(Mechanical Engineering, 165) ) Anja Belz, Roger Evans, Paul Piwek - Structural Analysis of Polymeric Composite Materials-CRC Press, Marcel Dekker (2003) PDF

Although great care has been taken to provide accurate and current information, neither the
author(s) nor the publisher, nor anyone else associated with this publication, shall be liable for
any loss, damage, or liability directly or indirectly caused or alleged to be caused by this book. The
material contained herein is not intended to provide specific advice or recommendations for any
specific situation.
Trademark notice: Product or corporate names may be trademarks or registered trademarks and
are used only for identification and explanation without intent to infringe.
Library of Congress Cataloging-in-Publication Data

A catalog record for this book is available from the Library of Congress.
ISBN: 0-8247-4717-8
This book is printed on acid-free paper.
Headquarters
Marcel Dekker, Inc., 270 Madison Avenue, New York, NY 10016, U.S.A.
tel: 212-696-9000; fax: 212-685-4540
Distribution and Customer Service

Marcel Dekker, Inc., Cimarron Road, Monticello, New York 12701, U.S.A.
tel: 800-228-1160; fax: 845-796-1772
Eastern Hemisphere Distribution

Marcel Dekker AG, Hutgasse 4, Postfach 812, CH-4001 Basel, Switzerland
tel: 41-61-260-6300; fax: 41-61-260-6333
World Wide Web

http://www.dekker.com
The publisher offers discounts on this book when ordered in bulk quantities. For more
information, write to Special Sales/Professional Marketing at the headquarters address above.
Copyright n 2004 by Marcel Dekker, Inc. All Rights Reserved.
Neither this book nor any part may be reproduced or transmitted in any form or by any means,
electronic or mechanical, including photocopying, microfilming, and recording, or by any
information storage and retrieval system, without permission in writing from the publisher.
Current printing (last digit):
10 9 8 7 6 5 4 3 2 1
PRINTED IN THE UNITED STATES OF AMERICA
Copyright © 2004 by Marcel Dekker, Inc. All Rights Reserved.

To my wife, Lisa,
my daughters, Stephanie and Jennifer,
my son, Daniel, my parents, James and Rita,
and my brothers and sisters, Mathew, Ann, David, and Claire.

Preface
This textbook is intended for use in a senior-year undergraduate or first-year

graduate course devoted to structural analyses of laminated polymer-matrix
composite materials and structures. Discussion is framed almost entirely at
the macromechanical (structural) level. Micromechanical issues and analyses
are discussed briefly but are not covered in detail. This allows an expanded
coverage of the structural response of composite beams and plates, as com-
pared to other introductory texts on composite materials. The text contains
ample material for a semester-based (15-week) course. I have used a similar
manuscript for several years to support two sequential quarter-based (10-
week) courses, supplementing this material with one or two laboratory
sessions. Since laboratory exercises depend heavily on the equipment and
materials available to the instructor, these lab sessions are not described in
this book.
It is assumed that the reader has already completed a sophomore- or
junior-level course devoted to the mechanics of isotropic solids. I have made
every effort to extend the concepts of this earlier coursework in a natural and
easily understandable way to the study of anisotropic composites.
Chapter 1 begins with a broad overview of the various types of
commercially available metal, ceramic, and polymer-based composite mate-
rials. This chapter also includes an overview of polymer and fibrous materials
and the manufacturing processes used to produce polymer composites and
v

structures. Chapter 2 is devoted to a review of force, stress, and strain tensors
and how these tensors may be transformed from one coordinate system to
another. Although these topics are normally covered in a course on the me-
chanics of isotropic solids, it has been my experience that a review of these
fundamental concepts is almost always required before they can be correctly
applied to the study of anisotropic composites.
Various material properties required to predict the structural perform-
ance of anisotropic composites are introduced in Chapter 3, and the three-
dimensional, anisotropic form of Hooke’s law is developed in Chapter 4. The
three-dimensional form of Hooke’s law is then reduced to the plane stress
(two-dimensional) form in Chapter 5 and applied to unidirectional laminated
composites. Rudimentary elements of plate theory are developed in Chapter 6
and combined with Hooke’s law, resulting in the analysis methodology
commonly known as classical lamination theory (CLT).
Chapter 7 describes composite failure modes and mechanisms, includ-
ing a qualitative description of composite fatigue behaviors and free-edge
effects. The chapter also presents methods of combining macroscopic failure
criteria with CLT to predict first-ply and last-ply failure loads. Chapter 8 is
devoted to statically determinate and indeterminate composite beams. The
chapter begins with the observation that CLT reduces to fundamental beam
theory (as studied during an earlier course on the mechanics of isotropic
solids) when isotropic properties and appropriate dimensions are assumed.
CLT is then used to predict the effective axial and bending rigidities of
composite beams with various cross sections (rectangular, I-, T-, hat-, and
box-beams).
Chapters 9 through 11 address composite plates. Discussion is limited
to symmetric rectangular composite plates, since this topic is extensive and
a complete discussion of nonrectangular and/or nonsymmetric composite
plates or shells deserves a separate text in itself. The equations that govern the
behavior of symmetric and rectangular composite plates are developed in
Chapter 9. Chapter 10 presents several closed-form solutions for specially
orthotropic composite plates. Chapter 11 presents approximate numerical
solutions for generally orthotropic plates (e.g., quasi-isotropic plates); this
includes solutions for the deflections of transversely loaded plates as well as
mechanical and/or thermal buckling due to in-plane loads. Three appendixes
that include material referenced in the main body of the book complete the
text, with the second briefly describing experimental methods used to measure
in-plane composite properties.
While many of these topics are covered in other introductory compo-
sites textbooks, I have included here certain pedagogical features that have, in
my experience, facilitated and enhanced an understanding of the concepts.
For example, the crucially important effects of environment—in particular,

the effects due to changes in temperature and/or moisture content—are
integrated throughout Chapters 3 through 11, rather than being confined to
later chapters, as is often the case in composite textbooks. With the exception
of Chapters 1 and 9, each chapter includes numerical example problems that
illustrate the concepts presented. A solutions manual for all homework
problems posed in the text is available for educators using the text in their
courses. I have also created a suite of computer programs that implement the
analyses discussed. These executable programs may be downloaded free of
charge from the following website:
http : ==depts:washington:edu=amtas=computer:html
These programs are meant to enhance the text and are referenced at ap-
propriate points throughout the book. Of course, composite computer pro-
grams are now widely available, both commercially and otherwise, so the
reader may opt to use resources other than those downloaded from the web-
site provided.
Preparation of this book has been a demanding and lengthy endeavor. I
sincerely hope that it will be a worthy addition to the composites literature.
Mark E. Tuttle

Acknowledgments
I have been fortunate to have worked with talented and inspirational

colleagues throughout my academic and professional career. I would espe-
cially like to acknowledge the lifelong influence of my two major academic
advisors, Professor Emeritus Halbert F. Brinson of the University of Houston
(formerly of Virginia Polytechnic Institute and State University) and Profes-
sor John B. Ligon of Michigan Technological University.
There are many others whom I would like to mention by name, but
space does not allow it. To my friends and colleagues at Battelle Columbus
Laboratories, Michigan Tech, Virginia Tech, the University of Washington,
NASA–Langley Research Center, the Boeing Company, and the Society for
Experimental Mechanics: thank you.
Particular thanks are extended to Mr. Rob Albers of the Boeing
Company, who read and gave helpful critiques of initial versions of Chapters
9 through 11. I would also like to thank the staff at Marcel Dekker, Inc.—in
particular Michael Deters, Production Editor, and John Corrigan, Acquis-
itions Editor—for their very professional and competent help throughout
preparation of the final manuscript. Finally, to the many undergraduate and
graduate students who have taken my composites courses or worked with me
during the pursuit of their degrees and who consequently struggled through
and ‘‘edited’’ manuscript versions of this textbook: thank you for your help
and patience.
Mark E. Tuttle
ix

Contents
Preface
Acknowledgments
1. Introduction
1. Basic Definitions
2. Polymeric Materials
3. Fibrous Materials
4. Commercially Available Forms
5. Manufacturing Processes
6. The Scope of This Book
References
2. Review of Force, Stress, and Strain Tensors

1. The Force Vector
2. Transformation of a Force Vector
3. Normal Forces, Shear Forces, and Free-Body
Diagrams
4. Definition of Stress

5. The Stress Tensor
6. Transformation of the Stress Tensor
7. Principal Stresses
8. Plane Stress
9. Definition of Strain
10. The Strain Tensor
11. Transformation of the Strain Tensor
12. Principal Strains
13. Strains Within a Plane Perpendicular to a Principal
Strain Direction
14. Relating Strains to Displacement Fields
15. Computer Programs 3DROTATE and
2DROTATE
Homework Problems
References
3. Material Properties
1. Anisotropic vs. Isotropic Materials
2. Material Properties That Relate Stress to Strain
3. Material Properties Relating Temperature to
Strain
4. Material Properties Relating Moisture Content to
Strain
5. Material Properties Relating Stress (or Strain) to
Failure
6. Predicting Elastic Composite Properties Based on
Constituents: The Rule of Mixtures
Homework Problems
References
4. Elastic Response of Anisotropic Materials

1. Strains Induced by Stress: Anisotropic Materials
2. Strains Induced by Stress: Orthotropic and
Transversely Isotropic Materials
3. Strains Induced by a Change in Temperature or
Moisture Content
4. Strains Induced by Combined Effects of Stress,
Temperature, and Moisture
Homework Problems
References

5. Unidirectional Composite Laminates Subject to Plane Stress
1. Unidirectional Composites Referenced to the
Principal Material Coordinate System
2. Unidirectional Composites Referenced to an
Arbitrary Coordinate System
3. Calculating Transformed Properties Using Material
Invariants
4. Effective Elastic Properties of a Unidirectional
Composite Laminate
5. Failure of Unidirectional Composites Referenced
to the Principal Material Coordinate System
6. Failure of Unidirectional Composites Referenced
to an Arbitrary Coordinate System
7. Computer Programs UNIDIR and UNIFAIL
Homework Problems
References
6. Thermomechanical Behavior of Multiangle Composite

Laminates
1. Definition of a ‘‘Thin Plate’’ and Allowable Plate
Loadings
2. Plate Deformations: The Kirchhoff Hypothesis
3. Principal Curvatures
4. Standard Methods of Describing Composite
Laminates
5. Calculating Ply Strains and Stresses
6. Classical Lamination Theory (CLT)
7. Simplifications Due to Stacking Sequence
8. Summary of CLT Calculations
9. Effective Properties of a Composite Laminate
10. Transformation of the ABD Matrix
11. Computer Program CLT
Homework Problems
References
7. Predicting Failure of a Multiangle Composite Laminate

1. Preliminary Discussion
2. Free-Edge Stresses
3. Predicting Laminate Failure Using CLT
4. Laminate First-Ply Failure Envelopes

5. Computer Program LAMFAIL
Homework Problems
References
8. Composite Beams
2. Comparing Classical Lamination Theory to
Isotropic Beam Theory
3. Types of Composite Beams Considered
4. Effective Axial Rigidity of Rectangular Composite
Beams
5. Effective Flexural Rigidities of Rectangular
Composite Beams
6. Effective Axial and Flexural Rigidities for
Thin-Walled Composite Beams
7. Statically Determinate and Indeterminate
Axially Loaded Composite Beams
8. Statically Determinate and Indeterminate
Transversely Loaded Composite Beams
9. Computer Program BEAM
Homework Problems
References
9. The Governing Equations of Thin-Plate Theory

2. The Equations of Equilibrium for Symmetric
Laminates
3. Boundary Conditions
4. Representing Arbitrary Transverse Loads as a
Fourier Series
References
10. Some Exact Solutions for Specially Orthotropic Laminates

1. Equations of Equilibrium for a Specially Orthotropic
Laminate
2. In-Plane Displacement Fields in Specially
Orthotropic Laminates
3. Specially Orthotropic Laminates Subject to Simple
Supports of Type S1
4. Specially Orthotropic Laminates Subject to Simple
Supports of Type S4

5. Specially Orthotropic Laminates With Two Simply
Supported Edges of Type S1 and Two Edges of
Type S2
6. The Navier Solution Applied to a Specially
Orthotropic Laminate Subject to Simple Supports
of Type S4
7. Buckling of Rectangular Specially Orthotropic
Laminates Subject to Simple Supports of
Type S4
8. Thermal Buckling of Rectangular Specially
Orthotropic Laminates Subject to Simple Supports
of Type S1
9. Computer Program SPORTHO
References
11. Some Approximate Solutions for Symmetric Laminates

2. In-Plane Displacement Fields
3. Potential Energy in a Thin Composite Plate
4. Symmetric Composite Laminates Subject to Simple
Supports of Type S4
5. Buckling of Symmetric Composite Plates Subject to
Simple Supports of Type S4
6. Computer Program SYMM
References
Appendixes
A. Finding the Cube-Root of a Complex Number
B. Experimental Methods Used to Measure In-Plane Properties
E11, E22, r12, and G12
C. Tables of Beam Deflections and Slopes

1
Introduction
A broad overview of modern composite materials is provided in this chapter.

The chapter begins with a definition of what is meant by the phrase ‘‘com-
posite material.’’ Separate sections devoted to polymeric materials, fibrous
materials, commercially available forms, and manufacturing techniques are
included. The chapter concludes with a section indicating the scope of the
remaining chapters.
1 BASIC DEFINITIONS
This textbook is devoted to a special class of structural materials often called
‘‘advanced’’ composites. Just what is a ‘‘composite material’’? A casual defi-
nition might be: ‘‘A composite material is one in which two (or more) mate-
rials are bonded together to form a third material.’’ Although not incorrect,
upon further reflection, it becomes clear that this definition is far too broad
because it implies that essentially all materials can be considered as ‘‘compo-
sites.’’ For example, the (nominal) composition of the 2024 aluminum alloy is
93.5% Al, 4.4% Cu, 0.6% Mn, and 1.5% Mg [1]. Hence, according to the
broad definition stated above, this common aluminum alloy could be consid-
ered as a ‘‘composite’’ because it consists of four materials (aluminum, cop-
per, manganese, and magnesium) bonded together at the atomic level to form
the 2024 alloy. In a similar sense, virtually all metal alloys, polymers, and

ceramics satisfy this broad definition of a ‘‘composite’’ because all of these
materials contain more than one type of elemental atom.
An important characteristic that is missing in the initial broad definition
is a consideration of physical scale. Another definition of a ‘‘composite mate-
rial,’’ which includes a reference to a physical scale appropriate for present
purposes, is as follows:
A composite material is a material system consisting of two (or more)
materials, which are distinct at a physical scale greater than about
1106 m (1 Am) and which are bonded together at the atomic and/or
molecular levels.
As a point of reference, the diameter of human hair ranges from about
30 to 60 Am. Objects of this size are easily seen with the aid of an optical
microscope. Hence, when composite materials are viewed under an optical
microscope, the distinct constituent materials (or distinct material phases)
that form the composite are easily distinguished. Structural composites
typically consist of a high-strength, high-stiffness reinforcing material, embed-
ded within a relatively low-strength, low-stiffness matrix material. Ideally, the
reinforcing and matrix materials interact to produce a composite whose
properties are superior to either of the two constituent materials alone.
Many naturally occurring materials can be viewed as composites. A
good example is wood and laminated wood products. Wood is a natural
composite, with a readily apparent grain structure. Wood exhibits a higher
stiffness and strength parallel to the grains than transverse to the grains.
In laminated wood products (which range from the large laminated beams
used in a church cathedral to a common sheet of plywood), relatively thin
layers of wood are adhesively bonded together. In plywood, the layers are
arranged such that the grain direction varies from one layer to the next.
Therefore, this laminated wood product has a high stiffness/strength in more
than one direction.
Although composites have been used in a variety of structural applica-
tions for centuries, modern (or ‘‘advanced’’) composites are a relatively recent
development, having been in existence for about 60 years. Modern composites
may be classified according to the size or shape of the reinforcing material
used. Four common classifications of reinforcements are:
Particulates, which are roughly spherical particles with diameters
typically ranging from about 1 to 100 Am
Whiskers, with lengths less than about 10 mm
Short (or ‘‘chopped’’) fibers, with a length ranging from about 10 to
200 mm
Continuous fibers, whose length are, in effect, infinite.

Whiskers, short fibers, and continuous fibers all have very small diam-
eters relative to their length; the diameter of these products ranges from about
5 to 200 Am.
Distinctly different types of composites can be produced using any of
the above reinforcements. For example, three types of composites based on
continuous fibers are shown in Fig. 1: unidirectional composites, woven
composites, and braided composites. In a unidirectional composite, all fibers
are aligned in the same direction and embedded within a matrix material.
In contrast, woven composites are formed by first weaving continuous fibers
into a fabric and then embedding the fabric in a matrix. Hence, a single layer
of a woven composite contains fibers in two orthogonal directions. In con-
trast, a single layer in a braided composite typically contains two or three
Figure 1 Different types of composites based on continuous fibers.

nonorthogonal fiber directions. Braided composites are then formed by
embedding the fabric in a matrix (additional discussion of these types of
composites is provided in Sec. 4).
As implied in Fig. 1, composite products based on continuous fibers are
usually produced in the form of thin layers. A single layer of these products is
called a lamina or ply. The thickness of a single ply formed using unidirec-
tional fibers ranges from about 0.12 to 0.20 mm, whereas the thickness of a
single ply of a woven or braided fabric ranges from about 0.25 to 0.40 mm.
Obviously, a single composite ply is quite thin. To produce a composite
structure with significant thickness, many plies are stacked together to form a
composite laminate. Conceptually any number of plies may be used in the
laminate, but in practice, the number of plies usually ranges from about 10
plies to (in unusual cases) perhaps as many as 200 plies. The fiber represents
the reinforcing material in these composites. Hence, the orientation of the
fibers is, in general, varied from one ply to the next, so as to provide high
stiffness and strength in more than one direction (as is the case in plywood). It
is also possible to use unidirectional, woven, and/or braided plies within the
same laminate. For example, it is common to use a woven or braided fabric as
the two outermost facesheets of a laminate, and to use unidirectional plies at
interior positions.
In all composites, the reinforcement is embedded within a matrix mate-
rial. The matrix may be polymeric, metallic, or ceramic. In fact, composite
materials are often classified on the basis of the matrix material used, rather
than the reinforcing material. That is, modern composites can be catego-
rized into three main types: polymer–matrix, metal–matrix, or ceramic–
matrix composite materials. Usually, the reinforcing material governs the
stiffness and strength of a composite. On the other hand, the matrix material
usually governs thermal stability. Polymeric–matrix composites are used in
applications involving relatively modest temperatures (service temperatures
of, say, 200jC or less). Metal–matrix composites are used at temperatures up
to about 700jC, while ceramic–matrix composites are used at ultra-high
temperatures (up to about 1200jC or greater). The matrix also defines several
additional characteristics of the composite material system. Some additional
roles of the matrix are:
To provide the physical form of the composite
To bind the fibers together
To protect the fibers from aggressive (chemical) environments or
mechanical damage (e.g., due to abrasion, for example)
To transfer and redistribute stresses between fibers, between plies,
and in areas of discontinuities in load or geometry.
To summarize the preceding discussion, there are many types of
composite materials, both natural and man-made. Composites can be clas-

sified according to the physical form of the reinforcing material (particu-
late, whisker, short fiber, or continuous fiber reinforcement), by the type of
matrix material used (metal, ceramic, or polymeric matrix), by the orienta-
tion of the reinforcement (unidirectional, woven, or braided), or by some
combination thereof. The temperature the composite material/structure will
experience in service often dictates the type of composite used in a given
application.
The primary focus of this textbook is the structural analysis of polymeric
composite materials and structures. Metal–matrix and ceramic–matrix com-
posites will not be further discussed, although many of the analysis methods
developed herein may be applied to these types of composites as well. Because
our focus is the structural analysis of polymeric composites, we will not be
greatly considered with the behavior of the individual constituent materials.
That is, we will not be greatly concerned with the behavior of an unreinforced
polymer, nor with the behavior of an individual reinforcing fiber. Instead, we
will be concerned with the behavior of the composite formed by combining
these two constituents. Nevertheless, a structural engineer who wishes to use
polymeric composites effectively in practice must understand at least the
rudiments of polymer and fiber science, in much the same way as an engineer
working with metal alloys must understand at least the rudiments of metal-
lurgy. Toward that end, a brief introduction to polymeric and fibrous mate-
rials is provided in Secs. 2 and 3, respectively. At the minimum, the reader
should become acquainted with the terminology used to describe polymeric
and fibrous materials because such terms have naturally been carried over to
the polymeric composites technical community.
2 POLYMERIC MATERIALS
A brief introduction to polymeric materials is provided in this section. This
introduction is necessarily incomplete. The reader interested in a more de-
tailed discussion is referred to the many available texts and/or web-based
resources devoted to modern polymers (e.g., see Refs. 1–4).
2.1 Basic Concepts

The term ‘‘polymer’’ comes from the Greek words poly (meaning ‘‘many’’)
and mers (meaning ‘‘units’’). Quite literally, a polymer consists of ‘‘many
units.’’ Polymer molecules are made up of thousands of repeating chemical
units and have molecular weights ranging from about 103 to 107.
As an illustrative example, consider the single chemical mer shown in
Fig. 2. This mer is called ethylene (or ethene), and consists of two carbon
atoms and four hydrogen atoms. The two lines between the carbon (C) atoms

Figure 2 The monomer ethylene.
indicate a double covalent bond,* whereas the single line between the hydro-
gen (H) and carbon atoms represents a single covalent bond. The chemical
composition of the ethylene mer is sometimes written as C2H4 or CH2jCH2.
Under proper conditions, the double covalent bond between the two
carbon atoms can be converted to a single covalent bond, which allows each
of the two carbon atoms to form a new covalent bond with a suitable neigh-
boring atom. A suitable neighboring atom would be a carbon atom in a
neighboring ethylene mer, for example. If ‘‘n’’ ethylene mers join together
in this way, the chemical composition of the resulting molecule can be
represented C2nH4n, where n is any positive integer. Hence, a ‘‘chain’’ of
ethylene mers joins together to form the well-known polymer, polyethylene
(literally, ‘‘many ethylenes’’), as shown in Fig. 3. The process of causing
a monomer to chemically react and form a long molecule in this fashion is
called polymerization.
The single ethylene unit is an example of a monomer (‘‘one mer’’). At
room temperature, a bulk sample of the ethylene monomer is a low-viscosity
fluid. If two ethylene monomers bond together, the resulting chemical entity
has two repeating units and is called a dimer. Similarly, the chemical entity
formed by three repeating units is called a trimer. The molecular weight of a
dimer is twice that of the monomer, the molecular weight of a trimer is three
times that of the monomer, etc. Prior to polymerization, most polymeric
materials exist as relatively low-viscosity fluids known as oligomers (‘‘a few
mers’’). The individual molecules within an oligomer possess a range of
molecular weights, typically containing perhaps 2–20 mers.
It should be clear from the above discussion that a specific molecular
weight cannot be assigned to a polymer. Rather, the molecules within a bulk
sample of a polymer are of differing lengths and hence exhibit a range in
molecular weight. The average molecular weight of a bulk sample of a polymer
is increased as the polymerization process is initiated and progresses. Another
* As fully described in an introductory chemistry text, a ‘‘covalent bond’’ is formed when two
atoms share an electron pair, so as to fill an incompletely filled valence level.

Figure 3 The polymer polyethylene.
measure of the ‘‘size’’ of the polymeric molecule is the degree of polymer-

ization, defined as the ratio of the average molecular weight of the polymer
molecule divided by the molecular weight of the repeating chemical unit
within the molecular chain.
The average molecular weight of a polymeric sample (or, equivalently,
the degree of polymerization) depends on the conditions under which it was
polymerized. Now, all physical properties exhibited by a polymer (e.g.,
strength, stiffness, density, thermal expansion coefficient, etc.) are dictated
by the average molecular weight. Therefore, a fundamental point that must be
appreciated by the structural engineer is that the properties exhibited by any
polymer (or any polymeric composite) depend on the circumstances under
which it was polymerized.
As a general rule, the volume of a bulk sample of a monomer decreases
during polymerization. That is, the bulk sample shrinks as the polymerization
process proceeds. This may have serious ramifications if the polymer is to be
used in structural applications. For example, if a fiber(s) is embedded within
the sample during the polymerization process (as is the case for some fiber-
reinforced polymeric composite systems), then shrinkage of the matrix causes
residual stresses to develop during polymerization. This effect contributes to
so-called cure stresses, which are present in most polymeric composites. As
will be seen in later chapters, cure stresses arise from two primary sources. The
first is shrinkage of the matrix during polymerization, as just described.
The second is stresses that arise due to temperature effects. In this case, the
composite is polymerized at an elevated temperature (say, 175jC) and then
cooled to room temperature (20jC). The thermal expansion coefficient of the
matrix is typically much higher than the fibers, and so during cooldown, the
matrix is placed in tension while the fiber is placed in compression. Cure
stresses due to shrinkage during cure and/or temperature effects can be quite
high relative to the strength of the polymer itself, and ultimately contribute
toward failure of the composite.

2.2 Addition vs. Condensation Polymers
Although in the case of polyethylene the repeat unit is equivalent to the
original ethylene monomer, this is not always the case. In fact, in many in-
stances, the repeat unit is derived from two (or more) monomers. A typical
example is nylon 6,6. A typical polymerization process for this polymer is
shown schematically in Fig. 4. Two monomers are used to produce nylon 6,6:
hexamethylene diamine (chemical composition: C6H16N2) and adipic acid
(chemical composition: CO2H(CH2)4CO2H). Note that the repeat unit of
nylon 6,6 (hexamethylene adipamide) is not equivalent to either of the two
original monomers.
A low-molecular-weight by-product (i.e., H2O) is produced during the
polymerization of nylon 6,6. This is characteristic of condensation polymers.
That is, if both a high-molecular-weight polymer as well as a low-molecular-
weight by-product are formed during the polymerization process, the polymer
is classified as a condensation polymer. Conversely, addition polymers are
those for which no by-product is formed, which implies that all atoms present
in the original monomer(s) occur somewhere within the repeat unit. Generally
speaking, condensation polymers shrink to a greater extent during the poly-
merization process than do addition polymers. Residual stresses caused by
shrinkage during polymerization are often a concern in structural compo-
sites, and hence difficulties with residual stresses can often be minimized if
an addition polymer is used in these applications.
Figure 4 The polymer nylon 6,6.

2.3 Molecular Structure
The molecular structure of a fully polymerized polymer can be roughly
grouped into one of three major types: linear, branched, or crosslinked poly-
mers. The three types of molecular structure are shown schematically in
Fig. 5.
Linear polymers can be visualized as beads on a string, where each bead
represents a repeat unit. It should again be emphasized that the length of these
‘‘strings’’ is enormous; if a typical linear molecule were scaled up to be 10 mm
in diameter, it would be roughly 4 km long. In a bulk sample, these long
macromolecules become entangled and twisted together, much like a bowl of
cooked spaghetti. Obviously, as the molecular weight (i.e., the length) of the
polymer molecule is increased, the number of entanglements is increased. At
the macroscopical scale, the stiffness exhibited by a bulk polymer is directly
related to its molecular weight and number of entanglements.
If all of the repeat units within a linear polymer are identical, the
polymer is called a homopolymer. Polyethylene is a good example of a linear
homopolymer. However, it is possible to produce linear polymers that consist
of two separate and distinct repeat units. Such materials are called copoly-
mers. In linear random copolymers, the two distinct repeat units appear
randomly along the backbone of the molecule. In contrast, for linear block
copolymers, the two distinct repeat units form long continuous segments
within the polymer chain. A good example of a common copolymer is
acrylonitrile–butadiene–styrene, commonly known as ‘‘ABS.’’
The second major type of polymeric molecular structure is the branched
polymer (see Fig. 5). In branched polymers, relatively short side chains are
bonded to the primary backbone of the macromolecule by means of a co-
valent bond. As before, the stiffness of a bulk sample of a branched polymer is
related to the number of entanglements between molecules. Because the
branches greatly increase the number of entanglements, the macroscopical
stiffness of a branched polymer will, in general, be greater than the macro-
scopical stiffness of a linear polymer of similar molecular weight. In many
branched polymers, the branches consist of the same chemical repeat unit
as the backbone of the molecular chain. However, in some cases, the branch
may have a distinctly different chemical repeat unit than the main backbone
of the molecule. Such materials are called graft copolymers.
Finally, the third major type of molecular structure is the crosslinked
or network polymer (see Fig. 5). During polymerization of such polymers,
a crosslink (i.e., a covalent bond) is formed between individual molecular
chains. Hence, once polymerization is complete, a vast molecular network
is formed. In the limit, a single ‘‘molecule’’ can no longer be identified. A
bulk sample of a highly crosslinked polymer may be thought of as a single
molecule.

Figure 5 Types of polymer molecular structure.

Linear and branched polymers can be visualized as a bowl of cooked
spaghetti. One can imagine that a single spaghetti noodle could be extracted
without damage from the bowl if the noodle were pulled slowly and carefully,
allowing the noodle to ‘‘slide’’ past its neighbors. In much the same way, an
individual molecule could also be extracted (at least conceptually) from a
bulk sample of a linear or branched polymer. However, this is not the case for
a fully polymerized crosslinked polymer. Because the ‘‘individual’’ molecular
chains within a crosslinked polymer are themselves linked together by
covalent bonds, the entire molecular network can be considered to be a
single molecule. Although regions of the chain in a crosslinked polymer may
slide past each other, eventually, relative motion between segments is limited
by the crosslinks between segments.
2.4 Thermoplastic vs. Thermoset Polymers

Suppose a bulk sample of a linear or branched polymer exists as a solid
material at room temperature and is subsequently heated. Due to the increase
in thermal energy, the average distance between individual molecular chains is
increased as temperature is increased. This results in an increase in molecular
mobility and a decrease in macroscopical stiffness. That is, as the molecules
move apart, both the forces of attraction between individual molecules as well
as the degree of entanglement decrease, resulting in a decrease in stiffness at
the macroscopical level. Eventually, a temperature is reached at which the
polymeric molecules can slide freely past each other and the polymer ‘‘melts.’’
Typically, melting does not occur at a single temperature, but rather over a
temperature range of about 15–20jC. A polymer that can be melted (i.e., a
linear or branched polymer) is called a thermoplastic polymer.
The molecular structure of a thermoplastic polymer may be amor-
phous or semicrystalline. The molecular structure of an amorphous ther-
moplastic is completely random (i.e., the molecular chains are randomly
oriented and entangled, with no discernible pattern). In contrast, in a semi-
crystalline thermoplastic, there exist regions of highly ordered molecular
arrays. An idealized representation of a crystalline region is shown in Fig. 6.
As indicated, in the crystalline region, the main backbone of the molecular
chain undulates back and forth such that the thickness of the crystalline
region is usually (about) 100 Å. The crystalline region may extend over an
area with a length dimension ranging from (about) 1000 to 10,000 Å.
Hence, the crystalline regions are typically platelike. The high degree of
order within the crystalline array allows for close molecular spacing, and
hence exceptionally high bonding forces between molecules in the crystal-
line region. At the macroscale, a semicrystalline thermoplastic typically
has a higher strength, stiffness, and density than an otherwise comparable

Figure 6 An idealized representation of a crystalline region in a thermoplastic
polymer.
amorphous thermoplastic. No thermoplastic is completely crystalline, how-

ever. Instead, regions of crystallinity are surrounded by amorphous regions,
as shown schematically in Fig. 7. Most semicrystalline thermoplastic are
10–50% amorphous (by volume).
As just described, a thermoplastic can be melted. In contrast, a cross-
linked polymer cannot be melted. If a crosslinked polymer is heated, it will
exhibit a decrease in stiffness at the structural level because the average
distance between individual segments of the molecular network is, in fact,
increased as temperature is increased. However, the crosslinks do not allow
indefinite relative motion between segments, and eventually limit molecular
motion. Therefore, a crosslinked polymer will not melt. Of course, if the
temperature is raised high enough, the covalent bonds that form both the
backbone of the molecular chains as well as the crosslinks are broken,
chemical degradation occurs, and the polymer is destroyed. A polymer that
cannot be melted (i.e., a crosslinked polymer) is called a thermoset polymer.
Three more or less distinct conditions are recognized during polymer-
ization of a thermoset polymer. The original resin or oligomer is typically a
low-viscosity, low-molecular-weight fluid, containing molecules with perhaps
2–10 repeat units. A thermoset resin is said to be A-staged when in this form.
As the polymerization process is initiated (by the introduction of a catalyst,
by an increase in temperature, by the application of pressure, or by some
combination thereof), molecular weight and viscosity increase rapidly. If the
polymerization process is then halted in some manner (say, by suddenly
reducing the temperature), the polymerization process will stop (or be

Figure 7 Overall molecular structure of a semicrystalline thermoplastic,
showing crystalline and amorphous regions.
dramatically slowed) and the polymer will exist in an intermediate stage. The
thermoset resin is said to be B-staged when in this form. If the polymerization
process is allowed to resume (say, by reheating) and continue until the
maximum possible molecular weight has been reached, the thermoset is said
to be C-staged (i.e., the polymer is fully polymerized).
Suppliers of composites based on thermoset polymers initially B-stage
their product and sell it to their customers in this form. This requires that the
B-staged composite be stored by the customer for months at low temperatures
(typically at temperatures below about 15jC or 0jF). Refrigeration is
required so that the thermosetting resin does not polymerize beyond the B-
stage during storage. The polymerization process is reinitiated and completed
(i.e., the composite is C-staged) during the final fabrication of a composite
part, typically through the application of heat and pressure. Most commer-
cially available thermoset composites are C-staged (or ‘‘cured’’) at a temper-
ature of either 120jC or 175jC.
In contrast, composites based on thermoplastic polymers do not require
refrigeration. In this case, the matrix is fully polymerized when delivered to
the customer, and may be stored for months or years without degradation.
Heat and pressure are applied during the final fabrication of a thermoplastic
composite part, but no chemical reaction occurs. That is, heat is applied

simply to soften/melt the thermoplastic matrix, and pressure is applied to
insure consolidation of the composite part. The temperature required to
soften/melt thermoplastic composites is usually 350jC or higher. Note that
the temperatures required to process thermoplastic composites are signifi-
cantly higher than those required to cure thermoset composites.
2.5 The Glass Transition Temperature
Stiffness and strength are physical properties of obvious importance to the
structural engineer. These properties are temperature-dependent for all
materials, but this is particularly true for polymers. The effect of temperature
on the stiffness of a polymer is summarized in Fig. 8. Thermoset and
thermoplastic polymers exhibit the same general behavior, except that at
high temperatures, thermoplastics melt whereas thermosets do not. All
polymers exhibit a decrease in stiffness near a characteristic temperature
called the glass transition temperature, Tg. At temperatures well below the Tg,
polymer stiffness decreases slowly with an increase in temperature. At
temperatures well below the Tg, polymers are ‘‘glassy’’ and brittle. In
contrast, at temperatures well above the Tg, all polymers are ‘‘rubbery’’
and ductile. Thus, the Tg denotes the transition between glassy and rubbery
behaviors. This transition is associated with a sudden increase in mobility of
segments within the molecular chain, and typically occurs over a range of 10–
15jC. At temperatures well below the Tg, the polymer molecules are closely
packed together and tightly bonded, and cannot easily slide past each other.
Figure 8 Effects of temperature on polymer stiffness.

Consequently, the polymer is ‘‘glassy’’ and exhibits high stiffness and strength
but relatively low ductility. Conversely, at temperatures well above the Tg, the
molecular spacing is increased such that the molecular chains (or segments of
those chains) are mobile and can readily slide past each other. Consequently,
the polymer is ‘‘rubbery’’ and exhibits relatively lower stiffness and strength
but higher ductility. As implied in Fig. 8, for amorphous thermoplastics, the
change in stiffness (and other physical properties) that occurs as the Tg is
reached may be one to two orders of magnitude. This astonishing decrease in
stiffness occurs over a temperature range of only a few degrees. A similar
change in properties occurs for semicrystalline thermoplastics and crosslinked
thermosets, although, in general, they are less pronounced. If temperature is
raised high enough, then a thermoplastic polymer will melt. The temperature
region at which melting occurs is denoted Tm in Fig. 8, although, as previously
discussed, a thermoplastic does not exhibit a unique melting temperature but
rather melts over a temperature range. Young’s modulus tends toward zero as
melting occurs, as implied in Fig. 8. Thermoset polymers cannot be melted,
although the polymer is destroyed if temperature is raised to an excessively
high level.
The Tg has been illustrated in Fig. 8 by demonstrating the change in
stiffness as temperature is increased. However, many other characteristic
physical properties (density, strength, thermal expansion coefficient, heat ca-
pacity, etc.) also change sharply at this transition. Hence, the Tg can be
measured by monitoring any of these physical properties as a function of
temperature. The Tg exhibited by a few common polymers is listed in Table 1.
Table 1 Approximate Glass Transition Temperatures for Some Common Polymers
Typical glass transition

temperature General character
at room temperature
Polymer jC jF (22jC or 70jF)
Silicone rubber 123 190 Rubbery

Polybutadiene 85 120 Rubbery
Polyisoprene 50 60 Rubbery
Nylon 6,6 50 122 Rigid
Polyvinyl chloride (PVC) 85 185 Rigid
Acrylonitrile–butadiene–styrene 90 195 Rigid
Polystyrene 100 210 Rigid
Polyester 150 300 Glassy
Epoxy 175 350 Glassy
Polyetheretherketone (PEEK) 200 400 Glassy
Polyetherimide 215 420 Glassy

Note that knowledge of the Tg allows an immediate assessment of the general
nature of the polymer at room temperature.
3 FIBROUS MATERIALS
Reinforcing fibers are the major strengthening element in all polymeric
composites. A brief introduction to these materials is presented in this section.
The reader interested in additional details is referred to Refs. 5 and 6.
Common continuous fiber materials are:
Glass
Aramid
Graphite or carbon
Polyethylene
Boron
Silicon carbide.
In all cases, the fiber diameters are quite small, ranging from about 5 to
12 Am for glass, aramid, or graphite fibers; from about 25 to 40 Am for
polyethylene fibers; and from about 100 to 200 Am for boron and silicon
carbide fibers.
Some of the terminologies used to describe fibers will be defined here.
The terms fiber and filament are used interchangeably. An end (also called a
strand) is a collection of a given number of fibers gathered together. If the
fibers are twisted, the collection of fibers is called a yarn. The ends are
themselves gathered together to form a tow (also called roving). The fibers are
usually coated with a size (also called a finish). The size is applied for several
reasons, such as:
To bind the fibers in the strand
To lubricate the fibers during fabrication
To serve as a coupling and wetting agent to insure a satisfactory
adhesive bond between the fiber and matrix materials.
There may be thousands of filaments in a single tow and, in fact, tow
sizes are often described in terms of thousands of filaments per tow. For
example, a ‘‘6k tow’’ implies that the tow consists of 6000 individual fibers.
Properties of several types of glass fibers, organic fibers, carbon fibers,
and silicon carbide fibers will be briefly described in the following subsec-
tions. It will be seen that:
Young’s modulus (stiffness) ranges from about 70 GPa for glass
fibers to 700 GPa (or higher) for carbon fibers.

Comparable tensile strengths can be obtained using any of the fibers
listed.
Specific gravity ranges from about 0.97 for polyethylene fiber to
about 2.5 for glass fibers.
Elongation at failure ranges from about 0.3% for some carbon
fibers to about 5% for glass fibers.
The mechanism responsible for these high stiffnesses and strengths

differs from one fiber type to another. For glass fibers, the process of draw-
ing to very small diameters simply reduces the number and size of flaws in
the material, thus increasing strength. For example, glass fibers with a
diameter of about 1 mm (0.04 in.) will commonly have a strength of about
170 MPa (25 ksi); but if this same glass is drawn to a diameter of 10 Am
(0.0004 in.), a strength of about 3450 MPa (500 ksi) will be achieved. For
organic fibers, strengthening is accomplished by stretching the fiber and
thereby aligning the polymer molecules. This produces fibers that are them-
selves anisotropic. For carbon or graphite fibers, strengthening is accom-
plished by aligning the basal planes of adjoining crystals, also producing
an anisotropic fiber.
3.1 Glass Fibers

Glass fibers are fabricated by melting glass marbles at a temperature of
about 1260jC (2300jF) and drawing the melt through platinum bushings
followed by a rapid cooldown and secondary drawing. A sizing is applied
to the fibers, which are then combined into a strand and wound onto a
spool. The two major types of glass fibers are called ‘‘E-glass’’ and ‘‘S-glass.’’
E-glass (alumino borosilicate) is so named (‘‘e’’lectrical glass) because of its
high electrical resistivity. S-glass (magnesium aluminosilicate) is so named
(‘‘s’’tructural glass) because of its high tensile strength. Glass fibers are
usually isotropic. Some mechanical and physical properties typical of E-glass
and S-glass fibers are listed in Table 2.
3.2 Aramid Fibers

The aramid polymer fiber produced by DuPont Corp. and marketed under
the trade name Kevlark is perhaps the most widely used organic fiber. This
fiber is based on poly( p-phenylene terephthalamide), which is a member of a
family of polymers called ‘‘aramids.’’ Aramid fibers are formed by a con-
densation/elongation process. The resulting fibers are highly anisotropic
because strong covalent bonds are formed in the fiber direction, whereas
weak hydrogen bonds are formed in the transverse direction. This chemical

Table 2 Typical Properties of Glass Fibers
Property E-glass S-glass
Specific gravity 2.60 2.50

Young’s modulus, GPa (Msi) 72 (10.5) 87 (12.6)
Tensile strength, MPa (ksi) 3450 (500) 4310 (625)
Tensile elongation, % 4.8 5.0
Coefficient of thermal expansion, 5.0 (2.8) 5.6 (3.1)
Am/m/jC (Ain./in./jF)
bonding arrangement results in the anisotropic behavior exhibited at the

macroscale. That is, aramid fibers have very high tensile strength, stiffness,
and toughness in the axial direction of the fiber, but relatively low tensile
strength and stiffness in the transverse direction.
Kevlar is commercially available in the following grades:
Kevlar 149: a high-performance, aerospace grade fiber with the
highest modulus of all Kevlar fibers
Kevlar 49: a high-performance, aerospace grade fiber with the
highest strength of all Kevlar fibers
Kevlar 129: a relatively inexpensive fiber with a lower strength and
stiffness than Kevlar 149 and 49, but with a higher percent elongation
Kevlar 29: a relatively inexpensive fiber with a strength and stiffness
lower than Kevlar 129, but a higher percent elongation.
Nominal mechanical and physical properties of Kevlar fibers are listed
in Table 3. Of particular interest is the negative coefficient of thermal
Table 3 Typical Properties of Kevlar Fibers (All Properties in Axial Direction of Fiber)
Kevlar Kevlar Kevlar Kevlar

Property 149 49 129 29
Specific gravity 1.44 1.44 1.44 1.44

Young’s modulus, 186 (27) 124 (18) 96 (13.9) 68 (9.8)
GPa (Msi)
Tensile strength, 3440 (500) 3700 (535) 3380 (490) 2930 (425)
MPa (ksi)
Tensile elongation, % 2.5 2.8 3.3 3.6
Coefficient of thermal 2.0 (1.1) 2.0 (1.1) 2.0 (1.1) 2.0 (1.1)
expansion, Am/m/jC
(Ain./in./jF)

expansion in the fiber direction. This implies that an increase in temperature
produces a decrease in the length of a Kevlar 49 fiber. This behavior produced
unexpected behavior during cooldown from cure in early applications of the
fiber. Further research indicates that Kevlar 49 has a high positive coefficient
of thermal expansion in the transverse direction, further demonstrating the
anisotropic nature of the fiber.
3.3 Graphite and Carbon Fibers

The terms ‘‘graphite’’ and ‘‘carbon’’ are often used interchangeably within
the composites community. The elemental carbon content of either type of
fiber is above 90%, and the stiffest and strongest fibers have carbon contents
approaching 100%. Some effort has been made to standardize these terms by
defining graphite fibers as those that have:
A carbon content above 95%
Been heat-treated at temperatures in excess of 1700jC (3100jF)
Been stretched during heat treatment to produce a high degree of
preferred crystalline orientation
A Young’s modulus on the order of 345 GPa (50 Msi).
Fibers that do not satisfy all of the above conditions are called carbon
fibers under this standard. However, as stated above, in practice, this defi-
nition is not widely followed, and the terms ‘‘graphite’’ and ‘‘carbon’’ are
often used interchangeably.
Both graphite and carbon fibers are produced by thermal decomposi-
tion of an organic (i.e., polymeric) fiber or ‘‘precursor’’ at high pressures and
temperatures. The three most common precursors are:
Polyacrylonitrile (‘‘PAN’’)
Pitch (a by-product produced during the petroleum distillation
process)
Rayon.
The PAN fiber is probably the most widely used.

Details of the specific steps followed during fabrication of a fiber are
proprietary and can only be described in a general manner. During the
fabrication process, the precursor is first drawn into a thread and then
oxidized at about 260jC (500jF) to form crosslinks and an extended carbon
network. The precursor is then subjected to a carbonization treatment, during
which noncarboneous atoms are driven off. This step typically involves
temperatures of approximately 700jC (1290jF), and is usually conducted
in an inert atmosphere. Finally, during the graphitization process, the fibers
are subjected to a combination of high temperature and tensile elongation.

The maximum temperature reached during this step determines, in large
part, the strength and/or stiffness that will be achieved. The graphite crystal
is highly anisotropic, with strong covalent bonds in the basal plane and
weak Van der Waals (‘‘secondary’’) bonds perpendicular to the basal plane.
High strengths and stiffnesses are attained by causing the basal planes to be
aligned in the fiber direction. This can be accomplished by controlled stretch-
ing of the precursor during fabrication.
Major developmental efforts have resulted in the ability to produce
fibers with a wide range of stiffnesses and strengths. Fibers are sometimes
classified in terms of the stiffness (i.e., elastic modulus). Mechanical and
physical properties typical of low-modulus, intermediate-modulus, and ultra-
high-modulus fibers are listed in Table 4.
3.4 Polyethylene Fibers

A high-strength, high-modulus polyethylene fiber called Spectrak was
developed at Allied Signal Technologies during the 1980s. Spectra is based
on ultra-high-molecular-weight polyethylene (UHMWPE). It has a specific
gravity of 0.97, meaning that it is the only reinforcing fiber available that is
lighter than water. Spectra is available in three classifications (Spectra 900,
1000, and 2000) and several grades are available in each class. Nominal
properties are listed in Table 5. The high specific strength of the fiber makes
it particularly attractive for tensile applications. The glass transition tem-
perature of UHMWPE is in the range of 20jC to 0jC, and hence the fiber
is in the rubbery state at room temperatures and exhibits time-dependent
(viscoelastic) behavior. This feature imparts outstanding impact resistance
and toughness, but may lead to undesirable creep effects under long-term
sustained loading. The melting temperature of the fiber is 147jC (297jF),
Table 4 Typical Properties of Commercially Available Graphite Fibers
Low Intermediate Ultra-high

Property modulus modulus modulus
Specific gravity 1.8 1.9 2.2

Young’s modulus, 230 (34) 370 (53) 900 (130)
GPa (Msi)
Tensile strength, 3450 (500) 2480 (360) 3800 (550)
MPa (ksi)
Elongation, % 1.1 0.5 0.4
Coefficient of thermal 0.4 (0.2) 0.5 (0.3) 0.5 (0.3)
expansion, Am/m/jC
(Ain./in./jF)

Table 5 Nominal Properties of Spectra Fibers
Property Spectra 900 Spectra 1000 Spectra 2000
Specific gravity 0.97 0.97 0.97

Young’s modulus, 70 (10) 105 (15) 115 (17)
GPa (Msi)
Tensile strength, 2600 (380) 3200 (465) 3400 (490)
MPa (ksi)
Elongation, % 3.8 3.0 3.0
Coefficient of thermal >70 (>38) >70 (>38) >70 (>38)
expansion, Am/m/jC
(Ain./in./jF)
and hence the use of polyethylene fibers is limited to relatively modest

temperatures. The thermal expansion coefficients of Spectra fibers have
apparently not been measured. The values listed in Table 5 are estimated
based on the properties of bulk high-molecular-weight polyethylene.
3.5 Boron Fibers

Boron fibers were one of the first high-performance fibers available for use in
composites. They are fabricated by depositing boron on a heated core using
the vapor deposition process. Both tungsten and carbon fiber cores have
been used. Boron fiber diameters range from 0.1 to 0.2 mm (0.004–0.008 in),
which is an order of magnitude larger than glass, aramid, or graphite fibers.
Boron fibers have a Young’s modulus of about 410 GPa (60 Msi) and a tensile
strength of about 3450 MPa (500 ksi). The combination of a large diameter
and high stiffness greatly restricts the bend radius to which the fiber may
be subjected. On the other hand, a large fiber diameter and high modulus
of elasticity contribute to excellent compressive performance for boron-
reinforced composites.
3.6 Ceramic Fibers

The single most outstanding feature offered by ceramic fibers is that they are
resistant to extremely high temperatures while still maintaining competitive
structural properties. An example is a silicon carbide fiber marketed under
the trade name SCS-Ultrak and fabricated by Specialty Materials, Inc.
(Lowell, MA), which can operate at temperatures up to 1200jC (2190jF).
This fiber has a modulus of 415 GPa (60 Msi), a strength of 5865 MPa
(850 ksi), and a specific gravity of 3.0. A second example is an aluminum–
boron–silica fiber fabricated by the 3M Company and marketed under the
trade name Nextelk. This fiber is capable of operating at temperatures up

to 1650jC (3000jF). It also exhibits excellent properties, with a modulus of
193 GPa (28 Msi), a strength of 2000 MPa (300 ksi), and a specific gravity of
3.03.
Because ceramic fibers are normally used at temperatures far in excess of
the useable temperature range of polymers, they are rarely used in polymeric
composites. Ceramic fibers will not be further discussed in this book.
4 COMMERCIALLY AVAILABLE FORMS

4.1 Discontinuous Fibers
Virtually all of the continuous fibers described in Sec. 3 are also available in
the form of discontinuous fibers. Discontinuous fibers are embedded within a
matrix, and may be randomly oriented (in which case the composite is
isotropic at the macroscale) or may be oriented to some extent (in which case
the composite is anisotropic at the macroscale). Orientation of discontinuous
fibers, if it occurs, is usually induced during the fabrication process used to
create the composite material/structure; fiber alignment often mirrors the
flow direction during injection molding, for example. Discontinuous fibers
are roughly classified according to length, as follows:
Milled fibers are produced by grinding the continuous fiber into very
short lengths. For example, milled graphite fibers are available with
lengths ranging from about 0.3 to 3 mm (0.0012–0.12 in.), and milled
glass fibers are available with lengths ranging from about 0.4 to 6 mm
(0.0016–0.24 in.).
Chopped fibers (or strands) have a longer length than milled fibers,
and composites produced using chopped fibers usually have higher
strengths and stiffnesses than those produced using milled fibers.
Chopped graphite fibers are available with lengths ranging from
about 3 to 50 mm (0.–2.0 in.), while chopped glass fibers are available
with lengths ranging from about 6 to 50 mm (0.24–2.0 in.).
In general, the mechanical properties of a composite produced using
discontinuous fibers (say, the strength or stiffness) are not as good as those
that can be obtained using continuous fibers. However, discontinuous fibers
allow the use of relatively inexpensive, high-speed manufacturing processes
such as injection molding or compression molding, and have therefore been
widely used in applications in which extremely high strength or stiffness is not
required.
One of the most widely used composites systems based on the use of
discontinuous fibers is known generically as ‘‘sheet molding compound’’
(SMC). In its most common form, SMC consists of chopped glass fibers

embedded within a thermosetting polyester resin. However, other resin sys-
tems (e.g., vinyl esters or epoxies) as well as other fibers (e.g., chopped graph-
ite or aramid fibers) are occasionally used in SMC material systems.
4.2 Roving Spools

Most continuous fiber types are available in the form of spools of roving
(i.e., roving wound onto a cylindrical tube and ultimately resembling a large
spool of thread). As mentioned in Sec. 3, roving is also known as tow. The
size of tow (or roving) is usually expressed in terms of the number of fibers
contained in a single tow. For example, a specific glass fiber might be avail-
able in the form of 2k, 3k, 6k, or 12k tow. In this case, the product is avail-
able in tows containing from 2000 to 12,000 fibers. Fibers purchased in this
form are usually ‘‘dry’’ and are subsequently combined with a polymer, metal,
or ceramic matrix during a subsequent manufacturing operation such as
filament winding or pultrusion.
4.3 Woven Fabrics

Most types of high-performance continuous fibers can be woven to form a
fabric. Weaving is accomplished using looms specially modified for use with
high-performance fibers, which are stiffer than those customarily used in the
textile industry. Woven fabrics are produced in various widths up to about
120 cm (48 in.), and are available in (essentially) infinite lengths. Two terms
associated with woven fabrics are:
The tow or yarn running along the length of the fabric is called the
warp. The warp direction is parallel to the long axis of the woven
fabric.
The tow or yarn running perpendicular to the warp is called the fill
tow (also called the weft or the woof tow). The fill direction is per-
pendicular to the warp direction.
Some common fabric weaves are shown schematically in Fig. 9. The
plane weave (Fig. 9a) is the simplest fabric pattern and is most commonly
used. The plane weave is produced by repetitively weaving a given warp tow
over one fill tow and under the next. The point at which a tow passes over/
under another tow is called a crossover point. The plane weave pattern results
in a very stable and firm fabric that exhibits minimum distortion (e.g., fiber
slippage) during handling.
A family of woven fabric patterns known as satin weaves provide better
drape characteristics than the plane weave pattern. That is, a satin weave is
more pliable and will more readily conform to complex curved surfaces than
plane weaves. In the crowfoot satin weave (Fig. 9b), one warp tow is woven

Figure 9 Some common woven fabrics used with high-performance fibers.
(From Ref. 7.)
over three successive fill tows and then under one fill tow. In the five-harness
satin weave, one warp tow passes over four fill tows and then under one fill
tow. Similarly, in an eight-harness satin weave, one warp tow passes over seven
fill tows and then under one fill tow.
The stiffness and strength of woven fabrics are typically less than that
achieved with unidirectional fibers. This decrease is due to fiber waviness.

That is, in any woven fabric, the tow is required to pass over/under one (or
more) neighboring tow(s) at each crossover point, resulting in a pre-existing
fiber waviness. Upon application of a tensile load, the fibers within a ply tend
to straighten, resulting in a lower stiffness than would be achieved if the ply
contained straight unidirectional fibers. Further, due to the weave pattern, the
fibers are not allowed to straighten fully and are subjected to bending stresses,
resulting in fiber failures at lower tensile loads than would otherwise be
achieved if the ply contained unidirectional fibers. This effect is most pro-
nounced in the case of simple weaves because each tow passes over/under
each neighboring tow. For simple weaves, the through-thickness distribu-
tion of tow in the warp and fill directions is identical. Consequently, the
strength and stiffness of simple weaves are usually identical in the warp and
fill directions.
In contrast, for satin weaves, the through-thickness distribution of tow
is inherently asymmetric. Referring to Fig. 9a, for example, for the five-
harness satin weave pattern, the warp tow is primarily within the ‘‘top’’ half of
the fabric (as sketched), whereas the fill tow is primarily within the lower half.
The asymmetrical through-thickness distribution of tow causes a coupling
between in-plane loading and bending deflections. That is, if a uniform tensile
load is applied to the midplane of a single layer of a satin weave fabric, the
fabric will not only stretch but will also deflect out of plane (i.e., bend).
Similarly, the crossover points are not symmetrically located with respect to
either the warp or fill directions. This causes a coupling between in-plane
loading and in-plane shear strain. That is, an in-plane shear strain is induced if
a uniform tensile load is applied to a single layer of a satin weave fabric [7].
A woven fabric is, in essence, a 2D structure consisting of orthogo-
nal warp and fill tows interlaced within a plane. Weaving or stitching sev-
eral layers of a woven fabric together can produce a woven structure with
a significant thickness. Structures produced in this fashion are called ‘‘3D
weaves.’’
4.4 Braided Fabrics

Note from Sec. 4.3 that woven fabrics contain reinforcing tow in two orthog-
onal directions—the warp and fill directions. In contrast, braided fabrics
typically contain tow oriented in two (or more) nonorthogonal directions.
Three common braiding patterns are shown in Fig. 10. It is apparent from
this figure that a braided fabric contains bias tow that intersects at a total
included angle 2a. The angle a is called either the braid angle or the bias
angle. While the braid angle can be varied over a wide range, there is always
some minimum and maximum possible value that depends on the width of
the tow and details of the braiding equipment used. Note that if a=45j,

Figure 10 Some common braided fabrics used with high-performance fibers.
(From Ref. 7.)

then the bias tows are in fact orthogonal and the braided fabric shown in Fig.
10a and b is equivalent to a woven fabric. A braided fabric is described using
the designation nn, where n is the number of tows between crossover points.
A 11-bias and a 22-bias braided fabric is shown in Fig. 10a and b,
respectively. A 11 triaxial braided fabric is shown in Fig. 10c. In this case, a
third axial tow is present.
Braided fabrics can be produced in tubular form or as a flat braided
fabric. A concise description of the equipment used to produce braided fab-
rics, as well as a discussion of the maximum and minimum braid angles that
can be achieved, is given in Ref. 8. It is also possible to braid 3D structures,
in which tows are braided to form a (fibrous) structure that is subsequently
infused with a matrix to form the composite. Applications include ‘‘I’’ or
‘‘T’’ cross sections, typically used with resin transfer molding (RTM) to pro-
duce composite stiffeners or beams. In contrast to structures produced using
fabrics (which may be unidirectional, woven, or braided fabrics), in a 3D
braided structure, there are no recognizable layers.
4.5 Preimpregnated Products or ‘‘Prepreg’’

As is obvious from the preceding discussion, at some point during the
fabrication of a polymer composite, the reinforcing fiber must be embedded
within the polymeric matrix. One approach is to combine the fiber and resin
during the manufacturing operation in which the final form of the composite
structure is defined. Three manufacturing processes in which this approach is
taken are filament winding (briefly described in Sec. 5.3), pultrusion (Sec. 5.4),
and resin transfer molding (Sec. 5.5).
An alternative approach is to combine the fiber and matrix in an
intermediate step, resulting in an intermediate product. In this case, either
individual tow or a thin fabric of tow (which may be woven or braided) is
embedded within a polymeric matrix and delivered to the user in this form.
Because the fibers have already been embedded within a polymeric matrix
when delivered, the fibers are said to have been ‘‘preimpregnated’’ with resin,
and products delivered in this condition are commonly known as ‘‘prepreg.’’
The method used to impregnate a large number of unidirectional tows
with resin is illustrated in Fig. 11 [9]. As indicated, tows delivered from a
large number of roving spools are arranged in a relatively narrow band. The
tows are passed through a resin bath and then wound onto a roll. An inert
backing sheet (called a scrim cloth) is placed between layers on the roll to
maintain a physical separation between layers and to aid during subsequent
handling and processing. The tows/fibers are subjected to various surface
pretreatments just prior to entering the resin bath. The pretreatments are
proprietary but are intended to cause good wetting of the fiber by the resin,

Figure 11 (a) Method used to impregnate unidirectional tows with resin. (b) A
3-in. wide roll of ‘‘prepreg tape.’’
which ultimately helps ensure good adhesion between the fibers and polymer
matrix in the cured composite. Products produced in this fashion are
commonly known as ‘‘prepreg tape’’ (Fig. 11b). Prepreg tape is available
in width ranging from about 75 to 1220 mm (3–48 in.). Prepreg fabrics,
produced using either woven or braided fabrics instead of unidirectional
tows, are produced using similar techniques and are also available in widths
ranging from about 75 to 1220 mm.
A variety of fabrication methods have been developed based on the
use of prepreg materials. A few such techniques will be described in Sec. 5.
The first commercially successful prepreg materials were based on B-
staged epoxy resins. As discussed in Sec. 2, in the B-staged condition, a
thermoset resin has been partially polymerized, resulting in relatively high
viscosity, which aids in handling B-staged prepreg materials. However, pre-
preg must be kept at low temperatures until used, otherwise the resin con-
tinues to polymerize and slowly harden. This required that the prepreg be
shipped to the user in a refrigerated condition (for small amounts, this is

often accomplished using insulated shipping containers and dry ice). Fur-
thermore, the user must keep the stock of prepreg on hand refrigerated until
used. Typically, storage temperatures are required to be 15jC (0jF) or
below. In practice, the prepreg material stock is removed from freezer, the
amount of prepreg necessary is removed from the roll of stock, and the re-
maining stock is returned to the freezer. Hence, the cumulative ‘‘out-time’’
that a given roll of prepreg stock has experienced (i.e., total amount of time a
roll has been out of a freezer) must be monitored and recorded. The need to
store thermoset prepreg in a refrigerated condition and to maintain accurate
records of cumulative out-time is a significant disadvantage because these
factors add significantly to the final cost of the composite structure.
More recently, prepreg materials based on thermoplastic resins have
become commercially available. In this case, the polymeric matrix is a fully
cured thermoplastic polymer, and hence the prepreg does not require refrig-
eration during shipping or storage, which is a distinct advantage.
Heat and/or pressure is applied during the final fabrication of a com-
posite based on prepreg materials. In the case of thermoset prepregs, heat
and pressure serve to complete the polymerization of the polymeric resin
(i.e., the composite is ‘‘C-staged’’). For thermoplastic prepreg, the objective
is not necessarily to complete the polymerization but rather to melt the ther-
moplastic matrix so as to consolidate individual plies within the laminate.
5 MANUFACTURING PROCESSES
Fiber-reinforced composites may be produced using metallic, ceramic, or
polymeric matrices. However, polymeric composites are the primary focus
of this book, and so techniques used to fabricate metal or ceramic composite
structures will not be discussed. Even with this limitation, a complete review
of the many different manufacturing processes used to produce polymeric
composite materials and structures is beyond the scope of this presentation.
Instead, only the most common manufacturing techniques will be described.
5.1 Layup
Many composites are produced using the tapes or fabrics discussed in Sec. 4.1.
These may be unidirectional tape, woven fabrics, or braided fabrics. These
products are all relatively thin. ‘‘Layup’’ simply refers to the process of
stacking several layers together, much like a deck of cards. Stacking several
layers together produces a laminate of significant thickness. The most direct
method of producing a multi-ply composite laminate is to simply stack the
desired number of layers of fabric by hand, referred to as ‘‘hand layup.’’ The
layers may consist of either ‘‘dry’’ fabrics (i.e., fabrics that have not yet been

impregnated with a resin) or prepreg materials. As will be discussed in later
chapters, fiber angles are typically varied from one ply to the next, so as to
insure adequate stiffness in more than one direction.
Whereas hand layup is simple and straightforward, it is labor-intensive
and therefore costly. It can also be very cumbersome if a large structure is
being produced, such as a fuselage panel intended for use in a modern
commercial aircraft. Therefore, various computer-controlled machines that
automate the process of assembling the ply stack using prepreg materials
have been developed. These include tow placement and tape placement
machines (see Fig. 12). In either case, a roll (or rolls) of prepreg material is
mounted on the head of a computer-controlled robot arm or gantry. The
appropriate number of layers of prepreg is placed on a tool surface auto-
matically and in the desired orientation. Although the capital costs of mod-
ern tow placement or tape placement machines may be very high, overall, this
approach is often less costly than hand layup if production quantities are
sufficiently high.
In the case of dry fabrics (which are usually either woven or braided
fabrics), the stack must be impregnated with a low-viscosity polymeric resin
following assembly of the fiber stack. Conceptually, this may be accomplished
by pouring liquidous resin over the dry fiber stack and using a squeegee or
Figure 12 A computer-controlled tape-laying machine, used to produce the

composite skin used in the vertical stabilizer for a Boeing 777 aircraft.
(Copyright n The Boeing Company.)

similar device to assist the resin to wet the fibers within the stack. This
technique is commonly used in the recreational boat-building industry, for
example. However, as can be imagined, it is very difficult to insure uniform
penetration of the resin and wetting of the fibers through the thickness of
the stack, to insure that no air pockets remain trapped in the stack, and to
avoid distortion of the fiber patterns while forcing resin into the fibrous
assembly. There are also potential health issues associated with continually
exposing workers to nonpolymerized resins. Hence, the technique of im-
pregnating a dry fiber stack using handheld tools such as squeegees is rarely
employed in industries requiring low variability in stiffness and strength and/
or high volumes, such as the aerospace or automotive industries, for exam-
ple. Alternate methods of impregnating a dry fiber stack with resin have been
developed, such as resin transfer molding (discussed in Sec. 5.5). These alter-
nate techniques result in a composite with a much more uniform matrix vol-
ume fraction and almost no void content.
A major advantage of using prepreg materials, of course, is that the
fibers have been impregnated with resin a priori. Therefore, it is much easier to
maintain the desired matrix volume fraction and to avoid entrapped air-
pockets. Further, prepreg material based on a B-staged thermoset are typi-
cally ‘‘tacky’’ (i.e., prepreg materials adhere to neighboring plies much like
common masking tape), and hence once a given ply has been placed in the
desired orientation, it is less likely to move or be distorted relative to neigh-
boring plies than is the case with dry fabrics.
5.2 Autoclave Process Cycles

Following layup (which may be accomplished using hand layup, automatic
tow placement or tape placement machines, or other techniques), the indi-
vidual plies must be consolidated to form a solid laminate. Usually, con-
solidation occurs through the application of pressure and heat. Although a
simple hot press can be used for this purpose, applying pressure and heat
using an autoclave produces highest-quality composites. An autoclave is
simply a closed pressure vessel that can be used to apply a precisely con-
trolled and simultaneous cycle of vacuum, pressure, and elevated tempera-
ture to the laminate during the consolidation process.
Although many variations exist, a typical assembly used to consolidate
a laminate using an autoclave is shown in Fig. 13. Some of the details of the
assembly are as follows:
The final shape of the composite is defined by a rigid tool. A simple
flat tool is shown in Fig. 13, but in practice, the tool is rarely flat
but instead mirrors the contour(s) desired in the final product. For
example, the surface of a tool used to produce the skin of an air-

Figure 13 Typical assembly used to consolidate a polymeric composite lami-
nate using an autoclave (expanded edge view).
plane wing would possess the curvature(s) necessary to provide an

aerodynamic surface. Various materials may be used to produce
the tool, including steel alloys, aluminum alloys, ceramics, or com-
posite materials.
The tool surface is coated with a release agent. Various liquid or wax
release agents are available, which are either sprayed or wiped onto
the surface. The purpose of the release agent is to prevent adherence
between the tool and the polymeric matrix.
A peel ply is placed next to both upper and lower surfaces of the
composite laminate. The release ply does not develop a strong bond
to the composite, and hence can be easily removed following con-
solidation. The peel ply may be porous or nonporous. Porous peel
plies allow resin to pass through the ply and be adsorbed by an
adjacent bleeder/breather cloth (see below). Note that the surface
texture of the consolidated laminate will be a mirror image of the
peel ply used. For example, Teflon-coated porous glass fabrics are
often used as peel plies, and these fabrics have a clothlike surface
texture. Hence, a composite laminate consolidated with such fab-
rics will exhibit a clothlike surface texture as well.
One or more layers of a breather/bleeder cloth is placed adjacent to
the porous peel ply. The bleeder cloth has the texture of a rather stiff
cotton ball. Its purpose is to allow any gases released to be vented
(hence the adjective breather), and also to adsorb any resin that
passed through the porous peel ply (hence the adjective bleeder). The
breather/bleeder is usually a glass, polyester, or jute cloth.

An edge dam is placed around the periphery of the laminate. The edge
dam is intended to maintain the position and resin content of the
laminate edges.
A pressure plate (also called a caul plate) is placed over the breather/
bleeder cloth. The pressure plate insures a uniform distribution of
pressures over the surface of the laminate.
The entire assembly is sealed within a vacuum film or bag. Often this
is a relatively thick (say, 5 mm) layer of silicone rubber. Sealant tape
is used to adhere the vacuum film to the tool surface, providing a
pressure-tight seal around the periphery of the vacuum film.
The volume within the vacuum bag is evacuated by means of a vac-
uum port, which is often permanently attached to the silicone rubber
vacuum film. The vacuum port often features a quick-disconnect
fitting, which allows for easy connection to a vacuum pump or line.
Following vacuum bagging of the laminate, the assembly is placed within
an autoclave, the autoclave is sealed, and the thermomechanical process cycle
that will consolidate the composite is initiated. A bagged composite laminate
being loaded into an autoclave is shown in Fig. 14.
The thermomechanical process cycle imposed using an autoclave varies
from one composite prepreg system to the next, and also depends on part
Figure 14 A vacuum-bagged composite skin used in the tail-section of a

Boeing 777 aircraft, about to be loaded into a large autoclave. (Copyright n The
Boeing Company.)

configuration (e.g., part thickness). Recall that if the prepreg is based on a B-
stage thermoset resin, then the autoclave is used to complete the polymer-
ization of the resin (i.e., the composite is C-staged). Alternatively, if the
prepreg is based on a thermoplastic, the pressure and heat applied during the
autoclave cycle soften the matrix and insure polymer flow across the ply
interfaces. The laminate is then solidified upon cooling.
A typical cure cycle, suitable for use with standard thermosetting resin
systems such as epoxies, is as follows:
Draw and hold a vacuum within the vacuum bag, resulting in a
pressure of roughly 100 kPa (14.7 psi) applied to the laminate. The
vacuum is typically maintained for about 30 min, and is intended to
remove any entrapped air or volatiles, and to hold the laminate in
place.
While maintaining a vacuum, increase the temperature from room
temperature to about 120jC (250jF), at a rate of about 2.8jC/min
(5jF/min). Maintain this temperature for 30 min. During this 30-min
dwell time, any remaining air or other volatiles are removed.
After 30 min, increase autoclave pressure from atmospheric to about
585 kPa (85 psi), at a rate of 21 kPa/min (3 psi/min). Release vacuum
when autoclave pressure reaches 138 kPa (20 psi).
When an autoclave pressure of 585 kPa is reached, increase the
temperature from 120jC to 175jC (350jF), at a rate of about 2.8jC/
min (5jF/min). Maintain temperature at 175jC for 2 hr. Polymer-
ization of the thermosetting resin matrix is completed during this 2-hr
dwell.
Cool to room temperature at a rate of about 2.8jC/min (5jF/min),
release autoclave pressure, and remove cured laminate from the
autoclave.
Process cycles used with thermoplastic prepregs are similar, except that
higher temperatures (500jC or higher) are involved.
5.3 Filament Winding

Filament winding is an automated process in which tow is wound onto a
mandrel at controlled position and orientation. A filament winder being used
to produce a small rocket motor case is shown in Fig. 15 [10]. During
operation, the mandrel rotates about its axis, and a fiber carriage simulta-
neously moves in a controlled manner along the length of the mandrel. The
angle at which fibers are placed on the mandrel surface is a function of
the mandrel diameter, rate of mandrel rotation, and translational speed of the
fiber carriage. The mandrel can include domed heads to accommodate fiber

Figure 15 A carbon-epoxy pressure vessel being produced using the filament-
winding process and unidirectional prepreg. The prepreg used in this case is
based on large 50k tows of Zoltek Panexo 35 carbon fiber. (a) A band of prepreg is
R
wound onto a mandrel, forming a Fhj fiber pattern. (b) Eventually the mandrel is
completely enclosed by one or more Fhj plies. (c) One or more 90j (‘‘hoop’’) plies
are often added to the cylindrical region to resist the high hoop stresses induced
in cylindrical pressure vessels. (Photos provided courtesy of Entec Composite
Machines Inc., Salt Lake City, UT.)

turnaround at the ends, or to wind the domes of a cylindrical pressure vessel
(as shown in Fig. 15).
If dry tows are used, then the tow must pass through a liquid resin
bath before being wound onto the mandrel. In this case, the process is often
referred to as ‘‘wet’’ winding. Often fiber tension provides sufficient com-
paction of the laminate, and so no additional external pressure is required.
If a thermosetting resin that cures at room temperature is used, following
completion of the winding operation, the structure is simply left in the
winder until polymerization is complete.
Of course, if prepreg tow is used, then the tow is already impregnated
with a resin and is not passed through a resin bath. This process is called
‘‘dry’’ winding. In this case, heat and pressure are normally required to com-
plete polymerization of the resin (in the case of a thermosetting polymer
matrix), or to cause resin flow and consolidation (in the case of a thermo-
plastic polymer matrix). The appropriate heat and pressure are usually
applied using an autoclave.
Filament winding machines are available in highly automated, nu-
merically controlled models (costing millions of dollars), but high-quality
filament winding can also be accomplished for simple shapes and patterns on
inexpensive gear/chain-driven machines similar to a lathe.
For simple wound shapes with open ends (such as tubes), the mandrel is
usually a simple solid cylinder whose surface has been coated with a release
agent. In this case, the mandrel is forced out of the internal cavity after
consolidation of the composite. Mandrel design and configuration become
more complex when a shape with restricted openings at the ends is produced
(such as the pressure tank shown in Fig. 15). In these cases, the mandrel must
somehow be removed after the part is consolidated. Several different types of
mandrel designs are used in these cases, including:
Soluble mandrels, which are made from a material that can be
dissolved in some fashion after the cure process is complete. In this
approach, the mandrel is cast and machined to the desired shape, the
composite part is filament wound over the mandrel, the part is cured,
and the mandrel is then simply dissolved. The wall of the composite
structure must obviously have at least one opening, such that the
dissolved (and now liquidous) mandrel material can be drained from
the internal cavity. Soluble mandrels can be made from metallic
alloys with suitably low-melting temperatures, eutectic salts, sand
with water-soluble binders, or various plasters.
Removable (or collapsible) mandrels, which resemble giant 3D puzzles.
That is, the entire mandrel can be taken apart piece by piece. The
composite structure being wound must have at least one wall opening,

which allows the mandrel pieces to be removed from the internal
cavity after cure. Obviously, the mandrel is designed such that no
single piece is larger than the available opening(s).
‘‘Inflatable’’ mandrels, which take on the desired shape when
pressurized and then are simply deflated and removed after winding
and consolidation.
Metal or polymer liners, which are actually a modification of the
inflatable mandrel concept. Liners can be described as metal or
polymer ‘‘balloons’’ and remain in the filament wound vessel after
cure. The liner does not contribute significantly to the strength or
stiffness of the structure. In fact, the wall thickness of the liner is often
so small that an internal pressure must be applied to the liner during
the winding process to avoid buckling of the liner wall. Metal liners
are almost always used in composite pressure vessels, where
allowable leakage rates are very low, or in filament wound chemical
storage tanks, where corrosive liquids are stored.
5.4 Pultrusion
Pultrusion is a fabrication process in which continuous tows or fabrics
impregnated with resin are pulled through a forming die, as shown schemati-
cally in Fig. 16 [10]. If dry tow or fabric is used, then the tow/fabric must
Figure 16 Sketch of a typical pultruder.

pass through a resin bath prior to entering the forming die. In this case, the
process is called ‘‘wet’’ pultrusion. If prepreg material is used, then there
is no need for a resin bath and the process is called ‘‘dry’’ pultrusion. The
cross-sectional shape is defined by the die and is therefore constant along
the length of the part. The principal attraction of pultrusion is that very
high production rates are possible, as compared to other composite manu-
facturing techniques.
5.5 Resin Transfer Molding

In the resin transfer molding process, a dry fiber preform is placed within a
cavity formed between two rigid molds, as shown in Fig. 17. The dry pre-
form may consist of a 3D braided structure, or may be made by stitching
together multiple layers of 2D woven or braided fabrics. Liquidous resin is
forced into the cavity under pressure via a port located in the upper or lower
mold halves. Air originally within the internal cavity (or other gases that
evolve during cure of the resin) is allowed to escape via one or more air
vents. Alternatively, a vacuum pump may be used to evacuate the internal
cavity, which also assists in drawing the resin into the cavity. When a vacuum
is used, the process is called ‘‘vacuum-assisted resin transfer molding’’
(VARTM). Both the upper and lower molds must be sufficiently rigid so
as to resist the internal pressures applied and to maintain the desired shape
of the internal cavity. Usually, the closed molds are placed within a press,
which provides a clamping pressure to assist in keeping the molds closed.
6 THE SCOPE OF THIS BOOK

A broad overview of modern composite materials has been provided in the
preceding sections. It should be clear from this discussion that modern
Figure 17 Picture/sketch of the RTM process.

polymeric composite material systems are a multidisciplinary subject, involv-
ing topics drawn from polymer chemistry, fiber science, surface chemistry and
adhesion, materials testing, structural analysis, and manufacturing tech-
niques, to name a few. It is simply not possible to cover all of these topics
in any depth in a single book. Accordingly, the material presented in this book
represents a small fraction of the scientific and technological developments
that have ultimately led to the successful use of modern composite material
systems. Specifically, the focus of this text is the structural analysis of
laminated, continuous-fiber polymeric composite materials and structures.
Having identified the structural analysis of laminated continuous-fiber
polymeric composites as our focus, we must make still another decision:
At what physical scale should we frame our analysis? The importance of
physical scale has already been discussed in Sec. 1 in conjunction with the
very definition of a ‘‘composite material.’’ Specifically, we have defined a
composite as a material system consisting of two (or more) materials, which
are distinct at a physical scale greater than about 1 Am and which are bonded
together at the atomic and/or molecular levels. Fibers commonly used in
polymeric composites possess diameters ranging from about 5 to 40 Am
(Sec. 3). Therefore, we could perform a structural analysis at a physical scale
comparable to the fiber diameter. Alternatively, laminated polymeric com-
posites consist of well-defined layers (called ‘‘plies’’) of fibers embedded in
a polymeric matrix. The thickness of these layers ranges from about 0.125
to 0.250 mm (Sec. 4). We could therefore elect to begin a structural analy-
sis at a physical scale comparable to the thickness of a single ply.
A distinction is drawn between structural analyses that begin at these
two different physical scales. Analyses that are framed at a physical scale
corresponding to the fiber diameter (or below) are classified as microme-
chanics analyses, whereas those framed at a physical scale corresponding to
a single ply thickness (or above) are classified as macromechanics analyses.
This distinction is comparable to the traditional distinction between metal-
lurgy and continuum mechanics. That is, metallurgy typically involves the
study of the crystalline nature of metals and metal alloys, and is therefore
framed at a physical scale roughly corresponding to atomic dimensions. A
metallurgist might attempt to predict Young’s modulus* of a given metal
alloy, based on knowledge of the constituent atoms and crystalline struc-
ture present in the alloy, for example. In contrast, continuum mechanics is
formulated at a much larger physical scale, such that the existence of indi-
* The definition of various material properties of interest to the structural engineer, such as
Young’s modulus, will be reviewed and discussed in greater detail in Chap. 3.

vidual atoms is not recognized. In continuum mechanics, a metal or metal
alloy is said to be ‘‘homogeneous’’ even though it actually consists of sev-
eral different atomic species. A structural engineer wishing to apply a solu-
tion based on continuum mechanics would simply measure Young’s
modulus exhibited by the metal alloy of interest, rather than try to predict
it based on knowledge of the atomic crystalline structure.
In much the same way, composite micromechanics analyses are con-
cerned with the predicting properties of composites based on the particular
fiber and matrix materials involved, the spacing and orientation of the fibers,
the adhesion (or lack thereof ) between fiber and matrix, etc. For example,
suppose that a unidirectional graphite–epoxy composite is to be produced by
combining graphite fibers with a known Young’s modulus (Ef) and an epoxy
matrix with a known Young’s modulus (Em). An analysis framed at a physi-
cal scale corresponding to the fiber diameter (i.e., a micromechanics analy-
sis) is required to predict the Young’s modulus that will be exhibited by the
composite (Ec) formed using these two constituents.
In contrast, composite macromechanics analyses are framed at a
physical scale corresponding to the ply thickness (or above). The existence
or properties of individual fibers or the matrix material are not recognized
(in a mathematical sense) in a macromechanics analysis. Instead, the ply is
treated as a homogenous layer whose properties is identical at all points,
although they differ in different directions. Details of fiber or matrix type,
fiber spacing, fiber orientation, etc., are represented in a macromechanics
analysis only indirectly, via properties defined for the composite ply as a
whole, rather than as properties of the individual constituents.
Micromechanics-based structural analyses will not be discussed in any
detail. A simple micromechanics model that may be used to predict ply stiff-
nesses based on knowledge of fiber and matrix properties, called the rule of
mixtures, will be developed in Sec. 6 in Chap. 3. However, the material de-
voted to micromechanics in this text is abbreviated and does not do justice
to the many advances made in this area. The lack of emphasis on microme-
chanical topics is not meant to imply that such analyses are unimportant.
Quite the contrary, micromechanics analyses are crucial during development
of new composite material systems because it is only through a detailed
understanding of the behavior of composites at this physical scale that new
and improved materials can be created. Micromechanics has been mini-
mized herein simply due to space restrictions. The reader interested in learn-
ing more about micromechanics is referred to several excellent texts that
cover this topic in greater detail, a few of which are Refs. 5, 7, 11, and 14.
Finally, then, the scope of this book is macromechanics-based structural
analysis of laminated, continuous-fiber polymeric composites.

REFERENCES
1. Rodriguez, F. Principles of Polymer Systems; 3rd Ed.; Hemisphere Publ. Co.:
New York, 1989; ISBN 0-89116-176-7.
2. Young, R.J.; Lovell, P.A. Introduction to Polymers; 2nd Ed.; Chapman and
Hall Publ. Co.: New York, 1991; ISBN 0-89116-176-7.
3. Strong, A.B. Plastics: Materials and Processing; 2nd Ed.; Prentice-Hall: Upper
Saddle River, NJ, 2000. ISBN 0-13-021626-7.
4. The Macrogalleria (http://www.psrc.usm.edu/macrog/index.htm), website
maintained by the University of Southern Mississippi and devoted to polymeric
materials.
5. Watt, W., Perov, B.V., Eds. Strong Fibres; Vol. 1. In Handbook of Composite
Materials; Kelly, A., Rabotnov, Y.N., Series Eds.; Elsevier Sci. Publ.: New
York, NY, 1985; ISBN 0-444-87505-0.
6. Donnet, J.-B.; Wang, T.K.; Peng, J.C.M.; Reboillat, M. Carbon Fibers; 3rd Ed.;
Marcel Dekker, Inc.: New York, NY, 1998; ISBN 0-8247-0172-0.
7. Cox, B.; Flanagan, G. Handbook of Analytical Methods for Textile Compo-
sites. NASA Contractor Report 4750. NASA-Langley Res. Ctr., Hampton,
VA, 1997.
8. Hasselbrack, S.A.; Pederson, C.L.; Seferis, J.C. Evaluation of carbon–fiber
reinforced thermoplastic matrices in a flat braid process. Polym. Compos. 1992,
13 (1), 38–46.
9. Kalpakjian, S. Manufacturing Processes for Engineering Materials; 3rd Ed.;
Addison-Wesley Longman, Inc.: Menlo Park CA, 1997; ISBN 0-201-82370-5.
10. Schwartz, M.M. Composite Materials Handbook; New York, NY, McGraw-Hill
Book Co.: 1983; ISBN 0-07-055743-8.
11. Hyer, M.W. Stress Analysis of Fiber-Reinforced Composite Materials; New
York, NY,McGraw-Hill Book Co.: 1998; ISBN 0-07-016700-1.
12. Herakovich, C.T. Mechanics of Fibrous Composites; John Wiley and Sons: New
York, NY, 1998; ISBN 0-471-10636-4.
13. Jones, R.M. Mechanics of Composite Materials; McGraw-Hill Book Co.: New
York, NY, 1975; ISBN 0-07-032790-4.
14. Hull, D. An Introduction to Composite Materials; Cambridge University Press:
Cambridge, UK, 1981; ISBN 0-521-23991-5.

2
Review of Force, Stress, and Strain Tensors
In this chapter, the fundamental definitions of force vectors, stress tensors,

and strain tensors are reviewed. The chapter begins with a discussion of force
vectors, since the concept of ‘‘force’’ is encountered in everyday life and is
therefore very intuitive. Separate sections devoted to stress tensors and strain
tensors are then presented. Certain parallels will be drawn between force
vectors and stress/strain tensors. An important underlying principal is that a
tensor cannot be described in a mathematical sense until a specific coordinate
system is selected for use. Also, a tensor cannot be properly described using
only a single component of the tensor, i.e., all components of a tensor must be
known in order to describe the tensor.
1 THE FORCE VECTOR

Forces can be grouped into two broad categories: surface forces and body
forces. Surface forces are those that act over a surface (as the name implies)
and result from direct physical contact between two bodies. In contrast, body
forces are those that act at a distance and do not result from direct physical
contact of one body with another. The force of gravity is the most common
type of body force. In this text, we are primarily concerned with surface forces;
the effects of body forces (such as the weight of a structure) will be ignored.

A force is a 3-D vector. A force is defined by a magnitude and a line of
action. In SI units, the magnitude of a force is expressed in Newtons,
abbreviated N, whereas in English units, the magnitude of a force is expressed
in pounds-force, abbreviated lbf. A force vector F acting at a point P and
referenced to a right-handed x–y–z coordinate system is shown in Fig. 1.
Components of F acting parallel to the x–y–z coordinate axes, Fx, Fyy, and
Fz, respectively, are also shown in the figure. The algebraic sign of each force
component is defined in accordance with the algebraically positive direction
of the corresponding coordinate axis. All force components shown in Fig. 1
are algebraically positive since each component ‘‘points’’ in the correspond-
ing positive coordinate direction.
The reader is likely to have encountered several different ways of ex-
pressing force vectors in a mathematical sense. Three methods will be de-
scribed here. The first is called vector notation and involves the use of unit
vectors. Unit vectors parallel to the x-, y-, and z-coordinate axes are typically
labeled ıˆ, j,̂ and kˆ, respectively, and by definition have a magnitude equal to
unity. A force vector F is written in vector notation as follows:
F ¼ Fx iˆ þ Fy jˆ þ Fz k̂ ð1Þ
The magnitude of the force is given by:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
jF j ¼ Fx2 þ Fy2 þ F 2z ð2Þ
Figure 1 A force vector F acting at point P. Force components Fx, Fy, and Fz
acting parallel to the x–y–z coordinate axes, respectively, are also shown.

A second method of defining a force vector is through the use of indicial
notation. In this case, a subscript is used to denote individual components of
the vectorial quantity:
F ¼ ðFx ; Fy ; Fz Þ
The subscript denotes the coordinate direction of each force component. One
of the advantages of indicial notation is that it allows a shorthand notation to
be used, as follows:
F ¼ Fi ; where i ¼ x; y; or z ð3Þ
Note that a range has been explicitly specified for the subscript i in Eq. (3).
That is, it is explicitly stated that the subscript i may take on values of x, y, or
z. Usually, however, the range of a subscript(s) is not stated explicitly but
rather is implied. For example, Eq. (3) is normally written simply as:
F ¼ Fi
where it is understood that the subscript i takes on values of x, y, and z.
The third approach is called matrix notation. In this case, individual
components of the force vector are listed within braces in the form of a column
array:
8 9
< Fx =
F ¼ Fy ð4Þ
: ;
Fz
Indicial notation is sometimes combined with matrix notation as follows:
F ¼ fFi g ð5Þ
2 TRANSFORMATION OF A FORCE VECTOR

One of the most common requirements in the study of mechanics is the need to
describe a vector in more than one coordinate system. For example, suppose
all components of a force vector Fi are known in one coordinate system (say,
the x–y–z coordinate system), and it is desired to express this force vector in a
second coordinate system (say, the xV–yV–zV coordinate system). In order to
describe the force vector in the new coordinate system, we must calculate the
components of the force parallel to the x V-, y V-, and zV-axes—that is, we must
calculate FxV, FyV, and FzV. The process of relating force components in one
coordinate system to those in another coordinate system is called the trans-
formation of the force vector. This terminology is perhaps unfortunate in the
sense that the force vector itself is not ‘‘transformed,’’ but rather our de-

scription of the force vector transforms as we change from one coordinate
system to another.
It can be shown (1,2) that the force components in the xV–yV–zV co-
ordinate system ( FxV, FyV, and FzV) are related to the components in the x–y–z
coordinate system ( Fx, Fy, and Fz) according to:
FxV ¼ cxVx Fx þ cx Vy Fy þ cxVz Fz ð6aÞ
Fy V ¼ cy Vx Fx þ cy Vy Fy þ cy Vz Fz
FzV ¼ czVx Fx þ czVy Fy þ czVz Fz
The terms ci Vj that appear in Eq. (6a) are called direction cosines and equal the
cosine of the angle between the axes of the new and original coordinate
systems. An angle of rotation is defined from the original x–y–z coordinate
system to the new xV–yV–zV coordinate system. The algebraic sign of the angle
of rotation is defined in accordance with the right-hand rule.
Equation (6a) can be succinctly written using the summation conven-
tion as follows:
FiV ¼ ci Vj Fj ð6bÞ
Alternatively, these three equations can be written using matrix notation as:
8 9 2 38 9
< FxV = cxVx cx Vy cxVz < Fx =
F ¼ 4 cy Vx cy Vy cy Vz 5 Fy ð6cÞ
: yV ; : ;
FzV czV x czVy czVz Fz
Note that although values of individual force components vary as we

change from one coordinate system to another, the magnitude of the force
vector [given by Eq. (2)] does not. The magnitude is independent of the
coordinate system used and is called an invariant of the force tensor.
Direction cosines relate unit vectors in the ‘‘new’’ and ‘‘old’’ coordinate
systems. For example, a unit vector directed along the xV-axis (i.e., unit vector
ıˆV) is related to the unit vectors in the x–y–z coordinate system as follows:
iVˆ ¼ cx Vx iˆ þ cxVy jˆ þ cx Vz k̂ ð7Þ

Since iˆ V is a unit vector, then in accordance with Eq. (2):
ðcx Vx Þ2 þ ðcxVy Þ2 þ ðcxVz Þ2 ¼ 1 ð8Þ

To this point, we have referred to a force as a vector. A force vector can
also be called a force tensor. The term ‘‘tensor’’ refers to any quantity that
transforms in a physically meaningful way from one Cartesian coordinate
system to another. The rank of a tensor equals the number of subscripts that
must be used to describe the tensor. A force can be described using a single

subscript, Fi, and therefore a force is said to be a tensor of rank one, or equiv-
alently, a first-order tensor. Equations (6a)–(6c) is called the transformation
law for a first-order tensor.
It is likely that the reader is already familiar with two other tensors: the
stress tensor, rij, and the strain tensor, eij. The stress and strain tensors will be
reviewed later in this chapter, but at this point, it can be noted that two
subscripts are used to describe stress and strain tensors. Hence, stress and
strain tensors are said to be tensors of rank two, or equivalently, second-order
tensors.
Example Problem 1
Given. All components of a force vector F are known in a given x–y–z co-
ordinate system. It is desired to express this force in a new xU–yU–zU
coordinate system, where the xU–yU–zU is generated from the original x–y–
z coordinate system by the following two rotations (see Fig. 2):
A rotation of h about the original z-axis (which defines an inter-
mediate xV–yV–zV coordinate system), followed by
A rotation of b about the xV-axis (which defines the final xU–yU–zU
coordinate system).
Problem. (a) Determine the direction cosines ciUj relating the original x–y–z
coordinate system to the new xU–yU–zU coordinate system; (b) obtain a general
expression for the force vector F in the xU–yU–zU coordinate system; and (c)
calculate numerical values of the force vector F in the xU–yU–zU coordinate
system if h=20j, b=60j, and Fx=1000 N, Fy=200 N, Fz=600 N.
Solution
Part (a). One way to determine the direction cosines ciUj is to rotate unit
vectors. In this approach, unit vectors are first rotated from the original x–y–z
coordinate system to the intermediate xV–yV–zV coordinate system, and then
from the xV–yV–zV system to the final xU–yU–zU coordinate system.
Define a unit vector I that is aligned with the x-axis:
I u ð1Þiˆ
That is, vector I is a vector for which Ix=1, Iy=0, and Iz=0. The vec-
tor I can be rotated to the intermediate xV–yV–zV coordinate system using
Eqs. (6a)–(6c):
IxV ¼ cxV x Ix þ cxV y Iy þ cxV z Iz
IyV ¼ cyV x Ix þ cyV y Iy þ cyV z Iz
IzV ¼ czV x Ix þ czV y Iy þ czV z Iz

Figure 2 Generation of the xW–yW–zW coordinate system from the x–y–z co-
ordinate system.

The direction cosines associated with a transformation from the x–y–z
coordinate system to the intermediate xV–yV–zV coordinate system can be
determined by inspection [see Fig. 2(a)] and are given by:
cxV x ¼ cosine ðangle between xV- and x-axesÞ ¼ cos h
cxV y ¼ cosine ðangle between xV- and y-axesÞ ¼ cosð90B hÞ ¼ sin h
cxV z ¼ cosine ðangle between xV- and z-axesÞ ¼ cosð90B Þ ¼ 0
cyV x ¼ cosine ðangle between the yV- and x-axesÞ ¼ cosð90B þ hÞ

¼ sin h
cyV y ¼ cosine ðangle between the yV- and y -axesÞ ¼ cos h
cyV z ¼ cosine ðangle between the yV- and z-axesÞ ¼ cosð90B Þ ¼ 0
czV x ¼ cosine ðangle between the zV- and x-axesÞ ¼ cosð90B Þ ¼ 0

czV y ¼ cosine ðangle between the zV- and y-axesÞ ¼ cosð90B Þ ¼ 0
czV z ¼ cosine ðangle between the zV- and z-axesÞ ¼ cosð0B Þ ¼ 1
Using these direction cosines:
IxV ¼ cxV x Ix þ cxV y Iy þ cxV z Iz ¼ ðcos hÞð1Þ þ ðsin hÞð0Þ þ ð0Þð0Þ ¼ cos h
IyV ¼ cyV x Ix þ cyV y Iy þ cyV z Iz ¼ ðsin hÞð1Þ þ ðcos hÞð0Þ þ ð0Þð0Þ ¼ sin h
IzV ¼ czV x Ix þ czV y Iy þ czV z Iz ¼ ð0Þð1Þ þ ð0Þð0Þ þ ð1Þð0Þ ¼ 0
Therefore, in the xV–yV–zV coordinate system, the vector I is written as:
I ¼ ðcos hÞiˆ V þ ðsin hÞjˆ V
Now define two additional unit vectors, one aligned with the original y-axis
(vector J ) and one aligned with the original z-axis (vector K ), i.e., let J=(1) ĵ
and K=(1) k̂. Transforming these vectors to the xV–yV–zV coordinate system,
again using the direction cosines listed above, results in:
J ¼ ðsin hÞiˆ V þ ðcos hÞjˆ V
K ¼ ð1Þ k̂ V
We now rotate vectors I , J, and K from the intermediate xV–yV–zV coordinate
system to the final xU–yU–zU coordinate system. The direction cosines associated
with a transformation from the xV–yV–zV coordinate system to the final xU–yU–zU
coordinate system are easily determined by inspection [see Fig. 2(b)] and are
given by:
cx Ux V ¼ 1 cx Uy V ¼ 0 cx Uz V ¼ 0
cy Ux V ¼ 0 cy Uy V ¼ cos b cy Uz V ¼ sin b
cz Ux V ¼ 0 cz Uy V ¼ sin b cz Uz V ¼ cos b

These direction cosines together with Eqs. (6a)–(6c) can be used to rotate the
vector I from the intermediate xV–yV–zV coordinate system to the final xU–yU–zU
coordinate system:
Ix W ¼ cxW xV IxV þ cxW yV IyV þ cxW zV IzV ¼ ð1Þðcos hÞ þ ð0Þðsin hÞ þ ð0Þð0Þ
IxW ¼ cos h
IyW ¼ cyWxV IxV þ cyWyV IyV þ cyWzV IzV ¼ð0Þðcos hÞ þ ðcos bÞðsin hÞ þ ðsin bÞð0Þ
IyW ¼ cos b sin h
IzW ¼ czWxV IxV þ czWyV IyV þ czWzV IzV ¼ð0Þðcos hÞ þ ðsin bÞðsin hÞ þ ðcosbÞð0Þ
IzW ¼ sin b sin h
Therefore, in the final xU–yU–zU coordinate system, the vector I is written as:
I ¼ ðcos hÞiˆ W þ ðcos b sin hÞjˆ W þ ðsin b sin hÞk̂ W ðaÞ
Recall that in the original x–y–z coordinate system, I is simply a unit vector
aligned with the original x-axis: I u (1)ıˆ. Therefore result (a) defines the
direction cosines associated with the angle between the original x-axis and
the final xU-, yU-, and zU-axes. That is:
cxWx ¼ cosine ðangle between xU- and x-axesÞ ¼ cos h
cyWx ¼ cosine ðangle between yU- and x-axesÞ ¼ cos b sin h
czWx ¼ cosine ðangle between zU- and x-axesÞ ¼ sin b sin h
A similar procedure is used to rotate the unit vectors J and K from the
intermediate xV–yV–zV coordinate system to the final xU–yU–zU coordinate system.
These rotations result in:
J ¼ ðsin hÞiˆ W þ ðcos b cos hÞjˆ W þ ðsin b cos hÞk̂ W ðbÞ

W W W
K ¼ ð0Þiˆ þ ðsin bÞjˆ þ ðcos bÞk̂ ðcÞ
Since vector J is a unit vector aligned with the original y-axis, J=(1)j,̂ result
(b) defines the direction cosines associated with the angle between the orig-
inal y-axis and the final xU-, yU-, and zU-axes:
cxWy ¼ cosine ðangle between xU- and y-axesÞ ¼ sin h
cyWy ¼ cosine ðangle between yU- and y-axesÞ ¼ cos b cos h
czWx ¼ cosine ðangle between zU- and y-axesÞ ¼ sin b cos h
Finally, result (c) defines the direction cosines associated with the angle
between the original z-axis and the final xU-, yU-, and zU-axes:
cxWz ¼ cosineð angle between xU- and z-axesÞ ¼ 0

cyWz ¼ cosine ðangle between yU-and z-axesÞ ¼ sin b
czWz ¼ cosine ðangle between zU-and z-axesÞ ¼ cos b
Assembling the preceding results, the set of direction cosines relating the
original x–y–z coordinate system to the final xU–yU–zU coordinate system can
be written as:
2 3 2 3
cxWx cxWy cxWz cos h sin h 0
4 cyWx cyWy cyWz 5 ¼ 4 cos b sin h cos b cos h sin b 5
czWx czWy czWz sin b sin h sin b cos h cos b
Part (b). Since the direction cosines have been determined, transformation
of force vector F can be accomplished using any version of Eqs. (6a)–(6c). For
example, using matrix notation, Eq. (6c):
8 9 2 38 9
< FxW = cxWx cxWy cxWz < Fx =
F W ¼ 4 cyWx cyWy cyWz 5 Fy
: y ; : ;
FzW czWx czWy czWz Fz
2 38 9
cos h sin h 0 < Fx =
¼ 4 cos b sin h cos b cos h sin b 5 Fy
: ;
sin b sin h sin b cos h cos b Fz
8 9 8 9
< FxW = < ðcos hÞFx þ ðsin hÞFy =
FyW ¼ ðcos b sin hÞFx þ ðcos b cos hÞFy þ ðsin bÞFz
: ; : ;
FzW ðsin b sin hÞFx þ ðsin b cos hÞFy þ ðcos bÞFz
Part (c). Using the specified numerical values and the results of Part (b):
8 9 8 9
< FxW = < ðcos 20B Þð1000 NÞ þ ðsin 20B Þ200 N =
FyW ¼ ðcos 60B sin 20B Þð1000 NÞ þ ðcos 60B cos 20B Þð200 NÞ þ ðsin 60B Þð600 NÞ
: ; : ;
FzW ðsin 60B sin 20B Þð1000 NÞ þ ðsin 60B cos 20B Þð200 NÞ þ ðcos 60B Þð600 NÞ
8 9 8 9
< FxW = < 1008 N =
F W ¼ 442:6 N
: y ; : ;
FzW 433:4 N
Using vector notation, F can now be expressed in the two different coordi-
nate systems as:
F ¼ ð1000 NÞiˆ þ ð200 NÞjˆ þ ð600 NÞk̂

or equivalently
F ¼ ð1008 NÞiˆ W þ ð442:6 NÞjˆ W þ ð433:4 NÞk̂ W

where ıˆ, j,̂ k̂ and ıˆ U, jˆ U, k̂ U are unit vectors in the x–y–z and xU–yU–zUcoordinate
systems, respectively. Force vector F drawn in the x–y–z and xU–yU–
zUcoordinate systems is shown in Fig. 3(a) and (b), respectively. The two
descriptions of F are entirely equivalent. A convenient way of (partially)
verifying this equivalence is to calculate the magnitude of the original and
transformed force vectors. Since the magnitude is an invariant, it is inde-
Figure 3 Force vector F drawn in the x–y–z and xW–yW–zW coordinate systems.

pendent of the coordinate system used to describe the force vector. Using
Eq. (2), the magnitude of the force vector in the x–y–z coordinate system is:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
jF j ¼ F x2 þ F y2 þ F z2 ¼ ð1000 NÞ2 þ ð200 NÞ2 þ ð600 NÞ2 ¼ 1183 N
The magnitude of the force vector in the xU–yU–zU coordinate system is:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 2
jF j ¼ ðFxWÞ þ FyW þðFzWÞ ¼ ð1008 NÞ2 þ ð442:6 NÞ2 þ ð433:4 NÞ2
2
¼ 1183N ðagreesÞ
3 NORMAL FORCES, SHEAR FORCES, AND FREE-BODY

DIAGRAMS
A force F acting at an angle to a planar surface is shown in Fig. 4. Since force
is a vector, it can always be decomposed into two force components, a normal
force component and a shear force component. The line-of-action of the nor-
mal force component is orthogonal to the surface, whereas the line-of-action
of the shear force component is tangent to the surface.
Internal forces induced within a solid body by externally applied forces
can be investigated with the aid of free-body diagrams. A simple example is
shown in Fig. 5, which shows a straight circular rod with constant diameter
subjected to two external forces of equal magnitude (R) but opposite direc-
tion. The internal force (F I, say) induced at any cross section of the rod can be
investigated by making an imaginary cut along the plane of interest. Suppose
an imaginary cut is made along plane a-a, which is perpendicular to the axis
Figure 4 A force F acting at an angle to a planar surface.

Figure 5 The use of free-body diagrams to determine internal forces acting on
planes a-a and b-b: (a) free-body diagram based on plane a–a–a–a perpendicular
to rod axis; (b) free-body diagram based on plane b–b–b–b, inclined at angle h
to rod axis.

of the rod. The resulting free-body diagram for the lower half of the rod is
shown in Fig. 5(a), where an x–y–z coordinate system has been assigned such
that the x-axis is parallel to the rod axis, as shown. On the basis of this free-
body diagram, it is concluded that an internal force F I = (R)ıˆ+(0)jˆ+(0)k̂
is induced at cross-section a-a. That is, only a normal force of magnitude R is
induced at cross-section a-a, which has been defined to be perpendicular to
the axis of the rod.
On the other hand, the imaginary cut need not be made perpendicular
to the axis of the rod. Suppose the imaginary cut is made along plane b-b,
which is oriented at an angle of h with respect to the axis of the rod. The
resulting free-body diagram for the lower half of the rod is shown in Fig. 5(b).
A new xV–yV–zV coordinate system has been assigned so that the xV-axis is
perpendicular to plane b-b and the zV-axis is coincident with the z-axis—that
is, the xV–yV–zV coordinate system is generated from the x–y–z coordinate
system by a rotation of h about the original z-axis. The internal force FI can
be expressed with respect to the xV–yV–zV coordinate system by transforming
FI from the x–y–z coordinate system to the xV–yV–zV coordinate system.
This coordinate transformation is a special case of the transformation
considered in Example Problem 1. The direction cosines now become (with
b=0j):
cxV x ¼ cos h cxV y ¼ sin h c xV z ¼ 0
cyV x ¼ sin h cyV y ¼ cos h cy V z ¼ 0
czV x ¼ 0 czV y ¼ 0 czV z ¼ 1
Applying Eqs. (6a)–(6c), we have:
8 9 2 38 9 2 38 9
< FxV = c xV x c xV y c xV z < F x = cos h sin h 0 < R =
F V ¼ 4 cyV x cyV y cyV z 5 Fy ¼ 4 sin h cos h 0 5 0
: y ; : ; : ;
FzV czV x czV y czV z Fz 0 0 1 0
8 9
< ðcos hÞR =
¼ ðsin hÞR
: ;
0
In the xV–yV–zV coordinate system, the internal force is F I = (R cos h)ıˆV(R sin
h)ĵV+(0)k̂V. Hence, by defining a coordinate system which is inclined to the
axis of the rod, we conclude that both a normal force (R cos h) and a shear
force (R sin h) are induced in the rod.
Although the preceding discussion may seem simplistic, it has been
included in order to demonstrate the following:
A specific coordinate system must be specified before a force vector
can be defined in a mathematical sense. In general, the coordinate

system is defined by the imaginary cut(s) used to form the free-body
diagram.
All components of a force must be specified to fully define the force
vector. Furthermore, the individual components of a force change
as the vector is transformed from one coordinate system to another.
These two observations are valid for all tensors, not just for force
vectors. In particular, these observations hold in the case of stress and strain
tensors, which will be reviewed in the following sections.
4 DEFINITION OF STRESS
There are two fundamental types of stress: normal stress and shear stress.
Both types of stress are defined as a force divided by the area over which
it acts.
A general 3-D solid body subjected to a system of external forces is
shown in Fig. 6(a). It is assumed that the body is in static equilibrium, that
is, it is assumed that the sum of all external forces is zero, SF i = 0. These
external forces induce internal forces acting within the body. In general,
the internal forces will vary in both magnitude and direction throughout the
body. An illustration of the variation of internal forces along a line within
an internal plane is shown in Fig. 6(b). A small area (DA) isolated from this
plane is shown in Fig. 6(c). Area DA is assumed to be ‘‘infinitesimally small.’’
That is, the area DA is small enough such that the internal forces acting over
DA can be assumed to be of constant magnitude and direction. Therefore,
the internal forces acting over DA can be represented by a force vector which
can be decomposed into a normal force, N, and a shear force, V, as shown
in Fig. 6(c).
Normal stress (usually denoted r) and shear stress (usually denoted s)
are defined as the force per unit area acting perpendicular and tangent to the
area DA, respectively. That is,
N V
r u lim s u lim ð9Þ
DA!0 DA DA!0 DA
Note that by definition, the area DA shrinks to zero: DA!0. Stresses r and s
are therefore said to exist ‘‘at a point.’’ Also, since internal forces generally
vary from point-to-point (as shown in Fig. 6), stresses also vary from point-
to-point.
Stress has units of force per unit area. In SI units, stress is reported in
terms of Pascals (abbreviated Pa), where 1 Pa=1 N/m2. In English units,
stress is reported in terms of pounds-force per square inch (abbreviated psi),

Figure 6 A solid 3-D body in equilibrium.

that is, 1 psi=1 lbf/in2. Conversion factors between the two systems of mea-
surement are 1 psi=6895 Pa, or equivalently, 1 Pa=0.1450103 psi. Com-
mon abbreviations used throughout this text are as follows:
1 103 Pa ¼ 1 kilo-Pascals ¼ 1 kPa 1 103 psi ¼ 1 kilo-psi ¼ 1 ksi
1 106 Pa ¼ 1 Mega-Pascals ¼ 1 MPa 1 106 psi ¼ 1 mega-psi ¼ 1 Msi
1 109 Pa ¼ 1 Giga-Pascals ¼ 1 GPa
5 THE STRESS TENSOR

A general 3-D solid body subjected to a system of external forces is shown in
Fig. 7(a). It is assumed that the body is in static equilibrium and that body
forces are negligible, that is, it is assumed that the sum of all external forces
is zero, SF i = 0. A free-body diagram of an infinitesimally small cube re-
moved from the body is shown in Fig. 7(b). The cube is referenced to an x–
y–z coordinate system, and the cube edges are aligned with these axes. The
lengths of the cube edges are denoted dx, dy, and dz. Although (in general)
internal forces are induced over all six faces of the cube, for clarity, the forces
acting on only three faces have been shown.
The force acting over each cube face can be decomposed into a normal
force component and two shear force components, as illustrated in Fig. 7(c).
Although each component could be identified with a single subscript (since
force is a first-order tensor), for convenience, two subscripts have been used.
The first subscript identifies the face over which the force is distributed,
while the second subscript identifies the direction in which the force is ori-
ented. For example, Nxx refers to a normal force component which is
distributed over the x-face and which ‘‘points’’ in the x-direction. Similarly,
Vzy refers to a shear force distributed over the z-face which ‘‘points’’ in the
y-direction.
Three stress components can now be defined for each cube face, in
accordance with Eq. (9). For example, for the three faces of the infinitesimal
element shown in Fig. 7:
Stresses acting on the +x-face:

Nxx Vxy Vxz
rxx ¼ lim sxy ¼ lim sxz ¼ lim
dy;dz!0 dydz dy;dz!0 dydz dy;dz!0 dydz
Stresses acting on the y-face:

Nyy Vyx Vyz
ryy ¼ lim syx ¼ lim syz ¼ lim
dx;dz!0 dxdz dx;dz!0 dxdz dx;dz!0 dxdz

Figure 7 Free-body diagrams used to define stress induced in a solid body.

Stresses acting on the +z-face:

Nzz Vzx Vzy
rzz ¼ lim szx ¼ lim szy ¼ lim
dx;dy!0 dxdy dx;dy!0 dxdy dx;dy!0 dxdy
Since three force components (and therefore three stress components)

exist on each of the six faces of the cube, it would initially appear that there
are 18 independent force (stress) components. However, it is easily shown
that for static equilibrium to be maintained (assuming body forces are neg-
ligible):
Normal forces acting on opposite faces of the infinitesimal element
must be of equal magnitude and opposite direction, and
Shear forces acting within a plane of the element must be orientated
either ‘‘tip-to-tip’’ [e.g., forces Vxz and Vzx in Fig. 7(c)] or ‘‘tail-to-
tail’’ (e.g., forces Vzy and Vyz) and be of equal magnitude. That is,
jVxy=Vyxj, jVxz=Vzxj, and jVyz=Vzyj.
These restrictions reduce the number of independent force (stress) com-
ponents from 18 to 6, as follows:
Independent force components Independent stress components

Nxx rxx
Nyy ryy
Nzz rzz
Vxy ð¼ Vyx Þ sxy ð¼ syx Þ
Vxz ð¼ Vzx Þ sxz ð¼ szx Þ
Vyz ð¼ Vzy Þ syz ð¼ szy Þ
An infinitesimal element showing all stress components is shown in

Fig. 8. We must next define the algebraic sign convention we will use to de-
scribe individual stress components. We first associate an algebraic sign with
each face of the infinitesimal element. A cube face is positive if the outward
unit normal of the face (that is, the unit normal pointing away from the
interior of the element) points in a positive coordinate direction; otherwise,
the face is negative. For example, face (ABCD) in Fig. 8 is a positive face,
while face (CDEF) is a negative face.

Figure 8 An infinitesimal stress element (all stress components shown in a
positive sense).
Having identified the positive and negative faces of the element, a stress
component is positive if:
The stress component acts on a positive face and points in a positive
coordinate direction, or if
The stress component acts on a negative face and points in a negative
coordinate direction
otherwise, the stress component is negative.
This convention can be used to confirm that all stress components
shown in Fig. 8 are algebraically positive. For example, to determine the
algebraic sign of the normal stress rxx which acts on face ABCD in Fig. 8, note
that (a) face ABCD is positive and (b) the normal stress rxx which acts on this

face points in the positive y-direction. Therefore, rxx is positive. As a second
example, the shear stress syz which acts on cube face CDEF is positive because
(a) face CDEF is negative and (b) syz points in the negative z-direction.
The preceding discussion shows that the state of stress at a point is de-
fined by six components of stress: three normal stress components and three
shear stress components. The state of stress is written using matrix notation
as follows:
2 3 2 3
rxx sxy sxz rxx sxy sxz
4 syx ryy syz 5 ¼ 4 sxy ryy syz 5 ð10Þ
szx szy rzz sxz syz rzz
To express the state of stress using indicial notation, we must first make
the following change in notation:
sxy ! rxy
sxz ! rxz
syx ! ryx
syz ! ryz
szx ! rzx
szy ! rzy
With this change, the matrix on the left side of the equality sign in Eq. (10)
becomes:
2 3 2 3
rxx sxy sxz rxx rxy rxz
4 syx ryy syz 5 ! 4 rxy ryy ryz 5
szx szy rzz rxz ryz rzz
which can be succinctly written using indicial notation as:
rij ; i; j ¼ x; y; or z ð11Þ
In Sec. 1, it was noted that a force vector is a first-order tensor since only
one subscript is required to describe a force tensor, Fi. From Eq. (11), it is clear
that stress is a second-order tensor (or equivalently, a tensor of rank two) since
two subscripts are required to describe a state of stress.
Example Problem 2
Given. The stress element referenced to an x–y–z coordinate system and
subject to the stress components shown in Fig. 9.

Figure 9 Stress components acting on an infinitesimal element (all stresses in
MPa).
Determine. Label all stress components, including algebraic sign.

Solution. The magnitude and algebraic sign of each stress component are
determined using the sign convention defined above. The procedure will be
illustrated using the stress components acting on face CDEF. First, note that
face CDEF is a negative face since an outward unit normal for this face points
in the negative y-direction. The normal stress which acts on face CDEF has a
magnitude of 50 MPa and points in the positive y-direction. Hence, this stress
component is negative and is labeled ryy=50 MPa. One of the shear stress
components acting on face CDEF has a magnitude of 75 MPa and points in
the positive x-direction. Hence, this stress component is also negative and is

labeled syx=75 MPa (or equivalently, sxy=75 MPa). Finally, the second
shear force component acting on face CDEF has a magnitude of 50 MPa and
points in the positive z-direction. Hence, this component is labeled syz=50
MPa (or equivalently, szy=50 MPa).
Following this process for all faces of the element, the state of stress
represented by the element shown in Fig. 9 can be written as:
2 3 2 3
rxx sxy sxz 100 MPa 75 MPa 30 MPa
4 syx ryy syz 5 ¼ 4 75 MPa 50 MPa 50 MPa 5
szx szy rzz 30 MPa 50 MPa 25 MPa
6 TRANSFORMATION OF THE STRESS TENSOR

In Sec. 5, the stress tensor was defined using a free-body diagram of an infini-
tesimal element removed from a 3-D body in static equilibrium. This concept
is again illustrated in Fig. 10(a), which shows the stress element referenced to
an x–y–z coordinate system.
Now, the infinitesimal element need not be removed in the orientation
shown in Fig. 10(a). An infinitesimal element removed from precisely the
same point within the body but at a different orientation is shown in Fig.
10(b). This stress element is referenced to a new xV–yV–zV coordinate system.
The state of stress at the point of interest is dictated by the external loads
applied to the body and is independent of the coordinate system used to
describe it. Hence, the stress tensor referenced to the xV–yV–zVcoordinate
system is equivalent to the stress tensor referenced to the x–y–z coordinate
system, although the direction and magnitude of individual stress compo-
nents will differ.
The process of relating stress components in one coordinate system to
those in another is called the transformation of the stress tensor. This
terminology is perhaps unfortunate in the sense that the state of stress itself
is not ‘‘transformed,’’ but rather our description of the state of stress trans-
forms as we change from one coordinate system to another.
It can be shown (1,2) that the stress components in the new xV–yV–zV
coordinate system (riVjV) are related to the components in the original x–y–z
coordinate system (rij) according to:
ri V j V ¼ ci V k cj V l rkl where i; j; k; l ¼ x; y; z ð12aÞ
or equivalently (using matrix notation):
½ri V j V ¼ ½ci V j ½rij ½ci V j T

Figure 10 Infinitesimal elements removed from the same point from a 3-D solid
but in two different orientations.

where [ciVj]T is the transpose of the direction cosine array. Writing in full
matrix form:
2 3 2 32 3
rx V x V r x V y V rxV zV c xV x c xV y c xV z rxx rxy rxz
4 ryV xV ryV yV ryV zV 5 ¼ 4 cyV x cyV y cyV z 54 ryx ryy ryz 5
rzV xV rzV yV rz V z V czV x czV y czV z rzx rzy rzz
2 3
c xV x c y V x c z V x
4 cxV y cyV y czV y 5 ð12bÞ
c xV z c y V z c z V z
As discussed in Sec. 2, the terms ciVj which appear in Eqs. (12a) and (12b)
are direction cosines and equal the cosine of the angle between the axes of the
x–y–z and xV–yV–zV coordinate systems. Recall that the algebraic sign of an
angle of rotation is defined in accordance with the right-hand rule and that
angles are defined from the x–y–z coordinate system to xV–yV–zV coordinate
system. Equations (12a) and (12b) are called the transformation law for a
second-order tensor.
If an analysis is being performed with the aid of a digital computer,
which nowadays is almost always the case, then matrix notation [Eq. (12b)]
most likely will be used to transform a stress tensor from one coordinate
system to another. Conversely, if a stress transformation is to be accom-
plished using hand calculations, then indicial notation [Eq. (12a)] may be the
preferred choice. To apply Eq. (12a), desired values are first specified for
subscripts iV and jV, and then the terms on the right side of the equality are
summed over the entire range of the remaining two subscripts, k and l. For
example, suppose we wish to write the relationship between rxVzV and the stress
components in the x–y–z coordinate system in expanded form. We first
specify that iV=xV and jV=zV, and Eq. (12a) becomes:
rxV zV ¼ cxV k czV l rkl where k; l ¼ x; y; z

We then sum all terms on the right side of the equality by cycling through the
entire range of k and l. In expanded form, we have:
rxV zV ¼ cxV x czV x rxx þ cxV x czV y rxy þ cxV x czV z rxz þ cxV y czV x ryx
þ cxV y czV y ryy þ cxV y czV z ryz þ cxV z czV x rzx þ cxV z czV y rzy
þ cxV z czV z rzz ð13Þ

Equations (12a) and (12b) show that the value of any individual stress
component riVjV varies as the stress tensor is transformed from one coordinate
system to another. However, it can be shown (1,2) that there are features of
the total stress tensor that do not vary when the tensor is transformed from
one coordinate system to another. These features are called the stress in-

variants. For a second-order tensor, three independent stress invariants exist
and are defined as follows:
First stress invariant ¼ H ¼ rii ð14aÞ

1
Second stress invariant ¼ U ¼ rii rjj rij rij ð14bÞ
2
1
Third stress invariant ¼ W ¼ rii rjj rkk 3rii rjk rjk þ 2rij rjk rki
6
ð14cÞ
Alternatively, by expanding these equations over the range i, j, k=x, y, z and

simplifying, the stress invariants can be written as:
First stress invariant ¼ H ¼ rxx þ ryy þ rzz ð15aÞ
Second stress invariant ¼ U ¼ rxx ryy þ rxx rzz þ ryy rzz

r2xy þ r2xz þ r2yz ð15bÞ
Third stress invariant ¼ W ¼ rxx ryy rzz rxx r2yz ryy r2xz
rzz r2xy þ 2rxy rxz ryz ð15cÞ
The three stress invariants are conceptually similar to the magnitude of a

force tensor. That is, the value of the three stress invariants is independent
of the coordinate used to describe the stress tensor, just as the magnitude of
a force vector is independent of the coordinate system used to describe the
force. This invariance will be illustrated in the following example problem.
Example Problem 3
Given. A state of stress referenced to an x–y–z coordinate is known to be:
2 3 2 3
rxx rxy rxz 50 10 15
4 ryx ryy ryz 5 ¼ 4 10 25 30 5ðksiÞ
rzx rzy rzz 15 30 5
It is desired to express this state of stress in an xU–yU–zU coordinate system,

generated by the following two sequential rotations:
(i) Rotation of h=20j about the original z-axis (which defines an
intermediate xV–yV–zV coordinate system), followed by
(ii) Rotation of b=35j about the xV-axis (which defines the final xU–yU–
zU coordinate system).
Problem. (a) Rotate the stress tensor to the xU–yU–zU coordinate system and
(b) calculate the first, second, and third invariants of the stress tensor using

both elements of the stress tensor referenced to the x–y–z coordinate system,
rij, and elements of the stress tensor referenced to the xU–yU–zU coordinate
system, riUjU.
Solution
Part (a). General expressions for direction cosines relating the x–y–z and
xU–yU–zU coordinate systems were determined as a part of Example Problem 1.
The direction cosines were found to be:
cxWx ¼ cos h
cxWy ¼ sin h
cxWz ¼ 0
cyWx ¼ cos b sin h
cyWy ¼ cos b cos h
cyWz ¼ sin b
czWx ¼ sin b sin h
czWy ¼ sin b cos h
czWz ¼ cos b
Since in this problem h=20j and b=35j, the numerical values of the direc-
tion cosines are:
cxWx ¼ cos ð20B Þ ¼ 0:9397
cxWy ¼ sin ð20B Þ ¼ 0:3420
cxWz ¼ 0
cyWx ¼ cos ð35B Þsin ð20B Þ ¼ 0:2802
cyWy ¼ cos ð35B Þcos ð20B Þ ¼ 0:7698
cyWz ¼ sin ð35B Þ ¼ 0:5736
czWx ¼ sin ð35B Þsin ð20B Þ ¼ 0:1962
czWy ¼ sin ð35B Þcos ð20B Þ ¼ 0:5390
czWz ¼ cos ð35B Þ ¼ 0:8192
Each component of the transformed stress tensor is now found through the
application of either Eq. (12a) or Eq. (12b). For example, if indicial notation
is used, stress component rxUzU can be found using Eq. (13):
rxWzW ¼ cxWx czWx rxx þ cxWx czWy rxy þ cxWx czWz rxz
þ cxWy czWx ryx þ cxWy czWy ryy þ cxWy czWz ryz
þ cxWz czWx rzx þ cxWz czWy rzy þ cxWz czWz rzz

rxW zW ¼ ð0:9397Þð0:1962Þð50 ksiÞ þ ð0:9397Þð0:5390Þð10 ksiÞ
þð0:9397Þð0:8192Þð15 ksiÞ þ ð0:3420Þð0:1962Þð10 ksiÞ
þð0:3420Þð0:5390Þð25 ksiÞ þ ð0:3420Þð0:8192Þð30 ksiÞ
þð0Þð0:1962Þð15ksiÞþð0Þð0:5390Þð30ksiÞþð0Þð0:8192Þð5 ksiÞ
rxWzW ¼ 28:95 ksi
Alternatively, if matrix notation is used, then Eq. (12b) becomes:
2 3 2 32 3
rxWxW rxWyW rxWzW 0:9397 0:3420 0 50 10 15
4ryWxW ryWyW ryWzW5 ¼40:2802 0:7698 0:57365410 25 30 5
rzWxW rzWyW rzWzW 0:1962 0:5390 0:8192 15 30 5
2 3
0:9397 0:2802 0:1962
4 0:3420 0:7698 0:53905
0 0:5736 0:8192
Completing the matrix multiplication indicated yields the following:
2 3 2 3
rxWxW rxWyW rxWzW 40:65 1:113 28:95
4 ryWxW ryWyW ryWzW 5 ¼ 4 1:113 43:08 10:60 5 ðksiÞ
rzWxW rzWyW rzWzW 28:95 10:60 13:72
Notice that the value of rxUzU determined through matrix multiplication is
identical to that obtained using indicial notation, as previously described. The
stress element is shown in the original and final coordinate systems in Fig. 11.
Part (b). The first, second, and third stress invariants will now be calculated
using components of both rij and riUjU. It is expected that identical values will
be obtained since the stress invariants are independent of the coordinate
system.
First stress invariant:
x–y–z coordinate system:
H ¼ rii ¼ rxx þ ryy þ rzz
H ¼ ð50 þ 25 5Þ ksi
H ¼ 70 ksi
xU–yU–zU coordinate system:
H ¼ ri Wi W ¼ rxWxW þ ryWyW þ rzWzW
H ¼ ð40:65 þ 43:08 13:72Þ
H ¼ 70 ksi
As expected, the first stress invariant is independent of the coordinate
system.

Figure 11 Stress tensor of Example Problem 3 referenced to two different
coordinates (magnitude of all stress components in ksi).
Second stress invariant:

1
U¼ rii rjj rij rij ¼ rxx ryy þ rxx rzz þ ryy rzz
2

r2xy þ r2xz þ r2yz
n
U ¼ ð50Þð25Þ þ ð50Þð5Þ þ ð25Þð5Þ ½ð10Þ2 þ ð15Þ2
o
þ ð30Þ2 ðksiÞ2
U ¼ 350ðksiÞ2
1
U¼ r W Wr W W ri Wj Wri Wj W
2 ii jj

U ¼ rxWxWryWyW þ rxWxWrzWzW þ ryWyWrzWzW r2xWyW þ r2xWzW þ r2yWzW

n
U ¼ ð40:65Þð43:08Þ þ ð40:65Þð13:72Þ þ ð43:08Þð13:72Þ
o
½ð1:113Þ2 þ ð28:95Þ2 þ ð10:60Þ2 ðksiÞ2
U ¼ 350ðksiÞ2
As expected, the second stress invariant is independent of the coordinate
system.
Third stress invariant:

1
W ¼ rii rjj rkk 3rii rjk rjk þ 2rij rjk rki
6
W ¼ rxx ryy rzz rxx r2yz ryy r2xz rzz r2xy þ 2rxy rxz ryz
W ¼ ½ð50Þð25Þð5Þ ð50Þð30Þ2 ð25Þð15Þ2 ð5Þð10Þ2
þ 2ð10Þð15Þð30Þ ðksiÞ3
W ¼ 65375ðksiÞ3
1
W ¼ ri Wi Wrj Wj WrkWkW 3ri Wi Wrj WkWrj WkW þ 2ri Wj Wrj WkWrkWi W
6
W ¼ rxWxWryWyWrzWzW rxWxWr2yWzW ryWyWr2xWzW rzWzWr2xWyW þ 2rxWyWrxWzWryWzW
W ¼ ½ð40:65Þð43:08Þð13:72Þ ð40:65Þð10:60Þ2 ð43:08Þ
ð28:95Þ2 ð13:72Þð1:113Þ2 þ 2ð1:113Þð28:95Þ
ð10:60Þ ðksiÞ3
W ¼ 65375ðksiÞ3
As expected, the third stress invariant is independent of the coordinate
system.
7 PRINCIPAL STRESSES
The definition of a stress tensor was reviewed in Sec. 5, and transformation of
a stress tensor from one coordinate system to another was discussed in Sec. 6.
It can be shown (1,2) that it is always possible to rotate the stress tensor to a
special coordinate system in which no shear stresses exist. This coordinate
system is called the principal stress coordinate system, and the normal stresses

that exist in this coordinate system are called the principal stresses. In most
textbooks, principal stresses are denoted r1, r2, and r3. However, as will be
discussed later (see Fig. 2 in Chap. 3) in this textbook, the labels ‘‘1,’’ ‘‘2,’’ and
‘‘3’’ will be used to label the axes in a special coordinate system called the
principal material coordinate system. Therefore, in this text, the axes asso-
ciated with the principal stress coordinate system will be labeled p1, p2, and p3
axes, and the principal stresses will be denoted rp1, rp2, and rp3.
Knowledge of the principal stresses induced in an isotropic structure is
of extreme importance, primarily because failure of isotropic materials (e.g.,
yielding and/or fracture) can often be directly related to the magnitude(s) of
the principal stresses. This is not the case for composite structures, however.
As will be seen in later chapters, failure of composite structures is not
governed by principal stresses, and hence the topic of principal stresses is
only of occasional importance to the composite engineer.
Principal stresses may be related to stress components in an x–y–z
coordinate system using the free-body diagram shown in Fig. 12. It is assumed
that plane ABC is one of the three principal planes (i.e., n = 1, 2, or 3) and
therefore no shear stress exists on this plane. The line-of-action of principal
stress rpn defines one axis of the principal stress coordinate system. The
direction cosines between this principal axis and the x-, y-, and z- axes are cpnx,
cpny, and cpnz, respectively. The surface area of triangle ABC is denoted AABC.
The normal force acting over triangle ABC therefore equals (rpnAABC). The
components of the normal force acting in the x-, y-, and z-directions equal
cpnxrpnAABC, cpnyrpnAABC, and cpnzrpnAABC, respectively.
The area of the other triangular faces are given by:
Area of triangle ABD=cpnxAABC.
Area of triangle ACD=cpnyAABC.
Area of triangle BCD=cpnzAABC.
Summing forces in the x-direction and equating to zero, we obtain:
cpn x rpn AABC rxx cpn x AABC sxy cpn y AABC sxz cpn z AABC ¼ 0
which can be reduced and simplified to:
ðrpn rxx Þcpn x sxy cpn y sxz cpn z ¼ 0 ð16aÞ
Similarly, summing forces in the y- and z-directions results in:
sxy cpn x þ ðrpn ryy Þcpn y syz cpn z ¼ 0 ð16bÞ
sxz cpn x syz cpn y þ ðrpn rzz Þcpn z ¼ 0 ð16cÞ
Equations (16a)–(16c) represent three linear homogeneous equations

which must be satisfied simultaneously. Since direction cosines cpnx, cpny, and
cpnz must also satisfy Eq. (8), and therefore cannot all equal zero, the solu-

Figure 12 Free-body diagram used to relate stress components in the x–y–z
coordinate system to a principal stress.
tion can be obtained by requiring that the determinant of the coefficients

of cpnx, cpny, and cpnz equal zero:

ðrpn rxx Þ sxy sxz

sxy ðrpn ryy Þ syz ¼ 0

sxz syz ðrpn rzz Þ
Equating the determinant to zero results in the following cubic
equation:
r3pn Hr2pn þ Urpn W ¼ 0 ð17Þ
where H, U, and W are the first, second, and third stress invariants, respec-
tively, and have been previously listed as Eqs. (14a)–(14c) and (15a)–(15c).

The three roots of the cubic equation (that is, the three principal stresses) may
be found by application of the standard approach (3), as follows.
Define:
1
a¼ 3U H2
3
1 3
b¼ 2H 9HU þ 27W
27
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
3 b b2 a3
A¼ þ þ
2 4 27
3 b b2 a3
B¼ þ
2 4 27
The three principal stresses [i.e., the three roots of Eq. (17)] are then given by:
8
> H
>
> AþBþ
>
> 3
>
< A þ B pffiffiffiffiffiffiffi A B H
>
rp1 ; rp2 ; rp3 ¼ þ 3 þ ð18Þ
>
> 2 2 3
>
>
>
> A þ B pffiffiffiffiffiffiffi A B H
>
: 3 þ
2 2 3
By convention, the principal stresses are numbered such that rp1 is the
algebraically greatest principal stress, whereas rp3 is the algebraically least.
That is, rp1 > rp2>rp3.
Two comments regarding the practical application of either Eq. (17) or
Eq. (18) are appropriate. First, many handheld calculators and computer
software packages feature standard routines to find the roots of nth-order
polynomials. Hence, the three roots of Eq. (17) (that is, the three principal
stresses) may often be found most conveniently through the use of these
standard calculator routines or software packages, rather than through
application of Eq. (18). The second comment is that if the principal stresses
are to be calculated through application of Eq. (18), then calculation of
constants A and B generally involves finding the cube root of a complex
number. The need to find the cube root of a complex number is encountered
infrequently, and hence the reader may not be aware of how to make such a
calculation. For convenience, a process that may be used to find the cube root
of a complex number has been included in Appendix A.
In any event, once the principal stresses are determined, the three sets of
direction cosines (which define the principal coordinate directions) are found
by substituting the three principal stresses given by Eq. (18) into Eqs. (16a)–
(16c) in turn. Since only two of Eqs. (16a)–(16c) are independent, Eq. (8) is

used as a third independent equation involving the three unknown constants
cpnx, cpny, and cpnz.
The process of finding principal stresses and direction cosines will be
demonstrated in the following example problem.
Example Problem 4
Given. A state of stress referenced to an x–y–z coordinate is known to be:
2 3 2 3
4 ryx ryy ryz 5 ¼ 4 10 25 30 5 ðksiÞ
rzx rzy rzz 15 30 5
Problem. Find (a) the principal stresses and (b) the direction cosines that
define the principal stress coordinate system.
Solution. This is the same stress tensor considered in Example Problem 3. As
a part of that problem, the first, second, and third stress invariants were found
to be:
H=70 ksi
U=350 (ksi)2
W=65375 (ksi)3
Part (a). Determining the principal stresses. In accordance with Eq. (17),
the three principal stresses are the roots of the following cubic equation:
r3 70r2 350r þ 65375 ¼ 0

As discussed earlier, the three roots of this equation can often be found most
conveniently using appropriate handheld calculators or software packages.
In this example solution, the roots will be found through application of Eq.
18. Following this process, we have:
1 1h i
a¼ 3U H2 ¼ 3ð350Þ ð70Þ2 ¼ 1983
3 3
1 3
b¼ 2H 9HU þ 27W
27
1 h i
¼ 2ð70Þ3 9ð70Þð350Þ þ 27ð65375Þ ¼ 31801
27
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u u
3 b b 2 a 3 t 48134
3 ð48134Þ2 ð1983Þ3
A¼ þ þ ¼ þ þ
2 4 27 2 4 27
p ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
¼ 3 15900 þ ið6010Þ

We find that constant A equals the cube root of the complex number:
z ¼ 15900 þ ið6010Þ
Following the process described in Appendix A, the modulus and argument
of this complex number are:
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
r ¼ a2 þ b2 ¼ ð15900Þ2 þ ð6010Þ2 ¼ 16998

1 6010
u ¼ tan ¼ tan1 ð0:37799Þ ¼ 0:3614 radðorÞ2:780 rad
15900
Since in this case:

a ¼ 15900 < 0
b ¼ 6010 > 0
it is clear that the argument u corresponds to an angle in the second quadrant
of the complex plane (refer to Fig. A.1). Hence, we select u=2.780 rad.
Applying Eq. A.3, we find:

pffiffiffiffiffiffiffiffiffiffiffiffi lnðrÞ n hui h u io
A ¼ 3 a þ ib ¼ exp cos þ i sin
3 3 3

p ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi lnð16998Þ
A ¼ 3 15900 þ ið6010Þ ¼ exp
3

2:780 2:780
cos þ i sin
3 3
A ¼ 25:71f0:6005 þ ið0:7996Þg
A ¼ 15:44 þ ið20:56Þ
Following an identical procedure, constant B is found to be:
B ¼ 15:44 ið20:56Þ
We now apply Eq. (18) to find:
8 8
>
> AþBþH >
>
54:21 ksi
>
> 3 >
>
>
> >
>
< A þ B pffiffiffiffiffiffiffi A B H <
rp1 ; rp2 ; rp3 ¼ þ 3 þ ¼ 27:72 ksi
> 2 2 3 >
>
> >
>
>
> A þ B pffiffiffiffiffiffiffi A B H >
>
>
: 3 þ >
:
2 2 3 43:51 ksi
Hence, rp1=54.21 ksi, rp2=43.51 ksi, and rp3=27.72 ksi.

Part (b). Determining the direction cosines. The first two of Eqs. (16a)–
(16c) and Eq. (8) will be used to form three independent equations in three
unknowns. We have:
ðrpn rxx Þcpn x sxy cpn y sxz cpn z ¼ 0
sxy cpn x þ ðrpn ryy Þcpn y syz cpn z ¼ 0
ðcpn x Þ2 þ ðcpn y Þ2 þ ðcpn z Þ2 ¼ 1
Direction cosines for rp1. The three independent equations become:
ð54:21 50Þcp1 x þ 10cp1 y 15cp1 z ¼ 0
10cp1 x þ ð54:21 25Þcp1 y 30cp1 z ¼ 0
ðcp1 x Þ2 þ ðcp1 y Þ2 þ ðcp1 z Þ2 ¼ 1
Solving simultaneously, we obtain:
cp1 x ¼ 0:9726 cp1 y ¼ 0:1666 cp1 z ¼ 0:1620
ð43:51 50Þcp2 x þ 10cp2 y 15cp2 z ¼ 0
10cp2 x þ ð43:51 25Þcp2 y 30cp2 z ¼ 0
cp2 x ¼ 0:05466 cp2 y ¼ 0:8416 cp2 z ¼ 0:5738
ð27:72 50Þcp3 x þ 10cp3 y 15cp3 z ¼ 0
10cp3 x þ ð27:72 25Þcp3 y 30cp3 z ¼ 0
cp3 x ¼ 0:8276 cp3 y ¼ 0:2259 cp3 z ¼ 0:5138
8 PLANE STRESS
A stress tensor is defined by six components of stress: three normal stress
components and three shear stress components. Now, in practice, a state of
stress often encountered is one in which all stress components in one co-
ordinate direction are zero. For example, suppose rzz=sxz=syz=0, as

shown in Fig. 13(a). Since the three remaining nonzero stress components
(rxx, ryy, and sxy) all lie within the x–y plane, such a condition is called a
state of plane stress. Plane stress conditions occur most often because of the
geometry of the structure of interest. Specifically, the plane stress condition
usually exists in thin, platelike structures. Examples include the web of an I-
beam, the body panel of an automobile, or the skin of an airplane fuselage.
Figure 13 Stress elements subjected to a state of plane stress.

In these instances, the stresses induced normal to the plane of the structure
are very small compared to those induced within the plane of the structure.
Hence, the small out-of-plane stresses are usually ignored, and attention is
focused on the relatively high stress components acting within the plane of
the structure.
Since laminated composites are often used in the form of thin plates or
shells, the plane stress assumption is widely applicable in composite structures
and will be used throughout most of the analyses discussed in this textbook.
Since the out-of-plane stresses are negligibly small, for convenience, an
infinitesimal stress element subjected to plane stress will usually be drawn
as a square rather than a cube, as shown in Fig. 13(b).
Results discussed in earlier sections for general 3-D state of stress will
now be specialized for the plane stress condition. It will be assumed that the
nonzero stresses lie in the x–y plane (i.e., rzz=sxz=syz=0). This allows the
remaining components of stress to be written in the form of a column array
rather than a 3 3 array:
2 3 8 9
rxx sxy 0 > r
< xx >=
6 7
4 sxy ryy 0 5 ! ryy
>
: >
;
0 0 0 sxy
Note that when a plane stress state is described, stress appears to be a first-
order tensor since (apparently) only three components of stress (rxx, ryy,
andsxy) need be specified in order to describe the state of stress. This is, of
course, not the case. Stress is a second-order tensor in all instances, and six
components of stress must always be specified in order to define a state of
stress. When we invoke the plane stress assumption, we have simply assumed
a priori that three stress components (rzz, sxz, and syz) are zero.
Recall that either Eq. (12a) or Eq. (12b) governs the transformation of a
stress tensor from one coordinate system to another. Equation (12b) is
repeated here for convenience:
2 3 2 32 3
rxV xV rxV yV rxV zV c xV x c xV y c xV z rxx rxy rxz
6r V V r V V r V V 7 6c V cyV y cyV z 7 6 7
4 yx yy yz 5 ¼ 4 yx 54 ryx ryy ryz 5
rzV xV rzV yV rzV zV czV x czV y czV z rzx rzy rzz
2 3
cxV x cyV x czV x
6c V c V c V 7
4 xy yy z y 5ðrepeatedÞ ð12bÞ
cxV z cyV z czV z
When transformation of a plane stress tensor is considered, it will be
assumed that the xV–yV–zV coordinate system is generated from the x–y–z

system by a rotation h about the z-axis. That is, the z- and zV-axes are co-
incident, as shown in Fig. 14. In this case, the direction cosines are:
cxV x ¼ cos ðhÞ
cxV y ¼ cos ð90B hÞ ¼ sin ðhÞ
cxV z ¼ cos ð90B Þ ¼ 0
cyV x ¼ cos ð90B þ hÞ ¼ sin ðhÞ
cyV y ¼ cos ðhÞ
cyV z ¼ cos ð90B Þ ¼ 0
czV x ¼ cos ð90B Þ ¼ 0
czV y ¼ cos ð90B Þ ¼ 0
czV z ¼ cos ð0B Þ ¼ 1
If we now (a) substitute these direction cosines into Eq. (12b), (b) label
the shear stresses using the symbol s rather r, and (c) note that rzz=
sxz=syz=0 by assumption, then Eq. (12b) becomes:
2 3 2 32 3
rxV xV sxV yV sxV zV cos h sin h 0 rxx sxy 0
6s V V r V V s V V7 6 76 7
4 yx yy y z 5 ¼ 4 sin h cos h 0 54 syx ryy 0 5
szV xV szV yV rzV zV 0 0 1 0 0 0
2 3
cos h sin h 0
6 7
4 sin h cos h 0 5
0 0 1
Completing the matrix multiplication indicated results in:
2 3
rxV xV s x V yV s xV z V
4 s yV x V r yV yV s yV z V 5
s z V xV s z V yV r zV zV
2 3
cos2 hrxx þ sin2 hryy þ 2cosh sin hsxy cosh sinhrxx þ cosh sinhryy þ ðcos2 h sin2 hÞsxy 0
6 7
¼6
4cosh sinhrxx þ cosh sinhryy þ ðcos h sin hÞsxy
2 2
sin2 hrxx þ cos2 hryy 2cosh sinhsxy 07
5
0 0 0
As would be expected, the out-of-plane stresses are zero: rzVzV=sxVzV=s yVzV=0.

