Вы находитесь на странице: 1из 640

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 specic advice or recommendations for any
specic situation.
Trademark notice: Product or corporate names may be trademarks or registered trademarks and
are used only for identication 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 oers 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 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, microlming, 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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Preface
This textbook is intended for use in a senior-year undergraduate or rst-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 briey 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 eort 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 brous materials
and the manufacturing processes used to produce polymer composites and
v
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 Hookes law is developed in Chapter 4. The
three-dimensional form of Hookes 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 Hookes 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
eects. The chapter also presents methods of combining macroscopic failure
criteria with CLT to predict rst-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 eective 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 deections 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 briey 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 eects of environmentin particular,
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
the eects due to changes in temperature and/or moisture contentare
integrated throughout Chapters 3 through 11, rather than being conned 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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 inuence 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,
NASALangley 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 sta at Marcel Dekker, Inc.in
particular Michael Deters, Production Editor, and John Corrigan, Acquis-
itions Editorfor their very professional and competent help throughout
preparation of the nal 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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Contents
Preface
Acknowledgments
1. Introduction
1. Basic Denitions
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. Denition of Stress
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
5. The Stress Tensor
6. Transformation of the Stress Tensor
7. Principal Stresses
8. Plane Stress
9. Denition 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 Eects of Stress,
Temperature, and Moisture
Homework Problems
References
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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. Eective 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. Denition of a Thin Plate and Allowable Plate
Loadings
2. Plate Deformations: The Kirchho Hypothesis
3. Principal Curvatures
4. Standard Methods of Describing Composite
Laminates
5. Calculating Ply Strains and Stresses
6. Classical Lamination Theory (CLT)
7. Simplications Due to Stacking Sequence
8. Summary of CLT Calculations
9. Eective 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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
5. Computer Program LAMFAIL
Homework Problems
References
8. Composite Beams
1. Preliminary Discussion
2. Comparing Classical Lamination Theory to
Isotropic Beam Theory
3. Types of Composite Beams Considered
4. Eective Axial Rigidity of Rectangular Composite
Beams
5. Eective Flexural Rigidities of Rectangular
Composite Beams
6. Eective 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
1. Preliminary Discussion
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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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
1. Preliminary Discussion
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
E
11
, E
22
, r
12
, and G
12
C. Tables of Beam Deections and Slopes
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
1
Introduction
A broad overview of modern composite materials is provided in this chapter.
The chapter begins with a denition of what is meant by the phrase com-
posite material. Separate sections devoted to polymeric materials, brous
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 de-
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 reection, it becomes clear that this denition 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 denition 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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
ceramics satisfy this broad denition 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 denition
is a consideration of physical scale. Another denition of a composite mate-
rial, which includes a reference to a physical scale appropriate for present
purposes, is as follows:
Acomposite material is a material systemconsisting of two (or more)
materials, which are distinct at a physical scale greater than about
110
6
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-stiness reinforcing material, embed-
ded within a relatively low-strength, low-stiness 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
stiness 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 stiness/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 classied according to the size or shape of the reinforcing material
used. Four common classications 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) bers, with a length ranging from about 10 to


200 mm

Continuous bers, whose length are, in eect, innite.


Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Whiskers, short bers, and continuous bers all have very small diam-
eters relative to their length; the diameter of these products ranges fromabout
5 to 200 Am.
Distinctly dierent types of composites can be produced using any of
the above reinforcements. For example, three types of composites based on
continuous bers are shown in Fig. 1: unidirectional composites, woven
composites, and braided composites. In a unidirectional composite, all bers
are aligned in the same direction and embedded within a matrix material.
In contrast, woven composites are formed by rst weaving continuous bers
into a fabric and then embedding the fabric in a matrix. Hence, a single layer
of a woven composite contains bers 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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
nonorthogonal ber 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 bers 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 bers 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 signicant 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 ber represents
the reinforcing material in these composites. Hence, the orientation of the
bers is, in general, varied from one ply to the next, so as to provide high
stiness 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 classied 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: polymermatrix, metalmatrix, or ceramic
matrix composite materials. Usually, the reinforcing material governs the
stiness and strength of a composite. On the other hand, the matrix material
usually governs thermal stability. Polymericmatrix composites are used in
applications involving relatively modest temperatures (service temperatures
of, say, 200jC or less). Metalmatrix composites are used at temperatures up
to about 700jC, while ceramicmatrix composites are used at ultra-high
temperatures (up to about 1200jC or greater). The matrix also denes 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 bers together

To protect the bers from aggressive (chemical) environments or


mechanical damage (e.g., due to abrasion, for example)

To transfer and redistribute stresses between bers, 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-
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
sied according to the physical form of the reinforcing material (particu-
late, whisker, short ber, or continuous ber 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. Metalmatrix and ceramicmatrix 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 ber. 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 eectively in practice must understand at least the
rudiments of polymer and ber 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 brous 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 brous 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. 14).
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 10
3
to 10
7
.
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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 C
2
H
4
or CH
2
jCH
2
.
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 C
2n
H
4n
, 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
uid. 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 uids known as oligomers (a few
mers). The individual molecules within an oligomer possess a range of
molecular weights, typically containing perhaps 220 mers.
It should be clear from the above discussion that a specic molecular
weight cannot be assigned to a polymer. Rather, the molecules within a bulk
sample of a polymer are of diering 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
Figure 2 The monomer ethylene.
* As fully described in an introductory chemistry text, a covalent bond is formed when two
atoms share an electron pair, so as to ll an incompletely lled valence level.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
measure of the size of the polymeric molecule is the degree of polymer-
ization, dened 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, stiness, density, thermal expansion coecient, 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 ramications if the polymer is to be
used in structural applications. For example, if a ber(s) is embedded within
the sample during the polymerization process (as is the case for some ber-
reinforced polymeric composite systems), then shrinkage of the matrix causes
residual stresses to develop during polymerization. This eect contributes to
so-called cure stresses, which are present in most polymeric composites. As
will be seen in later chapters, cure stresses arise fromtwo primary sources. The
rst is shrinkage of the matrix during polymerization, as just described.
The second is stresses that arise due to temperature eects. In this case, the
composite is polymerized at an elevated temperature (say, 175jC) and then
cooled to room temperature (20jC). The thermal expansion coecient of the
matrix is typically much higher than the bers, and so during cooldown, the
matrix is placed in tension while the ber is placed in compression. Cure
stresses due to shrinkage during cure and/or temperature eects can be quite
high relative to the strength of the polymer itself, and ultimately contribute
toward failure of the composite.
Figure 3 The polymer polyethylene.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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: C
6
H
16
N
2
) and adipic acid
(chemical composition: CO
2
H(CH
2
)
4
CO
2
H). 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., H
2
O) 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 classied 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 diculties with residual stresses can often be minimized if
an addition polymer is used in these applications.
Figure 4 The polymer nylon 6,6.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 stiness 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
acrylonitrilebutadienestyrene, 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 stiness 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
stiness of a branched polymer will, in general, be greater than the macro-
scopical stiness 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 dierent 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 identied. A
bulk sample of a highly crosslinked polymer may be thought of as a single
molecule.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Figure 5 Types of polymer molecular structure.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 stiness. 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 stiness 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 1520jC. 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 A

