Chou, 1986.)
1.5 r

1.0

0.5 I I I I I I J
0 10 20 30 40 50 60 70 80
Fiber volume fraction, Kf (%)
414 Three-dimensional textile structural composites

Poisson's ratio of three-dimensional composites based upon 4-step

braiding is considerably higher than that of a unidirectional com-
posite with the same fiber volume fraction. The Poisson's contrac-
tion can be minimized by introducing laid-in yarns in the axial
direction.
7.5.3 Macro-cell approach
The approach of the macro-cell model is different from that
of the unit cell. Instead of considering the smallest repeating unit in
a preform, the macro-cell is established for the entire cross-section
of the specimen. It takes into account the arrangements of the yarns
around the edges of the specimen. However, the most distinct
advantage of this approach perhaps is its capability of dealing with
specimens of 'thick' cross-sections, since the elastic properties are
derived from tensor transformations. In order to apply such a
model, it is necessary to have detailed knowledge of the fiber
geometric configurations.
In the following, the macro-cell model (Byun, Du and Chou
1991) is applied to the analysis of elastic properties of 2-step braided
fabric composites. The treatment is excerpted from Byun, Whitney,
Du and Chou (1991).
7.5.3.1 Geometric relations
Consider the 2-step braided composite depicted in Fig.
7.17. The variation in the braider yarn orientation along its length is
taken into account by introducing the average yarn orientation
angle. The average is identified by considering one braider yarn
which travels through the length of the macro-cell. For an m-
column by n-layer braided preform, a braider yarn travels m + (n +
l)/2 pitch lengths before it repeats its spatial position. Thus, for the
preform of Fig. 7.17 the repeating length is ten pitch lengths. Figure
7.32 shows the schematic view of the braider yarn location and

Fig. 7.32. Schematic view of a braider yarn extending through ten pitch
lengths. The numbers indicate the braider yarn carrier locations in Fig.
7.3. (After Byun et al. 1991.)
 Analysis of thermoelastic behavior of composites 415

orientation; the numbers indicate the positions of braider yarn

carriers in Fig. 7.3.
In order to identify the reinforcing direction of braider yarns, the
yarn segments generated due to the carrier movement from position
1 to 7' in Fig. 7.32 are projected onto the y-z (Fig. 7.33) and z-x
(Fig. 7.34) planes. Thus, all the yarn segments are identified
according to their directions with respect to the x—y—z coordinate.
Lhh Lby and Lhz denote the total projected length of braider
yarns which are inclined to the x, y and z axes, parallel to the x-y
plane, and parallel to the z-x plane, respectively. Then, from Fig.
7.17 and the parameters defined in Eqs. (7.12)-(7.14),

sbfb (7.44)
sin26)
Lby = 2[(n - l)Sm + 2(m - l)5 a cos 0] (7.45)
Lb2 = 2[2mSn + {n- l)5 a sin 6] (7.46)
where m and n denote the column and layer numbers, respectively.

Fig. 7.33 Projections of the yarn segment 11 '77' of Fig. 7.32 onto the y-z
plane. (After Byun et al. 1991.)

Fig. 7.34. Projections of the yarn segment 11'77' of Fig. 7.32 onto the x—z
plane. (After Byun et al 1991.)
416 Three-dimensional textile structural composites

The average angle between the braider yarn and the braid axis is
given by

whereL t ( = Lbi + Lby + L bz ) is the projected length of the braider

yarn onto the y-z plane. Based upon the average braider yarn
orientation angle, the total length of braider yarn (L b ) is approxi-
mated as

Lh = -^z (7.48)
sin a
The lengths of braider yarns inclined to the xyz axes and parallel to
the x-y and z-x planes can be obtained in a similar manner. Thus
the volumes of braider yarns of these three orientations are given by
Vhi = S2bfh(Ljsin a)
2
Vhy = S bfb(Lby/sina) (7.49)
2
Vbz = S bfh(Lbz/sin a)
The axial yarns have three different cross-sections as shown in Fig.
7.17. The total volume of the axial yarns is
Va = h[(m - l)(n - l)Sl sin(20) + 4(m - l)SaSn cos 6
+ 2(n - l)SaSm sin 6 + 4SmSn] (7.50)
Using Eqs. (7.15) and (7.16), the total macro-cell volume is
Vt = wt (7.51)
The fiber volume fractions of braider yarns of different orientations
can therefore be obtained from Eqs. (7.49)-(7.51); they are used
for evaluating the volume average of the stiffness constants.

7.5.3.2 Elastic constants

For the purpose of predicting the composite elastic pro-
perties, the yarns are treated as unidirectional composite rods. The
direction cosines between the reference coordinate system, xyzy and
the 123 coordinate system associated with the unidirectional com-
posite can be established by setting the 2 axis perpendicular to the z
axis (Fig. 7.35):
llx = cos /3 cos y l2x — —sin (3 l3x = —cos /3 sin y
lly = sin /3 cos y l2y = cos/? l3y = —sin jS sin y (7.52)
llz = sin y l2z — 0 l3z = cos y
 Analysis of thermoelastic behavior of composites All

Considering the average angle of braider yarn a instead of a in Fig.

7.35, the angles ft and y can be expressed in terms of a and the
aspect ratio of axial yarns, / a , as
V(1+/2)J (7.53)
-tf sin0 "I _t[ / a sin a ]
7
" t a n LV(cot2a + cos 2 0)J" tan l_V(l+/2cos2*)J
(7.54)
Using these direction cosines, the compliance matrix (S) of the
unidirectional composite (OO' in Fig. 7.35) referring to the 123
coordinate system can be transformed to that referring to the xyz
coordinate system:
S'ijmn = ipil s0", / , W, Hy p, q,V,S = 1, 2 , 3)
(7.55)
From symmetry conditions and using contracted notation, Eq.
(7.55) is reduced to a simple form:
'n = qmiqnjSmn (h j , m,n = 1-6) (7.56)
where qti denotes the element belonging to the /th row and yth
column of the transformation matrix (see Lekhnitskii 1963). For a
unidirectional composite with transverse isotropy, the compliance
matrix has five independent constants.
In order to determine the effective stiffness matrix of the
composite, the compliance matrix is inverted and then averaged

Fig. 7.35. Orientation of the braider yarn (OO'). (After Byun et al. 1991.)
418 Three-dimensional textile structural composites

over the macro-cell volume. The average should include all four
yarn orientations in Fig. 7.33, namely the axial yarn (13 = 0, y = 0),
braider yarn (BA) parallel to the x~y plane (/? = dr, y = 0), braider
yarn (BC) parallel to the z-x plane (/3 = 0, y = a) and inclined
braider yarn (CD) in the interior of the macro-cell (p =
f(&>fa)> Y = g(&>fa))- Thus, the effective stiffness of the composite
Q is
q=2(Q)»^ (7.57)
where (C,7)n and Vn/Vt are, respectively, the stiffness matrix and
volume fraction of the unidirectional composite for an individual
reinforcing direction. Finally, the stiffness matrix of the composite is
inverted to obtain the compliance matrix S». The engineering elastic
constants are then obtained from the compliance matrix. For
example, Exx = 1/Sn, Eyy = I/S2 2 anc* vxy = —S\ 2IS\2, etc.
Experimental measurements of the elastic properties of 2-step
braided composites have been reported by Byun et al. (1991). The
composites are the same as for Braid II given in Table 7.1.
Experimental observations of specimen cross-section confirm the
yarn shapes assumed in the analysis. Based upon the input data of
Table 7.1, the macro-cell model predicts the following composite
geometric parameters: thickness (t) = 7.1 mm, width (w) = 15.3 mm,
average braider angle (a) = 11.1°and fiber volume fraction (Vf) =
73.2%. Table 7.2 shows the comparison of elastic properties based
upon the macro-cell predictions and experiments.

Table 7.2. Comparisons of composite elastic properties from the

macro-cell model predictions and experiments. After Byun et al.
(1991).

Elastic constants Macro-cell model Experiment

Exx (GPa) 48.4 52.4(5*)

Eyy (GPa) 7.83
Ezz (GPa) 7.95
Gxy (Gpa) 2.58 1.45 (4*)
Gyz (GPa) 2.68
Gxz (GPa) 2.59
Vxy 0.33 0.53 (3*)
0.35
vxz 0.36
:
Number of tests.
 Structure-performance maps of composites 419

The analytical predictions deviate significantly from experimental

results for the in-plane shear modulus and Poisson's ratio. Some
reasons of uncertainty in the measurements of fabric composite
elastic properties are discussed in Section 7.7.

7.6 Structure-performance maps of composites

Considerable effort has been devoted by researchers to
evaluate the effectiveness of various reinforcement concepts.
However, the analyses and experiments performed on advanced
composites are usually reported for individual systems; it is thus
difficult to acquire a more comprehensive view. Chou and Yang
(1986) and Chou (1989), motivated by the concept of deformation
mechanism maps of Ashby, Gandi and Taplin (1979), Gandi and
Ashby (1979) and Frost and Ashby (1982) as well as the work of
Dow (1984) have integrated the results of studies in the modeling of
thermoelastic behavior of unidirectional laminated composites, as
well as two-dimensional (2-D) and three-dimensional (3-D) textile
structural composites. Through the construction of structure-
performance maps, the relative effectiveness and uniqueness of
various reinforcement concepts can be assessed. These maps
provide guidance in material selection for structural design, and in
identifying the needs of future work.
In order to assess the capability of various reinforcement con-
figurations with different fibers and matrix combinations, Chou and
Yang conducted parametric studies of the structure-performance
relationship. The geometric parameters considered include fiber
orientation in unidirectional laminated constructions, weaving para-
meters in two-dimensional fabrics, and braiding parameters in
three-dimensional constructions. The material parameters are fiber
and matrix thermoelastic properties. Four types of reinforcement
forms are presented below: laminated angle-plies based upon
unidirectional layers with the off-axis angle (6) ranging from 0° to
90°; [0°/90°] cross-plies; two-dimensional woven fabrics with gn
ranging from 2 (plain weave) to 8 (eight-harness satin); and 2-step
and 4-step braided composites. For 2-step braids, yarn linear
density, pitch length and aspect ratio are allowed to change. The
braiding angle between a fiber segment and braiding axis in the case
of 4-step braids varies from 15° to 35°. The analytical tools
employed in the construction of these maps include the lamination
theory for cross-ply and angle-ply laminates, the crimp and
bridging models (Chapter 6) for two-dimensional fabrics, the fiber
420 Three-dimensional textile structural composites

inclination model for 4-step braided composites, and the macro-cell

model for 2-step braided composites.
Several maps are presented here to illustrate the correlation
between reinforcement configurations and the thermoelastic be-
havior of composites. The fiber volume fractions of the composites
are assumed to be 73% for 2-step braided composites and 60% for
all other composites. Figures 7.36-7.39 present the thermoelastic
behavior of carbon, Kevlar and glass reinforced epoxy composites.
Figure 7.40 shows the variation of thermal expansion coefficients for
PEEK matrix composites. Figure 7.41 gives the elastic properties of
C, SiC and A12O3 fiber reinforced Mg matrix composites. Figure
7.42 demonstrates the elastic properties of glass matrix composites
reinforced with C, SiC and A12O3 fibers. The three-dimensional

Fig. 7.36. Exx vs. Eyy for carbon/epoxy ( • unidirectional angle-ply; A

two-dimensional woven; + three-dimensional braided), Kevlar/epoxy (A
unidirectional angle-ply; • two-dimensional woven; x three-dimensional
braided), and glass/epoxy ( • unidirectional angle-ply; O two-dimensional
woven; • three-dimensional braided) composites, (p) = plain weave;
(s) = eight-harness satin. (After Chou and Yang 1986.)

E vv (GPa)

10 25 50 100 150
30
25 Unidirectional angle-ply
— "Two-dimensional woven 150
20 Three-dimensional braided
- o° 100
15
[0°/90°]

10 +• 0 = 35

a. 50 OH
o o
) = 15 C
I*)'
I = 35°
8 = 35°
25

(P)

,±45° 90° 10
90° 90°

I
10 15 20 25 30
£v.v ( 1 0 6 PS1)
 Structure-performance maps of composites All

preforms discussed in Figs. 7.36-7.42 are based upon 4-step

braiding.
Figures 7.43 and 7.44 show the variations of elastic properties of
2-step braided composites of Kevlar/epoxy with fabric geometric
and processing parameters. The linear density ratio of axial and
braider yarns, the pitch length of braider yarns, and the aspect
ratios of axial and braider yarns are considered. The structure-
performance maps are constructed by starting with a set of values of
these parameters, and then varying each parameter independently
while keeping the other parameters at their original values. The
ranges of these values are denoted on the curves in Figs. 7.43 and

Fig. 7.37. Exx vs Gxy for carbon/epoxy ( • unidirectional angle-ply; A

two-dimensional woven; + three-dimensional braided), Kevlar/epoxy (A
unidirectional angle-ply; • two-dimensional woven; x three-dimensional
braided), and glass/epoxy ( • unidirectional angle-ply; O two -dimensional
woven; • three-dimensional braided) composites, (p) = plain weave;
(s) = eight-harness satin. (After Chou and Yang 1986.)

0.
O O
422 Three-dimensional textile structural composites

7.44 where the calculations are performed at equal intervals as

indicated by the symbols on each curve (Byun et al. 1991).
It is noted from Figs. 7.43 and 7.44 that the Young's moduli and
shear moduli are insensitive to the axial yarn aspect ratios between 1
and 3. The maximum volume fraction of axial yarns is achieved
when the yarn aspect ratio is around unity, which also gives the
maximum Exx and Gyz. Increases in the linear density ratio of axial
yarn to braider yarn result in an increase in axial yarn volume
fractions while the braider yarn volume fraction becomes smaller.
Since the stiffness increases of the 2-step braided composite in the
longitudinal and transverse directions are primarily due to the
contribution of the axial yarns and braider yarns, respectively,
the increase of the axial yarn volume fraction improves Exx. In the
meantime, Eyy and Gyz become smaller due to the reduction in
braider yarn volume fraction. Furthermore, the increase in the
aspect ratio of the braider yarns results in an increase in their
thickness, which in turn gives a larger volume of matrix pockets in the
composite. Since the total fiber volume fraction is reduced due to

Fig. 7.38. Exx vs. vxy for carbon/epoxy ( • unidirectional angle-ply, A

two-dimensional woven, + three-dimensional braided) composites. (After
Chou and Yang 1986.)
25 1
150
20

15 100

10

50

20
±45°

10

_J
0 0.1 0.5 5.0 1.5
 Structure-performance maps of composites 423

the increase in volume of the composite, all the components of

Young's moduli and shear moduli in Figs. 7.43 and 7.44 are reduced
as the braider yarn aspect ratio becomes bigger. Finally, longer
braider yarn pitch length gives a small volume fraction and orienta-
tion angle of braider yarns. Consequently, Eyy and Gyz are reduced
as the braider yarn pitch length increases.
Chou and Yang (1986) have compared the unique thermal and
elastic characteristics among various reinforcement configurations.
In general, the in-plane thermoelastic properties of unidirectional
lamina depend strongly on the fiber orientation. The unidirectional
reinforcement provides the highest elastic stiffness along the fiber
direction. The Young's moduli of off-axis unidirectional laminae are
lower than that of the unidirectional lamina.
The [0°/90°] cross-ply yields identical thermoelastic properties in 0°
and 90° orientations. Their in-plane shear rigidity is poor. The
longitudinal Young's modulus of an angle-ply laminate is lower than

Fig. 7.39. ccxx vs. ayy for carbon/epoxy ( • unidirectional angle-ply; A

two-dimensional woven; + three-dimensional braided), Kevlar/epoxy (A
unidirectional angle-ply; • two-dimensional woven; x three-dimensional
braided), and glass/epoxy ( # unidirectional angle-ply; O two-dimensional
woven; • three-dimensional braided) composites, (p) = plain weave; (s)
eight-harness satin. (After Chou and Yang 1986.)
a vv (10- 6 /°C)

-1 0 10 20 40 60
1
60
30 — Unidirectional angle-ply
— Two-dimensional woven
— Three-dimensional braided

40
20

20
10

10

Eight-harness 0
satin
-5 J L 1 -10
-5 0 10 20 30
a vv (10- 6 /°F)
424 Three-dimensional textile structural composites

that of a unidirectional lamina. But better transverse elastic

property and in-plane shear resistance can be achieved through the
stacking of the unidirectional laminae with different fiber orienta-
tions. For ±45° angle-ply, the in-plane stiffness drops to a mini-
mum, while the shear modulus reaches its maximum.
The two-dimensional biaxial woven fabric composites can provide
balanced in-plane thermoelastic properties within a single ply. They
behave similar to [0°/90°] cross-plies, although the fiber waviness
tends to reduce the in-plane efficiency of the reinforcements. As the
fabric construction changes from plain weave to eight-harness satin,
the frequency of crimp due to fiber cross-over is reduced, and the
fabric structure approaches that of [0°/90°] cross-plies.
The thermoelastic properties of braided composites also show a
strong dependence on fiber orientation. Three-dimensionally
braided composites have demonstrated good in-plane properties,
which are comparable to those of unidirectional angle-plies with the

Fig. 7.40. axx vs. ayy for carbon/PEEK ( • unidirectional angle-ply; A

two-dimensional woven; + three-dimensional braided), Kevlar/PEEK ( •
unidirectional angle-ply; • two-dimensional woven; x three-dimensional
braided), and glass/PEEK ( • unidirectional angle-ply; O two-dimensional
woven; • three-dimensional braided) composites. (After Chou and Yang
1986.)

a vv (10- 6 /°C)
-10 10 20 40 60

- 60
30 — — Unidirectional angle-p!y
— •— Two-dimensional woven
— — Three-dimensional braided

40
20

- 20
10

-5
-5 0
 Structure-performance maps of composites 425

same range of fiber orientation. The longitudinal Young's moduli

and in-plane shear rigidities of three-dimensional braided
composites with braiding angles ranging from 15° to 35° arebetter
than those of two-dimensional woven fabric composites. But
the transverse Young's moduli are lower and the major Poisson's
ratios higher than those of two-dimensional woven fabric compos-
ites. However, three-dimensional braided composites are unique in
providing both stiffness and shear rigidity along the thickness
direction. Also, because of the integrated nature of the fiber
arrangement, there are no interlaminar surfaces in three-
dimensional composites.
Furthermore, a comparison of their elastic behavior indicates that
the performance of 2-step braided composites is much more
versatile than that of 4-step braided composites. The presence

Fig. 7.41. Exx vs. Eyy carbon/magnesium ( • unidirectional angle-piy; A

two-dimensional woven; • three-dimensional braided), SiC/magnesium
(A unidirectional angle-ply; O two-dimensional woven; x three-
dimensional braided), and Al 2O3/magnesium ( • unidirectional angle-ply;
• two-dimensional woven; + three-dimensional braided) composites.
(After Chou and Yang 1986.)
Unidirectional angle-ply
— — Two-dimensional woven
— — Three-dimensional braided

Ey, (GPa)
50 100 150 200 250
40 1 1 i 1

- 250
15°
-
30 - I - 200

0°
o ^ Eightharness satin -
~ 150

' V^
~ 20 W0 = 15 Plain weave

A
- 100
nfl - 15 A
[0°/90°]
10 - \
A

A' Eight-harness satin V

Plain weave 90° - 50
—••
i 1
10 20 30 40
£vv(106psi)
426 Three-dimensional textile structural composites

of both axial and braider yarns in 2-step braids allows much

flexibility in the design of the preform microstructure. However,
such flexibility can be achieved in 4-step braided composites if
laid-in axial yarns are used.
The structure-performance maps can form the basis for material
selection and component design; these findings can be easily
extended to generate a wider range of information. Take the woven
fabric composite as an example; although the properties shown in
the maps are primarily along the filling and warp directions, the
off-axis properties can be readily obtained through proper tensor
transformation. Upon knowing these properties, it would be feas-

Fig. 7.42. Exx vs. Gxy for carbon/borosilicate glass ( • unidirectional

angle-ply; A two-dimensional woven; • three-dimensional braided),
SiC/borosilicate glass ( • unidirectional angle-ply; O two-dimensional
woven; + three-dimensional braided), and Al 2O3/borosilicate glass (A
unidirectional angle-ply; • two-dimensional woven; x three-dimensional
braided) composites and three-dimensional carbon/carbon composites
( • ) . (After Chou and Yang 1986.)

Unidirectional angle-ply
Two-dimensional woven
Three-dimensional braided

G,v (GPa)
0 10 20 50 100
1 1 1 1

40 _
M 0°
- 250

\ S s - e = 35
30 — - 200
Eight-harness sa
a. 0°
o O
Plain weave \ -— 150
20 - 9 = IS"»\ A ±45°
X 9
• \
e = 35°V\ -- 100
W
Eight-
L W \ \

harness-""/ \\
10 _ satin A
±45°
Plain weave p - 50

90° • —
n 1 1
5 10 15
6
Gxx (10 psi)
Structure—performance maps of composites 427

Fig. 7.43. The variations of Exx and Eyy with material and processing
parameters (intervals of the parameters, axial yarn aspect ratio: 1, pitch
length:2, linear density ratio: 0.05, braider yarn aspect ratio: 0.02). (After
B / . 1991.)

(0.2) (0 02) yA
A (5) Axial yarn i
aspect ratio i
8.0 - J(4) /
\
0 - 1 I Braider yarn
Linear 1 / aspect ratio
<X
O density ratio
""* (0.05)
~ 7.8-
i

7.6 - / Pitch length

/ 1
/

•
k(10)
(0.1)
7.4 . 1 1
1 , 1 . 1 1 1 1
45 46 47 48 49 50 51 52
£ „ (GPa)

Fig. 7.44. The variations of Gxy and Gyz with material and processing
parameters (intervals of the parameters, axial yarn aspect ratio: 1, pitch
length: 2, linear density ratio: 0.05, braider yarn aspect ratio: 0.02). (After
Byun etal. 1991.)
2.9
(0.2)
Linear density
ratio
\
2.8

. (0.02)
ft*
O Pitch length
(4) Braider yarn
aspect ratio

(10)

Axial yarn aspect ratio

I ,
2.5 2.6 2.7
G,v (GPa)
428 Three-dimensional textile structural composites

ible to tailor composite structures with various combinations of

reinforcement forms, or with different material combinations such
as hybrid unidirectional laminate, hybrid woven fabric structures, or
hybrid laid-in three-dimensional structures.
It should be noted that the analytical modeling techniques
employed in the construction of the structure-performance maps
assume 'defect-free' composites, i.e. perfect fiber/matrix interfacial
bonding, perfect fiber alignment, void-free matrix materials, etc. It
is understood that in actual composites defects are frequently
introduced in the fabrication and handling process. Limited studies
in this regard have been made, including the effect of fabrication
induced fiber distortion on the thermoelastic properties of two-
dimensional fabric composites (Yang and Chou 1989), the effects of
fiber/matrix interfacial debonding on the effective elastic properties
(Takahashi and Chou 1986), void content of as-fabricated polymeric
matrix composites (Yoshida, Ogasa and Hayashi 1986), cracking of
polymeric matrices in fabric composites (Ishikawa and Chou 1982),
and the effect of fiber bundle size and distribution on the behavior
of three-dimensional braided Al 2O3/Al-Li composites (Majidi,
Yang and Chou 1986). It is expected that with the advancement in
mathematical modeling and experimental techniques, performance
maps for strength and failure of various two- and three-dimensional
composites can also be constructed.

7.7 Mechanical properties of composites

This section summarizes the strength, fracture, and damage
tolerance behavior of three-dimensional fabric composites. The
material systems cited here include both polymer and metal based
composites.

7.7.1 Tensile and compressive behavior

Majidi, Yang, Pipes and Chou (1985) examined the
tensile and compressive behavior of 4-step braided composites of
alumina fiber in an aluminum-lithium matrix. The continuous,
polycrystalline a-alumina yarn (Fiber FP manufactured by the
Du Pont Co.) contains 210 filaments of approximately 20 jum dia-
meter. The properties of Fiber FP are: tensile strength = 1380 MPa,
tensile modulus = 345-79 GPa, elongation to failure = 0.4%, density
= 3.90g/cm3, and melting point = 2045°C (Dhingra, Champion
and Krueger 1975). The aluminum matrix is alloyed with 2-3 wt%
lithium for an enhanced chemical bond between the fiber and the
matrix.
 Mechanical properties of composites 429

Figure 7.45 depicts the tensile stress-strain curves of FP/Al-Li

composites of unidirectional laminates and three-dimensional
braided composites at V f =17% and 36%; the tensile behavior of
the pure Al-Li matrix is also given. A bilinear behavior is observed
for all composites; the 'knees' on the stress-strain curves occur at
about 0.02%. Yielding of the matrix appears to be responsible for
the bilinearity. Since the bilinear behavior has also been observed in
the unidirectional composites, it is believed to be a material
property rather than an effect caused by the braided structure. The
in situ strength of the matrix may well be higher than that measured
for the bulk material. Such a phenomenon has been discussed by
Kelly and Macmillan (1966). When the fiber spacing is very small
(<10,um), as is the case in the material studied by Majidi and
colleagues, the yield stress of the matrix is controlled by the
Orowan stress and it is higher than that of the bulk matrix. The
yield stress and work hardening increase with decreasing spacing
between the fibers. While the yield stress goes up, the strain at
which the matrix starts yielding in the composite drops for very
small fiber spacing (Kies 1962).
The ultimate tensile strengths of three-dimensional braided com-
posites with a braiding angle of about 20° are 189 MPa and
383 MPa for V f =\l% and 36%, respectively. These values are
slightly lower than those predicted for ±20% angle-ply laminates of

Fig. 7.45. Axial tensile stress-strain responses of (a) unidirectional

FP/Al-Li composite (Vf = 0.50), (b) three-dimensional FP/Al-Li compos-
ite (Vf = 0.36), and (c) the unreinforced matrix. (After Majidi and Chou
1987.)

S
430 Three-dimensional textile structural composites

the same fiber volume fraction. Microscopic examination of the

fracture surfaces reveals brittle fracture of fibers and considerable
deformation in the matrix between the fibers.
The measured initial Young's moduli are 97 GPa and 171 GPa for
Vf = 17% and 36%, respectively, which agree well with theoretical
predictions based upon the fiber inclination model (Yang, Ma and
Chou 1986). The secondary Young's moduli can be approximated
by assuming that the contribution of the matrix to the composite
modulus is negligible after yielding. The measured Poisson's ratios
are 0.30 (Vf= 17%) and 0.27 (Vf = 36%). It should be noted that a
certain degree of damage to brittle fibers often occurs in the
braiding process. Thus, the in situ fiber stiffness and strength
properties need to be estimated from, for instance, those measured
on unidirectional composites.
Figure 7.46 shows the compressive stress-strain behavior of the
same material as in Fig. 7.45(b). The curve demonstrates an initial
linear region up to a strain of about 0.15% followed by nonlinear
behavior. Other compressive properties include a failure strain of
about 1.8% and a major Poisson's ratio of 0.3. Kinking appears to
be the primary mode of failure in compression.
In transverse tension of three-dimensional braided FP/Al-Li
composites the stress-strain curve is highly nonlinear. The onset of
nonlinearity is at a strain of about 0.5%. The ultimate strength is

Fig. 7.46. Compressive stress-strain responses of FP/Al-Li composite.

(After Majidi et al. 1985.)
140

120

100

60 £

40

20

0
0.4 0.8 1.2 1.6 2.0 2.4
Strain (%)
 Mechanical properties of composites 431

considerably lower than that for the axial specimen and, thus,
significantly smaller than that of the unreinforced matrix material.
Transverse cracks initiate within the matrix-rich regions
between fiber bundles, most likely at microscopic voids in the
matrix.

7.7.2 Shear behavior

Majidi et al. (1985) have examined the shear behavior of
FP/Al-Li composites with a 4-step braided preform. Both intralami-
nar (in-plane) and interlaminar shear measurements are made. It
has been shown that the short beam shear test (ASTM Standards,
D2344-84 1987) causes premature failure by a flexural mode on the
tensile surface of the specimen even for a fairly small specimen
span-to-depth ratio. It is due to the high ratio of interlaminar shear
strength to tensile strength in three-dimensional braided compos-
ites. The two-rail shear test (ASTM Standards, D4255-83 1987) also
proved inadequate. The tests suitable for the measurement of the
shear strength of three-dimensional composites are the Iosipescu
shear test originally proposed by Iosipescu (1967) for isotropic
materials and applied to composite laminates by Walrath and
Adams (1983a&b), and the double-notch shear test.
The in-plane shear strength parallel to the braiding direction
measured with the Iosipescu shear test for Vf= 17% is 139.6 MPa,
which is comparable to the theoretical shear strength of 151 MPa for
[±20°] angle-ply laminates of the same material and fiber volume
fraction. Majidi, Remond and Chou (1987) reported the shear
properties of three-dimensional braided FP/Al-Li composite tubes,
with the tube axis parallel to the braiding axis and braiding
angles of approximately ±20°. The in-plane shear strengths meas-
ured from torsion tests are 141.8 MPa for V f =17% and
102.1 MPa for K f =36%. The shear moduli are 36.6 GPa and
39.0 GPa for V{=17% and 36%, respectively.
The interlaminar shear strength measured by the double-notched
shear test is 144.5 MPa. The calculated interlaminar shear strength
for 0° unidirectional laminates is 100.8 MPa. This improved inter-
laminar shear property in the thickness direction of three-
dimensional braided composites gives much improved fracture and
impact resistance over the conventional laminates. Some dfficulties
in the testing of three-dimensional textile structural composites
exist. Machining of specimens should be avoided as much as
possible because it destroys the integrated nature of the fiber
preform and thus results in lower strength. It is also difficult to
432 Three-dimensional textile structural composites

obtain meaningful readings of the shear strain and, hence, shear

modulus from the Iosipescu specimens. This is due to the relatively
small size of the region of pure shear, the highly non-homogeneous
three-dimensional braided structure based upon large bundles and
the small size of the strain gages used. All these factors indicate
the difficulties in the testing of textile structural composites in
general.

7.7.3 Fracture behavior

7.7.3.1 In-plane fracture

The fracture and toughness characteristics of unidirectional
and three-dimensional braided FP/Al-Li composites have been
examined by Majidi, Yang and Chou (1986, 1988). Metal matrix
composites, particularly those incorporating ceramic fibers, offer
very high strength and stiffness, but often significantly lower
fracture toughness than unreinforced metallic matrices. The
reduced toughness is due to the restriction of plastic deformation in
the presence of the stiff fibers and a strong fiber/matrix bond which
eliminates or restricts fiber debonding and pullout.
Majidi and colleagues have measured fracture toughness using
compact tension tests on the basis of the linear elastic fracture
mechanics approach and the notched three-point bend test (Tat-
tersal and Tappin 1966), which involves measurement of the
work of fracture (fracture surface energy averaged over the whole
fracture process).
Figures 7.47(a) and (b) show the load versus crack opening
displacement (COD) curves for repeated loading and unloading of a
three-dimensional braided composite and a unidirectional laminate
under compact tension tests. The stress intensity factors are
calculated from PQ and the maximum load Pmax indicated in Fig.
7.47(a). PQ is determined by drawing a straight line with a slope 5%
less than the slope of the linear part of the load-COD curve and
finding the corresponding load at the intersection of this line with
the curve. For Figs. 7.47(a) and (b), the crack propagation is
perpendicular to the braiding axis of the three-dimensional braided
composite and the fiber direction of the unidirectional laminate,
respectively. In both cases, the first part of the load-COD curve is
highly non-linear and, therefore, PQ is considerably lower than
Pm3LX. F°r subsequent loading cycles, however, the sharpened crack
removes the non-linearity in the curve and Po approaches Fmax.
Then, the stress intensity factors KQ and Kmax can be calculated
 Mechanical properties of composites 433

from PQ and Pmax, respectively (Annual Book of ASTM Standards,

E399 1978). Since the Kmax values are reasonably constant within
the range of crack length to specimen width ratio, Majidi, Yang and
Chou (1986) adopted the average Kmax as the critical stress intensity
factor, Kc. Remond (1987) has characterized the fracture behavior
of three-dimensional braided metal matrix composites with Vf =
36%. For the notch perpendicular to the principal reinforcement
axis, the unidirectional laminate (Xmax = 30.7 MPaVm) appeared
tougher than the three-dimensional braided composite (Kmax =
27.3 MPaVm)- For the notch parallel to the principal reinforce-
ment axis, the three-dimensional braided composite (KmsiX =
21.5 MPaVm) is tougher than the unidirectional laminate (Kmax =
19.0 MPaVm)- The [±20°] angle-ply laminate appears less tough than
the two other composites, the longitudinal toughness being Kmax =
24.6 MPaVm and the transverse toughness Kmax = 16.4 MPaVm.
Electron microscopy investigations of the fracture surfaces indi-
cate virtually no pull-out of the individual fibers. However, oc-
casionally the whole fiber bundle has been pulled out over a small
length of 1-2 mm. This is accompanied by some debonding between
the fiber bundle and the surrounding matrix. The mechanism of
crack propagation perpendicular to the braiding axis is believed to
be the fracture of fibers and eventual fracture of the fiber bundle
ahead of the crack tip. The above behavior differs greatly from that
of the unidirectional FP/Al-Li composite, which shows a rapid

Fig. 7.47. Load-crack opening displacement (COD) curves obtained from

compact tension tests on (a) three-dimensional braided FP/Al-Li compos-
ite (Vf=17%), and (b) unidirectional FP/Al-Li composite (Vf = 34%).
(After Majidi, Yang and Chou 1986.)

\
1 !f/; \
1
1
h fi
i 1: i i
i

0
0.0 0.4 0.8 1.2 1.6 2.0 2.4 0.0 0.4 0.8 1.2 1.6 2.0
COD (mm) C O D (mm)
(a) (a)
434 Three-dimensional textile structural composites

crack propagation from the outset and the fracture surface has a
flat, brittle appearance with no macroscopic dimples.
In the case of notched three-point bend tests, unidirectional
composites fracture in a much more brittle and less controlled
manner than braided composites, and the load-deflection curve
shows the sharp drop of load. The work of fracture, yf, which is
measured from the total energy absorbed for the complete fracture
of the specimen, or the area under the load-deflection curve, are
7.92 ± 1.27 kJ/m2 for the braided composites and 4.56 ± 0.44 kJ/m2
for unidirectional laminate for Vf= 17%.
The difference in the strength and fracture behavior between
textile structural composites and traditional laminated composites
has been further demonstrated for the FP/Al-Li composites by
Majidi, Yang and Chou (1986). In the case of unidirectional
composites, the contributions from fiber debonding and pull-out are
negligible since the critical load transfer length, /c, is only 0.34 mm.
This is calculated from the equation lc = afd/(2r) where af is the
fiber ultimate strength (1380 MPa), d is the fiber diameter (20jUm),
and R is assumed to be equal to the shear yield strength of the
matrix (—40 MPa). The strong fiber/matrix interface also reduces
the length on either side of the broken fiber over which the matrix
deforms plastically (see Cooper and Kelly 1967). For unidirec-
tional composites with V{ = 34% and matrix tensile strength of
160 MPa, this length is only 0.04 mm. Therefore, plastic deforma-
tion is severely restricted in the unidirectional system, and the lack
of fiber pull-out and the limited plastic deformation in the matrix
are responsible for the planar fracture and the low yf.
In three-dimensional braided composites, fibers are not
uniformly distributed in the matrix as in the unidirectional laminae,
and each fiber bundle can be regarded as an individual reinforce-
ment. The volume fraction of fibers within the bundle is approxi-
mately 50%, and the volume fraction of bundles in the composite is
approximately 40%. Using the diameter of 2 mm and tensile
strength of 586 MPa for the bundles, it is found that l c = 12.6 mm
and the length over which the matrix deforms plastically is 4 mm.
Although factors such as fiber inclination and interactions
among bundles are not considered above, these values illustrate the
beneficial effect of fiber clustering on the extent of matrix plastic
deformations and on the pull-out and debonding mechanisms. The
non-homogeneous microstructure, therefore, appears to be at least
partially responsible for the higher work of fracture of the three-
dimensional braided composites as compared with the unidirec-
tional laminate which shows a catastrophic planar fracture.
 Mechanical properties of composites 435

Majidi, Yang and Chou (1986) have also examined the effect of
thermal treatment on the fiber/matrix interface strength and, hence,
the fracture toughness of three-dimensional braided composites.
The fiber/matrix interface deteriorates after isothermal heating at
500°C. There is a decrease in the fracture load and an increase in
the amount of bundle pull-out which results in larger work of
fracture (yf = 20.30 ± 14.35 kJ/m2 after 72 hours of thermal treat-
ment). This reflects the weak nature of the interfacial reaction zone
which grows intergranularly towards the center of the fiber.
Guenon, Chou and Gillespie (1989) reported the in-plane frac-
ture toughness, Klc, of carbon/epoxy composites with a three-
dimensional orthogonal interlock fabric preform. The Klc values
for three-dimensional fabric composites are 28.56 MPaVm and
29.45 MPaVm in two principal material directions, which are higher
than that of laminates (21.22 MPaVm). The through-the-thickness
yarns are beneficial to the in-plane toughness by arresting and
deviating the crack. The interaction between a crack and in-
homogeneities, simulating fiber arrays, has been examined by
Fowser and Chou (1989, 1990a&b).

7.7.3.2 Interlaminar fracture

Traditional laminated composites exhibit low interlaminar
fracture toughness and are susceptible to delamination when
subjected to interlaminar stress concentrations. Improvements in
damage tolerance to date have focussed on utilizing tougher
matrices (Hunston 1984) or interleafing concepts (Masters 1987).
Through-the-thickness reinforcement provides an alternative ap-
proach to substantially increasing the resistance to delamination
(see Whitney, Browning and Hoogsteden 1982; Guess and Reedy
1985; Mignery, Tan and Sun 1985; Dexter and Funk 1986; Fowser
1986; Guenon, Chou and Gillespie 1987; Ogo 1987).
The orthogonal interlock fabric architecture (Fig. 7.10) retains the
in-plane performance while enhancing out-of-plane properties, by
including a small amount of through-the-thickness reinforcement.
The 'z direction' fibers are also known to be detrimental to the
in-plane tensile and compressive properties. The interlocking proc-
ess avoids the cutting of fibers, as it occurs in the stitching process.
However, it creates matrix pockets that reduce the volume fraction
of the in-plane fibers relative to the analogous two-dimensional
laminates. The z direction fibers also tend to be deformed in the
processing of the fabric composites.
Guenon, Chou and Gillespie (1989) have studied the effect of
fiber geometry on the interlaminar and in-plane fracture behavior of
436 Three-dimensional textile structural composites

orthogonal interlocked fabric composites. The experimental work is

based upon a T300/3501-6 carbon/epoxy system. Referring to Fig.
7.10, the fabric preform can be described as a [0°/90°] laminate in
which some through-the-thickness yarns are interlaced. The in-
plane yarns contain 6000 filaments per yarn and the through-the-
thickness yarns have 1000 filaments per yarn. The spacing between
two z direction yarns in both plate directions is 2.8 mm and the
plate contains about 13 z direction yarns/cm2. The total number of
plies is 27, with 14 and 13 plies in two mutually orthogonal
directions. The overall volume fraction is 50%, while the volume
fraction of the z direction is 1%.