The remaining stress components are:
rxV xV ¼ cos 2 ðhÞrxx þ sin 2 ðhÞryy þ 2cos ðhÞsin ðhÞsxy
ryV yV ¼ sin 2 ðhÞrxx þ cos 2 ðhÞryy 2cos ðhÞsin ðhÞsxy ð19Þ
sxV yV ¼ cos ðhÞsin ðhÞrxx þ cos ðhÞsin ðhÞryy þ ½cos 2 ðhÞ sin 2 ðhÞsxy

Figure 14 Transformation of a plane stress element from one coordinate sys-
tem to another.

Equation (19) can be written using matrix notation as:
8 9 2 3
< rx V x V >
> = cos2 ðhÞ sin2 ðhÞ 2cos ðhÞ sin ðhÞ
ryV yV ¼ 4 6 7
sin2 ðhÞ cos2 ðhÞ 2 cos ðhÞ sin ðhÞ 5
>
: >
;
sxV yV cos ðhÞsin ðhÞ cos ðhÞsin ðhÞ cos2 ðhÞ sin2 ðhÞ
8 9
< rxx >
> =
ryy ð20Þ
>
: >
;
sxy
It should be kept in mind that these results are valid only for a state
of plane stress. Transformation of a general (3-D) stress/tensor should be
performed using Eq. (12a) or (12b).
The 3 3 array that appears in Eq. (20) is called the transformation
matrix and is abbreviated as [T]:
2 3
cos2 ðhÞ sin2 ðhÞ 2cosðhÞsin ðhÞ
6 7
½T ¼ 6
4 sin2 ðhÞ cos2 ðhÞ 2cosðhÞsin ðhÞ 75 ð21Þ
cos ðhÞsin ðhÞ cos ðhÞ sin ðhÞ cos2 ðhÞ sin2 ðhÞ
The stress invariants [given by Eqs. (14a)–(14c) or Eqs. (15a)–(15c)]

are considerably simplified in the case of plane stress. Since by definition
rzz=sxz=syz=0, the stress invariants become:
First stress invariant ¼ H ¼ rxx þ ryy
Second stress invariant ¼ U ¼ rxx ryy s2xy ð22Þ
Third stress invariant ¼ W ¼ 0
The principal stresses equal the roots of the cubic equation previously
listed as Eq. (17). In the case of plane stress, this cubic equation becomes (since
W=0):
r3 Hr2 þ Ur ¼ 0 ð23Þ
Obviously, one root of Eq. (23) is r=0. This root corresponds to rzz
and for present purposes will be labeled rp3 although it may not be the alge-
braically least principal stress. Thus, in the case of plane stress, the z-axis is a
principal stress direction and rzz=rp3=0 is one of the three principal
stresses. Since the three principal stress directions are orthogonal, this im-
plies that the remaining two principal stress directions must lie within the
x–y plane.

Removing the known root from Eq. (23), we have the following qua-
dratic equation:
r2 Hr þ U ¼ 0 ð24Þ
The two roots of this quadratic equation (that is, the two remaining princi-
pal stresses, rp1 and rp2) may be found by application of the standard ap-
proach (3) and are given by:
1h pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi i
rp1 ; rp2 ¼ HF H2 4U ð25Þ
2
Substituting Eq. (22) into Eq. (25) and simplifying yields the following:
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
r r 2 ffi
rxx þ ryy xx yy
rp1 ; rp2 ¼ F þsxy 2 ð26Þ
2 2
The angle hp between the x-axis and either the p1 or p2 axis is given by:

1 2sxy
hp ¼ arctan ð27Þ
2 rxx ryy
Example Problem 5
Given. The plane stress element shown in Fig. 15(a).
Problem. (a) Rotate the stress element to a new coordinate system oriented
25j clockwise from the x-axis, and redraw the stress element with all stress
components properly oriented; (b) determine the principal stresses and
principal stress coordinate system, and redraw the stress element with the
principal stress components properly oriented.
Solution
Part (a). The following components of stress are implied by the stress
element shown (note that the shear stress is algebraically negative, in ac-
cordance with the sign convention discussed in Sec. 5):
rxx ¼ 70 MPa
ryy ¼ 15 MPa
sxy ¼ 50 MPa
The stress element is to be rotated clockwise. That is, the +xV-axis is rotated
away from the +y-axis. Applying the right-hand rule, it is clear that this is a
negative rotation:
h ¼ 25B

Figure 15 Plane stress elements associated with Example Problem 5.

Equation (20) becomes:
8 9 2 3
8 9
> cos2 ð25B Þ sin2 ð25B Þ 2cosð25B Þsinð25B Þ
< rxV xV >
= 6 7 < 70 =
ryV yV ¼ 6
4 sin2 ð25B Þ cos2 ð25B Þ 2cosð25B Þsinð25B Þ 7
5: 15 ;
>
:s V V > ;
B B B B B B
50
xy
cosð25 Þsinð25 Þ cosð25 Þsinð25 Þ cos ð25 Þ sin ð25 Þ
2 2
8 9 2 38 9 8 9
> 0:8214 0:1786 0:7660 > 70 > > 98:5 >
< rxV xV >
= 6 7
>
< >
= >
< >
=
ryV yV ¼ 6 4 0:1786 0:8214 7
0:7660 5 15 ¼ 13:5 MPa
>
: >
; >
> >
> > >
sxV yV : ; > : >
;
0:3830 0:3830 0:6428 50 11:1
The rotated stress element is shown in Fig. 15(b).

Part (b). The principal stresses are found through application of Eq. (26):
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

70 þ 15 70 15 2
rp1 ; rp2 ¼ F þð50Þ2 ¼ 42:5 F 57:1 MPa
2 2
rp1 ¼ 99:6MPa
rp2 ¼ 14:6MPa
The orientation of the principal stress coordinate system is given by Eq. (27):

1 2ð50Þ
hp ¼ arctan ¼ 31B
2 70 15
The stress element is shown in the principal stress coordinate system in Fig.
15(c).
9 DEFINITION OF STRAIN
All materials deform to some extent when subjected to external forces and/or
environmental changes. In essence, the state of strain is a measure of the
magnitude and orientation of the deformations induced by these effects. As
in the case of stress, there are two types of strain: normal strain and shear
strain.
The two types of strain can be visualized using the strain element
shown in Fig. 16. Imagine that a perfect square has been physically drawn
on a surface of interest. Initially, angle BABC is exactly p2 radians (i.e.,
initially BABC=90j) and sides AB and BC are of exactly equal lengths.
Now suppose that some mechanism(s) causes the surface to deform. The
mechanism(s) which causes the surface to deform need not be defined at

Figure 16 2-D element used to illustrate normal and shear strains (deforma-
tions shown are greatly exaggerated for clarity).
this point, but might be external loading (i.e., stresses), a change in tem-
perature, and/or (in the case of polymeric-based materials such as compo-
sites) the adsorption or desorption of water molecules. In any event, since
the surface is deformed, the initially square element drawn on the surface is
deformed as well. As shown in Fig. 16, point A moves to point AV and point
C moves to point CV. It is assumed that the element remains a parallelogram,
i.e., it is assumed that sides AVB and CVB remain straight lines after
deformation. This assumption is valid if the element is infinitesimally small.
In the present context, ‘‘infinitesimally small’’ implies that lengths AB and
CB are small enough such that the deformed element may be treated as a
parallelogram.
Normal strain exx is defined as the change in length of AB divided by the
original length of AB:
DAB
exx ¼ ð28Þ
AB
The change in length AB is given by:
DAB ¼ ðAVB ABÞ
From the figure, it can be seen that the projection of length AVB in the x-
direction, that is, length AUB, is given by:
AWB ¼ AVBcosðBAVBAÞ ð29Þ

If we now assume that BAVBA is ‘‘small,’’ then we can invoke the small-angle
approximation,* which states that if BAVBA is expressed in radians and is less
than about 0.1745 radians (about 10j), then:
sinðBAVBAÞcBAVBA tanðBAVBAÞcBAVBA cosðBAVBAÞc1
Based on the small-angle approximation, Eq. (29) implies that AUBc
AVB, and therefore the change in length of AB is approximately given by:
DABcðAUB ABÞ ¼ AUA
Equation (28) can now be written as:
AUA
exx ¼ ð30Þ
AB
In an entirely analogous manner, normal strain eyy is defined as the
change in length of CB divided by the original length of CB:
DCB
eyy ¼
CB
Based on the small-angle approximation, the change in length of CB
is approximately given by:
DCB ¼ ðC VB CBÞcC UC
and therefore:
C UC
eyy ¼ ð31Þ
CB
As before, the approximation for change in length CB is valid if angle
BCUBC is small.
Recall that the original element shown in Fig. 16 was assumed to be
perfectly square and, in particular, that angle BAVBC is exactly p2 radians (i.e.,
initially BAVBC=90j). Engineering shear strain is defined as the change in
angle BABC, expressed in radians:
cxy ¼ DðBABCÞ ¼ BAVBA þ BC VBC ð32Þ
The subscripts associated with a shear strain [e.g., subscripts xy in Eq. (32)]
indicate that the shear strain represents the change in angle defined by line
segments originally aligned with the x- and y-axes.
* The reader is encouraged to personally verify the ‘‘small-angle approximation.’’ For example,
use a calculator to demonstrate that an angle of 5j equals 0.08727 rad, and that sin(0.08727
rad)=0.08716, tan(0.08727 rad)=0.08749, and cos(0.08727 rad)=0.99619. Therefore, in this
example, the small angle approximation results in a maximum error of less than 1%.

As discussed in the following sections, it is very convenient to describe a
state of strain as a second-order tensor. However, in order to do so, we must
use a slightly different definition of shear strain. Specifically, tensoral shear
strain is defined as:
1
exy ¼ c ð33Þ
2 xy
Since engineering shear strain has been defined as the total change in
angle BAVBC, tensoral shear strain is simply half this change in angle. The
use of tensoral shear strain is convenient because it greatly simplifies the
transformation of a state of strain from one coordinate system to another.
The use of engineering shear strain is far more common in practice, how-
ever. In this text, tensoral shear strain will be used when convenient during
initial mathematical manipulations of the strain tensor, but all final results
will be converted to relations involving engineering shear strain.
Although strains are unitless quantities, normal strains are usually
reported in units of (length/length), and shear strains are usually reported
in units of radians. The values of a strain is independent of the system of units
used, e.g., 1 (m/m)=1 (in/in). Common abbreviations used throughout this
text are as follows:
1 106 meter=meter ¼ 1micrometer=meter ¼ 1 Am=m ¼ 1 Ain=in
1 106 radians ¼ 1 microradians ¼ 1 Arad
We must next define the algebraic sign convention used to describe
individual strain components. The sign convention for normal strains is very
straightforward and intuitive: a positive (or ‘‘tensile’’) normal strain is asso-
ciated with an increase in length, while a negative (or ‘‘compressive’’) normal
strain is associated with a decrease in length.
To define the algebraic sign of a shear strain, we first identify the
algebraic sign of each face of the infinitesimal strain element (the algebraic
sign of face was defined in Sec. 5). An algebraically positive shear strain cor-
responds to a decrease in the angle between two positive faces, or equiva-
lently, to a decrease in the angle between two negative faces.
The above sign conventions can be used to confirm that all strains
shown in Fig. 16 are algebraically positive.
Example Problem 6
Given. The following two sets of strain components:
exx ¼ 1000 Am=m exx ¼ 1000 Am=m
Set1 : eyy ¼ 500 Am=m Set2 : eyy ¼ 500 Am=m
cxy ¼ 1500 Arad cxy ¼ 1500 Arad

Figure 17 Strain elements associated with Example Problem 6 (not to scale).
Determine. Prepare sketches (not to scale) of the deformed strain elements

represented by the two sets of strain components.
Solution. The required sketches are shown in Fig. 17. Note that the only
difference between the two sets of strain components is that in Set 1, cxy is
algebraically positive, whereas in Set 2, cxy is algebraically negative.
10 THE STRAIN TENSOR

A general 3-D solid body is shown in Fig. 18(a). An infinitesimally small
cube isolated from an interior region of the body is shown in Fig. 18(b). The
cube is referenced to an x–y–z coordinate system, and the cube edges are
aligned with these axes.
Now assume that the body is subjected to some mechanism(s) which
causes the body to deform. The mechanism(s) which causes this deforma-
tion need not be defined at this point, but might be external loading (i.e.,
stresses), a change in temperature, the adsorption or desorption of water
molecules (in the case of polymeric-based materials such as composites), or
any combination thereof.
Since the entire body is deformed, the internal infinitesimal cube is
deformed into a parallelepiped, as shown in Fig. 18(c). It can be shown (1,2)
that the state of strain experienced by the cube can be represented as a

Figure 18 Infinitesimal elements used to illustrate the strain tensor.

symmetric second-order tensor, involving six components of strain: three
normal strains (exx, eyy, ezz) and three tensoral shear strains (exy, exz, eyz). These
six strain components are defined in the same manner as those discussed in
the preceding section. Normal strains exx, eyy, and ezz represent the change
in length in the x-, y-, and z-directions, respectively. Tensoral shear strains
exy, exz, and eyz represent the change in angle between cube edges initially
aligned with the (x-,y-), (x-,z-), and ( y-,z-) axes, respectively. Using matrix
notation, the strain tensor may be written as:
2 3 2 3
exx exy exz exx exy exz
4 eyx eyy eyz 5 ¼ 4 exy eyy eyz 5 ð34Þ
ezx ezy ezz exz eyz ezz
Alternatively, the strain tensor can be succinctly written using indicial nota-
tion as:
eij ; i; j ¼ x; y; or z ð35Þ
Note that if engineering shear strain is used, then Eq. (34) becomes
2 3 2 3
exx exy exz exx ðcxy =2Þ ðcxz =2Þ
6 7 6 ðcyz =2Þ 7
4 exy eyy eyz 5 ¼ 4 ðcxy =2Þ eyy 5
exz eyz ezz ðcxx =2Þ ðcyz =2Þ ezz
If engineering shear strain is used, the strain tensor cannot be written using
indicial notation [as in Eq. (35)] due to the 1/2 factor that appears in all off-
diagonal positions.
In Sec. 1, it was noted that a force vector is a first-order tensor since
only one subscript is required to describe a force tensor, Fi. The fact that
strain is a second-order tensor is evident from Eq. (35) since two subscripts
are necessary to describe a state of strain.
11 TRANSFORMATION OF THE STRAIN TENSOR

Since both stress and strain are second-order tensors, the transformation of
the strain tensor from one coordinate system to another is analogous to the
transformation of the stress tensor, as discussed in Sec. 6. For example, it
can be shown (1,2) that the strain components in the xV–yV–zV coordinate
system (eiVjV) are related to the components in x–y–z coordinate system (eij)
according to:
ei V j V ¼ ci V k cj V l ekl where k; l ¼ x; y; z ð36aÞ

Alternatively, using matrix notation, the strain tensor transforms accord-
ing to:
½ei V j V ¼ ½ci V j ½eij ½ci V j T
which expands as follows:

2 3 2 32 3
e xV xV e xV y V e xV z V c xV x c xV y c xV z exx exy exz
6e V V 7 6 7
eyV zV 5 ¼ 4 cyV x cyV y cyV z 54 eyx 6 7
4 yx ey V y V eyy eyz 5
e z V xV e zV y V e zV zV czV x czV y czV z ezx ezy ezz
2 3
6c V c V c V 7
4 xy yy zy5 ð36bÞ
The terms ci Vj which appear in Eqs. (36a) and (36b) are direction cosines
and equal the cosine of the angle between the axes of the xV–yV–zV and x–y–z
coordinate systems.
As was the case for the stress tensor, there are certain features of the
strain tensor that do not vary when the tensor is transformed from one co-
ordinate system to another. These features are called the strain invariants.
Three independent strain invariants exist and are defined as follows:
First strain invariant ¼ He ¼ eii ð37aÞ

1
Second strain invariant ¼ Ue ¼ eii ejj eij eij ð37bÞ
2
1
Third strain invariant ¼ We ¼ eii ejj ekk 3eii ejk ejk þ 2eij ejk eki ð37cÞ
6
Alternatively, by expanding these equations over the range i, j, k=x, y, z and
simplifying, the strain invariants can be written as:
First strain invariant ¼ He ¼ exx þ eyy þ ezz ð38aÞ

Second strain invariant ¼ Ue ¼ exx eyy þ exx ezz þ eyy ezz

e2xy þ e2xz þ e2yz ð38bÞ
Third strain invariant ¼ We ¼ exx eyy ezz exx e2yz eyy e2xz
ezz e2xy þ 2exy exz eyz ð38cÞ

Example Problem 7
Given. A state of strain referenced to an x–y–z coordinate is known to be:
2 3 2 3
exx exy exz 1000 Am=m 500 Arad 250 Arad
6 7 6 7
4 eyx eyy eyz 5 ¼ 4 500 Arad 1500 Am=m 750 Arad 5
ezx ezy ezz 250 Arad 750 Arad 2000 Am=m
It is desired to express this state of strain in an xU–yU–zU coordinate system,
generated by:
(i) Rotation of h=20j about the original z-axis (which defines an
intermediate xV–yV–zV coordinate system), followed by
(ii) Rotation of b=35j about the xV-axis (which defines the final xU–
yU–zU coordinate system).
(This coordinate transformation has been previously considered in
Example Problem 3 and is shown in Fig. 10.)
Problem
(a) Rotate the strain tensor to the xU–yU–zU coordinate system and
(b) Calculate the first, second, and third invariants of the strain tensor
using both elements of the strain tensor referenced to the x–y–z
coordinate system, eij, and elements of the strain tensor referenced
to the xU–yU–zU coordinate system, ei Uj U.
Solution
Part (a). General expressions for direction cosines relating the x–y–z and
xU–yU–zU coordinate systems were determined as a part of Example Problem 1.
Further, numerical values for the particular rotation h=20j and b=35j were
determined in Example Problem 3 and were found to be:
cxWx ¼ cos ð20B Þ ¼ 0:9397

cxWy ¼ sin ð20B Þ ¼ 0:3420
cxWz ¼ 0
cyWx ¼ cos ð35B Þsin ð20B Þ ¼ 0:2802
cyWy ¼ cos ð35B Þcos ð20B Þ ¼ 0:7698
cyWz ¼ sin ð35B Þ ¼ 0:5736
czWx ¼ sin ð35B Þsin ð20B Þ ¼ 0:1962

czWy ¼ sin ð35B Þcos ð20B Þ ¼ 0:5390
czWz ¼ cos ð35B Þ ¼ 0:8192
Each component of the transformed strain tensor can now be found through
application of Eq. (36a) or Eq. (36b). For example, setting iV=xU, jV=xU, and
expanding Eq. (36a), strain component exUxU is given by:
ex Ux U ¼ cx Ux cx Ux exx þ cx Ux cx Uy exy þ cx Ux cx Uz exz þ cx Uy cx Ux eyx þ cx Uy cx Uy eyy

þ cx Uy cx Uz eyz þ cx Uz cx Ux ezx þ cx Uz cx Uy ezy þ cx Uz cx Uz ezz
ex Ux U ¼ ð0:9397Þð0:9397Þð1000Þ þ ð0:9397Þð0:3420Þð500Þ
þð0:9397Þð0Þð250Þ þ ð0:3420Þð0:9397Þð500Þ
þð0:3420Þð0:3420Þð1500Þ þ ð0:3420Þð0Þð750Þ
þð0Þð0:9397Þð250Þ þ ð0Þð0:3420Þð750Þ
þð0Þð0Þð2000Þ
ex Ux U ¼ 1380 lm=m
Alternatively, if matrix notation is used, then Eq. (36b) becomes:
2 3 2 3
exUxU exUyU exUzU 0:9397 0:3420 0
6 7 6 7
4 eyUxU eyUyU eyUzU 5 ¼ 4 0:2802 0:7698 0:5736 5
ezUxU ezUyU ezUzU 0:1962 0:5390 0:8192
2 3
1000 500 250
6 7
4 500 1500 750 5
250 750 2000
2 3
0:9397 0:2802 0:1962
6 7
4 0:3420 0:7698 0:5390 5
0 0:5736 0:8192
Completing the matrix multiplication indicated, there results:
2 3 2 3
exUxU exUyU exUzU 1380 Am=m 727 Arad 91 Arad
6 7 6 7
4 eyUxU eyUyU eyUzU 5 ¼ 4 727 Arad 1991 Am=m 625 Arad 5
ezUxU ezUyU ezUzU 91 Arad 625 Arad 1129 Am=m
Notice that the value of exUxU determined through matrix multiplication is
identical to that obtained using indicial notation, as expected.

Part (b). The first, second, and third strain invariants will now be calcu-
lated using components of both eij and eiUjU. It is expected that identical values
will be obtained since the strain invariants are independent of the coordinate
system.
First strain invariant:

He ¼ eii ¼ exx þ eyy þ ezz
He ¼ ð1000 þ 1500 þ 2000Þ Am=m
He ¼ 4500Am=m ¼ :004500 m=m
He ¼ eiUiU ¼ exUxU þ eyUyU þ ezUzU
He ¼ ð1380 þ 1991 þ 1129Þ Am=m
He ¼ 4500Am=m ¼ 0:004500 m=m
As expected, the first strain invariant is independent of the coordinate
system.
Second strain invariant:

1
Ue ¼ eii ejj eij eij ¼ exx eyy þ exx ezz þ eyy ezz e2xy þ e2xz þ e2yz
2

Ue ¼ ð1000Þð1500Þ þ ð1000Þð2000Þ þ ð1500Þð2000Þ
½ð500Þ2 þ ð250Þ2 þ ð750Þ2 g ðAm=mÞ2
U ¼ 5:625 106 ðAm=mÞ2 ¼ 5:625 106 ðm=mÞ2
1
Ue ¼ eiUiU ejUjU eiUjU eiUjU
2

Ue ¼ exUxU eyUyU þ exUxU ezUzU þ eyUyU ezUzU e2xUyU þ e2xUzU þ e2yUzU
Ue ¼ fð1380Þð1991Þ þ ð1380Þð1129Þ þ ð1991Þð1129Þ

½ð727Þ2 þ ð91Þ2 þ ð625Þ2 g ðAm=mÞ2
Ue ¼ 5:625 106 ðAm=mÞ2 ¼ 5:625 106 ðm=mÞ2

As expected, the second strain invariant is independent of the coordi-
nate system.
Third strain invariant:

1
We ¼ eii ejj ekk 3eii ejk ejk þ 2eij ejk eki
6
We ¼ exx eyy ezz exx e2yz eyy e2xz ezz e2xy þ 2exy exz eyz
We ¼ ½ð1000Þð1500Þð2000Þ ð1000Þð750Þ2 ð1500Þð250Þ2 ð2000Þ
ð500Þ2 þ 2ð500Þð250Þð750Þ ðAm=mÞ3
W ¼ 2:031 109 ðAm=mÞ3 ¼ 2:031 109 ðm=mÞ3

1
We ¼ eiUiU ejUjU ekUkU 3eiUiU ejUkU ejUkU þ 2eiUjU ejUkU ekUiU
6
We ¼ exUxU eyUyU ezUzU exUxU e2yUzU eyUyU e2xUzU ezUzU e2xUyU þ 2exUyU exUzU eyUzU
We ¼ ½ð1380Þð1991Þð1129Þ ð1380Þð625Þ2 ð1991Þð91Þ2 ð1129Þ
ð727Þ2 þ 2ð727Þð91Þð625Þ ðAm=mÞ3
W ¼ 2:031 109 ðAm=mÞ3 ¼ 2:031 109 ðm=mÞ3
As expected, the third stress invariant is independent of the coordinate

system.
12 PRINCIPAL STRAINS
The definition of the strain tensor was reviewed in Sec. 10, and transfor-
mation of the strain tensor from one coordinate system to another was
discussed in Sec. 11. It can be shown (1,2) that it is always possible to rotate
the strain tensor to a special coordinate system in which no shear strains
exist. This coordinate system is called the principal strain coordinate system,
and the normal strains that exist in this coordinate system are called principal
strains. In most texts, the principal strains are denoted e1, e2, and e3.
However, as will be discussed later (see Fig. 2 in Chap. 3), in this text, the
axis labels ‘‘1,’’ ‘‘2,’’ and ‘‘3’’ will be used to refer to the principal material
coordinate system rather than the directions of principal strain. Therefore,

in this text, the axes associated with the principal strain coordinate system
will be labeled p1, p2, and p3 axes, and the principal strains will be denoted
ep1, ep2, and ep3.
Since both stress and strain are second-order tensors, the principal
strains can be found using an approach analogous to that used to find prin-
cipal stresses. Specifically, it can be shown (1,2) that the principal strains
must satisfy the following three simultaneous equations:
ðepn exx Þcpn x exy cpn y exz cpn z ¼ 0 ð39aÞ

exy cpn x þ ðepn eyy Þcpn y eyz cpn z ¼ 0 ð39bÞ
exz cpn x eyz cpn y þ ðepn ezz Þcpn z ¼ 0 ð39cÞ
Since direction cosines cpnx, cpny, and cpnz must also satisfy Eq. (8), and
therefore cannot all equal zero, the solution can be obtained by requiring
that the determinant of the coefficients of cpnx, cpny, and cpnz equal zero:

ðepn exx Þ exy exz

exy ðepn eyy Þ eyz ¼ 0

exz eyz ðepn ezz Þ
Equating the determinant to zero results in the following cubic
equation:
e3pn He e2pn þ Ue epn We ¼ 0 ð40Þ
where He, Ue, and We are the first, second, and third strain invariants,
respectively, and have been previously listed as Eqs. (37a)–(37c) and (38a)–
(38c). The three roots of the cubic equation (that is, the three principal
strains) may be found by application of the standard approach (3), as
follows:
Define:
1
a¼ 3Ue H2e
3
1 3
b¼ 2He 9He Ue þ 27We
27
3 b b2 a3
A¼ þ þ
2 4 27
3 b b2 a 3
B¼ þ
2 4 27

The three principal strains [i.e., the three roots of Eq. (40)] are then given by:
8
> He
>
> AþBþ
>
> 3
>
>
< AþB p ffiffiffiffiffiffi
ffi AB He
ep1 ; ep2 ; ep3 ¼ þ 3 þ ð41Þ
>
> 2 2 3
>
>
>
> AþB p ffiffiffiffiffiffi
ffi AB He
>
: 3 þ
2 2 3
By convention, the principal strains are numbered such that ep1 is the
algebraically greatest principal strain, whereas ep3 is the algebraically least.
That is, ep1>ep2>ep3.
It is appropriate to note that many handheld calculators and computer
software packages now feature standard routines to find the roots of nth-
order polynomials. Hence, the roots of Eq. (40) may often be found more
conveniently using these standard routines, rather than Eq. (41). The cal-
culation of constants A and B that appear in Eq. (41) often involves finding
the cube root of a complex number. The need to find the cube root of a
complex number is encountered infrequently, and hence the reader may not
be aware of how to make such a calculation. For convenience, a process that
may be used to find the cube root of a complex number has been included
in Appendix A.
Once the principal strains are determined, the three sets of direction
cosines (which define the principal coordinate directions) are found by
substituting the three principal strains given by Eq. (41) into Eqs. (39a)–
(39c) in turn. Since only two of Eqs. (39a)–(39c) are independent, Eq. (8) is
used as a third independent equation involving the three unknown constants,
cpnx, cpny, and cpnz.
The process of finding principal strains and direction cosines will be
demonstrated in the following example problem.
Example Problem 8
Given. A state of strain referenced to an x–y–z coordinate is known to be:
2 3 2 3
6 7 6 7
Problem. Find (a) the principal strains and (b) the direction cosines that
define the principal strain coordinate system.

Solution. This is the same strain tensor considered in Example Problem 7. As
a part of that problem, the first, second, and third strain invariants were found
to be:
He ¼ 0:004500 m=m
U ¼ 5:625 106 ðm=mÞ2
W ¼ 2:031 109 ðm=mÞ3
Part (a). Determining the principal strains.
In accordance with Eq. (40), the three principal strains are the roots
of the following cubic equation:
e3pn ð0:004500Þe2pn þ ð5:625 106 Þepn ð2:031 109 Þ ¼ 0
Following the standard procedure for finding the roots of a cubic equa-
tion, we have:
1 1h i
a¼ 3Ue H2e ¼ 3ð5:625 106 Þ ð0:004500Þ2
3 3
¼ 1:125 106
1 3
b¼ 2He 9He Ue þ 27We
27
1 h
b¼ 2ð0:004500Þ3 9ð0:004500Þð5:625 106 Þ
27
þ27ð2:031 109 Þ ¼ 3:435 1010
3 b b2 a3
A¼ þ þ
2 4 27
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u
u
t 3:435 10
3 10 ð3:435 1010 Þ2 ð1:125 106 Þ3
¼ þ þ
2 4 27
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
3
A ¼ 1:718 1010 þ ið1:524 1010 Þ
A ¼ 594:5 106 þ ið146:7 106 Þ

3 b b2 a 3
B¼ þ
2 4 27
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u
u
t
3 3:435 1010 ð3:435 1010 Þ2 ð1:125 106 Þ3
¼ þ
2 4 27

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
3
B¼ 1:718 1010 ið1:524 1010 Þ
B ¼ 594:5 106 ið146:7 106 Þ

8 8
>
> H >
> 2689Am=m
>
> AþBþ >
>
>
> 3 >
>
>
< >
<
A þ B pffiffiffiffiffiffiffi A B H
ep1 ; ep2 ; ep3 ¼ þ 3 þ ¼ 651Am=m
>
> 2 2 3 >
>
>
> >
>
>
>
> A þ B pffiffiffiffiffiffiffi A B H >
>
>
: 3 þ :
2 2 3 1160Am=m
Hence, ep1=2689 Am/m, ep2=1160 Am/m, and ep3=651 Am/m.

Part (b). Determining the direction cosines.
Equations (8), (39a), and (39b) will be used to form three independent
equations in three unknowns. We have:
ðepn exx Þcpn x exy cpn y exz cpn z ¼ 0
exy cpn x þ ðepn eyy Þcpn y eyz cpn z ¼ 0
ðcpn x Þ2 þ ðcpn y Þ2 þ ðcpn z Þ2 ¼ 1

Direction cosines for ep1. The three independent equations become:
ð2689 1000Þcp1 x 500cp1 y 250cp1 z ¼ 0

500cp1 x þ ð2689 1500Þcp1 y 750cp1 z ¼ 0

cp1 x ¼ 0:2872 cp1 y ¼ 0:5945 cp1 z ¼ 0:7511
ð1160 1000Þcp2 x 500cp2 y 250cp2 z ¼ 0

500cp2 x þ ð1160 1500Þcp2 y 750cp2 z ¼ 0

cp2 x ¼ 0:5960 cp2 y ¼ 0:5035 cp2 z ¼ 0:6256

ð651 1000Þcp3 x 500cp3 y 250cp3 z ¼ 0
500cp3 x þ ð651 1500Þcp3 y 750cp3 z ¼ 0

cp3 x ¼ 0:7481 cp3 y ¼ 0:6286 cp3 z ¼ 0:2128
13 STRAINS WITHIN A PLANE PERPENDICULAR TO A

PRINCIPAL STRAIN DIRECTION
It has been seen that a strain tensor is defined by six components of strain:
three normal strain components and three shear strain components. Now, in
practice, there are circumstances in which it is known a priori that both shear
strain components in one direction are zero: exz=eyz=0, say (or equivalently,
cxz=cyz=0). This implies that the z-axis is a principal strain axis. In these
instances, we are primarily interested in the strains induced within the x–y
plane, exz, eyy, and exy. Two different circumstances are encountered in which
it is known a priori that the z-axis is a principal strain axis.
In the first case, all three out-of-plane strain components in the z-
direction are known a priori to equal zero. That is, it is known a priori that
ezz=exz=eyz=0. Not only is the z-axis a principal strain axis in this case, but,
in addition, the principal strain equals zero: ezz=ep3=0. Since the three
remaining nonzero strain components (exx, eyy, and exy) all lie within the x–y
plane, it is natural to call this condition a state of plane strain. Plane strain
conditions occur most often because of the geometry of the structure of
interest. Specifically, the plane strain condition usually exists in internal
regions of very long (or very thick) structures. Examples include solid shafts
or long dams. In these instances, the strains induced along the long axis of the
structure are often negligibly small compared to those induced within the
transverse plane of the structure.
The second case in which the out-of-plane z-axis may be a principal axis
is when a structure is subjected to a state of plane stress. As has been discussed
in Sec. 8, the state of plane stress occurs most often in thin, platelike
structures. In this case, the z-axis is a principal strain axis, and ezz is again
one of the principal strains. However, in this second case, the out-of-plane
normal strain does not, in general, equal zero: ezz p 0.
It is emphasized that a state of plane stress usually, but not always,
causes a state of strain in which the z-axis is a principal strain axis. This point

will be further discussed in Chap. 4. It will be seen there that it is possible for a
material to exhibit a coupling between in-plane stresses and out-of-plane
shear strains. That is, in some cases, stresses acting within the x–y plane (rxx,
ryy, and/or sxy) can cause out-of-plane shear strains (exz and/or eyz). In these
instances, the out-of-plane z-axis is not a principal strain axis, although the
out-of-plane stresses all equal zero.
In any event, for present purposes, assume that it is known a priori that
the out-of-plane z-axis is a principal strain axis, and we are primarily inter-
ested in the strains induced within the x–y plane, exx, eyy, and exy. We will write
these strains in the form of a column array, rather than a 33 array:
2 3 2 3 8 9 8 9
exx exy 0 exx ðcxy =2Þ 0 < exx >
> = >
e
< xx > =
6 7 6 7
4 exy eyy 0 5 ¼ 4 ðcxy =2Þ eyy 0 5 ! eyy ¼ eyy
>
: >
; > : >
;
0 0 ezz 0 0 ezz exy cxy =2
Note that ezz does not appear in the column array. This is not of concern in the
case of plane strain since, in this case, ezz=0. However, in the case of plane
stress, it is important to remember that (in general) ezz p 0. Although in the
following chapters we will be primarily interested in strains induced within
the x–y plane, the reader is advised to remember that an out-of-plane strain
ezz is also induced by a state of plane stress.
The transformation of a general 3-D strain tensor has already been
discussed in Sec. 11. The relations presented there will now be simplified for
the case of transformation of strains within a plane.
Recall that either Eq. (36a) or Eq. (36b) governs the transformation of
a strain tensor from one coordinate system to another. Equation (36b) is
2 3 2 32 3
e xV xV e xV y V e xV z V c xV x c xV y c xV z exx exy exz
6e V V e V V e V V 7 6c V cyV y cyV z 7 6 7
4 yx yy yz 5 ¼ 4 yx 54 eyx eyy eyz 5
e z V xV e z V y V e z V z V czV x czV y czV z ezx ezy ezz
2 3
6 7
4 cxV y cyV y czV y 5 ðrepeatedÞð36bÞ
Assuming that the xV–yV–zV coordinate system is generated from the x–y–z
system by a rotation h about the z-axis, the direction cosines are:
cxV x ¼ cos ðhÞ
cxV y ¼ cos ð90B hÞ ¼ sin ðhÞ
cxV z ¼ cos ð90B Þ ¼ 0

cyV x ¼ cos ð90B þ hÞ ¼ sin ðhÞ
cyV y ¼ cos ðhÞ
cyV z ¼ cos ð90B Þ ¼ 0
czV x ¼ cos ð90B Þ ¼ 0
czV y ¼ cos ð90B Þ ¼ 0
czV z ¼ cos ð0B Þ ¼ 1
Substituting these direction cosines into Eq. (36b) and noting that by
assumption, exz=eyz=0, we have:
2 3 2 32 3
exV xV exV yV exV zV cos h sin h 0 exx exy 0
6e V V e V V e V V 7 6 76 7
4 yx yy y z 5 ¼ 4 sin h cos h 0 54 eyx eyy 0 5
ezV xV ezV yV ezV zV 0 0 1 0 0 ezz
2 3
cos h sin h 0
6 7
4 sin h cos h 05
0 0 1

2 3
e xV xV e x V yV exV zV
6 7
6 e yV x V e yV yV e yV z V 7
4 5
e z V xV e z V yV e zV zV
2 3
cos2 hexx þ sin2 heyy þ 2 cos h sin hexy cos h sin hexx þ cos h sin heyy þ ðcos2 h sin2 hÞexy 0
¼4cos h sin hexx þ cos h sin heyy þ ðcos2 h sin hÞexy
2
sin hexx þ cos2 heyy 2 cos h sin hexy
2
05
0 0 ezz
As would be expected, exVzV=eyVzV=0. The remaining strain components are:

exV xV ¼ cos2 ðhÞexx þ sin2 ðhÞeyy þ 2 cosðhÞsinðhÞexy
eyV yV ¼ sin2 ðhÞexx þ cos2 ðhÞeyy 2 cosðhÞsinðhÞexy ð42Þ
exV yV ¼ cosðhÞsinðhÞexx þ cosðhÞsinðhÞeyy þ ½cos ðhÞ þ sin ðhÞexy
2 2
ezV zV ¼ ezz
Tensoral shear strains were used in Eqs. (36a) and (36b) for mathemat-
ical convenience; that is, tensoral shear strains have been used so that rotation
of the strain tensor could be accomplished using the normal transformation
law for a second-order tensor. Since engineering shear strains are far more
commonly used in practice, we will now convert our final results, Eq. (42), to
ones which involve engineering shear strain (cxy). Recall from Sec. 9 that

exy ¼ 12 cxy . Hence, to convert Eq. (42), simply replace exy with 12 cxy every-
where, resulting in:
exV xV ¼ cos2 ðhÞexx þ sin2 ðhÞeyy þ cosðhÞsinðhÞcxy
eyV yV ¼ sin2 ðhÞexx þ cos2 ðhÞeyy cosðhÞsinðhÞcxy ð43Þ
c xV y V c xy
¼ cosðhÞsinðhÞexx þ cosðhÞsinðhÞeyy þ ½cos2 ðhÞ þ sin2 ðhÞ
2 2
ezV zV ¼ ezz
Equation (43) relates the components of strain in two different coordinate
systems within a single plane and will be used extensively throughout the re-
mainder of this text.
The first three of Eq. (43) can be written using matrix notation as:
8 9
> e xV xV > 2 3
>
< >
= cos2 ðhÞ sin2 ðhÞ 2 cosðhÞsinðhÞ
ey V y V 6 7
¼4 sin2 ðhÞ cos2 ðhÞ 2 cosðhÞsinðhÞ 5
>
> xy ;
c >
>
: V V
cosðhÞsinðhÞ cosðhÞ sinðhÞ cos2 ðhÞ sin2 ðhÞ
2 8 9
>
>
exx >
>
< =
eyy
ð44Þ
>
> >
: cxy >
;
2
Compare Eq. (44) with Eq. (20). In particular, note that the trans-
formation matrix, [T], which was previously encountered during the discus-
sion of plane stress in Sec. 8, also appears in Eq. (44).
The strain invariants [given by Eqs. (37a)–(37c) or Eqs. (38a)–(38c)] are
considerably simplified when the out-of-plane z-axis is a principal axis. Since
c
by definition exz ¼ c2xz ¼ eyz ¼ 2yz ¼ 0, the strain invariants become:
First strain invariant ¼ He ¼ exx þ eyy þ ezz

c2xy
Second strain invariant ¼ Ue ¼ exx eyy þ exx ezz þ eyy ezz ð45Þ
4
c4xy
Third strain invariant ¼ We ¼ exx eyy ezz ezz
4
The principal strains equal the roots of the cubic equation previously
listed as Eq. (40). Substituting Eq. (45) into Eq. (40) reduces to:
c2xy
e3pn ðexx þ eyy þ ezz Þe2pn þ ðexx eyy þ exx ezz þ eyy ezz Þepn
4
c2xy
ðexx eyy ezz ezz Þ¼0 ð46Þ
4

One root of Eq. (23) is epn=ezz. For present purposes, this root will be
labeled ep3 although it may not be the algebraically least principal strain. In
the case of plane strain, ep3=ezz=0.
Removing the known root from Eq. (46), we have the following qua-
dratic equation:
c2xy
e2pn ðexx þ eyy Þepn þ ðexx eyy Þ¼0
4
The two roots of this quadratic equation (that is, the two remaining prin-
cipal strains, ep1 and ep2) may be found by application of the standard ap-
proach (3) and are given by:
ffi
exx þ eyy e e 2 c 2
xx yy xy
ep1 ; ep2 ¼ F þ ð47Þ
2 2 2
The angle hpe between the x-axis and either the p1 or p2 axis is given by:

1 cxy
hpe ¼ arctan ð48Þ
2 exx eyy
Example Problem 9
Given. A state of plane strain is known to consist of:
exx ¼ 500 Am=m
eyy ¼ 1000 Am=m
cxy ¼ 2500 Arad
Problem. (a) Prepare a rough sketch (not to scale) of the deformed strain
element in the x–y coordinate system; (b) determine the strain components
which correspond to an xV–yV coordinate system, oriented 25j CCW from the
x–y coordinate system, and prepare a rough sketch (not to scale) of the
deformed strain element in the xV–yV coordinate system; and (c) determine
the principal strain components that exist within the x–y plane, and prepare
a rough sketch (not to scale) of the deformed strain element in the principal
strain coordinate system.
Solution
Part (a). A sketch showing the deformed strain element (not to scale) in the
x–y coordinate system is shown in Fig. 19(a). Note that:
The length of the element side parallel to the x-axis has increased
(corresponding to the tensile strain exx=500 Am/m).
The length of the element side parallel to the y-axis has decreased (cor-
responding to the compressive strain eyy=1000 Am/m).

Figure 19 Strain elements associated with Example Problem 9 (all deforma-
tions shown greatly exaggerated for clarity).

The angle defined by x–y axes has increased (corresponding to the
negative shear strain cxy=2500 Arad).
Part (b). Since the xV-axis is oriented 25j CCW from the x-axis, in
accordance with the right-hand rule, the angle of rotation is positive, i.e.,
h=+25j. Substituting this angle and the given strain components in Eq. (44):
8 9 2 3
>
>exV xV >
> cos2 ð25B Þ sin2 ð25B Þ 2 cosð25B Þsinð25B Þ
<e = 6 7
yV yV
¼6
4 sin2 ð25B Þ cos2 ð25B Þ 2cosð25B Þsinð25B Þ 7 5
>
>c V V>
: xy> ; B B B B B B
2 cosð25 Þsinð25 Þ cosð25 Þsinð25 Þ cos ð25 Þ sin ð25 Þ
2 2
8 9
>
> 500 > >
>
< >
=
1000

>
> 2500 > >
>
: >
;
2

8 9 8 9
> exV xV > > 725 Am=m >
>
< >
= < > >
=
ey V y V
¼ 225 Am=m
>
> > > >
: c xV y V >; > : >
;
2 1378 Arad
A sketch showing the deformed strain element (not to scale) in the xV–yV
coordinate system is shown in Fig. 19(b). Note that:
The length of the element side parallel to the xV-axis has decreased
(corresponding to the compressive strain exVxV=725 Am/m).
The length of the element side parallel to the yV-axis has increased (cor-
responding to the tensile strain eyVyV=225 Am/m).
The angle defined by the xV–yV axes has increased (corresponding to the
negative shear strain cxVyV=2756 Arad).
Part (c). The principal strains are found through application of Eq. (47):
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
500 1000 500 þ 1000 2 2500 2
ep1 ; ep2 ¼ F þ
2 2 2
ep1 ¼ 1208 Am=m
ep2 ¼ 1708 Am=m
The orientation of the principal strain coordinate system is given by Eq. (48):

1 2500
hpe ¼ arctan ¼ 29:5B
2 500 þ 1000

A sketch showing the deformed strain element (not to scale) in the principal
strain coordinate system is shown in Fig. 19(c). Note that:
The length of the element side parallel to the p1-axis has increased
(corresponding to the tensile principal strain ep1=1208 Am/m).
The length of the element side parallel to the p2-axis has decreased
(corresponding to the compressive principal strain ep2=1708
Am/m).
The angle defined by the principal strain axes has remained precisely
p/2 radians (i.e., 90j) since in the principal strain coordinate system,
the shear strain is zero.
14 RELATING STRAINS TO DISPLACEMENT FIELDS

Most analyses considered in this text begin with the consideration of the
displacement fields induced in the structure of interest. That is, mathematical
expressions that describe the displacements induced at all points within a
structure by external loading and/or environmental changes will be assumed
or otherwise specified. Strains induced in the structure will then be inferred
from these displacement fields.
In the most general case, three displacement fields are involved. Spe-
cifically, these are the displacements in the x-, y-, and z-directions, typically
denoted as the u-, v-, and w-displacement fields, respectively. In general, all
three displacement fields are functions of x, y, and z:
Displacements in the x-direction: u=u(x,y,z).
Displacements in the y-direction: v=v(x,y,z).
Displacements in the z-direction: w=w(x,y,z).
However, if the out-of-plane z-axis is a principal strain axis, then u and v
are (at most) functions of x and y only, while w is (at most) a function of z
only. In this case:
Displacements in the x-direction: u=u(x,y).
Displacements in the y-direction: v=v(x,y).
Displacements in the z-direction: w=w(z).
A detailed derivation of the relationship between displacements and
strains is beyond the scope of this review, and the interested reader is re-
ferred to Frederick and Chang (1) or Fung (2) for details. It can be shown
that the relationship between displacement fields and the strain tensor de-
pends upon the magnitude of derivatives of displacement fields (also called
displacement gradients). If displacement gradients are arbitrarily large, then

the associated level of strain is said to be finite, and each component of the
strain tensor is related nonlinearly to displacement gradients as follows:
" #
Bu 1 Bu 2 Bv 2 Bw 2
exx ¼ þ þ þ
Bx 2 Bx Bx Bx
" #
Bv 1 Bu 2 Bv 2 Bw 2
eyy ¼ þ þ þ
By 2 By By By
" #
Bw 1 Bu 2 Bv 2 Bw 2
ezz ¼ þ þ þ
Bz 2 Bz Bz Bz

Bu Bv Bu Bu Bv Bv Bw Bw
cxy ¼ þ þ þ þ
By Bx Bx By Bx By Bx By

Bu Bw Bu Bu Bv Bv Bw Bw
cxz ¼ þ þ þ þ
Bz Bx Bx Bz Bx Bz Bx Bz

Bw Bv Bu Bu Bv Bv Bw Bw
cyz ¼ þ þ þ þ
By Bz By Bz By Bz By Bz
The expressions listed above define what is known as Green’s strain tensor
(also known as the Lagrangian strain tensor).
In most cases encountered in practice, however, displacement gradients
are very small, and consequently the products of displacement gradients
are negligibly small and can be discarded. For example, it can usually be
assumed that:
2 2 2
Bu Bv Bw Bu Bu
c0 c0 c0 c0; etc:
Bx Bx Bx Bx By
When displacement gradients are very small, the level of strain is said to be
infinitesimal, and each component of the strain tensor is linearly related to
displacement gradients as follows:
Bu
exx ¼ ð49aÞ
Bx
Bv
eyy ¼ ð49bÞ
By
Bw
ezz ¼ ð49cÞ
Bz
Bv Bu
cxy ¼ þ ð49dÞ
Bx By

Bw Bu
cxz ¼ þ ð49eÞ
Bx Bz
Bw Bv
cyz ¼ þ ð49f Þ
By Bz
For most analyses considered in this text, we will assume that strains
are infinitesimal and are related to displacement fields in accordance with
Eqs. (49a)–(49f). The one exception occurs in Chap. 11, where it will be nec-
essary to include nonlinear terms in the strain–displacement relationships.
As stated above, most analyses begin with the consideration of the
displacement fields induced in a structure of interest. Strain fields implied by
these displacements are then calculated in accordance with Eqs. (49a)–(49f).
This process insures that strain fields are consistent with displacements. Con-
sider the opposite approach. Specifically, suppose that mathematical expres-
sions for strain fields are assumed, perhaps on the basis of engineering
judgment. In this case, it is possible that the assumed strain fields correspond
to physically unrealistic displacement fields. For example, displacement fields
inferred from assumed strain fields may imply that the solid body has voids
and/or overlapping regions, a physically unrealistic circumstance. A system
of six equations known as the compatibility conditions can be developed that
guarantee that assumed expressions for the six components of strain do,
in fact, correspond to physically reasonable displacement fields u(x,y,z),
v(x,y,z), and w(x,y,z). To develop the compatibility conditions, differentiate
Eq. (49d) twice, once with respect to x and once with respect to y. We obtain:
B2 cxy B3 u B3 v
¼ þ 2
BxBy BxBy 2 Bx By
From Eqs. (49a) and (49b), it is easily seen that:
B2 exx B3 u B2 eyy B3 v
¼ ¼
By2 BxBy2 Bx2 Bx2 By
Combining these results, we see that expressions for the strain components
exx, eyy, and cxy correspond to physically reasonable displacement fields (i.e.,
‘‘are compatible’’) only if they satisfy:
B2 cxy B2 exx B2 eyy
¼ þ ð50aÞ
BxBy By2 Bx2
Equation (50a) is the first compatibility condition. Following a similar pro-
cedure using Eqs. (49e) and (49f), we obtain:
B2 cyz B2 eyy B2 ezz
¼ þ ð50bÞ
ByBz Bz2 By2

B2 cxz B2 exx B2 ezz
¼ þ ð50cÞ
BxBz Bz2 Bx2
These are the second and third compatibility conditions. Next, the following
expressions are obtained using Eqs. (49a), (49d), (49e), and (49f), respectively:
B3 u B2 exx
¼
BxByBz ByBz
B3 u B2 cxy B3 v
¼ 2
BxByBz BxBz Bx Bz
B3 u B2 cxz B3 w
¼ 2
BxByBz BxBy Bx By
B3 w B3 v B2 cyz
þ ¼
Bx2 By Bx2 Bz Bx2
Combining these four expressions, we find that assumed expressions for

strain components exx, cxy, cxz, and cyz, are compatible if:

B2 exx B Bcxy Bcxz cyz
2 ¼ þ ð50dÞ
ByBz Bx Bz By Bx
This is the fourth compatibility condition. The final two compatibility con-
ditions are developed using a similar process and are given by:

B2 eyy B Bcyz Bcxz cxy
2 ¼ þ ð50eÞ
BxBz By Bx By Bz

B2 ezz B Bcyz Bcxz cxy
2 ¼ þ ð50f Þ
BxBy Bz Bx By Bz
15 COMPUTER PROGRAMS 3DROTATE AND 2DROTATE

A review of the force, stress, and strain tensors has been presented in this
chapter. These concepts will be applied routinely throughout the remainder
of this text, as we develop a macromechanics-based analysis of structural
composite materials and structures. It will be seen that the transformation
of stress and strain tensors is of particular importance. Indeed, nearly all analy-
ses of composite materials and structures presented herein require multiple
transformations of stress and strain tensors from one coordinate system to
another.
Two computer programs, 3DROTATE and 2DROTATE, that can be
used to perform transformations of force, stress, or strain tensors have been
developed to accompany this text. These programs can also be downloaded

at no cost from the following website: http://debts.washington.edu/amtas/
computer.html. Program 3DROTATE performs the calculations necessary
to transform a force, stress, or strain tensor from the x–y–z coordinate
system to the xU–yU–zU coordinate system, where the xU–yU–zU coordinate
system is generated from the x–y–z coordinate system by (up to) three
successive rotations. Derivation of the direction cosines that relate these two
coordinate systems is left as a student exercise, but are listed in Homework
Problem 2. Program 3DROTATE also calculates the angles between the x–y–z
and xU–yU–zU coordinate axes, invariants of the force, stress, or strain tensors,
and principal stresses and strains. All of the numerical results discussed in
Example Problems 1, 3, and 7 can be obtained through the use of program
3DROTATE.
The second program, 2DROTATE, can be used to rotate stresses within
a plane (as discussed in Sec. 8) and/or strains within a plane (as discussed
in Sec. 13). For the most part, thin platelike composite structures will be
considered in this textbook. Therefore, it can usually be assumed that the
direction normal to the surface of the composite is a direction of principal
stress or strain. Hence, most of the stress or strain transformations consid-
ered in this text involve rotations within a plane. Most of the numerical results
discussed in Example Problems 5 and 9 can be obtained through the use of
program 2DROTATE.
HOMEWORK PROBLEMS
In the following problems, the phrase ‘‘solve by hand’’ means that numerical
solutions should be obtained using a pencil, paper, and nonprogrammable
calculator. Solutions obtained by hand will then be compared to numerical
results returned by appropriate computer programs. This process will insure
understanding of the mathematical processes involved.
1. Solve part (c) of Example Problem 1 by hand based on the rotation

angles listed below. In each case, calculate the magnitude of the
transformed force vector. Confirm your calculations using program
3DROTATE.
(a) h=60j b=45j.

(b) h=60j b=45j.
(c) h=60j b=45j.
(d) h=60j b=45j.
(e) h=45j b=60j.

(f ) h=45j b=60j.
(g) h=45j b=60j.
(h) h=45j b=60j.
2. Consider an xVVV–yVVV–zVVV coordinate system, which is generated from

an x–y–z coordinate system by the following three rotations:
A rotation of about the original z-axis, which defines an intermediate

xV–yV–zV coordinate system [see Fig. 2(a)], followed by
A rotation of b about the xV-axis, which defines an intermediate xVV–
yVV–zVV coordinate system [see Fig. 2(b)], followed by
A rotation of w about the yVV-axis, which defines the final xVVV–yVVV–zVVV
coordinate system.
Show that the xVVV–yVVV–zVVV and x–y–z coordinate systems are related by
the following direction cosines:
2 3
cxVVV x cxVVV y cxVVV z
4 cyVVV x cyVVV y cyVVV z 5
czVVV x czVVV y czVVV z
2 3
cos w cos h sin w sin b sin h cos w sinh þ sin w sin b cos h sin w cos b
¼4 cos b sin h cos b cos h sin b 5
sin w cos h þ cos w sin b sin h sin w sin h cos w sin b cos h cos w cos b
3. The force vector discussed in Example Problem 1 is given by:
F ¼ 1000iˆ þ 200jˆ þ 600k̂
Using Eq. (6c), express F in a new coordinate system defined by three

successive rotations, as listed below, using the direction cosines listed
in Problem 2. In each case, compare the magnitude of the transformed
force vector to the magnitudes calculated in Example Problem 1. Solve
these problems by hand and then confirm your calculations using
program 3DROTATE.
(a) h=60j b=45j w=25j.

(b) h=60j b=45j w=25j.
(c) h=60j b=45j w=25j.
(d) h=60j b=45j w=25j.
(e) h=60j b=45j w=25j.

4. Solve Example Problem 3 by hand using the following rotation
angles:
(a) h=20j b=35j.

(b) h=20j b=35j.
(c) h=20j b=35j.
Confirm your calculations using program 3DROTATE.

5. Use Eq. (12a) to obtain an expression (in expanded form) for the fol-
lowing stress component [in each case, the expanded expression will be
similar to Eq. (13)]:
(a) rxVxV.
(b) rxVyV.
(c) ryVyV.
(d) ryVzV.
(e) rzVzV.
6. Use program 3DROTATE to determine the stress invariants for the

stress tensor listed below, and compare to those determined in Example
Problem 3. (Note: this stress tensor is similar to the one considered in
Example Problem 3 except that the algebraic sign of all three normal
stresses has been reversed.):
2 3 2 3
4 ryx ryy ryz 5 ¼ 4 10 25 30 5 ðksiÞ
rzx rzy rzz 15 30 5

Example Problem 3 except that the algebraic sign of all three shear
stresses has been reversed.):
2 3 2 3
4 ryx ryy ryz 5 ¼ 4 10 25 30 5 ðksiÞ
rzx rzy rzz 15 30 5


Example Problem 3 except that the algebraic sign of all stress compo-
nents has been reversed.):
2 3 2 3
4 ryx ryy ryz 5 ¼ 4 10 25 30 5 ðksiÞ
rzx rzy rzz 15 30 5
9. Use program 3DROTATE to determine the strain invariants for the

strain tensor listed below, and compare to those determined in Example
Problem 7. (Note: this strain tensor is similar to the one considered in
Example Problem 7 except that the algebraic sign of all shear strain
components has been reversed.):
2 3 2 3
6 7 6 7
Example Problem 7 except that the algebraic sign of all normal strain
components has been reversed.):
2 3 2 3
6 7 6 7
Example Problem 7 except that the algebraic sign of all strain compo-
nents has been reversed.):
2 3 2 3
6 7 6 7

REFERENCES
1. Frederick, D.; Chang, T.S. Continuum Mechanics; Scientific Publishers, Inc:
Cambridge, MA, 1972.
2. Fung, Y.C. A First Course In Continuum Mechanics; Prentice-Hall, Inc.: Engle-
wood Cliffs, NJ, 1969.
3. Consult any handbook of mathematical functions and tables, for example, CRC
Basic Mathematical Tables; Shelby, S.M. The Chemical Company: Cleveland,
OH, 1970.

3
Material Properties
In this chapter, various material properties required to predict the perform-

ance of composite structures are introduced. The chapter begins with a
general discussion of isotropic vs. anisotropic material behaviors. It will be
pointed out that most composites can be classified as either orthotropic or
transversely isotropic materials. Sections devoted to those material properties
of primary interest to structural engineers then follow. Specifically, separate
sections are presented, which describe material properties that allow an
engineer to:
Relate stress to strain
Relate temperature to strain
Relate moisture content to strain
Relate stress (or strain) to failure.
In each section, material properties will first be defined for anisotropic

materials. These general definitions will then be applied to the case of
composites (i.e., they will be specialized for the case of orthotropic or
transversely isotropic materials).
1 ANISOTROPIC VS. ISOTROPIC MATERIALS

The phrase material property refers to a measurable constant that is character-
istic of a particular material and can be used to relate two disparate quantities
117

of interest. Material properties that describe the ability of a material to
conduct electricity, to transmit (or reflect) visible light, to transfer heat, or to
support mechanical loading, to name but a few, have been defined. Material
properties of interest herein are those used by engineers during the design of
composite structures. Two specific examples are Young’s modulus, E and
Poisson’s ratio v. These two familiar material properties, which will be
reviewed and further discussed in Sec. 2, are used to relate the stress and
strain tensors.
The adjectives ‘‘anisotropic’’ and ‘‘isotropic’’ indicate whether a mate-
rial exhibits a single value for a given material property. More specifically, if
the properties of a material are independent of direction within the material,
then the material is said to be isotropic. Conversely, if the material properties
vary with direction within the material, then the material is said to be
anisotropic.
To clarify this statement, suppose that three test specimens are ma-
chined from a large block at three different orientations, as shown in Fig. 1.
The geometry of the three specimens is assumed to be identical, so that the only
difference between specimens is the original orientation of each specimen
within the ‘‘parent’’ block. Now suppose that the axial stiffness (i.e., Young’s
modulus E ) is measured for each specimen. Young’s modulus measured using
specimen 1 will be denoted Exx (i.e., subscripts are used to indicate the original
orientation of specimen 1 within the parent block). Similarly, Young’s
modulus measured using specimens 2 and 3 will be denoted Eyy and Ezz,
respectively.
If the parent block consists of an isotropic material, then Young’s
modulus measured for each specimen will be identical (to within engineering
accuracies)—for isotropic materials: Exx = Eyy = Ezz.
In this case, an identical value of Young’s modulus is measured in the x-,
y-, and z-directions, and is independent of direction within the material. In
contrast, if the parent block is an anisotropic material, a different Young’s
modulus will, in general, be measured for each specimen—for anisotropic
materials: Exx p Eyy p Ezz.
In this case, the value of Young’s modulus depends on the direction
within the material the modulus is measured. For anisotropic materials, a
similar dependence on direction can occur for any material property of
interest (Poisson’s ratio, thermal expansion coefficients, ultimate strengths,
etc).
It is the microstructural features of a material that determine whether it
exhibits isotropic or anisotropic behavior. Consequently, to classify a given
material as isotropic or anisotropic, one must first define the physical scale of
interest. For example, it is well known that metals and metal alloys are made
up of individual grains, and that the atoms that exist within these grains are

Figure 1 Illustration of method used to determine whether a material is iso-
tropic or anisotropic.
arranged in well-defined crystalline arrays. The most common crystalline

arrays are the body-centered cubic (BCC), the face-centered cubic (FCC), or
the hexagonal close-packed (HCP) structure [1]. Due to the highly ordered
and symmetrical atomic structures that exist within these arrays, an individual
grain exhibits different properties in different directions, and hence is aniso-
tropic. That is to say, if material properties are defined at a physical scale on

the order of a grain diameter or smaller, then all metals or metal alloys must be
defined as anisotropic materials.
It does not necessarily follow, however, that a metal or metal alloy will
exhibit anisotropic behavior at the structural level. This is because individual
grains are typically very small, and the orientation of the atomic crystalline
arrays usually varies randomly from one grain to the next. As a typical case, it
is not uncommon for a steel alloy to exhibit an average grain diameter of 0.044
mm (0.0017 in.), or roughly 8200 grains/mm3 (134 106 grains/in.3) [1]. If the
grains are randomly oriented, which is a common case, then at the structural
level (say, at a physical scale >1 mm), the steel alloy will exhibit isotropic
properties even though the constituent grains are anisotropic. Conversely, if a
significant percentage of grains is caused to be oriented by some mechanism
(such as cold rolling, for example), then the same steel alloy will be anisotropic
at the structural level.
Polymeric composites are anisotropic at the structural level, and the
microstructural features that lead to this anisotropy are immediately ap-
parent. Specifically, it is the uniform and symmetrical orientation of the
reinforcing fibers within a ply that leads to anisotropic behavior. As a simple
example, suppose two specimens are machined from a thin unidirectional
composite plate consisting of high-strength fibers embedded within a rela-
tively flexible polymeric matrix, as shown in Fig. 2a and b. Note that the
coordinate system used to describe the plate has been labeled the 1–2–3 axes.
In this case, the 1-axis is defined to be parallel to the fibers, the 2-axis is defined
to lie within the plane of the plate and is perpendicular to the fibers, and the 3-
axis is defined to be normal to the plane of the plate. Note that fibers are
arranged symmetrically about the 1–3 and 2–3 planes. The 1–2–3 coordinate
system will henceforth be referred to as the principal material coordinate
system. Referring to Fig. 2a, specimen 1 is machined such that the fibers are
aligned with the long axis of the specimen, whereas in specimen 2, the fibers
are perpendicular to the axis of the specimen. Obviously, Young’s modulus
measured for these two specimens will be quite different. Specifically, the
modulus measured for specimen 1 will approach that of the fibers, whereas the
modulus measured for specimen 2 will approach that of the polymeric matrix.
Therefore, E11> >E22, and Young’s modulus varies with direction within the
material, satisfying the definition of an anisotropic material.
The principal material coordinate system is not always aligned with the
fiber direction, as shown in Fig. 2c and d. In this case, the thin composite plate
is formed using a braided fabric. As discussed in Sec. 4 of Chap. 1, braided
fabrics contain fibers oriented in two (or more) nonorthogonal directions. The
three principal material coordinate axes lie within planes that are symmetrical
with respect to the fiber array. As before, the 1- and 2-axes lie within the plane
of the plate, and the 3-axis is defined normal to the plane of the plate.

Figure 2 Illustration of the principal material coordinate system for thin com-
posite laminates.
One of the most unusual features of anisotropic materials is that they

can exhibit coupling between normal stresses and shear strains, as well as
coupling between shear stress and normal strains. A physical explanation of
how this coupling occurs in the case of a unidirectional composite is presented
in Fig. 3. A specimen in which the unidirectional fibers are oriented at an angle
of 45j with respect to the x-axis is shown, and a small square element from the
gage region is isolated. Because the element is initially square and, in this
example, the fibers are defined to be at an angle of 45j, fibers are parallel to
diagonal AC of the element. In contrast, fibers are perpendicular to diagonal
BD. This implies that the element is stiffer along diagonal AC than along
diagonal BD.

Figure 3 A 45j off-axis composite specimen used to explain the origin of cou-
pling effects.
Now assume that a tensile stress is applied, causing the square element
(as well as the specimen as a whole) to deform. Because the stiffness is higher
along diagonal AC than along diagonal BD, the length of diagonal AC is
increased to a lesser extent than that of diagonal BD. Hence, the initially
square element deforms into a parallelogram, as shown in the figure. Note
that:
The length of the square element is increased in the x-direction
(corresponding to a tensile strain, exx).
The length of the square element is decreased in the y-direction
(corresponding to a compressive strain eyy, and associated with the
Poisson effect).
BDAB is no longer p/2 rad, which indicates that a shear strain cxy has
been induced.

Hence, in this example, the normal stress rxx has induced two normal strains
(exx and eyy) as well as a shear strain cxy. Hence, coupling exists between rxx
and cxy, as stipulated.
The couplings between normal stresses and shear strains (as well as
couplings between shear stresses and normal strains) will be explained in a
formal mathematical sense in Chap. 4. However, two important conclusions
can be drawn from the physical explanation shown in Fig. 3. First, note that
the specimen shown is subjected to normal stress rxx only. In particular, note
that no shear stress exists in the x–y coordinate system (sxy = 0). Con-
sequently, the in-plane principal stresses applied to the specimen are rp1 = rxx
and rp2 = ryy = 0, and the principal stress coordinate system is defined by the
x–y coordinate system. However, a shear strain does exist in the x–y
coordinate system (cxy p 0). Consequently, the x–y coordinate system is
not the principal strain coordinate system. We conclude, therefore, that the
principal stress coordinate system is not aligned with the principal strain
coordinate system. This is generally true for all anisotropic materials and is in
direct contrast to the behavior of isotropic materials because for isotropic
materials, the principal stress and principal strain coordinate systems are
always coincident.
Secondly, note that the physical argument used above to explain the
origin of the coupling effect hinges on the fact that the fiber direction differs
from the direction of the applied stress rxx. Specifically, the fibers are oriented
45j away from the direction of the applied stress rxx. If the fibers were aligned
with either the x- or y-axis, then a coupling between rxx and cxy would not
occur. We conclude that the unusual coupling effects exhibited by composites
only occur if stress and strain are referenced to a nonprincipal material
coordinate system.
Anisotropic materials are classified according to the number of planes of
symmetry defined by the microstructure. The principal material coordinate
axes lie within the planes of symmetry. For example, in the case of unidirec-
tional composites, three planes of symmetry can be defined: the 1–2 plane, the
1–3 plane, and the 2–3 plane. Composites fall within one of two classifications
of anisotropic behavior. Specifically, composites are either orthotropic mate-
rials or transversely isotropic materials (the distinction between orthotropic
and transversely isotropic materials will be further discussed in Sec. 2).
During the composite structural analyses discussed in this text, the composite
will be called ‘‘anisotropic’’ if the coordinate system of reference is a non-
principal material coordinate system. Use of the term ‘‘anisotropic’’ will
therefore signal the possibility of couplings between normal stresses and shear
strains, and couplings between shear stresses and normal strains. If, instead, a
structural analysis is referenced to the principal material coordinate system,
the composite will be called either orthotropic or transversely isotropic.

Although many kinds of material properties may be defined, in this text
we are only interested in those properties commonly used by structural
engineers. The properties needed to perform a structural analysis of compo-
site structures will be defined in the following sections. In each section, a
general definition of the material property will be given, suitable for use with
anisotropic materials. That is, the properties of a composite material when
referenced to a nonprincipal material coordinate system will be discussed first.
These general definitions will then be specialized to the principal material
coordinate system (i.e., they will be specialized for the case of orthotropic or
transversely isotropic composites).
Typical values of the properties discussed in this chapter measured at
room temperatures are listed for glass/epoxy, Kevlar/epoxy, and graphite/
epoxy in Table 3. These properties do not represent the properties of any
specific commercial composite material system, but rather should be viewed
as typical values. Due to ongoing research and development activities within
the industry, the properties of composites are improved more or less
continuously. Therefore, the properties listed in Table 3 may not reflect
those of currently available materials. The properties that appear in the table
will be used in example and homework problems throughout the remainder
of this text.
2 MATERIAL PROPERTIES THAT RELATE STRESS

TO STRAIN
Both stress and strain are second-order tensors, as discussed in Chap. 2. The
material properties used to relate the stress and strain tensors are inferred
from experimental measurements. Conceptually, two different experimental
approaches may be taken. In the first approach, the material of interest is
subjected to a well-defined stress tensor, and components of the resulting
strain tensor are measured. In the second approach, the material of interest is
subjected to a well-defined strain tensor, and the components of the resulting
stress tensor are measured. From an experimental standpoint, it is far easier to
impose a well-defined stress tensor than a well-defined strain tensor, and
hence the first approach is almost always used in practice.
Recall that there are two fundamental types of stress components:
normal stress and shear stress. As a consequence, two fundamental types of
tests are used to relate stress to strain—specifically, a test that involves the
application of a known normal stress component and a test involving
application of a known shear stress component. In either case, a stress tensor
is imposed in which five of the six stress components equal zero, and the
resulting six components of the strain tensor are measured.
Tests that involve application of a known normal stress component are
called uniaxial tests. In a typical case, a single normal stress component is

applied to a test specimen (say, rxx), while insuring that the remaining five
stress components are zero (ryy = rzz = sxy = sxz = syz = 0). The six com-
ponents of strain caused by rxx are measured, which allows the calculation of
various material properties relating a normal stress component to the strain
tensor.
In contrast, tests that involve application of a known shear stress
component are called pure shear tests. In a typical case, a single shear stress
component is applied to a test specimen (say, sxy), while insuring that the
remaining five stress components are zero (rxx = ryy = rzz = sxz = syz = 0).
The six components of strain caused by sxy are measured, which allows the
calculation of various material properties relating a shear stress component to
the strain tensor.
A detailed description of experimental methods used to impose a
specified state of stress is beyond the scope of the present discussion. It
should be mentioned, however, that some of the stress states discussed below
are very difficult to achieve in practice. For example, because composites are
usually produced in the form of thin platelike structures, it is very difficult to
impose well-defined stresses acting normal to the plane of the composite, or to
measure the strain components induced normal to the plane of the composite
by a given state of stress.
For present purposes, these very real practical difficulties will be
ignored. It will simply be assumed that the stress tensors discussed have been
induced in the test specimen, and that methods to measure the resulting strain
components involved are available. A number of international and industrial
organizations publish annual test standards that describe available exper-
imental arrangements in detail. Some of the best known standards are those
published by the American Society of Testing and Materials (ASTM). ASTM
tests standards that describe techniques to measure composite material
properties relevant to the present discussion are listed in Table 1. There are
also many composite test methods used routinely in industrial, governmental,
or university composite laboratories that have not as yet been standardized by
organizations such as the ASTM. New test methods are being developed
continuously, and the reader should be alert for new methods and test
standards as they become available.
2.1 Uniaxial Tests

Referring to Fig. 1, suppose a uniaxial test is conducted using specimen 1 (i.e.,
material properties are measured in the x-direction). As the test proceeds,
stress rxx is increased from zero to some maximal level, and the components
of strain induced as a result of this stress are measured. An idealized plot of
strain data collected during a uniaxial test of an anisotropic material is shown
in Fig. 4, where it is has been assumed that the magnitude of stress is relatively

Table 1 ASTM Test Standards for Determining Elastic Moduli of
Polymeric Composites and Related Standards
Designation Title
D3039 Standard Test Method for Tensile Properties of Polymer

Matrix Composite Materials
D5450 Standard Test Method for Transverse Tensile Properties
of Hoop Wound Polymer Matrix Composite Cylinders
D695 Standard Test Method for Compressive Properties of
Rigid Plastics
Polymer Matrix Composite Materials with Unsupported
Gage Section by Shear Loading
Unidirectional Polymer Matrix Composites Using
a Sandwich Beam
D5449 Standard Test Method for Transverse Compressive
Properties of Hoop Wound Polymer Matrix
Composite Cylinders
D3518 Standard Practice for In-Plane Shear Response
of Polymer Matrix Composite Materials by Tensile
Test of a +45j Laminate
D5379 Standard Test Method for Shear Properties of
Composite Materials by the V-Notched Beam Method
D4255 Standard Guide for Testing In-Plane Shear Properties
of Composite Laminates
D5448 Standard Test Method for In-Plane Shear Properties
of Hoop Wound Polymer Matrix Composite Cylinders
Related standards
D5687 Standard Guide for Preparation of Flat Composite Panels
with Processing Guidelines for Specimen Preparation
D638 Standard Test Method for Tensile Properties of Plastics
D882 Standard Test Method for Tensile Properties of Thin
Plastic Sheeting
D4018 Standard Test Methods for Properties of Continuous
Filament Carbon and Graphite Fiber Tows
D2343 Standard Test Method for Tensile Properties of Glass
Fiber Strands, Yarns, and Rovings Used in
Reinforced Plastics

Figure 4 Idealized plot of the six strain components caused by the application
a uniaxial stress rxx.
low, such that a linear relationship exists between the stress component rxx
and the resulting strains. Strains induced at high nonlinear stress levels,
including failure stresses, will be considered in Sec. 5. Note that for an
anisotropic material, stress rxx will induce all six components of strain: exx,
eyy, ezz, cxy, cxz, and cyz. This is not the case for isotropic materials; a uniaxial
stress rxx applied to an isotropic material will not induce any shear strains
(cxy = cxz = cyz = 0); furthermore, the transverse normal strains will be
identical (eyy = ezz). Hence, for anisotropic material, there is an unusual
coupling between normal stress and shear strain, which would not be expected
based on previous experience with isotropic materials.
As would be expected, as the magnitude of rxx is increased, the mag-
nitude of all resulting strain components is also increased. Because stress rxx
causes six distinct components of strain for an anisotropic material, six ma-
terial properties must be defined in order to relate rxx to the resulting strains.
Let us first consider material properties relating normal stress rxx to
normal strains exx, eyy, and ezz. The relationship between rxx and normal
strain exx is characterized by Young’s modulus Exx (also called the ‘‘modulus
of elasticity’’):
rxx
Exx u e ð1Þ
xx
Young’s modulus is simply the slope of the rxx vs. exx curve shown in
Fig. 4. In words, Young’s modulus is defined as ‘‘the normal stress rxx divided

by the resulting normal strain exx, with all other stress components equal to
zero.’’ Subscripts xx have been used to indicate the direction in which
Young’s modulus has been measured. Because we have restricted our
attention to the linear region of the stress–strain curve, Eq. (1) is only valid
at relatively low, linear stress levels.
The relationship between the two transverse strains (eyy and ezz) and exx
is defined by Poisson’s ratio:
eyy ezz
vxy u e vxz u e ð2Þ
xx xx
In words, Poisson’s ratio vxy (or vxz) is defined as ‘‘the negative of the trans-
verse normal strain eyy (or ezz) divided by the axial normal strain exx, both of
which are induced by stress rxx, with all other stresses equal to zero.’’
As before, subscripts have been used to indicate the uniaxial stress
condition under which Poisson’s ratio is measured. The first subscript
indicates the direction of stress, and the second subscript indicates the
direction of transverse strain. For example, in the case of vxy, the first
subscript x indicates that a uniaxial stress rxx has been applied, and the
second subscript y indicates that transverse normal strain eyy has been used to
calculate Poisson’s ratio.
Combining Eqs. (1) and (2), a relationship between rxx and transverse
strains eyy and ezz is obtained:
vxy vxz
eyy ¼ rxx ezz ¼ rxx ð3Þ
Exx Exx
Now consider material properties relating normal strain exx to shear
strains cxy, cxz, and cyz. Material properties relating normal strains to shear
strains were discussed by Lekhnitski [2] and are called ‘‘coefficients of mutual
influence of the second kind.’’ In this text, they will be denoted using the
symbol g, and are defined as follows:
cxy c cyz
gxx;xy u e gxx;xz u e xz gxx;yz u e ð4Þ
xx xx xx
In words, the coefficient of mutual influence of the second kind gxx,xy (or
gxx,xz, or gxx,yz) is defined as ‘‘the shear strain cxy (or cxz, or cyz) divided by the
normal strain exx, both of which are induced by normal stress rxx, when all
other stresses equal zero.’’
Subscripts have once again been used to indicate the stress condition
under which the coefficient of mutual influence of the second kind is
measured. The first set of subscripts indicates the direction of stress, and
the second set of subscripts indicates the shear strain used to calculate the
coefficient. For example, in the case of gxx,xy, the first two subscripts xx

indicate that a normal stress rxx has been applied, and the second two
subscripts xy indicate that cxy has been used to calculate the coefficient.
Combining Eqs. (1) and (4), a relationship between rxx and shear strain
cxy, cxz, or cyz is obtained:
gxx;xy gxx;xz gxx;yz

cxy ¼ e rxx cxz ¼ e rxx cyz ¼ e rxx ð5Þ
xx xx xx
Equations (1)–(5) define six properties measured in the x-direction,

using specimen 1. Referring again to Fig. 1, analogous results are obtained
when properties are measured in the y- and z-directions, using specimens 2
and 3:
Properties Measured Using Specimen 2 (syy Applied)

ryy 1
Eyy u e ðorÞ eyy ¼ ryy
yy Eyy
exx vyx
vyx u e ðorÞ exx ¼ ryy
yy Eyy
ezz vyz
vyz u e ðorÞ ezz ¼ ryy
yy Eyy ð6Þ
cxy gyy;xy
gyy;xy u e ðorÞ cxy ¼ e ryy
yy yy
cxz gyy;xz
gyy;xz u e ðorÞ cxz ¼ e ryy
yy yy
cyz gyy;yz
gyy;yz u e ðorÞ cyz ¼ e ryy
yy yy
Properties Measured Using Specimen 3 (szz Applied):

rzz 1
Ezz u e ðorÞ ezz ¼ rzz
zz Ezz
exx vzx
vzx u e ðorÞ exx ¼ rzz
zz Ezz
eyy vzy
vzy u e ðorÞ eyy ¼ rzz
zz Ezz ð7Þ
cxy gzz;xy
gzz;xy u e ðorÞ cxy ¼ e rzz
zz zz
cxz gzz;xz
gzz;xz u e ðorÞ cxz ¼ e rzz
zz zz
cyz gzz;yz
gzz;yz u e ðorÞ cyz ¼ e rzz
zz zz

2.2 Pure Shear Tests
If a pure shear stress (say, sxy) is applied to an anisotropic material, six
components of strain will be induced. An idealized plot of strain data
collected during a pure shear test of an anisotropic material is shown
schematically in Fig. 5, where it is assumed that magnitude of shear stress is
relatively low such that a linear relationship exists between the stress
component sxy and the resulting strains. Strains induced at high nonlinear
stress levels, including failure stresses, will be considered in Sec. 5. Once again,
the stress–strain response of an anisotropic material differs markedly from
that of an isotropic material. Specifically, for an anisotropic material, stress
sxy will induce all six components of strain: exx, eyy, ezz, cxy, cxz, and cyz. If an
isotropic material is subjected to a pure shear stress sxy, only one strain
component is induced (cxy); all other strain components are zero (exx =
eyy=ezz = cxz = cyz = 0). Hence, for anisotropic material, there is an unusual
coupling between shear stress and normal strain, as well as an unusual
coupling between shear stress in one plane (say, the x–y plane) and out-of-
plane shear strains (cxz and cyz). Neither of these coupling effects occurs in
isotropic materials.
As would be expected, as the magnitude of sxy is increased during the
test, the magnitude of the resulting strains is also increased. Because stress sxy
causes six distinct components of strain, six material properties must be
defined in order to relate sxy to the resulting strains.
Figure 5 Idealized plot of the six strain components caused by the application
a pure shear stress sxy.

Let us first consider material properties relating shear stress sxy to shear
strains cxy,cxz, and cyz. The relationship between sxy and shear strain cxy is
characterized by the shear modulus Gxy:
sxy
Gxy u c ð8Þ
xy
In words, the shear modulus is defined as ‘‘the shear stress sxy divided by
the resulting shear strain cxy, with all other stress components equal to zero.’’
Because we have restricted our attention to linear stress levels, Eq. (8) is only
valid at relatively low, linear shear stress levels.
The relationship between transverse strains (cxz, cyz) and cxy is charac-
terized by Chentsov coefficients, which will be denoted using the symbol l in
this text:
c cyz
lxy;xz u cxz lxy;yz u c ð9Þ
xy xy
In words, the Chentsov coefficient lxy,xz (or lxy,yz) is defined as ‘‘the

shear strain cxz (or cyz) divided by the shear strain cxy, both of which are
induced by shear stress sxy, with all other stresses equal to zero.’’ The first set
of subscripts indicates the stress component, and the second set of subscripts
indicates the out-of-plane shear strain used to calculate the Chentsov co-
efficient. For example, in the case of lxy,xz the subscripts xy indicate that a
pure shear stress sxy has been applied, and the second two subscripts xz
indicate that cxz has been used to calculate the coefficient.
A comparison between Eqs. (2) and (9) reveals that Chentsov coeffi-
cients are directly analogous to Poisson’s ratio. Poisson’s ratio is defined as a
ratio of normal strains caused by a normal stress, whereas Chentsov coef-
ficients are defined as a ratio of shear strains caused by a shear stress.
Combining Eqs. (8) and (9), a relationship between sxy and shear strain
cxz or cyz is obtained:
lxy;xz lxy;yz
cxz ¼ sxy cyz ¼ sxy ð10Þ
Gxy Gxy
Finally, consider material properties relating shear stress sxy to normal
strains exx, eyy, and ezz. Material properties relating shear stress to normal
strains were discussed by Lekhnitski [2] and are called ‘‘coefficients of mutual
influence of the first kind.’’ In this text, they will be denoted using the symbol
g, and are defined as follows:
exx eyy ezz
gxy;xx u c gxy;yy u c gxy;zz u c ð11Þ
xy xy xy
In words, the coefficient of mutual influence of the first kind gxy,xx (or
gxy,yy, or gxy,zz) is defined as ‘‘the normal strain exx (or eyy, or ezz) divided by

the shear strain cxy, both of which are induced by shear stress sxy, when all
other stresses equal zero.’’ The first set of subscripts indicates the stress
component applied, and the second set indicates the normal strain used to
calculate the coefficient. For example, in the case of gxy,xx, the first subscripts
xy indicate that shear stress sxy has been applied, and the second set of
subscripts xx indicates that exx has been used to calculate the coefficient.
Combining Eqs. (8) and (11), a relationship between sxy and normal
strain exx, eyy, or ezz is obtained:
gxy;xx gxy;yy gxy;zz
exx ¼ sxy eyy ¼ sxy ezz ¼ sxy ð12Þ
Gxy Gxy Gxy
Equations (8)–(12) define six properties measured when a pure shear
stress sxy is applied. Analogous material properties are defined during tests in
which pure shear sxz or syz is applied.
Properties Measured Using Pure Shear t xz
sxz 1
Gxz u c ðorÞ cxz ¼ sxz
xz Gxz
cxy lxz;xy
lxz;xy u c ðorÞ cxy ¼ sxz
xz Gxz
cyz lxz;yz
lxz;yz u c ðorÞ cyz ¼ sxz
xz Gxz
ð13Þ
exx gxz;xx
gxz;xx u c ðorÞ exx ¼ sxz
xz Gxz
eyy gxz;yy
gxz;yy u c ðorÞ eyy ¼ sxz
xz Gxz
ezz gxz;zz
gxz;zz u c ðorÞ ezz ¼ sxz
xz Gxz
Properties Measured Using Pure Shear t yz
syz 1
Gyz u c ðorÞ cyz ¼ syz
yz Gyz
cxy lyz;xy
lyz;xy u c ðorÞ cxy ¼ syz
yz Gyz
cxz lyz;xz
lyz;xz u c ðorÞ cxz ¼ syz
yz Gyz
gyz;xx ð14Þ
exx
gyz;xx u c ðorÞ exx ¼ syz
yz Gyz
eyy gyz;yy
gyz;yy u c ðorÞ eyy ¼ syz
yz Gyz
ezz gyz;zz
gyz;zz u c ðorÞ ezz ¼ syz
yz Gyz

2.3 Specialization to Orthotropic and Transversely Isotropic
Composites
As previously shown in Fig. 2a and b, for unidirectional composites, the
principal material coordinate system is defined by the fiber direction. That is,
the 1-axis is defined parallel to the fiber direction, the 2-axis is perpendicular
to the fibers and lies within the plane of the composite, and the 3-axis is
perpendicular to the fibers and lies out of plane. In other cases, such as the
braided composite shown in Fig. 2c and d, the 1–2–3 principal material
coordinate system is not aligned with the fiber direction, but is instead defined
by planes of symmetry associated with the fiber architecture involved. For all
composite fabrics based on continuous fibers and typically encountered in
practice (i.e., unidirectional, woven, or braided fabrics), the principal material
coordinate system is readily identified.
We will now consider those properties that are measured when the
composite is referenced to the principal material coordinate system. It will be
seen later that properties of an anisotropic composite (i.e., a composite
referenced to a nonprincipal material coordinate system) can always be
related to those measured relative to the 1–2–3 coordinate system. To simplify
our discussion, we will assume that the composite under consideration is a
unidirectional composite, and hence that the 1–2–3 axes are parallel and
perpendicular to the fibers.
A stress element representing a unidirectional composite subjected to
uniaxial tensile stress r11 is shown in Fig. 6. The deformed shape of the
element is also shown. Note that:
The element has increased in length in the l-direction, corresponding
to a tensile strain e11.
The element has decreased in width in the 2- and 3-directions,
corresponding to compressive strains e22 and e33, respectively.
The deformed element is a rectangular parallelepiped. That is, due to
the symmetrical distribution of fibers with respect to the 1-, 2-, and 3-
coordinate axes, in the deformed condition, all angles remained p/2
rad (90j). Hence, all shear strains equal zero (c12=c13=c23=0).
Applying Eqs. (1), (2), and (4), we have:
r11
E11 u e ð15aÞ
11
e22
v12 u e ð15bÞ
11
e33
v13 u e ð15cÞ
11
g11;12 ¼ g11;13 ¼ g11;23 ¼ 0 ð15dÞ

Figure 6 Deformations induced in a unidirectional composite by uniaxial stress
r11 (deformations are shown greatly exaggerated for clarity).
Because no shear strains are induced by r11, the coefficients of mutual

influence of the second kind all equal zero. This is only true when the
composite is referenced to the principal material coordinate system. That is,
uniaxial stress acting in a nonprincipal coordinate system will cause a shear
strain, as previously shown in Fig. 3, for example. Therefore, the coefficients
of mutual influence of the second kind do not equal zero for anisotropic
composites (i.e., if the composite is referenced to a nonprincipal material
coordinate system). Methods of calculating composite material properties in
nonprincipal coordinate systems will be presented in Chap. 4.
Similarly, material properties measured when stress r22 is applied are:
r22
E22 u e ð16aÞ
22
e11
v21 u e ð16bÞ
22
e33
v23 u ð16cÞ
e22
g22;12 ¼ g22;13 ¼ g22;23 ¼ 0 ð16dÞ

Once again, due to the symmetrical distribution of fibers, stress r22 does
not induce any shear strains, so the coefficients of mutual influence of the
second kind all equal zero.
As previously mentioned, due to the thin platelike nature of composites,
it is difficult in practice to apply a well-defined out-of-plane uniaxial stress r33,
or to measure the resulting normal strain induced in the out-of-plane direction
e33. Assuming that these practical difficulties are overcome, the material
properties measured when stress r33 is applied are:
r33
E33 u e ð17aÞ
33
e11
v31 u e ð17bÞ
33
e22
v32 u e ð17cÞ
33
g33;12 ¼ g33;13 ¼ g33;23 ¼ 0 ð17dÞ
For unidirectional composites, both the 2- and 3-axes are defined to be
perpendicular to the fibers, and hence properties measured in the 2- and 3-
directions are typically similar in magnitude. In fact, if the distribution of
fibers in the 2- and 3-directions is identical at the microlevel, then properties
measured in these directions will be equal: E22 = E33, v12 = v13, v21 = v31, and
v23 = v32. If this occurs, then the composite is classified as a transversely
isotropic material. In contrast, if the distribution of fibers differs in the 2- and
3-directions, or if the composite under consideration is a woven or braided
composite, then properties measured in the 2- and 3-directions will not be
identical and the composite is classified as an orthotropic material.
Optical micrographs showing the fiber distribution in the 2–3 plane for a
unidirectional graphite–polyimide laminate are shown in Fig. 7. Fig. 7a was
taken at a magnification of 150, and shows the fiber distribution in four
adjacent plies. The fiber angles are (from left to right) 0j, 45j, 90j, and 45j.
Fig. 7b was obtained for the same laminate but at higher magnification
(300), and shows fiber angles (from left to right) of 0j, 45j, and 90j. As
indicated, for this laminate, thin resin-rich zone exists between plies. The
thickness of the resin-rich zone varies from one laminate to the next, depend-
ing on the material system, stacking sequence, and processing conditions used
to produce the laminate. If the resin-rich zone is very thin (say, less than about
1/10 the ply thickness) and if the fibers are uniformly distributed within the
interior of each ply, the composite will respond as a transversely isotropic
material. If these conditions do not exist (if the thickness of the resin-rich zone
is an appreciable fraction of the ply thickness, or if the distribution of fibers in
the 2- and 3-directions differs substantially), then E33 will differ from E22, and
the composite will respond as an orthotropic material.

Figure 7 Optical micrographs of fibers within several plies of a [0/45/90/
45]2s graphite–polyimide (IM7/K3B) composite laminate. Note the resin-rich
zone between plies.
Let us now consider properties measured through the application of a

pure shear stress in the principal material coordinate system. A stress element
representing a unidirectional composite subjected to a pure shear stress s12 is
shown in Fig. 8. The deformed shape of the element is also shown. Note that:
The angle originally defined by the 1–2 axes has decreased, cor-
responding to a positive shear strain c12.
Figure 8 Deformations induced in a unidirectional composite by pure shear

stress s12.

Due to the symmetrical distribution of fibers with respect to the 1-, 2-,
and 3-coordinate axes, in the deformed condition, all remaining
angles remained p/2 rad (90j). Hence, the remaining two shear
strains equal zero (c13 = c23 = 0).
The length, width, and thickness of the element have not changed;
hence, all normal strain are zero (e11 = e22 = e33 = 0).
Applying Eqs. (8)–(10), we have:
s12
G12 u ð18aÞ
c12
l12;13 ¼ l12;23 ¼ 0 ð18bÞ
g12;11 ¼ g12;22 ¼ g12;33 ¼ 0 ð18cÞ
Because only c12 is induced by s12, the Chentsov coefficients as well as

the coefficients of mutual influence of the first kind are all equal to zero. This is
only true when the composite is referenced to the principal material coor-
dinate system. That is, a shear stress acting in a nonprincipal material
coordinate system will, in general, cause both normal strains and shear
strains. Therefore, neither the Chentsov coefficients nor the coefficients of
mutual influence of the first kind equal zero if the composite is referenced to a
nonprincipal material coordinate system. Methods of calculating composite
material properties in nonprincipal coordinate systems will be presented in
Chap. 4.
Once again, due to the thin platelike nature of composites, in practice, it
is difficult to apply well-defined out-of-plane shear stress s13 or s23, or to
measure the resulting shear strains induced in the out-of-plane direction c13 or
c23. Assuming that these practical difficulties were overcome, the material
properties measured when stress s13 is applied are:
s13
G13 u ð19aÞ
c13
l13;12 ¼ l13;23 ¼ 0 ð19bÞ
g13;11 ¼ g13;22 ¼ g13;33 ¼ 0 ð19cÞ
If the fibers are not uniformly distributed within the 2–3 plane, or if the
composite is based on woven or braided fabrics, then the composite will
behave as an orthotropic material and G12 p G13. If the composite is based on
a unidirectional fabric and fibers are uniformly distributed, then the compo-
site is transversely isotropic and G12 = G13.

Following an analogous process, material properties measured when s23
is applied are:
s23
G23 u ð20aÞ
c23
l23;12 ¼ l23;23 ¼ 0 ð20bÞ
g23;11 ¼ g23;22 ¼ g23;33 ¼ 0 ð20cÞ
A total of 12 material properties have been defined above for ortho-

tropic or transversely isotropic composites: three Young’s moduli (E11, E22,
and E33), six Poisson’s ratios (v12, v13, v21, v23, v31, and v32), and three shear
moduli ( G12, G13, and G23). However, it will be seen later that for orthotropic
composites, only nine of these 12 properties are independent, and for trans-
versely isotropic composites, only five of the 12 properties are independent.
Therefore, only nine material properties must be measured to fully character-
ize the elastic response of orthotropic composites; for transversely isotropic
composites, only five material properties must be measured.
The number of material properties required in most practical engineer-
ing applications of composite is reduced further still. For reasons that will be
explained later, it is usually appropriate to assume that a composite structure
is subjected to a state of plane stress. Ultimately, this means that we only
require material properties in one plane. Hence, whereas an orthotropic
composite possesses nine distinct elastic material properties (and a trans-
versely isotropic composite possesses five), in practice, only four of these
properties are ordinarily required: E11, E22, v12, and G12. Most of the ASTM
test standards listed in Table 1 describe techniques used to measure these
properties. Also, a brief summary of common experimental methods used to
measure in-plane properties is provided in Appendix B. Typical values for
several composite material systems are listed in Table 3.
As a final comment, an often overlooked fact is that the elastic proper-
ties of composites usually differ in tension and compression (in fact, this is true
for many materials, not just for composites). For example, for polymeric
composites, it is not uncommon for E22 measured in tension to differ by 10–
15% from that measured in compression. Materials that exhibit this behavior
are called ‘‘bimodulus materials.’’ Although it is possible to account for these
differences during a structural analysis (e.g., see Ref. 3), the bimodulus
phenomenon is a significant complication and will not be accounted for
herein. Throughout this text, it will be assumed that in-plane elastic properties
E11, E22, v12, and G12 are identical in tension and compression. The reader
should be aware that these differences usually exist, however. If in practice the
measured response of a composite structure differs from the predicted

behavior, the discrepancy may well be due to differences in elastic properties
in tension vs. compression.
3 MATERIAL PROPERTIES RELATING TEMPERATURE

TO STRAIN
If an unconstrained anisotropic composite is subjected to a uniform change
in temperature DT, six components of strain will be induced: exx T
, eTyy , ezz
T
,
T T T
cxy , cxz , and cyz. The superscript ‘‘T’’ has been used to indicate that these
strains are caused solely by a change in temperature. Note that three of these
strains are shear strains; for anisotropic materials, a change in temperature
will, in general, cause shear strains to develop. Strains induced solely by a
change in temperature are referred to as ‘‘free thermal strains’’ or simply
‘‘thermal strains.’’ Properties that relate strains to temperature change are
called coefficients of thermal expansion (CTEs).
As previously discussed, it is the microstructural features of a material
that determine whether it exhibits isotropic or anisotropic behavior. The
contention that a change in temperature will induce shear strains may seem
unusual (because isotropic materials do not exhibit such behavior), but can be
easily explained in the case of unidirectional composites. An initially square
unidirectional composite is shown in Fig. 9, where it has been assumed that
the fibers are oriented at an angle of 45j with respect to the x-axis. Because the
composite is initially square and, in this example, the fibers are defined to be at
an angle of 45j, fibers are parallel to diagonal AC and are perpendicular to
diagonal BD. Now, the coefficient of thermal expansion exhibited by high-
performance fibers is typically very low (or even slightly negative), whereas for
Figure 9 Deformations caused in a 45j unidirectional composite by a uniform

change in temperature DT (deformations are shown greatly exaggerated for
clarity).

most polymers, it is relatively high. For example, the coefficient of thermal
expansion of graphite fibers is about 1 Am/m jC, whereas for epoxies, it is on
the order of 30 Am/m jC. Therefore, assuming that the composite shown in
Fig. 9 consists of a graphite/epoxy system, then an increase in temperature
will cause a slight decrease in the length of diagonal AC, but will cause a
relatively large increase in the length of diagonal BD. Hence, the initially
square composite deforms into a parallelogram, as shown in the figure. The
fact that angle BDAB has increased reveals that a shear strain cxy (in this case,
a negative shear strain cxy) has been induced by the change in temperature DT.
Hence, there is a coupling between the change in temperature and shear
strains, as stipulated. Note that this physical explanation of the coupling
between a uniform change in temperature and shear strain indicates that this
coupling only occurs if strain is referenced to a nonprincipal material co-
ordinate system.
An idealized plot of the six strain components induced in an anisotropic
composite by a change in temperature is shown in Fig. 10. As would be
Figure 10 Idealized plot of the six strain components caused by a change in

temperature DT.

expected, as DT is increased, the magnitude of all strain components also
increases. For modest changes in temperature (say, DT<200 jC), the change
in temperature is linearly related to the resulting thermal strain components.
That is, the slopes of the six strain vs. DT curves shown in Fig. 10 are constant
at relatively low levels of DT. At high levels of DT, the slopes of the curves
typically increase. For polymeric composites, the temperature at which the
curves become nonlinear is related to the glass transition temperature Tg of
the polymeric matrix (the glass transition temperature of a polymer is
discussed in Sec. 2 of Chap. 1). Most composite structures are designed to
operate at temperatures below the Tg; hence, we will focus our attention on
the linear range shown in Fig. 10.
Because a change in temperature causes six strains to develop for
anisotropic composites, six coefficients of thermal expansion must be defined:
eTxx eTyy eTzz

axx u ayy u azz u
DT DT DT
ð21Þ
cTxy cT cTyz
axy u axz u xz ayz u
DT DT DT
Because we have limited our discussion to the linear range shown in Fig.
10, the coefficients of thermal expansion defined by Eq. (21) equal the slopes
of the corresponding strain vs. DT curves within the linear range. These
properties will henceforth be called linear coefficients of thermal expansion. In
SI units, they are usually reported in terms of Am/m jC or Arad/jC (for
normal or shear strains, respectively). In English units, they are usually
reported in terms of Ain./in. jF or Arad/jF. A CTE can be converted from
SI units to English units by multiplying by the factor 5/9. For example, a CTE
of 15 Am/m jC equals 8.3 Ain./in. jF.
Equation (21) can be easily rearranged and written in matrix form as
follows:
2 T 3 2 3
exx cTxy cTxz axx axy axz
6 7 6 7
6 T 7 6 7
6 cyx eTyy cTyz 7 ¼ ðDTÞ6 axy ayy ayz 7 ð22Þ
6 7 6 7
4 5 4 5
cTzx cTzy eTzz axz ayz azz
The reader should note that the strains caused by a change in temper-
ature can be transformed from one coordinate system to another, in exactly
the same way that mechanically induced strains are transformed. In partic-
ular, any of the strain transformation equations reviewed in Chap. 2 (e.g., Eq.
(36), Eq. (41), Eq. (43), Eq. (44), or Eq. (47)) can be used to transform
thermally induced strains from one coordinate system to another.

3.1 Specialization to Orthotropic and Transversely Isotropic
Composites
We will now apply these general definitions to the case of a unidirectional
composite referenced to the principal material coordinate system. It will be
seen later that CTEs of an anisotropic composite (i.e., a composite referenced
to a nonprincipal material coordinate system) can always be related to those
measured in the 1–2–3 coordinate system. Although unidirectional compo-
sites are used in the following discussion, a equivalent discussion applies to
woven or braided composites, when referenced to the principal material
coordinate system.
A unidirectional composite subjected to a uniform change in temper-
ature DT and referenced to the 1–2–3 coordinate system is shown in Fig. 11.
Because the fibers are distributed symmetrically with respect to the 1 - , 2-, and
3-coordinate axes, a change in temperature does not cause a shear strain to
develop. Hence, in the principal material coordinate system, Eq. (21)
becomes:
eT11 eT22 eT33
a11 ¼ a22 ¼ a33 ¼ a12 ¼ a13 ¼ a23 ¼ 0 ð23Þ
DT DT DT
As before, if the fibers are distributed uniformly throughout the 2–3
plane, then a22=a33 and the composite will be transversely isotropic. If the
fibers are not uniformly distributed, then a22 p a33, and the composite will be
orthotropic. Woven or braided composites are always orthotropic because
Figure 11 Deformations induced in a unidirectional composite by a change in

temperature DT (deformations are shown greatly exaggerated for clarity).

the distribution of fibers in the 2-direction differs substantially from that in
the 3-direction.
Experimental methods of measuring thermal expansion coefficients for
polymeric composites have not been standardized by organizations such as
the ASTM. In practice, the in-plane thermal expansion coefficients exhibited
by polymeric composites are most commonly determined through the use of
resistance foil strain gages. Strain gage manufacturers often provide recom-
mendations for measuring thermal expansion coefficients using strain gages
(e.g., see Ref. 4). Typical CTEs for several common polymeric composites are
included in Table 3.
4 MATERIAL PROPERTIES RELATING MOISTURE

CONTENT TO STRAIN
It is possible for water molecules to diffuse into (or out of ) the overall
molecular structure of polymeric materials. In other words, the moisture
content of polymeric-based materials, including composites, slowly varies as
the relative humidity of the surrounding atmosphere varies. The moisture
content of a polymer is usually expressed as a percentage by weight, and
typically ranges from f0% to as high as f5%.
From a structural point of view, the effects of a change in moisture
content are analogous to those caused by a change in temperature. For
example, a plot of strains as a function of moisture content would resemble
Fig. 10, except that DM would be plotted along the horizontal axis rather than
DT. Hence, if an unconstrained anisotropic composite is subjected to a
uniform change in moisture content DM, then six components of strain will
M M M M M M
be induced: exx , eyy , ezz , cxy , cxz , and cyz . The superscript ‘‘M’’ has been
used to indicate that these strains are caused solely by a change in moisture
content. Strains induced by a change in moisture content are sometimes
referred to as hygroscopic strains, and can be just as large as or larger than
those associated with a change in temperature.
In this text, it will be assumed that strain is linearly related to changes in
moisture content. Properties that relate strains to changes in moisture content
will be called linear coefficients of moisture expansion, abbreviated as
‘‘CMEs,’’ and will be denoted using the symbol b. Because a change in
moisture content causes six strains to develop for anisotropic composites, six
CMEs must be defined:
eM eM
yy eM
bxx u xx
byy u bzz u zz
DM DM DM
ð24Þ
cM
xy cM cM
yz
bxy u bxz u xz byz u
DM DM DM

The units of the CMEs are typically Am/m/%M or Arad/%M. Equa-
tion (24) can be easily rearranged and written in matrix form as follows:
2 3 2 3
eM
xx cM
xy cM
xz bxx bxy bxz
6 7 6 7
6 M 7 6 7
6 cyx eM cM 7 ¼ ðDMÞ6 bxy byy byz 7 ð25Þ
6 yy yz 7 6 7
4 5 4 5
cM
zx cM
zy
M
ezz bxz byz bzz
A comparison between Eqs. (24) and (21), or between Eqs. (25) and (23)
will reinforce the fact that strains induced by a change in moisture content are
analogous (in a mathematical sense) to those caused by a change in temper-
ature.
4.1 Specialization to Orthotropic and Transversely

Isotropic Composites
In the principal material coordinate system, there is no coupling between a
change in moisture content and shear strains. Hence, in the principal material
coordinate system, Eq. (24) becomes:
eM eM eM
b11 u 11
b22 u 22
b33 u 33
DM DM DM ð26Þ
b12 ¼ b13 ¼ b23 ¼ 0
If the composite is based on a unidirectional fabric and fibers are

distributed uniformly throughout the 2–3 plane, then b22 = b33 and the
composite will be transversely isotropic. If the fibers are not uniformly
distributed, or if the composite is based on woven or braided fabrics, then
b22 p b33 and the composite will be orthotropic. Recommended experimental
methods of measuring linear coefficients of moisture expansion are described
in the ASTM standard D5229 [5]. Typical CMEs for several common
polymeric composites are included in Table 3.
5 MATERIAL PROPERTIES RELATING STRESS (OR STRAIN)

TO FAILURE
Material properties that relate stress or strain to failure are measured during
either a uniaxial test or a pure shear test. These properties are referred to
collectively as material strengths. Before we begin our discussion of material
strengths for composite materials, let us briefly review the definitions of

material strengths customarily used by engineers to report the failure
response of conventional structural materials such as metals or metal alloys.
Three idealized plots of the axial strain exx measured during a uniaxial
test are shown in Fig. 12a–c. These figures are similar to Fig. 4, except that
now: (a) only one of the strain components caused by rxx has been plotted
(specifically, only exx has been plotted), and (b) data have been included at
high nonlinear stress levels, up to and including the stress level at which the
specimen fractures into two (or more) pieces. Materials that exhibit a stress–
strain response similar to Fig. 12a are called brittle materials. As indicated,
for a brittle material, the stress–strain curve is nearly linear at all stress levels,
up to and including the final fracture stress. In fact, a perfectly brittle material
exhibits no nonlinear behavior at all; stress is linearly related to strain at all
levels, up to final fracture. In contrast, Fig. 12b and c shows the stress–strain
behavior for ductile materials. The characteristic feature that distinguishes
the two material types is that a ductile material exhibits a far larger region of
nonlinear behavior prior to failure than does a brittle material. A second
distinction is that the maximum strain a ductile material can withstand prior
Figure 12 Idealized plots of the axial strain qxx caused by application of a tensile
uniaxial stress jxx. (a) Stress-strain plot for a brittle material. (b) Stress-strain plot
for a modestly ductile material. (c) Stress-strain plot for a highly ductile material.

Figure 12 Continued.

to final fracture is (in general) far higher than that of a brittle material. The
toughness of a material is defined as the area under the entire stress–strain
curve. Hence, brittle materials typically possess a low toughness when
compared with ductile materials.
Four common measures of material strengths (illustrated in Fig. 12c) in
terms of stress are the proportional limit rp the percent offset yield strength ry,
the ultimate strength ru, and the fracture strength rf. Although the axial strain
that corresponds to each of these stress levels can also be defined as a measure
of strength, typically only the percent offset yield strain ey and the fracture
strain ef are customarily used in this fashion. Many materials exhibit distinctly
different strengths in tension vs. compression. Hence, in this text, a super-
script ‘‘T’’ or ‘‘C’’ will be used to indicate whether a given strength has been
measured in tension or compression, respectively. For example, because Fig.
12 represents stress–strain curves measured in tension, the superscript asso-
ciated with all measures of strength that appear in these figures includes the
character ‘‘T.’’
The proportional limit rxxpT is defined as the maximum stress level at
which stress is linearly related to strain. However, for many materials, the
deviation from linearity is so gradual that it is difficult to precisely identify
the proportional limit based on experimental measurements. In these cases,
the stress level at which significant nonlinear behavior begins is defined based
on the ‘‘offset method.’’ A line parallel to the initial (linear) region of the
stress–stain curve is drawn on the stress–strain curve, but is offset along the
strain axis by some standard amount. The stress level at which this offset line
intersects the experimental stress–strain curve is defined as the percent offset
yield strength ryT xx . A strain offset of either 0.001 or 0.002 m/m (0.1% or 0.2%
strain, respectively) is most commonly used, and hence yield strengths
measured in this manner are reported as the 0.1% offset yield strength or
the 0.2% offset yield strength, respectively. The ultimate strength ruT xx is
defined as the maximum stress the material can withstand. The fracture stress
rfT
xx is defined as the stress that exists at final fracture.
Note from Fig. 12 that a given material may not exhibit all four types of
material strength. For example, for modestly ductile materials (Fig. 12b),
fracture usually occurs at the maximum stress level, and hence for this type of
material, there is no distinction between the ultimate strength and the fracture
strength. For brittle materials (Fig. 12a), there may be little or no nonlinear
behavior prior to fracture, and hence a yield strength is not defined and only
the fracture strength (or ultimate strength) would be reported.
All of the above definitions may be applied to the data collected during a
pure shear test as well. An idealized plot of the shear strain cxy measured
during a pure shear test of a modestly ductile material is shown in Fig. 13. This
figure is similar to Fig. 5, except that now: (a) only one of the strain

Figure 13 Idealized plot of the shear strain cxy caused by application of pure
fP fP
shear stress sxy, including the shear fracture stress and strains sxy and cxy , and
yP yP
the shear yield stress and strain sxy and cxy .
components (cxy) caused by sxy has been plotted, and (b) data have been
included at high nonlinear stress levels, including the shear stress level at
which the specimen fractures into two (or more) pieces.
Because anisotropic materials may exhibit different strengths in differ-
ent directions, strength must be measured in three orthogonal directions; the
x-, y-, and z-directions. It is possible for anisotropic materials, including many
composites, to exhibit brittle behavior in one direction (say, the x-direction)
but ductile behavior in other directions (the y-direction and/or z-direction).
5.1 Specialization to Unidirectional Composites

Strengths measured for composites are referenced to the principal material 1–
2–3 coordinate system. In the case of unidirectional composites, strengths are
measured parallel to the fibers (i.e., parallel to the 1-axis) and transverse to the
fibers (parallel to the 2- and 3-axes). As would be anticipated, the strength of a
composite in the 1-direction is determined primarily by the strength of the

fibers, whereas strengths in the 2- and 3-directions are determined primarily
by the strength of the matrix. Most fibers used in polymeric composites, such
as graphite or glass fibers, are nearly perfectly brittle materials. Therefore,
most unidirectional polymeric composites exhibit a brittle failure response in
the 1-direction. Prior to about 1980, most commercially available polymeric
composites were based on relatively brittle thermoset matrices such as
epoxies. Consequently, composite material systems available prior to about
1980 exhibited a brittle response in the 2- and 3-directions as well, and
exhibited a low overall toughness.
Substantial research efforts to develop tougher composite material
systems were conducted throughout the 1970s and 1980s, and new-generation
composite material systems with substantially increased toughness are now
commercially available. This increase in toughness was accomplished pri-
marily by replacing the brittle polymeric matrices with tougher (i.e., more
ductile) polymers or polymer blends. Two approaches were used. In the first
approach, the toughness of inherently brittle thermosets (such as epoxies) was
increased through the addition of a second ductile rubber phase. This resulted
in a class of toughened matrices called ‘‘rubber-toughened’’ epoxies. The
second approach was to replace the inherently brittle thermoset matrix with a
more ductile thermoplastic polymer.* Examples of this latter approach
include the use polyketone, polyamide, or polyimide polymers as matrix
materials. From a purely structural standpoint, the net result of these
developments is that most new composite material systems exhibit a ‘‘mod-
estly ductile’’ response to failure in the 2- and 3-directions, qualitatively
similar to Fig. 12b. Consequently, new-generation composites exhibit a
much-improved toughness relative to older systems.
Methods to characterize the strengths of composite for purposes of
structural design have not been widely standardized within the composites
industry. Therefore, in this text, composite strengths will be described
using terminology similar to that traditionally used with metals and metal
alloys.
A r11 vs. e11 curve measured for most unidirectional polymeric compo-
sites resembles Fig. 12a. That is, in the fiber direction, unidirectional compo-
sites exhibit nearly perfectly brittle behavior, although very modest nonlinear
behavior may begin to occur at stress levels approaching final fracture. Due to
the limited nonlinear response, it is not appropriate to define a yield stress in
the 1-direction. Throughout this text, it will be assumed that a uniaxial stress
in the fiber direction r11 causes failure due to fracture. That is, at failure, the
* The difference between a thermoset and a thermoplastic polymer is discussed in Sec. 2.4 of
Chap. 1.

specimen breaks into two or more pieces. The stress that causes fracture is
fT fC
denoted r11 or r11 , depending on whether r11 is tensile or compressive,
fT fC
respectively. Similarly, the strain at failure is denoted e11 or e22 . Because very
little nonlinear behavior occurs prior to failure, it is possible to relate the
fracture stress and strain using Young’s modulus. That is, for most compo-
sites, the measured fracture stress and measured fracture strain in the 1-
direction may be related (to within engineering accuracies) using Young’s
modulus:
11 i ðE11 Þðe11 Þ
rfT 11 i ðE11 Þðe11 Þ
fT
rfC fC
On the other hand, for most modern unidirectional polymeric compo-

sites, both the r22 vs. e22 and the r33 vs. e33 curves resemble Fig. 12b. That is,
the response measured in the 2- and 3-directions is usually ‘‘modestly
ductile.’’ In this case then, ‘‘failure’’ may be defined either on the basis of a
tensile yield stress/strain (i.e., on the basis of yielding), or a tensile fracture
stress/strain (i.e., on the basis of fracture).
In this text, the stress and strain values present at the onset of yielding
will be denoted (ryT yT yT yT yC yC yC yC
22 ; e22 ; r33 ; e33 Þ or ðr22 ; e22 ; r33 ; e33 Þ; where the superscript
‘‘T’’ or ‘‘C’’ has once again been used to indicate whether the yield stress/
strain is measured in tension or compression, respectively. Note that it is
possible to relate the tensile yield stress and strain (to within engineering
accuracy) using Young’s modulus:
ryT yT
22 i ðE22 Þðe22 Þ ryC yC
ryT yT
33 i ðE33 Þðe33 Þ ryC yC
fT fT
The stress and strain values present at fracture are denoted (r22 , e22 ,
fT
r33 ,e33 ) or ðr22 ; e22 ; r33 ; e33 Þ: Because most modern composites exhibit a
fT fC fC fC fC
modestly ductile response in the 2- and 3-directions, the fracture stress is less
than the value that would be calculated using a failure strain and Young’s
modulus:
22 < ðE22 Þðe22 Þ

rfT 22 < ðE22 Þðe22 Þ
fT
rfC fC
33 < ðE33 Þðe33 Þ

rfT 33 < ðE33 Þðe33 Þ
fT
rfC fC
Shear stress–strain curves (i.e., a plot of s12 vs. c12, s13 vs. c13, or s23 vs.
c23) for unidirectional composites typically exhibit a shape somewhere
between Fig. 12b and c. That is, the shear response is usually more ductile
than that measured for normal stress in the 2-direction, and in some cases may
be considered to be ‘‘highly ductile.’’ In the principal material coordinate

system (e.g., in the 1–2 coordinate system), the shear response is insensitive to
the algebraic sign of the shear stress. That is, in the 1–2 coordinate system, an
identical s12 vs. c12 curve will be measured regardless of whether s12 is positive
or negative.* As before, a ‘‘shear failure’’ may be defined on the basis of a
yield shear stress/strain or a shear fracture stress/strain, depending on
application. In this text, the shear stress and strain at yielding will be denoted
y y y y y y
s12 and c12 (or s13 and c13 , or s23 and c23), whereas the shear stress and strain at
f f f f f f
fracture will be denoted s12 and c12 (or s13 and c13 , or s23 and c23 ). As before,
the yield shear stress/strain may be related using the shear modulus, but the
shear fracture stress/strain may not:
sy12 iðG12 Þðcy12 Þ sy13 iðG13 Þðcy13 Þ sy23 iðG23 Þðcy23 Þ
sf12 < ðG12 Þðcf12 Þ sf13 < ðG13 Þðcf13 Þ sf23 < ðG23 Þðcf23 Þ
The preceding discussion is for orthotropic composites. Following the

discussion presented in Sec. 2, if the composite is transversely isotropic, then
the following strengths are related as indicated:
ryT yT
22 ¼ r33 ryC yC
22 ¼ r33
22 ¼ r33
rfT 22 ¼ r33
fT
rfC fC
sy12 ¼ sy13 sf12 ¼ sf13
Recommended methods of measuring in-plane composite strengths are

included in most of the ASTM test standards previously listed in Table 1.
Additional ASTM testing standards related specifically to failure of poly-
meric composites are listed in Table 2. Due to the thin platelike nature of
composites, it is difficult to apply well-defined out-of-plane stresses, or to
measure the resulting out-of-plane strains. Hence, in practice, the yield and
fracture stresses associated with the out-of-plane 3-direction are measured
infrequently. Typical yield and fracture strengths measured in the 1–2 plane
for three common polymeric composites at room temperatures are included in
Table 3. These properties do not represent the properties of any specific
commercial composite material system, but rather should be viewed as typical
values. In fact, due to ongoing research and development activities, the failure
* However, in a general nonprincipal material coordinate system, the shear strength is sensitive
to the algebraic sign of the shear stress. This important point will be further discussed in Secs. 5
and 6 of Chap. 5.

Table 2 ASTM Test Standards Related to Failure of Polymeric Composites
(see also Table 1)
Designation Title
D3479 Standard Test Method for Tension–Tension Fatigue of Polymer

Matrix Composite Materials
D5766 Standard Test Method for Open Hole Tensile Strength of
Polymer Matrix Composite Laminates
D2344 Standard Test Method for Apparent Interlaminar Shear Strength
of Parallel Fiber Composites by Short-Beam Method
D5528 Standard Test Method for Mode I Interlaminar Fracture
Toughness of Unidirectional Fiber-Reinforced Polymer Matrix
Composites
E1922 Standard Test Method for Translaminar Fracture Toughness of
Laminated Polymer Matrix Composite Materials
D2290 Standard Test Method for Apparent Tensile Strength of Ring or
Tubular Plastics and Reinforced Plastics by Split Disk Method
strengths of composites are improved more or less continuously. Therefore,

the properties shown in Table 3 may not reflect those of currently available
materials.
Note that the failure strengths listed in Table 3 represent values
measured at room temperatures. This will become an important factor in
later chapters. In particular, in Chap. 7, we will consider methods of
predicting failure of general composite laminates with multiple fiber angles.
As will be seen, failure predictions for a multiangle composite laminate
depend (in part) on the difference between the temperature at which the
laminate is consolidated (this temperature is called the ‘‘cure temperature’’
throughout this text) and the ‘‘service temperature’’ at which the failure
prediction is desired. For example, if a composite laminate is cured at a
temperature of 175jC and then cooled to room temperature (20jC), the
laminate has experienced a temperature change of 155jC during cooldown.
Failure predictions for the laminate are then based on the temperature
difference of 155jC, and on failure strengths measured at room temper-
ature. If, instead, a composite is cured at 175jC and then cooled to some
other service temperature, say, 100jC, then the laminate has experienced a
temperature change of 75jC and failure predictions for the laminate are
based on failure strengths measured at 100jC.
It is pertinent to point out that the matrix-dominated tensile strengths
exhibited by polymeric composites are often lower than the failure strength
of the polymeric matrix alone. For example, the tensile strength of a

Table 3 Nominal Material Properties for Common Unidirectional Composites
Property Glass/epoxy Kevlar/epoxy Graphite/epoxy
E11 55 GPa (8.0 Msi) 100 GPa (15 Msi) 170 GPa (25 Msi)
E22 16 GPa (2.3 Msi) 6 GPa (0.90 Msi) 10 GPa (1.5 Msi)
v12 0.28 0.33 0.30
G12 7.6 GPa (1.1 Msi) 2.1 GPa (0.30 Msi) 13 GPa (1.9 Msi)
fT
r11 1050 MPa (150 ksi) 1380 MPa (200 ksi) 1500 MPa (218 ksi)
fC
r11 690 MPa (100 ksi) 280 MPa (40 ksi) 1200 MPa (175 ksi)
yT
r22 45 MPa (5.8 ksi) 35 MPa (2.9 ksi) 50 MPa (7.25 ksi)
yC
r22 120 MPa (16 ksi) 105 MPa (15 ksi) 100 MPa (14.5 ksi)
fT
r22 55 MPa (7.0 ksi) 45 MPa (4.3 ksi) 70 MPa (10 ksi)
fC
r22 140 MPa (20 ksi) 140 Msi (20 ksi) 130 MPa (18.8 ksi)
y
s12 40 MPa (4.4 ksi) 40 MPa (4.0 ksi) 75 MPa (10.9 ksi)
f
s12 70 MPa (10 ksi) 60 MPa (9 ksi) 130 MPa (22 ksi)
a11 6.7 Am/m jC 3.6 Am/m jC 0.9 Am/m jC
(3.7 Ain./in. jF) (2.0 Ain./in. jF) (0.5 Ain./in. jF)
a22 25 Am/m jC 58 Am/m jC 27 Am/m jC
(14 Ain./in. jF) (32 Ain./in. jF) (15 Ain./in. jF)
b11 100 Am/m %M 175 Am/m %M 50 Am/m %M
(100 Ain./in. %M) (175 Ain./in. %M) (50 Ain./in. %M)
b22 1200 Am/m %M 1700 Am/m %M 1200 Am/m %M
(1200 Ain./in. %M) (1700 Ain./in. %M) (1200 Ain./in. %M)
nonreinforced bulk epoxy is commonly about 70 MPa (10 ksi), whereas from
Table 3, we see that graphite–epoxy typically possesses a matrix-dominated
tensile strength on the order of 50 MPa (7.25 ksi). Even more pronounced is
the reduction in tensile strain at fracture: for a nonreinforced bulk epoxy, the
tensile strain at fracture commonly ranges from about 1% to 5% (10,000–
50,000 Am/m), whereas for graphite–epoxy, the matrix-dominated tensile
fT
strain at fracture (e22 ) rarely exceeds about 0.7% (7000 Am/m). The relatively
low matrix-dominated strengths exhibited by polymeric composites can be
explained on the basis of micromechanics analyses [6–9]. Briefly, two factors
lead to low matrix-dominated tensile strengths. The first is thermal stresses
induced at the microlevel during cooldown from cure temperatures. Recall
that the thermal expansion coefficient of most high-performance fibers is very
low and, in fact, is often slightly negative. For example, thermal expansion
coefficients of glass, Kevlar, and graphite fibers are about 5, 2, and 0.5
Am/m jC, respectively (see Sec. 3 of Chap. 1). In contrast, the thermal ex-
pansion coefficient of polymers is quite high and usually exceeds 30 Am/m jC.
Consequently, during cooldown from cure temperatures, the matrix is re-
strained from thermal contraction by the fibers, leading to self-equilibrating

tensile stresses in the matrix and compressive stresses in the fibers. The
second factor is the stress-concentrating effect of the fibers. Because most
advanced composites are produced with a fiber volume fraction of about
0.65, the tensile stresses induced within the matrix surrounding a fiber are not
dictated strictly by the CTE mismatch between matrix and fiber but are also
influenced by the presence neighboring fibers. Together, these two factors
give rise to thermal stresses at the microlevel that are generally tensile in the
matrix and compressive in the fibers. The magnitude of the thermal stresses
induced in the fiber is very low relative to the fiber strength. However, the
magnitude of the tensile stresses induced in the matrix represents a sub-
stantial fraction of the tensile strength of the matrix alone. Numerical mi-
cromechanics analyses based on the finite element method have shown that
thermal matrix stresses can often be 50–60% of the bulk matrix tensile
strength [11–14]. Hence, these thermal stresses are responsible for the low
matrix-dominated tensile strengths and tensile strain at fractures exhibited by
composites. They also explain in a qualitative sense why the magnitudes of
matrix-dominated compressive strengths are invariably higher than matrix-
yC yT fC fT
dominated tensile strengths (i.e., r22 >r22 and r22 >r22 ).
6 PREDICTING ELASTIC COMPOSITE PROPERTIES BASED

ON CONSTITUENTS: THE RULE OF MIXTURES
Various material properties exhibited by composites at the structural level
have been described in preceding sections. These properties are usually
measured for a composite material of interest, using one or more of the
ASTM (or equivalent) test standards listed in Tables 1 and 2. However, in
practice, the need to predict composite material properties exhibited at the
structural level also arises. That is, in practice, there is a need to predict
composite properties at the structural level based on properties of the
individual constituent materials (i.e., the fiber and matrix). As a typical
example, suppose a new high-performance graphite fiber has recently been
developed, and properties of the fiber itself have been measured. Naturally,
the structural engineer is interested in determining whether this new fiber will
lead to improvements in composite material properties at the structural level.
The potential improvement in properties can, of course, be evaluated directly,
by embedding the new fiber in a polymeric matrix of interest and by
measuring the properties exhibited by the new composite material system.
However, creating and testing the new material system in this fashion is time-
consuming and expensive. A need to estimate the properties that will be
provided by the new fiber exists, so as to justify the time and money that will

be invested during the development of the composite material system based
on the new graphite fiber.
As described in Sec. 6 of Chap. 1, an analysis performed at a physical
scale corresponding to the fiber diameter is classified as a micromechanics
analysis. In the present instance, we wish to use a micromechanics-based
analysis to predict composite properties at the structural level. A simple
micromechanics model that can be used to make this prediction is called the
rule of mixtures and is developed as follows.
Consider the representative composite element shown in Fig. 14. As
indicated, the element consists of unidirectional fibers embedded within a
polymeric matrix. The principal material coordinate system, labeled the 1–
2–3 coordinate system, is defined by the fiber direction. It is assumed that
the fibers are evenly spaced, and that the matrix is perfectly bonded to the
fiber.
If a force F11 is applied to the element, as shown in Fig. 14a, the length of
the element is increased by an amount DL and the width of the element is
decreased by an amount DW. Force F11 is related to the average stress
imposed in the 1-direction by F11=r11A, where A is the cross-sectional area
of the element. Furthermore, the sum of forces present in the matrix and fibers
must equal the total applied force, which implies:
r11 A ¼ rf Af þ rm Am ð27Þ
where Af is the total cross-sectional area of the fibers presented within the
element and Am is the cross-sectional area of the matrix. The strain in the 1-
direction is associated with the change in length (DL), and is identical in fiber
and matrix because the fiber and matrix are assumed to be perfectly bonded.
That is:
DL
ef ¼ em ¼ e11 ¼
L
Stresses are assumed to be related to strains according to:
r11 ¼ e11 E11 ð28aÞ
rf ¼ ef Ef ¼ e11 Ef ð28bÞ
rm ¼ em Em ¼ e11 Em ð28cÞ
The expressions for stresses rf and rm are only approximate. In reality, a
triaxial state of stress is induced rather than a uniaxial stress state, as implied
by Eqs. (28b) and (28c), due to the mismatch in fiber and matrix properties as
well as the presence of adjacent fibers. Properly accounting for this (and
other) complicating factors requires a rigorous analysis that is beyond the
scope of the brief introduction presented here. Therefore, we will assume that

Figure 14 Representative composite element used to derive rule-of-mixtures
equations. (a) Composite element deformed by a load F11, acting parallel to the
fiber direction. (b) Composite element deformed by a load F22, acting perpen-
dicular to the fiber direction. (c) Composite element deformed by shear load F12.

fiber and matrix stresses are given by Eqs. (28b) and (28c) despite their
shortcomings. Substituting these expressions into Eq. (27) and rearranging,
we find:
Af Am
E11 ¼ Ef þ Em
A A
This expression allows us to predict E11 based on properties of the

constituents (Ef and Em) and the area fractions of fiber and matrix (Af/A)
and (Am/A). If no voids are present, then:
A ¼ Af þ Am
Usually, rule-of-mixtures expressions are written in terms of volume

fractions rather than area fractions. Volume fractions are given by:
Af Am A Af
Vf ¼ Vm ¼ ¼ ¼ ð1 Vf Þ
A A A
where Vf is the volume fraction of fibers and Vm = (1Vf) is the volume
fraction of matrix material. Consequently, the predicted value of E11 based on
the rule of mixtures approach is given by:
E11 ¼ Em þ Vf ðEf Em Þ ð29Þ
Polymeric composites used in practice are typically produced with a

fiber volume fraction Vf of about 0.65, although it can be lower (say, Vf =
0.30), depending on application and the manufacturing process used to
consolidate the composite. Equation (29) shows that if Ef Em (which is
usually the case), then to a first approximation, E11cVfEf. The value of E11
is dictated primarily by the fiber modulus Ef and fiber volume fraction Vf. E11
is therefore called a fiber-dominated property of the composite.
Now consider the Poisson effect exhibited by the composite element
shown in Fig. 14a. As per our normal definition, the average Poisson ratio is
defined as the negative of the transverse normal strain (e22) divided by axial
normal strain (e11), both of which are caused by r11:
e22
v12 ¼
e11
The transverse normal strain associated with the change in width of the
entire element (DW) is given by:
DW
e22 ¼
W

The change in width can also be written as the sum of the change in
width of the fibers present in the element DWf and the change in width of the
matrix present DWm. These are approximated as follows:
DWf ¼ WVf vf ef ¼ WVf vf e11
DWm ¼ WVm vm em ¼ WVm vm e11
where vf and vm are Poisson ratios of the fiber and matrix, respectively.
Hence, the transverse strain is given by:
DW
e22 ¼ ¼ ½Vf vf þ ðVm Þvm e11
W
Applying the definition of Poisson’s ratio for the composite as a whole,

we have:
e22
v12 ¼ ¼ V f vf þ V m vm
e11
Noting as before that Vm=(1Vf), the predicted value for Poisson’s

ratio v12 based on the rule of mixtures becomes:
v12 ¼ vm Vf ðvm vf Þ ð30Þ
Measurement of Poisson’s ratio of the matrix material vm is a straight-

forward matter. However, measuring Poisson’s ratio of the fiber vf is more
difficult due to the small fiber diameters involved. Experimentally measured
values of vf are often unavailable, even for fibers widely used in practice. The
data that are available imply that both vm and vf are algebraically positive,
and also that vm>vf. Hence, Eq. (30) implies that the composite Poisson ra-
tio v12 varies linearly with fiber volume fraction Vf, and that Poisson’s ratio
of the composite is less than that of the matrix (v12<vm) because usually
(vmvf)>0.
Assuming an identical fiber distribution in the 1–2 and 1–3 planes, then
an identical analysis can be conducted to predict Poisson’s ratio v13, which will
result in an identical expression: v13=v12.
Next, consider prediction of the transverse modulus E22 based on the
rule-of-mixtures approach. A composite element subjected to a force applied
perpendicular to the fibers, force F22, is shown in Fig. 14b. This force is related
to the average stress imposed in the 2-direction by F22 = r22A. We assume
that an identical and uniform stress r22 is induced in both the fiber and matrix.
Once again, this assumption is approximate at best; in reality, a triaxial state

of stress is induced in both the fiber and matrix. Based on this assumption, the
strains induced in the fiber and matrix perpendicular to the 1-axis are:
r22
ef ¼
Ef
r22
em ¼
Em
The transverse length represented by the fibers present in the element
equals VfW, whereas the transverse length represented by the matrix equals
VmW. Hence, the change in width W caused by the application of r22 is:
r22 r22
DW ¼ ðVf WÞ þ ðVm WÞ
Ef Em
The average transverse strain caused by r22 is:

DW Vf Vm
e22 ¼ ¼ r22 þ
W Ef Em
Young’s modulus E22 as predicted by the rule of mixtures therefore
becomes:
r22 1 Ef Em
E22 ¼ ¼ ¼
e22 E V f þ Ef Vm
Ef þ Em
Vf Vm m
As before, if no voids are present, then Vm=(1Vf), and we obtain:
Ef Em
E22 ¼ ð31Þ
Ef Vf ðEf Em Þ
For most polymeric composite material systems, Ef> >Em. Nevertheless,
Eq. (31) shows that E22 is dictated primarily by Em, and is only modestly
affected by the fiber modulus Ef. Indeed, even in the limit (i.e., as Ef !l), the
predicted value of E22 is only increased to:

Em
E22 Ef !l ¼
1 Vf
Because Vf is usually about 0.65, this result shows that E22 is still less
than three times the matrix modulus Em, even if the composite is produced
using a fiber whose stiffness is infinitely high (Ef !l). E22 is therefore called
a matrix-dominated property of the composite.
Assuming an identical fiber distribution in the 1–2 and 1–3 planes,
then an identical analysis can be conducted to predict Young’s modulus
in the 3-direction E33, resulting in an identical expression. Hence, E33=E22.

As before, E33 is dictated primarily by Em and is a matrix-dominated
property.
Now consider the shear modulus G12. An element subjected to a pure
shear force F12 is shown in Fig. 14c. This force is related to the average shear
stress according to F12=s12A. In a rule-of-mixture analysis, it is assumed that
an identical shear stress is induced in both the fibers and matrix regions. This
assumption is approximate at best. Nevertheless, on the basis of this
assumption, the shear strains induced in fiber and matrix are given by:
s12
cf ¼
Gf
s12
cm ¼
Gm
where Gf and Gm are the shear moduli of the fiber and matrix, respectively.
The total shear strain is given by:

s12 s12
c12 ¼ Vf cf þ Vm cm ¼ Vf þ Vm
Gf Gm
The shear modulus predicted by the rule of mixtures is then:

s12 1 Gf Gm
G12 ¼ ¼ ¼
c12 Gm Vf þ Gf Vm
Gf þ Gm
V f Vm
Assuming no voids are present, then Vm=(1Vf), and we obtain:

Gf Gm
G12 ¼ ð32Þ
Gf Vf ðGf Vm Þ
Comparing Eq. (32) with Eq. (31), it is seen that the shear modulus is
related to fiber and matrix properties in a manner similar to E22 and E33. The
value of G12 is dictated primarily by the shear modulus of the matrix Gm, and
is considered a matrix-dominated property. Assuming an identical fiber
distribution in the 1–2 and 1–3 planes, then G13=G12.
To summarize, the analysis presented above allows prediction of elastic
moduli E11, v12=v13, E22=E33, and G12=G13, based on knowledge of the
fiber modulus Ef, matrix modulus Em, and fiber volume fraction Vf. Although
not presented here, a rule-of-mixture approach can also be used to predict
thermal expansion coefficients a1 and a2, or moisture expansion coefficients b1
and b2 (e.g., see Chap. 3 of Ref. 11).
The analysis presented above is only one of several micromechanics-
based models that have been proposed. The rule of mixtures is certainly the
simplest approach, but unfortunately is often the least accurate. In general,

E11 and v12=v13 are reasonably well predicted by Eqs. (29) and (30), respec-
tively. However, matrix-dominated properties E22=E33 and G12=G13 are
generally underpredicted by Eqs. (31) and (32). The accuracy of these pre-
dictions is not high because many important factors have not been accounted
for. A partial listing of factors not accounted for include:
The more or less random distribution and spacing of fibers present in
a real composite
The triaxial state of stress induced in both matrix and fiber due to the
mismatch in fiber/matrix properties.
Differences in fiber distribution in the 1–2 and 1–3 planes
The adhesion (or lack thereof ) between fiber and matrix
Variations in fiber cross-sections from one fiber to the next
The presence of voids or other defects
The anisotropic nature of many high-performance fibers (e.g.,
Young’s modulus parallel and transverse to the long axis of the
fiber usually differs).
A rigorous closed-form analytical solution that accounts for all of these
factors (as well as others) is probably impossible to obtain. Consequently,
most advanced micromechanics analyses are performed numerically using
finite element methods. Because the primary objective of this book is to
investigate composite materials at the structural (i.e., macroscopic) level, only
the simple rule of mixtures is presented herein. The reader interested in
learning more about micromechanics analyses is referred to the many ex-
cellent texts that discuss this topic in greater detail, a few of which are listed
here as Refs. 6, 9–13.
HOMEWORK PROBLEMS
1. An orthotropic material is known to have the following elastic properties:
Exx ¼ 100 GPa Eyy ¼ 200 GPa Ezz ¼ 75 GPa

vxy ¼ 0:20 vxz ¼ 0:25 vyz ¼ 0:60
Gxy ¼ 60 GPa Gxz ¼ 75 GPa Gyz ¼ 50 GPa
gxx;xy ¼ 0:30 gxx;xz ¼ 0:25 gxx;yz ¼ 0:30
gyy;xy ¼ 0:60 gyy;xz ¼ 0:75 gyy;yz ¼ 0:20
gzz;xy ¼ 0:20 gzz;xz ¼ 0:05 gzz;yz ¼ 0:15
lxy;xz ¼ 0:10 lxy;yz ¼ 0:05 gxz;yz ¼ 0:10

(a) What strains are induced if a uniaxial tensile stress rxx=300 MPa is
applied?
(b) What strains are induced if a uniaxial tensile stress ryy=300 MPa is
applied?
(c) What strains are induced if a uniaxial tensile stress rzz=300 MPa is
applied?
(d) What strains are induced if a pure shear stress sxy=100 MPa is
applied?
(e) What strains are induced if a pure shear stress sxz=100 MPa is
applied?
(f ) What strains are induced if a pure shear stress syz=100 MPa is
applied?
2. An orthotropic material is known to have the following elastic properties:
E11 ¼ 100 GPa E22 ¼ 200 GPa E33 ¼ 75 GPa
v12 ¼ 0:20 v13 ¼ 0:25 v23 ¼ 0:60
G12 ¼ 60 GPa G13 ¼ 75 GPa G23 ¼ 50 GPa
(a) What strains are induced if a uniaxial tensile stress r11=300 MPa is
applied?
(b) What strains are induced if a uniaxial tensile stress r22=300 MPa is
applied?
(c) What strains are induced if a uniaxial tensile stress r33=300 MPa is
applied?
(d) What strains are induced if a pure shear stress s12=100 MPa is
applied?
(e) What strains are induced if a pure shear stress s13=100 MPa is
applied?
(f ) What strains are induced if a pure shear stress s23=100 MPa is
applied?
3. A tensile specimen is machined from an anisotropic material. The speci-

men is referenced to an x–y–z coordinate system, as shown in Fig. 15a.
The cross-section of the specimen is initially a ‘‘perfect’’ 55 mm square.
In addition, ‘‘perfect’’ 55 mm squares are drawn on the x–y and x–z
surfaces of the specimen, as shown. A uniaxial tensile stress rxx=700
MPa is then applied, causing the specimen to deform as shown in Fig.
15b–d. Determine the values of Exx, vxy, vxz, gxx,xy, gxx,xz, and gxx,yz that
correspond to these deformations.

Figure 15 Tensile specimen described in Problem 3 (deformations shown
greatly exaggerated for clarity). (a) Tensile specimen machined from an aniso-
tropic material. (b) Change in cross-section. (c) Change in dimensions on x–y
face. (d) Change in dimensions on x–z face.
Table 4 The [0]8 Specimen (Width=1.251 in.; Thickness=0.048 in.)
Load (1 bf) Axial strain (Ain./in.) Trans strain (Ain./in.)
0 0 0
260 192 61
630 454 146
1220 860 279
1910 1335 433
2600 1807 587
4100 2784 930

Table 5 The [90]16 Specimen (Width=
1.254 in.; Thickness=0.090 in.)
Load (1 bf) Axial strain (Ain./in.)
0 0
64 300
102 539
172 923
275 1489
385 2072
4. Load vs. strain data collected during two different composite tensile tests
are shown in Tables 4 and 5. Use linear regression to determine the
following properties for this composite material:
(a) Determine E11 and v12 using the data collected using the [0]8
specimen (Table 4).
(b) Determine E22 using the data collected using the [90]16
specimen (Table 5).
(c) Determine the value of v21 for this composite material system.
Figure 16 Tensile specimen described in Problem 5.

Figure 17 Deformed square described in Problem 6.
5. A thin tensile specimen is machined from a material with unknown

properties. A ‘‘perfect’’ square with dimensions 5 5 mm is drawn on
one surface of the specimen, as shown in Fig. 16. A tensile stress of 500
MPa is then applied, causing the square to deform. Determine Exx, vxy,
and gxx,xy for this material.
6. A ‘‘perfect’’ square with dimensions 11 mm is drawn on the surface of a

plate. The temperature of the plate is then uniformly increased by 300jC,
causing the square to deform as shown in Fig. 17. Determine the
corresponding strains exx, eyy, and exy, and coefficients of thermal
expansion axx, ayy, and axy.
REFERENCES
1. Dieter, G.E. Mechanical Metallurgy; New York, NY: McGraw-Hill Inc., 1986;
ISBN0-07-016893-8.
2. Lekhnitski, S.G. Theory of Elasticity of an Anisotropic Body; Holden-Day: San
Francisco, 1963.
3. Bert, C.W.; Reddy, J.N.; Reddy, V.S.; Chao, W.C. Analysis of thick rectangular
plates laminated of bimodulus composite materials. AIAA J. 1981, 19 (10), 1342–
1349.
4. Measurement of Thermal Expansion Coefficient, M-M Tech Note 513;
Measurement Group, Inc.: Raleigh, NC, USA (available at the Measure-
ment Group website at: http://www.measurementsgroup.com/guide/indexes/
tn_index.htm).
5. Standard Test Method for Moisture Absorption Properties and Equilibrium
Conditioning of Polymer Matrix Composite Materials, Test Standard 5229;

American Society for Testing and Materials: West Conshohocken, PA, USA
(may also be accessed via the ASTM website at: http://www.astm.org/).
6. Hyer, M.W. Stress Analysis of Fiber-Reinforced Composite Materials; McGraw-
Hill Book Co.: New York, NY; 1998; ISBN 0-07-016700-1.
7. Herakovich, C.T. Mechanics of Fibrous Composites; John Wiley and Sons: New
York, NY, 1998; ISBN 0-471-10636-4.
8. Hull, D. An Introduction to Composite Materials; Cambridge University Press:
Cambridge, Great Britain, 1981; ISBN 0-521-23991-5.
9. Gibson, R.F. Principles of Composite Material Mechanics; McGraw-Hill Inc.:
New York, NY, 1994; ISBN0-07-023451-5.