. The crystalline region may extend over an


area with a length dimension ranging from (about) 1000 to 10,000 A

.
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, stiness, and density than an otherwise comparable
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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
1050% 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 stiness 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
indenite 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 uid, containing molecules with perhaps
210 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 6 An idealized representation of a crystalline region in a thermoplastic
polymer.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 lowtemperatures
(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 nal 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 nal fabrication of a thermoplastic
composite part, but no chemical reaction occurs. That is, heat is applied
Figure 7 Overall molecular structure of a semicrystalline thermoplastic,
showing crystalline and amorphous regions.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 signi-
cantly higher than those required to cure thermoset composites.
2.5 The Glass Transition Temperature
Stiness 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 eect of temperature
on the stiness 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 stiness near a characteristic temperature
called the glass transition temperature, T
g
. At temperatures well below the T
g
,
polymer stiness decreases slowly with an increase in temperature. At
temperatures well below the T
g
, polymers are glassy and brittle. In
contrast, at temperatures well above the T
g
, all polymers are rubbery
and ductile. Thus, the T
g
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 T
g
, 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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Consequently, the polymer is glassy and exhibits high stiness and strength
but relatively lowductility. Conversely, at temperatures well above the T
g
, 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 stiness and strength
but higher ductility. As implied in Fig. 8, for amorphous thermoplastics, the
change in stiness (and other physical properties) that occurs as the T
g
is
reached may be one to two orders of magnitude. This astonishing decrease in
stiness 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 T
m
in Fig. 8, although, as previously
discussed, a thermoplastic does not exhibit a unique melting temperature but
rather melts over a temperature range. Youngs 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 T
g
has been illustrated in Fig. 8 by demonstrating the change in
stiness as temperature is increased. However, many other characteristic
physical properties (density, strength, thermal expansion coecient, heat ca-
pacity, etc.) also change sharply at this transition. Hence, the T
g
can be
measured by monitoring any of these physical properties as a function of
temperature. The T
g
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
Polymer jC jF
General character
at room temperature
(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
Acrylonitrilebutadienestyrene 90 195 Rigid
Polystyrene 100 210 Rigid
Polyester 150 300 Glassy
Epoxy 175 350 Glassy
Polyetheretherketone (PEEK) 200 400 Glassy
Polyetherimide 215 420 Glassy
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Note that knowledge of the T
g
allows an immediate assessment of the general
nature of the polymer at room temperature.
3 FIBROUS MATERIALS
Reinforcing bers are the major strengthening element in all polymeric
composites. Abrief introduction to these materials is presented in this section.
The reader interested in additional details is referred to Refs. 5 and 6.
Common continuous ber materials are:

Glass

Aramid

Graphite or carbon

Polyethylene

Boron

Silicon carbide.
In all cases, the ber diameters are quite small, ranging from about 5 to
12 Am for glass, aramid, or graphite bers; from about 25 to 40 Am for
polyethylene bers; and from about 100 to 200 Am for boron and silicon
carbide bers.
Some of the terminologies used to describe bers will be dened here.
The terms ber and lament are used interchangeably. An end (also called a
strand) is a collection of a given number of bers gathered together. If the
bers are twisted, the collection of bers is called a yarn. The ends are
themselves gathered together to form a tow (also called roving). The bers are
usually coated with a size (also called a nish). The size is applied for several
reasons, such as:

To bind the bers in the strand

To lubricate the bers during fabrication

To serve as a coupling and wetting agent to insure a satisfactory


adhesive bond between the ber and matrix materials.
There may be thousands of laments in a single tow and, in fact, tow
sizes are often described in terms of thousands of laments per tow. For
example, a 6k tow implies that the tow consists of 6000 individual bers.
Properties of several types of glass bers, organic bers, carbon bers,
and silicon carbide bers will be briey described in the following subsec-
tions. It will be seen that:

Youngs modulus (stiness) ranges from about 70 GPa for glass


bers to 700 GPa (or higher) for carbon bers.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

Comparable tensile strengths can be obtained using any of the bers


listed.

Specic gravity ranges from about 0.97 for polyethylene ber to


about 2.5 for glass bers.

Elongation at failure ranges from about 0.3% for some carbon


bers to about 5% for glass bers.
The mechanism responsible for these high stinesses and strengths
diers from one ber type to another. For glass bers, the process of draw-
ing to very small diameters simply reduces the number and size of aws in
the material, thus increasing strength. For example, glass bers 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 bers, strengthening is accomplished by stretching the ber and
thereby aligning the polymer molecules. This produces bers that are them-
selves anisotropic. For carbon or graphite bers, strengthening is accom-
plished by aligning the basal planes of adjoining crystals, also producing
an anisotropic ber.
3.1 Glass Fibers
Glass bers 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 bers, which are then combined into a strand and wound onto a
spool. The two major types of glass bers are called E-glass and S-glass.
E-glass (alumino borosilicate) is so named (electrical glass) because of its
high electrical resistivity. S-glass (magnesium aluminosilicate) is so named
(structural glass) because of its high tensile strength. Glass bers are
usually isotropic. Some mechanical and physical properties typical of E-glass
and S-glass bers are listed in Table 2.
3.2 Aramid Fibers
The aramid polymer ber produced by DuPont Corp. and marketed under
the trade name Kevlarkis perhaps the most widely used organic ber. This
ber is based on poly( p-phenylene terephthalamide), which is a member of a
family of polymers called aramids. Aramid bers are formed by a con-
densation/elongation process. The resulting bers are highly anisotropic
because strong covalent bonds are formed in the ber direction, whereas
weak hydrogen bonds are formed in the transverse direction. This chemical
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
bonding arrangement results in the anisotropic behavior exhibited at the
macroscale. That is, aramid bers have very high tensile strength, stiness,
and toughness in the axial direction of the ber, but relatively low tensile
strength and stiness in the transverse direction.
Kevlar is commercially available in the following grades:

Kevlar 149: a high-performance, aerospace grade ber with the


highest modulus of all Kevlar bers

Kevlar 49: a high-performance, aerospace grade ber with the


highest strength of all Kevlar bers

Kevlar 129: a relatively inexpensive ber with a lower strength and


stiness than Kevlar 149 and 49, but with a higher percent elongation

Kevlar 29: a relatively inexpensive ber with a strength and stiness


lower than Kevlar 129, but a higher percent elongation.
Nominal mechanical and physical properties of Kevlar bers are listed
in Table 3. Of particular interest is the negative coecient of thermal
Table 3 Typical Properties of Kevlar Fibers (All Properties in Axial Direction of Fiber)
Property
Kevlar
149
Kevlar
49
Kevlar
129
Kevlar
29
Specific gravity 1.44 1.44 1.44 1.44
Youngs modulus,
GPa (Msi)
186 (27) 124 (18) 96 (13.9) 68 (9.8)
Tensile strength,
MPa (ksi)
3440 (500) 3700 (535) 3380 (490) 2930 (425)
Tensile elongation, % 2.5 2.8 3.3 3.6
Coefficient of thermal
expansion, Am/m/jC
(Ain./in./jF)
2.0 (1.1) 2.0 (1.1) 2.0 (1.1) 2.0 (1.1)
Table 2 Typical Properties of Glass Fibers
Property E-glass S-glass
Specific gravity 2.60 2.50
Youngs 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,
Am/m/jC (Ain./in./jF)
5.0 (2.8) 5.6 (3.1)
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
expansion in the ber direction. This implies that an increase in temperature
produces a decrease in the length of a Kevlar 49 ber. This behavior produced
unexpected behavior during cooldown from cure in early applications of the
ber. Further research indicates that Kevlar 49 has a high positive coecient
of thermal expansion in the transverse direction, further demonstrating the
anisotropic nature of the ber.
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
ber is above 90%, and the stiest and strongest bers have carbon contents
approaching 100%. Some eort has been made to standardize these terms by
dening graphite bers 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 Youngs modulus on the order of 345 GPa (50 Msi).