(A) Mode I interlaminar fracture

Guenon, Chou and Gillespie (1989) have adopted two test
methods for Mode I interlaminar fracture, the double cantilever
beam (DCB) test and the 'tabbed DCB', which uses long aluminum
tabs bonded along both sides of the specimen to prevent the
deviation of crack propagation from a self-similar manner. Both
types of specimens are pin-loaded in tension in displacement-
controlled mode.
The load-deflection curves for the three-dimensional composite
specimens show a nonlinear unloading sequence and an appreci-
able permanent deformation after unloading. The crack tip did not
completely close after unloading. These features can be explained
by the crack closure process of the three-dimensional fabric
composite. Most of the z direction yarns do not break in the plane
of the crack. Instead, they fracture near the outer surface of the
specimen where they are curved by the weaving process, and then
debonded and pulled out. Figure 7.48 shows the fracture surface
with the z direction yarn protruding out of the plane of fracture.
During unloading, the pulled-out yarns do not resume their initial
locations and therefore progressively undergo compressive stresses
that lead to a nonlinear unloading behavior and a permanent
deflection of the specimen after a zero load is reached.
Two data reduction methods have been adopted by Guenon,
Chou and Gillespie (1989) for the three-dimensional fabric compos-
ites. These are the area method, based upon energy considerations,
and the compliance method, based upon the linear elastic beam
theory. The interlaminar critical strain energy release rate, Glc,
values from the area method are 0.307 kJ/m 2 (two-dimensional
regular DCB), 0.286 kJ/m2 (two-dimensional tabbed DCB) and
3.85 kJ/m2 (three-dimensional tabbed DCB). The compliance
 Mechanical properties of composites 437

method gives Glc values of 0.235 kJ/m2 (two-dimensional regular

DCB), 0.179 kJ/m2 (two-dimensional tabbed DCB) and 2.66 kJ/m2
(three-dimensional tabbed DCB). The compliance method only
takes into account the energy of crack initiation. In the case of
two-dimensional unidirectional laminates, the crack propagation
energy is generally equal to the initiation energy and therefore both
methods give similar results. In three-dimensional fabric composites,
the fracture, debonding and pull-out of z direction yarns as well as
the bridging of the crack by the z direction yarns dissipate energy.
Therefore, the area method gives a higher and more accurate result
of interlaminar fracture toughness.
The mode I delamination problem of three-dimensional or-
thogonal interlock fabric composites has been further examined by
Byun, Gillespie and Chou (1990b) using a finite element analysis.
The material systems and specimen geometries including two-
dimensional regular DCB, two-dimensional tabbed DCB and three-
dimensional tabbed DCB (see Guenon, Chou and Gillespie 1989)
are simulated in this numerical work. Specifically, the mode I
fracture behavior of carbon/epoxy composites is examined for

Fig. 7.48. Fracture surface of the orthogonal interlock fabric composite

showing a pulled-out yarn. (After Guenon, Chou and Gillespie 1989.)
438 Three-dimensional textile structural composites

various initial crack lengths. The strain energy release rates, Gr, are
evaluated based upon the crack closure method (Rybicki and
Kanninen 1977) to ascertain the influence of through-the-thickness
fibers on crack driving force. Byun and colleagues also have
considered the effect of progressive debonding of the z axis yarns on
the strain-energy release rate.
In the finite element model, the length of an element side is the
crack increment utilized in the study (Aa = 0.5562 mm). The initial
crack length is 25.4 mm and the locations of the through-the-
thickness fibers are 3Aa, 8A0, 13Aa and 18A«. A vertical unit
displacement of 1 mm is applied to simulate the displacement
controlled loading conditions used in the experimental work of
Guenon and colleagues. Three types of through-the-thickness fiber
debonding are modeled: perfect bonding, moderate bonding, where
the z axis fiber is debonded over 25% of the specimen thickness,
and complete debonding, where the load is carried by the fibers
only. The z axis yarns are assumed to be initially perfectly bonded;
partial or complete debonding does not occur until the crack front
passes the reinforcement. Also, fiber fracture is not considered by
Byun and colleagues.
Figure 7.49 demonstrates the effect of through-the-thickness
yarns and bonding conditions on the strain-energy release rate. The
strain-energy release rate for the two-dimensional laminate mono-
tonically decreases with increasing crack length as one would expect

Fig. 7.49. Numerical strain-energy release rates of two-dimensional lamin-

ated and three-dimensional orthogonal woven composites as functions of
crack length: — two-dimensional laminate; O fully debonded; + moder-
ately bonded; # perfectly bonded. (After Byun, Gillespie and Chou 1990.)

Aa 3Aa SAa 13Aa 18Aa

Crack length
 Mechanical properties of composites 439

under fixed grip conditions. The introduction of through-the-

thickness yarns reduces the local strain-energy release rate sig-
nificantly in the case of perfect bonding condition as the crack
approaches the first array of z axis yarns at 3Aa. The crack driving
force for interlaminar crack growth decreases because the load is
transferred to the z axis yarns. The crack opening displacement is
also reduced by the presence of the vertical fibers. As debonding
occurs, the decrease of strain-energy release rate is less significant
because the crack tip opening displacement increases.
In the numerical analysis, Byun and colleagues assume that the
crack continues to propagate without fiber fracture. Consider the
perfect bonding case of Fig. 7.49 when the crack propagates beyond
the first array of yarns to 4Aa. Tfie strain-energy release rate is
observed to increase slightly but is still significantly less than the
two-dimensional specimen. This is due to the increase in deforma-
tion of the z axis yarns as load transfer occurs, which results in an
increase in the crack tip opening displacement. Due to the presence
of the z axis yarns at 8Afl, the strain-energy release rate begins to
diminish as the crack tip approaches the next site of through-the-
thickness fibers where a second reduction in crack driving force is
observed. The process continues until the strain-energy release rate
is identically zero.
Figure 7.50 shows the tensile stress in the through-the-thickness
yarn as a function of crack length. As the crack approaches the first
yarn at 3Afl, the load in the fiber increases rapidly. Additional crack

Fig. 7.50. Tensile stresses in the z axis fiber arrays as functions of crack
length for perfect fiber-matrix bonding: — first array; O second array; •
third array (After Byun, Gillespie and Chou 1990.)

f§

haAha

Crack length
440 Three-dimensional textile structural composites

growth results in an asymptotic value of load in the first yarn as the

second yarn begins to carry load. The process continues until the
stress in the next array of yarns is zero. At this point, the applied
load is being carried exclusively by multiple z axis yarns bridging
the crack surface and the crack driving force is identically zero. The
information presented in Figs. 7.49 and 7.50 demonstrates the load
bearing and transferring mechanisms in the interlaminar fracture
process of a three-dimensional fabric composite. This information
enabled Byun and colleagues to predict the macroscopic critical
load for mode I interlaminar fracture. More importantly,
understanding of the load redistribution at the microscopic level is
beneficial to the design of fabric preform structure for enhanced
damage tolerance.

(B) Mode II interlaminar fracture

The mode II interlaminar fracture toughness of three-
dimensional orthogonal fabric composite has been assessed by Liu
and Chou (1989). The mode II fracture toughness tests are
performed on both three-dimensional composites and two-
dimensional laminates of the same carbon/epoxy system using
end-notch-flexural (ENF) specimens.
Byun, Gillespie and Chou (1989), following the work of Liu and
Chou (1989), have conducted a finite element analysis for evaluating
the mode II strain-energy release rate. Similar to the mode I
interlaminar fracture, the crack driving force for mode II interlami-
nar crack growth decreases as the crack approaches the z axis yarns
because the load is being transferred to these yarns.

7.7.4 Impact
Majidi and Chou (1986) reported the impact behavior of
both three-dimensional braided and unidirectional FP/Al-Li com-
posites. The average fiber volume fractions are 17% and 34% for
three-dimensional braided and unidirectional composites, respec-
tively. Instrumented drop-weight impact tests have been carried out
on un-notched impact panels, which do not require machining and,
hence, do not sustain damage to the integrated fiber structure.
Figure 7.51 compares the load-deflection traces obtained from
through-the-thickness penetration impact tests. The three-di-
mensional braided composites absorb significantly higher energy
and show larger deflection than the unidirectional composite during
both damage initiation and propagation stages. By definition, the
initiation energy is the area under the load-deflection curve up to
Mechanical properties of composites 441

Fig. 7.51. Load-deflection traces obtained from through-the-thickness

penetration impact tests, (a) Al-Li alloy; (b) three-dimensional braided
FP/Al-Li composite; (c) unidirectional FP/Al-Li composite. (After
Majidi and Chou 1986.)

10 15 20 25
Deflection (mm)

Fig. 7.52. Cross-sectional views of the FP/Al-Li composite specimens

impacted at 54 J of incident energy, (a) Three-dimensional braided,
Vf = 0.17; (b) three-dimensional braided, Vf = 0.36; (c) [±20°] angle-ply;
(d) unidirectional laminate. (After Majidi and Chou 1987.)
442 Three-dimensional textile structural composites

the maximum load. The total impact energy absorbed by the

three-dimensional braided composites is close to that of the matrix
material. The unidirectional composite fractures like a brittle
material with cracks propagating through the entire specimen, while
in the braided composite damage is restricted to a small region, and
the specimen shows a ductile type of behavior. A comparison of the
cross-sectional views of the FP/Al-Li composite specimen impacted
at 54 J of incident energy is shown in Fig. 7.52.
Additional information on the mechanical behavior of three-
dimensional fabric composites can be found in the work of Kregers
and Teters (1982), Ko and Pastore (1985), Crane and Camponeschi
(1986), Ko (1986), Yau, Ko and Chou (1986), Simonds, Stinchcomb
and Jones (1988), Ko (1989b), and Whitcomb (1989).
8 Flexible composites

8.1 Introduction
The term 'flexible composites' is used hereinafter to identify
composites based upon elastomeric polymers of which the usable
range of deformation is much larger than those of the conventional
thermosetting or thermoplastic polymer-based composites (Chou
and Takahashi 1987). The ability of flexible composites to sustain
large deformation and fatigue loading and still provide high
load-carrying capacity has been mainly analyzed in pneumatic tire
and conveyor belt constructions. However, the unique capability of
flexible composites is yet to be explored and investigated. This
chapter examines the fundamental characteristics of flexible
composites.
Besides tires and conveyor belts, flexible composites can be found
in a wide range of applications. Coated (with PVC, Teflon, rubber,
etc.) fabrics have been used for air- or cable-supported building
structures, tents, parachutes, decelerators in high speed airplanes,
bullet-proof vests, tarpaulin inflated structures such as boats and
escape slides, safety nets, and other inexpensive products. Hoses,
flexible diaphragms, racket strings, surgical replacements, geotex-
tiles, and reinforced membrane structures in general are examples
of flexible composites.
Following Chou (1989, 1990), the nonlinear elastic behavior of
three categories of materials is examined: pneumatic tires, coated
fabrics, and flexible composites containing wavy fibers. These
materials provide the model systems of analysis with elastic be-
haviors ranging from small to large deformations.
The performance characteristics of pneumatic tires are primarily
controlled by the anisotropic properties of the cord/rubber compos-
ite. The low modulus, high elongation rubber contains the air and
provides abrasion resistance and road grip. The high modulus, low
elongation cords carry most of the loads applied to the tire in
service. According to Walter (1978), the first quantitative study of
cord/rubber elastic properties in the tire industry was published in
Germany by Martin (1939), who analyzed bias ply aircraft tires
using thin shell theory to approximate toroidal tire behavior.
444 Flexible composites

Martin's analysis of the orthotropic composite elastic constants

assumes that the fibers are inextensible and the matrix stiffness is
negligibly small; this approach has been referred to here as the
classical netting analysis. Studies of the cord/rubber properties
became active worldwide in the 1960s as represented by the work of
Clark (1963a&b, 1964) in the USA, Gough (1968) in Great Britain,
Akasaka (1959-64) in Japan, and Biderman et al. (1963) in the
Soviet Union.
The existing analysis on tire mechanics is primarily based upon
the well developed anisotropic theory of rigid laminated composites
for small linear elastic deformation. Thus, the problems of vis-
coelasticity, strength behavior, fatigue and large non-linear be-
havior are often ignored.
In the case of coated fabrics, limited attention has been given to
the material stress-strain response to arbitrary loading paths and
histories. Experimental studies of the biaxial stress-strain behavior
can be found in the works of Skelton (1971), Alley and Fairslon
(1972) and Reindhardt (1976). Attempts have also been made by
Akasaka and Yoshida (1972) and by Stubbs and Thomas (1984) to
analytically model the elastic and inelastic properties of coated
fabrics under biaxial loading. Some of these results are briefly
recapitulated in this chapter.
Section 8.4 focusses on the understanding of the large nonlinear
deformation of flexible composites. To this end, model material
systems for analytical purposes need to be identified. The large
nonlinear deformation could originate from two sources, i.e.
matrix and fiber. In order to fully realize the ability of the
elastomeric matrix to sustain large deformation, the fibers must be
able to deform accordingly with the matrix. This can be achieved by
(a) using short fibers, (b) arranging continuous fibers in such an
orientation that they are allowed to rotate as the load increases, and
(c) using reinforcements in woven, knitted, braided, or other wavy
forms.
Possibility (c) is particularly interesting in that it utilizes the
waviness of the fibers. The gradual straightening of the wavy fibers
under external loading results in enhanced stiffness with an increase
in deformation. The linear and nonlinear elastic behavior of two- and
three-dimensional textile structural composites has been examined
by Ishikawa and Chou (1983), Chou (1985), and Chou and Yang
(1986) based on small deformation theory. The nonlinear finite
deformation analyses of flexible composites are presented in Chap-
ter 9.
 CordIrubber composites 445

8.2 Cord/rubber composites

Cord/rubber composites for pneumatic tires are examined
in this section from the viewpoint of the mechanics of anisotropic
materials. Cord/rubber composites are complex elastomeric com-
posites composed of (a) the rubber matrix of usually quite low
modulus and high extensibility, (b) the reinforcing cord of much
higher modulus and lower extensibility than the matrix, and (c) the
adhesive film which bonds the cord to the matrix. The combination
is subjected to (a) fluctuating loads, mostly tensile but on occasion
compressive, (b) temperatures as high as 125°C, and (c) moisture.
Obviously substantial stress develops at the cord-rubber interface.
Some of the discussions presented herein on the materials and
mechanics aspects of pneumatic tires are based upon the review
articles of Walter (1978) and Clark (1980).
The construction of tires involves calendering sheets of rubber
around an array of parallel textile cords to form a flat, essentially
two-dimensional anisotropic sheet. The cords usually have substan-
tial twist and often are made up of two or three oppositely twisted
yarns. These composite sheets are then assembled into various tire
configurations. Figure 8.1(a) shows the typical bias or angle-ply
design which utilizes two or more, usually an even number, of plies
laid at alternate diagonal angles to one another. Figure 8.1(b)
depicts a typical radial tire construction involving radially oriented
cords while the tread area is reinforced by a belt structure of
relatively small angle with respect to the tire center line. The radial
tire construction provides stiff longitudinal reinforcement for the
tread area (and, hence, is less subject to slip) and flexibility for the
vertical deflection. In the terminology of laminated composites, bias

Fig. 8.1. (a) Bias tire, (b) Radial ply tire. (After Clark 1980.)

(b)
446 Flexible composites

and radial tires can be categorized as laminates with [+0/ — 6] and

[+6/—6/90°] orientations with respect to the tire center line.

8.2.1 Rubber and cord properties

For relatively small strain, rubber may be treated as a
homogeneous and isotropic material. The Young's modulus, deter-
mined from the initial slope of the stress-strain curve, may be as
low as 0.69 MPa (100 psi) for non-reinforced (unfilled) elastomers to
as high as 689 MPa (100 000 psi) for highly vulcanized (high sulfur)
compounds such as ebonite. The Young's modulus of rubber is
affected by the conditions of physical testing (i.e. strain rate,
temperature, cyclic load history) and chemical vulcanization para-
meters (i.e. the compounding ingredients, state of cure) (see Clark
1980).
The assumption of negligible volume change of rubber leads to
the following values of Poisson's ratio (v), bulk modulus (K),
Young's modulus (E) and shear modulus (G):

(8.1)

Rubbers used in calendered plies of tires have E values of 5.51 MPa

(800 psi) for textile body ply, 20.67 MPa (3000 psi) for textile tread
ply and 13.78 MPa (2000 psi) for steel tread ply. The v value for
these materials is 0.49.
The Young's moduli for tire cords vary with cord constructions.
The following values are for belt ply: 109.55 GPa (15.9 x 106 psi) for
steel, 24.8 GPa (3.6xl0 6 psi) for Kevlar, and 11.02 GPa (1.6 x
106 psi) for rayon. The values for body ply are: 3.96 GPa (575 x
103psi) for polyester, and 3.45 GPa (500xl0 3 psi) for nylon.
Experiments have shown that textile cords can carry some load in
compression, although compressive loads are believed to be the
source of many textile failures and should be avoided whenever
possible.
Twisting of the cord is needed in order to provide adequate cord
fatigue life under service conditions. However, twisting of fiber into
tire cord can result in as much as a one-third decrease in tensile
Young's modulus for belt ply cords, and a one-half decrease in
Young's modulus for body ply cord. It has been predicted that the
axial Young's modulus of a single twisted fiber yarn is approx-
 Cord / rubber composites AA1

imately equal to 1/(1 + 4JT2R2T2) of that of the untwisted yarn. Here,

R and T denote yarn radius and twist (number of turns per unit
length), respectively (Hearle, Grosberg and Backer 1969). The
twisted and multi-plied cords should be considered as transversely
isotropic, although they are commonly approximated as isotropic.
Textile cords normally show substantial nonlinearity in their
stress-strain behavior. However, since the rubber behavior is
relatively elastic in the small strain range, and the cords in a
laminate are often aligned at an angle to the load direction, the
composite acts more like a linearly elastic solid than the cord itself.
Figure 8.2 shows the stress-strain curve of a tubular specimen using
rayon yarn in a rubber matrix. The fibers in this specimen are in
angle-ply arrangement. According to Clark (1980), most pneumatic
tires do not operate with strain much in excess of 10%.

8.2.2 Unidirectional composites

The linear elastic behavior of a unidirectional cord/rubber
composite can be easily deduced from the basic equations given in
Section 2.2. By assuming that

Ef»E
(8.2)
v m = 0.5

Fig. 8.2. Load-strain curve of a cylindrical tube with rayon yarns in a

rubber matrix. (After Clark 1980.)

Z
-o

Axial strain (%)

448 Flexible composites

The following relations can be obtained:

En = EtVf»E22
v21 = 0
£m (l + 1.3Vf)£m

12 = C2 — = (l +Vf) —

where cx and c2 denote two coefficients.

Akasaka (1989) considered the same assumptions as Eqs. (8.2)
and obtained the simpler form slightly different from Eqs. (8.3), with
the coefficients cx = % and c2 = 1. Then,

4£m
E22
3V
(8.4)
Gm E22
G12
V 4
Akasaka (1989) has noted that the relation of G l2 ~ E22/4- is
independent of cord volume fraction and has good predictability as
compared to existing formulas and experimental results (Walter
and Patel 1979; Clark 1980). Also, cx = c2 = \ has been used by
Jones (1975).
Based upon Eqs. (8.4), the variation of lamina transformed
reduced stiffness with cord off-axis angle 6 follows from Eqs. (2.16)
and can be approximated as (Akasaka and Hirano 1972):

E22 + En ! cos
4
Q22~ E22 + £ l ! sin
2 2
E22/4 + EEn sin 6cos 8
(8.5)
G12 « £22/2 + En sin2 6 cos2 6
<216~ £ u sin 6 cos3 6
<226 ~ Eu sin3 6 cos 6
When a unidirectional cord/rubber sheet is subjected to simple
tension, an interesting deformation behavior occurs, which is not
observed in most of the rigid composites. This can be elucidated by
using Eq. (2.17) for the relation between yxy and the applied oxx as
 CordIrubber composites 449

well as the approximations of Eqs. (8.4)

—2 sin # cos3 0
(2 - tan2 8)axx (8.6)
,22

Thus, the stretching-shear coupling vanishes at 0 — 54.7°, and

Yxy < 0 for 0 < 54.7° and yxy = 0 for 8 > 54.7°.

8.2.3 Laminated composites

The constitutive equations for the laminated cord/rubber
composites are of the same general form as Eqs. (2.25)-(2.30).
However, they can be simplified by using the approximated
expressions of Eqs. (8.5) for Q//- Also, the engineering elastic
constants, referring to the x-y coordinate system, for the angle-ply
laminated composite can be deduced. Using the results of Eqs.
(8.4), the following expressions of engineering elastic constants of a
± 6 laminate in terms of the properties of fiber and matrix as well as
the fiber volume fraction are obtained under the assumptions of
Ef»Em and vm = 0.5 (See Akasaka 1989, and Clark 1963a&b):

Exx = EiVi cos4 8 4G

1-V t
[EfV, sin d cos2 6 + 2G m /(l - Vf)]2
2

Eyy EfVfsin4
i - vf
[EfVf sin 6 cos2 0 + 2G m /(l - Vf)f
2

EfVt-cos4 6 + 4G m /(l - Vt)

Gxy = EfVf sin2 6 cos2 0 + -^~ (8.7)

_ EfVf sin2 6 cos2 6 + 2G m /(l - Vf)

Vxy
~ £ f V f sin 4 0 + 4G m /(l-V f )
_ EfVf sin2 0 cos2 6 + 2G m /(l - Vf)
V

The approach for obtaining Eqs. (8.7) based upon the assumptions
of ±6 cord angles and specially orthotropic symmetry is known as
the modified netting analysis.
450 Flexible composites

The classical netting analysis which assumes inextensible cords

(E{-»°°) simplifies Eqs. (8.7)
Exx = 4Gm(l - Vf)(cot4 6 - cot2 0 + 1)
Eyy = Exx(jt/2-6)
Gxy = EfVf sin2 6 cos2 0 + G m /(1 - Vf) (8.8)
2
vxy = cot 6
vyx = tan 2 6
Figures 8.3-8.5 show the results of analytical predictions based
upon Eqs. (8.5) for Exx, Gxyy and vxy, respectively, as functions of
the off-axis angle 6. These results coincide very well with the
experimental data, as reported by Clark (1963a&b) and based upon
£ n = 1440MPa and E22 = 6.9MPa. It is evident that Poisson's
ratios well in excess of one-half exist in cord/rubber composites.
Because of the incompressibility of the rubber matrix and the
relatively small volume change associated with the cord materials,
due to its high stiffness, it can be assumed that the cord/rubber
composite is incompressible. Thus, for small strain, exx + eyy +
ezz = 0, and
vxz = -EZZ/EXX = 1 + eyY/Exx = 1 - vx (8.9)
Figure 8.6 indicates the analytical results of Eq. (8.9) and the
experimental data of vxz as a function of 6 (Clark 1980) for

Fig. 8.3. Young's modulus, Exx, vs. cord angle, S, for a two-ply laminate.
— Eqs. (8.7); (x) experimental data. (After Clark 1963a.)

107 -

106
15 30 45 60 75 90
Cord angle, 6 (degrees)
 CordIrubber composites 451

£ n = 294MPa and £I22 = 6.6MPa. One of the solid lines is based

upon the vxy expression of Eqs. (8.7), and the simplifying expres-
sion of Eqs. (8.4), namely
En sin2 0 cos2 6 + E22/2
V (810)
=
The other solid line is based upon the vxy expression of Eqs. (8.8).

Fig. 8.4. Shear modulus, Gxy, vs. cord angle, 6, for a two-ply laminate. —
Eqs. (8.7); (x) experimental data. (After Clark 1963a.)

0.0
15 30 45 60

Cord angle, 9 (degrees)

Fig. 8.5. Poisson's ratio, vxy, vs. cord angle, 6, for a two-ply laminate.
Eqs. (8.7); (x) experimental data. (After Clark 1963a.)

2 -

15 30 45 60 75
Cord angle, 9 (degrees)
452 Flexible composites

It is interesting to note that for a range of 6 values, vxz is negative;

the laminate becomes thicker under axial load.
The interlaminar stresses ozzy RZX and rzy are not considered in
the classical lamination theory. These stresses and their correspond-
ing strains do exist in appreciable magnitude which promote a
reduction in the apparent stiffness of cord/rubber laminates. As a
result, the composite becomes more flexible and exhibits lower
natural frequencies of vibration and static buckling loads (Walter
1978).
Walter (1978) reviewed the work of Kelsey, who considered a
two-ply [±0] cord/rubber laminate, simulating the behavior of the
belt in a radial tire. Assuming the belt of finite width in the y
direction is loaded in the x (circumferential) direction, yyz vanishes
due to symmetry and ezz is assumed to be negligibly small. yxz is
maximum at the free edge of the belt and can be approximated, for
the case of inextensible cords (Zsf—»<»), by the simple expression

^ = exx(2 coi 2 d - l ) (8.11)

Equation (8.11) indicates that yxz vanishes when the two plies are
oriented at 6 = ±cot~ 1 V(l/2) = ±54.7°. The magnitude of yxz
decays exponentially away from the free edge and vanishes along
the belt center-line (y = 0). It is interesting to note that 6 = 54.7° is
also the angle for which the normal stress and shear strain are
uncoupled and each off-axis ply behaves as specially orthotropic.

Fig. 8.6. Poisson's ratio, vxz, vs. cord angle, 6, for a two-ply laminate.
Eqs. (8.8) and (8.10); (x) experimental data. (After Akasaka 1989.)

15 30 45 60 75 90
Cord angle, 9 (degrees)
 Cord I rubber composites 453

lnterlaminar shear strains have been observed by inserting

straight pins normal to the ply surface in a cord/rubber belt system
and observing its rotation under extensional load (Bohm 1966) or
by scribing a straight line on the edge of the specimen and
monitoring the rotation of the line under load. Figure 8.7 shows the
interlaminar shear strain measured by X-ray technique for a two-ply
polyester-rubber as a function of cord angle 0 (Lou and Walter
1978). The solid line is based upon the predictions of Eq. (8.11).
The importance of interlaminar shear decreases as the number of
plies increases.
Walter (1978) has presented values of the 18 elastic constants of
Aij, Bij and Dtj for bias, belted-bias and radial constructions; the
material combinations of nylon and rayon body plies with steel,
PVA and rayon belt plies are included. For the case of a specially
orthotropic laminate (A16 = A26 = Dl6 = D26 = Btj= 0) with respect
to the x—y axes, the out-of-plane flexural rigidities are
(EI)X=Anh2112 = £n/i3/12(l-VxyVyx)
(EI)y = A22h2112 = E22h3/ - vyxvxy) ^A2>
where / is the area moment of inertia, and h denotes ply thickness.

8.2.4 Cord loads in tires

According to Clark (1980) the key to good tire design is
long fatigue life. The loads on typical textile cords in pneumatic

Fig. 8.7. Interlaminar shear strain, yxz, vs. cord angle, 0, for a two-ply
polyester rubber. — Eq. (8.11); ( x ) experimental data. (After Lou and
Walter 1978.)

15 30 45 60 90
Cord angle, 6 (degrees)
454 Flexible composites

tires are extremely complex and the sources of loads can be

identified as follows: (a) inflation load, (b) vertical load, (c) steering
forces, (d) road irregularities, (e) camber, (f) speed, and (g) torque.
The tensile cord load due to inflation pressure can be predicted
with some certainty by considering the axisymmetric nature of
inflation and approximating the tire geometry as a thin toroidal
shell. However, this task is complicated by the fact that the tire
does not maintain a constant geometry during inflation. Further-
more, the membrane forces obtained from the thin shell analysis
may not adequately represent the force distributions in the bead
and tread regions. Figure 8.8 shows schematically the cross-section
of a pneumatic tire and the designation of the locations (Clark
1980).
The measurement of cord loads is important to the analysis and
design of tires. Various techniques have been employed; these
include the use of grid or elongation marks for outer plies, X-ray
photography relying on metal markers for inner plies, and resis-
tance foil strain gages imbedded in the tire for direct cord load
measurement in a tire under operating conditions. The force
transducers using resistance foil strain gages are much smaller than
the clip gages, the rubber-wire gages, or the liquid-metal
gages. Details of these measurement techniques can be found in
Clark and Dodge (1969), Patterson (1969), and Walter and Hall
(1969).
The measurements of tire cord loads have indicated that the loads

Fig. 8.8. Location description in a cord/rubber pneumatic tire. (After

Clark 1980.)
, Crown area

Shoulder area

Bead area
 Cordlrubber composites 455

induced by normal direct inflation account for about 10-15% of

cord strength. Another simple type of cord load is induced due to
the load carried by the tire. The cord load at a given location can
fluctuate fairly widely as the tire rolls. Also, the typical cord load
cycle varies with the locations on the tire, i.e. crown, sidewall or
shoulder region. Steering induces additional loads. Relatively small
amounts of steer could induce very large increases in the cord loads.
Figure 8.9 shows the basic characteristics of cord load fluctuation in
a rolling tire (Clark 1980). It should be noted that compressive cord
loads are possible. The characteristics of other cord loads due to
road irregularities, speed and torque are even more difficult to
quantify in a systematic manner.
The measurement of tire cord loads provides the basis of analysis
of the response of cord/rubber composites to the specified boundary
conditions. The netting theory, which only takes into account the
deformation of the cord and neglects completely the contribution of
the matrix rubber, was adopted in the earlier research on bias
constructions. The uncertainty of the orientation of the cord in the
net structure at different stress levels of inflation has limited the
applicability of this theory.
The theory of laminated composites has undoubtedly provided an
efficient means of analysis of cord/rubber composites. It is under-
stood that the theory has its limitations due to the following reasons:
(1) Textile cord strains of several per cent could develop at
some locations in the tire; even larger and nonlinear strains
could develop in the rubber.

Fig. 8.9. Basic characteristics of cord load fluctuation in a rolling tire.

(After Clark 1980.)
Maximum
cord load

o
U Inflation
cord load

Minimum
cord load

Position
456 Flexible composites

(2) Interlaminar deformations are not taken into account in the

theory, assuming plane stress condition.
(3) Cord/rubber composites usually exhibit bimodulus behavior
(Bert and Kumar 1981; Bert and Reddy 1982).
(4) The viscoelastic behavior is assumed to be small and is
often neglected in the analysis.
(5) Perfect cord/rubber interfacial bonds are assumed.
(6) The membrane forces in the bead and tread regions may be
very complex.
(7) Fatigue and hygrothermal loadings may also complicate the
problem.
However, in spite of its limitations, the lamination theory has been
applied with some success for investigating a number of tire
mechanics problems including stress analysis, obstacle enveloping,
treadwear, and vibration. It is thus an efficient tool based upon
linearly elastic, homogeneous and anisotropic material properties
for the representation of nonlinear viscoelastic, heterogeneous
calendered plies of cord/rubber tire composites (Walter 1978). The
large nonlinear deformation of flexible composites is treated in
Chapter 9.

8.3 Coated fabrics

Coated fabrics used in load bearing environments, for
instance, those for air- or cable-supported building structures, tents,
and inflated structures such as escape slides, must exhibit specific
mechanical properties. Some of the general requirements include
retaining flexibility over a wide temperature range, sufficient tensile
and tear strength, low air permeability, and sufficient dimensional
stability (Skelton 1971).
It has been recognized that coated fabrics generally exhibit
nonlinear stress-strain behavior due to straightening of the
crimped yarns under uniaxial or biaxial tension. As noted by
Akasaka (1989), the microscopic deformation behavior of the
woven yarns embedded in the matrix and subjected to membrane
loading is very complex. Thus, modeling of the strength behavior of
these materials requires reasonably precise knowledge of the
deformation of the yarns as a function of load configuration and
magnitude.
The linear elastic properties of laminates composed of coated
fabrics can be readily derived based upon the lamination theory of
Section 2.3. Akasaka and Yoshida (1972) presented explicit expres-
 Coated fabrics 457

sions for elastic moduli of laminates of coated fabrics; the analytical

predictions were compared with experimental data of laminates of
canvas.
Skelton (1971), among others, reported the biaxial stress-strain
behavior of coated orthogonal fabrics. It is concluded that the
stress-strain response at various stages of manufacture of coated
fabrics is dependent mainly on the crimp in the two sets of yarns.
Fig. 8.10. (a) A section of the fabric along warp yarns in off-loom (top),
heat set (middle) and coated (bottom) states, (b) A section of the fabric
along filling yarns in off-loom (top), heat set (middle) and coated (bottom)
states, (c) Surface feature of the fabric in heat set state. (After Skelton
1971 © ASTM. Reprinted with permission.)

(a)

(b)

(c)
458 Flexible composites

The balance of crimp is determined by the restraints imposed on the

fabric during the heat setting process, which precedes the coating
operation. If the fabric is set under tension in the warp direction,
the warp yarns tend to become straight and the yarns in the filling
direction become highly crimped. Thus, when such a fabric is
subjected to biaxial loading, it is almost inextensible in the warp
direction. Consequently, Skelton concluded that if a balanced fabric
is required with similar biaxial tensile behavior in the warp and
filling directions, the fabric must be heat set with both warp and
filling directions under restraint.
It is interesting to recapitulate the experimental observations of
Skelton (1971) for the biaxial testing of coated fabrics. Figures
8.10(a) and (b) show, respectively, the section views of a plain
weave fabric based upon high tenacity polyester. Since the fabric is
set under tension along the warp direction during heat setting, the
warp crimp is minimum and the filling crimp is relatively high.
Three stages, i.e. off-loom, heat set and coated state, are dem-
onstrated. Figure 8.10(c) shows the surface features of the fabric in
the heat set state.
Figure 8.11 shows the biaxial load-elongation curves for this
Fig. 8.11. Bi-axial load-elongation curves for a fabric; load ratio
(warp/fill) = 1:2. WL = warp direction, loom state; FL = filling direction,
loom state; WH = warp direction, heat set; FH =fillingdirection, heat set;
WC = warp direction, coated; FC = filling direction, coated. (After
Skelton 1971 © ASTM. Reprinted with permission.)

FH

-o
o

-3 0 10 15 20 25 30

Elongation (%)
 Nonlinear elastic behavior - incremental analysis 459

fabric with load ratio (warp/filling) = 1:2. The biaxial behavior can
be understood by bearing in mind that in the heat set state the
crimp is unbalanced; the warp yarns are essentially straight and the
filling yarns are highly crimped. Thus, according to Skelton, the
extension of the highly crimped direction of the filling yarns brings
about an increase in crimp and reduction in width in the warp
direction, in spite of the applied load in that direction. Conse-
quently, the load-elongation curve shows negative elongation in the
warp direction at low load level.
The elastic and inelastic responses of coated fabrics have been
studied by Stubbs and Thomas (1984) and Stubbs (1988) using a
space truss model. The model is capable of accounting for arbitrary
loading sequences.

8.4 Nonlinear elastic behavior - incremental analysis

The flexible composites discussed in this section are also
composed of continuous fibers in an elastomeric matrix. Because of
the low shear modulus of the matrix and the highly anisotropic
nature (Enȣ22) of the composites, their effective elastic pro-
perties are very sensitive to the fiber orientation. The geometric
nonlinearity of the flexible composite is mainly caused by the
reorientation of fibers. The material nonlinearity is also pro-
nounced in elastomeric composites under large deformation.
In order to fully realize the ability of the elastomeric matrix
composite to sustain large deformation, Takahashi and Chou (1986),
Takahashi, Kuo and Chou (1986), Chou and Takahashi (1987), and
Takahashi, Yano, Kuo and Chou (1987) have predicted the
nonlinear constitutive relation of flexible composites with sinusoid-
ally shaped fibers based upon a step-wise incremental analysis and
the classical lamination theory. In this section, the work of Chou
and Takahashi is recapitulated. Both fiber geometric nonlinearity
and matrix material nonlinearity have been taken into account.
Because of the superposition of the infinitesimal solutions from
lamination theory, the limitation of this approach is obvious.
However, being a well established analytical technique in the
composites field, the lamination theory does provide a convenient
tool for discerning the basic characteristics of flexible composites.
Comparisons are made between the analytical predictions and
experimental data for tire cord/rubber, and glass and
Kevlar/silicone-elastomer flexible composite laminae. Since com-
posites with fibers in wavy form have been used as a model system,
the geometric aspects of curved fibers are examined first.
460 Flexible composites

8.4.1 Geometry of wavy fibers

To demonstrate the effect of fiber extensibility from ge-
ometric design, a flexible composite composed of continuous fibers
with sinusoidal waviness in a ductile matrix is used as a model
system. Perfect bonding between the fibers and matrix is assumed.
The geometric relations among the wavelength (A), amplitude (a),
and fiber length (s) of a sinusoidally shaped fiber are identified first.
Then, two types of fiber arrangements are considered: the
iso-phase model, and the random-phase model. The fibers are
assumed to maintain the sinusoidal shape of which the geometric
parameters A, a and s vary with the increase of applied load.
The spatial position of a typical fiber in the xyz coordinates is
given by:
2JTX
y = a sin - (8.13)

where the parameters a and A of the curved fiber are shown in Fig.

Fig. 8.12. Geometrical relationships between aIk, s/X and 0max, where
0max is the maximum angle between the fiber and x axis. (After Chou and
Takahashi 1987.)

10 -
 Nonlinear elastic behavior - incremental analysis 461

8.12. The angle 6 between the tangent to the fiber and x axis is a
function of x:
dy 2na 2JTX
tan 6 = — = ——cos— (8.14)
dx A A
The length of fiber, ds, between x and x + dx is

ds = V(ck2 + dy2) = dxJ\l + c• c o s 2 ( ^ ) l (8.15)

where

Obviously, the maximum value of tan 6 occurs at

(8.16)
\ A, /

The fiber length, s} between x = 0 and A is given by

r A C 2
s= \ds=— A C2V ( l + c • cos P) dp (8.17)
J 2JT JO

By the use of an elliptic integral of the second kind,

1 I 2 •3 I2 • 32 • 5

(8.18)
where

k2 = T+^ (8.19)

Equation (8.18) can be rewritten as

A y/(l-k2)

-Tilf-fijf--)
By the use of Taylor expansion, we have
s (k2\ (k2\2 Ik2
r l + 2 (-) + 13(-) +90(¥
/ \ /A:2\5
+ 644 - +4708.5 - +•••• (8.21)
462 Flexible composites

In the following analysis, terms up to (k2/S)5 in Eq. (8.21) are

taken into account, and the range of a/k is limited to below 5. The
relationship between a/k and s/k is shown in Fig. 8.12 where 0max is
the maximum angle between the fiber and the x axis. For example,
for 0max = 2O°, a l l = 0.058 and s/k = 1.032. The curved fiber
composite with a/k = 0.10 can be extended up to 9.23% of its
original length only by the straightening of the fiber, if the matrix
stiffness is negligible.
Two kinds of arrangements of the curved fibers in the composite
have been considered: the iso-phase model and random-phase
model. The iso-phase model is defined in Fig. 8.13, where all the
fibers are in the same phase in the x direction. The distance
between the fibers in the y direction is assumed to be constant. In
the random phase model (Fig. 8.14), the axial locations of sinusoid-
al shaped fibers do not assume any regular pattern.

8.4.2 Axial tensile behavior

The nonlinear tensile stress-strain behavior of flexible
composites containing wavy fibers has been investigated according
to the iso-phase and random-phase models. The lamination theory
described in Section 2.3 is the basis of this analysis. The applied
load is parallel to the axes of the sinusoidally shaped fibers.

8.4.2.1 Iso-phase model

The linear elastic stress-strain relations are derived first.
Consider Fig. 8.13; each volume element between x and x + dx is

Fig. 8.13. Iso-phase model. (After Chou and Takahashi 1987.)

 Non-linear elastic behavior — incremental analysis 463

approximated by a unidirectional straight fiber composite, in which

fibers are inclined at an angle 6 to the x axis, as defined by Eq.
(8.14). The transformation of coordinates between the composite
reference axes (xyz) and the fiber local axes (LTz) is given by:

L\ I cos 8 sine 0
T I = I -sin 8 cos 8 0 (8.22)
0 0 1

The positive direction of 6 is defined in Fig. 8.13. Under the

uniaxial tension, oxx, Eq. (2.17) gives

e
xx = Sn o XJ

(8.23)

Yxy =

It is interesting to note the stretching-shear coupling represented by

516. Figure 8.15 shows schematically the yxy induced by an applied
stress axx.
The average tensile strain of the iso-phase composite, exx, is

.-If* £„ dx (8.24)

Fig. 8.14. Random-phase model. (After Chou and Takahashi 1987.)

464 Flexible composites

From Eqs. (8.14) and (8.23),

(8.25)

The effective Young's modulus of the iso-phase model in the x

direction is given by
(1 + c),3/2
E* =•

1 c - (1 + c) 3/2 )s 22 + ^ (2512 + 566)

*-* rr —

(8.26)
In a small volume element between x and x + dx, the tensile
strain of the fiber along its axial direction is expressed by
EL = sxx cos2 6 + eyy sin2 6 + yxy sin 0 cos 0 (8.27)
Substituting Eqs. (8.23) into Eq. (8.27) and integrating over s, the
average fiber axial strain is

et = - f e L ds = [(S n - S 12 )F(*) + S12]CT (8.28)

s =[
where
(8.29)

Fig. 8.15. Schematic illustration of the deformed shape of the iso-phase

model under uniaxial tension oxx. (After Chou and Takahashi 1987.)