4
Elastic Response of Anisotropic Materials
In this chapter, we will consider the strains induced in anisotropic materials

when subjected to arbitrary combinations of stress, uniform changes in
temperature, and/or uniform changes in moisture content. The chapter begins
with a consideration of the strains induced by stress under constant environ-
mental conditions. A ‘‘generalized’’ form of Hooke’s law, which relates strain
to stress for any anisotropic material, will be developed. Next, Hooke’s law
will be specialized for two particular types of anisotropy. First, for ortho-
tropic materials, and then for transversely isotropic materials.
Attention will then be focussed on strains caused by uniform changes in
temperature or moisture content. As before, relationships for anisotropic
materials will be developed first, and will then be specialized to the case of
orthotropic and transversely isotropic materials.
Finally, the strains induced by the combined effects of stress, temper-
ature, and moisture will be discussed.
1 STRAINS INDUCED BY STRESS: ANISOTROPIC

MATERIALS
A review of the stress and strain tensors has been provided in Chap. 2. Stress is
a symmetrical second-order tensor. In tensoral notation, stress is written as
167

rij, where subscripts i and j take on values of x, y, and z. Alternatively, the
stress tensor can be written using matrix notation as:
2 3
rxx rxy rxz
rij ¼ 4 ryx ryy ryz 5 ð1Þ
rzx rzy rzz
Because the stress tensor is symmetrical (i.e., ryx=rxy, rzx=rxz, and

rzy = ryz), only six independent stress components appear in Eq. (1).
Similarly, strain is a symmetrical second-order tensor eij and can be
written as:
2 3 2 3
exx exy exz ðexx Þ ðcxy =2Þ ðcxz =2Þ
6 7 6 ðcyz =2Þ 7
eij ¼ 4 eyx eyy eyz 5 ¼ 4 ðcyx =2Þ ðeyy Þ 5 ð2Þ
ezx ezy ezz ðc zx =2Þ ðc zy =2Þ ðe zz Þ
Only six independent strain components appear in Eq. (2). Also, the
tensoral shear strain components equal one-half the more commonly used
engineering shear strain components, that is,
cxy cyx c c cyz czy
exy ¼ eyx ¼ ¼ ; exz ¼ ezx ¼ xz ¼ zx ; eyz ¼ ezy ¼ ¼ ð3Þ
2 2 2 2 2 2
For any elastic solid, the strain and stress tensors are related as follows
(assuming temperature and moisture content remain constant):
eij ¼ Sijkl rkl ð4Þ
All subscripts that appear in Eq. (4) take on values of x, y, and z.

Equation (4) is called generalized Hooke’s law, and is valid for any elastic solid
under constant environmental conditions. It is seen that the strain and stress
tensors are related via the fourth-order compliance tensor Sijkl. Because strains
are unitless quantities, from Eq. (4), it is seen that the units of Sijkl are 1/
(stress) (i.e., either 1/Pa or 1/psi).
Because the compliance tensor is described using four subscripts and
because each subscript may take on three distinct value (e.g., x, y, or z), it
would initially appear that 34=81 independent terms appear within the
compliance tensor. However, due to symmetry of both the strain and stress
tensors, it will be shown below that the compliance tensor consists of (at most)
36 material constants.
It will be very convenient to express Eq. (4) using matrix notation.
However, because Sijkl is a fourth-order tensor (and hence can be viewed as
having ‘‘four dimensions’’), we cannot expand Sijkl as a two-dimensional

matrix directly. To expand Eq. (4), we must first define the components of
stress and strain using contracted notation, as follows:
exx ! e1 rxx ! r1
eyy ! e2 ryy ! r2
ezz ! e3 rzz ! r3
ð5Þ
cyz ¼ czy ! e4 ryz ¼ rzy ! r4
cxz ¼ czx ! e5 rxz ¼ rzx ! r5
cxy ¼ cyx ! e6 rxy ¼ ryx ! r6
Notice that the symmetry of the strain and stress tensors (cyz=cxz, etc.)
is embedded within the very definition of contracted notation. Also note that
the shear strain components (e4, e5, and e6) represent engineering shear strains,
rather than tensoral shear strains. Based on this change in notation, we can
now write Eq. (4) as:
ej ¼ Sij rj where i; j ¼ 16 ð6Þ
In contracted notation, the strain and stress tensors are expressed with a
single subscript (i.e., ei and rj), and hence in Eq. (6), they appear to be first-
order tensors. This is, of course, not the case. Both strain and stress are
second-order tensors. We are able to write them as using contracted notation
only because they are both symmetrical tensors. Similarly, contracted nota-
tion allows us to refer to individual components of the fourth-order com-
pliance tensor expressed using only two subscripts. We will henceforth refer to
Sij as the compliance matrix, and the use of contracted notation will be
implied.
Expanding Eq. (6), we have:
8 9 2 38 9
> e1 > S11 S12 S13 S14 S15 S16 > r1 >
>
> >
> >
> >
>
> e2 >> 6 S26 7>r > >
>
> >
> 6 S21 S22 S23 S24 S25 7>
> 2>>
<e >
> = 6 7>
>
<
>
>
6 S31 S32 S33 S34 S35 S36 7 r3 =
¼6 7
3
6S ð7Þ
>
> 4>
> e >
> 6 41 S42 S43 S44 S45 S46 7> r4 >
7> >
>
> > 6 > >
7> >
> e5 >
> > 4 S51
> S52 S53 S54 S55 > r5 >
S56 5>
> >
>
: >
> ; : >
> ;
e6 S61 S62 S63 S64 S65 S66 r6
In contracted notation, the compliance matrix has six rows and six
columns, so it is now clear that it consists of 36 independent material
constants (at most), as previously stated. Furthermore, through a consider-
ation of strain energy, it can be shown [1] that the compliance matrix must

itself be symmetrical. That is, all terms in symmetrical off-diagonal positions
must be equal:
S21 ¼ S12
S31 ¼ S13 S32 ¼ S23
S41 ¼ S14 S42 ¼ S24 S43 ¼ S34 ð8Þ
S51 ¼ S15 S52 ¼ S25 S53 ¼ S35 S54 ¼ S45
S61 ¼ S16 S62 ¼ S26 S63 ¼ S36 S64 ¼ S46 S65 ¼ S56
Hence, although the compliance matrix for an anisotropic composite

consists of 36 material constants, only 21 of these constants are independent.
Substituting the original strain and stress terms (defined in Eq. (5)) into
Eq. (7), we have:
8 9 2 38 9
> exx > S11 S12 S13 S14 S15 S16 > rxx >
>
> >
> 6 >
> >
>
>
> >
> 7>
> ryy >
>
>
> eyy >
> 6 S 21 S 22 S 23 S 24 S25 S 26 7>
> >
>
>
< e = 6S> 6 7> >
S32 S33 S34 S35 S36 7 rzz = <
zz 6
¼6
31
7 ð9Þ
> cyz > 7 r >
>
> >
> 6 S41 S42 S43 S44 S45 S46 7> >
> yz >
>
>
> > 6 7> >
>
> c > > 4 S51 S52 S53 S54 S55 S56 5> > >
>
> xz >
: >
;
>
> rxz >
: >
;
cxy S61 S62 S63 S64 S65 S66 rxy
The individual constants that appear in the compliance matrix can be
easily related to the material properties defined in Chap. 3 by invoking the
principal of superposition. That is, because we have restricted our attention to
linear elastic behavior, an individual component of strain caused by several
stress components acting simultaneously can be obtained by adding the strain
caused by each stress component acting independently. For example, the
strains exx caused by each stress component independently are given by:
1
rxx causes ðfrom Eq: 3:1Þ exx ¼ rxx
Exx
myx
ryy causes ðfrom Eq: 3:6Þ exx ¼ ryy
Eyy
mzx
rzz causes ðfrom Eq: 3:7Þ exx ¼ rzz
Ezz
gyz;xx
syz causes ðfrom Eq: 3:14Þ exx ¼ syz
Gyz
gxz;xx
sxz causes ðfrom Eq: 3:13Þ exx ¼ sxz
Gxz
gxy;xx
sxy causes ðfrom Eq: 3:12Þ exx ¼ sxy
Gxy

To determine the strain exx induced if all stress components act
simultaneously, simply add up the contribution to exx caused by each stress
individually, to obtain:

1 myx mzx gyz;xx
exx ¼ rxx þ ryy þ rzz þ syz
Exx Eyy Ezz Gyz
ð10aÞ
gxz;xx gxy;xx
þ sxz þ sxy
Gxz Gxy
Using an identical procedure, the remaining five strain components

caused by an arbitrary combination of stresses are:

mxy 1 mzy gxz;yy
eyy ¼ rxx þ ryy þ rzz þ syz
Exx Eyy Ezz Gyz
ð10bÞ
gxz;yy gxy;yy
þ sxz þ sxy
Gxz Gxy

mxz myz 1 gyz;zz
ezz ¼ rxx þ ryy þ rzz þ syz
Exx Eyy Ezz Gyz
ð10cÞ
gxz;zz gxy;zz
þ sxz þ sxy
Gxz Gxy

gxx;yz gyy;yz gzz;yz 1
cyz ¼ rxx þ ryy þ rzz þ syz
Exx Eyy Ezz Gyz
ð10dÞ
lxz;yz lxy;yz
þ sxz þ sxy
Gxz Gxy

gxx;xz gyy;xz gzz;xz lyz;xz
cxz ¼ rxx þ ryy þ rzz þ syz
Exx Eyy Ezz Gyz
ð10eÞ
1 lxy;xz
þ sxz þ sxy
Gxz Gxy

gxx;xy gyy;xy gzz;xy lyz;xy
cxy ¼ rxx þ ryy þ rzz þ syz
Exx Eyy Ezz Gyz
ð10f Þ
lxz;xy 1
þ sxz þ sxy
Gxz Gxy

Equation (10a)–(10f) can be assembled in matrix form:
8 9 2 38 9
> > 1 myx mzx gyz;xx gxz;xx gxy;xx > >
>
> > > xx >
> xx >
>
e > 6 Exx
> E E G G G 7>>
>
r >
>
>
>
> >
> 6 yy
zz
yz
xz
xy
7 >
> >
>
>
> >
> 6 m m 7 >
> >
>
>
> >
> 6 xy 1 zy g yz;yy g xz;yy g xy;yy 7 >
> >
>
>
> e >
yy > 6 7 >
> r yy >
>
>
> >
> 6 E E E G G G 7 >
> >
>
>
> >
> 6 xx
yy
zz
yz
xz
xy
7 >
> >
>
>
> > 6 mxz
> m 1 g g g 7 >
> >
>
>
> >
> 6 yz yz;zz xz;zz xy;zz 7 >
> >
>
< e zz = 6 E 7 < r zz =
6 xx Eyy Ezz Gyz Gxz Gxy 7
pc ¼ 6 7
>
> > 6 gxx;yz gyy;yz gzz;yz 1 lxz;yz lxy;yz 7> >
> cyz >
> > 6
> 7>>
> syz >>
>
>
> >
> 6 7 > >
>
> >
> 6 Exx Eyy Ezz Gyz Gxz Gxy 7> >
>
>
>
>
>
> >
> 6 7 >
> >
>
>
> >
> 6 g xx;xz g yy;xz g zz;xz l yz;xz 1 l xy;xz 7 >
> >
>
>
> c > 6
> 6 Exx 7 >
> s xz >
>
>
>
xz
>
> E E G G G 7 >
> >
>
>
> >
> 6 yy
zz
yz
xz
xy
7 >
> >
>
>
> >
> 4 g g g l l 1 5 >
> >
>
>
: cxy >; xx;xy yy;xy zz;xy yz;xy xz;xy >
: sxy >;
Exx Eyy Ezz Gyz Gxz Gxy
ð11Þ
By comparing Eqs. (9) and (11), it can be seen that the individual
components of the compliance matrix are directly related to the material
properties measured during uniaxial tests or pure shear tests:
1 1 1
S11 ¼ S22 ¼ S33 ¼
Exx Eyy Ezz
1 1 1
S44 ¼ S55 ¼ S66 ¼
Gyx Gxz Gxy
mxy myx mxz mzx
S21 ¼ S12 ¼ ¼ S31 ¼ S13 ¼ ¼
Exx Eyy Exx Ezz
myz mzy gxx;yz gyz;xx
S32 ¼ S23 ¼ ¼ S41 ¼ S14 ¼ ¼
Eyy Ezz Exx Gyz
gyy;yz gyz;yy gzz;yz gyz;zz
S42 ¼ S24 ¼ ¼ S43 ¼ S34 ¼ ¼
Eyy Gyz Ezz Gyz
gxx;xz gxz;xx gyy;xz gxz;yy ð12Þ
S51 ¼ S15 ¼ ¼ S52 ¼ S25 ¼ ¼
Exx Gxz Eyy Gxz
gzz;xz gxz;zz lyz;xz lxz;yz
S53 ¼ S35 ¼ ¼ S54 ¼ S45 ¼ ¼
Ezz Gxz Eyz Gxz
gxx;xy gxy;xx gyy;xy gxy;yy
S61 ¼ S16 ¼ ¼ S62 ¼ S26 ¼ ¼
Exx Gxy Eyy Gxy
gzz;xy gxy;zz lyz;xy lxy;yz
S63 ¼ S36 ¼ ¼ S64 ¼ S46 ¼ ¼
Ezz Gxy Eyz Gxy
lxz;xy lxy;xz
S65 ¼ S56 ¼ ¼
Gxz Gxy

Because the compliance matrix must be symmetric, Eq. (12) shows that
many of the properties of anisotropic materials are related through the
following inverse relationships:
mxz myx
¼
Exx Eyy
mxz mzx myz mzy
¼ ¼
Exx Ezz Eyy Ezz
gxx;yz gyz;xx gyy;yz gyz;yy gzz;yz gyz;zz
¼ ¼ ¼
Exx Gyz Eyy Gyz Ezz Gyz ð13Þ
gxx;xz gxz;xx gyy;yz gyz;yy gzz;xz gxz;zz lyz;xz lxz;yz
¼ ¼ ¼ ¼
Exx Gxz Eyy Gyz Ezz Gxz Gyz Gxz
gxx;xy gxy;xx gyy;xy gxy;yy gzz;xy gxy;zz lyz;xy lxy;yz
¼ ¼ ¼ ¼
Exx Gxy Eyy Gxy Ezz Gxy Gyz Gxy
lxz;xy lxy;xz
¼
Gxz Gxy
The inverse relationships are very significant from an experimental

point of view because they dramatically reduce the number of tests that must
be performed in order to determine the value of the many terms that appear
within the compliance matrix of an anisotropic composite. Specifically, if the
compliance matrix were not symmetrical, and hence if the inverse relation-
ships did not exist, then 36 tests would be required to measure all components
of the compliance matrix. The fact that the compliance matrix must be
symmetrical reduces the number of tests required to 21. Of course, this is still
a large number of tests. Fortunately, because the principal material coor-
dinate system of composites is readily apparent, the elastic properties of
composites are usually measured relative to the principal material coordinate
system rather than an arbitrary (nonprincipal) coordinate system. As dis-
cussed in Sec. 2, this further reduces the number of tests required. The 21
terms within the compliance matrix of an anisotropic composite can then be
calculated based on properties measured relative to the principal material
coordinate system.
Thus far, we have discussed Hooke’s law in the form of ‘‘strain–stress’’
relationships. That is, given values of the components of stress, we can
calculate the resulting strains using Eq. (9) or Eq. (11), for example. In
practice, we are often interested in the opposite problem. That is, a common
circumstance in practice is that the components of strain induced in a
structure have been measured, and we wish to calculate the stresses that
caused these strains. In this case, we need a ‘‘stress–strain’’ form of Hooke’s
law. The stress–strain form of Hooke’s law can be obtained by simply

inverting previous results. For example, inverting Hooke’s law given by Eq.
(6), we have:
ri ¼ Cij ej where i; j ¼ 16 ð14Þ
Cij is called the ‘‘stiffness matrix.’’* The stiffness matrix is the mathe-
matical inverse of the compliance matrix Cij=Sij1. In expanded form, Eq.
(14) is written as:
8 9 2 38 9
>
> r1 >>
C11 C12 C13 C14 C15 C16 > e1 >
> >
>
>r > > 6C 7>> >
>
>
> 2 >
>
> 6 21 C 22 C 23 C 24 C25 C 26 7 >
>
> e2 >>
>
>
>
<r = 6C> 6 7 >
< >
=
C C C C C 7 e
3
¼6 6 31 32 33 34 35 36
7 3
ð15Þ
> r4 > 7 e >
>
> >
> 6 C41 C42 C43 C44 C45 C46 7> >
> 4>>
>
> > 6 7> >
>
> r > > 4 C51 C52 C53 C54 C55 C56 5> > >
>
> 5>
: >
; > e5 >
>
: >
;
r6 C61 C62 C63 C64 C65 C66 e6
The stiffness matrix is symmetrical (C21=C12, C31=C13, etc.). The units
of each stiffness term are the same as stress (either Pa or psi).
2 STRAINS INDUCED BY STRESS: ORTHOTROPIC

AND TRANSVERSELY ISOTROPIC MATERIALS
As discussed in Sec. 2 of Chap. 3, many of the unusual couplings between
stress and strain exhibited by composites referenced to an arbitrary coordi-
nate system do not occur if the stress and strain tensors are referenced to the
principal material coordinate system. In this text, a material referenced to an
arbitrary (nonprincipal) coordinate system is called ‘‘anisotropic’’ whereas if
the same material is referenced to the principal material coordinate system, it
is called either an ‘‘orthotropic’’ or ‘‘transversely isotropic’’ material.
All of the following coupling terms are zero for orthotropic or trans-
versely isotropic materials:
Coefficients of mutual influence of the second kind:
g11;12 ¼ g11;13 ¼ g11;23 ¼ g22;12 ¼ g22;13 ¼ g22;23 ¼ g33;12
¼ g33;13 ¼ g33;23 ¼ 0
* The variable names assigned to the compliance and stiffness matrices in this chapter have
evolved over many years and are widely used within the structural mechanics community. The
reader should note that, unfortunately, the symbol ‘‘S’’ is customarily used to refer to the
compliance matrix, whereas the symbol ‘‘C’’ is customarily used to refer to the stiffness matrix.

Coefficients of mutual influence of the first kind:
g12;11 ¼ g12;22 ¼ g12;23 ¼ g13;11 ¼ g13;22 ¼ g13;33 ¼ g23;11
¼ g23;22 ¼ g23;33 ¼ 0
Chentsov coefficients:
l12;13 ¼ l12;23 ¼ l13;12 ¼ l13;23 ¼ l23;12 ¼ l23;13 ¼ 0
Because these coupling terms do not exist, Hooke’s law for orthotropic
or transversely isotropic materials is simplified considerably relative to that of
an anisotropic material. For an orthotropic material, Hooke’s law becomes
(compare with Eq. (11)):
8 9 2 38 9
> > 1 m21 m31 > >
>
> e11 >> > r11 >
>
7>
> > 0 0 0 > >
>
> >
> 6 > >
>
>
> >
> 6 E11 E22 E33 7>>
> >
>
>
> >
> 6 7>> >
>
>
> >
> 6 m21 1 m32 7>> >
>
> e22 >
> > 6 7>> r22 >
>
>
> >
> 6 E 0 0 0 7> >
>
>
> >
> 6 E22 E33 7>> >
>
7>
11
>
> >
> 6 >
> >
>
>
> >
> 6 7> >
>
>
> >
> 6 m13 m23 1 7>>
> >
>
>
> e >
> 6 0 0 0 7>> r >
>
< 33 = 6 E11 E22 E33 7< 33 =
6 7
> >
¼6 7
7> >
ð16Þ
> > 6 1 7> >
>
>
>
> c23 >> 6
>
> 6 0 0 0 0 0 7>> s23 >
>
>
>
>
>
>
> >
> 6 G23 7>> >
>
>
> >
> 6 7>> >
>
>
> >
> 6 7> >
>
> >
> 6 7>> >
> > s13 >
>
1
> > >
>
>
>
c 13 > 6
>
> 6
0 0 0 0
G13
0 7>
7>>
>
>
>
>
> >
> 6 7>> >
>
> > > >
>
> > 6
> 7
5>> >
>
>
> > 4
> 1 >
> >
>
: c12 >
> ; 0 0 0 0 0 > >
: s12 ;
G12
Alternatively, Eq. (16) may be written as:
8 9 2 38 9
> e11 > S11 S12 S13 0 0 0 > r11 >
>
> >
> > >
>
> >
> 6 7>>
>
>
>
>
>
> e >
> 6 S12 S22 S23 0 0 0 7 >
> r >
>
>
> 22 >
> 6 7>> 22 >
>
>
> >
> 6 7>> >
>
< e33 >
> = 6S 0 7 > >
6 13 S23 S33 0 0 7 r33 =
<
¼6
6
7
7 ð17Þ
>
> c23 >
> 6 0 0 0 S44 0 0 7> > s23 >
>
>
> >
> 6 7>> >
>
>
> >
> 6 7>> >
>
> c13 >
> > 6 0 > s13 >
> >
>
> >
> 4 0 0 0 S55 0 75>> >
>
>
> >
> >
> >
>
: ; : ;
c12 0 0 0 0 0 S66 s12

The fact that the compliance matrix must be symmetrical (S21=S12,
etc.) has been included in Eq. (17). Each compliance term in Eq. (17) is related
to the more familiar engineering properties as follows:
1 1 1
S11 ¼ S22 ¼ S33 ¼
E11 E22 E33
1 1 1
S44 ¼ S55 ¼ S66 ¼
G23 G13 G12
m12 m21
S21 ¼ S12 ¼ ¼ ð18Þ
E11 E22
m13 m31
S31 ¼ S13 ¼ ¼
E11 E33
m23 m32
S32 ¼ S23 ¼ ¼
E22 E33
It can be seen that only nine independent material constants exist for an
orthotropic material. The set of nine independent constants can be viewed as:
ðS11 ; S22 ; S33 ; S44 ; S55 ; S66 ; S12 ; S13 ; and S23 Þ
or, equivalently, as:
ðE11 ; E22 ; E33 ; m12 ; m13 ; m23 ; G12 ; G13 ; and G23 Þ
Equation (17) is the strain–stress form of Hooke’s law suitable for use
with orthotropic materials. To obtain a stress–strain relationship, Eq. (17) is
inverted, resulting in:
8 9 2 38 9
> r11 > C11 C12 C13 0 0 0 > e11 >
>
> >
> >
> >
>
> r22 >> 6 C12 C22 C23 0 7 > e22 > >
>
> >
> 6 0 0 7>
>
>
>
>
>
<r = 6C 7< =
33 6 13 C 23 C33 0 0 0 7 33e
¼6 7 ð19Þ
>
> s23 >> 6 0 0 0 C44 0 0 7>> c23 >>
>
>s > >
> 4 6 7>
> >
>
>
> 13 > 0 0 0 0 C55 0 5>> c > >
>
: >
; : 13 >
> ;
s12 0 0 0 0 0 C66 c 12
Individual components within the stiffness matrix for an orthotropic

material are related to the compliance terms as follows:
S22 S33 S223 S13 S23 S12 S33 S12 S33 S13 S22
C11 ¼ C12 ¼ C13 ¼
S S S
S11 S33 S213 S12 S13 S11 S23 S11 S22 S212
C22 ¼ C23 ¼ C33 ¼
S S S
1 1 1
C44 ¼ C55 ¼ C66 ¼
S44 S55 S66
ð20Þ

where:
S ¼ S11 S22 S33 S11 S223 S22 S213 S33 S212 þ 2S12 S13 S23
Alternatively, the stiffness terms may be calculated using the elastic
properties described in Sec. 2 of Chap. 3:
ðE22 m223 E33 ÞE211 ðm12 E22 þ m13 m23 E33 ÞE11 E22
C11 ¼ C12 ¼
X X
ðm12 m23 þ m13 ÞE11 E22 E33 ðE11 m13 E33 ÞE222
2
C13 ¼ C22 ¼
X X ð21Þ
ðm23 E11 þ m12 m13 E22 ÞE22 E33 ðE11 m212 E22 ÞE22 E33
C23 ¼ C33 ¼
X X
C44 ¼ G23 C55 ¼ G13 C66 ¼ G12
where:
X ¼ E11 E22 m212 E222 m213 E22 E33 m223 E11 E33 2m12 m13 m23 E22 E33
Hooke’s law for transversely isotropic materials is simplified further-
more because in this case, E22=E33, m12=m13, m21=m31, m23=m32, and
G12=G13. Also, it can also be easily shown that for transversely isotropic
E22
composites, G23 ¼ 2ð1þm . Hence, for transversely isotropic composites, Eq.
23 Þ
(16) reduces to:
8 9 2 38 9
> > 1 m21 m21 > >
>
> e >
> 0 0 0 >
> r >
>
>
>
11 >
> 6 E11 E22 E22 7>
>
11 >
>
> > 6 7> >
>
>
>
>
>
> 6 m 1 m 7>
> >
>
>
>
> >
> 6 7>
> >
>
0 7>
21 32
>
> e22 >
> 6 0 0 >
> r 22 >
>
>
> >
> 6 E11 E22 E22 7>
> >
>
>
> >
> 6 7>
> >
>
>
> >
> 6 m12 m 1 7>
> >
>
> > 6 23 7> r33 >
< e33 >
> = 6 E 0 0 0 7> < >
=
6 E22 E22 7
>
¼6
> 6
11
7
>
ð22Þ
2ð1 þ m23 Þ 7>
>
>
> c23 >
>
> 6 0 7 > s23 >
> >
>
>
> >
> 6
0 0 0
G22
0 7>
>
> >
>
>
> >
> 6 7>
> >
>
>
> >
> 6 >
7> >
>
> > 6
> 1 7> >
>
> c13 > 6 0 0 0 0 0 7> > s13 >>
>
> >
> 6 7>
> >
>
>
> >
> 6
G12
7>
> >
>
>
> >
> 4 5>
> >
>
>c > > 1 > >
>
: 12 ; 0 0 0 0 0 : s12 >
> ;
G12
which may also be written as:

8 9 2 38 9
> e11 > S11 S12 S12 0 0 0 > r11 >
>
> >
> >
> >
>
> >
e22 > 6 0 7 > r22 >>
>
> > 6 S12 S22 S23 0 0 7>> >
>
<e > = 6S >
0 7 r33 =
7 < >
33 6 12 S23 S22 0 0
¼6 7 ð23Þ
>
> c23 >
> 6 0 2ðS22 S23 Þ 0 7> s23 >
7> >
0 0 0
>
> >
> 6 >s >
> >
>
> c >
> 4 0 0 0 0 S66 0 5> > >
>
: 13 >
> ; >
:
13 >
;
c12 0 0 0 0 0 S66 s 12

It can be seen that only five independent material constants exist for a
transversely isotropic material. The set of five independent constants can be
viewed as:
ðS11 ; S22 ; S66 ; S12 ; and S23 Þ
or, equivalently, as:
ðE11 ; E22 ; m12 ; m23 ; and G12 Þ
Equation (23) is the strain–stress form of Hooke’s law suitable for use
with transversely isotropic composites. To obtain a stress–strain relationship,
Eq. (23) is inverted, resulting in:
8 9 2 38 9
>
> r11 >>
C11 C12 C12 0 0 0 > e11 >
> >
>
> >
> 6C 7>> >
>
>
> r >
> 6 C C 0 0 0 7 >
> e >
22 >
>
> 22 >
> 6
12 22 23
7 >
> >
>
>
< > 6
= 7 >
< >
=
r33 6 C 12 C 23 C 22 0 0 0 7 33 e
¼6 7
>
> s23 >> 6 0 0 2ðC22 C23 Þ 0 0 7 > c23 >
>
> >
> 6 0 7>>
>
>
>
>
>
> >
> 6 7 >
> >
>
>
> s 13 >
> 4 0 0 0 0 C 0 5 >
> c >
>
>
: >
;
66
>
:
13
>
;
s12 0 0 0 0 0 C66 c12
ð24Þ
Individual components within the stiffness matrix for a transversely
isotropic composite are related to the compliance terms, as follows:
S22 þ S33 S12 S11 S22 S212
C11 ¼ C12 ¼ C22 ¼
X X XðS22 S23 Þ
ð25Þ
S S11 S23
2
1
C23 ¼ 12 C66 ¼
XðS22 S23 Þ S66
where:
V ¼ S11 ðS22 þ S23 Þ 2S212
Alternatively, the stiffness terms may be calculated using the elastic
properties described in Sec. 2 of Chap. 3:
E211 ð1 m23 Þ m12 E11 E22 E22 ðE11 m212 E22 Þ
C11 ¼ C12 ¼ C22 ¼
X X Xð1 þ m23 Þ
E22 ðm23 E11 þ m12 E22 Þ
2
C22 C23 E22
C23 ¼ C44 ¼ ¼ C66 ¼ C12
Xð1 þ m23 Þ 2 2ð1 þ m23 Þ
ð26Þ
where:
X ¼ E11 ð1 m23 Þ 2m212 E22

Example Problem 1
The properties of a composite material are known to be:
E11 ¼ 170 GPa E22 ¼ 10 GPa E33 ¼ 8 GPa
m12 ¼ 0:30 m13 ¼ 0:35 m23 ¼ 0:40
G12 ¼ 13 GPa G13 ¼ 10 GPa G23 ¼ 8 GPa
Note that nine distinct material properties have been specified, indicat-
ing that this composite material is orthotropic. Determine the strains caused
by the following state of stress:
8 9 8 9
> r11 > > 350 MPa >
>
> >
> > >
> >
>
>
> r22 >> > 35 MPa > >
>
> >
> >
> >
>
< r = < 15 MPa >
> > > =
33
¼
>
> s23 >> > 30 MPa >
> >
>
> > >
> > >
>
>
> >
> >
> >
>
>
> s 13 >
> >
> 10 MPa >
>
: ; : ;
s12 25 MPa
Solution. Because the composite is orthotropic, strains are calculated using
Eq. (17), where each term within the compliance matrix is calculated using Eq.
(18):
1 1 5:88 1 1 100:0
S11 ¼ ¼ ¼ S22 ¼ ¼ ¼
E11 170 GPa 1012 Pa E22 10 GPa 1012 GPa
1 1 125:0 1 1 125:0
S33 ¼ ¼ ¼ 12 S44 ¼ ¼ ¼ 12
E33 8 GPa 10 Pa G23 8 GPa 10 Pa
1 1 100 1 1 76:9
S55 ¼ ¼ ¼ 12 S66 ¼ ¼ ¼ 12
G13 10 GPa 10 Pa G12 13 GPa 10 Pa
m12 0:030 1:76 m13 0:35 2:06
S21 ¼ S12 ¼ ¼ ¼ S31 ¼ S13 ¼ ¼ ¼
E11 170 GPa 1012 Pa E11 170 GPa 1012 Pa
m23 0:40 40:0
S32 ¼ S23 ¼ ¼ ¼ 12
E22 10 GPa 10 Pa
In this case, Eq. (17) becomes:

8 9 2 3
>
> e11 >> 5:88 1:76 2:06 0 0 0
>
> >
> 6 1:76 100:0
>
> e >
> 40:0 0 0 0 7
= 6 7
22
< 6 2:06 40:0
e33 125:0 0 0 0 7 1
¼6
6
7
> c23 >
> > 6 0 0 0 125:0 0 0 7 12
7 10 Pa
>
> >
> 4 0 5
>
> c13 >> 0 0 0 100:0 0
: ;
c12 0 0 0 0 0 76:9

8 9 8 9
>
> 350 >
> >
> 1966 Am=m >>
>
> 35 >> >
> >
>
> >
> >
> 2284 Am=m >>
>
< = < =
15 246 Am=m
ð10 PaÞ ¼
6
>
> 30 >
> >
> 3750 Arad >
>
>
> >
> >
> >
>
> 10 >
> >
> 1000 Arad >
>
>
: ; : ;
25 1923 Arad
Example Problem 2
The properties of a composite material are known to be:
E11 ¼ 25 Msi E22 ¼ E33 ¼ 1:5 Msi

m12 ¼ m13 ¼ 0:30 m23 ¼ 0:40
G12 ¼ G13 ¼ 2:0 Msi
Note that only five distinct material properties have been specified,
indicating that this composite material is transversely isotropic. Determine
the strains caused by the following state of stress:
8 9 8 9
>
> r11 >
> >
> 50 ksi >>
>
> > > >
>
> r22 >
>
>
>
> 5 ksi > >
< = > < >
=
r33 2 ksi
¼
> s23 >
> > > > 4 ksi > >
>
> > > >
>
> s13 >
>
>
>
> 1:5 ksi >
>
: ; > : >
;
s12 3:5 ksi
Solution. Because the composite is transversely isotropic, strains are calcu-

lated using Eq. (22). Individual terms within the compliance matrix are:
1 1 40:0 1 1 667
S11 ¼ ¼ ¼ S22 ¼ S33 ¼ ¼ ¼
E11 25 Msi 109 psi E22 1:5 Msi 109 psi
1 2ð1 þ m23 Þ 2ð1 þ 0:40Þ 1866
S44 ¼ ¼ ¼ ¼ 9
G23 E22 1:5 Msi 10 psi
1 1 500
S55 ¼ S66 ¼ ¼ ¼
G12 2:0 Msi 109 psi
m12 0:30 12:0
S12 ¼ S21 ¼ S13 ¼ S31 ¼ ¼ ¼ 9
E11 25 Msi 10 psi
m23 0:40 266
S23 ¼ S32 ¼ ¼ ¼
E22 1:5 Msi 109 psi

Equation (17) becomes:
8 9 2 3
>
> e11 >
> 40:0 12:0 12:0 0 0 0
>
> > 6
>
> e22 >
>
> 12:0 667 266 0 0 0 7
< = 6 6 12:0 266
7
7 1

e33 6 667 0 0 0 7
¼6
>
> c23 >
> 6 0 0 0 800 0 0 7 9
7 10 psi
>
> >
> 4 5
>
> c > 0 0 0 0 500 0
: 13 >;
c12 0 0 0 0 0 500
8 9 8 9
>
> 50 >> >
> 1966 Ain:=in: >
>
>
> > > >
>
> 5 >>
>
>
>
> 2203 Ain:=in: >
>
>
< = < =
2 596 Ain:=in:
ð10 psiÞ ¼
3
>
> 4 >> >
> 3200 Arad > >
>
> >
> >
> >
>
> 1:5 >
> >
> 750 Arad > >
>
: ; : ;
3:5 1750 Arad
Example Problem 3
An orthotropic composite is subjected to a state of stress that causes the
following state of strain:
8 9 8 9
>
> e11 >
> >
> 1500 Am=m >>
>
> > > >
>
> e22 >
>
>
>
>
> 2000 Am=m >>
>
< = < =
e33 1000 Am=m
¼
>
>
> > >
c23 > > 2500 Arad > >
>c >
> > >
> > 500 Arad >
> >
>
>
: 13 >
; : > >
;
c12 2000 Arad
Determine the stresses that caused these strains (use material properties
listed in Example Problem 1).
Solution. Because the composite is orthotropic, stresses are calculated using
Eq. (19). The stiffness matrix can be obtained by: (a) inverting the compliance
matrix determined as a part of Example Problem 1, (b) through the use of Eq.
(20), or (c) through the use of Eq. (21). All three methods are entirely
equivalent, and which procedure is selected for use is simply a matter of
convenience. Equation (21) will be used in this example:
X ¼ E11 E22 m212 E2

22 m13 E22 E33 m23 E11 E33 2m12 m13 m23 E22 E33
2 2
X ¼ fð170Þð10Þ ð0:30Þ2 ð10Þ2 ð0:35Þ2 ð10Þð8Þ ð0:40Þ2 ð170Þð8Þ

2ð0:30Þð0:35Þð0:40Þð10Þð8ÞgðGPaÞ2
X ¼ 1457ðGPaÞ2

ðE22 m223 E33 ÞE211 fð10 GPaÞ ð0:40Þ2 ð8 GPaÞgð170 GPaÞ2
C11 ¼ ¼
X 1457ðGPaÞ2
¼ 172:98 GPa
ðm12 E22 þ m13 m23 E33 ÞE11 E22
C12 ¼
X
fð0:30Þð10 GPaÞ þ ð0:35Þð0:40Þð8 GPaÞgð170 GPaÞð10 GPaÞ
¼
1457ðGPaÞ2
¼ 4:808 GPa
ðm12 m23 þ m13 ÞE11 E22 E33
C13 ¼
X
fð0:30Þð0:40Þ þ ð0:35Þð170 GPaÞgð10 GPaÞð8 GPaÞ
¼
1457ðGPaÞ2
¼ 4:387 GPa
ðE11 þ m213 E33 ÞE222 fð170 GPaÞ ð0:35Þ2 ð8 GPaÞgð10 GPaÞ
C22 ¼ ¼
X 1457ðGPaÞ2
¼ 11:602 GPa
ðm23 E11 þ m12 m13 E22 ÞE22 E33
C23 ¼
X
fð0:40Þð170 GPaÞ þ ð0:35Þð0:35Þð10 GPaÞgð10 GPaÞð8 GPaÞ
¼
1457ðGPaÞ2
¼ 3:792 GPa
ðE11 m212 E22 ÞE22 E33
C33 ¼
X
fð170 GPaÞ ð0:30Þ2 ð10 GPaÞgð10 GPaÞð8 GPaÞ
¼ ¼ 9:286 GPa
1457ðGPaÞ2
C44 ¼ G23 ¼ 8 GPa C55 ¼ G13 ¼ 10 GPa C66 ¼ G12 ¼ 13 GPa
Applying Eq. (19), the stresses are:
8 9 2 3
>
> r11 >
> 172:98 4:808 4:387 0 0 0
>
> >
> 6 4:808
>
> r >
> 11:602 3:792 0 0 0 7
= 6 7
22
< 6 4:387
r33 3:792 9:286 0 0 0 7
¼6
6
7ðGPaÞ
> s23 >
> > 6 0 0 0 8:0 0 0 7 7
>
> >
> 4 0
> s13 >
> > 0 0 0 10:0 0 5
: ;
s12 0 0 0 0 0 13:0

8 9 8 9
>
> 1500 Am=m >
> >
> 245:5 MPa >
>
>
> > > >
>
> 2000 Am=m >>
>
>
>
> 19:78 MPa >>
>
< = < =
1000 Am=m 10:29 MPa
¼
>
> 2500 Arad >
> > 20:0 MPa >
>
> > >
> >
>
>
>
>
> 500 Arad >
> >
> 5:00 MPa >>
>
: ; : ;
2000 Arad 26:0 MPa
Example Problem 4
A transversely isotropic composite is subjected to a state of stress that causes
the following state of strain:
8 9 8 9
>
> e11 >
> >
> 1250 Ain:=in: >
>
>
> > > >
>
> e22 >
>
>
>
> 1000 Ain:=in: >>
< = > < >
=
e33 500 Ain:=in:
¼
>
> c23 >
> > 2500 Arad >
>
> > > > >
>
>
> c13 >
>
>
>
>
> 1000 Arad > >
>
: ; : ;
c12 2000 Arad
Determine the stresses that caused these strains (use material properties
listed in Example Problem 2)
Solution. Because the composite is transversely isotropic, stresses are calcu-
lated using Eq. (24). The stiffness matrix can be obtained by: (a) inverting the
compliance matrix determined as a part of Example Problem 2, (b) through
the use of Eq. (25), or (c) through the use of Eq. (26). All three methods are
entirely equivalent, and which procedure is selected for use is simply a matter
of convenience. Equation (26) will be used in this example:
X ¼ E11 ð1 m23 Þ 2m212 E22
X ¼ ð25 MsiÞð1 0:40Þ 2ð0:30Þ2 ð1:5 MsiÞ ¼ 14:73 Msi
E211 ð1 m23 Þ ð25 MsiÞ2 ð1 0:40Þ

C11 ¼ ¼ ¼ 25:46 Msi
X 14:73 Msi
m12 E11 E22 ð0:30Þð25 MsiÞð1:5 MsiÞ

C12 ¼ ¼ ¼ 0:7637 Msi
X 14:73 Msi
2
E22 ðE11 m212 E22 Þ ð1:5 MsiÞ ð25 MsiÞ ð0:30Þ ð1:5 MsiÞ
C22 ¼ ¼
Xð1 þ m23 Þ 14:73 Msið1 þ 0:40Þ
¼ 1:809 Msi
E22 ðm23 E11 þ m212 E22 Þ
C23 ¼
Xð1 þ m23 Þ

ð1:5 MsiÞ ð0:40Þð25 MsiÞ þ ð0:30Þ2 ð1:5 MsiÞ
¼ ¼ 0:7372 Msi
14:73Msið1 þ 0:40Þ
E22 ð1:5 MsiÞ
C44 ¼ ¼ ¼ 0:5357 Msi
2ð1 þ m23 Þ 2ð1 þ 0:40Þ
C66 ¼ G12 ¼ 2:0 Msi

Applying Eq. (24), the stresses are:
8 9 2 3
>
> r11 >
> 25:46 0:7637 0:7637 0 0 0
>
> >
> 6 0:7637 1:809 0:07372
>
> r >
> 0 0 0 7
= 6 7
22
< 6 0:7637 0:7372
r33 1:809 0 0 0 7
¼6
6
7ðMsiÞ
>
> s23 >
> 6 0 0 0 0:5357 0 0 77
>
> >
> 4 0
>
> s13 >
> 0 0 0 2:0 0 5
: ;
s12 0 0 0 0 0 2:0
8 9 8 9
> 1250 Ain:=in: >
>
> >
>
>
>
>
32:97 ksi >
>
>
>
> 1000 Ain:=in: >
> >
> 3:13 ksi >>
>
< >
= < > >
=
500 Ain:=in: 2:60 ksi
¼
>
> 2500 Arad > > > > 1:34 ksi >>
> 1000 Arad >
> > > > 2:00 ksi >>
>
> >
> >
> >
>
: ; : ;
2000 Arad 4:00 ksi
3 STRAINS INDUCED BY A CHANGE IN TEMPERATURE

OR MOISTURE CONTENT
Material properties relating strains to a uniform change in temperature and a
uniform change in moisture content were defined in Secs. 3 and 4 of Chap. 3,
respectively. For anisotropic materials, strains caused by a change in temper-
ature are given by Eq. (3.22), repeated here for convenience:
2 3 2 3
eTxx eTxy eTxz axx axy axz
6 eT eTyz 7 6a ayz 7
4 yx eTyy 5 ¼ DT4 yx ayy 5 ðrepeatedÞð3:22Þ
eTzx eTzy eTzz azx azy azz
Similarly, strains caused by a change in moisture content are given by

Eq. (3.25), repeated here for convenience:
2 3 2 3
eM
xx cM
xy cM
xz bxx bxy bxz
6 cM 7 6 byz 7
4 yx eM
yy yz 5 ¼ ðDMÞ4 bxy
cM byy 5 ðrepeatedÞð3:25Þ
cM
zx cM
zy eM
zz bxz byz bzz

As before, the strain tensors must be symmetric. This allows the use of
contracted notation, and hence Eqs. (3.22) and (3.25) can be written in the
form of column arrays:
8 T 9 8 9 8 M9 8 9
>
> exx >> >
> axx >> >
> exx >
> >
> bxx >
>
>
> >
> >
> >
> >
> >
> >
> >
>
>
> eTyy >
> >
> ayy >> >
> eM >
> >
> byy >
>
>
> >
> >
> >
> >
> yy >
> >
> >
>
< eT >
> = <a >
> = 
> = 
> =
zz zz zz zz
¼ DT ðandÞ ¼ DT ð27Þ
> cyx >
> T
> > ayz >
> > > cyx >
> M
> > byz >
> >
>
> >
> >
> >
> >
> >
> >
> >
>
> cTxz >
> > > axz >
> > >
> cM >
> > bxz >
> >
>
> >
> >
> >
> >
> xz >
> >
> >
>
: cT >
> ; >
: >
; : cM >
> ; :b >
> ;
xy axy xy xy
In the case of an orthotropic material a12=a13=a23=b12=b13=

b23=0, and Eq. (27) becomes:
8 T 9 8 9 8 M9 8 9
> e11 > >
>a11 >> > e11 > > b11 >
>
> T > > >
> >
> >
> >
> >
> >
>
>
> >
> >
> >
> >
> M> > >
> >
>
>
> e >
> >
>a 22 >
> >
> e >
> >
> b22 >
>
>
>
22
>
> >
> >
> >
>
22
>
> >
> >
< eT = <a = =
33 33 33 33
>
> cT23 >
> >
> 0 > > >
> cM >
23 >
>
> 0 >>
>
> >
> >
> >
> >
> >
> >
> >
>
>
> > > > > > > >
> cT13 >
> >
> >0 >
>
> >
>
>
>
> cM >
>
13 >
>
>
> 0 >>
>
>
: T ; > >
> > > > > >
: > ; : M; : ;
c12 0 c12 0
In addition to these simplifications, for a transversely isotropic material
with symmetry in the 2–3 plane, a33=a22 and b33=b22. Hence, for a trans-
versely isotropic material, Eq. (28) becomes:
8 T 9 8 9 8 M9 8 9
> e11 > >
>a11 >> > e11 > > b11 >
>
> > > > > > > >
> eT >
> >
>
>
>
>
>
>
> > eM >
>
> >
>
>
>
>
>
>
>
>
> >
> >
>a 22 >
> >
> >
> >
> b22 >
>
>
>
22
>
> >
> > > 22
> > >
< eT = <a > = >
 
> =
33 22 33 22
>
> cT23 >
> >
> 0 > > >
> cM >
23 >
>
> 0 >>
>
> >
> >
> >
> >
> >
> >
> >
>
>
> >
> >
> > > > > >
> c13 >
>
T > >0 >
> >
>
>
>
M >
> c13 >
> > 0 >
>
> >
>
>
: T ; > >
> > > > > >
: > ; : M; : ;
c12 0 c12 0
4 STRAINS INDUCED BY COMBINED EFFECTS OF STRESS,

TEMPERATURE, AND MOISTURE
The strains induced by stress under constant environmental conditions for
anisotropic materials were discussed in Sec. 1, and a similar discussion for
the case of orthotropic and transversely isotropic materials was presented in

Sec. 2. Strains induced by a uniform change in temperature or moisture
content in the absence of stress was discussed for all three material
classifications in Sec. 3.
We will now consider the strains induced if all of these mechanisms
occur simultaneously. That is, we wish to consider the strains induced by the
combined effects of stress, a uniform change in temperature, and a uniform
change in moisture content. We will call this the total strain. Rigorously
speaking, the total strain tensor (eij) is a nonlinear coupled function of these
three mechanisms:
eij ¼ fðr; T; MÞ
The function f( ) is a nonlinear function of stress, temperature, and
moisture content, even though we have limited our attention to linear rela-
tionships between strain and these three mechanisms. That is, we have
defined:
Young’s modulus as the slope of linear region of the stress–strain
curve (Sec. 2 of Chap. 3)
The coefficient of thermal expansion (CTE) as the slope of the linear
region of the strain–DT curve (Sec. 3 of Chap. 3)
The coefficient of moisture expansion (CME) as the slope of the
linear region of the strain–DT curve (Sec. 4 of Chap. 3).
Despite these assumptions of linearity, the strain response may still be a
coupled function of stress, temperature, and moisture because a change in one
variable may cause a change in the other two. For example, for all polymer-
based materials, an increase in temperature will ordinarily cause a decrease in
Young’s modulus. Similarly, an increase in moisture content often causes an
increase in CTEs and a decrease in Young’s modulus.
These coupling effects are ignored throughout this text. It is assumed
that the strain response is an uncoupled function of stress, temperature, and
moisture. For example, we assume that Young’s modulus is measured under
some standard environmental condition (say, at room temperature and 0%
moisture content), and that subsequent changes in temperature or moisture
content are relatively modest such that Young’s modulus may be assumed to
remain constant. Based on these assumptions, the total strain tensor induced
in a structure is simply the sum of the strains induced by each of mechanism
acting independently:
eij ¼ erij þ eTij þ eM

ij ð30Þ
The superscripts j, T, and M used in Eq. (30) indicate that the individual
components of strain are caused by the application of stress, by a uniform

change in temperature, and by a uniform change in moisture content,
respectively.
Based on this assumption, for anisotropic materials, the total strain is
obtained by superimposing Eqs. (9) and (27):
8 9 2 38 9
>
> exx >
> S11 S12 S13 S14 S15 S16 >> rxx >>
>
> eyy >
> 6 S21 >
> ryy >>
>
> > S22 S23 S24 S25 S26 7
7> >
<e > = 6
6 S31
>
< >
=
S32 S33 S34 S35 S36 7 r
¼6 7
zz zz
> > 6 S41 S46 7
>
>
c yz >
> 6 S42 S43 S44 S45 7>
>
>
syz >>
>
>
> c >
> 4 S51 S52 S53 S54 S55 S56 5>
> s >
xz >
>
: xz >
; >
: >
;
cxy S61 S62 S63 S64 S65 S66 sxy
8 9 8 9
>
> axx >> >
> bxx >>
>
> ayy >> >
> byy >>
>
> >
> >
> >
< = =
azz
þDT þ DM zz
ð31Þ
>
> ayz >> >
> byz >>
>
> >
> >
> >
>
>
> a xz >
> >
> bxz >>
: ; : ;
axy bxy
Equation (31) allows the prediction of the strains induced by the

simultaneous effects of stress and uniform changes in temperature and/or
moisture content. In practice, the inverse problem is often encountered. That
is, a common circumstance is that the strains, the change in temperature, and
the change in moisture content have been measured, and we wish to calculate
stresses. This can be accomplished by inverting Eq. (31) according to the laws
of matrix algebra, resulting in:
8 9 2 3
>
> rxx >> C11 C12 C13 C14 C15 C16
>r >
> > 6
>
> yy >
> C21 C22 C23 C24 C25 C26 7
< = 6 6
7
rzz C31 C32 C33 C34 C35 C36 7
¼66 C41 C42 C43 C44 C45 C46 7
7
>
> syz >
> 6 7
>
> >
> 4 C51 C52 C53 C54 C55 C56 5
>
> sxz >>
: ;
sxy C61 C62 C63 C64 C65 C66
8 9
>
> exx DTaxx DMbxx > >
>
> >
>
>
> eyy DTayy DMbyy > >
>
> >
>
>
< >
ezz DTazz DMbzz =
ð32Þ
> cyz DTayz DMbyz >
> >
>
> >
>
>
> cxz DTaxz DMbxz > >
>
> >
>
>
: c DTa DMb > ;
xy xy xy
where the stiffness matrix Cij=Sij1, as discussed in Sec. 1.

Following an analogous procedure, the strains induced in an ortho-
tropic material by the combined effects of stress, a uniform change in
temperature, and/or a uniform change in moisture content can be found by
superimposing Eqs. (17) and (28):
8 9 2 38 9
>
> e11 >> S11 S13 0
S12 0 0 > > r11 >
>
>
> >
> 6 S12 >
> r22 >
>
>
> e >
> S23 0
S22 0 0 77>> >
>
= 6
22
< 6 S13 7 < =
e33 S33 0
S23 0 0 7 r33
¼6
6
>
> c23 >> 6 0 0 0 S44 0 0 77>> s >
>
> >
> > 23 >>
>
> c13 >>
4 0 0 0 0 S55 0 5> >
> s > >
: ; : 13 >;
c12 0 0 0 0 0 S66 s12
8 9 8 9
>
> a11 >
> >
> b11 >
>
>
> > > >
>
> a22 >
>
>
>
>
> b22 >
>
>
< = < =
a33 b33
þDT þ DM ð33Þ
>
> 0 >> >
> 0 >>
>
> >
> >
> >
>
> 0 >
> >
> 0 >>
>
: ; : ;
0 0
Inverting Eq. (33), we obtain:

8 9 2 3
> r11 > C11 C12 C13 0 0 0
>
> >
> 6
>
>
> r >
>
22 > 0 7
> 6 7
C C C 0 0
>
>
12 22 23
<r > = 6 6 C13 C23 C33 0 0 0 7
7
¼6 7
33
> > 6 0 7
>
>
s23 > 6 0
> 0 0 C44 0 7
>
> > 6
> 7
>
> s >
13 > 4 0 0 0 0 C 0 5
>
: >
;
55
s12 0 0 0 0 0 C66
8 9
>
> e11 DTa11 DMb11 > >
>
> >
>
>
> e22 DTa22 DMb22 > >
>
< >
=
e 33 DTa 33 DMb 33 ð34Þ
>
> c23 >
>
>
> >
>
>
> c >
>
>
: 13 >
;
c12
As discussed in preceding chapters, an implicit assumption in Eqs. (33)
and (34) is that the strain tensor, stress tensor, and material properties are all
referenced to the principal material coordinate system of the orthotropic
material (i.e., the 1–2–3 coordinate system). If an orthotropic is referenced to
a nonprincipal material coordinate system, then the relation between strain,
stress, temperature, and moisture content is given by Eq. (31) or Eq. (32).

Finally, the strains induced in a transversely isotropic material by the
combined effects of stress, a uniform change in temperature, and/or a uniform
change in moisture content can be found by superimposing Eqs. (23) and
(29):
8 9 2 38 9
>
> e11 >>
S11 S12 S12 0 0 0 > r11 >
> >
>
> >
> 6S 7>> >
>
>
> e >
> 6 S S 0 0 0 7 >
> r >
22 >
>
> 22 >
> 6
12 22 23
7>>
> >
>
> >
< e = 6 S12 S23 S22 7 < >
=
33 6 0 0 0 7 r 33
¼6 7
>
> c23 >> 6 0 0 2ðS22 S23 Þ 0 0 7 > s23 >
>
> >
> 6 0 7>>
>
>
>
>
>
> >
> 6 7 >
> >
>
>
> c >
> 4 0 0 0 0 S 0 5 >
> s 13 >
>
>
:
13
>
;
66
>
: >
;
c12 0 0 0 0 0 S66 s12
8 9 8 9
>
> a11 >
> >
> b11 >
>
>
> > > >
>
> a22 >
>
>
>
>
> b > >
< = < 22 >=
a22 b22
þDT þ DM ð35Þ
>
> 0 >> >
> 0 >>
>
> >
> >
> >
>
> 0 >
> >
> 0 >>
>
: ; : ;
0 0
Inverting Eq. (35), we have:

8 9 2 3
>
> r11 >> C11 C12 C13 0 0 0
>
> > 6
>
>
> r >
> C C C 0 0 0 7
= 6 7
22 12 22 23
< 6 C12 C23 C22
r33 0 0 0 7
¼66 0
7
>
> s 23 >
> 6 0 0 ðC 22 C23 Þ=2 0 0 77
>
> >
> 4 0
>
> s >
> 0 0 0 C66 0 5
: 13
;
s12 0 0 0 0 0 C66
8 9
>
> e11 DTa11 DMb11 > >
>
> >
>
> e22 DTa22 DMb22 > >
>
< =
e33 DTa22 DMb22
ð36Þ
>
> c23 >
>
>
> >
>
>
> c13 >
>
: ;
c12
Once again, Eqs. (35) and (36) are valid only if referenced to the
principal material coordinate system of the transversely isotropic material
(i.e., the 1–2–3 coordinate system). If a transversely isotropic material is
referenced to a nonprincipal material coordinate system, then the relation
between strain, stress, temperature, and moisture content is given by Eq. (31)
or Eq. (32).

HOMEWORK PROBLEMS
An anisotropic material with the following properties is considered in
problems 1–4:
Exx ¼ 100 GPa Eyy ¼ 200 GPa Ezz ¼ 75 GPa

vxy ¼ 0:20 vxz ¼ 0:25 vyz ¼ 0:60
Gxy ¼ 60 GPa Gxz ¼ 75 GPa Gyz ¼ 50 GPa
gxx;xy ¼ 0:30 gxx;xz ¼ 0:25 gxx;yz ¼ 0:30
gyy;xy ¼ 0:60 gyy;xz ¼ 0:75 gyy;yz ¼ 0:20
gzz;xy ¼ 0:20 gzz;xz ¼ 0:05 gzz;yz ¼ 0:15
lxy;xz ¼ 0:10 lxy;yz ¼ 0:05 gxz;yz ¼ 0:10

axx ¼ 5 Am=m jC ayy ¼ 10 Am=m jC azz ¼ 20 Am=m jC
axy ¼ 5 Arad=jC axz ¼ 15 Arad=jC ayz ¼ 25 Arad=jC
bxx ¼ 300 Am=m%M byy ¼ 60 Am=m%M bzz ¼ 1200 Am=m%M
bxy ¼ 150 Arad=%M bxz ¼ 1000 Arad=%M byz ¼ 350 Arad=%M
1. Calculate the compliance matrix, Sij.

2. Calculate the stiffness matrix, Cij=Sij 1. (Perform this calculation using a
suitable software package such as Maple, Matlab, Mathematica, etc.)
3. Consider the following stress tensor:
2 3 2 3
rij ¼ 4 ryx ryy ryz 5 ¼ 4 10 90 30 5 ðMPaÞ
rzx rzy rzz 25 30 25
(a) Calculate the strains induced by this stress tensor, assuming no

change in temperature or moisture content (i.e., assume
DT=DM=0).
(b) Calculate the strains induced by this stress tensor and a temper-
ature increase of 100jC (assume DM=0).
(c) Calculate the strains induced by this stress tensor and a 2%
increase in moisture content (assume DT=0).
(d) Calculate the strains induced by this stress tensor, a temperature
increase of 100jC, and a 2% increase in moisture content.

4. Consider the following strains:
exx ¼ 1500 Am=m eyy ¼ 2000 Am=m ezz ¼ 1750 Am=m

cxy ¼ 750 Arad cxz ¼ 500 Arad cyz ¼ 850 Arad
(a) Calculate the stress tensor that caused these strains, assuming no
change in temperature or moisture content (i.e., assuming
DT=DM=0).
(b) Calculate the stress tensor that caused these strains, if these strains
were caused by the simultaneous effects of stress and a temperature
decrease of 100jC (assume DM=0).
(c) Calculate the stress tensor that caused these strains, if these strains
were caused by the simultaneous effects of stress, a temperature
decrease of 100jC, and a 2% increase in moisture content.
An orthotropic material with the following properties is considered in
problems 5–8:
E11 ¼ 100 GPa E22 ¼ 200 GPa E33 ¼ 75 GPa

m12 ¼ 0:20 m13 ¼ 0:25 m23 ¼ 0:60
G12 ¼ 60 GPa G13 ¼ 75 GPa G23 ¼ 50 GPa
a11 ¼ 1 Am=mjC a22 ¼ 25 Am=mjC a33 ¼ 15 Am=mjC
b11 ¼ 100 Am=m%M b22 ¼ 600 Am=m%M b33 ¼ 1000 Am=m%M
5. Calculate the compliance matrix, Sij.

6. Calculate the stiffness matrix, Cij.
7. Consider the following stress tensor:
2 3 2 3
r11 r12 r13
75 10 25
6 7
rij ¼ 4 r21 r22 r23 5 ¼ 4 10 90 30 5 ðMPaÞ
25 30 25
r31 r32 r33
(a) Calculate the strains induced by this stress tensor, assuming no

change in temperature or moisture content (i.e., assume DT=
DM=0).
(b) Calculate the strains induced by this stress tensor and a temper-
ature increase of 100jC (assume DM=0).
(c) Calculate the strains induced by this stress tensor and a 2%
increase in moisture content (assume DT=0).
(d) Calculate the strains induced by this stress tensor, a temperature
increase of 100jC, and a 2% increase in moisture content.

8. Consider the following strains:
e11 ¼ 2000 Am=m e22 ¼ 3000 Am=m e33 ¼ 1500 Am=m
c12 ¼ 750 Arad c13 ¼ 1000 Arad c23 ¼ 1250 Arad
(a) Calculate the stress tensor that caused these strains, assuming no
change in temperature or moisture content (i.e., assuming
DT=DM=0).
(b) Calculate the stress tensor that caused these strains, if these strains
were caused by the simultaneous effects of stress and a temperature
decrease of 100jC (assume DM=0).
(c) Calculate the stress tensor that caused these strains, if these strains
were caused by the simultaneous effects of stress, a temperature
decrease of 100jC, and a 2% increase in moisture content.
REFERENCE
1. Jones, R.M. Mechanics of Composite Materials; Hemisphere Publ. Co.: New
York, NY, 1975; ISBN 0-89116-490-1.

5
Unidirectional Composite Laminates Subject
to Plane Stress
This chapter is devoted to the elastic behavior and failure response of

unidirectional composite laminates. In the present context, the term ‘‘uni-
directional’’ is meant to imply that although the laminates considered may
contain many plies, the principal material coordinate system in all plies is
oriented in the same direction. It is also assumed that the laminate is thin and
platelike, such that a state of plane stress exists.
Four primary topics are addressed in this chapter. First, in Sec. 1, the
elastic response of a unidirectional composite referenced to the principal ma-
terial coordinate system and subject to a plane stress state will be described.
This will lead to a so-called ‘‘reduced’’ form of Hooke’s law. Second, in Secs. 2
and 3, the elastic response of a unidirectional composite referenced to an
arbitrary (nonprincipal) coordinate system is discussed, which will lead to the
definition of the ‘‘transformed, reduced’’ form of Hooke’s law. The ‘‘effec-
tive’’ elastic properties of a unidirectional composite laminate are then
discussed in Sec. 4. We then turn our attention to the prediction of composite
failure. In Sec. 5, several macromechanics-based failure theories are defined
and used to predict the failure of a unidirectional laminate referenced to the
principal material coordinate system. These theories are then used to predict
the failure of a unidirectional composite laminate referenced to an arbitrary
(nonprincipal) coordinate system in Sec. 6.
193

1 UNIDIRECTIONAL COMPOSITES REFERENCED TO THE
PRINCIPAL MATERIAL COORDINATE SYSTEM
The strains induced in an orthotropic material subjected to a general 3-D
stress tensor, a uniform change in temperature, and/or a uniform change in
moisture content were described in Chap. 4. The strain response is summa-
rized by Eq. (33) of Chap. 4, repeated here for convenience:
8 9 2 38 9 8 9 8 9
> e11 > S11 S12 S13 0 0 0 > r11 > > a11 > > b11 >
>
> >
> > > > > > >
>
> > 6
> 7>>
>
>
>
>
>
>a >
> >
> >b >
>
> >
>
>
> e >
> 6 S S S 0 0 0 7 > r > > > > >
7> > > > > >
12 22 23 22 22 22
>
>
22
>
> 6 >
> >
> >
> >
> >
> >
>
<e = 6S 7 <
0 7 r33 = < a33 = < b33 =
33 6 13 S23 S33 0 0
¼6 7 þ DT þ DM
>
> c23 >> 6 0 0 0 S44 0 0 7 >
> s23 >> >
> 0 >> >
> 0 >>
>
> >
> 6 6 7 >
> >
> >
> >
> >
> >
>
>
> >
> 7 >
> >
> >
> >
> >
> >
>
> c 13 >
> 4 0 0 0 0 S55 0 5> > s13 >> >
> 0 >> >
> 0 >>
>
>
: >
; >
: >
; >
: >
; >
: >
;
c12 0 0 0 0 0 S66 s 12 0 0
Three mechanisms that contribute to the total strains appear in the

above equation: strains caused by stress, strains caused by a uniform change
in temperature (DT ), and strains caused by a uniform change in moisture
content (DM ). Equation (33) of Chap. 4 is valid for any orthotropic material,
as long as the strain tensor, stress tensor, and material properties are all
referenced to the principal 1–2–3 coordinate system. Let us now consider the
strains induced in an orthotropic material by a state of plane stress. Assuming
that r33=s23=s13=0, Eq. (33) of Chap. 4 becomes:
8 9 2 38 9 8 9 8 9
> e11 > S11 S12 S13 0 0 0 > r11 > > a11 > > b11 >
>
> >
> 6 >
> >
> >
> >
> >
> >
>
>
> >
> 7>> >
> >
> >
> >
> >
>
>
> e >
> 6 S12 S 22 S23 0 0 0 7 >
> r22 >
> >
> a 22 >
> >
> b 22 >
>
>
>
22
>
> 6 7 >
> >
> >
> >
> >
> >
>
<e = 6S 7 < = < a = 
> > 6 0 0 7 > 0 > > > > >
> c23 >
> > 6
>
0 0 S44 0 7>>
>
>
>
> > 0 >
>
> >
>
>
> 0 >
> >
>
>
> > 6 7 > > > > > >
> c13 >
> > 4 0
> 0 0 0 S55 0 5> > 0 >
> >
> > 0 >
>
> >
>
>
> 0 >
> >
>
>
: >
; >
: >
; >
: >
; >
: >
;
c12 0 0 0 0 0 S66 s12 0 0
ð1Þ
Note that Eq. (1) shows that in the case of plane stress, the out-of-plane shear
strains are always equal to zero (c23=c13=0). It is customary to write the
expressions for the remaining four strain components as follows:
8 9 2 38 9 8 9 8 9
> b
< e11 >= S11 S12 0 > < r11 >= < a11 >
> = < 11 >
> =
6 7
e22 ¼ 4 S12 S22 0 5 r22 þ DT a22 þ DM b22 ð2aÞ
>
: >
; >
: >
; >
: >
; >
: >
;
c12 0 0 S66 s12 0 0

and
e33 ¼ S13 r11 þ S23 r22 þ DTa33 þ DMb33 ð2bÞ
Equations (2a) and (2b) are called the reduced forms of Hooke’s law for
an orthotropic composite. They are only valid for a state of plane stress and
are called ‘‘reduced’’ laws because we have reduced the allowable stress tensor
from three dimensions to two dimensions. The 33 array in Eq. (2a) is called
the reduced compliance matrix. Note that despite the reduction from three to
two dimensions, we have retained the subscripts used in the original com-
pliance matrix. For example, the element that appears in the (3,3) position of
the reduced compliance matrix is labeled S66. The definition of each com-
pliance term is not altered by the reduction from three to two dimensions, and
each term is still related to the more familiar engineering constants (E11, E22,
v12, etc.) in accordance with Eq. (18) of Chap. 4.
Inverting Eq. (2a), we obtain:
8 9 2 38 9
< r11 >
> = Q11 Q12 0 > < e11 DTa11 DMb11 > =
6 0 7
r22 ¼ 4 Q12 Q22 5 e22 DTa22 DMb22 ð3Þ
>
: >
; >
: >
;
s12 0 0 Q66 c12
The 33 array that appears in Eq. (3) is called the reduced stiffness
matrix and equals the inverse of the reduced compliance matrix:
2 3 2 31
Q11 Q12 0 S11 S12 0
6Q 7 6
0 5u4 S12 0 7
4 12 Q22 S22 5 ð4Þ
0 0 Q66 0 0 S66
Note that we are now using a different symbol to denote stiffness. That
is, the original 3-D stiffness matrix was denoted Cij (as in Eq. (19) of Chap. 4,
for example), whereas the reduced stiffness matrix is denoted Qij. This change
in notation is required because individual members of the reduced stiffness
matrix are not equal to the corresponding members in the original stiffness
matrix. That is, Q11 p C11, Q12 p C12, Q22 p C22, and Q66 p C66. Relations
between Qij and Cij can be derived as follows. From Eq. (19) of Chap. 4, it can
be seen that for an orthotropic material subjected to an arbitrary state of
stress:
r11 ¼ C11 e11 þ C12 e22 þ C13 e33 ð5aÞ
and
r33 ¼ C13 e11 þ C23 e22 þ C33 e33 ð5bÞ

However, the reduced stiffness matrix relates stress to strain under conditions
of plane stress by definition. Setting r33=0 and solving Eq. (5b) for e33, we
have:
ðC13 e11 þ C23 e22 Þ

e33 ¼ ð6Þ
C33
The out-of-plane strain e33 must be related to in-plane strains e11 and e22 in
accordance with Eq. (6), otherwise, a state of plane stress does not exist in the
composite. Substituting Eq. (6) into Eq. (5a) and simplifying, we have:

C11 C33 C213 C12 C33 C13 C23

r11 ¼ e11 þ e22 ð7Þ
C33 C33
On the other hand, from Eq. (3), r11 is given by (with DT=DM=0):
r11 ¼ Q11 e11 þ Q12 e22 ð8Þ
Comparing Eqs. (7) and (8), it is immediately apparent that:
C11 C33 C213

Q11 ¼ ð9aÞ
C33
C12 C33 C13 C23
Q12 ¼ ð9bÞ
C33
Using a similar procedure, it can be shown that:
C22 C33 C223

Q22 ¼ ð9cÞ
C33
Q66 ¼ C66 ð9dÞ
In essence, the definition of elements of the Qij matrix differs from those of the
Cij matrix because the Qij matrix is defined for plane stress conditions only,
whereas Cij can be used for any stress state.
Elements of the reduced stiffness matrix may be related to the elements
of the compliance matrix by either substituting Eq. (20) of Chap. 4 in Eqs.
(9a)–(9d), or by simply performing the matrix inversion indicated in Eq. (4).
In either case, it will be found that:
S22 S12
Q11 ¼ Q12 ¼ Q21 ¼
S11 S22 S212 S11 S22 S212
ð10Þ
S11 1
Q22 ¼ Q66 ¼
S11 S22 S212 S66

Alternatively, the elements of the reduced stiffness matrix are related to
more familiar engineering constants as follows:
E211 m12 E11 E22
Q11 ¼ Q12 ¼ Q21 ¼
E11 m212 E22 E11 m212 E22
ð11Þ
E11 E22
Q22 ¼ Q66 ¼ G12
E11 m212 E22
Equations (1)–(11) were developed assuming that the composite is an

orthotropic material. Now consider the response of a transversely isotropic
composite subjected to a state of plane stress. As before, we assume that
r33=s23=s13=0. From Eq. (34) of Chap. 4, it is seen that the out-of-plane
shear strains equal zero (c23=c13=0), and the remaining four strains can be
written as:
8 9 2 38 9 8 9 8 9
< e11 >
> = S11 S12 0 > < r11 >
= < a11 >
> = >
< b11 >
=
6 7
e22 ¼ 4 S12 S22 0 5 r22 þ DT a22 þ DM b22
>
: >
; >
: >
; >
: >
; >
: >
;
c12 0 0 S66 s12 0 0
ð12aÞ
and
e33 ¼ S12 r11 þ S23 r22 þ DTa22 þ DMb22 ð12bÞ
Comparing Eq. (12a) with Eq. (2a), it is seen that the relationship
between in-plane strains (e11, e22, and c12) and in-plane stress components
(r11, r22, and s12) is identical for orthotropic and transversely isotropic
materials. In fact, identical results are obtained for the out-of-plane normal
strain as well since for a transversely isotropic material, S13=S12, a33=a22,
and b33=b22, and therefore Eq. (2b) is equivalent to Eq. (12b). Consequently,
Eq. (3) can also be applied to a transversely isotropic material. Equations
(9a)–(9d) are also still applicable except that for a transversely isotropic
material (with symmetry in the 2–3 plane), C33=C22 and C13=C12.
Inverting Eq. (12a), we obtain:
8 9 2 38 9
< r11 >
> = Q11 Q12 0 > < e11 DTa11 DMb11 > =
6 0 7
r22 ¼ 4 Q12 Q22 5 e22 DTa22 DMb22
>
: >
; >
: >
;
s12 0 0 Q66 c12
Since this result is identical to Eq. (3), it is seen once again that the relationship
between in-plane strains (e11, e22, and c12) and in-plane stress components
(r11, r22, and s12) is identical for orthotropic and transversely isotropic
materials.

Example Problem 1
Determine the strains induced in a unidirectional graphite–epoxy composite
subjected to the in-plane stresses shown in Fig. 1. Assume that DT=DM=0
and use material properties listed in Table 3 of Chap. 3.
Solution. The magnitude of each stress component is indicated in Fig. 1.
Based on the sign conventions reviewed in Sec. 5 of Chap. 2, the algebraic sign
of each stress component is:
r11 ¼ 200 MPa r22 ¼ 30 MPa s12 ¼ 50 MPa
Strains can be calculated using either Eq. (2a) or Eq. (12a). Since DT=
DM=0, we have:
8 9 2 38 9
< e11 >
> = S11 S12 0 > < r11 >
=
6 7
e22 ¼ 4 S12 S22 0 5 r22
>
: >
; >
: >
;
c12 0 0 S66 s12
Each term within the reduced compliance matrix is calculated using Eq. (18) of
Chap. 4:
2 1 m12 3
8 9 6 E 0
E11 78 9
>
< e11 >= 6
11
7>
< r11 >=
6 m 7
e22 ¼ 6 6
12 1
0 7 r22
7
>
: >
; 6 E11 E22 7>
: >
;
c12 4 1 5 s 12
0 0
G12
Figure 1 Unidirectional composite subjected to in-plane stresses (magnitudes

of in-plane stresses shown).

From Table 3 of Chap. 3:
E11 ¼ 170 GPa E22 ¼ 10 GPa
m12 ¼ 0:30 G12 ¼ 13 GPa
Using these values:
8 9 2 3
>
< e11 >= 5:88 1012 1:76 1012 0
6 7 1
e22 ¼ 4 1:76 1012 100 1012 0 5
>
: >
; Pa
c12 0 0 76:9 1012
8 9
< 200 10 >
> 6
=
30 106 ðPaÞ
>
: >
;
50 106
Completing the matrix multiplication indicated, we obtain:
8 9 8 9
> = >
< e11 > < 1230 Am=m > =
e22 ¼ 3350 Am=m
>
: ; >
> : >
;
c12 3850 Arad
Example Problem 2
Determine the strain induced in a unidirectional graphite–epoxy composite
subjected to
(a) The in-plane stresses shown in Fig. 1.
(b) A decrease in temperature DT=155jC.
(c) An increase in moisture content DM=0.5%.
Use material properties listed in Table 3 of Chap. 3.
Solution. This problem involves three different mechanisms that contribute
to the total strain induced in the laminate: the applied stresses, the temper-
ature change, and the change in moisture content. The total strains induced
by all three mechanisms can be calculated using either Eq. (2a) or Eq. (12a):
8 9 2 38 9 8 9 8 9
< e11 >
> = S11 S12 0 > < r11 >
= < a11 >
> = < b11 >
> =
6 7
e22 ¼ 4 S12 S22 0 5 r22 þ DT a22 þ DM b22
>
: >
; >
: >
; >
: >
; >
: >
;
c12 0 0 S66 s12 0 0
Numerical values for the stresses and reduced compliance matrix are given in
Example 1, and the linear thermal and moisture expansion coefficients for
graphite–epoxy are (from Table 3 of Chap. 3):
a11 ¼ 0:9 Am=m=o C
a22 ¼ 27 Am=m=o C

b11 ¼ 150 Am=m=%M
b22 ¼ 4800 Am=m=%M
Hence, Eq. 2(a) or Eq. (12a) becomes:
8 9 2 3
< e11 >
> = 5:88 1012 1:76 1012 0
6 7
e22 ¼ 4 1:76 1012 100 1012 0 5
>
: >
; 12
c12 0 0 76:9 10
8 9 8 9
6
< 200 10 >
> < 0:9 10 >
>
6
= =
30 106 þ ð155Þ 27 106 þ ð0:5Þ
>
: > > >
6; : ;
50 10 0
8 9
6
< 150 10 >
> =
4800 106
>
: >
;
0

8 9 8 9 8 9 8 9
< e11 >
> = > < 1230 Am=m > = > < 140 Am=m > = > < 75 Am=m >
=
e22 ¼ 3350 Am=m þ 4185 Am=m þ 2400 Am=m
>
: ; >
> : ; >
> : ; >
> : >
;
c12 3850 Arad 0 0
8 9
< 1445 Am=m >
> =
¼ 5135 Am=m
>
: >
;
3850 Arad
An implicit assumption in this problem is that the composite is free to expand
or contract, as dictated by changes in temperature and/or moisture content.
Consequently, neither DT nor DM affects the state of stress, but rather affects
only the state of strain. Conversely, if the composite is not free to expand or
contract, then a change in temperature and/or moisture content does con-
tribute to the state of stress, as illustrated in Example Problem 3.
Example Problem 3
A thin, unidirectional graphite–epoxy composite laminate is firmly mounted
within an infinitely rigid square frame, as shown in Fig. 2. The coefficients of
thermal and moisture expansion of the rigid frame equal zero.
The composite is initially stress-free. Subsequently, however, the com-
posite/frame assembly is subjected to a decrease in temperature DT=155jC
and an increase in moisture content DM=0.5%. Determine the stresses
induced in the composite by this change in temperature and moisture content.
Use the material properties listed in Table 3 of Chap. 3 and ignore the
possibility that the thin composite will buckle if compressive stresses occur.

Figure 2 Unidirectional composite laminate mounted within an infinitely rigid
frame.
Solution. According to the problem statement, the frame is ‘‘infinitely rigid’’

and is made of a material whose thermal and moisture expansion coefficients
equal zero. Consequently, the frame will retain its original shape, regardless
of the temperature change, moisture change, or stresses imposed on the frame
by the composite laminate. Furthermore, the composite is ‘‘firmly mounted’’
within the frame. Together, these stipulations imply that the composite does
not change shape, although changes in temperature and moisture content
have occurred. Consequently, the total strains experienced by the composite
equal zero:
8 9 8 9
< e11 >
> <0>
= > =
e22 ¼ 0
>
: : >
; >
> ;
c12 0
The stresses induced can be calculated using Eq. (3):

8 9 2 38 9
< r11 >
> = Q11 Q12 0 > < e11 DTa11 DMb11 >
=
6Q 0 7
r22 ¼ 4 12 Q22 5 e22 DTa22 DMb22
>
: >
; >
: >
;
s12 0 0 Q66 c12
The terms within the reduced stiffness matrix are calculated in accordance
with Eq. (11):
E211 ð170 GPaÞ2

Q11 ¼ ¼ ¼ 170:9 GPa
E11 m212 E22 170 GPa ð0:30Þ2 ð10 GPaÞ

m12 E11 E22 ð0:30Þð170 GPaÞð10 GPaÞ
Q12 ¼ Q21 ¼ ¼
E11 m12 E22 170 GPa ð0:30Þ2 ð10 GPaÞ
2
¼ 3:016 GPa
E11 E22 ð170 GPaÞð10 GPaÞ
Q22 ¼ ¼ ¼ 10:05 GPa
E11 m212 E22 170 GPa ð0:30Þ2 ð10 GPaÞ
Q66 ¼ G12 ¼ 13 GPa
Hence, in this case, Eq. (3) becomes:
8 9 2 3
>
< r11 > = 170:9 109 3:016 109 0
6 7
r22 ¼ 4 3:016 109 10:05 109 0 5
>
: >
;
s12 0 0 13:0 109
8 6
9
< 0 ð155Þ0:9 10 ð0:5 Þ 150 106 =

0 ð155Þ 27 106 ð0:5Þ 4800 106
: ;
0
8 9 8 9
>
< r11 >= > < 31:36 MPa >=
r22 ¼ 17:29 MPa
>
: >
; > : >
;
s12 0
The reader may initially view this example to be somewhat unrealistic.

After all, the frame has been assumed to be ‘‘infinitely rigid.’’ In reality, an
infinitely rigid material (i.e., one for which E!l) does not exist. Further-
more, it is assumed that the thermal expansion coefficient for the frame is zero,
which is also not valid for most real materials (the assumption that the
moisture expansion coefficient is zero is true for metals and metal alloys).
Despite these unrealistic assumptions, this example problem illustrates
a common occurrence in composite laminates. Specifically, most modern
composites are cured at an elevated temperature (for example, many
graphite–epoxy systems are cured at 175jCc350jF). The composite can
normally be considered to be stress- and strain-free at the cure temperature.
After the cure is complete, the composite is typically cooled to room
temperatures, say 20jCc70jF, which corresponds to a temperature de-
crease of DT=155jCc280jF. Also, the moisture content immediately
after the cure can usually be assumed to equal 0%, but will slowly increase
following exposure to normal humidity levels over subsequent days or
weeks. Although adsorption of moisture rarely causes a significant gain in
weight (most composites adsorb a maximum of 3–4% moisture by weight,
even if totally immersed in water), even this slight gain in moisture content

can nevertheless cause significant strains to develop. If a DT and/or DM
occur, and if the unidirectional composite is not free to expand or contract
as dictated by these changes, then thermal- and/or moisture-induced stresses
will develop.
2 UNIDIRECTIONAL COMPOSITES REFERENCED TO AN

ARBITRARY COORDINATE SYSTEM
A unidirectional composite referenced to two different coordinate systems
is shown in Fig. 3. In Fig. 3(a), the composite is referenced to the principal
material coordinate system (i.e., the 1–2 coordinate system), and in this case,
either Eqs. (2a) and (2b) or Eq. (3) may be used to relate strains and stresses
within the composite.
In Fig. 3(b), however, the composite is referenced to an arbitrary
(nonprincipal) x–y coordinate system. This is often called an ‘‘off-axis’’ spec-
imen since the specimen is referenced to an arbitrary x–y coordinate system
rather than the principal 1–2 coordinate system. Suppose we wish to relate
strains and stresses referenced to the x–y coordinate system. For example,
suppose we know the stresses rxx, ryy, and sxy, as well as the material
properties referenced to the principal material coordinate system (E11, E22,
m12, etc.), and wish to calculate strains exx, eyy, and cxy. In this case, neither
Eqs. (2a) and (2b) nor Eq. (3) can be used directly since they require that the
stresses and strains be referenced to the 1–2 coordinate system.
We can perform this calculation using a three-step process. Specifically,
we can:
(a) Transform the known stresses from the x–y coordinate system to
the 1–2 coordinate system (using Eq. 20 of Chap. 2, for example),
which will give us the stress components r11, r22, and s12 that
correspond to the known values of rxx, ryy, and sxy.
(b) Apply Eq. (2a) to obtain in-plane strains e11, e22, and c12.
(c) Transform the calculated strains (e11, e22, and c12) from the 1–2
coordinate system back to the x–y coordinate system, finally
obtaining the desired strains exx, eyy, and cxy.
This three-step process is a rigorously valid procedure. However, it will later
be seen that during the analysis of a multi-angle composite laminate, the
process of transforming strains/stresses from an arbitrary x–y coordinate
system to the 1–2 coordinate system (and vice versa) must be performed for
each ply in the laminate. Since this transformation is encountered so
frequently, it becomes cumbersome to apply the three-step process for every
ply. Instead, it is very convenient to simply develop a form of reduced Hooke’s
law suitable for use in an arbitrary x–y coordinate system. In effect, we will

Figure 3 A unidirectional composite referenced to two different coordinate
systems.

transform Hooke’s law from the 1–2 coordinate system to an arbitrary x–y
coordinate system.
To simplify our discussion, assume for the moment that no change in
environment occurs, i.e., assume that DT=DM=0. In this case, Eq. (2a)
becomes:
8 9 2 38 9
< e11 >
> = S11 S12 0 > < r11 >
=
6 7
e22 ¼ 4 S12 S22 0 5 r22 ð13Þ
>
: >
; >
: >
;
c12 0 0 S66 s12
Equation (13) can be rewritten as:
2 38 9 2 38 9
1 0 0 > e
< 11 > = S11 S12 0 >
< r11 >
=
6 7 6 7
40 1 05 e22 ¼ 4 S12 S22 0 5 r22 ð14Þ
>
: >
; >
: >
;
0 0 2 c12 =2 0 0 S66 s12
In Eq. (14), we have employed the so-called ‘‘Reuter matrix’’ (named

after the person who suggested this approach [1]) to, in effect, divide the shear
strain by a factor of (1/2) within the strain array. Let:
2 3 2 3
1 0 0 S11 S12 0
6 7 6 7
½R ¼ 4 0 1 0 5 ðandÞ ½S ¼ 4 S12 S22 0 5
0 0 2 0 0 S66
so that Eq. (14) can be written in the following abbreviated form:
8 9 8 9
< 11 >
> e
= < r11 >
> =
½R e22 ¼ ½S r22
>
: >
; >
: >
;
c12 =2 s12
The transformation of strains within a plane from an x–y coordinate
system to another xV–yV coordinate system was discussed in Sec. 13 of Chap. 2.
Adopting Eq. (44) of Chap. 2 for our use here (i.e., using axes labels 1 and 2,
rather than xV and yV, respectively), we have:
8 9 2 3
< 11 >
> e
= cos2 ðhÞ sin2 ðhÞ 2cosðhÞsinðhÞ
6 7
e22 ¼4 sin2 ðhÞ cos2 ðhÞ 2cosðhÞsinðhÞ 5
>
: >
;
c12 =2 cosðhÞsinðhÞ cosðhÞsinðhÞ cos2 ðhÞ sin2 ðhÞ
8 9 8 9
< exx >
> = < exx >
> =
eyy ¼ ½T eyy
: c =2 >
> ; : c =2 >
> ;
xy xy

Similarly, the transformation of stress within a plane was discussed in Sec. 8 of
Chap. 2, and Eq. (20) of Chap. 2 can be adopted as follows:
8 9 2 3
< r11 >
> = cos2 ðhÞ sin2 ðhÞ 2cosðhÞsinðhÞ
6 7
r22 ¼ 4 sin2 ðhÞ cos2 ðhÞ 2cosðhÞsinðhÞ 5
>
: >
;
s12 cosðhÞsinðhÞ cosðhÞsinðhÞ cos2 ðhÞ sin2 ðhÞ
8 9 8 9
>
< rxx >
= >
< rxx >=
ryy ¼ ½T ryy
>
: >
; >
: >
;
sxy sxy
As pointed out in Chap. 2, the identical transformation matrix, [T ], is used to
relate strains and stresses in the two coordinate systems:
2 3
cos2 ðhÞ sin2 ðhÞ 2cosðhÞsinðhÞ
6 7
½T ¼ 4 sin2 ðhÞ cos2 ðhÞ 2cosðhÞsinðhÞ 5
cosðhÞsinðhÞ cosðhÞsinðhÞ cos2 ðhÞ sin2 ðhÞ
Inserting these transformation relationships into Eq. (14), we have:
8 9 8 9
> e
< xx > = >
< rxx >
=
½R ½T ¼ eyy ¼ ½S ½T ryy ð15Þ
>
: >
; >
: >
;
cxy =2 sxy
To simplify Eq. (15), first multiply both sides of Eq. (15) by the inverse of the
Reuter matrix, [R]1, and then by the inverse of the transformation matrix,
[T ]1:
8 9 8 9
> e
< xx > = < rxx >
> =
eyy ¼ ½T 1 ½R 1 ½S ½T ryy ð16Þ
>
: >
; >
: >
;
cxy =2 sxy
where:
2 3
1 0 0
6 7
½R 1 ¼ 4 0 1 0 5
0 0 1=2
and
2 3
6 7
½T 1 ¼ 4 sin2 ðhÞ cos2 ðhÞ 2cosðhÞsinðhÞ 5

We next extract the factor of ( 1⁄2 ) from the shear strain within the strain array
using the [R]1 matrix:
8 9 8 9
< exx >
> = < rxx >
> =
½R 1 eyy ¼ ½T 1 ½R 1 ½S ½T ryy ð17Þ
>
: >
; >
: >
;
cxy sxy
Multiplying both sides of Eq. (17) by the [R] matrix, we arrive at our
final result:
8 9 8 9
< exx >
> = < rxx >
> =
eyy ¼ ½R½T 1 ½R1 ½S½T ryy ð18Þ
>
: >
; >
: >
;
cxy sxy
Equation (18) represents an ‘‘off-axis’’ version of Hooke’s Law. That is,
it relates the strains induced in the arbitrary x–y coordinate system (exx, eyy,
and cxy) to the stresses in the same x–y coordinate system (rxx, ryy, and sxy) via
material properties referenced to the principal 1–2 coordinate system (repre-
sented by the [S] matrix) and the fiber angle h (represented by the trans-
formation matrix [T ]). Completing the matrix algebra indicated, we obtain:
8 9 8 9
< exx >
> = > < rxx >=
eyy ¼ S ryy ð19Þ
>
: >
; >
: >
;
cxy sxy
where:

S ¼ ½R½T 1 ½R1 ½S½T
Equation (19) is known as transformed reduced Hooke’s law. It is called

‘‘transformed’’ because Eq. (19) has been transformed from the 1–2 coor-
dinate system to the x–y coordinate system, and ‘‘reduced’’ because we have
reduced the allowable state of stress from three dimensions to two dimen-
sions [i.e., Eq. (19) is only valid for a plane stress state]. Matrix [S ] is called
the transformed reduced compliance matrix.* In expanded form, Eq. (19) is
written as:
8 9 2 38 9
< exx >
> = S11 S12 S16 > < rxx >=
6 7
eyy ¼ 4 S21 S22 S26 5 ryy ð20Þ
>
: >
; >
: >
;
cxy S61 S62 S66 sxy
* In common parlance, the [

S] matrix is often called the ‘‘S-bar’’ matrix.

where:
S11 ¼ S11 cos4 h þ ð2S12 þ S66 Þcos2 h sin2 h þ S22 sin4 h

S12 ¼ S21 ¼ S12 cos4 h þ sin4 h þ ðS11 þ S22 S66 Þcos2 h sin2 h
S16 ¼ S61 ¼ ð2S11 2S12 S66 Þcos3 h sin h ð2S22 2S12 S66 Þ cos h sin3 h
S22 ¼ S11 sin4 h þ ð2S12 þ S66 Þcos2 h sin2 h þ S22 cos4 h
S26 ¼ S62 ¼ ð2S11 2S12 S66 Þcos h sin3 h ð2S22 2S12 S66 Þcos3 h sin h

S66 ¼ 2ð2S11 þ 2S22 4S12 S66 Þ cos2 h sin2 h þ S66 cos4 h þ sin4 h
ð21Þ
Three important observations should be made regarding the trans-
formed reduced compliance matrix. First, we have retained the original
subscripts used in our earlier discussion of 3-D states of stress and strain.
For example, the term that appears in the (3,3) position of the transformed
reduced compliance matrix is labeled S66 . Secondly, the [S ] matrix is
symmetric. Therefore, S 21 ¼ S 12 , S 61 ¼ S 16 , and S 62 ¼ S 26 , as indicated in
Eq. (21). Third, the [S] matrix is fully populated. That is, (in general) none of
the terms within the [S] matrix equal zero, in contrast to the [S] matrix where
four off-diagonal terms equal zero [see Eq. (2a)]. This simply reveals the
anisotropic nature of unidirectional composites. Following the convention
adopted earlier, a unidirectional composite laminate referenced to the princi-
pal 1–2 coordinate system is referred to as an orthotropic (or transversely
isotropic) material, whereas if the same material is referenced to an arbitrary
(nonprincipal) x–y coordinate system, it is called an anisotropic material.
Since neither S 16 nor S 26 is equal to zero for an anisotropic composite, a
coupling exists between shear stress and normal strains. That is, a shear stress
sxy will cause normal strains exx and eyy to occur, as indicated by Eq. (20). This
coupling does not occur for orthotropic or transversely isotropic composites
(or for that matter, for isotropic materials).
Let us now include thermal and moisture strains in the transformed,
reduced form of Hooke’s law. In the 1–2 coordinate system, in-plane thermal
strains are given by:
8 T 9 8 9
>
> e11 >
> >
> a11 >
>
< = < =
eT22 ¼ DT a22 ð22Þ
>
> > > >
: T > ; >
: >
;
c12 0
As has been previously discussed, in the 1–2 coordinate system, a
uniform change in temperature does not produce a shear strain, i.e., a12=0.
Now, thermally induced strains can be transformed from one coordinate

system to another in exactly the same way as mechanically induced strains are
transformed. That is, we can relate thermal strains in the 1–2 coordinate
system to the x–y coordinate system using the transformation matrix:
8 9 2 3
>
> eT11 >> cos 2
ðh Þ sin2
ð h Þ 2cos ðh Þsin ð h Þ
< = 6 7
eT22 ¼64 sin2 ðhÞ cos2 ðhÞ 2cosðhÞsinðhÞ 7 5
>
> >
: T > ;
c12 =2 cosðhÞsinðhÞ cosðhÞsinðhÞ cos2 ðhÞ sin2 ðhÞ
8 T 9
>
> e >
< xx > =
eTyy
>
> >
: T > ;
cxy =2
Inverting this expression, we have:

8 T 9 8 9
>
> exx > > >
> eT >
< = < 11 > =
eTyy ¼ ½T 1 eT22
>
> > > >
: T > ; : >
> ;
cxy =2 0
2 3
6 7
¼64 sin ðhÞ
2
cos2 ðhÞ 2cosðhÞsinðhÞ 7 5
8 T 9
>
> e >
< 11 > =
eT22 ð23Þ
>
> >
: T > ;
c12 =2
Substituting Eq. (22) in this result, completing the matrix multiplication
indicated, and simplifying the resulting expressions, we obtain:
8 9 8 9
>
> eT > > axx >
< xx > = >
< >
=
eTyy ¼ DT ayy ð24Þ
>
> > > >
: cT > ; >
: >
;
xy axy
where:
axx ¼ a11 cos2 ðhÞ þ a22 sin2 ðhÞ
ayy ¼ a11 sin2 ðhÞ þ a22 cos2 ðhÞ ð25Þ
axy ¼ 2cosðhÞsinðhÞða11 a22 Þ

In Sec. 6, we will define the properties of a unidirectional composite laminate
when referenced to an arbitrary x–y coordinate system. As further discussed
there, Eq. (25) defines the effective coefficients of thermal expansion for an
anisotropic composite laminate. Note that for anisotropic composites, a
coupling exists between a uniform change in temperature and shear strain.
That is, a change in temperature causes shear strain cTxy as well as normal
T
strains exx and eTyy.
In the 1–2 coordinate system, the in-plane strains caused by a uniform
change in moisture content are given by:
8 M9 8 9
< e11 >
> = < 11 >
> b
=
eM ¼ DM b ð26Þ
: M>
> 22
; >
:
22
>
;
c12 0
As was the case for thermal strains, we wish to express strains induced by DM
in an arbitrary x–y coordinate system. Using the identical procedure as
before, moisture strains in the x–y coordinate system are given by:
8 M9 8 9
> e b
< xx > = < xx >
> =
eM ¼ DM b yy ð27Þ
: M>
> yy
; >
: >
;
cxy bxy
where:
bxx ¼ b11 cos2 ðhÞ þ b22 sin2 ðhÞ

byy ¼ b11 sin2 ðhÞ þ b22 cos2 ðhÞ ð28Þ
bxy ¼ 2cosðhÞsinðhÞðb11 b22 Þ
In Sec. 6, we will define the effective properties of a unidirectional
composite laminate when referenced to an arbitrary x–y coordinate system.
As further discussed there, Eq. (28) defines the effective coefficients of
moisture expansion for an anisotropic composite laminate.
We can now calculate the strains induced by the combined effects of
stress, a uniform change in temperature, and a uniform change in moisture
content. Specifically, adding Eqs. (20), (24), and (27) together, we obtain:
8 9 2 38 r 9 8 9 8 9
> exx > S16 > xx > > axx > > bxx >
>
< >
= S11 S12 >
< >
= >
< >
= >
< >
=
6
eyy ¼ 4 S12 7
S22 S26 5 ryy þ DT ayy þ DM byy ð29Þ
>
> >
> >
> >
> >
> >
> >
> >
>
: ; S16 S26 S66 : ; : ; : ;
cxy sxy axy bxy
Equation (29) allows the calculation of the in-plane strains induced by
any combination of in-plane stresses, a uniform change in temperature, and a

uniform change in moisture content. If instead we have measured the total
strains induced by a known DT, DM, and unknown in-plane stresses, we can
calculate the stresses that caused these strains by inverting Eq. (29) to obtain:
8 9 2 38 exx DTaxx DMb 9
>
> rxx >
> Q Q Q >
> xx >
>
< = 11 12 16 < =
6 7 e DTa DMb
ryy ¼ 4 Q12 Q22 Q26 5 yy yy yy ð30Þ
>
> >
> >
> >
>
: ; : ;
sxy Q16 Q26 Q66 cxy DTaxy DMbxy
where:
Q11 ¼ Q11 cos4 h þ 2ðQ12 þ 2Q66 Þcos2 h sin2 h þ Q22 sin4 h

Q12 ¼ Q21 ¼ Q12 cos4 h þ sin4 h þ ðQ11 þ Q22 4Q66 Þcos2 h sin2 h
Q16 ¼ Q61 ¼ ðQ11 Q12 2Q66 Þcos3 h sin h

ðQ22 Q12 2Q66 Þ cos h sin3 h
Q22 ¼ Q11 sin4 h þ 2ðQ12 þ 2Q66 Þcos2 h sin2 h þ Q22 cos4 h
Q26 ¼ Q62 ¼ ðQ11 Q12 2Q66 Þcos h sin3 h

ðQ22 Q12 2Q66 Þcos3 h sin h

Q66 ¼ ðQ11 þ Q22 2Q12 2Q66 Þcos2 h sin2 h þ Q66 cos4 h þ sin4 h
ð31Þ

The [ Q] matrix is called the transformed, reduced stiffness matrix.* This name
again reminds us that we have reduced our analysis to the 2-D plane stress
case, and that we have transformed Hooke’s law from the 1–2 coordinate

system to an arbitrary x–y coordinate system. Note that the [ Q] matrix is (in
general) fully populated. This reflects the anisotropic nature of unidirectional
composites when referenced to a nonprincipal material coordinate system.

Also, the [ Q] matrix is symmetric, so that Q21 ¼ Q12 , Q61 ¼ Q16 , and Q62 ¼

Q26 , as indicated in Eq. (31).
The reader should note that the functional form of the equations that
define the elements of the transformed reduced stiffness matrix, i.e., Eq. (31),
is not identical to the functional form of the equations defining the elements of
the transformed reduced compliance matrix, Eq. (21). That is, Eq. (31) cannot
be transformed into Eq. (21) by a simple substitution of S11 for Q11, S12 for
Q12, S22 for Q22, etc. This difference in functional form is due to the fact that
we have defined both the stiffness and compliance matrices in terms of
* In common parlance, the [

Q] matrix is often called the ‘‘Q-bar’’ matrix.

engineering shear strains, rather than tensoral shear strains. The use of Eqs.
(29) and (30) to solve simple problems involving off-axis composite laminates
is demonstrated in Example Problems 4 to 6.
Example Problem 4
Determine the strains induced in the off-axis graphite–epoxy composite sub-
jected to the in-plane stresses shown in Fig. 4. Assume that DT=DM=0 and
use the material properties listed in Table 3 of Chap. 3.
Solution. This problem is analogous to Example Problem 1 except that we
are now considering the behavior of an off-axis composite. The magnitude of
each stress component is indicated in Fig. 4. Based on the sign conventions
reviewed in Sec. 5 of Chap. 2, the algebraic sign of each stress component is:
rxx ¼ 200 MPa ryy ¼ 30 MPa sxy ¼ 50 MPa
Fiber angles are measured from the +x-axis to the +1-axis (or equivalently,
from the +y-axis to the +2-axis). In accordance with the right-hand rule, the
fiber angle in Fig. 4 is algebraically positive: h=+30j.
Strains are calculated using Eq. (29). Since DT=DM=0, we have:
8 9 2 38 r 9
>
> exx >
> S S S >
> xx >
>
< = 11 12 16 < =
6
eyy ¼ 4 S12 S22 S26 5 ryy 7
>
> >
> >
> >
>
: ; : ;
cxy S16 S26 S66 sxy
Figure 4 A 30j off-axis composite subjected to in-plane stresses (magnitudes

of in-plane stresses shown).

Recall that the elements of the reduced compliance matrix, the [S] matrix,
were calculated as a part of Example Problem 1. Each term within the
transformed reduced compliance matrix, the [S ] matrix, can therefore be
calculated via a straightforward application of Eq. (21). The calculation of
S 11, for example, proceeds as follows:
S11 ¼ S11 cos4 h þ ð2S12 þ S66 Þcos2 h sin2 h þ S22 sin4 h

S11 ¼ 5:88 1012 cos4 ð30B Þ

þ 2 1:76 1012 þ 76:9 1012 cos2 ð30B Þsin2 ð30B Þ

þ 100:0 1012 sin4 ð30B Þ

S11 ¼ 23:32 1012 Pa1
Calculating the remaining elements of the S matrix in similar fashion and

applying Eq. (29), we find:
8 9 2 3
< xx >
> e
= 23:32 1012 4:327 1012 33:72 1012
6
eyy ¼ 4 4:327 1012 7
70:38 1012 47:79 1012 5
:c >
> ;
xy 33:72 1012 47:79 1012 101:3 1012
8 9
> < 200 106 >
=
1
30 106 ðPaÞ
Pa > : >
;
50 106

8 9 8 9
>
< exx > < 6220 Am=m >
= > =
eyy ¼ 1144 Am=m
>
: >
; >
: >
;
cxy 10375 Arad
Example Problem 5
Determine the strains induced in the off-axis graphite–epoxy composite
subjected to (a) the in-plane stresses shown in Fig. 4, (b) a decrease in
temperature DT=155jC, and (c) an increase in moisture content
DM=0.5%. Use the material properties listed in Table 3 of Chap. 3.
Solution. This problem involves three different mechanisms that contribute
to the total strain induced in the laminate: the applied stresses, the temper-

ature change, and the change in moisture content. The total strains induced by
all three mechanisms can be calculated using Eq. (29):
8 9 2 38 9 8 9 8 9
> exx > S11 S12 S16 > rxx > > axx > > bxx >
>
< >
= 6 > > > > > >
7< = < = < =
eyy ¼ 6
4 S12 S22 7
S26 5 ryy þ DT ayy þ DM b yy
>
> >
> >
> >
> >
> >
> >
> >
>
: ; : ; : ; : ;
cxy S16 S26 S66 s xy a xy bxy
Numerical values for the stresses and transformed reduced compliance matrix
are given in Example 4. The linear thermal and moisture expansion coef-
ficients for graphite–epoxy, referenced to the x–y coordinate system, are
calculated using Eqs. (25) and (28), respectively.
The thermal expansion coefficients are:

axx ¼ a11 cos2 ðhÞ þ a22 sin2 ðhÞ ¼ 0:9 106 cos2 ð30jÞ

þ 27 106 sin2 ð30jÞ
axx ¼ 6:1 Am=m=o C

ayy ¼ a11 sin2 ðhÞ þ a22 cos2 ðhÞ ¼ 0:9 106 sin2 ð30jÞ

þ 27 106 cos2 ð30jÞ
ayy ¼ 20:0 Am=m=o C
axy ¼ 2 cosðhÞsinðhÞða11 a22 Þ ¼ 2 cosð30jÞsinð30jÞ

0:9 106 27 106
axy ¼ 24:2 Arad=o C
The moisture expansion coefficients are:

bxx ¼ b11 cos2 ðhÞ þ b22 sin2 ðhÞ ¼ 150 106 cos2 ð30jÞ

þ 4800 106 sin2 ð30jÞ
bxx ¼ 1313 Am=m=%M

byy ¼ b11 sin2 ðhÞ þ b22 cos2 ðhÞ ¼ 150 106 sin2 ð30jÞ

þ 4800 106 cos2 ð30jÞ
byy ¼ 3638 Am=m=%M
bxy ¼ 2 cosðhÞsinðhÞðb11 b22 Þ ¼ 2 cosð30jÞsinð30jÞ

150 106 4800 106
bxy ¼ 4027 Arad=%M

Hence, Eq. (29) becomes:
8 9 2 3
> 23:32 1012 4:327 1012 33:72 1012
< exx >=
6
eyy ¼ 4 4:327 1012 7
70:38 1012 47:79 1012 5
>
: >
;
cxy 33:72 1012 47:79 1012 101:3 1012
8 9 8 9
> 200 106 > > 6:1 106 >
< = < 6
=
30 106 þ ð155Þ 20:0 10 þ ð0:5Þ
>
: >
; : 24:2 106 >
> ;
50 10 6
8 6
9
< 1313 10 >
> =
3638 106
>
: >
;
4027 106
8 9 8 9 8 9 8 9
>
> exx >> > 6220 Am=m > > 946 Am=m > > 657 Am=m >
< = > < = >
> < = >
> < >
=
eyy ¼ 1144 Am=m þ 3100 Am=m þ 1819 Am=m
>
> > > > > > > >
:c > ; > : ; >
> : ; >
> : >
;
xy 10375 Arad 3751 Arad 2014 Arad
8 9
>
> 5931 Am=m > >
< =
¼ 137 Am=m
>
> >
>
: ;
8638 Arad
An implicit assumption in this problem is that the composite is free to expand
or contract, as dictated by changes in temperature and/or moisture content.
Consequently, neither DT nor DM affects the state of stress, but rather affects
only the state of strain. Conversely, if the composite is not free to expand or
contract, then a change in temperature and/or moisture content does con-
tribute to the state of stress, as illustrated in Example Problem 6.
Example Problem 6
A thin off-axis graphite–epoxy composite laminate is firmly mounted within
an infinitely rigid square frame, as shown in Fig. 5. The coefficients of thermal
and moisture expansion of the rigid frame equal zero.
The composite is initially stress-free. Subsequently, however, the com-
posite/frame assembly is subjected to a decrease in temperature DT=155jC
and an increase in moisture content DM=0.5%. Determine the stresses
induced in the composite by this change in temperature and moisture content.
Ignore the possibility that the thin composite will buckle if compressive
stresses occur.

Figure 5 Off-axis composite laminate mounted within an infinitely rigid frame.
Solution. As discussed in Example Problem 3, the situation described in this

problem is very often encountered in composite materials, despite somewhat
unrealistic assumptions regarding an ‘‘infinitely rigid’’ frame with zero
thermal expansion coefficients. Since the frame is ‘‘infinitely rigid’’ and is
made of a material whose thermal and moisture expansion coefficients equal
zero, the frame will retain its original shape. Since the off-axis composite is
‘‘firmly mounted’’ within the frame, the composite does not change shape,
although changes in temperature and moisture content have occurred.
Consequently, the total strains experienced by the composite equal zero:
8 9 8 9
> 0
< exx > < >
= > =
eyy ¼ 0
:c >
> : >
; > ;
xy 0
The stresses induced can be calculated using Eq. (30):

8 9 2 38 exx DTaxx DMbxx 9
< rxx >
> = Q11 Q12 Q16 >< >
=
ryy ¼ 6
4 Q12

Q22
7
Q26 5 eyy DTayy DMbyy
>
: >
; >
: >
;
sxy Q16 Q26 Q66 cxy DTaxy DMbxy
Recall that the elements of the reduced stiffness matrix, the [ Q] matrix, were
calculated as a part of Example Problem 3. Each term within the transformed

reduced stiffness matrix, the [ Q ] matrix, can therefore be calculated via a

straightforward application of Eq. (31). Calculation of Q11 , for example,
proceeds as follows:

Q11 ¼ Q11 cos4 h þ 2ðQ12 þ 2Q66 Þ cos2 h sin2 h þ Q22 sin4 h

Q11 ¼ 170:9 109 cos4 ð30jÞ

þ 2 3:016 109 þ 2 13:0 109 cos2 ð30jÞsin2 ð30jÞ

þ 10:05 109 sin4 ð30jÞ

Q11 ¼ 107:6 109 Pa

Calculating the remaining elements of the [ Q] matrix in similar fashion, and
using the thermal and moisture expansion coefficients referenced to the x–y
coordinate system (calculated as a part of Example Problem 5), we find:
8 9 2 3
> r 107:6 109 26:1 109 48:1 109
< xx >=
6 7
ryy ¼ 4 26:1 109 27:2 109 21:5 109 5
>
: >
;
sxy 48:1 109 21:5 109 36:0 109
8 9
> 0 ð155Þ 6:1 106 ð0:5Þ 1313 106 >
< =
0 ð155Þ 20:0 106 ð0:5Þ 3638 106
>
: >
;
0 ð155Þ 24:2 106 ð0:5Þ 4027 106
8 9 8 9
> r 19:0 MPa >
< xx >= >
< =
ryy ¼ 5:04 MPa
>
: >
; >
: >
;
sxy 21:1 MPa
3 CALCULATING TRANSFORMED PROPERTIES USING

MATERIAL INVARIANTS
The stresses and strains in a unidirectional composite referenced to an
arbitrary x–y coordinate system may be related using either Eq. (29) or Eq.
(30). Equation (29) involves the use of the transformed reduced compliance
matrix, [S ], and individual elements within the [S ] matrix are calculated in
accordance with Eq. (21). Alternatively, Eq. (30) involves the use of the

transformed reduced stiffness matrix, [ Q], and individual elements of the [ Q]
matrix are calculated in accordance with Eq. (31).

Now, both Eqs. (21) and (31) involve trigonometric functions raised to a
power (i.e., sin4 h, cos4 h, cos h sin3 h, etc.). These equations can be simplified
somewhat through the use of the following trigonometric identities:
1
sin4 h ¼ ð3 4 cos 2h þ cos 4hÞ
8
1
cos4 h ¼ ð3 þ 4 cos 2h þ cos 4hÞ
8
1
cos h sin3 h ¼ ð2 sin 2h sin 4hÞ ð32Þ
8
1
cos2 h sin2 h ¼ ð1 cos 4hÞ
8
1
cos3 h sin h ¼ ð2 sin 2h þ sin 4hÞ
8
For example, substituting these identities into Eq. (21) and simplifying
result in:
S11 ¼ US1 þ US2 cos 2h þ US3 cos4 h
S12 ¼ S21 ¼ US4 US3 cos 4h
S16 ¼ S61 ¼ US2 sin 2h þ 2US3 sin 4h ð33Þ
S22 ¼ US1 US2 cos 2h þ US3 cos 4h
S26 ¼ S62 ¼ US2 sin 2h 2US3 sin 4h
S66 ¼ US5 4US3 cos 4h
The terms UiS which appear in Eq. (33) are called compliance invariants and
are defined as follows:
1
US1 ¼ ð3S11 þ 3S22 þ 2S12 þ S66 Þ
8
1
US2 ¼ ðS11 S22 Þ
2
1
US3 ¼ ðS11 þ S22 2S12 S66 Þ ð34Þ
8
1
US4 ¼ ðS11 þ S22 þ 6S12 S66 Þ
8
1
US5 ¼ ðS11 þ S22 2S12 þ S66 Þ
2
The superscript S is used to indicate that these quantities are calculated using
members of the reduced compliance matrix [S]. They are called compliance

invariants because they define the elements of the compliance matrix that are
independent of coordinate system. In this sense, compliance invariants are
analogous to stress and strain invariants, which were discussed in Secs. 6 and
11 of Chap. 2, respectively.
In a similar manner, substituting the trigonometric identities listed as
Eq. (32) into Eq. (31) and simplifying result in:
Q11 ¼ UQ Q Q
1 þ U2 cos 2h þ U3 cos 4h
Q12 ¼ Q21 ¼ UQ Q
4 U3 cos 4h
1
Q16 ¼ Q61 ¼ UQ sin 2h þ UQ3 sin 4h ð35Þ
2 2
Q22 ¼ UQ Q Q
1 U2 cos 2h þ U3 cos 4h
1
Q26 ¼ Q62 ¼ UQ sin 2h UQ3 sin 4h
2 2
Q66 ¼ UQ Q
5 U3 cos 4h
where the stiffness invariants, UiQ, are defined as:

1
UQ
1 ¼ ð3Q11 þ 3Q22 þ 2Q12 þ 4Q66 Þ
8
1
UQ
2 ¼ ðQ11 Q22 Þ
2
1
UQ
3 ¼ ðQ11 þ Q22 2Q12 4Q66 Þ ð36Þ
8
1
UQ
4 ¼ ðQ11 þ Q22 þ 6Q12 4Q66 Þ
8
1
UQ
5 ¼ ðQ11 þ Q22 2Q12 þ Q66 Þ
8
The superscript Q is used to indicate that these quantities are calculated using
members of the reduced stiffness matrix [ Q]. The reader should note that the
functional forms of the stiffness invariants defined in Eq. (36) are not identical
to those of the compliance invariants defined in Eq. (34). The difference in
functional form can be traced to the use of engineering shear strain rather
than tensoral shear strain.
A comparison of Eqs. (21) and (33) reveals that the use of compliance
invariants does indeed simplify the calculation of elements of the [S] matrix,
although, mathematically, the two equations are entirely equivalent. Sim-
ilarly, a comparison of Eqs. (31) and (35) shows that the use of the stiffness
invariants simplifies the calculation of the terms within the [Q] matrix. At this
point, this simplification may seem to be a trivial matter since, in practice,

elements of the [S ] and [Q] matrices are calculated with the aid of a digital
computer and do not require hand calculation. However, in Chap. 6, it will be
shown that in special circumstances, the use of stiffness invariants leads to a
convenient method of transforming the stiffness of multi-angle composite
laminates from one coordinate system to another. Hence, in these special
cases, the use of the invariant formulation is advantageous. Further discus-
sion of the invariant approach will be deferred to Chap. 6.
Example Problem 7
Problem. Use the material invariants [i.e., Eqs. (33) and (34)] to calculate the
transformed reduced compliance matrix for a 30j graphite–epoxy laminate.
Use the material properties listed in Table 3 of Chap. 3.
Solution. From Example Problem 1, the reduced compliance matrix for this
material system is:
2 3 2 3
S11 S12 0 5:88 1012 1:76 1012 0
4 S12 S22 0 5 ¼ 4 1:76 1012 100:0 1012 5 1
0
Pa
0 0 S66 0 0 76:9 1012
The compliance invariants may be calculated using these values and
Eq. (34):
1 1h
US1 ¼ ð3S11 þ 3S22 þ 2S12 þ S66 Þ ¼ 3 5:88 1012
8 8
i
12
þ3 100:0 10 þ 2 1:76 1012 þ 76:9 1012
¼ 48:9 1012
1 1
US2 ¼ ðS11 S22 Þ ¼ 5:88 1012 100:0 1012
2 2
¼ 47:1 1012
1 1h
US3 ¼ ðS11 þ S22 2S12 S66 Þ ¼ 5:88 1012 þ 100:0 1012
8 8
i
2 1:76 1012 76:9 1012 ¼ 4:06 1012
1 1h
US4 ¼ ðS11 þ S22 þ 6S12 S66 Þ ¼ 5:88 1012 þ 100:0 1012
8 8
i
þ6 1:76 1012 76:9 1012 ¼ 2:296 1012
1 1h
US5 ¼ ðS11 þ S22 2S12 þ S66 Þ ¼ 5:88 1012 þ 100:0 1012
2 2 i

2 1:76 1012 þ 76:9 1012 ¼ 93:17 1012

Using the first equation of Eq. (33) and setting h=30j:

S11 ¼ US1 þ US2 cos 2h þ US3 cos 4h ¼ 48:9 1012

þ 47:1 1012 cosð60jÞ þ 4:06 1012 cosð120jÞ
¼ 23:3 1012 Pa1

Similarly, using the second equation of Eq. (33) and setting h=30j:
S12 ¼ S21 ¼ US4 US3 cos 4h

¼ 2:296 1012 4:06 1012 cosð120jÞ
¼ 4:33 1012 Pa1

The additional terms within the [S] matrix are calculated using the rest of
Eq. (33). A summary of our results is:
2 3
23:3 1012 4:33 1012 33:7 1012
6 7 1
½S ¼ 6
4 4:33 10
12
70:4 1012 47:8 1012 7 5 Pa
33:7 1012 47:8 1012 101:3 1012
Note that the [S] matrix is identical to that calculated in Example Problem 4.
4 EFFECTIVE ELASTIC PROPERTIES OF A UNIDIRECTIONAL

COMPOSITE LAMINATE
The definitions of common engineering material properties were reviewed in
Chap. 3. In this section, these concepts will be used to define the ‘‘effective’’
properties of a unidirectional composite laminates referenced to an arbitrary
x–y coordinate system.
4.1 Effective Properties Relating Stress to Strain

Let us first consider the elastic properties measured during uniaxial tests.
Consider the unidirectional composite laminate subjected to a uniaxial stress
rxx, as shown in Fig. 6. The strains induced in this laminate can be determined
using Eq. (29). Assuming that DT=DM=0 (and hence that thermal and
moisture strains are zero), and also noting that by definition, ryy=sxy=0, Eq.
(29) becomes for this case:
8 9 2 38 9
e
< xx >
> = S11 S12 S16 <>rxx>
=
6 7 0
eyy ¼ 4 S12 S22 S26 5
: >
> ; : >
> ;
cxy S16 S26 S66 0

Figure 6 Unidirectional composite laminate subjected to uniaxial stress rxx.
In-plane strains caused by uniaxial stress rxx are therefore given by:
exx ¼ S11 rxx ð37aÞ
eyy ¼ S12 rxx ð37bÞ
cxy ¼ S16 rxx ; ð37cÞ
In Sec. 2 of Chap. 3, Young’s modulus was defined as ‘‘the normal stress
rxx divided by the resulting normal strain exx, with all other stress components
equal zero.’’ Applying this definition to the unidirectional laminate shown in
Fig. 6, Young’s modulus in the x-direction is given by:
rxx rxx 1
Exx ¼ ¼ ¼ ð38aÞ
exx S11 rxx S11
Inserting the relation for Q11 listed in Eq. (22), we have:
1
Exx ¼ ð38bÞ
S11 cos4 h þ ð2S12 þ S66 Þ cos2 h sin2 h þ S22 sin4 h
Since each of the compliance terms (S11, S12, etc.) can also be related to the
more familiar engineering constants using Eq. (18) of Chap. 4, Young’s
modulus can also be written as:
1
Exx ¼ ð38cÞ
cos ðhÞ
4
1 2m12 sin4 h
þ cos h sin h þ
2 2
E11 G12 E11 E22

In Sec. 2 of Chap. 3, Poisson’s ratio vxy was defined as ‘‘the negative of
the transverse normal strain eyy divided by the axial normal strain exx, both of
which are induced by stress rxx, with all other stresses equal zero.’’ Poisson’s
ratio for the unidirectional laminate shown in Fig. 6 is therefore given by:
eyy S12
vxy ¼ ¼ ð39aÞ
exx S11
Using Eq. (22), this can be written as:

S12 cos4 h þ sin4 h þ ðS11 þ S22 S66 Þ cos2 h sin2 h
vxy ¼ ð39bÞ
S11 cos4 h þ ð2S12 þ S66 Þ cos2 h sin2 h þ S22 sin4 h
or equivalently, using Eq. (18) of Chap. 4:

v12 4 1 1 1
cos h þ sin h
4
þ cos2 h sin2 h
E11 E11 E22 G12
vxy ¼ ð39cÞ
cos4 ðhÞ 1 2v12 sin4 ðhÞ
þ cos 2 ðhÞ sin2 ðhÞ þ
E11 G12 E11 E22
In Sec. 2 of Chap. 3, the coefficient of mutual influence of the second kind

gxx,xy was defined as ‘‘the shear strain cxy divided by the normal strain exx, both
of which are induced by normal stress rxx, when all other stresses equal zero.’’
For a unidirectional composite laminate, gxx,xy is therefore given by:
cxy S16
gxx;xy ¼ ¼ ð40aÞ
exx S11
which may be written as:
ð2S11 2S12 S66 Þ cos3 h sin h ð2S22 2S12 S66 Þ cos h sin3 h
gxx;xy ¼
S11 cos4 h þ ð2S12 þ S66 Þ cos2 h sin 2 h þ S22 sin4 h
ð40bÞ
or equivalently:

2 2v12 1 2 2v12 1
þ cos3 ðhÞ sin ðhÞ þ cos ðhÞ sin2 ðhÞ
E11 E11 G12 E22 E11 G12
gxx;xy ¼
cos4 ðhÞ 1 2v12 sin4 ðhÞ
þ cos2 ðhÞ sin2 ðhÞ þ
E11 G12 E11 E22
ð40cÞ

An identical procedure can be employed to define the properties
measured during a uniaxial test in which only ryy is applied. In this case,
Eq. (29) becomes:
8 9 2 38 9
> 0 >
< exx >
= S11 S12 S16 > < =
eyy ¼ 6 7
4 S12 S22 S26 5 ryy
>
: >
; >
: >
;
cxy S16 S26 S66 0
In-plane strains are:
exx ¼ S12 ryy ð41aÞ
eyy ¼ S22 ryy ð41bÞ
cxy ¼ S26 ryy ð41cÞ
These strains can be used to define the Young’s modulus Eyy, Poisson’s ratio
vyx, and coefficient of mutual influence of the second kind gyy,xy:
1
Eyy ¼ ð42aÞ
S22
S12
vyx ¼ ð42bÞ
S22
S26
gyy;xy ¼ ð42cÞ
S22
If desired, these relations can be expanded in terms of compliances referenced
to the 1–2 coordinate system using Eq. (22) or written in terms of measured
engineering properties using Eq. (18) of Chap. 4.
Next, consider the effective material properties measured during a pure
shear test. A composite laminate subjected to pure shear stress sxy is shown in
Fig. 7. Assuming that DT=DM=0, Eq. (29) becomes:
8 9 2 38 9
< exx >
> = S11 S12 S16 > < 0 > =
eyy ¼ 6 4 S12 S22 S26 5 0
7
>
: >
; >
: >
;
cxy S16 S26 S66 sxy
Hence, the strains caused by a pure shear stress are given by:
exx ¼ S16 sxy ð43aÞ
eyy ¼ S26 sxy ð43bÞ
cxy ¼ S66 sxy ð43cÞ
In Sec. 2 of Chap. 3, the shear modulus was defined as ‘‘the shear stress
sxy divided by the resulting shear strain cxy, with all other stress components

Figure 7 Unidirectional composite subjected to pure shear stress sxy.
equal zero.’’ Applying this definition to the laminate shown in Fig. 7, the shear
modulus referenced to the x–y coordinate axes is given by:
sxy 1
Gxy ¼ ¼ ð44Þ
cxy S66
As before, this expression can be expanded in terms of compliances
referenced to the 1–2 coordinate system using Eq. (21) or written in terms of
measured engineering properties using Eq. (18) of Chap. 4.
The coefficient of mutual influence of the first kind gxy,xx (or gxy,yy) was
defined as ‘‘the normal strain exx (or eyy) divided by the shear strain cxy, both
of which are induced by shear stress sxy, when all other stresses equal zero.’’
For a unidirectional composite laminate, the coefficient of mutual influence of
the first kind gxy,xx is therefore given by:
exx S16
gxy;xx ¼ ¼ ð45aÞ
cxy S66
while gxy,yy is given by:
eyy S26
gxy;yy ¼ ¼ ð45bÞ
cxy S66
Chentsov coefficients were defined in Sec. 2 of Chap. 3 as ‘‘the shear
strain cxz (or cyz) divided by the shear strain cxy, both of which are induced by

shear stress sxy, with all other stresses equal zero.’’ For a thin composite
laminate, the principal material coordinate system lies within the plane of the
laminate, and hence there is no coupling between a shear stress acting within
the x–y plane (sxy) and out-of-plane shear strains (cxz or cyz). Consequently,
Chentsov coefficients are always equal to zero for thin composite laminates.
4.2 Effective Properties Relate Temperature or Moisture

Content to Strain
As discussed in Sec. 3 of Chap. 3, the linear coefficients of thermal expansion
are measured by determining the strains induced by a uniform change in
temperature and forming the following ratios:
eTxx eTyy cTxy

axx ¼ ayy ¼ axy ¼ ð46Þ
DT DT DT
The superscript T is included as a reminder that the strains involved are those
caused by a change in temperature only. The strains induced in a unidirec-
tional laminate subjected to a change in temperature can be determined using
Eq. (29). Assuming that rxx=ryy=sxy=DM=0, Eq. (29) becomes for this
case:
8 9 8 9
>
< exx >= >
< axx > =
eyy ¼ DT ayy ð47Þ
>
:c > ; >
: >
;
xy a xy
Hence, the thermal expansion coefficients for a unidirectional laminate are

given by Eq. (25), repeated here for convenience:

ayy ¼ a11 sin2 ðhÞ þ a22 cos2 ðhÞ
axy ¼ 2 cosðhÞ sinðhÞða11 a22 Þ
Similarly, the linear coefficient of moisture expansion is measured by

determining the strains induced by a uniform change in moisture content and
forming the following ratios:
eM eM cM
bxx ¼ xx
byy ¼
yy
bxy ¼
xy ð48Þ
DM DM DM
The superscript M is included as a reminder that the strains involved are those
caused by a change in moisture only. The strains induced in a unidirectional
laminate subjected to a change in moisture content can be determined using

Eq. (29). Assuming that rxx=ryy=sxy=DT=0, Eq. (29) becomes for this
case:
8 9 8 9
> e b
< xx >= < xx >
> =
eyy ¼ DM byy ð49Þ
>
: >
; >
: >
;
cxy bxy
Hence, the moisture expansion coefficients for a unidirectional laminate are
given by Eq. (28), repeated here for convenience:
byy ¼ b11 sin2 ðhÞ þ b22 cos2 ðhÞ
bxy ¼ 2 cosðhÞ sinðhÞðb11 b22 Þ
Example Problem 8
Plot the effective properties listed below for a unidirectional hj graphite–
epoxy laminate, for all fiber angles ranging 0j V h V 90j:
(a) Effective Young’s moduli, Exx and Eyy.
(b) Effective Poisson’s ratio, vxy and vyx.
(c) Effective shear modulus Gxy.
(d) Coefficients of mutual influence of the first kind, gxy,xx and gxy,yy.
(e) Coefficients of mutual influence of the second kind, gxx,xy and gyy,xy.
(f) Coefficients of thermal expansion, axx, ayy, and axy.
(g) Coefficients of moisture expansion, bxx, byy, and bxy.
Use the material properties listed in Table 3 of Chap. 3.
Solution. Plots of the effective elastic properties for unidirectional hj graph-
ite–epoxy laminates are presented in Figs. 8–14.
5 FAILURE OF UNIDIRECTIONAL COMPOSITES REFERENCED

TO THE PRINCIPAL MATERIAL COORDINATE SYSTEM
Fundamental material strengths for a unidirectional composite were dis-
cussed in Sec. 5 of Chap. 3. Recall that material strengths are measured under
simple states of stress, usually either a uniaxial stress state or a pure shear
stress state. Also, high-performance fibers are often very brittle, whereas
modern polymeric matrices are fairly ductile. Consequently, most fiber-
reinforced polymeric composites exhibit brittle behavior in the 1-direction,
qualitatively similar to Fig. 12(a) of Chap. 3, but relatively ductile behavior in
the 2- and 3-directions, qualitatively similar to Fig. 12(b) and c of Chap. 3. A
brief summary of experimental methods used to measure properties in the 1–2
plane is provided in Appendix B.

Figure 8 A plot of the effective Young’s moduli Exx and Eyy for unidirectional
graphite–epoxy laminates and fiber angles ranging over 0j V h V 90j.
Figure 9 A plot of the effective Poisson ratios vxy and vyx for unidirectional

Figure 10 A plot of the effective shear modulus Gxy for unidirectional
Figure 11 A plot of the effective coefficients of mutual influence of the first

kind gxy,xx and gxy,yy for unidirectional graphite–epoxy laminates and fiber
angles ranging over 0j V h V 90j.

Figure 12 A plot of the effective coefficients of mutual influence of the second
kind gxx,xy and gyy,xy for unidirectional graphite–epoxy laminates and fiber
angles ranging over 0j V h V 90j.
Figure 13 A plot of the effective coefficients of thermal expansion axx, ayy, and
axy for unidirectional graphite–epoxy laminates and fiber angles ranging over
0j V h V 90j.

Figure 14 A plot of the effective coefficients of moisture expansion bxx, byy,
and bxy for unidirectional graphite–epoxy laminates and fiber angles ranging
over 0j V h V 90j.
In this section, we will discuss failure criteria that are commonly used to
predict failure of composites under general 3-D or 2-D states of stress. It will
be assumed that the composite is brittle in the 1-direction, but ductile in the
2- and 3-directions. That is, in the fiber direction, ‘‘failure’’ is assumed to
involve fracture, whereas transverse to the fibers, ‘‘failure’’ is assumed to
involved yielding, defined on the basis of a % strain offset. Both orthotropic
and transversely isotropic composites will be considered. For the orthotropic
case, failure predictions will be based on the combinations of the following
fundamental material strengths:
fT
Fracture stress in the 1-direction: r11 fC
, r11 .
yT yC yT yC y y
Yield stress in the 2- and 3-directions: r22 , r22 , r33 , r33 , s12, s13,
y
s23.
If the composite is transversely isotropic, then the number of independent
material strengths involved is reduced, since in this case:
y
ryT yT
22 ¼ r33 ryC yC
22 ¼ r33 sy12 ¼ s13
Recall that all of these strengths may vary with temperature.

The need for ‘‘failure criteria’’ in engineering analysis and design is often
misunderstood. In essence, the objective of any failure criterion is to account
for potential coupling effects of individual stress components on the yielding
and/or fracture phenomenon. This statement applies to both anisotropic and
isotropic materials. To explain what is meant by the phrase ‘‘coupling effects
between individual stress components,’’ consider the two different tests of
unidirectional composite shown in Fig. 15. A composite subjected to uniaxial
stress r11 is shown in Fig. 15(a). This is, of course, the very state of stress used
fT
during the measurement of the fundamental material strength r11 (as
Figure 15 An illustration of what is meant by ‘‘coupling effects’’ of stress on

the failure phenomenon.

discussed in Sec. 5 of Chap. 3). In this case then, we do not need to invoke any
failure criterion to predict when failure occurs: failure occurs when r11 is
fT
increased to r11 , by definition. However, consider a more general state of
stress, such as the state of plane stress shown in Fig. 15(b). In this test, two
additional components of stress (r22 and s12) are applied prior to the appli-
cation of r11. It is assumed that the combination of r22 and s12 does not cause
failure prior to the application of r11. While maintaining r22 and s12 at
constant values, stress r11 is increased until failure occurs. It is for this more
general state of stress that a failure criterion is required. That is, does the
application r22 and/or s12 change the value of r11 at which failure occurs?
‘‘Coupling effects’’ refer to the fact that the application of r22 and/or s12 often
does alter the value of r11 necessary to cause failure. If the test represented by
Fig. 8(b) is conducted and r11p r11 fT
at failure, then a coupling effect has oc-
curred, i.e., the presence of r22 and s12 has changed the value of r11 necessary
to cause failure. Conversely, if the test depicted in Fig. 15(b) is performed and
fT
r11=r11 at failure, then no coupling has occurred. Experimental measure-
ments have shown that the coupling phenomenon is much more significant
in some materials than it is in others. This is unfortunate because it implies
that it is not possible to develop a ‘‘universal’’ failure criterion that can be
applied to all materials. Furthermore, there is no way of predicting a priori
whether coupling effects are pronounced for a given material or not. For
metals, the general trend is that coupling effects are less pronounced in brittle
materials (such as cast irons) than in ductile materials (such as aluminum
alloys). The question as to whether this general trend holds in the case of
composites is complicated by the fact that composites are (usually) brittle in
the fiber direction but ductile transverse to the fiber. It is generally accepted
that coupling effects do exist in composites, but at the present state of the art,
one is well advised to perform experimental measurements to evaluate the
level of coupling for each composite material system of interest.
Many failure criteria applicable to composites have been proposed in
the literature. Three of the most common will be described in the following
subsections: the maximum stress failure criterion, the Tsai–Hill failure
criterion, and the Tsai–Wu failure criterion. As will be seen, the maximum
stress criterion does not account for coupling effects, whereas potential
coupling effects are accounted for in the Tsai–Hill and Tsai–Wu criteria.
5.1 The Maximum Stress Failure Criterion

According to this criterion, a given state of stress will not cause a unidirec-
tional composite to fail if all of the following nine inequalities are satisfied:
1*rfC
11 < r11 < r11
fT ð50Þ

ðandÞ
1*ryC
22 < r22 < ryT
22
ðandÞ
1*ryC
33 < r33 < ryT
33
ðandÞ
js12 j < sy12
ðandÞ
js13 j < sy13
ðandÞ
js23 j < sy23
According to the maximum stress failure criterion, failure (i.e., fracture in the
fiber direction or yield-like behavior transverse to the fiber) is predicted strictly
on the basis of individual stress components. Thus, failure is assumed to be
independent of any coupling effects between individual stress components.
In the case of plane stress (r33=s13=s23=0), the maximum stress
failure criterion reduces to the following five inequalities:
1*rfC
11 < r11 < r11
fT
ðandÞ
1*ryC yT
22 < r22 < r22
ð51Þ
ðandÞ
js12 j < sy12
The maximum stress failure criterion is most commonly applied in the
form of Eq. (51) since in most cases, an individual composite ply can be
assumed to be in a state of plane stress.
5.2 The Tsai–Hill Failure Criterion

The von Mises yield criterion is widely used to predict yielding of isotropic
metals and metal alloys.* In 1950, Hill [2] proposed a modified version of the
* The von Mises yield criterion is also mathematically equivalent to the ‘‘octahedral shear
stress’’ and ‘‘distortional energy’’ yield criteria.

von Mises criterion for use with orthotropic metals. Subsequently, Tsai [3]
applied this method to predict failure of unidirectional polymeric composites,
and the resulting theory is now known within the polymeric composites
community as either the ‘‘Tsai–Hill’’ failure criterion or as the ‘‘quadratic’’
failure criterion. For general 3-D states of stress, the Tsai–Hill criterion
predicts that failure of an orthotropic composite will not occur if the following
inequality is satisfied:
ðr11 Þ2 ðr22 Þ2 ðr33 Þ2 ðs23 Þ2 ðs13 Þ2 ðs12 Þ2

fT 2 2 2 y 2 þ y 2 þ y 2
þ þ þ
r11 ryT ryT s23 s13 s12
22 33
2 3
6 1 1 1 7
r11 r22 4 2 þ 2 2 5
fT
r11 ryT
22 ryT
33
2 3 ð52Þ
6 1 1 1 7
r11 r33 4 2 2 þ 2 5
fT
r11 ryT
22 ryT
33
2 3
6 1 1 1 7
r22 r33 4 2 þ 2 þ 2 5 < 1
fT
r11 yT yT
r22 r33
In the case of plane stress conditions (r33=s13=s23=0), the Tsai–Hill

criterion reduces to:
2 3
ðr11 Þ2 ðr22 Þ2 ðs12 Þ2 6 1 1 1 7
fT 2 þ 2 þ y 2 r11 r22 4 fT 2 þ 2 2 5 < 1 ð53Þ
r11 yT
r22 s12 r11 yT
r22 yT
r33
It is interesting to note that according to the Tsai–Hill failure criterion,

failure of orthotropic composites is sensitive to the out-of-plane strength term
yT
(r33 ), even for plane stress conditions.
yT yT
If the composite is transversely isotropic (that is, if r33 =r22 ), then Eq.
(53) reduces to:
ðr11 Þ2 ðr22 Þ2 ðs12 Þ2 r11 r22

fT 2 þ 2 þ y 2 fT 2 < 1 ð54Þ
r11 ryT s12 r11
22

A potential advantage of the Tsai–Hill failure criterion is that coupling
effects between individual stress components are accounted for, in contrast
with the maximum stress failure criterion. On the other hand, most com-
posites exhibit significantly different failure strengths in tension and com-
pression (as indicated in Table 3 of Chap. 3), and so a shortcoming of the
Tsai–Hill failure criterion is that it does not directly account for these dif-
ferences. That is, an implicit assumption of the Tsai–Hill criterion (as well as
the original von Mises criterion) is that failure strengths in tension and
fT yT
compression have equal magnitudes. Hence, only tensile strengths r11 , r22 ,
yT
and r33 appear in Eqs. (53) and (54). Differences in tensile and compressive
strengths can be accounted for ‘‘artificially’’ in the Tsai–Hill criterion by
using the appropriate compressive strength if a stress component involved is
compressive. Suppose, for example, that a failure prediction is required for a
transversely isotropic composite subjected to three stress components r11,
r22, and s12, and also that r11 is tensile but r22 is compressive. In such a case,
the differences in tensile/compressive strengths can be accounted for by using
fT
the tensile strength in the 1-direction, r11 , but the compressive strength in
yC
the 2-direction, r22 .
While the Tsai–Hill criterion can be modified in this way to account for
differences in tensile and compressive strengths, it would be ideal if a failure
criterion was available that accounts for both coupling effects and differences
in tensile and compressive strengths ‘‘automatically.’’ One such criterion is
the Tsai–Wu criterion, described in the next paragraph.
5.3 The Tsai–Wu Failure Criterion

Tsai and Wu [4] developed their criterion by postulating that the strength of a
unidirectional composite can be treated mathematically as a tensoral quan-
tity, in much the same way as stress or strain tensors. For general 3-D states of
stress, the Tsai–Wu criterion predicts that failure will not occur if the
following inequality is satisfied:
X1 r11 þ X2 r22 þ X3 r33 þ X11 r211 þ X22 r222 þ X33 r233 þ X44 s223
ð55Þ
þX55 s213 þ X66 s212 þ 2X12 r11 r22 þ 2X13 r11 r33 þ 2X23 r22 r33 < 1
Most of the constants that appear in this inequality (i.e., X1, X2, X3, X11, etc.)
fT
can be determined based on fundamental strength measurements (i.e., r11 ,
fC yT yC
r11 , r22 , r22 , etc.). First, consider a uniaxial strength measurement in which
only stress r11 is applied (that is, a test in which r11 p 0, r22 =r33
=s23=s13=s12=0). Stress r11 is increased monotonically from zero until

fT
failure occurs. If r11 is tensile, then at the moment of failure, r11=r11 , and the
Tsai–Wu criterion reduces to:
fT 2
11 þ X11 r11 ¼ 1
X1 rfT
fC
Conversely, if r11 is compressive, then at the moment of failure, r11=r11
fC
(where the measured compressive strength, r11 , is treated as an algebraically
positive number), and the Tsai–Wu criterion reduces to:
fC 2
11 þ X11 r11 ¼ 1
X1 rfC
Solving for X1 and X11, we find:

1 1 1
X1 ¼ fC X11 ¼ ð56Þ
rfT
11 r 11 rfT fC
11 r11
Similarly, if failure is measured during two uniaxial stress tests in which only
r22 is applied, we find:
1 1 1
X2 ¼ X22 ¼ ð57Þ
ryT
22 ryC
22 ryT yC
22 r22
Using measurements obtained during two tests in which only r33 is applied:
1 1 1
X3 ¼ X33 ¼ ð58Þ
ryT
33 ryC
33 ryT yC
33 r33
Three additional constants are determined using measured shear strengths:

2 2 2
1 1 1
X44 ¼ X55 ¼ X66 ¼ ð59Þ
sy23 sy13 sy12
Only three coefficients remain to be determined, X12, X13, and X23. Several
methods of determining these coefficients have been suggested, but thus far,
no one technique has gained widespread acceptance. Two methods that have
been proposed will be discussed here.
Conceptually, the most straightforward approach is through the use of
additional biaxial testing. For example, X12 can be determined by conducting
a biaxial test in which r11=r22=r and r33=s23=s13=s12=0. The magnitude
of biaxial stresses (r) is increased until failure occurs (i.e., increased until
either fracture or yielding occurs). For simplicity, let us assume that failure
occurs due to yielding and denote the onset of yielding using the superscript y.
At the moment of failure then, the stresses applied are r11=r22=ry and

r33=s23=s13=s12=0. Substituting these values into the Tsai–Wu criterion
and solving for X12 results in:
1 h i
X12 ¼ 2
1 ðX1 þ X2 Þry ðX11 þ X22 Þðry Þ2 ð60Þ
2 ð ry Þ
At least conceptually, X13 and X23 can also be determined in a similar manner.
Two additional biaxial tests to failure would be required, where in one test,
r11=r33=r, and in the second test, r22=r33=r. These data would then
allow the calculation of X13 and X23, respectively. In practice, however, these
tests would be very difficult to perform. Since composites are usually quite
thin, it is especially difficult to apply well-defined out-of-plane stress compo-
nents (i.e., r33, s13, or s23). Hence, in most instances, determining X13 or X23 in
this manner is impractical.
A second approach is to assume that X12, X13, and X23 can be calculated
as follows:
1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1
X12 ¼ X11 X22 ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 fC yT yC
2 rfT
11 r11 r22 r22
X13 ¼ X11 X33 ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð61Þ
2 fC yT yC
2 rfT
11 r11 r33 r33
X23 ¼ X22 X33 ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 yT yC yT yC
2 r22 r22 r33 r33
The basis of this approach is that if Eq. (61) is enforced and isotropic
yT yC yT yC
strengths are assumed
p ffiffi
ffi
fT
(i.e., if r11 fC
=r11 =r22 =r22 =r33 =r33 =ry, and sy12
¼ sy13 ¼ sy23 ¼ ry = 3), then the Tsai–Wu criterion reduces to the original von
Mises criterion for isotropic materials. This approach holds some intellectual
appeal since it ‘‘makes sense’’ that a failure criterion proposed for use with an
orthotropic material should reduce to a well-known isotropic yield criterion if
isotropic strengths are assumed. It is also a convenient assumption since X12,
X13, and X23 are now calculated using fundamental strength data and hence
the need to perform any additional testing is avoided. However, there is little
data available to assess the validity of these assumptions and so the accuracy
of failure predictions obtained using this approach is unknown.
As discussed earlier, in most practical applications, composites are
subjected to a state of plane stress within the 1–2 plane. In this case, the
Tsai–Wu criterion reduces to:
X1 r11 þ X2 r22 þ X11 r211 þ X22 r222 þ X66 s212 þ 2X12 r11 r22 < 1 ð62Þ
Hence, in the plane stress case, six constants are involved, five of which can
fT fC yT
be calculated using readily available strength data (r11 , r11 , r22 , etc.). Only
one problematic coefficient remains, X12. This term can be determined using

an off-axis specimen (which is, in effect, a biaxial test). For example, suppose a
uniaxial stress rxx is applied to a unidirectional composite specimen in which
the fibers are oriented at h=45j with respect to the direction of loading.
Under these conditions, the stresses in the 1–2 coordinate system are easily
calculated:
8 9 2 38 9
< r11 = cos2 ð45jÞ sin2 ð45jÞ 2cosð45jÞsinð45jÞ < rxx >
> =
6 7
r22 ¼ 4 sin2 ð45jÞ cos2 ð45jÞ 2cosð45jÞsinð45jÞ 5 0
: ; >
: >
;
s12 cosð45jÞsinð45jÞ cosð45jÞsinð45jÞ cos2 ð45jÞ sin2 ð45jÞ 0
or
8 9 8 9 8 9
>
> r11 >
> > rxx cos2 ð45jÞ > > rxx =2 >
< = >
< >
= > < >
=
r22 ¼ rxx sin2 ð45jÞ ¼ rxx =2
>
> >
> > > > >
: ; >
: >
; > : >
;
s12 rxx cosð45jÞsinð45jÞ rxx =2
The strength of the 45j off-axis specimen is measured by increasing

stress rxx until failure occurs. Let us assume that failure occurs due to yielding
y
and denote the stress level at which failure occurs as rxx=rxx . At failure, the
y
ply stresses are r11=r22=s12=rxx/2. Substituting these stresses into Eq.
(62) and solving for X12, we find:
1
X12 ¼ y 2
4 ryxx 2ðX1 þ X2 Þ þ ryxx ðX11 þ X22 þ X66 Þ ð63Þ
2ðrxx Þ
While this example has been based on a 45j off-axis specimen, a similar
approach can be used with any hj off-axis specimen.
From an analytical standpoint, the Tsai–Wu failure criterion is an
improvement over the other two failure criteria considered. First, unlike the
maximum stress failure criterion, the coupling effects between individual
stress components are accounted for in the Tsai–Wu criterion. Second, unlike
the Tsai–Hill criterion, differences in tensile and compressive strengths are
automatically and naturally accounted for via the X1, X11, X2, X22, X3, and
X33 terms.
6 FAILURE OF UNIDIRECTIONAL COMPOSITES REFERENCED

TO AN ARBITRARY COORDINATE SYSTEM
In this section, the three failure criteria introduced in Sec. 5 will be used to
predict failure of unidirectional composites subjected to a state of plane
stress, where stress components rxx, ryy, and sxy are referenced to an arbi-
trary x–y coordinate system. There are, of course, an infinite number of
different combinations of rxx, ryy, and sxy that (collectively) define a state of

plane stress. For illustrative purposes, two simple stress states will be con-
sidered: first, a state of uniaxial stress (i.e., rxx p 0, ryy=rxy=0), and second,
a state of pure shear (sxy p 0, rxx=ryy=0).
Numerical results for a unidirectional graphite–epoxy composite will be
used to facilitate these comparisons. The following failure strengths are taken
from Table 3 of Chap. 3 and are typical for graphite–epoxy at room temper-
ature:
11 ¼ 1500 MPa
rfT ryT
22 ¼ 50 MPa sy12 ¼ 75 MPa
11 ¼ 1200 MPa
rfC ryC
22 ¼ 100 MPa
6.1 Uniaxial Stress

An off-axis composite ply subjected to a uniaxial stress rxx has been
previously shown in Fig. 6. The stresses induced in the 1–2 coordinate system
by stress rxx can be determined using Eq. (20) of Chap. 2:
8 9 2 38 9
>
> r11 >
> cos2 ðhÞ sin2 ðhÞ 2 cosðhÞsinðhÞ > > rxx >
>
< = 6 7< =
r22 ¼ 6 4 sin 2
ð h Þ cos 2
ðh Þ 2 cosð h Þsin ð h Þ 7 0
5>
>
> >
> > >
>
: ; : ;
s12 cosðhÞsinðhÞ cosðhÞsinðhÞ cos ðhÞ sin ðhÞ
2 2 0
or equivalently:
r11 ¼ rxx cos2 ðhÞ
r22 ¼ rxx sin2 ðhÞ ð64Þ
s12 ¼ rxx cosðhÞsinðhÞ
6.1.1 Maximum Stress Criterion

Substituting Eq. (64) into Eq. (50), we obtain:
rfC rfT
11
< rxx < 11
cos ðhÞ
2 cos2 ðhÞ
ðandÞ
ryC ryT
22
< rxx < 22 ð65Þ
sin ðhÞ
2
sin2 ðhÞ
ðandÞ
sy12
jrxx j <
cosðhÞsinðhÞ
According to the maximum stress criterion, failure will not occur if
these five inequalities are satisfied. The predicted tensile and compressive

fT fC
failure strengths, rxx and rxx , respectively, for a hj off-axis graphite–epoxy
laminate are therefore the smallest values returned by the following expres-
sions:*
Tensile strength:
rfT 1500 MPa

xx ¼
rfT ¼ ð66aÞ
11
cos2 ðhÞ cos2 ðhÞ
ryT 50 MPa
xx ¼
rfT ¼ ð66bÞ
22
sin2 ðhÞ sin2 ðhÞ

sy12 75 MPa
rfT
¼ ¼ ð66cÞ
xx
cosðhÞsinðhÞ cosðhÞsinðhÞ
Compressive strength:
rfC 1200 MPa

xx ¼
rfC ¼ ð67aÞ
11
cos2 ðhÞ cos2 ðhÞ
ryC 100 MPa

xx ¼
rfC ¼ ð67bÞ
22
sin ðhÞ
2
sin2 ðhÞ

sy12 75 MPa
rfC
¼ ¼ ð67cÞ
xx
cosðhÞsinðhÞ cosðhÞsinðhÞ
Note that the superscript fT or fC has been used to denote the failure strength
of the unidirectional composite. It should be understood that in this context,
‘‘failure’’ may represent fracture of the fibers or yielding of the matrix.
Equations (66a)–(66c) and (67a) (67b) (67c) were used to create the failure
envelope for a unidirectional graphite–epoxy laminate shown in Fig. 16.
Equations (66a)–(66c) and (67a) (67b) (67c) bound the ‘‘safe’’ region. The
reader should note the following:
The failure envelope shown in Fig. 16 is valid for a uniaxial state of
stress only. Specifically, Fig. 16 is valid only if:
rxx p 0
ryy ¼ rzz ¼ sxy ¼ sxz ¼ syz ¼ 0
As will be seen later, failure envelopes for other states of stress differ
substantially from Fig. 16.
* As before, compressive strength is treated as an algebraically positive number.