Fibers that do not satisfy all of the above conditions are called carbon
bers under this standard. However, as stated above, in practice, this de-
nition is not widely followed, and the terms graphite and carbon are
often used interchangeably.
Both graphite and carbon bers are produced by thermal decomposi-
tion of an organic (i.e., polymeric) ber 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 ber is probably the most widely used.
Details of the specic steps followed during fabrication of a ber are
proprietary and can only be described in a general manner. During the
fabrication process, the precursor is rst 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 o. This step typically involves
temperatures of approximately 700jC (1290jF), and is usually conducted
in an inert atmosphere. Finally, during the graphitization process, the bers
are subjected to a combination of high temperature and tensile elongation.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
The maximum temperature reached during this step determines, in large
part, the strength and/or stiness 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 stinesses are attained by causing the basal planes to be
aligned in the ber direction. This can be accomplished by controlled stretch-
ing of the precursor during fabrication.
Major developmental eorts have resulted in the ability to produce
bers with a wide range of stinesses and strengths. Fibers are sometimes
classied in terms of the stiness (i.e., elastic modulus). Mechanical and
physical properties typical of low-modulus, intermediate-modulus, and ultra-
high-modulus bers are listed in Table 4.
3.4 Polyethylene Fibers
A high-strength, high-modulus polyethylene ber called Spectrak was
developed at Allied Signal Technologies during the 1980s. Spectra is based
on ultra-high-molecular-weight polyethylene (UHMWPE). It has a specic
gravity of 0.97, meaning that it is the only reinforcing ber available that is
lighter than water. Spectra is available in three classications (Spectra 900,
1000, and 2000) and several grades are available in each class. Nominal
properties are listed in Table 5. The high specic strength of the ber 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 ber
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 eects under long-term
sustained loading. The melting temperature of the ber is 147jC (297jF),
Table 4 Typical Properties of Commercially Available Graphite Fibers
Property
Low
modulus
Intermediate
modulus
Ultra-high
modulus
Specific gravity 1.8 1.9 2.2
Youngs modulus,
GPa (Msi)
230 (34) 370 (53) 900 (130)
Tensile strength,
MPa (ksi)
3450 (500) 2480 (360) 3800 (550)
Elongation, % 1.1 0.5 0.4
Coefficient of thermal
expansion, Am/m/jC
(Ain./in./jF)
0.4 (0.2) 0.5 (0.3) 0.5 (0.3)
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
and hence the use of polyethylene bers is limited to relatively modest
temperatures. The thermal expansion coecients of Spectra bers 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 bers were one of the rst high-performance bers available for use in
composites. They are fabricated by depositing boron on a heated core using
the vapor deposition process. Both tungsten and carbon ber cores have
been used. Boron ber diameters range from 0.1 to 0.2 mm (0.0040.008 in),
which is an order of magnitude larger than glass, aramid, or graphite bers.
Boron bers have a Youngs 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 stiness greatly restricts the bend radius to which the ber may
be subjected. On the other hand, a large ber diameter and high modulus
of elasticity contribute to excellent compressive performance for boron-
reinforced composites.
3.6 Ceramic Fibers
The single most outstanding feature oered by ceramic bers is that they are
resistant to extremely high temperatures while still maintaining competitive
structural properties. An example is a silicon carbide ber 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 ber has a modulus of 415 GPa (60 Msi), a strength of 5865 MPa
(850 ksi), and a specic gravity of 3.0. A second example is an aluminum
boronsilica ber fabricated by the 3M Company and marketed under the
trade name Nextelk. This ber is capable of operating at temperatures up
Table 5 Nominal Properties of Spectra Fibers
Property Spectra 900 Spectra 1000 Spectra 2000
Specific gravity 0.97 0.97 0.97
Youngs modulus,
GPa (Msi)
70 (10) 105 (15) 115 (17)
Tensile strength,
MPa (ksi)
2600 (380) 3200 (465) 3400 (490)
Elongation, % 3.8 3.0 3.0
Coefficient of thermal
expansion, Am/m/jC
(Ain./in./jF)
>70 (>38) >70 (>38) >70 (>38)
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 specic gravity of
3.03.
Because ceramic bers are normally used at temperatures far in excess of
the useable temperature range of polymers, they are rarely used in polymeric
composites. Ceramic bers will not be further discussed in this book.
4 COMMERCIALLY AVAILABLE FORMS
4.1 Discontinuous Fibers
Virtually all of the continuous bers described in Sec. 3 are also available in
the form of discontinuous bers. Discontinuous bers 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
bers, if it occurs, is usually induced during the fabrication process used to
create the composite material/structure; ber alignment often mirrors the
ow direction during injection molding, for example. Discontinuous bers
are roughly classied according to length, as follows:

Milled bers are produced by grinding the continuous ber into very
short lengths. For example, milled graphite bers are available with
lengths ranging from about 0.3 to 3 mm (0.00120.12 in.), and milled
glass bers are available with lengths ranging fromabout 0.4 to 6 mm
(0.00160.24 in.).

Chopped bers (or strands) have a longer length than milled bers,
and composites produced using chopped bers usually have higher
strengths and stinesses than those produced using milled bers.
Chopped graphite bers are available with lengths ranging from
about 3 to 50 mm(0.2.0 in.), while chopped glass bers are available
with lengths ranging from about 6 to 50 mm (0.242.0 in.).
In general, the mechanical properties of a composite produced using
discontinuous bers (say, the strength or stiness) are not as good as those
that can be obtained using continuous bers. However, discontinuous bers
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 stiness is not
required.
One of the most widely used composites systems based on the use of
discontinuous bers is known generically as sheet molding compound
(SMC). In its most common form, SMC consists of chopped glass bers
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
embedded within a thermosetting polyester resin. However, other resin sys-
tems (e.g., vinyl esters or epoxies) as well as other bers (e.g., chopped graph-
ite or aramid bers) are occasionally used in SMC material systems.
4.2 Roving Spools
Most continuous ber 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 bers
contained in a single tow. For example, a specic glass ber 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 bers. Fibers purchased in this
formare usually dry and are subsequently combined with a polymer, metal,
or ceramic matrix during a subsequent manufacturing operation such as
lament winding or pultrusion.
4.3 Woven Fabrics
Most types of high-performance continuous bers can be woven to form a
fabric. Weaving is accomplished using looms specially modied for use with
high-performance bers, which are stier 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) innite 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 ll