Oxx
 Nonlinear elastic behavior - incremental analysis 465

Here, the relations among s, x and A, Eqs. (8.14)-(8.20), and the

elliptic integral are used in the derivations.

8.4.2.2 Random-phase model

In the case of the iso-phase model, the stretching-shear
coupling constants 516 and S26 do not vanish. This coupling effect
could be eliminated through the random positioning of wavy fibers
along the x-direction:
y=a sin(2^(x - d)/k) (8.30)
where d is the translation of the fiber in the x direction. A random
distribution of d ( 0 < d < A ) is assumed in this model. That is, in
each infinitesimal section, dx, fibers with any arbitrary
orientation angles exist with the same probability. Therefore, it is
assumed that exx is uniform throughout the sample under uniaxial
tension. The stress in a fiber segment depends on its orientation, 6:
2jta 2jta
— ^ tan 0 < ——
A A
By these assumptions, the classical lamination theory can again be
applied.
The stress-strain relations of a unidirectional lamina consisting of
straight fibers are given by Eq. (2.13) with the reduced stiffness Qtj
given by Eqs. (2.14). The transformed stress-strain relations of an
off-axis lamina, referring to the x-y coordinate system, are given by
Eqs. (2.15) and (2.16). The small element of the random-phase
composite situated between the sections at x and x + dx is treated as
a laminate with different orientations. The fibers with the orienta-
tion angle 6 which lies in the range defined by

^ ) (8.31)
\ A I

have the probability dx/A.

Therefore, the stress-strain relation of the laminate can be
rewritten as

(8.32)
\%J \C?6 C2*6 CtJ\yx
where
2mn{d)Ax (8.33)
A Jo
466 Flexible composites

The average stiffness constants of Eq. (8.33) are

Cn = ( 1 ^ 3 / 2 [Gn(l + 0 + (G« + 2Q66)c

• - c

=
(l + cf2 [{Qn + Ql2 ~ I \
= (1 H [(Gn + G22 - 2Q12 - 2G66)

C*6= C | 6 = 0

Inversion of Eq. (8.32) leads to

(8.35)
66/ \^xy/
where
S*n = (C^CSs - C2*62)/D
52*2 = (Cr i C 6 * 6 -C 1 *6 2 )/£>
(8.36)

<
— *"11*- 22 ^ 6 6 -'12»-'66

Following Eqs. (2.9), the Young's modulus and Poisson's ratio in

the x direction of the random-phase model are given by:
E*xx=\IS*n
(
v*_ c*/c* 8-3 7 )
If the random-phase model is subjected to uniaxial tension, axx,
 Nonlinear elastic behavior - incremental analysis 467

the strain components are

£
xx — °xxlE'xx

xx (8.38)

The strain of the fiber in its axial direction is calculated by

substituting Eqs. (8.38) into Eq. (8.27) and averaging over s (Eq.
(8.28))
eyy (8.39)
where F(k) is given by Eq. (8.29).

8.4.2.3 Nonlinear tensile stress-strain behavior

The nonlinear axial (X direction) tensile stress-strain
behavior of the flexible composite is examined using the stepwise
incremental analysis of Petit and Waddoups (1969). Consider an
incremental tensile strain Aexx, applied on either the iso-phase or
random-phase model. Here, Aexx = A///; A/ and / are the
incremental length and the current length, respectively. Using the
initial Young's modulus Exx, the first stress increment, Aoxxy is
calculated by the linear elastic relation:
Aoxx = E*xxAexx (8.40)
where the expressions of Exx are given by Eqs. (8.26) and (8.37) for
the iso-phase and random-phase models, respectively. The nth
stress increment is added to the previous stress state after n - \
increments to determine the current total stress:
(a = (On-i + (AO« (8.41)
For the iso-phase model, the average tensile strain increment of
the fiber along its axial direction, Ae£, is obtained by substituting
Eq. (8.40) into Eq. (8.28):
Aet = [(S u - S12)F(k) + 512] Aaxx (8.42)
For the random-phase model, the transverse strain increment, Aeyy,
is determined from Aexx and v*y:
Aeyy = -v*xyAexx (8.43)
Then, the tensile strain increment of the fiber is calculated by
substituting Aexx and Aeyy into Eq. (8.39):
Ae*L = (&exx - Aeyy)F(k) + Aeyy (8.44)
468 Flexible composites

The total strain, referring to the current specimen length, after n

increments is

(8-45)
;=i /=!w/

Replacing A/ by the infinitesimal increment, dl, it follows:

f'd/ . /
exx = — = In y = ln(l + exx) (8.46)

Here, exx is the tensile strain referred to the initial specimen length

(8.47)
'o

In the range of large strain, the use of exx is more convenient than
the summation of Aexx. From Eq. (8.46)
exx = exp(exx) - 1 (8.48)
Then, the total strain, after the nth increment, in the axial direction
(exx), transverse direction (eyy) and the fiber (el) are given by:

(£,,)„ = exp[2 (AO,1 - 1 (8.49)

L/=l J

(eyy)n = e x p [ i (Ae w ),1 - 1 (8.50)

L/=l J

l (8.51)

Finally, the change of fiber shape under loading needs to be taken

into account. Due to the tensile loading in the x direction, the
wavelength of the curved fiber is changed to
X = Xo(l + exx) (8.52)
where A and Ao are, respectively, the current and initial values of
the wavelength, and the total strain exx is given by Eq. (8.49). The
current value of the fiber length is
s = so(l + et) (8.53)
where so is the initial fiber length and E£ is the total fiber strain
given by Eq. (8.51).
 Nonlinear elastic behavior - incremental analysis 469

In order to determine the shape of the fiber, it is assumed that the

fiber maintains a sinusoidal waviness during deformation while
varying its amplitude (a) and wavelength (A). The current value of
the amplitude, a, can be determined by Fig. 8.12 from the given
current values of A and s. The values of k2 = c/(l + c), c =
(ijtalX)2, Exx and v*y after the nth step are determined from the
current values of A and a, and these values are used in the (n + l)th
step of the incremental analysis. Eqs. (8.41) and (8.49) give the
uniaxial tensile stress-strain relation of the flexible composite.
The elastic constants of fibers (Chamis 1984) and matrices
(Modern Plastics Encyclopedia 1983) used in the numerical calcula-
tions of Chou and Takahashi (1987) are shown in Table 8.1. Linear
elastic stress-strain relations are assumed for glass and Kevlar
fibers. Rubber elasticity (James and Guth 1943; Treloar 1973) is
assumed for PBT and the other elastomeric polymers:

(8.54)

where ££, is the initial Young's modulus of the matrix, and a is the
extension ratio:
a=l + exx (8.55)
The secant Young's modulus of the matrix, £m, is determined from
the current tensile strain, exxy (Eqs. (8.49) and (8.55)):

£
m = -r^- = ^ r ( l + A ) (8-56)
dexx 3 V a
Numerical examples of the stress-strain relations predicted by the
incremental analysis are shown in Figs. 8.16 and 8.17. The results
indicate that Kevlar is less effective than glass fiber in contributing

Table 8.1. Elastic constants and elongations (Chou and Takahashi 1987)

EL ET ELET eb(%)

(GPa) (GPa) (GPa)

Glass 72.52 29.7 0.22 4
fiber
Kevlar 151.6 4.13 2.89 0.35 0.35 3.5
PBT 2.156 0.77 0.4 50-300
matrix

Isotropic relation G = E/2(l + v) is assumed.

470 Flexible composites

Fig. 8.16. Comparisons of the effects of glass and Kevlar fibers on the
tensile stress (axx)-strmn (exx) curves for an iso-phase composite at
various £^. Rubber elasticity is assumed for the matrix. Crosses (x) show
average fiber axial tensile strain; e* reaches 4% and 3.5% for glass and
Kevlar fibers, respectively. vm = 0.4, Vf = 50%, fl/A = 0.1. (After Chou
and Takahashi 1987.)

Glass
Kevlar
l
/
/

/
0
E — in GPa
m = 10
OH 2 - /
O /

/ /
a i —_ 1 . /'I

y
^ ^
h
0.1 I'i
5 10 15
Strain, exx (%)

Fig. 8.17. Tensile stress (a^J-strain (exx) curves of Kevlar/PBT polymer

composites predicted by using the iso-phase (solid line) and random-phase
(dotted line) models. £ ° = l G P a , vm = 0.4 and Vf=50%. Crosses (x)
show average fiber axial tensile strain; £* reaches 3.5%. (After Chou and
Takahashi 1987.)

5 10 15
Strain, £„ (%)
 Nonlinear elastic behavior - incremental analysis 471

to the stiffness of curved fiber composites, because the transverse

Young's modulus of Kevlar is lower than that of glass. After the
wavy fibers are stretched, however, Kevlar becomes increasingly
more effective with regard to stiffness and strength (Fig. 8.16). For
a given wavy fiber composite, the random-phase model predicts
higher Young's modulus and lower elongation than those of the
iso-phase model (Fig. 8.17).

8.4.3 Transverse tensile behavior

The transverse tensile behavior of wavy fiber composites
has been analyzed for both iso-phase and random-phase models by
Kuo, Takahashi and Chou (1988). The lamination theory is again
the basis of the incremental analysis.

8.4.3.1 lso-phase model

Consider the small volume element situated between y and
y + Ay in the iso-phase model as shown in Fig. 8.13. It is assumed
that the transverse stress oyy is uniformly distributed along one
wavelength A. Then an element of the size dy dx can be treated as
an off-axis unidirectional lamina. From Eq. (2.17) and plane stress
condition, the strain components in this element are
= =
Exx — d\2Oyy yy S22oyy y xy S26oyy (8.57)

Then the transverse strain averaged over the wavelength of the

iso-phase model is

eyy (k (8.58)

The effective Young's modulus in the y direction is

(l + c)

+ cf 2 - 1 - | ) s n + (l + C-)s22 + ^ (2512 + S66)

(8.59)
Following the approach of Section 8.4.2.1, the average tensile strain
along the fibers due to transverse tension is obtained by substituting
Eqs. (8.57) into Eq. (8.27) and averaging over the length s:
+ Sn]ayy (8.60)
F(k) is given by Eq. (8.29).
472 Flexible composites

Kuo, Takahashi and Chou (1988) also analyzed the transverse

tensile behavior based upon the constant strain assumption. This
assumption is validated by the observation during transverse tension
experiments that the elongation of the specimen is uniform through-
out its width away from the specimen ends. Although the
constitutive relations are not of the same form for constant stress
and constant strain analyses, the numerical calculations in Kuo,
Takahashi and Chou (1988) yield the same result. This is the direct
consequence of the approaches, namely the stress (or strain) is
considered in the average sense along the x direction.

8.4.3.2 Random-phase model

The transverse Young's modulus and minor Poisson's ratios
are given by
E*vv = l/S*2*
V
* _ _ o * /J n * (8.61)
yx — >J12/ 22

Fig. 8.18. Comparisons between theoretical predictions and experimental

data of transverse tension of an iso-phase model. Specimen initial
a/k = 0.05-0.07 and V{= 1.337% for Thornel-300/silicone elastomer com-
posites. (After Kuo, Takahashi and Chou 1988.)

0.6
O Specimen 1
# Specimen 2
all = 0.03
0.5 all = 0.05
a/l = 0.09

0.4

40 50
 Nonlinear elastic behavior - incremental analysis 473

Under the transverse stress, oyyy the strain components are

£xx = -(v*x/ E;y)oyy
eyy = oyy/Eyy (8.62)

Again, the average tensile strain along the fiber is obtained from
Eq. (8.39).
Figure 8.18 depicts the comparison between theoretical curves
and experimental data of an iso-phase model under transverse
tension. The experimental material system reported by Kuo,
Takahashi and Chou (1988) is based upon Sylgard 184 silicone
elastomer reinforced with Thornel-300 carbon fiber. A loose fiber
bundle contains 1000 filaments, with a filament diameter of 7 jum.
The specimen fabrication technique follows that given by Luo and
Chou (1988). The initial aIA values of the specimens are in the
range of 0.05-0.07. The fiber volume fraction is very low, about
1.34%.
 Nonlinear elastic finite deformation of
flexible composites

9.1 Introduction
Flexible composites, which are described in Chapter 8,
behave very differently from conventional rigid polymer composites in
the following ways:
(1) Flexible composites are highly anisotropic (i.e. longitudinal
elastic modulus/transverse elastic modulus » 1). Figure 9.1
compares the normalized effective Young's modulus
(Exx/E22) vs. fiber orientation for two types of unidirec-
tional composites. The upper curve obtained from Kevlar-
49/silicone elastomer shows that the stiffness of the elas-
tomeric composite lamina is very sensitive to the fiber
orientation. At a 5° off-axis fiber orientation, for example, a
1° change in fiber angle causes the effective stiffness to
change by 53%. The lower curve obtained from Kevlar-
49/epoxy shows less than 7% change at the same off-axis
angle.
(2) Flexible composites show low shear modulus and hence
large shear distortion, which allows the fibers to change
their orientations under loading.
(3) Flexible composites have a much larger elastic deformation
range than that of conventional rigid polymer composites.
Thus, the geometric changes of the configuration (i.e. area,
direction, etc.) need to be taken into consideration.
(4) The nonlinear elastic behavior with stretching-shear cou-
pling, due to material and geometrical effects, is pro-
nounced in flexible composites under finite deformation.
Therefore, the conventional linear elastic theory, based on the
infinitesimal strain assumption for rigid matrix composites, may no
longer be applicable to elastomeric composites under finite
deformation.
The theories of non-linear and finite elasticity made a major
advancement during the Second World War, in response to the
development of the rubber industry. M. Mooney, in 1940, advanced
his well-known strain-energy function. Rivlin and colleagues (for
example, Rivlin 1948a&b; Rivlin and Saunders 1951; Ericksen and
 Introduction 475

Rivlin 1954), in a series of publications starting in 1948, successfully

predicted the large deformation of rubber-like incompressible
isotropic material. These works have greatly enhanced and stimu-
lated the development of nonlinear finite elasticity. The fundamen-
tal aspects of finite elasticity can be found in advanced text books
(for example, Truesdell 1966; Fung 1977; Malvern 1969; Spencer
1972; Lai, Rubin and Krempl 1978).
To predict the large deformation of fiber reinforced rubber
material, Adkins and Rivlin (1955) treated the nonlinear, aniso-
tropic, and finite deformation problem by using the 'ideal fiber
reinforced material theory'; the assumptions of volume incompres-
sibility and fiber inextensibility are basic to the analysis. Further
developments of this theory can be found in the work of Rivlin
(1964), Pipkin and Rogers (1971) and Spencer (1972). Difficulties
often arise in applying this theory to composites with complicated
fiber geometries and in cases where the extension of the fibers
cannot be neglected.

Fig. 9.1. Variation of effective Young's modulus with fiber orientation.

(After Luo, 1988.)
10 000

1000
faj

Kevlar/silicone elastomer

20 40 60 80
Fiber orientation (degrees)
476 Nonlinear elastic finite deformation

The ability of flexible composites to sustain large deformation and

fatigue loading, and still provide high load carrying capacity, has
been mainly analyzed in textile cord/rubber composites and coated
fabrics. However, most of the existing analyses on the mechanics of
pneumatic tires are primarily based on the composite lamination
theory for small linear deformation. Chou (1989) has provided a
review of the mechanics of flexible composites.
In recent years, the constitutive relation of biological materials
has been a subject of considerable research interest. A variety of
biological materials are incompressible, viscoelastic, and anisotro-
pic; they often demonstrate nonlinear behavior with a large
deformation range (Fung 1981). For instance, Aspden (1986)
considered the influences of fiber reorientation in biological mate-
rials during finite deformation by using a fiber orientation distribu-
tion function and assuming that the fiber carries only axial tension.
However, the finite deformation and the rigid body rotation of
fibers, as well as the shear property of the matrix material, which
greatly influence the fiber reorientation during deformation, are not
adequately considered in the analysis. Humphrey and Yin (1987)
presented a constitutive model based upon a pseudostrain-energy
function, and compared the theoretical analysis with both uniaxial
and biaxial experimental results. The parameters used in the energy
function are dependent on the experimental data; the fiber spatial
arrangements, which are responsible for the geometric nonlinearity,
are ignored in their analysis.
Various response functions have been proposed to represent the
experimentally determined nonlinear stress-strain curves in prin-
cipal material directions. Petit and Waddoups (1969) employed the
increment method. Hahn (1973) and Hahn and Tsai (1973) used the
complementary energy density to derive the stress-strain relation,
which is nonlinear in shear but linear in tensile properties. Jones
and Morgan (1977) used an orthotropic material model in which the
nonlinear mechanical properties are functions of the elastic energy
density. The nonlinear elastic behavior of textile structural com-
posites has been examined by Ishikawa and Chou (1983). How-
ever, these analyses are restricted to a small strain range.
In an effort to provide a rigorous treatment of the finite
deformation problem, two analytical approaches, considering both
geometric and material nonlinearities, have been employed in this
chapter to predict the constitutive relations of flexible composites
(R. S. Rivlin, private communication, 1986; Luo and Chou, 1988a,
1990a&b).
 Introduction 477

(1) In the first method (Section 9.3), a closed form repre-

sentation of the constitutive equations has been derived
based on the Lagrangian description. The strain-energy
density is assumed to be a function of the Lagrangian strain
components referring to the initial principal material coord-
inate X (Fig. 9.2a).
(2) In the second approach (Section 9.4), a nonlinear constitu-
tive relation has also been developed based upon the
Eulerian description where the deformed configuration of
the composite is used as the reference state. A stress-
energy function, referring to the moving principal material

Fig. 9.2. A rectangular element of composite lamina before and after

loading (a) in the Lagrangian system, (b) in the Eulerian system. (After
Luo and Chou 1988a.)

Deformed body

Fiber

(a)

t',X2

Deformed body

Fiber
478 Nonlinear elastic finite deformation

coordinate x (Fig. 9.2b), provides the basis for deriving the

constitutive relations; and an iterative calculation method is
employed.
The constitutive relations obtained from Sections 9.3 and 9.4
have been applied to study the nonlinear elastic behavior of flexible
composites with wavy fibers in Section 9.5. Section 9.2, which is
based upon Luo (1988), provides the basis for the theoretical
treatment of this chapter.

9.2 Background

9.2.1 Tensor notation

Some brief descriptions of the notations and operations of
tensors are shown in this section. These are taken from various
sources, including Fung (1965, 1977), Rivlin (1970) and Lai, Rubin
and Krempl (1978). It is not intended to provide a comprehensive
coverage of tensor analysis. Only the subjects that are relevant to
the present work are described. For simplicity, only Cartesian
tensors are used, and thus the distinction between contra variance
and covariance disappears and all indices of the tensor components
can be written as subscripts. Furthermore, tensors are printed in
bold-faced letters.

Einstein summation convention

The following three equations have the same meaning:

= ailX1 + ai2X2 + 0/3*3 + * ' * + 0«n *n (9.1)

The first line of Eq. (9.1) follows the rule of Einstein summation.
Here, / is known as the dummy index, which repeats once, denoting
a summation with respect to that index over its range, and i is a free
index, which appears once in every term of the equation, assuming
the numbers of 1, 2 or 3. The following are two other examples:
(1) For the two vectors a = a,e,, and b = biei (/ = 1,2,3), the
scalar product is defined by
3 3

c = a • b = 2 E «A( e, • e,) - a,&y(e, • e,) (9.2)

,-=ly = l
 Background 479

(2) For the matrices a = [atj\ and b = [Z>,y] (i = 1,2,3, ; = 1,2,3),

the product of these two matrices is

[c,y] = ab = (9.3)

Kronecker delta
The Kronecker delta 6 is defined as

fl for i=j
(9.4)
10 for

or
[du 8 1 0 0
°2i $22 <523 = 0 1 0 (9.5)
d3l 6 32 O33_ 0 0 1
The following relations are useful:
(1) <5,y = 3 (9.6)
(2) 6itnTmJ = 7-0 (9.7)
(3) For the mutually perpendicular unit vectors e l5 e2 and e3,
(9.8)

Permutation symbol
The permutation symbol is defined as

{ 1
— 1
0
if ijk is even permutation of 1, 2, 3
if ijk is odd permutation of 1, 2, 3
otherwise (i.e. if two of the indices are equal)
(9.9)
where the even and odd permutations are indicated as

Even permutation Odd permutation

480 Nonlinear elastic finite deformation

The following relations are also used in this chapter:

vectors a = a & , and b = b f i i , then
(1) For the two
a X b = eijkaibjek
.s normal to .he plan, c o n t a i n i
where the unit vector . ,

mn
det(m) = m22
(9.H)

used infiniteelasticity. The d - j j ^ a f t j r e l a t ^ ^ ^ ^

undeformed and \ and D (Fig. 9.3). To
considered as a m'appi be twofixedrectangu-
find the transformation u wiui — original and deformed
lar Cartesian coordinates position of a —*tedc particle P
configurations, respective! Position veCt°f X " *
inside the domain £>o is ' 1; this partide assumes
 Background 481

figuration can be described mathematically by the coordinate

transformation between Xj and xt.
The coordinate system Xl-X2-X3 is chosen as the reference
system. The description of deformation, of which the independent
variable is the particle position vector X in the original state, is
known as the Lagrangian description. The reference system X is
known as a Lagrangian coordinate. The transformation equation in
terms of Xj is
xi=xi(XuX2>X3) (9.12)
where the specified function XT (Xu X2y X3) is assumed to be
continuous and differentiate. It follows, then:

dx^-^dXj^XtjdXj (9.13)

where the dummy, or repeating, index,7, denotes a summation over

its range. The matrix form of Eq. (9.13) can be written as
[dx] = [g][dX], where the deformation gradient matrix [g] is

The strain tensor associated with the Lagrangian system is called

the Lagrangian strain (£,y), also known as the Green's or St.
Venanfs strain, and it is defined in matrix form as

where [g]T is the transpose of the deformation gradient matrix [g],

and [6] is the Kronecker delta. The explicit form of Eq. (9.15) is
given by
= 1 / 3*i 3*! dx2 dx2
11
2\3X9X 3XdX
\ldxx dx> 3x2 dx2 \
X2aX2 dX2aX2 I
= 1 /dxx dxi dx2 dx2
12 21
2 \3Xl 3X2 3XX 3X2
On the other hand, the coordinate system x1-x2-x3 can be chosen
as the reference system. Then the description, in which the
independent variable is the particle position vector x in the de-
482 Nonlinear elasticfinitedeformation

formed state, is known as the Eulerian description, and the

reference system x is known as an Eulerian coordinate. The
transformation equation in terms of xt is Xt — Xt (xXy x2, x3). Then,
[dX] = [ g H d x ] (9.17)
where [g]" 1 is the inverse matrix of [g] with the components

e^ 1 = —z_J^°A£j/ZJ_ (9 18")
dXj det g
where [co(g,y)]T is the transpose of the cofactor matrix of gijy and
det g is the determinant of [g].
The strain tensor associated with the Eulerian system (in terms of
the deformed configuration) is termed the Eulerian strain (e,y), which
is also known as AlmansVs strain for large deformation and Cauchy's
strain for infinitesimal deformation (Fung 1977). It is defined as

or

x dxx

) (9.20)

(axlax1 ax2ax2\
2el2 = 2e2l= - I — +- -—
V ox i dx2 dXi dx2
Equations (9.16) and (9.20) can be rewritten in terms of the
displacement vectors U and u, which are associated with the
coordinate systems X and x (Fig. 9.3), respectively, and can be
expressed as
U=u=x-X (9.21)
Then, the alternate formulas for Lagrangian and Eulerian strain
tensors are

and
(923)
M^^^) -
1 ldui duL_ duJi duh\
 Background 483

If the displacement gradients are sufficiently small, the quadratic

terms in Eqs. (9.22) and (9.23) can be neglected in comparison with
the linear terms. Then, the Lagrangian and Eulerian strain tensors
are reduced to the linear forms, and both are equal to the strain
(e,y) for infinitesimal deformation

En = ~ •+- (9.24)

The force acting per unit area is known as stress. In the case of
finite deformation, the area and normal direction of a surface of an
undeformed element may be quite different from those of the same
surface in the deformed state. Thus, the stress can be defined by the
force per either undeformed or deformed area; the former is known
as the Piola-Kirchhoff stress or Lagrangian stress (n,y), and the
latter is known as the Cauchy's stress or Eulerian stress (a,y).
For the purpose of illustration, consider a rectangular element
ABCD of unit thickness, which is deformed into A'B'C'D' under a
uniaxial load P (Fig. 9.4), and neglect the dimensional change in the
thickness direction. The Eulerian stress is defined as the force per
unit deformed area,

orr =-CD' (9.25)

The Lagrangian stress is defined as the force, which is acting on the

deformed surface, divided by the original surface area (correspond-

Fig. 9.4. A force P acting on a deformable body.

Undeformed body
Deformed body
B / c /
B' /c
P P

A' D'
4 D
484 Nonlinear elasticfinitedeformation

ing to the deformed area),

P
n x (9.26)
* ~~cb
From Eqs. (9.25) and (9.26),
oxxC'D' = UxxCD = P (9.27)
The general relations between the two stress descriptions can be
found by analyzing the deformation of a generic two-dimensional
element as shown in Fig. 9.5. Let QR, with area 6A and normal N,
be the edge surface of the element OQR in the undeformed state.
The corresponding edge surface in the deformed state is Q'R' with
area da and normal n. The coordinates xx-x2 are fixed on the
deformed element. Also let the surface force vector per unit area of
the deformed surface {Q'R') be f (traction), then the total surface
force acting on Q'R' is ida. The nominal traction, F, acting on the
undeformed edge surface (QR) is so defined that
F6A = (da = P (9.28)

Fig. 9.5. The correspondence between Lagrangian stress and Eulerian

stress.

Undeformed element OQR

0,2

Deformed element O'Q'R'

 Relations based on the Lagrangian description 485

where P is the actual force acting on the surface element Q'R' in

the deformed state. By neglecting the body force and assuming that
the displacement rate with respect to time is very small, the
following relation from force equilibrium can be obtained:
f^Ojtn, (9.29)
where f and nj denote the components of f and n, respectively.
Similarly,
i? = iyv} (9.30)
where Ft and Nj denote the components of F and N, respectively.
Substituting Eqs. (9.29) and (9.30) into Eq. (9.28) and applying
the relation

/iy^^ (9.31)

the result is
a = (detg)-1gn (9.32)
or
^(detgr^JI*, (9.33)
where detg is the determinant of g. The inversion of Eq. (9.32)
gives
n = (detg)g' 1 a (9.34)
or
nji=(detg)gjk 1 a k i (9.35)
Note that the Eulerian stress tensor is symmetric (i.e. otj = ay7)
whereas the Lagrangian stress tensor (Eq. (9.34)) is not symmetric.
However, the following quantity

PAB =
~dt n B i = (det Z)gA>gBi °i> (9 36)
.
is symmetric, and PAB is known as the second Piola-Kirchhoff stress
tensor, or simply the Kirchhoff stress.

9.3 Constitutive relations based on the Lagrangian description

9.3.1 Finite deformation of a composite lamina

A basic element in a flexible composite is assumed to be a
thin lamina consisting of straight, parallel continuous elastic fibers
486 Nonlinear elastic finite deformation

embedded in an elastic matrix which can sustain large deformation.

It is also assumed that the lamina is homogeneous on a scale much
larger than that of the inter-fiber spacing. Then, the flexible
composite lamina can be treated as a homogeneous two-
dimensional orthotropic elastic continuum. In this section, the
constitutive equations for such an element under finite deformation
are derived based on the Lagrangian description.
Figure 9.2(a) illustrates a unidirectional flexible composite lamina
under a finite deformation, where the initial fiber orientation is at
an angle 6O with respect to the Xx axis. The rectangular Cartesian
coordinates l-t are along the initial fiber and transverse directions,
respectively. Under loading, the rectangular element ABCD in the
undeformed lamina is deformed into a quadrilateral element
A'B'C'D' in the deformed lamina. There is an angle A6 between
AD and A'D1. Corresponding to this change, the current fiber
orientation /' is at an angle 6 with respect to the Xx axis, and

6 = eo + A6 (9.37)
Because of the low shear modulus of the matrix and the highly
anisotropic nature (En» E22) of flexible composites, A0 may be
quite large and the effective elastic properties of the composite
become very sensitive to the fiber orientation. The geometric
nonlinearity of a flexible composite is mainly caused by the
reorientation of fibers. The material nonlinearity is also pro-
nounced in elastomeric composites under large deformation.
The deformation of the basic element ABCD (Fig. 9.2a) is
further examined in Fig. 9.6. Let the rectangular Cartesian coordin-

Fig. 9.6. Deformation of a rectangular element of a composite lamina,

referring to the principal material coordinate system. (After Luo and Chou
1990b.)
't, X

A A'
d/
 Relations based on the Lagrangian description 487

ate system X coincide with the initial principal material coordinates

l-t, where the axis Xx is parallel to /. Here, a quantity with an over
bar refers to the initial principal material coordinates. Then, the
Lagrangian strain matrix in the system X is written as
(9-38)
The deformation of the element shown in Fig. 9.6 can be
expressed in terms of these Lagrangian strain components. Let the
line elements AD — dlo and AB = dto in the undeformed lamina;
also A'D' = dl and A'B' = dt in the deformed lamina. Then, the
following relations can be found:
2 E n = [(dl) 2 -(dl o ) 2 ]/(dl o ) 2
2E22 = [(dt) 2 -(dto) 2 ]/(dto) 2 (9.39)
2E12 = -sin(A0)V(l + 2 £ n ) V ( l + 2E22)
where
A(j) = LB'A'D' - LBAD = (f) - Ji/2 (9.40)

9.3.2 Constitutive equations for a composite lamina

9.3.2.1 Strain-energy function

Rivlin (1959) made the following remarks concerning the
strain-energy function: 'It was realized that the physical properties
of an elastic material can be characterized by a strain-energy
function and that this cannot depend on the nine displacement
gradients in a completely arbitrary fashion. And also if the material
has symmetry, the dependence of the strain-energy on these strain
components cannot be arbitrary either.' Following R. S. Rivlin
(private communications, 1986-7), the strain-energy density treated
here is assumed to be a function of the Lagrangian strain com-
ponents referring to the principal material coordinates. In the
two-dimensional case, referring to Eq. (9.39), the strain-energy per
unit volume is written as
W = W(En,E22,E12) (9.41)
Since W is unchanged by the following permutation, namely,
Xx—*—Xi and x1->—Jc 1

or
X2-+-X2 and x2-> -x2 (9.42)
488 Nonlinear elastic finite deformation

the strain-energy function must be an even function of El2. Then,

Eq. (9.41) is rewritten as
W = W(En,E22,E212) (9.43)
Finally, the strain-energy per unit volume of the undeformed
lamina is assumed in the following fourth-order polynomial form:
W = \CXXE\ + \CnxE\ + \CnilE\ + Cl2ExE2
~%C222E2 + -\C2222E2 faEf, + ^CfftfaEf,
(9.44)
where Cijy Cijky and Cijkt are elastic constants. The short-hand
notations are used, namely, Ex = En, E2 = E22y and E6 = 2El2.

9.3.2.2 General constitutive equations for a unidirectional lamina

The stress matrix referring to the material principal coordi-
nate system X1-X2, is given in terms of W (Rivlin 1970),

(9.45)
dW

Using Eqs. (9.38) and (9.43), it follows that

_ dW\
ljl
28ip\3Ejp +
dEpj)

r { Wngngn + W22gi2gj2 (9.46)

where

= dW
dW =
- CnEu + CinEn + Cmi£n + Cl2E2
3E
nEU
dW
dW
22 = - ^ = C22E22 + C222E\2 + C2222EL + C12£u (9.47)
ot22

dW
n = - ^ = 4 ( C 6 6 E 1 2 + 4C 6 6 6 6 £? 2 )
6E12
 Relations based on the Lagrangian description 489

To derive the general constitutive equations with reference axes

other than the principal material directions, a two-dimensional
rectangular Cartesian coordinate system Xx-X2 is chosen in the
plane of the lamina. The angle between Xt and Xt is 0o (Fig. 9.7).
Let [a] be an orthogonal transformation matrix,

rcos0 o -sin0ol
Lsin 6O cos 6O J
and [X] = [a] • [X]. Then, the transformation relations for the
deformation gradient and Lagrangian strain between coordinate
systems X and X are:

[£] = [a]T[g][a] (9.49)

and
[E] = [a]T[E][a] (9.50)
With Eq. (9.48), Eq. (9.50) yields
Eu = 5£n(l + cos 20O) + El2 sin 20O + \E 2 2 {\ - cos 20O)
E22 = \En{l - cos 20O) - E12 sin 20O + \E22(\ + cos 20O)
E12 = E2X = \{E22 - Eu) sin 20O + E12 cos 20O
(9.51)
The stress matrix referring to the coordinate system X is
[°]= [a][o][a]T
(9.52)

With Eqs. (9.46) and (9.49), Eqs. (9.52) are expressed as

On = -7-—gipgjq{ ap\aq\ Wn + ap2aq2W22

(9.53)
and
n>/ = gip{aplanWn + ap2aj2W22 + h(aPiaJ2 + ap2an)Wn}
(9.54)
Then, from Eq. (9.48), Eq. (9.54) is given in the following explicit
490 Nonlinear elastic finite deformation

form:

n n = [gnc2 + gi2cs]Wn + [gns2 ~ gi2cs\ W22

n + [g22c2 - g2lcs]W22

(9.55)
n 1 2 = [g22cs + g2lc2] Wu + [g2ls2 - g22cs] W22

= [gncs + gi2s2]Wn + [gl2c2 - gncs]W22

1 2
2gu(c -s2)]W12
where c = cos 6O and s = sin 0O.
Equations (9.55) are the general constitutive equations for a
composite lamina under finite deformation, where the deformation
gradients, giJf represent the geometric nonlinearity influenced by
the configuration changes of the lamina. The nonlinear expressions
of Wij (Eqs. (9.47)) represent the material nonlinearity of the
composites. If the deformation of the composite lamina is in-
finitesimal (i.e. gij = dij) and only the linear terms (i.e. C,7) remain
in the expression of Wijy Eqs. (9.55) can be easily reduced to the
familiar linear stress-strain equation used for rigid composites.
For a specific deformation, the deformation gradient matrix, [g],
is calculated from Eq. (9.14); the Lagrangian strain referring to the
principal material coordinates, [E], is obtained from Eqs. (9.15)
and (9.51); and Wtj are obtained by introducing [E] into Eqs. (9.47).
Then, the corresponding Lagrangian stresses, [II], can be deter-
mined from Eqs. (9.55). In the following Sections 9.3.2.3-9.3.2.5,
this procedure is illustrated by some specific examples.

9.3.2.3 Pure homogeneous deformation

Consider the rectangular lamina of Fig. 9.7; its edges are
parallel to the axes of the coordinate system X. The lamina is
subjected to a pure homogeneous deformation with principal
extension ratios Xx and A2 defined along the axes of the coordinate
system X. The deformation is described by
x
\ AI-AI
(9.56)
x2 — A2^V2
 Relations based on the Lagrangian description 491

Consequently, referring to Eqs. (9.14) and (9.15),

and

From Eqs. (9.51) and (9.58), the following can be obtained:

En = J[(A? + A l - 2 ) + (A?- X2)cos20O]
£22 = i[(Af + A| - 2) - (A? - Al) cos 20O] (9.59)

Then, the components of the Lagrangian stress are obtained from

Eqs. (9.55) and (9.57) as
n n = lx{c2Wxl + s2W22 - csW12)
n22 = X2(s2Wn + c2W22 + csW12)
(9.60)
n 1 2 = A2[cs(Wi! - W22) + \{c2 - 52)W12]

n 2 1 = ^[csiWu - W22) + Kc2 - s2)Wn]

where Wn, W22 and W12 are given by Eqs. (9.47) and (9.59).

Fig. 9.7. Pure homogeneous deformation. (After Luo and Chou 1990b.)

Undeformed
element

\ X(XUX2)

x(k[Xi,X2X2)
t,x2
Deformed element
492 Nonlinear elasticfinitedeformation

9.3.2.4 Simple shear

Suppose that the rectangular lamina, with its edges parallel
to the axes of the coordinate system X, is subjected to a simple
shear of amount K in the direction of the Xx axis (Fig. 9.8). Then,
the deformation is described by
KX2
(9.61)

For this deformation, referring to Eqs. (9.14), (9.15) and (9.51),

LU 1J LA A J

and
£12 = l[K2 + (2K sin 20O - K2 cos 20O)]
£22 = i[# 2 - (2K sin 20O - K2 cos 20O)] (9.63)
2
£12 = l[K sin 20O + 2/C cos 20O]
Then, the components of the Lagrangian stress are obtained from
Eqs. (9.55) and (9.62):
n n = [c2 + Kcs]Wu + [s2 - Kcs]W22
i 2
+ [-cs + 2K(c -s2)]Wl2
n 2 2 = s2Wn + c2W22 + csWl2 (9.64)
2 2
n , 2 = csWn - csW22 + x2{c - s )Wl2
n 2 1 = [cs + Ks2]Wn + [Kc2 - cs]W22 + [Kcs + \{c2 - s2)) Wl2

Fig. 9.8. Simple shear deformation. (After Luo and Chou 1990b.)

KX2,X2)

t,X2

+~ X,
Deformed element
 Relations based on the Lagrangian description 493

Considering the example of 6O = 45°,Eqs. (9.63) yield

(9.65)

Then Eqs. (9.64) give

n u = i{(i + K)wu - K)w22 - w12}

(9.66)

K)Wn - l)W22 + KWX2)

Figure 9.9 shows the theoretical prediction of the stress-strain

relation (Eqs. (9.64)) for Kevlar/silicone elastomer laminae with
various initial fiber orientations under simple shear deformation. The
elastic constants used in the analysis are shown in Table 9.1 (Luo
1988). The result shows that the simple shear properties of a
composite lamina under finite deformation are significantly influenced
by the fiber orientation. Figure 9.10 gives the comparison between
analytical predictions and experimental results of a 0° specimen
under simple shear. Figure 9.11 shows the same comparison on a

Fig. 9.9. Theoretical predictions of simple shear deformation of

Kevlar/silicone elastomer composite laminae for various initial fiber
orientations. (After Luo and Chou 1990b.)

OH
S

0.05 0.10 0.15 0.20 0.25

Shear deformation, K
494 Nonlinear elastic finite deformation

Table 9.1. Elastic constants of Kevlar-

491silicone flexible composites {After Luo,
1988).

Si, (MPa) 1 0.114 x 1 0 - '

C
(MPa) - ' 0
(MPa)1 -69.9 x10 ~ 6
S22 (MPa) 1 0.306
•$2222 (MPa) 3 0.563
566 (MPa) -1 0.387
"J6666 (MPa)"3 77.5 x 10"'
(MPa) -2 3.43 x 10"6
"^2266 (MPa)"3 56.3 x 1 0 '
cn (MPa) 8.6 x 10'
0
^1111 (MPa)
c,2 (MPa) -1.3
c22 (MPa) 2.77
£- 2222 (MPa) -12.5
c66 (MPa) 2.55
(-6666 (MPa) -2.45

Fig. 9.10. Comparisons between theoretical predictions and experimental

data of simple shear response of 0°Kevlar/silicone elastomer composite
laminae. (After Luo and Chou 1990b.)
1.0

0.8 -

O • Experiments
— Theory

0.0 0.1 0.2 0.3 0.4 0.5 0.6

Shear deformation, K
 Relations based on the Lagrangian description 495

90° specimen; the experimental data, which are obtained from

three-rail shear tests (Whitney, Daniel and Pipes 1982), are lower
than the predicted values at large shear deformation. This is caused
by the edge effects, and fiber pull-out from clamped edges for 90°
specimens at large deformation.