Figure 16 Failure envelope for a unidirectional graphite–epoxy laminate
subjected to a uniaxial stress rxx, based on the maximum stress criterion.
The mode of failure depends on whether rxx is tensile or compressive,

and also on the fiber angle h:
If rxx is tensile, then:
Matrix failure is predicted for: 90j<h<33.7j and 33.7j<h
<90j.
Shear failure is predicted for: 33.7j<h<2.9j and 2.9j<h
<33.7j.
Fiber failure is predicted for: 2.9j<h<2.9j.
If rxx is compressive, then:
Matrix failure is predicted for: 90j<h<53.1j and 53.1j<h
<90j.
Shear failure is predicted for: 53.1j<h<3.6j and 3.6j<h
<53.1j.
Fiber failure is predicted for: 3.6j<h<3.6j.
Fiber failures are predicted for only a very narrow range of fiber angles. This
implies that failure of a unidirectional composite subjected to a uniaxial state
of stress will almost always occur due to matrix or shear failures, rather than

fiber failure. In general, fiber failure will only occur when the composite is
tested under carefully controlled laboratory conditions in which the uniaxial
stress is aligned with the fiber direction to within a few degrees. Also, note that
the fiber angle at which a change from shear failure to matrix failure occurs
differs in tension and compression.
6.1.2 Tsai–Hill Criterion

According to the Tsai–Hill criterion, failure of an orthotropic composite
subjected to plane stress conditions is governed by Eq. (53), whereas failure of
a transversely isotropic composite is governed by Eq. (54). For present
purposes, we have assumed that the composite is transversely isotropic.
Substituting Eq. (64) into Eq. (54), the Tsai–Hill failure criterion predicts
that failure will not occur if the following inequality is satisfied:
8 91=2
< cos2 ðhÞcos2 ðhÞ sin2 ðhÞ sin4 ðhÞ cos2 ðhÞsin2 ðhÞ >
> =
rxx < fT 2 þ 2 þ f 2
>
: r11 s12 >
;
ryT
22
ð68Þ
As previously noted, the Tsai–Hill criterion does not automatically account
for differences in tensile and compressive strengths. A failure envelope for a
unidirectional graphite–epoxy composite will be generated using tensile or
compressive strengths, as appropriate. Thus, the tensile strength predicted by
the Tsai–Hill criterion is:
8 91=2
> =
rfT ¼ þ þ
xx
>
: rfT
2 2
sf12
2 >
;
11 ryT
22
Similarly, the compressive strength predicted by the Tsai–Hill criterion is:

8 91=2
> =
xx ¼
rfC fC 2 þ 2 þ f 2
>
: r11 s12 >
;
ryC
22
Substituting the strength values that have been assumed for graphite–epoxy,
we have:
( )1=2
cos2 ðhÞ cos2 ðhÞ sin2 ðhÞ sin4 ðhÞ cos2 ðhÞsin2 ðhÞ
rxx ¼
fT
þ þ
ð1500 MPaÞ2 ð50 MPaÞ2 ð75 MPaÞ2
ð69aÞ

( )1=2
cos2 ðhÞ cos2 ðhÞ sin2 ðhÞ sin4 ðhÞ cos2 ðhÞsin2 ðhÞ
xx ¼
rfC þ þ
ð1200 MPaÞ2 ð100 MPaÞ2 ð75 MPaÞ2
ð69bÞ
A failure envelope based on Eqs. (69a) and (69b) is shown in Fig. 17. As
before, it is important to realize that this failure envelope is valid for a uniaxial
state of stress only. Failure envelopes obtained using the Tsai–Hill criterion
but for other states of stress differ substantially from Fig. 17.
6.1.3 Tsai–Wu Criterion

The Tsai–Wu criterion for the case of plane stress is given by Eq. (62).
f 2
rxx X11 cos4 ðhÞ þ X22 sin4 ðhÞ þ cos2 ðhÞ sin2 ðhÞðX66 þ 2X12 Þ

þrfxx X1 cos2 ðhÞ þ X2 sin2 ðhÞ 1 ¼ 0 ð70Þ

subjected to a uniaxial stress rxx, based on the Tsai–Hill criterion.

The constants that appear in Eq. (70) are calculated using the strength
properties that have been assumed for graphite–epoxy and Eqs. (56)–(59),
as appropriate:
1 1 1 1 106
X1 ¼ fT
fC ¼ ¼
r11 r11 1500 MPa 1200 MPa 6000 Pa
1 1 1015
X11 ¼ ¼ ¼
rfT fC
11 r11 ð1500 MPaÞð1200 MPaÞ 1800 Pa2
1 1 1 1 106
X2 ¼ ¼ ¼
ryT
22 ryC
22
50 MPa 100 MPa 100 Pa
1 1 1015
X22 ¼ ¼ ¼
ryT yC
22 r22
ð50 MPaÞð100 MPaÞ 5 Pa2
2 2
1 1 1012
X66 ¼ y ¼ ¼
s12 75 MPa 5625 Pa2
As previously discussed, there is no widely accepted technique used to

calculate X12. For present purposes, X12 will be calculated in accordance with
Eq. (61):
15 pffiffiffiffiffi !
1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 1015 10 10 15
X12 ¼ X11 X22 ¼ ¼ 10
2 2 1800 Pa2 5 Pa2 600 Pa2
Substituting these values into Eq. (70), we obtain:

( 15 4
f 2 1015 cos4 ðhÞ 10 sin ðhÞ
rxx 2
þ 2
þ cos2 ðhÞsin2 ðhÞ
1800 Pa 5 Pa
" 15 pffiffiffiffiffi #)
1012 10 10
ð71Þ
5625 Pa2 300 Pa2

106 cos2 ðhÞ 106 sin2 ðhÞ
þ rfxx þ 1¼0
6000 Pa 100 Pa
Equation (71) is a second-order polynomial in the unknown failure stress,
f
rxx . For any given fiber angle h, there will be two roots to this equation. The
predicted tensile strength equals the algebraically positive root, whereas the

predicted compressive strength equals the negative root. For example, for a
fiber angle h=30j, Eq. (71) becomes:

2 4417 1020 2375 1012
rfxx þ r f
xx 1¼0
Pa2 Pa
The two roots of this expression are found to be (125.9106 Pa, 179.7106
Pa). Hence, the strengths predicted by the Tsai–Wu criterion for a 30j
graphite–epoxy specimen are:
xx ¼ 125:9 MPa
rfT
xx ¼ 179:7 MPa
rfC
A failure envelope based on the Tsai–Wu criterion for a unidirectional
graphite–epoxy laminate subjected to a uniaxial state of stress is shown in
Fig. 18. This figure is analogous to those obtained using the maximum stress
criterion and the Tsai–Hill criterion (Figs. 16 and 17, respectively). As before,

subjected to a uniaxial stress rxx, based on the Tsai–Wu criterion.

it is important to realize that this failure envelope is valid for a uniaxial state of
stress only. Failure envelopes based on the Tsai–Wu criterion but for other
states of stress differ substantially from Fig. 18.
6.1.4 Comparison
The failure envelopes for uniaxial stress obtained on the basis of the three
failure criteria considered are compared directly in Fig. 19, and an expanded
view of just the first quadrant is presented in Fig. 20. It is apparent that similar
predictions are obtained on the basis of all three criteria, although a
significant numerical difference occurs at low fiber angles, near the region
at which the failure mode shifts from fiber failure to shear matrix failure.
However, one should not conclude that the failure criterion described above
always leads to similar predictions. In fact, depending on the state of stress
considered, the predicted failure envelopes may differ substantially. One
stress state that exhibits this effect is the state of pure shear stress, considered
in the following subsection.
Figure 19 Comparison of the failure envelopes for a unidirectional graphite–

epoxy laminate subjected to a uniaxial stress rxx, obtained using the maximum
stress, Tsai–Hill, and Tsai–Wu failure criteria.

Figure 20 Comparison of the failure envelopes (first quadrant only) for a
unidirectional graphite–epoxy laminate subjected to a uniaxial stress rxx,
obtained using the maximum stress, Tsai–Hill, and Tsai–Wu failure criteria.
6.2 Pure Shear Stress States

It was mentioned in Sec. 5 of Chap. 3 that the shear strength of composites is
sensitive to the algebraic sign of the shear stress when referenced to a
nonprincipal material coordinate system. This sensitivity would not be
expected based on previous experience with isotropic materials since the
shear strength of isotropic materials is not sensitive to algebraic sign. We are
now in a position to explain this phenomenon. An off-axis composite ply
subjected to a pure shear stress state is shown in Fig. 21. The stresses induced
in the 1–2 coordinate system can be determined using Eq. (20) of Chap. 2:
8 9 2 38 9
> r11>
< = 6 cos 2
ðh Þ sin 2
ð h Þ 2 cos ð h Þsinð h Þ >0>
7< =
r22 ¼ 6 4 sin ðhÞ
2
cos2 ðhÞ 2 cosðhÞsinðhÞ 7 5> 0 >
>
: > ; : ;
s12 cosðhÞ sinðhÞ cosðhÞ sinðhÞ cos2 ðhÞ sin2 ðhÞ sxy
or equivalently:
r11 ¼ 2sxy cosðhÞ sinðhÞ

Figure 21 Unidirectional composite subjected to pure shear stress sxy.
r22 ¼ 2sxy cosðhÞ sinðhÞ

s12 ¼ sxy cos2 ðhÞ sin2 ðhÞ ð72Þ
6.2.1 Maximum Stress Criterion

rfC rfT
11
< sxy < 11
ð73aÞ
2cosðhÞsinðhÞ 2cosðhÞsinðhÞ
(and)
ryC ryT
22
< sxy < 22
ð73bÞ
2cosðhÞsinðhÞ 2cosðhÞsinðhÞ
(and)
sy12
sxy < ð73cÞ
cos2 ðhÞ sin2 ðhÞ
Equations (73a)–(73c) will now be used to generate a failure envelope for a

graphite–epoxy laminate subjected to pure shear. With reference to Eqs.
(73a)–(73c), the following subtleties in these calculations should be noted:
Over the range 0j<h<90j, both the cosine and sine functions return
algebraically positive values. Consequently, for this range, a positive

shear stress sxy will induce a tensile value r11 and a compressive value
for r22. However,
Over the range 90j<h<0j, the cosine function returns a positive
value, whereas the sine function returns a negative value. Over this
range, a positive shear stress sxy will induce a compressive stress r11
but tensile stress r22.
Of course, if a negative shear stress sxy is applied rather than a positive shear
stress, then the algebraic signs of all stress components are reversed. These
subtleties are important during the application of Eqs. (73a)–(73c) because
composite strengths typically differ in tension and compression.
With these observations in mind, the following equations may be used
to generate a failure envelope for a graphite–epoxy laminate subjected to pure
shear, based on the strength properties previously listed:
Positive shear strengths:
For 0j<h<90j:
rfT 1500 MPa
sPxy ¼ 11
¼ ð74aÞ
2 cosðhÞ sinðhÞ 2 cosðhÞ sinðhÞ
ryC 100 MPa
sPxy ¼ 22
¼ ð74bÞ
2 cosðhÞ sinðhÞ 2 cosðhÞ sinðhÞ
sy12 75 MPa
sPxy ¼ ¼ ð74cÞ
cos ðhÞ sin ðhÞ
2 2
cos ðhÞ sin2 ðhÞ
2
For 90j<h<0j:
rfC 1200 MPa
sPxy ¼ 11
¼ ð74dÞ
2 cosðhÞsinðhÞ 2 cosðhÞsinðhÞ
ryT 50 MPa
sPxy ¼ 22
¼ ð74eÞ
sy12 75 MPa
sPxy ¼ ¼ ð74f Þ
cos ðhÞ sin ðhÞ cos ðhÞ sin2 ðhÞ
2 2 2
Negative shear strengths:

For 0j<h<90j:
rfC 1200 MPa
xy ¼
sN ¼ ð75aÞ
11
ryT 50 MPa
xy ¼
sN ¼ ð75bÞ
22

sy12 75 MPa
xy ¼
sN ¼ ð75cÞ
cos2 ðhÞ sin2 ðhÞ cos2 ðhÞ sin2 ðhÞ
For 90j<h<0j:
rfT 1500 MPa
xy ¼
sN ¼ ð75dÞ
11
ryC 100 MPa
xy ¼
sN ¼ ð75eÞ
22
sy12 75 MPa
xy ¼
sN ¼ ð75f Þ
cos ðhÞ sin ðhÞ cos ðhÞ sin2 ðhÞ
2 2 2
Equations (74a)–(74f ) and (75a)–(75f ) were used to create the failure

envelope shown in Fig. 22. Note the following.
The failure envelope shown in Fig. 22 is valid for a pure shear stress
state only. Failure envelopes for other states of stress differ sub-
stantially (for example, compare Figs. 16 and 22, both of which are
based on the maximum stress failure criterion).

subjected to a pure shear stress sxy, based on the maximum stress criterion.

None of the curves shown in Fig. 22 is associated with fiber failure.
The magnitudes of the critical shear stress values returned by Eqs.
(74a), (74d), (75a), or (75d) are all large enough that they do not ap-
pear in Fig. 22 due to the scale used for the vertical axis. These results
indicate that failure of a unidirectional graphite–epoxy laminate sub-
jected to a pure shear stress state will never occur due to fiber failure.
The shear strength of an off-axis unidirectional composite depends
on algebraic sign of the shear stress. For example, a 45j specimen is
predicted to have a positive shear strength of 100 MPa and a negative
shear strength of 50 MPa.
The predicted mode of failure depends on whether sxy is positive or
negative as well as on fiber angle h.
If sxy is positive, then:
Shear failure is predicted for 90j<h<73.2j.
Matrix failure is predicted for 73.2j<h<16.8j.
Shear failure is predicted for 16.8j<h<26.6j.
Shear failure is predicted for 63.4j<h<90j.
If sxy is negative, then:
Shear failure is predicted for 90j<h<63.4j.
Shear failure is predicted for 26.6j<h<16.8j.
Shear failure is predicted for 73.2j<h<90j.
6.2.2 Tsai–Hill Criterion

8 2 3 9
>
< >
6 2 1 1 7 cos ðhÞ þ sin ðhÞ =
4 4
s2xy 4 cos2 ðhÞsin2 ðhÞ4 2 þ 2 y 2 5 þ y 2 <1
>
: rfT 2 s12 s12 >
;
11 ryT
22
ð77Þ
Equating the left-hand side to unity and solving for sxy:
8 91=2
>
> >
>
>
> >
>
< 1 =
sxy ¼
>
> cos4 ðhÞ þ sin4 ðhÞ >
>
> >
: 4 cos ðhÞsin ðhÞ ðrfT Þ2 þ ðryT Þ2 2ðsy Þ2 þ
2
> 2 2 1 1
y 2 >
;
11 22 12 s 12
ð78Þ

Recall that the Tsai–Hill criterion does not automatically account for differ-
ences in tensile and compressive stresses. Therefore, to predict shear strengths
using Eq. (78), the failure strengths used must be selected according to
whether r11 and r22 are positive or negative.
Positive shear strengths:
For 90j<h<0j, both r11 and r22 are positive; therefore, use
fT yT
r11 and r22 .
For 0j<h<90j, r11 is positive and r22 is negative; therefore,
yC
use r11fT and r22 .
Negative shear strengths:
For 90j<h<0j, both r11 and r22 are negative; therefore, use
fC yC
r11 and r22 .
For 0j<h<90j, r11 is negative and r22 is positive; therefore,
fC yT
use r11 and r22 .
A failure envelope based on these failure strengths and Eq. (78) is shown in
Fig. 23.
Figure 23 Failure envelope for a unidirectional graphite–epoxy laminate sub-

jected to a pure shear stress sxy, based on the Tsai–Hill criterion.

6.2.3 Tsai–Wu Criterion
n h
s2xy X66 cos4 ðhÞ þ sin4 ðhÞ þ 2 cos2 ðhÞsin2 ðhÞ 2X11 þ 2X22 ð79Þ
io
4X12 X66 þ 2sxy cosðhÞsinðhÞ½X1 X2 < 1
Numerical values for constants X1, X2, X11, etc., were calculated in Sec. 6.1.3
(based on strengths assumed for graphite–epoxy). Substituting these values in
the left-hand side of Eq. (79) and equating to unity, we obtain:
(
1012 4
s2xy 2
cos ðhÞ þ sin4 ðhÞ þ cos2 ðhÞsin2 ðhÞ
5625Pa
" pffiffiffiffiffi 15 #)
67 1015 10 10 61 106
sxy cosðhÞsinðhÞ ¼1
150Pa2 75Pa2 3000
ð80Þ
Equation (80) is a second-order polynomial in the unknown shear failure

f
stress, sxy . For a given fiber angle, h, there are two roots to this equation. The
fP
predicted positive shear strength, sxy , equals the algebraically positive root,
fN
whereas the predicted negative shear strength, sxy , equals the negative root.
For example, for a fiber angle h=30j, Eq. (80) becomes:
2 202:8 1018
88:05 1010

f
sxy sxy
f
1¼0
Pa2 Pa
The two roots of this expression are found to be (95.22 106 Pa, 51.8 106
Pa). Hence, the shear strengths predicted by the Tsai–Wu criterion for a 30j
graphite–epoxy laminate are:
xy ¼ 95:2 MPa
sfP
xy ¼ 51:8 MPa
sfN
A failure envelope based on the Tsai–Wu criterion for a unidirectional

graphite–epoxy laminate subjected to a pure shear stress state is shown in
Fig. 24. This figure is analogous to those obtained using the maximum stress
criterion and the Tsai–Hill criterion (Figs. 22 and 23, respectively).
6.2.4 Comparisons
The failure envelopes for pure shear stress obtained on the basis of the three
failure criteria considered are compared directly in Fig. 25, and an expanded

Figure 24 Failure envelope for a unidirectional graphite–epoxy laminate sub-
jected to a pure shear stress sxy, based on the Tsai–Wu criterion.
view of just the first quadrant is presented in Fig. 26. The difference between
predictions obtained using the three failure criteria is more striking in pure
shear than was the case in uniaxial stress (e.g., compare Figs. 20 and 26).
Predictions on the basis of the Tsai–Hill and Tsai–Wu criteria are similar,
although some difference exists. The maximum stress criterion predicts local
maximums in shear strength near fiber angles of h=27j and 64j. These
maximums are associated with the previously noted change in failure mode,
from a shear failure mode to a matrix failure mode (or vice versa).
7 COMPUTER PROGRAMS UNIDIR AND UNIFAIL

The results derived in this chapter will be used extensively throughout the
remainder of this text. As is already abundantly clear, the calculations
associated with any thermomechanical analysis of an anisotropic composite
are tedious and time-consuming if performed using a hand calculator.
Consequently, most composite analyses are performed with the aid of a
digital computer.
Two computer programs have been developed to supplement the ma-
terial presented in this chapter: UNIDIR and UNIFAIL. These programs

Figure 25 Comparison of the failure envelopes for a unidirectional graphite–
epoxy laminate subjected to pure shear stress sxy, obtained using the Maximum
Stress, Tsai–Hill, and Tsai–Wu failure criteria.
can be downloaded at no cost from the following website: http://depts.

washington.edu/amtas/computer.html.
The analyses that can be performed with the aid of these programs will
be discussed in the following subsections. Both programs require the user to
provide various numerical values required during the calculations performed.
The user must define these values using a consistent set of units. For example,
program UNIDIR requires the user to input elastic moduli, thermal expan-
sion coefficients, and moisture expansion coefficients for the composite
material system of interest. Using the properties listed in Table 3 of Chap. 3
and based on the SI system of units, the following numerical values would be
input for graphite–epoxy:
E11 ¼ 170 109 Pa E22 ¼ 10 109 Pa v12 ¼ 0:30
G12 ¼ 13 109 Pa
a11 ¼ 0:9 106 m=mo C a22 ¼ 27:0 106 m=mo C
b11 ¼ 150:0 106 m=m%M b22 ¼ 4800 106 m=m%M

Figure 26 Comparison of the failure envelopes (first quadrant only) for a
unidirectional graphite–epoxy laminate subjected to a uniaxial stress rxx,
obtained using the maximum stress, Tsai–Hill, and Tsai–Wu failure criteria.
If the analysis requires the user to input numerical values for stresses, then
stresses must be input in pascals (not in MPa). A typical value would be
rxx=200106 Pa. If, instead, the analysis requires the user to input numerical
values for strains, then strains must be input in m/m (not in Am/m). A typical
value would be exx=2000106 m/m=0.002000 m/m. All temperatures
would be input in jC.
In contrast, if the English system of units was used, then the following
numerical values would be input for the same graphite–epoxy material
system:
E11 ¼ 25:0 106 psi E22 ¼ 1:5 106 psi v12 ¼ 0:30
G12 ¼ 1:9 106 psi
a11 ¼ 0:5 106 in:=in:o F a22 ¼ 15 106 in:=in:o F
b11 ¼ 150:0 106 in:=in:%M b22 ¼ 4800 106 in:=in:%M
If the analysis requires the user to input numerical values for stresses, then
stresses must be input in psi (not in ksi). A typical value would be rxx=30,000
psi. If, instead, the analysis requires the user to input numerical values for

strains, then strains must be input in in./in. (not in Ain./in.). A typical value
would be exx=2000106 in./in.=0.002000 in./in. All temperatures would be
input in jF.
7.1 Program UNIDIR

Program UNIDIR may be used to predict the elastic behavior of unidirec-
tional composites and is based on the material presented in 1 Secs. 2 Secs. 3
Secs. 4. Two different types of analyses may be performed. The program may
be used in either of the following:
Calculating total strains (exx, eyy, cxy) caused by a specified com-
bination of stresses (rxx, ryy, sxy), a uniform temperature change
(DT), and a uniform change in moisture content (DM). Calculations
performed as a part of Example Problems 1, 2, 4, and 5 are typical of
this type of analysis.
Calculating stresses (rxx, ryy, sxy) caused by a specified combination
of total strains (exx, eyy, cxy), a uniform temperature change (DT), and
a uniform change in moisture content (DM). Calculations performed
as a part of Example Problems 3 and 6 are typical of this type of
analysis.
The program also determines the effective properties of a unidirectional
composite based on the definitions described in Sec. 4. An implicit assumption
in these calculations is that all material properties (E11, a11, b11, etc.) input by
the user correspond to the temperature and moisture content dictated by DT
and DM.
7.2 Program UNIFAIL

Program UNIFAIL may be used to obtain failure predictions for unidirec-
tional composites based on the maximum stress, Tsai–Hill, or Tsai–Wu
failure criteria introduced in Sec. 5. Two different types of analyses may be
performed. The program may be used in either of the following:
Calculating predicted uniaxial and shear strengths of a unidirectional
laminate with a specified fiber angle h.
Generating a data file that can subsequently be used to produce
failure envelopes for unidirectional composites subjected to several
types of plane stress conditions.
Note that the program UNIFAIL itself does not create a failure envelope.
Rather, the program creates a file (named Envelop.txt) that contains the
stress(es) predicted to cause failure of a unidirectional composite as a function
of fiber angle, based on the particular failure criterion specified by the user. A

failure envelope may then be created using a second software package to
‘‘import’’ the data generated by the program UNIFAIL and then plotting
failure stress vs. fiber angle. For example, any of the failure envelopes
presented in Sec. 6 may be easily recreated in this way. As before, an implicit
fT yT
assumption is that failure strengths (r11 , a22 , etc.) input by the user corre-
spond to the values exhibited by the composite at the temperature and mois-
ture content of interest. Also, in the present context, ‘‘failure’’ may represent
fracture of the fibers or yielding of the matrix.
HOMEWORK PROBLEMS
Notes: (a) In the following problems, the phrase ‘‘by hand calculation’’ means
that solutions are to be obtained using a calculator, pencil, and paper. (b)
Computer programs UNIDIR and/or UNIFAIL are referenced in many of
the following problems. As described in Sec. 7, these programs can be
downloaded from the following website: http://depts.washington.edu/
amtas/computer.html.
1. Calculate the reduced compliance matrix for the materials listed below,
first by hand calculation and then using program UNIDIR. Use the
material properties listed in Table 3 of Chap. 3.
(a) Glass/epoxy.
(b) Kevlar/epoxy.
(c) Graphite/epoxy.
2. Calculate the reduced stiffness matrix for the materials listed below, first
by hand calculation and then using program UNIDIR. Use the material
properties listed in Table 3 of Chap. 3.
(a) Glass/epoxy.
(b) Kevlar/epoxy.
(c) Graphite/epoxy.
3. A thin unidirectional glass/epoxy composite laminate is simultaneously

subjected to a uniform temperature change DT=175jC, an increase in
moisture content DM=0.5%, and the following in-plane stresses:
r11 ¼ 350 MPa

r22 ¼ 40 MPa
s12 ¼ 60 MPa

Determine the resulting strains (e11, e22, and c12), first by hand calcu-
lation and then using program UNIDIR. Use the material properties
listed in Table 3 of Chap. 3.
4. Repeat Problem 3 for a unidirectional Kevlar/epoxy composite lami-

nate.
5. Repeat Problem 3 for a unidirectional graphite/epoxy composite lami-

nate.
6. A thin unidirectional glass/epoxy composite laminate is simultaneously

subjected to a uniform temperature change DT=275jF and an un-
known plane stress state. The following strains are measured as a result
(moisture content remains constant):
e11 ¼ 1250 Ain:=in:

e22 ¼ 2000 Ain:=in:
c12 ¼ 0
Determine the stresses (r11, r22, and s12), first by hand calculation and
then using program UNIDIR. Use the material properties listed in
Table 3 of Chap. 3.

nate.

nate.
9. A square unidirectional glass/epoxy composite laminate with dimen-

sions 11 m is clamped between four infinitely rigid walls, as shown in
Fig. 27. Material properties are listed in Table 3 of Chap. 3. Initially, the
clamped composite is stress-free, but the temperature is subsequently
decreased by 100jC. The thermal expansion coefficient of the rigid walls
is zero, so the rigid walls do not expand or contract.
(a) Calculate the stresses (r11, r22, s12) induced by this change in
temperature, first by hand calculation and then using
program UNIDIR.
(b) Predict whether the composite will fail based on the maxi-
mum stress failure criterion.

Figure 27 Clamped composite laminate considered in Problems 9, 10, and 11.

nate.

nate.
12. A 11 m square unidirectional glass/epoxy composite laminate is

placed within a cavity defined by four rigid walls, as shown in Fig. 28.
An initial gap of 0.050 mm exists between all edges of the ply and the
walls. The composite ply adsorbs 1.5% moisture, causing the ply to
expand and completely fill the cavity. Temperature remains constant.
The rigid walls do not adsorb moisture, and hence do not expand or
contract. Assuming the ply does not buckle, calculate the stresses (r11,
r22, s12) caused by the change in moisture content.

nate.

nate.
15. A perfectly square unidirectional glass/epoxy composite laminate is

mounted in a frame consisting of four infinitely rigid frame members,

Figure 28 A composite laminate placed within a cavity defined by four rigid
walls (considered in Problems 12, 13, and 14).
as shown in Fig. 29(a). The frame members are pinned at each corner.
Since the laminate is perfectly square, the angle defined by corners ABC
is initially 90j (precisely).
A force F is then applied to two diagonal corners, as shown in Fig. 29(b).

After F is applied, angle ABC is measured to be 89.50j (precisely). Both
temperature and moisture content remain constant. What stresses (r11,
r22, s12) are induced in the panel?

nate.

nate.
18. Create a plot of the following effective properties for a unidirectional

glass/epoxy composite, with fiber angles ranging from 90j to +90j.
(a) Exx and Eyy.

(b) vxy and vyx.

Figure 29 Unidirectional composite laminate considered in Problems 15, 16,
and 17.
(c) Gxy.
(d) gxy,xx and gxy,yy.
(e) gxx,xy and gyy,xy.
(f ) axx, ayy, and axy.
(g) bxx, byy, and bxy.
Suggested solution procedure: use program UNIDIR (repeatedly) to

calculate the required properties in increments of 5j (that is, calcu-
late for h=0j, 5j, 10j, 15j, 20j, etc.) and then plot these calcu-
lations.

nate.

20. (a) Using hand calculation, predict the positive and negative shear
strengths for a unidirectional 45j glass/epoxy composite laminate
based on the maximum stress failure criterion.
(b) Use program UNIFAIL to predict the positive and negative shear
strengths for a unidirectional 45j glass/epoxy composite laminate
based on the maximum stress, Tsai–Hill, and Tsai–Wu failure cri-
teria. Compare these results with your calculations obtained in part
(a).

nate.

nate.
23. On the same graph, plot failure envelopes for a unidirectional glass/
epoxy composite laminate for the following two conditions:
(a) Unidirectional stress: rxx, ryy=sxy=0

(b) Biaxial normal stress: ryy=rxx/10, sxy=0
Use the Tsai–Hill failure criterion. Suggested solution procedure: use

program UNIFAIL (twice) to generate data files corresponding to the
specified loading conditions, and then plot these data files.

nate.

nate.
REFERENCES
1. Reuter, R.C. Concise property transformation relations for an anisotropic lami-
na. J. Compos. Mater., Vol. 5, April 1971, 270–272.
2. Hill, R. The Mathematical Theory of Plasticity; New York, Oxford University
Press, 1998.
3. Tsai, S.W. Strength theories of filamentary structures. In: Fundamental Aspects
of Fiber Reinforced Plastic Composites; Schwartz, R.T., Schwartz, H.S., eds;
Wiley Interscience: New York, 1968; 3–11.
4. Tsai, S.W.; Wu, E.M. A general theory of strength for anisotropic materials. J.
Compos. Mater., Vol. 5, January 1971, 58–80.

6
Thermomechanical Behavior of Multiangle
Composite Laminates
Thin composite laminates in which the principal material coordinate system

of all plies is aligned were considered in Chap. 5. The behavior of laminates,
wherein the alignment of the principal material system varies from one ply to
the next, will be considered in this chapter. In essence, we will combine the
results of Chap. 5 with traditional ‘‘thin-plate theory.’’ This combination will
result in an analysis technique commonly known as ‘‘classical lamination
theory’’ (CLT).
1 DEFINITION OF A ‘‘THIN PLATE’’ AND ALLOWABLE PLATE

LOADINGS
A ‘‘thin plate’’ with in-plane dimensions a and b and thickness t is shown
schematically in Fig. 1. The plate can be considered ‘‘thin’’ if the plate
thickness is less than about one-tenth the in-plane dimensions (i.e., if t<a/
10 and t<b/10). An x–y–z coordinate system is defined as indicated. Note that
the origin of the x–y–z coordinate system is positioned at the geometrical
center of the plate, such that the midplane (or midsurface) of the plate lies
within the plane z = 0. Consequently, the plate exists within the space defined
by the planes z = t/2 and z = +t/2.
We will assume that the thin plate is subjected to plane stress conditions.
Therefore, we will only consider plate loadings that result in a plane stress
265

Figure 1 A ‘‘thin plate’’ with in-plane dimensions a and b and thickness t.
state within the plate. Furthermore, in this chapter, we will only consider
uniformly distributed loads. That is, we will assume that the loads are constant
and uniformly distributed along the edge of the plate. The more general case
in which loads vary along the edge of the plate will be considered in Chap. 10.
Six types of uniformly distributed loads that give rise to plane stress
conditions within the x–y plane are shown in Fig. 2. All load components are
shown in an algebraically positive sense. Because the line-of-action of all load
vectors shown in Fig. 2 lies within the x–y plane, these load components are
often referred to as in-plane loads.
First, consider load components Nxx, Nyy, and Nxy. Two subscripts are
used to identify these load components. The algebraic sense of each compo-
nent is interpreted in a manner analogous to that previously used to identify
the algebraic sense of individual stress components (discussed in Sec. 2.5).
That is, the first subscript indicates the face of the plate a given load acts upon,
whereas the second subscript indicates the line of action of the load. A positive
load is one that:
Acts on a positive face and points in a positive coordinate direction,
or
Acts on a negative face and points in a negative coordinate direction.
The algebraic sense of normal loads Nxx and Nyy (Fig. 2a) is readily
apparent and intuitive: A positive (tensile) normal load is one that tends to
cause the plate to stretch. The algebraic sense of shear loads Nxy and Nyx (Fig.
2b) is not as immediately apparent, but application of the sign convention just

Figure 2 Schematic of allowable plate loadings: (a) in-plane normal forces Nxx
and Nyy, (b) in-plane shear forces Nxy and Nyx, (c) bending moment Mxx, (d)
bending moment Myy, (e) in-plane torques Mxy and Myx.

described will confirm that the shear loads shown in Fig. 2b are indeed
positive. That is, two of the shear loads shown are acting on a positive face and
point in a positive coordinate direction, whereas two of the shear loads shown
act on a negative face and point in a negative coordinate direction. Because
individual load components are not allowed to vary spatially (i.e., loads are
assumed to be constant and uniformly distributed along each edge of the
plate), statical equilibrium requires that the shear loads acting along the x-
edge and y-edge be orientated tip-to-tip and tail-to-tail, as shown in Fig. 2b.
Furthermore, the magnitude of the shear loads must be identical, jNxyj =
jNyxj. These requirements are also analogous to those of shear stresses acting
on an infinitesimal stress element, as discussed in Sec. 2.5. It is emphasized
that Nxx, Nyy, and Nxy are all defined as distributed loads, expressed in units of
force/plate length (such as N/m or lbf/in.).
The remaining loads shown in Fig. 2 (Mxx, Myy, and Mxy) are bending
moments (or torques) distributed along the edge of the plate. Load compo-
nents Mxx and Myy are uniformly distributed bending moments acting along
the x-edge and y-edge of the plate, respectively, as shown in Fig. 2c and d.
These loads are shown in an algebraically positive sense. The subscripts
assigned to Mxx and Myy may seem puzzling at first because, for example, Mxx
represents a bending moment acting about the y-axis. However, Mxx is
directly related to the distribution of rxx through the thickness of the plate,
as will be shown below. Because Mxx arises due to the distribution of rxx (or
vice versa), it is customary to use the same subscripts for both entities.
Unfortunately, the convention used to assign an algebraic sign to distributed
bending moments varies from one author to the next, and in fact can be
arbitrarily chosen. The sign convention used herein is most commonly used in
the study of composite plates. An algebraically positive distributed bending
moment is defined as one that tends to cause tensile stresses in the positive z-
face of the plate and compressive stresses in the negative z-face. Referring to
Fig. 2c and noting that the positive z-direction is downward as drawn, it is
seen that Mxx tends to cause tensile stresses in the positive z-face (i.e., the
lower face) of the plate. Hence, Mxx is positive as drawn. A similar
observation holds for Myy, as shown in Fig. 2d.
Finally, loads Mxy and Myx are defined as uniformly distributed in-
plane moments (or torques) acting along neighboring edges of the plate, as
shown in Fig. 2d. It will be shown below that Mxy and Myx are directly related
to in-plane shear stresses sxy and syx, respectively. Because Mxy and Myx arise
due to the distribution of sxy and syx, it can be shown that for statical
equilibrium to be maintained, jMxyj = jMyxj. An algebraically positive
distributed torque is defined as one that tends to cause a positive shear stress
in the positive z-face (i.e., the lower face) of the plate.
Recall that the units of an applied moment or torque are force–length
(such as Nm or lbfin). Because Mxx, Myy, and Mxy all represent uniformly

distributed moments acting along the plate edge, they are all expressed in units
of force–length/plate length (such as Nm/m or lbfin/in).
It is expected that most readers will have considered the behavior of
isotropic prismatic beams during earlier studies. It is therefore instructive to
contrast the definitions just given for a thin plate, as well as the loads applied
thereon, to those encountered in fundamental beam theory. As previously
shown in Fig. 1, a ‘‘thin plate’’ is defined as a structure whose thickness, t, is
much less than the in-plane dimensions a and b. That is, t<<a,b. In contrast, a
beam is a structure for which two dimensions are small compared to the third.
Hence, a beam can be described as a structure in which one of the in-plane
dimensions, say width b, is of the same order as the thickness t. Hence, the
thin plate shown in Fig. 1 is ‘‘converted’’ to a beam if we allow bct<<a. In
this way, we describe a beam with rectangular cross-section bt and length a,
as shown in Fig. 3. The beam is called ‘‘prismatic’’ if the cross-section remains
constant along the length of the beam (i.e., if b and t remain constant along
length a).
Regarding the description of applied loading, in the case of thin plates,
all loading conditions are specified in terms of distributed loads, as described
in preceding paragraphs. For example, in SI units, Nxx is expressed in terms of
Newtons per meter, whereas Mxx is expressed in terms of Newton meter per
meter. In contrast, in fundamental beam theory, point loads are often
specified. For a beam, a normal load is often expressed in terms of Newtons,
whereas bending moments (or torques) are expressed in terms of Newton
meter. Loads that correspond to Nxx and Mxx, when applied to a beam, have
b b
also been shown in Fig. 3. They have been denoted as N xx and M xx , where the
superscript ‘‘b’’ is used to denote that these loads are defined in the sense
traditionally used in beam theory and therefore have different units than the
Figure 3 A prismatic beam with rectangular cross-section (compare with Fig. 1).

distributed loads used in plate theory. Because the width of the beam is b, the
b b
two load definitions are related according to Nxx =bNxx and Mxx =bMxx. As
mentioned above, the sign convention used to define an algebraically positive
bending moment varies from one author to the next, and can be arbitrarily
selected. The bending moment applied to the beam shown in Fig. 3 is
considered to be positive according to the sign convention used throughout
this text. This corresponds to the convention most commonly used in the
study of composite plates. However, according to the sign convention used in
many textbooks devoted to fundamental beam theory, the bending moment
shown in Fig. 3 would be considered as negative. Hence, the sign convention
used to describe bending moments in this and other textbooks devoted to
composites differs from the sign convention used in many textbooks devoted
to beam theory. The reader must simply be aware of this potential source of
confusion, and carefully note which convention has been used when compar-
ing the results described in this text to those developed elsewhere.
Let us now return to the topic of thin plates. We wish to relate the
external distributed loads applied to the plate to the resulting internal stresses.
An edge view of a plate loaded only by distributed load Nxx and moment Mxx
is shown in Fig. 4a. A free-body diagram of a section of the plate is shown in
Fig. 4b. The free-body diagram has been drawn showing the distributed load
Nxx and distributed moment Mxx on the left-hand side, and the resulting
internal stress rxx on the right-hand side. The plate is assumed to be in statical
equilibrium. Therefore, SF=0, and the force per unit width associated with
the distribution of stress rxx through the thickness of the plate must be exactly
balanced by the distributed load Nxx. Let the free-body diagram have a width
of ‘‘1’’ and consider an incremental strip of height dz. The cross-sectional area
dA of this strip is dA=1dz. The incremental force dFxx associated with the
stress acting over this thin strip is dFxx=dNxx1=rxxdA=rxxdz. We can now
relate the total distributed force Nxx acting on the left-hand side of the free-
body diagram to the distribution of rxx acting on the right-hand side by
simply ‘‘adding up’’ the forces acting over all incremental strips; that is, we
integrate over the thickness of the plate:
Z þt=2 Z t=2
Nxx ¼ dNxx ¼ rxx dz ð1aÞ
t=2 t=2
In an entirely equivalent manner, we can relate distributed forces Nyy and Nxy
to stresses ryy and sxy, respectively:
Z t=2
Nyy ¼ ryy dz ð1bÞ
t=2
Z t=2
Nxy ¼ sxy dz ð1cÞ
t=2

Figure 4 Edge view of a thin plate subjected to loads Nxx and Mxx.
Now consider moment Mxx. As before, the plate is assumed to be in

statical equilibrium, and hence moments acting about the y-axis must sum to
zero: SMy=0. Again consider an incremental strip of height dz, which is
located a distance z from the midsurface. The incremental distributed moment
dMxx contributed by Nxx acting over the incremental strip is dMxx=dNxxz=
rxxzdz. We can obtain the total moment acting on the right-hand side of the
free-body diagram by integrating over the thickness of the plate:
Z t=2
Mxx ¼ rxx zdz ð2aÞ
t=2
In an entirely equivalent manner, we can relate moments Myy and Mxy to

stresses ryy and sxy, respectively:
Z t=2
Myy ¼ ryy zdz ð2bÞ
t=2
Z t=2
Mxy ¼ sxy zdz ð2cÞ
t=2

Equations (1a) (1b) (1c) (2a) (2b) (2c) show that the uniformly distributed
loads and moments applied to the plate edge are directly related to the stresses
within the plate. Distributed loads Nxx, Nyy, and Nxy are commonly called
stress resultants, and moments Mxx, Myy, and Mxy are commonly called
moment resultants.
2 PLATE DEFORMATIONS: THE KIRCHHOFF HYPOTHESIS

Let us now consider the deformation of a thin, flat plate. Fig. 5 represents a
(magnified) view of the edge of the plate in both the ‘‘initial’’ and ‘‘deformed’’
positions. The positive x-direction is to the right, the positive y-direction is out
of the plane of the figure, and the positive z-direction is downward, which is
consistent with the original definition shown in Fig. 1. Although we will
eventually apply our results to a composite laminate, for the moment, we will
not consider the existence of individual plies, and therefore the ply interfaces
are not shown in the figure. If the flat plate is loaded and/or is subjected to a
change in environment, it will be deformed and (in general) will become
curved, as shown in the figure. We will base our analysis on the Kirchhoff
hypothesis, which states that a ‘‘straight line that is initially perpendicular to
the midplane of the plate remains straight and perpendicular to the midplane
of the plate after deformation.’’ For example, let us consider straight line b–o–
d. This line is shown in Fig. 5, in the sketch of both the initial and deformed
positions of the plate. In accordance with the Kirchhoff hypothesis, line b–o–
Figure 5 Initial and deformed positions of a flat plate.

d has been drawn perpendicular to the midplane in both the initial and
deformed positions. In effect, we have assumed that out-of-plane shear strains
are zero (cxz=cxz=0), which is equivalent to saying that we have assumed
that the z-axis is a principal strain axis.
We are interested in describing the displacement of an arbitrary point
‘‘c,’’ which is located within the thickness of the plate at some distance zc from
the midsurface. Point ‘‘c’’ is shown in Fig. 5, and lies along line b–o–d. Denote
the displacement of point ‘‘o’’ in the x-direction and z-direction as distances
uo and wo, respectively. From the figure, it can be seen that the distance point
‘‘c’’ has moved in the x-direction, uc, is approximately given by:
uc iuo zc sin a ð3Þ
where a is the angle formed by the plate midplane and the x-axis in the
deformed condition. Equation (3) is approximate because we have ignored
any change in plate thickness (i.e., we have ignored any change in distance zc
that may have occurred during deformation of the plate). If we now further
assume that angle a is small, then we can simplify Eq. (3) using the small-angle
approximation, which states that if a is expressed in radians and is less than
about 0.1745 rad (about 10j), then:
sin aia tan aia cos ai1 ð4Þ
Based on this assumption, Eq. (3) can be written as:
uc iuo zc a ð5Þ
Now, from Fig. 5, it can be seen that tan a= dw/dx. Applying the small-angle
approximation once more, we can say that tan a i a i dw=dx. Substituting
this result in Eq. (5), we obtain:
dw
uc ¼ u o z c ð6Þ
dx
To summarize, we have expressed the displacement in the x-direction of
arbitrary point ‘‘c’’ (which we have called distance uc) as a function of the
displacement in the x-direction of a point on the plate midsurface (distance
uo), the position of point c with respect to the midsurface (length zc), and the
slope of the plate midsurface dw/dx.
We will also require an expression for the displacement of point ‘‘c’’ in
the y-direction. We will denote the displacement of point ‘‘c’’ in the y-
direction as vc. Using a procedure that is entirely equivalent to that just
described, it can be shown that:
dw
vc ¼ vo zc ð7Þ
dy
Equations (6) and (7) represent the displacements of point ‘‘c’’ in the x-
direction and y-direction, respectively, and follow directly from the Kirchhoff

hypothesis. We can now determine the infinitesimal in-plane strains at point
‘‘c’’, in accordance with Eq. (49) of Chap. 2:
Bu Bv Bv Bu
exx ¼ eyy ¼ cxy ¼ þ ð8Þ
Bx By Bx By
Substituting Eqs. (6) and (7) into Eq. (8), we obtain the following expressions
for the strains induced at point ‘‘c’’:
Buo B2 w
ecxx ¼ zc 2
Bx Bx
Bvo B2 w
ecyy ¼ zc 2 ð9Þ
By By
Buo Bvo B2 w
ccxy ¼ þ 2zc
By Bx BxBy
Let:
Buo
eoxx ¼ ð10aÞ
Bx
Bv o
eoyy ¼ ð10bÞ
By
Buo Bvo
coxy ¼ þ ð10cÞ
By Bx
B2 w
jxx ¼ ð10dÞ
Bx2
B2 w
jyy ¼ 2 ð10eÞ
By
B2 w
jxy ¼ 2 ð10f Þ
BxBy
where eoxx ; eoyy ; and coxy are the in-plane strains that exist at the midplane of the
plate. The terms jxx, jyy, and jxy are called midplane curvatures, and represent
the rate of change of the slope of the midplane of the plane.
The reader is likely to have encountered the concept of midplane cur-
vatures during earlier studies of fundamental beam theory. Unfortunately,
the algebraic sign convention used to define curvatures varies from one author
to the next. The sign convention used throughout this text and defined by Eqs.
(10a–10f ) is most commonly used in the study of composite plates. However,
in many textbooks devoted to beam theory, curvatures are defined using the
opposite sign convention. For example, in beam theory curvature, jxx is often
defined as jxx=+B2w/Bx2, rather than jxx=B2w/Bx2 as indicated above.
Also, in some textbooks devoted to plate theory, jxy is defined as jxy=+B2w/
BxBy, rather than as jxy=2B2w/BxBy as indicated in Eq. (10f ). These

unfortunate deviations from one author to the next have developed over
many years, and a universal agreement on algebraic signs or even the
fundamental definition of jxy is not likely to occur for the foreseeable future.
The reader must simply be aware of these potential sources of confusion, and
carefully note that the convention has been used when comparing the results
described in this text to those developed elsewhere. Returning now to the
discussion of through-thickness strains, note from Fig. 5 that point ‘‘c’’ is
located at an arbitrary distance z from the neutral surface. We will therefore
discontinue the use of the subscript ‘‘c’’ in Eq. (9). Substituting Eqs. (10a–10f)
in Eq. (9), we obtain:
exx ¼ eoxx þ zjxx

eyy ¼ eoyy þ zjyy ð11Þ
cxy ¼ coxy þ zjxy
Equation (11) can be conveniently written in matrix form as:

8 9 8 o 9 8 9
>
> exx >
> >
> exx >
> >
> jxx >>
< = < = < =
eyy ¼ eoyy þ z jyy ð12Þ
>
> >
> > > > >
: ; > : co >; >
: >
;
cxy xy j xy
Equation (12) is the primary result we require for present purposes from
classical thin-plate theory. It allows us to calculate the infinitesimal in-plane
strains (exx,eyy,cxy) induced at any position z through the thickness of the
plate, based on the midplane strains (eoxx ; eoyy ; coxy ) and midplane curvatures
(jxx,jyy,jxy). Note that this result is based strictly on the Kirchhoff hypoth-
esis. We have made no assumptions regarding the mechanism(s) that caused
the flat plate to deform. Hence, Eq. (12) is valid if the plate is deformed by a
change in temperature, a change in moisture content, externally applied
mechanical loads, or any combination thereof. Also, we have made no
assumptions regarding material properties. Equation (12) is therefore valid
for isotropic, transversely isotropic, orthotropic, or anisotropic thin plates.
Example Problem 1
A thin plate with a thickness of 1 mm is subjected to mechanical loads, a
change in temperature, and a change in moisture content. Strain gages are
used to measure the surface strains induced in the plate. They are found to be:
at z ¼ t=2 ¼ 0:5 mm : exx ¼ 250 Am=m; eyy 1500 Am=m; cxy ¼ 1000 Arad
at z ¼ þt=2 ¼ þ0:5 mm : exx ¼ 250 Am=m; eyy 1100 Am=m; cxy ¼ 800 Arad
What midplane strains and curvatures are induced in the plate?

Solution. To solve this problem, we simply apply Eq. (12) to both surfaces of
the plate. For example, using the measured strains for exx, we have:
at z ¼ t=2 ¼ 0:0005 m : exx ¼ 250 Am=m ¼ eoxx ð0:0005Þjxx
at z ¼ þt=2 ¼ þ0:0005 m : exx ¼ 250 Am=m ¼ eoxx þ ð0:0005Þjxx
Solving simultaneously, we find:

eoxx ¼ 0 Am=m; jxx ¼ 0:50 rad=m
Using a similar approach utilizing the measured values for eyy and cxy, we
find:
eoyy ¼ 1300 Am=m; jyy ¼ 0:40 rad=m
coxy ¼ 900 Arad; jxy ¼ 0:20 rad=m
3 PRINCIPAL CURVATURES
In Sec. 2, we invoked the Kirchhoff hypothesis, according to which it is
assumed that a straight line that is initially perpendicular to the midplane of
the plate remains straight and perpendicular to the midplane after deforma-
tion. The Kirchhoff hypothesis has ultimately allowed us to calculate the in-
plane strains referenced to the x–y coordinate system (exx, eyy, and cxy)
induced at any position z through the thickness of a thin plate, using either
Eq. (11) or Eq. (12). These equations are valid for any combination of
midplane strains (eoxx, eoyy, and coxy) and midplane curvatures (jxx, jyy, and jxy).
In this section, we will consider a special case. Specifically, we will
consider a state of deformation in which the midplane strains are zero: eoxx ¼
eoyy ¼ coxy ¼ 0. In this special case, Eq. (12) becomes:
8 9 8 9
< exx >
> = < jxx >
> =
eyy ¼ z jyy ð13Þ
>
: >
; >
: >
;
cxy jxy
This state of deformation is known as pure bending. When a thin plate is in a

state of pure bending, all midplane strains are zero (eoxx ¼ eoyy ¼ coxy ¼ 0) and
the midplane of the plate is called the neutral surface.
Equation (13) gives the in-plane strains referenced to the x–y coordinate
system. Referring to Fig. 6, suppose we wish to express these strains relative to
a new x V–y Vcoordinate system, obtained by rotating through angle a about the
z-axis. As noted in Sec. 2, the Kirchhoff hypothesis implies that cxz=cyz=0.
Therefore, the z-axis is a principal strain axis. Consequently, we can rotate in-
plane strains from the x–y coordinate system to the new x V–y V coordinate

Figure 6 In-plane coordinate system xV–y V, obtained by rotating through angle
a about the z-axis.
system using Eq. (44) of Chap. 2 (developed in Sec. 13 of Chap. 2), repeated
here for convenience:
8 9 2 38 9
>
> ex Vx V >> cos2 ðaÞ sin2 ðaÞ 2cosðaÞ sinðaÞ >
> exx >
>
>
< >
= 6 7>
< >
=
ey Vy V 6 7 eyy
¼6 sin2 ðaÞ cos2 ðaÞ 2cosðaÞ sinðaÞ 7 ð2:44Þ
>
> c >
> 4 5>
> c > >
>
: x Vy V >
; : xy >
> ;
2 cosðaÞ sinðaÞ cosðaÞ sinðaÞ cos ðaÞ sin ðaÞ
2 2
2
Substituting Eq. (13) into Eq. (44) of Chap. 2, we can write:
8 9 8 9
< ex Vx V = < jx Vx V =
ey Vy V ¼ z jy Vy V ð14Þ
:c ; : ;
x Vy V jx Vy V
where:
8 9 2 38 9
>
> jx Vx V >
> cos2 ðaÞ sin2 ðaÞ 2cosðaÞ sinðaÞ >
> jxx >
>
< = 6 7 jyy =
<
jy Vy V
¼6 sin2 ðaÞ cos2 ðaÞ 2cosðaÞ sinðaÞ 7 ð15Þ
> 4
> jx Vy V >
>
5> j >
>
: ; : xy >;
2 cosðaÞ sinðaÞ cosðaÞ sinðaÞ cos ðaÞ sin ðaÞ
2 2
2
Note that midplane curvatures in the x V–yV coordinate system are related to
curvatures in the x–y coordinate system by means of the familiar transforma-
tion matrix [T]. This reveals that midplane curvatures can be treated as a
second-order tensor, and can be transformed from one coordinate system to

another in exactly the same way as the strain tensor (or stress tensor) is
transformed.
The in-plane principal strains and the orientation of the principal strain
coordinate system can also be determined using Eqs. (47) and (48) of Chap. 2,
respectively, repeated here for convenience:
ffi
exx þ eyy exx þ eyy 2 cxy 2
ep 1 ; ep 2 ¼ F þ ð2:47Þ
2 2 2

1 cxy
hpe ¼ arctan ð2:48Þ
2 exx eyy
Substituting Eq. (13) into Eq. (47) of Chap. 2, we find that for pure bending,
the in-plane principal strains are given by:
2 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 3
2 ffi
ðj þ j Þ ðj þ j Þ j 2
e p 1 ; e p 2 ¼ z4 5
xx yy xx yy xy
F þ ð16Þ
2 2 2
Substituting Eq. (13) into Eq. (48) of Chap. 2, we find that the orientation of
the principal strain coordinate system is given by:

1 jxy
hpe ¼ arctan ð17Þ
2 jxx jyy
Noting that jxx, jyy, and jxy are midplane values, Eq. (16) shows that
principal strains are linear functions of z. In contrast, Eq. (17) shows that,
for the case of pure bending, the orientation of the principal strain coordinate
system is constant and does not vary with through-thickness position, even
though the principal strains do vary with z.
A simplified expression for the principal strains is obtained by writing
Eq. (16) as:
ep1 ¼ jp1
ep2 ¼ jp2 ð18Þ
where jp1 and jp2 are called principal curvatures and are given by:
2 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 3
2 ffi
ðj þ j Þ ðj þ j Þ j 2
jp1 ; jp2 ¼ z4 5
xx yy xx yy xy
F þ ð19Þ
2 2 2
For the case of pure bending, the principal curvatures occur in the same
coordinate system as the principal strains. Hence, Eq. (17) gives the orienta-
tion of the coordinate system in which the principal curvatures exist. Because
shear strain is zero in the principal strains coordinate system, jp1p2=0 as well.
A physical interpretation of the preceding results can be obtained
through sketches of deformed strain elements parallel to the x–y plane, as

was done in Sec. 2.13 (in particular, refer to Sample Problem 2.9). A thin plate
reference to an x–y–z coordinate system is shown in Fig. 7a. A rectangular 3D
element cut out of this plate by two pairs of planes parallel to the x–z and y–z
planes is also shown. The dimensions of the element in the x-direction and y-
direction are dx and dy, respectively, whereas the height of the element equals
the plate thickness t. Assuming this plate is subjected to pure bending, then the
strains induced at any position z, relative to the x–y coordinate system, can be
calculated using Eq. (13).
Consider as representative examples three 2D strain elements parallel to
the x–y plane and located at positions defined by:
z=t/2 (element a–b–c–d, shown in Fig. 7b)
z=0 (element e–f–g–h, shown in Fig. 7c)
z=+t/2 (element i–j–k–l, shown in Fig. 7d).
In each case, we imagine a 2D strain element whose sides are parallel

to the x-axis and y-axis prior to deformation. As the plate is deformed, the
length of the element sides increases or decreases, in accordance with the
algebraic sign of strains exx and eyy, and the angle between adjacent faces of
Figure 7 Illustration of strains induced at the three through-thickness positions

z = t/2, 0, and +t/2 by pure bending (deformations shown greatly exag-
gerated for clarity).

the element changes from p/2 rad (i.e., 90j), in accordance with the algebraic
sign of cxy.
First, consider strain element i–j–k–l, located at z=+t/2 (Fig. 7d). For a
state of pure bending, the strains induced at this through thickness position
are given by Eq. (13):
exx jz ¼ t=2 ¼ tjxx =2
eyy jz ¼ t=2 ¼ tjyy =2
cxy jz ¼ t=2 ¼ tjxy =2
Assume for illustrative purposes that all curvatures are positive (jxx,jyy,
jxy > 0). This implies that all strains induced at z=+t/2 are algebraically
positive. A sketch of a deformed element that corresponds to these assump-
tions is shown (not to scale) in Fig. 7d. In the deformed condition, the lengths
of the element sides have increased because exx and eyy are positive, and angle
j–i–l has decreased because cxy is positive.
Now consider strain element e–f–g–h, located at the midplane of the
plate z=0. Because we have assumed a state of pure bending, the strains at the
midplane are zero, and consequently element e–f–g–h is not deformed, as
shown in Fig. 7c.
Finally, consider strain element a–b–c–d, located at z=t/2 (Fig. 7b).
Using Eq. (13), the strains induced at this position are:
exx jz ¼ t=2 ¼ tjxx =2
eyy jz ¼ t=2 ¼ tjyy =2
cxy jz ¼ t=2 ¼ tjxy =2
Because we have already assumed that all midplane curvatures are positive,
these results show that all strains induced at z=t/2 are algebraically
negative. A sketch of the deformed element that corresponds to this condition
is shown (not to scale) in Fig. 7b. Note that in this case, the lengths of the
element sides have decreased because exx and eyy are negative, and angle b–a–d
has increased because cxy is negative.
The deformed 2D strains elements shown in Fig. 7b–d are assembled to
create a sketch of the entire 3D element in Fig. 8. Note that, in accordance
with the Kirchhoff hypothesis, the four line segments that define the vertical
edges of the element (line segments a–e–i, b–f–j, c–g–k, and d–h–l) remain
straight lines after deformation. However, the transverse planes are no longer
plane after deformation. For example, plane b–j–h–c has been twisted during
deformation of the plate. Inspection of Figs. 7b–d and 8 reveals that trans-
verse planes do not remain plane after deformation due to curvature jxy. That
is, if jxy p 0, shear strain cxy varies with through-thickness position z, in
accordance with Eq. (13). It is this through-thickness variation in cxy that
leads to twisting of the transverse planes. For this reason, jxy is known as the
twist curvature.

Figure 8 A 3D strain element assembled from the 2D deformed elements
shown in Fig. 7b–d (deformations shown greatly exaggerated for clarity).
We will now repeat this process for a rectangular 3D element referenced

to the principal strain coordinate system, as shown in Fig. 9a. Once again, we
assume that the plate is subjected to pure bending. The principal strains
induced at any position z can therefore be calculated using Eq. (18). We
consider three 2D strain elements located at through-thickness positions
z=t/2, 0, and +t/2. A 2D sketch of the deformed strain elements at these
three positions is shown in Fig. 9b–d. Because the element is aligned with the
principal strain coordinate system, no shear strain is induced in any element
(i.e., all corner angles equal p/2 rad before and after deformation). Assuming,
for illustrative purposes, that both principal strains are positive (jp1,jp2>0),
the principal strains induced at z=+t/2 are tensile (Fig. 9d), whereas the
principal strains induced at z=t/2 are compressive (Fig. 9b). No deforma-
tions occur at z=0 because we have assumed pure bending and the midplane
is therefore the neutral surface. The deformed 2D strains elements shown in
Fig. 9b–d are assembled to create a sketch of the deformed 3D element in
Fig. 10. As before, the four line segments that define the vertical edges of the
element (line segments a–e–i, b–f–j, c–g–k, and d–h–l) remain straight lines
after deformation. However, in contrast to Fig. 8, the planes in which these
line segment lie remain plane after deformation. Twisting of these transverse
planes does not occur. When referenced to the coordinate system in which the
principal curvatures exist, the transverse planes of the strain element simply
rotate about the neutral surface.
A summary of the results presented in this section is as follows. We have
found that curvatures can be treated as second-order tensors, and can be
rotated from one coordinate system to another using the same process as that

Figure 9 Illustration of principal strains induced at the three through-thickness
positions z=t/2, 0, and +t/2 by pure bending; compare with Fig. 7 (deform-
ations shown greatly exaggerated for clarity).
Figure 10 A 3D strain element assembled from 2D deformed elements

referenced to the principal strain coordinate system; compare with Fig. 8
(deformations shown greatly exaggerated for clarity).

used to transform the strain or stress tensors. For a thin plate governed by the
Kirchhoff hypothesis, midplane curvatures transform according to Eq. (15).
In general, three midplane curvatures are induced in a thin plate: jxx, jyy, and
jxy. Curvature jxy is called the twist curvature because it represents a twisting
of a plane transverse to the midplane of the plate. The principal curvatures jp1
and jp1 are the maximum and minimum curvatures, respectively, induced at a
given point in a plate. Equation (17) gives the orientation of the coordinate
system in which the principal curvatures exists, and no twisting occurs in this
coordinate system (the twist curvature equals zero in the principal coordinate
system). For a thin plate in pure bending, the orientation of the principal
strain coordinate system is constant through the thickness of the plate, and
the principal curvatures are induced in this coordinate system.
The reader should note that the results in this section are valid for the
special case of pure bending. Some of the results presented above are not valid
for the case of general nonuniform plate bending. For example, if the mid-
plane is not the neutral surface (i.e., if eoxx,eoyy,coxy p 0), then it can be shown
that the orientation of the principal strain coordinate system is not constant
but rather varies as a function of z. However, even in this more general case,
midplane curvatures transform according to Eq. (15), and principal curva-
tures are given by Eq. (19).
A more detailed discussion of principal strains and curvatures under
general conditions will not be presented because these topics are not of
immediate interest. The results presented in this section for pure bending will
be applied in Chap. 8, where the topic of composite beams is considered.
4 STANDARD METHODS OF DESCRIBING COMPOSITE

LAMINATES
A magnified edge view of a thin composite laminate that contains n plies is
shown in Fig. 11. The figure is similar to the edge view of a thin plate shown in
Fig. 4, except that now the ply interface positions are shown. The thickness of
ply k will be denoted tk. The origin of the x–z axes lies at the geometrical
midsurface of the laminate, and so the outer surfaces of the laminate exist at
z=t/2 and z=+t/2, where t equals total thickness of the laminate. Total
thickness of the laminate equals the sum of all ply thicknesses: t ¼ Snk ¼ 1 tk .
Note that a ply interface does not necessarily exist at the midplane of the
laminate, as indicated in Fig. 11.
We will require a method of specifying the coordinate position of each
ply interface with respect to the laminate midplane. By convention, we will
denote the coordinate position of the outermost laminate surface in the
negative z-direction as position z0 (i.e., z0 u t/2). Note that z0 is always an
algebraically negative number. The coordinate position of the interface

Figure 11 An edge view of an n-ply laminate showing ply interface positions.
between plies 1 and 2 is denoted z1, and z1=z0+t1. Similarly, the coordinate
position of the interface between plies 2 and 3 is denoted z2, and z2=z1+t2,
etc. For an n-ply laminate, the outermost surface of the laminate in the
positive z-direction is will be labeled zn; obviously, zn=+t/2. Note that in all
cases, zn is an algebraically positive number. Also note that the total thickness
of the laminate equals (znz0), and the thickness of an individual ply k is
tk=(zkzk1). For example, the thickness of ply 2 is t2=(z2z1).
We also need a method of consistently describing the stacking sequence
of a composite laminate. That is, we need to develop a method of indicating
the orientation of the principal material coordinate system of each ply with
respect to the x-axis, and the order in which they appear. As discussed in
previous chapters, a ply may contain unidirectional fibers, or may consist of a
woven or braided fabric. In these latter two cases, there are two or more fiber
directions present within each ply, although the orientation of the principal
material coordinate system is always evident due to the symmetrical pattern of
the fiber architecture. For simplicity in the following discussion, it will be
assumed that all plies are composed of unidirectional fibers. In this case, the
angle between the principal material coordinate system and the x-axis is
equivalent to the angle between the fibers and the x-axis. Hence, in the
discussion to follow, we will simply refer to the ‘‘fiber angle’’ in each ply. It
should be understood that this angle actually refers to the orientation of the
principal material coordinate system. This terminology is adopted simply

because the phrase ‘‘fiber angle’’ is more concise than the phrase ‘‘principal
material coordinate system angle.’’
To describe the stacking sequence of a laminate, we list fiber angles
within square brackets ‘‘[’’ and ‘‘]’’. The fiber angle (in degrees) of ply 1 is
listed first, followed by the fiber angle of ply 2, ply 3, etc. Each fiber angle will
be separated by a slash ‘‘/’’. For example, a four-ply laminate consisting of
plies with fiber angles of 0j, 45j, 20j, and 90j is shown in Fig. 12a. This
laminate is denoted [0/45/20/90]T. The subscript ‘‘T’’ has been used to
indicate that the ‘‘total’’ laminate has been described (i.e., a fiber angle is listed
for all plies within the laminate within the square brackets). In practice, it is
common to encounter laminates with 10, 20, 30, or (in unusual cases) even
hundreds of plies. In such cases, it becomes very tedious to list all fiber angles
within the laminate. Fortunately, for many reasons (some of which will be
described later in this chapter), composite laminates are usually designed with
some systematic pattern of fiber angles, which allows us to abbreviate the
listing of ply fiber angles that appear within the laminate. It is easiest to
introduce these abbreviations with a series of examples. Consider the eight-
ply laminate shown in Fig. 12b. In this case, the fiber angles are (starting from
ply 1) 0j, +45j, 45j, 90j, 90j, 45j, +45j, and 0j. This is an example of a
symmetrical laminate because the fiber angles are symmetrical about the
laminate midplane. This laminate is denoted [0/F45/90]s. The subscript ‘‘s’’
indicates that the four fiber angles listed appear symmetrically about the
midplane, and hence a total of eight plies exist within the laminate, even
though only four angles are listed.
A nine-ply laminate containing fiber angles 0j, 30j, 60j, 10j, 45j, 10j,
60j, 30j, and 0j is shown in Fig. 12c. This laminate is symmetrical about the
geometrical midplane, but because an odd number of plies is present, the
midplane passes through the center of the 45j ply (ply 5). This laminate is
denoted [0/30/60/10/45]s. That is, a bar is used to indicate that the midplane
passes through the 45j ply, and hence ‘‘4 1/2’’ plies exist symmetrically about
the midplane of this laminate.
A 10-ply laminate containing fiber angles 20j, 30j, 30j, 20j, 0j, 0j,
20j, 30j, 30j, and 20j is shown in Fig. 12d. This laminate is symmetrical
about the midplane, but also contains a symmetrical pattern within both
halves of the laminate. In this case, the laminate is denoted [(20/30)s/0]s. The
subscript ‘‘s’’ appears twice: first to indicate that fiber angles 20j and 30j
appear symmetrically within one-half of the laminate, and the second to
indicate that the entire laminate is symmetrical about the midplane.
A final example is the 10-ply laminate shown in Fig. 12e. In this case, the
fiber angles are 20j, 30j, 20j, 30j, 0j, 0j, 30j, 20j, 30j, and 20j. This
laminate is denoted [(20/30)2/0]s, where the subscript ‘‘2’’ indicates that the
fiber pattern listed within the parentheses occurs twice. Note that this
laminate is similar but not identical to that shown in Fig. 12d.

Figure 12 Edge view of several composite laminates illustrating stacking
sequences: (a) [0/45/20/90]T, (b) [0/F45/90]s, (c) [0/30/60/10/45]s, (d)
[(20/30)s/0]s, (e) [(20/30)2/0]s.

Sample Problem 2
A ½(0/F30/90)2/45/20s laminate is fabricated using a graphite–epoxy material
system. Each ply has a thickness of 0.125 mm. Determine the number of plies
in the laminate, the total laminate thickness, and the z-coordinate of each ply
interface.
Solution. An ordered listing of all fiber angles that appear in the laminate is
as follows:
½0j; 30j; 30j; 90j; 0j; 30j; 30j; 90j; 45j; 20j; 45j; 90j; 30j; 30j; 0j; 90j; 30j; 30j; 0
z z z
ply 1 ply 10 ply 19
ðmidplaneÞ
The laminate contains a total of 19 plies, the fiber angles appear symmetrically
about the midplane of the laminate, and the midplane passes through the
center of the 20j ply. Because all plies are made of the same composite
material system, they all have the same thickness. The total laminate thickness
is therefore t = 19 (0.125 mm)=2.375 mm.
Ply interface positions are:
z0=t/2=1.1875 mm z1=z0+t1=1.0625 mm z2=z1+t2=0.9375 mm

z3=z2+t3=0.8125 mm z4=z3+t4=0.6875 mm z5=z4+t5=0.5625 mm
z18=z17+t18=1.0625 mm z19=z18+t19=1.1875 mm
Note that the total laminate thickness equals the difference between z19
and z0, as expected: t=z19z0=1.1875 mm(1.1875 mm)=2.375 mm.
5 CALCULATING PLY STRAINS AND STRESSES

The theory developed to this point allows calculation of the elastic strains and
stresses present at any through-thickness position within a multiangle com-
posite laminate subjected to known midplane strains and curvatures. A sum-
mary of how strains and stresses are calculated is as follows.
Laminate Description: A composite laminate is described by specifying:
The laminate stacking sequence (i.e., the number of plies within a
laminate and the fiber angles of each ply)
The material properties and thickness of each ply.

Note that the plies are not necessarily all of the same material type. For
example, some plies within a laminate may be of graphite–epoxy whereas
others may be glass–epoxy.
Once the stacking sequence and thickness of each ply have been
specified, the total laminate thickness and interface positions throughout
the laminate may be determined, as previously illustrated in Fig. 11. Also, the
transformed reduced stiffness matrix ½ Q can be calculated for each ply in
accordance with Eq. (31) of Chap. 5.
Ply Strains: Strains are calculated using the laminate strains and curvatures
(eoxx ; eoyy ; coxy) and (jxx,jyy,jxy), respectively, in accordance with the Kirchhoff
hypothesis (Eq. (12)). For example, the strains induced in a distance zk from
the laminate midplane are:
8 9 8 o 9 8 9
> e
< exx > = >
< xx > = >
< jxx >
=
eyy ¼ eoyy þ zk jyy
>
: ;
> >
: o > ; >
: >
;
cxy z¼z cxy jxy
k
Note that these strains are referenced to the x–y coordinate system. If desired,
these strains can be rotated from the x–y coordinate system to the ‘‘local’’ 1–2
coordinate system for each ply (defined by the ply fiber angle) using Eq. (44) of
Chap. 2:
8 9 8 9
< e11 =
:
e22
; j
c12 =2 z ¼ zk
¼ ½Tk
< exx =
:
eyy
; j
cxy =2 z¼zk
2 38 9
j
cos2 ðhÞ sin2 ðhÞ 2cosðhÞ sinðhÞ e
< xx >
> =
6 7
¼4 sin2 ðhÞ cos2 ðhÞ 2cosðhÞ sinðhÞ 5 eyy
>
: >
;
cosðhÞ sinðhÞ cosðhÞ sinðhÞ cos2 ðhÞ sin2 ðhÞ k cxy =2 z ¼ zk
Ply Stresses: Once ply strains are determined, ply stresses are calculated
using Hooke’s law, as discussed in Sec. 5.2. For example, the stresses induced
at a distance zk from the laminate midplane are calculated using Eq. (30) of
Chap. 5:
8 9 2 3 8 9
> rxx > Q11 Q12 Q16 > exx DTaxx DMbxx >
>
< >
= >
< >
=
6 7
ryy 6
¼ 4 Q12 Q22 Q26 5 7 eyy DTayy DMbyy
> > > >
>
: ;
> >
: ;
>
sxy z¼z k
Q16 Q26 Q66 z¼z cxy DTaxy DMbxy z ¼ z
k k
Note that the material properties used in this calculation (specifically, ½ Q , aij,
and bij) are properties of the ply that exists at position zk, and in particular are

functions of the ply fiber angle h. Because fiber angle generally varies from one
ply to the next, these material properties also vary from one ply to the next.
If desired, stresses can be rotated from the x–y coordinate system to the
‘‘local’’ 1–2 coordinate system for each ply using Eq. (20) of Chap. 2:
8 9 8 9
j j
< r11 = < rxx =
r ¼ ½Tz¼zk ryy
: 22 ; : ;
s12 z¼zk sxy z¼zk
A numerical example that illustrates these calculations is presented in
the following Sample Problem.
Sample Problem 3
Assume that the panel considered in Sample Problem 1 is actually an eight-ply
[0/30/90/30]s graphite–epoxy laminate. Assume that the laminate was
initially flat and stress-free (i.e., ignore possible preexisting stresses/strains
due to temperature and/or moisture changes). Determine the strains and
stresses induced at each ply interface. Use material properties listed in Table 3
of Chap. 3, and assume that the thickness of each ply is 0.125 mm.
Solution. From Sample Problem 1, the midplane strains and curvatures are:
eoxx ¼ 0 Am=m jxx ¼ 0:50 rad=m
eoyy ¼ 1300 Am=m jyy ¼ 0:40 rad=m
coxy ¼ 900 Arad jxy ¼ 0:20 rad=m
To determine ply interface positions, first note that the total laminate
thickness is:
t ¼ ð8 pliesÞð0:125 mmÞ ¼ 1:0 mm ¼ 0:001 m
A total of nine ply interface positions must be determined because there are
eight plies in the laminate. Following the numbering scheme discussed in Sec.
4 and referring to Fig. 11, ply interface positions are:
z0=t/2=(0.001 m)=0.000500 m
z1=z0+t1=0.000500 m+0.000125 m=0.000375 m
z2=z1+t2=0.000375 m+0.000125 m=0.000250 m
z3=z2+t3=0.000250 m+0.000125 m=0.000125 m
z4=z3+t4=.000125 m+0.000125 m=0.000000 m
z5=z4+t5=0.000000 m+0.000125 m=0.000125 m
z6=z5+t6=0.000125 m+0.000125 m=0.000250 m
z7=z6+t7=0.000250 m+0.000125 m=0.000375 m
z8=z7+t7=0.000375 m+0.000125 m=0.000500 m

Strain Calculations. Strains are calculated using Eq. (12), and can be
determined at any through-thickness position. Usually, strains of greatest
interest are those induced at the ply interface locations. For example, strains
present at the outer surface of ply 1 (i.e., strains present at zo=0.000500 m)
are:
8 9 8o 9 8 9 8 9 8 9
j
>e e j > 0 > 0:50 rad=m >
< xx>
= >
< xx >
= >
< xx>= >
< >
= >
< =
eyy ¼ eoyy þ z0 jyy ¼ 1300 106 m=m þ ð0:000500 mÞ 0:40 rad=m
: >
> ; : o >
> ; : >
> ; >
> >
> >
: >
;
: ;
cxy z¼z0 cxy jxy 900 106 m=m 0:20 rad=m
8 9 8 9
j
< exx = < 250 Am=m >
> =
eyy ¼ 1500 Am=m
: ; >
: >
;
cxy z¼z 1000 Arad
0
Similarly, strains present at the interface between plies 1 and 2 (i.e., strains
present at z1=0.000375 m) are:
8 9 8 o 9 8 9 8 9 8 9
j
> e > e > j 0 0:50 rad=m >
< xx >= >
< xx > = >
< xx >= >
< >
= >
< =
eyy ¼ eoyy þ z1 jyy ¼ 1300 106 m=m þ ð0:000375 mÞ 0:40 rad=m
>
: > ; >
> > > >
: > ; >
: >
; >
: >
;
cxy z¼z1 : coxy ; jxy 900 106 m=m 0:20 rad=m
8 9 8 9
j < 188 Am=m >

< exx = > =
eyy ¼ 1450 Am=m
: ; > >
cxy z¼z1 : 975 Arad ;
Strains present at all remaining interfaces are calculated in exactly the same
fashion. Strains calculated at all ply interfaces are summarized in Table 1 and
are plotted in Fig. 13. Note that all three strain components (exx, eyy, and cxy)
are predicted to be linearly distributed through the plate thickness. This linear
distribution is a direct consequence of the Kirchhoff hypothesis, which is a
good approximation as long as the plate is ‘‘thin.’’ In fact, identical strain
distributions would be predicted for any thin plate subjected to the midplane
strains and curvatures specified in Sample Problem 1. For example, we would
predict the identical strains if an aluminum plate were under consideration
rather than a laminated composite plate.
The strains listed in Table 1 and plotted in Fig. 13 are referenced to the
global x–y coordinate system. As will be seen, knowledge of ply strains
referenced to the local 1–2 coordinate system (defined by the fiber angle within
each ply) is often required. Transformation of the strain tensor from one

Table 1 Ply Interface Strains in a [0/30/90/30]s Graphite-Epoxy Laminate
Subjected to the Midplane Strains and Curvatures Discussed in Sample
Problem 1
z-coordinate (mm) exx (Am/m) eyy (Am/m) cxy (Arad)
0.500 250 1500 1000

0.375 188 145 975
0.250 125 1400 950
0.125 62 1350 925
0.0 0 1300 900
0.125 62 1250 875
0.250 125 1200 850
0.375 188 1150 825
0.500 250 1100 800
Strains are referenced to the x–y coordinate system.
coordinate system to another was reviewed in Chap. 2 and, in particular,

strains can be rotated from the x–y coordinate system to the 1–2 coordinate
system using Eq. (44) of Chap. 2. In practice, strains are usually calculated at
both the ‘‘top’’ and ‘‘bottom’’ interface for each ply. Example calculations for
plies 1 and 2 are listed below:
Ply 1. Because h1=0j, the x–y and 1–2 coordinate systems are coincident,
and therefore the description of the strain tensor is identical in both coor-
dinate systems. This can be confirmed through application of Eq. (44) of
Chap. 2:
Top interface:
8 9 ply 1 2 38 9
j j
> cos2 h1 sin2 h1 2 cos h1 sin h1 > exx > ply 1
< e11 > = 6 7< =
e22 ¼6
4 sin 2
h cos 2
h 2 cos h sin h 7
15
e yy
>
: >
;
1 1 1
>
: >
;
c12 =2 z ¼ z0 cos h1 sin h1 cos h1 sin h1 cos h1 sin h1
2 2 cxy =2 z ¼ z0
8 9 ply 1 2 38 9
j j
> 250 Am=m > ply1
< e11 > = 1 0 0 > < =
6 7
e22 ¼ 40 1 0 5 1500 Am=m
>
: >
; >
: >
;
c12 =2 z ¼ z0 0 0 1 ð1000 AradÞ=2 z ¼ z0
8 9 ply 1 8 9 ply 1
j < 250 Am=m >

j
< e11 >
> = > =
e22 ¼ 1500 Am=m
>
: >
; >
: >
;
c12 =2 z ¼ z0 1000 Arad z ¼ z0

Figure 13 Through-thickness strain plots dictated by the midplane strains and
curvatures discussed in Sample Problem 1. Strains are referenced to the x–y
coordinate system.

Bottom interface:
8 9 ply 1 2 38 9 ply 1
j j
> cos2 h1 sin2 h1 2cos h1 sin h1
< e11 > = 6 7
>
< exx >=
e22 ¼6
4 sin2 h1 cos2 h1 2cos h1 sin h1 7
5>
eyy
>
: >
; : >
;
c12 =2 z ¼ z1 cos h sin h cos h1 sin h1
2 2 cxy =2 z ¼ z1
1 1 cos h1 sin h1
8 9 ply 1 2 38 9
j j
> 188 Am=m > ply 1
< e11 > = 1 0 0 > < =
6 7
e22 ¼ 4 0 1 0 5 1450 Am=m
>
: >
; >
: >
;
c12 =2 z ¼ z1 0 0 1 ð975 AradÞ=2 z ¼ z1
8 9 ply 1 8 9 ply 1
j < 188 Am=m >

j
< e11 >
> = > =
e22 ¼ 1450 Am=m
>
: >
; >
: >
;
c12 z ¼ z1 975 Arad z ¼ z1
Ply 2. In this case, h2=30j and consequently the description of strain in the
x–y and 1–2 coordinate systems differs substantially. Applying Eq. (44) of
Chap. 2, we have:
Top interface:
8 9 ply 2 2 38 9
j j
> e > cos2 h2 sin2 h2 2 cos h2 sin h2 > exx > ply 2
< 11 = 6 7 < =
e22 ¼6
4 sin2 h2 cos2 h2 2 cos h2 sin h2 7
5
eyy
>
: >
; >
: >
;
c12 =2 z ¼ z1 cos h sin h cos h sin h cos2 h sin2 h cxy =2 z ¼ z1
2 2 2 2 2 2
8 9 ply 2 2 3
j
> cos2 ð30jÞ sin2 ð30jÞ 2 cosð30jÞ sinð30jÞ
< e11 > = 6 7
e22 ¼6
4 sin2 ð30jÞ cos2 ð30jÞ 2 cosð30jÞ sinð30jÞ 7
5
>
: >
;
c12 =2 z ¼ z1 cosð30jÞ sinð30jÞ cosð30jÞ sinð30jÞ cos ð30jÞ sin ð30jÞ
2 2
8 9
j
> 188 Am=m > ply 2
< =
1450 Am=m
>
: >
;
ð975 AradÞ=2 z ¼ z1
8 9 ply 2 2 38 9
j j
> 188 Am=m > ply 2
< e11 > = 0:750 0:250 0:866 > < =
6 7
e22 ¼ 4 0:250 0:750 0:866 5 1450 Am=m
>
: >
; >
: >
;
c12 =2 z ¼ z1 0:433 0:433 0:500 ð975 AradÞ=2 z ¼ z1
8 9 ply 2 8 9 ply 2
j < 200 Am=m >

j
< e11 >
> = > =
e22 ¼ 1463 Am=m
>
: >
; >
: >
;
c12 z ¼ z1 931 Arad z ¼ z1

Bottom interface:
8 9 ply 2 2 38 9
j j
> e > cos2 h2 sin2 h2 2 cos h2 sin h2 > exx > ply 2
< 11 = 6 7 e =
<
e22 ¼6
4 sin2 h2 cos2 h2 2 cos h2 sin h2 75>
yy
>
: >
; :c =2>;
c12 =2 z ¼ z2 cos h sin h cos h2 sin h2 cos2 h2 sin2 h2 xy
z ¼ z2
2 2
8 9ply 2 2 3
> cos2 ð30jÞ sin2 ð30jÞ 2 cosð30jÞ sinð30jÞ
< e11 > = 6 7

e22 ¼6
4 sin2 ð30jÞ cos2 ð30jÞ 2 cosð30jÞ sinð30jÞ 7
5
>
: ;
>
c12 =2 z ¼ z2 cosð30jÞ sinð30jÞ cosð30jÞ sinð30jÞ cos2 ð30jÞ sin2 ð30jÞ
8 9
j
> 125 Am=m > ply 2
< =
1400 Am=m
>
: >
;
ð950 AradÞ=2 z ¼ z2
8 9 ply 2 2 38 9
j j
> 125 Am=m > ply 2
< e11 > = 0:750 0:250 0:866 > < =
6 7
e22 ¼ 4 0:250 0:750 0:866 5 1400 Am=m
>
: >
; >
: >
;
c12 =2 z ¼ z2 0:433 0:433 0:500 ð950 AradÞ=2 z ¼ z2
8 9 ply 2 8 9 ply 2
j < 155 Am=m >

j
>
< e11 >
= > =
e22 ¼ 1430 Am=m
>
: >
; >
: >
;
c12 z ¼ z2 846 Arad z ¼ z2
Ply strains referenced to the local 1–2 coordinate systems at all interface
locations are summarized in Table 2 and plotted in Fig. 14. Comparing Figs.
12 and 13, it is apparent that the through-thickness strain distributions no
longer appear linear or continuous when referenced to the 1–2 coordinate
system. This is of course illusionary, in the sense that strains appear to be dis-
continuous only because the coordinate system used to describe the through-
thickness strain is varied from one ply to the next.
Stress Calculations. Because strains are now known at all ply interface
positions, we can calculate stresses at these locations using Eq. (30) of Chap.
5, with DT=DM=0. During these calculations, we will require the trans-
formed reduced stiffness matrix for each ply. Using graphite–epoxy material
properties from Table 2 of Chap. 3 and Eqs. (11) and (31) of Chap. 5, we find:
For 0j plies:
2 3
170:9 109 3:016 109 0
6 7
Q 0j plies ¼ 6
4 3:016 10 10:05 10
9 9
0 7ðPaÞ
5
0 0 13:00 10 9

Table 2 Ply Interface Strains in a [0/-30/90/30]s Graphite-Epoxy Laminate
Problem 1
Ply number z-coordinate (mm) e11 (Am/m) e22 (Am/m) c12 (Arad)
Ply 1 0.500 250 1500 1000

0.375 188 1450 975
Ply 2 0.375 200 1463 931
0.250 155 1430 846
Ply 3 0.250 1400 125 950
0.125 1350 63 925
Ply 4 0.125 691 596 1686
0.000 715 585 1576
Ply 5 0.000 715 585 1576
0.125 738 574 1466
Ply 6 0.125 1250 62 875
0.250 1200 125 850
Ply 7 0.250 26 1299 506
0.375 71 1267 421
Ply 8 0.375 188 1150 825
0.500 250 1100 800
Strains are referenced to the 1–2 coordinate system local to individual plies.
For 30j plies:

2 3
107:6 109 26:06 109 48:3 109
6 7
Q 30j plies
¼6
4 26:06 10
9
27:22 109 21:52 109 7
5ðPaÞ
48:3 109 21:52 109 36:05 109
For 90j plies:
2 3
10:05 109 3:016 109 0
6 7
Q 90j plies
¼6
4 3:016 10
9
170:9 109 0 7ðPaÞ
5
0 0 13:00 109
For 30j plies:
2 3
107:6 109 26:06 109 48:3 109
6 7
Q 30j plies ¼ 6
4 26:06 10
9
27:22 109 21:52 109 7
5ðPaÞ
48:3 109 21:52 109 36:05 109

Figure 14 Through-thickness strain plots dictated by the midplane strains and
curvatures discussed in Sample Problem 1. Strains are referenced to the 1–2
coordinate system.
Stresses present at the outer surface of ply 1 (i.e., strains present at
z0=0.000500 m) can now be calculated:
8 9 ply 1 2 3 8 9
j j j
>
< rxx >= Q11 Q12 Q16 ply 1 > < exx >
=
6 7
ryy ¼ 4 Q12 Q22 Q26 5 eyy
>
: >
; >
:c > ;
sxy z ¼ z0 Q Q Q z ¼ z0 xy z ¼ z0
16 26 66
8 9 ply 1 2 38 9
j
> 170:9 109 3:016 109 0 > 250 Am=m >
<rxx >= 6 7< =
ryy ¼6
4 3:016 10
9
10:05 109 0 7 1500 Am=m
5>
: >
> ; : >
;
sxy z ¼ z0 0 0 13:00 109 1000 Arad
8 9 ply 1 8 9
j
< rxx >
> = < 38:2 MPa >
> =
ryy ¼ 14:3 MPa
>
: >
; >
: >
;
sxy z ¼ z0 13 MPa
To calculate stresses at the interface between plies 1 and 2 (i.e., at z1 =
0.000375 m), we must specify whether we are interested in the stresses within
ply 1 or ply 2. That is, according to our idealized model, a ply interface is
treated as a plane of discontinuity in material properties. Ply 1 ‘‘ends’’ at z =
z1(), whereas ply 2 ‘‘begins’’ at z=z1(+). Hence, the stresses within ply 1 at z =
z1() are:
9 ply 1 2 Q 3
j
8 Q12 Q16 ply 1 8 9
j j
> r > 11 > exx >
< xx = 6 7 < =
ryy ¼64
Q12 Q22 Q26 7
5 e yy
>
: >
; >
: >
;
sxy z ¼ z1 Q16 Q26 Q66 cxy z ¼ z1
z ¼ z1
8 9 ply 1 2 38 9
j
> 170:9 109 3:016 109 0 188 Am=m >
<rxx >= 6 7
>
< =
ryy ¼6
4 3:016 10
9
10:05 109 0 7 1450 Am=m
5>
>
: ; > : >
;
sxy z ¼ z1 0 0 13:00 109 975 Arad
8 9 ply 1 8 9
j
>
< rxx >
= < 27:8 MPa >
> =
ryy ¼ 14:0 MPa
>
: >
; >
: >
;
sxy z ¼ z1 12:7 MPa
The stresses within ply 2 (a 30j ply) at z=z1(+) are:
8 9 ply 2 2 3 8 9
j j j
> Q11 Q12 Q16 ply 2 > exx >
< rxx >= 6 7 < =
ryy ¼6 Q
4 12 Q Q 7
26 5 e yy
>
: >
;
22
>
: >
;
sxy z ¼ z1 Q16 Q26 Q66 z ¼ z1 cxy z ¼ z1

8 9 ply 2 38 9
j
>
2 107:6 109 26:06 109 48:3 109 > 188 Am=m >
<rxx >= 6 7
>
< >
=
ryy ¼6 21:52 109 7 1450 Am=m
4 26:06 10 27:22 109
9
> 5>
: sxy >
; > 975 Arad >
9 :
>
;
z ¼ z1 48:3 109 21:52 109 36:05 10
8 9 ply 2 8 9
j
>
< rxx >
= >
< 29:5 MPa >
=
ryy ¼ 13:6 MPa
>
: >
; >
: >
;
Stresses are calculated at all remaining ply interfaces in exactly the same
fashion. Ply interface stresses are summarized in Table 3 and are plotted in
Fig. 15. Obviously, stresses are not linearly distributed through the thickness
of the laminate, even when referenced to the global x–y coordinate system. In
general, all stress components exhibit a sudden discontinuous change at all ply
interface positions. The abrupt
change
in stresses at ply interfaces is due to the
discontinuous change in the Q matrix from one ply to the next. In turn, the

discontinuous change in Q occurs because the fiber angle (in general)
changes from one ply to the next. Indeed, in this example problem, the same
fiber angle occurs in only two adjacent plies (namely, plies 4 and 5, both of
which have a fiber angle of 30j), and inspection of Fig. 15 shows that the
Table 3 Ply Interface Stresses in a [0/30/90/30]s Graphite-Epoxy Laminate

Subjected to the Midplane Strains and Curvatures Discussed in Sample Problem 1
Ply z-coordinate rxx ryy sxy Q A

number (mm) (MPa) (MPa) (MPa) (MPa) (MPa)2
Ply 1 0.500 38.2 14.3 13.0 23.9 715

0.375 27.8 14.0 12.7 13.8 550
Ply 2 0.375 29.3 13.6 13.0 15.7 567
0.250 22.7 14.4 10.1 8.3 429
Ply 3 0.250 2.97 239. 12.4 242 556
0.125 3.44 231 12.0 234 651
Ply 4 0.125 73.0 55.0 59.4 128 487
0.000 77.2 54.7 60.4 132 575
Ply 5 0.000 77.2 54.7 60.4 132 575
0.125 81.4 54.5 61.4 136 666
Ply 6 0.125 4.40 214 11.4 218 812
0.250 4.90 205 11.0 210 884
Ply 7 0.250 3.82 17.6 1.20 21.4 65.8
0.375 10.4 18.4 4.03 28.8 175
Ply 8 0.375 35.5 12.1 10.7 47.6 315
0.500 46.0 11.8 10.4 57.8 435
Stresses are referenced to the x–y coordinate system.

Figure 15 Through-thickness stress plots predicted for a [0/30/90/30]s
graphite-epoxy laminate subjected to the midplane strains and curvatures
discussed in Sample Problem 1. Stresses are referenced to the x–y coordinate
system.

interface between plies 4 and 5 is the only interface for which the stresses do
not change abruptly.
It has been mentioned that the linear strain distributions shown in Fig.
13 would be the same for any thin plate, regardless of the material the plate is
made of. The same statement cannot be made for stress distributions. In
general, through-thickness stress distributions for isotropic plates (e.g., an
isotropic aluminum plate) are linear and continuous, unless high nonlinear
stresses occur, in which case the stress distribution may not be linear but will
nevertheless be continuous. In contrast, the stress distributions in laminated
composite plates are usually discontinuous. The only conditions under which
a linear and continuous stress distribution is encountered is when: (a) the
laminate is subjected to elastic stress/strain levels, and (b) when the Q matrix
does not vary from one ply to the next (i.e., for unidirectional laminates in
which the fiber angle does not vary from one ply to the next).
Knowledge of ply stresses referenced to the local 1–2 coordinate system
(defined by the fiber angle within each ply) is often required. Transformation
of the stress tensor from one coordinate system to another was reviewed in
Chap. 2 and, in particular, stresses can be rotated from the x–y coordinate
system to the 1–2 coordinate system using Eq. (20) of Chap. 2. Typically,
stresses are calculated at both the ‘‘top’’ and ‘‘bottom’’ interfaces for all plies.
For example, rotation of the ply stresses that exist within ply 2 at the interface
between plies 1 and 2 (i.e., at z=z1=0.375 mm) proceeds as follows:
8 9ply 2 2 38 9
<r11= cos2 h2 sin2 h2 2 cos h2 sin h2 < rxx =ply 2

6 7
r22 ¼4 sin2 h2 cos2 h2 2 cos h2 sin h2 5 ryy
: ; : ;
s12 z ¼ z1 cos h2 sin h2 cos h2 sin h2 cos2 h2 sin2 h2 sxy z ¼ z1
8 9ply 2 2 3
< r11 = cos2 ð30jÞ sin2 ð30jÞ 2 cosð30jÞ sinð30jÞ
6 7
r ¼4 sin2 ð30jÞ cos2 ð30jÞ 2 cosð30jÞ sinð30jÞ 5
: 22 ;
s12 z ¼ z1 cosð30jÞ sinð30jÞ cosð30jÞ sinð30jÞ cos2 ð30jÞ sin2 ð30jÞ
8 9ply 2
< 29:3 MPa =
13:6 MPa
: ;
13:0 MPa z ¼ z1
8 9ply 2 2 38 9
< r11 = 0:750 0:250 0:866 < 29:3 MPa =ply 2

r ¼ 4 0:250 0:750 0:866 5 13:6 MPa
: 22 ; : ;
s12 z ¼ z1 0:433 0:433 0:500 13:0 MPa z ¼ z1
8 9ply 2 8 9ply 2
< r11 = < 29:8 MPa =
r22 ¼ 14:1 MPa
: ; : ;
s12 z ¼ z1 12:1 MPa z ¼ z1

Ply interface stresses referenced to local 1–2 coordinate systems are
summarized in Table 4 and plotted in Fig. 16. Once again, stresses are not
linearly distributed through the thickness of the laminate, and instead exhibit
a sudden discontinuous change at all ply interface positions.
Stress invariants can be used to confirm that the ply stresses referenced
to the x–y coordinate system and listed in Table 3 are equivalent to the ply
stresses referenced to the 1–2 coordinate system, as listed in Table 4. The
concept of ‘‘stress invariants’’ was discussed in Chap. 2. The stress invariants
for the case of plane stress are given by Eq. (22) of Chap. 2, repeated here for
convenience:
First stress invariant ¼ Q ¼ rxx þ ryy
Second stress invariant ¼ A ¼ rxx ryy s2xy ðrepeatedÞ ð2:22Þ
Third stress invariant ¼ C ¼ 0
For plane stress conditions, the third stress invariant always equals zero, and
so C cannot be used to evaluate whether two plane stress states are equivalent.
The first and second stress invariants, Q and A, respectively, have been
calculated using the ply stress components referenced to both the x–y and 1–2
Table 4 Ply Interface Stresses in a [0/-30/90/30]s Graphite-Epoxy Laminate

Problem 1
Ply z-coordinate r11 r22 s12 Q A

number (mm) (MPa) (MPa) (MPa) (MPa) (MPa)2
Ply 1 0.500 38.2 14.3 13.0 23.9 715

0.375 27.8 14.0 12.7 13.8 550
Ply 2 0.375 29.8 14.1 12.1 15.7 567
0.250 22.2 13.9 11.0 8.3 429
Ply 3 0.250 239 2.97 12.4 242 556
0.125 231 3.44 12.0 234 651
Ply 4 0.125 120 8.08 21.9 128 487
0.000 124 8.04 20.5 132 575
Ply 5 0.000 124 8.04 20.5 132 575
0.125 128 8.00 19.1 136 666
Ply 6 0.125 214 4.40 11.4 218 812
0.250 205 4.88 11.0 210 884
Ply 7 0.250 8.31 13.1 6.58 21.4 65.8
0.375 15.9 13.0 5.47 28.8 175
Ply 8 0.375 35.5 12.1 10.7 47.6 315
0.500 46.0 11.8 10.4 57.8 435
Stresses are referenced to the 1–2 coordinate system.

Figure 16 Through-thickness stress plots predicted for a [0/30/90/30]s
graphite–epoxy laminate subjected to the midplane strains and curvatures
discussed in Sample Problem 1. Stresses are referenced to the 1–2 coordinate
system.

coordinate systems. Values calculated for Q and A are included in the last two
columns of both Tables 3 and 4. Identical values are obtained in all cases,
indicating the equivalence of the ply stress states described using the two
different coordinate systems.
6 CLASSICAL LAMINATION THEORY (CLT)

Stress and moment resultants were introduced in Sec. 1. As was discussed, a
thin plate subjected to any combination of stress and moment resultants will
experience a state of plane stress. Deformations of a thin plate were then
considered in Sec. 2. There the Kirchhoff hypothesis was invoked, which
allows us to calculate the in-plane strains induced at any location through the
thickness of a thin plate. In this section, we will combine the material presented
in Secs. 1 and 2, as well as certain material presented in Chap. 5. This will lead
to the ability to relate stress and moment resultants to the resulting strains (and
hence stresses) induced within a thin composite plate. This combination of
analysis tools is commonly known as classical lamination theory (CLT).
Stress and moment resultants represent the mechanical loads applied to
a laminate. Obviously then, stress and moment resultants will induce strains
within the laminate. However, strains may also be induced by environmental
factors as well, as discussed in earlier chapters. Of particular importance for
polymeric composite laminates are strains due to a change in temperature
(DT ) and/or strains due to a change in moisture content (DM ). To simplify
our discussion, we will first develop CLT by assuming that constant environ-
mental conditions exist (i.e., we will initially assume DT=DM=0). We will
then consider how to account for a change in temperature and/or a change in
moisture content.
6.1 Constant Environmental Conditions

The stresses rxx induced in a thin composite laminate are related to stress
resultant Nxx in accordance with Eq. (1a), repeated here for convenience:
Z t=2
Nxx ¼ rxx dz ðrepeatedÞ ð1aÞ
t=2
The composite laminate consists of n plies, and the fiber angle may vary from
one ply to the next. The stresses in any ply (say, in ply number k) are related to
ply strains in accordance with Eq. (30) of Chap. 5, which, for DT=DM=0,
becomes:
8 9 2 3 8 9
< rxx >
> = Q11 Q12 Q16 > < exx >
=
6 7
ryy ¼ 4 Q21 Q22 Q26 5 eyy ð20Þ
>
: >
; >
: >
;
sxy k Q Q Q cxy k
61 62 66 k

The subscript k in Eq. (20) indicates that the stresses, transformed reduced
stiffness matrix, and strains are all for ply number k, where 1 V k V n and
n = number of plies in the laminate.
From Eq. (20), the stress rxx induced in ply k is:

ðrxx Þk ¼ Q11 exx þ Q12 eyy þ Q16 cxy k
Substituting this relationship into Eq. (1a), we have:
Z t=2

Nxx ¼ Q11 exx þ Q12 eyy þ Q16 cxy k dz ð21Þ
t=2
The strains induced in ply k can be related to the midplane strains and
curvatures via the Kirchhoff hypothesis, in accordance with Eq. (11) or Eq.
(12). Substituting Eqs. (11) and (12) into Eq. (21), we obtain:
Z t=2 n o
Nxx ¼ Q11 eoxx þ Q12 eoyy þ Q16 coxy þ zQ11 jxx þ zQ12 jyy þ zQ16 jxy dz
t=2
ð22Þ
We cannot integrate Eq. (22) directly because the integrand is a discontinuous
function of z. That is, the transformed reduced stiffness terms Q11 , Q12 , and
Q16 are all directly related to the ply material properties and fiber angle h (see
Eq. (31) of Chap. 5). Because the ply material and/or fiber angle may change
from one ply to the next, the transformed reduced stiffness terms also change,
and hence are discontinuous functions of z. Note, however, that the midplane
strains and curvatures are not functions of z, but instead are constants for a
given laminate. Hence, they may be brought out from under the integral sign.
Equation (15) can therefore be broken into six individual integrals:
Z t=2 Z t=2

Nxx ¼ exxo
Q11 k dz þ eyy
o
Q12 k dz
t=2 t=2
Z Z
t=2 t=2
þ coxy Q16 k
dz þ jxx Q11 k
dz ð23Þ
t=2 t=2
Z t=2
þ jyy z Q12 k dz þ jxy z Q16 k dz
t=2
Because the transformed stiffness terms are constant over each ply thickness,
each of the six integrals in Eq. (23) can be evaluated in a ‘‘piecewise’’ fashion:
Z Z Z
z1 z2 z3
Nxx ¼ exx Q11 1 dz þ Q11 2 dz þ Q11 3 dz þ : : :
o
z0 z1 z2
Z Z
zn1 zn
þ Q11 n1
dz þ Q11 n
dz
zn2 zn1

Z Z Z
z1 z2 z3
þeoyy Q12 1
dz þ Q12 2 dz þ Q12 3
dz þ : : :
z0 z1 z2
Z Z
zn1 zn
þ Q12 n1
dz þ Q12 n
dz
zn2 zn1
Z Z Z
z1 z2 z3
þcoxy Q16 1
dz þ Q16 2
dz þ Q16 3 dz þ : : :
z0 z1 z2
Z Z
zn1 zn
þ Q16 n1
dz þ Q16 n
dz
zn2 zn1
Z Z Z
z1 z2 z3
þjxx Q11 1
zdz þ Q11 2 zdz þ Q11 3 zdz þ : : :
z0 z1 z2
Z Z
zn1 zn
þ Q11 n1
zdz þ Q11 n
zdz
zn2 zn1
Z Z Z
z1 z2 z3
þjyy Q12 1
zdz þ Q12 2
zdz þ Q12 3 zdz þ : : :
z0 z1 z2
Z Z
zn1 zn
þ Q12 n1 zdz þ Q12 n zdz
zn2 zn1
Z Z Z
z1 z2 z3
þjxy Q16 1
zdz þ Q16 2
zdz þ Q16 3
zdz þ : : :
z0 z1 z2
Z Z
zn1 zn
þ Q16 n1
zdz þ Q16 n
zdz ð24Þ
zn2 zn1
Although Eq. (24) may appear daunting at first, closer inspection reveals that
evaluation of Eq. (24) is actually a simple matter. All integrals that appear in
Eq. (24) are of one of the following two forms, both of which are easily
evaluated:
Z zk
dz ¼ ðzk zk1 Þ
zk1
or
Z zk
1 2
zdz ¼ ðz z2k1 Þ
zk1 2 k
Hence, evaluating all integrals that appear in Eq. (24), we obtain:
n
Nxx ¼ eoxx Q11 1 ½z1 z0 þ Q11 2 ½z2 z1 þ Q11 3 ½z3 z2 þ : : :
o
þ Q11 n ½zn zn1
n
þeoyy Q12 1 ½z1 z0 þ Q12 2 ½z2 z1 þ Q12 3 ½z3 z2 þ : : :

o
þ Q12 n ½zn zn1
n
þcoxy Q16 1 ½z1 z0 þ Q16 2 ½z2 z1 þ Q16 3 ½z3 z2 þ : : :
o
þ Q16 n ½zn zn1
1 n
þ jxx Q11 1 z21 z20 þ Q11 2 z22 z21 þ Q11 3 z23 z22
2
2 o
þ ::þ Q
: z z2
11 n n n1
1 n
þ jyy Q12 1 z21 z20 þ Q12 2 z22 z21 þ Q12 3 z23 z22
2
2 o
þ ::þ Q
: z z2
12 n n n1
1 n
þ jxy Q16 1 z21 z20 þ Q16 2 z22 z21 þ Q16 3 z23 z22
2
o
þ . . . þ Q16 n z2n z2n1 ð25Þ
Equation (25) can be simplified substantially by defining the following terms:

n
A11 ¼ Q11 1 ½z1 z0 þ Q11 2 ½z2 z1 þ Q11 3 ½z3 z2 þ : : :
o
þ Q11 n ½zn zn1
n
A12 ¼ Q12 1 ½z1 z0 þ Q12 2 ½z2 z1 þ Q12 3 ½z3 z2 þ : : :
o
þ Q12 n ½zn zn1
n
A16 ¼ Q16 1 ½z1 z0 þ Q16 2 ½z2 z1 þ Q16 3 ½z3 z2 þ : : :
o
þ Q16 n ½zn zn1
1 n
B11 ¼ Q11 1 z21 z20 þ Q11 2 z22 z21 þ Q11 3 z23 z22 þ : : :
2
o
þ Q11 n z2n z2n1
1 n
B12 ¼ Q12 1 z21 z20 þ Q12 2 z22 z21 þ Q12 3 z23 z22 þ : : :
2
o
þ Q12 n z2n z2n1
1 n
B16 ¼ Q16 1 z21 z20 þ Q16 2 z22 z21 þ Q16 3 z23 z22 þ : : :
2
o
þ Q16 n z2n z2n1

With these definitions, Eq. (25), becomes:
Nxx ¼ A11 eoxx þ A12 eoyy þ A16 coxy þ B11 jxx þ B12 jyy þ B16 jxy ð26aÞ
Following an entirely analogous procedure for stress resultants Nyy and Nxy,
it can be shown that:
Nyy ¼ A21 eoxx þ A22 eoyy þ A26 coxy þ B21 jxx þ B22 jyy þ B26 jxy ð26bÞ
Nxy ¼ A61 eoxx þ A62 eoyy þ A66 coxy þ B61 jxx þ B62 jyy þ B66 jxy ð26cÞ
where:
X
n
Aij ¼ Qij k ðzk zk1 Þ ð27aÞ
k¼1
1X n 2
Bij ¼ Q z z2k1 ð27bÞ
2 k¼1 ij k k
and i, j = 1, 2, or 6. Because subscripts i and j may take on one of three values,

both Aij and Bij can be written as 3 3 matrices. Also, recall that the
transformed reduced stiffness matrix is symmetrical (see Eq. (31) of Chap. 5).
Hence, both Aij and Bij are also symmetrical:
2 3 2 3
A11 A12 A16 A11 A12 A16
6 7 6 7
Aij ¼ 4 A21 A22 A26 5 ¼ 4 A12 A22 A26 5
A61 A62 A66 A16 A26 A66
2 3 2 3
B11 B12 B16 B11 B12 B16
6 7 6 7
Bij ¼ 4 B21 B22 B26 5 ¼ 4 B12 B22 B26 5
B61 B62 B66 B16 B26 B66
Equation (26a) (26b) (26c) can be written in matrix form as follows:
8 o 9
>
> e > >
> xx >
8 9 2 3>
>
> e o >>
>
> > >
> >
N A A A B B B < o >
yy
< xx = 11 12 16 11 12 16 =
6 7 cxy
Nyy ¼ 4 A12 A22 A26 B12 B22 B26 5 ð28Þ
>
: >
; >
> jxx >>
Nxy A16 A26 A66 B16 B26 B66 > >
>
>
>
>
>
> j >
>
>
:
yy >
;
jxy
To summarize our results to this point, Eq. (28) relates the stress resultants
applied to a composite laminate to the resulting midplane strains and
curvatures via the Aij and Bij matrices. The values of each term within the
Aij and Bij matrices depend on the material properties and fiber angle of each
ply (i.e., they depend on terms within the Qij matrix) as well as the stacking

sequence (i.e., the distance zk of each ply from the laminate midplane), in
accordance with Eq. (27a) (27b). In practice then, if the midplane strains and
curvatures induced in a laminate under constant environmental conditions
are measured, then the stress resultants that caused these strains and
curvatures can be calculated using Eq. (28).
This entire process must now be repeated for the moment resultants.
The stresses rxx induced in a thin composite laminate are related to moment
resultant Mxx in accordance with Eq. (2a), repeated here for convenience:
Z t=2
Mxx ¼ rxx zdz ðrepeatedÞ ð2aÞ
t=2
Substituting the expression for rxx from Eq. (20), we obtain:

Z t=2

Mxx ¼ Q11 exx þ Q12 eyy þ Q16 cxy k zdz ð29Þ
t=2
Each strain that appears in Eq. (29) can be related to the midplane strains and
curvatures via the Kirchhoff hypothesis. Hence, substituting either Eq. (11) or
Eq. (12), we have:
Z t=2 n o
Mxx ¼ zQ11 eoxx þ zQ12 eoyy þ zQ16 coxy þ z2 Q11 jxx þ z2 Q12 jyy þ z2 Q16 jxy dz
t=2
ð30Þ
Equation (30) is similar to Eq. (22). Once again, this integral cannot be
evaluated directly because the integrand is a discontinuous function of z.
However, (a) noting that the midplane strains and curvatures are not
functions of z and can be brought outside the integral sign, and then (b)
evaluating the integral in a ‘‘piecewise’’ fashion through the thickness of the
laminate, we obtain:
Z Z Z
z1 z2 z3
Mxx ¼ exx Q11 1 zdz þ Q11 2 zdz þ Q11 3 zdz þ : : : :
o
z0 z1 z2
Z Z
zn1 zn
þ Q11 n1
zdz þ Q11 n
zdz
zn2 zn1
Z z1 Z z2 Z
z3
þeoyy Q12 1
zdz þ Q12 2
zdz þ Q12 3
zdz þ : : : :
z0 z1 z2
Z Z
zn1 zn
þ Q12 n1
zdz þ Q12 n
zdz
zn2 zn1
Z Z Z
z1 z2 z3
þcoxy Q16 1
zdz þ Q16 2
zdz þ Q16 3 zdz þ : : : :
z0 z1 z2

Z Z
zn1 zn
þ Q16 n1 zdz þ Q16 n
zdz
zn2 zn1
Z Z Z
z1 z2 z3
þjxx Q11 1
z dz þ Q11
2
2
z dz þ Q11 3
2
z2 dz þ : : : :
z0 z1 z2
Z Z
zn1 zn
þ Q11 n1 z dz þ Q11
2
n
2
z dz
zn2 zn1
Z Z Z
z1 z2 z3
þjyy Q12 1
z2 dz þ Q12 2
z2 dz þ Q12 3 z2 dz þ : : : :
z0 z1 z2
Z Z
zn1 zn
þ Q12 n1 z dz þ Q12
2
n
2
z dz
zn2 zn1
Z Z Z
z1 z2 z3
þjxy Q16 1
z2 dz þ Q16 2
z2 dz þ Q16 3 z2 dz þ : : : :
z0 z1 z2
Z Z
zn1 zn
þ Q16 n1 z2 dz þ Q16 n z2 dz ð31Þ
zn2 zn1
The piecewise integrals that appear in Eq. (31) are of one of the following two
forms, both of which are easily evaluated:
Z zk
1
zdz ¼ z2k z2k1
zk1 2
or
Z zk
1 3
z2 dz ¼ zk z3k1
zk1 3
Hence, evaluating all integrals, we obtain:
1 o n
Mxx ¼ exx Q11 1 z21 z20 þ Q11 2 z22 z21 þ Q11 3 z23 z22
2
2 o
þ: : : : þ Q z z2
11 n n n1
1 n

þ eoyy Q12 1 z21 z20 þ Q12 2 z22 z21 þ Q12 3 z23 z22
2
o
þ : : : þ Q12 n z2n z2n1
:
1 n
þ coxy Q16 1 z21 z20 þ Q16 2 z22 z21 þ Q16 3 z23 z22
2
o
þ : : : þ Q16 n z2n z2n1
:

1 n
þ jxx Q11 1 z31 z30 þ Q11 2 z32 z31 þ Q11 3 z33 z32
3
3 o
þ: : : : þ Q z z3
11 n n n1
1 n

þ jyy Q12 1 z31 z30 þ Q12 2 z32 z31 þ Q12 3 z33 z32
3
3 o
þ :::þ Q
: z z3
12 n n n1
1 n
þ jxy Q16 1 z31 z30 þ Q16 2 z32 z31 þ Q16 3 z33 z32
3
o
þ : : : : þ Q16 n z3n z3n1 ð32Þ
The first three quantities on the right-hand side of the equality sign involve the
previously defined terms B11, B12, and B16. We now define three new terms,
associated with the last three quantities:
1 n
D11 ¼ Q11 1 z31 z30 þ Q11 2 z32 z31 þ Q11 3 z33 z32 þ : : : :
3
o
þ Q11 n z3n z3n1
1 n
D12 ¼ Q12 1 z31 z30 þ Q12 2 z32 z31 þ Q12 3 z33 z32 þ : : : :
3
o
þ Q12 n z3n z3n1
1 n
D16 ¼ Q16 1 z31 z30 þ Q16 2 z32 z31 þ Q16 3 z33 z32 þ : : : :
3
o
þ Q16 n z3n z3n1
Hence, Eq. (32) can be written in the following simplified form:

Mxx ¼ B11 eoxx þ B12 eoyy þ B16 coxy þ D11 jxx þ D12 jyy þ D16 jxy ð33aÞ
Following an entirely equivalent procedure for Myy and Mxy, it can be shown
that:
Myy ¼ B21 eoxx þ B12 eoyy þ B26 coxy þ D21 jxx þ D22 jyy þ D26 jxy ð33bÞ
Mxy ¼ B61 eoxx þ B62 eoyy þ B66 coxy þ D61 jxx þ D62 jyy þ D66 jxy ð33cÞ
The Bij terms that appear in Eqs. (33a) (33b) (33c) have been previously
encountered and are given by Eq. (27b). The new terms Dij are given by:
1X n
Dij ¼ Qij k z3k z3k1 ð34Þ
3 k¼1

The Dij terms can be written as a symmetrical 3 3 matrix:
2 3 2 3
D11 D12 D16 D11 D12 D16
6 7 6 7
Dij ¼ 4 D21 D22 D26 5 ¼ 4 D12 D22 D26 5
D61 D62 D66 D16 D26 D66
Equations (33) can be written in matrix form as follows:
8 o 9
>
> e > >
> xx >
8 9 2 3>
>
> e o >>
>
> > >
< Mxx > B11 B12 B16 D11 D12 D16 > < o >
yy
= =
6 7 cxy
Myy ¼ 4 B12 B22 B26 D12 D22 D26 5 ð35Þ
>
: >
; >
> >
Mxy B16 B26 B66 D16 D26 D66 > > jxx > >
>
>
> jyy >>
>
> >
>
: ;
jxy
Equation (35) relates the moment resultants applied to a composite
laminate to the resulting midplane strains and curvatures via the Bij and Dij
matrices. The value of each term within the Bij and Dij matrices depends on the
material properties and fiber angle of each ply (i.e., they depend on terms
within the Qij matrix) as well as the stacking sequence (i.e., the distance zk of
each ply from the laminate midplane), in accordance with Eqs. (27b) and (34).
In practice then, if the midplane strains and curvatures induced in a laminate
under constant environmental conditions are measured, then the moment
resultants that caused these strains and curvatures can be calculated using Eq.
(35).
It is customary to combine Eqs. (28) and (35) and express them together
in matrix form:
8 9 2 38 o 9
>
> Nxx >> A11 A12 A16 B11 B12 B16 > > exx >
>
>
> >
> 6 7 >
> >
>
>
> N >
> 6 A A A B B B 7 >
>
26 > eyy >
o >
>
> yy >
> 6 12 22 26 12 22
7>> >
>
> >
< Nxy = 6 A 7 < >
=
6 16 A26 A66 B 16 B 26 B 66 7 o
cxy
¼6 7 ð36Þ
>
> Mxx > > 6 B11 B12 B16 D11 D12 D16 7> > jxx >
>
>
> >
> 6 7 > >
>
> >
> 6 7>> >
> jyy >
>
>
> Myy > > 4 B12 B22 B26 D12 D22 D26 5> > >
>
>
: >
; >
: >
;
Mxy B16 B26 B66 D16 D26 D66 jxy
Equation (36) will sometimes be written in abbreviated form as:

o
N A B e
¼
M B D j
The 6 6 array that appears in Eq. (36) is called the ‘‘ABD matrix.’’ Because
each of the individual matrices that make up the total ABD matrix is in itself
symmetrical (e.g., A12=A21, B12=B21, D12=D21, etc.), the entire ABD matrix
is also symmetrical.

It should be noted that the above results are applicable to thin laminates
fabricated using any combination of ply materials. Because the Aij, Bij, and Dij
matrices are each calculated based on a summation over all plies, and
individual ply properties (represented by the Qij matrix) are embedded within
these summations, both ply material type and fiber angle can vary from one
ply to the next. Hence, the ABD matrix for any thin plate can be calculated
using (Eq. (27a), (27b), and (34). For example, the ABD matrix for ‘‘hybrid’’
laminates (i.e., laminates fabricated using two different prepreg material
systems) are calculated using (Eq. (27a), (27b), and (34).
The Aij matrix relates in-plane stress resultants to in-plane midplane
strains. For this reason, the Aij terms are called extensional stiffnesses.
Similarly, the Dij matrix relates moment resultants to midplane curvatures,
and elements within the Dij matrix are therefore called bending stiffness. The
Bij matrix relates in-plane stress resultants to midplane curvatures, and also
relates moment resultants to the in-plane midplane strains. The Bij terms are
called coupling stiffnesses. For an isotropic plate, the coupling stiffnesses are
always zero.
The stress and moment resultants can be thought of as ‘‘stress-like’’
quantities because they are directly related to the stresses through the
thickness of the laminate via Eqs. (1) and (2). On the other hand, the midplane
strains and curvatures are ‘‘strain-like’’ quantities because they can be used to
calculate the strains at any position through the thickness of the laminate via
Eqs. (11) and (12). Hence, Eq. (36) relates ‘‘stress-like’’ quantities to ‘‘strain-
like’’ quantities, and in this sense can be thought of as ‘‘Hooke’s law’’ for a
composite laminate.
Equation (36) is in convenient form if we measure midplane strains and
curvatures and wish to calculate the stress and moment resultants that caused
these strains and curvatures. Suppose, instead, that the stress and moment
resultants are known and we wish to calculate the midplane strain and
curvatures that will be caused by these known loads. In this case, we must
invert Eq. (36) to obtain a relationship of the form:
o

e A B 1 N
¼ ð37Þ
j B D M
In this text, the inverse of the ABD matrix will be called the abd matrix:
" #
a b A B 1
¼
b d B D
Methods of inverting the [ABD] matrix analytically are discussed in
several composite texts, including Refs. 1, 2, and 3. However, in practice, the
ABD matrix is most often inverted numerically with the aid of a digital

computer because many commercial software packages (e.g., MATLAB,
Maple, Mathematica, etc.) that can invert a 66 matrix routinely are
available nowadays.
Written out in full, Eq. (37) is:
8 o 9 2 38 9
> exx > a11 a12 a16 b11 b12 b16 > Nxx >
>
> >
> 6 > >
>
> o >> 7>>
>
>
>
>
>
> e >
yy >
6 a 12 a 22 a 26 b 21 b 22 b 26 7 >
> N yy >
>
>
> > 6 7 > >
< o = 6 a16 a26 a66 b61 b62 b66 7< Nxy >
> > =
cxy 6 7
¼6 7 ð38Þ
>
> jxx >> 6 b11 b21 b61 d11 d12 d16 7> Mxx >
>
> >
> 6 7>>
>
>
>
>
>
> > 6 7> M >
> jyy >
> > 4 b12 b22 b62 d12 d22 d26 5>
> > yy >
> >
>
>
: >
; >
: >
;
jxy b16 b26 b66 d16 d26 d66 M xy
The reader should carefully inspect the subscripts used in Eq. (38). Note that
the [abd] matrix is symmetrical. Furthermore, the individual 3 3 matrices
that appear in the upper left-hand quadrant and lower right-hand quadrant of
the [abd] matrix, aij and dij, respectively, are also symmetrical. However, the
3 3 matrix that appears in the upper right-hand quadrant is not symmetrical
(b12 p b21, b16 p b61, and b26 p b62). Also, the 3 3 matrix in the lower left-
hand quadrant is the transpose of the 3 3 matrix that appears in the upper
right-hand quadrant.
Example Problem 4
Determine the [ABD] and [abd] matrices for a [30/0/90]T graphite-epoxy
laminate. Use material properties listed for graphite-epoxy in Table 3 of
Chap. 3, and assume that each ply has a thickness of 0.125 mm.
Solution. A side view of the laminate is shown in Fig. 17. The total laminate
thickness t = 3 (0.125 mm) = 0.375 mm. Because all three plies are of the
same material, the thickness of each ply is identical: t1 = t2 = t3 = 0.125 mm.
Figure 17 Side view of the [30/0/90]T laminate considered in Sample Problem 4.

Note that because an odd number of plies are used, the origin of the x–y–z
coordinate system exists at the midplane of ply 2. The ply interface
coordinates can be calculated as:
z0 ¼ t=2 ¼ ð0:375 mmÞ=2 ¼ 0:1875 mm ¼ 0:0001875 m

z1 ¼ z0 þ t1 ¼ 0:1875 mm þ 0:125 mm ¼ 0:0625 mm ¼ 0:0000625 m
z2 ¼ z1 þ t2 ¼ 0:0625 mm þ 0:125 mm ¼ 0:0625 mm ¼ 0:0000625 m
z3 ¼ z2 þ t3 ¼ 0:0625 mm þ 0:125 mm ¼ 0:1875 mm ¼ 0:0001875 m
We will also require the transformed reduced stiffness matrix for each ply.
Elements of the [Q]k matrices are calculated using Eq. (31) of Chap. 5* and are
equal to:
For ply 1 (the 30j ply):
2 3
Q11 Q12 Q16
6 7
6 7
Q 30j ply
¼ 6 Q12 Q22 Q26 7
4 5
Q16 Q26 Q66
2 3
107:6 109 26:06 109 48:13 109
6 7
¼6
4 26:06 10
9
27:22 109 21:52 109 7
5ðPaÞ
48:13 109 21:52 109 36:05 109
2 3
Q11 Q12 Q16
6 7
6 7
Q 0B ply
¼ 6 Q12 Q22 Q26 7
4 5
Q16 Q26 Q66
2 3
107:9 109 3:016 109 0
6 7
¼64 3:016 10 10:05 10
9 9
0 7ðPaÞ
5
0 0 13:00 109

* The Q matrix for a 30j graphite-epoxy ply was calculated as a part of Example Problem 5.6
of Chap. 5.

2 3
Q11 Q12 Q16
6 7
6 7
Q 90B ply ¼ 6 Q12 Q22 Q26 7
4 5
Q16 Q26 Q66
2 3
10:05 109 3:016 109 0
6 7
¼64 3:016 10 170:9 10
9 9
0 7ðPaÞ
5
0 0 13:00 109
We can now calculate each member of the Aij, Bij, and Dij matrices, in
accordance with (Eq. (27a), (27b), and (34), respectively.
. Using Eq. (27a), element A11 is calculated as follows:
X
3
A11 ¼ Q11 k
ðzk zk1 Þ
k¼1

A11 ¼ Q11 1 ðz1 z0 Þ þ Q11 2 ðz2 z1 Þ þ Q11 3 ðz3 z2 Þ

A11 ¼ 107:6 109 ð:0000625 þ 0:0001875Þ þ 170:9 109

ð0:0000625 þ 0:0000625Þ þ 10:05 109
ð0:0001875 0:0000625Þ
A11 ¼ 36:07 106 Pa m
The remaining elements of the Aij matrix are found in similar fashion:
2 3
36:07 4:012 6:016
6 7
Aij ¼ 4 4:012 26:02 2:690 5 106 ðPa mÞ
6:016 2:690 7:756
. Using Eq. (27b), element B11 is calculated as follows:
1X 3 2
B11 ¼ Q z z2k1
2 k ¼ 1 11 k k
1 h i
B11 ¼ Q11 1 z21 z20 þ Q11 2 z22 z21 þ Q11 3 z23 z22
2
1 h n o
B11 ¼ 107:6 109 ð:0000625Þ2 þ ð0:0001875Þ2
2
n o
þ 170:9 109 ð0:0000625Þ2 ð0:0000625Þ2
n oi
þ 10:05 109 ð0:0001875Þ2 ð0:0000625Þ2

B11 ¼ 1:524 103 Pa m2
The remaining elements of the Bij matrix are found in similar fashion:
2 3
1:524 0:3601 0:7521
6 7
Bij ¼ 4 0:3601 2:245 0:3362 5 103 Pa m2
0:7521 0:3362 0:3601
In passing, in this example, it appears that B12 is numerically equal to B66. This
is not true, in general. In this problem, the apparent numerical equivalence is
due to the fact that only four significant digits have been used. Nevertheless,
for laminates produced using a single material system, it is often (but not
always) the case that B12 c B66. This common occurrence can be traced to the
fact the functional form and magnitude of Q 12 and Q 66 are similar (see Eq.
(31) of Chap. 5). Because B12 and B66 are directly related to Q12 and Q66,
respectively, their values are often nearly identical. Also, in Sec. 6.2, it will be
seen that all elements within the Bij matrix are zero for symmetrical laminates.
Hence, for symmetrical laminates, these two terms are, in fact, numerically
equal, that is, B12=B66=0 for symmetrical laminates.
. Using Eq. (34), element D11 is calculated as follows:
1X 3
D11 ¼ Q11 k z3k z3k 1
3 k¼1
1 h i
D11 ¼ Q11 1 z31 z30 þ Q11 2 z32 z31 þ Q11 3 z33 z32
3
1 h n o
D11 ¼ 107:6 109 ð:0000625Þ3 ð0:0001875Þ3
3
n o
þ 170:9 109 þ ð0:0000625Þ3 ð0:0000625Þ3
n oi
þ 10:05 109 ð0:0001875Þ3 ð0:0000625Þ3
D11 ¼ 0:2767 Pa m3
The remaining elements of the Dij matrix are found in similar fashion:
2 3
0:2767 0:0620 0:1018
6 7
Dij ¼ 6
4 0:0620 2:513 0:0455 7
5 Pa m
3
0:1018 0:0455 0:1059

The [ABD] matrix can now be assembled:
2 3
36:07 106 4:012 106 6:016 106 1524 360:1 752:1
6 7
6 4:012 106 26:02 106 2:690 106 360:1 336:2 7
2245
6 7
6 7
6 6:016 106 2:690 106 7:756 106 752:1 336:2 360:1 7
½ABD ¼ 6
6
7
7
6 1524 360:1 752:1 0:2767 0:0620 0:1018 7
6 7
6 360:1 336:2 0:0455 7
4 2245 0:0620 2:513 5
752:1 336:2 360:1 0:1018 0:0455 0:1059
The [abd] matrix is obtained by inverting the [ABD] matrix, and is found
to be:
2 3
3:757 108 1:964 109 1:038 108 1:440 104 3:905 106 8:513 105
6 7
6 1:964 109 1:037 107 4:234 108 1:866 105 6:361 104 4:268 104 7
6 7
6 7
6 5 7
6 1:038 108 4:234 108 2:004 107 3:661 104 3:251 104 1:851 10 7
6 7
½abd ¼ 6 7
6 1:440 104 1:866 105 3:661 104 7:064 3:122 102 4:572 7
6 7
6 7
6 3:905 106 6:361 104 3:251 104 3:122 102 3:620 7
6 6:429 7
4 5
8:513 105 4:268 104 1:851 105 4:572 3:620 17:41
Example Problem 5
A [30/0/90]T graphite-epoxy laminate is subjected to the following stress and
moment resultants:
Nxx ¼ 50 kN=m Nyy ¼ 10 kN=m Nxy ¼ 0 N=m
Mxx ¼ 1 N m=m Myy ¼ 1 N m=m Mxy ¼ 0 N m=m
Determine the following quantities caused by these stress and moment
resultants:
(a) Midplane strains and curvatures
(b) Ply strains relative to the x–y coordinate system
(c) Ply stresses relative to the x–y coordinate system.
Use material properties listed for graphite-epoxy in Table 3 of Chap. 3
and assume that each ply has a thickness of 0.125 mm.
Solution. Note that this is the same laminate considered in Example
Problem 4. A side view of the laminate appears in Fig. 17.
(a) Midplane strains and curvatures. The [abd] matrix for this laminate
was calculated as a part of Example Problem 4. Hence, midplane strains

and curvature may be obtained through application of Eq. (38), which
becomes:
8 o 9 2 3
>
> exx >
> 3:757 108 1:964 109 1:038 108 1:440 104 3:905 106 8:513 105
>
> >
> 6 7
>
> >
> 6 7
>
> o > 6 1:964 109 1:037 107 4:234 108 1:866 105 6:361 104 4:628 104 7
> eyy >
> >
> 6 7
>
> >
> 6 7
< co >
> = 6 1:038 108 4:234 108 2:004 107 3:661 104 3:251 104 1:851 105 7
6 7
xy
¼6 7
>
> > 6 1:440 104 7
>
> jxx >
>
> 6 1:866 105 3:661 104 7:064 3:122 102 4:572 7
>
> >
> 6 7
>
> > 6
> 7
>
> j >
> 6 3:905 106 6:361 104 3:251 104 3:122 102 6:429 3:620 7
>
>
yy
>
> 4 5
>
: >
;
jxy 8:513 105 4:628 10 4
1:851 10 5
4:572 3:620 17:41
8 9
>
> 50 103 >>
>
> >
>
>
> >
3 >
>
> 10 10 > >
>
> >
>
>
> >
>
< 0 =

>
> >
>
>
> 1 >
>
>
> >
>
>
> >
>
>
> 1 >
>
>
> >
>
: ;
0
Completing this matrix multiplication, we obtain:
8 o 9 8 9
> exx > > 2039 Am=m >
>
> > > >
> o >
> > >
> >
> 518 Am=m >
>
>
>
> e >
> >
> >
>
>
<
yy > >
= < >
o 55 Arad =
cxy ¼
>
> jxx > > >
> 14:48 m1 >>
>
> >
> >
> >
>
> >
> >
> 1 >
>
>
>
> j >
> > 0:096 m >
; > >
yy
: : ;
jxy 1:323 m1
(b) Ply strains relative to the x–y coordinate system. Ply strains may now be
calculated using Eq. (12). For example, strains present at the outer surface of
ply 1 (i.e., strains present at z0 = 0.0001875 m) are:
8 9 8 o 9 8 9 8 9
8 9 > 2038 106 m=m
j
> e > 14:48 rad=m >
< exx >
= < xx >
> = < jxx = >
< >
= >
< =
eyy ¼ eoyy þ z0 jyy ¼ 518 106 m=m þ ð 0:0001875mÞ 0:096 rad=m
>
:c > ; >
: o ; > : ; >
> >
> >
: >
;
jxy : ; 1:328 rad=m
xy
z ¼ z0
cxy 55 106 m=m
8 9
j
8 9
< exx = < 677 Am=m >
> =
e ¼ 536 Am=m
: yy ; >
: >
;
cxy 194 Arad
z ¼ z0
Strains calculated at the remaining ply interface positions are summarized in

Table 5.