tow (also called the weft or the woof tow). The ll 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 ll 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 rm fabric that exhibits minimum distortion (e.g., ber
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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
over three successive ll tows and then under one ll tow. In the ve-harness
satin weave, one warp tow passes over four ll tows and then under one ll
tow. Similarly, in an eight-harness satin weave, one warp towpasses over seven
ll tows and then under one ll tow.
The stiness and strength of woven fabrics are typically less than that
achieved with unidirectional bers. This decrease is due to ber waviness.
Figure 9 Some common woven fabrics used with high-performance fibers.
(From Ref. 7.)
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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
ber waviness. Upon application of a tensile load, the bers within a ply tend
to straighten, resulting in a lower stiness than would be achieved if the ply
contained straight unidirectional bers. Further, due to the weave pattern, the
bers are not allowed to straighten fully and are subjected to bending stresses,
resulting in ber failures at lower tensile loads than would otherwise be
achieved if the ply contained unidirectional bers. This eect 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 ll directions is identical. Consequently, the
strength and stiness of simple weaves are usually identical in the warp and
ll directions.
In contrast, for satin weaves, the through-thickness distribution of tow
is inherently asymmetric. Referring to Fig. 9a, for example, for the ve-
harness satin weave pattern, the warp towis primarily within the top half of
the fabric (as sketched), whereas the ll tow is primarily within the lower half.
The asymmetrical through-thickness distribution of tow causes a coupling
between in-plane loading and bending deections. That is, if a uniformtensile
load is applied to the midplane of a single layer of a satin weave fabric, the
fabric will not only stretch but will also deect out of plane (i.e., bend).
Similarly, the crossover points are not symmetrically located with respect to
either the warp or ll 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 ll tows interlaced within a plane. Weaving or stitching sev-
eral layers of a woven fabric together can produce a woven structure with
a signicant 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 directionsthe warp and ll 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 gure 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,
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Figure 10 Some common braided fabrics used with high-performance fibers.
(From Ref. 7.)
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 at 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 (brous) 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 stieners 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 ber must be embedded
within the polymeric matrix. One approach is to combine the ber and resin
during the manufacturing operation in which the nal form of the composite
structure is dened. Three manufacturing processes in which this approach is
taken are lament winding (briey described in Sec. 5.3), pultrusion (Sec. 5.4),
and resin transfer molding (Sec. 5.5).
An alternative approach is to combine the ber 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 bers have already been embedded within a polymeric matrix
when delivered, the bers 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/bers 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 ber by the resin,
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
which ultimately helps ensure good adhesion between the bers 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 (348 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 rst 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
Figure 11 (a) Method used to impregnate unidirectional tows with resin. (b) A
3-in. wide roll of prepreg tape.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 signicant disadvantage because these
factors add signicantly to the nal 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 nal 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 dierent 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 signicant 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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
impregnated with a resin) or prepreg materials. As will be discussed in later
chapters, ber angles are typically varied from one ply to the next, so as to
insure adequate stiness 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
suciently 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 ber stack. Conceptually, this may be accomplished
by pouring liquidous resin over the dry ber 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.)
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
similar device to assist the resin to wet the bers within the stack. This
technique is commonly used in the recreational boat-building industry, for
example. However, as can be imagined, it is very dicult to insure uniform
penetration of the resin and wetting of the bers through the thickness of
the stack, to insure that no air pockets remain trapped in the stack, and to
avoid distortion of the ber patterns while forcing resin into the brous
assembly. There are also potential health issues associated with continually
exposing workers to nonpolymerized resins. Hence, the technique of im-
pregnating a dry ber stack using handheld tools such as squeegees is rarely
employed in industries requiring low variability in stiness and strength and/
or high volumes, such as the aerospace or automotive industries, for exam-
ple. Alternate methods of impregnating a dry ber 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
bers 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 nal shape of the composite is dened by a rigid tool. A simple


at tool is shown in Fig. 13, but in practice, the tool is rarely at
but instead mirrors the contour(s) desired in the nal product. For
example, the surface of a tool used to produce the skin of an air-
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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, Teon-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 sti
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.
Figure 13 Typical assembly used to consolidate a polymeric composite lami-
nate using an autoclave (expanded edge view).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

An edge damis 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 lm or bag. Often this


is a relatively thick (say, 5 mm) layer of silicone rubber. Sealant tape
is used to adhere the vacuum lm to the tool surface, providing a
pressure-tight seal around the periphery of the vacuum lm.

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 lm. The vacuum port often features a quick-disconnect
tting, which allows for easy connection to a vacuum pump or line.
Following vacuumbaggingof the laminate, the assemblyis placedwithin
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.)
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
conguration (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 ow across the ply
interfaces. The laminate is then solidied 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 lament 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 ber carriage simulta-
neously moves in a controlled manner along the length of the mandrel. The
angle at which bers are placed on the mandrel surface is a function of
the mandrel diameter, rate of mandrel rotation, and translational speed of the
ber carriage. The mandrel can include domed heads to accommodate ber
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Figure 15 A carbon-epoxy pressure vessel being produced using the filament-
winding process and unidirectional prepreg. The prepreg used in this case is
basedonlarge 50k tows of Zoltek Panex
o
R
35carbon fiber. (a) Aband of prepreg is
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.)
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 ber tension provides sucient 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 ow 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
lament 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 conguration 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 dierent 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 lament 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(orcollapsible)mandrels, whichresemblegiant 3Dpuzzles.


That is, the entire mandrel can be taken apart piece by piece. The
composite structure being woundmust have at least one wall opening,
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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).

Inatable mandrels, which take on the desired shape when


pressurized and then are simply deated and removed after winding
and consolidation.

Metal or polymer liners, which are actually a modication of the


inatable mandrel concept. Liners can be described as metal or
polymer balloons and remain in the lament wound vessel after
cure. The liner does not contribute signicantly to the strength or
stiness 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 lament 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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 dened 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 ber 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 suciently 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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
polymeric composite material systems are a multidisciplinary subject, involv-
ing topics drawn frompolymer chemistry, ber 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 scientic and technological developments
that have ultimately led to the successful use of modern composite material
systems. Specically, the focus of this text is the structural analysis of
laminated, continuous-ber polymeric composite materials and structures.
Having identied the structural analysis of laminated continuous-ber
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 denition of a composite material. Specically, we have dened 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 ber diameter. Alternatively, laminated polymeric com-
posites consist of well-dened layers (called plies) of bers 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 dierent physical scales. Analyses that are framed at a physical scale
corresponding to the ber diameter (or below) are classied as microme-
chanics analyses, whereas those framed at a physical scale corresponding to
a single ply thickness (or above) are classied 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 Youngs 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 denition of various material properties of interest to the structural engineer, such as
Youngs modulus, will be reviewed and discussed in greater detail in Chap. 3.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 dierent atomic species. A structural engineer wishing to apply a solu-
tion based on continuum mechanics would simply measure Youngs
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
ber and matrix materials involved, the spacing and orientation of the bers,
the adhesion (or lack thereof ) between ber and matrix, etc. For example,
suppose that a unidirectional graphiteepoxy composite is to be produced by
combining graphite bers with a known Youngs modulus (E
f
) and an epoxy
matrix with a known Youngs modulus (E
m
). An analysis framed at a physi-
cal scale corresponding to the ber diameter (i.e., a micromechanics analy-
sis) is required to predict the Youngs modulus that will be exhibited by the
composite (E
c
) 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 bers 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 dier in dierent directions. Details of ber or matrix type,
ber spacing, ber orientation, etc., are represented in a macromechanics
analysis only indirectly, via properties dened 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 sti-
nesses based on knowledge of ber 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-ber polymeric composites.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 carbonber
reinforced thermoplastic matrices in a at braid process. Polym. Compos. 1992,
13 (1), 3846.
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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
2
Review of Force, Stress, and Strain Tensors
In this chapter, the fundamental denitions 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 specic 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 eects of body forces (such as the weight of a structure) will be ignored.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
A force is a 3-D vector. A force is dened 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 xyz coordinate system is shown in Fig. 1.
Components of F acting parallel to the xyz coordinate axes, F
x
, F
yy
, and
F
z
, respectively, are also shown in the gure. The algebraic sign of each force
component is dened 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 dierent ways of ex-
pressing force vectors in a mathematical sense. Three methods will be de-
scribed here. The rst 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 , j