9.3.2.5 Simple shear superposed on simple extension

A rectangular lamina, with the edges parallel to the axes of
the rectangular Cartesian coordinate system X, is first subjected to
the pure homogeneous deformation described by Eqs. (9.56),
followed by a simple shear of magnitude K. There are two cases for
the direction of the shear deformation: (1) parallel to the Xl axis,
and (2) parallel to the X2 axis; both are discussed in the following:

Case 1
Figure 9.12(a) illustrates this deformation, which can be
specified as
X\ — A\X\ ~\~ K.A2X2
(9.67)
=
X2 A^A- 2

Fig. 9.11. Comparisons between theoretical predictions and experimental

data of simple shear response of 90° Kevlar/silicone elastomer composite
lamina. (After Luo and Chou 1990b.)
1.50 1
O• Experiments
Theory
1.25

OH 1.00

0.75

0.50

0.25

0.00
0.00 0.02 0.04 0.06 0.08
Shear deformation, K
496 Nonlinear elastic finite deformation

From Eqs. (9.14), (9.15) and (9.51),

:A,A 2 I
+ D-iJ
(9.68)
and
- 2 -A! - A | cos 2 0 O

£22 = i[(A ? + A| - 2 + A|A:2) - (A? - A| - k22K2) cos 20 O

- 2ArA,A2 sin 20O] (9.69)
2 2
£12 = i[(Al - A + A2,*: ) sin 2QO + IKXJ^ cos 2do]
The components of the Lagrangian stress are obtained from Eqs.

Fig. 9.12. Simple shear superposed on simple extension. (After Luo and
Chou 1990b.)

Undeformed . 1, X\
element f

X2 1 Deformed element x (A, X , , K A , Xx + A 2 X 2 )

t, X.
Undeformed
element

(b)
 Relations based on the Lagrangian description 497

(9.55) and (9.68):

+ Kk2cs]Wn + [k x s 2 - Kk2cs]W22
2 2
+ [-A x cs + \Kk2{c - s )]Wl2

n 2 2 = A2[s2Wn + c2W22 + csW12]

n 1 2 = k2[csWn - csW22 + \{c2 - s2)W12] (9.70)

II 21 = [Ajcs + KX2s2]Wn + [/CA2c2 - Xxcs]W

+ [KX2cs + Ui(c2 ~ s2)]W12

Case 2
Figure 9.12(b) shows the deformation defined by
x =
\ AXA i
(9.71)
x2 — KAiJC i 4" A2JC2

For this deformation, referring to Eqs. (9.14), (9.15) and (9.51),

(9.72)
and

£n = i[(Af + A| - 2 + A2AT2) + (A2 - A^ + A2/C2) COS 2 6 O

+ 2A:A!A2sin20o]

fi22 = 4 [(Ai + A| - 2 + A2AT2) - (A2 - Al + A? /C2) COS 2 0 O

0 0 ] (9.73)

= 4[(Ai - * ? + A2A:2) sin 20 O + 2 / a i A 2 cos 20O]

The components of the Lagrangian stress obtained from Eqs. (9.55)

and (9.72) are:

n 2 2 = [A252 + KXxcs]Wu + [A2C2 - Kk1cs]W22

2 2
+ [X2CS + kKX^c - s )]W12
(9.74)
]W22

n 2 1 = A,[csWu - csW22 + \{c2 - s2)Wl2]

498 Nonlinear elasticfinitedeformation

Figure 9.13 illustrates an off-axis specimen under uniaxial tension.

It is understood that the clamping of the specimen at the ends
induces a local non-uniform strain field. However, if the length-to-
width ratio of the specimen is sufficiently large, a uniform state of
stress and strain prevails at the center of the specimen (Pagano and
Halpin 1968), and the central lines of the specimen remain straight
in the Xx direction. Then, this deformation corresponds to Case 1,
namely, 'simple shear superposed on simple extension'. The uni-
axial loading condition can be described as n n ¥=0, n 22=0 and
n 21 =0. Then, from Eqs. (9.74),

u + s2W22 - csW12]

(9.75)
0 = csWn - csW22 + \{c2 - s2)W12

where Wtj are obtained by from Eqs. (9.68) and (9.47). The three
unknowns XXy k2 and K in Eqs. (9.75) can be solved from these
equations. Figure 9.14 shows the comparison between analytical
predictions and experimental results for the off-axis response of
Kevlar/silicone elastomer laminae under simple tension. The fiber
initial orientations are 10°, 30° and60°.The elastic constants used in
the calculation are shown in Table 9.1.

Fig. 9.13. Off-axis specimens of flexible composite laminae (a) without

loading, (b) with loading. The 15° one is tirecord/rubber, and the 10° and
30° ones are Kevlar/silicone elastomer. (After Luo and Chou 1988a.)

(u) (u)
 Relations based on the Lagrangian description 499

9.3.3 Constitutive equations of flexible composite laminates

9.3.3.1 Constitutive equations

The analytical methodology developed in Section 9.3.2 for
composite laminae is applied to study the constitutive relations of
laminated flexible composites (Fig. 9.15) under finite plane defor-
mation (Luo and Chou 1989). The stress resultant in Lagrangian
description (Ny) is defined as
rh/2
= ]UtJdz (9.76)
J-h/2

where h is the initial laminate thickness. Ntj so defined gives the

total force in the i direction per unit length of the undeformed
laminate.
Assume that the laminate is composed of n layers of unidirec-
tional laminae. By neglecting the interlaminar shear deformation,
the deformation gradient, gi} (Eq. (9.14)), has the same value for all
the layers; this is also true for Etj (Eq. (9.15)). For an arbitrary A:th
lamina within the laminate, let 8(ok) be the fiber orientation angle

Fig. 9.14. Comparisons between theoretical predictions and experimental

results on 10°, 30° and 60° off-axis stress-strain response of Kevlar/silicone
elastomer composite laminae.

o • Experiments do = 10°
— Theory 9 o

6-

4 •

2-
./ z X n,,

e0 = 30°

_B—9—8 60°

2 4 6 8 10

Strain, Ai — 1
500 Nonlinear elasticfinitedeformation

with respect to the coordinate Xlt and a\P the values given by Eq.
(9.48) for 6O = 6(ok). Also, for the kth lamina, let E\V be the values
of Eij given by Eqs. (9.51), and W\p the values of Wtj given by Eqs.
(9.47). Then, from Eqs. (9.54) and (9.76) the following can be
derived:

(9.77a)
If the laminae are identical in thickness, t> then

?a$+a (9.77b)
Equations (9.77) are the general constitutive equations for
flexible composite laminates under finite deformations. The ap-
plications of the constitutive relations are exemplified in the
following.

9.3.3.2 Homogeneous deformation

The homogeneous deformation of a composite laminate
(Fig. 9.16) is defined by
X\ —
(9.78)

Fig. 9.15. A composite laminate. (After Luo and Chou 1989.)

h/2 // X

h2

/ \

/ / n
/ / k
3
2
1
 Relations based on the Lagrangian description 501

where Xx and A2 are the extension ratios in the Xx and X2

directions, respectively. Thus, referring to Eqs. (9.14) and (9.15),
for all laminae:

t ]'
The Lagrangian strains referring to the principal material coordin-
ate for the kth lamina are obtained from Eqs. (9.51)
Eft = i(Aj cos2 0<*> + A| sin2 6™ - 1)
2
Eg A sin2 0<*> + A2 cos2 0<*> - 1) (9.80)
?
Eg = i ~ A ) sin 0(*> cos 0^>
Then, the components of the Lagrangian stress resultant are
obtained from Eq. (9.77b)

" cos 26(ok)) - W(£ sin 2 6

k=\

+ Wg(l + cos 2d(ok)) + W & sin 20 (9.81)

= | A2 2 {(W^ - W^) sin cos

sin 20<*> + Wg cos 20

k=l

Fig. 9.16. A symmetric flexible composite laminate under a uniaxial load.

(After Luo and Chou 1989.)
A
J, l

Undeformed laminate
Deformed laminate
\

A'l
502 Nonlinear elastic finite deformation

where W { \ \ W(£ and W& are obtained from Eqs. (9.47) and
(9.80) with two variables Ax and A2.

9.3.3.3 Simple extension of a symmetric composite laminate

(A) Tensile stress-strain relation

The state of homogeneous deformation is assumed for a
symmetric composite laminate with fiber orientation sequences of
+ 6o/—6J—d J + do under unidirectional tension. Because En and
£"22 a r e e v e n functions of 0O, and E12 is an odd function of 6O (Eqs.
(9.80)), Eqs. (9.47) become

W(22e) (9.82)
6)
-w\ -
2
Then, Eqs. (9.81) can be reduced to

+ Wg\l - cos 20O) - W[e2> sin 20o}

h (9-83)
N22 = \A2{w[V(i-cosieo)

+ H\1 + cos 20O) + W(tf > sin 26O}

M 2 = N21 = 0
where h is the thickness of the laminate; W\p, obtained from Eqs.
(9.47) and (9.80), is a function of Xl and A2. With the uniaxial
loading condition and the values of elastic constants, the two
unknowns, Xx and A2, in Eqs. (9.83) can be solved.
For example, let 0O = 45°, from Eqs. (9.47), (9.80) and (9.83),
NJh = AX{C66(A? - A22) + } C W A ? - Al)3}
N22/h = 0
= (D - 4C66)A? - (D + 4C 66 )A| (9.84)
2
+ i ( C m + C222)(A? + Ai-2)
+ l6(Cim + C'2222)(A1 + A2 — 2) + C6666(A2 — Aj)
where D = C n + 2C 12 + C
 Relations based on the Lagrangian description 503

Figure 9.17 shows the comparison between the theoretical predic-

tions and experimental results of the stress-strain relation of [±45°]s
Kevlar/silicone elastomer composite laminates under uniaxial load.
Reasonable agreement has been found.

(B) Effective Poisson's ratio

The Poisson's ratio is defined as the negative ratio of the
strain in the Xj direction to the strain in the Xt direction due to an
applied stress in the Xt direction. The Poisson's ratio of a symmetric
composite laminate was derived by Posfalvi (1977) based upon a
finite deformation consideration. Although experimental results of
large deformation were presented, the comparison of theory with
experiments was still limited to the small deformation range.
From the above analysis the effective Poisson's ratio in the finite
deformation range can be readily predicted. For example, for a
[+0 O /-0 O ] S laminate under unidirectional load, the effective
Poisson's ratio at a given strain level can be determined from Eqs.
(9.79) as

Al-1
(9.85)

Fig. 9.17. Comparisons between theoretical predictions and experimental

data of stress-strain response of a [±45°]s Kevlar/silicone elastomer
composite laminate under uniaxial load. (After Luo and Chou 1989.)

V) 1 -

0.0 0.1 0.2

Strain, Ai — 1
504 Nonlinear elastic finite deformation

where the relation between kx and A2 can be obtained from Eqs.

(9.83) with N22 = 0.
The approximate order of the ratio E22/ Eu can be obtained by
neglecting the non-linear terms (i.e. Cm, . . . , C 6666 , etc.) in the
expressions of Eqs. (9.47) for Wtj, Then
(9 86)
t
Ai— 1 A
-
where
A = Cn cos2 6O sin2 0O + C12(sin4 do + cos 4 0O)
+ C 22 cos2 6O sin2 6O — 4C 66 cos2 6O sin2 6O
B = Cn sin4 6O + 2C12 cos2 6O sin2 6O + C 22 cos4 6O
+ 4C 66 cos2 0 o sin2 0o
For example, for 6O = 45°, Eq. (9.86) yields
A = (Cn + 2C12 + C22)-4C66

Since the shear modulus C 66 for flexible composites is relatively

small, it can be assumed A/B ~ 1. Then, Eq. (9.85) becomes

^ = - 1 (9.88)
£11

Furthermore, if the flexible composite is very stiff in the fiber

direction (i.e. Cu»Cij y ij¥= 11) and 8o^0, Eq. (9.86) becomes

fs.Mzi __•!»£. (9.89)

En Af — 1 sin 6O
The results of Eq. (9.89) can also be derived by using the 'ideal
fiber reinforced material theory' (Adkins and Rivlin 1955).
Figure 9.18 gives the comparison between theoretical predictions
and experimental results of the ratio A2/Ax for [±0 o ] s
Kevlar/silicone elastomer composite laminates under uniaxial load.
The initial fiber orientations are 15°, 30° and 45°. Very good
agreement has been found.
Also, using the definition of Posfalvi (1977), the current Poisson's
ratio at a given strain level can be derived from Eq. (9.86) as
 Relations based on the Lagrangian description 505

Referring to Fig. 9.16, the current fiber orientation, 6 ( k \ of the

kth lamina, can be expressed in terms of k lf A2 and the initial fiber
orientation 0^' as

= A,
(9.91)
a cos 0 J ,'' A
A,

where kx and A2 are obtained by solving Eqs. (9.83).

9.3.4 Determination of elastic constants

In Section 9.3.2.1, the strain-energy per unit volume of an
undeformed lamina is assumed in a polynomial form (Eq. (9.44)).
The elastic constants in the strain-energy expression need to be
determined experimentally. Some experimental methods for char-
acterizing these constants are summarized below (Luo 1988).

9.3.4.1 Tensile properties

The constants Cn, C n l , Cnll, C22, C222, C2222, a n d C12
are associated with the tensile behavior of flexible composites and
are determined by unidirectional tensile tests. Consider a composite
lamina under a unidirectional load (i.e. I l n ^ O , n 2 2 = 0 and

Fig. 9.18. Comparisons between theoretical predictions and experimental

results of the ratio A2/Al of [±0 O ] S Kevlar/silicone elastomer composite
laminates under uniaxial load. (After Luo and Chou 1989.)

- 0 o = 15

Experiment — Theory
0.2 j [_ _|
1.0 1.0 1.2 1.0 1.0
A,
506 Nonlinear elastic finite deformation

n 1 2 = 0). For do = 0°, Eqs. (9.47) and (9.60) yield

A2 — 1 \ 3 /A 2 —
- I + C12I -

For 0O = 9O°, Eqs. (9.47) and (9.60) yield

/A2 2- -l\
l\3
C22221
2222\ - I T+LC
i2
2 ; v 2

2 / \ 2
]2 _ 1 \ 3 / i 2 _

n n , Ax and A2 are measured experimentally from both 6O = 0° and

90° unidirectional tensile tests. The constants C n , C 12 , and C 22 in
Eqs. (9.92) and (9.93) are related to the initial slope of these
experimental curves of n n / A , vs. (A2 — l)/2. The constants C m and
C222 are the nonlinear terms associated with the bi-modulus
properties of the composite; and the constants C U 1 1 and C2222 are
the fourth-order nonlinear terms in Eq. (9.44). Cm, C222, C i m ,
and C22222 are determined by fitting the theoretical curves of Eqs.
(9.92) and (9.93) to the longitudinal and transverse experimental
curves of Hu/ki vs. (A2 - l)/2, respectively.

9.3.4.2 Shear properties

C66 and C 6666 are the elastic constants associated with the
shear properties. Two test methods have been used for characteriz-
 Relations based on the Lagrangian description 507

ing the shear behavior: (1) three-rail 0°simple shear, and (2) simple
tension of [±45°]s specimen.
First, consider the simple shear test in which the applied shear
force is parallel to the fiber direction. From Eqs. (9.64), for 0 = 0,

= K(C22E2 + C222E2 + C2222E2) + C66E6 + C6666El

= K(C22 + 2C222K + lC2222K4) + C^K + C 6 6 6 6 K
(9.94)

Since the values of C22, C222 and C2222 are already known from
tensile property measurements, C66 and C g ^ can be determined by
fitting the experimental data of U2l vs. K.
Next, consider the tensile test using [±45°]s specimens. For a
[±45°]s laminate specimen under a tensile load, Eqs. (9.84) can be
rewritten as

Nn/hX, = CM. - A|) + iC6666(A? - A!)3 (9.95)

By measuring h, Nu, Xx and A2, the curve of Nn/hkx vs. (A? — Ai)
can be determined experimentally. Then a curve fitting method can
be used to identify the constants C66 and C6666. Experiments using
both simple shear and tensile tests on Kevlar-49/silicone elastomer
composites have yielded comparable results of the elastic constants
as shown in Table 9.1.
The tensile experiment on [±45] s specimens has been quite often
used to determine the shear modulus of conventional rigid polymer
composites (see ASTM Standard D 3518-76). The basic equation
for this experiment is

oxxl2 = Gl2(exx-eyy) (9.96)

where the engineering stress (oxx) and strains (exx and eyy) are
measured experimentally.
In order to compare Eq. (9.95) with Eq. (9.96) the following
relations are introduced:

Ax = 1 + exx

A2=l + eyy (9.97)

Nn =Uxx 1 =oxxA2
2hkx~ 2 Ax~ 2 Ax
508 Nonlinear elastic finite deformation

Substitution of Eqs. (9.97) into Eq. (9.95) gives

°~f= J~ [Cee[(exx — eyy) + (<& - <&)]

Z* J- ~i &yy

+ C6666[(exx - £yy) + (e2xx - E2yy)f (9.98a)

For linear elastic materials (i.e. C f ^ = 0) under small deformation,

and by neglecting the higher order terms of strain, Eq. (9.98a) can
be rewritten as

~ y ~ 7~7 [C66(£xX ~ Eyy)\ (9.98b)

Z* 1. ~l &yy

Since the initial shear modulus G12 = C 66 , the difference between

Eqs. (9.96) and (9.98b) is the geometric factor (1 + exx)/(l + £yy).
Obviously, for infinitesimal deformation (1 + exx)/(l + eyy) = 1 and
Eqs. (9.96) and (9.98b) are identical. However, if the deformation
is not infinitesimal, G12 determined from Eq. (9.96) may not be
accurate because of the omission of the geometric factor. For
instance, let the strain £ ^ = 0 . 0 2 and use the relation of Eq.
(9.98a); the error is (1 + exx)l{l + eyy) = 4.1%.
The elastic constants of Kevlar-49/silicone elastomer obtained
from the above methods are listed in Table 9.1. The higher order
elastic constants (i.e. Cm and Cuii) are determined by a regression
curve fitting to the experimental data. Thus, they are valid only
within the strain level at which they are obtained experimentally.

9.4 Constitutive relations based on the Eulerian description

In the above, the Lagrangian system has been used to
derive the closed form constitutive equations for flexible compos-
ites, based upon a strain-energy function, for both lamina and
laminate. These equations can be used to predict the nonlinear
elastic behavior of flexible composites under different cases of finite
deformation. It should be mentioned that the Lagrangian stress,
defined as force per undeformed area, is a nominal stress and the
real force equilibrium is established in the deformed or contem-
porary configuration. Furthermore, the anisotropic elastic pro-
perties of the composite always refer to the deformed configuration.
For example, the current Young's modulus describes the stiffness in
the current fiber direction which rotates during deformation.
Therefore, in some cases, it is convenient to use the deformed body
as the reference to describe the constitutive relation.
 Constitutive relations based on the Eulerian description 509

In this section, a nonlinear constitutive relation has also been

developed based upon the Eulerian description where the deformed
configuration of the composite is used as the reference state (Luo
and Chou 1988a). A stress-energy function, referring to the current
principal material coordinate x (Fig. 9.2b), provides the basis for
deriving the constitutive relations.

9.4.1 Stress -energy function

In finite elasticity, the energy densities in terms of either
the Eulerian or Lagrangian stresses are not unique referring to a
fixed coordinate; this can be demonstrated through the considera-
tion of a 'rigid-body rotation' (Fung 1969). As an example, consider
a bar which is subjected to a simple tension and rotating about the z
axis. At one instant, the bar is parallel to the x axis so that oxx i=0
and oyy = 0. At another instant, when the bar becomes parallel to
the y axis, the stress state is given by oxx = 0 and oyy =£0. Thus a
rigid-body rotation changes the stress tensor, even though the state
of stress in the bar remains unchanged. A complementary energy
function referring to a fixed coordinate may be defined based upon
the second Piola-Kirchhoff stress tensor PAB. However, as indi-
cated in Eq. (9.36), the second Piola-Kirchhoff stress still involves
the displacement gradient. Thus, the use of complementary energy
in terms of /^y does not really make the constitutive relation simpler.
In order to establish the stress-energy function, a moving
Eulerian coordinate system is introduced in this section. Figure
9.2(b) illustrates a unidirectional flexible composite lamina under a
finite deformation in the Eulerian system. Unlike Fig. 9.2(a), here
the deformed configuration has been chosen as the reference state,
and the rectangular element A'E'F'D' in the deformed body is
considered. The sides A'D' and A'E' coincide with the current
principal material coordinate system V-t' or xx-x2, with /' referring
to the current fiber direction. The underline of a quantity refers to
the current principal material coordinates xl-x2. Thus, the element
AEFD corresponds to the element A'E'F'D' in the undeformed
state. One may assume that the rectangle A'E'F'D' undergoes two
stages of deformation in restoring to its initial shape AEFD. These
stages are illustrated in Fig. 9.19. First, A'E'F'D' becomes a
smaller rectangle A"E"F"D" by removing the normal stresses; then
it reverses to AEFD by removing the shear stress.
The deformations depicted in Fig. 9.19 can be related to the
Eulerian strain components. Let the line elements AD = d/o and
AE = dto in the undeformed lamina (Fig. 9.19c); also define
510 Nonlinear elastic finite deformation

A'D' = d/ and A'E' = dt in the deformed lamina (Fig. 9.19a). Then,

the physical significance of the Eulerian strains can be explained as
2e n = [(dl) 2 -(dl o ) 2 ]/(dl) 2
2e22 = [(dt) 2 -(dto) 2 ]/(dt) 2 (9.99)
2e12 = siny 12(V(l - 2* n )V(l - 2e22))
where etj are the Eulerian strains referring to the current principal
material coordinates xx-x2y and y12 is the angular deviation from a
right-angle as shown in Fig. 9.19.
The stress-energy per unit area of the deformed lamina
(A'E'F'D') is assumed to be a function of the Eulerian stress
components referring to the current principal material coordinate

Fig. 9.19. Illustration of the deformation of a rectangular element in the

Eulerian system. (After Luo and Chou 1988a.)

V -+ E'

26
dt
Si 21

A' A'

26
E" E"

26
26 (b)

A" D"

(c)

A
d/ 0 D
 Constitutive relations based on the Eulerian description 511

Xy-x2> namely W* = W*(gu, o22, o\2). The following expression is

adopted:

• S2266gjg26 (9.100)

where gt are the Eulerian stresses referring to the current principal

material coordinates xl-x2. Also, the short-handed notations are
used, i.e. gl = on, o2 = o22 and g6 = gl2. Sijy Sijk and Sijkl are the
compliance constants. Equation (9.100) is similar to the expressions
of Hahn and Tsai (1973) in their mathematical forms. However, due
to the finite deformation, it should be mentioned that: (1) The
Eulerian coordinate x used here is a moving coordinate, which is
chosen to coincide with the current fiber longitudinal and transverse
directions, /' and t''. Therefore, the energy function satisfies the test
of rigid-body rotation. (2) The Eulerian stresses, giJ} used in the
energy function are the current stress state of the deformed lamina.

9.4.2 General constitutive equations

The complementary energy per unit volume of a deformed
a lamina is W* = g^e^ — W(f/7). Here, W(e,7) = g^Se^ is the strain-
energy density. Then,

dW
<5W* =CT,,(5e,,+ eHdOjj 6eh (9.101)
' ' ' ' 3Sij '
Since

dW
(9.102)
"" Be,'

the following can be obtained from Eq. (9.101):

dW* = eijdgij (9.103)

or

dw*
Si,=—— (9.104)
512 Nonlinear elasticfinitedeformation

Substituting Eq. (9.100) into Eq. (9.104), the Eulerian strain

components referring to the coordinates xx-x2 are obtained as

£i = SnQi + S\nQ\ + Simgi + Sl2o2 + SX66gi

e2 = S22g2 + S222g22 + S2222g\ + ^12^1 + 2S2266g2gl (9.105)

S(,= S(,(,g(, + •J6666£J6 + 251g6cx1or6 + 2S226(,g2g6

Here ex — en, e2 = e22y e6 = 2eX2. The choice of compliance con-

stants in Eq. (9.100) is made on the following basis. First, S n , 522,
S12 and 566 are associated with the linear deformation. Second, the
terms Sn and 5222 are adopted for representing the bi-modulus
behavior in the axial and transverse directions, respectively. Third,
the nonlinear deformations are represented by 51 1U , 52222 a n d 56666.
Lastly, the greatest uncertainty involves the coupling terms between
the normal and shear deformations. Unlike in rigid composites, the
coupling effects may not be negligible in flexible composites. Two
terms, 5166 and 52266, are retained to represent the interactions
between axial and shear deformations in Eqs. (9.105).
Having established the constitutive relations with respect to the
principal material coordinates xx-x2y the general constitutive rela-
tions of a composite lamina referring to the fixed coordinates xx-x2
(Fig. 9.2b) can be derived from Eq. (9.105) and the tensor
transformation relation,

[e] = [r]T[S][r][(j] = [S*][a] (9.106)

where

•S,,
S12 S22 + S222g2 + S2222g2
2S166CT6 2S 2 2 6 6 g 2 CT 6

c2 s2 2cs
2 2
{a} = \o2\, and [T] = s c —2cs
2
—cs cs c — s

Here, c = cos 6 and s = sin 6, where 6 denotes the current fiber

orientation angle. Also, et and OT are, respectively, the Eulerian
stress and strain referring to the coordinates xx-x2. The full
 Constitutive relations based on the Eulerian description 513

expression of [S*] in Eq. (9.106) is

[S*] =
_S 6 *i S62
c2s2Su + (c 4 + s4)Sl2
+s%2 - +c2s2S22 -

i r2 2o _ . 2 2 r
T t A O22 C A ^66 —c s S csS66
c3sSn — cs(c2 — s2)Sl2 -s2)Sl 2 cs(c2 — s2)S12
-cs3S22-\cs{c2-s2)S 3c _1_I ^2 2\c
C SS22 + 2CS\C S )^6fy
2c3sSu-2cs(c2-s2)Sl2
-2cs3S22-cs(c2-s2)S('66
lcs(c2 - s2)Sl2
-cs(c2-s2)See $,6
2c2s2Su - 4c2s2Sl2
+2c2s2S22 + \{c2 - s2)Sb6 -
c2s2 2c3s
2 2
cs c2s2 2cs3
c 3s cs 2cV_
c2s2 -2cs3~
+ 5222^1) c2s2 c2s2 -2c3s
—cs -c3s -c33ss2
2 2
cs -c2s2 —cs(c 2 — s2)
-cs2 —c s cs(c2-s2)
_ - l2cs{c2 - s2) x2cs{c2-s2) h{c2-s2)
-3c3s c3s - 2cs3 c4 - 5cV
3 3
+ Stf^Of, 2c 5 — 2cs 3cs3 -s4 + 5c252
c4 - 2c2s2 -s4 + 2c2s2 3c3s - 3cs3
-Acs3 2cs3-2c3s -2s4 + (
+ S2266o2O(, 2cs3 — 2c3s 4c3s 4
2c — 6c 5 2 2

4
_ -s + 3cV c - 3cV
4
Acs3 - 4c3s
(9.107)
The stresses in the current principal material directions, g h are
also obtained as
(9.108)
514 Nonlinear elasticfinitedeformation

Referring to Fig. 9.2(b) the current fiber orientation angle is

9 = 0o + A6 (9.109)
The fiber reorientation angle (AS) due to finite deformation can be
determined as follows. First, the angles DAD' and EAE' are
defined as a and /?, respectively. Then,
(9.110)
Here, the symmetric part of A6, (a + /3)/2, equals y12/2; the
antisymmetric part of A0, (a — p)/2, is defined as co. It is
understood that co is the rigid-body rotation, which is independent
of the coordinate system but dependent on the boundary condi-
tions. If co can be expressed in terms of the strain tensors, from
Eqs. (9.99) and (9.110)
2e12
A0=-sin" 1

Introducing Eq. (9.111) into Eq. (9.109), the current fiber orienta-
tion angle 6 is expressed as a function in terms of the strain tensor.
Then the general constitutive relations can be completely deter-
mined from Eqs. (9.106) and (9.111). The following are two
illustrative examples.

9.4.3 Pure homogeneous deformation

The pure homogeneous deformation, with principal exten-
sion ratios At and A2 defined along the axes of the fixed coordinate
system X, is shown in Fig. 9.7 and described by Eqs. (9.56).
Referring to Eqs. (9.14) and (9.20),

"1 0"
r[ g I] = fAi
|_0
0"l
A 2 J'
x; (9.112)
o1
and

0
2[e] = (9.113)
0
 Constitutive relations based on the Eulerian description 515

Eq. (9.113) gives

1

(9.114)

- 2e22)

Referring to Fig. 9.7 the current fiber orientation can also be

written as

d = tan = tan" 1 -f- (9.115)

UiCosflJ
Substituting Eqs. (9.114) into Eq. (9.115),

(9.116)

The substitution of Eq. (9.116) into Eq. (9.106) results in three

independent equations. It is known from Eq. (9.113) that e12 = 0.
Thus, by giving any two values of the following five variables in Eq.
(9.106): stresses ou o2y o6> and strains elf e2 (or Ax and A2), the
remaining three can be solved.
Finally, it is worth noting that from Eqs. (9.35) and (9.112), the
Lagrangian stresses can be written in terms of the Eulerian stresses
as

(9.117)
n 1 2 = A 2 a 12

9.4.4 Simple shear superposed on simple extension

The deformation of 'simple shear superposed on simple
extension' (Case 1) is shown in Fig. 9.12(a) and expressed by Eqs.
(9.67). Using Eqs. (9.14) and (9.20), it can be found
"l K~

" [o *£]• Ai

0
1
(9.118)
516 Nonlinear elastic finite deformation

and
A:
A2
2[e] = (9.119)
K
A! Lu,/ \A2/J_
Invert the above equation to obtain

-2en)
A9 = (9.120)
-2eu)(l-2e22)~'

1 - 2eu
Also, referring to Fig. 9.12(a), the current fiber orientation can be
expressed as
A2 tan 8O
0 = tan- 1 f-'l=tan-
I =ian
1
\-.— (9.121)
A, tan 6O
The substitution of Eq. (9.121) into Eq. (9.106), results in three
independent equations. If the values are known for any three of the
following six variables: stresses alt a2y o6, and strains ely e2> e 6 (or
Ax, A2 and K), the remaining three can be solved.
As mentioned in Section 9.3.2.5, for an off-axis specimen
under uniaxial loading, with a length/width ratio » 1 , the central
lines of the middle section of the specimen remain straight in the
loading direction. Then, this deformation can be referred to as
'simple shear superposed on simple extension (Case 1)'. Using the
uniaxial loading conditions, au =£0, a 22 = o 2 \ = 0, and Eqs. (9.106),
(9.120) and (9.121), the deformation parameters k x , A2, and K (or
etj) can be solved.
Figure 9.20 shows the comparison between analytical predictions
and experimental results for the off-axis responses of Kevlar/
elastomer composites under simple tension based upon the
Eulerian approach. The fiber initial orientations are 10°, 30° and
60°. The same comparisons for tirecord/rubber composite speci-
mens are shown in Fig. 9.21. The fiber initial orientations are 15°,
30° and 60° in this case. The predicted results are based upon Eqs.
(9.106), (9.120) and (9.121) and an iterative calculation method.
The elastic constants are shown in Table 9.1.
 Constitutive relations based on the Eulerian description 517

9.4.5 Determination of elastic compliance constants

The compliance constants in Eq. (9.100) can be determined
experimentally (Luo and Chou 1988a). The second-order constants
(Sn, S22, S u and S66) are based on the linear behavior. The other
constants are obtained by fitting the theoretical curves to ex-
perimental data. For example, for unidirectional tensile test in the
Xi direction (i.e. ox ¥= 0 and g2= g6 = 0), Eq. (9.105) becomes

Snig\ + Snng\ (9.122)

where the underline denotes the current principal material coordin-

ate. Then, Sn is obtained from the initial slope of the experimental
Qy-ex curve (i.e. Sn = I/Young's modulus). S m (which reflects
bi-modulus behavior) and 51U1 are determined by fitting the
theoretical curves to experimental data in both tension and com-
pression. With the unidirectional load applied in the x2 direction,
S22> 5222 and 52222> can be determined by the same procedures as in
the X1 direction.

Fig. 9.20. Comparisons between theoretical predictions and experimental

results of 10°, 30° and 60° off-axis stress-strain response
of Kevlar/silicone
elastomer composite laminae (Eulerian description). (After Luo and Chou
1988a.)

o.
S

0.0 2.5 5.0 7.5 0.0

Strain, e„(%)
518 Nonlinear elastic finite deformation

The remaining constants are related to shear (S 66 and 56666) and

stretching-shear coupling (S166 and 52266). If S166 and S22(y6 are
negligibly small, the shear constants can be determined experimen-
tally as described in Section 9.3.4.2, with proper stress and strain
transformations from the Lagrangian system into the Eulerian
system as described in Section 9.2.2.
The shear constants including the stretching-shear coupling listed
in Table 9.1 are obtained by off-axis tensile tests at various fiber
off-axis angles. For the unidirectional tensile condition (oll¥ z0,
o22 = a 12 = 0), Eq. (9.106) can be rewritten as
[c4Sn + 2c2s2S12 + s4S22]on/(cs)-en/(cs)
= S66g6 + S6666gl " (3c/s)S166oi + (*s lc)S2266gl (9.123)
where c = cos# and s = sin 6, g6 = csou. The fiber orientation
angle, 6, and the stress-strain relations are measured experimen-
tally. In Eq. (9.123), there are four unknown constants, 5 66 , 5 6666 ,
5 166 and 52266- The relations between O 6 and the values of Eq.
(9.123) which are determined by experimental measurements of ou,

Fig. 9.21. Comparisons between theoretical predictions and experimental

results of 15°, 30° and 60° off-axis stress-strain response of tirecord/rubber
composite laminae (Eulerian description). (After Luo and Chou 1988a.)

2.5 5.0 7.5 10.0

Strain, exx (%)
 Flexible composites reinforced with wavy fibers 519

e n , 0 and the elastic constants related to the tensile properties (Sn,

Sl2 and S^)- 566 is t h e initial slope of the stress-strain curve. Given
sufficient experimental data (the number of specimens with different
initial fiber orientations should be larger than the number of
unknown constants), the remaining compliance constants S6666, S166
and S2266 can be determined by a regression technique to fit the
theoretical curve of Eq. (9.123) to the experimental curves.

9.5 Elastic behavior of flexible composites reinforced with

wavy fibers

9.5.1 Introduction
In Chapter 8, an iso-phase model for flexible composites
containing sinusoidally shaped fibers (Fig. 8.13) is presented and the
analysis of the elastic behavior of such composites is based upon a
step-wise incremental technique and the classical lamination theory.
Being a well established analytical technique, the lamination theory
does provide a convenient tool for describing the basic characteris-
tics of flexible composites. However, because of the use of super-
position techniques for nonlinear finite deformation problems, the
limitation of incremental analysis is obvious. In an effort to provide
a rigorous treatment, Luo and Chou (1988a&b, 1990b) applied the
constitutive models based upon the Lagrangian (Section 9.3) and
Eulerian descriptions (Section 9.4) to study the nonlinear elastic
behavior of flexible composites with wavy fibers under finite
deformation.
The deformation of the iso-phase flexible composite (see Fig.
8.13) is best understood by examining a representative element
which contains a full wavelength of the sinusoidal curve (Fig. 9.22).
This element is further divided into sub-elements along the x1 axis.
Each sub-element of the composite between xx and XX + AXX is
approximated by an off-axis unidirectional fiber composite, in which
fibers are inclined at an angle 6^ to the xx axis. Referring to Eq.
(8.14), the initial fiber orientation of sub-element (n), for example,
is given as

, lina
n —— cos

+ tan~' — cos — ^ (9.124)

\ A A /J
520 Nonlinear elasticfinitedeformation

It is also assumed that the stress and strain of a sub-element are

homogeneous under axial loading. This assumption is supported by
the photoelastic analysis (Luo and Chou 1988a). Figure 9.23 is a
photoelastic view of a flexible composite sample under longitudinal
loading; it shows that relatively uniform strain is maintained in
distinct regions along the longitudinal direction. It should be noted
that although all the experimental data collected are based upon a
Kevlar-49/silicone elastomer system, the photograph shown in Fig.
9.23 is based upon graphite fiber as a reinforcement materials, so
better contrast between the fiber and matrix in the photograph can
be achieved.
Based upon the above assumptions, the analysis for the iso-phase
model consists of two steps: (1) The constitutive relation of an
off-axis sub-element under finite deformation is examined based
upon the analysis developed in Sections 9.3 and 9.4. (2) The total
deformation of the composite is the summation of the deformations
of all these sub-elements.

9.5.2 Longitudinal elastic behavior based on the Lagrangian

approach
Under the uniaxial tensile force Fx in the xx direction, the
following plane stress condition of the flexible composite is

Fig. 9.22. Deformation of a sub-element of a flexible composite containing

sinusoidally shaped fibers under longitudinal tension. (After Luo and Chou
1990b.)

Deformed
sub-element

Undeformed
sub-element
 Flexible composites reinforced with wavy fibers 521

assumed:

(9.125)

=o
where Ao is the initial cross-sectional area perpendicular to the x1
axis.
Figure 9.22 shows the deformation of a typical sub-element
PQQ'P' pqq'p' represents the configuration in the deformed state.
Due to the iso-phase fiber arrangement, the edge qq' remains
perpendicular to the Xx axis. Let x\n) be the coordinates of an
arbitrary particle in PQQ'P1', and x\ n) be the corresponding
local coordinates of this particle in pqq'p'. Here, the superscript
refers to the sub-element \n). This deformation is specified as

(9.126)

Equations (9.126) specify a deformation equivalent to Case 2 of

Section 9.3.2.5, namely 'simple shear superposed on simple exten-
sion'. Using Eqs. (9.74) and the stress boundary condition of Eqs.

Fig. 9.23. A photoelastic view of a flexible composite lamina under

longitudinal tension. (After Luo and Chou 1988a.)
522 Nonlinear elastic finite deformation

(9.125), the following can be obtained:

0 = n 2 2 = s2Wxl + c2W22 + csW12 (9.127)

2 2
0 = n 2 1 = csWn - csW22 + \{c - s )W12
Here, Wtj is a function of K(n\ k[n) and A£°, and it is given by Eqs.
(9.47) and (9.73). The initial fiber orientation 0(on) is given by Eq.
(9.124). Then, the three unknowns, K(n), A(/° and A2n) can be solved
from Eqs. (9.127). It is interesting to note that H 12 does not vanish,
and it can be found from Eqs. (9.74)
n12 = K ™ n n (9.128)
The current fiber orientation, 0(n), of the sub-element (n) (Fig.
9.22) is

6(n) = tan"1[A™ + ^ tan 0gl)l (9.129)

The average extension ratio of the wavelength in the Xx direction

can be derived as

l^y^i"1 (9-130)

9.5.3 Longitudinal elastic behavior based on the Eulerian

approach
Under the uniaxial tension force Fx in the longitudinal
direction, the following stress states in the Eulerian system are
assumed:
oxl = FxIAy O22=OL2 = 0 (9.131)
where A is the area of the section perpendicular to the longitudinal
direction in the deformed state. From Eqs. (9.126) the deformation
of the sub-element (n) can be written in the Eulerian system as
1

(9.132)
 Flexible composites reinforced with wavy fibers 523

Using Eqs. (9.18) and (9.20), it can be found

1
0
1
[g]" = K(")
(9.133)

and

2[e] =
MGF)'-©! (9.134)
i \2

Invert the above equation to obtain

(9.135)

where A(xrt) and A^n) are the extension ratios of the sub-element (n) in
the longitudinal and transverse directions, respectively; and K(n) =
tan O (n) as shown in Fig. 9.22. The current fiber orientation
is obtained from Eq. (9.129). Then, X[n\ X(2n), and K{n) can
be determined by an iterative calculation from Eqs. (9.106),
(9.131) and (9.134). Also, the average extension of a wave-
length in the longitudinal direction, Xlf can be determined by Eq.
(9.130).
The predictions of the longitudinal constitutive relations based
upon the Lagrangian and Eulerian approaches are compared to
experimental results and an incremental analysis in the finite
deformation range (Luo, Kuo and Chou 1988). The model compos-
ite system consists of silicone elastomer reinforced with sinusoidally
shaped Kevlar fibers (a/A = 0.09). Due to fiber waviness, the
volume fraction may vary among the sub-elements. An average
524 Nonlinear elastic finite deformation

fiber volume fraction Vf = 9% is used in the calculation. Also, the

elastic constants are given in Table 9.1. Figure 9.24 compares the
analytical predictions with experimental results. The heavy solid
line indicates theoretical predictions of the Lagrangian approach
(Luo and Chou 1988b, 1990b); the thin solid line indicates
theoretical predictions based upon the Eulerian approach (Luo and
Chou 1988a, 1990a); and the dotted line is from the incremental
analysis (Kuo, Takahashi and Chou 1988). Experimental results are
also presented.
Furthermore, the local strains in the sub-element can be pre-
dicted directly from Eqs. (9.127). The current fiber orientation
angle of the sub-element is given by Eq. (9.129). These results show
that the maximum local tensile strain of the fiber occurs at the

Fig. 9.24. Stress-strain relations of Kevlar/silicone elastomer composite

laminae containing sinusoidally shaped fibers for a/A = 0.09. (After Luo
and Chou 1990b.)