(c) Ply stresses relative to the x–y coordinate system. The Q matrix for all
plies was calculated as a part of Example Problem 4. Ply stresses may now be

Table 5 Ply Interface Strains in a [30/0/90] Graphite-Epoxy Laminate Caused
by the Stress and Moment Resultants Specified in Example Problem 5
0.1875 677 536 194

0.0625 1133 524 28
0.0625 2943 512 137
0.1875 4753 500 303
calculated using Eq. (30) of Chap. 5, with DT = DM = 0. The stresses present

at the outer surface of ply 1 (i.e., at z=z0) are:
8 9 ply 1 2 3
j j j
>
< rxx >= Q11 Q12 Q16 ply 1 8 9
< exx =
6 7
ryy ¼ 4 Q12 Q22 Q26 5 eyy
>
: >
; :c ;
sxy z ¼ z0 xy
Q16 Q26 Q66 z ¼ z0 z ¼ z0
8 9 ply 1 2 38 9
j
> 107:6 109 26:06 109 48:13 109 > 677 106 >
< rxx >= < =
6 7
ryy ¼ 4 26:06 109 27:22 109 21:52 109 5 536 106
>
: >
; >
: >
;
sxy z ¼ z0 48:13 109 21:52 109 36:05 109 194 106
8 9 ply 1 8 9
j < 77:5 MPa >

< rxx >
> = > =
ryy ¼ 28:1 MPa
>
: >
; >
: >
;
Stresses calculated at remaining ply interface positions are summarized
in Table 6.
Table 6 Ply Interface Stresses in a [30/0/90]T Graphite-Epoxy Laminate Caused

by the Stress and Moment Resultants Specified in Example Problem 5
Ply number z-coordinate (mm) rxx (MPa) ryy (MPa) sxy (MPa)
Ply 1 0.1875 77.5 28.1 37.1

0.0625 109.7 15.9 44.3
Ply 2 0.0625 192.1 1.85 0.366
0.0625 501.5 3.73 1.78
Ply 3 0.0625 28.0 78.6 1.78
0.1875 46.3 71.1 3.93

6.2 Including Changes in Environmental Conditions
Recall that we simplified the analysis leading up to Eq. (36) by assuming that
DT=DM=0. We will now consider how to predict the behavior of a laminate
subjected to a change in temperature and/or moisture content as well as
external mechanical loads.
To begin, the stresses in any ply (say, in ply number k) are related to ply
strains in accordance with Eq. (30) of Chap. 5:
8 9 2 3 8 9
>
> rxx >
> Q11 Q12 Q16 > > exx DTaxx DMbxx > >
< = 6 7 < =
ryy ¼6 Q Q Q 7 e DTa DMb ðrepeatedÞ ð5:30Þ
> 4 21 26 5 > yy yy yy
>
> >
22
> >
>
: ; : ;
sxy k Q Q Q cxy DTaxy DMbxy k
61 62 66 k
Stress rxx in ply k is given by:

rxx ¼ Q11 fexx DTaxx DMbxx g þ Q12 eyy DTayy DMbxx
ð39Þ
þ Q16 cxy DTaxy DMbxy
Stress resultant Nxx is related to rxx via Eq. (1a). Substituting Eq. (39) into Eq.
(1a), we have:
Z t=2

Nxx ¼ Q11 exx þ Q12 eyy þ Q16 cxy k dz
t=2
Z t=2
DT Q11 axx þ Q12 ayy þ Q16 cxy k
dz ð40Þ
t=2
Z t=2
DM Q11 bxx þ Q12 byy þ Q16 bxy k
dz
t=2
The first integral on the right-hand side of the equality sign is identical to Eq.
(21), and after evaluation (using the same techniques as previously described)
will result in Eq. (26a). The second and third integrals were not previously
encountered because they involve DT and DM, which were previously
assumed to equal zero. Using methods similar to those used previously, it
can be shown that the second integral may be written as:
Z t=2

DT Q11 axx þ Q12 ayy þ Q16 cxy k dz
t=2
X
n
¼ DT Q11 axx þ Q12 ayy þ Q16 axy k ½zk zk1
k¼1
T
This quantity is called a thermal stress resultant, and will be denoted Nxx .
That is,
Xn
NTxx ¼ DT Q11 axx þ Q12 ayy þ Q16 axy k ½zk zk1 ð41aÞ
k¼1

Similarly, the third integral in Eq. (40) can be evaluated to give the moisture
stress resultant, denoted NM xx:
X
n n o
NMxx ¼ DM Q b
11 xx þ Q b
12 yy þ Q b
16 xy k ½z k z k1 ð42aÞ
k¼1
Hence, after evaluating all integrals, Eq. (40) may be written as:
Nxx ¼ A11 eoxx þ A12 eoyy þ A16 coxy þ B11 jxx þ B12 jyy þ B16 jxy
ð43aÞ
NTxx NM
xx
This result should be compared to Eq. (26a). It will be seen that the inclusion
of temperature and/or moisture changes has resulted in the addition of two
new terms (NTxx and NM xx); otherwise, our earlier results remain unchanged.
If an analogous procedure is now followed for the remaining stress and
moment resultants, using Eqs. (1b), (1c), (2a), (2b), and (2c), five additional
thermal stress/moment resultants and five additional moisture stress/moment
resultants will be identified, as follows:
X n
NTyy u DT Q12 axx þ Q22 ayy þ Q26 axy k ½zk zk1 ð41bÞ
k¼1
Xn
NTxy u DT Q16 axx þ Q26 ayy þ Q66 axy k ½zk zk1 ð41cÞ
k¼1
DT Xn
MTxx u Q11 axx þ Q12 ayy þ Q16 axy k z2k z2k1 ð41dÞ
2 k¼1
DT Xn
MTyy u Q12 axx þ Q22 ayy þ Q26 axy k z2k z2k1 ð41eÞ
2 k¼1
DT X n
MTxy u Q16 axx þ Q26 ayy þ Q66 axy k z2k z2k1 ð41f Þ
2 k¼1
Xn n o
NM
yy u DM Q12 bxx þ Q22 byy þ Q26 bxy k ½zk zk1 ð42bÞ
k¼1
Xn n o
xy u DM
NM Q16 bxx þ Q26 byy þ Q66 bxy k ½zk zk1 ð42cÞ
k¼1
Xn n o
DM
xx u
MM Q11 bxx þ Q12 byy þ Q16 bxy k z2k z2k1 ð42dÞ
2 k¼1
n n o
DM X
MM
yy u Q12 bxx þ Q22 byy þ Q26 bxy k z2k z2k1 ð42eÞ
2 k¼1
n n o
DM X
MM
xy u Q16 bxx þ Q26 byy þ Q66 bxy k z2k z2k1 ð42f Þ
2 k¼1

In each case, the corresponding thermal and moisture resultants will be
subtracted from the right-hand side of Equations (26a) (26b) (26c) and
(33a) (33b) (33c)) and (26a) (26b) (26c) and (33a) (33b) (33c). Finally, the
response of a composite laminate subjected to mechanical loads, a change in
temperature, and a change in moisture content can be written in a form similar
to Eq. (36):
8 9 2 38 9 8 T 9 8 M 9
> Nxx > A11 A12 A16 B11 B12 B16 > eoxx > > > Nxx > > >
> Nxx > >
>
> >
> > > > > > >
>
> >
> 6 7>>
>
>
>
>
>
>
>
>
>
>
>
>
> M >
>
>
>
> N >
yy > 6A A A B B B 7 >
26 > eyy >
o > >
> N T
>
yy > >
> N yy >>
>
> > 6
>
12 22 26 12 22
7>> > >
> > > >
> > >
>
>
< Nxy >= 6 7>< >
= >
< T >= >
< M > =
6A16 A26 A66 B16 B26 B66 7 cxy o N N
¼6 7 xy xy
6 7
>
>Mxx > > 6B11 B12 B16 D11 D12 D16 7> > jxx >> >
> MTxx >> >
> MMxx >
>
>
> >
> 6 7>> >
> >
> >
> >
> >
>
>
> > > > > > > >
>Myy >
> > 6
> 4 B B B D D D 7>
5 >
> j >
>
>
>
>
> M T >>
>
>
>
> M M> >
yy >
>
> >
> 12 22 26 12 22 26 >
>
yy
>
> > yy > > >
: ; : ; > > T >
: > >
; > M>
: >
;
Mxy B16 B26 B66 D16 D26 D66 j xy M xy M xy
ð44Þ
Equation (44) will sometimes be abbreviated as:

o ( T ) ( M )
N A B e N N
¼
M B D j MT MM
Equation (44) is comparable to Eq. (36), except we have now included the
effects due to a change in temperature and/or moisture content. Equation (44)
can be viewed as ‘‘Hooke’s law’’ for a composite laminate, in the sense that it
may be used to relate stresslike quantities (i.e., stress and moment resultants)
to strainlike quantities (i.e., midplane strains and curvatures). Inverting Eq.
(44), we obtain:
8 o 9 2 38 9
> exx > a11 a12 a16 b11 b12 b16 > Nxx þ NTxx þ NM xx >
>
> >
> 6 7>
> >
>
> > 6 a12 >
> N þ NT þ NM >
>
>
> eoyy >
>
> 6 a22 a26 b21 b22 b26 7
7>
> yy >
>
>
>
> >
> 6 7>
>
yy yy >
>
> >
< co = 6 a16 >
< N þ NT þ NM > =
6 a26 a66 b61 b62 b66 7
7 xy
xy
¼6 7
xy xy
ð45Þ
>j >
> > 6 b11 d16 7> M þ MT þ MM >
>
> xx >
> 6 b21 b61 d11 d12 7>
>
> xx xx
>
xx >
>
>
> >
> 6 7>
> >
>
> j >
> 6 b12 d26 7>
> M >>
þ þ yy >
> yy
> 4 b22 b62 d12 d22 >
5> M M T
M >
>
: >
; :
yy yy
M ;
>
jxy b16 b26 b66 d16 d26 d66 M xy þ M T
xy þ M xy
where, as before:

1
a b A B
¼
b d B D

A subtlety embedded fact within the preceding discussion is that most
composites are subjected to a significant state of stress prior to the application
of any external mechanical loading. That is, most modern composite material
systems are cured at an elevated temperature (common cure temperatures are
either 120jC or 175jC), and are nominally stress-free at the cure temperature.
Once the polymerization process is complete, the composite is cooled to room
temperatures (say, 20jC) and, consequently, the composite experiences a
uniform change in temperature of DT = 100jC or 155jC during cooldown.
In general, this change in temperature results in thermal stress and/or moment
resultants to develop, causing thermal stresses within all plies of the laminate.
These thermal stresses can be quite high, and contribute toward failure of the
laminate.*
A further complicating factor is related to measurement of strains. In
most practical situations, strain measurement devices (e.g., resistance foil
strain gages) are bonded to a composite material or structure after cooldown to
room temperature. Hence, in practice, the reference state of a strain measure-
ment device mounted on a laminate at room temperature does not necessarily
correspond to the stress-free (or strain-free) state of the composite. This
complication will be further explored in Chap. 7. At this point, it will simply
be noted that a significant difficulty arises when prediction of nonlinear
behavior (or more generally, the prediction of composite failure) is required
based on measured laminate strains.
Example Problem 6
A [30/0/90]T graphite-epoxy laminate is cured at 175jC and then cooled to
room temperature (20 jC). Determine:
(c) Ply stresses relative to the x–y coordinate system
which are induced during cooldown. Use material properties listed for
graphite-epoxy in Table 3 of Chap. 3, assume that each ply has a thickness
of 0.125 mm, and assume no change in moisture content (i.e., assume DM=0).
* Determination of the ‘‘stress-free temperature’’ is actually more complex than is implied here.
It is true that thermal stresses begin to develop as cooldown begins, but because polymeric
materials exhibit viscoelastic characteristics at these elevated temperatures, the matrix will
creep, initially relieving thermal stresses somewhat. As temperature is decreased further, the
viscoelastic nature of the matrix is rapidly decreased, and thermal stresses develop as described.
A second factor is that all polymers exhibit some shrinkage during the polymerization process
(see Sec. 1.2), and this shrinkage results in additional stresses similar to thermal stresses. As a
rule of thumb, the stress-free temperature is often estimated to be 20–50jC below the final cure
temperature. Nevertheless, this complication will be ignored in this text; it will be assumed that
the final cure temperature defines the stress-free temperature.

Solution. Note that this is the same laminate considered in Sample Problem
4. A side view of the laminate appears in Fig. 17.
(a) Midplane strains and curvatures. The laminate has experienced a
change in temperature DT = (20–175) = 155jC and, consequently, is sub-
jected thermal stress and moment resultants. However, no external loads are
applied and there has been no change in moisture content; therefore, the stress
and moment resultants and the moisture stress and moment results are zero:
8 9 8 M 9 8 9
>
> Nxx >> >
> Nxx >> >
> 0>>
>
> >
> >
> >
> >
> >
>
>
> >
> >
> M >
> >
> >
>
>
> N >
> >
> N >
> >
> 0 >
> >
yy yy
>
> > > > > > >
>
<Nxy = < NM = < 0 > > =
xy
¼ ¼
>
> Mxx >> >
> MM > >
xx > > 0>>
>
> >
> >
> >
> >
> >
>
>
> >
> >
> >
> >
> >
>
>
> M >
> > > M M
> yy > >
> > >
> >
0 >
>
>
yy
> >
> >
:M ; : MM ; : 0 > > > ;
xy xy
The effective thermal expansion coefficients for each ply are calculated using
Eq. (25) of Chap. 5, repeated here for convenience:
ayy ¼ a11 sin2 ðhÞ þ a22 cos2 ðhÞ ð5:25Þ
axy ¼ 2cosðhÞsinðhÞða11 a22 Þ
From Table 3 of Chap. 3, the thermal expansion coefficients for graphite-
epoxy (relative to the 1–2 coordinate system) are a11 = 0.9 Am/m jC and
a22 = 27Am/m jC. Therefore:
aðxx
1Þ
¼ ð0:9 Am=m B CÞ cos2 ð30B Þ þ ð27 Am=m B CÞ sin2 ð30B Þ
¼ 6:08 Am=m B C
aðyy1Þ ¼ ð0:9 Am=m B CÞ sin2 ð30B Þ þ ð27 Am=m B CÞ cos2 ð30B Þ
¼ 20:0 Am=m B C
aðxy1Þ ¼ 2cosð30Þ sinð30Þ½ð0:9 27ÞAm=m B C ¼ 24:2 Arad=B C
aðxx
2Þ
¼ 0:9 Am=m B C
¼ 27:0 Am=m B C
aðxy2Þ ¼ 2cosð0B Þ sinð0B Þ½ð0:9 27ÞAm=m B C ¼ 0 Arad=B C

aðxx
3Þ
¼ 27:0 Am=m B C
¼ 0:9 Am=m B C
aðxy3Þ ¼ 2cosð90B Þ sinð90B Þ½ð0:9 27ÞAm=m B C ¼ 0 Arad=B C
Both the ply interface positions as well as the Qij matrices for each ply were
calculated as a part of Example Problem 4. Hence, we now have all the
information needed to calculate the thermal stress and moment resultants,
using Eqs. (41a)–(41f). For example, Eq. (41a) is evaluated as follows:
Xn
NTxx u DT Q11 axx þ Q12 ayy þ Q16 axy k ½zk zk1
k¼1
n
NTxx ¼ DT Q11 axx þ Q12 ayy þ Q16 axy 1 ½z1 z0

þ Q11 axx þ Q12 ayy þ Q16 axy 2 ½z2 z1
o
þ Q11 axx þ Q12 ayy þ Q16 axy 3 ½z3 z2
n
NTxx ¼ ð155Þ 107:6 109 6:08 106 þ 26:06 109 20:0 106

þ 48:13 109 24:2 106 ð0:0625 þ 0:1875Þ 103

þ 170:9 109 0:9 106 þ 3:016 109 27:0 106

þð0Þð0Þ ð0:0625 þ 0:0625Þ 103 þ 10:05 109 27 106
o
þ 3:016 109 0:9 106 þ ð0Þð0Þ ð0:1875 0:0625Þ 103
NTxx ¼ 4060 N=m

The remaining thermal stress and moment resultants are calculated in similar
fashion, eventually resulting in:
8 T 9 8 9
>
> Nxx >> > 4060 N=m >
>
> T > >
> > >
> >
>
>
> > > 7360 >
>
>
> N yy >
> >
> N=m >
>
>
> > > >
< NT = < 2860 N=m >
> > =
xy
¼
>
> MTxx >> >
> 0:62 N m=m >
>
>
> >
> >
> >
>
>
> >
> >
> >
> Myy >
>
T > > > 0:62 N m=m > >
>
>
> > > >
: MT > ; : ;
xy 0:36 N m=m

We can now calculate midplane strains and curvature using Eq. (45), which
becomes:*
8 o 9 2 3
>
> e > 3:757 108 1:964 109 1:038 108 1:440 104 3:905 106 8:513 105
> xx >
> >
> 6 7
> >
> 6 1:964 109
>
>
> eoyy >
> >
> 6 1:037 107 4:234 108 1:866 105 6:361 104 4:628 104 77
>
> >
> 6 7
>
> > 6 7
< co > = 6 1:038 108 4:234 108 2:004 107 3:661 104 3:251 104 1:851 105 7
xy 6 7
¼6 7
>
> >
> 6 1:440 104 1:866 105 3:661 104 3:122 102 4:572 7
>
> j xx >
> 6 7:064 7
>
> >
> 6 7
>
> >
> 6 7
>
> jyy >> 6 3:905 106 6:361 104 3:251 104 3:122 102 6:429 3:620 7
>
> >
> 4 5
>
: >
;
jxy 8:513 105 4:628 104 1:851 105 4:572 3:620 17:41
8 9
> 4060 >
>
> >
>
>
> >
>
>
> 7360 >
>
>
>
>
> >
>
< 2860 >
> =

>
> >
> 0:62 >
> >
>
>
> >
>
>
> >
> 0:62 >
> >
>
>
: >
;
0:36
8 o 9 8 9
> exx > > 285 Am=m >
>
> o > > >
> >
>
>
> >
> >
> 1424 Am=m >
>
>
> e >
yy > >
> >
>
< o = < 908 Arad >
> > > =
cxy ¼
>
> jxx >> >
> 2:16 m1 >>
>
> >
> >
> >
>
>
> >
> >
> 1 >
>
>
> j yy > >
> :
> 10:9 m >
>
: ; 1
;
jxy 9:4 m
ply 1 (i.e., strains present at zo=0.0001875 m) are:
8 9 8 o 9 8 9 8 9
6
< 285 10 m=m >
> e e
< xx > >
< xx > >
< xx >= >
j
= = =
eyy ¼ eoyy þ z0 jyy ¼ 1424 106 m=m
>
: ;
> : o >
> ; >
: >
; >
: >
;
c xy z ¼ z0
cxy jxy 908 106 m=m
8 9
< 2:16 rad=m >
> =
þ ð0:0001875 mÞ 10:9 rad=m
>
: >
;
9:4 rad=m
* The [abd] matrix for a [30/0/90]T graphite-epoxy laminate was calculated in Sample Problem 3.

8 9 8 9
< exx = < 120 Am=m =
eyy ¼ 3468 Am=m
: c ; : ;
xy
z ¼ z0 2672 Am=m
Strains calculated at the remaining ply interface positions are summa-
rized in Table 7.
(c) Ply stresses relative to the x–y coordinate system. Ply stresses may now
be calculated using Eq. (30) of Chap. 5, with DM=0. The stresses present at
the outer surface of ply 1 are (i.e., at z=z0):
8 9ply 1 2 3ply 1 8 9
> rxx > Q11 Q12 Q16 > exx DTaxx >
>
< >
= >
< =
>
6 7
ryy ¼6 Q
4 12
Q22 Q26 7
5 eyy DTayy
>
> >
> >
> >
>
: ; : ;
sxy z ¼ z0
Q16 Q26 Q66
z ¼ z0
cxy DTaxy z ¼ z0
8 9ply 1 2 3
> 107:6 109 26:06 109 48:13 109
< rxx >
=
6 7
ryy ¼ 4 26:06 109 27:22 109 21:52 109 5
>
: >
;
sxy z¼z0 48:13 109 21:52 109 36:05 109
8 9
>
> ½ð120Þ ð155Þð6:08Þ 106 >>
< =
½ð3468Þ ð155Þð20:0Þ 10 6
>
> >
>
: ;
½ð2672Þ ð155Þð24:2Þ 106
8 9ply 1 8 53 MPa 9
< rxx = >
< >
=
ryy ¼ 5:2 MPa
: ; > >
sxy z ¼ z0 : 4:8 MPa ;
Stresses calculated at the remaining plies and ply interface positions are
summarized in Table 8.
Table 7 Ply Interface Strains in a [30/0/90]T Graphite-Epoxy

Laminate Caused by a Cooldown from 175jC to 20jC
0.1875 120 3468 2672

0.0625 150 2100 1500
0.0625 420 750 320
0.1875 690 620 860

Table 8 Ply Interface Stresses in a [30/0/90]T Graphite-Epoxy
Laminate Caused by a Cooldown from 175jC to 20jC
Ply 1 0.1875 53 5.2 4.8

0.0625 3.1 0.53 21
Ply 2 0.0625 43 20 19
0.0625 85 33 4.1
Ply 3 0.0625 35 140 4.1
0.1875 37 92 11
Example Problem 7
A [30/0/90]T graphite-epoxy laminate is cured at 175 jC and cooled to room
temperature (20 jC). Initially, the moisture content of the laminate is zero.
However, the laminate is subjected to a humid environment for several weeks,
resulting in an increase of moisture content of 0.5% (by weight). Determine:
(c) Ply stresses relative to the x–y coordinate system
which are present following the increase in moisture content. Use material
properties listed for graphite-epoxy in Table 3 of Chap. 3, and assume that
each ply has a thickness of 0.125 mm.
Solution. Note that this is the same laminate considered in Sample Problem
6, and the midplane strains and curvatures, ply strains, and ply stresses that
will be induced immediately upon cooldown by the change in temperature
have already been calculated. These quantities will all be modified due to the
slow diffusion of water molecules into the epoxy matrix.
(a) Midplane strains and curvatures. The laminate has experienced a change
in moisture content DM=+0.5% and, consequently, is subjected moisture
stress and moment resultants. The effective moisture expansion coefficients
for each ply are calculated using Eq. (28) of Chap. 5, repeated here for
convenience:

byy ¼ b11 sin2 ðhÞ þ b22 cos2 ðhÞ ð5:28Þ
bxy ¼ 2cosðhÞsinðhÞðb11 b22 Þ

From Table 3 of Chap. 3, the moisture expansion coefficients for graphite-
epoxy (relative to the 1–2 coordinate system) are b11=150 Am/m %M and
b22=4800 Am/m %M. Therefore:
bð1Þ B B
xx ¼ ð150 Am=m %MÞ cos ð30 Þ þ ð4800 Am=m %MÞ sin ð30 Þ
2 2
¼ 1310 Am=m %M
bð1Þ
yy ¼ ð150 Am=m %MÞsin2 ð30B Þ þ ð4800 Am=m %MÞcos2 ð30B Þ
¼ 3640 Am=m %M
bð1Þ
xy ¼ 2cosð30Þsinð30Þ½ð150 4800ÞAm=m %M ¼ 4030 Arad=%M

bð2Þ 2 B 2 B
¼ 150 Am=m %M
bð2Þ
yy ¼ ð150 Am=m %MÞ sin2 ð0B Þ þ ð4800 Am=m %MÞ cos2 ð0B Þ
¼ 4800 Am=m %M
bð2Þ
xy ¼ 2cosð0B Þ sinð0B Þ½ð150 4800ÞAm=m %M ¼ 0 Arad=%M
bð3Þ B B
2 2
¼ 4800 Am=m %M
bð3Þ
yy ¼ ð150 Am=m %MÞ sin2 ð90B Þ þ ð4800 Am=m %MÞ cos2 ð90B Þ
¼ 150 Am=m %M
bð3Þ
xy ¼ 2cosð90B Þsinð90B Þ½ð150 4800Þ Am=m %M ¼ 0 Arad=%M
Both the ply interface positions as well at the [Q] matrices for each ply were
calculated as a part of Example Problem 4. Hence, we now have all the
information needed to calculate the moisture stress and moment resultants,
using Eqs. (42a)–(42f). For example, Eq. (42a) is evaluated as follows:
X3 n o
NMxx ¼ DM Q11 bxx þ Q12 byy þ Q16 bxy k ½zk zk1
k¼1
n
xx ¼ DM
NM Q11 bxx þ Q12 byy þ Q16 bxy 1 ½z1 z0

þ Q11 bxx þ Q12 byy þ Q16 bxy 2 ½z2 z1
o
þ Q11 bxx þ Q12 byy þ Q16 bxy 3 ½z3 z2

6
xx ¼ ðþ0:5Þ ð½ð107:6 10 Þð1312 10 Þ
NM 9
þð26:06 109 Þð3638 106 Þ þ ð48:13 109 Þð4027 106 Þ

½ð0:0625 þ 0:1875Þ 103 Þ þ ð½ð170:9 109 Þð150 106 Þ
þð3:016 109 Þð4800 106 Þ þ ð0Þð0Þ½ð0:0625 þ 0:0625Þ
103 Þ þ ð½ð10:05 109 Þð4800 106 Þ þ ð3:016 109 Þ

ð150 106 Þ þ ð0Þð0Þ½ð0:1875 0:0625Þ 103 Þ
NTxx ¼ 8190 N=M
The remaining thermal stress and moment resultants are calculated in similar
fashion, eventually resulting in:
8 9
> NTxx > 8 9
>
> >
> > 8190 N=m >
>
> T >
> > >
>
> > > > 8460 N=m > >
> Nyy >
> > >
> >
>
>
>
>
>
> > > >
T >
< N = < 233 N=m >
> =
xy
¼
>
> T > > 0:05 N m=m >
> Mxx >
> > >
> >
>
>
>
>
>
> T >
> >
> >
>
>
> M >
> >
> 0:05 N m=m >
>
>
> yy >
> >
: >
;
>
: MT ; >
xy
0:03 N m=m
We can now calculate midplane strains and curvatures using Eq. (45), which
becomes:*
8 o 9 2 3
> exx > 3:757 108 1:964 109 1:038 108 1:440 104 3:905 106 8:153 105
>
> >
>
>
> > 6 7
>
>
> eoyy >
>
>
>
6 1:964 109 1:037 107 4:234 108 1:866 105 6:361 104 4:628 104 7
>
> > 6 7
< o > = 66 1:038 108
4:234 108 2:004 107 3:661 104 3:251 104
7
1:851 105 7
cxy
¼6
6
7
7
>
> jxx >> 6 1:440 10 4
1:866 105 3:661 104
7:064 3:122 102 4:572 7
>
> >
> 6 7
>
> >
> 6 7
>
> jyy >> 4 3:905 10 6
6:361 104 3:251 104
3:122 102
6:429 3:620 5
>
> >
>
: ;
jxy 8:513 105 4:628 104 1:851 105 4:572 3:620 17:41
8 9
> 4060 þ 8190 >
>
> >
> 7360 þ 8460 >
> >
>
>
> >
>
>
> >
< 2860 233 > =

>
> 0:62 þ 0:05 >
>
>
> >
>
>
> >
>
>
> 0:62 0:05 >
>
>
: >
;
0:36 þ 0:03
* The [abd] matrix for a [30/0/90]T graphite-epoxy laminate was calculated in Sample Problem
3, and the thermal stress and moment resultants were calculated in Sample Problem 5.

8 o 9 8 9
> exx > > 18 Am=m >
>
> > > >
> o >
> >
>
>
>
> 509 Am=m >
>
>
>
> e >
> >
> >
>
>
<
yy >
= >
< >
=
o
cxy ¼ 420 Arad
>
> jxx >> >
> 1:0 m1 >>
>
> >
> >
> >
>
>
> >
> >
> 1 >
>
>
> j yy >
> >
> 5:0 m >
>
: ; : ;
jxy 4:4 m1
ply 1 (i.e., strains present at z0=0.0001875 m) are:
8 9 8 o 9 8 9 8 9
> e > exx > > jxx > > 18 106 m=m >
< xx >
= >
< >
= >
< = >
> < >
=
eyy ¼ eoyy þ z0 jyy ¼ 509 106 m=m þ ð0:0001875 mÞ
>
: ;
> >
>
: o >
>
;
>
>
:
> >
; >
> :
>
>
;
c xy z¼z0 cxy jxy 420 106 m=m
8 9
> 1:0 rad=m >
>
< >
=
5:0 rad=m
>
> >
>
: ;
4:4 rad=m
8 9 8 9
> 206 Am=m >
< exx >
= >
< =
eyy ¼ 1450 Am=m
> ;
: cxy > >
: >
;
z¼z0 1240 Am=m
Strains calculated at the remaining ply interface positions are summa-

rized in Table 9.
Table 9 Ply Interface Strains in a [30/0/90]T Graphite-Epoxy Laminate

Caused by the Combined Effects of Cooldown from 175jC to 20jC
and an Increase in Moisture Content of +0.5%
0.1875 206 1450 1240

0.0625 80 820 690
0.0625 44 190 150
0.1875 170 440 400

(c) Ply stresses relative to the x–y coordinate system. Ply stresses may now be
calculated using Eq. (30) of Chap. 5. The stresses present at the outer surface
of ply 1 are (i.e., at z=z0):
8 9ply 1 2 3ply 1 8 9
> rxx > Q11 Q12 Q16 exx DTaxx DMbxx >
>
< =
>
6 7
>
< =

ryy ¼6 Q
4 12 Q Q 7
26 5
e yy DTa yy DMb yy
> > > ;
>
22
>
: ;
> :
sxy z ¼ z0 Q16 Q26 Q66 z ¼ z0 cxy DTaxy DMbxy z ¼ z0
2 3
8 9ply 1 107:6 109 26:06 109 48:13 109
< xx =
r
6 7
ryy ¼6
4 26:06 10
9
27:22 109 21:52 109 7
5
: ;
sxy z ¼ z0 48:13 109 21:52 109 36:05 109
8 9
> ½ð206Þ ð155Þð6:08Þ ð0:5Þð1312Þ 106 >
>
< >
=
½ð1450Þ ð155Þð20:0Þ ð0:5Þð3638Þ 106
>
> >
: ½ð1240Þ ð155Þð24:2Þ ð0:5Þð4027Þ 106 >
;
8 9ply 1 8 25 MPa 9
> > >
< rxx >
= >
< >
=

ryy ¼ 2:4 MPa
: sxy >
> ; >
> >
: 2:2 MPa >
;
z ¼ z0
Stresses calculated at the remaining plies and ply interface positions are
summarized in Table 10.
A comparison of the results obtained in Example Problems 6 and 7 leads
to the following observation: The initial ply stresses and strains caused by
cooldown from cure temperatures to room temperatures are partially relieved
Table 10 Ply Interface Stresses in a [30/0/90]T Graphite-Epoxy Laminate

Caused by the Combined Effects of Cooldown from 175jC to 20jC
and an Increase in Moisture Content of +0.5%
Ply 1 0.1875 25 2.2 2.2

0.0625 1.4 0.24 9.9
Ply 2 0.0625 20 9.3 9.0
0.0625 40 15 1.9
Ply 3 0.0625 16 65 1.9
0.1875 17 43 5.2
Stressed are referenced to the x–y coordinate system.

by the subsequent adsorption of moisture. Although the interaction between
temperature and moisture effects obviously depends on the details of the
situation (material properties involved, stacking sequence, magnitudes of DT
and DM, etc.), this observation is often true. That is, the thermal stresses
predicted to develop in a multiangle laminate during cooldown are usually
predicted to be relieved somewhat by subsequent adsorption of moisture.
7 SIMPLIFICATIONS DUE TO STACKING SEQUENCE

Eqs. (44) and (45) summarize the response of a multiangle composite laminate
due to the combined effects of uniform mechanical loading, uniform changes
in temperature, and/or uniform changes in moisture content. The primary
objective of this section is to show that these equations may be substantially
simplified through proper selection of the laminate stacking sequence. Before
these simplifications are discussed, however, it is illustrative to consider the
simplest case of all-specifically, let us consider Eqs. (44) and (45) when applied
to a plate of total thickness t made from an isotropic material.
Recall that for isotropic materials, all properties are independent of
direction. For present purposes, let:
E11 ¼ E22 ¼ E
m12 ¼ m21 ¼ m
G12 ¼ G
a11 ¼ a22 ¼ a
b11 ¼ b22 ¼ b
Also recall that for isotropic materials, only two of the elastic moduli are
independent. That is:
E
G¼
2ð1 þ mÞ
If these interrelations between material properties are enforced, then Eq. (44)
reduces to:
2 3
A11 A12 0 0 0 0 8 o 9 8 T9 8 M9
8 9 N > >N >
> Nxx >6 7>> exx >
> >
> > > >
>
> >
>6 A12 7>> >
> >
> >
>
> >
> >
>
>
> >
>6 A11 0 0 0 0 7>>
> eo >
>
> >
>
> T>
>
>
>
>
>
M>
>
>
> >
>6 7> > > N > > N >
>
>
Nyy >
>6 A11 A12 7>>
>
yy >
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
> >
>6 0 7 > > > > > >
< Nxy =6 0 0 0 0 7 < co = < 0 = < 0 >
> > > > > =
6 2 7 (46)
¼6 7
xy

>
> M >
> 6 0 0 0 D11 D12 0 7>> >
> >
> >
> >
> 0 >
>
>
> xx >
> 6 7> > >
> jxx > > 0 >
> >
> >
>
>
> >
> 6 7> > > > > >
>
> >
> 6 0 0 0 D12 D11 0 7>>
>j >
>
>
> >
>
>
>
>
>
>
>
>
>
>
>
>
>
>
Myy >
> 6 7> > > >
7> yy > > 0 > > >
> > > > > > > 0 >
: ; 64

5 >
> > >
> > >
> >
> >
>
Mxy D11 D12 : ; : ; : ;
0 0 0 0 0 jxy 0 0
2

where:
Et mEt
A11 ¼ A12 ¼ mA11 ¼
1 m2 1 m2
Et3 mEt3
D11 ¼ D22 ¼ D12 ¼ mD11 ¼
12ð1 m2 Þ 12ð1 m2 Þ
ðD11 D12 Þ Et3

D66 ¼ ¼
2 24ð1 þ mÞ

Eta Etb
NT ¼ DT NM ¼ DM
ð1 mÞ ð1 mÞ
The constant D11 is often called the flexural rigidity of an isotropic plate.
Taking the inverse of Eq. (46), we find:
8 o 9 2a 38 9
> exx > 11 a12 0 0 0 0 >
> Nxx þ NT þ NM > >
>
> >
> 6 7>> >
>
> > 7> M>
>
>
>
o >> 6 a12 a11 0 0 0 0 >
> N þ N T
þ N >
>
> eyy >
> > 6 7>> yy >
>
< o > = 6 2ða11 a12 Þ
7>< >
=
cxy ¼ 6
6
0 0 0 0 0 7
7 N xy
ð47Þ
> > 6 0 7> >
> jxx >
> > 6 0 0 d11 d12 0 7> Mxx >
>
> >
> 6 7>>
>
>
>
>
>
> j >
> 6 0 7>> >
>
>
> yy >
> 4 0 0 d12 d11 0 5>> Myy >
>
: ; >
: >
;
jxy 0 0 0 0 0 2ðd11 d12 Þ Mxy
where:
1 1 m m
a11 ¼ ¼ a12 ¼ ma11 ¼
A11 ð1 m2 Þ Et A11 ð1 m2 Þ Et
1 12 m 12m
d11 ¼ d22 ¼ ¼ d12 ¼ md11 ¼ ¼
D11 ð1 m2 Þ Et3 D11 ð1 m2 Þ Et3
24ð1 þ mÞ
d66 ¼ 2ðd11 d12 Þ ¼
Et3
Comparing Eqs. (44) and (45) with Eqs. (46) and (47), it is apparent that
multiangle composite laminates may exhibit unusual coupling effects, as
compared to the more familiar behavior of isotropic plates. For example,
referring to Eq. (45), it can be seen that application of a normal stress resultant
Nxx will (in general) induce a midplane shear strain cjxy and curvatures jxx,

jyy, and jxy, due to the presence of the a16, b11, b12, and b16 terms, respectively.
Physically, these means that a uniform in-plane uniaxial loading will cause in-
plane shear strains as well as out-of-plane curvatures in a composite plate (i.e.,
the plate will bend). These couplings do not exist for isotropic panels, as
indicated by Eq. (47).
Unusual couplings between thermal resultants and laminate strains and
curvatures also exist, and are immediately apparent in practice. As previously
discussed, most modern composite material systems are cured at elevated
temperatures and are subsequently cooled to room temperatures. Therefore,
thermal stress and moment resultants (NijT and MijT, respectively) develop
during cooldown. Equation (45) shows that the thermal stress resultants NijT
and MijT will cause curvatures jxx, jyy, and jxy to develop upon cooldown.
Physically, this means that (in general) a composite laminate that is flat at the
elevated cure temperature will bend and warp as it is cooled to room
temperature. Coupling effects due to moisture stress and moment resultants,
NijT and MijT, are analogous to those associated with thermal stress and
moment resultants. Thus, even if a composite laminate is cured and used at
the same temperature (so that NijT=MijT=0), the laminate may still bend or
warp if the surrounding humidity causes the moisture content of the laminate
to change with time.
Because coupling effects greatly complicate the design of composite
structures, it is of interest to determine whether these coupling effects can be
reduced or eliminated. It will be seen that it is indeed possible to reduce or
eliminate many of these coupling effects through proper selection of the
laminate stacking sequence. Common stacking sequences used to eliminate
coupling effects are described in separate sections below.
7.1 Symmetrical Laminates

A symmetrical laminate is one that possesses both geometrical and material
symmetry about the midplane. In a symmetrical laminate, plies located
symmetrically about the laminate midplane are of the same material, have
the same thickness, and have the same fiber angle. Several examples of
symmetrical stacking sequences have been previously shown in Fig. 12. For
a symmetrical n-ply laminate, the material and fiber angle used in ply 1 is
identical to that used in ply n, the material and fiber angle used in ply 2 is
identical to that used ply n-1, etc.
Use of a symmetrical stacking sequence results in three major simpli-
fications to Eqs. (44) and (45). Specifically, for a symmetrical laminate:
All coupling stiffnesses equal zero (Bij=0).
All thermal moment resultants equal zero (MijT=0).
All moisture moment resultants equal zero (MijM=0).

To demonstrate that coupling stiffnesses are zero for a symmetrical
laminate, consider the coupling stiffness B11. From Eq. (27b), B11 is given by:
1X n
B11 ¼ Q11 k z2k z2k1
2 k¼1
In expanded form, B11 is given by:
1 n
B11 ¼ Q11 1 z21 z20 þ Q11 2 z22 z21 þ Q11 3 z23 z22
2

þ : : : þ Q11 n2 z2n2 z2n3 þ Q11 n1 z2n1 z2n2 ð48Þ

þ Q11 n z2n z2n1 g
Because the laminate is assumed to be symmetrical, it must be that:

Q11 1 ¼ Q11 n

Q11 2 ¼ Q11 n1
ð49aÞ
Q11 3 ¼ Q11 n2

etc:
Also, due to symmetry, the ply interface positions are located symmetrically
about the midplane, and hence (recalling that z0<0 and zn>0):
z0 ¼ zn
z1 ¼ zn1
z2 ¼ zn2 ð49bÞ

etc:
Together, the relations listed as Eqs. (49a) and (49b) imply that for any
symmetrical laminate:

Q11 1 z21 z20 ¼ Q11 n z2n z2n1

Q11 2 z22 z21 ¼ Q11 n1 z2n1 z2n2 ð50Þ
2
Q11 3 z3 z22 ¼ Q11 n2 z2n2 z2n3

etc:

Substituting Eq. (50) into Eq. (48), it is seen that B11 = 0. Similar results may
be demonstrated for all other coupling stiffnesses, and hence Bij = 0 for any
symmetrical laminate, as stated.
To demonstrate that thermal moment resultants are zero for a symmet-
T
rical laminate, consider the thermal moment resultant Mxx . From Eq. (42d),
T
Mxx is given by:
DT Xn
MTxx ¼ Q11 axx þ Q12 ayy þ Q16 axy k z2k z2k1
2 k¼1
T
In expanded form, Mxx is given by:
(
DT
MTxx ¼ Q11 axx þ Q12 ayy þ Q16 axy 1 z21 z20
2

þ Q11 axx þ Q12 ayy þ Q16 axy 2 z22 z21

þ Q11 axx þ Q12 ayy þ Q16 axy 3 z23 z22
þ...... ð51Þ

þ Q11 axx þ Q12 ayy þ Q16 axy n2 z2n2 z2n3

þ Q11 axx þ Q12 ayy þ Q16 axy n1 z2n1 z2n2
)
2
þ Q11 axx þ Q12 ayy þ Q16 axy n zn z2n1
Because the laminate is assumed to be symmetrical, it must be that:

2
Q11 axx þ Q12 ayy þ Q16 axy z z20 ¼ Q11 axx þ Q12 ayy þ Q16 axy n z2n z2n1
1 1

Q11 axx þ Q12 ayy þ Q16 axy 2 z22 z21 ¼ Q11 axx þ Q12 ayy þ Q16 axy n1 z2n1 z2n2

Q11 axx þ Q12 ayy þ Q16 axy 3 z23 z22 ¼ Q11 axx þ Q12 ayy þ Q16 axy n2 z2n2 z2n3
(52)
T
Substituting Eq. (52) into Eq. (51), it is seen that Mxx =0 for a symmetrical
T T
laminate. Similar results may be demonstrated for Myy and Mxy , and
hence all thermal moment resultants equal zero for any symmetrical lami-
nate, as stated. Therefore, a symmetrical laminate will not bend or warp
when subjected to a uniform change in temperature. In particular, a sym-
metrical laminate will not warp or bend during cooldown from the cure
temperature.

An identical procedure may be used to prove that all moisture moment
resultants are zero for a symmetrical laminate.
In summary then, for symmetrical laminates, Eqs. (44) and (45) reduce
to:
8 9 2 38 eo 9 8 T 9 8 M 9
> Nxx > A11 A12 A16 0 0 0 > > xx >> >
> Nxx >> >
> Nxx >>
>
> >
> >
> >
> >
> >
> >
> >
>
> >
> 6 0 7>7> o >
> >
> T >
> >
> M> >
>
>
> N yy >
> 6 A12 A22 A26 0 0 >
> e >
yy > >
> N >
yy > >
> N yy >
>
>
< N = 6A> 6 7>
< >
= >
< >
= >
< >
=
A26 A66 0 0 0 7 c o NT
N M
¼6 7
xy 16
6 7
xy xy xy
>
> Mxx > > 6 0 0 0 D11 D12 D16 7> > jxx >> >
> 0 > > >
> 0 > >
>
> >
> 6 7>
> >
> >
> >
> >
> >
>
>
> Myy > > 4 0 >
5> >
> >
> >
> >
> >
>
>
> >
> 0 0 D 12 D 22 D 26 >
> j >
yy > >
> 0 >
> >
> 0 >
>
: ; >
: >
; >
: >
; >
: >
;
Mxy 0 0 0 D16 D26 D66 jxy 0 0
(53a)
8 o 9 2 a11 38 9
a12 a16 0 0 0 > > Nxx þ NTxx þ NM >
>
>
> exx >
> 6 >
>
xx >
>
> eo >
> > 0 0 7
0 7>> þ T
þ M >
>
>
> >
> 6 a12 a22 a26 >
> N yy N yy Nyy >>
>
< yy >
= 6 7>
< >
=
coxy ¼ 6 a a26 a66 0 0 0 7 þ þ
T M
6 16 7 N N N
6 0 7
xy xy xy ð53bÞ
>
> > d16 7> >
> jxx >
> > 6
> 6
0 0 d11 d12 >
>
7>
Mxx >
>
>
> jyy >
> > 4 0 >
d26 5> >
>
>
: >
; 0 0 d12 d22 >
> M >
>
jxy >
:
yy
>
;
0 0 0 d16 d26 d66 Mxy
Due to these dramatical simplifications, symmetrical laminates are
almost always used in practice. In those rare circumstances in which the use
of a nonsymmetrical laminate is required for some reason, it is best to place
the nonsymmetrical ply (or plies) at or near the laminate midplane, which
will minimize the coupling stiffnesses and thermal and moisture moment
resultants.
It is noted in passing that unidirectional composite laminates [h]n are
symmetrical and hence Bij=MijT=MijM=0 for unidirectional laminates. In
T
addition, for the case of [0j]n or [90j]n laminates, A16=A26=D16=D26=Nxy
=Nxy ¼ 0.
M
7.2 Cross-Ply Laminates

Composite laminates that contain plies with fiber angles of 0j or 90j only are
called ‘‘cross-ply’’ laminates. Inspection of Eq. (31) of Chap. 5 reveals that for
any 0j or 90j ply, Q16=Q26 ¼ 0. From (Eq. (27a) (27b) and (34) it is seen that
A16, A26, B16, B26, D16, and D26 all involve a summation of terms involving Q16
and Q26 , and hence all of these terms also equal zero for cross-ply laminates.
Furthermore, from Eq. (25) of Chap. 5, it is seen that axy=0 for any 0j or 90j
ply, and from Eq. (38) of Chap. 5 that bxy=0 for any 0j or 90j ply. From Eqs.
T
(42c), (42d), (42e), and (42f), it is seen that Nxy ¼ Mxy
T
¼ Nxy
M
¼ MxyM
¼ 0 for

cross-ply laminates (because Q16 ¼ Q26 ¼ axy ¼ bxy ¼ 0Þ . For cross-ply
laminates then, Eqs. (44) and (45) reduce to:
8 9 2 38 eo 9 8 T 9 8 M 9
> Nxx > A11 A12 0 B11 B12 0 > > xx >> > > Nxx >> >
> Nxx >>
>
> >
> 6 >
> o > > >
> >
> >
> >
>
>
> >
> 7>
> >
> >
> T >> >
> M >>
>
> N yy >
> 6 12
A A 22 0 B 12 B 22 0 7>
> e yy >
> >
> N yy >
> >
> N yy >
>
> >
<N = 6 0 6 7>
< >
= >
< >
= >
< >
=
0 A66 0 0 B66 7 co 0 0
xy
¼66 7 xy
>
>
>
Mxx >>
> 6 B11 B12 0 D11 D12 0 7 7>
> jxx >> >
> Mxx >
T
> > > Mxx >
M
>
>
> > 6
> 7>
>
>
>
> >
> > T >
> > >
> > M>
> >
>
>
> M >
yy > 4 B B 0 D D 0 5>
> jyy >> >
> Myy >> >
> Myy >>
>
: >
;
12 22 12 22 >
> >
> >
> >
> >
> >
>
Mxy 0 0 B66 0 0 D66 : jxy ; : 0 ; : 0 ;
ð54aÞ
and
8 o 9 2 38 N þ NT þ NM 9
>
> exx >
> a11 a12 0 b11 b12 0 >
> xx xx >>
> o >
> > 6 >
>
xx
>
M >
>
>
>
>
> a12 a22 0 b21 b22 0 7 >
> N þ N T
þ N >
>
> yy >
>
e > 6 7>
7>
yy yy yy >
>
< o > = 66 0 0 a66 0 0 b66 7
>
< Nxy
>
=
cxy
¼6
6b
7 ð54bÞ
>
>
> jxx >
>
> 6 11 b21 0 d11 d12 0 77>
> Mxx þ MTxx þ MM >
xx >
>
> >
> 6 7>
>
>
>
>
>
>
> > 4 b12 5> >
> jyy > b22 0 d12 d22 0 > þ þ >
T M
> >
> >
>
M yy M yy M yy >
>
: ; : ;
jxy 0 0 b66 0 0 d66 Mxy
If a cross-ply laminate is also symmetrical, then all remaining coupling
stiffnesses equal zero, as well as all remaining thermal and moisture moment
resultants. Hence, for symmetrical cross-ply laminates, Eqs. (54a) and (54b)
are further simplified to:
8 9 2 38 eo 9 8 T 9 8 M 9
> Nxx > A11 A12 0 0 0 0 > > xx >> >
> Nxx >> >
> Nxx >>
>
> >
> >
> >
> >
> >
> >
> >
>
>
> >
> 6 7
0 7>> o >
> >
> T >
> >
> M> >
>
> N yy >
> 6 A12 A22 0 0 0 >
> e >
yy > >
> N >
yy > >
> N yy >
>
>
< > 6 7> > > > > >
Nxy = 6 0 0 A66 0 0 0 7< co = < 0 = < 0 =
¼6 6 0
7 xy
>
>
>
Mxx > >
> 6 0 0 D11 D12 0 7 > jxx >
7> > >
> 0 > > >
> 0 > >
>
> >
> 6 7>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
> Myy > > 4 0 0 0 D D 0 5>
> j >
> >
> 0 >
> >
> 0 >
>
>
: >
;
12 22 >
> yy >
> : >
> >
> : >
> >
>
Mxy 0 0 0 0 0 D66 : ; ; ;
jxy 0 0
ð55aÞ
(and)
8 o 9 2 38 9
> e > a11 a12 0 0 0 0 > > Nxx þ NTxx þ NM >
>
>
> xx >
> >
>
xx >
>
>
> eo >> 6 0 0 0 7
0 7>> þ þ >
>
6 a12 a22 T M
>
> >
> >
>
Nyy N yy N yy >
>
< yy = 6 0 7< =
cxy ¼ 6
o
6
0 a66 0 0 0 7
7
Nxy
ð55bÞ
>
> > 6 0
xx > 0 0 d11 d12 0 7> >
>
>
j >
> 6 7>
>
>
Mxx >
>
>
>
> jyy >> 4 0 0 0 d12 d22 0 5>> >
>
>
: >
; >
>
M yy >
>
jxy : ;
0 0 0 0 0 d66 Mxy

Note from Eqs. (55a) and (55b) that symmetrical cross-ply laminates do not
exhibit any coupling stiffnesses. That is:
A16 ¼ A26 ¼ D16 ¼ D26 ¼ Bij ¼ 0
or, equivalently,
a16 ¼ a26 ¼ d16 ¼ d26 ¼ bij ¼ 0
Laminates that do not possess these coupling stiffnesses are called specially
orthotropic laminates. The fact that symmetrical cross-ply laminates are
specially orthotropic will become an important factor in Chaps. 9 and 10,
where the mechanical response of such laminates subjected to varying loads
will be considered.
7.3 Balanced Laminates

A laminate is ‘‘balanced’’ if, for every ply with fiber angle h, there exists a
second ply whose fiber angle is h. The two plies must be otherwise identical,
(i.e. they must be composed of the same material and have the same thickness).
Inspection of Eq. (31) of Chap. 5 reveals that the Q16 and Q26 terms for these
balanced plies will always be of equal magnitude but opposite algebraic sign
(i.e., Q16 jh ¼ Q16 jh and Q26 jh ¼ Q26 jh Þ. Consequently, from Eq. (27a),
A16=A26=0 for a balanced laminate. Further, from Eqs. (25) and (28) of
Chap. 5, it is seen that axy and bxy for the two balanced plies will always be of
equal magnitude but opposite algebraic sign (i.e., axy jh ¼ axy jh and bxy jh
¼ bxy jh Þ . Consequently, from Eqs. (42c) and (43), Nxy T
=NM xy=0 for a
balanced laminate. Equations (44) and (45) are therefore simplified to:
8 9
8 9 2 3 eoxx > 8 NT 9 8 NM 9
> Nxx > A11 A12 0 B11 B12 B16 > >
> > >
> > xx >> > xx >
>
> > > o > > > > > >
>
> Nyy >
> > 6 A12 A22
> 0 B B B 7>
>
> e >
>
>
>
>
> N T >>
>
>
>
> N M >>
yy >
>
> > 6 12 22 26 7> yy > > yy > > >
<N > = 6 0 7>
< >
= >
< >
= >
< 0 =
>
xy 6 0 A66 B16 B26 B66 7 cxy o 0
¼6 7
>
> Mxx > > 6 B11 B12 B16 D11 D12 D16 7> jxx > > MTxx > > MMxx >
>
> >
> 6 7>
>
>
>
>
>
>
>
>
>
>
>
>
>
> M>
>
>
>
> M >
yy > 4 B12 B22 B26 D12 D22 D26 > 5> >
> >
> M T >
> >
> M >
>
>
: >
; >
> j >
yy > >
> yy >
> >
> yy >
>
Mxy B16 B26 B66 D16 D26 D66 : j ; > > : T ; : M ;
xy
M xy M xy
ð56aÞ
and
8 9
8 o 9 2 a11 a12 0 b11 b12
3 Nxx þ NT þ NM >
b16 >> >
> exx > >
>
xx xx >
>
>
> > 6 7> M >
> eo >
> >
> a12 a22 0 b21 b22 >
> yy
b26 7> N þ N T
þ N yy >
>
>
>
> > 6 7>
yy
>
< o >
yy
= 6
6 b66 7
>
< Nxy >
=
cxy ¼ 6 0 0 a66 b61 b62
7 ð56bÞ
> 6b d16 7
> jxx > > Mxx þ Mxx þ Mxx >
7>
T M
> >
> 6 11 b12 b61 d11 d12 >
>
> > 6 7>
> >
>
>
> jyy >
>
> 4 b12 > Myy þ MT þ MM >
d26 5>
> >
: ; b22 b62 d12 d22 > yy >
>
jxy >
>
yy
>
>
b16 b26 b66 d16 d26 d66 : M þM þM ;
T M
xy xy xy

If a balanced laminate is also symmetrical, then Eqs. (56a) and (56b) are
simplified to:
8 9 2 3 8 T 9 8 M9
>
> Nxx >> A11 A12 0 0 0 0 8 eo 9 > > Nxx >
> >
> Nxx >>
> > 6A > > > > > >
> > 0 7 > > > > > >
xx
>
> Nyy >> 6 12 A22 0 0 0 7>> o >> >
> >
Nyy >
T >
> NM >
>
>
> >
> 6 7>> e >
> >
> >
> >
> yy >
>
< Nxy >
> = 6 0 0 A66 0 0 0 7 > yy > >
< o = < 0 = < 0 >> > =
6 7 cxy
¼6 7
>
> M >
> 6 0 0 0 D11 D12 D16 7> jxx > > 0 > > 0 >
7> > > > > >
xx
>
> >
> 6 >
> >
> >
> >
> >
> >
>
>
> yy >
M >
> 6 D26 7 >
> j >
yy > >
> >
> >
> >
>
>
> > 4 0 0 0 D12 D22 5:> >
; >
>
0 >
> >
>
0 >
>
>
: Mxy ;> j xy >
: 0 ; : 0 >
> > ;
0 0 0 D16 D26 D66
ð57aÞ
and
8 o 9 2 a11 38 9
a12 0 0 0 0 > Nxx þ NTxx þ NM
xx >
>
> exx >> >
> >
>
>
> o >
> 6 a12 7> >
>
> e > 6
yy >
a22 0 0 0 0 7>
>
> Nyy þ NTyy þ NM >
yy >
>
>
< o > = 6 0 7> >
cxy ¼ 6 0 a66 0 0 0 7< N
=
6 7 xy
ð57bÞ
> > 6 0 7
d16 7> >
>
> j xx >
> 6 0 0 d11 d12 >
> Mxx >
>
>
> >
> 6 7>
> >
>
>
> jyy >> 4 0 0 0 d12 d22 d26 5>
> M >
>
: ; >
:
yy >
;
jxy 0 0 0 d16 d26 d66 Mxy
7.4 Balanced Angle-Ply Laminates

The definition of a ‘‘balanced’’ laminate was given in Sec. 7.4. A balanced
angle-ply laminate is really just a special class of balanced laminates; it is
discussed in a separate subsection because of an additional simplification that
occurs.
All plies in an angle-ply laminate have a fiber angle of the same
magnitude. That is, all plies within an angle-ply laminate have a fiber angle
of either+h or h, where the value h is the same for all plies. In general, for an
angle-ply laminate, the number of plies with fiber angle+h may differ from
the number of plies with fiber angle h. However, a balanced angle-ply
laminate must have an equal number of plies with fiber angle+h and h, so as
to satisfy the preceding definition of a balanced laminate. Carefully note the
distinction between a balanced laminate and balanced angle-ply laminate. A
balanced laminate may involve more than one ‘‘distinct’’ fiber angle. For
example, a [35/65/-35/-65]T laminate is balanced (although not symmetrical),
and A16=A26=0 for this laminate. In contrast, a balanced angle-ply laminate
may involve only one ‘‘distinct’’ angle. A stacking sequence of either [35/-35/

35/-35]T or [65/65/-65/-65]T result in balanced angle-ply laminates, for
example.
The following simplifications occur for a balanced angle-ply laminate:
B11 ¼ B22 ¼ B66 ¼ MTxx ¼ MTyy ¼ MTxx ¼ MTyy ¼ 0
In addition, the simplifications that exist for any balanced laminate

T M
(A16=A26=Nxy =Nxy =0) also occur for a balanced angle-ply laminate.
Equations (44) and (45) are therefore simplified to:
8 9 2 3 8 T 9 8 M9
> Nxx > A11 A12 0 0 0 B16 8 eo 9 > > Nxx >> >
> Nxx >>
>
> >
> >
> xx >> >
> >
> >
> >
>
> > 6 7 > > > > > M >>
>
> Nyy >
>
> 6
A12 A22 0 0 0 B26 7>> o > > >
> N T >
> >
> N >
>
7> eyy > > > > >
yy yy
<N >
> = 6 >
< >
= >
< >
= >
< >
=
6 0 0 A66 B16 B26 0 7 o 0 0
¼6 7 cxy
xy
6 0 7
>
> xx >
M >
> 6 0 B16 D11 D12 D16 7>> jxx >> >
> 0 > > >
> 0 > >
>
> > 6 7>> >
> >
> >
> >
> >
>
> Myy >
> > 4 0 D26 5>>
> jyy >> >
> > >
> >
> >
>
>
> >
> 0 B26 D12 D22 : >
; > 0 >
> >
> 0 >
>
: ; j xy
>
: >
; >
: >
;
Mxy B16 B26 0 D16 D26 D66 T
Mxy M
Mxy
ð58aÞ
and
2 38 9
8 o 9 a11 a12 0 0 0 b16 > Nxx þ NTxx þ NM
xx >
> e > > >
>
>
xx >
> 6 7>> >
>
> eo >
> 6a a22 0 0 0 b26 7>> Nyy þ NTyy þ NM >
>
>
> yy >
> 6 12 7>> yy >
>
< o > = 6 0 0 a66 b16 b62 7
b66 7
>
< Nxy
>
=
cxy ¼ 66 7 ð58bÞ
>
> > 6 0 d16 7 > >
> xx > 7> >
j 0 b61 d11 d12 Mxx
> > 6
> > >
> jyy >
> > 6 7>> >
>
>
: >
; 4 0 0 b62 d12 d22 d26 5>>
> M yy >
>
>
jxy >
: >
;
b 16 b26 0 d16 d26 d66 Mxy þ Mxy þ Mxy
T M
If a balanced angle-ply laminate is also symmetrical, then Eqs. (58a) and (58b)
are simplified still further to:
8 9 2 3 8 9 8 9
Nxx > A11 A12 0 0 0 0 8 o 9 > NTxx > > NM xx >
>
> > e
> xx > > > > >
>
> >
> 6 7> > > > > >
T > > M>
>
>
> Nyy >
> 6 A12 A22 0 0 0 0 7>>
> o >>
>
>
>
> N >
eyy > > yy > > yy >
>
>
>
>
> N >
>
>
> >
> 6 7>
> > > > > >
>
< Nxy = 6 7< = < = < =
6 0 0 A66 0 0 0 7 co 0 0
¼6 7 xy
>
> Mxx >
> 6 0 D16 7> jxx > > 0 > > 0 >
>
> >
> 6 0 0 D11 D12 7>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
> >
> 6 D26 7>
> >
> >
> >
> >
> >
> yy >
> M > 4 0 0 0 D12 D22 >
5: j yy > >
; > 0 > > > > 0 > >
>
: ; jxy >
: >
; : > >
;
Mxy 0 0 0 D16 D26 D66 0 0
ð59aÞ

and
8 o 9 2a 38 9
> e > 11 a12 0 0 0 0 >
> Nxx þ NTxx þ NM xx >
>
>
> xx > > 6 7>> >
>
>
> > 0 7> M >
eoyy > 6 a12 a22 0 0 0 > þ þ >
T
>
> >
> 7> N N N yy >
> 6 > >
yy
>
> > yy
>
< o > = 66 0 0 a66 0 0 0 7
7>< Nxy
>
=
cxy
¼6
6 0
7 ð59bÞ
>
>
> jxx >>
> 6 0 0 d11 d12 d16 7
7>> Mxx >
>
>
> >
> 6 7>>
>
>
>
>
>
> jyy >>
> 6 7 >
> >
>
>
> > 4 0 0 0 d12 d22 d26 5>> M yy >
>
: ; >
: >
;
jxy 0 0 0 d16 d26 d66 Mxy
7.5 Quasi-Isotropic Laminates

A quasi-isotropic laminate is one that satisfies the following conditions:
Three or more distinct fiber angles must be present within a laminate.
The number of distinct fiber angles will be denoted m, and hence if a
laminate is quasi-isotropic, then m z 3.
The m distinct fiber angles must appear at equal increments of (180/
m) degrees.
An equal number of plies must be present at each of the m distinct
fiber angles.
It can be shown (see Ref. 2) that if the three conditions are met, then
members of the Aij matrix for the laminate are related as follows:
1
A11 ¼ A22 A66 ¼ ðA11 A12 Þ A16 ¼ A26 ¼ 0
2
Now, these same relations between extensional stiffness also hold for an
isotropic plate (see Eq. (46)). Hence, laminates that satisfy the above
conditions are called quasi-isotropic laminates. Also, for a quasi-isotropic
T
laminate, Nxy ¼ Mxy T
¼ Nxy
M
¼ MxyM
¼ 0 . In this case, Eqs. (44) and (45)
reduce to:
8 9 2 38 o 9 8 T 9 8 M 9
>
> N >
> A11 A12 0 B11 B12 B16 > > e > > Nxx > > N >
> xx
> >
> > xx > > >
> > > xx >
>
> >
> 6 7>>
> eo > >
>
>
>
>
>
>
>
>
>
>
>
>
> Nyy >
> 6 A12 A22 0 B12 B22 B26 7>> >
> >
> N T >
> >
> M >>
>
>
> >
> 6 7>> yy >
> >
> yy >
> >
> N yy >
>
>
> >
> 6 7>> o > > >
> >
> >
> >
>
< Nxy = 6 A11 A12 7< = < = < 0 =
6 B66 7 cxy 0
¼6 0 0
2
B16 B26
7
>
> >
> 6 7> j > > MTxx > > >
>
>
M xx >
> 6 B11 B12 B16 D11 D12 D16 7>> xx > > >
> >
> >
> MMxx >
>
>
> >
> 6 7>>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
> >
Myy > 6
> 7 >
> > >
> > MT > > > > MM > >
>
>
> > 4 B12 B22 B26 D12 D22 D26 5>> j yy >
> >
> yy >
> >
> yy >
>
>
> >
> >
> >
> >
: >
; >
: >
;
: ; B16 B26 B66 D16 D26 D66 : ; 0 0
Mxy jxy
ð60aÞ

2 38 9
8 o 9 a11 a12 0 b11 b12 b16 > > Nxx þ NTxx þ NM >
>
> e xx > >
>
xx >
>
>
> >
> 6a 7 >
> þ T
þ M >
>
>
> eoyy >
> 6 12 a22 0 b21 b22 b26 7>> N yy N yy N yy >
>
>
> > 6 7> >
< o > = 6 0 0 2ða11 a12 Þ b61 b62 b66 7 >
< N >
=
cxy 6 7 xy
¼6 7
>
> jxx >> 6 b11 b21 b61 d11 d12 d16 7> Mxx þ MTxx þ MM xx >
>
> >
> 6 7>>
>
>
>
>
>
> j >
yy > 6b b22 b62 d12 d22 7
d26 5>> þ T
þ M >>
> >
: j ; 4 12 >
> M yy M M >
>
>
>
yy yy
>
>
xy
b16 b26 b66 d16 d26 d66 : Mxy ;
ð60bÞ
The simplest possible quasi-isotropic laminate contains three plies, oriented

at equal increments of (180j/3)=60j. For example, [0/60/60]T or [60/0/
60]T laminates are quasi-isotropic. Probably the most common quasi-
isotropic laminate involves four distinct fiber angles (m=4). These laminates
must have ply angles oriented at increments of (180j/4)=45j. Typical
stacking sequences in this case are [0/45/90/45]T or [45/0/45/90]T. Al-
though the extensional stiffnesses for these laminates are ‘‘quasi-isotropic’’,
they still exhibit coupling stiffnesses (i.e., Bij p 0); furthermore, the bending
stiffnesses are not isotropic (e.g., D11 p D22).
If a quasi-isotropic laminate is also symmetrical (e.g., [0/45/90/45]s),
then the coupling stiffnesses Bij=0, and all remaining thermal moment
resultants and moisture moment resultants equal zero. Hence, Eqs. (60a)
and (60b) are simplified still further:
8 9 2 38 o 9 8 T 9 8 M 9
> Nxx > A11 A12 0 0 0 0 > exx > > Nxx > > Nxx >
>
> >
> >
> >
> > > > >
>
> >
>6A 0 7 >
> o >
> >
>
>
>
> >
T >
>
> M>
>
>
>
> Nyy >
>6 12 A11 0 0 0 7> > > > > Nyy >
>
> >
>6 7>
> eyy >> >
> Nyy >
> >
> >
>
>
< Nxy >
=6 A11 A12 7>
< o >
> > >
< 0 >
= >
> >
= > < 0 >
>
=
6 0 0 0 0 7
0 7 cxy
¼6
6 2 7>
>
> Mxx >
> 6 0 > jxx >> >
> 0 >> >
> 0 >>
>
> >
> 6 0 0 D11 D12 D16 7
7>
> >
> >
> >
> >
> >
>
>
> >
> 6 7>
> >
> >
> >
> >
> >
>
>
> Myy >
> 4 0 5>
> j >
> >
> 0 >
> >
> 0 >
>
>
> >
> 0 0 D12 D22 D26 >> yy >
> > > > >
: ; : ; >
: >
; > : >
;
Mxy 0 0 0 D16 D26 D66 jxy 0 0
ð61aÞ
8 o 9 2 38 9
>
> exx >> a11 a12 0 0 0 0 > Nxx þ NTxx þ NM xx >
>
> > > >
>
> o >>
> 6a 0 7
>
> M >
>
> yy >
> e > 6 12 a11 0 0 0 7>> N þ NT
þ N >
yy >
>
> > 6 7>> yy yy >
>
< co > = 6 0 0 2ða11 a12 Þ 0 0 0 7 >
< Nxy
>
=
xy 6 7
¼6 7 ð61bÞ
>
> j >
xx > 6 0 0 0 d11 d12 d16 7> Mxx >
>
> >
> 6 7>>
>
>
>
>
>
> > 6 d26 7 > >
> jyy >
> > 4 0
>
0 0 d12 d22 5>>
>
Myy >
>
>
>
> > > >
: jxy >; 0 0 0 d16 d26 d66 : Mxy ;

Again, note that only the in-plane extensional stiffnesses are ‘‘isotropic’’; the
bending stiffnesses for a symmetrical quasi-isotropic laminate are not related
in the same manner as in an isotropic plate.
It is noted in passing that symmetrical quasi-isotropic laminates are not
specially orthotropic, in contrast with symmetrical cross-ply laminates. That
is, for symmetrical quasi-isotropic laminates, D16,D26,d16,d26 p 0.
8 SUMMARY OF CLT CALCULATIONS

At this point, it is helpful to summarize the calculations steps involved in a
composites analysis based on classical lamination theory. Two different
analysis requirements are commonly encountered. In the first case, the loads
applied to the composite laminate are known, and the objective of the analysis
is to determine the ply stresses and strains that will be induced by these loads.
This situation is usually encountered during the design process. For example,
suppose the wing of a new airplane is being designed, and the skin is to be
fabricated using a composite laminate. Typically, the loads that must be
supported by the skin, temperatures extremes that will be encountered in
service, and the moisture content that may develop over long times are known
(having been calculated on the basis of separate analyses), and are supplied as
input data to the design engineer. The objective is to design the laminate such
that the ply stresses and strains induced by the given mechanical and
environmental loads will remain within safe limits.
A second case is when the strains induced in a laminate are known, and
the objective of the analysis is to determine the loads that caused these strains.
This situation is commonly encountered during stress analysis of an existing
structure or prototype. For example, suppose a prototype of the wing
described in the preceding paragraph has been built, and a stress analyst is
assigned to evaluate its performance. During the evaluation process, the
entire wing structure is subjected to various aerodynamic maneuvers and/or
other loading conditions expected to be encountered during the service life of
the aircraft. The strains, temperatures, and changes in moisture content are all
measured. The objective is now to deduce the loads that caused these
measured strains, as well as to determine the strains and stresses induced
within individual plies.
A slightly different calculation path is used in these two different
scenarios. The calculation procedures will be summarized in Secs. 8.1 and 8.2.
8.1 A CLT Analysis When Loads Are Known

The calculation steps followed when the applied loads are known are
summarized below:

1. Define the problem:
(a) Specify the number of different ply materials to be used in the
laminate.
(b) Specify the elastic properties for each ply material used (i.e.,
specify E11, E22, v12, G12, a11, a22, b11, and b22 for each ply
material).
(c) Specify the laminate description (i.e., specify the number of plies
n and the ply material and fiber angle for each ply).
(d) Specify the mechanical and environmental loads applied to the
laminate (specify Nxx, Nyy, Nxy, Mxx, Myy, Mxy, DT, and DM).
2. Calculate the ABD matrix:

(a) Calculate the reduced stiffness matrix Qij for each material used
in the laminate (using Eq. (11) of Chap. 5).
(b) Calculate the transformed reduced stiffness matrix Qij for each
ply based on the appropriate reduced stiffness matrix and ply
fiber angle (using Eq. (31) of Chap. 5).
(c) Calculate the Aij, Bij, and Dij matrices, using (Eq. (27a), (27b),
and (34), respectively.
(d) Assemble the ABD matrix.
3. Calculate the inverse of the ABD matrix: abd=ABD1.

4. Calculate the thermal and moisture stress and moment resultants:
(a) Calculate the effective thermal and moisture expansion coef-
ficients for each ply, using Eqs. (25) and (28) of Chap. 5,
respectively.
(b) Calculate the thermal and moisture stress and moment resultants
using Eqs. (42a) (42b) (42c) (42d) (42e) (42f) (43), respectively.
5. Calculate the midplane strains and curvatures induced in the
laminate, using Eq. (45).
6. For each ply:
(a) Calculate the ply strains in the x–y coordinate system, using Eq.
(12). Strains may be calculated at any desired position z, but most
often they are calculated at the ply interface positions (i.e., at
positions z0,z1,z2, z3,. . ., zn). Strains are also often transformed to
the local 1–2 coordinate system, defined by the fiber angle within
each ply.
(b) Calculate the ply stresses in the x–y coordinate system, using Eq.
(30) of Chap. 5. Once again, whereas stresses may be calculated
at any desired position z, most often they are calculated at the ply
interface positions (i.e., at positions z0, z1,z2,z3, . . ., zn). Stresses

are also often transformed to the local 1–2 coordinate system,
defined by the fiber angle within each ply.
8.2 A CLT Analysis When Midplane Strains and Curvatures are

Known
The calculation steps followed when midplane strains and curvatures are
known are summarized below:
1. Define the problem:
(a) Specify the number of different ply materials to be used in the
laminate.
(b) Specify the elastic properties for each ply material used (i.e.,
specify E11, E22, v12, G12, a11, a22, b11, and b22 for each ply
material).
(c) Specify the laminate description (i.e., specify the number of plies n
and the ply material and fiber angle for each ply).
(d) Specify the midplane strains and curvatures and environmental
loads experienced by the laminate (specify ejxx, ejyy, cjxy, jxx, jyy,
jxy, DT, and DM).
2. Calculate the ABD matrix:
(a) Calculate the reduced stiffness matrix Qij for each material used
in the laminate (using Eq. (11) of Chap. 5).
(b) Calculate the transformed reduced stiffness matrix Q ij for each
ply based on the appropriate reduced stiffness matrix and ply
fiber angle (using Eq. (31) of Chap. 5).
(c) Calculate the Aij, Bij, and Dij matrices, using (Eq. (27a), (27b), and
(34), respectively.
(d) Assemble the ABD matrix.
3. Calculate the thermal and moisture stress and moment resultants:
(a) Calculate the effective thermal and moisture expansion coef-
ficients for each ply, using Eqs. (25) and (28) of Chap. 5,
respectively.
(b) Calculate the thermal and moisture stress and moment resultants
using Eqs. (42a–42f) (43), respectively.
4. Calculate the applied stress and moment resultants, using Eq. (44).
5. For each ply:
(a) Calculate the ply strains in the x–y coordinate system, using Eq.
(12). Strains may be calculated at any desired position z, but most
often they are calculated at the ply interface positions (i.e., at

positions z0, z1, z2,z3, . . . ,zn). Strains are also often transformed
to the local 1–2 coordinate system, defined by the fiber angle
within each ply.
(b) Calculate the ply stresses in the x–y coordinate system, using Eq.
(30) of Chap. 5. Once again, whereas stresses may be calculated at
any desired position z, most often they are calculated at the ply
interface positions (i.e., at positions z0,z1,z2,z3, . . . ,zn). Stresses
are also often transformed to the local 1–2 coordinate system,
defined by the fiber angle within each ply.
9 EFFECTIVE PROPERTIES OF A COMPOSITE LAMINATE

The definitions of common engineering material properties were reviewed in
Chap. 3. In this section, these concepts will be used to define the ‘‘effective’’
properties of multiangle composite laminates. Before we begin this discus-
sion, it is pertinent to note that most of the properties defined in Chap. 3 do
not take into account the peculiar coupling effects exhibited by general
composite laminates. As a typical example, consider the definition of the
thermal expansion coefficient. As discussed in Sec. 3.3, the thermal expansion
coefficient is defined as the ratio of a thermal strain divided by the temperature
change that caused that strain. Six different thermal expansion coefficients
may be defined for an anisotropic material (see Eq. (21) of Chap. 3). For
example, if a uniform temperature change (DT ) causes a normal strain in the
T
x-direction (exx ), then the thermal expansion coefficient associated with the x-
direction is:
eTxx
axx ¼
DT
However, this definition does not anticipate the out-of-plane coupling effects
exhibited by general composite laminates. That is, for a general composite
laminate, a temperature change DT causes both midplane strain ejxx and
midplane curvature jxx, and consequently eTxx varies through the laminate
thickness. In this case then, what strain should be used to calculate axx?
A reasonable approach is to define axx based on the thermal strain
induced at the laminate midplane, axx=ejxx /DT, regardless of whether or not
a curvature is induced as well. A second thermal property can then be defined
as the ratio of midplane curvature to temperature change (jxx/DT ).
Although several new properties representing unusual coupling effects
can be defined in this manner, this topic will not be pursued here. Recall that
coupling effects occur in composite laminates with arbitrary stacking sequen-
ces because Bij p 0. Symmetrical laminates are almost always used in practice

so as to insure that Bij = 0. Consequently, these couplings are rarely
encountered in practice. The complications due to out-of-plane coupling
effects present in general laminates will be ignored in this section. The effective
properties for symmetrical composite laminates are discussed in the following
subsections.
9.1 Effective Properties Relating Stress to Strain

The standard definitions of material properties used to relate stress to strain
were discussed in Sec. 3.2. Three types of properties are measured during
uniaxial tests: Young’s modulus, Poisson’s ratios, and coefficients of mutual
influence of the second kind. Although these properties are usually measured
using uniaxial tests, they can also be measured based on strains induced by
pure bending. Properties measured for isotropic materials during uniaxial tests
are identical to those measured during pure bending tests. For example, for an
isotropic material, the value of Young’s modulus measured during a uniaxial
test is identical to that measured in pure bending. However, for composite
materials, the properties measured during uniaxial tests differ substantially
from those measured during pure bending tests. For example, Young’s mod-
ulus of a composite as measured during a uniaxial test is substantially different
than that measured in pure bending. Therefore, we must distinguish between
the two. In the following discussion, those properties measured through
application of in-plane loads are called ‘‘extensional’’ properties. In contrast,
properties during a pure bending test are called ‘‘flexural’’ properties.
Extensional Properties
Consider a symmetrical composite laminate subjected to uniaxial loading
Nxx, as shown in Fig. 18. The in-plane strains induced in this laminate can be
determined using classical lamination theory, in accordance with Eq. (45).
Assuming DT=DM=0 (and hence that thermal and moisture stress and
moment resultants are zero), and also noting that by definition Nyy=Nxy=
Myy=Myy=Mxy=0, Eq. (45) becomes (for symmetrical laminates):
8 o 9 2 38 9
> exx > a11 a12 a12 0 0 0 > Nxx >
>
> > > >
>
>
> eoyy >
>
>
>
6 a12 a22 a26 0
6
0 0 7>
7
>
>
> 0 >
>
>
>
< o = 6a 0 7 < =
cxy ¼ 6 16 a26 a66 0 0
7 0
> jxx > 6 0 0 d11 d12 d16 7
>
> >
> 6 0 7>>
>
0 > >
>
>
> >
> 4 0 0 0 d d d 5 >
>
26 > 0 >
>
>
: j yy >
; 12 22
: >
;
jxy 0 0 0 d16 d26 d66 0
The midplane strains caused by uniaxial loading are therefore given by:
eoxx ¼ a11 Nxx ð62aÞ

Figure 18 A symmetrical composite laminate subjected to uniaxial load Nxx.
eoyy ¼ a12 Nxx ð62bÞ

coxy ¼ a16 Nxx ð62cÞ
Recalling that Nxx is defined as a constant load/unit plate length (with units of
N/m or lbf/in.), the effective (or nominal) normal stress rxx applied to the
laminate is given by rxx =Nxx/t, where t is the total laminate thickness.
In Sec. 3.2, Young’s modulus was defined as ‘‘the normal stress rxx
divided by the resulting normal strain exx, with all other stress components
equal to zero.’’
Applying the standard definition to the laminate shown in Fig. 18, the
effective extensional Young’s modulus in the x-direction is given by:
ex rxx ðNxx =tÞ 1
Exx ¼ ¼ ¼ ð63Þ
o
exx ða11 Nxx Þ ta11
The superscript ‘‘ex’’ is used to denote that this property is Young’s modulus
measured in extension.
In Sec. 3.2, Poisson’s ratio vxy was defined as ‘‘the negative of the
transverse normal strain eyy divided by the axial normal strain exx, both of
which are induced by stress rxx, with all other stresses equal to zero.’’ The
effective Poisson’s ratio in extension for the laminate shown in Fig. 18 is given
by:
eoyy a12 Nxx a12
xy ¼
m ex ¼ ¼ ð64Þ
eoxx a11 Nxx a11

Once again, a superscript ‘‘ex’’ has been used to indicate that this Poisson’s
ratio is measured in extension.
The coefficient of mutual influence of the second kind gxx,xy was
defined as ‘‘the shear strain cxy divided by the axial normal strain exx, both
of which are induced by normal stress rxx, when all other stresses equal
zero.’’ For a composite laminate, the effective coefficient of mutual influence
ex
of the second kind gxx;xy is therefore given by:
coxy a16 Nxx a16
ex
gxx;xy ¼ ¼ ¼ ð65Þ
eoxx a11 Nxx a11
An identical procedure can be employed to define properties measured
during a uniaxial test in which only ryy =Nyy/t is applied. In this case, Eq.
(45) becomes:
8 o 9 2 38 9
> exx > a11 a12 a12 0 0 0 > 0 >
>
> > > >
> o >
> >
> 6 a12 7>>N > >
>
> eyy >
> 6
a22 a26 0 0 0 7>> yy >
>
< o >
> = 6a 7>>
<
>
>
cxy ¼ 6 a26 a66 0 0 0 7 0 =
6 16 7
> 6 0 d16 7
>
> jxx >
>
> 6 0 0 d11 d12 7>> 0 > >
>
> >
> 6 7>>
>
>
>
>
>
> > 4 0 5 > >
> jyy >
: >
;
0 0 d12 d22 d26 > >
:
0 >
>
;
jxy 0 0 0 d16 d26 d66 0
Midplane strains induced are therefore:
eoxx ¼ a12 Nyy ð66aÞ
eoyy ¼ a22 Nyy ð66bÞ
coxy ¼ a26 Nyy ð66cÞ

These strains can be used to define the effective extensional Young’s mod-
ex ex
ulus, Eyy , Poisson’s ratio, myx , and coefficient of mutual influence of the sec-
ex
ond kind, g yy,xy :
ex 1
Eyy ¼ ð67aÞ
ta22
a12
ex
myx ¼ ð67bÞ
a22
a26
ex
gyy;xy ¼ ð67cÞ
a22
Next, consider the effective material properties measured during a pure
shear test. A symmetrical composite laminate subjected to pure shear loading

Nxy is shown in Fig. 19. Because the load Nxy is applied within the plane of
the laminate, properties measured during a pure shear test are included in
this discussion of ‘‘extensional’’ properties, even though there is no common
counterpart measured in a test involving out-of-plane loading. Assuming
DT=DM=0, Eq. (45) becomes:
8 o 9 2 38 9
> e > a11 a12 a12 0 0 0 > 0 >
>
> xx >
> >
> >
>
>
> > 6 7> >
>
> eoyy >
>
> 6
a12 a22 a26 0 0 0 7>> 0 >
>
< o >
> = 6 7>> >
>
cxy ¼ 6 a a26 a66 0 0 0 7 Nxy =
<
6 16 7
> > 6 0 d16 7
>
> jxx >> 6 0 0 d11 d12 7>> 0 >
>
>
> >
> 6 7>> >
>
>
> yy >
> j >
>
4 0 0 0 d12 d22 d26 5>>
>
> 0 >
>
>
>
: ; : ;
jxy 0 0 0 d16 d26 d66 0
Hence, the midplane strains caused by pure shear loading are given by:
eoxx ¼ a16 Nxy ð68aÞ
eoyy ¼ a26 Nxy ð68bÞ
coxy ¼ a66 Nxy ð68cÞ
The effective (or nominal) shear stress sxy applied to the laminate is given by
sxy=Nxy/t, where t is the total laminate thickness.
In Sec. 3.2, the shear modulus was defined as ‘‘the shear stress sxy
divided by the resulting shear strain cxy, with all other stress components
Figure 19 A symmetrical composite laminate subjected to shear load Nxy.

equal to zero.’’ Applying this definition to the laminate shown in Fig. 19, the
effective shear modulus referenced to the x–y coordinate axes is given by:

ex sxy Nxy =t 1
Gxy ¼ o ¼ ¼ ð69Þ
cxy a66 Nxy ta66
The coefficient of mutual influence of the first kind gxy,xx (or gxy,yy) was
defined as ‘‘the normal strain exx (or eyy) divided by the shear strain cxy, both
of which are induced by shear stress sxy, when all other stresses equal zero.’’
For a composite laminate, the effective coefficient of mutual influence of the
ex
first kind gxy;xx is therefore given by:
eoxx a16 Nxx a16
ex
gxy;xx ¼ ¼ ¼ ð70aÞ
coxy a66 Nxy a66
ex
whereas gxy;yy is given by:
eoyy a26
ex
gxy;yy ¼ ¼ ð70bÞ
coxy a66
Recall that during the derivation of classical lamination theory, we

assumed that a state of plane stress exists within a thin composite laminate.
This assumption implies that out-of-plane shear strains (cjxz and cjyz ) are equal
to zero. Consequently, Chentsov coefficients, which were defined in Sec. 3.2,
are always equal to zero for thin composite laminates.
The effective properties of a laminate obey the inverse relations fol-
lowed by any anisotropic plate, which are defined in Eq. (13) of Chap. 4. In
particular:
ex ex ex ex ex ex
mxy myx gxx;xy gxy;xx gyy;xy gxy;yy
ex ¼ ex ex ¼ ex ex ¼ ex
Exx Eyy Exx Gxy Eyy Gxy
Cross-ply, balanced, and balanced angle-ply laminates were discussed in
Secs. 7.2, 7.3, and 7.4, respectively. Recall that a16=a26=0 for these types of
stacking sequences. Therefore, coefficients of mutual influence of the first
and second kind always equal zero for cross-ply, balanced, and balanced
angle-ply laminates. Finally, in the case of a quasi-isotropic laminate (dis-
ex
cussed in Sec. 7.4), it will be found that the effective in-plane moduli Exx =
ex ex ex ex
Eyy , gxy =gyx , and Gxy are related in the same manner as for an isotropic
plate:
ex
ex E
Gxy ¼ xx
2 1 þ mxy
ex

Flexural Properties
All of the effective properties described in Sec. 9.1.1 are determined by
subjecting the composite laminate to a load whose line of action lies within
the plane of the plate, and by measuring the resulting strains. Properties
measured in this way are called effective extensional properties. In this
section, we consider properties measured by applying a pure bending moment
and measuring the resulting strains. Properties measured under a state of pure
bending are called effective flexural properties. Effective flexural properties do
not follow from the fundamental definitions of material properties reviewed
in Sec. 3.2. Rather, effective flexural properties represent the response of a
structure subjected to bending, and are defined in analogy to those exhibited
by an isotropic plate. As discussed in Sec. 7, the bending stiffnesses (i.e., the Dij
matrix) for an isotropic plate are given by:
Et3
D11 ¼ D22 ¼
12ð1 m2 Þ
mEt3
D12 ¼ mD11 ¼
12ð1 m2 Þ
ðD11 D12 Þ Et3
D66 ¼ ¼
2 24ð1 þ mÞ
Also, for an isotropic plate, the inverse of the Dij matrix (i.e., the dij matrix) is
given by:
1 12
d11 ¼ d22 ¼ ¼
D11 ð1 m2 Þ Et3
m 12m
d12 ¼ md11 ¼ ¼
D11 ð1 m2 Þ Et3
24ð1 þ mÞ
d66 ¼ 2ðd11 d12 Þ ¼
Et3
Thus, for isotropic plates, both the Dij and dij matrices can be calculated
directly from material properties E and v and the plate thickness t. Given these
results, one might anticipate that the Dij and dij matrices for a composite
laminate could be calculated in a similar manner, using the total laminate
ex ex ex ex
thickness t and effective extensional properties Exx , Eyy , gxy , and gyx . This is
not the case. The bending stiffnesses of a composite laminate (i.e., the Dij
matrix) cannot, in general, be calculated on the basis of effective extensional
properties. For example, in general:
ex
E t3
D11 p xx
2
12 1 mex
xy

for a composite laminate. In essence, this cannot be done because the flexural
properties of a laminate are dictated, in part, by the laminate stacking
sequence. In contrast, the effective extensional properties are independent of
the stacking sequence. Hence, effective flexural properties cannot be directly
related to effective extensional properties.
As stated above, effective flexural properties are defined for a state of
pure bending. Note from the above discussion that for an isotropic plate:
12
d11 ¼ d12 ¼ md11
Et3
The effective Young’s moduli and Poisson ratios exhibited by a composite
fl fl fl fl
laminate in flexure, denoted Exx , E yy , gxy , and gyx , can therefore be defined as
follows:
fl 12 fl 12
Exx ¼ Eyy ¼
d11 t3 d22 t3
d12 d12
fl
mxy ¼ fl
myx ¼
d11 d22
Although the effective flexural Young’s moduli and Poisson ratios differ from
the corresponding effective extensional properties, direct substitution will
shown that the flexural properties obey the inverse relations:
fl fl
mxy myx
fl
¼ fl
Exx Eyy
It is also interesting to note that d11 p d22 for a quasi-isotropic laminate.
fl fl
Consequently, Exx p Eyy ; mxyfl
p myx
fl
; even for a quasi-isotropic laminate.
The effective flexural coefficient of mutual influence of the second kind
fl
gxx;xy can be defined as ‘‘the midplane shear strain cjxx divided by the midplane
normal strain ejxx, both of which are induced by moment resultant Mxx, when
all other stress resultants equal zero.’’ For a composite laminate, the effective
fl
flexural coefficient of mutual influence of the second kind gxx,xy is therefore
given by:
coxy d16 Mxx d16
gflxx;xy ¼ ¼ ¼
eoxx d11 Mxx d11
An identical procedure can be employed to define the effective flexural
fl
coefficient of mutual influence of the second kind gyy,xy , a property measured
during a test in which only Myy is applied:
coxy d26 Myy d26
gflyy;xy ¼ ¼ ¼
eoyy d22 Myy d22

It is mentioned in passing that flexural coefficients of mutual influence of the
second kind are not encountered during study of isotropic plates because
d16=d26=0 for isotropic plates. The effective flexural properties will be very
useful during the study of composite beams, considered in Chap. 8.
9.2 Effective Properties Relating Temperature or Moisture

Content to Strain
As discussed in Sec. 3.3, linear coefficients of thermal expansion are measured
by determining the strains induced by a uniform change in temperature, and
eTxx eTyy cTxy

axx ¼ ayy ¼ axy ¼ ð71Þ
DT DT DT
The superscript ‘‘T’’ is included as a reminder that the strains involved are
those caused by a change in temperature only. The midplane strains and
curvatures induced by a uniform temperature change may be calculated
according to Eq. (45), which becomes (for a symmetrical laminate and for
DM ¼ Nxx ¼ Nyy ¼ Nxy ¼ Mxx ¼ Myy ¼ Mxy ¼ 0):
8 o 9 2 38 T 9
> > a a a 0 0 0 >
> N >
>
>
e xx > 11 12 12
> xx > >
> o >
> > 6 a12 a22 a26 0
> 0 0 7>> NT >
> >
>
>
> e >
yy > 6 7 > >
< o >
> = 6 7>>
<
yy >
>
6 a a a 0 0 0 7 T =
cxy ¼ 6 16 26 66
7 Nxy ð72Þ
> 6 0 0 d11 d12 d16 7
>
> jxx >>
> 6 0 7>>
> 0 > >
>
>
> >
> 6 7> >
>
>
> jyy >
> 4 0
> 0 0 d12 d22 d26 5> >
>
> 0 >
>
>
>
: ; >
: >
;
jxy 0 0 0 d16 d26 d66 0
Substituting the midplane strains indicated by Eq. (72) into Eq. (71), the
effective linear thermal expansion coefficients for a general laminate are:
1 h i
axx ¼ a11 NTxx þ a12 NTyy þ a16 NTxy
DT
1 h i
ayy ¼ a12 NTxx þ a22 NTyy þ a26 NTxy ð73aÞ
DT
1 h i
axy ¼ a16 NTxx þ a26 NTyy þ a66 NTxy
DT
It is noted in passing that these results are for a symmetrical laminate and can
therefore be inverted to give:

NTxx ¼ DT A11 axx þ A12 ayy þ A16 axy

NTyy ¼ DT A12 axx þ A22 ayy þ A26 axy ð73bÞ

NTxy ¼ DT A16 axx þ A26 ayy þ A66 axy
The effective linear coefficient of moisture expansion is measured by
determining the strains induced by a uniform change in moisture content, and
eM eM
yy cM
xy
bxx ¼ xx
byy ¼ bxy ¼ ð74Þ
DM DM DM
The superscript ‘‘M’’ in the above equations is included as a reminder that the
strains involved are those caused by a change in moisture only. The strains
induced by a change in moisture content (only) may be calculated according to
Eq. (45), which becomes (for a symmetrical laminate and for DT=Nxx=Nyy
=Nxy=Mxx=Myy=Mxy=0):
8 o 9 2 38 M 9
> exx > a11 a12 a12 0 0 0 >
>
> >
> > Nxx >>
>
> > 6 7>> M> >
>
> eyy >
o
>
> 6
a12 a22 a26 0 0 0 7>> N >
yy >
>
< o > = 6 7>> >
>
6
cxy ¼ 6 a16 a26 a66 0 0 0 7< NM =
7 xy ð75Þ
> 6 0 d16 7
> jxx >
> >
> 6 0 0 d11 d12 7>> 0 > >
>
> > 6 7>> >
>
> jyy >
> > 4 0
> 0 0 d12 d22 d26 5>>
> 0
>
>
>
>
: >
; >
: >
;
jxy 0 0 0 d16 d26 d66 0
Substituting the midplane strains indicated by Eq. (75) into Eq. (74), the
effective linear moisture expansion coefficients for a general laminate are:
1 h i
bxx ¼ a11 NM þ a 12 N M
þ a 16 N M
DM xx yy xy
1 h i
byy ¼ a12 NMxx þ a22 Nyy þ a26 Nxy
M M
ð76aÞ
DM
1 h i
bxy ¼ a16 NM þ a N
26 yy
M
þ a NM
66 xy
DM xx
Because these results are for a symmetrical laminate, they can be inverted to
give:

xx ¼ DM A11 bxx þ A12 byy þ A16 bxy
NM

yy ¼ DM A12 bxx þ A22 byy þ A26 bxy
NM ð76bÞ

xy ¼ DM A16 bxx þ A26 byy þ A66 bxy
NM

Sample Problem 8
Determine the effective extensional and flexural moduli, thermal expansion
coefficients, and moisture expansion coefficients for a [30/0/90]s graphite-
epoxy laminate. Use material properties listed for graphite-epoxy in Table 3
of Chap. 3, and assume that each ply has a thickness of 0.125 mm.
Solution. As described in this section, effective moduli are calculated using
various elements of the [abd] matrix. A six-ply symmetrical laminate is
considered in this problem. The total laminate thickness is t=6(0.000125
m)=0.000750 m. Using methods discussed in Sec. 6, the [ABD] matrix is
determined to be:
2 3
72:2 106 8:02 106 12:0 106 0 0 0
6 8:02 106 52:0 106 5:38 106 0 7
6 0 0 7
6 7
6 12:0 106 5:38 106 15:5 106 0 7
6 0 0 7
½ABD ¼ 6 7
6 0 0 0 4:23 0:676 1:19 7
6 7
6 0:676 0:988 0:532 7
4 0 0 0 5
0 0 0 1:19 0:532 1:03
Because the laminate is symmetrical, all elements of the Bij matrix are zero, as
expected. We obtain the [abd] by inverting the [ABD] numerically:
2 3
16:0 109 1:23 109 12:0 109 0 0 0
6 7
6 1:23 109 20:0 109 6:0 109 0 0 0 7
6 7
6 7
6 7
6 12:0 109 6:0 109 75:8 109 0 0 0 7
½abd ¼ 6
6
7
1 1 7
6 0 0 0 3:51 10 2:92 102 3:92 10 7
6 7
6 7
6 0 0 0 2:92 102 1:408 6:96 101 7
4 5
1
0 0 0 3:92 10 6:96 101 1:79
The effective extensional moduli of the laminate can now be calculated using
Eqs. (63)–(70b):
ex 1 1
Exx ¼ ¼ ¼ 83:3 GPa
ta11 ð0:000750Þð16:0 109 Þ
ex 1 1
Eyy ¼ ¼ ¼ 66:7 GPa
ta22 ð0:000750Þð20 109 Þ
ex 1 1
Gxy ¼ ¼ ¼ 17:6 GPa
ta66 ð0:000750Þð75:8 109 Þ

a12 1:23 109
mex ¼ ¼ ¼ 0:077
xy
a11 16:0 109

a12 1:23 109
mex ¼ ¼ ¼ 0:061
yx
a22 20 109
a16 12 109
xx;xy ¼
gex ¼ ¼ 0:75
a11 16 109
a26 6:0 109
yy;xy ¼
gex ¼ ¼ 0:30
a22 20 109
a16 12:0 109
xy;xx ¼
gex ¼ ¼ 0:16
a66 75:8 109
a26 6:0 109

xy;yy ¼
gex ¼ ¼ 0:079
a66 75:8 109
Effective flexural properties are found to be:
12 12
E flxx ¼ ¼ ¼ 81:0 GPa
d11 t 3
ð0:351Þð0:000750Þ3
12 12
E flyy ¼ ¼ ¼ 20:2 GPa
d22 t3 ð1:408Þð0:000750Þ3
d12 2:92 102

mflxy ¼ ¼ ¼ 0:083
d11 0:351
d12 2:92 102
mflyx ¼ ¼ ¼ 0:021
d22 1:408
d16 0:392
gflxx;xy ¼ ¼ ¼ 1:12
d11 0:351
d26 0:696
gflyy;xy ¼ ¼ ¼ 0:49
d22 1:408
Note that the values of extensional properties are quite different from anal-
ogous flexural properties.
The thermal stress resultants associated with a given change in temper-
ature must be determined in order to calculate the effective thermal expansion
coefficients. Numerically speaking, any change in temperature can be used,
but for present purposes, a unit change in temperature will be assumed (i.e.,

DT=1). Using (42a–(42f), the thermal stress resultants associated with DT=1
are:
NTxx jDT¼1 ¼ 52:3 N=m NTyy jDT¼1 ¼ 94:9 N=m NTxy jDT¼1 ¼ 36:9 N=m
The effective thermal expansion coefficient axx can now be calculated in

accordance with Eq. (73a):
1 h i
DT
1
axx ¼ 16:0 109 ð52:3Þ 1:23 109 ð94:9Þ
ð 1Þ
12:0 109 ð36:9Þ
axx ¼ 1:16 Am=m B C
Using an equivalent procedure:
ayy ¼ 2:06 Am=m B C axy ¼ 4:00 Arad=o C
Finally, moisture stress resultants associated with a given change in moisture

content must be determined in order to calculate the effective moisture
expansion coefficients. Numerically speaking, any change moisture content
can be used, but for present purposes, a unit change in content will be assumed
(i.e., DM=1). Using Eq. (43), the moisture stress resultants associated with
DM=1% are:
xx jDM¼1 ¼ 32; 800 N=m Nyy jDM¼1 ¼ 33; 800 N=m

NM xy jDM¼1 ¼ 930 N=m
M
NM
Applying Eqs. (76a) and (76b), we find:
bxx ¼ 494 Am=m %M byy ¼ 643 Am=m %M bxy ¼ 667 Arad %M
10 TRANSFORMATION OF THE ABD MATRIX

The Aij, Bij, and Dij matrices all involve a summation over the thickness of the
laminate in accordance with (Eq. (27a) (27b) and (34), repeated here for
convenience:
X
n
Aij ¼ Qij k
ðzk zk1 Þ ðrepeatedÞ ð27aÞ
k¼1

1X n
Bij ¼ Qij k z2k z2k1 ðrepeatedÞ ð27bÞ
2 k¼1
1X n
Dij ¼ Qij k z3k z3k1 ðrepeatedÞ ð34Þ
3 k¼1
As discussed in Sec. 5.2, Qij may be calculated using Eq. (31) of Chap. 5, which
involves use of the Qij matrix and various trigonometrical functions raised to
a power (e.g., cos4 h, cos2 h sin2h, sin4 h, etc.). Alternatively, as discussed in
Sec. 5.3, Qij may be calculated using Eq. (35) of Chap. 5, which involves the
use of material invariants UiQ and trigonometrical functions, whose argu-
ments involve fiber angles multiplied by a constant (i.e., cos 2h, cos 4h, sin 2h,
or sin 4h).
Although either approach is mathematically equivalent, in some cir-
cumstances, the use of material invariants (Eq. (35) of Chap. 5) can be
advantageous. Specifically, if all plies within the laminate are of the material
type, then use of Eq. (35) of Chap. 5 leads to the ability to easily transform the
ABD matrix from one coordinate system to another.
To aid in our development, define the following ‘‘geometry factors,’’
which are related to the fiber angles and ply interface positions:
Xn Xn
VA
0 ¼ ð z k z k1 Þ ¼ t V A
1 ¼ cos 2hk ðzk zk1 Þ
k¼1 k¼1
X
n X
n
2 ¼
VA sin 2hk ðzk zk1 Þ 3 ¼
VA cos 4hk ðzk zk1 Þ ð77Þ
k¼1 k¼1
X
n
4 ¼
VA sin 4hk ðzk zk1 Þ ¼ t
k¼1
1Xn 1X n
VB0 ¼ z2 z2k1 ¼ 0 VB1 ¼ cos 2hk z2k z2k1
2 k¼1 k 2 k¼1
1X n 1X n
VB2 ¼ sin 2hk z2k z2k1 VB3 ¼ cos 4hk z2k z2k1 ð78Þ
2 k¼1 2 k¼1
1X n
VB4 ¼ sin 4hk z2k z2k1
2 k¼1
1X n t3 1X n
0 ¼
VD z3k z3k1 ¼ 1 ¼
VD cos 2hk z3k z3k1
3 k¼1 12 3 k¼1
1X n 1X n
2 ¼
VD sin 2hk z3k z3k1 VD
3 ¼ cos 4hk z3k z3k1 ð79Þ
3 k¼1 3 k¼1

1X n
4 ¼
VD sin 4hk z3k z3k1
3 k¼1
Next, consider the steps necessary to calculate the A11 term. In this case, Eq.
(27a) becomes:
X
n
A11 ¼ Qij k ðzk zk1 Þ ð80Þ
k¼1
Using the invariant formulation the expression for Q 11 listed in Eq. (35) of
Chap. 5 may be substituted, resulting in:
n n
X o
A11 ¼ UQ Q Q
1 þ U2 cos 2h þ U3 cos 4h ðzk zk1 Þ ð81Þ
k
k¼1
where U1Q, U2Q, and U3Q are the stiffness invariants defined in Eq. (36) of
Chap. 5. If all plies within the laminate are composed of the same material,
then stiffness invariants are constant for the laminate and only fiber angle h
varies from one ply to the next. Therefore, Eq. (81) can be rewritten as:
X
n X
n
A11 ¼ UQ
1 ðzk zk1 Þ þ UQ
2 cos 2hk ðzk zk1 Þ
k¼1 k¼1
ð82Þ
Q
X
n
þ U3 cos 4hk ðzk zk1 Þ
k¼1
The geometry factors VA A A

0 , V1 , and V3 appear in Eq. (82), and hence the
equation may be written as:
A11 ¼ UQ Q A Q A
1 V0 þ U2 V1 þ U3 V3
A
ð83aÞ
Following an identical procedure, the remaining elements of the Aij matrix are
given by:
A22 ¼ UQ Q A Q A
1 V0 U2 V1 þ U3 V3
A
ð83bÞ
A12 ¼ A21 ¼ UQ Q A
4 V0 U3 V3
A
ð83cÞ
A66 ¼ UQ A
5 V0 UQ A
3 V3 ð83dÞ
1
A16 ¼ A61 ¼ UQ VA þ UQ A
3 V4 ð83eÞ
2 2 2
1
A26 ¼ A62 ¼ UQ VA UQ A
3 V4 ð83f Þ
2 2 2

Analogous procedures can be used to calculate members of the Bij and Dij
matrices. It can be shown that elements of the Bij matrix are given by:
B11 ¼ UQ Q B Q B
1 V0 þ U2 V1 þ U3 V3
B
ð84aÞ
B22 ¼ UQ B
1 V0 UQ Q B
2 V1 þ U3 V3
B
ð84bÞ
B12 ¼ B21 ¼ UQ Q B
4 V0 U3 V3
B
ð84cÞ
B66 ¼ UQ Q B
5 V0 U3 V3
B
ð84dÞ
1
B16 ¼ B61 ¼ UQ VB þ UQ B
3 V4 ð84eÞ
2 2 2
1
B26 ¼ B62 ¼ UQ VB UQ B
3 V4 ð84f Þ
2 2 2
Members of the Dij matrix are given by:
D11 ¼ UQ Q D Q D
1 V0 þ U2 V1 þ U3 V3
D
ð85aÞ
D22 ¼ UQ D
1 V0 UQ Q D
2 V1 þ U3 V3
D
ð85bÞ
D12 ¼ D21 ¼ UQ Q D
4 V0 U3 V3
D
ð85cÞ
D66 ¼ UQ Q D
5 V0 U3 V3
D
ð85dÞ
1
D16 ¼ D61 ¼ UQ VD þ UQ VD ð85eÞ
2 2 2 4
1 Q D
D26 ¼ D62 ¼ U2 V2 UQ D
3 V4 ð85f Þ
2
Let us now consider transformation of the ABD matrix from one
coordinate system to another. A multiangle composite laminate referenced
to an x–y coordinate system is shown in Fig. 20. It is assumed that the ABD
matrix for this laminate has been calculated and is known based on fiber
angles referenced to the x–y coordinate system. Now suppose that a different
coordinate system is of interest, the x V–yV coordinate system, orientated h
degrees counterclockwise from the original x–y coordinate system. The
transformed stiffness matrices referenced to the x V–yV coordinate system will
U U U
be labeled Aij, Bij, and Dij.
Inspection of Fig. 20 shows that a ply with fiber angle hk relative to the
original x-axis will form an angle (hk/) relative to the xV-axis. Element A 11
may therefore be calculated using Eqs. (83a)–(83f) by substituting (hk/) for
hk. Hence:
X n X
n
A 11 ¼ U1Q ðzk zk1 Þ þ UQ
2 ½cos 2ðhk /Þðzk zk1 Þ ð86Þ
k¼1 k¼1
Xn
þUQ
3 ½cos 4ðhk /Þðzk zk1 Þ
k¼1

Figure 20 A general composite laminate, showing the ‘‘original’’ x–y
coordinate system and the ‘‘new xV–yV coordinate system.
Recalling the general trigonometrical identity:

cosða bÞ ¼ cos a cos b þ sin a sin b
Equation (86) can be written as:
Q A
A11 ¼ UQ 1 V0 þ U2 V1 cosð2/Þ þ V2 sinð2/Þ
A A
ð87aÞ
þUQ 3 V3 cosð4/Þ þ V4 sinð4/Þ
A A
The geometry factors ViA have been defined in Eq. (77). A similar procedure
can be applied to all remaining elements of the Aij matrix as well as the Bij
and Dij matrices:
Q A
A22 ¼ UQ 1 V0 U2 V1 cosð2/Þ þ V2 sinð2/Þ
A A
A ð87bÞ
þUQ 3 V3 cosð4/Þ þ V4 sinð4/Þ
A
Q A
A12 ¼ A21 ¼ UQ 4 V0 U3 V3 cosð4/Þ þ V sinð4/Þ
A A
ð87cÞ

A66 ¼ UQ A Q
5 V0 U3 V3 cosð4/Þ þ V4 sinð4/Þ
A A
ð87dÞ
1 Q A
A16 ¼ A61 ¼ U2 V2 cosð2/Þ VA1 sinð2/Þ
2 ð87eÞ
þUQ
3 V4 cosð4/Þ V3 sinð4/Þ
A A
1 Q A
A26 ¼ A62 ¼ U2 V2 cosð2/Þ VA1 sinð2/Þ
2 ð87f Þ
UQ
3 V4 cosð4/Þ V3 sinð4/Þ
A A

B11 ¼ U1Q VB0 þ U2Q V1B cosð2/Þ þ VB2 sinð2/Þ
ð88aÞ
þ U3Q V3B cosð4/Þ þ V4B sinð4/Þ

B22 ¼ U1Q V0B U2Q V1B cosð2/Þ þ V2B sinð2/Þ
ð88bÞ
þU3Q V3B cosð4/Þ þ V4B sinð4/Þ

B12 ¼ B21 ¼ U Q V0B U3Q V3B cosð4/Þ þ V4B sinð4/Þ ð88cÞ

B66 ¼ U5Q V B U3Q V3B cosð4/Þ þ V4B sinð4/Þ ð88dÞ
1 Q B
B16 ¼ B61 ¼ U2 V2 cosð2/Þ V1B sinð2/Þ
2
ð88eÞ
þU3Q V4B cosð4/Þ V3B sinð4/Þ
1 Q B
B26 ¼ B62 ¼ U2 V2 cosð2/Þ V1B sinð2/Þ
2 ð88f Þ
U3Q V4B cosð4/Þ V3B sinð4/Þ

D11 ¼ U1Q V0D þ U2Q V1D cosð2/Þ þ V2D sinð2/Þ
ð89aÞ
þU3Q V3D cosð4/Þ þ V4D sinð4/Þ

D22 ¼ U Q V0D U2Q V1D cosð2/Þ þ V2D sinð2/Þ
ð89bÞ
þU3Q V3D cosð4/Þ þ V4D sinð4/Þ

D12 ¼ D21 ¼ U4Q V D U3Q V3D cosð4/Þ þ V4D sinð4/Þ ð89cÞ

D66 ¼ U5Q V0D U3Q V3D cosð4/Þ þ V4D sinð4/Þ ð89dÞ
1 Q D
D16 ¼ D61 ¼ U2 V2 cosð2/Þ V1D sinð2/Þ
2 ð89eÞ

þU3Q V4D cosð4/Þ V3D sinð4/Þ
1 Q D
D26 ¼ D62 ¼ U V2 cosð2/Þ V1D sinð2/Þ
2 2 ð89f Þ

U3Q V4D cosð4/Þ V3D sinð4/Þ
It is again emphasized that all new results presented in this section (in
particular, Eqs. (82a)–(89f ) are valid only if all plies within the laminate are
of the same material type. In many cases, this is a severe restriction. These
equations cannot be used to calculate the ABD matrix for a hybrid composite

laminate, for example. On the other hand, an advantage of the invariant
approach is the ability to easily rotate the ABD matrix from one coordinate
system to another using Eqs. (88a)–(89f), so in the proper circumstances, the
invariant approach is convenient.
11 COMPUTER PROGRAM CLT

The computer program CLT has been developed to supplement the material
presented in this chapter. This program can be downloaded at no cost from
the following website: http://depts.washington.edu/amtas/computer.html.
Program CLT can be used to recreate all numerical results discussed in
the Example Problems presented in this chapter. In essence, the calculation
steps described in Sec. 8 are implemented in this program. The user may select
two different analysis paths. One analysis path corresponds to the case in
which the loads applied to the laminate are specified, whereas the second
corresponds to the case in which midplane strains and curvatures applied to
the laminate are specified. In either case, the user must provide various
numerical values required during the calculations performed. The user must
define these values using a consistent set of units. For example, the user must
input elastic moduli, thermal expansion coefficients, and moisture expansion
coefficients for the composite material system(s) of interest. Using the
properties listed in Table 3 of Chap. 3 and based on the SI system of units,
the following numerical values would be inputted for graphite-epoxy:
E11 ¼ 170 109 Pa E22 ¼ 10 109 Pa m12 ¼ 0:30 G12 ¼ 13 109 Pa
a11 ¼ 0:9 106 m=m B C a11 ¼ 27:0 106 m=m B C
b11 ¼ 150:0 106 m=m %M b22 ¼ 4800 106 m=m %M
If the analysis requires the user to input numerical values for stress and
moment resultants, then stress resultants must be inputted in Newtons per
meter, and moment resultants must be inputted in Newton meters per meter.
Typical value would be Nxx=150103 N/m and Mxx=5 N m/m. If, instead,
the analysis requires the user to input numerical values for midplane strains
and curvatures, then strains must be inputted in meters per meter (not in Am/
m) and curvatures must be inputted in per meter. Typical value would be
ejxx=2000106 m/m=0.002000 m/m and jxx=0.5 m1. All temperatures
would be inputted in degrees Celsius. Ply thicknesses must be inputted in
meters (not millimeters). A typical value would be t k =0.000125 m
(corresponding to a ply thickness of 0.125 mm).

In contrast, if the English system of units were used, then the following
numerical values would be inputted for the same graphite-epoxy material
system:
E11 ¼ 25:0 106 psi E22 ¼ 1:5 106 psi m12 ¼ 0:30 G12 ¼ 1:9 106 psi
a11 ¼ 0:5 106 in=in B F a11 ¼ 15 106 in=in B F
b11 ¼ 150:0 106 in=in %M b22 ¼ 4800 106 in=in %M

Stress resultants must be inputted in pound-forces per inch, and moment
resultants must be inputted in pound-force inches per inch. Typical value
would be Nxx=1000 lbf/in. and Mxx=1 lbf in./in. If, instead, the analysis
requires the user to input numerical values for midplane strains and cur-
vatures, then strains must be inputted in inches per inch and curvatures
must be inputted in per inch. Typical value would be ejxx=2000 106 in./
in. = 0.002000 in./in. and jxx = 0.01 in1. All temperatures would be inputted
in degrees Fahrenheit. Ply thicknesses must be inputted in inches. A typical
value would be tk = 0.005 in.
HOMEWORK PROBLEMS
Notes: (a) In the following problems, the phrase ‘‘by hand calculation’’
means that solutions are to be obtained using a calculator, pencil, and paper.
(b) The computer programs UNIDIR and CLT are referenced in some of
the following problems. As described in Sec. 11, these programs can be
downloaded from the following website: http://depts.washington.edu/amtas/
computer.html.
1. Three-element strain gage rosettes are mounted on opposite sides of a [0/

F30]s graphite-epoxy laminate, as shown in Fig. 21. An individual ply
has a thickness of 0.005 in. The laminate is then subjected to an unknown
system of forces. The strains measured by strain gage rosette 1 are:
exx ¼ 2000 Ain:=in: eyy ¼ 500 Ain:=in: cxy ¼ 1000 Arad
Similarly, the strains measured by strain gage rosette 2 are:
Determine by hand calculation:
(a) Midplane strains and curvatures induced in the laminate

(b) Strains (exx,eyy,cxy) induced at the interface between plies 1
and 2.

Figure 21 Edge view of the composite laminate described in Problem 1.
2. Imagine that a new room temperature cure graphite-epoxy prepreg

system has been developed. A [0/F30]s laminate is produced at room
temperature using this material, and is therefore initially stress-free and
strain-free. Strain gage rosettes are mounted on opposite sides of the
laminate, as shown in Fig. 21. The laminate is then subjected to an
unknown system of forces, whereas temperature and moisture content
remain constant. The strains measured by rosette 1 are:
Similarly, the strains measured by rosette 2 are:

exx ¼ 1000 in:=in: eyy ¼ 2000 Ain:=in: cxy ¼ 750 Arad

Obtain numerical values for all elements of the Q matrices for plies 1
and 2 using program UNIDIR and properties listed in Table 3 of Chap.
3. Assume that an individual ply has a thickness of 0.005 in. Then
determine the following by hand calculation:

and 2
(c) Stresses (rxx,ryy,sxy) induced in ply 1, at the interface between
plies 1 and 2
(d) Stresses (rxx,ryy,sxy) induced in ply 2, at the interface between
plies 1 and 2.

3. Repeat Problem 2 for a glass-epoxy system.
4. Repeat Problem 2 for a Kevlar-epoxy system.
5. Imagine that a new room temperature cure graphite-epoxy prepreg
system has been developed. A [0/F30]s laminate is produced at room
temperature using this material, and is therefore initially stress-free and
strain-free. Strain gage rosettes are mounted on opposite sides of the
laminate, as shown in Fig. 21. The laminate is then subjected to an
unknown system of forces and a temperature increase of 250jF (mois-
ture content remain constant). The strains measured by rosette 1 are:
Similarly, the strains measured by rosette 2 are:
Obtain numerical values for all elements of the [Q] matrices for plies 1
and 2 using program UNIDIR and properties listed in Table 3 of Chap.
3. Assume that an individual ply has a thickness of 0.005 in. Then
determine the following by hand calculation:

and 2
(c) Stresses (rxx,ryy,sxy) induced in ply 1, at the interface between
plies 1 and 2
(d) Stresses (rxx,ryy,sxy) induced in ply 2, at the interface between
plies 1 and 2.
6. Repeat Problem 5 for a glass-epoxy system.

7. Repeat Problem 5 for a Kevlar-epoxy system.
8. An engineer is designing a structure that involves a [0/10/90]s graphite-
epoxy composite laminate that will be cured at 350jF. During service,
the structure must support a load of 1000 lbf, and will experience a
change of temperature of 150jF in a dry environment. Based on the cure
temperature and expected service conditions, the engineer predicts that
the laminate will experience the following midplane strains and curva-
tures:
eoxx ¼ 1500 Ain:=in:
eoyy ¼ 2200 Ain:=in:
coxy ¼ 1000 Arad

jxx ¼ jyy ¼ jxy ¼ 0

Obtain numerical values for all elements of the Q matrix for ply 3 using
program UNIDIR and properties listed in Table 3 of Chap. 3. Assume
that an individual ply has a thickness of 0.005 in. Then determine the
stresses induced in ply 3 relative to the 1-2 coordinate system by hand
calculation.
9. Repeat Problem 8 for a glass-epoxy structure.

10. Repeat Problem 8 for a Kevlar-epoxy structure.
11. A [0/F30/90]s graphite-epoxy laminate is cured at 350jF and then
cooled to room temperature (70jF). Use hand calculation to determine
the following thermal resultants induced during cooling (use program
UNIDIR and properties
listed in Table 3 of Chap. 3 to determine
elements of the Q matrix as necessary):
T
(a) Nxx
T
(b) Nyy
T
(c) Nxy
(d) MxxT
(e) MTyy
T:
(f) Mxy
12. Repeat Problem 11 for a glass-epoxy structure.

13. Repeat Problem 11 for a Kevlar-epoxy structure. Note the following:
A [0/F30/90]s graphite-epoxy laminate will be considered in Problems

14–18. Assuming a ply thickness of 0.125 mm and using properties listed
in Table 3 of Chap. 3 (SI units), the [Q] matrix for ply 2 is:
2 3
10:76E10 2:606E10 4:813E10
6 7
4 2:606E10 2:722E10 2:152E10 5
4:813E10 2:152E10 3:605E10
The [ABD] matrix for this laminate is:
2 3
9:906E7 1:454E7 0 0 0 0
6 1:454E7 5:885E7 0 7
6 0 0 0 7
6 7
6 0 0 2:452E7 0 0 0 7
6 7
6 0:7521 7
6 0 0 0 11:89 1:032 7
6 7
4 0 0 0 1:032 1:628 0:3362 5
0 0 0 0:7521 0:3362 1:864

The [abd] matrix for this laminate is:
2 3
1:048E 8 0:2558E 8 0 0 0 0
6 0:2558E 8 1:763E 8 0 0 0 70
6 7
6 7
6 0 0 4:078E 8 0 0 0 7
6 7
6 9:029E 2 5:160E 2 2:713E 2 7
6 0 0 0 7
6 7
6 0 0 0 5:160E 2 0:6674 9:958E 2 7
4 5
0 0 0 2:713E 2 9958E 2 0:5655
If the laminate were cured at 175jC and then cooled to room tempera-
tures (20jC), the thermal resultants induced during cooldown would be:
NTxx ¼ 8607 N=m
NTyy ¼ 21; 825 N=m
NTxy ¼ MTxx ¼ MTyy ¼ MTxy ¼ 0
If the moisture content of the laminate immediately after cure is 0%, but
over the course of several months is increased to 0.5%, then the
moisture resultants induced by this slow moisture adsorption are:
xx ¼ 6092 N=m
NM
yy ¼ 6098 N=m
NM
xy ¼ Mxx ¼ Myy ¼ Mxy ¼ 0

NM M M M
14. A [0/F30/90]s graphite-epoxy laminate is cured at 175jC and then

cooled to room temperatures (20jC). Use hand calculation to determine
the following:
(a) Midplane strains and curvatures induced during cooldown

(b) Strains (exx,eyy,cxy) induced in ply 2 during cooldown
(c) Stresses (rxx,ryy,sxy) induced in ply 2 during cooldown
(d) Stresses (r11,r22,s12) induced in ply 2 during cooldown
15. A [0/F30/90]s graphite-epoxy laminate is cured at 175jC and then

cooled to room temperatures (20jC). Although the moisture content
immediately after cure was 0%, the laminate is stored in a humid
environment and, over the course of several months, the moisture
content is increased to 0.5%. Use hand calculation to determine the
following:
(a) Midplane strains and curvatures induced after moisture
content increased to 0.5%

(b) Strains (exx,eyy,cxy) induced in ply 2 after moisture content
increased to 0.5%
(c) Stresses (rxx,ryy,sxy) induced in ply 2 after moisture content
increased to 0.5%
(d) Stresses (r11,r22,s12) induced in ply 2 after moisture content
increased to 0.5%
16. A [0/F30/90]s graphite-epoxy laminate is subjected to the following
loads:
Nxx ¼ Nyy ¼ Nxy ¼ 10; 000 N=m Mxx ¼ Myy ¼ Mxy ¼ 0
Ignoring thermal and moisture effects, use hand calculation to determine
the following:
(b) Strains (exx, eyy, cxy) induced in ply 2
(c) Stresses (rxx, ryy,sxy) induced in ply 2
(d) Stresses (r11,r22, s12) induced in ply 2.
17. A [0/F30/90]s graphite-epoxy laminate is cured at 175jC and then cooled
to room temperatures (20jC). The moisture content immediately after
cure was 0%. However, the laminate is stored in a humid environment
and, over the course of several months, the moisture content is increased
to 0.5%. The laminate is then subjected to the following loads:
Nxx ¼ Nyy ¼ Nxy ¼ 10; 000 N=m Mxx ¼ Myy ¼ Mxy ¼ 0
Use hand calculation to determine the following:
(b) Strains (exx,eyy,cxy) induced in ply 2
(c) Stresses (rxx,ryy,sxy) induced in ply 2
(d) Stresses (r11,r22,s12) induced in ply 2.
18. A 10 10 in.2 [0/F30/90]s graphite-epoxy laminate is supported between
three infinitely rigid walls and rollers, as shown in Fig. 22. A load
Nxx=7500 N/m is applied to the plate. Ignoring thermal effects,
moisture effects, and the possibility of buckling, use hand calculation
to determine Nxx, Nyy, Nxy, ejxx, ejyy, and cjxx.
19. A [20/65/25]s graphite-epoxy laminate is cured at 175jC and then
cooled to room temperatures (20jC). Moisture content remains at 0%.
The following loads are then applied:
Nxx ¼ 30 kN=m Nyy ¼ 7 kN=m Nxy ¼ 0

Nxx ¼ 10 N m=m Myy ¼ Mxy ¼ 0

Figure 22 A [0/F30/90]s graphite-epoxy laminate described in Problem 18.
Using properties listed in Table 3 of Chap. 3 and assuming ply thick-

nesses of 0.125 mm:
(a) Determine all ply strains and stresses using program CLT.
(b) Prepare plots similar to Fig. 13, showing the through-thickness
variation of strains exx, eyy, and cxy.
(c) Prepare plots similar to Fig. 14, showing the through-thickness
variation of strains e11, e22, and c12.
(d) Prepare plots similar to Fig. 15, showing the through-thickness
variation of stresses rxx, ryy, and sxy.
(e) Prepare plots similar to Fig. 16, showing the through-thickness
variation of stresses r11, r22, and s12.
20. Repeat Problem 19 for a glass-epoxy laminate.
21. Repeat Problem 19 for a Kevlar-epoxy laminate.
22. A [20/65/-25]s ‘‘hybrid’’ laminate is cured at 175jC and then cooled to
room temperatures (20jC). Plies 1 and 6 are graphite-epoxy, plies 2 and 5
are glass-epoxy, and plies 3 and 4 are Kevlar-epoxy. The following loads
are then applied:
Nxx ¼ 30 kN=m Nyy ¼ 7 kN=m Nxy ¼ 0

Nxx ¼ 10 N m=m Myy ¼ Mxy ¼ 0
Moisture content remains constant at 0%. Using properties listed in

Table 3 of Chap. 3 and assuming that the graphite-epoxy, glass-epoxy,

and Kevlar-epoxy plies have thicknesses of 0.125, 0.200, and 0.15 mm,
respectively:
(a) Determine all ply strains and stresses using program CLT.
(b) Prepare plots similar to Fig. 13, showing the through-thickness
variation of strains exx, eyy, and cxy.
(c) Prepare plots similar to Fig. 14, showing the through-thickness
variation of strains e11, e22, and c12.
(d) Prepare plots similar to Fig. 15, showing the through-thickness
variation of stresses rxx, ryy, and sxy.
(e) Prepare plots similar to Fig. 16, showing the through-thickness
variation of stresses r11, r22, and s12.
REFERENCES
1. Jones, R.M. Mechanics of Composite Materials; Hemisphere Publishing Cor-
poration: New York, NY. ISBN 0-89116-490-1, 1975.
2. Tsai, S.W.; Hanh, H.T. Introduction to Composite Material; Technomic Pub-
lishing Co. ISBN 0-87762-288-4, 1980.
3. Halpin, J.C. Primer on Composite Materials: Analysis; Technomic Publishing
Co. ISBN 87762-349-X, 1984.

7
Predicting Failure of a Multiangle
Composite Laminate
Ideally, the objective of this chapter would be to describe the analytical tools
and/or methodologies available to accurately predict the yielding and fracture
of multiangle composite laminates under general thermomechanical loading
conditions. Unfortunately, we will not be able to reach this objective. As will
be seen, predicting the fracture of multiangle composite laminates under
general loading conditions and with a high degree of accuracy is still beyond
the state of the art, despite extensive research efforts undertaken over the past
several decades. Of course, fracture predictions for the ‘‘simple’’ case of an
isotropic metallic structure under general loading conditions are not com-
pletely reliable either, despite more than a century of effort. In any case,
predicting the fracture of polymeric composite structure has proven to be a
formidable challenge, and methods of predicting this phenomenon remain an
active area of research.
The difficulties encountered are many and varied. Some of the most
common factors involved are summarized in the following introductory
section, and a more detailed discussion of several of these factors will be
presented in following sections. However, the reader should be aware from
the outset that composite fracture predictions are currently a blend of
engineering art and science. The reader is advised to keep abreast of new
literature in this area.
375

1 PRELIMINARY DISCUSSION
1.1 Yielding and Fracture of Isotropic Metals vs.
Unidirectional Polymeric Composites
Let us begin our discussion of the difficulties posed by composites by
contrasting the initial nonlinear behavior of isotropic metals and metal alloys
to that of a unidirectional composite laminate. As is well known, the initial
nonlinear deformation (i.e., yielding) of a metal or metal alloy can be ex-
plained based on the formation and subsequent motion and coalescence of
various types of imperfections within the crystalline atomic structure of the
metal/metal alloy (e.g., dislocations). Nonlinear behavior of polymeric
composites (here loosely termed ‘‘yielding’’) can also be explained on the
basis of atomic and/or molecular mechanisms. However, a composite consists
of at least two constituents (the reinforcing fiber and the polymeric matrix),
and the atomic and/or molecular structure of these two constituents differs
substantially. Furthermore, the molecular structure of the polymeric matrix
in the immediate vicinity of the fiber/matrix interface usually differs from the
molecular structure at positions removed from the fiber. That is, the molec-
ular structure developed during polymerization of a polymer in bulk differs
from that developed in the region immediately adjacent to the surface of the
fiber. The region near the fiber is often referred to as the interphase, rather
than the ‘‘interface,’’ and the mechanical properties exhibited by the polymer
in the interphase differ from bulk properties. Hence, from a continuum-
mechanics point of view, three more or less distinct ‘‘materials’’ can be
defined: the fiber, the matrix, and the fiber–matrix interphase.
Consider the case of graphite–epoxy. As discussed in Sec. 3 of Chap. 1,
graphite is a highly ordered crystalline structure, and the high strength and the
high stiffness exhibited by a graphite fiber at the macroscale are achieved by
aligning the basal planes of the graphite crystal with the axis of the fiber. The
macroscopical stress–strain response of an individual graphite fiber is almost
perfectly linear up to final fracture. In contrast, epoxy is a cross-linked ther-
moset and generally does not possess a crystalline structure at the atomic level
(see Sec. 2 of Chap. 1). The stress–strain response of epoxy becomes nonlinear
when stresses are high enough to cause segments of the overall molecular
structure to slide relative to one another. At the macroscopical level, epoxy
exhibits a ductile stress–strain response if cross-link density is low (i.e.,
relatively low stress levels can cause segments of the molecular structure to
slide past one another), but becomes increasingly brittle as cross-link density
is increased. The macroscopical stress–strain response of the interphase is
difficult to measure (or even define), but is likely somewhere between that of
the fiber and matrix. The stress necessary to cause dislocations or other
imperfections to develop or move within the crystalline lattice of the graphite

fiber are far higher than those necessary to cause molecular segments to slide
in the epoxy matrix. The initial nonlinear deformations exhibited by a graph-
ite–epoxy composite are therefore almost entirely initiated within the poly-
meric matrix.
As a metal or metal alloy is loaded beyond the yield point, the crystalline
imperfections coalesce to form microcracks, often located (at least initially) at
grain boundaries. As loading is further increased (and/or if loading fluctuates
with time, as in fatigue loading), these microcracks grow in size until they may
be observed with a low-power microscope or with the naked eye. The principal
stresses present in the isotropic structure ultimately govern the rate at which
cracks form and grow. Furthermore, the orientation of the principal stress
coordinate system governs the direction of crack propagation. The final frac-
ture of metals and metal alloys occurs when the crack reaches a critical length,
at which point the crack generally propagates at a high rate of speed and the
metal fractures.
In a composite, the fracture process is also initiated when one or more
microcracks are formed. Although a crack may form in either the graphite
fiber or the epoxy matrix, the ‘‘imperfections’’ that lead to the formation of a
crack in these two mediums are wholly different. In addition, the stress levels
necessary to cause crack growth in a graphite fiber are at least an order of
magnitude higher than that necessary to cause crack growth in epoxy. Con-
sequently, cracks are far more likely to form in the polymeric matrix than in
the fiber. The orientation of matrix crack(s) that forms in a unidirectional
laminate is invariably related to the fiber direction (e.g., cracks form either
parallel or perpendicular to the fibers) and is independent of the principal
stress and principal stress coordinate system when defined at the macroscale.
The following types of cracks are observed in unidirectional composite
laminates:
Matrix cracks. These are cracks that occur in the polymeric matrix, at
some distance from the fiber–matrix interface. Matrix cracks gen-
erally occur in planes either parallel or perpendicular to the fiber
direction.
Fiber–matrix debonding. In this case, the crack has formed in the
interphase region, and a (nonplanar) crack extends around the peri-
phery of the fiber.
Fiber cracks. These are cracks that occur in the fiber itself. Fiber
cracks almost always occur in a plane perpendicular to the axis of the
fiber, and extend across the entire width of the fiber.
Another complication is time dependency. Under most conditions,
yielding and crack growth in metals and metal alloys, at least those used in
load-bearing structural applications, can be considered to be independent of

time at room temperatures.* For example, if a tensile stress is applied to a
metal structure and initially causes yielding, no further yielding occurs if the
tensile stress is held constant for long times. Similarly, if a tensile stress that
causes some (noncatastrophic) amount of crack growth is applied, no further
crack growth occurs if the tensile stress is held constant for long times.y
Graphite fibers are also time-independent. In contrast, yielding and crack
growth in polymers are time-dependent phenomena, even at room temper-
atures. An increase in temperature and/or an increase in moisture content
further accentuates the time dependency. Hence, polymer–matrix composites
are time-dependent materials. If a tensile load is applied to a unidirectional
[90j]n composite and held constant for a long period of time, the resulting
strain will slowly increase (i.e., the composite exhibits a creep response).
Similarly, if a tensile stress is applied and held constant, the composite may
eventually fail due to slow crack growth (often called a ‘‘creep-to-rupture’’
failure). Although the time-dependent behavior exhibited by modern poly-
meric composites for short times at room temperature is usually minimal,
these effects must be considered if a composite structure is intended for years
or tens of years of service, or if the structure will be exposed to elevated
temperatures.
1.2 Failure of Multiangle Composite Laminates

The preceding discussion concerned the failure of unidirectional composites.
Recall that several macromechanics-based failure criteria were discussed and
applied to unidirectional composites in Secs. 5 and 6 of Chap. 5. Predictions
obtained using these criteria are based on fracture/yield strengths measured
during simple uniaxial and pure shear tests. Although these macromechanical
criteria cannot be used to capture the details of crack coalescence and growth,
they may nevertheless be used to predict the failure of a unidirectional com-
posite laminate subjected to an arbitrary state of plane stress (assuming time-
dependent effects are not pronounced) with an acceptable degree of accuracy.
Predicting the failure of multiangle composite laminates is far more difficult
than for unidirectional composites, however. One difficulty is that a single
macroscopical crack or flaw has no measurable impact on the overall
mechanical response of a multiangle composite laminate. This is in direct
* Roughly, yielding and crack growth in metals/metal alloys become time-dependent if the
temperature exceeds one-half the melting temperature, measured on an absolute scale (Kelvin
or Rankine).
y
The significance and importance of crack growth, as related to yielding and fracture of metallic
structures, began to be recognized in the 1930s and 1940s, and have ultimately led to the
development of the branch of engineering known as fracture mechanics [1].

contrast to the case of a unidirectional composite (e.g., a [90]n laminate), in
which the final fracture is ultimately governed by a single (or, at most, a few)
crack or flaw. Hence, in the case of a multiangle laminate, the successful
prediction of a single-ply fracture does not necessarily lead to a successful
prediction of overall laminate fracture. As will be seen, the final fracture of a
multiangle laminate does not occur until hundreds or even thousands of
cracks and other flaws have formed.
A second factor is that multiangle laminates are subject to failure modes
that do not exist in unidirectional laminates. For example, multiangle lam-
inates are subject to delamination failures. A delamination occurs when the
bond between adjacent plies fails, such that a crack forms in a plane parallel to
the plies. The initiation of delamination failures is often attributed to free-
edge stresses, discussed in Sec. 2. As will be seen, free-edge stresses occur
whenever adjacent plies possess differing Poisson ratios or coefficients of
mutual influence.
Thirdly, pre-existing thermal and/or moisture stresses occur in multi-
angle laminates due to a mismatch in effective thermal expansion and
moisture expansion coefficients from one ply to the next. As discussed in
previous chapters, thermal and moisture stresses are substantial and contrib-
ute toward the failure of individual plies and the final fracture of the laminate
as a whole.
The many difficulties encountered when attempting to predict the
failure of a multiangle laminate can be summarized by considering the
damage induced within an eight-ply symmetrical quasi-isotropic [0/45/90/
45]s laminate subjected to tensile loading. A highly idealized (but more or
less representative) response is shown in Fig. 1. It is assumed that the laminate
is defect-free prior to loading, which implies that any pre-existing thermal
and/or moisture stresses are not high enough to have caused any ply failures
or other defects. It is also assumed that the laminate is tested under conditions
in which time-dependent factors are not an issue. A uniaxial tensile load Nxx
(or, equivalently, a tensile effective stress r̄xx ¼ Nxx =t) is applied and steadily
increased until the final laminate fracture occurs.
An axial strain exx is, of course, induced as r̄xx is increased from zero.
The initial slope of the r̄xx vs: exx represents the effective Young’s modulus of
the laminate, Ēxx, and can be predicted using classical lamination theory
(CLT), as discussed earlier. As r̄xx is increased, individual ply stresses are
increased as well. Eventually, ply stresses are increased to the point that ply
stresses are no longer linearly related to ply strains (i.e., the ply ‘‘yields’’) (see
Sec. 5 of Chap. 3). At somewhat higher load levels, cracks begin to form in one
(or more) ply. For a [0/45/90/45]s laminate subjected to uniaxial tensile
loading, the first plies to yield and then crack are almost always the 90j plies;
that is, the 90j plies yield when r̄xx ¼ Nxx =t has been increased to a critical

Figure 1 An idealized stress–strain plot for a [0/45/90/45]s laminate,
showing the evolution of internal damage.
level. The difference between the load level at which yielding is initiated and
the level at which cracks begin to form within the 90j plies depends on the
level of ductility exhibited by the particular composite material system. In this
text, ply yielding is considered to be a form of composite failure, and the
effective stress level necessary to cause yielding of the 90j plies is called the
first-ply failure stress. Hence, cracks begin to form in the 90j plies at load
levels above the first-ply failure stress. Note that if the laminate were a
unidirectional [90]n laminate, then final fracture would occur as soon as a
single crack formed within the 90j plies. Because the laminate is instead a [0/
45/90/45]s laminate and because no fractures have yet occurred within the
F45j or 0j plies, the laminate as a whole can support higher stress/strain
levels. Further, experimentally, it has been shown that many cracks form
within the 90j plies as the load is increased, and the distance between cracks is
(approximately) constant. The characteristic spacing between cracks within

the 90j plies can be predicted on the basis of so-called shear lag models. A
number of researchers have developed shear lag models. A detailed discussion
of such models will not be presented here, but the interested reader should
refer to Refs. 2–5 for representative example analyses.
Once the 90j plies have cracked/fractured, they can no longer contrib-
ute fully to the effective stiffness of the laminate. Hence, it would be expected
that the slope of the r̄xx vs: exx curve (i.e., the effective Young’s modulus Exx)
would decrease as cracks form in the 90j plies. Although such a decrease has
been observed in practice, it is often barely discernible (a pronounced decrease
in slope has been shown in Fig. 1 for illustrative purposes; in practice, the
change in slope is far less than that implied in the figure).* The failure of
the 90j plies has little impact on the effective Young’s modulus because the
stiffness of these plies in the x-direction, relative to the stiffness of the F45j
and 0j plies, is very low even before failure occurs.
As the effective stress level is further increased, cracks eventually begin
to form within the F45j plies. Once again, many cracks form within the F45j
plies, and these cracks tend toward a characteristic spacing that may be
predicted based on a shear lag model. At this elevated stress level, extensive
matrix cracking has been induced within the 90j and F45j plies, and yet no
significant damage has yet occurred within the 0j plies. Hence, the laminate as
a whole can support still higher effective stress levels. An additional decrease
in slope (i.e., a decrease in apparent Young’s modulus) occurs as cracks
develop in the F45j plies.
Thus far, the cracks that have formed within the 90j and F45j plies all
lie within planes perpendicular to the x–y plane. As the effective stress is
increased further still, delamination failures begin to develop. That is, matrix
cracks begin to form between plies, and these new matrix cracks lie within
planes that are parallel to the x–y plane. The initiation of delamination is
often attributed to free-edge stresses, which will be further discussed in Sec. 2.
The delaminated regions grow in size as the stress is increased and eventually
coalesce, such that a delaminated region may extend across the entire width of
the specimen. At still higher effective stress levels, matrix cracks begin to form
within the 0j plies (often referred to as ‘‘splitting’’). These cracks lie within a
plane perpendicular to the x–y plane.
The final laminate fracture is precipitated by fiber failures within the 0j
plies. The effective stress level at which final fracture occurs is often called the
last-ply failure stress. At the final fracture, the laminate usually fractures into
many fragments due to the extensive and pre-existing matrix cracks and
* It is interesting to note, however, that a pronounced decrease in the effective Poisson ratio vxy
occurs due to transverse cracking of the 90j plies [4].

delamination that occurred at lower stress levels, as well as the large amount
of energy release associated with fiber failure.
Two other forms of damage encountered in composites are fiber–matrix
debonding and fiber microbuckling. In the case of fiber–matrix debonding, a
crack forms around the periphery of a fiber, so that the load can no longer be
transferred from the matrix to the fiber. In the case of fiber microbuckling, the
fibers within a ply that experiences compressive stresses in the fiber direction
buckle. This reduces the compressive stiffness exhibited by the ply and
ultimately leads to failure of the fibers due to the bending stresses induced.
Whereas the sequence of damage events depicted in Fig. 1 is for the
particular case of a [0/45/90/45]s laminate subjected to uniaxial tensile
loading, this sequence of events is more or less representative of all multiangle
laminates. That is, multiangle laminates subjected to a monotonically increas-
ing but otherwise arbitrary loading condition typically experience yielding
and significant internal damage at load levels far below that required to cause
final fracture. It should now be clear that ‘‘damage’’ refers to many failure
events including matrix cracks, delaminations, fiber–matrix debonding, fiber
microbuckling, fiber cracks, etc. Hence, a fundamental question arises during
the design of composite structures: What is ‘‘failure’’ and what is a ‘‘safe’’
effective stress level? There is no single answer to this question, and the
definition of ‘‘failure’’ as well as selection of a design failure stress often
depends on details of the structural application. A partial list of the factors
involved includes:
The intended service life of the structure. A structure intended for
years of service (such as a highway bridge or an airplane fuselage) will
generally be designed to a much more conservative failure stress level
compared to a structure intended for a single use (such as a rocket
motor case).
The cyclical nature of the applied loading. A structure subjected to
cyclical fatigue loading will generally be designed to a more
conservative failure stress level compared to a structure that will
experience static or slowly varying loads during service.
The consequences of structural failure. A structure whose failure will
result in the loss of life and/or extensive property damage will
generally be designed to a more conservative failure stress level
compared to a structure whose failure is of lesser consequence.
In practice, the maximum allowed effective stress level used during the
design process is also related to our ability to accurately predict the loads
induced by all possible service conditions. This factor is governed by the
complexity of the structure being designed and/or the service environment.
For example, compare the design of a simple pressure vessel intended to store

a pressurized gas (the body of a fire extinguisher, perhaps) to that of a pres-
surized airplane fuselage. In the case of a fire extinguisher, the primary service
loading involved is due to the internal gas pressure, and over most of the
service life of this structure the internal pressure varies slowly, if at all. Hence,
in this case, it is possible to predict the loads/stresses experienced by the
structure during service quite accurately. On the other hand, although a pres-
surized composite airplane fuselage may be (roughly) thought of as a ‘‘pres-
sure vessel,’’ the service loading conditions the structure must withstand are
very complex and difficult to quantify precisely. In this case, the ‘‘pressure
vessel’’ must accommodate not only internal pressures but also direct
aerodynamic loading and loads transferred from the wings and landing gear
to the fuselage. Furthermore, the fuselage will experience distinctly different
loading conditions during takeoff, climb, cruising, descent, landing, and other
aircraft maneuvers. The design of an aircraft fuselage must also accommodate
the need for passenger/cargo doors, windows, wing attachments, etc. Taken
together, these many factors imply that it is very difficult to precisely predict
the loads/stresses induced in a fuselage under all conditions that will be
encountered in service and, consequently, the design of a fuselage is based on
relatively conservative design philosophies.
Again, what is a ‘‘failure’’? Referring to Fig. 1, at least two limiting
design philosophies may be defined. First, the laminate failure stress may be
defined as the effective stress level required to initiate a ply failure of any kind.
‘‘Ply failure’’ is meant to imply either yielding or fiber fractures in one or more
plies. The laminate failure stress or load defined in this manner is called the
first-ply failure stress or first ply failure load. This is a conservative design
approach (in some cases overly conservative) because for most multiangle
laminates, it will be found that first-ply failure is initiated due to yielding of
one or more plies rather than fiber fracture. Most new-generation polymeric
composite material systems are relatively ductile, and ply yielding occurs at
stress levels well below that necessary to cause any matrix cracks or fiber
fractures. Hence, the ultimate load-carrying capacity of most multiangle
laminates is well above the first-ply failure load. Still, a conservative design
philosophy is appropriate when failure of the composite structure will result
in the loss of life and/or extensive property damage; thus, in many instances,
the use of the first-ply failure approach may be appropriate.
The second limiting philosophy is to define the laminate failure stress as
the effective stress level that causes final catastrophic fracture; in this case, the
laminate failure stress is defined as the last-ply failure stress. Obviously, this
latter definition is far less conservative and should be employed with caution
and only under special circumstances. The design of any composite structure
based on last-ply design philosophies should incorporate a generous factor of
safety.