, and k

, respectively, and by denition have a magnitude equal to


unity. A force vector F is written in vector notation as follows:
F F
x

i F
y

j F
z

k
1
The magnitude of the force is given by:
F j j

F
2
x
F
2
y
F
2
z
_
2
Figure 1 A force vector F acting at point P. Force components F
x
, F
y
, and F
z
acting parallel to the xyz coordinate axes, respectively, are also shown.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
A second method of dening 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 F
x
; F
y
; F
z

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 F
i
; where i x; y; or z
3
Note that a range has been explicitly specied 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 F
i
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 formof a column
array:
F
F
x
F
y
F
z
_
_
_
_
_
_
4
Indicial notation is sometimes combined with matrix notation as follows:
F F
i
f g
5
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 F
i
are known in one coordinate system (say,
the xyz coordinate system), and it is desired to express this force vector in a
second coordinate system (say, the xVyVzV 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 xV-, yV-, and zV-axesthat is, we must
calculate F
xV
, F
yV
, and F
zV
. 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-
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 xVyVzV co-
ordinate system ( F
xV
, F
yV
, and F
zV
) are related to the components in the xyz
coordinate system ( F
x
, F
y
, and F
z
) according to:
F
xV
c
xVx
F
x
c
xVy
F
y
c
xVz
F
z
6a
F
yV
c
yVx
F
x
c
yVy
F
y
c
yVz
F
z
F
zV
c
zVx
F
x
c
zVy
F
y
c
zVz
F
z
The terms c
i 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 dened from the original xyz coordinate
system to the new xVyVzV coordinate system. The algebraic sign of the angle
of rotation is dened in accordance with the right-hand rule.
Equation (6a) can be succinctly written using the summation conven-
tion as follows:
F
i V
c
i Vj
F
j
6b
Alternatively, these three equations can be written using matrix notation as:
F
xV
F
yV
F
zV
_
_
_
_
_
_

c
xVx
c
xVy
c
xVz
c
yVx
c
yVy
c
yVz
c
z
V
x
c
zVy
c
zVz
_
_
_
_
F
x
F
y
F
z
_
_
_
_
_
_
6c
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
V) is related to the unit vectors in the xyz coordinate system as follows:

iV c
xVx

i c
xVy

j c
xVz

k 7
Since i

V is a unit vector, then in accordance with Eq. (2):


c
xVx

2
c
xVy

2
c
xVz

2
1 8
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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
subscript, F
i
, and therefore a force is said to be a tensor of rank one, or equiv-
alently, a rst-order tensor. Equations (6a)(6c) is called the transformation
law for a rst-order tensor.
It is likely that the reader is already familiar with two other tensors: the
stress tensor, r
ij
, and the strain tensor, e
ij
. 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 xyz co-
ordinate system. It is desired to express this force in a new xUyUzU
coordinate system, where the xUyUzU is generated from the original xy
z coordinate system by the following two rotations (see Fig. 2):

A rotation of h about the original z-axis (which denes an inter-


mediate xVyVzV coordinate system), followed by

A rotation of b about the xV-axis (which denes the nal xUyUzU


coordinate system).
Problem. (a) Determine the direction cosines c
iUj
relating the original xyz
coordinate system to the new xUyUzU coordinate system; (b) obtain a general
expression for the force vector F in the xUyUzU coordinate system; and (c)
calculate numerical values of the force vector F in the xUyUzU coordinate
system if h=20j, b=60j, and F
x
=1000 N, F
y
=200 N, F
z
=600 N.
Solution
Part (a). One way to determine the direction cosines c
iUj
is to rotate unit
vectors. In this approach, unit vectors are rst rotated fromthe original xyz
coordinate system to the intermediate xVyVzV coordinate system, and then
from the xVyVzV system to the nal xUyUzU coordinate system.
Dene a unit vector I that is aligned with the x-axis:
I u 1

i
That is, vector I is a vector for which I
x
=1, I
y
=0, and I
z
=0. The vec-
tor I can be rotated to the intermediate xVyVzV coordinate system using
Eqs. (6a)(6c):
I
x
V c
x
V
x
I
x
c
x
V
y
I
y
c
x
V
z
I
z
I
y
V c
y
V
x
I
x
c
y
V
y
I
y
c
y
V
z
I
z
I
z
V c
z
V
x
I
x
c
z
V
y
I
y
c
z
V
z
I
z
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Figure 2 Generation of the xWyWzW coordinate system from the xyz co-
ordinate system.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
The direction cosines associated with a transformation from the xyz
coordinate system to the intermediate xVyVzV coordinate system can be
determined by inspection [see Fig. 2(a)] and are given by:
c
x
V
x
cosine angle between x
V
- and x-axes cos h
c
x
V
y
cosine angle between x
V
- and y-axes cos90
B
h sin h
c
x
V
z
cosine angle between x
V
- and z-axes cos90
B
0
c
y
V
x
cosine angle between the y
V
- and x-axes cos90
B
h
sin h
c
y
V
y
cosine angle between the y
V
- and y-axes cos h
c
y
V
z
cosine angle between the y
V
- and z-axes cos90
B
0
c
z
V
x
cosine angle between the z
V
- and x-axes cos90
B
0
c
z
V
y
cosine angle between the z
V
- and y-axes cos90
B
0
c
z
V
z
cosine angle between the z
V
- and z-axes cos0
B
1
Using these direction cosines:
I
x
V c
x
V
x
I
x
c
x
V
y
I
y
c
x
V
z
I
z
cos h1 sinh0 00 cos h
I
y
V c
y
V
x
I
x
c
y
V
y
I
y
c
y
V
z
I
z
sinh1 cos h0 00 sinh
I
z
V c
z
V
x
I
x
c
z
V
y
I
y
c
z
V
z
I
z
01 00 10 0
Therefore, in the xVyVzV coordinate system, the vector I is written as:
I cos h