O • Experiments
— Lagrangian description
40 — Eulerian description
— - Incremental analysis

30

••& 2 0

10

3 6 9 12

Strain, Xt - 1 (%)
 Flexible composites reinforced with wavy fibers 525

region where the initial fiber orientation angle equals zero (i.e.
Xx = ±A/4). The maximum local shear strain of the composites
occurs in the region where the initial fiber orientation is a maximum
(i.e. Xx = 0, A/2). Hence, the strength of the flexible composites
may be determined by the maximum tensile strain at Xx = XIA and
the maximum shear strain at Xl = 0.
 References

Chapter 1
Ashby, M. F. (1987) 'Technology on the 1990s: advanced material and predictive
design', Phil. Trans. R. Soc. London, A322, 393-407.
Bunsell, A. R. and Harris, B. (1974) 'Hybrid carbon and glass fiber composites',
Composites, 5, 157-64.
Chou, T. W. (1989) 'Flexible composites', J. Mat. Sci., 2A, 761-83.
Chou, T. W. and Kelly, A. (1976) 'What we do not know about fiber composites',
Mat. Sci. Engr., 25,35.
Chou, T. W., Kelly, A. and Okura, A. (1985) 'Fiber reinforced metal matrix
composites', Composites, 16, 177.
Chou, T. W., McCullough, R. L. and Pipes, R. B. (1986) 'Composites', Sci. Am.,
255, 192-203.
Clark, J. P. and Flemings, M. C. (1986) 'Advanced materials and the economy', Sci.
Am., 255, 50-7.
Compton, W. D. and Gjostein, N. A. (1986) 'Materials for ground transportation',
Sci. Am., 255,92-100.
Congress of the United States, Office of Technology Assessment (1988) New
Structural Materials Technologies, OTA-E-352, Washington, D.C.
Gordon, J. E. (1988) The Science of Structures and Materials, Scientific American
Library, New York.
Humphrey, J. D. and Yin, F. C. P. (1987) 'A new constitutive formulation for
characterizing the mechanical behavior of soft tissues', Biophys. J., 52, 563-70.
Kelly, A. (1985) 'Composites in context', Comp. Sci. Tech., 23, 171-200.
Kelly, A. (1987a) 'An outline of trends in materials science and processing', Mat.
Sci. Engr., 85, 1-13.
Kelly, A. (1987b) 'Composites for the 1990's', Phil. Trans. R. Soc. London, A322,
409-423.
McCullough, R. L. (1985) 'Generalized combining rules for predicting transport
properties of composite materials', Comp. Sci. Tech., 22, 3-21.
Mignery, L. A., Tan, T. M. and Sun, C. T. (1985) The use of stitching to suppress
delamination in laminated composites, ASTM STP876, American Society for
Testing and Materials, Philadelphia, PA, pp. 371-85.
Mody, P. B. and Majidi, A. P. (1987) 'Metal and ceramic matrix composites: the
heat is on', Composites in Manufacturing, 3, 1-5.
Nardone, V. C. and Prewo, K. M. (1988) 'Tensile performance of carbon-reinforced
glass', /. Mat. Sci., 23, 168-80.
Port, O., King, R. W. and Hawkins, C. (1988) 'Materials that think for themselves',
Business Week, December 5, pp. 166-7.
Rogers, C. A., ed. (1988) Smart Materials, Structures, and Mathematical Issues,
Technomic Pub. Co., Lancaster.
Sousa, L. J. (1988) Problems and Opportunities in Metals and Materials: An
Integrated Perspective, US Department of the Interior, Washington, D.C.
 References 527

Steinberg, M. A. (1986) 'Materials for aerospace', Sci. Am., 255, 66-72.

Sun, C. T. (1989) 'Intelligent tailoring of composite laminates', Carbon, 27, 679-87.
Takagi, T. (1989) 'A concept of intelligent materials in Japan', in Proceedings of
International Workshop on Intelligent Materials, The Society of Non-Traditional
Technology, Tokyo, Japan, pp. 1-10.
Vinson, J. R. and Chou, T. W. (1975) Composite Materials and Their Use in
Structures, Elsevier-Applied Science, London.

Chapter 2
Akoz, A. Y. and Tauchert, T. R. (1972) 'Thermal stresses in an orthotropic elastic
semispace', /. Appl. Mech., 39, 87-90.
Akoz, A. Y. and Tauchert, T. R. (1978) 'Thermoelastic analysis of a finite
orthotropic slab'. /. Mech. Eng. Sci., 20, 65-71.
Ashton, J. E., Halpin, J. C. and Petit, P. H. (1969) Primer on Composite Materials:
Analysis, Technomic, Westport, Connecticut.
Carlsson, L. A. and Pipes, R. B. (1987) Experimental Characterization of Advanced
Composite Materials, Prentice-Hall, Englewood Cliffs, New Jersey.
Carslaw, H. S. and Jaeger, J. C. (1959) Conduction of Heat in Solids, Clarendon
Press, Oxford.
Chamis, C. C. (1983) NASA Tech. Memo 83320 (presented at the 38th Annual
Conference of the Society of Plastics Industry (SPI), Houston, TX, Feb. 1983).
Chang, Y. P. (1977) 'Analytical solution for heat conduction of anisotropic media in
infinite, semi-infinite and two-plane bounded regions', Int. J. Heat and Mass
Transfer, 20, 1019.
Chawla, K. K. (1987) Composite Materials Science and Engineering, Springer-
Verlag, New York.
Cheng, C. M. (1951) 'Resistance to thermal shock', /. Am. Rocket Soc, 21, 147-53.
Chou, T. W. (1989a) 'Flexible composites', /. Mat. Sci., 24, 761-83.
Chou, T. W. (1989b) 'Elastic properties of laminates', Concise Encyclopedia of
Composite Materials, Pergamon Press, Oxford, p. 159.
Christensen, R. M. (1979) Mechanics of Composite Materials, Wiley-Interscience,
New York.
Chu, H. S., Weng, C. I. and Chen, C. K. (1983) 'Transient response of a composite
straight fin', ASMEJ. Heat Transfer, 105, 307-11.
Fukunaga, H. and Chou, T. W. (1988a) 'On laminate configurations for
simultaneous failure', /. Comp. Mat., 22, 271.
Fukunaga, H. and Chou, T. W. (1988b) 'Simplified design techniques for laminated
cylindrical pressure vessels under stiffness and strength constraints', /. Comp.
Mat., 22, 1156-69.
Halpin, J. C. (1984) Primer on Composite Materials Analysis, Technomic Pub. Co.,
Lancaster, Pennsylvania.
Halpin, J. C. and Tsai, S. W. (1967) Environmental Factors in Composite Materials
Design, Air Force Materials Laboratory Technical Report 67-423.
Hsu, P. W. and Herakovich, C. T. (1977) 'A perturbation solution for interlaminar
stresses in bidirectional laminates', Composite Materials Testing and Design (4th
Conference), ASTM STP 617, American Society for Testing and Materials,
Philadelphia, pp. 296-316.
Hsu, P. W. and Herakovich, C. T. (1977) 'Edge effects in angle-ply composite
laminates',/. Comp. Mat., 11,422-8.
Huang, S. C. and Chang, Y. P. (1980) 'Heat conduction in unsteady, periodic, and
steady states in laminated composites', ASME J. Heat Transfer, 102, 742-8.
528 References

Jones, R. M. (1975) Mechanics of Composite Materials, McGraw-Hill, New York.

Jost, W. (1960) Diffusion, Academic Press, New York.
Katayama, K., Saito, A. and Kobayashi, N. (1974) 'Transient heat conduction in
anisotropic solids', Proceedings of the International Conference on Heat and Mass
Transfer, Tokyo, p. 137.
Kingery, W. D. (1955) 'Factors affecting thermal stress resistance of ceramic
materials', J. Am. Ceramic. Soc, 38, 3-5.
Lo, K. H., Christensen, R. M. and Wu, E. M. (1977a) 'A higher-order theory of
plate deformation. Part 1: Homogeneous plates', J. Appl. Mech., 44, 663-8.
Lo, K. H., Christensen, R. M. and Wu, E. M. (1977b) 'A higher-order theory of
plate deformation. Part 2: Laminated plates', /. Appl. Mech., 44, 669-76.
Luo, J. and Sun, C. T. (1989) 'Global-local methods for thermoelastic stress analysis
of thick fiber-wound cylinders', Proceedings of the Fourth Technical Conference on
Composite Materials, American Society for Composites, Technomic Pub. Co.,
Lancaster, Pennsylvania, pp. 535-44.
Mossakowska, Z. and Nowacki, W. (1958) 'Thermal stresses in transversely isotropic
bodies', Archiv. Mech. Stosow, 10, (4), 569-603.
Noda, N. (1983) 'Transient thermal stress problem in a transversely isotropic finite
circular cylinder under three-dimensional temperature field', J. Thermal Stresses,
6, 57-71.
Nomura, S. and Chou, T. W. (1986) 'Heat conduction in composite materials due to
oscillating temperature field', Int. J. Engr. Sci., 24, 643.
Ozisik, M. N. (1980) Heat Conduction, John Wiley and Sons, Inc., New York.
Pipes, R. B., Vinson, J. R. and Chou, T. W. (1976) On the hygrothermal response
of laminated composite systems. /. Comp. Mat., 10, 129-48.
Poon, K. C. and Chang, Y. P. (1978) 'Transformation of heat conduction problems
from anisotropic to isotropic', Heat and Mass Transfer, 5, 215.
Reddy, J. N. (1984) 'A simple higher-order theory for laminated composite plates',
J. Appl. Mech., 51, 745-52.
Reissner, E. (1945) 'Transverse shear deformation on the bending of elastic plates',
J. Appl. Mech., 2, (2), 69-77.
Rosen, B. W. (1973). 'Stiffness of fibre composite materials', Composites, 4, 16-25.
Sharma, B. (1958). 'Thermal stresses in transversely isotropic semi-infinite elastic
solid',/. Appl. Mech., 25, 86-8.
Singh, A. (1960) 'Axisymmetric thermal stresses in transversely isotropic bodies',
Archiv. Mech. Stosow., 39, (3), 287-304.
Stein, M. and Jegley, D. C. (1987) 'Effects of transverse shearing on the cylindrical
bending, vibration, and buckling of laminated plates', AIAA J., 25, (1), 123-9.
Sugano, Y. (1979) 'Transient thermal stress in a transversely isotropic finite circular
cylinder due to an arbitrary internal heat-generation', Int. J. Engr. Sci., 17,
729-39.
Sun, C. T. and Li, S. (1988) 'Three-dimensional effective elastic constants for thick
laminates', /. Comp. Mat., 22, 629-39.
Takahashi, K. and Chou, T. W. (1988) 'Transverse elastic moduli of unidirectional
fiber composites with interfacial debonding', Met. Trans. AIME, 19A, 129.
Takeuti, Y. and Noda, N. (1978). lA general treatise on the three-dimensional
thermoelasticity of curvilinear aeolotropic solids', /. Thermal Stresses, 1, 25-39.
Tauchert, T. R. and Akoz, A. Y. (1974) 'Thermal stresses in an orthotropic elastic
slab due to prescribed surface temperatures', J. Appl. Mech., 41, 222-8.
Tsai, S.-W. and Hahn, H. T. (1980) Introduction to Composite Materials,
Technomic, Westport, Connecticut.
 References 529

Van Dyke, M. (1975) Perturbation Methods in Fluid Mechanics, The Parabolic Press,
Stanford, California.
Vinson, J. R. and Chou, T. W. (1975) Composite Materials and Their Use in
Structures, Elsevier-Applied Science, London.
Wang, H. S. and Chou, T. W. (1985) 'Transient thermal stress analysis of a
rectangular orthotropic slab', /. Comp. Mat., 19,424-42.
Wang, H. S. and Chou, T. W. (1986) 'Transient thermal behavior of a thermally and
elastically orthotropic medium', AlAA J., 2Ay 664-72.
Wang, H. S., Pipes, R. B. and Chou, T. W. (1986) 'Thermal transient stresses due to
rapid cooling in thermally and elastically orthotropic medium', Met. Trans. A,
17A, 1051-5.
Wang, Y. R. and Chou, T. W. (1988) 'Three-dimensional analysis of transient
interlaminar thermal stress of laminated composites', Symposium on Mechanics of
Composite Materials ASME AMD, 92, 185-92.
Wang, Y. R. and Chou, T. W. (1989) 'Three-dimensional analysis of transient
interlaminar thermal stress of laminated composites', /. Appl. Mech., 56, 601.
Wang, Y. R. and Chou, T. W. (1991). 'Thermal shock resistance of laminated
ceramic matrix composites', /. Mat. Sci. in press.
Whitney, J. M. (1972) 'Stress analysis of thick laminated composite and sandwich
plates', /. Comp. Mat., 6, 426—40.
Whitney, J. M. and Pagano, N. J. (1970) 'Shear deformation in heterogeneous
anisotropic plates', /. Appl. Mech., 37, (4), 1031-6.
Whitney, J. M. and Sun, C. T. (1973) 'A higher order theory for extensional motion
of laminated composites', /. Sound and Vibration, 30, 85-97.

Chapter 3
Aveston, J. and Kelly, A. (1973) 'Theory of multiple fracture of fibrous composites',
/. Mat. Sci., 8,352-62.
Aveston, J. and Kelly, A. (1980) 'Tensile first cracking strain and strength of hybrid
composites and laminates', Phil. Trans. Royal Soc. London Series A, 294, 519-34.
Aveston, J., Cooper, G. and Kelly, A. (1971) 'Single and multiple fracture', in The
Properties of Fibre Composites, Conference Proceedings, National Physical
Laboratory, IPC Science and Technology Press Ltd., pp. 15-26.
Bader, M. G., Bailey, J. E., Curtis, P. T. and Parvizi, A. (1979) 'The mechanism of
initiation and development of damage in multi-axial fibre-reinforced plastics on
laminates', in Mechanical Behavior of Materials, K. J. Miller and R. F. Smith,
eds., Pergamon Press, Oxford, pp. 227-39.
Bailey, J. E., Curtis, P. T. and Parvizi, A. (1979) 'On the transverse cracking and
longitudinal splitting behavior of glass and carbon fibre reinforced epoxy cross-ply
laminates and the effect of Poisson and thermally generated strain', Proc. Royal
Soc. London, Series A, 366, 599-623.
Bailey, J. E. and Parvizi, A. J. (1981) /. Mat. Sci., 16, 649.
Bjeletich, J. G., Crossman, F. W. and Warren, W. J. (1979) 'The influence of
stacking sequence on failure modes in quasi-isotropic graphite-epoxy laminates', in
Failure Modes in Composites - IV, J. R. Cornie and F. W. Crossman, eds.,
American Institute of Mining, Metallurgical and Petroleum Engineers, New York,
p. 118.
Bradley, W. L. and Cohen, R. N. (1985) 'Matrix deformation and fracture in
graphite reinforced epoxies', in Delamination and Debonding of Materials, W. S.
Johnson, ed., ASTM STP 876, pp. 389-410.
530 References

Budiansky, B., Hutchinson, J. W. and Evans, A. G. (1986) 'Matrix fracture in

fiber-reinforced ceramics', /. Mech. Phys. Sol., 34, 167-89.
Burgel, B., Perry, A. J. and Scheider, W. R. (1970) 'On the theory of fiber
strengthening',/. Mech. Phys. Sol., 18, 101-14.
Carrara, A. S. and McGarry, F. J. (1968) 'Matrix and interface stresses in a
discontinuous fiber composite model', J. Comp. Mat., 2, 222-43.
Chen, C. H. (1973) 'Tension of a composite bar with fibre discontinuities and soft
inter-fibre material', Fibre Sci. Tech., 6, 1.
Chen, P. E. (1971) 'Strength properties of discontinuous fiber composites', Polymer
Eng. Sci., 11,51-6.
Chi, Z. F. and Chou, T. W. (1983) 'An experimental study of the effect of
prestressed loose carbon strands on composite strength', /. Comp. Mat., 17,
196-209.
Chi, Z. F., Chou, T. W. and Shen, G. (1984) 'Determination of single fibre strength
distribution from fibre bundle testings', /. Mat. Sci., 19, 3319-24.
Coleman, B. D. (1958) 'On the strength of classical fibres and fibre bundles', /.
Mech. Phys. Sol., 7,60.
Crossman, F. W. and Wang, A. S. D. (1982) ASTM Symposium on Damages in
Composite Materials: Basic Mechanisms, Accumulation, Tolerance, and
Characterization, K. L Reifsnider, ed., ASTM STP 775, ASTM, Philadelphia.
Crossman, F. W., Warren, W. J., Wang, A. S. D. and Law, G. E. (1980) 'Initiation
and growth of transverse cracks and edge delamination in composite laminates, II.
Experimental correlation', /. Comp. Mat., 14, 88-108.
Daniels, H. E. (1945) 'The statistical theory of the strength of bundles of threads I',
Proc. Royal Soc. London Series, A, 183, 405.
Dhingra, A. K. (1980) 'Alumina fibre FP\ Phil. Trans. R. Soc. London, (A), 294,
411-17.
Epstein, B. (1948) 'Statistical aspects of fracture problems', /. Appl. Phys., 19, 140.
Fichter, B. W. (1969) 'Stress concentration around broken filaments in a filament-
stiffened sheet', NASA TN D-5453.
Fichter, B. W. (1970) 'Stress concentrations infilament-stiffenedsheets of finite
length', NASA TN D-5947.
Fukuda, H. and Kawata, K. (1976a) 'On the stress concentration factor in fibrous
composites', Fibre Sci. Tech., 9, 189.
Fukuda, H. and Kawata, K. (1976b) 'Strength estimation of unidirectional
composites', Trans. Japan Soc. Comp. Mat., 2, 59.
Fukuda, H. and Kawata, K. (1977) 'On the strength distribution of unidirectional
fibre composites', Fibre Sci. Tech., 10, 53.
Fukuda, H. and Kawata, K. (1980) 'Stress distribution of laminates including
discontinuous layers', Fibre Sci. Tech., 13, 255-67.
Fukunaga, H., Peters, P. W. M., Schulte, K. and Chou, T. W. (1984) 'Probabilistic
failure strength analyses of graphite/epoxy cross-ply laminates', /. Comp. Mat.,
18, 339.
Garrett, K. W. and Bailey, J. E. (1977a) /. Mat. Sci., 12, 157.
Garrett, K. W. and Bailey, J. E. (1977b) /. Mat. Sci., 12, 2189.
Goree, J. G. and Gross, R. S. (1979) 'Analysis of a unidirectional composite
containing broken fibers and matrix damage', Eng. Fracture Mech., 13, 563-78.
Goree, J. G. and Gross, R. S. (1980) 'Stresses in a three-dimensional unidirectional
composite containing broken fibers', Eng. Fracture Mech., 13, 395-405.
Gucer, D. E. and Gurland, J. (1962) 'Comparison of the statistics of two fracture
modes',/. Mech. Phys. Sol., 10, 365.
 References 531

Harlow, D. G. (1979) 'Properties of the strength distribution for composite

materials', Composite Materials: Testing and Design {Fifth conference), ASTM
STP 674, S. W. Tsai, ed., American Society for Testing and Materials, pp.
484-501.
Harlow, D. G. and Phoenix, S. L. (1978a) 'The chain-of-bundles probability model
for the strength of fibrous materials I: analysis and conjectures', /. Comp. Mat.,
12, 195.
Harlow, D. G. and Phoenix, S. L. (1978b) 'The chain-of-bundles probability model
for the strength of fibrous materials II: a numerical study of convergence', /.
Comp. Mat, 12,314.
Harlow, D. G. and Phoenix, S. L. (1979) 'Bounds on the probability of failure of
composite materials', Int. J. Fracture, 15, 321-36.
Harlow, D. G. and Phoenix, S. L. (1981a) 'Probability distributions for the strength
of composite materials I: two level bounds', Int. J. Fracture, 17, 347-72.
Harlow, D. G. and Phoenix, S. L. (1981b) 'Probability distributions for the strength
of composite materials II: a convergent sequence of tight bounds', Int. J. Fracture,
17, 601-30.
Hedgepeth, J. M. (1961) 'Stress concentrations in filamentary structures', NASA TN
D-882.
Hedgepeth, J. M. and Van Dyke, P. (1967) 'Local stress concentrations in imperfect
filamentary composite materials', /. Comp. Mat., 1,294.
Henstenburg, R. B. and Phoenix, S. L. (1989) 'Interfacial shear strength studies
using the single-filament-composite Test II: a probability model and Monte-Carlo
simulation', Polymer Composites, 10, 389-408.
Hikami, F. and Chou, T. W. (1984a) 'A probabilistic theory for the strength of
discontinuous fiber composites', /. Mat. ScL, 19, 1805.
Hikami, F. and Chou, T. W. (1984b) 'Statistical treatment of transverse crack
propagation in aligned composites', AIAA J., 22, 1485.
Hikami, F. and Chou, T. W. (1990) 'Explicit crack problem solutions of
unidirectional composites: elastic stress concentrations', AIAA J., 28, 499-505.
Ji, X. (1982) 'On the hybrid effect and fracture mode of interlaminated hybrid
composites', Proceedings of the Fourth International Conference on Composite
Materials, Tokyo, p. 1137.
Ji, X., Liu, X. R. and Chou, T. W. (1985) 'Dynamic stress concentration factors in
unidirectional composites',J. Comp. Mat., 19, 269-75.
Kelly, A. (1973) Strong Solids, Clarendon Press, Oxford.
Kelly, A. (1976) 'Composites with brittle matrices', in Frontiers in Materials Science,
L. E. Murr and C. Stein, eds., Marcel Dekker Inc., New York, pp. 335-64.
Kelly, A. and Nicholson, R. B. (eds.) (1971) Strengthening Methods in Crystals,
Elsevier, London.
Kies, J. A. (1962) US Naval Research Laboratory, Report No. 5752.
Kirkpatrick, E. G. (1974) Introductory Statistics and Probability for Engineering,
Science and Technology, Prentice-Hall, Englewood Cliffs, New Jersey.
Kulkarni, S. V., Rosen, B. W. and Zweben, C. (1973) 'Load concentration factors
for circular holes in composite laminates', /. Comp. Mat., 7, 387.
Lei, S. C. (1986) 'A stochastic model for the damage growth during the transverse
cracking process in composite laminates', Ph.D. Thesis, Drexel University.
Lipson, S. G. and Lipson, H. (1981) Optical Physics, 2nd edn., Cambridge
University Press.
McCartney, L. N. (1987) 'Mechanics of matrix cracking in brittle-matrix fibre-
reinforced composites', Proc. Royal Soc. London, Series A, 409, 329-50.
532 References

Manders, P., Bader, M. and Chou, T. W. (1982) 'Monte Carlo simulation of the
strength of composite fiber bundles', Fiber Sci. Tech., 17, 183.
Manders, P. W. and Chou, T. W. (1983a) 'Variability of carbon and glass fibers, and
the strength of aligned composites', /. Reinforced Plastics & Composites, 2, 43.
Manders, P. W. and Chou, T. W. (1983b) 'Enhancement of strength in composites
reinforced with previously-stressed fibers', /. Comp. Mat., 17, 26.
Manders, P. W., Chou, T. W., Jones, F. R. and Rock, J. W. (1983) 'Statistical
analysis of multiple fracture in 079070° glass fibre/epoxy resin laminates', /. Mat.
Sci., 18,2876-89.
Metcalfe, A. G. and Schmitz, G. K. (1964) 'Effect of length on the strength of glass
fibers', Proc. ASTM, 64, 1075.
Mileiko, S. T. (1969) 'The tensile strength and ductility of continuous fibre
composites', /. Mat. Sci., 4, 974.
Mills, G. J. and Dauksys, R. J. (1973) 'Effect of prestressing boron/epoxy prepreg
on composite strength properties', AIAA J., 11, 1459.
Netravali, A. N., Henstenburg, R. B., Phoenix, S. L. and Schwartz, P. (1989)
'Interfacial shear strength studies using the single-filament-composite Test I:
experiments on graphite fibers in epoxy', Polymer Composites, 10, 226-41.
Oh, K. P. (1979) 'A Monte Carlo study of the strength of unidirectional
fiber-reinforced composites', /. Comp. Mat., 13, 311.
Pagano, N. J. and Pipes, R. B. (1971) 'The influence of stacking sequence on
laminate strength', /. Comp. Mat., 5, 50-7.
Parratt, N. J. (1960) 'Defects in glass fibers and their effects on the strength of plastic
mouldings', Rubber and Plastics Age, March 1960.
Parvizi, A. (1979). 'Transverse cracking in glass fibre reinforced plastic composites',
Ph.D. Thesis, University of Surrey.
Parvizi, A. and Bailey, J. E. (1978) 'On multiple transverse cracking in glass fiber
epoxy cross-ply laminates', /. Mat. Sci., 13, 2131.
Parvizi, A., Garrett, K. W. and Bailey, J. E. (1978) 'Constrained cracking in glass
fiber-reinforced epoxy cross-ply laminates', /. Mat. Sci., 13, 195.
Peters, P. W. M. and Chou, T. W. (1987) 'On cross-ply cracking in glass- and
glass-epoxy laminates', Composites, 18, 40.
Phoenix, S. L. (1974) 'Probabilistic strength analysis of fiber bundle structures', Fibre
Sci. Tech., 7, 15.
Phoenix, S. L. (1979) 'Statistical aspects of failure of fibrous composites', Composite
Materials: Testing and Design (Fifth Conference), ASTM STP 674, S. W. Tsai, ed.,
American Society for Testing and Materials, pp. 455-83.
Phoenix, S. L., Schwartz, P. and Robinson IV, H. H. (1988) 'Statistics for the
strength and lifetime in creep-rupture of model carbon/epoxy composites', Comp.
Sci. Tech., 32,81-120.
Phoenix, S. L. and Smith, R. L. (1983) 'A comparison of probabilistic techniques for
the strength of fibrous materials under local load-sharing among fibers', Int. J. Sol.
Structures, 19, 479-96.
Phoenix, S. L. and Taylor, H. M. (1973) 'The asymptotic strength distribution of a
general fiber bundle', Adv. Appl. Prob., 5, 200.
Pipes, R. B. and Pagano, N. J. (1970) 'Interlaminar stresses in composite laminates
under uniaxial extension', /. Comp. Mat., 4, 538-48.
Reifsnider, K. L., Henneke, E. G., Stinchcomb, W. W. and Duke, J. C. (1983)
'Damage mechanics and NDE of composite laminates', in Mechanics of Composite
Materials - Recent Advances (Z. Hashin and C. T. Herakovich, eds.), Pergamon,
New York, pp. 399-420.
 References 533

Rosen, B. W. (1964) 'Tensile failure of fibrous composites', AlAA J., 2, 1985.

Rosen, B. W. (1970) Thermomechanical properties of fibrous composites', Proc.
Royal Soc. London, Series A, 319, 79-94.
Russell, A. J. and Street, K. N. (1985) 'Moisture and temperature effects on the
mixed-mode delamination fracture of unidirectional graphite-epoxy', in
Delamination and Debonding of Materials, W. S. Johnson, ed., ASTM STP 876,
pp. 349-70.
Scop, P. M. and Argon, A. S. (1967) 'Statistical theory of strength of laminated
composites', /. Comp. Mat., 1, 92.
Scop, P. M. and Argon, A. S. (1969) 'Statistical theory of strength of laminated
composites IF, /. Comp. Mat., 3, 30.
Smith, R. L. (1980) 'A probability model for fibrous composites with local
load-sharing', Proc. Royal Soc. London, Series A, 372, 539-53.
Smith, R. L. (1982) 'A note on a probability model for fibrous composites', Proc.
Royal Soc. London, Series A, 382, 179-82.
Smith, R. L. and Phoenix, S. L. (1981) 'Asymptotic distributions for the failure of
fibrous materials under series-parallel structure and equal load sharing', /. Appl.
Mech., 48, 75-82.
Smith, R. L., Phoenix, S. L., Greenfield, M. R., Henstenburg, R. B. and Pitt, R. E.
(1983) 'Lower-tail approximations for the probability of failure of three-
dimensional fibrous composites with hexagonal geometry', Proc. Royal Soc.
London, Series A, 388, 353-91.
Spiegel, M. R. (1961) Statistics, Schaum's Outline Series, McGraw-Hill,
New York.
Takao, Y., Taya, M. and Chou, T. W. (1981) 'Stress field due to a cylindrical
inclusion with constant axial eigenstrain in an infinite elastic body', ASME J. Appl.
Mech., 48, 853-8.
Talreja, R. (1985) 'A continuum mechanics characterization of damage in composite
materials', Proc. Royal Soc, London, Series A, 399, 195-216.
Talreja, R. (1986) 'Stiffness properties of composite laminates with matrix cracking
and interior delamination', Eng. Fract. Mech., 25, 751-62.
Talreja, R. (1987) Fatigue of Composite Materials, Technomic Pub. Co., Lancaster,
Pennsylvania.
Talreja, R. (1989) 'Fatigue of composites', in Concise Encyclopedia of Composite
Materials, A. Kelly, ed., Pergamon Press, Oxford, pp. 77-81.
Van Dyke, P. and Hedgepeth, J. M. (1969) 'Stress concentrations from single-
filament failures in composite materials', Textile Res. J., July, p. 618.
Vinson, J. R. and Chou, T. W. (1975) Composite Materials and Their Use in
Structures, Applied Science Publishers, London.
Wang, A. S. D. (1984) 'Fracture mechanics of sublaminate cracks in composite
materials', Comp. Techl. Rev., 6, 45-62.
Wang, A. S. D. (1987) 'Strength, failure, and fatigue analysis of laminates'.
Engineering Materials Handbook, 1, 236-51, ASM International, Metals Park,
Ohio.
Wang, A. S. D., Chou, P. C. and Lei, S. C. (1984) 'A stochastic model for the
growth of matrix cracks in composite laminates', /. Comp. Mat., 18, 239-54.
Wang, A. S. D. and Crossman, F. W. (1977) 'Some new results on edge effects in
symmetric composite laminates', J. Comp. Mat., 11,92-102.
Wang, A. S. D. and Crossman, F. W. (1980) 'Initiation and growth of transverse
cracks and edge delamination, I. An energy method', /. Comp. Mat., 14, 71-87.
Wang, A. S. D., Kishore, N. N. and Li, C. A. (1985) 'On crack development in
534 References

graphite-epoxy [0 2 /90 Js laminates under uniaxial tension', Comp. Sci. Tech., 23,
1-31.
Wang, A. S. D., Slomiana, M. and Bucinell, R. B. (1985) 'Delamination crack
growth in composite laminates', in Delamination and Debonding of Materials, W.
S. Johnson, ed., ASTM STP 876, pp. 135-67.
Weibull, W. (1939a) 'A statistical theory of the strength of materials', Ing.
Vetenskaps Akad. Handl., no. 151.
Weibull, W. (1939b) 'The phenomenon of rupture in solids', Ing. Vetenskaps Akad.
Handl., no. 153.
Weibull, W. (1951) 'A statistical distribution function of wide applicability', /. Appl.
Mech., 18,293.
Zender, G. W. and Deaton, J. W. (1963) 'Strength offilamentarysheets with one or
more fibers broken', NASA TN D-1609.
Zweben, C. (1968) 'Tensile failure of fiber composites', AIAA J., 6, 2325.
Zweben, C. (1974) 'An approximate method of analysis for notched unidirectional
composites', Eng. Frac. Mech., 6, 1.
Zweben, C. and Rosen, B. W. (1970) 'A statistical theory of material strength with
application to composite materials',/. Mech. Phys. Sol., 18, 189.

Chapter 4
Akasaka, T. (1974) 'A practical method of evaluating the isotropic elastic constants
of glass mat reinforced plastics', Comp. Mat. Struct. (Japan), 3, 21-2.
Anderson, R. M. and Lavengood, R. E. (1968) 'Variables affecting strength and
modulus of short fiber composites', Soc. Plastic. Engrs. J., 24, 20.
Arridge, R. G. C. (1963) 'Orientation effects in fibre reinforced composites where
the modulus of the fibres is no more than an order of magnitude greater than that
of the matrix', Proc. 18th Ann. Tech. Conf. Reinf. Plastics Div., Soc. Plastics
Industry, Sec. 4-A, February 1963.
Bader, M. G., Chou, T. W. and Quigley, J. (1979) 'On the strength of discontinuous
fiber composites with polymeric matrices', in New Developments and Applications
in Composites, D. Wilsdorf, ed., TMS-AIME, New York.
Baker, R. M. and MacLaughlin, T. F. (1971) 'Stress concentrations near a
discontinuity infibrouscomposites', /. Comp. Mat. 5, 492.
Bert, C. W. (1979) 'Composite-material mechanics: prediction of properties of
planar-random fiber composites'. Presented at the 34th Annual Conference of the
Reinforced Plastics/Composites Institute, New Orleans, Louisiana, January
29-February 2.
Beran, M. (1965) Nuovo Cimento Ser. X, 35, 771.
Beran, M. and Molyneax, J. (1966) Quart. Appl. Math., 24, 107.
Blumentritt, B. F., Vu, B. T. and Cooper, S. L. (1974) 'The mechanical properties
of oriented discontinuous fibre-reinforced thermoplastics I. Unidirectional fiber
orientation', Polymer Eng. Sci., 14, 633-45.
Blumentritt, B. G., Vu, B. T. and Cooper, S. L. (1975) 'Mechanical properties of
discontinuous fiber reinforced thermoplastics, II. Random-in-plane fiber
orientation', Polymer Eng. Sci., 15, 428-36.
Bowyer, W. H. and Bader, M. G. (1972) 'On the reinforcement of thermoplastics by
imperfectly aligned discontinuous fibers', /. Mat. Sci., 7, 1315.
Budiansky, B. (1965) 'On the elastic moduli of some heterogeneous materials', J.
Mech. Phys. Sol., 13, 223-7.
Budiansky, B. (1970) J. Comp. Mat., 4, 286.
 References 535

Burgel, B., Perry, A. J. and Schneider, W. R. (1970) 'On the theory of fibre
strengthening',/. Mech. Phys. Sol., 18, 101-14.
Carrara, A. S. and McGarry, F. J. (1968) 'Matrix and interface stresses in a
discontinuous fiber composite model', /. Comp. Mat., 2, 222-43.
Chamis, C. C. and Sendeckyj, G. P. (1968) 'Critique on theories predicting
thermoelastic properties of fibrous composites', /. Comp. Mat., 2, 232.
Chang, C. I., Conway, H. D. and Weaver, T. C. (1972) 'The elastic constants and
bond stresses for a three-dimensional composite reinforced by discontinuous
fibers', Fibre Sci. Tech., 5, 143-62.
Chen, P. E. (1971) 'Strength properties of discontinuous fiber composites', Polymer
Eng. Sci., 11,51.
Chen, P. E. and Lavengood, R. E. (1969) 'Stress fields around multiple inclusions',
Monsanto/Washington University, ONR/ARPA Association, HPC 68-60, AD
846907, January, 1969.
Chen, P. E. and Lewis, T. B. (1970) 'Stress analysis of ribbon reinforced
composites', Polymer Eng. Sci., 10, 43.
Chou, T. W. and Kelly, A. (1976) 'Fiber composites', in 'Challenges and
opportunities in materials science and engineering', Mat. Sci. Eng., 25, 35.
Chou, T. W. and Kelly, A. (1980) 'Mechanical properties of composites', Ann. Rev.
Mat. Sci., 10,229.
Chou, T. W. and Nomura, S. (1980) 'On the thermoelastic behavior of short fiber
and hybrid composites', Proceedings of the Third International Conference on
Composite Materials, Pergamon Press, New York, pp. 69-80.
Chou, T. W. and Nomura, S. (1981) 'Fiber orientation effects on the thermoelastic
properties of short-fiber composites', Fibre Sci. Tech., 14, 279.
Chou, T. W., Nomura, S. and Taya, M. (1980) 'A self-consistent approach to the
elastic stiffness of short-fiber composites', /. Comp. Mat., 14, 178.
Christensen, R. M. (1971) Theory ofViscoelasticity, Academic Press, New York.
Christensen, R. M. and Waals, F. M. (1972) 'Effective stiffness of randomly oriented
fibre composites',/. Comp. Mat., 6, 518-32.
Christensen, R. M. (1979) Mechanics of Composite Materials, Wiley-Interscience,
New York.
Conway, H. D. and Chang, C. I. (1971) 'The effective elastic constants and bond
stresses for a fiber reinforced elastic sheet', Fibre Sci. Tech., 5, 249-60.
Cook, J. (1968) 'The elastic constants of an isotropic matrix reinforced with
imperfectly oriented fibres', Brit. J. Appl. Phys. (/. Phys. D), Series 2, 1, 799-812.
Cox, H. L. (1952). 'The elasticity and strength of paper and other fibrous materials',
Brit. J. Appl. Phys., 3, 72-9.
Curtis, P. T., Bader, M. G. and Bailey, J. E. (1978) 'The stiffness and strength of a
polyamide thermoplastic reinforced with glass and carbon fibres', /. Mat. Sci., 13,
377.
Dederichs, P. H. and Zeller, R. (1973). 'Variational treatment of the elastic
constants of disordered materials', Z. Physik, 259, 103.
Edwards, H. and Evans, N. P. (1980) 'A method for the production of high quality
aligned short fibre mats and their composites', Proceedings of the Third
International Conference on Composite Materials, Pergamon Press, New York, pp.
1620-35.
Eimer, C. Z. (1971) 'The viscoelasticity of multi-phase media'. Arch. Mech., 23,
3-15.
Eshelby, J. D. (1957) 'The determination of the elastic field of an ellipsoidal
inclusion and related problems', Proc. Royal Soc, A241, 376—96.
536 References

Eshelby, J. D. (1961) 'Elastic inclusions and inhomogeneity', in Progress in Solid