2 FREE-EDGE STRESSES
2.1 The Origins of Free-Edge Stresses
A thin multiangle (and/or multimaterial) composite laminate subjected to a
uniaxial load Nxx is shown in Fig. 2. The thickness and width of the laminate
are denoted t and 2b, respectively. The unloaded edges of the laminate, de-
fined by y = Fb, are called ‘‘free edges’’ because no external forces are applied
to these edges. The phrase ‘‘free-edge stresses’’ (also called ‘‘interlaminar
stresses’’) refers to the stresses induced at and near a free edge. As will be seen,
a highly three-dimensional state of stress is, in general, induced near a free
edge. Furthermore, these free-edge stresses are developed because of a mis-
match in the properties of adjacent plies within a laminate.
Before discussing free-edge stresses directly, consider the ply stresses
that would be predicted on the basis of classical lamination theory. The ma-
terial and/or fiber angle is assumed to vary from one ply to the next for the
laminate shown in Fig. 2. In other words, the effective material properties of
each ply (in particular, the effective Poisson’s ratio vxy and the effective
coefficient of mutual influence of the second kind gxx,xy) varies from one ply to
the next. Due to these variations, the state of stress predicted for any ply based
on a CLT analysis will, in general, include all three in-plane stress compo-
nents. That is, the state of stress that is predicted for ply k on the basis of CLT
generally includes stress components (rxx, ryy, and sxy)k. Hence, even for a
Figure 2 A thin multiangle (and/or multimaterial) composite laminate subjected

to a uniaxial load Nxx.

simple tensile specimen, a uniaxial load in the x-direction (Nxx) is predicted to
cause stresses ryy and sxy, as well as stress rxx.
Referring again to Fig. 2, note that although it may be possible for ryy
and sxy to exist in the central regions of the laminate, neither ryy nor sxy can
possibly exist at the free edge because neither Nyy nor Nxy is applied to the
laminate. Therefore, even if ryy and sxy are induced in the central regions of
the laminate, as y ! Fb, it must be that ryy ! 0 and sxy ! 0. Hence, CLT
predictions must be incorrect, at least for regions near the free edge.
To further investigate these observations, we will consider CLT pre-
dictions for a very simple laminate, where it is easily seen that the stresses
predicted by CLT are reasonable for interior regions of the laminate, but are
invalid near a free edge. This example will also give some physical insights into
why a three-dimensional state of stress exists near a free edge.
Consider a hybrid three-ply [0j]3 laminate subjected to uniaxial tensile
load Nxx (only), as shown in Fig. 3a. The laminate is assumed to consist
of three 0j plies, but the material used in plies 1 and 3 differs from that used
ð1Þ ð2Þ ð1Þ ð2Þ ð1Þ ð2Þ
in ply 2. Furthermore, assume that E11 ¼ E11 ; E22 ¼ E22 ; G12 ¼ G12 , but
ð1Þ ð2Þ
that v12 > v12 . In this hypothetical case then, a mismatch in Poisson’s ratio
exists between plies, but all other material properties are exactly identical. In
addition, the coefficient of mutual influence of the second kind equals zero
for all three laminates because the 1–2 axes are coincident with the x–y axis for
all three plies. We assume that the thickness of each ply is identical (say,
t1 = t2 = t3 = tp), so that the total laminate thickness is t = 3tp.
Based strictly on physical reasoning, it is clear that all plies will
experience a tensile axial stress and strain (rxx and xx) because the overall
loading applied to the laminate (Nxx) has been assumed to be tensile. If the
three plies are firmly bonded together, it is reasonable to assume that the tensile
loading will cause a uniform contraction in the transverse y-direction due to
the Poisson effect. On the other hand, if the three plies were not bonded
together, plies 1 and 3 would contract to a greater extent than ply 2 because it
has been assumed that Poisson’s ratio of the material used in these plies is
ð1Þ ð2Þ
greater than of ply 2: v12 > v12 . Thus, if the plies were not bonded together,
then the magnitude of the compressive transverse strains induced in plies 1
and 3 would be greater than that in ply 2: |eyy|plies 1 and 3>|eyy|ply 2. The
transverse contractions that would occur if the three plies were not bonded
together are shown schematically (and highly exaggerated) in Fig. 3b.
Of course, in reality, the three plies are bonded together, and hence we
assume that all three plies contract by an equal amount: jeyy jplies 1 and 3 ¼ jeyy jply 2 :
Because ply 2 is forced to contract to a greater extent than it would otherwise,
in the bonded case, it is expected that a transverse compressive stress will be
induced in ply 2. On the other hand, plies 1 and 3 contract to a lesser extent

Figure 3 A hybrid three-ply 0j laminate subjected to a uniaxial tensile loading.
It is assumed that plies 1 and 3 have a Poisson ratio that is numerically greater
than ply 2, but all other properties are identical.

that they would otherwise, and hence it is expected that a transverse tensile
stress will be induced in plies 1 and 3.
CLT calculations confirm these expectations. Using the procedures
described in Chap. 6 (and assuming that DT = DM = 0), it can be shown*
that the stresses induced in each ply under these conditions are:
n h io
ð2Þ ð2Þ ð1Þ
3E11 E22 v12 2v12 þ v12 Nxx
rxx jplies 1 and 3 ¼ hn o n oi2 ð1aÞ
ð1Þ ð2Þ
tp 9E11 E22 v12 þ 2 v12
h i
ð1Þ ð2Þ
v12 v12 E22 Nxx
ryy jplies 1 and 3 ¼ hn o n oi2 ð1bÞ
ð1Þ ð2Þ
tp 9E11 E22 v12 þ 2 v12
sxy jplies 1 and 3 ¼0 ð1cÞ

n h io
ð1Þ ð2Þ ð1Þ
3E11 E22 v12 2v12 þ v12 Nxx
rxx jply 2 ¼ hn o n oi2 ð2aÞ
ð1Þ ð2Þ
tp 9E11 E22 v12 þ 2 v12
h i
ð1Þ ð2Þ
2 v12 v12 E22 Nxx
ryy jply 2 ¼ hn o n oi2
ð1Þ ð2Þ
tp 9E11 E22 v12 þ 2 v12
ð2bÞ
¼ 2ryy jplies 1 and 3
sxy jply 2 ¼ 0 ð2cÞ

It is interesting to note that Eqs. (1a) and (2a) show that the axial stress
rxx induced in plies 1 and 3 is not precisely equal to that induced in ply 2
ð1Þ ð2Þ
because v12 p v12 . Of greater immediate interest, however, is that Eqs. (1b)
and (2b) predict that a transverse stress ryy is induced in all three plies, again
ð1Þ ð2Þ
because v12 p v12 . Hence, a load in the x-direction (Nxx) is predicted to cause
a normal stress in the y-direction. The transverse stress induced in ply 2 is
predicted to be compressive (as expected) and with a magnitude twice as high
as the tensile transverse stress induced in plies 1 and 3. These transverse
stresses occur solely because of the mismatch in Poisson’s ratio. That is, if
ð1Þ ð2Þ
v12 ¼ v12 then from Eqs. (1b) and (2b), the transverse stress in all three plies
becomes ryy = 0.
* Because this is a relatively simple laminate, Eqs. (1) and (2) can be confirmed by hand
calculations.

The physical reasoning described above is simple and compelling;
furthermore, the expected algebraic sign of the transverse stresses expected
in each ply is confirmed by CLT calculations, lending credence to the CLT
approach. Nevertheless, as pointed out earlier, a transverse stress ryy cannot
possibly exist at the free edge of the laminate because no transverse load Nyy is
applied to the laminate.
A free-body diagram of an individual ply can now be used to understand
why free-edge stresses develop and are three-dimensional. A cross-section
within the y–z plane of the three-ply laminate and a free-body diagram of a
section removed from ply 3 is shown in Fig. 4. A force Fyy (associated with
stress ryy) acts on the left side of the free-body diagram. However, an
Figure 4 A free-body diagram of a section removed from ply 3 of the hybrid

three-ply laminate shown in Fig. 3.

equilibrating force Fyy cannot be present on the right side of the free-body
diagram because no load is applied to the free edge of the laminate. Therefore,
a shear force Vzy must develop on the interior surface of ply 3 (i.e., the ‘‘lower’’
surface of ply 3, as shown in Fig. 4), so as to maintain force equilibrium in the
y-direction (SFy = 0). The fact that a shear force Vzy is present implies, of
course, that a shear stress syz = szy is present near the free edge. Note, how-
ever, that it must be that Vzy ! 0 as y !Fb because no shear force can exist at
the free edge; Vzy can only be nonzero at interior regions of ply 3. Also, the
normal force acting on the left side Fyy is acting through the centroid of the
free-body diagram, whereas the shear force on the right side Vzy is present on
the lower surface. These two forces therefore induce a bending moment about
the x-axis because they are not colinear. The only way that this bending
moment can be reacted so as to satisfy moment equilibrium (i.e., so as to
maintain SMx = 0) is if a force acting normal to the x–y plane ( Fzz) also
develops. The fact that a normal force Fzz is present implies, of course, that a
normal stress rzz is present. Numerical and analytical solutions (discussed
below) indicate that Fzz (and hence rzz) reaches a maximum value at the free
edge ( y =Fb), as shown schematically in Fig. 4. Because the only external
force applied to the laminate as a whole is Nxx, internal force Fzz (or, equiva-
lently, stress rzz) must be self-equilibrating: mby ¼ 0 rzz dxdy ¼ 0. This implies
that rzz must undergo a change in algebraic sign at regions near the free edge.
If, for example, rzz is tensile at y = b, it must become compressive at some
distance from the free edge so as to maintain static equilibrium.
This simple example illustrates that free-edge stresses rzz and szy are
caused by a uniaxial tensile loading Nxx. The magnitude and distribution of
rzz and szy cannot be determined solely on the basis of the equations of
equilibrium. Nevertheless, the fact that they must exist can be appreciated
through consideration of the free-body diagram shown in Fig. 4. For the
simple hybrid [0j]3 laminate considered, the coefficient of mutual influence of
the second kind gxx,xy=0, and therefore according to CLT, a uniaxial load
Nxx does not cause an in-plane shear stress (sxy) to be induced at interior
regions in any ply. However, for more general laminates, gxx,xy p 0, and a
uniaxial loading Nxx will cause both ryy and sxy at interior regions of the
laminate. In these cases, the state of stress near a free edge involves all six
components of stress (rxx, ryy, rzz, syz, sxz, and sxy). Once again, it can be said
that free-edge stresses develop because of a mismatch in the properties of
adjacent plies within a laminate.
2.2 Numerical and Analytical Studies of Free-Edge Stresses

Free-edge stresses have a profound influence on the delamination failure of
multiangle composite laminates, and have therefore been the topic of a great

many research papers. The existence of free-edge strains in multiangle
composite laminates has been well documented experimentally [5–8]. Meth-
ods used to predict free-edge stresses may be roughly grouped into numerical
solutions [8–14] and approximate analytical solutions [15–21]. With the
exception of Ref. 9 (summarized below), all of the numerical solutions
referenced here were based on the finite-element method.
A detailed discussion of the numerical methods and/or approximate
analytical tools that have been used to study free-edge stresses is beyond the
scope of this book. The results of these analyses will simply be summarized,
and the interested reader is referred to the original references if additional
details are desired. The studies that have led to our current understanding of
free-edge stresses will be discussed more or less in chronological order.
Pipes and Pagano [10] presented the first numerical study of free-edge
stresses in 1970. Their analysis was based on the method of finite differences.
They investigated free-edge stresses induced in four-ply symmetrical angle-ply
laminates having the general stacking sequence [h/h]s and subjected to a
remote uniform axial extension, which implies that, at remote locations, the
laminate is subjected to a uniform axial tensile strain exx. The following
properties, which were typical of transversely isotropic graphite–epoxy
material systems at the time, were assumed in their study:
E11 ¼ 138 GPa ð20 MsiÞ

E22 ¼ E33 ¼ 14:5 GPa ð2:1 MsiÞ
G12 ¼ G23 ¼ G13 ¼ 5:86 GPa ð0:85 MsiÞ
v12 ¼ v23 ¼ v31 ¼ 0:21
The thickness of each ply was denoted ho. In this text, the total laminate
thickness has been denoted t; thus, in the Pipes and Pagano study, the total
laminate thickness was t = 4ho. The width of the laminate was denoted 2b.
Laminates with the three different width-to-laminate thickness ratios of 2, 4,
and 6 were studied (i.e., laminates were studied with width-to-ply thickness
ratios of b/h = 4, 8, and 12, respectively). Free-edge stresses were found to be
independent of width-to-laminate thickness ratio over the range considered,
and most of the results presented in Ref. 9 are for laminates with b/ho = 8 (i.e.,
for laminates whose width is four times as great as their thickness).
A summary of ply stresses calculated by Pipes and Pagano for a [45/
45]s laminate, at z = hþo (i.e., just within the +45j ply), is presented in Fig. 5.
In this figure, the ply stresses are plotted vs. normalized position y/b within
the laminate, where the position y/b = 0 corresponds to the centerline of the
laminate and y/b = 1.0 corresponds to the free edge (see Fig. 2). Using the
material properties listed above, stresses predicted by a CLT analysis for

Figure 5 Ply stresses calculated for a graphite–epoxy [45/45]s laminate, at
the through-thickness position z = h+ 0 . (From Ref. 9, with permission from
Sage Publications Ltd.)
the 45j plies in a [45/45]s laminate subjected to a uniform axial strain exx are
as follows:
rxx
¼ 20:4 GPa ð2:96 MsiÞ
exx
ryy ¼ 0
sxy
¼ 7:93 GPa ð1:15 MsiÞ
exx

Because CLT is based on the assumption of plane stress, for a CLT
calculation, the remaining stress components are zero by assumption:
rzz=sxz=syz=0. Referring to Fig. 5, at interior regions of the laminate
(say, for y/b<0.5), the Pipes and Pagano results are in excellent agreement
with CLT predictions. However, at regions near the free edge (for y/b>0.5),
the state of stress becomes three-dimensional and finite-difference results do
not agree with CLT calculations. The distribution of each of the six possible
components of stress as predicted by the Pipes and Pagano analysis may be
summarized as follows:
As the free edge is approached, the normalized axial stress rxx/exx is
predicted to increase very slightly from the CLT value, reaching a
maximum value of about 21.0 GPa (3.05 Msi) at y/bc0.7 It then
decreases to a finite value of about 15.9 GPa (2.3 Msi) at the free edge.
A normalized transverse stress ryy/exx is predicted to develop at y/
bc0.5, reaching a maximum value of about 1.3 GPa (0.18 Msi) at y/
bc0.85. It then decreases back to zero at the free edge as it must to
satisfy equilibrium at that point.
A normalized out-of-plane stress rzz/exx is predicted to occur in
regions very near the free edge; rzz/exx is essentially zero for y/
b<0.70. As y/b ! 1.0, rzz/exx is initially compressive and then
becomes tensile, reaching a maximum tensile value of about 2.8 GPa
(0.40 Msi) at the free edge.
As the free edge is approached, the predicted normalized in-plane
shear stress sxy/exx increases very slightly from the value predicted by
CLT, reaching a maximum value of about 8.41 GPa (1.22 Msi) at y/
bc0.7 It then decreases to zero at the free edge as it must to satisfy
equilibrium at that point.
The predicted normalized out-of-plane shear stress sxz/ezz is
essentially zero for y/b<0.5. However, it increases very rapidly as
y/b!1.0 and, in fact, Pipes and Pagano speculated that sxz is singular
at the free edge, although they could not prove this conclusively due
to the numerical nature of their study. As will be further discussed
below, subsequent analytical studies have shown that free-edge
stresses are in fact singular at the free edge.
The predicted normalized out-of-plane shear stress syz/ezz is almost
nonexistent throughout the entire width of the laminate. Values of
syz/ezz less than 0.6 GPa (0.09 Msi) were predicted for y/b>0.75, but
then decrease to zero at the free edge (these predictions are barely
discernable in Fig. 5).
It should be noted that these are not general results. Rather, these
predicted stress distributions are for the specific case of a [45/45]s graphite–

epoxy laminate subjected to a remote uniform axial strain exx, based on the
ply properties listed above. Different stress distributions would be predicted if
a different stacking sequence were involved, if the laminate were subject to
different loading conditions, and/or if different material properties were used.
Nevertheless, the pioneering study of Pipes and Pagano showed that a three-
dimensional state of stress generally exists within a narrow region near an
unloaded edge of a composite laminate. The width of the region over which
appreciable free-edge stresses develop was estimated to be equal to or less than
the total thickness of the laminate t. The free-edge shear stress sxz was of
particular concern because the numerical analysis indicated that this stress
component might be singular near the free edge.
The computing power, processing speed, and availability of digital
computers were rapidly increasing at about the time the Pipes and Pagano
analysis appeared in the literature. The advent of powerful digital computers
led to rapid advances in numerical structural analysis methods, particularly in
the finite-element technique. Several numerical analyses devoted to free-edge
stresses and based on the use of the finite-element technique appeared in the
literature following the Pipes and Pagano study (Refs. 11–13) are representa-
tive examples. Although predictions based on finite-element methods were in
general agreement with the original finite-difference study by Pipes and
Pagano, there was growing consensus that a singular stress field exists at
the free edge [16–18]. That is, as finite-element analyses became more refined,
allowing greater mesh densities to be used to study the free-edge problem, the
magnitude of predicted free-edge stresses became larger and larger. The fact
that individual stress components (in particular, the values of rzz and sxz) did
not converge to a stable value as mesh densities were increased near the free
edge is indicative of a singular stress field, although this conclusion could not
be proven rigorously on the basis of numerical studies alone.
Wang and Choi [20,21] and Zwiers et al. [22] eventually confirmed the
singular nature of free-edge stress fields on the basis of analytical studies. The
approach taken by Wang and Choi [20,21] is based on the use of complex
variable stress potentials, which is an analytical approach originated primar-
ily by Lekhnitski. In contrast, the approach taken by Zwiers et al. [22] is based
on the use of complex variable displacement potentials, which is an analytical
approach originated primarily by Stroh. Although either approach may be
taken, most analytical studies of free-edge stresses (e.g., Refs. 19,20,23–25)
have been based on the use of Lekhnitski’s complex stress potentials. These
analytical studies have shown conclusively that stresses induced near a free
edge are singular in nature and can be written in general form as:
f ðy; zÞ
rij ¼ fðy; zÞrd ¼ ð3Þ
rd

where f ( y,z) represents a generalized function that depends on laminate
stacking sequence, material properties, and applied loading, and r represents
the distance from the intersection between the interface between two plies and
the free edge. It can be shown [16] that the exponent d depends on both
material properties as well as the difference in fiber angles between adjacent
plies in the laminate. Also, d must be a positive number bounded by 0<d<1.
Hence, as the free edge is approached (i.e., as r!0), the fraction 1/rd!l and,
in accordance with Eq. (3), the stresses rij are predicted to be infinitely high.
That is, free-edge stresses are singular.
Exponent d is called the ‘‘strength’’ of the singularity. Because d
depends on both material properties and the difference in fiber angles between
adjacent plies in a laminate, a universal value for d does not exist. Rather, d
must be determined for each case of interest. Wang and Choi [20] calculated d
for [Fh]s laminates, using the material properties used in the original Pipes
and Pagano finite-difference analysis (listed above). A plot of d as a function
of fiber angle h is presented in Fig. 6. For these conditions, d is at a maximum
for hc51j, whereas d = 0 if h = 0j or 90j. This latter result is as expected
because if h = 0j or 90j, the two adjacent plies have identical fiber angles and
therefore there is no mismatch in ply material properties.
Wang and Choi recalculated free-edge stresses for the same laminate
and loading conditions considered by Pipes and Pagano, but included the
stress singularity represented by Eq. (3) in their solution. They found that
inclusion of the stress singularity has little impact on the distribution of sxz
near the free edge, but substantially alters the predicted distribution of the
out-of-plane normal stress rzz. In fact, in the original Pipes and Pagano finite-
difference solution, rzz was predicted to be tensile at the free edge, whereas in
the Wang and Choi analysis, rzz was predicted to be compressive. In both
analyses, the shear stress sxz was considered to be the ‘‘dominant’’ free-edge
stress because it exists over a relatively wider area and is singular at the free
edge. Although the magnitude of rzz is generally less than sxz (except at the
free edge, where both are singular) and exists over a narrower region, rzz is
believed to play a central role in delamination failure. If rzz is tensile at the free
edge, it will tend to separate plies, contributing to initial delamination and
subsequent growth of the delaminated area. In contrast, if rzz is compressive,
it will tend to hold plies together, inhibiting delamination.
Once the singular nature of free-edge stresses had been clarified on the
basis of analytical studies, it became possible to define special ‘‘singular
elements’’ for use in finite-element studies of free-edge stresses. The stress
singularity represented by Eq. (3) is embedded within the shape function used
to define these singular elements, following an approach suggested by Stern
[27]. For example, Nailadi et al. [15] used singular elements in a finite-element
study of free-edge stresses, and compared their predictions to displacements

Figure 6 Strength of the free-edge stress singularity in [Fh]s graphite–epoxy
laminates. (From Ref. 20, with permission from ASME.)
measured at the free-edge using Moiré interferometry. These authors also

considered a [45/45]s graphite–epoxy laminate subjected to axial tensile
strain, although the material properties used in this study differed slightly
from those in the earlier analyses of Pipes and Pagano and Wang and Choi.
They concluded that stress component rzz is compressive at the free edge, in
agreement with the Wang and Choi analysis. Hence, it appears that the
predicted distribution of the out-of-plane normal stress rzz is particularly
sensitive to inclusion of the stress singularity in the problem formulation.

Three important points should be emphasized. First, the distribution of
free-edge stresses depends on material property and stacking sequence.
Because the number of different combinations of materials and stacking
sequences is infinitely large, it is not possible to tabulate solutions for free-
edge stresses for all laminates that may be encountered in practice; a new
analysis is required for each new combination. Second, most analyses that
have appeared in the literature have focused on free-edge stresses caused in
rectangular test specimens subjected to uniaxial tensile loading. In practice,
composite structures are required to support combinations of tensile, com-
pressive, and shear loading, and free-edge stresses may occur at bolt holes,
cutouts, or other geometrical discontinuities. Therefore, results of studies that
have appeared in the literature may not be directly applicable during the
design of a composite structure, even if the same material and stacking
sequence are involved. Finally, as discussed by Herakovich et al. [9], the
predicted singular nature of free-edge stresses is, in reality, an artifact of
available structural analysis methods. That is, all of the analyses described
above are based on the assumption of linear–elastic material behavior and,
furthermore, all analyses assume that a distinct interface exists between plies.
Neither of these assumptions is rigorously valid. A distinct interface does not
exist in real composite laminates. Although it is a certainty that high stress
levels and stress gradients are induced near a free edge, in reality, high stress
levels cause a nonlinear material response, and therefore the magnitude of
free-edge stresses is large but finite. This observation is, of course, borne out
by experiment because if free-edge stresses were truly singular, then a
composite laminate would delaminate immediately upon application of any
external load, regardless of the magnitude of the load.
3 PREDICTING LAMINATE FAILURE USING CLT

The many factors that must be considered in order to predict the failure of a
multiangle composite laminate have been discussed in Secs. 1 and 2. Factors
that must be considered at any load level include the existence of thermal and/
or moisture stresses, time-dependent material behavior, and singular (in a
mathematical sense) free-edge stresses. As loads are applied and increased,
nonlinear behavior (‘‘yielding’’) eventually occurs in one or more plies. As
loads are further increased, a multitude of distinct damage ‘‘events’’ (matrix
cracking, delaminations, fiber microbuckling, etc.) occurs. It is the integrated
effect of all of these factors that governs the final fracture of a composite
laminate. Because of these complications, the prediction of fracture of
polymeric composites has proven to be a formidable challenge, and methods
of predicting fracture under general thermomechanical loading conditions
remain an active area of research.

Nevertheless, the practicing engineer requires some method of estimat-
ing the load-carrying capacity of a composite laminate for purposes of
preliminary design. Toward that end, classical lamination theory can be used
to estimate when failure of a multiangle laminate will occur. In essence, CLT
is used to predict the ply stresses and/or strains that will be induced by a
thermomechanical loading of interest. Once ply stresses/strains are known,
one of the macromechanics-based failure criteria developed in Sec. 5 of Chap.
5 is used to predict failure of individual plies, ultimately leading to a failure
prediction for the laminate as a whole. CLT can be used to predict the first-ply
failure of a multiangle laminate with a reasonable degree of accuracy. Recall
that, in this text, first-ply failure is meant to imply the initiation of either
yielding or fiber fracture in one or more plies. This philosophy results in a
conservative estimate of the load-carrying capacity of a composite laminate
because (in general) a multiangle laminate can withstand a considerable
increase in loading beyond first-ply failure. It is also possible to estimate
last-ply failure on the basis of a CLT analysis, although with a substantial
decrease in accuracy. Methods of predicting first-ply and last-ply failure using
CLT are discussed in Secs. 3.1 and 3.2.
3.1 Predicting First-Ply Failure

According to the first-ply failure design philosophy, laminate failure is
considered to occur when a given combination of loading, temperature
changes, and/or moisture changes causes nonlinear behavior to develop in
any ply. The first ply failure stress for the specific case of a [0/45/90/45]s
laminate subjected to a uniaxial tensile loading was discussed in Sec. 1 and
illustrated schematically in Fig. 1. This same concept can be extended to
laminates with different stacking sequences and/or to more general thermo-
mechanical loading conditions. Although several different solution paths may
be followed, one approach is illustrated in Fig. 7. The flow diagram shown in
this figure is based on the assumption that the laminate of interest is subjected
to a uniform change in temperature and/or moisture content as well as some
combination of uniform external loads (Nxx, Nyy, Nxy, Mxx, Myy, Mxy). It is
further assumed that the relative magnitude of each load component is
known. That is, if Nxx=1, then it is assumed that the magnitudes of the
remaining loads can be related to Nxx (i.e., Nyy=k1|Nxx|, Nxy=k2|Nxx|,
Mxx=k3|Nxx|, Myy=k4|Nxx|, Mxy=k5|Nxx|). The relative magnitudes of each
load component (i.e., constants ki) are referred to as ‘‘unit loads’’ in Fig. 7.
The calculation steps shown in Fig. 7 can be summarized as follows:
Step 1. The first step is to define the problem. Ply properties,
laminate stacking sequence, unit loads applied to the laminate, stress-
free temperature (often assumed to be the cure temperature), service

Figure 7 A flow diagram illustrating calculation of the first-ply failure load
using CLT.

temperature, moisture content, and the failure criterion to be
employed must all be specified.
Step 2. The [ABD] matrix for the laminate is determined. The re-
duced stiffness matrix Qij for each material used in the laminate is
calculated (using Eq. (11) of Chap. 5), and then the transformed
reduced stiffness matrix Qij for each ply is determined (using Eq. (31)
of Chap. 5). This allows calculation of the Aij, Bij, and Dij matrices,
using Eqs. (20a), (20b), and (27) of Chap. 6, respectively, which are
then assembled to form the [ABD] matrix.
Step 3. Thermal and moisture stress and moment resultants cor-
responding to the specified DT and DM are calculated using Eqs. (25)
and (28) of Chap. 5, and Eqs. (34) and (35) of Chap. 6.
Step 4. Ply stresses and strains in the global x–y coordinate system
are determined. This involves calculation of the midplane strains
and curvatures using Eq. (37) of Chap. 6, calculation of the ply
strains in the x–y coordinate system using Eq. (12) of Chap. 6, and
calculation of the ply stresses in the x–y coordinate system using Eq.
(30) of Chap. 5.
Step 5. If a stress-based failure criterion has been selected for use
(such as the maximum stress, Tsai–Hill, or Tsai–Wu failure cri-
terion), then ply stresses are rotated to the local 1–2 coordinate
system defined by each ply angle using Eq. (20) of Chap. 2. Alter-
natively, if a strain-based failure criterion has been selected for use
(strain-based failure criteria have not been discussed in this text), the
ply strains are rotated to the local 1–2 coordinate system using Eq.
(44) of Chap. 2.
Steps 6 and 7. The failure criterion is applied to each ply, and the ply
‘‘nearest’’ to failure is identified. The unit loads are then increased,
and steps 4 and 5 are repeated iteratively, until the first ply failure
condition is reached. Details of the iteration process depend on the
failure criterion selected for use.
Step 8. Results are outputted once the first-ply failure condition has
been identified.
The use of CLT to predict first-ply failure according to the flow diagram
shown in Fig. 7 is illustrated in the following Example Problem.
Example Problem 1
Predict the first-ply failure load for a [0/30/60]s graphite–epoxy laminate
subjected to a uniaxial tensile load Nxx as well as a change in temperature
corresponding to cooldown from cure temperatures (175jC) to room temper-
atures (20jC). Assume no change in moisture content, and that the laminate is

stress-free at the cure temperature.* Base the prediction on the maximum
stress failure criterion, use properties for graphite–epoxy listed in Table 3 of
Chap. 3, and assume that each ply has a thickness of 0.125 mm.
Solution. The problem statement implies that:
Nyy ¼ Nxy ¼ Mxx ¼ Myy ¼ Mxy ¼ DM ¼ 0
The following properties for graphite–epoxy are taken from Table 3 of
Chap. 3:
E11 ¼ 170 GPa E22 ¼ 10 GPa v12 ¼ 0:30 G12 ¼ 13 GPa
rf11T ¼ 1500 MPa ryT
22 ¼ 50 MPa sy12 ¼ 75 MPa
11 ¼ 1200 MPa
rfC ryC
22 ¼ 100 MPa
a11 ¼ 0:9 Am=m C a22 ¼ 27 Am=m B C
B
Employing CLT calculations discussed in previous chapters, a change in

temperature of DT = 155jC (alone) will cause the following stresses in each
ply:
For the 0j plies (plies 1 and 6), a DT = 155jC will cause:
r11 ¼ 55:54 MPa r22 ¼ 28:36 MPa s12 ¼ 22:83 MPa
For the 30j plies (plies 2 and 5), a DT=155jC will cause:
r11 ¼ 29:59 MPa r22 ¼ 24:79 MPa s12 ¼ 0 MPa
For the 60j plies (plies 3 and 4), a DT=155jC will cause:
r11 ¼ 55:54 MPa r22 ¼ 28:36 MPa s12 ¼ 22:83 MPa
The additional stresses caused by the external load Nxx are superimposed with
the initial thermal strains. A CLT analysis gives the following predicted
stresses:
For the 0j plies (plies 1 and 6):
r11 ¼ ð2:750 103 ÞNxx 55:54 MPa
r22 ¼ ð0:5193 102 ÞNxx þ 28:36 MPa
s12 ¼ ð0:1748 103 ÞNxx þ 22:83 MPa
r11 ¼ ð1:112 103 ÞNxx þ 29:59 MPa
r22 ¼ ð0:1206 103 ÞNxx þ 24:79 MPa
s12 ¼ ð0:2646 103 ÞNxx
* As discussed in Sec. 5 of Chap. 6, the stress-free temperature is likely to be lower than the cure
temperature. For simplicity, it is assumed in this text that the final cure temperature defines the
stress-free temperature.

r11 ¼ ð0:2098 103 ÞNxx 55:54 MPa
r22 ¼ ð0:1760 103 ÞNxx þ 28:36 MPa
s12 ¼ ð0:8985 102 ÞNxx 22:83 MPa
These results can be used to determine the stresses induced by the
combined effects of DT = 155jC and any value of Nxx, assuming a linear–
elastic response. For example, if Nxx = 100 kN/m, then the stress induced in
the fiber direction in plies 1 and 6 is: r11 = (2.750103)(100103)55.54
MPa=219.5 MPa. Because the maximum stress failure criterion is used in
this sample problem, the value of Nxx necessary to cause stresses to reach
critical levels in any ply can be determined directly using the specified failure
stresses:
1500 106 þ 55:54 106
xx ¼
For fiber failure : N fail ¼ 565:7 103 N=m
2:75 103
50 106 28:36 106
xx ¼
For matrix failure : N fail ¼ 416:7 103 N=m
0:5193 102
75 106 22:83 106
xx ¼
For shear failure : N fail ¼ 559:7 103 N=m
0:1748 103
1500 106 29:59 106
xx ¼
For fi ber failure : N fail ¼ 1322 103 N=m
1:112 103
50 106 24:79 106
xx ¼
0:1206 103
75 106
xx ¼
0:2646 103
1200 106 þ 55:54 106

xx ¼
For fi ber failure : N fail ¼ 5455 103 N=m
0:2098 103
50 106 28:36 106

xx ¼
0:1760 103
75 106 þ 22:83 106

xx ¼
0:8985 102

The first-ply failure load equals the lowest value of Nxx that causes the stress in
any ply to reach a critical level. Examination of these results shows that first-
ply failure is predicted to occur when Nxx=123.0 kN/m due to a matrix
yielding failure in the 60j plies (plies 3 and 4). Equivalently, the effective stress
at first-ply failure is:
Nxx 123:0 kN=m
rxx ¼ ¼ ¼ 164 MPac24 ksi:
t 6ð0:125 mmÞ
3.2 Predicting Last-Ply Failure

Last-ply failure calculated using CLT is based on the so-called ‘‘ply-discount’’
scheme. A flow diagram showing typical steps in the analysis is presented in
Fig. 8. The process begins with calculations similar to those described for a
first-ply failure analysis. The first step is to define the problem to be considered
(e.g., specify materials properties, laminate description, unit loads, tempera-
ture and/or moisture changes, and the ply failure criterion to be applied).
Standard CLT calculations are then performed to determine the stresses and
strains induced in all plies. Unit loads are increased until first-ply failure is
predicted, based on the ply failure criterion selected for use. Once a ply is pre-
dicted to have failed, the elastic properties of that ply are ‘‘discounted’’ (i.e.,
changed). The change in elastic properties represents the fact that the ply has
either undergone yielding or has fractured. An unresolved question is: In what
manner should the change in ply properties be modeled? The traditional
approach is based on the assumption that the ply has suffered a matrix crack;
therefore, the stiffnesses of a failed ply are reduced in a discrete step. As an
example of typical values, the stiffnesses of a failed ply could be related to the
intact values according to:
Efailed
11 ¼ E11
vfailed
12 ¼ v12
Efailed
22 ¼ 0:3E22
Gfailed
12 ¼ 0:3G12
failed
In this example, the fiber-dominated properties of the failed ply (E11 and
failed
v12 ) are not reduced at all from those of an intact ply. In contrast, the
failed failed
matrix-dominated properties (E22 and G12 ) are reduced to 30% of those
of an intact ply. These estimates reflect the fact that matrix fractures occur
well before fiber fractures and, in fact, a significant number of fiber fractures
do not ordinarily occur until the last-ply failure load is reached. Matrix-
dominated stiffnesses are not reduced to zero because even a cracked ply will
contribute to some extent to the overall laminate stiffness.
Although a discrete reduction in ply stiffness has been the traditional
approach, it is by no means clear that this technique can be used to model the

Figure 8 A flow diagram illustrating calculation of the last-ply failure load using
CLT.

nonlinear behavior of composites based on new and toughened matrix
materials. As discussed in Chap. 3, prior to about 1980, most commercially
available high-performance composites were based on relatively brittle
polymer matrices. In that case, the assumption that plies fracture before
any significant nonlinear (i.e., yielding) behavior occurs may be appropriate.
However, the ductility of currently available composites has been greatly
enhanced since the 1980s, which means that most systems exhibit a pro-
nounced nonlinear response prior to fracture. In this context then, first-ply
failure is defined as the load at which ply yielding first occurs, and the stiffness
of yielded plies should be reduced in a smooth manner to represent this
phenomenon, rather than by a discrete reduction in ply stiffness.
In any event, once a ply or plies is predicted to fail and ply properties are
reduced accordingly, all CLT calculations are repeated as indicated in Fig. 8.
Note that the [ABD] matrix must be recalculated because the properties of one
or more failed plies have been changed. This process is simply repeated until
all plies in the laminate are predicted to have failed, at which point the last-ply
failure load has been reached.
Although last-ply failure analyses have been described routinely in the
composites literature, the reader is cautioned that such analyses may be
significantly in error. As already discussed, significant errors in prediction
may occur at load levels beyond the first-ply failure load because the ductile
nature of modern composite material systems is often not properly modeled.
A second factor is that as loading is increased, an extensive number of damage
events (matrix cracks, delaminations, etc.) eventually occur within the
laminate, even for those based on modern toughened matrices, as previously
shown in Fig. 1. As damage extends throughout the laminate, the Kirchhoff
hypothesis becomes less and less valid. That is, once extensive matrix cracking
and/or delaminations have occurred, it is unreasonable to assume that a
straight line originally normal to the laminate midplane remains straight and
normal, thus invalidating the Kirchhoff hypothesis upon which CLT is based.
Once significant damage has occurred, it may not be possible to relate ply
stresses and strains to stress and moment resultants using the laminate [ABD]
and [abd] matrices.
The reader is likely to encounter a last-ply failure analysis during future
studies of composite materials and/or structures. The above comments have
been included here to urge the reader to view such analyses with a healthy level
of skepticism.
4 LAMINATE FIRST-PLY FAILURE ENVELOPES

Methods used to determine the first-ply failure load for a laminate subjected to
a specific single combination of stress and moment resultants were discussed

in Sec. 1 of Chap. 3. Obviously, there are an infinite number of combinations
of stress and moment resultants that will cause first-ply failure. Conceptually,
those combinations of stress and moment resultants that (collectively) cause
first-ply failure define a laminate failure surface. It is difficult to visualize such
a surface, however, because there are a total of six stress and moment
resultants involved and hence the failure surface involves six ‘‘dimensions.’’
However, a plane that intersects the failure surface can be visualized by
considering only two of the six stress resultants. For example, combinations of
Nxx and Nyy that cause first-ply failure can be calculated using the approach
described in Sec. 1 of Chap. 3, while assuming that the remaining four
resultants are zero (Nyy=Mxx=Myy=Mxy=0). A plot of the (Nxx,Nyy)
combinations calculated under these assumptions is called a first-ply failure
envelope. A typical first-ply failure envelope for a [0/45/90/45]s graphite–
epoxy laminate based on the maximum stress failure criterion is shown in
Fig. 9. Three curves are shown in the figure. In one case, thermal and moisture
stresses and strains are ignored (i.e., this curve was generated by assuming that
DT = DM = 0). In a second case, the thermal stresses/strains caused during
Figure 9 First-ply failure envelope for a [0/45/90/45]s graphite–epoxy

laminate based on the maximum stress failure criterion and combinations of
Nxx and Nyy (Nxy = Mxx = Myy = Mxy = 0).

cooldown are included in the analysis (this curve was generated by assuming
that DT=155jC, DM=0). This curve illustrates the deleterious effects of
thermal stresses and strains on first-ply failure loads, and reinforces the fact
that thermal stresses and strains should always be considered during design of
a composite structure, especially during consideration of predicted first-ply
failure loads. Finally, the third curve includes both thermal effects (based on
DT=155jC) as well as a change in moisture content of 1.0% (i.e., DM=1.0).
As mentioned earlier, an increase in moisture content often tends to relieve
pre-existing thermal stresses. This trend is evident in Fig. 9 because first-ply
failure loads for the analysis that include both thermal and moisture effects are
slightly higher than the analysis that includes only thermal effects.
It should be kept in mind that the first-ply failure load is a conservative
design philosophy, and that the failure surfaces shown in Fig. 9 do not
represent load combinations that cause laminate fracture. Rather, these
curves represent the loads that cause the first instance ply failure on any kind,
which in most cases is the onset of yielding in one or more plies.
Comparable two-dimensional failure envelopes can be developed based
on any two of the six stress and moment resultants. For example, failure
envelopes for a [0/45/90/45]s graphite–epoxy laminate based on any combi-
nation of (Nxx,Nxy), (Nxx,Mxx), and (Mxx, Myy) are shown in Figures 10–12,
respectively. As before, three curves are shown and are based on the
maximum stress failure criterion: one in which temperature and moisture
effects are neglected, one in which thermal stresses/strains caused during
cooldown of the laminate are included, and a third in which both thermal and
moisture effects are included.
Failure envelopes of distinctly different appearances may be obtained if
a different failure criterion is used. For example, failure envelopes obtained
using the maximum stress, Tsai–Hill, and Tsai–Wu failure criteria for
combinations of Mxx and Myy are compared in Fig. 13. For this loading
condition, the Tsai–Wu criterion gives the most conservative first-ply failure
envelope, whereas the maximum stress criterion gives the least conservative
failure envelope.* Unfortunately, the only way to determine which failure
envelope best represents the failure response of a composite laminate is to
compare predictions to experimental measurement. In general, for composite
laminates based on ductile polymeric matrices, it would be expected that the
Tsai–Hill or Tsai–Wu criterion would provide the best prediction, whereas for
brittle composites the maximum stress criterion may be best.
* The failure envelope based on the Tsai–Wu criterion was generated by calculating the
coupling term X12 using Eq. (61) of Chap. 5. A significantly different failure envelope may be
obtained if X12 were determined on the basis of a biaxial test or by some other means (see Sec.
5.3 of Chap. 5).

Figure 10 First-ply failure envelope for a [0/45/90/45]s graphite–epoxy lam-
inate based on the maximum stress failure criterion and combinations of Nxx
and Nxy (Nyy=Mxx=Myy=Mxy=0).
5 COMPUTER PROGRAM LAMFAIL

The computer program LAMFAIL has been developed to supplement the
material presented in Secs. 3 and 4 of this chapter. This program can also be
downloaded at no cost from the following website: http://depts.washington.
edu/amtas/computer.html.
Program LAMFAIL can be used to either:
Perform a first-ply failure analysis, as described in Sec. 1 of Chap. 3
and illustrated in Fig. 7; or
Generate first-ply failure envelopes as discussed in Sec. 4 and
illustrated in Figs 9–13.
The program prompts the user to input all information necessary to
perform the analysis. Properties of up to five different materials may be
defined. As previously discussed, an implicit assumption is that moduli and
failure strengths inputted by the user correspond to the values exhibited by the
composite at the temperature and moisture content of interest. Also, in the
present context, ‘‘failure’’ may represent fracture of the fibers (rf11T ; rf11C ) or

laminate based on the maximum stress failure criterion and combinations of Nxx
and Mxx (Nyy = Nxy = Myy = Mxy = 0).
yielding of the matrix (ryT yC y

22 ; r22 ; s12 ). The user specifies the number of plies
within the laminate and material type and fiber angle for each ply. The user
also selects from the maximum stress, Tsai–Hill, or Tsai–Wu failure crite-
rion.*
If a first-ply failure analysis for a particular combination of stress and
moment resultants is to be performed, the program prompts the user for ‘‘unit
loads’’ (see Fig. 7). Alternatively, if the program is to be used to generate a
failure envelope, then the particular pair of resultants to be considered during
the analysis is specified by the user (i.e., any two of the six stress resultants
Nxx, Nyy, Nxy, Mxx, Myy, Mxy). Note that program LAMFAIL itself does not
create a failure envelope. Rather, the program creates a file (named Envel-
op.txt) that contains the stress resultant pairs predicted to cause first-ply
failure of a composite laminate specified by the user. A failure envelope may
* If the Tsai–Wu criterion is selected, then the coupling strength term X12 is calculated ac-
cording to Eq. (61) of Chap. 5.

laminate based on combinations of Mxx and Myy (Nxx = Nyy = Nxy = Mxy = 0).
then be created using a second software package to ‘‘import’’ the data

generated by program LAMFAIL. The first few lines of the file Envelop.txt
created during a typical analysis are shown in Table 1. In this case, the analysis
was for a [0/45/90/45]s graphite–expoxy laminate, using properties listed in
Table 3 of Chap. 3. The program had been directed to apply the maximum
stress failure criterion, and to base the analysis on stress resultant pair Nxx
and Nyy. The first few lines of the file represent header information. Notice
that the file contains the ply number predicted to fail and the failure type (i.e.,
fiber, matrix, or shear), in addition to the combinations of Nxx and Nyy
predicted to cause failure. Table 1 shows 12 such combinations of Nxx and
Nyy. The complete file Envelop.txt contains a total of 2001 pairs. Thus, a first-
ply failure envelope may be created by importing and plotting the Nxx and Nyy
pairs.
File Envelop.txt differs slightly if either the Tsai–Hill or Tsai–Wu
criterion is selected for use. Specifically, if either of these criteria is selected,
then file Envelop.txt will not contain a column with the heading ‘‘failure type’’
because these criteria do not distinguish a particular failure mode in the same
sense as does the maximum stress failure criterion. The few 25 lines of the file

Figure 13 First-ply failure envelope for a [0/45/90/45]s graphite–epoxy lam-
inate for combinations of Mxx and Myy, based on the maximum stress, Tsai–Hill,
and Tsai–Wu criteria (Nxx = Nyy = Nxy = Mxy = 0, DT = 155jC, DM= 1%). Fig-
ure 13 A [0/F30]s laminate loaded in biaxial tension, as described in Problem 1.
Envelop.txt created during an analysis for a [0/45/90/45]s graphite–epoxy

laminate in which the Tsai–Hill criterion was employed are shown in Table 2.
HOMEWORK PROBLEMS
Computer programs CLT and LAMFAIL are used in the following prob-
lems. These programs can be downloaded from the following website: http://
depts.washington.edu/amtas/computer.html.
1. Assume that a new room temperature cure graphite–epoxy prepreg

system with properties listed in Table 3 of Chap. 3 has been developed. A
[0/F30]s laminate is produced using this material, and is therefore
initially stress-free and strain-free at room temperature. The room
temperature laminate is then subjected to the biaxial tensile loads shown
in Fig. 14 (i.e., Nyy = Nxx/2 and Nxy = Mxx = Myy = Mxy = 0). Use the

Table 1 First Few Lines of the File Envelop.txt, Created by Program LAMFAIL
During an Analysis of a [0/45/90/45]s Graphite–Epoxy Laminate
**Program LAMFAIL***
Laminate failure predictions based on the maximum stress failure
criterion
Stress-free temperature=175.0j
Service temperature=20.0j
Change in temperature=155.0j
Change in moisture content=1.00%
Nxx Nyy Failed ply Failure type

0.1590855E+06 0.0000000E+00 3 Matrix
0.1590665E+06 0.3187704E+03 3 Matrix
0.1590474E+06 0.6387443E+03 3 Matrix
0.1590282E+06 0.9599286E+03 3 Matrix
0.1590090E+06 0.1282330E+04 3 Matrix
0.1589897E+06 0.1605956E+04 3 Matrix
0.1589703E+06 0.1930813E+04 3 Matrix
0.1589508E+06 0.2256908E+04 3 Matrix
0.1589313E+06 0.2584248E+04 3 Matrix
0.1589117E+06 0.2912841E+04 3 Matrix
0.1588920E+06 0.3242693E+04 3 Matrix
0.1588723E+06 0.3573813E+04 3 Matrix

The maximum stress failure criterion had been selected for use, and stress resultants Nxx
and Nyy were considered. The file also contains the number of the ply predicted to fail and
the failure type (i.e., fiber, matrix, or shear) for each combination of Nxx and Nyy predicted
to cause failure.
following process to determine the tensile loads Nxx and Nyy necessary to
cause first-ply failure, according to the maximum stress failure criteria:
(a) Use program CLT to determine the stresses (r11,r22,s12)

induced in each ply by unit loads Nxx=1, Nyy=1/2 (i.e.,
k1=0.5).
(b) Determine which ply (or plies) is closest to the failure con-
dition dictated by the maximum stress failure criterion, using
the failure strengths listed for graphite/epoxy in Table 3 of
Chap. 3

Table 2 First Few Lines of the File Envelop.txt, Created by Program LAMFAIL
During an Analysis of a [0/45/90/45]s Graphite–Epoxy Laminate
**Program LAMFAIL***
Laminate failure predictions based on the Tsai–Hill failure
criterion
Stress-free temperature=175.0j
Service temperature=20.0j
Change in temperature=155.0j
Change in moisture content=1.00%
Nxx Nyy Failed ply

0.1565825E+06 0.0000000E+00 3
0.1565915E+06 0.3138105E+03 3
0.1566003E+06 0.6289169E+03 3
0.1566091E+06 0.9453267E+03 3
0.1566178E+06 0.1263047E+04 3
0.1566263E+06 0.1582084E+04 3
0.1566348E+06 0.1902446E+04 3
0.1566430E+06 0.2224140E+04 3
0.1566512E+06 0.2547174E+04 3
0.1566593E+06 0.2871555E+04 6
0.1566672E+06 0.3197291E+04 3
0.1566750E+06 0.3524387E+04 3

The Tsai–Hill failure criterion had been selected for use, and stress resultants Nxx and Nyy
were considered. The file also contains the number of the ply predicted to fail for each
combination of Nxx and Nyy predicted to cause failure.
(c) Calculate the increase in unit loads that will cause first-ply ply
failure.
(d) Confirm the results of part (c) using program LAMFAIL.
2. Repeat Problem 1 for a room temperature cure glass–epoxy system

3. Repeat Problem 1 for a room temperature cure Kevlar–epoxy system.
4. Repeat Problem 1 for the following load condition:
Nyy ¼ Nxx =2 Nxy ¼ 5Nxx Mxx ¼ Myy ¼ Mxz ¼ 0
5. Repeat Problem 1, except assume that the laminate is produced using a
graphite–epoxy system cured at 175jC (350jF).
glass–epoxy system cured at 175jC (350jF).

Figure 14 A [0/F30]s laminate loaded in biaxial tension, as described in
Problem 1.

Kevlar–epoxy system cured at 175jC (350jF).
8. Three different first-ply failure envelopes for a [0/45/90/45]s graphite–
epoxy laminate are shown in Fig. 9. The three points listed below lie on
these failure envelopes. In each case, use program CLT to determine
which ply(ies) is predicted to fail, according to the maximum stress
failure criterion:
(a) Nxx=327 kN/m, Nyy=0 kN/m (this point lies on the

DT=DM=0 curve).
(b) Nxx=79.4 kN/m, Nyy=0 kN/m (this point lies on the DT=
155jC, DM=0 curve).
(c) Nxx = 159 kN/m, Nyy = 0 kN/m (this point lies on the
DT = 155jC, DM = 1% curve).
9. Three different first-ply failure envelopes for a [0/45/90/45]s graphite–

failure criterion:
(a) Nxx=0 kN/m, Nyy=32 7kN/m (this point lies on the

DT=DM=0 curve).
(b) Nxx=0 kN/m, Nyy=79.4 kN/m (this point lies on the
DT=155jC, DM=0 curve).

(c) Nxx=0 kN/m, Nyy=159 kN/m (this point lies on the
DT=155jC, DM=1% curve).
10. Three different first-ply failure envelopes for a [0/45/9045]s graphite–

failure criterion:
(a) Nxx=115.2 kN/m, Nyy=212.1 kN/m (this point lies on the

DT= DM=0 curve).
(b) Nxx=205.2 kN/m, Nyy=91.6 kN/m (this point lies on the
DT= 155jC, DM=0 curve).
(c) Nxx=172.7 kN/m, Nyy=154.6 kN/m (this point lies on the
DT=155jC, DM=1% curve).
11. Use program LAMFAIL to create first-ply envelopes for a [0/F45]s

graphite–epoxy laminate subjected to different combinations of Nxx and
Myy, based on the maximum stress, Tsai–Hill, and Tsai–Wu failure
criteria. Assume that the laminate was cured at 175jC and will be used at
room temperatures (20jC) during service, and that it will not experience
an increase in moisture content (DM = 0%). Use a plotting format sim-
ilar to Fig. 13.
REFERENCES
1. Broek, D. Elementary Engineering Fracture Mechanics; Kluwer Academic Pub-
lishers: Hingham, MA, 1986, ISBN 90-247-2580-1.
2. Reifsnider, K.L., Ed. Fatigue of Composite Materials; Elsevier Sci. Publ.: New
York, NY, 1991.
3. Selvarathinam, A.S.; Weitsman, Y.J. Shear-lag analysis of transverse cracking
and delamination in cross-ply carbon-fibre/epoxy composites under dry, sat-
urated and immersed fatigue conditions. Compos. Sci. Technol. 1999, 59 (14),
2115–2123.
4. Surgeon, M.; Vanswijgenhoven, E.; Wevers, M.; van der Biest, O. Transverse
cracking and Poisson’s ratio reduction in cross-ply carbon fibre-reinforced
polymers. J. Mater. Sci. 1999, 34 (22), 5513–5517.
5. Joffe, R.; Varna, J. Analytical modeling of stiffness reduction in symmetric and
balanced laminates due to cracks in 90j layers. Compos. Sci. Technol. 1999, 59
(11), 1641–1652.
6. Pipes, R.B.; Daniel, I.M. Moiré analysis of the interlaminar shear edge effect in
laminated composites. J. Compos. Mater. 1971, 5, 255.
7. Oplinger, D.W.; Parker, B.S.; Chiang, F.-P. Edge-effect studies in fiber-re-
inforced laminates. Exp. Mech. 1974, 14 (9), 347–354.

8. Czarnek, R.; Post, D.; Herakovich, C.T. Edge effects in composites by Moiré
interferometry. Exp. Tech. 1983, 7 (1), 18–21.
9. Herakovich, C.T.; Post, D.; Buczek, M.B.; Czarnek, R. Free edge strain con-
centrations in real composite laminates: experimental–theoretical correlation.
J. Appl. Mech. 1985, 52, 787–793.
10. Pipes, R.B.; Pagano, N.J. Interlaminar stresses in composite laminates under
uniform axial extension. J. Compos. Mater. 1970, 4, 538–548.
11. Rybicki, E.F. Approximate three-dimensional solutions for symmetric laminates
under in-plane loading. J. Compos. Mater. 1971, 5, 354–360.
12. Wang, A.S.D.; Crossman, F.W. Some new results on edge effects in symmetric
composite laminates. J. Compos. Mater. 1977, 11, 92–106.
13. Hsu, P.W.; Herakovich, C.T. Edge effects in angle-ply composite laminates.
J. Compos. Mater. 1977, 11, 422–428.
14. Herakovich, C.T. On the relationship between engineering properties and de-
lamination of composite materials. J. Compos. Mater. 1981, 15, 336–348.
15. Nailadi, C.L.; Adams, D.O.; Adams, D.F. An experimental and numerical in-
vestigation of the free edge problem in composite laminates. Proceedings of the
1998 SEM Spring Conference, Houston, TX, June 1–3, pp. 59–62.
16. Puppo, A.H.; Evensen, H.A. Interlaminar shear in laminated composites under
generalized plane stress. J. Compos. Mater. 1970, 4, 204–220.
17. Pagano, N.J.; Pipes, R.B. The influence of stacking sequence on laminate
strength. J. Compos. Mater. 1971, 5, 50–55.
18. Pagano, N.J. On the calculation of interlaminar normal stress in composite
laminate. J. Compos. Mater. 1974, 8, 65–81.
19. Pagano, N.J. Free-edge stress fields in composite laminates. Int. J. Solids Struct.
1978, 14, 401–406.
20. Wang, S.S.; Choi, I. Boundary-layer effects in composite laminates: Part 1. Free-
edge stress singularities. J. Appl. Mech. 1982, 49, 541–548.
21. Wang, S.S.; Choi, I. Boundary-layer effects in composite laminates: Part 2. Free-
edge stress solutions and basic characteristics. J. Appl. Mech. 1982, 49, 549–
560.
22. Zwiers, R.I.; Ting, T.C.T.; Spilker, R.L. On the logarithmic singularity of free-
edge stress in laminated composites under uniform extension. J. Appl. Mech.
1982, 49, 561–569.
23. Ting, T.C.T. Anisotropic Elasticity Theory and Applications; Oxford University
Press: New York, NY, 1996, ISBN 0-19-507447-5.
24. Kassapoglou, C.; Lagace, P.A. An efficient method for the calculation of in-
terlaminar stresses in composite materials. J. Appl. Mech. 1986, 53, 744–750.
25. Kassapoglou, C.; Lagace, P.A. Closed form solutions for the interlaminar stress
field in angle-ply and cross-ply laminates. J. Compos. Mater. 1987, 21, 292–308.
26. Becker, W. Closed-form solutions for the free-edge effect in cross-ply laminates.
Compos. Struct. 1993, 26 (1–2), 39–45.
27. Stern, M. Families of consistent conforming elements with singular derivative
fields. Int. J. Numer. Methods Eng. 1979, 14, 409–421.

8
Composite Beams
Composite beams subjected to axial, transverse, or combined loads are

discussed in this chapter. Composite beams with rectangular and various
thin-walled cross sections are considered. The analysis presented allows the
determination of the effective axial and flexural rigidities of a composite
beam. Composite beams can be analyzed using the same techniques used to
study the behavior of statically determinate or indeterminate isotropic beams,
once the effective rigidities are determined.
It is expected that readers of this text will have previously studied a subject
known as ‘‘mechanics of materials’’ (also called ‘‘strength of materials’’). One
of the primary topics considered in mechanics of materials is the structural
behavior of a prismatic beam. A prismatic beam is defined as a long, slender,
and initially straight structural member whose cross section does not vary
along its length. Typical problem objectives are to determine the stresses,
strains, and deflections induced in a prismatic beam by various types of
external loads. Many excellent textbooks are available to support such
studies, a few of which are listed here as Refs. (1–5). In virtually all of these
references, it is assumed that the beam is composed of an isotropic material or

at most is a beam in which different layers of isotropic materials are bonded or
bolted together to form the overall beam.* In contrast, in this chapter, we will
consider the behavior of prismatic beams produced using laminated aniso-
tropic composite materials.
The traditional pedagogical approach used in textbooks devoted to
mechanics of materials, including Refs. (1–5), is to separate the types of
loading commonly applied to prismatic beams/bars into three categories:
Axial loads (i.e., loads acting parallel to the long axis of the beam,
which tend to cause the beam to elongate or compress).
Bending loads (i.e., loads acting transverse to the long axis of the
beam, which tend to cause the initially straight beam to bend).
Torques (i.e., moments acting about the long axis of the beam, which
tend to cause the beam to twist about its long axis).
Having made this classification, the stresses, strains, and deflections
induced in a prismatic beam by each category of external load are considered
in separate sections or chapters. It should be clear that this is an artificial
distinction made purely for pedagogical purposes. In actual practice, a
prismatic beam or shaft may well be subjected to axial load(s), to bending
load(s), and/or to torque(s) simultaneously. In such a case, the stresses,
strains, and deflections induced are due to the combined effects of all load
components.
Composite beams subjected to axial loads and/or bending loads will be
discussed in this chapter. Composite beams subjected to torques are not
considered in this text. As will be seen, the analysis of composite beams
subjected to axial loads and bending loads follows directly from Classical
Lamination Theory (CLT). A note of caution regarding free-edge stresses is
therefore appropriate. As explained in Chap. 7, free-edge stresses occur
because of a mismatch in the effective properties exhibited by adjacent plies
in a multiangle composite laminate. A CLT analysis does not account for
free-edge stresses, and hence in regions dominated by free-edge stresses, the
ply stresses and strains predicted on the basis of CLT are invalid. As a rule of
thumb, the region over which significant free-edge stresses occur has a width
approximately equal to the laminate thickness. Now, by definition, a ‘‘beam’’
is a structure whose width is similar in magnitude to its thickness (i.e., the
dimensions of the beam cross section are much less than the beam length).
* A beam composed of two or more isotropic materials is sometimes called a ‘‘composite’’

beam; in such cases, the use of the term ‘‘composite’’ refers to the fact that the beam is
composed of two or more isotropic materials. This is in contrast to the sort of ‘‘composite’’
beam discussed in this text, in which each individual composite ply is anisotropic.

Consequently, free-edge effects may significantly influence the ply stresses
and strains induced in a composite beam. Despite these concerns, CLT will
be used in this chapter to predict the response of composite beams to axial
and bending loads. The effects of free-edge stresses will be ignored. The error
in predicted response caused by neglecting free-edge stresses is difficult to
estimate, in part because the magnitude of free-edge stresses depends on
both stacking sequence and elastic properties of the material system used
to fabricate the beam. In general, it is expected that free-edge stresses will
have little impact on predicted beam stiffness, but may dominate beam failure
loads.
2 COMPARING CLASSICAL LAMINATION THEORY

TO ISOTROPIC BEAM THEORY
Before we begin our discussion of composite beams, it is instructive to
compare classical lamination theory (CLT) to results from isotropic beam
theory. CLT is based on the behavior of a ‘‘thin plate.’’ A plate with in-plane
dimensions a and b and thickness t has been shown previously in Fig. 1 of
Chap. 6. The plate can be considered ‘‘thin’’ if the plate thickness is less than
about 1/10 the in-plane dimensions, i.e., if t < a/10 and t < b/10. As already
discussed in Sec. 1 of Chap. 6, the description of a thin plate can be
‘‘converted’’ to the description of a beam if we let the x-direction define the
beam axis and allow the in-plane width of the plate (dimension b) to approach
the plate thickness: b c t. In this way, we describe a beam with rectangular
cross section bt and length a, as previously shown in Fig. 3 of Chap. 6.
Recall that six types of loads are considered in CLT: three stress
resultants (Nxx, Nyy, and Nxy) and three moment resultants (Mxx, Myy, and
Mxy). For present purposes, we will only consider two of these loads: stress
resultant Nxx and moment resultant Mxx. That is, to compare CLT to iso-
tropic beam theory, we will specify that Nyy = Nxy = Myy = Mxy = 0. A
further distinction must also be made regarding the manner in which loads are
described in CLT vs. beam theory. In CLT, all stress and moment resultants
are assumed to be uniformly distributed along the edge of a plate and are
described in units corresponding to this assumption. For example, stress
resultant Nxx is described in units of N/m (or lbf/in.), and moment resultant
Mxx is described in units of N m/m (or lbf in./in.). In contrast, in beam theory,
‘‘point’’ or ‘‘concentrated’’ loads are usually specified. In beam theory, an
axial load (corresponding to Nxx) is described in units of N (or lbf ), whereas a
bending moment (corresponding to Mxx) is described in units of N m (or lbf
in.). In this textbook, the superscript ‘‘b’’ is used to differentiate between loads
as defined in beam theory vs. loads as defined in CLT. Thus, an axial load and
b b
bending moment applied to a beam will be denoted Nxx and Mxx , respec-

tively. Since the width of the beam is b, the two load definitions are related
b b
according to Nxx = Nxx /b and Mxx = Mxx /b.
We are now ready to compare CLT with beam theory. From Sec. 6.2 of
Chap. 6, the midplane strains and curvatures induced in a general multiangle
composite laminate subjected to a combination of mechanical loads, a uni-
form change in temperature, and a uniform change in moisture content are
given by Eq. (45) of Chap. 6, repeated here for convenience:
8 9 2 38 9
>
> eoxx >
> a11 a12 a16 b11 b12 b16 > > Nxx þ NTxx þ NM xx >>
>
> >
> 6 > >
>
> o >> 7>>
> þ T
þ M > >
>
>
> e > 6 a a a b b b 7 > N N N >
> yy > 7> >
12 22 26 21 22 26 yy yy yy
> > 6
> >
> >
< co = 6 a 7 < M > =
xy 6 16 a 26 a 66 b 61 b 62 b 66 7 N xy þ N T
xy þ N xy
¼6 7
>
> jxx >> 6 b11 b21 b61 d11 d12 d16 7> Mxx þ MTxx þ MM >
>
> >
> 6 7>>
> xx >
>
>
>
> >
> 6 7>> >
>
>
> yy >
j >
> 4 b b b d d d 5 >
> M þ M T
þ M yy >
M
>
>
> >
12 22 62 12 22 26 >
>
yy yy >
: ; : M þ MT þ MM > ;
jxy b16 b26 b66 d16 d26 d66 xy xy xy
Since our current objective is to compare CLT to isotropic beam theory, we

will simplify our discussion by ignoring the effects of changes in temperature
and moisture content (i.e., let NijT=MijT=NijM=MijM=0). For an isotropic
beam or plate, substantial simplifications occur in the [abd] matrix, as was
already discussed in Sec. 7 of Chap. 6. Noting that we have already specified
that Nyy=Nxy=Myy=Mxy=0, for an isotropic beam, Eq. (45) of Chap. 6
reduces to:
8 o 9 2 38 b 9
> e > a11 a12 0 0 0 0 > Nxx =b >
>
> xx >
> 6 >
> >
>
>
> o > > a a11 0 0 0 0 7>> 0 > >
>
> eyy >> 6 7> >
7> >
12
< o >
> = 6 >
< >
cxy ¼ 66 0 0 2ða11 a12 Þ 0 0 0 7 0 =
7
>
> j > 6
xx > 6 0 0 0 d11 d12 0 7> b >
7> M =b >
>
> >
> 6 7>> xx > >
>
> >
> 4 0 5 >
> >
>
> jyy >
> > 0 0 d12 d11 0 > 0 >
> >
: ; : ;
jxy 0 0 0 0 0 2ðd11 d12 Þ 0
where for an isotropic material, the elements of the [abd] matrix reduce to (see
Sec. 7 of Chap. 6):
1 v
a11 ¼ a12 ¼
Et Et
12 12v
d11 ¼ d12 ¼
Et3 Et3
24ð1 þ vÞ
d66 ¼ 2ðd11 d12 Þ ¼
Et3

b
Let us first consider the midplane strains induced by Nxx alone (that is, let
b b
Mxx = 0). From the above result, the axial midplane strain induced by Nxx
(only) is given by:
b
Nbxx 1 Nxx
exx ¼ ða11 Þ
o
¼
b Et b
Since the cross-sectional area of the rectangular beam is A = (t)(b), we can
write this result as:
Nbxx rxx
eoxx ¼ ¼ ð1aÞ
AE E
b
The quantity rxx = Nxx /A is simply the uniaxial stress induced in a prismatic
isotropic beam subjected to an axial load. Hence, CLT reduces to the uniaxial
form of Hooke’s Law for an isotropic material. The product AE is known as
the axial rigidity of an isotropic beam.
Similarly, the transverse midplane strain is given by:
Nb v Nb vNbxx vrxx
eoyy ¼ ða12 Þ xx ¼ xx
¼ ¼ ¼ veoxx ð1bÞ
b Et b AE E
Hence, for uniform axial loading, the transverse strain eoyy is related to the
o
axial strain exx via Poisson’s ratio, as expected for an isotropic beam sub-
jected to a state of uniaxial stress. Finally, no midplane shear strain is pre-
dicted (also as expected for an isotropic beam):
coxy ¼ 0 ð1cÞ
It is mentioned in passing that the stress tensor induced in the beam is
uniaxial (i.e., rxx p 0, ryy = rzz = sxy = sxz = syz = 0). That is, the principal
stresses are rp1 = rxx, rp2 = rp3 = 0, and the principal stress coordinate sys-
tem is coincident with the x–y–z coordinate system. Similarly, since shear
strain is zero, the principal strain coordinate system is coincident with the
o
x–y–z coordinate system and ep1 = exx , ep2 = eoyy . The principal stress and
principal strain coordinate systems are therefore coincident, which is always
the case for isotropic materials.
b
Now consider the midplane strains and curvatures induced by Mxx
b
alone (that is, let Nxx = 0). In this case, CLT predicts that all midplane strains
o o o
are zero (exx = eyy = cxy = 0). Therefore, the midplane represents the neutral
surface, as expected for an isotropic beam subjected to pure bending. The
midplane curvatures are:
b b
Mxx Mxx
jxx ¼ d11 jyy ¼ d12 jxy ¼ 0
b b
The fact that the twist curvature jxy is zero implies that the x–z and y–z planes
are the principal planes of curvature, as expected for a prismatic isotropic

beam with symmetric rectangular cross section. Substituting the reduced
forms for d11 and d12 for an isotropic material, we find:

12 Mbxx 12v Mbxx
jxx ¼ 3 jyy ¼
Et b Et3 b
The moment of inertia for a rectangular cross section of width b and height t is
I = bt3/12. Therefore the above expressions can be written as:
Mbxx
jxx ¼ ð2aÞ
EI
vMbxx
jyy ¼ ð2bÞ
EI
Equations (2a) and (2b) are the well-known moment-curvature equations for
isotropic beams. The product EI is known as the flexural rigidity of an
isotropic beam. Together, Eqs. (2a) and (2b) imply that jyy = –vjxx, which
shows that anticlastic bending of an isotropic beam with symmetric rectan-
gular cross section is predicted by CLT and, further, that jyy is related to jxx
via Poisson’s ratio, as expected.
The through-thickness variation in normal strains exx and eyy is given by:
b
Mxx
exx ¼ exx þ zjxx ¼ z
o
EI
b
vMxx
eyy ¼ eoyy þ zjyy ¼ z
EI
Through-thickness variation in stresses can be calculated using Eq. (30) in
Chap. 5, which becomes (assuming isotropic properties and also that DT =
DM = 0):
2 3
8 9 E vE 8 9
> rxx > 6 ð1 v2 Þ ð1 v2 Þ 0 7> exx >
>
> >
> 6 7 >
> >
>
>
< >
= 6 vE 7>< >
=
6 E 7
ryy ¼ 6 0 7 yye
>
> >
> 6 ð1 v Þ ð1 v Þ
2 2 7> >
>
> >
> 6 7>> >
>
: ; 4 E 5>:
c
>
;
sxy 0 0 xy
2ð1 þ vÞ
2 38 9
E vE > zMbxx >
6 ð1 v2 Þ ð1 v2 Þ 0 7>> >
6 7>> EI > >
>
6 vE 7<> >
=
6 E 7 zvM
¼6 0 7 b
6 ð1 v2 Þ ð1 v2 Þ 7> xx
>
6 7>>
> EI > >
>
4 E 5 >
> >
>
0 0 : ;
2ð1 þ vÞ 0

Upon completing the matrix multiplication indicated above, we find:
8 9 8 b 9
>
> rxx >
> > Mxx z >
< = >< >
=
I
ryy ¼
>
> >
> > 0 >
: ; >: >
;
sxy 0
Hence, the CLT analysis predicts that ryy and sxy are both zero, and that the
b
axial stress rxx induced in an isotropic beam by Mxx (only) is given by the
familiar flexure formula:
Mbxx z
rxx ¼ ð3Þ
I
As before, since only an axial stress is induced, the stress tensor is
everywhere uniaxial and principal stresses are rp1 = rxx, rp2 = rp3 = 0. The
principal stress coordinate system is coincident with the x–y–z coordinate
system. Similarly, since shear strain is zero, the principal strain coordinate
o
system is coincident with the x–y–z coordinate system and ep1 = exx , ep2 = eoyy .
We again conclude that the principal stress and principal strain coordinate
systems are coincident, which is always the case for isotropic materials.
Based on the above, we conclude that the analysis represented by CLT
can be used to represent an isotropic beam with rectangular cross section
b b
subjected to an axial load Nxx and bending moment Mxx , both of which are
b
constant along the length of the beam. If Nxx = 0, then the loading condition
represented by CLT corresponds to a state of pure bending. Now, there is an
additional fundamental result from traditional beam theory that cannot be
recovered by applying CLT to an isotropic beam. Specifically, traditional
beam theory allows one to calculate the shear stresses induced in a beam.
Recall that shear stresses sxz are given by the so-called shear formula (1–5):
VQ
sxz ¼
Ib
where V is the shear force present at a specified beam cross section, Q is the
first moment of an area about the neutral axis (for a rectangular beam
Q = b[(t2/4)–z2]/2), I is the area moment of inertia of the entire cross section
about the neutral axis, and b is the width of the beam. The reason that CLT
cannot be used to predict shear stress sxz is due to the fundamental
b
assumption that the bending moment Mxx is constant along the length of
the beam. That is, recall from earlier studies [see, for example, Refs. (1–5)] that
b
the shear force V is related to the bending moment Mxx according to:
dMbxx
V¼
dx

b b
Since Mxx is assumed constant in a CLT analysis, V = dMxx /dx = 0. There-
fore a CLT analysis implies that shear stress sxz = 0 due to the assumed
loading conditions.
The discussion presented in this section has been intended to show that
CLT is entirely consistent with traditional isotropic beam theory. The reader
should carefully note, however, that in this discussion, CLT has been
specialized to the case of isotropic beams. The behavior of composite beams
will be discussed in following sections. It will be seen that in some ways, the
behavior of composite and isotropic beams is similar, but in others, they are
quite different. In particular, for composite beams, the flexure formula [Eq.
(3)] is only valid if the beam is oriented in a specific way with respect to the
applied loads. Also, for composite beams, the principal stress and principal
strain coordinate systems are not (in general) coincident.
3 TYPES OF COMPOSITE BEAMS CONSIDERED

The types of composite beams considered in this chapter are summarized in
b b
Fig. 1. An externally applied axial load Nxx , bending moment Mxx , and
transverse distributed load q(x) are also shown. Both the beam cross section
and applied loads are referenced to an x–y–z coordinate system, where the
Figure 1 Composite beams with various cross sections. (a) Rectangular beam,
plies orthogonal to plane of loading. (b) Rectangular beam, plies parallel to plane
of loading. (c) I-beam. (d) T-beam. (e) Hat-beam. (f) Box-beam.

y- and z-axes are orthogonal to the long axis of the beam. Bending moment
Mxxb and transverse load q(x) act within the x–z plane, and hence the x–z
plane is called the plane of loading.
The type of ‘‘composite beam’’ implied during earlier sections is shown
in Fig. 1(a). In this case, the beam cross section is rectangular, and all ply
interfaces are orthogonal to the plane of loading. In contrast, a composite
beam with rectangular cross section but in a decidedly different orientation
with respect to the plane of loading is shown in Fig. 1(b). In this case, the ply
interfaces are parallel to the plane of loading. The effective axial rigidity
exhibited by a composite beam is identical in either orientation. As will be
seen, however, the effective flexural rigidity of a rectangular composite beam
will differ substantially depending on whether the beam is orientated as shown
in Fig. 1(a) or (b).
It is of course possible to manufacture and use composite beams with
rectangular cross sections. However, in practice, it is far more common to use
thin-walled composite beams, such as those shown in Fig. 1(c)–(f ). Beams
with thin-walled cross sections are more commonly used because they provide
far higher flexural rigidities per unit weight than a solid rectangular beam.
The method used to study thin-walled composite beams herein is similar
to that described by Swanson (8). The general approach is to approximate the
cross sections shown in Fig. 1(c)–(f ) as an assembly of flat rectangular
laminates. The stresses and strains induced in each region of the cross section
will then be determined based on CLT. In all cases, the plane of loading is de-
fined as the x–z plane. Those regions of a beam cross section that are parallel
to the plane of loading will be called the ‘‘web laminate’’ (or ‘‘web lami-
nates’’), whereas those regions that are orthogonal to the plane of loading will
be called the ‘‘flange laminate’’ (or ‘‘flange laminates’’). Thus, for example, a
composite I-beam has one web laminate and two flange laminates, whereas a
composite box-beam has two flange laminates and two web laminates.
Several restrictions are placed on the types of composite beams consid-
ered. First and foremost, only beams in which all flange and web laminates are
produced using symmetric stacking sequences will be considered. This restric-
tion is imposed to avoid complications due to thermal and moisture effects.
That is, if a beam is produced using nonsymmetric web or flange laminates,
then substantial thermal and/or moisture moment resultants (MijT and MijM)
are induced, which may lead to substantial warping of the beam, even before
application of any external load. For this reason, composite beams used in
practice are almost always produced using laminates based on a symmetric
stacking sequence. As discussed in Sec. 7.1 of Chap. 6, for a symmetric
stacking sequence, MijT = MijM = 0, and hence this complication is eliminated.
Secondly, we require that all beam cross sections possess both geometric
and material symmetry about the plane of loading (the x–z plane). Having

already stipulated that all laminates are symmetric, then the requirement of
geometric and material symmetry about the x–z plane is achieved automati-
cally for solid rectangular composite beams and in these cases requires no
further discussion. However, symmetry about the x–z plane implies addi-
tional restrictions for thin-walled cross sections involving multiple web
laminates. For cross sections with a single web laminate (an I-beam or
T-beam), symmetry about the x–z plane is automatically satisfied if the web
laminate is symmetric. However, for cross sections with two web laminates (a
hat- or box-beam), the two symmetric web laminate must represent a set of
fiber angles that are symmetric about the x–z plane, as well as being symmetric
about the local web laminate midplanes.
Note that symmetry about the x–z plane places no restrictions on the
flange laminates. Thus, the top and bottom flanges in an I-, hat-, or box-beam
are allowed to have a different stacking sequence and thickness, although they
must both be symmetric about their respective midplanes (the stacking
sequences and widths of the two bottom flanges in a hat-beam must be
identical, however). The beam cross section may therefore be nonsymmetrical
about the x–y plane.
In practice, it can be problematic to produce thin-walled composite
beams that feature the symmetries just described. Difficulties may arise due to
the manufacturing process used to produce the beam. Methods of producing
a thin-walled composite beam using unidirectional pre-preg tapes will be
discussed to illustrate the practical difficulties encountered.
First, consider the processes that could be used to produce a composite
T-beam. Two possibilities are shown in Fig. 2. In Fig. 2(a), the cross section is
formed by first curing the flange and web laminates separately. The two
laminates are then bonded together in a second operation to form the desired
T-cross section, as indicated. An advantage to this approach is that the
stacking sequences used in the web and flange laminates are completely
independent. Since any stacking sequence can be used in either laminate, it
is easy to produce a T-beam with the required symmetric stacking sequence in
both web and flange laminates in this manner and to insure symmetry about
the x–z plane. However, a distinct disadvantage is that the web and flange
laminates are joined solely by the adhesive bond—no continuous fibers cross
the junction between web and flange. Hence, this approach is likely to result in
a relatively low-strength T-beam. A more common method of producing a
composite T-beam is illustrated in Fig. 2(b). In this case, the plies that exist
within the web laminate are extended into (and become part of ) the flange
laminate. Additional plies, which span the width of the flange, are added to
complete the flange laminate. During layup, an internal ‘‘v-shaped’’ cavity is
formed near the web–flange junction. This cavity is filled with some filler
material (often pre-impregnated unidirectional tow) prior to curing the

Figure 2 Methods to produce a composite T-beam using unidirectional pre-preg
tape. (a) Forming a T-cross section by bonding together web and flange lam-
inates. (b) Forming a T-cross section by extending web plies into the flange.
T-beam. This second approach of producing a T-beam has the distinct ad-
vantage of providing continuous fibers across the junction between the web
and flange. However, the stacking sequences used in the web and flange
laminates are no longer independent. Since the flange laminate is required to
be symmetric (at least for the analysis presented in this section), the fiber
angles used in the web and inner portions of the flange laminates must be
repeated on the outer surface of the flange. Furthermore, if the web is pro-
duced using unidirectional pre-preg tape, then it is impossible to produce a
T-beam that has both a symmetric web and a symmetric flange laminate,

unless only 0j or 90j plies are used in the web. This difficulty may not be
immediately obvious, but is illustrated in Fig. 3. The figure shows the
formation of a symmetric 2-ply web laminate, that is, a [hj]s web laminate
(where h p 0j or 90j). As indicated, extending the web plies into the flange
results in a single flange ply in which the fiber angle is +hj on one side of the
flange–web junction, but –hj on the other side. Hence, it is not possible to add
additional plies across the entire width of the flange to produce both a sym-
metric laminate, unless web fiber angles are restricted to either h = 0j or 90j.
In some instances, it may not be acceptable to restrict web angles to 0j or 90j
since this restriction results in a web laminate with relatively low shear stiff-
ness. A similar difficulty is encountered when producing a beam with two web
laminates such as a hat- or box-beam. A single ply that becomes part of both
web laminates and both flange laminates in a box-beam is shown in Fig. 4. In
this case, it is not possible to produce web laminates that are symmetric about
the x–z plane since a web ply with a fiber angle of +hj on one side of the x–z
plane becomes a ply with a fiber angle of –hj on the other side of the x–z plane.
Figure 3 Illustration of why unidirectional fabrics or pre-pregs cannot be used

to produce web and flange laminates that are both symmetric for T- or I-beams.

Figure 4 Illustration of why unidirectional fabrics or pre-pregs cannot be used
to produce web laminates that are symmetric about the x–z plane.
As before, the only way this can be avoided (if the beam is produced using
unidirectional pre-preg tape) is to use fiber angles of h = 0j or 90j in the web,
which may lead to web laminates with unacceptably low shear stiffness.
These difficulties in achieving symmetry are avoided if the web is
produced using a woven or braided pre-preg fabrics rather than unidirec-
tional pre-preg, as shown in Fig. 5. Recall from Sec. 4 of Chap. 1 that a single
ply of a woven or braided fabric features two or three fiber directions, oriented
symmetrically about the warp direction. Since each individual woven or
braided ply features a symmetric Fhj fiber pattern, plies extending from the
web into the flange (as in a T-beam) or completely around the circumference
of the cross section (as in a box-beam) retain the identical F hj fiber pattern in
both web and flange laminates. Note that the use of a woven or braided ply to
form the web laminate(s) does not preclude the use of additional unidirec-
tional plies in the flange laminate(s).
Finally, objectionable levels of thermal warping may still occur for some
thin-walled composite beams, even though they satisfy all of the symmetry
requirements described above. Thermal warping may occur because we have
allowed the beam cross section to be nonsymmetric about the x–y plane.
Suppose, for example, that an I-beam is produced in which the (symmetric)
top flange is produced using a very different stacking sequence than the
(symmetric) bottom flange. This implies that the effective thermal expansion
coefficient axx of the top flange may be very different than that of the bottom
flange. Consequently, if the beam experiences a uniform change in temper-

Figure 5 The use of woven or braided fabrics or pre-preg avoids the difficulties
shown in Figs. 3 and 4.

ature, then the top flange will tend to expand or contract at a different rate
than the bottom flange, causing the beam to warp. The symmetric web
laminate will tend to restrict such warping, but nevertheless, this effect may
be significant depending on details of a given beam design. This possibility will
not be addressed in this text, although the theory presented can be easily
modified to account for this effect.
4 EFFECTIVE AXIAL RIGIDITY OF RECTANGULAR

COMPOSITE BEAMS
b
A rectangular composite beam subjected to axial load Nxx is shown in Fig. 6.
The width and thickness of the beam cross section are labeled b and t,
respectively. We will now use CLT to predict the mechanical response of this
beam to pure axial loading. As mentioned earlier, significant free-edge stresses
may be present (depending on the stacking sequence and material system used
to produce the beam) but will be neglected in the following discussion.
Since the beam is assumed to be symmetric and the only applied load
b
is Nxx , the midplane strains and curvatures induced in a symmetric compo-
site beam can be predicted using Eq. (45) of Chap. 6, which becomes (for
Nyy=Nxy=Mxx=Myy=Mxy=bij=MijT=MijM=0):
8 o 9 2 38 b 9
> exx > a11 a12 a16 0 0 0 >
> ðNxx =bÞ þ NTxx þ NM
xx >
>
>
> >
> > >
>
> > 6 7>
> >
>
>
> eoyy >
>
> 6
a12 a22 a26 0 0 0 >
7>
> N T
yy þ N M
yy
>
>
>
< o >
> = 6 7>
< >
=
cxy ¼ 6 a
6 16
a26 a66 0 0 0 7
7 N T
xy þ N M
xy ð4Þ
> > 6 0 d16 7
> jxx >
> > 6 0 0 d11 d12 7>
>
> 0 >
>
>
>
> >
> 6 7>
> >
>
> yy >
> j > 4 0 0 0 d12 d22 d26 5>
> 0 >
>
>
: >
; >
> >
>
jxy 0 0 0 d16 d26 d66 : ;
0
Note that thermal and moisture stress resultants are (in general) present in
the symmetric composite beam, even though the only externally applied
b
load is Nxx . As discussed in earlier chapters, thermal and moisture result-
ants contribute substantially to the strains and stresses induced in each ply
and may lead to premature failure of a composite beam if not properly
accounted for.
The preexisting ply stresses and strains are of course changed upon
b
application of an external axial load Nxx . From Eq. (4), it is easy to see
b
that the incremental change in midplane strains caused by Nxx (only) is
given by:
a11 b a12 b a16 b
eoxx ¼ N eoyy ¼ N coxy ¼ N ð5Þ
b xx b xx b xx

Figure 6 A composite beam with cross-section bt, subjected to axial load
b
Nxx .
The effective extensional Young’s modulus, Poisson’s ratio, and coefficient of

mutual influence of the second kind for a composite laminate were all defined
as follows in Sec. 9.1.1 of Chap. 6:
ex 1 a12 a16
E xx ¼ ex
vxy ¼ ex
gxx;xy ¼
ta11 a11 a11
Substituting these effective properties into Eq. (5), we find:
Nbxx Nbxx rxx

eoxx ¼ ex ¼ ex ¼ ex ð6aÞ
ðtbÞE xx AE xx E xx
! !
Nbxx ex rxx
eoyy ¼ vxy
ex
ex ¼ vxy ex ¼ vxy
ex o
exx ð6bÞ
AE xx E xx
! !
Nbxx rxx
coxy ¼ ex
gxx;xy ex ¼ ex
gxx;xy ex ¼ gxx;xy
ex
eoxx ð6cÞ
AE xx E xx
From Eq. 6(a), we see that the effective axial rigidity of a rectangular com-
posite beam can be written as:
ex ex
AE xx ¼ ðtbÞE xx ð7aÞ

where A is the cross-sectional area of the beam and E ex xx is the effective
extensional Young’s modulus of the laminate (defined in Sec. 9.1.1 of Chap.
6). Alternatively, the effective axial rigidity may be written as:

ex b
AE xx ¼ ð7bÞ
a11
Although the beam shown in Fig. 6 is oriented such that ply interfaces are
orthogonal to the x–z plane, the analysis is also applicable to a beam oriented
such that ply interfaces are parallel to the x–z plane, as shown in Fig. 1(b).
Hence, Eqs. (7a) and (7b) give the effective axial rigidity of a symmetric
rectangular composite beam, regardless of ply orientation.
Note that Eqs. (6a)–(6c) are directly analogous to similar relations for
an isotropic beam, as given by Eqs. (1a) and (1b). Since only an axial load is
applied, the effective stress tensor is uniaxial (i.e, rxx p 0; ryy ¼ rzz ¼ sxy
¼ sxz ¼ syz ¼ 0Þ. That is to say, the effective principal stresses are rp1 = rxx ;
rp2 = rp3 ¼ 0, and the effective principal stress coordinate system is coinci-
dent with the x–y–z coordinate system. However, for a composite beam, an
axial stress jxx will (in general) cause a shear strain coxy due to the presence of
a16 (or equivalently, due to the presence of Dxx,xy
ex
). Therefore, the principal
strain coordinate system is not, in general, coincident with the effective prin-
cipal stress coordinate system.
Recall from Sec. 7 of Chap. 6 that a16=0 for the following commonly
used stacking sequence:
Cross-ply.
Balanced.
Balanced angle-ply.
Quasi-isotropic.
If an axially loaded symmetric composite beam is produced using any of

these stacking sequences, then shear strain coxy is zero and the effective
principal stress and principal strain coordinate systems are coincident.
5 EFFECTIVE FLEXURAL RIGIDITIES OF RECTANGULAR

COMPOSITE BEAMS
An initially straight beam subjected to an external bending moment of equal
magnitude at either end is said to be in state of pure bending. The flexural
rigidity of a beam is defined in pure bending. In this section, the effective
flexural rigidities of composite beams with rectangular cross sections will be
determined. In order to do so, we must consider two different cases. In the first
case, we consider a beam oriented such that ply interfaces are orthogonal to

the plane of loading, as previously shown in Fig. 1(a). In the second case, we
consider a beam for which ply interfaces are parallel to the plane of loading, as
was shown in Fig. 1(b).
5.1 Effective Flexural Rigidity of Rectangular Composite

Beams with Ply Interfaces Orthogonal to the Plane
of Loading
A composite beam oriented such that ply interfaces are orthogonal to the
plane of loading is shown in Fig. 7. The height and width of the beam cross
section are denoted t and b, respectively. We will now use CLT to predict the
mechanical response of this beam in pure bending. As mentioned earlier,
significant free-edge stresses may be present (depending on the stacking
Figure 7 A composite beam with ply interfaces orthogonal to the x–z plane,
subjected to pure bending. (a) A composite beam subjected to pure bending.
(b) Cutaway view of composite beam at cross-section A–A.

sequence and material system used to produce the beam) but will be neglected
in the following discussion.
Since we have assumed that the beam is symmetric and subjected to
b
Mxx only, the midplane strains and curvatures can be predicted using Eq.
(45) of Chap. 6, which becomes (for bij ¼ Nxx ¼ Nyy ¼ Nxy ¼ Myy ¼ Mxy ¼
MTij ¼ MM ij ¼ 0 and Mxx=Mxx /b):
b
8 o 9 2 38 T 9
>
> exx >>
a11 a12 a16 0 0 0 >
> Nxx þ NM
xx >
>
> o >
> > 6 a12 7>> NT þ NM > >
>
> e > 6> a22 a26 0 0 0 7>> >
yy >
> yy >
> > 7>> Tyy >
>
< = 6a 7 N þ NM =
<
coxy ¼ 6
6 16
a26 a66 0 0 0
7 xy xy ð8Þ
> > 6 0 d16 7
> xx >
> j > 6 0 0 d11 d12 7>> Mbxx =b > >
>
> >
> 6 7>> >
>
>
>
> j yy >
>
>
4 0 0 0 d12 d22 d26 5>>
>
> 0
>
>
>
>
: ; : ;
jxy 0 0 0 d16 d26 d66 0
Thermal and moisture stress resultants NijT and NijM are (in general)
present in the symmetric composite beam. As discussed in earlier chapters,
thermal and moisture stress resultants contribute substantially to the strains
and stresses induced in each ply and may lead to premature failure of a
composite beam if not properly accounted for. The thermal and moisture
stress resultants do not influence the initial elastic response of the beam,
however. From Eq. (8), it is easy to see that midplane strains are not changed
b
upon the application of Mxx since bij=0. Hence, the midplane of the beam
cross section represents the neutral surface in the sense that application of
b
Mxx does not contribute to (or alter) preexisting midplane strains. As noted in
Sec. 2, the neutral surface for an isotropic beam with rectangular cross section
is also coincident with the midplane.
From Eq. (8), the midplane curvatures are given by:
b b b
Mxx Mxx Mxx
jxx ¼ d11 jyy ¼ d12 jxy ¼ d16 ð9Þ
b b b
The fact that the twist curvature jxy p 0 (in general) shows that the x–z and
y–z planes are not the principal planes of curvature. This is in direct contrast
to an isotropic beam since for a prismatic isotropic beam with symmetric rect-
angular cross section, the x–z and y–z planes represent the principal planes of
curvature.
The effective flexural Young’s modulus, Poisson’s ratio, and coefficient
of mutual influence of the second kind for a composite laminate were all
defined as follows in Sec. 9.1 of Chap. 6:
fl 12 d12 d16
E xx ¼ fl
vxy ¼ fl
gxx;xy ¼
d11 t3 d11 d11

Substituting these effective properties into Eq. (9), we find:

12 b
Mxx 12vxy
fl b
Mxx fl
12gxx;xy Mbxx
jxx ¼ fl 3 jyy ¼ fl 3
jxy ¼ fl 3
E xx t b E xx t b E xx t b
The moment of inertia for a rectangular cross section of width b and height t
is I = bt3/12. Therefore the above expressions can be written as:
b
Mxx
jxx ¼ fl I
ð10aÞ
E xx
fl
vxy Mbxx
jyy ¼ fl I
ð10bÞ
E xx
fl b
gxx;xy Mxx
jxy ¼ fl I
ð10cÞ
E xx
Equations (10a)–(10c) are the moment-curvature equations for a composite

beam with ply orientation orthogonal to the plane of loading and should be
compared with analogous results for isotropic beams [Eqs. (2a) and (2b)]. The
product E flxx I represents the effective flexural rigidity for a composite beam
in this orientation. That is, the effective flexural rigidity of a rectangular
composite beam with ply orientation orthogonal to the plane of loading can
be written as:

fl bt3 E xx fl
b
IE xx ¼ ¼ ð11Þ
12 d11
where E flxx is the effective flexural Young’s modulus of the laminate (defined
in Sec. 9.1.2 of Chap. 6).
Equations (10a) and (10b) also imply jyy = –vxyfl jxx, which shows that
a composite beam with rectangular cross section will exhibit anticlastic
bending in the same manner as an isotropic beam. Equations (10a) and
fl
(10c) imply jxy = gxx,xy jxx, which shows that a twist curvature is induced in a
fl
composite laminate, unless gxx,xy = 0. A midplane twist curvature jxy does
fl
not occur for isotropic beams subjected to pure bending. Recalling that gxx,xy
fl
= d16/d11, it is seen that g xx,xy is zero only for specially orthotropic laminates
(i.e., if d16=0). This rarely occurs for composite beams used in practice. The
only common stacking sequences that lead to a specially orthotropic beam are
unidirectional [0]n or [90]n stacks or symmetric cross-ply [0/90]ns stacks. The
first two stacking sequences are essentially never used in practice because they
possess significant strength and stiffness in one direction only. Cross-ply
beams may be used on occasion but suffer from very low shear stiffness.
Beams produced using any other stacking sequence, for example, a symmetric

quasi-isotropic stacking sequence, will exhibit a twist curvature when sub-
jected to a pure bending.
Based on the above definitions, we see that through-thickness variation
in strains exx, eyy, and cxy induced by pure bending is given by:
b
Mxx
exx ðzÞ ¼ zjxx ¼ z ð12aÞ
E flxx I
!
fl Mb
vxy xx
eyy ðzÞ ¼ zjyy ¼ z fl
ð12bÞ
E xx I
!
fl
gxx;xy Mbxx
cxy ðzÞ ¼ zjxy ¼ z fl ð12cÞ
E xx I
Strains are predicted to vary as linear functions of z, as expected from results

in earlier chapters. Note, however, that ply stresses rxx, ryy, and sxy are not
linear functions of z. It is not possible to develop a ‘‘flexure formula’’ for use
with composite beams with ply orientation orthogonal to the plane of
loading. Rather, stresses must be calculated using Hooke’s Law for an
anisotropic material and based on the ply strain distributions given by Eq.
(11), as discussed in earlier chapters.
5.2 Effective Flexural Rigidity of Rectangular Composite

Beams with Ply Interfaces Parallel to the Plane of Loading
We now consider a composite beam that is oriented such that the ply
interfaces are parallel to the plane of loading. Such a beam in pure bending
is shown in Fig. 8. We will continue to label the beam cross section using the
same symbols as in earlier sections. Hence, the beam depth is denoted b,
whereas the beam width (which equals the thickness of the composite
laminate) is denoted t (compare Figs. 7 and 8). As drawn in Fig. 8, plies are
numbered from left to right. That is, the outermost surface of ply 1 exists at
zV=–t/2, whereas the outermost surface of ply n exists at zV=+t/2. Ply fiber
angles are referenced to the +x V-axes and are measured positive from the
+x V-axis towards the +yV-axis, in accordance with the right-hand rule.
Note that the bending moment is referenced to the x–y–z coordinate
system, whereas the ply stacking sequence is referenced to a ‘‘new’’ x V–y V–z V
coordinate system. As is apparent from Fig. 8, the +z V- and +yV-axes are
coincident, as are the +x- and +x V-axes. The +y- and +zV-axes are parallel,
but the +zV-direction is opposite to the +y-direction. The reader should
carefully consider the two coordinate systems shown since they represent a
change in nomenclature from our earlier discussion. This is a subtle but
b
potentially confusing change. For example, the bending moment Mxx shown

Figure 8 A composite beam with ply interfaces parallel to the x–z plane,
subjected to pure bending. (a) A composite beam subjected to pure bending.
(b) Cutaway view of composite beam at cross-section A-A.
in Fig. 8 causes a very different state of stress in the composite beam than does
b
the bending moment shown in Fig. 7 (which is also denoted Mxx ). This
difference is of course due to the change in orientation of the composite beam.
Throughout this chapter, we assume that the stacking sequence is
symmetric. Hence, both the beam cross section as well as beam material
b
properties are symmetric about the x–z plane. Since the moment Mxx is
constant along the length of the beam, all beam cross sections must deform in
an identical manner. Under these conditions, the long axis of the beam must
deform into a circular arc with a radius of curvature rxx, as shown in Fig. 9.
Hence, beam cross sections that are initially plane and perpendicular to the
axis of the beam must remain plane and perpendicular to the deformed axis of
the beam following loading. These are precisely the same conditions encoun-

Figure 9 Deformations induced in a composite beam with ply interfaces
parallel to the plane of loading, subjected to pure bending (deformations shown
greatly exaggerated). (a) A composite beam subjected to pure bending. (b)
Deformed beam.

tered in pure bending of isotropic beams with symmetric rectangular cross
sections. Hence, many conclusions based on isotropic beam theory are also
applicable to the composite beam shown in Figs. 8 and 9. In particular, the
neutral axis must pass through the centroid of the beam cross section, and
axial normal strains are given by:
exx ¼ zjxx ð13Þ
where curvature jxx = 1/rxx. Hence, axial strain is a maximum at z = Fb/2
and is zero at z = 0, the neutral axis. A subtle point is that in this case, the
curvature jxx represents bending in the x–z plane (the plane of loading) and is
not comparable to the curvature jxx discussed previously. The curvature pre-
viously discussed occurs in the xV–z V plane and would now be labeled jx Vx V.
Since we have assumed both a symmetric stacking sequence and
symmetric rectangular cross section, the axial strain induced at any position
z will be uniform across the width of the beam. Hence, the axial strain given by
Eq. (13) represents the strain induced at the midplane of the laminate:
eoxx ¼ eox Vx V ¼ zjxx ð14Þ
We also note that for the loading condition shown in Fig. 8:
Ny Vy V ¼ Nx Vy V ¼ Mx Vx V ¼ My Vy V ¼ Mx Vy V ¼ 0:
Therefore, Eq. (45) of Chap. 6 becomes:
8 o 9 2 38 9
> ex Vx V > a11 a12 a16 0 0 0 > Nx Vx V þ NxTVx V þ NxMVx V >
>
> > > >
> o >
> > 6 a12 a22 a26 0
> 7>> >
>
>
> e > 6 0 0 7>> N T
þ N M >
>
> y Vy V >
< > 6
= 7<>
> y Vy V y Vy V >
>
=
cx Vy V ¼ 6
o a
6 16
a26 a66 0 0 0 7
7 N T
x Vy V þ N M
x Vy V
> 6 d16 7
>
> jx VxV >>
> 6 0 0 0 d11 d12 7>> 0 >
>
>
> >
> 6 7>>
>
>
>
>
>
> >
jy Vy V > 4 0 0 0 d12 d22 d26 5>> 0 >
>
>
: >
; >
: >
;
jx Vy V 0 0 0 d16 d26 d66 0
Our current objective is to evaluate the effective flexural rigidity of the

composite beam, so we ignore strains caused by thermal or moisture stress
resultants. Hence, the midplane strain caused by application of stress
resultant NxVxV=Nxx is:
eox Vx V ¼ eoxx ¼ a11 Nx Vx V ¼ a11 Nxx
Combining this result with Eq. (14), we find:

jxx
Nxx ¼ z ð15Þ
a11
Now consider a free-body diagram showing the internal distribution of stress
resultant Nxx at arbitrary cross-section A–A, as shown in Fig. 10. Since the

Figure 10 Free-body diagram used to relate external bending moment Mob to
internal stress resultant NxVxV.
beam is in static equilibrium, the externally applied moment Mob must be

balanced by the distribution of internal stress resultant Nxx. Recalling that the
units of Nxx are force-per-length (e.g., N/m), the incremental axial force
acting over an infinitesimal strip of height dz, located at an arbitrary distance
z from the neutral axis, is given by:
dNbxx ¼ Nxx dz
Since this force acts at distance ‘‘z’’ from the neutral axis, the incremental
internal moment associated with this force is:
dMbxx ¼ dNbxx z ¼ Nxx zdz
The total moment is obtained by integrating over the height of the beam:
þb=2
m m
þb=2
jxx 2
Mbxx ¼ Nxx zdz ¼ z dz
b=2 b=2 a11
Neither jxx nor a11 is a function of z, so they can be removed from the integral
sign. Completing the integration indicated, we find:
jxx b3
Mbxx ¼
12a11
The effective extensional Young’s modulus is given by Eq. (63) of Chap. 6,
ex 1
E xx ¼
ta11

b
Our expression for Mxx can therefore be written as:
ex
jxx E xx tb3
Mbxx ¼
12
Noting that the area moment of inertia about the y-axis for the rectangular
cross section is I = tb3/12, we can rearrange this result to find:
Mbxx
jxx ¼ ex ð16Þ
E xx I
Equation (16) is a moment-curvature equation for a composite beam with

ex
ply orientation parallel to the plane of loading. The product E xx I represents
the effective flexural rigidity of a composite beam in this orientation. That is,
the effective flexural rigidity of a rectangular composite beam with ply
orientation parallel to the plane of loading can be written as:

ex
ex
tb3 E xx b3
IE xx ¼ ¼ ð17Þ
12 12a11
ex
where E xx is the effective extensional modulus of the laminate (defined in
Sec. 9.1.1 of Chap. 6). Equation (17) should be compared with our earlier
expression for the effective flexural rigidity of a composite beam with plies
orthogonal to the plane of loading Eq. (11). It is seen that the flexural rigidity
of a composite beam with plies parallel to the plane of loading is dominated by
ex
extensional stiffnesses (i.e., E xx or a11). In contrast, the flexural rigidity of a
beam with plies orthogonal to the plane of loading is dominated by flexural
fl
stiffnesses (i.e., E xx or d11).
Anticlastic bending will occur for the composite beam shown in Fig. 10.
Normal strains eyy=ez Vz V will vary linearly with distance from the neutral axis.
Assuming a positive bending moment, then in regions below the neutral axis
(i.e., for z = y V > 0), the transverse normal strains will be compressive,
whereas in regions above the neutral axis (for z = y V<0), these strains will be
tensile. These transverse normal strains were neglected during our earlier
development and hence are not included in CLT. These strains are not of
immediate interest, so they will not be further discussed here.
Next, consider the state of stress induced in the beam. Recall from Sec.
9.1 of Chap. 6 that the effective (or nominal) normal stress is given by
rx Vx V = rxx = Nxx/t. Substituting this definition into Eq. (15) and rearranging,
we have:
rxx ta11 rxx
jxx ¼ ¼ ex
z zE xx

Combining this result with Eq. (16) and solving for effective stress, we
have:
Mbxx z
rxx ¼ ð18Þ
I
Equation (18) is analogous to the flexure formula for isotropic beams [Eq. (3)]
and can be used to determine the effective stress at any position z. As would be
expected, the effective stress rxx is at a maximum at the outer surfaces of the
beam (at z = F t/2) and is zero at the neutral surface (at z = 0). Once the
effective stress rxx has been calculated at a point of interest, then the stresses
and strains present in individual plies can be determined using the standard
CLT analysis methods.
6 EFFECTIVE AXIAL AND FLEXURAL RIGIDITIES FOR

THIN-WALLED COMPOSITE BEAMS
In this section, the effective axial and flexural rigidities of thin-walled
composite beams will be considered. The beam cross section must conform
to the symmetries described in Sec. 3. The process used to determine effective
rigidities is similar for all cross sections. A composite box-beam will be used to
illustrate the process.
Referring to Fig. 11, a box-beam is modeled using four symmetric
rectangular laminates. The two web laminates are required to have identical
stacking sequences, but the two symmetric flange laminates may differ. The
thickness of the top and bottom flange laminates is labeled ttf and tbf,
respectively. The width of both flanges is labeled b. The two identical web
laminates have a thickness of tw and height h. Thus, the overall width and
depth of the beam cross section are (b) and (h+ttf+tbf), respectively.
To define the axial rigidity, we must determine the axial strains exx
b
induced when the beam is subjected to an axial load Nxx (only). Further,
b
Nxx must be applied such that uniform axial strains are induced; that is, the
b
beam must not bend. This is equivalent to saying that the line of action of Nxx
must pass through the centroid of the beam cross section. It is known a priori
that the centroid lies along within the x–z plane since we have required
symmetry about this plane. However, we have not required symmetry about
the x–y plane. Hence, the position of the centroid relative to either outer
surface of the beam cross section (that is, coordinates zt and zb shown in Fig.
11) is initially unknown. Note that since the z-axis is positive downwards, in
all cases, zt<0 and zb>0. Our first objective is to determine either distance zt
or zb. In the following derivation, we will determine distance zb. Note that
once zb is known, zt is also known since zt = zb–(h+tbf+ttf).

Figure 11 Cross section of a composite box-beam. The positive x-axis is out
of the plane of the figure.
b
Sketches of a composite box-beam subjected to axial load Nxx (only)
b
are shown in Fig. 12. In this figure, it is assumed that the line of action of Nxx
does pass through the centroid. Consequently, a uniform axial strain exx is
induced at an arbitrary cross section of the beam, as shown in Fig. 12(a). Since
strain is uniform, the stress resultants induced in each segment of the beam
can be written as:
exx ex
Nxx jw ¼ w ¼ tw E xx jw exx ð19aÞ
a11
exx ex
Nxx jbf ¼ bf ¼ tbf E xx jbf exx ð19bÞ
a11
exx ex
Nxx jtf ¼ tf ¼ ttf E xx jtf exx ð19cÞ
a11
Symbols ‘‘w,’’ ‘‘bf,’’ and ‘‘tf’’ have been used to denote variables associated
with the web laminate, bottom flange laminate, and top flange laminate,
respectively. For example, E xxex
jw represents the effective extensional modulus
of the web laminates.

b
Figure 12 Side view of a composite box-beam subjected to axial load Nxx . The
b
line-of-action of Nxx is assumed to pass through the centroid, located distance
zb from the bottom surface of the beam. Consequently, (a) uniform axial strains
are induced at an arbitrary cross section, and (b) uniform stress resultants are
induced in each segment of the beam wall.
The stress resultants given by Eqs. (19a)–(19c) are indicated in the free-
body diagram shown in Fig. 12(b). To determine the location of the centroidal
axis, we require that the moment associated with the internal stress resultants
vanish. Summing moments about the y-axis and equating to zero (SMy=0),
we obtain:
ð2Þ½ðNxx jw ÞðhÞðzb tbf h=2Þ½ðNxx jbf ÞðbÞðzb tbf =2Þ½ðNxx jtf ÞðbÞ
ðzb tbf h ttf =2Þ ¼ 0
Substituting [Eqs. (19a), (19b), (19c)] and solving for distance zb, we find:
2 3
ex ex ex
2A E
1 6 w xx w ð h þ 2t Þ þ t A E þ A E ð2t þ 2h þ t Þ
bf

bf bf xx
bf
tf xx
tf
b tf
7
zb ¼ 4 ex ex ex 5ð20aÞ
2 2A E þ A E þ A E
w xx bf xx tf xx
w bf tf

where
Aw = tw h = cross-sectional area of the web laminate.
Abf = tbf b = cross-sectional area of the bottom flange laminate.
Atf = ttf b = cross-sectional area of the top flange laminate.
Noting that the effective extensional Young’s moduli can be written as
ex ex ex
E xx Aw ¼ 1=tw aw
11, E xx Abf ¼ 1=tbf a11, and E xx Atf ¼ 1=ttf a11, this result can also
bf tf
be expressed as:
( tf )
1 2habf 11 a11 ðh þ 2tbf Þ þ ba11 tbf a11 þ a11 ð2tbf þ 2h þ ttf Þ
tf w bf
zb ¼ ð20bÞ
11 a11 þ ba11 ða11 þ a11 Þ
2habf
2 tf w bf tf
Equations (20a) and (20b) gives the distance from the centroid to the bottom
surface of the composite box-beam. If the bottom and top flange laminates are
ex ex
identical (i.e., if Abf=Atf, tbf=ttf, and E xx Abf ¼ E xx Atf ), then Eqs. (20a) and
(20b) reduces to zb=tbf+h/2, as would be expected.
Let us now determine the effective axial rigidity of the composite box-
beam. Since a uniform axial strain (exx) is induced, the stress resultants present
in each segment of the beam to the axial strain are given by:
exx ex
Nxx jw ¼ w ¼ tw E xx w exx
a11
exx ex
Nxx bf ¼ bf ¼ tbf E xx bf exx ð21Þ
a
11
exx ex
Nxx tf ¼ tf ¼ ttf E xx tf exx
a11
b
Also, the total load applied to the beam as a whole (Nxx ) must equal the
sum of the forces acting in each beam segment:
Nbxx ¼ 2ðNxx jw ÞðhÞ þ ðNxx jbf ÞðbÞ þ ðNxx jtf ÞðbÞ
Define the effective stress applied to the beam as the force applied to the beam
divided by the area of the beam cross section:

Nbxx 2ðNxx w ÞðhÞ þ ðNxx bf ÞðbÞ þ ðNxx tf ÞðbÞ
rbeam ¼ ¼ ð22Þ
Abeam 2Aw þ Abf þ Atf
Substituting Eq. (21) into Eq. (22), we find:
2
ex ex ex 3
2Aw E xx þ Abf E xx þ Atf E xx
rbeam ¼ exx 4 w bf tf 5
2Aw þ Abf þ Atf

We can now define the effective Young’s modulus of the beam as:

ex ex ex
2Aw E xx þ Abf E xx þ Atf E xx
beam
E xx ¼ w bf tf
ð23Þ
2Aw þ Abf þ Atf
The effective axial rigidity of the composite box-beam is simply the effective
beam Young’s modulus multiplied by the cross-sectional area:

beam ex ex ex
AE xx ¼ 2Aw E xx þ Abf E xx þ Atf E xx ð24aÞ
w bf tf
Alternatively, the effective axial rigidity of the box-beam can be written as:
beam 2h b b
AE xx ¼ w þ bf þ tf ð24bÞ
a11 a11 a11
If the bottom and top flange laminates are identical (i.e., if Abf=Atf, tbf=ttf,
ex
and E xx jbf=E xx
ex
jtf), then Eqs. (24a) and (24b) reduce to:

ex ex h b
AE beam ¼ 2 Aw E xx þ Abf E xx ¼ 2 w þ bf
w bf a11 a11
Next, consider the effective flexural rigidity of a composite box-beam.
Recall that the flexural rigidity of a beam is defined under a condition of pure
bending. A side view of a composite box-beam subjected to pure bending is
shown in Fig. 13(a). Since the beam cross section is symmetric about the x–z
plane, the long x-axis of the beam must deform into a circular arc with a
radius of curvature rxx, as shown. Hence, cross sections of the box-beam that
are initially plane and perpendicular to the axis of the beam must remain plane
and perpendicular to the deformed axis of the beam following loading. These
are precisely the same conditions encountered in pure bending of a rectan-
gular composite beam with ply orientation parallel to the plane of loading
(discussed in Sec. 5.2), as well as in isotropic beams with symmetric rectan-
gular cross sections. Axial normal strains vary linearly with z and are given by:
exx ¼ zjxx ð25Þ
where curvature jxx=1/rxx. We assume that the axial strain induced at any
position z is uniform across the width of the web and flange laminates; that is,
we assume that exx is not a function of y.
Our first objective is to determine the location of the neutral axis. That
is, we wish to determine either distance zt or zb, previously shown in Fig. 11. In
the following derivation, we will determine distance zb, the distance from the
neutral axis to the bottom surface of the beam. Note that this is an arbitrary
decision, and a comparable derivation based on distance zt can easily be
developed.

Figure 13 Side view of a composite box-beam subjected to pure bending. (a)
Composite box-beam deformed into a circular arc of radius rxx. (b) Distribution
of internal stress resultants NxxAw, NxxAbf, and NxxAtf.

First consider the web laminates. We note that for the web laminate, the
ply interfaces are parallel to the x–z plane. Also, we have assumed that the
strain exx is not a function of y. Hence, Eq. (25) indicates that the midplane
strains induced in the web laminates vary with z:

eoxx w ¼ zjxx ð26Þ
Since the web laminates are symmetric, the stress resultant induced at any
point in the webs, Nxx w , is given by:

eoxx w ¼ aw
11 Nxx w ð27Þ

Equating Eqs. (26) and (27) and solving for Nxx w , we have:

jxx ex
Nxx w ¼ z w ¼ ztw E xx w jxx
a11

This result shows that Nxx w varies linearly with z over the web laminate, as
implied in Fig. 13(b). The axial load induced over an incremental strip of
width dz is:

dNw xx ¼ Nxx w dz

Hence, the total axial load that supported the web, Nbxx w , is given by:

m m
zb tbf zb tbf
b
ex
Nxx ¼ Nxx w dz ¼ tw E xx w jxx zdz
zb htbf zb htbf
Evaluating this integral, we find:

ex
tw E xx jxx h i
b
Nxx ¼ w
ðzb tbf Þ2 ðzb h tbf Þ2 ð28Þ
w 2
Now consider the bottom flange laminate. Referring to Fig. 11, note that the
midplane of the bottom flange is located at z = zb–tbf/2. Substituting this
value into Eq. (25), the strain induced at the midplane of the bottom flange
laminate is given by:
tbf

eoxx ¼ zb jxx ð29Þ
bf 2
Since the bottom flange laminate is symmetric, the stress resultant induced in
the bottom flange, Nxx bf , is:

Nxx
o
exx ¼ a11 Nxx ¼
bf bf
ex
ð30Þ
bf bf
tbf E xx
bf

Equating Eqs. (29) and (30) and solving for Nxx bf , we find:
tbf
ex
Nxx ¼ zb tbf E xx jxx ð31Þ
bf 2 bf

This result gives the value of Nxx bf induced at the midplane of the bottom
laminate, shown
in Fig. 13(b). The total axial force supported
by the bottom
flange, Nbxx bf, is obtained by multiplying stress resultant Nxx bf by the width of
the bottom flange, b:
tbf
ex
Nbxx ¼ bNxx ¼ zb Abf E xx jxx ð32Þ
bf bf 2 bf
Note that the cross-sectional area of the bottom flange, Abf=btbf, has been
used in Eq. (32).
Finally, we must determine the total axial force supported by the top
flange. From Fig. 11, we see that the midplane of the top flange laminate exists
at z=(zb–tbf–h–ttf/2). Based on this value, expressions analogous to Eqs. (28)–
(32) can be obtained for the top flange laminate. We find that the total axial
load supported by the top flange is given by:
ttf
ex
Nbxx ¼ zb tbf h Atf E xx jxx ð33Þ
tf 2 tf
All equations necessary to determine the position of the neutral axis

have now been developed. The box-beam is modeled as four rectangular lam-
inates (two web laminates, the bottom flange laminate, and the top flange
laminate) and is in pure bending. Hence, the axial loads present in all four
laminates must sum to zero:

2Nbxx þ Nbxx þ Nbxx ¼ 0
w bf tf
Substituting Eqs. (28), (32), and (33) and solving for zb, we obtain:
2 3
ex ex ex
16 2A E ðh þ 2t Þ þ t A E þ A E ð2t þ 2h þ t Þ
w xx bf

bf bf xx
bf
tf xx
tf
b tf
7
zb ¼ 4 w
ex ex ex
5 ð34Þ
2 2A E þ A E þ A E
w xx bf xx tf xx
w bf tf
Equation (34) is identical to Eq. (20a). Hence, we conclude that in pure

bending, the neutral axis passes through the centroid of the beam cross
section. This same conclusion holds true for isotropic beams in pure bending.
Now that the position of the neutral axis has been identified, the
effective flexural rigidity of the composite box-beam can be determined. First,

consider the moment supported by a web laminate. The incremental moment
associated with the force dNbxx w located distance z from the neutral axis is
given by:

ex
dMbxx ¼ dNbxx z ¼ tw E xx jxx z2 dz
w w w
The total moment supported by a web laminate is then given by:

m
zb tbf
ex
Mbxx ¼ tw E xx jxx z2 dz
w zb htbf w
Evaluating this integral, we find that the moment supported by a web

0 1
ex
tw E xx jxx h i

Mbxx ¼ @ w A ðzb tbf Þ3 ðzb h tbf Þ3 ð35Þ
w 3
Now consider the moment supported by the bottom flange laminate.

Since the midplane of the bottom flange laminate is not located at the neutral
axis of the overall beam cross section, the flange laminate will experience both
a midplane strain and a midplane curvature. The beam cross section must
remain plane and, consequently, the curvature experienced by the bottom
flange is identical to that experienced by the beam as a whole: jxx bf = jxx. The
moment resultant associated with this curvature is:

fl
Mxx ¼ jxx =d11
bf
¼ t3bf E xx jxx =12
bf bf
This moment resultant contributes to the total moment supported by the

bottom flange. In addition, the stress resultant induced Nxxjbf also contributes
to the moment supported by the bottom flange and may be of equal or greater
importance. The bottom flange laminate is located at z = (zb–tbf/2). Sub-
stituting this coordinate into Eq. (25), we find that the midplane strain
induced in the bottom flange laminate is

eoxx ¼ ðzb tbf =2Þjxx
bf

This midplane strain is related to stress resultant Nxx bf according to:

eoxx ¼ abf
11 N xx
bf bf

Equating and solving for Nxx bf , we find:

jxx ex
Nxx ¼ ðzb tbf =2Þ bf ¼ ðzb tbf =2Þ tbf E xx jxx
bf a11 bf
This stress resultant acts at a distance

(zb–tbf/2) from the neutral axis, and so
the moment contributed by N is obtained by multiplying by this distance.
xx bf
Both Nxx bf and Mxx bf are resultants (i.e., loads per unit length) and must be
multiplied by the width of the flange to determine the bending moment. Based
on the above discussion, the total moment supported by the bottom flange
laminate is:
2 3
btbf E xx
fl

ex
Mbxx ¼ 4ðzb tbf =2Þ2 btbf E xx þ bf 5
jxx
bf bf 12
This expression can be simplified by noting that the area of the bottom flange
is Abf = btbf and, further, that the area moment of inertia of the bottom flange
(taken about the local midplane of the bottom flange) is Ibf=btbf3/12:
h i
b 2 ex fl
Mxx ¼ ðzb tbf =2Þ Abf E xx þ Ibf E xx jxx ð36Þ
bf bf bf
In passing, it might be anticipated that Eq. (36) could be further simplified by

expressing the area moment of inertia of the bottom flange laminate about the
neutral axis of the beam through application of the parallel axis theorem. It
turns out that this is not the case. That is, recall that through the application of
the parallel axis theorem, the moment of inertia of the bottom flange laminate
about the neutral axis of the beam cross section is given by:

Ibf ¼ ðzb tbf =2Þ2 Abf þ Ibf
na
Comparing this expression with Eq. (36), it is seen that the parallel axis
theorem cannot be used to advantage because the effective extensional
ex
modulus of the bottom flange, Exx bf , differs from the effective flexural
fl
modulus of the bottom flange, E xx bf
.
Finally, we consider the moment supported by the top flange. Following
an equivalent process, we conclude that the moment support by the top flange
h i
b 2 ex fl
Mxx ¼ ðzb tbf h ttf =2Þ Atf E xx þ Itf E xx jxx ð37Þ
tf tf tf

The total moment applied to the beam, Mob, must equal the sum of the
moment components supported by the two web laminates, the bottom flange
laminate, and the top flange laminate:

Mbo ¼ 2Mbxx þ Mbxx þ Mbxx
w bf tf
Also, by definition, the effective flexural rigidity of the box-beam is related to

Mob according to:
Mbo
EI ¼
jxx
Substituting Eqs. (35)–(37), the effective flexural rigidity of a composite box-
beam is given by:

ex
2tw E xx h i
EI ¼ w
ðzb tbf Þ3 ðzb h tbf Þ3
3

ex
þ ðzb tbf =2Þ2 Abf E xx ð38Þ
bf

ex fl fl
þðzb tbf h ttf =2Þ2 Atf E xx þ Ibf E xx þ Itf E xx
tf bf tf
If a box-beam is considered for which the bottom and top flange laminates
ex
are identical (i.e., if Abf = Atf, tbf = ttf, and E xx bf = E xx tf ), then Eq. (38)
ex
reduces to:
2 3
ex
ðtbf þ hÞ2 Abf E xx
ex fl
EI ¼ 2Iw E xx þ 24 bf
þ Ibf E xx 5
w 4 bf
where Iw is the area moment of inertia of a web laminate, taken about the
neutral axis of the box-beam.
This concludes our analysis of a composite box-beam. To summarize,
we have restricted our analysis to composite box-beams that possess certain
material and geometric symmetries. Specifically, both the flange and web
laminates must be produced using a symmetric stacking sequence, and the
beam cross section must be symmetric about the plane of loading, the x–z
plane. This implies that the two web laminates must be produced using an
identical symmetric stacking sequence, such that the set of fiber angles
represented by the two web laminates is symmetric about the x–z plane.
The top and bottom flange laminates must also be symmetric, but they need
not be identical. Therefore the beam cross section need not be symmetric
about the y–z plane. Having defined a beam cross section in this way, then the

location of the centroid (which also defines the location of the neutral axis in
pure bending) is given by Eqs. (20a) and (20b) or Eq. (34). The effective axial
rigidity of the box-beam is given by Eqs. (24a) and (24b), and the effective
flexural rigidity is given by Eq. (38). As will be discussed in following sections,
once the effective axial and flexural rigidities of the composite box-beam have
been determined, then deflection of the beam caused by axial and/or trans-
verse loads can be determined using the methods traditionally employed to
study isotropic beams.
The general approach described above can also be applied to the other
thin-walled cross sections previously discussed in Sec. 3 and shown in Fig. 1. A
summary of results for these additional thin-walled cross sections is provided
in Tables 1–3.
Example Problem 1
Three box-beams are constructed using unidirectional graphite–epoxy pre-
preg tape with the material properties listed in Table 3 of Chap. 3 and ply
thickness of 0.125 mm. The first box-beam is shown in Fig. 14(a). In this case,
the web and flange laminates are produced using an identical 8-ply [0/90]2s
stacking sequence. The second beam is shown in Fig. 14(b). In this case,
additional plies are added to the top flange. Hence, both web laminates and
the bottom flange laminate are produced using an 8-ply [0/90]2s stacking
sequence, but the top laminate is produced using a 20-ply [(0/90)2s/F45]s
stacking sequence. Finally, the third beam is produced using [0/90]2s web
laminates and [(0/90)2s/F45]s top and bottom flange laminates. In all three
cases, the overall beam width and depth is 30 and 50 mm, respectively.
Determine the position of the centroid and the effective axial and flexural
rigidities for:
(a) The beam shown in Fig. 14(a).
(b) The beam shown in Fig. 14(b).
(c) The beam shown in Fig. 14(c).
Solution
Part (a). Since both webs and flanges are produced using an 8-ply stacking
sequence, laminate thicknesses are:
tw ¼ tbf ¼ ttf ¼ 8ð0:125 mmÞ ¼ 1 mm
The length of the bottom and top flange laminates equals the overall width
of the beam (30 mm). The length of the web laminates is:
h ¼ 50 mm tbf ttf ¼ 50mm 2ð1mmÞ ¼ 48mm

Table 1 Centroid Location of Some Thin-Walled Composite Beams with
Symmetric Web and Flange Laminates

Table 2 Effective Axial Rigidities of Some Thin-Walled Composite Beams
with Symmetric Web and Flange Laminates

Table 3 Effective Flexural Rigidities of Some Thin-Walled Composite Beams
with Symmetric Web and Flange Laminates
Figure 14 The composite box-beams considered in Example Problem 1 (not to
scale). (a) Composite box-beam with identical stacking sequence in all web and
flange laminates. (b) Composite box-beam with additional plies added to the top
flange laminate. (c) Composite box-beam with additional plies added to both top
and bottom flange laminates.
We can now calculate the area properties of the web and flange laminates:
Aw ¼ ðhÞðtw Þ ¼ ð48 mmÞð1 mmÞ ¼ 48 106 m2
Abf ¼ Atf ¼ ðbÞðtbf Þ ¼ ð30 mmÞð1 mmÞ ¼ 30 106 m2
ðbÞðtbf Þ3 ð30 mmÞð1 mmÞ3
Ibf ¼ Itf ¼ ¼ ¼ 2:5 1012 m4
12 12
The effective moduli for the web and flange laminates are calculated using the
process described in Sec. 9.1 of Chap. 6 and are found to be:

ex ex ex
E xx ¼ E xx ¼ E xx ¼ 90:4 Gpa
w bf tf

fl fl
E xx ¼ E xx ¼ 120 Gpa
bf tf
From Table 1, for a composite box-beam, the location of the centroid is given
by:
2 3
ex ex ex
16 2A E ðh þ 2t Þ þ t A E þ A E ð2t þ 2h þ t Þ
w xx bf

bf bf xx
bf
tf xx
tf
bf tf
7
zb ¼ 4 w
ex ex ex
5
2 2Aw E xx þ Abf E xx þ Atf E xx
w bf tf
Substituting all known values and completing the indicated calculations, we

find:
zb ¼ 25 mm
This result is as anticipated since the bottom and top flange laminates are
identical and the total depth of the beam is 50 mm.

From Table 2, for a composite box-beam, the effective axial rigidity is:

ex ex ex
ðAE Þ ¼ 2Aw E xx þ Abf E xx þ Atf E xx
w bf tf
Substituting all known values and completing the indicated calculations,

we find:
ðAE Þ ¼ 14:1106 N
From Table 3, for a composite box-beam, the effective flexural rigidity
is:

ex
2tw E xx h i
ðIE Þ ¼ w
ðzb tbf Þ3 ðzb h tbf Þ3
3

bf ex
þ ðzb tbf =2Þ2 Abf E xx þ ðzb tbf h ttf =2Þ2 Atf E xx
tf

fl fl
þIbf E xx þ Itf E xx
bf tf
Substituting all known values and completing the indicated calculations,

we find:

IE ¼ 4:92 103 N m2
Part (b). The web and bottom flanges are produced using the same 8-ply
stacking sequence considered in Part (a), but now the top flange is a 20-ply
laminate. Adjusting our previous calculations, we find:
tw ¼ tbf ¼ 8ð0:125 mmÞ ¼ 1 mm

ttf ¼ 20ð0:125 mmÞ ¼ 2:5 mm
h ¼ 50 mm tbf ttf ¼ 50 mm 1 mm 2:5 mm ¼ 46:5 mm
Aw ¼ ðhÞðtw Þ ¼ ð46:5 mmÞð1 mmÞ ¼ 46:5 106 m2
Abf ¼ ðbÞðtbf Þ ¼ ð30 mmÞð1 mmÞ ¼ 30 106 m2
Atf ¼ ðbÞðttf Þ ¼ ð30 mmÞð2:5 mmÞ ¼ 75 106 m2
ðbÞðtbf Þ3 ð30 mmÞð1 mmÞ3
Ibf ¼ ¼ ¼ 2:5 1012 m4
12 12
ðbÞðttf Þ3 ð30 mmÞð2:5 mmÞ3
Itf ¼ ¼ ¼ 39:1 1012 m4
12 12

ex ex
E xx ¼ E xx ¼ 90:4 GPa
w bf

fl
E xx ¼ 120 GPa
bf

ex
E xx ¼ 83:3 GPa
tf

fl
E xx ¼ 94:0 GPa
tf
Based on the appropriate expression taken from Table 1, we find that the
centroid is located at:
zb ¼ 29:4 mm
Note that the increased thickness of the top flange laminate has had the effect
of moving the centroid 4.4 mm towards the top flange. The new axial and
flexural rigidities are:
ðAE Þ ¼ 17:4 106 N

ðIE Þ ¼ 6:35 103 N m2
The increased thickness of the top flange laminate has increased the effective
axial and flexural rigidities by 23% and 29%, respectively.
Part (c). For this cross section, the web laminates are produced using the
same 8-ply stacking sequence considered in Part (a), but now, both the bottom
and top flanges consist of a 20-ply laminate. We find:
tw ¼ 8ð0:125 mmÞ ¼ 1 mm
ttf ¼ tbf ¼ 20ð0:125 mmÞ ¼ 2:5 mm
h ¼ 50 mm tbf ttf ¼ 50 mm 2:5 mm 2:5 mm ¼ 45 mm
Aw ¼ ðhÞðtw Þ ¼ ð45 mmÞð1 mmÞ ¼ 45 106 m2
Abf ¼ Atf ¼ ðbÞðtbf Þ ¼ ð30 mmÞð2:5 mmÞ ¼ 75 106 m2
ðbÞðtbf Þ3 ð30 mmÞð2:5 mmÞ3
Ibf ¼ Itf ¼ ¼ ¼ 39:1 1012 m4
12 12

ex
E xx ¼ 90:4 GPa
w

ex ex
bf tf

fl fl
bf tf
Employing the appropriate expressions from Tables 1–3, we find:

zb ¼ 25:0 mm

ðAE Þ ¼ 20:6 106 N
ðIE Þ ¼ 8:43 103 N m2
Since the top and bottom flanges are identical, the centroid is centered in the
overall beam cross section, as was the case in Part (a). There is now a further
increase in effective rigidities. Compared to the values calculated in Part (a),
the increased thickness of the bottom and top flange laminates has increased
the effective axial and flexural rigidities by 46% and 71%, respectively.
7 STATICALLY DETERMINATE AND INDETERMINATE

AXIALLY LOADED COMPOSITE BEAMS
Deflections of composite beams subjected to axial loads (only) are considered
in this section. Beam problems may be divided into two categories: those
involving statically determinate beams and those involving statically indeter-
minate beams. For a statically determinate beam, all unknown reaction forces
can be determined through application of the equations of equilibrium. In
contrast, for a statically indeterminate beam, the reaction forces are deter-
mined through the equations of equilibrium and a consideration of beam
deflections.
A statically determinate composite beam is shown in Fig. 15(a). As in-
dicated, the beam is clamped at one end and subjected to three known axial
loads NBb , NCb , and NDb . In addition, the effective axial rigidity changes dis-
cretely at points B and C. In practice, a discrete change in effective axial ri-
gidity is readily accomplished by adding or removing a ply (or plies) from the
stacking sequence along the length of the beam. The internal force present
in each beam segment can be determined through the use of the free-body
diagrams shown in Fig. 15(b)–(d) and the appropriate equation of equilib-
rium (SFx = 0):

b
Nxx ¼ NBb þ NCb þ NDb
AB

b
Nxx ¼ NCb þ NDb
BC

Nxxb
¼ NDb
CD
The change in length of a beam caused by axial loading is called the

elongation, uo. The elongation induced within each segment of the beam
shown in Fig. 15 is given by:

b

Nxx LAB
uo ¼ AB
AB ðAE xx ÞAB

Figure 15 A statically determinate composite beam with changing effective
axial rigidity subjected to multiple axial load. (a) Composite Beam. (b) Free-body
diagram used to determine reaction force Rxx = NBb+NCb+NDb. (c) Free-body
b
diagram used to determine internal load in segment BC equals Nxx = NCb+NDb.
(d) Free-body diagram used to determine internal load in segment CD equals
b
Nxx = NDb.

b

Nxx LBC
uo ¼ BC
BC ðAE xx ÞBC

b

Nxx LCD
uo ¼ CD
CD ðAE xx ÞCD
where LAB, LBC, and LCD represent the length of beam segments AB, BC, and
CD, respectively. The elongation of the bar as a whole is simply the sum of the
elongation induced over each segment:

b
X Nxx Li
uo ¼ i
ð39Þ
i ðAE xx Þi
A statically indeterminate composite beam is shown in Fig. 16. In this

case, the beam is clamped at both ends and subjected to two known axial loads

Figure 16 A statically indeterminate composite beam subjected to two known
intermediate loads. (a) Original structure. (b) Free-body diagram.
NBb and NCb. An unknown reaction force is induced at each end of the bar,
Rbxx jA and RbxxjD. Note that both reaction forces have been assumed to act in
the +x-direction in Fig. 16. Enforcing the appropriate equation of equilib-
rium (SFx = 0), we find:

Rbxx þ NbB þ NbC þ Rbxx ¼ 0 ð40Þ
A D
Since Rxxb
jA and Rxx
b
jD are unknown, a second independent equation is
required in order to solve for the reaction forces. The remaining equations
of equilibrium cannot be used to provide the necessary second independent
equation since only axial loads are present. The second independent equation
is obtained by requiring that the beam satisfy known boundary conditions.
That is, since the beam is clamped at both ends, the total elongation of the
beam must be zero. By requiring that the total elongation is zero, we develop a
second independent equation, known as an equation of compatibility. Treating

reaction force Rbxx jD as a redundant force, applying Eq. (39), and equating the
total elongation to zero, we find:

b
NBb þ NCb þ Rxx LAB NCb þ Rbxx LBC

uo ¼ D
þ D
AD ðAE xx ÞAB ðAE xx ÞBC
ð41Þ
b
Rxx LCD
þ D
¼0
ðAE xx ÞCD
Equations (40) and (41) represent two independent equations in two un-
knowns, Rbxx jA and Rbxx jD. Once these equations are solved to determine the
reaction forces, then the axial force induced in any segment of the beam may
also be determined.
The analysis of statically determinate and indeterminate composite
beams will be illustrated in Example Problems 2 and 3, respectively.
Example Problem 2
Assume that the composite beam shown in Fig. 15 is a graphite–epoxy box-
beam, produced using a pre-preg system with the material properties listed in
Table 3 of Chap. 3 and a ply thickness of 0.125 mm. Over segment AC, the
beam has the cross section previously shown in Fig. 14(c). That is, the two web
laminates are produced using an 8-ply [0/90]2s stacking sequence, whereas the
top and bottom flange laminates are produced using a 20-ply [(0/90)2s/ F 45]s
stacking sequence. Over segment CD, the beam has the cross section
previously shown in Fig. 14(a). Thus, over this segment, both the web and
flange laminates are produced using an 8-ply [0/90]2s stacking sequence. The
length of each beam segment is as follows:
LAB ¼ 250 mm
LBC ¼ 500 mm
LCD ¼ 250 mm
Thus, the total length of the beam is 1.0 m. Finally, the applied loads are:
NBb ¼ 15 kN
NCb ¼ 10 kN
NDb ¼ 5 kN
Determine the total beam elongation produced by this loading condition.

Solution. The effective axial rigidities provided by the beam cross sections
involved were determined as a part of Example Problem 1 and found to be:

ðAE xx Þ ¼ ðAE xx Þ ¼ 20:6 106 N
AB BC

ðAE xx Þ ¼ 14:1 106 N
CD
The elongation can be determined through the application of Eq. (39):

b
X Nxx Li
uo ¼ i
i ðAE xx Þi

b b b
Nxx LAB Nxx LBC Nxx LCD
uo ¼ AB
þ BC
þ CD
ðAE xx ÞAB ðAE xx ÞBC ðAE xx ÞCD
ð30 000NÞð0:25mÞ ð15000NÞð0:50mÞ ð5000NÞð0:25mÞ

uo ¼ þ þ
20:6 106 N 20:6 106 N 14:1 106 N
uo ¼ 0:817 mm
Example Problem 3
Assume that the composite beam shown in Fig. 16 is a graphite–epoxy box-
beam, produced using a pre-preg system with the material properties listed in
Table 3 of Chap. 3 and a ply thickness of 0.125 mm. Over segment AC, the
beam has the cross section previously shown in Fig. 14(c). That is, the two web
laminates are produced using an 8-ply [0/90]2s stacking sequence, whereas the
top and bottom flange laminates are produced using a 20-ply [(0/90)2s/ F 45]s
stacking sequence. Over segment CD, the beam has the cross section
previously shown in Fig. 14(a). Thus, over this segment, both the web and
flange laminates are produced using an 8-ply [0/90]2s stacking sequence. The
length of each beam segment is as follows:
LAB ¼ 250 mm
LBC ¼ 500 mm
LCD ¼ 250 mm
b
Thus, the total length of the beam is 1.0 m. Finally,
loadsbNB = 15 kN and
b b
NC = 10 kN. Determine the reaction forces Rxx A and Rxx D .

Solution. The reaction forces are determined through solving Eqs. (40) and
(41). The effective axial rigidities provided by the beam cross sections involved
were determined as a part of Example Problem 1 and found to be:

ðAE xx Þ ¼ ðAE xx Þ ¼ 20:6 106 N
AB BC

ðAE xx Þ ¼ 14:1 106 N
CD
Using the values provided, Eq. (41) becomes:

25 kN þ Rbxx ð0:25 mÞ 10 kN þ Rbxx ð0:5 mÞ Rbxx ð0:25 mÞ
D
þ D
þ D
¼0
ð20:6 106 NÞ ð20:6 106 NÞ ð14:1 106 NÞ
Solving for the redundant force:

Rbxx ¼ 10:1 kN
D
b

The fact that Rxx D
is negative implies that it acts in the direction opposite to
that shown in Fig. 16(b).
b
The unknown reaction force Rxx A
can now be determined through the
application of Eq. (40):

Rbxx þ NbB þ NbC þ Rbxx ¼ Rbxx þ 15 kN þ 10 kN 10:1 kN ¼ 0
A D A

Rbxx ¼ 14:9 kN
A
b

Reaction Rxx A
is also negative, implying that it acts the direction opposite to
that shown in Fig. 16(b).
8 STATICALLY DETERMINATE AND INDETERMINATE

TRANSVERSELY LOADED COMPOSITE BEAMS
Deflections associated with bending of composite beams are considered in this
section. Three types of external loads that lead to beam bending are shown in
Fig. 17. A distributed load, q(x), is a load distributed over a length of the beam
and is specified in units of either Newtons/meter or pounds-force/inch (N/m
or lbf/in.). A point load, P, is a load acting transverse to the axis of the beam
and applied at a specific cross section. A point load is usually specified in units
of either Newtons or pounds-force (N or lbf ). Finally, a concentrated moment
(also called a couple), Mob , is a moment applied at a particular cross section

Figure 17 Illustration of external transverse loads that lead to beam bending.
and is specified in units of Newton-meters or pounds-force-in. (N m or lbf in.).

As in earlier sections, a superscript ‘‘b’’ has been used to differentiate between
a concentrated moment and a moment resultant, Mxx. In the following
discussion, the external beam loads shown in Fig. 17 will be referred to as
transverse loads.
In general, the external transverse forces applied to a beam induce
internal forces at all cross sections of a beam. To determine beam deflections,
we must determine these internal forces. In order to do so, the manner in
which the beam is supported must first be defined. The three most common
beam support conditions are illustrated in Fig. 18. A pinned support is one that
restrains the beam from deflection in the x- and z-directions, but does not
prevent rotation. In contrast, a roller support restrains the beam from
deflection in the z-direction, but does not restrain deflections in the x-
direction nor does it restrict rotation. Finally, a clamped (or fixed ) support
does not allow the beam to rotate or deflect in the x- or z-directions.
Some combinations of beam support conditions occur so frequently
that they are given special names. A simply supported beam is shown in
Fig. 19(a). In this case, one end of the beam is restrained by a pin support,
while the other is restrained by a roller support. A cantilever beam is shown in
Fig. 19(b) and is defined as a beam clamped at one end and free at the other. A
Figure 18 Common beam support conditions. (a) Pinned support. (b) Roller
support. (c) Clamped support.