i
V
sin h

j
V
Now dene 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) j

and K=(1) k

. Transforming these vectors to the xVyVzV coordinate system,


again using the direction cosines listed above, results in:
J sin h

i
V
cos h

j
V
K 1

k
V
We now rotate vectors I , J, andK from the intermediate xVyVzV coordinate
systemtothe nal xUyUzU coordinate system. The directioncosines associated
with a transformation fromthe xVyVzV coordinate systemto the nal xUyUzU
coordinate system are easily determined by inspection [see Fig. 2(b)] and are
given by:
c
xUx
V 1 c
xUy
V 0 c
x Uz
V 0
c
y Ux
V 0 c
y Uy
V cos b c
y Uz
V sin b
c
z Ux
V 0 c
z Uy
V sin b c
z Uz
V cos b
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
These direction cosines together with Eqs. (6a)(6c) can be used to rotate the
vector I from the intermediate xVyVzV coordinate system to the nal xUyUzU
coordinate system:
I
x
W c
x
W
x
V I
x
V c
x
W
y
V I
y
V c
x
W
z
V I
z
V 1cos h 0sin h 00
I
x
W cos h
I
y
W c
y
W
x
V I
x
V c
y
W
y
V I
y
V c
y
W
z
V I
z
V 0cosh cosbsinh sinb0
I
y
W cos b sin h
I
z
W c
z
W
x
V I
x
V c
z
W
y
V I
y
V c
z
W
z
V I
z
V 0cosh sinbsinh cosb0
I
z
W sin b sin h
Therefore, in the nal xUyUzU coordinate system, the vector I is written as:
I cos h

i
W
cos b sin h

j
W
sin b sin h

k
W
a
Recall that in the original xyz coordinate system, I is simply a unit vector
aligned with the original x-axis: I u (1). Therefore result (a) denes the
direction cosines associated with the angle between the original x-axis and
the nal xU-, yU-, and zU-axes. That is:
c
x
W
x
cosine angle between xU- and x-axes cos h
c
y
W
x
cosine angle between yU- and x-axes cos b sin h
c
z
W
x
cosine angle between zU- and x-axes sin b sin h
A similar procedure is used to rotate the unit vectors J and K from the
intermediatexVyVzV coordinatesystemtothenal xUyUzU coordinatesystem.
These rotations result in:
J sin h

i
W
cos b cos h

j
W
sin b cos h

k
W
b
K 0

i
W
sin b

j
W
cos b

k
W
c
Since vector J is a unit vector aligned with the original y-axis, J=(1)j

, result
(b) denes the direction cosines associated with the angle between the orig-
inal y-axis and the nal xU-, yU-, and zU-axes:
c
x
W
y
cosine angle between xU- and y-axes sin h
c
y
W
y
cosine angle between yU- and y-axes cos b cos h
c
z
W
x
cosine angle between zU- and y-axes sin b cos h
Finally, result (c) denes the direction cosines associated with the angle
between the original z-axis and the nal xU-, yU-, and zU-axes:
c
x
W
z
cosine angle between xU- and z-axes 0
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
c
y
W
z
cosine angle between yU-and z-axes sin b
c
z
W
z
cosine angle between zU-and z-axes cos b
Assembling the preceding results, the set of direction cosines relating the
original xyz coordinate system to the nal xUyUzU coordinate system can
be written as:
c
x
W
x
c
x
W
y
c
x
W
z
c
y
W
x
c
y
W
y
c
y
W
z
c
z
W
x
c
z
W
y
c
z
W
z
_
_
_
_

cos h sin h 0
cos b sin h cos b cos h sin b
sin b sin h sin b cos h cos b
_
_
_
_
Part (b). Since the direction cosines have been determined, transformation
of force vector Fcan be accomplished using any version of Eqs. (6a)(6c). For
example, using matrix notation, Eq. (6c):
F
x
W
F
y
W
F
z
W
_
_
_
_
_
_

c
x
W
x
c
x
W
y
c
x
W
z
c
y
W
x
c
y
W
y
c
y
W
z
c
z
W
x
c
z
W
y
c
z
W
z
_
_
_
_
F
x
F
y
F
z
_
_
_
_
_
_

cos h sin h 0
cos b sin h cos b cos h sin b
sin b sin h sin b cos h cos b
_
_
_
_
F
x
F
y
F
z
_
_
_
_
_
_
F
x
W
F
y
W
F
z
W
_
_
_
_
_
_

cos hF
x
sin hF
y
cos b sin hF
x
cos b cos hF
y
sin bF
z
sin b sin hF
x
sin b cos hF
y
cos bF
z
_
_
_
_
_
_
Part (c). Using the specied numerical values and the results of Part (b):
F
x
W
F
y
W
F
z
W
_
_
_
_
_
_

cos 20
B
1000N sin20
B
200N
cos 60
B
sin20
B
1000N cos 60
B
cos 20
B
200N sin60
B
600N
sin60
B
sin20
B
1000N sin60
B
cos 20
B
200N cos 60
B
600N
_
_
_
_
_
_
F
x
W
F
y
W
F
z
W
_
_
_
_
_
_

1008N
442:6N
433:4N
_
_
_
_
_
_
Using vector notation, F can now be expressed in the two dierent coordi-
nate systems as:
F 1000 N

i 200 N

j 600 N

k
or equivalently
F 1008 N

i
W
442:6 N

j
W
433:4 N

k
W
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
where , j

, k

and U, j

U, k

U are unit vectors in the xyz and xUyUzUcoordinate


systems, respectively. Force vector F drawn in the xyz and xUyU
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 xyz and xWyWzW coordinate systems.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
pendent of the coordinate system used to describe the force vector. Using
Eq. (2), the magnitude of the force vector in the xyz coordinate system is:
F j j

F
2
x
F
2
y
F
2
z
_

1000N
2
200N
2
600N
2
_
1183 N
The magnitude of the force vector in the xUyUzU coordinate system is:
F j j

F
x
W
2
F
y
W
_ _
2
F
z
W
2
_

1008N
2
442:6N
2
433:4N
2
_
1183N agrees
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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 aaaa perpendicular
to rod axis; (b) free-body diagram based on plane bbbb, inclined at angle h
to rod axis.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
of the rod. The resulting free-body diagram for the lower half of the rod is
shown in Fig. 5(a), where an xyz 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)+(0)j

+(0)k

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 dened 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 xVyVzV 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-axisthat
is, the xVyVzV coordinate system is generated from the xyz coordinate
system by a rotation of h about the original z-axis. The internal force F
I
can
be expressed with respect to the xVyVzV coordinate system by transforming
F
I
from the xyz coordinate system to the xVyVzV 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):
c
x
V
x
cos h c
x
V
y
sin h c
x
V
z
0
c
y
V
x
sin h c
y
V
y
cos h c
y
V
z
0
c
z
V
x
0 c
z
V
y
0 c
z
V
z
1
Applying Eqs. (6a)(6c), we have:
F
x
V
F
y
V
F
z
V
_
_
_
_
_
_

c
x
V
x
c
x
V
y
c
x
V
z
c
y
V
x
c
y
V
y
c
y
V
z
c
z
V
x
c
z
V
y
c
z
V
z
_
_
_
_
F
x
F
y
F
z
_
_
_
_
_
_

cos h sin h 0
sin h cos h 0
0 0 1
_
_
_
_
R
0
0
_
_
_
_
_
_

cos hR
sin hR
0
_
_
_
_
_
_
In the xVyVzV coordinate system, the internal force is F
I
= (R cos h)V(R sin
h)j

V+(0)k

V. Hence, by dening 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 specic coordinate system must be specied before a force vector


can be dened in a mathematical sense. In general, the coordinate
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
system is dened by the imaginary cut(s) used to form the free-body
diagram.