Mechanics, I. N. Sneddon and R. Hill, eds., vol. 2, North-Holland, Amsterdam,
p. 89.
Favre, J. P. and Perrin, J. (1972) 'Carbon fibre adhesion to organic matrices', /. Mat.
ScL, 7, 1113.
Friedrich, K. (1985) 'Microstructural efficiency and fracture toughness of short
fiber/thermoplastic matrix composites', Comp. Sci. Tech., 22,43-74.
Friedrich, K. (1989) 'Fractographic analysis of polymer composites', in Application
of Fracture Mechanics to Composite Materials, Composite Material Series, vol. 6,
Klaus Friedrich, ed., Elsevier, Amsterdam, p. 425.
Friedrich, K. and Karger-Kocsis, J. (1989) 'Fractography and failure mechanisms of
unfilled and short fiber reinforced semi-crystalline thermoplastics', in Fractography
and Failure Mechanisms of Polymers and Composites, A. C. Roulin-Moloney,
ed., Elsevier-Applied Science, London, pp. 437-94.
Friedrich, K., Schulte, K., Horstenkamp, G. and Chou, T. W. (1985) 'Fatigue
behavior of aligned short carbon fiber reinforced polyimide and polyethersulfone
composites', /. Mat. ScL, 20, 3353.
Fukuda, H. and Chou, T. W. (1981a) 'An advanced shear-lag model applicable to
discontinuous fiber composites', /. Comp. Mat., 15, 79.
Fukuda, H. and Chou, T. W. (1981b) 'A probabilistic theory for the strength of
short-fiber composites',/. Mat. Sci., 16, 1088.
Fukuda, H. and Chou, T. W. (1982) 'A probabilistic theory of the strength of
short-fiber composites and variable fiber length and orientation', /. Mat. Sci., 17,
1003.
Fukuda, H. and Kawata, K. (1974) 'On Young's modulus of short fibre composites',
Fibre Sci. Tech., 7,207-22.
Fukuda, H. and Kawata, K. (1977) 'On the strength distribution of unidirectional
fibre composites', Fibre Sci. Tech., 10, 53.
Haener, J. and Ashbaugh, N. (1967) 'Three-dimensional stress distribution in a
unidirectional composite', /. Comp. Mat., 1, 54-63.
Hahn, H. T. (1978) 'Stiffness and strength of discontinuous fiber composites', in
Composite Materials in the Automotive Industry, S. V. Kulkarni, C. H. Zweben
and R. B. Pipes, eds., ASME, New York, pp. 85-109.
Hale, D. K. and Kelly, A. (1972) 'Strength of fibrous composite materials', in
Annual Review of Materials Science, vol. 2, R. A. Huggins, ed., Annual Review,
Inc., Palo Alto, California, p. 405.
Halpin, J. C. (1969) 'Stiffness and expansion estimates for oriented short fiber
composites',/. Comp. Mat., 3, 732-4.
Halpin, J. C. (1984) Primer on Composite Materials: Analysis, Technomic Publishing
Co., Inc., Lancaster, Pennsylvania.
Halpin, J. C , Jerine, K. and Whitney, J. M. (1971) 'The laminate analogy for 2 and
3 dimensional composite materials', J. Comp. Mat., 5, 36—49.
Halpin, J. C. and Pagano, N. J. (1969) 'The laminate approximations for randomly
oriented fibrous composites', /. Comp. Mat., 3, 720-4.
Hancock, P. and Cuthbertson, J. (1970) 'The effect of fibre length and interfacial
bond in glassfibre-epoxyresin composites, /. Mat. Sci., 15, 762-8.
Hashin, Z. (1965a) 'On elastic behavior of fibre reinforced materials of arbitrary
transverse phase geometry', /. Mech. Phys. Sol, 13, 179.
Hashin, Z. (1965b) 'Viscoelastic behavior of heterogeneous media', /. Appl. Mech.,
32, 630-6.
 References 537

Hashin, Z. (1968) 'Assessment of the self-consistent scheme approximation:

conductivity of particulate composites',/. Comp. Mat., 2,284-300.
Hashin, Z. (1969) 'The inelastic inclusion problem', Int. J. Eng. Sci., 7, 11-36.
Hashin, Z. (1972) Theory of Fiber Reinforced Materials, NASA CR-1974.
Hashin, Z. and Rosen, B. W. (1964) 'The elastic moduli of fiber-reinforced
materials', J. Appl. Mech., 31, 223-32.
Hashin, Z. and Shtrikman, S. (1962) /. Appl. Phys., 30, 3125.
Hashin, Z. and Shtrikman, S. (1963) 'A variational approach to the theory of the
elastic behavior of multiphase materials', /. Mech. Phys. Sol, 11, 127-40.
Hermans, J. J. (1967) 'The elastic properties of fibre reinforced materials when the
fibers are aligned', Proc. Konigl. Nederl, Akad, van Weteschappen Amsterdam,
Series B, 70, 1-9.
Hikami, F. and Chou, T. W. (1984a) 'A probabilistic theory for the strength of
discontinuous fiber composites', J. Mat. Sci., 19, 1805.
Hikami, F. and Chou, T. W. (1984b) 'Statistical treatment of transverse crack
propagation in aligned composites', AIAA J., 22, 1485-90.
Hikami, F. and Chou, T. W. (1990) 'Explicit crack problem solutions of
unidirectional composites: elastic stress concentrations', AIAA J., 28, 499-505.
Hill, R. (1952) 'The elastic behavior of a crystalline aggregate', Proc. Phys. Soc,
A65, 349.
Hill, R. (1965a) 'Theory of mechanical properties of fiber-strengthened materials -
III. Self-consistent model', J. Mech. Phys. Sol, 13, 189-98.
Hill, R. (1965b) 'A self-consistent mechanics of composite materials', /. Mech. Phys.
Sol, 13, 213-25.
Hori, M. and Yonezawa, F. (1975) 'Statistical theory of effective electrical, thermal
and magnetic properties of random heterogeneous materials IV, /. Math. Phys.,
16. 352.
Hsu, P. L., Yau, S. S. and Chou, T. W. (1986) 'Stress-corrosion cracking and its
propagation in aligned short-fiber composites', J. Mat. Sci., 21, 3703.
Ishikawa, H., Chou, T. W. and Taya, M. (1982) 'Prediction of failure modes in
unidirectional short fiber composites', /. Mat. Sci., 17, 832.
Jackson, P. W. and Cratchley, D. (1966) /. Mech. Phys. Sol, 14, 49.
Kacir, L. and Narkis, M. (1975) Polymer Eng. Sci., 15, 525.
Kardos, J. (1973) 'Structure property relations in short-fiber reinforced plastics',
CRC Crit. Rev. Solid State Sci., August, pp. 419-50.
Kelly, A. (1971) 'Reinforcement of structural materials by long strong fibres', Met.
Trans., 3,2313.
Kelly, A. (1973) Strong Solids, 2nd edn., Clarendon Press, Oxford.
Kelly, A. and Davies, G. J. (1965) 'The principles of the fibre reinforcement of
metals', Met. Rev., 10, 1.
Kelly, A. and Tyson, W. R. (1965a) 'Fibre-strengthened materials', in High Strength
Materials, V, F. Zackay, ed., J. Wiley and Sons, Inc., New York, p. 578.
Kelly, A. and Tyson, W. R. (1965b) 'Tensile properties of fiber-reinforced metals:
copper/tungsten and copper/molybdenum', J. Mech. Phys. Sol, 13, 329.
Kerner, E. H. (1956) 'The elastic and thermoelastic properties of composite media',
Proc. Phys. Soc, B69, 808-13.
Kilchinskii, A. A. (1965) 'On the model for determining thermoelastic characteristics
of fiber reinforced materials', Prikladnaia Mekhanika, 1, 1.
Kilchinskii, A. A. (1966) 'Approximate method for determining the relation
between the stresses and strains for reinforced materials of the fiber glass type',
538 References

Thermal Stresses in Elements of Construction, Naukova Kumka, Kiev,

6, 123.
Knibbs, R. H. and Morris, J. B. (1974) 'The effects of fibre orientation on the
physical properties of composites', Composites, 5, 209-18.
Kroner, E. (1958) 'Berechnung der Elastichen Konstanten des Vielkristalls aus den
Konstanten des Eikristalls', Z. Physik, 151, 504.
Kroner, E. (1967) 'Elastic moduli of perfectly disordered composite materials', J.
Mech. Phys.Sol, 15,319.
Kroner, E. (1972) Statistical Continuum Mechanics, CISM Courses and Lectures no.
92, Udine, Springer-Verlag, Wien.
Kroner, E. (1977) 'Bounds for effective elastic moduli of disordered materials', /.
Mech. Phys. Sol, 25, 137.
Lavengood, R. E. (1972) 'Strength of short-fiber reinforced composites', Polymer
Eng. 5ci., 12,48.
Laws, N. (1973) 'On the thermostatics of composite materials', /. Mech. Phys. Sol,
21,9.
Laws, N. and McLaughlin, R. (1978) 'Self-consistent estimates for the viscoelastic
creep compliances of composite materials', Proc. Royal Soc, A359, 251-73.
Lee, L. H. (1969) 'Strength-composition relationships of random short glass
fiber-thermoplastics composites', Polymer Eng. Sci., 9, 213-24.
Lees, J. K. (1968) 'A study of the tensile modulus of short fiber reinforced plastics',
Polymer Eng. Sci., 8, 186-94
Levin, V. M. (1967) Inzh. Zh. Mekk. Tverd. Tela, No. 1, p. 88.
MacLaughlin, T. F. (1966) 'Effect of fiber geometry on stress in fiber reinforced
composite materials', Exp. Mech., 6, 481-92.
MacLaughlin, T. F. (1968) 'A photoelastic analysis of fiber discontinuities in
composite materials', /. Comp. Mat., 2, (1), 44.
McNally, D. (1977) 'Short fiber orientation and its effect on the properties of
thermoplastic composite materials', Polymer Plast. Tech. Eng., 8, 101-54.
Mandell, J. F., Grande, D. H., Tsiang, T.-H. and McGarry, F. J. (1986) 'Modified
microdebonding test for direct in situ fiber/matrix bond strength determination in
fiber composites', in Composite Materials: Testing and Design {Seventh
Conference), ASTM STP 893, J. M. Whitney, ed., American Society for Testing
and Materials, Philadelphia, pp. 87-108.
Manders, P. W. and Chou, T. W. (1982) 'The strength of aligned short-fiber carbon,
glass, and hybrid carbon/glass composites', Proceedings of the Fourth International
Conference on Composite Materials, The Japan Society for Composite Materials,
Tokyo, pp. 1075-82.
Manera, M. (1971) 'Elastic properties of randomly oriented short fiber-glass
composites',/. Comp. Mat., 11,235-47.
Miller, B., Muri, P. and Rebenfeld, L. (1987) Comp. Sci. Tech., 28, 17.
Muki, R. and Sternberg, E. (1969) 'On the diffusion of an axial load from an infinite
cylindrical bar embedded in an elastic medium', Int. J. Sol. Struct., 5, 587.
Muki, R. and Sternberg, E. (1970) 'Elastostatic load-transfer to a half-space from a
partially embedded axially loaded rod', Int. J. Sol. Struct., 6, 69.
Muki, R. and Sternberg, E. (1971) 'Load-absorption by a discontinuous filament in a
fiber-reinforced composite', Z. Angew. Math. Phys., 22, 809.
Mura, T. (1982) Micromechanics of Defects in Solids, Martinus Nijhoff Publishers,
The Hague.
 References 539

Nicolais L. (1975) 'Mechanics of composites', Polymer Eng. Sci., 15, 137-49.

Nielsen, L. E. and Chen, P. E. (1968) 'Young's modulus of compositesfilledwith
randomly oriented fibers', J. Mat., 3, 352-8.
Nomura, S. and Chou, T. W. (1980) 'Bounds of effective thermal conductivity of
short-fiber composites', /. Comp. Mat., 14, 120.
Nomura, S. and Chou, T. W. (1981) 'Effective thermoelastic constants of
short-fiber composites', Int. J. Eng. Sci., 19, 1.
Nomura, S. and Chou, T. W. (1984) 'Bounds of elastic moduli of multiphase
short-fiber composites', /. Appl. Mech., 51, 540.
Nomura, S. and Chou, T. W. (1985) 'The viscoelastic behavior of short-fiber
composite materials', Int. J. Eng. Sci., 23, 193.
Outwater, J. O., Jr. (1956) 'The mechanics of plastics reinforcement in tension',
Mod. Plast., 33, 156-62.
Pakdemirli, E. and Williams, J. G. (1969) 'Metal fibre reinforced thermoplastics and
the role of adhesion efficiency', /. Mech. Eng. Sci., 11, 68-75.
Piggott, M. R. (1987) 'Debonding and friction atfibre-polymerinterfaces I: Criteria
for failure and sliding', Comp. Sci. Tech., 30, 295-306.
Piggott, M. R. and Dai, S. R. (1988) 'Debonding and friction at fibre-polymer
interfaces II: microscopic model experiments, Comp. Sci. Tech., 31, 15-24.
Reuss, A. (1929) Zeit Angew Math. U. Mech., 9, 49.
Richter, H. (1980) 'Single fibre and hybrid composites with aligned discontinuous
fibres in polymer matrix', Proceedings of the Third International Conference on
Composite Materials, Pergamon Press, New York, pp. 387-98.
Riley, V. R. and Reddaway, J. L. (1968) /. Mat. Sci., 3, 41.
Rosen, B. W. (1964) 'Tensile failure of fibrous composites', AIAA J., 2, 1985.
Rosen, B. W. and Shu, L. S. (1971) 'On some symmetry conditions for three
dimensional fibrous composites', /. Comp. Mat., 5, 279.
Rosen, W. and Hashin, Z. (1970) 'Effective thermal expansion coefficients and
specific heats of composite materials', Int. J. Eng. Sci., 8, 157-73.
Schapery, R. A. (1967) 'Stress analysis of viscoelastic composite materials', /. Comp.
Mat., 1,228-67.
Schapery, R. A. (1968) 'Thermal expansion coefficients of composite materials based
on energy principles', J. Comp. Mat., 2, 380-404.
Schapery, R. A. (1974) Composite Materials, G. P. Sendeckyj, ed., Vol. 2,
Academic Press, New York.
Schierding, R. G. and Deex, O. D. (1969) 'Factors influencing the properties of
whisker-metal composites', /. Comp. Mat., 3, 618-29.
Stowell, E. Z. and Liu, T. S. (1961) J. Mech. Phys. Sol, 9, 242.
Takao, T., Chou, T. W. and Taya, M. (1982) 'Effective longitudinal Young's
modulus of misoriented short fiber composites', J. Appl. Mech., 49, 536.
Takao, Y., Taya, M. and Chou, T. W. (1981) 'Stress field due to cylindrical inclusion
with constant axial eigenstrain in an infinite body', /. Appl. Mech., 48, 853.
Takao, Y., Taya, M. and Chou, T. W. (1982) 'Effects of fiber-end cracks on the
stiffness of aligned short-fiber composites', Int. J. Sol. Struct., 8, 723.
Taya, M. and Chou, T. W. (1981) 'On two kinds of ellipsoidal inhomogeneities in
infinite elastic body: an application to a hybrid composite', Int. J. Sol. Struct., 17,
553.
Taya, M. and Chou, T. W. (1982) 'Prediction of the stress-strain curve of a short
fiber reinforced thermoplastics', /. Mat. Sci., 17, 2801.
540 References

Tsai, S. W. and Pagano, N. J. (1968) 'Invariant properties of composite materials', in

Composite Materials Workshop, S. W. Tsai, J. C. Halpin and N. J. Pagano, eds.,
Technomic Pub. Co., Stamford, CT, pp. 233-53.
van de Poel, C. (1958) 'On the rheology of concentrated dispersions', Rheol. Acta, 1,
198-205.
Vinson, J. R. and Chou, T. W. (1975) Composite Materials and Their Use in
Structures, Elsevier-Applied Science, London.
Voigt, W. (1889) Ann. Phys., 33, 573.
Wadsworth, N. J. and Spilling, I. (1968) Brit. J. Appl. Phys., 1, 1049.
Warren, F. and Norris, C. B. (1953) 'Mechanical properties of laminate design to be
isotropic', Forest Products Laboratory, Madison, WI, Report No. 1841, May 1953.
Wilczynki, A. P. (1978) 'Random directional reinforcement theory', Fibre Sci.
Tech., 11, 19-22.
Willis, J. R. (1977) 'Bounds and self-consistent estimates for the overall properties of
anisotropiccomposites',/. Mech. Phys. Sol., 25, 185-202.
Wu, C.-T. D. and McCullough, R. L. (1977) 'Constitutive relationships for
heterogeneous materials', in Developments in Composite Materials—1, G. S.
Holister, ed., Applied Science Publishers, London, p. 119.
Yates, B., Overy, M. J., Sargent, J. P., McCalla, B. A., Kingston-Lee, D. M.,
Phillips, L. N. and Rogers, K. F. (1978) /. Mat. Sci., 13, 433.
Yau, S. S. and Chou, T. W. (1989) 'Low temperature performance of short fiber
reinforced thermoplastics', in Test Methods and Design Allowables for Fiber
Composites, vol. 2, C. C. Chamis, ed., ASTM STP 1003, p. 45.
Zeller, R. and Dederichs, P. H. (1973) 'Elastic constants of polycrystals', Phys. Stat.
Sol., b55, 831.

Chapter 5
Adam, T., Fernando, G., Dickson, R. F., Reiter, H. and Harris, B. (1989) 'Fatigue
life prediction for hybrid composites', Int. J. Fatigue, 11, 233-7.
Adams, D. F. (1975) 'A scanning electron microscopic study of hybrid composite
impact response', /. Mat. Sci., 10, 1591—602.
Adams, D. F. and Miller, A. K. (1975) 'An analysis of the impact behavior of hybrid
composite materials', Mat. Sci. Eng., 19, 245-60.
Adams, D. F. and Miller, A. K. (1976) 'The influence of transverse shear on the
static flexure and Charpy impact response of hybrid composite materials', /. Mat.
Sci., 11, 1697-710.
Adams, D. F. and Zimmerman, R. S. (1986) 'Static and impact performance of PE
fiber/graphite fiber hybrid composites', SAMPE J., Nov./Dec, pp. 10-16.
Arrington, M. and Harris, B. (1978) 'Some properties of mixed fibre CFRP',
Composites, 9, 149-52.
Aveston, J. and Kelly, A. (1980) 'Tensile first cracking strain and strength of hybrid
composites and laminates', Phil. Trans. Royal Soc. London, A294, 519-34.
Aveston, J. and Sillwood, J. M. (1976) 'Synergistic fibre strengthening in hybrid
composites', /. Mat. Sci., 11, 1877.
Bader, M. G. and Manders, P. W. (1978) 'Failure strain enhancement in
carbon/glass fiber hybrid composites', Proceedings of the Third International
Conference on Composite Materials, vol. 3, Toronto.
 References 541

Bader, M. G. and Manders, P. W. (1981a) 'The strength of hybrid glass/carbon fiber

composites, part I, failure strain enhancement and failure mode', /. Mat. Sci., 16,
2233-45.
Bader, M. G. and Manders, P. W. (1981b) The strength of hybrid glass/carbon fiber
composites, part II, a statistical model', /. Mat. Sci., 16, 2246-56.
Beaumont, P. W. R., Riewald, P. G. and Zweben, C. (1974) 'Methods for
improving the impact resistance of composite materials', in Foreign Object Impact
Damage to Composites, ASTM STP 568, American Society of Testing and
Materials, Philadelphia, pp. 134-58.
Bucci, R. J., Mueller, L. N., Schultz, R. W. and Prohaska, J. L. (1987) 'ARALL
laminates - results from a cooperative test program', in Advanced Materials
Technology 87, Proceedings 32nd International SAMPE Symposium, vol. 32,
Society for the Advancement of Material and Process Engineering, Corina, CA,
pp. 902-16.
Bunsell, A. R. (1976) Letter to the Editor, Composites, 7, 158.
Bunsell, A. R. and Harris, B. (1974) 'Hybrid carbon and glass fibre composites',
Composites, 5, 157-64.
Chamis, C. C , Hanson, M. P. and Serafini, T. T. (1972) 'Impact resistance of
unidirectional fiber composites', Composite Materials: Testing and Design (Second
Conference), ASTM STP 497, American Society for Testing and Materials,
Philadelphia, pp. 324-49.
Chamis, C. C. and Lark, R. F. (1978) 'Non-metallic hybrid composites: analysis,
design, application and fabrication', in Hybrids and Selected Metal-Matrix
Composites: A State-of-the-Art Review, W. J. Renton, ed., AIAA, New York,
pp. 13-51.
Chamis, C. C. and Sinclair, J. H. (1979) 'Micromechanics of intraply hybrid
composites: elastic and thermal properties', in Modern Developments in Composite
Materials and Structures, J. R. Vinson, ed., The American Society of Mechanical
Engineers, New York, pp. 253-67.
Chan, W. S., Rogers, C. and Aker, S. (1976) 'Improvement of edge delamination
strength using adhesive layers', in Composite Materials Testing and Design, 7th
Conference, J. M. Whitney, ed., ASTM STP893, American Society for Testing
and Materials, Philadelphia, pp. 266-85.
Chen, J. L. and Sun, C. T. (1989) 'Modeling of orthotropic elastic-plastic properties
of ARALL laminates', Comp. Sci. Tech., 36, 321-38.
Chou, T. W. and Kelly, A. (1980a) 'Mechanical properties of fiber composite
materials', in Annual Review of Materials Science, vol. 10, Annual Review, Inc.,
Palo Alto, pp. 229-59.
Chou, T. W. and Kelly, A. (1980b) 'The effect of transverse shear on the
compressive strength of fiber composites', /. Mat. Sci., 15, 327.
Chou, T. W., Nomura, S. and Taya, M. (1980) 'A self-consistent approach to the
elastic stiffness of short-fiber composites',/. Comp. Mat., 14, 178.
Chou, T. W., Steward, B. and Bader, M. G. (1979) 'On the compression strength of
glass-epoxy composites', in New Developments and Applications in Composites,
D. Wilsdorf, ed., TMS-AIME, New York.
Dorey, G., Sidey, G. R. and Hutchings, J. (1978) 'Impact properties of carbon
fibre/Kevlar 49 fibre hybrid composites', Composites, 9, 25-32.
Eshelby, J. D. (1957) 'The determination of the elastic field of an ellipsoidal
inclusion, and related problem', Proc. Royal Soc. London, 241, 376-96.
Fernando, G., Dickson, R. F., Adam, T., Reiter, H. and Harris, B. (1988) 'Fatigue
542 References

behaviour of hybrid composites: I carbon/Kevlar hybrids', J. Mat. Sci., 23,

3732-43.
Fischer, S. and Marom, G. (1987) 'The flexural behavior of aramid fiber hybrid
composite materials', Comp. Sci. Tech., 28, 1—24.
Fukuda, H. (1983a) 'Mechanics of hybrid composites, part I', Trans. Japan Soc.
Composite Mat., 9, 76-80.
Fukuda, H. (1982b) 'Mechanics of hybrid composites, part IF, Trans. Japan Soc.
Composite Mat., 9, 118-23.
Fukuda, H. (1983c) 'Mechanics of hybrid composites, part IIP. Trans. Japan Soc.
Composite Mat., 9, 153-9.
Fukuda, H. and Chou, T. W. (1981) 'Stress concentrations around a discontinuous
fiber in a hybrid composite sheet', Trans. Japan Soc. Composite Mat., 7, 37-42.
Fukuda, H. and Chou, T. W. (1982a) 'Monte Carlo simulation of the strength of
hybrid composites', /. Comp. Mat., 16, 357.
Fukuda, H. and Chou, T. W. (1982b) 'A statistical approach to the strength of
hybrid composites', Proceedings of the Fourth International Conference on
Composite Materials, Japan Society for Composite Materials, Tokyo, pp. 1145-52.
Fukuda, H. and Chou, T. W. (1983) 'Stress concentration in a hybrid composite
sheet',/. Appl. Mech., 50, 845-8.
Fukunaga, H., Chou, T. W. and Fukuda, H. (1984) 'Strength of intermingled hybrid
composites', /. Reinforced Plastics Composites, 3, 145-60.
Fukunaga, H., Chou, T. W. and Fukuda, H. (1989) 'Probabilistic strength analyses
of interlaminated hybrid composites', Comp. Sci. Tech., 35, 331.
Fukunaga, H., Chou, T. W., Peters, P. M. W. and Schulte, K. (1984a) 'Probabilistic
failure strength analyses of graphite/epoxy cross-ply laminates', /. Comp. Mat.,
18, 339-56.
Fukunaga, H., Chou, T. W., Schulte, K. and Peters, P. M. W. (1984b) 'Probabilistic
initial failure strength of hybrid and non-hybrid laminates', J. Mat. Sci., 19, 3546.
Gruber, M. B., Overbeeke, J. L. and Chou, T. W. (1982) 'A reusable sandwich
beam concept for composite compression test', /. Comp. Mat., 16, 162-71.
Gruber, M. B. and Chou, T. W. (1983) 'Elastic properties of intermingled hybrid
composites', Polymer Composites, 4, 265-9.
Hancox, N. L. and Wells, H. (1973) 'Izod impact properties of carbon-fibre/glass-
fibre sandwich structures', Composites, 4, 26-30.
Harlow, D. G. (1983) 'Statistical properties of hybrid composites', Proc. Royal Soc,
A389, 67-100.
Harlow, D. G. and Phoenix, S. L. (1978a) 'The chain-of-bundles probability model
for the strength of fibrous materials I: analysis and conjectures', J. Comp. Mat.,
12, 195.
Harlow, D. G. and Phoenix, S. L. (1978b) 'The chain-of-bundles probability model
for the strength of fibrous materials II: a numerical study of convergence', J.
Comp. Mat., 12,314.
Harris, S. J. and Bradley, P. D. (1976) Proceedings of the First International
Conference on Composite Materials, pp. 327-35.
Harris, B. and Bunsell, A. R. (1975) 'Impact properties of glass-fibre/carbon fibre
hybrid composites', Composites, 6, 197-201.
Hayashi, T. (1972) 'On the improvement of mechanical properties of composites by
hybrid composition', Proceedings of the Eighth International Reinforced Plastics
Conference, paper 22.
 References 543

Hedgepeth, J. M. (1961) 'Stress concentration infilamentarystructures', NASA TN

D-882.
Jang, B. Z., Chen, L. C , Wang, C. Z., Lin, H. T. and Zee, R. H. (1989) 'Impact
resistance and energy absorption mechanisms in hybrid composites', Comp. Sci.
Tech., 34,305-35.
Ji, X. (1982) 'On the hybrid effect and fracture mode of interlaminated hybrid
composites', in Proceedings of the Fourth International Conference on Composite
Materials, T. Hayashi, K. Kawata and S. Umekawa, eds., Japan Society for
Composite Materials, Tokyo, pp. 1137-44.
Ji, X., Hsiao, G. C. and Chou, T. W. (1981) 'A dynamic explanation of the hybrid
effect',/. Comp. Mat., 15,443-61.
Kalnin, I. L. (1972) 'Evaluation of unidirectional glass-graphite fiber/epoxy resin
composites', Composite Materials Testing and Design {Second Conference), ASTM
STP 497, American Society for Testing and Materials, Philadelphia,
pp.551-63.
Kenaga, D., Doyle, J. F. and Sun, C. T. (1987) 'The characterization of
boron/aluminum composite in nonlinear range as an orthotropic elastic-plastic
material', /. Comp. Mat., 21, 516-31.
Kirk, J. N., Munro, M. and Beaumont, P. W. R. (1978) 'The fracture energy of
hybrid carbon and glass composites', /. Mat. Sci., 13, 2197-204.
Kretsis, G. (1987) 'A review of the tensile, compressive, flexural and shear
properties of hybrid fiber-reinforced plastics', Composites, 18, 13-23.
McColl, I. R. and Morley, J. G. (1977) 'Crack growth in hybrid fibrous composites',
/. Mat. Sci., 12, 1165-75.
McCullough, R. L. and Peterson, J. M. (1977) 'Property optimization analysis for
multicomponent (hybrid) composites', in Developments in Composite Materials -
1, G. S. Holister, ed., Applied Science Publisher, London.
Marissen, R. (1984) 'Flight simulation behavior of aramid reinforced aluminum
laminates (ARALL)', Eng. Fracture Mech., 19, 261-77.
Marissen, R., Trautmann, K. H., Foth, J. and Nowack, H. (1984) 'Microcrack
growth in aramid reinforced aluminum laminates (ARALL)', in Fatigue 84,
Proceedings 2nd International Conference on Fatigue and Fatigue Thresholds, C. J.
Beevers, ed. vol. II. EMAS Ltd, Warley, UK, pp. 1081-91.
Marom, G. and Chen, E. J. H. (1987) 'Asymmetric hybrid composite: a design
concept to improve flexural properties of Kevlar aramid composites', /. Comp.
Sci. Tech., 29, 161-8.
Marom, G., Fischer, S., Tuler, F. R. and Wagner, H. (1978) 'Hybrid effects in
composites',/. Mat. Sci., 13, 1419-26.
Mori, T. and Tanaka, K. (1973) 'An average stress in matrix and average elastic
energy of materials with misfitting inclusions', Ada Met., 21, 571-4.
Mueller, L. N., Prohaska, J. L. and Davis, J. W. (1985) 'ARALL (aramid aluminum
laminates): introduction of a new composite material', Proceedings AIAA
Aerospace Engineering Conference, AIAA, New York, AIAA paper no. 85-0846.
Nomura, S. and Chou, T. W. (1984) 'Bounds of elastic moduli of multiphase
short-fiber composites',/. Appl. Mech., 51, 540.
Phillips, L. N. (1976) 'The hybrid effect-does it exist?' Composites, 7, 7-8.
Pitkethly, M. J. and Bader, M. G. (1987) 'Failure modes of hybrid composites
consisting of carbon fiber bundles dispersed in a glass fiber epoxy resin matrix', /.
Phys. D: Appl. Phys., 20, 315-22.
544 References

Renton, W. J. (ed.) (1978) Hybrids and Selected Metal—Matrix Composites: A

State-of-the-Art Review, AIAA, New York.
Rosen, B. W. (1964) Tensile failure of fibrous composites', AIAA J., 2, 1985.
Rybicki, E. and Kanninen, M. (1978) 'Fracture mechanics of non-metallic hybrid
composites', in Hybrids and Selected Metal-Matrix Composites: A State-of-the-Art
Review, W. J. Renton, ed., AIAA, New York, pp. 53-65.
Summerscales, J. and Short, D. (1978) 'Carbon fibre and glass fibre hybrid
reinforced plastics', Composites, 9, 157-66.
Sun, C. T. and Luo, J. (1985) 'Failure loads for notched graphite/epoxy laminates
with a softening strip', Comp. Sci. Tech., 22, 121-33.
Sun, C. T. and Norman, T. L. (1988) 'Design of laminated composite with
controlled-damage concept', Proceedings of American Society for Composites, 3rd
Technical Conference, Technomic Pub. Co., Lancaster, PA, pp. 485-9.
Sun, C. T. and Rechak, S. (1988) 'Effect of adhesive layers on impact damage in
composite laminates', in Composite Materials Testing and Design, 8th Conference,
ASTM STP 972, J. D. Whitcomb, ed., American Society for Testing and
Materials, Philadelphia.
Takahashi, K. and Chou, T. W. (1987) 'Non-linear deformation and failure behavior
of carbon/glass hybrid laminate', /. Comp. Mat., 21, 396—420.
Taya, M. and Chou, T. W. 'On two kinds of ellipsoidal inhomogeneities in an
infinite elastic body: an application to a hybrid composite', Int. J. Sol. Struct., 17,
553-63.
Vogelesang, L. B. and Gunnink, J. W. (1986) 'ARALL, a materials challenge for
the next generation of aircraft', Mat. & Design, 7, 287-300.
Wagner, H. D. and Marom, G. (1982) 'On composition parameters for hybrid
composite materials', Composites, 13, 18.
Walton, P. L. and Majumdar, A. J. (1975) Composites, 6, 209-16.
Wells, H. and Hancox, N. L. (1971) 'Stiffening and strengthening GRP beams with
CFRP', Composites, 2, 147-51.
Yau, L. N. and Chou, T. W. (1989) 'Analysis of hybrid effect in unidirectional
composites under longitudinal compressions', Composite Structures, 12, 27-37.
Zweben, C. (1977) 'Tensile strength of hybrid composites', /. Mat. Sci., 12, 1325-37.

Chapter 6
Bishop, S. M. (1989) 'Strength and failure of woven carbon-fibre reinforced plastics
for high performance applications', in Textile Structural Composites, T. W. Chou
and F. K. Ko, eds., Elsevier Science Publishers B.V., Amsterdam, pp. 173-207.
Byun, J. H. and Chou, T. W. (1989) 'Modeling and characterization of textile
structural composites: a review', /. Strain Analysis, 2A, 253—62.
Chang, L. W., Yau, S. S. and Chou, T. W. (1987) 'Notched strength of woven fabric
composites with moulded-in holes', Composites, 18, 233-41.
Chou, T. W. (1985) 'Characterization and modeling of textile structural composites:
an overview', Proceedings of the European Conference on Composite Materials,
AEMC, Bordeaux, pp. 133-7.
Chou, T. W. (1986) 'Strength and failure behavior of textile structural composites',
Proceedings of the American Society for Composites First Technical Conference,
Technomic Pub. Co., Lancaster, PA, p. 104.
Chou, T. W. (1989a) 'Properties of woven fabric composites', in Encyclopedia of
Materials Science and Engineering and Concise Subject Encyclopedias, Pergamon
Press, Oxford, p. 292.
 References 545

Chou, T. W. (1989b) 'Mechanics of two-dimensional woven fabric composites', in

Mechanical Behavior and Properties of Composite Materials, Technomic
Pub. Co., Lancaster, PA, pp. 131-50.
Chou, T. W. and Ishikawa, T. (1989) 'Analysis and modeling of two-dimensional
fabric composites', in Textile Structural Composites, T. W. Chou and F. K. Ko,
eds., Elsevier Science Publishers B.V., Amsterdam, pp. 209—64.
Chou, T. W. and Ko, F. K. (1989) Textile Structural Composites, Elsevier Science
Publishers B.V., Amsterdam.
Curtis, P. T. and Bishop, S. M. (1984) 'An assessment of the potential of woven
carbon fibre-reinforced plastics for high performance applications', Composites,
15, 259-65.
Douglass, W. A. (1964) Braiding and Braiding Machinery, Centrex Pub. Co.,
Eindhoven.
Dow, N. F. (1969) Triaxial Fabric, U.S. Patent 3446251, May 1969.
Dow, N. F. (1982) 'Studies of woven fabric reinforced composites for automotive
applications', Tech. Final Rep., MSC TFR 1301/8101, Materials Science Corp.,
Springhouse, Pennsylvania.
Dow, N. F. and Tranfield, G. (1970) 'Preliminary investigations of feasibility of
weaving triaxial fabrics (Doweave)', Text. Res. /., 40, 986-98.
Ghasemi Nejhad, M. N. and Chou, T. W. (1990a) 'Compression behavior of woven
carbonfibre-reinforcedepoxy composites with moulded-in and drilled holes',
Composites, 21, 33-40.
Ghasemi Nejhad, M. N. and Chou, T. W. (1990b) 'A model for the prediction of
compressive strength reduction of composite laminates with molded-in holes', /.
Comp. Mat. 24,236-55.
Hahn, H. T. and Tsai, S. W. (1973) 'Nonlinear elastic behavior of unidirectional
composite laminate',/. Comp. Mat., 7, 102-18.
Hearle, J. W. S. (1989) 'Mechanics of yarns and nonwoven fabrics', in Textile
Structural Composites, T. W. Chou and F. K. Ko, eds., Elsevier Science
Publishers B.V., Amsterdam, pp. 27-65.
Hearle, J. W. S., Grosberg, P. and Backer, S. (1969) Structural Mechanics of Fibers,
Yarns, and Fabrics, vol. 1, Wiley-Interscience, New York.
Ishikawa, T. (1981) 'Anti-symmetric elastic properties of composite plates of satin
weave cloth', Fiber Sci. Tech., 15, 127-45.
Ishikawa, T. and Chou, T. W. (1982a) 'Elastic behavior of woven hybrid
composites', /. Comp. Mat., 16, 2-19.
Ishikawa, T. and Chou, T. W. (1982b) 'Stiffness and strength behavior of woven
fabric composites', J. Mat. Sci., 17, 3211-20.
Ishikawa, T. and Chou, T. W. (1982c) 'Stiffness and strength properties of woven
fabric composites', in Proceedings of the Fourth International Conference on
Composite Materials, Japan Society for Composite Materials, Tokyo, pp. 489-96.
Ishikawa, T. and Chou, T. W. (1983a) 'In-plane thermal expansion and thermal
bending coefficients of fabric composites', J. Comp. Mat., 17, 92-104.
Ishikawa, T. and Chou, T. W. (1983b) 'One-dimensional analysis of woven fabric
composites', AIAA J., 21, 1714.
Ishikawa, T. and Chou, T. W. (1983c) 'Nonlinear behavior of woven fabric
composites',/. Comp. Mat., 17, 399-413.
Ishikawa, T. and Chou, T. W. (1983d) Thermoelastic analysis of hybrid fabric
composite',/. Mat. Sci., 18,2260-8.
Ishikawa, T., Koyama, K. and Kobayashi, S. (1977) 'Elastic moduli of carbon-epoxy
composites and carbon fibers', /. Comp. Mat., 11, 332—44.
546 References

Ishikawa, T., Matsushima, M., Hayashi, Y. and Chou, T. W. (1985) 'Experimental

confirmation of the theory of elastic moduli of fabric composites', /. Comp. Mat.,
19, 443-58.
Jones, R. M. (1975) Mechanics of Composite Materials, McGraw-Hill, New York.
Kimpara, I., Hamamoto, A. and Takehana, M. (1977) Trans. Japan Soc. Comp.
Mat., 3,21.
Lekhnitskii, S. G. (1963) Theory of Elasticity of an Anisotropic Elastic Body,
Holden-Day, San Francisco.
Lord, P. R. and Mohamed, M. H. (1982) Weaving: Conversion of Yarn to Fabric,
2nd edn, Merrow Publishing Company, Durham, UK.
Mody, P. B., Chou, T. W. and Friedrich, K. (1988) 'Effect of testing conditions and
microstructure on the sliding wear of graphite fiber/PEEK matrix composites', /.
Mat. ScL, 23, 4319-30.
Mody, P. B., Chou, T. W. and Friedrich, K. (1989) 'Abrasive wear behavior of
unidirectional and woven graphite fiber/PEEK composites', in Test Methods and
Design Allowables for Fiber Composites, ASTM STP 1003, C. C. Chamis, ed.,
American Society for Testing and Materials, Philadelphia, Pennsylvania, p. 75.
Rogers, K. F., Kingston-Lee, D. M., Phillips, L. N., Yates, B., Chandra, M. and
Parker, S. F. H. (1981) 'The thermal expansion of carbonfibre-reinforcedplastics,
part 6. The influence of fibre weave in fabric reinforcement', J. Mat. Sci., 16,
2803-18.
Rogers, K. F., Phillips, L. N., Kingston-Lee, D. M., Yates, B., Overy, M. J.,
Sargent, J. P. and McCalla, B. A. (1977) 'The thermal expansion of carbon
fibre-reinforced plastics, part 1. The influence of fibre type and orientation', /.
Mat. ScL, 12, 718-34.
Rosen, B. W., Chatterjee, S. N. and Kibler, J. J. (1977) 'An analysis model for
spatially oriented fiber composites', ASTM STP 617, Composite Materials: Testing
and Design (Fourth Conference), American Society for Testing and Materials,
Philadelphia, pp. 243-54.
Scardino, F. L. (1989) 'An introduction to textile structures and their behavior', in
Textile Structural Composites, T. W. Chou and F. K. Ko, eds., Elsevier Science
Publishers B.V., Amsterdam, pp. 1-26.
Scardino, F. L. and Ko, F. (1981) 'Triaxial woven fabrics Part I: behavior under
tensile, shear, and burst deformation', Text. Res. J., 51, 80-9.
Schulte, K., Reese, E. and Chou, T. W. (1987) 'Fatigue behavior and damage
development in woven fabric and hybrid fabric composites', Proceedings of the
Sixth International Conference and Second European Conference on Composite
Materials, vol. 4, Elsevier Applied Science, London, pp. 89-99.
Schwartz, P. (1984) A mathematical analysis of a fabric having non-orthogonal
interfacings using strain energy methods', Fiber Sci. Tech., 20, 273-82.
Schwartz, P., Fornes, R. E. and Mohamed, M. H. (1982) 'An analysis of the
mechanical behavior of triaxial fabrics and the equivalency of conventional
fabrics', Text Res. J., 52, 388-94.
Schwartz, P., Rhodes, T. and Mohamed, M. H. (1982) Fabric Forming Systems,
Noyes Publications, Park Ridge, New Jersey.
Skelton, J. (1971) Triaxially woven fabrics - their structure and properties, Text.
Res.J., 41,637-47.
Thomas, D. G. B. (1971) An Introduction to Warp Knitting, Merrow, Watford, UK.
Wray, G. R. and Vitols, R. (1982) 'Advances in stitch-bonding, warp- and
weft-knitting systems, and automated knitwear manufacture', in Contemporary
Textile Engineering, F. Happey, ed., Academic Press, London, pp. 375-409.
 References 547

Yang, J. M. and Chou, T. W. (1986) 'Performance optimization of woven fabric

composites for printed circuit boards', in Electronic Packaging Materials Science,
II, Symposia Proceedings vol. 72, Materials Research Society, Pittsburgh, pp.
163-73.
Yang, J. M. and Chou, T. W. (1987) 'Performance maps of textile structural
composites', Proceedings of the Sixth International Conference on Composite
Materials, vol. 5, Elsevier Applied Science, London, p. 579.
Yang, J. M. and Chou, T. W. (1989) Thermo-elastic analysis of triaxial woven fabric
composites', in Textile Structural Composites, T. W. Chou and F. K. Ko, eds.,
Elsevier Science Publishers B.V., Amsterdam, pp. 265-77.
Yates, B., Overy, M. J., Sargent, J. P., McCalla, B. A., Kingston-Lee, D. M.,
Phillips, L. N. and Rogers, K. F. (1978) 'The thermal expansion of carbon
fibre-reinforced plastics, part 2. The influence of fibre volume fraction', /. Mat.
ScL, 13,433-40.
Yau, S. S. and Chou, T. W. (1988) 'Strength of woven fabric composites with drilled
and molded holes', in Composite Materials Testing and Design {Eighth
Conference), ASTM STP 972, J. D. Whitcomb, ed., ASTM, Philadelphia, pp.
423-37.
Yokoyama, A., Fujita, A., Kobayashi, H., Hamada, H. and Maekawa, Z. (1989) 'A
new braiding process - robotised braiding mechanism', in Materials and
Processing - Move into the 90's, S. Benson, T. Cook, E. Trewin and R. M.
Turner, eds., Elsevier, Amsterdam, pp. 87-99.
Zweben, C. and Norman, J. C. (1976) 'Kevlar 49/Thornel 300 hybrid fabric
composites for aerospace application', SAMPE Quarterly, 1, 1-10.