Figure 19 Common combinations of beam support conditions. (a) Simply sup-
port beam. (b) Cantilevered beam. (c) Overhanging beam.
third common configuration is an overhanging beam, shown in Fig. 19(c). This

condition is similar to a simply supported beam except that the roller support
is at an interior cross section, such that one end of the beam is free.
Having specified the support conditions and transverse loading applied
to a beam, the next step is to determine the reaction forces. In essence, reaction
forces are external loads applied to the beam by the beam support(s).
Determination of these forces begins by enforcing the six equations of
equilibrium. Since we have assumed that all loads act within the x–z plane
(the plane of loading), three equations of equilibrium are satisfied automati-
cally: SFy = SMx = SMz = 0. In this section, we do not consider any axial
loads; therefore SFx=0 is also satisfied automatically. Consequently, only
two equations of equilibrium are available for use. Specifically, to remain in
equilibrium, it is required that all forces acting in the z-direction and all
moments acting about the y-axis sum to zero:
X
Fz ¼ 0
X
My ¼ 0
Beam problems may now be divided into two main categories: statically
determinate beams and statically indeterminate beams. For a statically
determinate beam, the unknown reaction forces are determined solely
through the application of the equations of equilibrium. Since we only have
two remaining equilibrium equations, it is apparent that a statically determi-
nate beam can have (at most) two unknown reaction forces. Free-body
diagrams for simply supported, cantilevered, and overhanging beams are
shown in Fig. 20. All reaction forces are shown in an algebraically positive
sense. In each case, only two unknown reaction forces are involved, and hence
each of these beams is statically determinate. In contrast, if a beam problem
involves three (or more) reaction forces, then the beam is statically indeter-

Figure 20 Free-body diagrams for common statically determinate beams. (a)
Simply supported beam. (b) Cantilevered beam. (c) Overhanging beam.
minate and the reaction forces are determined through the application of the
equations of equilibrium and consideration of beam deflections. An example
of a statically indeterminate beam is shown in Fig. 21. In this case, the beam is
clamped at one end, supported by a roller support at the other end, and
subjected to a known transverse point load P. A total of three unknown
reaction forces exist in this case (R(A) (B) b
z , Rz , and Mo ), as indicated in the
accompanying free-body diagram. The three unknown reaction forces cannot

Figure 21 Free-body diagram for a beam statically indeterminate to the first
degree.
be determined solely on the basis of equilibrium since only two equations of

equilibrium are available for use. The difference between the number of
unknown reaction forces and the number of available equations of equilib-
rium is called the degree of indeterminacy; the beam shown in Fig. 21 is
statically indeterminate to the first degree. One way to determine reaction
force in this case is to enforce the requirement that the beam z-deflection at the
roller support must equal zero, in addition to enforcing the two equations of
equilibrium. This leads to third independent equation, a so-called equation of
compatibility. Hence, together, the two equations of equilibrium and the
equation of compatibility represent three independent equations and can be
used to solve for the three unknown reaction forces.
In any event, assume for the moment that all reaction forces have been
determined for a beam of interest, either through application of the equations
of equilibrium or through an appropriate combination of the equations of
equilibrium and compatibility. The reaction forces and the applied transverse
loads represent external forces applied to the beam, and these induce internal
b
forces within the beam. Internal forces may consist of a shear force Vxx and
b b b
bending moment Mxx. In general, Vxz and Mxx both vary along the length of
the beam. Note that during the earlier discussion presented in Secs. 5 and 6, we
considered composite beams subjected to pure bending—in these earlier
b
sections, the internal bending moment Mxx did not vary along the length of
b
the beam. Further, in pure bending, no shear force exists: Vxz = 0. We now
wish to determine transverse beam deflections under more general loading
b b
conditions in which both Vxz and Mxx exist. This is often called a state of
b b
nonuniform bending. Rigorously speaking, both Vxz and Mxx contribute to
b
transverse beam deflections. However, the deflections due to Vxz are less
b
significant than the deflections due to Mxx and can usually be ignored. That is,
b
we will assume that the contribution of shear force Vxz to the total deflection
of composite beams is negligibly small. In essence, we assume Secs. 5 and 6

remain valid for the case of nonuniform bending, even though these equations
were developed assuming a state of pure bending. This same assumption is
made during the study of isotropic beams. The moment-curvature equation
for pure bending is:
b
Mxx
jxx ¼ ð42Þ
IE xx
The effective flexural rigidity (IE xx) of composite beams has been developed in
previous sections. For rectangular beams with plies orthogonal to the plane of
loading, (IE xx ) was developed in Sec. 5.1; for rectangular beams with plies
parallel to the plane of loading, (IE xx) was developed in Sec. 5.2; and for thin-
walled beams, (IExx) was developed in Sec. 6. Also, for a state of pure bending,
the midplane curvature jxx is related to transverse beam deflections according
to:
d2 w
jxx ¼ ð43Þ
dx2
where w = w(x) represents the z-deflection of the composite beam midplane.*
Combining Eqs. (42) and (43), we find:
d2 w Mb
2
¼ xx ð44Þ
dx IE xx
Most textbooks devoted to mechanics of materials (Refs. [1–5] for example)
describe several techniques to determine transverse beam deflections. These
include direct integration of the governing equation [i.e., direct integration of
Eq. (44)], the moment-area method, the method of superposition, the techniques
based on the use of discontinuity functions, or the use of Castigliono’s Theorem
(which is a technique based on energy methods). Two techniques will be briefly
discussed here: direct integration of the governing equations and the method
of superposition.
Direct integration of the governing differential equation is perhaps the
most straightforward approach. Once all external forces are known (includ-
ing reaction forces), then it is a simple matter to express the internal bending
b b
moment as a function of x: Mxx = Mxx (x). Hence, the beam deflection w(x)
is obtained by simply integrating Eq. (44) twice. Constants of integration
are determined by enforcing known beam support conditions (i.e., known
boundary conditions).
The method of superposition is based on the observation that Eq. (44)
is a linear differential equation with constant coefficients. Consequently, the
* For pure bending, Eq. (43) is exact and is derived in any text devoted to mechanics of
materials, including Refs. [1–5] for example.

deflection of a beam subjected to several different load components acting
simultaneously can be found by superimposing (i.e., adding together) the
deflections caused by each load component acting separately. In practice then,
tables of beam deflections caused by simple loading conditions and for
commonly encountered beam support conditions are generated, and the
deflections caused by a combination of two or more loads are obtained by
adding together the deflection caused by each load component. Tables for
deflections induced by common individual load components applied to
cantilevered and simply supported beams are provided in Appendix C, Tables
C.1 and C.2, respectively. Much more extensive tables, which provide
solutions for other types of loading and support conditions, are available in
engineering handbooks, for example, Ref. [7].
Three example problems will now be presented to illustrate the process
of determining composite beam deflections. The deflections of a statically
determinate composite beam will be obtained by direct integration of Eq. (44)
in Example Problem 4. The method of superposition is then applied to a
statically determinate composite beam subjected to a combination of loads in
Example Problem 5. Finally, the method of superposition is used to determine
the reaction forces and deflections for a statically indeterminate beam in
Example Problem 6.
Example Problem 4
A cantilevered graphite–epoxy box-beam subjected to a linearly increasing
distributed load qðxÞ ¼ ðqo =LÞx is shown in Fig. 22(a):
(a) Obtain analytical expressions giving the reaction forces and
transverse beam deflections.
(b) Assume that the beam has the box cross section previously shown
in Fig. 14(c) and length L=1 m. Obtain numerical values for re-
action forces and plot transverse beam deflections if qo=200 N/m.
Solutions
Part (a). A free-body diagram of the beam is shown in Fig. 22(b). Since the
beam is clamped, two reaction forces may exist at the left-hand end: Rz and
Mob . The first step is to determine the magnitude and algebraic sign of these
unknown reaction forces. The beam is statically determinate since only two
unknown reaction forces are present. The reaction forces are initially assumed
to act in an algebraically positive sense, as shown in Fig. 22(b).
To determine Rz and Mob, we first replace the distributed load q(x) with
the statically equivalent force qo L/2, located at x = 2L/3. This statically
equivalent force is shown in Fig. 22(b). By summing (a) forces in the z-
direction, (b) moments about the y-axis, and (c) equating both to zero (SFz =

Figure 22 The cantilevered beam considered in Example Problem 4. (a) A
cantilevered beam of length L subjected to a linearly increasing distributed load
q(x) = (qo/L)x. (b) Free-body diagram used to determined the unknown reaction
forces (distributed load qo).
SMy = 0), we obtain analytical expressions giving the reaction forces at the
clamped end:
qo L qo L2
Rz ¼ Mbo ¼ ðaÞ
2 3
Note that if qo>0, then the bending moment Mob is algebraically
negative and would act in the sense opposite to that shown in Fig. 22(b).
The internal shear and bending moment induced at any cross section located
at arbitrary position x can now be determined using the free-body diagram
shown in Fig. 23. Summing forces in the z-direction, we find:
qo 2
b
Vxz ðxÞ ¼ ðL x2 Þ
2L

Figure 23 Free-body diagram used to determine internal shear and bending
b b
moment, Vxz and Mxx , respectively, acting at an arbitrary cross section located
at position x.
b
A so-called shear force diagram is created by plotting Vxz (x), as shown in
Fig. 24. Similarly, summing moments about an axis passing through the left-
hand side of the free-body diagram shown in Fig. 23 and parallel to the y-axis,
we find:
qo
Mxxb
ðxÞ ¼ ð2L3 þ 3L2 x x3 Þ
6L
b
A so-called bending moment diagram is created by plotting Mxx (x), as shown
b
in Fig. 24. Substituting the above expression for the bending moment Mxx (x)
into Eq. (44), we find:
d2 w qo
¼ ð2L3 þ 3L2 x x3 Þ
dx2 6LðIE xx Þ
Integrating once results in:

dw qo 3L2 x2 x4
¼ 2L3 x þ þ C1
dx 6LðIE xx Þ 2 4
where C1 is a constant of integration. Since the beam is clamped at the left end,
the slope must equal zero there: dw/dx = 0 at x = 0. Enforcing this boundary
condition, we find that the constant of integration must equal zero:
C1 ¼ 0
Performing a second integration and simplifying, we obtain:
qo x2
w¼ ð20L3 10L2 x þ x3 Þ þ C2
120LðIE xx Þ

Figure 24 Shear and bending moment diagrams for the cantilevered beam
considered in Example Problem 4.
where C2 is a second constant of integration. Since the beam is clamped

at the left end, the deflection must also equal zero there: w = 0 at x = 0.
Enforcing this second boundary condition, we find that the constant of inte-
gration must again equal zero: C2 = 0 Hence, we find that beam deflections
are given by:
q o x2
w¼ ð20L3 10L2 x þ x3 Þ ðbÞ
120LðIE xx Þ
Note that this result is included in the list of solutions tabulated in Appendix
C, Table C.1.
Part (b). The effective flexural rigidity of the box-beam shown in Fig.
14(c) was determined as a part of Example Problem 1 and was found to be:
ðIE xx Þ ¼ 8:43 103 N m2

Substituting this and the other specified numerical values (L = 1 m, qo = 200
N/m) into expression (b) above, we find that the transverse deflections are
predicted to be (in millimeters):

wðxÞ ¼ ð0:1977x2 Þ 20L3 10L2 x þ x3
A plot of these beam deflections is shown in Fig. 25. Note that both the de-
flection and slope are zero at the clamped end, as dictated by this boundary
condition. A maximum deflection of 2.17 mm is predicted to occur at the
free end.
Example Problem 5
Suppose the cantilevered graphite–epoxy beam considered in Example Prob-
lem 4 is subjected to both a bending moment Mob applied at the free end and a
linearly increasing distributed load qðxÞ ¼ ðqo =LÞx , as shown in Fig. 26.
Perform the following:
(a) Use the method of superposition to obtain an analytical expression
giving the predicted deflection w(z).
(b) Plot numerical values of the predicted deflections, assuming the
following beam length and loading:
L¼1m
qo ¼ 200 N=m
Mbo ¼ 40 N m
Figure 25 Deflections predicted for the graphite–epoxy composite box-beam

considered in Example Problem 4.

Figure 26 Cantilevered graphite–epoxy beam considered in Example Problem 5.
Solution
Part (a). Deflections induced by a concentrated moment Mob and distrib-
uted load q(x) when acting separately are included in Table C.1. The deflec-
tions due to Mob acting alone are given by:
Mbo x2

wðxÞ ¼
Mbo 2ðIE xx Þ
The deflections due to the distributed load are given by:

qo x 2

wðxÞ ¼ ð20L3 10L2 x þ x3 Þ
qðxÞ 120LðIE xx Þ
Applying the principle of superposition, the beam deflections caused by both

load components acting simultaneously are simply the sum of these two
solutions:

Mbo x2 q o x2
wðxÞ ¼ wðxÞ b þ wðxÞ ¼ þ ð20L3 10L2 x þ x3 Þ
Mo qðxÞ 2ðIE xx Þ 120LðIE xx Þ
Part (b). The distributed load q(x) will tend to deflect the beam downwards
(i.e., in the positive z-direction), whereas Mob will tend to deflect the beam
upwards (in the negative z-direction).
Using the values specified for beam dimensions, material properties,
and loads, deflections are given (in millimeters) by:
wðxÞ ¼ ð0:1977x2 Þð20 10x þ x3 Þ 2:372x2
which can be rearranged as:
wðxÞ ¼ 1:582x2 1:977x3 þ 0:1977x5

A plot of predicted deflections is shown in Fig. 27. As before, both the
deflection and slope are zero at the left end, as required by the clamped
boundary condition. A locally maximum positive (downwards) deflection of
0.160 mm occurs at an axial position of x = 0.563 m. The globally maximum
deflection occurs at the free end, where a negative (upward) deflection of
0.198 mm occurs.
Example Problem 6
Consider a composite beam supported as previously shown in Fig. 21. (a)
Obtain analytical expressions for the reaction forces. (b) Assume that the
beam has the box cross section previously shown in Fig. 14(c) and length
L=1 m. Obtain numerical values for reaction forces and plot transverse beam
deflections if a = 3 L/4 = 0.75 m and P = 1000 N.
Solution
Part (a). The beam shown in Fig. 21 is statically indeterminate to the first
degree since there are three unknown reaction forces. Therefore one of the
unknown reaction forces must be selected to be a redundant force. Anyone of
ðAÞ ðBÞ
the reaction forces (Rz , Rz , or Mob) can be treated as the redundant force.
ðBÞ
For present purposes, Rz will be treated as the redundant.
This problem can be solved based on the method of superposition, as
summarized in Fig. 28. The first step is to determine the beam deflections that
would occur if the redundant force was removed. Removal of redundant
ðBÞ
reaction force Rz implies that the roller support at the right end is removed,
as shown in Fig. 28(b). The process of obtaining beam deflections when the
redundant force is removed is often called the reduced problem. From Appen-
Figure 27 Deflections predicted for the beam considered in Example Problem 5.

Figure 28 Summary of the method of superposition applied to the indetermi-
nate beam considered in Example Problem 4.
dix C Table C.1 we find that beam deflections over the range a V x V L
associated with the reduced problem are given by:
Pa2

wðxÞ ¼ ð3x aÞ
reduced 6ðIE xx Þ
The next step is to determine beam deflections that would be caused if

only the redundant force was applied, as shown in Fig. 28(c). From Table C.1,
ðBÞ
we find that the beam deflections associated with the redundant force Rz
are given by:
ðBÞ
Rz x2

wðxÞ ¼ ð3L xÞ
ðBÞ
Rz 6ðIE xx Þ
This expression is valid for any axial position 0 V x V L.

We now superimpose the deflections associated with the reduced and
redundant problems. In this case, we have, for a V x V L:
h i
1
wðxÞ ¼ wðxÞ þ wðxÞ ¼ Pa2 ð3x aÞ þ RzðBÞ x2 ð3L xÞ ðaÞ
reduced
ðBÞ
Rz 6ðIE xx Þ
We require that the total beam deflection equals zero at the right end
(at x=L) since the beam is supported by a roller at that point. Hence:
1 h i
wðx ¼ LÞ ¼ Pa2 ð3L aÞ þ RðBÞ 2
z ðLÞ ð3L LÞ ¼ 0
6ðIE xx Þ
which reduces to:

2RðBÞ
z L ¼ Pa ða 3LÞ
3 2
ðbÞ

Expression (b) represents an equation of compatibility. Only one such equa-
tion is necessary in this problem since the beam is indeterminate to the first
degree.
Next, we enforce the equations of equilibrium, SFz = 0 and SMy = 0,
resulting in (respectively):
RðAÞ ðBÞ
z þ Rz ¼ P ðcÞ
RðBÞ
z L þ Mo ¼ Pa
b
ðdÞ
where Eq. (d) was obtained by summing moments about the left end of the
beam (at x = 0).
Expressions (b), (c), and (d) represent three simultaneous equations in
terms of the three unknown reaction forces. They can be rewritten in matrix
form as:
2 38 ðAÞ 9 8 2 9
0 2L3 0 > > Rz > > >
> Pa ða 3LÞ >>
6 >
7> > >
> > >
>
6 7< ðBÞ = < =
61 1 0 7 ¼ P
6 7> z > >
R
>
4 5>
> >
> > > >
>
: b >
> ; > : >
;
0 L 1 Mo Pa
Solving this system of equations results in:

P
RðAÞ
z ¼ ða3 þ 2L3 3a2 LÞ ðeÞ
2L3
Pa2
RðBÞ
z ¼ ð3L aÞ ðf Þ
2L3
Pa
Mbo ¼ 2 ða2 þ 2L2 3aLÞ ðgÞ
2L
Part (b). Based on the specified numerical values, the reaction forces are:
ð1000NÞ h i
RðAÞ
z ¼ ð0:75mÞ 3
þ 2ð1mÞ 3
3ð0:75mÞ 2
ð1mÞ ¼ 367N
2ð1mÞ3
ð1000NÞð0:75mÞ2
RðBÞ
z ¼ ½3ð1mÞ ð0:75mÞ ¼ 633N
2ð1mÞ3
ð1000NÞð0:75mÞ h i
Mbo ¼ 2
ð0:75mÞ2 þ 2ð1mÞ2 3ð0:75mÞð1mÞ
2ð1mÞ
¼ 117Nm

ðAÞ ðBÞ
Note that the calculated values for Rz , Rz , and Mob are all algebraically
negative, which indicates that these reaction forces are acting in a sense
opposite to that shown in Fig. 21.
To plot beam deflections, note that expression (a), developed in the first
part of this example problem, gives the deflection over the range a V x V L.
Hence, we need an additional expression for beam deflections that is valid
over 0 V x V a. From Table C.1, we find that beam deflections over the range
0 V x V a associated with the reduced problem are given by:
Px2

wðxÞ ¼ ð3a xÞ
reduced 6ðIE xx Þ
As before, from Table C.1, we find that the beam deflections associated
with the redundant force R (B)
z are given by:
ðBÞ
Rz x2

wðxÞ ¼ ð3L xÞ
ðBÞ
Rz 6ðIE xx Þ
This result is valid for any axial position 0 V x V L. Superimposing these

two results, we obtain an expression for the total beam deflection, valid for
0 V x V a:
h i
1
wðxÞ ¼ wðxÞ þ wðxÞ ¼ Px2 ð3a xÞ þ RzðBÞ x2 ð3L xÞ ðhÞ
reduced
ðBÞ
Rz 6ðIE xx Þ
Figure 29 Deflections predicted for the beam considered in Example Problem 6.

Recall from Example Problem 1 that the effective flexural rigidity for this
beam is ( IE xx) = 8.43103 N m2. Substituting this value (and the other
numerical values specific to this problem) into expressions (a) and (h) and
plotting over the appropriate ranges, results in Fig. 29. As dictated by the
specified boundary conditions, the beam deflection is zero at either end of the
beam, and the slope of the beam is zero at the left end. A maximum positive
(downward) deflection of 0.94 mm is predicted to occur at the axial location
x = 0.64 m.
9 COMPUTER PROGRAM BEAM

The computer program BEAM has been developed to supplement the material
presented in 4 Secs. 5 Secs. 6 of this chapter. This program can also be down-
loaded at no cost from the following website: http://depts.washington.edu/
amtas/computer.html.
Program BEAM can be used to calculate the centroidal location,
effective axial rigidity, and effective flexural rigidity of a composite beam
with either a rectangular cross section or with any of the thin-walled cross
sections shown in Tables 1–3. The program prompts the user to input all
information necessary to perform these calculations. Properties of up to five
different materials may be defined. The user must input various numerical
values using a consistent set of units. For example, the user must input elastic
moduli for the composite material system(s) of interest. Using the properties
listed in Table 3 of Chap. 3 and based on the SI system of units, the following
numerical values would be input for graphite–epoxy:
E11 ¼ 170 109 Pa E22 ¼ 10 109 Pa v12 ¼ 0:30

G12 ¼ 13 10 Pa9
Since 1 Pa = 1 N/m2, all lengths must be input in meters. For example, ply
thicknesses must be input in meters (not millimeters). A typical value would
be tk = 0.000125 m (corresponding to a ply thickness of 0.125 mm). Similarly,
if an I-beam that involves a 50-mm-wide flange laminate was under consid-
eration, then the width of the flange must be input as 0.050 m.
If the English system of units was used, then the following numerical
values would be input for the same graphite–epoxy material system:
E11 ¼ 25:0 106 psi E22 ¼ 1:5 106 psi v12 ¼ 0:30
G12 ¼ 1:9 10 psi
6
All lengths would be input in inches. A typical ply thickness would be

tk = 0.005 in., and for an I-beam, a typical flange width would be 2.0 in.

Having determined the effective axial and flexural rigidities of a
composite beam of interest, then beam deflections caused by axial loading
or transverse loading can be calculated using the methods discussed in Secs. 7
and 8, respectively.
HOMEWORK PROBLEMS
Notes: Computer programs CLT and BEAM are used in the following
problems. These programs can be downloaded from the following website:
http://depts.washington.edu/amtas/computer.html.
1. A [(02/F30/F45/F60)3/902]s composite beam with rectangular cross

section is shown in Fig. 30. Assume this laminate is produced using a
graphite–epoxy pre-preg with a thickness of 0.125 mm and properties
listed in Table 3 of Chap. 3:
(a) Use the program CLT to determine numerical values of the

[abd] matrix for this stacking sequence.
(b) Use hand calculations to determine: (i) the effective axial
rigidity of the beam, (ii) the effective flexural rigidity of the
beam when ply interfaces are orthogonal to the plane of
loading [Fig. 30(a)], and (iii) the effective flexural rigidity of
the beam when ply interfaces are parallel to the plane of load-
ing [Fig. 30(b)].
(c) Use the program BEAM to determine the effective axial and
flexural rigidities of the beam and compare these results with
your hand calculations obtained in step (b).
2. Repeat Problem 1, except assume that the beam is produced using a

glass/epoxy pre-preg with a thickness of 0.150 mm.
Kevlar/epoxy pre-preg with a thickness of 0.125 mm.
4. Repeat Problem 1, except assume that the 0j, +30j, and 30j plies are
produced using a glass/epoxy pre-preg with a thickness of 0.150 mm,
whereas the remaining plies are produced using a graphite–epoxy pre-
preg with a thickness of 0.125 mm.
5. A composite I-beam is shown in Fig. 31. Assume the beam is produced
using a graphite–epoxy pre-preg with a thickness of 0.125 mm and
properties listed in Table 3 of Chap. 3. Also, the stacking sequences of
the top flange laminate, web laminate, and bottom flange laminate are
[(02/902/02)2/452/0/–452]s, [02/902/02]2s, and [(02/902/02)2/452/0/–452]s,
respectively (note that Fig. 31 implies that the web plies extend into the
top and bottom flange laminates).

Figure 30 Composite beam with rectangular cross section described in
Problem 1 (not all ply interfaces shown).
(a) Determine the centroid location, effective axial rigidity, and

effective flexural rigidity of the beam using hand calculation
and appropriate expressions from Tables 1–3. Determine the
effective elastic moduli involved using program CLT.
(b) Use the program BEAM to calculate the centroid location,
effective axial rigidity, and effective flexural rigidity of the
beam, and compare these results with your hand calculations
obtained in step (a).

Figure 31 Composite I-beam described in Problem 5 (not all ply interfaces
shown).

glass/epoxy pre-preg with a thickness of 0.150 mm.
Kevlar/epoxy pre-preg with a thickness of 0.125 mm.
8. Repeat Problem 5, except assume that the stacking sequences of the top
and bottom flange laminates are [(02/902/02)2/452/0/–452]s and [02/902/
02]2s, respectively.
9. Refer to Example Problem 4, part (b). Suppose the maximum deflection
of the cantilevered beam must be reduced by a factor of 2. That is, the
maximum deflection must be reduced to 1.085 mm (or less). Deflections
will be reduced by adding 0j plies to the top and bottom flange laminates
of the box-beam. The stacking sequence of the flange laminates will then
be of the type [(0/90)2s/F45/0n]s. Determine the value of n necessary to
reduce deflections to the desired level.
10. Refer to Example Problem 5, what maximum deflection occurs if a
negative bending moment is applied, i.e., if Mob = 40 N m?
REFERENCES
1. Gere, J.M.; Timoshenko, S.P. Mechanics of Materials, 4th Ed.; PWS Publishing
Co.: Boston, MA, ISBN 0-534-93429-3, 1997.
2. Craig, R.R. Mechanics of Materials. John Wiley and Sons: New York, NY, ISBN
0-471-50284-7, 1996.
3. Hibbeler, R.C. Mechanics of Materials, 4th Ed.; Prentice Hall: Upper Saddle
River, New Jersey, ISBN 0-13-016467-4, 2000.

4. Bedford, A.; Liechti, K.M. Mechanics of Materials. Prentice Hall: Upper Saddle
River, New Jersey, ISBN 0-201-89552-8, 2000.
5. Benham, P.P.; Crawford, R.J.; Armstrong, C.G. Mechanics of Engineering
Materials, 2nd Ed.; Longman Group Limited: Essex, ISBN 0-582-25164-8, 1996.
6. Timoshenko, S.; Goodier, J.N. Theory of Elasticity, 3rd Ed.; McGraw-Hill Book
Company: New York, NY, Section 124, ISBN 07-064720-8, 1970.
7. Roark, R.J.; Young, W.C. Formulas for Stress and Strain, 6th Ed.; McGraw-Hill
Book Company: New York, NY, ISBN 0-07-072541-1, 1989.
8. Swanson, S.R. Introduction to Design and Analysis with Advanced Composite
Materials. Prentice-Hall Inc.: Upper Saddle River, New Jersey, ISBN 0-02-
418554-X,1997.

9
The Governing Equations of Thin-Plate Theory
In earlier chapters we required that stress and moment resultants applied to

the edge of a thin plate were constant and uniform. In this and following
chapters we will relax this requirement. That is, we consider problems in
which loads vary along the edge(s) of a plate. To predict stress, strain, or
deflections induced in a thin plate by varying loads we must first develop
equations that govern the behavior of a thin composite plate. These are called
the ‘‘governing equations,’’ and are derived in this chapter. It turns out that
the governing equations for composite laminates with arbitrary stacking
sequences become quite lengthy. Therefore we limit our discussion to
symmetric composite laminates, which results in a substantial simplification
of the governing equations. This limitation is reasonable, as most composites
used in practice are symmetric.
Thin-plate theory as applied to isotropic plates was developed throughout the
19th and 20th centuries and is now very well established. Many textbooks
devoted to isotropic plates have been published; two typical examples are
Timoshenko and Woinowsky-Krieger (1) and Ugural (2). Although thin plate
theory is also applicable to anisotropic plates, relatively few texts devoted to
487

this topic have appeared. Two of the best known references describing the
application of thin plate theory to composite laminates (or other anisotropic
plates) are those by Whitney (3) and Turvey and Marshall (4).
A complete exposé of thin-plate theory as applied to composites is
beyond the scope of this textbook. Rather, the objective of this and following
chapters is to introduce the fundamental equations that govern the behavior
of thin composite plates and to present solutions to a few selected problems.
It is hoped that this discussion will prepare the reader for more advanced
studies in this area, including those presented in Whitney (3) and Turvey and
Marshall (4).
Thin plate problems may involve many different types of boundary
conditions. In this text, discussion is limited to rectangular composite lam-
inates with symmetric stacking sequences, subjected to simply supported
boundary conditions along all four edges. The range of problems considered
herein is therefore not as extensive as is presented elsewhere. The reader
interested in a more detailed discussion of symmetric or nonsymmetric
rectangular panels subject to alternate boundary conditions and/or a dis-
cussion of other panel shapes (e.g., elliptical composite panels) is referred to
Whitney (3) and Turvey and Marshall (4).
The rudiments of thin plate theory have already been applied in Chapter
6. Specifically, during the development of Classical Lamination Theory
(CLT) we made several assumptions, two of which were:
All plies within a thin laminate are subjected to a state of plane stress
(rzz = sxz = syz = 0), and
The Kirchhoff hypothesis is valid: a straight line that is initially
perpendicular to the midplane of a thin plate is assumed to remain
straight and perpendicular to the midplane after deformation.
These assumptions are central to classical thin plate theory, and we will
continue to make these two assumptions throughout the analyses presented
in this and following chapters. However, in Chapter 6 we also assumed that
the external loads applied to the laminate (i.e., stress and moment resultants
Nxx, Nyy, Nxy, Mxx, Myy, Mxy) were constant and uniformly distributed along
the edge of the plate. Because the stress and moment resultants applied to
the edge of the laminate were assumed to be uniform and constant, the
resultants induced at all interior regions of the laminate were also constant
and identically equal to the edge loads. Therefore there was no need to
distinguish between the resultants applied to the edge of a laminate and the
resultants induced at any interior point.
In reality, uniform edge loads rarely occur in practice. That is, the loads
applied to a composite laminate in a real structure typically vary over the
length and/or width of the laminate. In this chapter we will consider

conditions in which nonuniform resultants are applied to the edge of the plate.
Because the edge loads will now be allowed to vary, the stress and moment
resultants induced at internal regions of the laminate will also vary and do not
necessarily equal the stress and moment resultants applied along the edge of
the laminate. We must therefore be careful to distinguish between the stress
and moment resultants applied to the edge of the laminate and the resulting
stress and moment resultants induced at internal regions of the laminate.
We will also include an additional type of loading in our analysis.
Specifically, we will include the possibility that a distributed load acts per-
pendicular to the surface of the laminate. This transverse load can be vi-
sualized as a transverse pressure, and will be denoted q(x,y).
A formal mathematical definition of the stress and moment resultants
applied to the edge of the laminate will be given in Sec. 3. However, we will
introduce some of the nomenclature used to describe external edge loads here.
Examples of externally applied loads that vary over the edge of a laminate are
shown schematically in Figs. 1 and 2. As in earlier chapters, we assume the
laminate is rectangular with plate edges parallel to the x- and y-axes. A sketch
showing stress resultants Nxx acting on opposite edges of a thin laminate is
presented in Fig. 1a. Because Nxx is a load applied to the two plate edges that
are parallel to the y-axis, Nxx cannot vary as a function of x. Therefore along
the plate edge the externally applied resultant Nxx is either a constant or, at
most, a function of y only:
Nxx ¼ Nxx ðyÞ:
There is no reason to expect that an identical distribution of loading is present
on opposite sides of laminate. Hence we must distinguish between the
resultant Nxx( y) applied to the ‘‘negative’’ laminate edge (i.e., the edge whose
outward normal ‘‘points’’ in the negative x-direction) and the resultant
Nxx( y) applied to the ‘‘positive’’ laminate edge. Therefore the load applied
(x)
to the negative x-edge will be labeled Nxx ( y), whereas the load applied to
(+x)
the positive x-edge will be labeled Nxx ( y). Although not shown in Fig. 1a,
stress resultant Nxy and moment resultants Mxx and Mxy may also be applied
to the two laminate edges parallel to the y-axis. These additional edge loads
will be labeled:
Resultants acting on the Resultants acting on the

negative x-edge: positive x-edge:
(x) (+x)
Nxy ( y) Nxy ( y)
(x) (+x)
Mxx ( y) Mxx ( y)
(x) (+x)
Mxy ( y) Mxy ( y)

Figure 1 Illustration of varying stress resultants acting on opposite edges of a
(+)
thin plate. (a) Variation of stress resultants Nxx (y). (b) Variation of stress
(+)
resultants Nyy (x).
It is emphasized that the resultants applied to the edges parallel to the y-axis
are either constants or are functions of y only. This will become an important
point in following discussions.
In a similar manner, a stress resultant Nyy may be applied to the two
plate edges parallel to the x-axis, as shown in Fig. 1b. In this case the plate
edge is parallel to the x-axis, and hence the stress resultants applied along
these edges are functions of x only. Stress and moment resultants Nyx, Myy,

Figure 2 Illustration of varying transverse load q(x,y) acting over the surface of
a thin plate and perpendicular to the x–y plane and the resulting out-of-plane
shear stress resultants V (Fx) (Fy)
xz ( y) and V yz (x).
and Myx also act on these plate edges. As before, we must distinguish between
loads applied to the negative and positive y-edge:
Resultants acting on the Resultants acting on the

negative y-edge: positive y-edge:
N (y)
yy (x) N (+y)
yy (x)
N (y)
yx (x) N (+y)
yx (x)
M (y)
yy (x) M (+y)
yy (x)
(y) (+y)
M yx (x) M yx (x)
Because in our earlier analysis we assumed that stress resultants were

constant and uniformly distributed along the edge of the plate, there was no
need to distinguish between shear stress resultants acting on adjacent faces.
That is, an assumption throughout Chapters 6 and 7 was that Nxy = Nyx. We
can no longer make this assumption. For example, the variation of the shear
stress resultant over the finite length of the positive x-face of the laminate,
(+x)
Nxy ( y), is independent of the variation of the shear stress resultant applied
(+y)
over the finite length of the positive y-face, Nyx (x). Similarly, the variation
(+x)
of the twisting moment applied to the positive x-face, Mxy ( y), is now
(+y)
independent of the twisting moment applied to the positive y-face, Myx (x).

Exceptions to this statement occur at the four corners of the laminate. For
example, at the corner (x = a, y = b):
NðxyþxÞ ðbÞ ¼ NðyxþyÞ ðaÞ and MðxyþxÞ ðbÞ ¼ MðyxþyÞ ðaÞ

A distributed transverse force acting over the surface of the laminate,
q(x,y), is shown in Fig. 2. The transverse force q(x,y) has units of force per area,
such as N/m2 = Pascals or lbf/in2 = psi. Because q(x,y) is a distributed force
acting in the z-direction, shear stress resultants Vxz( y) and/or Vyz(x) must be
present along one or more edges of the laminate, so as to maintain static
equilibrium (i.e., to maintain AFz = 0). Vxz (Fx)
( y) and Vyz(Fy)
(x) are given by:
Z t=2 Z t=2
V ðxz
FxÞ
¼ sðxzFxÞ dz V ðyzFyÞ ¼ sðyzFyÞ dz ð1Þ
t=2 t=2
(Fx)
The reader should immediately object to the inclusion of q(x,y), Vxz ( y),
ðFyÞ
and V yz ðxÞ in our analysis. After all, they violate our assumption of plane
stress. That is, the presence of q(x,y) implies that stress rzz will be induced in
the plate, and furthermore Eq. (1) implies that shear stresses sxz and syz will
also be induced. These objections are entirely valid. If a transverse load q(x,y)
is applied then a 3-D state of stress is, in fact, induced in the plate. However, if
the plate is thin, then the magnitudes of the out-of-plane stresses are far lower
than the magnitudes of the in-plane stresses: rzz, sxz, syz<<rxx, ryy, sxy.
Hence, the stresses induced in the plate satisfy the plane stress assumption
approximately. The inclusion of the out-of-plane transverse load q(x,y) while
still invoking the plane stress assumption is a fundamental discrepancy in thin
plate theory. In the discussion to follow we will simply ignore this inconsist-
ency.
To summarize, a total of five types of externally-applied distributed
loads on each of the four edges of a rectangular laminate are allowed in thin-
plate theory: one normal stress resultant, one in-plane shear stress resultant,
one out-of-plane shear stress resultant, one bending moment resultant, and
one twisting moment resultant. All of these loads may vary along each
laminate edge. In addition, a transverse loading q(x,y) may be applied to
the surface of the laminate. A laminate subjected to all external edge loads
considered in thin plate theory is shown in Fig. 3.
Thus far we have described the stress and moment resultants that may
be specified along the edges of a rectangular laminate. Specified edge loading
is a type of boundary condition. However, in many practical instances the loads
applied along the plate edges are unknown and hence cannot be used as
specified boundary conditions. Rather, in many instances midplane displace-
ments along the plate edges are known. Specified midplane displacements

Figure 3 A thin plate subjected to stress resultants and transverse loading that
vary along the length and width of the plate.
represent a second type of boundary condition. Types of displacement

boundary conditions have already been discussed in Chapter 8. For example,
in beam theory it is common to specify that one (or both) end(s) of a beam is
(are) ‘‘clamped.’’ If a beam end is clamped then that end of the beam cannot
move up/down, and the slope of the beam midplane must equal zero at the
clamped end. Hence, when one specifies that a beam end is clamped, one has
specified something about the displacement field in the beam, rather than
something about the applied loading.
Similar displacement boundary conditions can also be defined for thin
plates. Hence to define a thin-plate problem of interest we must specify the
boundary conditions along each edge of the plate. The specified boundary
conditions may involve (a) the loads applied to each edge of the plate, (b) the
displacements imposed on each edge of the plate, or (c) some combination
thereof. Boundary conditions that are consistent with the assumptions of
thin-plate theory will be discussed in detail in Sec. 3.
The equations that govern the behavior of thin composite laminates will
be developed in the following two sections. First, the equations of equilibrium
will be developed in Sec. 2. Boundary conditions that are consistent with thin-
plate theory will then be developed in Sec. 3. These equations and boundary
conditions are quite general and govern the solution to any problem involving
a thin plate, including composite laminates. Unfortunately, it is often im-
possible to obtain exact solutions to the equations of equilibrium that also
satisfy the prevailing boundary conditions. Difficulties in obtaining exact
solutions occur for either isotropic plates or anisotropic composite plates,
and arise from two different sources. First, the boundary conditions encoun-

tered in a problem may preclude obtaining an exact solution. For example, an
exact solution can be found for a flat rectangular isotropic plate subjected to
transverse loading if all four edges are simply supported, but an exact solution
cannot be found for an identical flat rectangular isotropic plate subjected to
transverse loading if all four edges are clamped. Because boundary conditions
are independent of material properties, the difficulties in obtaining exact
solutions for certain boundary conditions are encountered for either isotropic
or anisotropic composite plates.
The second source of difficulty has to do with the anisotropic nature of
composites and is not encountered during the study of isotropic plates.
Specifically, exact solutions cannot be found for laminates that exhibit the
following couplings terms:
in-plane normal-shear coupling (i.e., laminates for which A16,
A26 p 0),
bending-twisting coupling (i.e., laminates for which D16, D26 p 0),
and/or
coupling between in-plane loads and out-of-plane twisting (i.e.,
laminates for which B16, B26 p 0).
All of these coupling terms exist for a composite laminate with an arbitrary
stacking sequence, and hence an exact solution cannot, in general, be found
for composite laminates with an arbitrary stacking sequence.
Still another complication associated with environmental factors exists
for nonsymmetric laminates. Recall from Chapter 6 that nonsymmetric lam-
inates will bend/warp if subjected to a change in temperature and/or moisture
content, because Bij, MijT, MijM p 0 for nonsymmetric laminates. For example,
following cure at an elevated temperature a nonsymmetric laminate is likely to
be warped at room temperature and may warp further upon exposure to
changes in humidity over time. It is for this reason that nonsymmetric lam-
inates are rarely used in practice. An analysis of a nonsymmetric laminate
must therefore account for the displacements due to externally applied loads
as well as displacements (e.g., warping) due to changes in temperature or
moisture content. Because the intention here is to provide a brief introduction
to thin plate theory as applied to composites, nonsymmetric laminates will not
be considered.
In summary then, the equations of equilibrium and boundary condi-
tions that will be developed in the following sections of this Chapter are valid
for any symmetric composite laminate. However, exact solutions to these
governing equations can only be found for certain combinations of stacking
sequences and boundary conditions. Some available exact solutions to the
governing equations will be discussed in Chapter 10. Fortunately, techniques
are also available that can be used to obtain approximate solutions for those

laminates and boundary conditions that do not admit exact solutions.
Although approximate, the accuracy of these methods is usually quite good
and are suitable for most engineering applications. These approximate
solution techniques will be discussed in Chapter 11.
2 THE EQUATIONS OF EQUILIBRIUM FOR SYMMETRIC

LAMINATES
The externally applied edge loads shown in Fig. 3 induce internal stress and
moment resultants. In general, these internal resultants vary throughout the
interior regions of the laminate. Ultimately, we wish to relate these varying
internal stress and moment resultants to the externally applied loads that
induce them. In order to do so we must first develop the equations of equi-
librium for a thin plate. In essence, the equations of equilibrium dictate how
internal stress and moment resultants may vary over the length and width of
the laminate such that static equilibrium is maintained.
The equations of equilibrium will be developed in Sec. 2.1 by applying
the equations of statics; that is, by requiring that the sum of all forces and
moments equal zero. As will be seen, the equations developed in this manner
are partial differential equations involving the internal stress and moment
resultants, transverse loading q(x,y), and the out-of-plane displacement field,
w(x,y). These equations will then be converted to a form more useful for
composite studies in Sec. 2.2. This will involve expressing the internal stress
and moment resultants in terms of elements of the [ABD] matrix and midplane
displacement fields, uo(x,y), vo(x,y), and w(x,y).
2.1 The Equations of Equilibrium Expressed in Terms of

Internal Stress and Moment Resultants, Transverse
Loading, and Out-of-Plane Displacements
Consider a thin laminate subjected to some combination of edge loads and/or
edge displacement that induce stress and moment resultants at internal
regions of the laminate. The laminate is assumed to be flat prior to application
of external edge loads or displacement. In particular, the laminate is assumed
to be symmetric and hence Bij =0. Also, because the laminate is symmetric it
will not warp prior to the application of external loads if subjected to a change
in temperature and/or moisture content (MijT = MijM = 0). An infinitesimal
element removed from an internal region of such a laminate is shown in
Fig. 4. For clarity, only the midsurface of the element is shown in Fig. 4. The
element thickness is actually equal to the total laminate thickness, t. The in-
plane dimensions are dx and dy. The in-plane normal and shear stress
resultants are shown in Fig. 4a, while the moment resultants, out-of-plane

Figure 4 Internal stress resultants, moment resultants, and transverse loading
acting on an infinitesimal element. (a) In-plane normal and shear stress result-
ants. (b) Moment resultants, out-of-plane shear stress resultants and trans-
verse loading.
shear resultants, and transverse loading are shown in Fig. 4b (it should be
understood that the resultants and transverse loads shown in Fig. 4a,b are
applied simultaneously to the element). Once again, it is important to realize
that the resultants shown are induced at an interior region represented by the
infinitesimal element and do not necessarily equal the external loads applied
along the edges of the laminate. A superscript (*) will be used to denote
resultants present at interior regions of the laminate. Also, because the
element is infinitesimal we do not need to distinguish between shear resultants
on adjacent faces, i.e., N*yx =N*xy .
All resultants are assumed to vary slightly across the infinitesimal
length and width of the element. For example, the stress resultant Nxx* acting
on the negative x-face of the element is assumed to vary slightly over distance

dx, such that a stress resultant N *xx+(BN *xx/Bx) dx is present on the positive
x-face of the element. The equations of equilibrium can be developed using
the free-body diagrams shown in Figs. 5 and 6 by applying the standard
equations of statics. That is, by requiring that the sum of all forces and
moments acting on the infinitesimal element equate to zero:
X X X X X X
Fx ¼ Fy ¼ Fx ¼ Mx ¼ My ¼ Mz ¼ 0:
Let us first sum forces in the x-direction, based on the free-body diagram
shown in Fig. 5. Only those resultants that contribute toward forces in the x-
direction are shown in this figure. The element has been deflected out of the
original x–y plane by the transverse loading. A view of the x–y plane is shown
in Fig. 5a, while a view of the x–z plane is shown in Fig. 5b. In the deflected
condition the left and right side of the element form angles a and (a+(Ba/Bx)
dx), respectively, with the original x–y plane. Consider the x-directed force
associated with Nxx* , acting on the left side of the element. Recall that the units
of a stress resultant are (force/length), so to convert Nxx * to a force we must
Figure 5 Free-body diagram of an infinitesimal element, showing only those re-

sultants that contribute toward forces in the x-direction. (a) x-y plane. (b) x-z
plane.

Figure 6 Free-body diagram of an infinitesimal element, showing only those
resultants that contribute toward forces in the y-direction: (a) x–y plane; (b) y–z
plane.
multiply it by the distance over which it acts, namely, by distance dy. Fig. 5b
shows that the line-of-action of Nxx* is not precisely parallel to the x-axis, due
to the out-of-plane deflection. The component of the force associated with
* that is parallel to the x-axis is therefore given by:
Nxx
ðN*
xx ÞðdyÞcosa
We now assume that angle a is small (less than about 10j or 0.1745 rad), such
that the small-angle approximation can be applied: cos a c1. With this
assumption the component of the force associated with Nxx * that is parallel
to the x-axis reduces to:
ðN*
xx ÞðdyÞ
In effect, invoking the small-angle approximation implies that we will ignore
the fact that Nxx
* is not precisely parallel to the x-axis during the summation of
forces in the x-direction. In exactly the same way, the x-directed forces acting
on the other three sides of the element are given by:
x-directed force acting on the positive x-face:

BNxx*
*þ
Nxx dx ðdyÞ
Bx
x-directed force acting on the negative y-face:
ðNxy
*ÞðdxÞ
x-directed force acting on the positive y-face:

BNxy*
*þ
Nxy dy ðdxÞ
By

Summing the force components acting on all four sides and equating to zero
(SFx =0):

BNxx
* BNxy
*
ðNxx*Þdy þ Nxx*þ dx dy Nxy * dx þ Nxy *þ dy dx ¼ 0
Bx By
Simplifying, we obtain:
* BNxy
BNxx *
þ ¼0 ð2aÞ
Bx By
Eq. (2a) is the first equation of equilibrium. It shows that if Nxx * does
indeed vary in the x-direction at an interior point in the plate, then N*xy must
also vary in the y-direction at a comparable rate, so that static equilibrium is
maintained.
A second free-body diagram showing only those stress resultants whose
line-of-action is parallel to the y-axis is shown Fig. 6. As before, the x–y plane
of the element is shown in Fig. 6a while the y–z plane is shown in Fig. 6b.
Summing forces in the y-direction, invoking the small angle approximation,
and equating to zero (SFy =0), we find:

BNyy* BNxy
*
Nyy * dx þ Nyy *þ dy dx Nxy * dy þ Nxy *þ dx dy ¼ 0
By Bx
which simplifies to:
BNyy
* BNxy*
þ ¼0 ð2bÞ
By Bx
Eq. (2b) is the second equation of equilibrium.
Let us now consider forces parallel to the z-axis. Due to out-of-plane
deflection of the plate midsurface the stress resultants shown in Figs. 5 and 6
have components in the z-direction. For example, from Fig. 5b the compo-
nent of the force associated with Nxx
* that is parallel to the z-axis is given by:
ðNxx
*ÞðdyÞsina
The small angle approximation (where a is expressed in radians) is:
Bw
sinacac
Bx
* is:
Hence, the z-directed component of the force associated with Nxx
Bw
ðNxx
*ÞðdyÞ
Bx

In exactly the same way, the z-directed forces acting on the other three
sides of the element shown in Fig. 5 are given by:
z-directed force acting on the positive x-face:

BNxx
* Bw B2 w
*þ
Nxx dx þ 2 dx ðdyÞ
Bx Bx Bx
z-directed force acting on the negative y-face:
Bw
Nxy
* ðdxÞ
Bx
z-directed force acting on the positive y-face:

BNxy
* Bw B2 w
*þ
Nxy dy þ dy ðdxÞ
By Bx BxBy
Adding these results, the resultants shown in Fig. 5 result in a component of

force in the z-direction given by:

Bw BN*xx Bw B2 w
ðN*xx Þ ðdyÞ þ N*xx þ dx þ 2 dx ðdyÞ
Bx Bx Bx Bx

Bw BN*xy Bw B2 w
N*xy ðdxÞ þ N*xy þ dy þ dy ðdxÞ
Bx By Bx BxBy
When the algebra indicated is completed, two higher-order terms appear:

BN*xx B2 w BN*xy B2 w
ðdxÞ2 dy and dxðdyÞ2
Bx Bx2 By BxBy
Neglecting these higher-order terms, the above expression simplifies to:

B2 w BN*xx Bw B2 w BN*xy Bw
N*xx 2 þ þ N*xy þ dxdy ðaÞ
Bx Bx Bx BxBy By Bx
Following an identical procedure, the stress resultants shown in Fig. 6

represent a force in the z-direction given by:

B2 w BN*yy Bw B2 w BN*xy Bw
N*yy 2 þ þ N*xy þ dxdy ðbÞ
By By By BxBy Bx By

The vertical shear resultants and transverse loading act directly in the z-
direction, as previously shown in Fig. 3b. Using a procedure similar to that
described above, the sum of these forces in the z-direction is:

BVxz
* BVyz*
½Vxz*dy þ Vxz*þ dx dy Vyz * dx þ Vyz *þ dy dx
Bx By
þ ½qðx; yÞdxdy

* BVyz
BVxz *
þ þ qðx; yÞ dxdy ðcÞ
Bx By
We are now ready to sum all forces in the z-direction. Adding expressions (a),
(b), and (c), equating the resulting sum to zero (SFz =0), and rearranging,
there results:

B2 w B2 w B2 w BNxx* BNxy * Bw
* 2 þ Nyy
Nxx * 2 þ 2Nxy * þ þ
Bx By BxBy Bx By Bx

BNyy * BNxy* Bw BVxz * BVyz *
þ þ þ þ þ qðx; yÞ ¼ 0
By Bx By Bx By
Notice that the terms within the two sets of parenthesis have been previously
shown to equal zero, in accordance with Eqs. (2a) and (2b). Hence these terms
may be dropped, and our third equation of equilibrium becomes:
B2 w B2 w B2 w * BVyz
BVxz *
*
Nxx þ N *
yy þ 2N *
xy þ þ þ qðx; yÞ ¼ 0 ð2cÞ
Bx2 By2 BxBy Bx By
Eqs. (2a), (2b), and (2c) represent the requirement that all forces in the x-, y-,
and z-directions, respectively, sum to zero.
Now consider moment equilibrium about the x-axis. A free-body dia-
gram showing the resultants that contribute to the moment about the x-axis is
shown in Fig. 7a. Summing moments and equating to zero (SMx =0), we
have:

BMyy * BMxy*
Myy* dx Myy *þ dy dx þ Mxy * dy Mxy *þ dx dy
By Bx

BVyz
* BVxz
* dy
þ Vyz *þ dy dxðdyÞ þ Vxz *þ dx dy
By Bx 2

dy dy
½Vxz *dy þ ½qðx; yÞdxdy ¼0
2 2

Figure 7 Free-body diagrams used to sum moments about the x-, y-, and z-
axes. (a) Infinitesimal element used to sum moments about the x-axiz. (b)
Infinitesimal element used to sum moments about the y-axis. (c) Infinitesimal
element used to sum moments about the z-axis.


BMyy* BMxy * BVyz
* 2 BVxz
* dx
dxdy dxdy þ Vyz
* dxdy þ dxðdyÞ þ
By Bx By Bx 2
qðx; yÞ
ðdyÞ2 þ dxðdyÞ2 ¼ 0
2
The last three terms in this equation contain the higher-order factor
(dy)2. Hence dropping the last three terms and simplifying, we obtain:
BMxy* BMyy*
*¼
Vyz þ ð3aÞ
Bx By
A free-body diagram showing the stress and moment resultants that
contribute to the moment about the y-axis is shown in Fig. 7b. Summing
moments (SMy = 0), we have:

BMxx* BMxy*
½Mxx
*dy Mxx *þ dx dy þ Mxy * dx Mxy *þ dy dx
Bx By

BVxz* BVyz
* dx
þ Vxz *þ dx dyðdxÞ þ Vyz *þ dy dx
Bx By 2

dy dx
Vyz* dx þ ½qðx; yÞdxdy ¼0
2 2
Upon completing the algebra indicated, a negligible higher-order term
appears (in this case dx2). Neglecting all terms that include this factor and
simplifying, we obtain:
BMxx* BMxy*
*¼
Vxz þ ð3bÞ
Bx By
Lastly, a free-body diagram showing the stress resultants that contrib-

ute to the moment about the z-axis is shown in Fig. 7c. Summing moments
(SMz =0), we obtain:

dy BNxx
* dy
ðNxx
*dyÞ Nxx *þ dx dy
2 Bx 2

BNxy* BNxy*
þ Nxy *þ dx dy dx Nxy *þ dy dx dy
Bx By

BNyy
* dx dx
þ Nyy *þ dy dx Nyy* dx ¼0
By 2 2

Simplifying:
* dxdy2 BNxy
BNxx * 2 BNxy
*
þ dx dy dxdy2 þ Nxy
* dx2 dy dxdy2
Bx 2 Bx By
BNyy
* dx2 dy
¼0
By 2
All terms in this equation contain a higher-order term (either dx2 or dy2 ).
Hence all terms on the left side of the equality are negligible, and the
summation of moments about the z-axis provides no new information.
In summary, we have developed five useful equations through a
summation of forces and moments. The first three, Eqs. (2a)–(2c), were
developed by requiring that the sum of forces in the x-, y-, and z-directions
sum to zero, while the fourth and fifth equations, Eqs. (3a) and (3b), were
developed by requiring that the sum of moments about the x- and y-axes sum
to zero. Note that Eqs. (3a) and (3b) relate the out-of-plane shear stress
resultants, Vxz* and Vyz* , to the moment resultants Mxx* , Myy* and Mxy * , and
that the shear stress resultants also appear in Eq. (2c). This means that only
three of the five equations we have developed are independent. The fact that only
three equations are independent is fundamentally due to the assumption of
plane stress, because this assumption implies there are only three independent
stress components. It is customary to substitute Eqs. (3a) and (3b) into Eq.
(2c), resulting in:
B2 w B2 w B2 w B2 Mxx* B2 Myy* B2 Mxy

*
*
Nxx þ N *
yy þ 2N *
xy þ þ þ 2
Bx 2 By 2 BxBy Bx 2 By 2 BxBy ð4Þ
þ qðx; yÞ ¼ 0
Eqs. (2a), (2b), and (4) represent three independent equations of

equilibrium for a thin plate subjected to a state of plane stress. If a thin
plate is in static equilibrium then the distribution of internal stress and mo-
ment resultants induced at any interior region of a rectangular plate must
satisfy these equations, which guarantee that SFi =0 and SMi =0. Because
these equations will be referenced repeatedly in the following discussion, the
three equations of equilibrium are collected here and renumbered for con-
venience:
* BNxy
BNxx *
þ ¼0 ð5aÞ
Bx By
BNyy
* BNxy*
þ ¼0 ð5bÞ
By Bx

B2 w B2 w B2 w B2 Mxx* B2 Myy* B2 Mxy
*
*
Nxx þ N *
yy þ 2N *
xy þ þ þ 2
Bx 2 By 2 BxBy Bx 2 By 2 BxBy ð5cÞ
þ qðx; yÞ ¼ 0
These results have been developed strictly on the basis of static

equilibrium. Note that we have not specified what mechanism caused the
internal resultants, and so Eqs. (5a)–(5c) are valid regardless of whether the
internal resultants were caused by external edge loading, by enforced edge
displacements, or by any combination thereof. Also, we have made no
assumptions regarding material properties. Therefore Eqs. (5a)–(5c) are valid
for any thin plate, including isotropic metallic plates or anisotropic composite
plates. We have, however, made one important assumption: we have assumed
that the angles a and h, defined by the slope of the deflected midplane and
shown in Figs. 5 and 6, are relatively small so that the small angle approx-
imation can be invoked.
The third equation of equilibrium, Eq. (5c), can be simplified for cases in
which the loads applied to the plate result in a maximum out-of-plane
deflection less than about half the laminate thickness: w(x,y)jmax< t/2. In
such a case the midplane slopes (i.e., angles a and h) are exceedingly small and
consequently the z-directed component of the in-plane stress resultants (Nxx * ,
* , and Nxy
Nyy * ) can be neglected. If the in-plane resultants are ignored during
the summation of forces in the z-direction [setting (Nxx * dy)Bw/Bx =0, for
example] and the preceding derivation is repeated, the equations of equili-
brium simplify as follows:
* BNxy
BNxx *
þ ¼0 ð6aÞ
Bx By
BNyy
* BNxy*
þ ¼0 ð6bÞ
By Bx
B2 Mxx* B2 Myy* B2 Mxy
*
þ þ2 þ qðx; yÞ ¼ 0 ð6cÞ
Bx 2 By 2 BxBy
Eqs. (6a)–(6c) are valid if the maximum out-of-plane displacement is less than
about half the plate thickness: w(x,y)jmax< t/2.
2.2 The Equations of Equilibrium Expressed in Terms of the

[ABD] Matrix, Transverse Loading, and Midplane
Displacement Fields
In Chapter 6 we related stress and moment resultants to midplane strains and
curvatures, temperature changes, and changes in moisture content. This

relationship is summarized by Eq. (44) in Chapter 6. We have since limited our
discussion to symmetric laminates (Bij = MTij =MijM) and are now using the
superscript ‘‘*’’ to denote internal stress and moment resultants. Conse-
quently, Eq. (44) is modified slightly to become:

o T M
N* A 0 e N N
¼ ð7Þ
M* 0 D j 0 0
In this equation:
N* and M* represent internal stress and moment resultants
associated with externally applied forces,
NT represents thermal stress resultants associated with a uniform
through-thickness change in temperature, DT, and
NM represents moisture stress resultants associated with a uniform
through-thickness change in moisture content, DM.
Recall that only uniform changes in temperature and/or moisture
content are considered in this text. Consequently, NT and NM should be
viewed as constants, i.e., they are not functions of x, y, or z.
Expanding the first of the six equations represented by Eq. (7), we have:
* ¼ A11 eoxx þ A12 eoyy þ A16 coxy NTxx NM

Nxx xx
We will now express the midplane strains and curvatures that appear in this
expression in terms of midplane displacement gradients. Before we do so,
however, the discussion presented in Sec. 14 of Chapter 2 should be reiterated.
In particular, a distinction between finite strains and infinitesimal strains was
made at that point. Basically, strains can be considered to be infinitesimal
when displacement gradients are small, so that the square of any displacement
gradient can be neglected; (Bw/Bx)2 c 0, for example. In this chapter we have
already assumed that the slope of the deflected plate midplane is small (i.e.,
Bw/Bx and Bw/By have already been assumed to be small), which allowed us
to apply the small angle approximation. Consequently, we will continue to
treat strains as infinitesimal strains. As we will see, this assumption will
ultimately lead to the conclusion that in-plane displacement fields uo(x,y) and
vo(x,y) (as well as in-plane forces) are independent of the transverse load,
q(x,y). Rigorously speaking, this conclusion is incorrect. That is, if a thin plate
is subjected to a transverse loading then in-plane displacement fields and/or
in-plane forces will change, reflecting a dependence on transverse load.
However, if displacement gradients are small, then the changes in-plane
displacements or forces are also small and can usually be ignored. In effect,
the assumption of infinitesimal strains has eliminated (in a mathematical
sense) the coupling between transverse loads and in-plane displacements/

forces. Occasionally, this leads to predictions that are nonintuitive. This will
be pointed out at appropriate points in following sections (in particular, see
Sample Problem 2 in Sec. 4 of Chapter 10).
Although the interdependence between out-of-plane and in-plane dis-
placements is ignored throughout most of this text, there is one important
exception. This interdependence is included in Chapter 11, during formula-
tion of an approximate analysis technique known as the Ritz method. The
interdependence must be included there so as to obtain buckling predictions.
In any event, we now assume that strains are infinitesimal and can
therefore be expressed in terms of midplane displacement fields in accordance
with Eqs. 49(a)–(49f) of Chapter 2:
Buo Bvo Buo Bvo
eoxx ¼ eoyy ¼ coxy ¼ þ
Bx By By Bx
B2 w B2 w B2 w
Bx2 By2 BxBy
Therefore, the internal stress resultant Nxx* can be written:

Buo Bvo Buo Bvo
* ¼ A11
Nxx þ A12 þ A16 þ NTxx NM
Bx By By Bx xx

Given this result, the quantity BN * is easily obtained:
xx
Bx
2
BNxx
* B2 uo B2 vo B uo B 2 vo
¼ A11 þ A12 þ A16 þ
Bx Bx2 BxBy BxBy Bx2
Note that this quantity appears as a term in the first equation of equilibrium,
Eq. (5a). Also note that neither the thermal nor moisture stress resultant is
involved in this result, because as previously noted NT and NM are constants
and therefore BNTxx =Bx ¼ BNM xx =Bx ¼ 0. Following an analogous procedure,
the second term in Eq. (5a) is found to be:
2
BNxy* B2 uo B 2 vo B uo B 2 vo
¼ A16 þ A26 þ A66 þ
By BxBy By2 By2 BxBy
Adding these two results, Eq. (5a) can be written in terms of elements of the
[ABD] matrix and midplane displacements as follows:
B 2 uo B 2 vo B 2 uo B2 vo B2 vo
A11 þ ðA12 þ A66 Þ þ 2A16 þ A16 þ A26
Bx 2 BxBy BxBy Bx 2 By2
ð8aÞ
B2 uo
þA66 ¼0
By2

Following an identical procedure, the second and third equations of equilib-
rium [Eqs. (5b) and (5c)] can be written:
B 2 uo B2 uo B2 vo B 2 uo B 2 vo
A16 þ ð A 12 þ A66 Þ þ 2A 26 þ A26 þ A22
Bx2 BxBy BxBy By2 By2
ð8bÞ
B2 vo
þ A66 ¼0
Bx2

2
Buo Bvo Buo Bvo B w
A11 þ A12 þ A16 þ Nxx Nxx
T M
Bx By By Bx Bx2

2
Buo Bvo Buo Bvo B w
þ A12 þ A22 þ A26 þ NTyy NM
Bx By By Bx yy
By2
ð8cÞ

2
Buo Bvo Buo Bvo B w
þ 2 A16 þ A26 þ A66 þ Nxy Nxy
T M
Bx By By Bx BxBy
B4 w B4 w B4 w
D11 4D16 4D 26
Bx4 Bx3 By BxBy3
B4 w B4 w
2ðD12 þ 2D66 Þ D22 4 þ qðx; yÞ ¼ 0
Bx By
2 2 By
Eqs. (8a)–(8c) represent the equations of equilibrium for a symmetric

laminate in terms of the [ABD] matrix, transverse loading, and midplane
displacement fields. These are valid for any thin symmetric laminate as long as
out-of-plane displacements are not excessive. If out-of-plane displacements
are very small [i.e., if w(x,y)jmax< t/2] then the equations of equilibrium given
by Eqs. (6a)–(6c) are applicable, which become:
B 2 uo B 2 vo B 2 uo B 2 vo B 2 vo
A11 þ ð A 12 þ A66 Þ þ 2A 16 þ A16 þ A26
Bx2 BxBy BxBy Bx2 By2
ð9aÞ
B2 uo
þA66 ¼0
By2
B 2 uo B2 uo B2 vo B 2 uo B 2 vo
A16 þ ðA12 þ A66 Þ þ 2A26 þ A26 þ A22
Bx 2 BxBy BxBy By 2 By2
ð9bÞ
B 2 vo
þ A66 ¼0
Bx2

B4 w B4 w B4 w B4 w
D11 4D 16 4D 26 2 ðD 12 þ 2D 66 Þ
Bx 4 Bx3 By BxBy3 Bx2 By2
ð9cÞ
B4 w
D22 þ qðx; yÞ ¼ 0
By 4
3 BOUNDARY CONDITIONS
A summary of the material presented in Secs. 1 and 2 is as follows. We
consider a thin rectangular plate of in-plane dimensions a b, whose edges are
parallel to the x- and y- axes. The external edge loads considered in thin plate
theory were introduced in Sec. 1. A total of five types of externally applied
distributed loads may be present on each of the four edges of a rectangular
laminate: one normal stress resultant, one in-plane shear stress resultant, one
out-of-plane shear stress resultant, one bending moment resultant, and one
twisting moment resultant. Each of these loads is, at most, a function of the
coordinate direction tangent to the plate edge. In addition, a transverse
loading q(x,y) may be applied to the surface of the laminate. Fig. 3 provides a
summary of all externally applied loads considered in thin plate theory.
These externally applied edge loads induce a distribution of internal
stress and moment resultants at all interior regions of the plate. Generally,
internal stress and moment resultants are functions of both x and y and
vary throughout the plate. The distribution of internal stress and moment
resultants were investigated in Sec. 2.1 by considering free-body diagrams of
an infinitesimal element removed from an interior point within the plate.
Requiring that the sum of all forces and moments equate to zero resulted in
the equations of equilibrium, summarized as Eqs. (5a)–(5c). If the max-
imum out-of-plane displacement is less than half the laminate thickness,
then a simplified version of the equations of equilibrium is applicable,
summarized as Eqs. (6a)–(6c). A mathematically equivalent form for a
symmetric composite laminate was obtained in Sec. 2.2, where the equations
of equilibrium were written in terms of elements of the [ABD] matrix and
midplane displacement fields, summarized as Eqs. (8a)–(8c) and Eqs. (9a)–
(9c).
In this section we will formally define the boundary conditions of the
plate. That is, we wish to precisely define what conditions exist along each of
the four edges of the rectangular plate. Actually, we have already begun our
discussion of boundary conditions, in the sense that the external stress and
moment resultants that may be applied along the edges of the laminate were
described in Sec. 1. However, in many instances the loads applied along the
plate edges are unknown and hence cannot be used as specified boundary
conditions. Rather, the midplane displacements along the edges are known,
where displacements of the midplane in the x-, y, and z-directions are
denoted uo(x,y), vo(x,y), and w(x,y), respectively, as in earlier chapters.
Hence, two categories of boundary conditions can be defined. We can either
specify components of the edge displacements, or we can specify components
of the edge loads. Boundary conditions involving specified displacements are
called geometric (or kinematic) boundary conditions, while boundary con-
ditions involving specified edge loads are called static (or natural) boundary
conditions.
3.1 Geometric (Kinematic) Boundary Conditions

Geometric boundary conditions are those that dictate some feature of mid-
plane displacements along a plate edge. Each of the four edges of the plate is
characterized by a normal and tangential direction. For example, for the
positive x-edge shown in Fig. 3 the x-direction is normal to the edge while the
y-direction is tangent to the edge. Along this edge we may specify values for
displacements in the x-, y-, and z-directions, as well as the slope of the
laminate midplane measured in the normal direction. Because this edge is
parallel to the y-axis, the displacements imposed along this edge are either
constant or, at most, functions of y. That is, we may specify values of uo(+x)( y),
vo(+x)( y), w (+x)( y), and Bw(+x)( y)/Bx along the edge x = a. As was the case
for edge loads (discussed in Sec. 1), there is no reason to assume that the
displacements on opposite edges are identical. Therefore the superscript (+x)
has been used to indicate that these displacements are imposed along the
positive x-edge of the laminate.
Note that by specifying a particular value of the midplane slope at the
boundary, (Bw(+x)( y)/Bx), we have specified only one of several possible gra-
dients in the three displacement fields. Initially, one might suspect that, for
reasons of mathematical symmetry perhaps, other displacement field gra-
dients should also be specified. For example, perhaps Buo(+x)( y)/By or
Bvo(+x)( y)/Bz should be specified along x=a as well. These other gradients
need not be considered for two reasons. First, in thin plate theory the
boundary conditions are defined for the plate edge at the midplane. By
definition the midplane has a thickness of zero. This fundamental definition
precludes any midplane boundary conditions with a z-dependency; i.e., the
midplane displacement field gradients Buo(+x)(y)/Bz, Bvo(+x)(y)/Bz, or
Bw(+x)( y)/Bz are undefined. This leaves us with possible gradients in the x-
or y-directions. With the exception of (Bw(+x)( y)/Bx), a specified gradient in
geometric boundary conditions along a plate edge in the x- or y-directions
results in an over-specified problem. For example, having specified displace-
ments in the x-direction along the edge x = a, namely, u o(+x)( y), the value

of Buo(+x)( y)/By is also specified as a consequence. A similar consideration of
the other possible gradients results in a similar conclusion.
Recalling that displacements at any arbitrary interior point of the plate
are denoted uo(x,y), vo(x,y), and w(x,y), the geometric boundary conditions
that may be specified along each of the four edges of the plate are:
For edge x= a: For edge x = 0:

ðþxÞ ðxÞ
u0 ða; yÞ ¼ uo ð yÞ u0 ð0; yÞ ¼ uo ð yÞ
ðþxÞ ðxÞ
v0 ða; yÞ ¼ vo ð yÞ v0 ð0; yÞ ¼ vo ð yÞ
ðþxÞ
Bwða; yÞ Bw ð yÞ Bwð0; yÞ BwðxÞ ð yÞ
¼ ¼
Bx Bx Bx Bx
wða; yÞ ¼ wðþxÞ ð yÞ wð0; yÞ ¼ wðxÞ ð yÞ
For edge y= b: For edge y =0:

ðþyÞ ðyÞ
u0 ðx; bÞ ¼ uo ðxÞ u0 ðx; 0Þ ¼ uo ðxÞ
ðþyÞ ðyÞ
v0 ðx; bÞ ¼ vo ðxÞ v0 ðx; 0Þ ¼ vo ðxÞ
ðþyÞ ðyÞ
Bwðx; bÞ Bw ðxÞ Bwðx; 0Þ Bw ðxÞ
¼ ¼
By By By By
wðx; bÞ ¼ wðþyÞ ðxÞ wðx; 0Þ ¼ wðyÞ ðxÞ
3.2 Static (Natural) Boundary Conditions

We will again use the positive x-edge to demonstrate possible static boundary
conditions. Referring to Fig. 3, note that five external stress and moment
(+x)
resultants may be specified along this edge: N xx ( y), N (+x)
xy ( y), Vxz
(+x)
(+x) (+x) (+x) (+x) (+x)
( y), Mxx ( y), and Mxy ( y). Values for Nxx ( y), Nxy ( y), and Mxx
( y) can be specified independently, without consideration of any other stress
(+x)
or moment results. However, the remaining two resultants, Vxz ( y) and
(+x)
Mxy ( y), are not independent and must be considered together. The in-
(+x) (+x)
terdependence between Vxz ( y) and Mxy ( y) was first noted by Kirchhoff
in 1850 and occurs because (as far as static equilibrium is concerned) the
(+x)
twisting moment Mxy ( y) can be replaced by a statically equivalent couple
involving two vertical shear resultants. This is illustrated in Fig. 8. The
twisting moments acting on two adjacent elements of length dy are shown
in Fig. 8a. The element on the right is subjected to a twisting moment
(+x)
M xy ( y)dy, whereas the element on the left is subjected to a twisting moment
ðþxÞ ðþxÞ
[Mxy (y)+(dMxy ðyÞ/dy) dy] dy. These twisting moments may be replaced
by an equivalent couple involving two vertical shear resultants, as shown in
Fig. 8b. The element on the right is subjected to two vertical shear resultants of

Figure 8 Representing the twisting moment resultant acting along the edge
x = a with statically equivalent vertical shear stress resultants. (a) Twisting
moment resultants acting on two adjacent elements of length dy along the edge
x = a. (b) Vertical shear resultants acting on adjacent elements along the edge
x = a which are statically equivalent to the twisting moments shown in (a).
(+x)
magnitude Mxy ( y), acting in the directions shown and separated by
distance dy. Similarly, the element on the left is subjected to two vertical
ðþxÞ ðþxÞ
shear resultants of magnitude [Mxy (y)+(dMxy ( y)/dy)dy], acting in the
directions shown and separated by distance dy. Summing the shear force
components present at the interface between the two infinitesimal elements,
ðþxÞ ðþxÞ ðþxÞ ðþxÞ
we obtain {[Mxy ( y)+(dMxy ( y)/dy)dy]Mxy ( y)}=(dMxy ( y)/dy)dy.
(+x)
Hence the variation of the twist moment M xy ( y) along the edge x =a is
ðþxÞ
statically equivalent to a vertical shear resultant of magnitude dMxy ( y)/dy.
Adding this statically equivalent vertical shear resultant component to the
(+x)
external vertical shear that is actually applied to the edge, Vxz ( y), we define
the ‘‘Kirchhoff’’ shear resultant acting along the edge x= a:
ðþxÞ
dMxy ðyÞ
V ðþxKÞ
xz ðyÞ ¼ V ðþxÞ
xz ðyÞ þ ð10Þ
dy

The superscript ‘‘+xK’’ is used to denote the Kirchhoff shear resultant
(sometimes called the ‘‘effective’’ shear resultant) acting along the positive
x-edge.
Now, the equations of equilibrium (developed in Sec. 2) must be
satisfied at all points within the laminate, including the laminate edge. There-
fore the distribution of the external vertical shear resultant, V (+x)
xz ( y), must
satisfy Eq. (3b). Along the edge x=a this equation becomes:
ðþxÞ ðþxÞ
BMxx BMxy
V ðþxÞ
xz ¼ þ
Bx By
Recalling that all boundary conditions along the positive x-edge are either
ðþxÞ
constant or are functions of y only, it must be that dMxx /dx=0. Therefore
along the edge x= a:
ðþxÞ
BMxy
V ðþxÞ ¼ ð11Þ
xz
By
(+x)
Eq. (11) shows that the variation of the externally applied shear V xz ( y)
along the edge x =a is intimately related to the variation of the externally
(+x)
applied twist moment Mxy ( y) along this same edge, and that the functional
form of these two resultants cannot be specified independently.* Suppose, for
example, that the laminate is subjected to a vertical shear resultant that varies
(+x)
linearly over the edge x = a. That is, suppose Vxz ( y)= Ay+B, where A and
B are constants. Under this circumstances, Eq. (11) shows that a twist
moment must also be present in order for static equilibrium to be maintained.
ðþxÞ
Furthermore, the twist moment must vary according to dMxy ( y)/dy=
ðþxÞ
Ay+B, which implies that Mxy ðyÞ ¼ A=2 ðy2 Þ þ By þ C , where C is a
constant of integration.
(+x)
As a second case of interest, suppose that V xy is constant, i.e., assume
(+x)
A = 0 and hence that Vxz ( y)= B. In this case the rate of change in the twist
ðþxÞ
moment must equal dMxy ( y)/dy=B; otherwise, Eq. (11) is not satisfied and
static equilibrium is not maintained. This shows that it is not possible to apply
constant vertical shear only along x = a; if a constant vertical shear is applied,
then a corresponding twist moment that varies over the length of the plate
edge must also be present. However, the converse is admissible. That is, if a
(+x)
constant twist moment Mxy = B is applied, then the vertical shear must
ðþxÞ ðþxÞ
equal zero: dMxy ( y)/dy=V xz ( y)=0.
* It is noted that a similar interdependence occurs between the shear force and bending mo-
ment present in a prismatic beam. That is, from fundamental beam theory, the shear force pres-
ent at any cross section within a prismatic beam is related to the bending moment according to
V = dM/dx.

Substituting Eq. (11) into Eq. (10), we find that the effective Kirchhoff
shear resultant acting along the edge x=a can be written in two equivalent
ways:
ðþxÞ
ðþxKÞ ðþxÞ dMxy ðyÞ
V xz ðyÞ ¼ V xz ðyÞ þ
dy
or ð12Þ
ðþxÞ
ðþxKÞ dMxy ðyÞ
V xz ðyÞ ¼ 2
dy
Note that the Kirchhoff shear resultant differs from the applied vertical shear
(+xK)
only if there is variation in the twisting moment. That is, V xz ( y) differs
ðþxÞ
from Vxz ( y) only if dMxy ( y)/dy p 0. In effect, the Kirchhoff shear
(+x)
resultant is used to specify both the vertical shear and twist moment acting
along the edge of the plate.
Recalling that stress and moment resultants at any arbitrary interior
point of the plate are denoted Nxx * (x,y), Nyy
* (x,y), Nxy
* (x,y), Nyx * (x,y),
Mxx* (x,y), Myy
* (x,y), Mxy
* (x,y), Myx
* (x,y), Vxz
* (x,y), and Vyz
* (x,y), the static
boundary conditions that may be specified along the edges x= a are as
follows:
For edge x= a:
ðþxÞ
*ða; yÞ ¼ Nxx ðyÞ
Nxx
*ða; yÞ ¼ NðþxÞ
Nxy xy ðyÞ
*ða; yÞ ¼ MðþxÞ
Mxx xx ðyÞ
ðþxÞ ðþxÞ
BMxx
*ða; yÞ BMxy
*ða; yÞ dMxy ðyÞ dMxy ðyÞ
þ2 ¼ V ðþxÞ
xz ðyÞ þ ¼2
Bx By dy dy
Static boundary conditions along the edge x= a have been discussed above.
Following an analogous procedure, static boundary conditions that may
be present along the remaining three edges of the rectangular plate are as
follows:
For edge x= 0:
ðxÞ
*ð0; yÞ ¼ Nxx ðyÞ
Nxx
ðxÞ
*ð0; yÞ ¼ Nxy ðyÞ
Nxy
ðxÞ
*ð0; yÞ ¼ Mxx ðyÞ
Mxx
ðxÞ ðxÞ
BMxx
*ð0; yÞ BMxy
*ð0; yÞ dMxy ðyÞ dMxy ðyÞ
þ2 ¼ V ðxÞ
xz ðyÞ þ ¼2
Bx By dy dy

For edge y= b:
ðþyÞ
*ðx; bÞ ¼ Nyy ðxÞ
Nyy
ðþyÞ
*ðx; bÞ ¼ Nyx ðxÞ
Nyx
ðþyÞ
*ðx; bÞ ¼ Myy ðxÞ
Myy
ðþyÞ ðþyÞ
BMyy
*ðx; bÞ BMyx
*ðx; bÞ dM yx ðxÞ dM yx ðxÞ
þ2 ¼ V ðþyÞ
yz ðxÞ þ ¼2
By Bx dx dy
For edge y= 0:
ðyÞ
*ðx; 0Þ ¼ Nyy ðxÞ
Nyy
ðyÞ
*ðx; 0Þ ¼ Nyx ðxÞ
Nyx
ðyÞ
*ðx; 0Þ ¼ Myy ðxÞ
Myy
ðyÞ ðyÞ
BMyy
*ðx; 0Þ BMyx
*ðx; 0Þ dM yx ðxÞ dM yx ðxÞ
þ2 ¼ V ðyÞ
yz ðxÞ þ ¼2
By Bx dx dy
3.3 Combinations of Geometric and Static Boundary

Conditions
Two fundamental categories of boundary conditions were defined in the
preceding subsections: Geometric Boundary Conditions and Static Boundary
Conditions. If either a geometric or static boundary condition is specified to
equal zero, then that condition is called a homogeneous condition. In contrast,
if a condition is required to take on a nonzero value, then the condition is
called an inhomogeneous condition. Examples of homogenous and inhomo-
geneous geometric conditions along the edge x =a are uo(a,y) =0 and
uo(a,y)=1 mm, respectively. Examples of homogeneous and inhomogeneous
static conditions along the edge x =a are Nxx(a,y)=0 and Nxx(a,y) =1000 N/
m, respectively.
Four potential conditions exist in each category for each of the four
edges of a rectangular laminate. The possible geometric and static condition
for each condition must be viewed as complementary pairs. That is, for each
condition we may specify either a geometric requirement or a static require-
ment, but we cannot specify both. For example, suppose a problem is
considered in which it is stipulated that displacements in the x-direction
along the edge x =a are zero. That is, we specify that uo(a,y)=0. Note that
this requirement is an example of a homogeneous geometric boundary

condition. In such a case the geometric boundary condition is specified and is
therefore known (i.e., uo(a,y)=0), whereas the corresponding static boun-
dary condition is unknown. In fact, the question becomes, ‘‘what stress
resultant N (+x)
xx ( y) must exist along the edge x =a, so as to maintain the
boundary condition uo(a,y)=0?’’
Conversely, suppose in a different problem it is stipulated that the stress
resultant acting normal to the edge x=a must equal some constant value. For
example, suppose we specify a boundary condition in which N (+x) xx ( y) =1000
N/m. This is an example of an inhomogeneous static boundary condition, and
in this case it is the geometric boundary condition that is unknown. In fact, the
question becomes, ‘‘what displacements uo(a,y) exist along the edge x =a,
(+x)
given that the edge loading is known to be Nxx ( y)=1000 N/m?’’
As stated above, there are four complementary pairs of boundary
conditions for each edge of a rectangular plate. For each pair we may specify
either a geometric requirement or a static requirement, but we cannot specify
both. The conditions that must be specified for each edge of a rectangular plate
are summarized below:
The following conditions can be specified for the edge x=a:
Geometric Condition Static Condition
ðþxÞ ðþxÞ
uo ða; yÞ ¼ uo ðyÞ ðORÞ *ða; yÞ ¼ Nxx ðyÞ
Nxx
ðþxÞ ðþxÞ
vo ða; yÞ ¼ vo ðyÞ ðORÞ *ða; yÞ ¼ Nxy ðyÞ
Nxy
Bwða; yÞ BwðþxÞ ðyÞ ðþxÞ
¼ ðORÞ *ða; yÞ ¼ Mxx ðyÞ
Mxx
Bx Bx
BMxx
*ða; yÞ BMxy
*ða; yÞ
wða; yÞ ¼ wðþxÞ ðyÞ ðORÞ þ2
Bx By
ðþxÞ
dMxy ðyÞ
¼ VðþxÞ
xz ðyÞ þ
dy
ðþxÞ
dMxy ðyÞ
¼2
dy
The following conditions can be specified for the edge x =0:
ðxÞ ðxÞ
uo ð0; yÞ ¼ uo ðyÞ ðORÞ *ð0; yÞ ¼ Nxx ðyÞ
Nxx
ðxÞ ðxÞ
vo ð0; yÞ ¼ vo ðyÞ ðORÞ *ð0; yÞ ¼ Nxy ðyÞ
Nxy
Bwð0; yÞ BwðxÞ ðyÞ ðxÞ
¼ ðORÞ *ð0; yÞ ¼ Mxx ðyÞ
Mxx
Bx Bx
BMxx
*ð0; yÞ BMxy
*ð0; yÞ
wð0; yÞ ¼ wðxÞ ðyÞ ðORÞ þ2
Bx By
ðxÞ
dMxy ðyÞ
¼ VðxÞ
xz ðyÞ þ
dy
ðxÞ
dMxy ðyÞ
¼2
dy
The following conditions can be specified for the edge y =b:
ðþyÞ ðþyÞ
uo ðx; bÞ ¼ uo ðxÞ ðORÞ *ðx; bÞ ¼ N yx ðxÞ
Nyx
ðþyÞ ðþyÞ
vo ðx; bÞ ¼ vo ðxÞ ðORÞ *ðx; bÞ ¼ N yy ðxÞ
Nyy
Bwðx; bÞ BwðþyÞ ðxÞ ðþyÞ

¼ ðORÞ *ðx; bÞ ¼ M yy ðxÞ
Myy
By By
BMyy
*ðx; bÞ BMyx
*ðx; bÞ
wðx; bÞ ¼ wðþyÞ ðxÞ ðORÞ þ2
By Bx
ðþyÞ
dM yx ðxÞ
¼ V ðþyÞ
yz ðxÞ þ
dx
ðþyÞ
dM yx ðxÞ
¼2
dx
The following conditions can be specified for the edge y =0:
ðyÞ ðyÞ
uo ðx; 0Þ ¼ uo ðxÞ ðORÞ *ðx; 0Þ ¼ N yx ðxÞ
Nyx
ðyÞ ðyÞ
vo ðx; 0Þ ¼ vo ðxÞ ðORÞ *ðx; 0Þ ¼ N yy ðxÞ
Nyy
Bwðx; 0Þ BwðyÞ ðxÞ ðyÞ
¼ ðORÞ *ðx; 0Þ ¼ M yy ðxÞ
Myy
By By
BMyy
*ðx; 0Þ BMyx
*ðx; 0Þ
wðx; 0Þ ¼ wðyÞ ðxÞ ðORÞ þ2
By Bx
ðyÞ
ðyÞ dM yx ðxÞ
¼ Vyz ðxÞ þ
dx
ðyÞ
dM yx ðxÞ
¼2
dx
The number of different combinations that may be defined is enormous.
Four conditions must be specified to define the boundary conditions along

an edge. Because there are two possibilities in each case, a total of 24 =16
possible combinations exist along each edge. Because there are four edges, a
total of 164=65,536 different combinations of boundary conditions can be
defined for a thin rectangular plate. However, some combinations are
encountered so frequently that they have been given special names. Specifi-
cally, free edges, uniformly loaded edges, simply supported edges, and clamped
edges are encountered very frequently in practice. These particular edge
conditions will be illustrated below by considering the positive x-edge of
the plate.
3.3.1 Free Edge
A free edge is defined as an edge that is entirely free of external loading. Hence
if the edge x =a is a free edge, then the following four homogeneous static
boundary conditions must be satisfied:
For x=a:
*ða; yÞ ¼ 0
Nxx
*ða; yÞ ¼ 0
Nxy
*ða; yÞ ¼ 0
Mxx
BMxx
*ða; yÞ BMxy
*ða; yÞ
þ2 ¼0
Bx By
Analogous boundary conditions may be used to specify that any of the
remaining three edges are free edges.
Each of the internal stress and moments resultants can be expressed in
terms of the [ABD] matrix and midplane displacement fields. Recalling that
we have limited our discussion to symmetric laminates (Bij =M ijT=M ijM =0)
the boundary conditions for a free edge may also be written:
For x=a:

Buo ða; yÞ Bvo ða; yÞ Buo ða; yÞ Bvo ða; yÞ
A11 þ A12 þ A16 þ Nxx
T
Nxx
M
¼0
Bx By By Bx

A16 þ A26 þ A66 þ Nxy
T
Nxy
M
¼0
Bx By By Bx
B2 wða; yÞ B2 wða; yÞ B2 wða; yÞ

D11 D12 2D16 ¼0
Bx2 By2 BxBy
B3 wða; yÞ B3 wða; yÞ B3 wða; yÞ B3 wða; yÞ
D11 4D16 ðD 12 þ 4D 66 Þ 2D26 ¼0
Bx3 Bx2 By BxBy2 By3

The results listed above are for the edge x =a. Comparable conditions are
imposed if any of the other three edges of the plate are free edges.
3.3.2 Simply Supported Edges
A simply supported boundary condition is one in which the out-of-plane
deflection and bending moment are zero along an edge. Hence the x=a edge
of a rectangular plate is simply supported if w(a,y)=0 and Mxx(a,y)=0. Note
that these are homogeneous geometric and static boundary conditions,
respectively. The corresponding boundary condition in beam theory is a
‘‘pinned support,’’ as discussed in Sec. 8 of Chapter 8. Recall that for the case
of a pinned support as defined in beam theory only two requirements are
involved (w=Mx =0), whereas in thin plate theory four conditions must be
specified along each plate edge. Hence for thin plates four distinct combina-
tions of geometric and static boundary conditions may be classified as a
‘‘simple support.’’ The possible simple-support boundary conditions are
often numbered S1 through S4, following a numbering scheme introduced
by Almroth (5), and are defined along the edge x=a as follows:
For x=a:
ðþxÞ ðþxÞ
S1 : wða; yÞ ¼ 0 Mxx
*ða; yÞ ¼ 0 uo ða; yÞ ¼ uo ð yÞ vo ða; yÞ ¼ vo ð yÞ
ðþxÞ ðþxÞ
*ða; yÞ ¼ 0 Nxx
S2 : wða; yÞ ¼ 0 Mxx *ða; yÞ ¼ Nxx ð yÞ vo ða; yÞ ¼ vo ð yÞ
ðþxÞ ðþxÞ
* ða; yÞ ¼ 0
S3 : wða; yÞ ¼ 0 Mxx uo ða; yÞ ¼ uo ð yÞ *ða; yÞ ¼ Nxy
Nxy ð yÞ
ðþxÞ ðþxÞ
* ða; yÞ ¼ 0
S4 : wða; yÞ ¼ 0 Mxx * ða; yÞ ¼ Nxx
Nxx ð yÞ Nxy
*ða; yÞ ¼ Nxy ð yÞ
Note that a ‘‘simple support’’ is by definition a mixture of geometric and

static boundary conditions. Also, depending on the values specified for
ðþxÞ ðþÞ ðþxÞ ðþÞ
[uo ( y) or Nxx ( y)] and [vo ( y) or Nxy ( y)], either homogeneous or
inhomogeneous boundary conditions may be involved. The conditions
involving stress and moment resultants can once again be expressed in
terms of elements of the [ABD] matrix and midplane displacement fields.
For example, the four conditions that (collectively) define a simple support
of type S4 for a symmetric laminate along the edge x=a can be written:
S4 simple support, for x=a:
wða; yÞ ¼ 0
B2 wða; yÞ B2 wða; yÞ B2 wða; yÞ

D11 þ D12 2D16 ¼0
Bx 2 By 2 BxBy

A11 þA12 þA16 þ
Bx By By Bx
ðþxÞ
NTxx NM
xx ¼ N xx

A16 þ A26 þ A66 þ
Bx By By Bx
ðþxÞ
NTxy NM
xy ¼ Nxy
The other types of simple support can also be expressed in terms of

midplane displacements following a comparable procedure. Analogous
conditions may be imposed if any of the other three edges of a plate are
simply supported.
3.3.3 Clamped Edges
A clamped edge (also called a fixed edge) is one in which the out-of-plane
deflection and slope are zero along the edge. Hence the x=a edge of a rect-
angular plate is clamped if two homogenous geometric boundary conditions
are satisfied: w(a,y)=0 and Bw(a,y)/Bx=0. Once again, the term ‘‘clamped
end’’ is used during the study of beams, as discussed in Sec. 8 of Chapter 8.
However, during the study of beams only two geometric requirements are
necessary to define a clamped boundary condition, whereas in the case of thin
plates four conditions are required. Hence, four distinct combinations can be
classified as a ‘‘clamped edge.’’ The possible clamped boundary conditions
are often numbered C1 through C4 and are defined as follows:
For x=a:
Bwða; yÞ ðþxÞ ðþxÞ
C1 : wða; yÞ ¼ 0 ¼ 0 uo ða; yÞ ¼ uo ðyÞ vo ða; yÞ ¼ vo ðyÞ
Bx
Bwða; yÞ ðþÞ ðþxÞ
C2 : wða; yÞ ¼ 0 ¼ 0 Nxx
*ða; yÞ ¼ Nxx ðyÞ vo ða; yÞ ¼ vo ðyÞ
Bx
Bwða; yÞ ðþxÞ ðþÞ
C3 : wða; yÞ ¼ 0 ¼ 0 uo ða; yÞ ¼ uo ðyÞ *ða; yÞ ¼ Nxy ðyÞ
Nxy
Bx
Bwða; yÞ ðþÞ ðþÞ
C4 : wða; yÞ ¼ 0 ¼ 0 Nxx
*ða; yÞ ¼ Nxx ðyÞ Nxy
*ða; yÞ ¼ Nxy ðyÞ
Bx
A clamped boundary may or may not involve a mixture of geometric and

static conditions, depending on the values specified for [uo(+x)( y) or Nxx
(+)
( y)]
(+x) (+)
and [vo ( y) or Nxy ( y)]. Also, these conditions may be either homoge-

neous or inhomogeneous. In those cases in which the internal stress or
moment resultants are required to take on specific values at the boundary
(e.g., condition C4), the internal stress and moment resultant can be
expressed in terms of elements of the [ABD] matrix and midplane displace-
ment fields, as demonstrated above.
4 REPRESENTING ARBITRARY TRANSVERSE LOADS AS A

FOURIER SERIES
In thin plate theory the distributed transverse loading is allowed to vary in any
manner over the x–y plane, i.e., q=q(x,y). However, as will be seen it is
particularly easy to obtain a solution if the transverse loading is distributed
according to a double sinusoidal variation in x and y:
px py
qðx; yÞ ¼ qo sin sin ð13Þ
a b
where qo is a constant and equals the magnitude of the distributed load at the
center of the plate, i.e., at x=a/2 and y=b/2. Solutions obtained for this
particular distributed transverse loading will be described in Secs. 2 and 3 of
Chapter 10.
Of course, a double sinusoidal variation is just one of an infinite number
of possible transverse loads that may be encountered in practice. However, it
can be shown that any arbitrary function q(x,y) can be represented in terms of
a double Fourier series:
X
l X
l mpx npy
qðx; yÞ ¼ qmn sin sin ð14Þ
m¼1 n¼1
a b
Therefore if a solution is obtained for the simple sinusoidal variation given by

Eq. (13), then a solution for an arbitrary transverse loading can be obtained
by representing the arbitrary load as the sum of a series of double sinusoidal
terms, in accordance with Eq. (14). The constant coefficients qmn that appear
in Eq. (14) are determined based on the functional form of q(x,y) and
correspond to a particular combination of m and n. To determine the values
of qmn, multiply Eq. (14) by the factor sin(nVpy/b)dy, and integrate from 0 to b:
Z Z b X !
b
nVpy l X l mpx npx nVpy
qðx; yÞsin dy ¼ qmn sin sin sin dy
0 b 0 m¼1 n¼1
a b b
ð15Þ

It can be shown that:
Z b
npy nVpy
sin sin dy ¼ 0; if n p nV
0 b b
Z b
npy nVpy b
sin sin dy ¼ ; if n ¼ nV
0 b b 2
Consequently, upon performing the integration on the right side of Eq. (15),
only one term for the range 0<n<l remains, specifically, the term for which
n=nV. Hence Eq. (15) becomes:
Z b mpx
nVpy bX l
qðx; yÞsin dy ¼ qmnVsin ð16Þ
0 b 2 m¼1 a
We now repeat this process for the variation in x. That is, multiply Eq. (16) by
the factor sin(mVpx/a)dx and integrate from 0 to a:
Z aZ b
mVpx nVpy
qðx; yÞsin sin dxdy
0 0 a b
Z a !
b X l mpx mVpx ab
¼ qmnVsin sin dx ¼ qm Vn V
0 2 m¼1
a a 4
Solving this result for the coefficient qmVnV, we find:

Z aZ b
4 mVpx nVpy
qmVnV ¼ qðx; yÞsin sin dxdy ð17Þ
ab 0 0 a b
Eq. (17) allows calculation of the constant coefficients, and hence a given
distribution of transverse load q(x,y) can be expressed as the double Fourier
series shown in Eq. (14).
Eq. (14) involves a summation over an infinite number of terms (as
m,n ! l). Of course, it is impossible to use an infinite number of terms;
in practice a finite number of terms must be used. As the number of terms
used is increased, the series representation given by Eq. (14) resembles the ac-
tual load q(x,y) more and more closely. That is, the series representation
converges to the actual loading q(x,y) as the number of terms is increased. In
practice then, one should use the maximum number of terms that is reason-
ably possible, to ensure a reasonable series representation of the applied load
q(x,y).
Three types of commonly encountered transverse loads will be used to
illustrate the preceding discussion. First, consider the case of a constant
uniform transverse load:
qðx; yÞ ¼ qo

where qo is the magnitude of the uniformly distributed load. Substituting
q(x,y) = qo into Eq. (17), we find the coefficients in the Fourier series ex-
pansion associated with specified value of m and n to be:
Z Z
4qo a b mpx npy 16qo
qmn ¼ sin sin dxdy ¼ 2 ; m; n ¼ odd integers
ab 0 0 a b p mn
Z aZ b mpx npy
4qo
qmn ¼ sin sin dxdy ¼ 0; m; n ¼ even integers
ab 0 0 a b
Hence a constant transverse load q(x,y)=qo can be written:
16qo X
l X l
1 mpx npy
qðx; yÞ ¼ sin sin ðm; n ¼ 1; 3; 5; . . .Þ ð18Þ
p2 m ¼ 1 n ¼ 1 mn a b
How well Eq. (18) describes a constant transverse loading depends on the
number of terms used. A normalized plot of the series representation given by
Eq. (18) along the plate centerline y=b/2 is presented in Fig. 9. Curves are
shown based on a 9-term expansion (i.e., m,n=1,3,5), a 64-term expansion
(m,n=1, 3, 5,. . ., 15), and a 169-term expansion (m,n=1, 3, 5,. . ., 25). The
series clearly converges toward a constant loading as the number of terms
used is increased, although even with 169 terms the series expansion repre-
sents a constant transverse loading in only an approximate sense.
As a second example, consider the case of a transverse force P, uni-
formly distributed over an internal rectangular region of dimensions ai bi,
as shown in Fig. 10. The center of the internal region is located at x=n and
y=g, as shown. It is emphasized that P is defined as a force, with units of
Newtons or pounds-force, for example. Eq. (17) becomes in this case:
Z nþai =2 Z gþbi =2 mpx npy
4P
qmn ¼ sin sin dxdy
abai bi nai =2 gbi =2 a b
Evaluating this integral we find:

16P mpn npg mpai npbi
qmn ¼ sin sin sin sin ð19Þ
p2 mnai bi a b 2a 2b
Together, Eq. (14) and (19) define the Fourier series expansion of a force P,
uniformly distributed over the internal region ai bi. To illustrate this series
expansion, consider the following specific example. Assume that the interior
region is centered in the middle of the plate, i.e., let n=a/2, g=b/2. Further,

Figure 9 A normalized plot of Eq. (18) along the plate centerline defined by
y=b/2.
Figure 10 A rectangular plate with in-plane dimensions a b, subjected to a

force P uniformly distributed over an interior region of dimensions ai bi. The
interior region is centered at the point x=n, y=g.

assume that the length and width of the interior region equals one half the
length and width of the plate, i.e., let ai =a/2, bi =b/2. For this specific
example then, Eq. (19) becomes:
qmn ¼
64P
p2 mnab
sin ðmp2 Þsinðnp2 Þsinðmp4 Þsinðnp4 Þ; m; n ¼ odd integers
qmn ¼ 0; m; n ¼ even integers
Note that coefficients associated with even integers are zero only because we
let n=a/2, g=b/2 in this example. This would not be true if the interior region
were not centered on the plate. Substituting this result in Eq. (14), we find:
l
64P X l X
qðx; yÞ ¼ ð Þ ð Þ ð Þ ð Þ
1
p2 ab m¼1 n¼1 mn
sin
mp
2
sin
np
2
sin
mp
4
sin
np
4
ð20Þ
sinð
a Þ ð b Þ
mpx mpy
sin ; m; n ¼ odd integers
As before, how well Eq. (20) describes a load P uniformly distributed over the
interior region depends on the number of terms used. A normalized plot of the
series representation given by Eq. (20) along the plate centerline y=b/2 is
presented in Fig. 11. Curves are shown based on a 9-term expansion (i.e.,
m,n=1, 3, 5), a 64-term expansion (m,n=1, 3, 5,. . ., 15), and a 169-term
Figure 11 A normalized plot of Eq. (20) along the plate centerline defined by
y = b/2.

expansion (m,n=1,3,5,. . .,25). The series clearly converges toward a constant
loading over the central region of the plate and approaches zero elsewhere.
As a third example, consider a concentrated transverse force P, located
at x=n and y=g. The Fourier coefficients for this case can be obtained from
Eq. (19) by allowing ai !0, bi !0, i.e., the internal area over which load P acts
is allowed to shrink to a point. It can be shown that in the limit Eq. (19)
becomes:
qmn ¼
4P
ab
sinð Þ ð Þ
mpn
a
sin
mpg
b
ð21Þ
Together, Eqs. (14) and (21) define the Fourier series expansion of a
concentrated force P applied at the point x=n and y=g. As a specific
example, consider the case in which the concentrated load is applied at the
center of the plate, i.e., at n=a/2, g=b/2. In this case Eq. 21 becomes:
8
< 4P ð1Þ½ðmþnÞ=21 ; m; n ¼ odd integers
qmn ¼ ab
:
0; m; n ¼ even integers
Combining this result with Eq. (14), we find that the Fourier series expansion
for a concentrated load applied at the center of the plate is given by:
4P Xl X l
qðx; yÞ ¼
ab m ¼1 n ¼1 ð Þ ð Þ
sin
mpx
a
sin
npy
b
ð1Þ½ðmþnÞ=21 ð22Þ
where m,n=odd integers. As in the earlier examples, as the number of terms

used in the series representation given by Eq. 22 is increased the distribution of
q(x,y) resembles a concentrated force P more and more closely.
REFERENCES
1. Timoshenko, S.; Woinowsky-Krieger, S. Theory of Plates and Shells; McGraw-
Hill Book Co.: New York, NY, ISBN 0-07-0647798.
2. Ugural, A.C. Stresses in Plates and Shells; McGraw-Hill Book Co.: New York,
NY, ISBN 0-07-065730-0.
3. Whitney, J.M. Structural Analysis of Laminated Anisotropic Plates; Technomic
Pub Co.: Lancaster, PA, ISBN 87762-518-2.
4. Turvey, G.J., Marshall, I.H., Eds.; Buckling and Postbuckling of Composite Plates;
Chapman and Hall: New York, NY, 1995.
5. Almroth, B.O. Influence of edge conditions on the stability of axially compressed
cylindrical shells. AIAA J. 1966, 4 (1), 134–140.

10
Some Exact Solutions for Specially
Orthotropic Laminates
The equations that govern the behavior of a symmetric composite plate

subjected to loads and/or displacements that vary along the edges of the plate
were derived in Chap. 9. In this chapter, exact solutions to the governing
equations will be obtained for a special class of symmetric laminates called
‘‘specially orthotropic’’ laminates. The solutions presented allow calculation
of the deflections caused by the combined effects of in-plane loads, transverse
loads, and changes in temperature or moisture content. Buckling caused by
in-plane loads and/or temperature changes is also discussed.
1 EQUATIONS OF EQUILIBRIUM FOR A SPECIALLY

ORTHOTROPIC LAMINATE
The equations of equilibrium for a symmetric composite laminate were
presented in Sec. 2.2 of Chap. 9. Boundary conditions consistent with thin-
plate theory were then discussed in Sec. 3 of Chap. 9. Unfortunately, exact
solutions to the equations of equilibrium that satisfy specified boundary con-
ditions often cannot be found. Sometimes, the difficulty in obtaining an exact
solution is because of the boundary conditions involved; difficulties of this
sort are encountered for both isotropic and anisotropic plates. In other cases,
the difficulties are strictly because of the anisotropic nature of composites.
527

We have previously restricted our discussion to symmetric stacking
sequences, and so Bij ¼ MTij ¼ MM ij = 0 in all cases considered. In addition,
in this chapter, we will further restrict our discussion to the so-called specially
orthotropic laminates. A specially orthotropic laminate is one for which A16
¼ A26 ¼ D16 ¼ D26 ¼ NTxy ¼ NM xy = 0. It turns out that this is a very re-
strictive limitation. Indeed, a review of the stacking sequences presented in
Sec. 7 of Chap. 6 will reveal that there are only three stacking sequences that
eliminate these coupling stiffnesses and thermal/moisture resultants and can
therefore be classified as ‘‘specially orthotropic.’’ These are:
Unidirectional [0j]n laminates (described in Sec. 7.1 of Chap. 6).
Unidirectional [90j]n laminates (described in Sec. 7.1 of Chap. 6).
Symmetric cross-ply laminates, e.g., [(0/90)n]s (described in Sec. 7.2 of
Chap. 6).
It is mentioned in passing that the definition of a ‘‘specially orthotropic
laminate’’ is inconsistently used in the literature, and is sometimes relaxed
from the definition just described. Specifically, the definition of a specially
orthotropic laminate may be relaxed if the laminate is symmetric and is
subjected to in-plane loading only. In problems of this type, the coupling
stiffnesses D16 and D26 play no role. Therefore researchers studying the
behavior of symmetric composites subjected to in-plane loads and displace-
ments (only) often specify that a specially orthotropic laminate is one in which
A16 ¼ A26 ¼ NTxy ¼ NM xy = 0, but place no restriction on the values of D16
or D26. For example, Lekhnitski [1] and Savin [2] have obtained solutions that
allow calculation of the in-plane stresses induced in regions near elliptical,
rectangular, or triangular holes in anisotropic composite plates, and these
solutions are valid if the laminate is symmetric and subjected to in-plane loads
only. The solutions by Lekhnitski and Savin are said to be applicable to
specially orthotropic laminates, but in fact only require that A16 ¼ A26 ¼
NTxy ¼ NM xy = 0. Hence both symmetric balanced laminates and symmetric
angle-ply laminates (see Sec. 7 of Chap. 6) are specially orthotropic under this
relaxed definition. Nevertheless, in this chapter, we consider situations in
which the composite laminate is subjected to both in-plane and out-of-plane
loading, and the therefore the bending stiffnesses of the laminate plays an
important role. Therefore we must maintain the more restrictive definition
and stipulate that a specially orthotropic laminate must have A16 ¼ A26 ¼
D16 ¼ D26 ¼ NTxy ¼ NM xy = 0.
For a specially orthotropic laminate, the equations of equilibrium for
a symmetric laminate [i.e., Eq. (8) of Chap. 9] are simplified further still,
and become:
@ 2 uo @ 2 vo @ 2 uo
A11 þ ðA12 þ A66 Þ þ A66 ¼0 ð1aÞ
@x 2 @x@y @y2

@ 2 uo @ 2 vo @ 2 uo
ðA12 þ A66 Þ þ A22 þ A 66 ¼0 ð1bÞ
@x@y @y2 @x2

2
@uo @vo @ w
A11 þ A12 NTxx NM
@x @y xx
@x2

2
@uo @vo @ w
þ A12 þ A22 NTyy NM
@x @y yy
@y2 ð1cÞ

2
@uo @vo @ w
þ 2 A66 þ
@y @x @x@y
@4w @4w @4w
D11 2ðD12 þ 2D66 Þ 2 2 D22 4 þ qðx; yÞ ¼ 0
@x4 @x @y @y
In the following sections, we will first obtain solutions for plate
deflections under relatively simple loading conditions. It will then be seen
that solutions for simple loading condition form the basis for predicting
displacements caused by complex transverse loads.
There are, in essence, two techniques that can be used to obtain exact
solutions for problems involving arbitrary transverse loads. Both methods
place limitations on the type of boundary conditions that can be considered.
The first approach was developed by Navier in about 1820, and is known as
the Navier solution [3]. The Navier solution is applicable if all four edges of
the specially orthotropic rectangular plate are simply supported. The Navier
solution applied to the case of a simply supported specially orthotropic panel
subjected to arbitrary transverse loading will be described in Sec. 6. The
second exact solution technique was developed by Levy in 1899, and is known
as the Levy solution [3]. In this case, two opposite edges of the plate must be
simply supported, while the boundary conditions on the remaining two edges
are arbitrary and may be clamped or free, for example. The Levy solution
technique will not be discussed in this textbook, and the interested reader is
referred to Refs. 1–5 for a discussion of this method.
A final preliminary comment is that thin composite plates (or in fact any
thin-walled structure) are prone to buckling if subjected to compressive in-
plane loads and/or high shear loads. Buckling of specially orthotropic
laminates is discussed in Secs. 7 and 8, respectively.
2 IN-PLANE DISPLACEMENT FIELDS IN SPECIALLY

ORTHOTROPIC LAMINATES
The overall goal in this chapter is to predict deflections induced in specially
orthotropic composite plates that are simply supported along all four edges.

As discussed in Sec. 3 of Chap. 9, four types of simple supports may be defined
for a thin plate. For the edge x=a, these are:
* ða; yÞ ¼ 0
S1: wða; yÞ ¼ 0 Mxx uo ða; yÞ ¼ uðþxÞ ðyÞ vo ða; yÞ ¼ vðþxÞ ðyÞ
o o
S2: wða; yÞ ¼ 0 M* ða; yÞ ¼ 0

xx
*
N ða; yÞ ¼ N
xx
ðþxÞ
ðyÞ
xx vo ða; yÞ ¼ vðþxÞ
o ðyÞ
* ða; yÞ ¼ 0
S3: wða; yÞ ¼ 0 Mxx uo ða; yÞ ¼ uðþxÞ ðyÞ N* ða; yÞ ¼ NðþxÞ ðyÞ
o xy xy
* ða; yÞ ¼ 0
S4: wða; yÞ ¼ 0 Mxx * ða; yÞ ¼ NðþxÞ ðyÞ
Nxx * ða; yÞ ¼ NðþxÞ ðyÞ
Nxy
xx xy
Hence the definition of a simply supported edge requires that either in-plane
displacements, in-plane stress resultants, or some combination thereof must
be specified along the edge, in addition to the requirement that out-of-plane
displacements and bending moments vanish along the edge.
The following sequence of events is assumed to occur during fabrication
and assembly of a simply supported composite plate. We assume that the
laminate is cured at an elevated temperature, and that the laminate is stress-
and strain-free at the cure temperature.* Following cure, the laminate is
cooled to room temperature, and therefore midplane displacements (as well as
thermal stress resultants and associated ply strains and stresses) are induced
during cooldown to room temperature. The laminate is then trimmed to the
desired dimensions a b and assembled in a surrounding structure that
provides simple supports along all four edges. According to this scenario then,
midplane displacements have already been induced within the laminate prior
to assembly in the simple supports. Whether additional in-plane displace-
ments subsequently occur, due to application of q(x,y), a further change in
temperature, and/or a change in moisture content, depends on the type of
simple supports involved. That is, the development of additional in-plane
displacements depends on whether simple supports of type S1, S2, S3, or S4
are imposed along each edge of the plate.
All problems considered here will be based on the following. First, we
assume that opposite edges of the plate are subjected to the same type of
simple support. For example, if the edge x = 0 is subjected to the type S1
simple support, then by assumption the edge x = a is also subjected to type
S1 supports. Second, we assume that stress resultants Nxx, Nyy, and/or Nxy
applied to opposite edges of the plate (if any) are identical and uniformly
distributed along the edge. This loading condition is precisely equivalent to
*As mentioned in Section 6.2 of Chap. 6, in practice the stress- and strain-free temperature is often
20–50jC below the final cure temperature. This complication has been ignored throughout this
text, and it is assumed that the laminate is stress- and strain-free at the cure temperature.

that assumed during the development of CLT in Chap. 6. Therefore we can
develop general expressions giving the in-plane displacement fields caused by
any combination of Nij, DT, and/or DM using a standard CLT analysis. We
will then specialize these general expressions to correspond to type S1, S2, S3,
or S4 simple supports.
To begin this process, we relate midplane strains and curvatures to stress
resultants using Eq. (45) of Chap. 6. For a symmetric specially orthotropic
laminate subjected to uniform in-plane stress resultants, a uniform change in
temperature, and/or a uniform change in moisture content, Eq. (45) of Chap.
6 becomes:
8 o 9 2 38 9
> exx > a11 a12 0 0 0 0 > Nxx þ NTxx þ NMxx >
>
> > > >
> eoyy >
> > 6 a12 a22 0
> 0 0 0 7
>
> Nyy þ NT þ NM >
> >
>
>
< o = 6> 6 7 >
< yy yy >
=
cxy ¼ 6 0 0 a 0 0 0 7
66 7 Nxy
> jxx > 6 0 d11 d12 0 7> 7 >
>
> >
> 6 0 0 > 0 >
>
> >
> 4 0 0 0 d d 0 5>>
>
>
>
>
>
: j yy >
; 12 22 >
: 0 >
;
jxy 0 0 0 0 0 d66 0
As in earlier chapters, we assume infinitesimal strain levels. Therefore mid-
plane strains are related to midplane displacements according to Eq. (10) of
Chap. 6:
@uo @vo @uo @vo
eoxx ¼ eoyy ¼ coxy ¼ þ
@x @y @y @x
Consequently, midplane displacement fields are given by:
@uo
¼ a11 ðNxx þ NTxx þ NM
xx Þ þ a12 ðNyy þ Nyy þ Nyy Þ
T M
ð2aÞ
@x
@vo
¼ a12 ðNxx þ NTxx þ NM
xx Þ þ a22 ðNyy þ Nyy þ Nyy Þ
T M
ð2bÞ
@y
@uo @vo
þ ¼ a66 ðNxy Þ ð2cÞ
@y @x
Integrating Eq. (2a) with respect to x, we find:

h i
uo ðx; yÞ ¼ a11 ðNxx þNTxx þNM
xx Þþ a 12 ðNyy þ N T
yy þ N yy x þ f1 ðyÞ þ k1
M
Þ
Where f1( y) is an unknown function of y (only) and k1 is an unknown constant

of integration. Similarly, integrating Eq. (2b) with respect to y we find:
h i
vo ðx; yÞ ¼ a12 ðNxx þNTxx þNMxx Þ þ a 22 ðN yy þ NT
yy þ N yy yþf2 ðxÞ þ k2
M
Þ

where f2(x) is an unknown function of x (only) and k2 is a second unknown
constant. Without a loss in generality, we assume that midplane displace-
ments are zero at the origin (i.e., let uo = vo = 0 at x = y = 0), and con-
sequently we conclude k1 = k2 = 0. Substituting the above expressions for
uo(x,y) and vo(x,y) into Eq. (2c), we find:
@uo @vo @f1 @f2
þ ¼ þ ¼ a66 Nxy
@y @x @y @x
Because the terms that appear on the right side of the equality (i.e., a66 and
Nxy) are known constants, it follows that f1 and f2 must be at most linear
functions of y and x, respectively:
f1 ðyÞ ¼ k3 y
f2 ðxÞ ¼ k4 x
Hence we can write:
k3 þ k4 ¼ a66 Nxy
Constants k3 and k4 can take on any value as long as they sum to the product
(a66Nxy). To determine particular values convenient for our use, we now
require that the infinitesimal rotation vector in the x–y plane, xxy (which
represents rigid body motion of the plate), is zero. The infinitesimal rotation
vector is given by [see Ref. 3]:

1 @uo @vo
xxy ¼
2 @y @x
Requiring that xxy=0 leads to:
1
k3 ¼ k4 ¼ a66 Nxy
2
Combining the preceding results, we conclude that in-plane midplane dis-
placement fields induced in a symmetric specially orthotropic composite panel
by the combination of uniform in-plane stress resultants, a uniform change in
te mperature, and/or a uniform change in moisture contents are given by:
h i 1
uo ðx; yÞ ¼ a11 ðNxx þ NTxx þ NM
xx Þ þ a 12 ðNyy þ N T
yy þ Nyy x þ
M
Þ
2
ða66 Nxy Þy
1
vo ðx; yÞ ¼ ða66 Nxy Þx þ a12 ðNxx þ NTxx þ NM
xx Þ
2
þ a22 ðNyy þ NTyy þ NMyy Þy

These expressions can be simplified through the use of the effective thermal
expansion coefficients of the laminate, defined by Eq. (73a) of Chap. 6, and the
effective moisture expansion coefficients, defined by Eq. (76a) of Chap. 6. For
a specially orthotropic laminate, these become:
1 h i 1 h i
axx ¼ a11 NTxx þ a12 NTyy ayy ¼ a12 NTxx þ a22 NTyy axy ¼ 0
DT DT
1 h i 1 h i
bxx ¼ a11 NMxx þ a12 Nyy
M
byy ¼ a12 NMxx þ a22 Nyy
M
bxy ¼ 0
DM DM
Hence we see that the midplane displacement fields can be written as:
1
uo ðx; yÞ ¼ a11 Nxx þ a12 Nyy þ DTaxx þ DMbxx x þ ða66 Nxy Þy ð3aÞ
2
1
vo ðx; yÞ ¼ ða66 Nxy Þx þ a12 Nxx þ a22 Nyy þ DTayy þ DMbyy y ð3bÞ
2
Note that the displacement fields are independent of the transverse load
q(x,y). As pointed out in Sec. 2 of Chap. 9, uo(x,y) and vo(x,y) are predicted to
be independent of q(x,y) because we have assumed displacement gradients are
small, such that gradients squared can be ignored [e.g., (@w/@x)2 c 0]. Analy-
ses that account for large displacement gradients are not considered in this
text. If we had included large gradients in our analysis, then expressions for
uo(x,y) and vo(x,y) corresponding to Eqs. (3a) and (3b) would depend on
transverse load q(x,y).
Direct substitution of Eqs. (3a) and (3b) will reveal that these equations
satisfy the equations of equilibrium, Eqs. (1a) and (1b). In the following
sections, we will use these expressions to specify in-plane displacement fields
associated with simple supports of type S1 through S4.
3 SPECIALLY ORTHOTROPIC LAMINATES SUBJECT

TO SIMPLE SUPPORTS OF TYPE S1
In this section, we consider the specially orthotropic plate shown in Fig. 1.
The plate is assumed to be rectangular with thickness t and in-plane
dimensions a b. All four edges of the plate are subject to simple supports
of type S1, where it is assumed that the laminate was mounted in the structure
that imposes type S1 supports following cooldown from the cure temperature
to room temperature. After assembly, the plate is subjected to a uniform
change in temperature and a transverse loading that varies over the x–y plane
according to:
px py
a b

Figure 1 Thin rectangular plate of thickness t and in-plane dimensions a b,
subjected to a transverse load q(x,y) = qo {sin[(px)/a]}{sin[(py)/b]}. All four
edges of the plate are subject to simple supports of type S1.
We will not consider a change in moisture content (i.e., let DM = 0), although
from earlier discussion it should be clear that a change in moisture content
can be accounted for (in a mathematical sense) using the same techniques
used to model uniform changes in temperature.
Let us define DT c as the change in temperature from cure to room
temperature:
DT c ¼ TRT TC
where TRT = room temperature and TC = the cure temperature. For a

symmetric specially orthotropic laminate, the in-plane displacements caused
by DT c can be calculated using Eqs. (3a) and (3b) with Nxx = Nyy = Nxy = 0:
uco ðx; yÞ ¼ ðaxx DT c Þx ð5aÞ

vco ðx; yÞ ¼ ayy DT c y ð5bÞ
These displacements are induced before assembly of the simply supported

plate. Note that these in-plane displacement fields satisfy the first two
equations of equilibrium, Eqs. (1a) and (1b). A type S1 simple support will
simply maintain these displacements during subsequent loading and/or

temperature changes. Therefore type S1 boundary conditions for all four
edges of the plate become:
For x ¼ 0 For x ¼ a :
wð0; yÞ ¼ 0 wða; yÞ ¼ 0
* ð0; yÞ ¼ D @2w @2w * ða; yÞ ¼ D @2w @2w

Mxx 11 þ D12 2 ¼ 0 Mxx 11 þ D12 2 ¼ 0
@x 2 @y @x2 @y
ðxÞ ðþxÞ
uo ð0; yÞ ¼ uo ðyÞ ¼ 0 uo ða; yÞ ¼ uo ðyÞ ¼ ðaxx DT c Þa
ðxÞ ðþxÞ
vo ð0; yÞ ¼ vo ðyÞ ¼ ayy DT c y vo ða; yÞ ¼ vo ðyÞ ¼ ayy DT c y
For y ¼ 0 For y ¼ b :
wðx; 0Þ ¼ 0 wðx; bÞ ¼ 0
* ðx; 0Þ ¼ D @ w
2
@ w 2
* ðx; bÞ ¼ D @ w þ D @ w ¼ 0
2 2
Myy 11 þ D 12 ¼ 0 M 11 12
@x2 @y2 yy
@x2 @y2
ðyÞ ðþxÞ
uo ðx; 0Þ ¼ uo ðxÞ ¼ ðaxx DT Þx
c
uo ðx; bÞ ¼ uo ðxÞ ¼ ðaxx DT c Þx
ðyÞ ðþxÞ
vo ðx; 0Þ ¼ vo ðxÞ ¼ 0 vo ðx; bÞ ¼ vo ðyÞ ¼ ayy DT c b
We wish to determine the out-of-plane displacement field w(x,y) that

satisfies these boundary conditions as well as the equations of equilibrium,
when the plate is subjected to a transverse loading q(x,y) and/or a further
change in temperature. Guided by the functional form of the transverse
pressure [i.e., Eq. (4)], we assume that the out-of-plane displacement field is
given by:
px py
wðx; yÞ ¼ c sin sin ð6Þ
a b
where c is an unknown constant. Substituting this assumed form into the
boundary conditions will reveal that they are identically satisfied for any value
of c. Hence the value of constant c must be determined by enforcing the third
equation of equilibrium, Eq. (1c). We perform the following operations:
(a) substitute Eqs. (4), (5a), (5b), and (6) into the third equation of
equilibrium, Eq. (1c);
T T
(b) write the thermal stress resultants Nxx and Nyy in terms of effec-
tive thermal expansion coefficients [using Eq. (73b) of Chap. 6]; and
then
(c) solve the resulting expression for constant c.

Following this process we find:
qo
c¼
D 11 2 D 22 ð DT c
DT Þ 1 1
p4 4 þ 2 2 ðD12 þ 2D66 Þ þ 4 þ 2 2
A11 axx þ A12 ayy þ 2 A12 axx þ A22 ayy
a a b b p a b
ð7aÞ
Notice that the temperature change as defined in earlier chapters (DT ) ap-
pears in Eq. (7a). That is:
DT ¼ ðcurrent temperatureÞ ðcure temperatureÞ
Also note that if the temperature is not changed following assembly in the
type S1 simple supports, then DT = DT c, and the effects of temperature cancel
in Eq. (7a).
Results from thin-plate theory are often expressed in terms of the so-
called plate aspect ratio, R = a/b. Equation (7a) can be rewritten using the
aspect ratio as follows:
qo R4 b4
c¼
ðDT DT Þa2
c
p4 D11 þ 2R2 ðD12 þ 2D66 Þ þ R4 D22 þ a xx A11 þ R2
A 12 þ a yy A 12 þ R 2
A22
p2
ð7bÞ
The predicted out-of-plane deflection is obtained by combining either
Eq. (7a) or (7b) with Eq. (6). Using Eq. (7b) for example, we have:
qo R4 b4 sinðpx=aÞsinðpy=bÞ
wðx; yÞ ¼
ðDT c DT Þa2
p4 D11 þ 2R2 ðD12 þ 2D66 Þ þ R4 D22 þ axx A 11 þ R 2
A 12 þ ayy A12 þ R2
A22
p2
ð8Þ
Equations (5a), (5b), and (8) give the predicted displacement field induced in
the plate and represent the solution to this problem.
To summarize, we have considered a symmetric specially orthotropic
plate subjected to type S1 simple-supports. We have assumed that the plate is
mounted within simple supports while at room temperature. The laminate has
therefore likely experienced a change in temperature prior to assembly be-
cause modern composites are typically cured at an elevated temperature. The
change in temperature associated with cooldown to room temperature is
represented by DT c. After assembly, the plate is subjected to a sinusoidally
varying transverse loading and/or the temperature is changed away from
room temperature. The resulting displacement fields are given by Eqs. (5a),
(5b), and (8). A typical application of this solution is discussed in Sample
Problem 1.

As a closing comment, it should be kept in mind that we have not yet
considered the possibility of buckling. The solution presented here is not valid
if the change in temperature is such that the resulting in-plane stress resultants
are compressive and have a magnitude large enough to cause buckling. The
phenomenon of buckling induced by a change in temperature is called
‘‘thermal buckling’’ and will be discussed in Sec. 8.
Sample Problem 1
A [02/90)2]s graphite–epoxy laminate is cured at 175jC and then cooled to
room temperature (20jC). After cooling, the flat laminate is trimmed to in-
plane dimensions of 300 150 mm and mounted in an assembly that provides
type S1 simple supports along all four edges. The x axis is defined parallel to
the 300 mm edge (i.e., a = 0.3 m; b = 0.15 m). The laminate is then subjected
to a transverse pressure given by q(x,y) = 40 {sin[(px)/a]}{sin[(py)/b]} (kPa)
and a uniform temperature change. No change in moisture content occurs
(DM = 0). Plot the maximum out-of-plane displacement as a function of
temperature, over the range 50jC < T < 20jC. Use the properties listed for
graphite–epoxy in Table 3 of Chap. 3, and assume each ply has a thickness of
0.125 mm.
Solution. The rectangular plate is a 12-ply laminate with total thickness
t = 12(0.125 mm) = 1.5 mm and aspect ratio R = a/b = (0.3 m)/(0.15 m) =
2.0. Out-of-plane displacements are given by Eq. (8), and hence elements of
the [ABD] matrix are required. Based on the properties listed in Table 3 of
Chap. 3 for graphite–epoxy and the specified stacking sequence, the [ABD]
matrix is:
2 3
1:76 106 4:52 106 0 0 0 0
6 7
6 4:52 106 95:6 106 0 0 0 0 7
6 7
6 0 0 19:5 106 0 0 0 7
½ABD ¼ 6
6
7
7
6 0 0 0 40:1 0:848 0 7
6 7
4 0 0 0 0:848 10:8 0 5
0 0 0 0 0 3:66
where the units of Aij are Pa m and the units of Dij are Pa m3.
We also require the effective thermal expansion coefficients. Based on
the properties listed in Table 3 of Chap. 3 for graphite–epoxy and the specified
stacking sequence, these are:
axx ¼ 0:29 lm=m jC ayy ¼ 2:44 lm=m jC

Because the laminate is specially orthotropic, the effective shear thermal
expansion coefficient is zero (axy = 0). The cooldown from the cure temper-
ature to room temperature is:
DT c ¼ ð20jCÞ ð175jCÞ ¼ 155jC
Following assembly, the temperature ranges from room temperature to as
low as 50jC. Therefore:
255jC < DT < 155jC
Note that the temperatures to be considered are all at or below room tem-
perature. Because the effective thermal expansion coefficients are algebrai-
cally positive, if the laminate were not constrained by the S1 simple supports,
it would tend to contract as temperature is lowered. Because it is in fact
constrained by the simple supports, the in-plane stress resultants that develop
as temperature is lowered tend to be tensile. Therefore thermal buckling is not
of concern. Substituting all known values, Eq. (8) becomes:
324 px py
wðx; yÞ ¼
sin sin ðmetersÞ
88:6ðDT c DTÞ 0:30 0:15
p4 278 þ
p2
This expression can be used to calculate the out-of-plane displacement in-
duced at any point (x,y) over the surface of the plate. The maximum out-of-
Figure 2 Maximum out-of-plane displacement for the graphite–epoxy plate

considered in Sample Problem 1 as a function of temperature.

plane displacement occurs at the center of the plate, i.e., at x = a/2 = 0.15 m
and y = b/2 = 0.075 m. Because DT c = 20jC–175jC = 155jC, and tem-
peratures ranging from 50jC < T < 20jC are to be considered, the quantity
(DT cDT ) ranges from 70jC z (DT cDT ) z 0jC in this problem. A plot
of maximum out-of-plane displacement as a function of temperature is shown
in Fig. 2. At room temperature (20jC), a maximum deflection of 12 mm is
predicted. As would be expected, the plate becomes stiffer as the temperature is
decreased, in the sense that out-of-plane displacements are decreased because
of the in-plane tensile loads that develop as temperature is decreased. At the
lowest temperature considered (50jC), a maximum deflection of 3.7 mm is
predicted.
4 SPECIALLY ORTHOTROPIC LAMINATES SUBJECT

In this section, we consider the specially orthotropic plate shown in Fig. 3. As
in the preceding section, the plate is rectangular with thickness t and in-plane
dimensions a b. However, we now assume that each edge of the plate is
subject to simple supports of type S4, rather than type S1. Hence displacement
Figure 3 Thin rectangular plate of thickness t and in-plane dimensions a b,

py
subjected to a transverse load q(x,y) = qosin(px
a )sin( b ). All four edges of the
plate are subject to simple supports of type S4 (compare with Figure 1).

fields along the plate edges are not required to take on any specified value.
Rather, we assume that a uniform normal stress resultant is applied to each
edge, as shown in Fig. 3. Note that shear loading is not considered: Nxy =
Nyx = 0. The plate is also subjected to a uniform change in temperature, DT,
and a transverse loading that varies over the x–y plane according to:
px py
a b
We will not consider any change in moisture content, i.e., DM = 0. Also, we
will not consider the possibility of buckling in this section, although buckling
is a possibility if either of the in-plane normal stress resultants is compressive.
Buckling under type S4 simple supports will be discussed in Sec. 9.
The boundary conditions that define type S4 simple support were
discussed in Sec. 3.3.2 of Chap. 9. Because the stress resultants applied along
each edge are uniform and constant, we can write:
* ð0; yÞ ¼ N* ða; yÞ ¼ N
Nxx xx xx
N*yy ðx; 0Þ ¼ Nyy

* ðx; bÞ ¼ N
yy
N*xy ð0; yÞ ¼ Nxy

* ða; yÞ ¼ N* ðx; 0Þ ¼ N* ðx; bÞ ¼ 0
yx yx
Because we have limited consideration to symmetric specially orthotropic

laminates, A16 ¼ A26 ¼ D16 ¼ D26 ¼ Bij ¼ NTxy ¼ NM xy ¼ Mij ¼ Mij = 0.
T M
Therefore the boundary conditions can be written as:
For x = 0,a:
wð0; yÞ ¼ wða; yÞ ¼ 0 ð10aÞ
* ð0; yÞ ¼ Mo ða; yÞ ¼ D @2w @2w

Mxx 11 þ D12 ¼0 ð10bÞ
xx
@x2 @y2
* ð0; yÞ ¼ N* ða; yÞ ¼ A @uo þ A @vo NT ¼ N

Nxx ð10cÞ
11 12 xx
xx
@x @y xx

* ð0; yÞ ¼ N* ða; yÞ ¼ A @uo @vo
Nxy 66 þ ¼0 ð10dÞ
xy
@y @x
For y = 0,b:
wðx; 0Þ ¼ wðx; bÞ ¼ 0 ð11aÞ

* ðx; 0Þ ¼ M* ðx; bÞ ¼ D @ w þ D @ w ¼ 0
2 2
Myy 12 22 ð11bÞ
yy
@x2 @y2
* ðx; 0Þ ¼ N* ðx; bÞ ¼ A @uo þ A @vo NT ¼ N
Nyy ð11cÞ
12 22 yy
yy
@x @y yy

* * @uo @vo
Nyx ðx; 0Þ ¼ Nyx ðx; bÞ ¼ A66 þ ¼0 ð11dÞ
@y @x
Using Eqs. (3a) and (3b) for the assumed conditions (i.e., Nxy = DM = 0), the
in-plane displacements fields are:

uo ðx; yÞ ¼ a11 Nxx þ a12 Nyy þ DT axx x ð12aÞ

vo ðx; yÞ ¼ a12 Nxx þ a22 Nyy þ DT ayy y ð12bÞ
Let us confirm that these equations satisfy the appropriate boundary con-
ditions. Substituting Eqs. (12a) and (12b) into boundary condition (10c) and
rearranging, we find that the following expression must be satisfied:
Nxx ðA11 a11 þ A12 a12 Þ þ Nyy ðA11 a12 þ A12 a22 Þ
þ DTðA11 axx þ A12 ayy Þ NTxx ¼ Nxx
Because the laminate is specially orthotropic,

2 3
A11 A12 0
½A ¼ 4 A12 A22 0 5
0 0 A66
2 3
A22 A12
6 ðA A A2 Þ ðA A A2 Þ 0 7
6 11 22 12 11 22 12 7
6 A12 7
1 6 A11 7
½a ¼ ½A ¼ 6 0 7
6 ðA11 A22 A212 Þ ðA11 A22 A212 Þ 7
6 7
4 1 5
0 0
A66
Therefore by direct substitution we find:
ðA11 a11 þ A12 a12 Þ ¼ 1
ðA11 a12 þ A12 a22 Þ ¼ 0
Also, from Eq. (73b) of Chap. 6, we find (for A16 = 0):
NTxx ¼ DTðA11 axx þ A12 ayy Þ

Hence the boundary condition represented by Eq. (10c) is satisfied identically
by the in-plane displacement fields. A similar process can be used to confirm
that the boundary condition given by Eq. (11c) is also satisfied. Both Eqs.
(10d) and (11d) are satisfied as well, because @uo/@y = @vo/@x = 0.
Now consider out-of-plane displacements w(x,y). Guided by the func-
tional form of the transverse pressure [i.e., Eq. (9)], we once again assume the
out-of-plane displacement field is given by:
px py
a b
where c is an unknown constant. Substituting this assumed form into
boundary conditions [Eqs. (10a), (10b), (11a), and (11b)] will reveal that they
are identically satisfied for any value of c. Hence the value of constant c must
be determined by enforcing the third equation of equilibrium. Substituting
Eqs. (12a), (12b), and (13) into the third equation of equilibrium, Eq. (1c),
and solving for constant c, we find:
qo
c¼
ð14aÞ
1 2 1 1 1
p 4 D11 þ 2 2 ðD12 þ 2D66 Þ þ 4 D22 þ 2 2 Nxx þ 2 2 Nyy
4
a ab b p a pb
Using the definition of the plate aspect ratio, R = a/b, this result can also be
written as:
qo R 4 b4
c¼
ð14bÞ
a2
p4 D11 þ 2R2 ðD12 þ2D66 Þ þ R4 D22 þ 2
Nxx þ Nyy R2
p
The predicted out-of-plane deflection is obtained by combining either Eq.

(14a) or (14b) with Eq. (13). Using Eq. (14b) for example, we have:
wðx; yÞ ¼
ð15Þ
a2
p D11 þ 2R ðD12 þ 2D66 Þ þ R D22 þ 2 Nxx þ Nyy R
4 2 4 2
p
Equations (12a), (12b), and (15) give the predicted displacement fields in-
duced in the plate and represent the solution to this problem.
To summarize, we have found the displacement fields induced in a
symmetric specially orthotropic type S4 simply supported plate subjected to a
sinusoidally varying transverse load, a uniform change in temperature DT,
and uniform stress resultants Nxx and Nyy. A typical application of this
solution is discussed in Sample Problem 2. It should be kept in mind that we

have not considered the possibility of buckling in this section. The solution we
have obtained is not valid if Nxx and/or Nyy are compressive and have
magnitudes large enough to cause buckling. Buckling under type S4 simple
supports will be discussed in Sec. 9.
Sample Problem 2
A [(02/90)2]s graphite–epoxy laminate is cured at 175jC and then cooled to
plane dimensions of 300150 mm and mounted in an assembly that provides
the 300-mm edge (i.e., a = 0.3 m; b = 0.15 m). The laminate is then subjected
to a uniform in-plane tensile loading (i.e., Nxx = Nyy) and transverse pressure
given by q(x,y) = 40 {sin [(px)/a]} {sin[(py)/b]} (kPa). Temperature remains
constant and no change in moisture content occurs (DM = 0).
(a) Plot the out-of-plane displacements induced along the centerline
defined by y = 0.075 m, if in-plane loads Nxx = Nyy = 50 kN/m
are applied.
(b) Plot the maximum out-of-plane displacement as a function of in-
plane loads, over the range 0<(Nxx = Nyy)<70 kN/m.
(c) Compare the maximum out-of-plane displacement caused by the
specified transverse load at room temperature if the plate is sub-
ject to
(i) type S1 simple supports (as discussed in Sec. 3), and
(ii) type S4 simple supports, with Nxx = Nyy = 0
Use the properties listed for graphite–epoxy in Table 3 of Chap. 3, and assume
Solution. A [(02/90)2]s graphite–epoxy laminate was also considered in
Sample Problem 1, and numerical values for the [ABD] matrix are listed
there. As before, the 12-ply laminate has a total thickness t = 1.5 mm and
aspect ratio R = a/b = 2.0. Using these laminate stiffnesses, dimensions, and
the specified transverse loading, Eq. (15) becomes:
2 3
6 324 7 px py
wðx; yÞ6
4 7sin sin ðmetersÞ
0:090 5 0:3 0:15
p 278 þ 2 Nxx þ 4Nyy
4
p
This expression can be used to calculate the out-of-plane displacement

induced at any point (x,y) over the surface of the plate.

Part (a). Using Nxx ¼ Nyy = 50 kN/m and y = 0.075 m, out-of-plane dis-
placements are:
px
wðx; yÞ ¼ ½1:30 103 sin ðmetersÞ
0:3
A plot of these displacements over the length of the plate (0<x<0.3m is
shown in Fig. 4(a). Displacements are zero at the edges defined by x = 0, 0.3 m,
as dictated by the specified boundary conditions. As would be expected
because of symmetry, out-of-plane displacement is maximum at the center
of the plate, and equals 1.3 mm for the loading considered.
Part (b). The maximum out-of-plane displacement occurs at the center of
the plate, i.e., at x = a/2 = 0.15 m and y = b/2 = 0.075 m, and at this point
the maximum out-of-plane displacement is given by:
2 3
6 324 7
wjmax ¼ 6
4 7 ðmetersÞ
0:090 5
p4 278 þ 2 Nxx þ 4Nyy
p
A plot of maximum out-of-plane displacement as a function of in-plane tensile
loads is shown in Fig. 4b. As would be intuitively expected, the plate is
stiffened by the application of in-plane loading. That is, the maximum out-of-
plane displacement is decreased as in-plane tensile loads are increased. A
maximum deflection of 12 mm occurs when Nxx = Nyy = 0, whereas the max-
imum deflection is reduced to 0.96 mm when Nxx = Nyy = 70 kN/m.
Part (c). This same panel and transverse loading was considered in
Sample Problem 1, except type S1 simple supports were assumed. Thus in
Sample Problem 1 in-plane displacements were fixed and were not allowed
to change when the transverse load was applied. In contrast, type S4 simple
supports are assumed in this problem; in-plane stress resultants are specified
rather than in-plane displacements.
Referring to the results presented in these two sample problems, we
find that identical deflections are predicted, despite the differences in
boundary conditions. That is, a maximum deflection of 12 mm is predicted
at room temperature for type S1 condition, and an identical 12 mm
deflection is predicted if Nxx = Nyy = 0 for type S4 conditions. This result
may seem nonintuitive and (rigorously speaking) is incorrect. That is, for
type S4 boundary conditions, a transverse loading will cause a change in in-
plane displacements. Therefore one might anticipate that the out-of-plane
displacement for type S4 conditions would be increased, relative to type S1
conditions. However, the relative increase is very small if displacement
gradients are small. Thus the relative increase in out-of-plane displacements

Figure 4 Out-of-plane displacements induced in the graphite–epoxy plate
considered in Sample Problem 2. (a) Out-of-plane displacements induced along
centerline y = 0.075 m when Nxx = Nyy = 50 kN/m, (b) Maximum out-of-plane
displacements as a function of in-plane loads Nxx = Nyy.

for type S4 conditions is not predicted because we have based our analysis
on infinitesimal strains. The consequences of the infinitesimal strain as-
sumption were alluded to in Sec. 2.2 of Chap. 9. It was noted there that this
assumption ultimately leads to the conclusion that in-plane displacement
fields uo(x,y) and vo(x,y) are independent of the transverse load, q(x,y). The
comparison between the results of Sample Problems 1 and 2 presented here
is an illustration of this independence.
Of course, results for the two different boundary conditions are iden-
tical because we have considered the case in which Nxx = Nyy = 0. If Nxx
and/or Nyy p 0, then the transverse displacements for a plate supported by
type S4 simple supports is quite different from that of a plate supported by
type S1 supports.
5 SPECIALLY ORTHOTROPIC LAMINATES WITH TWO

SIMPLY SUPPORTED EDGES OF TYPE S1 AND TWO
EDGES OF TYPE S2
In this section, we consider the specially orthotropic plate shown in Fig. 5. As
in the preceding section, the plate is rectangular with thickness t and in-plane
Figure 5 Thin rectangular plate of thickness t and in-plane dimensions a x b,

subjected to a transverse load q(x,y) = qo {sin[(px)/a]}{sin[(py)/b]}. Edges x =
0,a are subject to simple supports of type S2, whereas edges y = 0,b are subject
to simple support of type S1 (compare with Figs. 1 and 3).

dimensions a b. We now assume that the two edges x = 0,a are subjected to
simple supports of type S2. That is, we specify that these two edges are
subjected to known stress resultants Nxx and known displacements vo. In
contrast, the two edges y = 0,b are subject to simple supports of type S1, which
means that we require these edges to maintain known in-plane displacements
uo and vo. The plate is also subjected to a uniform change in temperature, DT,
and a transverse loading that varies over the x–y plane according to:
px py
qðx; yÞ ¼ qo sin ð16Þ
a b
We will not consider a change in moisture content, i.e., DM = 0. Also, we will

not consider the possibility of buckling in this section, although buckling is a
possibility if Nxx is compressive or if the change in temperature DT tends to
cause the laminate to expand.
As in earlier sections, we assume the laminate was cured at an elevated
temperature and cooled to room temperature prior to assembly in a sur-
rounding structure that provides simple supports. Consider the midplane
strains vo(x,y) induced during cooldown. Using Eq. (3b), this can be written
(with Nxx = Nyy = Nxy = DM = 0):
vo ðx; yÞ ¼ ðDT c ayy Þy ð17Þ
We can use this expression to specify known values of vo along all four edges
of the plate. Because a known stress resultant Nxx is applied to the two edges
x = 0 and x = a, all quantities necessary to define the type S2 simple support
along these two edges are known. To define the type S1 simple supports along
the edges y = 0,b, we must specify known values of uo. Toward that end,
equate Eqs. (3b) and (17):
1
ða66 Nxy Þx þ a12 Nxx þ a22 Nyy þ DT ayy y ¼ ðDT c ayy Þy
2
This relation must be satisfied along the edges y = 0,b. Substituting y = 0, it is
seen that Nxy must vanish (Nxy = 0). Substituting y = b, and solving for Nyy,
we obtain:
ayy ðDT c DTÞ a12 Nxx
Nyy ¼ ð18Þ
a22
Equation (18) gives the stress resultant Nyy that must be provided by the
simple supports along the y = 0,b so as to maintain the stipulated value of vo
represented by Eq. (17). Note that if temperature does not change from room
temperature (i.e., if DT = DT c ), and if no stress resultant Nxx is applied (i.e.,
if Nxx = 0), then Nyy = 0, as would be expected.