All components of a force must be specied to fully dene 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 dened 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 innitesimally 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 dened as the force per unit area acting perpendicular and tangent to the
area DA, respectively. That is,
r u lim
DA!0
N
DA
s u lim
DA!0
V
DA
9
Note that by denition, 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/m
2
. In English units,
stress is reported in terms of pounds-force per square inch (abbreviated psi),
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Figure 6 A solid 3-D body in equilibrium.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
that is, 1 psi=1 lbf/in
2
. Conversion factors between the two systems of mea-
surement are 1 psi=6895 Pa, or equivalently, 1 Pa=0.145010
3
psi. Com-
mon abbreviations used throughout this text are as follows:
1 10
3
Pa 1kilo-Pascals 1kPa 1 10
3
psi 1kilo-psi 1ksi
1 10
6
Pa 1Mega-Pascals 1 MPa 1 10
6
psi 1mega-psi 1Msi
1 10
9
Pa 1Giga-Pascals 1GPa
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 innitesimally small cube re-
moved from the body is shown in Fig. 7(b). The cube is referenced to an x
yz 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 identied with a single subscript (since
force is a rst-order tensor), for convenience, two subscripts have been used.
The rst subscript identies the face over which the force is distributed,
while the second subscript identies the direction in which the force is ori-
ented. For example, N
xx
refers to a normal force component which is
distributed over the x-face and which points in the x-direction. Similarly,
V
zy
refers to a shear force distributed over the z-face which points in the
y-direction.
Three stress components can now be dened for each cube face, in
accordance with Eq. (9). For example, for the three faces of the innitesimal
element shown in Fig. 7:
Stresses acting on the +x-face:
r
xx
lim
dy;dz!0
N
xx
dydz
_ _
s
xy
lim
dy;dz!0
V
xy
dydz
_ _
s
xz
lim
dy;dz!0
V
xz
dydz
_ _
Stresses acting on the y-face:
r
yy
lim
dx;dz!0
N
yy
dxdz
_ _
s
yx
lim
dx;dz!0
V
yx
dxdz
_ _
s
yz
lim
dx;dz!0
V
yz
dxdz
_ _
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Figure 7 Free-body diagrams used to define stress induced in a solid body.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Stresses acting on the +z-face:
r
zz
lim
dx;dy!0
N
zz
dxdy
_ _
s
zx
lim
dx;dy!0
V
zx
dxdy
_ _
s
zy
lim
dx;dy!0
V
zy
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 innitesimal 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 V
xz
and V
zx
in Fig. 7(c)] or tail-to-
tail (e.g., forces V
zy
and V
yz
) and be of equal magnitude. That is,
jV
xy
=V
yx
j, jV
xz
=V
zx
j, and jV
yz
=V
zy
j.
These restrictions reduce the number of independent force (stress) com-
ponents from 18 to 6, as follows:
Independent force components Independent stress components
N
xx
r
xx
N
yy
r
yy
N
zz
r
zz
V
xy
V
yx
s
xy
s
yx

V
xz
V
zx
s
xz
s
zx

V
yz
V
zy
s
yz
s
zy

An innitesimal element showing all stress components is shown in


Fig. 8. We must next dene the algebraic sign convention we will use to de-
scribe individual stress components. We rst associate an algebraic sign with
each face of the innitesimal 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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Having identied 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 conrm that all stress components
shown in Fig. 8 are algebraically positive. For example, to determine the
algebraic sign of the normal stress r
xx
which acts on face ABCDin Fig. 8, note
that (a) face ABCD is positive and (b) the normal stress r
xx
which acts on this
Figure 8 An infinitesimal stress element (all stress components shown in a
positive sense).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
face points in the positive y-direction. Therefore, r
xx
is positive. As a second
example, the shear stress s
yz
which acts on cube face CDEFis positive because
(a) face CDEF is negative and (b) s
yz
points in the negative z-direction.
The preceding discussion shows that the state of stress at a point is de-
ned 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:
r
xx
s
xy
s
xz
s
yx
r
yy
s
yz
s
zx
s
zy
r
zz
_
_
_
_

r
xx
s
xy
s
xz
s
xy
r
yy
s
yz
s
xz
s
yz
r
zz
_
_
_
_
10
To express the state of stress using indicial notation, we must rst make
the following change in notation:
s
xy
! r
xy
s
xz
! r
xz
s
yx
! r
yx
s
yz
! r
yz
s
zx
! r
zx
s
zy
! r
zy
With this change, the matrix on the left side of the equality sign in Eq. (10)
becomes:
r
xx
s
xy
s
xz
s
yx
r
yy
s
yz
s
zx
s
zy
r
zz
_
_
_
_
!
r
xx
r
xy
r
xz
r
xy
r
yy
r
yz
r
xz
r
yz
r
zz
_
_
_
_
which can be succinctly written using indicial notation as:
r
ij
; i; j x; y; or z 11
In Sec. 1, it was noted that a force vector is a rst-order tensor since only
one subscript is required to describe a force tensor, F
i
. FromEq. (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 xyz coordinate system and
subject to the stress components shown in Fig. 9.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Determine. Label all stress components, including algebraic sign.
Solution. The magnitude and algebraic sign of each stress component are
determined using the sign convention dened above. The procedure will be
illustrated using the stress components acting on face CDEF. First, note that
face CDEFis 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 r
yy
=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
Figure 9 Stress components acting on an infinitesimal element (all stresses in
MPa).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
labeled s
yx
=75 MPa (or equivalently, s
xy
=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 s
yz
=50
MPa (or equivalently, s
zy
=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:
r
xx
s
xy
s
xz
s
yx
r
yy
s
yz
s
zx
s
zy
r
zz
_
_
_
_

100MPa 75MPa 30MPa
75MPa 50MPa 50MPa
30MPa 50MPa 25MPa
_
_
_
_
6 TRANSFORMATION OF THE STRESS TENSOR
In Sec. 5, the stress tensor was dened using a free-body diagram of an inni-
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 xyz coordinate system.
Now, the innitesimal element need not be removed in the orientation
shown in Fig. 10(a). An innitesimal element removed from precisely the
same point within the body but at a dierent orientation is shown in Fig.
10(b). This stress element is referenced to a new xVyVzV 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 xVyVzVcoordinate
system is equivalent to the stress tensor referenced to the xyz coordinate
system, although the direction and magnitude of individual stress compo-
nents will dier.
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 xVyVzV
coordinate system (r
iVjV
) are related to the components in the original xyz
coordinate system (r
ij
) according to:
r
i
V
j
V c
i
V
k
c
j
V
l
r
kl
where i; j; k; l x; y; z 12a
or equivalently (using matrix notation):
r
i
V
j
V c
i
V
j
r
ij
c
i
V
j