Chapter 7
Annual Book of ASTM Standards, Part 10, Practice E399, 'Plane-strain fracture
toughness of metallic materials', American Society for Testing and Materials,
Philadelphia, PA.
Ashby, M. F., Gandi, C. and Taplin, D. M. R. (1979) 'Fracture-mechanism maps
and their construction for F.C.C. metals and alloys', Ada MetalL, 27, 699-729.
ASTM Standards and Literature References for Composite Materials, D 4255-83,
'In-plane shear properties of composite laminates', American Society for Testing
and Materials, Philadelphia, PA.
ASTM Standards and Literature References for Composite Materials, D 2344-84,
'Apparent interlaminar shear strength of parallel fiber composites by short-beam
method', American Society for Testing and Materials, Philadelphia, PA.
Brunnschweiler, D. (1954) 'The structure and tensile properties of braids', /. Text.
Instrum., 45, T55-T77.
Byun, J. H., Leach, B. S., Stroud, S. S. and Chou, T. W. (1990a) 'Structural
characteristics of three-dimensional angle-interlock woven fabric preforms', in
Processing of Polymers and Polymeric Composites, ASME, MD-vol. 19 American
Society for Mechanical Engineers, New York, pp. 177.
Byun, J. H., Du, G. W. and Chou, T. W. (1991) 'Analysis and modeling of 3-D
textile structural composites', ACS Symposium Series 457, American Chemical
Society, Washington D.C., pp. 22-33.
Byun, J. H., Gillespie, J. W. and Chou, T. W. (1989) 'Mode II delamination of
three-dimensional textile structural composites', Proceedings of the American
Society for Composites, 4th Technical Conference, Technomic Pub. Co.,
Lancaster, PA, pp. 287-96.
548 References

Byun, J. H., Gillespie, J. W. and Chou, T. W. (1990b) 'Mode I delamination of a

three-dimensional fabric composite', /. Comp. Mat., 24, 497.
Byun, J. H., Whitney, T. J., Du, G. W. and Chou, T. W. (1991) 'Analytical
characterization of two-step braided composites' /. Comp. Mat. in press.
Chou, T. W. (1989) 'Structure-performance maps', in Encyclopedia of Materials
Science and Engineering and Concise Subject Encyclopedias, Pergamon Press,
Oxford, p. 261.
Chou, T. W., McCullough, R. L. and Pipes, R. B. (1986) 'Composites', Sci. Am.,
254, 193-203.
Chou, T. W. and Yang, J. M. (1986) 'Structure-performance maps of polymeric,
metal and ceramic matrix composites', Metall. Trans. A, 17A, 1547-9.
Cole, P. M. (1988) 'Three-dimensional structures of interlocked strands', U.S.
Patent 4,737,399.
Cooper, G. A. and Kelly, A. J. (1967) 'Tensile properties of fiber-reinforced
metals: fracture mechanics', Mech. Phys. Solids, 15, 279-97.
Crane, R. M. and Camponeschi, E. T. (1986) 'Experimental and analytical
characterization of multidimensionally braided graphite/epoxy composites', Exp.
Mech., 26,259.
Dexter, H. B. and Funk, J. G. (1986) 'Impact resistance and interlaminar fracture
toughness of through-the-thickness reinforced graphite epoxy', AIAA Paper
86-1020-CP, pp. 700-9.
Dhingra, A. K., Champion, A. R. and Krueger, W. H. (1975) 'Fiber FP reinforced
aluminum and magnesium composites', in Proceedings of the First Metal Matrix
Composite Workshop, Paper Number C-501, Institute for Defense Analyses,
Washington, D.C., September.
Dow, N. F. (1984) Proceedings of the Fiber Society /SAMPE Conference on High
Performance Textile Structures, Philadelphia College of Textile and Science,
Philadelphia, PA.
Du, G. W., Popper, P. and Chou, T. W. (1989) 'Analysis and automation of
two-step braiding', FIBER-TEX 88, NASA Conference Publication no. 3038, pp.
217-33.
Du, G. W., Popper, P. and Chou, T. W. (1991) 'Analysis of 3D textile preform for
multidirectional reinforcement of composites', /. Mat. Sci. in press.
Florentine, R. (1982) 'Apparatus for weaving a three-dimensional article', U.S.
Patent 4,312,261.
Fowser, S. W. (1986) 'The behavior of orthogonal fabric composites', M.S. thesis,
University of Delaware.
Fowser, S. W. and Chou, T. W. (1989) 'Simplified Green's functions for mode I and
II cracks', Int. J. Fracture, 39, 301-21.
Fowser, S. W. and Chou, T. W. (1990a) 'Integral equations solution for reinforced
mode I cracks opened by internal pressure', /. Appl. Mech., in press.
Fowser, S. W. and Chou, T. W. (1990b) 'Numerical integration of Green's functions
for an edge-loaded infinite strip', Computers and Structures, in press.
Frost, H. J. and Ashby, M. F. (1982) Deformation-Mechanism Maps, Pergamon
Press, Oxford.
Gandi, C. and Ashby, M. F. (1979) 'Fracture mechanism maps for materials which
cleave: F.C.C. and H.C.P. metals and ceramics', Acta Metall., 27,
1565-1602.
Guenon, V. A., Chou, T. W. and Gillespie, J. W. (1987) 'Interlaminar fracture
toughness of a three-dimensional fabric composite', Proceedings of the Society of
 References 549

Manufacturing Engineers, EM87-551, 1-17, Society of Manufacturing Engineers,

Dearborn, Michigan.
Guenon, V. A., Chou, T. W. and Gillespie, J. W. (1989) 'Toughness properties of a
three-dimensional carbon-epoxy composite', /. Mat. ScL, 24, 4168-75.
Guess, T. R. and Reedy, Jr., E. D. (1985) 'Comparison of interlocked fabric and
laminated fabric Kevlar 49/epoxy composites', /. Comp. Tech. Res., 7,
136-42.
Hearle, J. W. S., Grosberg, P. and Backer, S. (1969) Structural Mechanics of Fibers,
Yarns, and Fabrics, vol. 1, Wiley-Interscience, New York, p. 80.
Hunston, D. H. (1984) Comp. Tech. Rev., 6, 176.
Iosipescu, N. (1967) 'New accurate procedures for single shear testing of metals', /.
Mat., 2,537-66.
Ishikawa, T. and Chou, T. W. (1982) 'Stiffness and strength behavior of woven
fabric composites', /. Mat. ScL, 17, 3211-20.
Kelly, A. and Macmillan, N. H. (1986) Strong Solids, 3rd edn, Clarendon Press,
Oxford.
Kies, J. A. (1962) 'Maximum strains in the resin of fiberglass composites', U.S.
Naval Research Laboratory Report NRL 5752.
Ko, F. K. (1989a) Three-dimensional fabrics for composites', in Textile Structural
Composites, T. W. Chou and F. K. Ko, eds., Elsevier Science Publishers B.V.,
Amsterdam, pp. 129-71.
Ko, F. K. (1989b) 'Preform fiber architecture for ceramic-matrix composites',
Ceramic Bull., 68, 401-14.
Ko, F. K. (1986) 'Tensile strength and modulus of a 3-D braided composite' in
Composite Materials: Testing and Design, ASTM STP 893, American Society for
Testing and Materials, Philadelphia, PA, p. 392.
Ko, F. K. and Pastore, C. M. (1985) 'Structure and properties of an integrated 3-D
fabric for structural composites', ASTM STP 864, American Society for Testing
and Materials, Philadelphia, PA, p. 428.
Ko, F. K., Pastore, C. M., Yang, J. M. and Chou, T. W. (1986) 'Structure and
properties of multilayer, multidirectional warp knit fabric reinforced composites',
in Composites "86: Recent Advances in Japan and the United States, Japan Society
for Composite Materials, Tokyo.
Ko, F. K., Soebroto, H. B. and Lei, C. (1988) '3-D net shaped composites by the
2-step braiding process', Proceedings of 33rd International SAMPE Symposium,
vol. 33, SAMPE International Business Office, Covina, California, pp. 912-21.
Kregers, A. F. and Teters, G. A. (1982) 'Structural model of deformation of
anisotropic three-dimensionally reinforced composites', Mech. Comp. Mat., 1, 14.
Lekhnitskii, S. G. (1963) Theory of Elasticity of an Anisotropic Elastic Body,
Holden-Day, San Francisco.
Li, W. and El Shiekh, A. (1988) 'The effect of processes and processing parameters
on 3-D braided preforms for composites', SAMPE Quarterly, 19, 22-8.
Li, W., Kang, T. J., and El Shiekh, A. (1988) 'Structural mechanics of 3-D braided
preforms for composites, part I: Geometry of fabric produced by 4-step process',
in Proceedings of Fiber-Tex'87 Conference, NASA Conference Publication.
Liu, C. H. and Chou, T. W. (1989) 'Mode II interlaminar fracture toughness of
three-dimensional textile structural composites', Proceedings of the 4th Japan-
U.S. Conference on Composite Materials, Technomic Pub. Co., 981.
Ma, C. L., Yang, J. M. and Chou, T. W. (1986) 'Elastic stiffness of three-
dimensional braided textile structural composites', in Composite Materials, Testing
550 References

and Design (Seventh Conference), ASTM STP 893, American Society for Testing
and Materials, Philadelphia, PA, pp. 404-21.
Majidi, A. P. and Chou, T. W. (1986) 'Impact tolerance of braided alumina fiber
reinforced aluminum composites', Proceedings of the 31st International SAMPE
Symposium, SAMPE International Business Office, Covina, California.
Majidi, A. P. and Chou, T. W. (1987) 'Structure-reliability studies of three-
dimensionally braided metal matrix composites', in Proceedings of the Sixth
International and Second European Conference on Composite Materials, vol. 2,
Elsevier Applied Science, London.
Majidi, A. P., Remond, O. G. and Chou, T. W. (1987) 'The effect of fiber
architecture on the mechanical performance of metal matrix composites',
Proceedings of the 2nd Annual Conference of the American Society for Composites,
Technomic Pub. Co., Lancaster, Pennsylvania, p. 371.
Majidi, A. P., Yang, J. M. and Chou, T. W. (1986) 'Toughness characteristics of
three-dimensionally braided AI203/Al-Li composites', in Interfaces in Metal-
Matrix Composites, A. K. Dhingra and S. G. Fishman, eds., The Metallurgical
Society, Warrendale, PA, pp. 27-44.
Majidi, A. P., Yang, J. M. and Chou, T. W. (1988) 'Mechanical behavior of
three-dimensional braided metal-matrix composites', Testing Technology of Metal
Matrix Composites, ASTM STP 964, American Society for Testing and Materials,
Philadelphia, PA, p. 31.
Majidi, A. P., Yang, J. M., Pipes, R. B. and Chou, T. W. (1985) 'Mechanical
behavior of three-dimensional woven fiber composites', in Proceedings of the Fifth
International Conference on Composite Materials, The Metallurgical Society of
AIME, Warrendale, PA, pp. 1247-65.
Masters, J. E. (1987) 'Characterization of impact development in graphite epoxy
laminates', ASTM, STP 948, American Society for Testing and Materials,
Philadelphia, PA, pp. 238-58.
Mignery, L. A., Tan, T. M. and Sun, C. T. (1985) 'The use of stitching to suppress
delamination in laminated composites', ASTM STP 876, American Society for
Testing and Materials, Philadelphia, PA, pp. 371-85.
Ogo, Y. (1987) 'The effect of stitching on in-plane and interlaminar properties of
carbon-epoxy fabric laminates', M.S. Thesis, University of Delaware,
Peirce, F. T. (1937) 'The geometry of cloth structure', J. Text. Instrum., 28,
T45-T96.
Popper, P. and McConnell, R. F. (1988) 'Complex shaped braided structures', U.S.
Patent 4,719,837.
Remond, G. O. (1987) 'Characterization and modeling of 3-D braided metal matrix
composites', M.S. Thesis, University of Delaware.
Rybicki, E. F. and Kanninen, M. F. (1977) 'A finite element calculation of stress
intensity factors by a modified crack closure integral', Eng. Fracture Mech., 9,
931-8.
Simonds, R. A., Stinchcomb, W. and Jones, R. M. (1988) 'Mechanical behavior of
braided composite materials', ASTM STP 972, American Society for Testing and
Materials, Philadelphia, PA, p. 438.
Steeger, USA, Inc. (1989) Production Program, Spartanburg, South Carolina.
Takahashi, K. and Chou, T. W. (1986) 'Modeling of the interfacial behavior of
flexible composites', in Interfaces in Metal-Matrix Composites, A. K. Dhingra and
S. G. Fishman, eds., The Metallurgical Society, Warrendale, PA, pp. 45-59.
 References 551

Tattersall, H. G. and Tappin, G. J. (1966) /. Mat. Sci. 1, 296.

Walrath, D. E. and Adams, D. F. (1983a) 'The Iosipescu shear test as applied to
composite materials', Exp. Mech., 23, 105-10.
Walrath, D. E. and Adams, D. F. (1983b) 'Analysis of the stress state in an
Iosipescu shear test specimen', Technical Report UWME-DR-301-102-1,
University of Wyoming, Laramie, Wyoming.
Weller, R. D. (1985) 'AYPEX: a new method of composite reinforcement braiding,
3-D Composite Materials, NASA Conference Publication 2420.
Whitcomb, J. D. (1989) 'Three dimensional stress analysis of plain weave
composites', NASA Technical Memorandum 101672.
Whitney, J. M., Browning, C. E. and Hoogsteden, W. (1982) /. Reinf. Plastics
Comp., 1, 297.
Whitney, T. J. (1988) 'Analytical characterization of 3-D textile structural
composites using plane stress and 3-D lamination analogies', M.S. Thesis,
University of Delaware.
Whitney, T. J. and Chou, T. W. (1988) 'Modeling of elastic properties of 3-D textile
structural composites', Proceedings of the American Society for Composites, Third
Technical Conference, Technomic Pub. Co., Lancaster, Pennsylvania, p. 427.
Whitney, T. J. and Chou, T. W. (1989) 'Modeling of 3-D angle-interlock textile
structural composites',/. Comp. Mat., 23, 890-911.
Yang, J. M. and Chou, T. W. (1989) 'Thermo-elastic analysis of triaxial woven fabric
composites', in Textile Structural Composites, T. W. Chou and F. K. Ko, eds.,
Elsevier Science Publishers B.V., Amsterdam, pp. 265-77.
Yang, J. M., Ma, C. L. and Chou, T. W. (1986) 'Fiber inclination model of
three-dimensional textile structural composites', /. Comp. Mat., 20, 472-84.
Yau, S. S., Ko, F. and Chou, T. W. (1986) 'Flexural and axial compressive failures
of three-dimensionally braided composite I-beams', Composites, 17, 227.
Yoshida, H., Ogasa, T. and Hayashi, R. (1986) 'Statistical approach to the
relationship between ILSS and void content of CFRP', Comp. Sci. Tech., 25,
3-18.

Chapter 8
Akasaka, T. (1959-64) Various Reports/Bulletins, Faculty of Science and
Engineering, Chuo University, Tokyo.
Akasaka, T. (1989) 'Flexible composites', in Textile Structural Composites T. W.
Chou and F. Ko, eds., Elsevier Science Publishers, Amsterdam, pp. 279-330.
Akasaka, T. and Hirano, M. (1972) Comp. Mat. Struc, 1, 70.
Akasaka, T. and Yoshida, N. (1972) Proc. Intl. Conf. Mech. Behavior of Mater.,
Kyoto, Japan, vol. 5, pp. 187-97.
Alley, V. A. and Fairslon, R. W. (1972) 'Experiment investigation of strains in a
fabric under biaxial and shear forces', /. Aircraft, 9, 55.
Bert, C. W. and Kumar, M. (1981) 'Experiments on highly nonlinear elastic
composites', Proc. NCKU/AAS Int-Sym. in Eng. Sci. and Mech., National Chen
Kung Univ., Taiwan, vol. 2, pp. 1269-83.
Bert, C. W. and Reddy, J. N. (1982) 'Mechanics of bimodular composite structures',
in Mechanics of Composite Materials: Recent Advances, Proceedings of the
IUTAM Symposium, Virginia Polytechnic Institute, pp. 323-37.
552 References

Biderman, V. I., Gusliter, R. L., Sakharov, S. P., Nenakhov, B. V., Seleznev, I. I.

and Tsukerberg, S. M. (1963) 'Automobile tires, construction, design, testing and
usage', NASA, TT F-12, 382, 1969. (Original publication in Russian, State
Scientific and Technical Press for Chemical Literature, Moscow).
Bohm, F. (1966) 'Mechanik des Gurtelreifens', Ing-Arch., 35, 82-101.
Chamis, C. C. (1984) 'Simplified composite micromechanics equations for hygral,
thermal and mechanical properties', SAMPE Quarterly, 15, 14-23.
Chou, T. W. (1985) 'Characterization and modeling of textile structural composites:
an overview', Proceedings of First European Conference on Composite Materials,
A.E.M.C., Bordeaux, France, pp. 133-8.
Chou, T. W. (1989) 'Review:flexiblecomposites', /. Mat. ScL, 2A, 761-83.
Chou, T. W. (1990) 'Flexible composites', in International Encyclopedia of
Composites, VCH Publishers, New York.
Chou, T. W. and Takahashi, K. (1987) 'Nonlinear elastic behavior offlexiblefiber
composites', Composites, 18, 25.
Chou, T. W. and Yang, J. M. (1986) 'Structure-performance maps of textile
structural composites in polymeric, metal and ceramic matrices', Met. Trans. A,
17A, 1547.
Clark, S. K. (1963a) 'The plane elastic characteristics of cord-rubber laminates',
Textile Res. J., 33, 295-313.
Clark, S. K. (1963b) 'Internal characteristics of orthotropic laminates', Textile Res.
J., 33,935-53.
Clark, S. K. (1964) 'A review of cord-rubber elastic characteristics', Rubber Chem.
Tech., 37, 1365-90.
Clark, S. K. (1980) 'The role of textiles in pneumatic tires', in Mechanics of Flexible
Fibre Assembles, J. W. S. Hearle, J. J. Thwaites and J. Amirbayat, eds., Sijthoff
and Noordhoff, The Netherlands.
Clark, S. K. and Dodge, R. N. (1969) 'A load transducer for tire cord', SAE Paper
690521, Society of Automotive Engineers, Warrendale, PA.
Gough, V. E. (1968) 'Stiffness of cord and rubber constructions', Rubber Chem.
Tech., 41,988-1021.
Hearle, J. W. S., Grosberg, P. and Backer, S. (1969) Structural Mechanics of Fibers,
Yarns and Fabrics, vol. 1, Wiley-Interscience, New York.
Ishikawa, T. and Chou, T. W. (1983) 'Nonlinear behavior of woven fabric
composites', J. Comp. Mat., 17, 399.
James, H. M. and Guth, E. (1943) 'Theory of the elastic properties of rubber', /.
Chem. Phys., 11,455-81.
Jones, R. E. (1975) Mechanics of Composite Materials, McGraw-Hill, New York.
Kuo, C. M., Takahashi, K. and Chou, T. W. (1988) 'Effects of fiber waviness on the
nonlinear elastic behavior offlexiblecomposites', /. Comp. Mat., 12, 1004.
Lou, A. Y. C. and Walter, J. D. (1978) Tnterlaminar shear strain measurements in
cord-rubber composites', paper presented at SESA meeting, Wichita, Kansas,
May.
Luo, S. Y. and Chou, T. W. (1988) 'Finite deformation and nonlinear elastic
behavior offlexiblecomposites',/. Appl. Mech., 55, 149-55.
Modern Plastics Encyclopedia, Engineering Data Bank, McGraw-Hill, Inc., New
York.
Patterson, R. G. (1969) 'The measurement of cord tensions in tires', Rubber Chem.
Echnology, 42, 812.
Petit, P. H. and Waddoups, M. E. (1969) 'A method of predicting the nonlinear
behavior of laminated composites',/. Comp. Mat., 3,2-19.
 References 553

Reinhardt, H. W. (1976) 'On the biaxial testing and strength of coated fabrics', Exp.
Mech., 11,71.
Skelton, J. (1971) 'The biaxial stress-strain behavior of fabrics for air-supported
tents',/. Mat., J.M.L.S.A., 6, 656.
Stubbs, N. (1988) 'Elastic and inelastic response of coated fabrics to arbitrary loading
paths', in Textile Structural Composites T. W. Chou and F. K. Ko, eds., Elsevier
Science Publishers, Amsterdam, pp. 331-54.
Stubbs, N. and Thomas, S. (1984) 'A nonlinear elastic constitutive model for coated
fabrics', in Mechanics of Material, S. Nernat-Nasser, eds., vol. 3, Elsevier Science
Publishers, BV, Amsterdam, pp. 157-68.
Takahashi, K. and Chou, T. W. (1986) 'Modeling of the interfacial behavior of
flexible composites', in Interfaces in Metal-Matrix Composites, A. K. Dhingra and
S. G. Fishman, eds., The Metallurgical Society, Warrendale, PA, pp. 45-59.
Takahashi, K., Kuo, C. M. and Chou, T. W. (1986) 'Nonlinear elastic constitutive
equations offlexiblefiber composites', in Composites '86: Recent Advances in
Japan and the United States, Japan Society for Composite Materials, Tokyo, p.
389.
Takahashi, K., Yano, T., Kuo, C. M. and Chou, T. W. (1987) 'Effect of fiber
waviness on elastic moduli of fiber composites', Trans. Japan Fiber Soc, 43, 376.
Treloar, L. R. G. (1973) 'The elasticity and related properties of rubbers', Rep.
Prog. Phys., 36,755-826.
Walter, J. D. (1978) 'Cord-rubber tire composites: theory and application', Rubber
chem. Tech., 51,524.
Walter, J. D. and Hall, G. L. (1969) 'Cord load characteristics in bias and
belted-bias tires', SAE Paper 690522, Society of Automotive Engineers,
Warrendale, PA.
Walter, J. D. and Patel, H. P. (1979) "Approximate expressions for the elastic
constants of cord-rubber laminates', Rubber Chem. Tech., 52, 710-24.

Chapter 9
Adkins, J. E. and Rivlin, R. S. (1955) 'Large elastic deformation of isotropic
materials X. Reinforcements by inextensible cords', Phil. Trans. Royal Soc.
London, (A), 248, 201-23.
Aspden, R. M. (1986) 'Relation between structure and mechanical behavior of
fibre-reinforced composite materials at large strains', Proc. Royal Soc. London,
(A), 406, 287-98.
ASTM Standard D3518-76 (1982) 'Practice for in-plane shear stress-strain response
of unidirectional reinforced plastics' American Society for Testing and Materials,
Philadelphia.
Chou, T. W. (1989) 'Flexible composites', /. Mat. ScL, 24, 761-83.
Ericksen, J. L. and Rivlin, R. S. (1954) 'Large elastic deformations of homogeneous
anisotropic materials', /. Rational Mech. Analysis, 3 (3) 281-301.
Fung, Y. C. (1965) Foundations of Solid Mechanics, Prentice-Hall Inc., Englewood
Cliffs, NJ.
Fung, Y. C. (1977) A First Course in Continuum Mechanics, Prentice-Hall, Inc.,
Englewood Cliffs, N.J.
Fung, Y. C. (1981) Biomechanics: Mechanical Properties of Living Tissues, Springer
Verlag, New York.
Hahn, H. T. (1973) 'Nonlinear behavior of laminated composites', /. Comp. Mat., 7,
257-71.
554 References

Hahn, H. T. and Tsai, S. W. (1973) 'Nonlinear elastic behavior of unidirectional

composite laminae', /. Comp. Mat., 7, 102-18.
Humphrey, J. D. and Yin, F. C. P. (1987) 'A new constitutive formulation for
characterizing the mechanical behavior of soft tissues', Biophys. J., 52, 563-70.
Ishikawa, T. and Chou, T. W. (1983) 'Nonlinear behavior of woven fabric
composites',/. Comp. Mat., 17, 399-413.
Jones, R. S. and Morgan, H. S. (1977) 'Analysis of nonlinear stress-strain behavior
of fiber-reinforced composite materials', AIAA J., 15, 1669-76.
Kuo, C. M., Takahashi, K. and Chou, T. W. (1988) 'Effect of fiber waviness on the
nonlinear elastic behavior of flexible composites', /. Comp. Mat., 12, 1004.
Lai, W. M., Rubin, D. and Krempl, E. (1978) Introduction to Continuum
Mechanics, Pergamon Press, Oxford.
Luo, S. Y. (1988) 'Theoretical modeling and experimental characterization of
flexible composites', Ph.D. Dissertation, University of Delaware, Newark,
Delaware.
Luo, S. Y. and Chou, T. W. (1988a) 'Finite deformation and nonlinear elastic
behavior of flexible composites',/. Appl. Mech., 55, 149-55.
Luo, S. Y. and Chou, T. W. (1988b) 'Constitutive relations of flexible composites
under finite elastic deformation', in Mechanics of Composite Materials - 1988, G.
J. Dvorak and N. Laws, eds., ASME, AMD, New York, vol. 92, pp. 209-16.
Luo, S. Y. and Chou, T. W. (1989) 'Elastic behavior of laminated flexible
composites under finite deformation', in Micromechanics and Inhomogeneity - The
Toshio Mura Anniversary Volume, G. J. Weng, M. Taya and H. Abe, eds.,
Springer-Verlag, New York, pp. 243-56.
Luo, S. Y. and Chou, T. W. (1990a) 'Modeling of the nonlinear elastic behavior of
elastomeric flexible composites', in Composites: Chemical and Physicochemical
Aspects, T. L. Vigo and B. J. Kinzig eds., VCH Publishers, New York, in press.
Luo, S. Y. and Chou, T. W. (1990b) 'Finite deformation of flexible composites',
Proc. Royal Soc. London, (A), 429, 569-86.
Luo, S. Y., Kuo, C. M. and Chou, T. W. (1988) 'Theoretical modeling and
experimental characterization of flexible composites', Proceedings of the Fourth
Japan-United States Conference on Composite Materials, Technomic Pub. Co.,
Lancaster, Pennsylvania, pp. 885-74.
Malvern, L. E. (1969) Introduction to the Mechanics of a Continuous Medium,
Prentice-Hall, Inc., Englewood Cliffs.
Pagano, N. J. and Halpin, J. C. (1968) 'Influence of end constraint in the testing of
anisotropic bodies',/. Comp. Mat., 2, 18—31.
Petit, P. H. and Waddoups, M. E. (1969) 'A method of predicting the nonlinear
behavior of laminated composites', /. Comp. Mat., 3, 2-19.
Pipkin, A. C. and Rogers, T. G. (1971) 'Plane deformation of incompressible
fiber-reinforced materials', /. Appl. Mech., 38, 634-40.
Posfalvi, O. (1977) 'The Poisson ratio for rubber-cord composites', Rubber Chem.
Tech., 50,224-32.
Rivlin, R. S. (1948a) 'Large elastic deformation of isotropic materials I.
Fundamental Concepts', Phil. Trans. Royal Soc. London, (A), 240, 459-90.
Rivlin, R. S. (1948b) 'Large elastic deformation of isotropic materials IV. Further
developments of the general theory', Phil. Trans. Royal Soc. London, (A), 241,
379-97.
Rivlin, R. S. (1959) 'Mathematics and rheology, the 1958 Bingham Medal Address',
Phys. Today, 12, (5), 32-6.
 References 555

Rivlin, R. S. (1964) 'Networks of inextensible cords', in Nonlinear Problems of

Engineering, W. F. Ames, ed., Academic Press, New York.
Rivlin, R. S. (1970) 'Nonlinear continuum theories in mechanics and physics and
their applications', in Centro Internazionale Matematico Estivo, Ciclo, II and R. S.
Rivlin, eds., Edizioni Cremonese, Roma.
Rivlin, R. S. and Saunders, D. W. (1951) 'Large elastic deformation of isotropic
materials XII. Experiments on the deformation of rubber', Phil. Trans. Royal Soc.
London, (A), 243, 251-98.
Spencer, A. J. M. (1972) Deformation of Fibre Reinforced Materials, Clarendon
Press, Oxford.
Truesdell, C. (1966) Elements of Continuum Mechanics, Springer-Verlag, New York.
Whitney, J. M., Daniel, I. M. and Pipes, R. B. (1982) Experimental Mechanics of
Fiber Reinforced Composite Materials, The Society for Experimental Stress
Analysis, Brookfield Center, Connecticut.
 Author index

Adam, T. 277, 278, 540, 541 418, 422, 427, 437, 438, 439, 440,
Adams, D. F. 277, 431, 540, 551 544, 547, 548
Adkins, J. E. 475, 504, 553
Akasaka, T. 191, 444, 448, 449, 452, Camponeschi, E. T. 442, 548
456, 534, 551 Carlsson, L. A. 29, 527
Aker, S. 275, 541 Carrara, A. S. 85, 170, 530, 535
Akoz, A. Y. 54, 55, 78, 527, 528 Carslaw, H. S. 75, 527
Alley, V. A. 444, 551 Chamis, C. C. 33, 177, 231, 249, 250,
Anderson, R. M. 183, 534 251, 277, 469, 527, 535, 541, 552
Argon, A. S. 115, 118,533 Champion, A. R. 428, 548
Arridge, R. G. C. 187, 534 Chan, W. S. 275, 541
Arlington, M. 231, 540 Chandra, M. 333, 546
Ashbaugh,N. 170, 536 Chang, C. I. 182, 535
Ashby, M. F. 1, 2, 419, 526, 547, 548 Chang, L. W. 372, 544
Ashton, J. E. 29, 527 Chang, Y. P. 54, 527, 528
Aspden, R. M. 476, 553 Chatterjee, S. N. 362, 546
Aveston, J. 82, 84, 134, 135, 136, 231, Chawla, K. K. 29, 527
248, 251, 254, 256, 257, 258, 276, Chen, C. H. 94, 530
529, 540, 541 Chen, C. K.54,527
Chen, E. J. H. 278, 543
Backer, S. 286, 390, 392, 447, 545, 549, Chen, J. L. 275, 541
552 Chen, L. C. 276, 277, 543
Bader, M. G. 80, 98, 133, 136, 144, 204, Chen, P. E. 85, 94, 176, 187, 224, 225,
205, 207, 208, 216, 222, 226, 230, 229, 530, 535, 539
231, 261, 275, 278, 529, 534, 535, Cheng, C. M. 73, 527
540,541,543 Chi, Z. F. 106, 107, 108, 109, 110, 111,
Bailey, J. E. 136, 137, 138, 139, 140, 112, 114, 166, 167, 168, 530
144, 149, 207, 216, 230, 529, 530, Chou, P. C. 150, 533
532, 535 Chou, T. W. 1, 6, 8, 9, 10, 19, 21, 29,
Baker, R. M. 229, 534 33, 40, 46, 48, 54, 55, 58, 62, 65, 66,
Beaumont, P. W. R. 276, 277, 541, 543 67, 68, 69, 72, 73, 74, 76, 77, 78, 80,
Beran, M. 195, 534 81, 85, 88, 90, 91, 92, 93, 94, 96, 97,
Bert, C. W. 191,456,534,551 103, 106, 107, 108, 109, 110, 111,
Biderman, V. I. 444, 552 112, 113, 114, 133, 136, 145, 146,
Bishop, S. M. 371, 372, 544, 545 147, 148, 149, 150, 166, 167, 169,
Bjeletich, J. G. 160, 529 171, 172, 174, 175, 176, 178, 179,
Blumentritt, B. F. 182, 225, 534 180, 181, 184,191, 193, 195, 197,
Bohm, F. 453, 552 198, 199, 200, 201, 202, 203, 206,
Bowyer, W. H. 204, 205, 216, 226, 534 207, 208, 209, 210, 211, 212, 213,
Bradley, P. D. 231, 542 214, 215, 216, 217, 222, 223, 225,
Bradley, W. L. 165, 529 226, 229, 230, 231, 232, 233, 234,
Browning, C. E. 435, 551 238, 239, 241, 242, 243, 244, 246,
Brunnschweiler, D. 392, 547 247, 248, 249, 251, 258, 259, 261,
Bucci, R. J. 275, 541 262, 263, 265, 266, 267, 268, 269,
Bucinell, R. B. 158, 160, 165, 534 271, 272, 278, 285, 288, 300, 302,
Budianski, B. 136, 178, 198, 530, 534, 303, 307, 308, 313, 314, 316, 317,
535 318, 319, 320, 321, 322, 324, 325,
Bunsell, A. R. 21, 231, 254, 256, 268, 326, 330, 332, 334, 335, 336, 337,
526,541,542 338, 341, 342, 344, 346, 347, 348,
Burgel, B. 85, 170, 530, 535 349, 350, 353, 354, 355, 357, 358,
Byun, J. H. 302, 384, 386, 414, 415, 417 359, 360, 364, 365, 366, 367, 368,

556
 Author index 557

Chou, T. W. (cont.) Dodge, R. N. 454, 552

369, 370, 371, 372, 373, 374, 376, Dorey, G. 277, 541
377, 378, 384, 386, 387, 389, 390, Douglass, W. A. 294, 295, 296, 545
391, 392, 393, 394, 396, 397, 398, Dow, N. F. 358, 419, 545, 548
399, 400, 401, 402, 406, 407, 408, Doyle, J. F. 275, 543
410, 412, 413, 414, 415, 417, 418, Du, G. W. 376, 377, 378, 379, 380, 381,
419, 420, 421, 422, 423, 424, 425, 389, 391, 392, 393, 394, 396, 397,
426, 427, 428, 429, 430, 431, 432, 398, 399, 400, 401, 403, 404, 414,
433, 434, 435, 436, 437, 438, 439, 547, 548
440, 441, 442, 443, 444, 459, 460, Duke, J. C. 81,532
462, 463, 464, 469, 470, 471, 472,
473, 476, 477, 480, 486, 491, 492, Edwards, H. 202, 535
493, 494, 495, 496, 498, 499, 500, Eimer, C. Z. 200, 535
501,503,505,509,510,517,518, El Shiekh, A. 378, 379, 382, 549
519, 520, 521, 523, 524, 526, 527, Epstein, B. 104, 530
528, 529, 530, 531, 532, 533, 534, Ericksen, J. L. 474, 553
535, 536, 537, 538, 539, 540, 541, Eshelby, J. D. 178, 179, 249, 535, 536,
542, 543, 544, 545, 546, 547, 548, 541
549,550,551,552,553,554 Evans, A. G. 136, 530
Christensen, R. M. 29, 51, 179, 191, Evans. N. P. 202, 535
200, 527, 528, 535
Chu, H. S. 54, 527 Fairslon, R. W. 444, 551
Clark, J. P. 16, 17, 526 Favre,J. P. 170,536
Clark, S. K. 444, 445, 447, 448, 449, Fernando,G. 278, 540, 541
450,451,453,454,455,552 Fichter, B. W. 85, 89, 94, 530
Cohen, R. N.165,529 Fischer, S. 231, 278, 542, 543
Cole, P. M. 377, 548 Flemings, M. C. 16, 17, 526
Coleman, B. D. 102, 103, 104, 113, 530 Florentine, R. 377, 548
Compton, W. D. 13, 14, 526 Fornes, R. E. 358, 546
Conway, H. D. 182, 535 Foth, J. 275, 543
Cook, J. 182, 183,535 Fowser, S. W. 435, 548
Cooper, G. A. 82, 84, 134, 135, 136, Friedrich, K. 192, 227, 228, 230, 369,
434, 529, 548 370, 371, 536, 546
Cooper, S. L. 182, 225, 534 Frost, H. J. 419, 548
Cox, H. L. 170, 182, 187, 216, 226, 535 Fujita, A. 297, 547
Crane, R. M. 442, 548 Fukuda, H. 85, 94, 118, 119, 133, 134,
Cratchley, D. 224, 537 170, 171, 172, 174, 176, 183, 184,
Crossman, F. W. 136, 137, 150, 151, 205, 206, 210, 215, 217, 222, 223,
158, 160, 530, 533 225, 226, 231, 233, 234, 238, 239,
Curtis, P. T. 136, 137, 138, 139, 140, 241, 242, 243, 258, 259, 260, 261,
144, 207, 216, 230, 371, 529, 535, 262, 263, 265, 266, 267, 268, 269,
545 271,272,530,536
Cuthbertson, J. 204, 536 Fukunaga, H. 46, 150, 242, 258, 259,
260, 261, 262, 263, 268, 269, 271,
Dai, S. R. 170, 539
272, 527, 530, 542
Daniel, I. M. 495, 555
Fung, Y. C. 475, 476, 478, 482, 509, 553
Daniels, H. E. 104, 530
Funk, J. G. 435, 548
Dauksys, R. J. 166, 532
Daviess, G. J. 202, 537
Davis, J. W. 275, 543 Gandi, C. 419, 547, 548
Deaton, J. W. 94, 534 Garrett, K. W. 136, 137, 149, 530, 532
Dederichs, P. H. 179, 535, 540 Ghasemi Nejhad, M. N. 373, 545
Deex, O. D. 184, 539 Gillespie, J. W. 435, 436, 437, 438, 439,
Dexter, H. B. 435, 548 440, 547, 548, 549
Dhingra, A. K. 98, 428, 530, 548 Gjostein, N. A. 13, 14, 526
Dicken, R. F. 278, 540, 541 Gordon, J. E. 20, 526
558 Author index