We can now calculate the displacement uo implied by these conditions,
using Eq. (3b):

ayy ðDT c DTÞ a12 Nxx

uo ðx; yÞ ¼ a11 Nxx þ a12 þ DT axx x
a22
This expression can be written as:

a212 a12
uo ðx; yÞ ¼ Nxx a11 þ ayy ðDT DTÞ þ axx DT x
c
ð19Þ
a22 a22
Equations (17) and (19) can be used to specify the displacement boundary
conditions along all four edge of the plate. A summary of all boundary con-
ditions associated with the problem considered in this section is:
For x = 0,a
wð0; yÞ ¼ wða; yÞ ¼ 0 ð20aÞ
* ð0; yÞ ¼ M* ða; yÞ ¼ D @ w þ D @ w ¼ 0
2 2
Mxx 11 12 ð20bÞ
xx
@x2 @y2
* ð0; yÞ ¼ N* ða; yÞ ¼ A @uo þ A @vo NT ¼ N
Nxx ð20cÞ
11 12 xx
xx
@x @y xx
vo ð0; yÞ ¼ vo ða; yÞ ¼ ðDT c ayy Þy ð20dÞ
For y = 0,b:
wðx; 0Þ ¼ wðx; bÞ ¼ 0 ð21aÞ
* ðx; 0Þ ¼ M* ðx; bÞ ¼ D @ w þ D @ w ¼ 0
2 2
Mxx 12 22 ð21bÞ
yy
@x2 @y2
uo ðx; 0Þ ¼ uo ðx; bÞ

a212 a12
¼ Nxx a11 þ ayy ðDT DTÞ þ axx DT x
c
ð21cÞ
a22 a22
vo ðx; 0Þ ¼ 0 vo ðx; bÞ ¼ ðDT c ayy Þb ð21dÞ
Now consider out-of-plane displacements w(x,y). Guided by the func-
tional form of the transverse pressure [i.e., Eq. (16)], we once again assume the
out-of-plane displacement field is given by:
px py
wðx; yÞc sin sin ð22Þ
a b
where c is an unknown constant. Substituting this assumed form into
boundary conditions Eqs. (20a), (20b), (21a), and (21b) will reveal that they

are identically satisfied for any value of c. Hence the value of constant c must
be determined by enforcing the third equation of equilibrium. Substituting
Eqs. (19), (17), and (22) into the third equation of equilibrium, Eq. (1c), and
solving for constant c, we find:
qo
c¼
D11 2ðD12 þ 2D66 Þ D22 Nxx 1 A12 ayy ðA12 A22 A212 Þ
p4 4 þ þ 4 þ 2 þ ðDT c
DT Þ
a a2 b2 b p a2 A11 b2 p2 b2 A11
ð23Þ
written as:
qo R4 b4
c¼
R b 4 2
1 A12 ayy ðA12 A22 A212 Þ
p4 D11 þ 2R2 ðD12 þ 2D66 Þ þ R4 D22 þ 2 Nxx 2 þ ðDT c DT Þ
p R A11 A11
ð24Þ
The predicted out-of-plane deflection is obtained by combining either Eq.

(23) or (24) with Eq. (22). Using Eq. (24), for example, we have:
wðx; yÞ ¼
R4 b2 1 A12 ayy ðA12 A22 A212 Þ

p4 D11 þ 2R2 ðD12 þ 2D66 Þ þ R4 D22 þ 2 Nxx 2 þ ðDT c DT Þ
p R A11 A11
ð25Þ
Equations (17), (19), and (25) give the predicted displacement fields induced
in the plate and represent the solution to this problem.
To summarize, we have found the displacement fields induced in a
symmetric specially orthotropic plate subjected to a sinusoidally varying
transverse load and a uniform change in temperature DT. The two edges x =
0,a are subject to simple supports of type S2, whereas the two edges y = 0,b
are subject to simple supports of type S1. A typical application of this solution
is discussed in Sample Problem 3. It should be kept in mind that the
possibility of buckling has not been considered. Buckling is a possibility if
Nxx is compressive or if the change in temperature DT tends to cause the
laminate to expand.
Sample Problem 3A
[(02/90)2]s graphite–epoxy laminate is cured at 175jC and then cooled to

type S2 simple supports along the two edges x = 0 and x = a, and type
S1 simple supports along the two edges y = 0 and y = b. The x axis is
defined parallel to the 300-mm edge (i.e., a = 0.3 m; b = 0.15 m). The
laminate is then subjected to a uniform in-plane tensile loading Nxx along x =
0,a and a transverse pressure given by q(x,y) = 40 {sin[(px)/a]}{sin[(py)/b]}
(kPa).
(a ) Plot the maximum out-of-plane displacement as a function of tensile
load over the range 0 <Nxx < 70 kN/m, assuming temperature
remains constant at room temperature.
(b) Plot the maximum out-of-plane displacement as a function of
temperature, over the range 50jC<T<20jC, assuming a con-
stant in-plane tensile load Nxx = 50 kN/m.

Sample Problem 1, and numerical values for the [ABD] matrix and effective
thermal expansion coefficients are listed there. As before, the 12-ply laminate
has a total thickness t = 1.5 mm and aspect ratio R = a/b = 2.0. Using these
laminate stiffnesses, dimensions, and the specified transverse loading, Eq. (25)
becomes:
2 3
6 324 7 px py
wðx; yÞ6
4 7
5sin 0:3 sin 0:15 ðmetersÞ
0:099 82:5
p4 278 þ Nxx þ 2 ðDTc DTÞ
p2 p
This expression can be used to calculate the out-of-plane displacement

induced at any point (x,y) over the surface of the plate.
Part (a). The maximum out-of-plane displacement occurs at the center of
the plate, i.e., at x = a/2 = 0.15 m and y = b/2 = 0.075. Because the plate
remains at room temperature, DT = DT c, under these conditions the
maximum out-of-plane displacement is given by:
324
wjmax ¼ ðmetersÞ
0:099
p 278 þ 2 Nxx
4
p
A plot of maximum out-of-plane displacement as a function of Nxx is shown

in Fig. 6(a). As would be intuitively expected, the plate is stiffened as Nxx is

Figure 6 Out-of-plane displacements induced in the graphite–epoxy plate
considered in Sample Problem 3. (a) Maximum out-of-plane displacements at
room temperature as a function of in-plane load Nxx. (b) Maximum out-of-plane
displacements as a function of temperature (Nxx = 50 kN/m).

increased. A maximum deflection of 12 mm occurs when Nxx = 0, whereas
the maximum deflection is reduced to 3.4 mm when Nxx = 70 kN/m.
Part (b). As before, the maximum out-of-plane displacement occurs at x =
a/2 = 0.15 m and y = b/2 = 0.075. Because a constant in-plane tensile load
Nxx = 50 kN/m is applied, the maximum out-of-plane displacement is given
by:
324
wjmax ¼
ðmetersÞ
82:5
p4 781 þ 2 ðDT c DTÞ
p
A plot of maximum out-of-plane displacement as a function of temperature is
shown in Fig. 6(b). At room temperature (20jC), a maximum deflection of 4.3
mm is predicted. As would be expected, the plate becomes stiffer as the
temperature is decreased. Out-of-plane displacements are decreased because
of the in-plane tensile load Nyy that develops as temperature is decreased. At
the lowest temperature considered (50jC), a maximum deflection of 2.4 mm
is predicted.
6 THE NAVIER SOLUTION APPLIED TO A SPECIALLY

ORTHOTROPIC LAMINATE SUBJECT TO SIMPLE
SUPPORTS OF TYPE S4
In Sec. 4, we developed the solution for a specially orthotropic laminate
subjected to homogenous simple supports of type S4 along all four edges,
uniform in-plane loads Nxx and Nyy, a uniform temperature change DT, and a
sinusoidal transverse loading given by:
px py
qðx; yÞ ¼ qo sin sin
a b
In-plane and transverse displacements caused by this thermomechanical
loading are given by Eqs. (12a), (12b), and (15).
Now recall from Sec. 4 of Chap. 9 that any transverse loading may be
represented in terms of the double-Fourier series given by Eq. (14) of Chap. 9,
Xl X l mpx npy
qðx; yÞ ¼ qmn sin sin ð9:14Þ
m¼1 n¼1
a b
Thus any transverse loading can be viewed as the sum of a large number of
sinusoidal load components. The displacements caused by an arbitrary
transverse loading applied to a plate with simple supports of type S4 can
therefore be obtained using the same approach as that used in Sec. 4. The only

differences are that the transverse load is given by Eq. 14 of Chap. 9 and the
transverse deflections w(x,y) are assumed to be of the form:
Xl X l mpx npy
wðx; yÞ ¼ cmn sin sin ð26Þ
m¼1 n¼1
a b
where cmn are unknown constants to be determined using the equations of

equilibrium. The assumed form and solution for in-plane displacements
uo(x,y) and vo(x,y) remain unchanged and are given by Eqs. (12a) and
(12b). Substituting Eqs. (26) and Eq. (14) of Chap. 9 into the third equation
of equilibrium, Eq. (1c), and equating coefficients (following the same
procedures as used in Sec. 4), we obtain:
qmn a4 b4
cmn ¼ " 2 n #
ab o
4 2 4 2 2
p4 D11 ðmbÞ þ 2ðD12 þ 2D66 ÞðmnabÞ þ D22 ðnaÞ þ Nxx ðmbÞ þ Nyy ðnaÞ
p
ð27aÞ
Using the definition of the plate aspect ratio, R=a/b, this result can also be
written as:
qmn R4 b4
cmn ¼ o
a2 n
p4 D11 m4 þ 2ðD12 þ 2D66 ÞðmnRÞ2 þ D22 ðnRÞ4 þ Nxx m2
þ N yy ðnRÞ 2
p2
ð27bÞ
Substituting this result into Eq. (26) completes the solution to the problem.
To summarize, the midplane displacements induced in a symmetric specially
orthotropic laminate subjected to an arbitrary transverse loading given by
Eq. (15) of Chap. 9, constant and uniform in-plane loads Nxx and Nyy, a
uniform temperature change DT, and homogenous simple supports of type S4
along all four edges are given by:

A22 Nxx A12 Nyy

uo ðx; yÞ ¼ þ axx DT x ð28aÞ
A11 A22 A212

A11 Nyy A12 Nxx

vo ðx; yÞ ¼ þ ayy DT y ð28bÞ
A11 A22 A212
R4 b4 X
l Xl
qmn sinðmpx=aÞsinðnpy=bÞ
wðx; yÞ ¼ o
p4 m¼1 n¼1 a2 n
D11 m4 þ 2ðD12 þ 2D66 ÞðmnRÞ2 þ D22 ðnRÞ4 þ 2 Nxx m2 þ Nyy ðnRÞ2
p
ð28cÞ

The method of using a double-Fourier series expansion to represent the
transverse loading and out-of-plane displacements was first suggested by
Navier in about 1820, and is known as the Navier solution. An application of
this technique is illustrated in Sample Problem 4. The solution presented here
is for a plate subjected to type S4 simple supports, but the same approach may
be used to study simply supported plates with type S1 supports (as in Sec. 3),
or plates with mixed simple supports (as in Sec. 5).
As previously mentioned, a second technique known as the Levy
solution [1–5] can also be used to obtain exact solutions for symmetric
specially orthotropic laminates. While the Navier solution requires that all
four edges be simply supported, the Levy solution only requires that two
opposite edges of the plate are simply supported; the boundary conditions on
the remaining two edges are arbitrary and may be clamped or free, for
example. The Levy solution technique will not be discussed in this textbook,
and the interested reader is referred to the references cited for a discussion of
this approach.
Sample Problem 4
plane dimensions of 300150 mm and mounted in an assembly that provides
the 300 mm edge (i.e., a = 0.3 m; b = 0.15 m). The laminate is then subjected
to a uniform in-plane tensile loading Nxx = Nyy = 50 kN/m and uniform
transverse loading q(x,y) = 100 kPa. The temperature remains constant and
no change in moisture content occurs (DM = 0). Plot the out-of-plane dis-
placements induced along the centerline defined by y = 0.075 m. Use the
properties listed for graphite–epoxy in Table 3 of Chap. 3, and assume each
ply has a thickness of 0.125 mm.
aspect ratio R = a/b = 2.0.
The double-Fourier series expansion of a uniform transverse loading
was discussed in Sec. 4 of Chap. 9. The coefficients in the Fourier series
expansion were found to be:
16qo
qmn ¼ ; m; n ¼ odd integers
p2 mn
Combining these coefficients with Eq. (28c) allows prediction of out-of-plane
displacements. A plot of these displacements along the centerline of the plate

Figure 7 Out-of-plane displacements induced in the graphite–epoxy plate con-
sidered in Sample Problem 4, along centerline y = 0.075 m.
defined by y = b/2 = 0.075 m is shown in Fig. 7. Curves are shown based on a

1-term expansion (i.e., m,n = 1), a 4-term expansion (m,n = 1,3), and a 9-
term expansion (m,n = 1,3,5). As would be expected due to symmetry, out-of-
plane displacement is maximum at the center of the plate. The solution
rapidly converges. The maximum displacement predicted on the basis of a 1-,
4-, and 9-term expansion equals 5.27, 4.71, and 4.78 mm, respectively. If 100
terms were used (m,n = 1,3,. . .,19), the maximum predicted displacement is
4.77 mm.
7 BUCKLING OF RECTANGULAR SPECIALLY ORTHOTROPIC

LAMINATES SUBJECT TO SIMPLE SUPPORTS OF TYPE S4
This section is devoted to the phenomenon known as plate ‘‘buckling.’’ A
brief discussion of what the term ‘‘buckling’’ means in the context of a thin
plate is in order. Consider an initially flat symmetric laminate subjected to
ðxÞ ðþxÞ ðyÞ
constant and uniform boundary edge loads Nxx ¼ Nxx = Nxx and Nyy =
ðþy Þ
Nyy = Nyy (shear resultants are assumed zero: Nxy = Nyx = 0). As we have
seen in preceding sections, coupling stiffnesses Bij = 0 for all symmetric lam-
inates. Hence, according to our earlier analyses, we would not expect these
edge loads to cause out-of-plane displacements because in-plane loads and
out-of-plane displacements are (apparently) uncoupled for a symmetric
laminate.

This statement is always true if both Nxx and Nyy are tensile; a flat
symmetric laminate subjected to tensile edge loads Nxx and/or Nyy will remain
flat regardless of the magnitude of load (unless, of course, the loads are high
enough to cause fracture). However, suppose that either Nxx or Nyy (or both)
are compressive. If the magnitude of the compressive load(s) is (are) relatively
low, then the initially flat laminate remains flat; that is, only in-plane
displacements uo(x,y) and vo(x,y) are induced and out-of-plane displacement
remain zero: w(x,y)=0. However, as the compressive load(s) is (are) in-
creased, the laminate may exhibit a sudden out-of-plane displacement, w(x,y)
p 0. This is the phenomenon is known as ‘‘buckling.’’
The load level at which out-of-plane displacements initially occur is
called the critical buckling load. In general, a thin plate does not collapse at the
critical buckling load, and can support a further increase in load. However,
out-of-plane displacements rapidly increase as in-plane loads are increased
beyond the initial buckling load, signaling imminent structural failure.
The out-of-plane displacements that occur at the onset of buckling
exhibit a characteristic pattern. This characteristic pattern is called the
buckling mode. A typical buckling mode is illustrated in Fig. 8. A three-
dimensional view of the buckling mode is shown in Fig. 8(a), while a 2-D view
(which can be thought of as a ‘‘topographical map’’ of out-of-plane displace-
ments) is shown in Fig. 8(b). In the discussion to follow, predicted buckling
modes will be illustrated in a form similar to Fig. 8(b).
The buckling mode exhibited by a given laminate depends on the
stiffness of the laminate, plate aspect ratio, applied loading, and the boundary
conditions. A particular buckling mode is described in terms of the number of
points of relative maximum out-of-plane displacements in the x and y
directions. For the buckling mode shown in Fig. 8, there are two points of
relative maximum displacement in the x direction, whereas there is only one
point of relative maximum displacement in the y direction. Hence the pattern
shown in Fig. 8 is called ‘‘mode [2,1].’’
In this section, we consider buckling of an initially flat rectangular
specially orthotropic laminate subjected to simple supports of type S4 and
ðxÞ ðþxÞ ðyÞ ðþyÞ
uniform edge loads Nxx ¼ Nxx ¼ Nxx and Nyy ¼ Nyy = Nyy, at least one
of which is compressive. If the laminate was cured at an elevated temperature
and then cooled to room temperature, then pre-existing thermal stress re-
sultants associated with cooldown, NTxx and NTyy , are present. It is assumed
that no transverse loading is applied and that no change moisture content
occurs: q(x,y) = DM = 0.
Because the applied edge loads are constant and uniform, prior to
buckling the internal stress resultants at all points within the laminate are also
constant and uniform, and are equal to the external edge loads. Hence, prior

Figure 8 A typical mode [2,1] buckling mode, caused in a rectangular plate by
a uniaxial compressive loading Nxx.
to buckling, the internal stress resultants at all points within the laminate are
uniform and given by:
@uo @vo
N*xx ðx; yÞ ¼ A11 þ A12 NTxx ¼ NðxÞ ðþxÞ
xx ¼ Nxx ¼ Nxx ð29aÞ
@x @y
* ðx; yÞ ¼ A @uo þ A @vo NT ¼ NðyÞ ¼ NðþyÞ ¼ N
Nyy ð29bÞ
12 22 yy
@x @y yy yy yy

* ðx; yÞ ¼ A @uo @vo
Nxy 66 þ ¼0 ð29cÞ
@y @x
It is emphasized that Eqs. (29a), (29b), and (19c) are valid prior to buckling.
After buckling has occurred and significant out-of-plane displacements have

developed, the internal stress and moment resultants may no longer be
constant and uniform, and may not equal the applied edge loads.
The remaining type S4 simple support boundary conditions are:
Along x = 0:
wðx; yÞ ¼ 0 ð30aÞ
* ð0; yÞ ¼ D @ 2 wð0; yÞ @ 2 wð0; yÞ

Mxx 11 þ D12 ¼0 ð30bÞ
@x 2 @y2
Along x = a:
wða; yÞ ¼ 0 ð31aÞ
* ða; yÞ ¼ D @ wða; yÞ
2
@ wða; yÞ
2
Mxx 11 þ D12 ¼0 ð31bÞ
@x2 @y2
Along y = 0:
wðx; 0Þ ¼ 0 ð32aÞ
* ðx; 0Þ ¼ D @ 2 wðx; 0Þ @ 2 wðx; 0Þ

Myy 12 þ D22 ¼0 ð32bÞ
@x 2 @y2
Along y = b:
wðx; bÞ ¼ 0 ð33aÞ
* ðx; bÞ ¼ D @ wðx; bÞ
2
@ wðx; bÞ
2
Myy 12 þ D22 ¼0 ð33bÞ
@x 2 @y2
The equation of equilibrium governing out-of-plane displacements, Eq. (1c),
is [with q(x,y) = DM = 0]:

2
@uo @vo @ w
A11 þ A12 NTxx
@x @y @x2

2
@uo @vo @ w
þ A12 þ A22 Nyy
T
@x @y @y2
ð34Þ

2
@uo @vo @ w @4w
þ 2 A66 þ D11 4
@y @x @x@y @x
@4w @4w
2ðD12 þ 2D66 Þ D 22 ¼0
@x2 @y2 @y4
Equations (29a), (29b), and (29c) are valid up to the onset of buckling; that is,
we assume that internal stress resultants are constant and uniform as the

buckling phenomenon is approached. Hence substituting Eqs. (29a), (29b),
and (29c) into Eq. (34), we have:
@4w @4w @4w
D11 þ 2ðD 12 þ 2D 66 Þ þ D22
@x4 @x2 @y2 @y4
2 2 ð35Þ
@ w @ w
¼ Nxx þ Nyy
@x2 @y2
Notice that thermal stress and moment resultants do not appear in Eq. (35).
This reveals that the buckling response of symmetric specially orthotropic
laminates is independent of temperature, when subject to simple supports of
type S4. As will be discussed in Sec. 8, if the laminate were subject to
boundary conditions that constrain in-plane displacements, e.g., simple
supports of type S1, then buckling may be caused by a temperature change.
We now assume out-of-plane displacements are given by:
mpx npy
a b
where m and n are positive integers and c is an unknown constant representing
the maximum out-of-plane deflection. It is easy to show that Eq. (36) satisfies
the boundary conditions, Eqs. (30a), (30b), (31a), (31b), (32a), (32b), and
(33a), (33b), so the next step is to evaluate the conditions under which the
equation of equilibrium governing out-of-plane deflections is satisfied. Sub-
stituting Eq. (36) into Eq. (35) and simplifying, we find that in order for
equilibrium to be maintained, it is required that:
mp 4 mp 2 np 2 np 4
D11 þ2ðD12 þ 2D66 Þ þD22
a a b b
mp 2 np 2
¼ Nxx Nyy ð37Þ
a n
The unknown constant c has canceled out and does not appear in the
requirement for equilibrium, Eq. (37). Referring to Eq. (36), note that integers
m and n dictate the spatial variation of the out-of-plane displacement field,
i.e., m and n define how w(x,y) varies with x and y, and consequently define
the buckling mode. Constant c represents the magnitude of w(x,y). Hence
while we are able to determine m and n and thus predict the buckling mode
on the basis of Eq. (37), we cannot predict the magnitude of out-of-plane
displacements.
Although Eq. (37) is valid for any combination of Nxx and Nyy, it is
convenient to rearrange Eq. (37) for three different loading conditions. As
explained above, Nxx and/or Nyy must be compressive to cause buckling.
Assume for the moment that a constant transverse tension Nyy z 0 is applied,

in which Nxx case must be compressive to cause buckling. Solving Eq. (37) for
Nxx, we find:
"
p2
Nxx ¼ 2 2 2 D11 ðmbÞ4 þ 2ðD12 þ 2D66 ÞðmnabÞ2
m a b
ð38Þ
2 2 #
na b
þ D22 ðnaÞ4 þ Nyy
p
written as:
"
p2
Nxx ¼ D11 m4 þ 2ðD12 þ 2D66 ðmnRÞ2
ðmaÞ2
# ð39Þ
4 naR 2
þ D22 ðnRÞ þ Nyy
p
Hence, given some value for Nyy z 0 and assumed integer values for m and n,
either Eq. (38) or (39) can be used to calculate a corresponding value for Nxx.
Because m and n can be any combination of positive integers, there are an
infinite number of values for Nxx that satisfy Eq. (38) or (39). The critical
c
buckling load, denoted Nxx , corresponds to the particular combination of m
and n that leads to the value of Nxx with lowest magnitude. The combination
of m and n that correspond to this lowest load define the predicted critical
buckling mode.
The following observations are based on inspection of Eq. (38) or (39).
Because we have assumed for the moment that Nyy z 0, all variables that
appear on the right side of the equality sign are algebraically positive. Hence
c c
Nxx must be algebraically negative, i.e., the critical buckling load Nxx is
predicted to be compressive, as would be expected. Secondly, the minimum
magnitude of Nxx c
will always correspond to n = 1 because Nyy z 0. Finally,
note that if the constant transverse load Nyy is increased, then the magnitude
c
of Nxx is increased. That is, a transverse tension will cause an increase in the
critical buckling load.
As a second loading condition of interest, let us now assume that a
constant Nxx z 0 is applied, and that a compressive load in the y direction
causes buckling. Solving Eq. (37) for Nyy, we find:
"
p2
Nyy ¼ 2 4 2 D11 ðmbÞ4 þ 2ðD12 þ 2D66 ÞðmnabÞ2
n ab
# ð40Þ
4 mab 2
þ D22 ðnaÞ þ Nxx
p

Using the plate aspect ratio, R = a/b, this result can be written as:
"
p2
Nyy ¼ 2
D11 m4 þ 2ðD12 þ 2D66 ÞðmnRÞ2
ðnbR2 Þ
#
ma 2
4
þ D22 ðnRÞ þ Nxx ð41Þ
p
As before, m and n can be any combination of positive integers, and therefore

there are an infinite number of values for Nyy that satisfy Eq. (40) or (41). The
c
critical buckling load Nyy corresponds to the particular combination of m
and n that leads to the value of Nyy with lowest magnitude, and the
combination of m and n that correspond to this load define the predicted
buckling mode. Because it has been assumed that Nxx z 0, m = 1 in all cases,
and if the constant tensile load Nxx is increased, the magnitude of the critical
c
buckling load Nyy will increase.
Finally, consider a third loading condition in which buckling is caused
by the simultaneous increase in both Nxx and Nyy. Further, assume the two
loads are linearly related. That is, assume Nyy = kNxx, where k is a known
constant. Substituting this relation into Eq. (37) and solving for Nxx, we find:
h i
4 2 4
p2 D11 ðmbÞ þ 2ðD12 þ 2D66 ÞðmnabÞ þ D22 ðnaÞ
Nxx ¼ h i ð42Þ
ðabÞ2 ðmbÞ2 þ kðnaÞ2
As before, m and n can be any combination of positive integers, and therefore

there are an infinite number of values for Nxx that satisfy Eq. (42). The critical
c c
buckling condition, defined by the two simultaneous loads, Nxx and Nyy =
c
kNxx , corresponds to the particular combination of m and n that leads to the
value of Nxx (and Nyy) with lowest magnitude. The combination of m and n
that correspond to this load condition defines the predicted buckling mode.
Note that if k < 0, then Nxx and Nyy are of opposite algebraic signs. This
implies that there are two distinct buckling load conditions, one in which
(Nxx < 0, Nyy > 0), and a second in which (Nxx > 0, Nyy < 0).
Numerical examples illustrating buckling predictions for symmetric
specially orthotropic simply supported laminates will be discussed in Sample
Problems 5 and 6. Two important concluding comments are made in passing.
First, from the preceding discussion, the reader may have inferred that
buckling is only caused by Nxx, by Nyy, or by some combination thereof.
This is not the case. Other forms of loading may cause buckling, e.g., a shear
load Nxy. In this chapter, only buckling caused by Nxx and/or Nyy is con-
sidered. The reader interested in buckling caused by other types of loading is
referred to Refs. 4 and 5.

Second, recall that the analysis presented above does not allow pre-
diction of the magnitude of out-of-plane displacements induced at buckling.
Buckling of thin plates is rarely catastrophic, and in general finite out-of-
plane displacements occur at and above the critical buckling load. That is, a
thin plate does not collapse once the buckling load is reached and can support
a further increase in load, albeit at a reduced level of stiffness. Methods to
predict the magnitude of out-of-plane displacement for load levels at or above
the initial buckling load are known as ‘‘post-buckling’’ analyses, and are not
discussed in this text.
Sample Problem 5
the 300 mm edge (i.e., a = 0.3 m; b = 0.15 m).
c
(a) Predict the critical buckling load Nxx and mode for this laminate, if
0 V Nyy V 400 kN/m.
c
(b) Predict the critical buckling load Nyy and mode for this laminate, if
0 V Nxx V 400 kN/m.
aspect ratio R = a/b = 2.0.
Part (a). It is noted that n = 1, because Nyy z 0. Equation (39) becomes in
this case:
" 2 #
p2 0:6
Nxx ¼ 40:1m þ 65:34m þ 172:8 þ Nyy
4 2
0:09m2 p
A plot of the critical buckling load for 0 V Nyy V 400 kN/m is presented in
c
Fig. 9(a). As expected, Nxx is increased as Nyy is increased. A change in
buckling mode also occurs as Nyy is increased. The plate buckles in mode
[2,1] over the range 0 V Nyy < 35 kN/m, in mode [3,1] over the range 35 kN/
m V Nyy < 150 kN/m, and in mode [4,1] over the range 150 kN/m V Nyy
< 400 kN/m. These buckling modes are illustrated in Fig. 9(b,c,d),
respectively.

Figure 9 Buckling response of the [(02/90)2]s graphite–epoxy plate considered
in Sample Problem 5.
Part (b). It is noted that m = 1, because Nxx z 0. Equation (41) becomes in

this case:

p2 4 0:09
Nyy ¼ 40:1 þ 65:34n 2
þ 172:8ðnÞ þ Nxx
0:36n2 p2
A plot of the critical buckling load for 0 V Nxx V 400 kN/m is presented in
c
Fig. 10(a). As expected, Nyy is increased as Nxx is increased. A change in

Figure 10 Buckling response of the [(02/90)2]s graphite–epoxy plate consid-
ered in Sample Problem 5.
buckling mode also occurs as Nxx is increased. The plate buckles in mode [1,1]
over the range 0 V Nxx < 72 kN/m, and in mode [1,2] over the range 72 kN/m
V Nxx < 400 kN/m. These buckling modes are illustrated in Fig. 10(b,c),
respectively.
c
Note that the magnitudes of Nyy calculated in part (b) are far lower than
c
those calculated for Nxx in part (a). This pronounced difference is largely
because of the stacking sequence involved. For the [(02/90)2]s laminate under

consideration, 8 of 12 plies are 0j plies (i.e., plies with fibers parallel to the x
axis). Hence the resistance to buckling because of a compressive load Nxx is
far higher than resistance to buckling because of a compressive Nyy.
Sample Problem 6
A structure is being designed that will involve a [(02/90)2]s graphite–epoxy
laminate with a width (in the y direction) of 150 mm. The length of the panel
(in the x direction) has not yet been established, and could be anywhere from
150 to 750 mm. During service, the panel will be subjected to a compressive
load Nxx (only), and simple supports of type S4 along all four edges. Buckling
is therefore of concern. Predict the buckling load and mode for the panel, for
any panel length ranging from 150 to 750 mm.

there. As before, the 12-ply laminate has a total thickness t = 1.5 mm.
Buckling loads and modes will be predicted using Eq. (39). According to the
problem statement, b = 0.15 m, and 0.15 m < a < 0.75 m. The plate aspect
Figure 11 Predicted buckling loads and modes as a function of aspect ratio for
the [(02/90)2]s graphite–epoxy panel considered in Sample Problem 6.

ratio therefore varies over 1 V R V 5. Because transverse loading is zero
(Nyy = 0), n = 1 and Eq. (39) becomes:
p2 h i
Nxx ¼ 2
40:1m4 þ 16:34ðmRÞ2 þ 10:8R4
ðmaÞ
A plot of the predicted critical buckling load over the specified range in aspect
ratio is presented in Fig. 11. The buckling mode is predicted to increase as the
aspect ratio increases: mode [1,1] is predicted over the range 1 <R < 1.96,
mode [2,1] is predicted over the range 1.96 < R < 3.40, mode [3,1] is predicted
for 3.40 <R < 4.81, and mode [4,1] is predicted for 4.81 < R < 5.00. Still
higher buckling modes would occur at higher aspect ratios.
The predicted buckling load generally decreases with aspect ratio,
although a local maximum in the buckling load occurs at each aspect ratio
corresponding to a change in mode shape. At an aspect ratio R = 1 (i.e., for a
c
square plate), buckling is predicted to occur at Nxx = 29.5 kN/m.
8 THERMAL BUCKLING OF RECTANGULAR SPECIALLY

ORTHOTROPIC LAMINATES SUBJECT TO SIMPLE
SUPPORTS OF TYPE S1
Buckling caused by direct application of uniform external edge loads Nxx
and/or Nyy was considered in the Sec. 7. The analysis was conducted for a
specially orthotropic laminate subjected to inhomogeneous simple supports
of type S4. The initially flat symmetric laminate was subjected to uniform
ðxÞ ðþxÞ ðyÞ ðþyÞ
boundary edge loads Nxx ¼ Nxx and Nyy = Nyy . Because we specified
in-plane edge loads, we did not specify in-plane edge displacements. In es-
sence, edge displacements were allowed to vary with changes in the specified
edge loads.
We now wish to consider buckling caused by a different mechanism.
Namely, we wish to consider buckling caused by a change in temperature. In
this case, we will specify inhomogeneous simple supports of type S1. That is,
we will specify in-plane edge displacements, but will not specify in-plane edge
loads. These boundary conditions were also considered in Sec. 3. As was
discussed there, we assume that pre-existing midplane displacements (as well
as ply strains) are induced prior to assembly of the laminate in the simply
supported configuration. That is, midplane displacements are induced during
cooling from an elevated cure temperature to room temperature. For a
symmetric specially orthotropic laminate, the midplane displacements caused
by cooling are given by Eq. (5a), (5b):
uco ðx; yÞ ¼ ðaxx DT c Þx
vco ðx; yÞ ¼ ðayy DT c Þy

where axx and ayy are the effective thermal expansion coefficients of the
laminate and DT c is the change in temperature from cure to room temper-
ature. Therefore boundary conditions along all four edges are:
For x ¼ 0 : For x ¼ a :
wð0; yÞ ¼ 0 wða; yÞ ¼ 0
* ð0; yÞ ¼ 0
Mxx * ða; yÞ ¼ 0
Mxx
ðxÞ ðþxÞ
uo ð0; yÞ ¼ uo ðyÞ0 uo ða; yÞ ¼ uo ðyÞ ¼ ðaxx DT c Þa
ðxÞ ðþxÞ
vo ð0; yÞ ¼ vo ðyÞ ¼ ðayy DT c Þy vo ð0; yÞ ¼ vo ðyÞ ¼ ðayy DT c Þy
For y ¼ 0 : For y ¼ b :
wðx; 0Þ ¼ 0 wðx; bÞ ¼ 0
* ðx; 0Þ ¼ 0
Myy * ðx; bÞ ¼ 0
Myy
ðyÞ ðþyÞ
uo ðx; 0Þ ¼ uo ðxÞ ¼ ðaxx DT c Þx uo ðx; bÞ ¼ uo ðxÞ ¼ ðaxx DT c Þa
ðyÞ ðþyÞ
vo ðx; 0Þ ¼ vo ðxÞ ¼ 0 vo ðx; bÞ ¼ vo ðyÞ ¼ ðayy DT c Þb
After assembly, we assume the laminate is subjected to a uniform change in

temperature. The change in temperature is referenced to the strain-free tem-
perature (assumed to be the cure temperature) and is represented by DT =
(current temperature) (cure temperature). If the laminate was not simply
supported and instead was free to expand or contract, then a uniform change
in temperature would simply cause a change in midplane displacements and
no external edge loads would result. Because the laminate is instead subject to
simple supports of type S1, no changes in midplane displacements are allowed
to occur and external edge loads develop as temperature changes. Depending
on the magnitude of the temperature change, these thermally induced edge
loads may cause the laminate to buckle.
Although the edge loads are thermally induced, they are external
mechanical loads nevertheless and consequently the analysis presented in
Sec. 7 is still applicable. The onset of thermal buckling occurs if Eq. (37) is
satisfied, repeated here for convenience:
mp 4 mp 2 np 2 np 4
D11 þ2ðD12 þ 2D66 Þ þD22
a a b b
mp 2 np 2
¼ Nxx Nyy
a b

In the present case, loads Nxx and Nyy are caused by a deviation from room
temperature, and are given by:
Nxx ¼ ðDT c DT ÞðA11 axx þ A12 ayy Þ ð43aÞ
Nyy ¼ ðDT c DT ÞðA12 axx þ A22 ayy Þ ð43bÞ

Substituting Eqs. (43a) and (43b) into Eq. (37), solving for the temperature
difference, and using the definition of the plate aspect ratio, R = a/b, we
obtain:
2 3
p2 6 D11 m4 þ 2ðD12 þ 2D66 ÞðmnRÞ2 þ D22 ðnRÞ4 7
ðDT c DT Þ ¼ 4 m 2 5 ð44aÞ
b2 R4 2
n ðA12 āxx þ A22 āyy Þ þ ðA11 āxx þ A12 āyy Þ
R
It is convenient to express the temperature differences involved as:
DTc DT ¼ TRT T
where TRT=room temperature (i.e., the temperature at which the laminate
was mounted in the simple support fixture); T = current temperature.With
this change in notation we have:
2 3
p2 6 D11 m4 þ 2ðD12 þ 2D66 ÞðmnRÞ2 þ D22 ðnRÞ4 7
T¼ 4 m 2 5TRT ð44bÞ
b2 R4 2
n ðA12 āxx þ A22 āyy Þ þ ðA11 āxx þ A12 āyy Þ
R
Equations (44a) and (44b) are entirely equivalent, and both are based on the
assumption that the simply supported laminate is assembled at room temper-
ature, TRT. In practice, it is conceptually simplest to use Eq. (44b). By
specifying a particular laminate and room temperature, all variables on the
right side of the equality sign in Eq. (44b) are known, except for variables m
and n. Because m and n can be any combination of positive integers, there are
an infinite number of values of temperature T that will satisfy Eq. (44b). The
temperature at which thermal buckling will occur, denoted Tbk, corresponds
to the particular combination of m and n that leads to the value of T of lowest
magnitude. The combination of m and n that correspond to this lowest
temperature define the predicted critical thermal buckling mode.
Now, based on the results of Sample Problem 6, one might anticipate
that the thermal buckling mode exhibited would vary as a function of aspect
ratio. That is, based on earlier analyses, one would expect that the values of m
and n that correspond to the critical thermal buckling load would vary with
aspect ratio. Equations (44a) and (44b) allow for such dependence. However,
if physically reasonable material properties are used during the evaluation of
Eqs. (44a) and (44b), then numerical experiments show that m = n = 1 in all

cases. Thus a thermal buckling mode [1,1] is always predicted when realistic
material properties are used (at least for the simply supported boundary
conditions considered in this section), regardless of the aspect ratio or
stacking sequence considered.
Assuming that this observation always holds true and that m = n = 1,
Eq. (44b) reduces to:
2 3
p2 6 D11 þ 2ðD12 þ 2D66 ÞR2 þ D22 R4 7
Tbk ¼ 4 5 þ TRT
b2 R 4 1
ðA12 axx þ A22 ayy Þ þ 2 ðA11 axx þ A22 ayy Þ
R
This latter result shows that the thermal buckling temperature decreases with
an increase in aspect ratio. In the limit (i.e., as R!l), the thermal buckling
temperature becomes:
p2 D22
Tbk ¼ þ TRT
b2 ðA12 axx þ A22 ayy Þ
Sample Problem 7
Two [02/90)2]s graphite–epoxy laminates are cured at 175jC and then cooled
to room temperature (20jC). After cooling, one laminate is trimmed to in-
plane dimensions of 300150 mm, whereas the second is trimmed to in-plane
dimensions of 3000150 mm. Both laminates are then mounted in assemblies
that provide type S1 simple supports along all four edges, and subjected to a
uniform increase in temperature. Determine the temperature at which each
plate will buckle. Use the properties listed for graphite–epoxy in Table 3 of
Chap. 3, and assume each ply has a thickness of 0.125 mm.
Sample Problem 1, and numerical values for the [ABD] matrix and effective
thermal expansion coefficients are listed there. The aspect ratios involved in
this problem are:
R ¼ 300 mm=150 mm ¼ 2
and
R ¼ 3000 mm=150 mm ¼ 20:
The temperature at which thermal buckling is predicted to occur is obtained
through application of Eq. (44b). It is found that both laminates are predicted
to buckle in mode [1,1]. For the laminate with aspect R = 2, thermal buckling
is predicted to occur when temperature is raised to 51jC, whereas for the
laminate with aspect ratio R = 20, thermal buckling is predicted at a
temperature of 40jC.

9 COMPUTER PROGRAM SPORTHO
The computer program SPORTHO has been developed to supplement the
material presented in this chapter. This program can also be downloaded
at no cost from the following website: http://depts.washington.edu/amtas/
computer.html.
Program SPORTHO is applicable to symmetric specially orthotropic
laminates. The program can be used to calculate the transverse deflections
according to the analyses presented in Secs. 3 through 6, to calculate buckling
loads according to the analysis presented in Sec. 7, or to calculate the
temperature at which thermal buckling will occur, according to the analysis
presented in Sec. 8. The user is prompted to input all information necessary to
perform these calculations. Properties of up to five different materials may be
defined. Numerical values must be defined using a consistent set of units in all
cases. For example, the user must input elastic moduli for the composite
material system(s) of interest. Using the properties listed in Table 3 of Chap. 3
and based on the SI system of units, the following numerical values would be
input for graphite–epoxy:
E11 ¼ 170 109 Pa E22 ¼ 10 109 Pa v12 ¼ 0:30 G12 ¼ 13 109 Pa
Because 1 Pa = 1 N/m2, all other lengths must be input in meters. For ex-
ample, ply thickness must be input in meters (not millimeters). A typical value
would be tk = 0.000125 m (corresponding to a ply thickness of 0.125 mm). In-
plane plate dimensions must also be input in meters.
If the English system of units were used, then the following numerical
values would be input for the same graphite–epoxy material system:
E11 ¼ 25:0106 psi E22 ¼ 1:5106 psi v12 ¼ 0:30 G12 ¼ 1:9106 psi
In this case, all lengths would be input in inches.
REFERENCES
1. Lekhnitskii, S.G. Anisotropic Plates, translated by S. W. Tsai and T. Cheron,
Taylor and Francis Books Ltd, London, UK, ISBN 0-677-20670-4, 1968.
2. Savin, G.N. Stress Concentration Around Holes; New York: Pergamon Press,
1961.
3. Timoshenko, S.; Woinowsky-Krieger, S. Theory of Plates and Shells; New York,
NY: McGraw-Hill Book Co (ISBN 0-07-0647798), 1987.
4. Whitney, J.M. Structural Analysis of Laminated Anisotropic Plates; Lancaster,
PA: Technomic Pub Co (ISBN 87762-518-2).
5. Buckling and Postbuckling of Composite Plates. Turvey, G.J.; Marshall, I.H.,
Eds; Chapman and Hall: New York, NY, 1995.

11
Some Approximate Solutions for Symmetric
Laminates
Exact solutions for many symmetric composite laminates commonly used in

practice cannot be found. In this chapter, methods of obtaining approximate
numerical solutions for such laminates are discussed. The approximate so-
lutions presented allow prediction of the deflections caused by the combined
effects of in-plane loads, transverse loads, and changes in temperature or
moisture content. Buckling caused by in-plane loads and/or temperature
changes is also discussed.
The equations of equilibrium for a thin symmetric composite laminate were
derived in Sec. 2 of Chap. 9 based on a summation of forces and moments.
Boundary conditions consistent with thin-plate theory were then discussed in
Sec. 3 of Chap. 9. It turns out that exact solutions to these equations and
boundary conditions can only be obtained if A16 = A26 = D16 = D26 =
N Txy = N M
xy = 0; that is, exact solutions are only available for specially
orthotropic laminates. A few exact solutions for simply supported and
symmetric specially orthotropic laminates were presented in Chap. 10.
Unfortunately, many stacking sequences widely used in practice are not
specially orthotropic. For example, symmetric quasi-isotropic laminates are
not specially orthotropic because D16,D26 p 0 for this stacking sequence.
571

Hence, the solutions presented in Chap. 10 are not rigorously valid for many
laminates encountered in practice. Fortunately, approximate numerical
solutions are available that are suitable for use with any laminate including
symmetric quasi-isotropic laminates. A brief introduction to these approx-
imate solutions techniques will be presented in this chapter. Readers inter-
ested in a more detailed discussion are referred to Refs. (1,2).
It is appropriate to clarify the distinction between ‘‘exact’’ and ‘‘approx-
imate’’ solutions in the present context. In particular, confusion may arise in
that exact solutions often involve a series with an infinite number of terms (as
in the Navier solutions presented in Sec. 6 of Chap. 10, for example). A
numerical evaluation of these solutions will contain some level of error
because a finite number of terms must obviously be used during any practical
application. However, the source of error in these cases arises from an in-
ability to describe the applied loading in an exact analytical sense, rather than
the solution itself. Furthermore, in a practical application, the number of
terms used can be increased to obtain numerical results that satisfy all bound-
ary conditions and the equations of equilibrium to any desired number of
significant figures.
The distinction between an exact and approximate solution does not
refer to the need to use a finite number of terms during a numerical
evaluation. Rather, the distinction refers to the rigor with which the boundary
conditions are satisfied. Recall from Sec. 3 of Chap. 9 that it is necessary to
specify four boundary conditions along each edge of a thin plate. Further-
more, recall that any boundary condition can be classified as either a
geometric boundary condition or a static (also called a ‘‘natural’’) boundary
condition. Geometric boundary conditions are those that dictate some
feature of midplane displacements along a plate edge, whereas static bound-
ary conditions are those that dictate some external load applied along a plate
edge. Many common edge conditions consist of a combination of geometric
and static boundary conditions. For example, a simply supported edge is
one in which out-of-plane deflection is zero along the edge (a geometric
condition) and the bending moment is zero along the edge (a static con-
dition); the remaining two conditions may be either geometric or static.
Consequently there are four ‘‘types’’ of simply supported edges (see Sec. 3.3.2
of Chap. 9).
Now, during derivation of an exact solution all four boundary con-
ditions along each edge of the plate are accounted for during derivation of the
solution and are satisfied exactly. For example, in the solutions for simply
supported specially orthotropic plates discussed in Chap. 10 all geometric and
static boundary conditions are satisfied exactly by the solutions presented. In
contrast, in an approximate solution, only the geometric boundary conditions
are specified and enforced directly. Static boundary conditions are not en-

forced directly. Therefore, solutions obtained using approximate analysis
techniques may, or may not, satisfy the static boundary conditions.
The approximate solutions to be discussed in this chapter are based on
the principle of minimum potential energy. Solution techniques that follow
from this principle are known as energy methods. Consider a solid body
subjected to some external loading. Briefly stated, the principle of minimum
potential energy states that for all possible displacement fields that satisfy
given boundary conditions, the displacement field that actually occurs is one
in which the potential energy of the solid body assumes a minimum value.
This statement is exact. In fact, it is possible to rederive the equations of
equilibrium and associated boundary conditions (and, hence, all of the results
presented in Chaps. 9 and 10) based on this principle.
Energy methods will not be used to rederive the equations of equilib-
rium and associated boundary conditions in this textbook; the reader
interested in this derivation is referred to the text by Whitney (1). Rather,
the reason energy methods are of interest here is that they form the basis for
approximate numerical solution techniques that are based on the principle of
minimum potential energy. The two most common approximate numerical
techniques are the Ritz method and the Galerkin method. The Ritz method will
be developed and applied to several problems in this chapter. The Galerkin
method will not be discussed, and the reader interested in this method is
referred to Refs. (1,2).
The essential elements of the Ritz method are as follows. The potential
energy of a solid body is denoted P. The mathematical form of P depends
on details of the problem under consideration and will be discussed in later
sections. At this point, suffice it to say that the potential energy P of a solid
body can be calculated if (a) the elastic properties of the body are known,
(b) the external forces applied to the body are known, and (c) the resulting
displacement fields induced in the body are known. During a typical
structural analysis, the elastic properties and the external forces are known,
but the displacement fields are not (in fact, the objective during a structural
analysis is often to determine the displacement fields induced by some
specified loading). During application of the Ritz method it is assumed, in
effect, that the functional form of the displacement fields are known. For
general thin-plate problems the midplane displacement fields are assumed to
be of the form:
M1 X
X N1
uo ðx; yÞ ¼ amn Umn ðx; yÞ ð1aÞ
m ¼ 1n ¼ 1
M2 X
X N2
vo ðx; yÞ ¼ bmn Vmn ðx; yÞ ð1bÞ
m ¼ 1n ¼ 1

X
M3 X
N3
wðx; yÞ ¼ cmn Wmn ðx; yÞ ð1cÞ
m ¼ 1n ¼ 1
where amn, bmn, and cmn are unknown constants; and Umn, Vmn, and Wmn are
known functions that vary over x and y. In general, the number of terms used
to describe each displacement field may differ (e.g., M1 does not necessarily
equal M2 or M3). Also, the number of terms used to describe the variation in x
and y may differ (e.g., M1 does not necessarily equal N1).
Having assumed the functional form for the displacement fields as
represented by Eqs. (1a)–(1c), then the potential energy P of an elastic plate
subjected to specified external loading can be calculated. Note that the values
of constants amn, bmn, and cmn in Eqs. (1a)–(1c) effectively define the
magnitudes of the displacement fields. Hence, as constants amn, bmn and/or
cmn are increased or decreased (and assuming that elastic properties and
external forces remain constant), the potential energy of the body is increased
or decreased accordingly. The principle of minimum potential energy states
that the displacement field actually adopted by a solid body is one in which the
potential energy is a minimum. This condition is therefore defined by the
following criteria:

BP m ¼ 1; 2; . . . ; M1
¼0 ð2aÞ
Bamn n ¼ 1; 2; . . . ; N1

BP m ¼ 1; 2; . . . ; M2
¼0 ð2bÞ
Bbmn n ¼ 1; 2; . . . ; N2

BP m ¼ 1; 2; . . . ; M3
¼0 ð2cÞ
Bcmn n ¼ 1; 2; . . . ; N3
Equations (2a)–(2c) lead to (M1 N1)+(M2 N2)+(M3 N3) equations
that must be satisfied simultaneously. Hence, by solving these equations for
constants amn, bmn, and cmn, the magnitudes of the displacement fields given by
Eqs. (1a)–(1c) that correspond to the minimum potential energy are known
and the problem is solved. The validity of the solution obtained hinges on
whether Eqs. (1a)–(1c) adequately represent the displacement fields actually
induced in the structure.
It is seen therefore that solutions obtained using the Ritz method are
based on the functions Umn, Vmn, and Wmn that appear in Eqs. (1a)–(1c).
These functions are more-or-less arbitrarily selected, but must posses two
important characteristics: (a) they must be continuous and differentiable to at
least the second order, and (b) they must satisfy the geometric boundary
conditions. This latter characteristic is the source of the approximate nature
of the Ritz analysis. That is, the mathematical forms of functions Umn(x,y),
Vmn(x,y), and Wmn(x,y) are selected to satisfy the prevailing geometric
boundary conditions, but the static boundary conditions are not considered

during this selection. Hence, these functions may, or may not, satisfy the pre-
vailing static boundary conditions. If the functions Umn(x,y), Vmn(x,y), and
Wmn(x,y) do not satisfy both geometric and static boundary conditions, then
the solution obtained using the Ritz approach will be approximate. On the
other hand, if the forms selected for Umn, Vmn, and Wmn do happen to satisfy
both geometric and static boundary conditions, then the solution obtained
using the Ritz method is exact.
The challenge of identifying functional forms for Umn, Vmn, and Wmn
that satisfy both geometric and static boundary conditions will be illustrated
for a simply supported plate. Functions Umn, Vmn, and Wmn are usually
selected to be separable functions of x and y. That is, function Wmn, for
example, is usually selected to be of the form:
Wmn ¼ Xm ðxÞYn ðyÞ
For simply supported plates, it is common to assume that Xm and Yn are
sinusoidal functions of x and y:

mpx npy
Xm ðxÞ ¼ sin Yn ðyÞ ¼ sin ð3Þ
a b
where a and b are the length and width of a rectangular composite plate,
respectively. Thus, based on this selection, Eq. (1c) becomes:
XM3 X N3
mpx npy
m ¼ 1n ¼ 1
a b
These functional forms for Xm and Yn are appropriate for simply supported
plates because they satisfy the geometric boundary conditions, regardless of
the magnitudes of cmn. That is, Eq. (4) gives w(x,y) = 0 for x = 0,a and y = 0,
b for all m,n. They do not necessarily satisfy the static boundary conditions for
a simply supported plate, however. Recall from Sec. 3 of Chap. 9 that the
bending moment must vanish along a simply supported edge, resulting in the
following static boundary condition along the two edges x = 0, a (a
comparable condition must be satisfied along edges y=0,b):
B2 w B2 w B2 w
D11 þ D12 2 2D16 ¼ M*xx ¼ 0
Bx 2 By BxBy
Upon substituting the assumed displacement field [Eq. (4)] into this static
boundary condition, the first two terms lead to:
" #
B2 w XM3 X N3 mp 2 mpx npy
D11 2 ¼ D11 cmn sin sin ¼ 0;
Bx m ¼ 1n ¼ 1
a a b
for x ¼ 0; a

" 2 #
B2 w X
M3 X
N3
np mpx npy
D12 2 ¼ D12 cmn sin sin ¼ 0;
By m ¼ 1n ¼ 1
b a b
for x ¼ 0; a
However, the third term leads to:
" #
B2 w X
M3 XN3
mp np mpx npy
2D16 ¼ 2D16 cmn cos cos
BxBy m ¼ 1n ¼ 1
a b a b
Evaluating this third term along x=0, a:

B2 w 2p2 D16 X
M3 XN3
npy
2D16 ¼ ðmnÞ cmn cos ; for x ¼ 0; a
BxBy ab m ¼ 1 n ¼ 1 b
Hence, the static boundary condition associated with simply supported

edges x = 0,a is satisfied only if:
M3 X
X N3 npy
ðmnÞcmn cos ¼ 0; for 0 V y V b ð5Þ
m ¼ 1n ¼ 1
b
If only a single term is used to describe the displacement field (i.e., if M3=
N3=1), then the static boundary condition is clearly not satisfied even ap-
proximately because:
py
c11 cos p 0; for all 0 V y V b
b
However, if the number of terms used is increased (i.e., as M3 and/or N3 are
increased), then Eq. (5) may be satisfied more and more exactly, through
proper selection of the values of constants cmn. For example, if four terms are
used (i.e., if M3=N3=2), then the static boundary condition along edges
x=0,a is satisfied if:
py py
2py 2py
c11 cos þ 2c12 cos þ 2c21 cos þ 4c22 cos ¼ 0;
b b b b
for all 0 V y V b
It is now possible to satisfy the static boundary condition along edges x = 0,
a exactly by setting c21 = c11/2 and c22 = c12/2. Of course, selecting
constants that satisfy these requirements may not lead to the displacement
field that represents the state of minimum potential energy because constants
c11, c12, c21, and c22 must also satisfy the static boundary conditions along the
edges y = 0, b. Still, it is apparent that by increasing the number of terms
used to describe the displacement field, it is possible to satisfy the static
boundary conditions along all four edges more and more exactly.

In general then, the validity of a solution based on the Ritz approach is
increased as the number of terms is increased. Assuming functions Xm and Yn
are selected to satisfy the geometric boundary conditions, then a Ritz analysis
will converge toward the exact solution as M3 and/or N3 are increased. It is
also appropriate to note that the difficulty in obtaining an exact solution is due
to the D16 term. If D16 = 0 (i.e., if the laminate were specially orthotropic),
then the geometric and static boundary conditions associated with the simply
supported edges x = 0,a would be satisfied exactly by the assumed sinusoidal
form for Xm and Yn.
A relatively general description of how the Ritz approach is applied has
been presented in the preceding paragraphs. We will now specialize this
approach for application to the specific problems discussed in this chapter.
First, as in previous chapters, we limit our discussion to rectangular sym-
metric laminates, simply supported along all four edges. Second, for all prob-
lems considered herein, the in-plane displacement fields uo(x,y) and vo(x,y)
that exist prior to application of q(x,y) can be deduced based on a CLT
analysis (as discussed in the next section) and are known a priori. Hence, for
present purposes, there is no need to express uo(x,y) or vo(x,y) in terms of the
double series as listed as Eqs. (1a,b) nor to determine the magnitudes of in-
plane displacement fields using Eqs. (2a,b). For the problems considered
herein, only w(x,y) is unknown. The out-of-plane displacement field will be
expressed using the double series listed as Eq. (1c), and the magnitude of out-
of-plane displacements will be determined through application of Eq. (2c).
Functions Xm and Yn will be assumed to be sinusoidal functions as listed as
Eq. (3); thus, the assumed form for w(x,y) is given by Eq. (4). To simplify
nomenclature, let M3!M and N3!N; thus, w(x,y) will be written:
XM XN
mpx npy
m ¼ 1n ¼ 1
a b
Although not discussed in this text, solutions are also available based on
alternate (nonsinusoidal) function forms for Xm and Yn. Alternate forms
include polynomials in x and y:
Xm ðxÞ ¼ ðx2 axÞ2 xm1

Yn ðyÞ ¼ ðy2 ayÞ2 yn1
forms involving other trigonometric functions:

2mpx
Xm ðxÞ ¼ 1 cos
a

2npy
Yn ðyÞ ¼ 1 cos
b

or forms involving so-called beam functions:

km x km x km x km x
Xm ðxÞ ¼ cm cos cm cosh þ sin sinh
a a a a

kn x kn x kn x kn x
Yn ðxÞ ¼ cn cos cn cosh þ sin sinh
b b b b
These alternate forms may be used to model other boundary conditions such
as a clamped edge. The reader interested application of these alternate forms
is referred to Refs. [1,2,4].
2 IN-PLANE DISPLACEMENT FIELDS

In this section, we will derive expressions that give the midplane displacement
fields induced in a symmetric laminate for specific loading conditions. The
derivation is based on the following sequence of events. First, assume the
laminate is cured at an elevated temperature and that the laminate is stress-
and strain-free at the cure temperature.* Following cure, the laminate is
cooled to room temperature. The laminate is then trimmed to the desired final
rectangular dimensions a b. Finally, the laminate is subjected to constant
and uniformly distributed stress resultants (Nxx, Nyy, and/or Nxy) along all
four edges and/or a further change in temperature. Although we will not
consider the possibility of a uniform change in moisture content directly, from
previous discussion, it should be clear that changes in moisture content could
be accounted for (in a mathematical sense) using the same techniques used to
model uniform changes in temperature.
The sequence of events described above are precisely those assumed
during the development of CLT in Chap. 6. Because the externally applied
stress resultants are uniformly distributed along each edge, the stress result-
ants induced at all interior points are equal to the edge loads:
* ð0; yÞ ¼ Nxx
Nxx * ða; yÞ ¼ Nxx
N*yy ðx; 0Þ ¼ Nyy

* ðx; bÞ ¼ Nyy
N*xy ð0; yÞ ¼ Nxy

* ða; yÞ ¼ Nyx
* ðx; 0Þ ¼ Nyx
* ðx; bÞ ¼ Nxy
* As mentioned in Sec. 6.2 of Chap. 6, in practice, the stress- and strain-free temperature is often
20–50jC below the final cure temperature. This complication has been ignored throughout this
text, and it is assumed that the laminate is stress- and strain-free at the cure temperature.

For a symmetric laminate subjected to uniform in-plane stress resultants and
a change in temperature, Eq. (45) of Chap. 6 becomes:
8 0 9 2 38 9
>
> exx >> a11 a12 a16 0 0 0 > > Nxx þ NTxx >>
>
> > 6
> 7>> >
>
>
> >
> 6 7 >
> >
>
>
> 0 >> 6 7 >
> >
>
>
> e >
> 6 a 12 a 22 a 26 0 0 0 7>> N yy þ N T
>
>
>
>
yy
>
> 6 7 >
>
yy
>
>
>
> 0 > >
> 6 7>>
> >
>
>
> > >
6
< cxy = 6 a16 a26 a66 0 0 0 7< Nxy þ Nxy >
7 > T
=
¼6 7
>
> > 6 7> >
>
> jxx >>
>
6 0
6 0 0 d11 d12 d16 7 7>
>
> 0 >
>
>
>
> >
> 6 7 >
> >
>
>
> > 6
> 7 >
> >
>
>
> >
> 6 7 >
> >
>
>
> >
jyy > 6 0 0 0 d12 d22 d26 7> > 0 >
>
>
> >
> 4 5>> >
>
>
> >
> >
> >
>
: ; : ;
jxy 0 0 0 d16 d26 d66 0
As in earlier chapters, we assume infinitesimal strain levels, and therefore
midplane strains are related to midplane displacements according to Eq. (10)
of Chap. 6:
Buo Bvo Buo Bvo
e0xx ¼ e0yy ¼ c0xy ¼ þ
Bx By By Bx
Consequently, midplane displacement fields are related to the stress and ther-
mal resultants as follows:
Buo
¼ a11 ðNxx þ NTxx Þ þ a12 ðNyy þ NTyy Þ þ a16 ðNxy þ NTxy Þ ð7aÞ
Bx
Bvo
¼ a12 ðNxx þ NTxx Þ þ a22 ðNyy þ NTyy Þ þ a26 ðNxy þ NTxy Þ ð7bÞ
By
Buo Bvo
þ ¼ a16 ðNxx þ NTxx Þ þ a26 ðNyy þ NTyy Þ þ a66 ðNxy þ NTxy Þ ð7cÞ
By Bx
Integrating Eq. (7a) with respect to x, we find:
h i
uo ðx; yÞ ¼ a11 Nxx þ NTxx þ a12 Nyy þ NTyy þ a16 Nxt þ NTxy x
þ f1 ðyÞ þ k1
where f1( y) is an unknown function of y (only); and k1 is an unknown constant
of integration. Similarly, integrating Eq. (7b) with respect to y, we find:
h i
vo ðx; yÞ ¼ a12 Nxx þ NTxx þ a22 Nyy þ NTyy þ a26 Nxy þ NTxy y
þ f2 ðxÞ þ k2
where f2(x) is an unknown function of x (only); and k2 is a second unknown
constant. Without a loss in generality, we assume that midplane displacements

are zero at the origin (i.e., at x = y = 0) and, consequently, k1 = k2 = 0.
Substituting these expressions for uo(x,y) and vo(x,y) into Eq. (7c), we find:
Buo Bvo Bf1 Bf2
þ ¼ þ ¼ a16 ðNxx þ NTxx Þ þ a26 ðNyy þ NTyy Þ
By Bx By Bx
þ a66 ðNxy þ NTxy Þ
Because all terms on the right side of the equality are known constants, it
follows that f1 and f2 must be at most linear function of y and x, respectively:
f1 ðyÞ ¼ k3 y
f2 ðxÞ ¼ k4 x
Hence, we can write:
k3 þ k4 ¼ a16 ðNxx þ NTxx Þ þ a26 ðNyy þ NTyy Þ þ a66 ðNxy þ NTxy Þ
Because all quantities that appear in this relation are constants, k3 and k4 can
take on any value as long as they sum to the expression on the right side of the
equality. To determine particular values that satisfy this expression, we now
require that the infinitesimal rotation vector in the x–y plane xxy (which
represents rigid body motion of the plate) is zero. The infinitesimal rotation
vector is given by (3):

1 Buo Bvo
xxy ¼
2 By Bx
Requiring that xxy=0 leads to:
1h i
k3 ¼ k4 ¼ a16 Nxx þ NTxx þ a26 Nyy þ NTyy þ a66 Nxy þ NTxy
2
Combining the preceding results, we conclude that the in-plane midplane
displacements induced in a symmetric composite panel by the combination of
uniform in-plane stress resultants and a change in temperature are given by:
h i
uo ðx; yÞ ¼ a11 Nxx þ NTxx þ a12 Nyy þNTyy þ a16 Nxy þ NTxy x
1h i
þ a16 Nxx þ NTxx þ a26 Nyy þ NTyy þ a66 Nxy þ NTxy y
2
1h i
vo ðx; yÞ ¼ a16 Nxx þ NTxx þ a26 Nyy þ NTyy þ a66 Nxy þ NTxy x
2
h i
þ a12 Nxx þ NTxx þ a22 Nyy þ NTyy þ a26 Nxy þ NTxy y

These expressions can be simplified through the use of the effective thermal
expansion coefficients of the laminate, defined by Eqs. (73a) of Chap. 6 and
1 h i
DT
1 h i
ayy ¼ a12 NTxx þ a22 NTyy þ a26 NTxy ð6:73aÞ
DT
1 h i
axy ¼ a16 NTxx þ a26 NTyy þ a66 NTxy
DT
Hence, we see that the midplane displacement fields can be written:
h i
uo ðx; yÞ ¼ a11 Nxx þ a12 Nyy þ a16 Nxy þ DT axx x
1h i
þ a16 Nxx þ a26 Nyy þ a66 Nxy þ DT axy y
2
ð7Þ
1h i
vo ðx; yÞ ¼ a16 Nxx þ a26 Nyy þ a66 Nxy þ DT axy x
2
h i
þ a12 Nxx þ a22 Nyy þ a26 Nxy þ DT ayy y
Note that Eqs. (7a)–(7c) are valid only if both stress resultants and temper-
ature changes are uniform.
In the following sections, we will use these in-plane displacement fields
to obtain solutions based on the Ritz method for simply supported composite
plates. Recall from Sec. 3 of Chap. 9 that four distinct combinations of
geometric and static boundary conditions, numbered S1 through S4, can be
defined as a ‘‘simple support.’’ The distinction between the different types of
simple supports has to do with the boundary condition assumed for the in-
plane displacement fields. For example, to define a simple support of type S1,
one specifies known values of in-plane displacements, whereas to define a
simple support of type S4, one specifies known values of in-plane stress
resultants. For the problems considered here, we are able to calculate the
midplane displacement field induced by a specified combination of edge loads
(or vice versa). This is possible because we have assumed all stress resultants
applied at the edge of the plate are constant and uniform. Of course, if the
stress resultants were not uniform but rather varied along the plate edges, then
it would be more difficult (and, in most cases, impossible) to determine
associated in-plane displacement fields.
Because we have limited discussion to cases in which stress resultants
applied to the edges are constant and uniform, the midplane displacement

fields given by Eqs. (7a)–(7c) can be used to obtain solutions for any of the
four types of simple-support boundary conditions, S1 through S4. Only one
type of simple support will be considered in this chapter because this
discussion is intended to be a brief introduction to the Ritz method.
Specifically, type S4 simple supports will be assumed throughout the remain-
der of this chapter. Thus, solutions will be obtained using the Ritz method for
problems in which uniform in-plane stress resultants Nxx, Nyy, and/or Nxy are
applied to the edge of the plate. Although not discussed herein, similar ana-
lyses can be performed assuming type S1, S2 or S3 simple supports, or com-
binations thereof.
3 POTENTIAL ENERGY IN A THIN COMPOSITE PLATE

In this section, we will develop the equations necessary to calculate the po-
tential energy of a thin elastic plate subjected to a combination of loads and
uniform temperature changes. The possibility of a uniform change in mois-
ture content will not be considered although from previous discussion, it
should be clear that changes in moisture content can be accounted for (in
a mathematical sense) using the same techniques used to model uniform
changes in temperature. Type S4 boundary conditions are assumed for all
four edges of the plate.
Two energy terms will be encountered in the following discussion. First,
we will consider the work done when a transverse load q(x,y) is applied to a
thin plate. This energy term is denoted W. Recall that the fundamental
definition of ‘‘work’’ is force multiplied by the distance through which it
travels. Also, recall that the transverse load applied to a plate q(x,y) has been
defined using units of force/area. The product ( q)(dx)(dy) represents the
transverse force provided by q(x,y) acting over an infinitesimal element of area
(dx)(dy). The distance through which this force travels equals the out-of-plane
deflection of the plate at that point, w(x,y). Therefore, the total work done by
the transverse load acting over the entire surface of the plate is given by:
ZZ
W¼ ½qðx; yÞ½wðx; yÞdxdy ð8Þ
Secondly, we will consider the strain energy within the plate in the deformed
condition. This energy term is denoted U. A general expression giving the
strain energy within a linear elastic solid body subjected to an arbitrary state
of stress is:
ZZZ
1
U¼ ðrxx exx þ ryy eyy þ rzz ezz þ syz cyz þ sxz cxz þ sxy cxy Þdxdydz
2

Note that evaluation of strain energy involves integration over the entire vol-
ume of the body. In our case, we wish to calculate the strain energy within a
thin symmetric composite laminate subjected to a state of plane stress. We
will therefore specialize the above general expression of strain energy for
present purposes. First, because plane stress is assumed (rzz = syz = sxz = 0),
we can immediately discard terms involving these stress components:
ZZZ
1
U¼ ðrxx exx þ ryy eyy þ sxy cxy Þdxdydz ð9Þ
2
Next, the stresses in ply k of the laminate are given by Eqs. (30) of Chap. 5,
which become (for DM = 0):
8 9 2 3 8 9
< rxx = Q11 Q12 Q16 < exx DTaxx =
ryy ¼ 4 Q12 Q22 Q26 5 eyy DTayy
: ; : ;
rxy k Q16 Q26 Q66 k cxy DTaxy k
Substituting these expressions for ply stresses into Eq. (9) and rearranging,
we find:
ZZ Z h
1 ðkÞ ðkÞ ðkÞ ðkÞ ðkÞ ðkÞ
U¼ Q11 e2xx þ 2Q12 exx eyy þ 2Q16 exx cxy þ 2Q26 eyy cxy þ Q22 e2yy þ Q66 c2xy
2

ðkÞ ðkÞ ðkÞ ðkÞ ðkÞ
DT aðkÞ Q
xx 11 þ a Q
yy 12 þ a xy 16 exx
Q
ð10Þ
DT aðkÞ Q
xx 12 þ a Q
yy 22 þ a xy 26 eyy
Q
i
DT aðkÞ Q
xx 16 þ a Q
yy 26 þ a xy 66 cxy dxdydz
Q
We now invoke the Kirchhoff hypothesis, which allows us to relate ply strains
at any through-thickness position z to midplane strains and curvatures, in
accordance with Eqs. (12)of Chap. 6, repeated here for convenience:
exx ¼ e0xx þ zjxx

eyy ¼ e0yy þ zjyy
cxy ¼ c0xy þ zjxy

Substitution of Eqs. (12) of Chap. 6 into Eq. (10) results in:
ZZZ n h i
1 2
U¼ Q11 e0xx þ 2ze0xx jxx þ z2 j2xx
2

h i
þ2Q12 e0xx e0yy þ ze0xx jyy þ ze0yy jxx þ z2 jxx jyy
h i
þ2Q16 e0xx c0xy þ ze0xx jxy þ zc0xy jxx þ z2 jxx jxy
h i
þ2Q26 e0yy c0xy þ ze0yy jxy þ zc0xy jyy þ z2 jyy jxy
h 2 i
þQ22 e0yy þ2ze0yy jyy þ z2 j2yy
h 2 i ð11Þ
þQ66 c0xy þ2zc0xy jxy þ z2 j2xy
h i
DT axx Q11 þ ayy Q12 þ axy Q16 e0xx þ zjxx
h i
DT axx Q12 þ ayy Q22 þ axy Q26 e0yy þ zjyy
h i o
DT axx Q16 þ ayy Q26 þ axy Q66 c0xy þ zjxy dxdydz
Next, integrate Eq. (11) over the thickness of the laminate, i.e., over the range
t/2 V z V t/2. During this process, a number of integrals will be encoun-
tered that were previously evaluated in Chap. 6. A few specific examples are:
Z t=2
Q11 dz; which after integration becomes A11
t=2
Z t=2
Q11 zdz; which after integration becomes B11
t=2
Z t=2
Q11 z2 dz; which after integration becomes D11
t=2
Z t=2
DT axx Q11 þ ayy Q12 þ axy Q16 dz;

t=2
which after integration becomes NTxx

Z t=2
DT axx Q11 þ ayy Q12 þ axy Q16 zdz;

t=2
which after integration becomes MTxx
Hence, after integration Eq. (11) can be written:

ZZ 2
1 2
U¼ A11 e0xx þ 2A12 e0xx e0yy þ A22 e0yy þ 2 A16 e0xx þ A26 e0yy c0xy
2

þ A66 c2xy þ 2B11 e0xx jxx þ 2B12 e0xx jyy þ e0yy jxx

þ 2B16 e0xx jxy þ c0xy jxx þ 2B26 e0yy jxy þ c0xy jyy
þ 2B22 e0yy jyy þ 2B66 c0xy jxy þ D11 j2xx þ 2D12 jxx jyy
ð12Þ
þ 2 D16 jxx þ D26 jyy jxy þ D22 j2yy þ D66 j2xy NTxx e0xx

NTyy e0yy NTxy c0xy MTxx jxx MTyy jyy MTxy jxy dxdy
T
Because we have limited discussion to symmetric laminates, Bij = Mxx =
T T
Myy = Mxy = 0 in all cases considered. Equations (12) therefore simplifies
to:
ZZ 2
1 2
U¼ A11 e0xx þ 2A12 e0xx e0yy þ A22 e0yy þ2 A16 e0xx þ A26 e0yy c0xy
2

þ A66 c2xy þ D11 j2xx þ 2D12 jxx jyy þ 2 D16 jxx þ D26 jyy jxy ð13Þ

þ D22 j2yy þ D66 j2xy NTxx e0xx NTyy e0yy NTxy c0xy dxdy
We have now developed expressions for the two energy terms necessary for
our purposes: the work W done by a transverse load applied to a laminate
[Eq. (8)] and the strain energy U within a deformed laminate [Eq. (13)].
We wish to form an appropriate combination of these terms so as to
represent the total potential energy of a symmetric composite laminate. An
itemized conceptual description of how U and W are combined is
presented below. Mathematical implementation of these concepts for the
particular class of problems considered in this text is then discussed in
separate subsections. The reader is urged to carefully consider the follow
conceptual description before considering the mathematical formulation
that follows.
Step 1. We assume that in-plane stress resultants Nij are applied to the
laminate first, before the application of any other load(s) that cause bending.
Our first step is therefore to calculate the strain energy within a laminate
subjected to Nij only even if other loads are involved in the problem under
consideration. We will label this component of strain energy UI. Calculation
of UI is straightforward because we have limited our discussion to symmetric
laminates. That is, for a symmetric laminate, stress resultants Nij are solely
responsible for the development of midplane strains and do not cause
curvatures to develop. If we had included nonsymmetric laminates in our
analysis, then Bij p 0. If this were the case, then Nij would also contribute
to curvatures; furthermore, both Mij and q(x,y) would contribute to

midplane strains. These coupling effects would greatly complicate our
analysis. Because symmetry has been assumed, the laminate remains flat
following application of Nij. Therefore, the only loads that contribute to
midplane strains are stress resultants Nij. The strain energy associated with Nij
only (i.e., the strain energy component UI) can therefore be obtained from Eq.
(13) simply by setting jxx=jyy= jxy=0. Ultimately our expression for UI
will involve in-plane displacements u(x,y) and v(x,y), but will not involve
out-of-plane displacements w(x,y).
Step 2. Next, we calculate strain energy associated with midplane curvatures

only. That is, we ignore any preexisting midplane strains and base our
calculation of strain energy based solely on curvatures jxx, jyy, and jxy. We
will label this component of the strain energy UII. Because the laminate is
symmetric, midplane curvatures are caused by the combined effects of the
transverse load q(x,y) and/or bending moment resultants Mij, but are
independent of stress resultants Nij. Strain energy component UII can be
obtained from Eq. (13) simply by setting e0xx =e0yy =e0xy =0. Ultimately, our
expression for UII will involve out-of-plane displacements w(x,y), but will not
involve in-plane displacements u(x,y) or v(x,y).
Step 3. As described above, in step 1, we calculate the strain energy UI

associated with stress resultants Nij prior to the application of any load(s)
that cause the laminate to bend. Of course, when transverse loads and/or
bending moments are subsequently applied, the laminate will bend. Now,
once bending has occurred, then calculation of strain energy as performed
in step 1 is incomplete. That is, as the laminate begins to bend, the in-plane
strains that exist prior to bending will change, resulting in a change in the
strain energy associated with stress resultants Therefore, in step 3, we
calculate the change in strain energy caused by bending and associated
with in-plane stress resultants. We will label this component of the strain
energy UIII. We assume that the stress resultants Nij that exist prior to
bending remain constant during bending. That is, we assume that only in-
plane strains change as bending occurs. The change in in-plane strains will
be related to out-of-plane displacements w(x,y). Hence, our expression for
UIII will involve w(x,y), but will not involve in-plane displacements u(x,y)
or v(x,y).
Step 4. In step 4, we calculate the work W done by the transverse load

q(x,y), in accordance with Eq. (6). The mathematical form of our expression
for W will obviously depend on the nature of the transverse load. Thus, for
example, the mathematical form of W for a uniform transverse load,
q(x,y)=qo, will differ from the mathematical form of W if the transverse
load varies sinusoidally: q(x,y)= qosin(px/a)sin(py/b).

Step 5. Finally, in step 5, we form the desired expression for the total
potential energy P in the composite plate. The total potential energy is given
by: P = UI+UII+UIIIW.
3.1 Evaluation of Strain Energy Component UI

Strain energy component UI is obtained from Eq. (11) by setting jxx=jyy=
jxy=0:
ZZ 2
1 2
UI ¼ A11 e0xx þ 2A12 e0xx e0yy þ A22 e0yy þ 2 A16 e0xx þ A22 e0yy c0xy
2

þ A66 cxy Nxx exx Nyy eyy Nxy cxy dxdy
2 T 0 T 0 T 0
As in earlier chapters, we assume infinitesimal strain levels and therefore

midplane strains are related to midplane displacements according to Eq. (10)
of Chap. 6:
Buo Bvo Buo Bvo
e0xx ¼ e0yy ¼ c0xy ¼ þ
Bx By By Bx
With this substitution, we have:
ZZ (
1 Buo 2 Buo Bvo Bvo 2
UI ¼ A11 þ 2A12 þ A22
2 Bx Bx By By

Buo Bvo Buo Bvo Buo Bvo 2
þ 2 A16 þ A26 þ þ A66 þ ð14Þ
Bx By By Bx By Bx
)
Buo Bvo Buo Bvo
Nxx
T
Nyy
T
Nxy
T
þ dxdy
Bx By By Bx
Equation (14) gives the strain energy component UI for any in-plane
displacement fields uo(x,y) and vo(x,y). We will now integrate this expression
using the displacement fields induced by uniform stress resultants and a
uniform change in temperature, as given by Eqs. (7). To avoid a very lengthy
expression, we make the following change in notation:
h i
C1 ¼ a11 Nxx þ a12 Nyy þ a16 Nxy þ DT axx
h i
C2 ¼ a12 Nxx þ a22 Nyy þ a26 Nxy þ DT ayy ð15Þ
1h i
C3 ¼ a16 Nxx þ a26 Nyy þ a66 Nxy þ DT axy
2

With this change in notation, Eqs. (7) can be written:
uo ðx; yÞ ¼ C1 x þ C3 y
vo ðx; yÞ ¼ C3 x þ C2 y
Substituting these expressions into Eq. (14), we have:

Z Z
1 b an
UI ¼ A11 ðC1 Þ2 þ 2A12 ðC1 ÞðC2 Þ þ A22 ðC2 Þ2
2 0 0
þ 2ðA16 C1 þ A26 C2 Þð2C3 Þ þ A66 ð2C3 Þ2 NTxx ðC1 Þ
o
NTyy ðC2 Þ NTxy ð2C3 Þ dxdy
Evaluating this definite integral, we find:

h
UI ¼ A11 ðC1 Þ2 þ 2A12 ðC1 ÞðC2 Þ þ A22 ðC2 Þ2

þ 4 A16 C1 C3 þ A26 C2 C3 þA66 C23
i
NTxx C1 NTyy C2 2NTxy C3 ðabÞ
Substituting Eqs. (73b) of Chap. 6, which give the thermal stress resultants
in terms of elements of the [A] matrix, and Eq. (5), we find that UI can be
written:

UI ¼ C1 Nxx þ C2 Nyy þ 2C3 Nxy ðabÞ ð16Þ
3.2 Evaluation of Strain Energy Component UII

As previously discussed, for symmetric laminates, strain energy component
UII can be obtained from Eq. (13) by setting e0xx ¼ e0yy ¼ e0xy =0:
ZZ n
1
UII ¼ D11 j2xx þ 2D12 jxx jyy þ 2ðD16 jxx þ D26 jyy Þjxy þ D22 j2yy
2
o
þ D66 j2xy dxdy
Midplane curvatures are related to out-of-plane displacements according to

Eq. (10) of Chap. 6:
B2 w B2 w B2 w
Bx2 By2 BxBy

With this substitution, we have:
Z Z ( 2 2 2 2
1 B w B w B w
UII ¼ D11 þ2D 12
2 Bx2 Bx2 By2
2 2
2
B w Bw B w
þ 4 D16 þ D26
Bx2 By 2 BxBy
ð17Þ
2 2 2 2 )
Bw B w
þ D22 þ4D66 dxdy
By2 BxBy
We will now integrate Eq. (17) for the class of problems considered in this
text. In all problems, we consider simply supported plates and assume out-
of-plane displacements are given by Eq. (6), repeated here for convenience:
XM XN
mpx npy
wðx; yÞ ¼ cmn sin sin
m ¼ 1n ¼ 1
a b
Because Eq. (6) will be used in all problems considered, we will integrate
Eq. (17) based on this displacement field. The following derivatives appear
in Eq. (17):
X X
B2 w M M
mp 2 mpx npy
¼ cmn sin sin
Bx2 m ¼ 1 n ¼ 1
a a b
XM XN 2
B2 w np mpx npy
¼ c mn sin sin
By2 m ¼ 1 n ¼ 1
b a b
XM XN
B2 w mp np mpx npy
¼ cmn cos cos
BxBy m ¼ 1 n ¼ 1 a b a b
Consider the first term under the integral sign in Eq. (17). Upon substituting
the expression for (B2w/Bx2) listed above, this term becomes:
Z Z 2 2
1 b a B w
D11 dxdy
2 0 0 Bx2
Z Z " 2 #
1 b a XM XN
mp mpx npy
¼ D11 cmn sin sin
2 0 0 m ¼ 1 n ¼ 1
a a b
" 2 #
XM XN
ip ipx jpy
cij sin sin dxdy
i ¼ 1 j ¼ 1
a a b

The evaluation of this integral is greatly simplified by noting the following
identities:
Z a
mpx ipx
sin sin dx ¼ 0; if m p i
0 a a
Z b
npy ipy
sin sin dy ¼ 0; if n p j
0 b b
Hence, during integration of this term, we need only consider:

Z bZ 2
1 a
B2 w
D11 dxdy
2 0 0 Bx2
Z Z " 4 #
1 b a X
M X
M
mp mpx 2 npy
¼ D11 c2mn 2
sin sin dxdy
2 0 0 m ¼ 1n ¼ 1
a a b
After integration and evaluation, this term becomes:

Z Z 2 XM XN
1 b a
B2 w p4 b
D11 dxdy ¼ D11 m4 c2mn
2 0 0 Bx2 8a3 m ¼ 1 n ¼ 1
The following terms also appear in Eq. (17) and are evaluated in a similar
manner:
Z Z XM XN
1 b a
B2 w B2 w p4
2D12 dxdy ¼ D12 m2 n2 c2mn
2 0 0 Bx2 By2 4ab m ¼ 1 n ¼ 1
Z Z 2 XM XN
1 b a
B2 w p4 a
D22 dxdy ¼ D 22 n4 c2mn
2 0 0 By2 8b3 m ¼ 1 n ¼ 1
Z Z 2 XM XN
1 b a
B2 w p4
4D66 dxdy ¼ D66 m2 n2 c2mn
2 0 0 BxBy 2ab m ¼ 1 n ¼ 1
The remaining terms in Eq. (17) involve D16 and D26. Upon substituting the
appropriate derivatives, the first of these becomes:
Z Z 2 2
1 b a B w B w
4D16 dxdy
2 0 0 Bx 2 BxBy

Z Z " #
1 b a XM X N
mp 2 mpx npy
¼ 4D16 cmn sin sin
2 0 0 m ¼ 1 n ¼ 1
a a b
" #
XM XN
ip jp ipx jpy
cij cos cos dxdy
i ¼ 1 j ¼ 1
a b a b
This integration does not simplify as readily as those considered above. We

make use of the following identities:
8
Z > 0; if ðm þ iÞ is even
a
mpx ipx <
sin cos dx ¼ 2ma
0 a a >
: ; if ðm þ iÞ is odd
p½m2 i2
8
Z > 0; if ðn þ jÞ is even
b
npx jpx <
sin cos dy ¼ 2nb
0 b b >
: ; if ðn þ jÞ is odd
p½n2 j2
On the basis of these identities, we find after integration and evaluation:

Z Z
1 b a
B2 w B2 w
4D16 dxdy
2 0 0 Bx2 BxBy
X
M XN X M X N
2p2 m2 nij
¼ 2 D16 cmn cij ðMIÞðNJ Þ
a m ¼ 1n ¼ 1i ¼ 1j ¼ 1
½m2 i 2 ½n2 j 2
where:
ðMIÞ ¼ ½ð1Þm ð1Þi 1

ðNJ Þ ¼ ½ð1Þn ð1Þ j 1
Notice that:
8
< 0; if ðm þ iÞ is even
ðMI Þ ¼
:
2; if ðm þ iÞ is odd
8
< 0; if ðn þ jÞ is even
ðNJ Þ ¼
:
2; if ðn þ jÞ is odd

The final term that appears in Eq. (17) is evaluated in a similar manner and
becomes:
Z Z 2 2
1 b a B w B w
4D26 dxdy
2 0 0 By2 BxBy
X
M XN X M X N
2p2 mn3 ij
¼ 2 D26 cmn cij ðMIÞðNJ Þ
b m ¼ 1n ¼ 1i ¼ 1j ¼ 1
½m2 i 2 ½n2 j 2
We have now integrated all terms that appear in Eq. (17). Combining these
results and rearranging, the integrated form of Eq. (17) can be written:
XM XN 4
p 2 bm4 2m2 n2 an4

UII ¼ cmn D11 þ ðD12 þ 2D 66 Þ þ D 22
m ¼ 1n ¼ 1
8 a3 ab b3
XM X N
ij
2p mncmn
2
cij ð18Þ
i ¼ 1j ¼ 1
ð m 2 i 2 Þ ð n2 j 2 Þ

m2 n2
2 D16 þ 2 D26 ðMIÞðNJÞ
a b
3.3 Evaluation of Strain Energy Component UIII

Strain energy component UIII represents the change in strain energy associ-
ated with stress resultants Nij caused by bending. We assume that the stress
resultants Nij that exist prior to bending remain constant; only in-plane strains
are changed during bending. We must therefore evaluate the change in in-
plane strains caused by the development of out-of-plane displacements
w(x,y). The change in midplane strain e0xx may be determined via Fig. 1.
The figure shows an element of length dx, which represents an infinitesimal
element of the midplane that has already been deformed by stress resultants Nij
during step 1, as previously discussed. The length dx is further increased to
length dxV if the plate is deflected out of plane. From Fig. 1, we see that:
" 2 #1=2 " 2 #1=2
Bw Bw
dxV ¼ dx þ 2
dx 2
¼ dx 1 þ
Bx Bx
The quantity within the square bracket and raised to the 1/2 power can be
expanded in terms of a binomial power series expansion. A general statement
of this series expansion is given by (5):

1 1 1 2 1 1 3 3
ð1 þ nÞ1=2 ¼ 1 þ n n þ n þ ...
2 2 4 2 4 6

Figure 1 Sketch used to determine change in in-plane strain caused by out-of-
plane deflections.
We adopt this general formula for our use by retaining only the first two terms
and letting n = (Bw/Bx)2. Hence:
" 2 #1=2 2
Bw 1 Bw
1þ c1 þ
Bx 2 Bx
With this approximation, the new length of the element is given by:
" #
1 Bw 2
dx V ¼ dx 1 þ
2 Bx
The change in-plane strain e0xx caused by out-of-plane displacement w(x,y)

is labeled e0;b 0;b
xx . Based on the above, exx is given by:
2
dx V dx 1 Bw
exx ¼
0;b
¼ ð19aÞ
dx 2 Bx
Using a similar approach, the change in in-plane strains e0yy and c0xy caused by
out-of-plane displacement w(x,y) are given by:
2
1 Bw
yy ¼
e0;b ð19bÞ
2 By

Bw Bw
c0;b ¼ ð19cÞ
xy
Bx By
We can now calculate the strain energy associated with the change in in-plane
strains caused by bending. We assume in-plane stress resultants Nij remain
constant as bending develops. Consider, for example, the force represented
by stress resultant Nxx acting over an infinitesimal element with length and
width dx and dy. The x-directed force is (Nxxdy), and the distance through
which this force moves as bending develops equals e0;bxx dx. The incremental
strain energy associated with this force and caused by bending is therefore
dUIII=(Nxxdy)(e0;bxx dx). Analogous expressions hold for stress resultants Nyy
and Nxy. Hence, strain energy UIII is given by:
ZZ h
UIII ¼ xx þ Nyy eyy þ Nxy cxy dxdy
Nxx e0;b 0;b 0;b
ð20Þ
Equation (20) represents a general expression for UIII. Recall that during
application of the Ritz method, we are able to specify geometric boundary
conditions directly, but are not able to specify static boundary conditions, at
least directly. Therefore, in present form, Eq. (20) is inconvenient for use with
the Ritz method. That is, we wish to express stress resultants Nxx, Nyy, and
Nxy in terms of displacement fields, which will ultimately allow us to specify
geometric boundary conditions that represent known values of Nxx, Nyy, and
Nxy. From Eq. (44) of Chap. 6, we can write (for Bij = DM = 0)
Nxx ¼ A11 e0xx þ A12 e0yy þ A16 c0xy NTxx

Buo Bvo Buo Bvo
¼ A11 þ A12 þ A16 þ NTxx
Bx By By Bx
Similarly,

Buo Bvo Buo Bvo
Nyy ¼ A12 þ A22 þ A26 þ NTyy
Bx By By Bx

Buo Bvo Buo Bvo
Nxy ¼ A16 þ A26 þ A66 þ NTxy
Bx By By Bx
Substituting these expressions as well as Eqs. (19a,b,c) into Eq. (20), we
obtain:
Z Z
1 Buo Bvo Buo Bvo
UIII ¼ A11 þ A12 þ A16 þ NTxx
2 Bx By By Bx

Bw 2 Buo Bvo Buo Bvo
þ A12 þ A22 þ A26 þ Nyy
T
Bx Bx By By Bx
2
Bw Buo Bvo Buo Bvo
þ2 A16 þ A26 þ A66 þ NTxy
By Bx By By Bx

Bw Bw
dxdy ð21Þ
Bx By
We will now integrate Eq. (21) for the class of problems considered in this
text. We assume the plate is simply supported and that out-of-plane displace-
ments are given by Eq. (6). We also assume the laminate is symmetric and
subjected to uniform stress resultants Nij and/or a temperature change DT.
For these conditions, the in-plane displacement fields are given by Eqs. (7).
Substituting Eqs. (6) and (7) into Eq. (21) [and utilizing the simplifying
change in notation introduced as Eqs. (15)], we have:
Z Z "
1 b a
UIII ¼ A11 C1 þ A12 C2 þ 2A16 C3 NTxx
2 0 0
X
M X
N !2
mp mpx npy
cmn cos sin
m ¼ 1n ¼ 1
a a b
n o
þ A12 C1 þ A22 C2 þ 2A26 C3 NTyy
X
M X
N !2
np mpx npy
cmn sin cos
m ¼ 1n ¼ 1
b a b
n o
þ 2 A16 C1 þ A26 C2 þ 2A66 C3 NTxy
( )
X
M XN
mp mpx npy
cmn cos sin
m ¼ 1n ¼ 1
a a b
M X
X N !#
jp ipx jpy
cij sin cos dxdy
i ¼ 1j ¼ 1
b a b
Integration of this expression is simplified through the use of the trigono-

metric identities listed in preceding section. We obtain:
X N 2
X
M
p m2 b
UIII ¼ c2mn A11 C1 þ A12 C2 þ 2A16 C3 NTxx
m ¼ 1n ¼ 1
8 a

n2 a
þ A12 C1 þ A22 C2 þ 2A26 C3 Nyy T

b
h i
þ 2m2 ncmn A16 C1 þ A26 C2 þ 2A66 C3 NTxy
XM X N

j
cij ðMIÞðNJ Þ
i ¼ 1j ¼ 1
ðm2 i 2 Þðn2 j 2 Þ
This result can be further simplified by substituting Eq. (73b) of Chap. 6,

which give the thermal stress resultants in terms of elements of the [A] matrix,
and Eq. (5). We finally find:
X
M XN 2
p 2 m2 b n2 a
UIII ¼ cmn Nxx þ Nyy þ 2m2 ncmn Nxy
m ¼ 1n ¼ 1
8 a b
N
ð22Þ
XM X
j
cij ðMI ÞðNJ Þ
i ¼ 1j ¼ 1
As an aside, it is interesting to note that the change in in-plane strains

given by Eqs. (19a,b,c) are similar to the nonlinear terms that appear in
Green’s strain tensor, mentioned in Sec. 14 of Chap. 2. Recall that Green’s
strain tensor represents a definition of ‘‘finite’’ strains, which must be ac-
counted for during analyses involving large displacement gradients. For ex-
ample, finite strain exx is defined as:
" #
Bu 1 Bu 2 Bv 2 Bw 2
exx ¼ þ þ þ
Bx 2 Bx Bx Bx
2
Because we have incorporated the term e0;b xx = 1/2(Bw/Bx) during our
calculation of UIII, it would be natural to conclude that our analysis is valid
for finite strain levels. This conclusion would be incorrect. Our analysis is
based on infinitesimal strains, despite the inclusion of Eq. (18) during cal-
culation of UIII. To perform an analysis based on energy methods that ac-
counts for finite strain levels, we would need to develop new expressions
comparable to Eqs. (16), (18), and (20) (i.e., our current expressions for UI,
UII, and UIII, respectively), based on the Green strain tensor. In turn, this
would require a new derivation of results from Chap. 6; that is, an analysis
based on finite strains would require a rederivation of CLT. Such an analysis
is beyond the scope of this textbook and will not be discussed.
3.4 Evaluation of Work Done by Transverse Loads

The work done by the transverse load q(x,y) is denoted W and is calculated
in accordance with Eq. (8). Only simply supported plates are considered in

this text, and w(x,y) is assumed to be given by Eq. (6) in all cases. Therefore,
the work done by transverse loads is given by:
ZZ n ( )
o X M XN
mpx npx
W¼ qðx; yÞ cmn sin sin dxdy ð23Þ
m ¼ 1n ¼ 1
a b
Integration of Eq. (23) depends on the functional form of transverse load

q(x,y). Problems involving various types of transverse loads will be consid-
ered in the following sections. Equation (23) will be integrated as needed
during the discussion to follow.
4 SYMMETRIC COMPOSITE LAMINATES SUBJECT TO

SIMPLE SUPPORTS OF TYPE S4
In this section, the transverse deflections of simply supported symmetric
composite panels will be predicted on the basis of a Ritz analysis. The gen-
eral approach is to first obtain an expression for the total potential energy
P of the plate, given by:
P ¼ UI þ UII þ UIII W
In general, P is a function of the elastic properties of the plate, plate di-
mensions, and midplane displacement fields uo(x,y), vo(x,y), and w(x,y).
For the problems considered herein, in-plane displacement fields uo(x,y) and
vo(x,y) are known, while the out-of-plane displacement field w(x,y) is un-
known. The out-of-plane displacement field is assumed to be of the form:
X
M X
N
mpx npy
wðx; yÞ ¼ cmn sin sin
m ¼ 1n ¼ 1
a b
The magnitude of out-of-plane deflections are obtained by applying the prin-

ciple of minimum potential energy, which requires:

BP m ¼ 1; 2; . . . ; M
¼0
Bcmn n ¼ 1; 2; . . . ; N
This process leads to (M N) equations that must be satisfied simulta-
neously. Hence, by solving these equations for constants cmn, the magnitude
of out-of-plane displacements that corresponds to the state of minimum po-
tential energy is determined and the problem is solved.
Equations for strain energy components UI, UII, and UIII are identical
for all problems considered herein and were developed in Sec. 3. The work
done by the transverse load W depends on the nature of the applied load.
Solutions for a few common transverse loads are presented in the follow-
ing subsections.

4.1 Deflections Due to a Uniform Transverse Load
Consider a composite plate subjected to a constant and uniform transverse
load, q(x,y)=qo. In this case, Eq. (23) becomes:
Z bZ a ( XM XN )
mpx npx
W ¼ qo cmn sin sin dxdy
0 0 m ¼ 1n ¼ 1 a b
During integration of this expression, we note the following:

8
Z a >
< 0; if m is even
mpx
sin dx ¼
0 a >
:
2a
; if m is odd
mp
8
Z b >
< 0; if n is even
npx
sin dy ¼ 2b
0 b >
: ; if n is odd
np
Hence, the work done by a uniform transverse load can be written:
N i
X
M X abqo cmn h m
ih
n
W¼ ð1Þ 1 ð1Þ 1 ð24Þ
m ¼ 1n ¼ 1
p2 mn
The total potential energy can now be obtained by combining Eqs. (16), (18),
(22), and (24):

P¼ C1 Nxx þ C2 Nyy þ 2C3 Nxy ðabÞ
f g
X
M XN 4
p 2 bm4 2m2 n2 an4

cmn D 11 þ ðD 12 þ 2D66 Þ þ D22
m ¼ 1n ¼ 1
8 a3 ab b3
þ
X N
M X

ij m2 n2
2p2 mncmn cij D16 þ 2 D26 ðMI ÞðNJ Þ
i ¼ 1j ¼ 1
ðm 2 i 2 Þðn 2 j 2Þ a 2 b
f g
N 2
ð25Þ
X
M X p m2 b n2 a
c2mn Nxx þ Nyy
m ¼ 1n ¼ 1
8 a b
þ
X X
M N
j

þ2m2 ncmn Nxy cij ðMI ÞðNJ Þ
i ¼ 1j ¼ 1
(
)
X
M N
X

abqo cmn m n
ð1Þ 1 ð1Þ 1
m ¼ 1n ¼ 1
p2 mn
The four individual energy components UI, UII, UIII and W are shown within
the large braces in Eq. (25). This expression is unwieldy so a change in

notation will be made to facilitate inspection of the mathematical structure of
P. Toward that end, define the following constants:
p4 b 2p4 p4 a
F1 ¼ D11 F2 ¼ ðD12 þ 2D66 Þ F3 ¼ D22
8a3 8ab 8b3
p2 b p2 a 2p2
F4 ¼ Nxx F5 ¼ Nyy F6 ¼ 2 D16
8a 8b a
2p2 abqo
F7 ¼ 2 D26 F8 ¼ 2Nxy F9 ¼
b p2
Note that these terms are all known constants for a given laminate and
loading condition. On the basis of these definitions, Eq. (25) can be rear-
f g
ranged as follows:

cmn F1 m þ F2 m n þ F3 n þ F4 m þ F5 n
2 4 2 2 4 2 2

F9
þ cmn ð1Þm 1 ð1Þn 1
X
M X
N mn
P ¼ UI þ (
XM X N ð26Þ
m ¼ 1n ¼ 1 cij
þ cmn
i ¼ 1j ¼ 1
ðm i Þðn2 j 2 Þ
2 2

)
F6 ðm nijÞ þ F7 ðmn ijÞ þ F8 ðm njÞ ðMIÞðNJÞ
3 3 2
To further explore the Ritz method, we must now expand the expression for
P, based on some specified values of M and N. In general, the accuracy of the
Ritz approach is improved as M and N are increased. Although not required,
it is usual practice to let M=N, which means that the number of terms with x-
and y- dependency in Eq. (6) is identical. Often, 100 terms or more
(M=N=10, or more) are necessary to obtain a reasonable convergence of
the Ritz solution. Writing down the expanded form of P based on values of M
and N as high as 10 is obviously untenable. For purposes of illustration, we
will expand P using M=N=2, which will allow us to explore the essential
elements of the Ritz analysis.
Hence, expanding our Eq. (26) based on M=N=2, we find:
P ¼ UI þ c211 ðF1 þ F2 þ F3 þ F4 þ F5 Þ
þ c212 ðF1 þ 4F2 þ 16F3 þ F4 þ 4F5 Þ

ð27Þ
þ c221 ð16F1 þ 4F2 þ F3 þ 4F4 þ F5 Þ
þ 4c222 ð4F1 þ 4F2 þ 4F3 þ F4 þ F5 Þ

40
þ c11 c22 ð2F6 þ 2F7 þ F8 Þ
9
40
c12 c21 ð2F6 þ 2F7 þ F8 Þ 4c11 F9
9
Two important features of Eq. (27) should be noted. First, the expression for
P is a second-order polynomial in terms of the unknown coefficients cmn.
Second, the term involving F9 is a linear function of cmn. Referring to the
definitions of F1 through F9 listed above, it is seen that F9 is the only term
related to the transverse load qo. Hence, that portion of the total expression
for P which is related to the transverse loading is a linear function of the
unknown coefficients cmn. These two characteristics always hold for P,
regardless of the values of M and N or the nature of the transverse loading.
That is, P is always a second-order polynomial in cmn, and terms involving
the transverse load are always linear functions of cmn.
Proceeding with the Ritz analysis, we now apply the principle of min-
imum potential energy. That is, we wish to identify the particular values of cmn
dictated by:
BP
¼0
Bcmn
Because M=N=2 in this example, we must take four partial derivatives

of P, each of which will represent an independent equation that is then
equated to zero. For example, taking the derivative of P with respect to c11
and equating to zero, we have:
BP 40
¼ 2ðF1 þ F2 þ F3 þ F4 þ F5 Þc11 þ ð2F6 þ 2F7 þ F8 Þc22 4F9 ¼ 0
Bc11 9
Three additional equations are also formed (BP/Bc12=BP/Bc21=BP/

Bc22=0). The four equations can be represented using matrix notation as
shown in Fig. 2(a). These equations may be easily solved for coefficients cmn by
multiplying both sides of the equation by the inverse of the [4 4] array, as
shown in Fig. 2(b). Once coefficients cmn have been determined, the out-of-
plane deflections can be calculated using Eq. (6) and the problem is solved.
Referring to the definitions of F1 through F9 listed above, it is seen that
normal stress resultants Nxx and Nyy appear only in terms F4 and F5, re-
spectively, while the shear stress resultant Nxy appears only in term F8.
Inspection of the [4 4] array shown on the left side of the equality in Fig. 2(a)
reveals that F4 and F5 appear only along the main diagonal of the matrix,
whereas F8 appears only in off-diagonal positions. This pattern always oc-
curs, regardless of the value of the value of M and N or the nature of the
transverse loading—the normal stress resultants appear only along the main

Figure 2 Summary of the solution obtained for a simply supported laminate subjected to a uniform transverse
pressure using a Ritz analysis and M = N = 2. (a) The set of simultaneous equations obtained by enforcing
BP/Bcmn= 0, for M = N = 2. (b) Solving the set of simultaneous equations shown in part (a).
diagonal, while the shear stress resultant appears only in off-diagonal posi-
tions. In summary, Fig. 2 represents the solution for a simply supported
symmetric composite panel subjected to a uniform transverse pressure qo,
a uniform in-plane stress resultants Nxx, Nyy, and Nxy, and a uniform change
in temperature, DT, based on M = N = 2. Depending on details of a specific
problem, the number of terms required to insure convergence may be subs-
tantially higher than only four terms; it is not uncommon to require 100 terms
or more (M = N = 10 or more). Obviously, the solution process presented
above is rarely (if ever) performed by hand calculation. Rather, a computer-
based routine is typically used to expand Eq. (26) [or, equivalently, Eq. (25)]
for specified values of M and N, to perform the required partial differen-
tiation, to determine the inverse of the resulting [M N] matrix, and to
complete the final matrix multiplication that gives the coefficients cmn. Typical
solutions obtained on the basis of the Ritz approach are illustrated in the
following three sample problems.
Sample Problem 1
A [(F45/0)2]s graphite-epoxy laminate is cured at 175jC and then cooled to
plane dimensions of 300150 mm and is mounted in an assembly that
provides type S4 simple supports along all four edges. The x axis is defined
parallel to the 300-mm edge (i.e., a = 0.3 m, b = 0.15 m). The laminate is then
subjected to a uniform transverse load q(x,y) = 30 kPa. No in-plane loads are
applied (i.e., Nxx = Nyy = Nxy = 0). Determine the maximum out-of-plane
displacement based on a Ritz analysis and plot the out-of-plane displacement
field. Use the properties listed for graphite-epoxy in Table 3 of Chap. 3 and
assume each ply has a thickness of 0.125 mm.
Solution. Based on the properties listed in Table 3 of Chap. 3 for graphite-
epoxy, the [ABD] matrix for a [(F45/0)2]s laminate is:
2 3
145:2 106 35:3 106 0 0 0 0
6 35:3 106 64:8 106 0 0 0 0 7
6 7
6 0 0 50:2 10 6
0 0 0 7
6
½ABD ¼ 6 7
6 0 0 0 22:3 7:97 2:20 7 7
4 0 0 0 7:97 14:3 2:20 5
0 0 0 2:20 2:20 10:8
where the units of Aij are Pa m and the units of Dij are Pa m3. Notice that
neither D16 nor D26 equals zero; hence, the laminate is generally ortho-
tropic. The 12-ply laminate has a total thickness t = 1.5 mm and the aspect
ratio R = a/b = 2.0.
The computer program SYMM (described in Sec. 6) can be used to
perform the required Ritz analysis. Several analyses were performed using

increasing values of M (and N ) to evaluate whether the solution has con-
verged to a reasonably constant value. Solutions were obtained using values
of M (and N ) ranging from 1 through 10 (i.e., analyses were performed in
which the number of terms used to describe the displacement field ranged
from 1 through 100). Maximum predicted displacement is plotted as a func-
tion of M and N in Fig. 3. As indicated, the maximum displacement con-
verges to a value of 8.03 mm when M=N=10.
A contour plot of out-of-plane displacements predicted using M=N=
10 is shown in Fig. 4. As would be expected, the maximum displacement
occurs at the center of the plate (i.e., at x = 150 mm, y = 75 mm). Careful
examination of these contours will reveal that the contours are very slightly
distorted. This distortion (which is barely discernible in Fig. 4) occurs be-
cause the plate is generally orthotropic. That is, for a [(F45/0)2]s laminate
D16, D26 p 0. However, for this problem, the magnitudes of D16 and D26
(relative to D11 and D22) are very small. Specifically, for the laminate
considered in this problem D16/D11 = D26D11 = 0.0986 and D16/D22 =
D26 D22 = 0.153. Consequently, distortion of out-of-plane displacements is
very slight. The out-of-plane displacement induced by a uniform transverse
load applied to a laminate with relatively higher values of D16 and D26 is
considered in Sample Problem 3. As will be seen, the distortion of displace-
ment contours is much more pronounced in that case due to the relatively
higher values of D16 and D26.
Figure 3 Convergence of predicted plate deflections based on a Ritz analysis

as M and N are increased from 1 to 10 (i.e., as the number of terms is increased
from 1 to 100).

Figure 4 A contour plot of out-of-plane displacements for the [(F45/0)2]s
laminate considered in Sample Problem 1.
Sample Problem 2
The [(F45/0)2]s graphite-epoxy laminate described in Sample Problem 1 is
again subjected to type S4 simple supports along all four edges and a uniform
transverse load q(x,y) = 30 kPa. However, the plate is now also subjected to
uniform in-plane stress resultants Nxx = Nyy. Use a Ritz analysis to deter-
mine the maximum out-of-plane displacement for 0 < N xx = N yy<
100 kN/m.
Solution. As was the case for Sample Problem 1, solutions for this prob-
lem can be obtained using program SYMM (described in Sec. 6). Multiple
solutions were obtained using the specified range in Nxx and Nyy, and 100
terms were used in the displacement field in all cases.
Results are summarized in Fig. 5. The maximum displacement occurs
at the center of the plate (i.e., at x = 150 mm, y = 75 mm). As would be
expected, the in-plane tensile stress resultants tend to reduce out-of-plane
displacement. For Nxx = Nyy = 0, a maximum out-of-plane displacement of
8.03 mm is predicted. In contrast, if tensile stress resultants Nxx = Nyy = 100
kN/m are applied, the maximum out-of-plane displacement is reduced to
0.71 mm.

Figure 5 Predicted maximum deflections of the plate considered in Sample
Problem 2 (M=N=10 in all cases).
Sample Problem 3
A [25j]12 graphite-epoxy laminate is trimmed to in-plane dimensions of
300150 mm and is mounted in an assembly that provides type S4 simple
supports along all four edges. The laminate is then subjected to a uniform
transverse load q(x,y) = 30 kPa. No in-plane loads are applied (i.e., Nxx=
Nyy = Nxy = 0). Determine the maximum out-of-plane displacement based
on a Ritz analysis and plot the out-of-plane displacement field. Use the
properties listed for graphite-epoxy in Table 3 of Chap. 3, and assume each
ply has a thickness of 0.125 mm.
Solution. Note that the plate has an aspect ratio R = 150/300 = 2.0, as was
the case for the laminates considered in Sample Problems 1 and 2. A rather
unusual fiber angle of 25j has been selected for consideration in this prob-
lem because it results in high relative values of D16 and D26, resulting in an
interesting distortion of the predicted out-of-plane displacement field. Spe-
cifically, for this laminate:
D16 =D11 ¼ 0:370 D16 =D22 ¼ 2:21
D26 =D11 ¼ 0:126 D26 =D22 ¼ 0:755

These relative values of D16 and D26 are quite high, at least as compared
to those exhibited by the [(F45/0)2]s laminate considered in Sample Prob-
lem 1.
A solution for this problem was obtained using program SYMM, using
M=N=10. A maximum displacement of 16.1 mm is predicted to occur at the
center of the plate. A contour plot of out-of-plane displacements is shown
in Fig. 6. Distortion of the displacement field due to the generally ortho-
tropic nature of the [25j]12 panel is obvious, especially when compared to the
very slightly distorted pattern for a [(F45/0)2]s laminate, previously shown in
Fig. 4.
4.2 Deflections Due to a Sinusoidal Transverse Load

Consider a simply supported symmetric composite plate subjected to a trans-
verse load that varies sinusoidally over the surface of the plate:
px py
qðx; yÞ ¼ qo sin sin
a b
Figure 6 A contour plot of out-of-plane displacements for the [25]12 laminate

considered in Sample Problem 3.

In this case, the work done by the transverse load, calculated using Eq. (23),
is given by:
Z bZ a n o(X M X N mpx npx
)
px py
W ¼ qo sin sin cmn sin sin dxdy
0 0 a b m¼1 n¼1
a b
During integration of this expression, we note the following:

8
Z a px mpx < 0; if m p 1
sin sin dx ¼ a
0 a a : ; if m ¼ 1
2
8
Z b py npx < 0; if n p 1
>
sin sin dy ¼ b
0 b b >
: ; if n ¼ 1
2
Hence, the work done by a sinusoidal transverse load is simply:
qo abc11
W¼
4
The total potential energy is P=UI+UII+UIII+W. Our earlier expressions
for UI, UII, and UIII are not altered by the change in transverse load and
are given by Eqs. (16), (18), and (22), respectively. Hence, the total potential
energy is:

f g
X
M XN 4
p 2 bm4 2m2 n2 an4

cmn D 11 þ ðD 12 þ 2D66 Þ þ D22
m ¼ 1n ¼ 1
8 a3 ab b3
XM X N
ij
þ 2p2 mncmn cij ð28Þ
i ¼ 1j ¼ 1
½ m2 i 2 ½n 2 j 2

m2 n2
D16 þ D26 ðMI ÞðNJ Þ
a2 b2
f g
X
M XN 2
p 2 m2 b n2 a
cmn Nxx þ Nxx
m ¼ 1n ¼ 1
8 a b
þ N

XM X
j
þ 2m ncmn Nxy
2
cij ðMI ÞðNJ Þ
i ¼ 1j ¼ 1

qo abc11

4

As before, the next step is to apply the principle of minimum potential
energy:

BP m ¼ 1; 2; . . . ; M
¼0
Bcmn n ¼ 1; 2; . . . ; N
This process leads to (M N) equations (similar in form to those shown in
Fig. 2) that must be satisfied simultaneously. Hence, by solving these equa-
tions for constants cmn, the out-of-plane displacement caused by a transverse
load that varies sinusoidally over the surface of the plate can be calculated
using Eq. (6), and the problem is solved.
4.3 Deflections Due to a Transverse Load Distributed Over

an Interior Region
A composite panel subjected to a transverse force P uniformly distributed
over an internal rectangular region of dimensions ai bi was previously
shown in Fig. 10 of Chap. 9. The center of the interior region is located at
x = n and y = g. As discussed in Sec. 4 of Chap. 9, this loading can be
expressed in terms of a Fourier series expansion:
l
16P X l X 1 mpn npg mpai
qðx; yÞ ¼ 2 sin sin sin
p ai bi m ¼ 1 n ¼ 1 mn a b 2a

npbi mpx npy
sin sin sin
2b a b
Substituting the above into the expression representing the work done by
transverse loads W [Eq. (23)] and integrating, it will be found:
N
4abP X M X
cmn mpn npg mpai npbi
W¼ sin sin sin sin
p2 ai bi m¼1 n¼1 mn a b 2a 2b
The total potential energy is P=UI+UII+UIII+W. Our earlier expressions

for UI, UII, and UIII are not altered by the change in transverse load and are
given by Eqs. (16), (18), and (22), respectively. Hence, the total potential
energy is:

f g
N 4
M X
X
p bm4 2m2 n2 an4

c2mn 3
D11 þ ðD12 þ 2D66 Þ þ 3 D22
m ¼ 1 n¼1
8 a ab b
XM X N
ij
þ 2p2 mncmn cij
i ¼ 1j ¼ 1
½ m 2 i 2 ½n2 j 2
2

m n2
2 D16 þ 2 D26 ðMI ÞðNJ Þ
a b

f g
XM X N 2
p 2 m2 b n2 a
cmn Nxx þ Nyy
m ¼1n ¼1
8 a b
þ
X N
M X

j ð29Þ
i ¼ 1j ¼ 1
( )
N
4abP X M X
cmn mpn npg mpai npbi
sin sin sin sin
p2 ai bi m ¼ 1 n ¼ 1 mn a b 2a 2b
As before, the next step is to apply the principle of minimum potential energy:

BP m ¼ 1; 2; . . . ; M
¼0
Bcmn n ¼ 1; 2; . . . ; N
This process leads to (M N ) equations (similar in form to those shown

in Fig. 2) that must be satisfied simultaneously. Hence, by solving these
equations for constants cmn, the out-of-plane displacements caused by a
transverse force P uniformly distributed over an interior rectangular region
of dimensions ai bi can be calculated using Eq. (6), and the problem is
solved.
4.4 Deflections Due to a Transverse Point Load

The work done by a concentrated point load P applied at x = n and y = g can
be obtained by allowing ai ! 0, bi ! 0 in Eq. (29). In the limit, we obtain:
XM X N
mpn npg
W¼P cmn sin sin
m ¼ 1n ¼ 1
a b
The total potential energy is:

f g
X N 4
M X
p bm4 2m2 n2 an4

c2mn D11 þ ðD12 þ 2D 66 Þ þ D22
m ¼1n ¼ 1
8 a3 ab b3
X N
M X
þ ij
2p2 mncmn cij
i ¼1j ¼ 1
½ m2 i 2 ½n 2 j 2
2

m n2
2 D16 þ 2 D26 ðMI ÞðNJ Þ
a b

f g
XM X N 2
p 2 m2 b n2 a
cmn Nxx þ Nyy
m ¼1n ¼1
8 a b
þ
X N
M X

j
i ¼1j ¼ 1
(
)
N
M X
X
mpn npg
P cmn sin sin ð30Þ
m ¼1n ¼1
a b
We apply the principle of minimum potential energy:

BP m ¼ 1; 2; . . . ; M
¼0
Bcmn n ¼ 1; 2; . . . ; N
This process leads to (M N) equations (similar in form to those shown in
Fig. 2) that must be satisfied simultaneously. Hence, by solving these equa-
tions for constants cmn, the out-of-plane displacements caused by a concen-
trated load P applied at x = n and y = g can be calculated using Eq. (6), and
the problem is solved.
5 BUCKLING OF SYMMETRIC COMPOSITE PLATES SUBJECT

In this section, buckling of type S4 simply supported symmetric composite
panels will be considered. A brief explanation of what is meant by ‘‘buckling’’
was presented in Sec. 4 of Chap. 10. In essence, the term buckling refers to the
fact that (under the proper circumstances) in-plane stress resultants Nxx, Nyy,
and/or Nxy can cause out-of-plane displacements w(x,y). A coupling between
in-plane loads and out-of-plane displacement is of course expected for non-
symmetric laminates because Bij p 0 in this case. We have limited discussion
to symmetric laminates, however, so we are considered with a coupling be-
tween in-plane loads and out-of-plane displacement that is not predicted by
the CLT analysis developed in Chap. 6.
The buckling phenomenon can be explained on the basis of the prin-
ciple of minimum potential energy as follows. Based on our earlier discus-
sion, the total potential energy P of an initially flat plate subjected to in-
plane loading is given by:
P ¼ UI þ UII þ UIII
Note that the work term W no longer appears in our expression for P be-
cause we assume that no transverse loads are applied. General expressions
for strain energy components UI, UII, UIII for a composite panel have been

presented as Eqs. (14), (17), and (21), respectively. Inspection of these equa-
tions reveals that UI is a function of in-plane displacements uo(x,y) and
vo(x,y), that UII is a function of out-of-plane displacements w(x,y), and that
UIII is a function of uo, vo, and w. If an initially flat plate is subjected to
relatively low levels of in-plane loads, then only in-plane displacements oc-
cur; that is, the plate remains flat because at low load levels, the flat con-
figuration corresponds to the state of minimum potential energy. In this case,
UI is the only nonzero strain energy component: UII = UIII = 0 because
w(x,y) = 0. However, as the magnitude of in-plane loading is increased, then
the configuration that corresponds to the state of minimum potential energy
may no longer be flat but rather bent; the configuration that corresponds to
the state of minimum potential energy involves out-of-plane displacement
w(x,y) and the plate ‘‘buckles.’’ Once bucking occurs, then UI, UII, UIII p 0.
In Secs. 3 and 4, we developed expressions for the potential energy of a
simply supported symmetric composite panel subjected to both in-plane and
transverse loads. These earlier results can be adopted for present purposes by
simply discarding those terms involving the transverse load. Hence, from Eq.
(26), we can immediately write:
"
XM X N
P ¼ UI þ c2mn F1 m4 þ F2 m2 n2 þ F3 n4 þ F4 m2 þ F5 n2
m ¼ 1n ¼ 1
(
X N
M X
cij
þcmn F6 ðm3 nijÞ þ F7 ðmn3 ijÞ ð31Þ
i ¼ 1j ¼ 1
ðm 2 i 2 Þðn2 j 2 Þ

)#
þF8 ðm2 njÞ ðMI ÞðNJ Þ
The constants F1 through F8 were defined in Sec. 4, in conjunction with Eq.

(26). Note that one constant, namely, F9, no longer appears. This term has
been dropped because it represents a transverse load that has now been
assumed to be zero. We can predict the onset of buckling by applying the
principle of minimum potential energy:

BP m ¼ 1; 2; . . . ; M
¼0
Bcmn n ¼ 1; 2; . . . ; N
This process leads to (M N) equations that must be satisfied simulta-

neously. For purposes of illustration, consider the set of equations that are
obtained using M = N = 2. In this case, we obtain the same set of equa-
tions as those previously illustrated in Fig. 2(a), except that now F9 = 0.
Referring to Fig. 2(a), we see that in the present case, all terms on the right
side of the equality are zero.

We must now specify the loading condition of interest. That is, we must
specify the loading conditions under which buckling is to be predicted. There
are many possibilities although we have limited discussion in this chapter to
type S4 simple supports. For example, we may wish to determine the value
of Nxx that will cause buckling, given some constant values for Nyy and Nxy.
Alternatively, we may wish to determine the value of Nxy that will cause
buckling, given some constant values for Nxx and Nyy. A third possibility is a
loading situation in which all three resultants are increased proportionately
(i.e., Nyy = k1Nxx and Nxy = k2Nxx, where k1 and k2 are known constants). In
this case, we are interested in buckling caused by a simultaneous increase in
Nxx, Nyy, and Nxy.
For illustrative purposes, let us assume that we are interested in
buckling caused by uniaxial loading, i.e., for the loading condition Nxx p 0,
Nyy = Nxy = 0. In this case, we have:
p2 a
F5 ¼ Nyy ¼ 0 F8 ¼ 2Nxy ¼ 0
8b
Also, F9=0, because no transverse load is applied. For this situation, the set
of equations shown in Fig. 2(a) reduce to:
2 3
40
6 2ðF1 þ F2 þ F3 þ F4 Þ 0 0
9
ð2F6 þ 2F7 Þ 7
6 7
6 40 7
6 0 2ðF þ 4F þ 16F þ F Þ ð2F þ 2F Þ 0 7
6 1 2 3 4 6 7 7
6 9 7
6 40 7
6 ð2F6 þ 2F7 Þ 2ð16F1 þ 4F2 þ F3 þ 4F4 Þ 7
6 0 0 7
6 9 7
4 40 5
ð2F6 þ 2F7 Þ 0 0 8ð4F1 þ 4F2 þ 4F3 þ F4 Þ
9
8 9 8 9
c11 > > 0>
>
> > > >
> >
>
>
> > >
>
> >
>
>
>
>
= >
< c12 >
> <0> =
¼
>
> c > > > >0> >
> 21 >
> > > > > >
>
> >
> > >
>
: ; > : >
> ;
c22 0
The unknown buckling load is contained in the term F4 = p2bNxx/8a. We

can therefore rearrange the above expression by bringing all terms involving
F4 to the right side of the equality:
2 3
40
6 2ðF1 þ F2 þ F3 Þ 0 0
9
ð2F6 þ 2F7 Þ 7
6 7
6 40 7
6 0 2ðF1 þ 4F2 þ 16F3 Þ ð2F6 þ 2F7 Þ 0 7
6 7
6 9 7
6 40 7
6 ð2F6 þ 2F7 Þ 2ð16F1 þ 4F2 þ F3 Þ 7
6 0 0 7
6 9 7
4 40 5
ð2F6 þ 2F7 Þ 0 0 8ð4F1 þ 4F2 þ 4F3 Þ
9

8 9 8 98 9
> c11 > > 2 0 0 0 >> c11 >
>
> >
> >
> >
> >
> >
>
>
> > > > > >
<c > = >
< 0 2 0 0
>
= >
< >
c12 =
12
¼ F4
>
> >
> >
> >> >
>
>
>
c 21 >
>
>
>
>
> 0 0 8 0 > >
>
>
>
>
>
>
c21 >
>
>
>
: ; : ;: ;
c22 0 0 0 8 c22
8 98 9
> p2 b > ð32Þ
>
> >
>>
> c11 >
>
>
> 0 0 0 >
>>
> >
>
>
> 4a >
>>c >
< =>
< 12 > =
¼ Nxx p 2
b
> 0 0 0 >>
>
> a >
>>
> c21 >
>
>
>
> >
>>
> >
>
>
> p b
2 >
>>
: >
;
: 0 0 0 ; c22
a
Equation (32) is in the form of a so-called generalized eigenvalue problem. In
general, there are (M N) eigenvalues that satisfy a generalized eigenvalue
problem; for each eigenvalue, there exists a corresponding eigenvector. For
this particular example, the eigenvalues corresponds to Nxx, and the eigen-
vector corresponds to the column matrix containing coefficients cij. Because
we have assumed M = N = 2, there are four values of Nxx that will satisfy
Eq. (32). The eigenvalue with lowest magnitude represents the critical
c
buckling load Nxx , and the coefficients cij that correspond to this eigenvalue
represent the critical buckling mode because wðx; yÞ ¼ SM m ¼ 1 Sn ¼ 1 cmn sin
N
(mkx/a) sin (mky/b). As in all analyses based on the Ritz method, several
predictions of buckling load should be obtained using increased values of M
and N to insure convergence of the predicted buckling load and mode.
Equation (32) represents the generalized eigenvalue problem for the
case Nxx p 0, Nyy = Nxy = 0, and M = N = 2. As a second example, consider
buckling caused by a pure shear load. That is, assume Nxy p 0, Nxx=Nyy=0.
We now have F4 = F5 = F9 = 0. Following a process identical to that
described above, we arrive at the following generalized eigenvalue problem:
2 3
40
6 2ðF1 þ F 2 þ F 3 Þ 0 0 ð2F 6 þ 2F 7 Þ 7
6 9 7
6 40 7
6 0 2ðF1 þ 4F2 þ 16F3 Þ ð2F6 þ 2F7 Þ 0 7
6 7
6 9 7
6 40 7
6 ð2F6 þ 2F7 þ F8 Þ 2ð16F1 þ 4F2 þ F3 Þ 7
6 0 0 7
6 9 7
4 40 5
ð2F6 þ 2F7 Þ 0 0 8ð4F1 þ 4F2 þ 4F3 Þ
9
8 9 8 98 c 9
>
> c11 >> > 80 >> 11 >
>
> >
> >
> 0 0 0 >>
> >
>
>
>
>
>
>
>
>
>
> 9 >
>
>
>
>
>
>
>
>
>
>
>
>
> >
> >
> 80 >> >
0 > => >
c c
< 12
= < 0 0 < 12
=

>
¼ Nxy 9 ð33Þ
c > > >> c21 >
> > > 80
> 21 >
> > >
> 0 0 0 > >
>
>
>
>
>
>
>
>
> >
> >
> 9 >
>>
> >
>
>
> >
> >
> >
>>
> >
>
>
: >
; : 80 0 0 0 ;: > >
;
c22 9 c22

Once again, there will be four eigenvalue/eigenvector pairs that satisfy Eq.
(33) because we have assumed M = N = 2. The eigenvalue with lowest
c
magnitude represents the critical buckling shear load Nxy and the corre-
sponding eigenvector represents the critical buckling mode.
6 COMPUTER PROGRAM SYMM

The solutions described in this chapter are implemented in a computer
program called SYMM. This program can be downloaded at no cost from
the following website: http://depts.washington.edu/amtas/computer.html.
Program SYMM is based on the Ritz method and is applicable to any sym-
metric composite laminate. The program prompts the user to input all in-
formation necessary to perform these calculations. Properties of up to five
different materials may be defined. The user must input various numerical
values using a consistent set of units. For example, the user must input elastic
moduli for the composite material system(s) of interest. Using the properties
listed in Table 3 of Chap. 3 and based on the SI system of units, the following
numerical values would be input for graphite-epoxy:
E11 ¼ 170 109 Pa E22 ¼ 10 109 Pa v12 ¼ 0:30
G12 ¼ 13 10 Pa
9
Because 1 Pa=1 N/m2, all lengths must be input in meters. For example, ply
thicknesses must be input in meters (not millimeters). A typical value would
be tk=0.000125 m (corresponding to a ply thickness of 0.125 mm). Similarly,
if an analysis of a plate with a length and width of 500300 cm were being
performed, then the length and width of the plate must be input as 5.00 and
3.00 m, respectively.
If the English system of units were used, then the following numerical
values would be input for the same graphite-epoxy material system:
E11 ¼ 25:0 106 psi E22 ¼ 1:5 106 psi v12 ¼ 0:30
G12 ¼ 1:9 106 psi

All lengths would be input in inches. A typical ply thickness might be
tk=0.005 in., and the length and width of a plate might be 36 and 20 in.,
for example.
REFERENCES
1. Whitney, J.M. Structural Analysis of Laminated Anisotropic Plates. Technomic
Pub Co.: Lancaster, PA, ISBN 87762-518-2.

2. Turvey, G.J.; Marshall, I.H., Eds. Buckling and Postbuckling of Composite
Plates. Chapman and Hall: New York, NY, 1995.
3. Fung, Y.C. A First Course in Continuum Mechanics. Prentice-Hall: Englewood
Cliffs, NJ, 1969.
4. Timoshenko, S.; Woinowsky-Krieger. Theory of Plates and Shells. McGraw-Hill
Book Co.: New York, NY, 1987. ISBN 0-07-0647798.
5. The binomial series expansion can be found in most reference books devoted to
mathematical tables and formulas. See, for example, Korn, G.A.; Korn, T.M.
Mathematical Handbook for Scientists and Engineers. 2nd ed. McGraw-Hill
Book Co.: New York, NY, 1968. Table E-6.

Appendix A
Finding the Cube-Root of a Complex Number
The process of finding the principal stresses and principal strains is dis-
cussed in Section 7 and 12 of Sec. 2, respectively. The approach described
in these sections ultimately leads to the requirement of calculating the
cube root of a complex number. The nth-root of a complex number can
be determined using ‘‘DeMoivre’s theorem,’’ which is explained in many
advanced calculus books. A complete review of this topic is beyond the
scope of the present discussion. The explanation given below is a
specialization of DeMoivre’s theorem to find the cube root of a complex
number.
Any complex number z can be written as:
z ¼ h þ iv
where:
pffiffiffiffiffiffi
iu i
h ¼ the ‘‘real part’’ of z
v ¼ the ‘‘imaginary part’’ of z
The modulus (r) and argument (u) of a complex number are defined as:
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
r ¼ h2 þ v 2 ðA:1Þ
617

u ¼ tan1 ðv=hÞ ðA:2Þ
The relationship between the modulus, argument, and components h
and v can be visualized by plotting the number z in the complex plane, as
shown in Fig. 1.
Based on the above definitions, DeMoivre’s theorem can be used to
show that the cube root of a complex number z = h+iv is given by:

p ffiffiffi p ffiffiffiffiffiffiffiffiffiffiffiffi lnðrÞ n h u i h u io
3
z ¼ h þ iv ¼ exp
3
cos þ isin ðA:3Þ
3 3 3
Two important cautions should be noted:
The value of u used in Eq. (A.3) must be expressed in radians (not in
degrees).
Referring to Fig. 1, note that the function tan1(v/h) corresponds to
two different arguments (that is, two different angles), which occur in
opposite quadrants. For example, assume that both h and v are
positive for some complex number z = h+iv (this is the situation
illustrated in Fig. 1). Referring to Fig. 1, it is clear that in this case,
argument u is an angle in the first quadrant of the complex plane:
p
0 V u radians
2
In contrast, consider a different complex number z V, where z V= 1* z=
h iv. The modulus r for complex numbers z and zV [calculated using Eq.
Figure 1 The number z plotted in the complex plane, showing the real part (h),
imaginary part (v), the modulus (r), and the argument (u).

(A.1)] is identical. However, the argument of z V, angle uV say, is an angle in the
third quadrant of the complex plane:
3p
p rad V uV V radians
2
Hence, when calculating the argument of a complex number using Eq. (A.2),
one must insure that the angle returned by the tan1( ) function corresponds
to the proper quadrant, as determined by the algebraic signs of h and v.

Appendix B
Experimental Methods Used to Measure
In-Plane Properties E11, E22, v12, and G12
An abbreviated discussion of tests used to measure tensile in-plane proper-

ties E11, E22, v12, and G12 is provided in this Appendix. The fact that this
discussion is limited to tensile test methods is significant and should be
noted by the reader. First, since E11, E22, and v12 often differ in tension and
compression, in practice both tensile and compressive values should be
measured. Specimen geometries used during compression testing generally
differ from the tensile test specimen geometries described herein. Secondly,
it is more-or-less implied that the specimens described herein are produced
using pre-preg tape. As opposed to resin-transfer molding or filament wind-
ing, for example. Third, material properties can also be inferred from flex-
ure tests, and these will not be discussed herein. The reader interested in a
more detailed discussion of the experimental methods used to measure com-
posite properties should consult Refs. (1–3) or the appropriate ASTM test
standards (many of which are listed in Tables 1 and 2 of Chapter 3).
The most widely used tensile test specimen geometry is based on ASTM
Standard 3039 ‘‘Test Method for Tensile Properties of Polymer Matrix Com-
posite Materials.’’ Typical specimens that conform to this standard are
shown in Figs. 1–4. As indicated, tensile specimens are flat and straight-sided
with rectangular cross section. Very often adhesively bonded end tabs are
used. The [0]n specimen, the ‘‘off-axis’’ [h]n specimen, and the [90]n specimen
621

Figure 1 A [0]n tensile specimen, showing typical dimensions and adhesively
bonded end tabs.
shown in Figs. 1–3 are equipped with adhesively bonded end tabs. As the
use of bonded end-tabs increases specimen preparation time and cost, they
are only used when necessary. In general, end tabs are used when
the fracture stress or strain is to be measured, or if
it is found that the specimen fails within the grip region if bonded
end tabs are not used
End tabs used with high-strength specimens (such as the [0]n or [h]n specimens
shown in Figs. 1 and 2, respectively) must be beveled at a shallow angle to
avoid specimen failure at the end of the tab. The shallow bevel provides for a
smooth transfer of load from the grip region to the gage region of the

Figure 2 A [h]n tensile specimen, showing typical dimensions and adhesively
bonded end tabs.
specimen. The bevel angle is usually about 5–7j, although angles as high as
30j are used on occasion. Use of a shallow bevel angle becomes less important
when testing low-strength specimens, and in fact end tabs with a bevel angle
of 90j are often used when testing [90]n specimens (Figs. 3).
As a general rule all tensile test specimens are long and narrow. That is,
they have a high ‘‘aspect ratio.’’ Aspect ratio equals specimen length (where
specimen length is defined as the tab-to-tab distance) to specimen width.

Figure 3 A [90]n tensile specimen, showing typical dimensions and adhesively
bonded end tabs.
However, high-strength composite specimens must have an exceptionally

high aspect ratio. For example, the aspect ratios of the [0]n and [h]n specimens
shown in Figs. 1 and 2 are 9.3 and 12.0, respectively. In contrast, the aspect
ratio of the relatively low-strength [90]n specimen (Fig. 3) is 5.0.
Bonded end tabs are often not necessary when testing symmetric
and balanced laminates, such as the [F45]ns specimen shown in Fig. 4. In
these cases so-called ‘‘friction’’ tabs may be used. A friction tab is held in
place by the pressure of the grips. Occasionally, an abrasive paper (such
as emery cloth) is placed between the tab and surface of the specimen to
enhance friction.
End tabs (either bonded or friction) are most often made using a
cross-ply [0/90]ns E-glass/polymer composite, although end tabs made using
steel or aluminum have also been successfully used.

Figure 4 A [F45]ns tensile specimen, showing typical dimensions.
Strains induced during a tensile test are measured using either bonded
resistance strain gages or extensometers. Properties E11 and v12 are measured
using a [0]n specimen while E22 is measured using a [90]n specimen. It is
possible to measure v21 as well, using a [90]n specimen. However, for most
advanced unidirectional composites the numerical value of v21 (the so-called
‘‘minor’’ Poisson ratio) is very small and is at least an order of magnitude
smaller than the ‘‘major’’ Poisson ratio v12. For example, for graphite/epoxy
the major Poisson ratio v12c0.3 whereas the minor Poisson ratio v12c0.01.
The small value of v21 can lead to a relatively high measurement error.

Consequently, v21 is rarely measured in practice. If the value of v21 is required
it is calculated using the inverse relation:
E22
v21 ¼ v12
E11
Several techniques by which the shear modulus G12 can be measured
using a tensile specimen are available. One approach is based on an off-axis
[h]n specimen. Referring to Eq. (38c) in Chapter 5, the axial modulus Exx of
such a specimen is given by:
1
Exx ¼
cos ðhÞ
4
1 2v12 sin4 ðhÞ
þ cos2 ðhÞsin2 ðhÞ þ
E11 G12 E11 E22
Solving this equation for G12, we obtain:
Exx E11 E22 cos2 ðhÞsin2 ðhÞ

G12 ¼ ðB:1Þ
E11 E22 Exx E22 cos4 ðhÞ þ 2v12 E22 cos2 ðhÞsin2 ðhÞ E11 sin4 ðhÞ
Thus if Exx is measured for a [h]n specimen (where h is known), and assuming
that E11, v12, and E22 have also been measured and are known, then the shear
modulus G12 can be calculated using Eq. (B.1). For example, if h = 45j (the
most common case) then Eq. (B.1) reduces to:
Exx E11 E22

G12 ¼ ðB:2Þ
4E11 E22 Exx ½E22 þ 2v12 E22 E11
An alternate method is based on the use of a [F45]ns specimen. In this

case the axial and transverse strains (exx and eyy) induced by a uniaxial tensile
stress (rxx) are measured, usually using biaxial strains gages. For this stacking
sequence it can be shown (4) that the shear stress (s12) and shear strain (c12)
induced in the +45j plies are given by:
s12 Aþ45j ¼ rxx =2

c12 Aþ45j ¼ ðexx eyy Þ
In contrast, the shear stress and strain induced in the 45j plies are given
by:
s12 A45j ¼ rxx =2
c12 A45j ¼ ðexx eyy Þ

Hence for the [F45]ns stacking sequence the shear modulus is given by:
s12 rxx
G12 ¼ ¼
c12 2ðexx eyy Þ
REFERENCES
1. Whitney, J.M.; Daniel, I.M.; Pipes, R.B. Experimental Mechanics of Fiber
Reinforced Composite Materials, 2nd Ed; Bethel, CT: SEM Monograph 4 So-
ciety for Experimental Mechanics (ISBN 0-912053-01-1).
2. Carlsson, L.A.; Pipes, R.B. Experimental Characterization of Advanced Com-
posite Materials, 2nd Ed; Lancaster, PA: Technomic Pub Co. (ISBN 1-56676-
433-5), 1997.
3. Manual on Experimental Methods of Mechanical Testing of Composites, 2nd Ed;
Jenkins, C.H., Ed; Bethel, CT: Society for Experimental Mechanics ISBN 0-
88173-284-2, 1998.
4. Rosen, B.W. A simple procedure for experimental determination of the longi-
tudinal shear modulus of unidirectional composites. J Compos Mater 1972, 6,
552.

Appendix C
Tables of Beam Deflections and Slopes
Tables of typical beam deflections and slopes can be found on the following
pages.
629

Table 1 Deflection and Slopes of Cantilever Beams

Table 1 Continued
Source: Ref. 1.

Table 2 Deflections and Slopes of Simply Supported Beams

Table 2 Continued
Source: Ref. 1.

(Mechanical Engineering, 165) ) Anja Belz, Roger Evans, Paul Piwek - Structural Analysis of Polymeric Composite Materials-CRC Press, Marcel Dekker (2003) PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(Mechanical Engineering, 165) ) Anja Belz, Roger Evans, Paul Piwek - Structural Analysis of Polymeric Composite Materials-CRC Press, Marcel Dekker (2003) PDF

Uploaded by

Copyright:

Available Formats

Although great care has been taken to provide accurate and current information, neither the

Library of Congress Cataloging-in-Publication Data

This book is printed on acid-free paper.

Distribution and Customer Service

Eastern Hemisphere Distribution

World Wide Web

Copyright n 2004 by Marcel Dekker, Inc. All Rights Reserved.

Current printing (last digit):

PRINTED IN THE UNITED STATES OF AMERICA

Copyright © 2004 by Marcel Dekker, Inc. All Rights Reserved.

Copyright © 2004 by Marcel Dekker, Inc. All Rights Reserved.

This textbook is intended for use in a senior-year undergraduate or ﬁrst-year

Copyright © 2004 by Marcel Dekker, Inc. All Rights Reserved.

Copyright © 2004 by Marcel Dekker, Inc. All Rights Reserved.

Copyright © 2004 by Marcel Dekker, Inc. All Rights Reserved.

I have been fortunate to have worked with talented and inspirational

Copyright © 2004 by Marcel Dekker, Inc. All Rights Reserved.

2. Review of Force, Stress, and Strain Tensors

Copyright © 2004 by Marcel Dekker, Inc. All Rights Reserved.

4. Elastic Response of Anisotropic Materials

Copyright © 2004 by Marcel Dekker, Inc. All Rights Reserved.

6. Thermomechanical Behavior of Multiangle Composite

7. Predicting Failure of a Multiangle Composite Laminate

Copyright © 2004 by Marcel Dekker, Inc. All Rights Reserved.

9. The Governing Equations of Thin-Plate Theory

10. Some Exact Solutions for Specially Orthotropic Laminates

Copyright © 2004 by Marcel Dekker, Inc. All Rights Reserved.

11. Some Approximate Solutions for Symmetric Laminates

Copyright © 2004 by Marcel Dekker, Inc. All Rights Reserved.

A broad overview of modern composite materials is provided in this chapter.

Copyright © 2004 by Marcel Dekker, Inc. All Rights Reserved.

Copyright © 2004 by Marcel Dekker, Inc. All Rights Reserved.

Figure 1 Different types of composites based on continuous fibers.

Copyright © 2004 by Marcel Dekker, Inc. All Rights Reserved.

Copyright © 2004 by Marcel Dekker, Inc. All Rights Reserved.

2.1 Basic Concepts

Copyright © 2004 by Marcel Dekker, Inc. All Rights Reserved.

Copyright © 2004 by Marcel Dekker, Inc. All Rights Reserved.

measure of the ‘‘size’’ of the polymeric molecule is the degree of polymer-

Copyright © 2004 by Marcel Dekker, Inc. All Rights Reserved.

Figure 4 The polymer nylon 6,6.

Copyright © 2004 by Marcel Dekker, Inc. All Rights Reserved.

Copyright © 2004 by Marcel Dekker, Inc. All Rights Reserved.

Copyright © 2004 by Marcel Dekker, Inc. All Rights Reserved.

2.4 Thermoplastic vs. Thermoset Polymers

Copyright © 2004 by Marcel Dekker, Inc. All Rights Reserved.

amorphous thermoplastic. No thermoplastic is completely crystalline, how-

Copyright © 2004 by Marcel Dekker, Inc. All Rights Reserved.

Copyright © 2004 by Marcel Dekker, Inc. All Rights Reserved.

Figure 8 Effects of temperature on polymer stiffness.

Copyright © 2004 by Marcel Dekker, Inc. All Rights Reserved.

Table 1 Approximate Glass Transition Temperatures for Some Common Polymers

Typical glass transition

Silicone rubber 123 190 Rubbery

Copyright © 2004 by Marcel Dekker, Inc. All Rights Reserved.

Copyright © 2004 by Marcel Dekker, Inc. All Rights Reserved.

The mechanism responsible for these high stiﬀnesses and strengths

3.1 Glass Fibers

3.2 Aramid Fibers

Copyright © 2004 by Marcel Dekker, Inc. All Rights Reserved.

Property E-glass S-glass

Specific gravity 2.60 2.50

bonding arrangement results in the anisotropic behavior exhibited at the