Goree, J. G. 85, 530 Hori, M. 195, 537

Gough, V. E. 444, 552 Horstenkamp, G. 230, 536
Grande, D. H. 170, 538 Hsiao, G. C. 233, 243, 244, 246, 247,
Greenfield, M. R. 119, 132, 533 248, 543
Grosberg, P. 286, 390, 392, 447, 545, Hsu, P. L. 70, 230, 527, 537
549, 552 Hsu, P. W. 70, 527
Gross, R. S. 85, 530 Huang, S. C. 54, 527
Gruber, M. B. 251,278,542 Humphrey, J. D. 20, 476, 526, 554
Gucer, D. E. 115,530 Hunston, D. H. 435, 547
Guenon, V. A. 435, 436, 437, 438, 548, Hutchings, J. 277, 541,543
549 Hutchinson, J. W. 136, 530
Guess, T. R. 435, 549
Gunnink, J. W. 275, 544
Gurland, J. 115, 530 Iosipescu, N. 431, 549
Gusliter, R. L. 444, 552 Iremonger, M. J. 229, 537
Guth, E. 469, 552 Ishikawa, H. 230, 537
Ishikawa, T. 298, 300, 302, 303, 307,
Haener, J. 170, 536 308,313,314,316,317,318,319,
Hahn, H. T. 29, 191, 367, 476, 528, 536, 320, 321, 322, 324, 325, 326, 330,
545, 553 332, 334, 335, 336, 337, 338, 341,
Hale, D. K. 202, 536 342, 344, 346, 347, 348, 349, 350,
Hall, G. L. 454, 553 353, 354, 355, 358, 366, 367, 368,
Halpin, J. C. 29, 33, 178, 181, 188, 191, 428, 476, 545, 546, 549, 552, 554
192, 194, 498, 527, 536, 554
Hamada, H. 297, 547
Jackson, P. W. 224, 537
Hamamoto, A. 322, 546
Jaeger, J. C. 75, 527
Hancock, P. 204, 536
James, H. M. 469, 552
Hancox, N. L. 231,542,544
Jang, B. Z. 276, 277, 543
Hanson, M. P. 277, 541
Jegley, D. C. 46, 528
Harlow, D. G. 114, 115, 118, 119, 120,
Jerine, K. 188, 191, 192, 194, 536
121, 122, 123, 124, 231, 263, 268,
Ji, X. 96, 97, 98, 233, 243, 244, 246, 247,
270,531,542
248, 531, 543
Harris, B. 21, 254, 256, 268, 278, 526,
Jones, F. R. 136, 145, 146, 147, 148,
540, 541, 542
149, 532
Harris, S. J. 231,542
Jones, R. E. 448, 552
Hashin, Z. 178, 179, 180, 195, 198, 200,
Jones, R. M. 29, 39, 362, 367, 442, 528,
536, 537, 539
546, 550
Hawkins, C. 26, 526
Jones, R. S. 476, 554
Hayashi, R. 428, 551
Jost, W. 62, 528
Hayashi, T. 231,542
Hayashi, Y. 316, 317, 318, 319, 546
Hearle, J. W. S. 286, 298, 390, 392, 447, Kalnin, I. L. 231, 257, 543
545, 549, 552 Kang, T. J. 379, 382, 549
Hedgepeth, J. M. 85, 89, 91, 94, 96, 120, Kanninen, M. F. 231, 438, 544, 550
131,233,238,246,531,533,543 Kardos, J. 188, 192, 204, 225, 537
Henneke, E. G. 81, 532 Karger-Kocsis, J. 192, 227, 536
Henstenburg, R. B. 84, 113, 119, 132, Kacir, L. 202, 537
531, 532, 533 Katayama, K. 54, 528
Herakovich, C. T. 70, 527 Kawata, K. 85, 94, 118, 119, 133, 134,
Hermans, J. J. 177, 178, 181, 537 170, 176, 183, 184, 530, 536
Hikami, F. 85, 88, 90, 91, 92, 93, 94, Kelly, A. 1, 17, 18, 19, 20, 27, 81, 82,
205, 209, 210, 211, 212, 213, 214, 84, 103, 104, 134, 135, 136, 169,
215,216,531,537 202, 203, 204, 231, 232, 248, 251,
Hill, R. 178, 537 254, 256, 257, 258, 276, 278, 429,
Hirano, M. 448, 551 434, 526, 529, 531, 535, 536, 537,
Hoogsteden, W. 435, 551 540,541,548,549
 Author index 559

Kenaga, D. 275, 543 Luo, S. Y. 475, 476, 477, 478, 480, 486,
Kerner, E. H. 178, 198, 537 491, 492, 493, 494, 495, 496, 498,
Kibler, J. J. 363, 546 499, 500, 501, 503, 505, 509, 510,
Kies, J. A. 137, 139, 429, 531, 549 517,518,519,520,521,523,524,
Kilchinskii, A. A. 177, 178, 537 552, 554
Kimpara, I. 322, 546
King, R. W. 26, 526 Ma, C. L. 406, 407, 408, 410, 412, 413,
Kingery, W. D. 73, 528 430, 549, 551
Kingston-Lee, D. M. 199, 200, 333, 540, MacLaughlin, T. F. 229, 534, 538
546, 547 Macmillan, N. H. 429, 549
Kirk, J. N. 276, 543 Maekawa, Z. 297, 547
Kirkpatrick, E. G. 98, 531 Majidi, A. P. 14, 15, 428, 429, 430, 431,
Kishore, N. N. 158, 533 432, 433, 434, 435, 440, 441, 526,
Knibbs, R. H. 187, 538 550
Ko, F. K. 285, 358, 374, 379, 387, 389, (See also Parvizi, A)
442, 545, 546, 549, 551 Majumder, A. J. 231, 554
Kobayashi, H. 297, 547 Mandell, J. F. 170, 538
Kobayashi, N. 54, 528 Manders, P. W. 80, 103, 109, 110, 112,
Kregers, A. F. 442, 549 113, 133, 136, 145, 146, 147, 148,
Krempl, E. 475, 478, 554 149, 166, 168, 202, 231, 261, 275,
Kretsis, G. 278, 543 532, 538, 540, 541
Kroner, E. 178, 179, 538 Manera, M. 191, 538
krueger, W. H. 428, 548 Marissen, R. 275, 543
Kulkarni, S. V. 94, 531 Marom, G. 231, 278, 542, 543, 544
Kumar, M. 456, 551 Masters, J. E. 435, 550
Kuo, C. M. 459, 471, 472, 473, 523, 524, Matsushima, M. 316, 317, 318, 319, 546
552, 553,554 McCalla, B. A. 199, 200, 333, 540, 546,
547
McCartney, L. N. 136, 531
Lai, W. M. 475, 478, 554 McColl, I. R. 276, 543
Lark, R. F. 231,541 McConnell, R. F. 377, 550
Lavengood, R. E. 183, 225, 229, 534, McCullough, R. L. 1, 6, 8, 9, 24, 179,
535, 538 279, 281, 282, 283, 284, 374, 387,
Law, G. E. 136, 137, 150, 158, 530 390, 526, 540, 542, 548
Laws, N. 198, 200, 538 McGarry, F. J. 85, 170, 530, 535, 538
Leach, B. S. 384, 386, 547 McLaughlin, R. 200, 538
Lee, L. H. 225, 538 McNally, D. 225, 538
Lees, J. K. 182, 205, 224, 225, 538 Metcalfe, A. G. 114, 532
Lei, C. 379, 549 Mignery, L. A. 24, 435, 526, 550
Lei, S. C. 150,155,531,533 Mileiko, S. T. 84, 532
Lekhnitskii, S. G. 311, 417, 546, 549 Miller, A. K. 277, 540
Levin, V. M. 198, 538 Miller, B. 170, 538
Lewis, T. B.229, 535 Mills, G. J. 166, 532
Li, C. A. 158, 533 Mody, P. B. 14, 15, 369, 370, 371, 526,
Li, S. 52, 528 546
Li, W. 378, 379, 382, 549 Mohamed, M. H. 289, 290, 291, 293,
Lin, H. T. 276, 277, 543 358, 546
Lipson, H. 110,531 Molyneax, J. 195, 534
Lipson, S. G. 110,531 Morgan, H. S. 476, 554
Liu, C. H. 440, 549 Mori, T. 249, 543
Liu, T. S. 224, 539 Morley, J. G. 276, 543
Liu, X. R. 96, 97, 531 Morris, J. B. 187, 538
Lo, K. H. 51,52, 53, 528 Mossakowska, Z. 54, 528
Lord, P. R. 289, 546 Mueller, L . N . 275, 541,543
Lou, A. Y. 453, 473, 552 Muki, R. 229, 538, """
Luo, J. 52, 273, 274, 528, 544 Munro, M. 276, 543
560 Author index

Mura, T. 178, 538 Pipes, R. B. 1, 6, 8, 9, 29, 55, 58, 62, 65,

Muri, P. 170, 538 66, 67, 151, 160, 374, 387, 390, 428,
431, 495, 526, 527, 528, 532, 548,
Nardone, V. C. 20, 526 550, 555
Narkis, M. 202, 537 Pipkin, A. C. 475, 554
Nenakhov, B. V. 444, 552 Pitkethly, M. J. 275, 543
Netravali, A. N. 84, 532 Pitt, R. E. 119, 132, 533
Nicholson, R. B. 81, 531 Poon, K. C. 54, 528
Nicolais, L. 188, 539 Popper, P. 376, 377, 378, 389, 390, 392,
Nielsen, L. E. 187, 539 393, 394, 396, 397, 398, 399, 400,
Noda, N. 55, 528 401, 548, 550
Nomura, S. 54, 178, 179, 180, 181, 184, Port, O. 26, 526
191, 193, 195, 197, 198, 199, 200, Posfalvi, O. 503, 504, 554
201,249,528,535,539,541,543 Prewo, K. M. 20, 526
Norman, J. C. 345, 346, 355, 547 Prohaska, J. L. 275,541,543
Norman, T. L. 275, 544
Norris, C. B. 191, 540 Ouigley, J. 208, 222, 534
Nowack, H. 275, 544
Nowacki, W. 54, 528 Rebenfeld, L. 170, 538
Rechak, S. 275, 544
Ogasa, T. 428, 551 Reddaway, J. L. 204, 539
Ogo, Y. 435, 550 Reddy, J. N. 52, 456, 528, 551
Oh, K. P. 132, 532 Reedy, Jr., E. D. 435, 549
Okura, A. 1, 526 Reese, E. 368, 546
Outwater, J. O., Jr. 204, 539 Reifsnider, K. L. 81, 532
Overbeeke, J. L. 278, 542 Reinhardt, H. W. 444, 553
Overy, M. J. 199, 200, 333, 540, 546, Reissner, E. 49, 528
547 Reiter, H. 278, 540, 541
Ozisik, M. N. 61, 528 Remond, O. G. 431, 433, 550
Renton, W. J. 232, 544
Pagano, N. J. 49, 151, 160, 188, 191, Reuss, A. 179, 186, 187, 539
498, 529, 532, 536, 540, 554 Rhodes, T. 289, 290, 291, 293, 546
Pakdemirli, E. 187, 539 Richer, H. 202, 539
Parker, S. F. H. 333, 546 Riewald, P. G. 277, 541
Parratt, N. J. 115,532 Riley, V. R. 204, 539
Parvizi, A. 136, 137, 138, 139, 140, 144, Rivlin, R. S. 474, 475, 476, 478, 487,
149, 529, 532 488, 504, 533, 554, 555
(See also Majidi, A. P.) Robinson, H. H. 119, 132, 532
Pastore, C. M. 389, 442, 549 Rock, J. W. 136, 145, 146, 147, 148,
Patel, H. P. 448, 553 149, 532
Patterson, R. G. 454, 552 Rogers, C. A. 26, 275, 526, 541
Peirce, F. T. 342, 550 Rogers, K. F. 200, 333, 540, 546, 547
Perrin, J. J. 170, 536 Rogers, T. G. 475, 554
Perry, A. J. 85, 170, 530, 535 Rosen, B. W . 32, 94, 103, 115, 116, 117,
Peters, P. W. M. 150, 258, 259, 260, ,us, 134, 175, 178, 191, 198,362,
262, 530, 532, 542 527, 528, 531, 533, 534, 536, 539,
Peterson, J. M. 279, 281, 282, 283, 284, 544, 546
543 Rubin, D. 475, 478, 554
Petit, P. H. 29, 467, 476, 527, 552, 554 Russel, A. J. 165, 533
Phillips, L. N. 200, 277, 333, 540, 543, Rybicki, E. F. 231, 438, 544, 550
546, 547
Phoenix, S. L. 81, 84, 104, 106, 113, Saito A 54 528
114, 115, 118, 119, 120, 121, 122, Sakharov, S. P. 444, 552
123, 124, 127, 129, 130, 132, 263, Sargent, J. P. 199, 200, 333, 539, 540,
268, 270, 530, 532, 533, 542 546, 547
Piggott, M. R. 170, 539 Saunders, D. W. 474, 555
 Author index 561

Scardino, F. L. 285, 286, 292, 294, 298, Takao, T. 229, 230, 539, 540
299, 358,546 Takao,Y. 85, 533
Schapery, R. A. 198, 200, 539 Takehana, M. 322, 546
Schierding, R. G. 184, 539 Takeuti, Y. 55, 85, 528
Schmitz, G. K. 114, 532 Talreja, R. 81, 533
Schneider, W. R. 85, 170, 530, 535, 539 Tan, T. M. 24, 435, 526, 550
Schulte, K. 150, 230, 258, 259, 260, 262, Tanaka,K. 249, 543
368, 530, 536, 542, 546 Taplin, D. M. R. 419, 547
Schultz, R. W. 275, 541 Tappin, G. J. 432, 551
Schwartz, P. 84, 119, 132, 289, 290, 291, Tattersall, H. G. 432, 551
293, 358, 532, 546 Tauchert, T. R. 54, 78, 527, 528
Scop, P. M. 115, 118, 533 Taya, M. 85, 178, 180, 229, 230, 249,
Seleznev, I. I. 444, 552 533,535,539,540,541,544
Sendeckyj,G. P. 177,535 Teters, G. A. 442, 549
Serafini, T. J. 277, 541 Thomas, D. G. B. 292, 293
Sharma, B. 58, 528 Thomas, S. 444, 459, 546, 553
Shen, G. 106, 107, 108, 109, 110, 114, Tranfield, G. 358, 545
530 Trautman, K. H. 275, 543
Short, D. 231,544 Treloar, L. R. G. 469, 553
Shtrikman, S. 178, 195, 537 Truesdell, C. 475, 555
Shu, L. S. 191,539 Tsai, S. W. 29, 33, 191, 367, 476, 527,
Sidey, G. R. 277, 542 528, 540, 545,554
Sillwood, J. M. 231,540 Tsiang, T.-H. 170, 538
Simonds, R. A. 442, 550 Tsukerberg, S. M. 444, 552
Sinclair, J. H. 249, 250, 251, 541 Tuler, F. R. 231, 543
Singh, A. 56, 528 Tyson, W. R. 202, 203, 204, 537
Skelton, J. 358, 444, 456, 457, 458, 459,
546,553 van de Poel, C. 178, 540
Slomiana, M. 158, 160, 165, 534 Van Dyke, M. 69, 529
Smith, R. L. 106, 119, 127, 129, 130, Van Dyke, P. 85, 89, 94, 136, 533
131,132,532,533 Vinson, J. R. 1, 10, 29, 33, 40, 48, 58,
Soebroto, H. B. 379, 549 62, 65, 66, 67, 81, 203, 527, 528,
Sousa, L. J. 11,526 529, 533, 540
Spencer, A. J. M. 475, 555 Vitols, R. 292, 546
Spiegel, M. R. 98, 533 Vogelesang, L. B. 275, 544
Spilling, I. 170, 540 Voigt, W. 179, 186, 187, 540
Stein, M. 46, 528 Vu, B. T. 182, 225, 534
Steinberg, M. A. 15, 16, 527
Sternberg, E. 229, 538 Waals, F. M. 191,535
Steward, B. 278, 541 Waddoups, M. E. 467, 476, 552, 554
Stinchcomb, W. W. 81, 442, 532, 550 Wadsworth, N. J. 170, 540
Stowell, E. Z. 224, 539 Wagner, H. D. 231, 543, 544
Street, K. N. 165, 533 Walrath, D. E. 431,551
Stroud, S. S. 384, 386, 547 Walter, J. D. 443, 445, 448, 452, 453,
Stubbs, N. 444, 459, 553 454, 456, 552, 553
Sugano,Y. 55, 528 Walton, P. L. 231,544
Summerscales, J. 231, 544 Wang, A. S. D. 81, 136, 137, 150, 151,
Sun, C. T. 22, 23, 24, 52, 273, 274, 275, 152, 154, 155, 156, 158, 159, 160,
435, 526, 527, 528, 529, 541, 543, 161, 163, 164, 165, 530, 533, 534
544,550 Wang, C. Z. 276, 277, 543
Wang, H. S. 55, 76, 78, 529
Takagi, T. 25, 527 Wang, Y. R. 55, 68, 69, 72, 73, 74, 529
Takahashi, K. 33, 278, 428, 443, 459, Warren, F. 191,540
460, 462, 463, 464, 467, 470, 471, Warren, W. J. 136, 137, 150, 158, 160,
472, 473, 524, 528, 544, 550, 552, 169, 529, 530
553, 554 Weaver, T. C. 182, 535
562 Author index

Weibull, W. 101, 102, 147, 534 423, 424, 425, 427, 428, 430, 431,
Weller, R. D. 377, 551 432, 433, 434, 435, 444, 547, 548,
Wells, H. 231, 542, 544 549,550,551,552
Weng, C. I. 54, 527 Yano, T. 459, 553
Whitcomb, J. D. 442, 551 Yates, B. 199, 200, 333, 540, 546, 547
Whitney, J. M. 46, 49, 51, 52, 188, 191, Yau, L. N. 278, 544
192, 194, 435, 495, 529, 536, 551, Yau, S. S. 230, 372, 442, 537, 540, 544,
555 547,551
Whitney, T. J. 379, 384, 406, 414, 548, Yin, F. C. P. 20, 476, 526, 554
551 Yokoyama, A. 297, 547
Wilczynki, A. P. 191,540 Yonezawa, F. 195, 537
Williams, J. G. 187, 539 Yoshida, H. 428, 551
Wills, J. R. 195, 540 Yoshida, N. 456, 551
Wood, W. G. 229, 537
Wray, G. R. 292, 546 Zee, R. H. 276, 277, 543
Wu, C.-T. D. 179, 540 Zeller, R. 179, 535, 540
Wu, E. M. 51,53,528 Zender, G. W. 85, 94, 534
Zimmerman, R. S. 277, 540
Yang, J. M. 302, 357, 358, 359, 360, 364, Zweben, C. 94, 118, 231, 262, 263, 265,
365, 366, 389, 402, 406, 407, 408, 277, 345, 346, 355, 531, 534, 541,
410, 412, 413, 419, 420, 421, 422, 544, 547
 Subject index

abrasion 369 ceramic fiber 10

abrasive wear 369 ceramic matrix composite 5, 19
acoustic emission 254, 322 chain stitching 375, 387
adhesive layer 23 chain-of-links model 106, 116, 119, 256
adhesive mechanism 369 characteristic distribution function 121,
Almansi's strain 482 123
angle-ply 44, 45, 190, 420, 421, 422, 423, Charpy impact test 277
424, 425, 426, 431 circular braider 295
anisotropy 30, 33, 474, 476 class frequency 98
antisymmetry 45, 333, 514 clip gage 454
aramid aluminum laminate 275 close packing 391
aramid fiber 169 coated fabric 443, 456, 476
area method 436 coefficient of variation 100, 103
arithmetic mean 99 cofactor matrix 482
averaging method 186, 187, 224, 405, Coleman factor 105
416 collocation approximation 200
axial yarn 374, 377 compact tension test 432
compatibility 255
complementary energy density 476, 509
balanced fabric 340, 345, 372, 458 compliance 34, 36, 185, 186, 198, 302,
basket braid 295 307, 312, 315, 323, 340, 345, 351,
basket weave 357 352, 367, 412, 417, 418, 517
beat-up 289, 291 compliance method 436
bending-stretching coupling 43, 45, 304 composite
Bernoulli distribtuion 101 A1 2O3/aluminum/silicon 225
bimodulus 456, 506, 512 Al 2 O 3 /borosilicate glass 426
binomial distribution 101 A12O3/magnesium 425
biological material 20, 476 boron/epoxy 183
bond strength 205 carbon/borosilicate glass 426
bonded ends 170 carbon/carbon 7, 19, 426
bound approach 177, 178, 180, 186, 197, carbon/epoxy 23, 137, 200, 307, 313,
247, 249, 302, 307, 313, 317, 342, 412, 420, 421, 422, 423, 435, 436
355 carbon/Kevlar/epoxy 345, 355, 372
braid pattern 379, 390 carbon/magnesium 425
braider yarn 374, 377, 415
carbon/PEEK371,424
braiding 9, 287, 294, 374, 377, 420, 421
FP/Al-Li431,432, 440, 442
braiding angle 396, 413, 414, 425
glass/epoxy 118, 137, 144, 178, 183,
braiding parameter 402
192, 196, 197, 198, 225, 420, 421,
bridging fiber 165, 208, 209, 210, 218,
423
220, 253, 439 glass/nylon 224
bridging model 302, 314, 317, 323, 327, glass/PEEK 424
329, 348, 352, 367 glass/polyimide 324, 367
bright-field illumination 146 Kevlar/epoxy 39, 372, 420, 421, 423
buckling load 452
Kevlar/PEEK 424
bulk modulus 30, 446
Kevlar/silicone elastomer 474, 493,
bundle strength 104, 106
494, 498, 499, 503, 505, 517, 524
polyester/rubber 453
carbon fiber 33, 99, 114, 169, 413 SiC/borosilicate glass 72, 73, 426
Cartesian tensor 478 SiC/magnesium 425
Cauchy's stress/strain 482, 483 tirecord/rubber 498, 518
563
564 Subject index

composite rod 406, 416 initiation 158

compression 97, 373, 428 mixed mode 164, 165
compression molding 14 mode 1 164
conductivity 11 propagation 158
consolidation 393 desorption of moisture 64
constitutive relation 33, 39, 43, 48, 50, diamond braid 295
55, 58, 198, 279, 300, 301, 327, 409, dielectric constant 24, 194
449, 476, 485, 490, 499, 509 differential Poisson's effect 254
constrained cracking 136, 142 differential thermal expansion 256
constrained failure 252 diffusion 24, 195
contact pressure 369, 371 dimensional stability 286, 291
continuous fiber 8, 80, 187, 196, 200, direction cosine 416
201,285 discontinuous fiber 169, 234
continuum 81, 136,480 dispersion 99, 253
contravariance 478 displacement potential 78
cord/rubber composite 443, 445, 476 dobby loom 289, 383
corrosion resistance 11 double cantilever beam 436
covariance 478 double-notched shear test 431
covering factor 298 Doweave 356
crack 3, 91, 93, 135, 138, 141, 142, 146, drop-weight impact test 277, 440
205, 428, 439 dynamic stress concentration 94, 166,
arrestor 265 243
density 140, 156
opening displacement 432 edge delamination 22
spacing 140, 149, 252, 276 effective elastic property 409, 412
crack tip 89, 153, 161, 433, 436 effective flaw 150, 151, 152, 155, 156,
creep 18 160, 162
creep compliance 200, 201 effective laminar elastic property 410
crimp 457 effective relaxation modulus 200
crimp model 302, 308, 317, 319, 327, eigen value 121
328, 348, 349 eigenvector expansion 241, 268
critical stress intensity factor 433 Einstein summation 478
critical zone 207, 208, 209, 218 elastic property 30, 31, 34, 43, 47, 176.
cross-link 4 188, 189, 193, 318, 412, 414, 416,
cross-ply 44, 56, 81, 134, 137, 255, 304, 418,420,421,505
313, 315, 336, 340, 423 elastomer 18
crystalline defect 10 elastomeric composite 443, 474
cumulative distribution function 101, electrical conductivity 24, 194
102, 119, 122, 124, 147, 259, 264, ellipsoidal inclusion 179, 195
270 elliptic integral 461
cumulative frequency 99 end-notch-flexural specimen 440
cumulative probability of survival 103 energy absorption 20, 134
cumulative strength distribution 102, energy method 160
105, 168 environmental effect 230
curvature 41, 49, 300, 306, 327 epoxy 4
curved fiber 459 equal load shearing 115
equilibrium equation 69, 71, 78, 86, 89,
damage 80, 81 95, 172, 234, 255, 485
damage model 208 equivalent inclusion method 249
damage tolerance 285, 372, 428, 435 Euler's constant 212
dark-field illumination 146 Eulerian description 477, 480
debonding 146, 227, 229, 254, 275, 432, Eulerian stress/strain 483, 510, 511, 512,
434, 439 515
deformation gradient matrix 490 extensional stiffness 43, 412
deformation mechanism map 419
delamination 23, 46, 157, 162 failure criterion 319
 Subject index 565

failure mechanism 224, 227 flexural stress 51, 52, 53

failure mode 81, 257 force dipole 93
failure probability 168, 260, 270 force resultant 41, 42, 61, 64
failure sequence 270 effective, hygroscopic 58, 62
failure strain 136, 202, 208, 249, 252, effective, thermal 58
260, 264 four-step braiding 374, 375, 377, 379,
failure stress 23, 117,252 406
fancy braider 297 Fourier series 236, 246
fatigue 231, 275, 277, 453, 476 Fourier transformation 91
fiber fracture 227, 275, 276, 368, 428, 432
aspect ratio 169, 170, 180, 183, 196, mechanics 80, 150
199, 200, 393, 396, 422, 427 strength 202
breakage 85, 118, 233 surface energy 143
bundle size 243, 428 toughness 164, 228, 231, 432
bundle strength 81, 105, 117 free edge 23, 46, 158
critical length 203, 204, 209, 215, 217, free edge delamination 160, 165
225, 227, 228, 434 frequency distribution 98
density 394 friction 135, 368
end 116, 169, 170, 208, 209, 218 friction coefficient 204
end gap 209, 210, 212, 213, 214, 215
entanglement 286
fracture 98, 117, 227 gage length 103, 104, 110, 113, 125, 167,
fracture spacing 83, 84 265
fragmentation test 170 gamma function 103, 265
ineffective length 115, 117, 118, 167, geometric mean 99
175, 208, 259, 263, 264 glass 169, 191
length 84, 97, 203, 207, 223 glass fiber 183, 469
link 133 glass-ceramic matrix 10
orientation 22, 160, 183, 424, 474 global-local method 52
packing 286, 378, 390, 394 gradient matrix 481
prestressing 166 Green's function 179
pull-out 170, 205, 227, 432, 433, 434, Green's strain 481
437, 495
spacing 87, 89, 212, 429
strength 81, 84, 97, 102, 106, 108, 116, Halpin-Tsai equation 181, 189
125 harness 289
undulation 320 heat conduction 61, 194
volume fraction 29, 32, 81, 83, 180, heat flux 74, 195
195, 200, 217, 255, 260, 272, 280, heat setting 458
434 heddle 289
fiber inclination model 405, 406, 430 helical angle 286
fiber-fiber interaction 171 heterogeneous interlacing 339, 348
figured braid 297 hexagonal array 94
filament 169, 390 high elongation fiber 233, 247
filament winding 141 high modulus fiber 233, 234, 239, 241,
filling yarn 289, 311, 327, 335, 350, 356, 247
384, 458 higher order plate theory 46, 49
finite deformation 19, 443, 459, 474, 475 histogram 98
finite elasticity 474 homogeneity 30, 177
finite element analysis 169, 249, 274, homogeneous deformation 490, 500,
313, 322, 437, 440 502,514
first cracking strain 251 homogeneous interlacing 339, 343, 348,
first ply failure 259, 260 352, 354
flat braider 295 Hooke's law 106
flaw size 157, 163 horn gear 294, 377
flexible composite 8, 19, 20, 443, 474 hybrid 22, 23, 27, 428
566 Subject index

hybrid composite 10, 178, 231, 335 knitting 287, 292, 374, 376, 389
hybrid effect 241, 251, 272, 300, 302 Kronecker delta 479
hygroscopic effect 62, 65
hygroscopic expansion coefficient 57
hygrothermal effect 54, 55, 198 Lagrangian description 477, 480
Lagrangian stress/strain 483, 487, 492,
496,497,501,515
impact 23, 275, 291, 440, 442 laid-in yarn 294, 298, 375, 384, 389, 407,
impact resistance 231, 276 414, 426
impregnation 287 lamina 30, 41,45, 358, 407
in-plane fracture 432 laminate 8, 22, 39, 58
inclination angle 393 laminated composite 29, 449
inclined lamina 410 lamination analogy 188, 191, 405
incompressibility 450, 476 lamination theory 251, 300, 310, 409,
incremental analysis 459, 467, 476, 519 462
infinitesimal strain 474 Laplace transform 95, 246
influence function 235, 236, 246 Legendre polynomial 89, 91
injection molded composite 182, 191 linear programming 284
intelligent composite 19, 24, 25 link 102, 266
interface 7, 82, 138, 163, 170, 203, 434
load sharing 209, 210, 214, 216, 254
interfacial bond 135, 177, 216, 227, 456
load transfer 4, 106, 169, 171, 176, 183,
interfacial debonding 33, 85, 428
209, 218, 227, 252, 253, 315
interfacial strength 170, 204, 209, 228
load-COD curve 432
interfacial stress 18, 83, 170, 171, 203,
load-deflection curve 434, 436, 440
252 load-extension curve 205
interlacing 287, 288, 292, 296, 306 load-strain curve 106, 107, 108, 109,
interlaminar fracture 435 114, 447
mode I 436 load-time curve 277
mode II 440 local load shearing 118, 120
interlaminar shear modulus 46, 251 lock stitching 375, 387
interlaminar shear strength 285, 431 loom 288, 376
interlaminar stress 22, 41, 67, 73, 153, loose bundle 105, 110, 168, 170
163, 452 loose bundle test 113
interlaminated composite 20, 231, 233, low elongation fiber 233, 247
244, 254 low modulus fiber 233, 234, 239, 241,
interlock 406 247
interlooping 287
intermingled hybrid 231, 233, 249, 342
interphase 7, 227 machine cycle 379, 403
interply 231 macro-cell 405, 414, 418
interwoen 232
Magnaweave 377
intralaminar flaw 150
magnetic permeability 24, 195
intraply 231
material principal coordinates 31, 33, 38,
Iosipescu shear test 431 39, 46, 56
iso-phase model 460, 462, 471, 519, 520
matrix cracking 23, 84, 209, 320
iso-strain 202, 250, 302, 341, 352
matrix deformation 165, 224, 227
iso-stress 250, 302, 343
matrix inversion 61, 301, 313, 412
isotropy 30, 31, 177, 292, 357
matrix rich region 387
Izod impact test 277
maypole 294, 375
mean bundle strength 104
Jacquard 289, 383 mean fiber strength 103
median 99
median failure strain 260
Kevlar fiber 33, 399, 400, 446, 469 median strength 126, 271
kinking 430 membrane force 456
Kirchhoff hypothesis 40, 300 metal matrix composite 5, 172
knee behavior 20, 137, 146, 319, 429 microcrack 144, 208, 227
 Subject index 567

microstructure-performance relationship permutation 479

19,22 perturbation 69, 70, 179, 224
microvoid 227 photoelastic 229, 520
misorientation 178, 182, 216 physical property 169, 194
mixed interlacing 339 pick 289
mixed mode crack 163 Piola/Kirchhoff stress 483, 509
mode I crack 163 pitch length 395, 396, 414, 427
moisture absorption 62 plain braid 295
moisture concentration 62, 66 plane strain 30, 77, 163
moisture diffusion 54, 61, 66 plane stress 77, 183
moisture diffusion coefficient 62 plastic deformation 85, 135, 176, 211,
mold-in hole 372 229, 432, 434
moment resultant 41, 42, 61, 64, 300, plastic stress concentration factor 211
320 plate theory 29, 41
effective, hygroscopic 59, 63 ply termination 22
effective, thermal 59 PMMA 225
Monte-Carlo simulation 115, 125, 132, pneumatic tire 476
150, 157, 263, 265, 266 Poisson distribution 101
mosaic model 302, 317, 327, 335, 336 Poisson's effect 224, 255
multi-axial warp knit 9, 287, 389, 405 Poisson's ratio 17, 30, 139, 187, 251,
multi-filament failure 88, 92 279,304,311,361,412,419,430,
multi-functional composite 25 446, 503
multi-layer warp knit 376 polycarbonate 191
multiple cracking 145, 146, 154, 252, polycrystalline ar-alumina fiber 428
253, 275, 276 polyester 446
multiple fracture 20, 83, 84, 110, 134, polyethylene 225
135, 263 polyethylene terephthalate 191
polyimide 4
natural frequency 46, 452 polyurethane 94
near-net-shape 285, 374 population 98
negative thermal expansion 17 potential energy 52
netting analysis 444, 449, 455 Prandtl's matching principle 71
nonwoven fabric 298 probabilistic initial failure strength 257
nonlinear behavior 18, 278, 366, 430, probabilistic theory 149, 205, 217
432, 436, 443, 459, 467, 474, 476, probabilistic ultimate failure strength
490, 504, 509 262
normal distribution 101, 104 probability 80, 98, 100, 102
notch 89, 94, 433 probability density function 101, 103,
notch sensitivity 231 104, 116, 117, 162, 185, 207, 212,
notched strength 371 217, 218, 222
notched three-point bend test 434 processing window 389, 398, 400
pseudostrain-energy function 476
off-axis angle 38, 311, 410, 474
Omniweave 377 quadratic mean 99
open packing 391 quasi-isotropy 190, 191
optimization 46, 233, 279
orientation factor 216, 224, 225 random mat 225
Orowan stress 429 random number 132, 133
orthotropy 33, 36, 38, 46, 56, 61, 68, 69 random phase model 460, 465, 472
rapier 291
parametric study 73, 419 rayon 446
particulate-filled composite 25, 178 recycle 14, 18
PBT 469 reduced stiffness constant 36, 37
PEEK (polyetheretherketone) 4, 369 reed 291
percolative mechanism 24 regular braid 295
performance optimization 282 relative frequency 98
permeability 456 relaxation modulus 200
568 Subject index

repeating unit 314, 339 standard braid 295

residual stress 56, 79, 372 standard deviation 99, 101, 103, 105,
resin transfer modling 14, 400 110, 133
resistivity 279 staple fiber 285
rigid-body rotation 509 statistical analysis 80, 145, 195
risk of rupture 147 steady-state 66, 67, 75, 195
root-mean-square 99, 183, 186 stiffness 35, 47, 48, 186, 198, 301, 302,
roving 169 303, 306, 312, 315, 323, 352, 409,
rule-of-mixtures 81, 178, 180, 202, 205, 411, 416, 418, 474
213, 215, 217, 224, 256, 257, 261, stitch density 387
272 stitching 22, 23, 287, 374, 375, 387, 405
stochastic process 132
sample 98 stochastic strength 133
sandwich laminate 253 strain 40, 41,300, 327
scale parameter 102, 148, 259 strain concentration 263, 264, 268
scanning electron micrograph 191 strain gage 454
self-consistent approach 177, 178, 180, strain-displacement relation 50, 77
189, 195, 198, 200, 249 strain-energy 135, 276, 405, 406
self-similar crack 160 strain-energy function 474, 487
series model 306 strain-energy release rate 137, 153, 154,
shape parameter 102, 109, 111, 113, 119, 161, 162, 436
129, 148,259,261,265 strand 169
shear modulus 30, 31, 89, 95, 171, 304, strength 11, 80, 102, 104, 106, 115, 201,
311, 361, 419, 422, 446, 474 213, 219, 256, 279, 428
shear-lag analysis 85, 89, 94, 116, 136, stress concentraion 29, 80, 85, 88, 91,
140, 169, 170, 171, 177, 210, 233, 93, 94, 96, 133, 158, 166, 201, 204,
266 205, 207, 209, 210, 213, 214, 217,
shear-lag effect 153 229, 233, 238, 266, 372
shear-lag zone 153 stress intensity factor 432
shedding 289 stress redistribution 85, 120, 269
shell theory 443 stress resultant 44, 300, 320
shielding effect 153 stress singularity 72
short beam shear test 431 stress transfer 169
short fiber 8, 113,444 stress-energy function 509
short-fiber composite 169, 182, 187, 192, stress-log life curve 278
202, 216, 225 stress-strain curve 20, 104, 138, 146,
shuttle 290 254, 268, 276, 324, 429, 476
silicon carbide 5 stress-strain relation 19, 29, 30, 35, 36,
silicon nitride 5, 183 37, 56, 69, 77, 82, 146, 247, 267,
single fiber test 110, 170 319, 366, 430, 502
single fracture 82, 84 stress-transfer zone 153
sinusoidal curve 519 stretching-shear coupling 449, 474, 518
size effect 119, 126, 131, 252, 262 stuffed basic weave 357
sliding distance 369 superplastic 27
sliding velocity 369, 371 superposition 75
sliding wear 369 surface energy 135
smart composite 25 surface fatigue 369
softening strip 273 survivability 168
solid braid 287, 374, 382 synergistic effect 233, 254
specific heat 61, 73
spherical inclusion 196, 198, 200 tear strength 456
splitting 138, 139, 141,202 tensile behavior 257, 372, 428, 439, 462,
spun yarn 285 478, 505
square array 94 tensor
St Venant's strain 481 stress/strain 34, 47
stacking sequence 44, 45, 160, 268, 271, transformation 405, 414
408 ternary composite 279
 Subject index 569

textile composite 285 two-rail shear test 431

braided 422, 423, 424, 425, 426 two-step braiding 374, 375, 377
four-step braided 406, 428
two-step briaded 414, 418 ultimate strain 84, 269, 276
woven 371, 422, 423, 424, 425, 426 ultimate strength 82, 141, 202, 221, 268,
textile preform 8, 285, 287, 376 430
thermal bending coefficient 327, 331 unidirectional composite 35, 447
thermal conductivity 24, 29, 61, 73, 194, unit cell 300, 358, 395, 402, 405, 406,
195, 279 407, 408
thermal diffusion 54
thermal effect 4, 54, 57, 61, 76, 153, 198, variance 100
variational approach 186, 189, 195, 198
251, 327
thermal expansion coefficient 57, 73, viscoelasticity 200, 201, 456, 476
void content 428
139, 194, 198, 256, 279, 301, 302,
vulcanization 446
327, 328, 331, 348, 352, 361, 363,
365 warp 289, 311, 318, 327, 335, 350, 356,
thermal shock 30, 73 384, 385, 458
thermal stress 68, 77, 198 warp knitting 293, 389
thermoelastic property 198, 300, 354, wavy fiber 424, 444, 460, 478, 519
405, 423 weakest link 117, 122, 123
thermoplastic 4, 10, 18, 191, 205 weakest link model 117, 259
thermoset 4, 18 wear mechanism 368, 371
thick-section composite 46, 285 wear rate 369, 371
three-dimensional constitutive relation wear resistance 18, 369
46,47
three-dimensional fiber array 215 plain 191, 291, 300, 308, 316, 324, 332,
three-dimensional heat conduction 61 367,420,421,423,458
three-dimensional random fiber model satin 291, 300, 314, 315, 318, 323, 324,
226 332,340,367,420,421,423
three-dimensional stress 150, 160 twill 291, 300
three-dimensional textile composite 285, weaving 287, 300, 374, 382
374, 419 weft 289, 384, 385
three-rail shear test 495 weft knitting 292, 389
through-the-thickness fiber 23, 249, 376, Weibull distribution 101, 105, 111, 119
382, 435 128, 148, 156, 259, 264, 269
tire 443, 445, 452 Weibull parameter 125, 130, 264
track plate 294 wetting 8
traction 484 whisker 3, 169, 183
transformation matrix 47, 185 whitening 137, 146
transient thermal stress 67, 68, 73, 78 work of fracture 253, 434, 435
transport property 24 woven fabric 202, 288 375, 420, 421
transpose 481 angle-interlock 9, 287, 375, 383
transverse crack spacing 139 biaxial, nonorthogonal 365
transverse cracking 80, 126, 136, 138, orthogonal, three-dimensional 375,
150, 153, 366, 367, 431 387, 435
transverse cracking constraint 142 orthogonal, two-dimensional 9, 289,
transverse deformation 29, 46, 48, 49, 356, 375
143, 255, 325, 471 triaxial 292, 302, 356
transverse isotropy 31, 33, 177, 195, 197,
250,311,312,410,417 yarn
transverse Poisson's ratio 31 carrier 374, 377, 379
transverse shear modulus 31 jamming 298, 390, 398, 402, 403, 406
transverse Young's modulus 31, 250 linear density 394, 400, 422, 427
trapezoidal rule 239 packing 402
twist 41, 45, 286, 390, 446 Young's modulus 30, 31, 46, 106, 141,
two-dimensional textile composite 285, 173, 180, 250, 304, 311, 361, 365,
419 412, 422, 446, 474