T
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Figure 10 Infinitesimal elements removed from the same point from a 3-D solid
but in two different orientations.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
where [c
iVj
]
T
is the transpose of the direction cosine array. Writing in full
matrix form:
r
x
V
x
V r
x
V
y
V r
x
V
z
V
r
y
V
x
V r
y
V
y
V r
y
V
z
V
r
z
V
x
V r
z
V
y
V r
z
V
z
V
_
_
_
_

c
x
V
x
c
x
V
y
c
x
V
z
c
y
V
x
c
y
V
y
c
y
V
z
c
z
V
x
c
z
V
y
c
z
V
z
_
_
_
_
r
xx
r
xy
r
xz
r
yx
r
yy
r
yz
r
zx
r
zy
r
zz
_
_
_
_

c
x
V
x
c
y
V
x
c
z
V
x
c
x
V
y
c
y
V
y
c
z
V
y
c
x
V
z
c
y
V
z
c
z
V
z
_
_
_
_
12b
As discussed in Sec. 2, the terms c
iVj
which appear in Eqs. (12a) and (12b)
are direction cosines and equal the cosine of the angle between the axes of the
xyz and xVyVzV coordinate systems. Recall that the algebraic sign of an
angle of rotation is dened in accordance with the right-hand rule and that
angles are dened from the xyz coordinate system to xVyVzV 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 rst specied 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 r
xVzV
and the stress
components in the xyz coordinate system in expanded form. We rst
specify that iV=xV and jV=zV, and Eq. (12a) becomes:
r
x
V
z
V c
x
V
k
c
z
V
l
r
kl
where k; l 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:
r
x
V
z
V c
x
V
x
c
z
V
x
r
xx
c
x
V
x
c
z
V
y
r
xy
c
x
V
x
c
z
V
z
r
xz
c
x
V
y
c
z
V
x
r
yx
c
x
V
y
c
z
V
y
r
yy
c
x
V
y
c
z
V
z
r
yz
c
x
V
z
c
z
V
x
r
zx
c
x
V
z
c
z
V
y
r
zy
c
x
V
z
c
z
V
z
r
zz
13
Equations (12a) and (12b) show that the value of any individual stress
component r
iVjV
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-
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
variants. For a second-order tensor, three independent stress invariants exist
and are dened as follows:
First stress invariant H r
ii
14a
Second stress invariant U
1
2
r
ii
r
jj
r
ij
r
ij
_ _
14b
Third stress invariant W
1
6
r
ii
r
jj
r
kk
3r
ii
r
jk
r
jk
2r
ij
r
jk
r
ki
_ _
14c
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 H r
xx
r
yy
r
zz
15a
Second stress invariant U r
xx
r
yy
r
xx
r
zz
r
yy
r
zz
r
2
xy
r
2
xz
r
2
yz
_ _
15b
Third stress invariant W r
xx
r
yy
r
zz
r
xx
r
2
yz
r
yy
r
2
xz
r
zz
r
2
xy
2r
xy
r
xz
r
yz
15c
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 xyz coordinate is known to be:
r
xx
r
xy
r
xz
r
yx
r
yy
r
yz
r
zx
r
zy
r
zz
_
_
_
_

50 10 15
10 25 30
15 30 5
_
_
_
_
ksi
It is desired to express this state of stress in an xUyUzU coordinate system,
generated by the following two sequential rotations:
(i) Rotation of h=20j about the original z-axis (which denes an
intermediate xVyVzV coordinate system), followed by
(ii) Rotation of b=35j about the xV-axis (which denes the nal xUyU
zU coordinate system).
Problem. (a) Rotate the stress tensor to the xUyUzU coordinate system and
(b) calculate the rst, second, and third invariants of the stress tensor using
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
both elements of the stress tensor referenced to the xyz coordinate system,
r
ij
, and elements of the stress tensor referenced to the xUyUzU coordinate
system, r
iUjU
.
Solution
Part (a). General expressions for direction cosines relating the xyz and
xUyUzU coordinate systems were determined as a part of Example Problem 1.
The direction cosines were found to be:
c
x
W
x
cos h
c
x
W
y
sin h
c
x
W
z
0
c
y
W
x
cos b sin h
c
y
W
y
cos b cos h
c
y
W
z
sin b
c
z
W
x
sin b sin h
c
z
W
y
sin b cos h
c
z
W
z
cos b
Since in this problem h=20j and b=35j, the numerical values of the direc-
tion cosines are:
c
x
W
x
cos 20
B
0:9397
c
x
W
y
sin 20
B
0:3420
c
x
W
z
0
c
y
W
x
cos 35
B
sin 20
B
0:2802
c
y
W
y
cos 35
B
cos 20
B
0:7698
c
y
W
z
sin 35
B
0:5736
c
z
W
x
sin 35
B
sin 20
B
0:1962
c
z
W
y
sin 35
B
cos 20
B
0:5390
c
z
W
z
cos 35
B
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 r
xUzU
can be found using Eq. (13):
r
x
W
z
W c
x
W
x
c
z
W
x
r
xx
c
x
W
x
c
z
W
y
r
xy
c
x
W
x
c
z
W
z
r
xz
c
x
W
y
c
z
W
x
r
yx
c
x
W
y
c
z
W
y
r
yy
c
x
W
y
c
z
W
z
r
yz
c
x
W
z
c
z
W
x
r
zx
c
x
W
z
c
z
W
y
r
zy
c
x
W
z
c
z
W
z
r
zz
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
r
x
W
z
W 0:93970:196250ksi 0:93970:539010ksi
0:93970:819215ksi 0:34200:196210ksi
0:34200:539025ksi 0:34200:819230ksi
00:196215ksi00:539030ksi00:81925ksi
r
x
W
z
W 28:95ksi
Alternatively, if matrix notation is used, then Eq. (12b) becomes:
r
x
W
x
W r
x
W
y
W r
x
W
z
W
r
y
W
x
W r
y
W
y
W r
y
W
z
W
r
z
W
x
W r
z
W
y
W r
z
W
z
W
_
_
_
_

0:9397 0:3420 0
0:2802 0:7698 0:5736
0:1962 0:5390 0:8192
_
_
_
_
50 10 15
10 25 30
15 30 5
_
_
_
_

0:9397 0:2802 0:1962


0:3420 0:7698 0:5390
0 0:5736 0:8192
_
_
_
_
Completing the matrix multiplication indicated yields the following:
r
x
W
x
W r
x
W
y
W r
x
W
z
W
r
y
W
x
W r
y
W
y
W r
y
W
z
W
r
z
W
x
W r
z
W
y
W r
z
W
z
W
_
_
_
_

40:65 1:113 28:95
1:113 43:08 10:60
28:95 10:60 13:72
_
_
_
_
ksi
Notice that the value of r
xUzU
determined through matrix multiplication is
identical to that obtained using indicial notation, as previously described. The
stress element is shown in the original and nal coordinate systems in Fig. 11.
Part (b). The rst, second, and third stress invariants will nowbe calculated
using components of both r
ij
and r
iUjU
. It is expected that identical values will
be obtained since the stress invariants are independent of the coordinate
system.
First stress invariant:
xyz coordinate system:
H r
ii
r
xx
r
yy
r
zz
H 50 25 5 ksi
H 70ksi
xUyUzU coordinate system:
H r
i
W
i
W r
x
W
x
W r
y
W
y
W r
z
W
z
W
H 40:65 43:08 13:72
H 70ksi
As expected, the rst stress invariant is independent of the coordinate
system.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Second stress invariant:
xyz coordinate system:
U
1
2
r
ii
r
jj
r
ij
r
ij
_ _
r
xx
r
yy
r
xx
r
zz
r
yy
r
zz
r
2
xy
r
2
xz
r
2
yz
_ _
U
_
5025 505 255 10
2
15
2
30
2

_
ksi
2
U 350ksi
2
xUyUzU coordinate system:
U
1
2
r
i
W
i
Wr
j
W
j
W r
i
W
j
Wr
i
W
j
W
_ _
U r
x
W
x
Wr
y
W
y
W r
x
W
x
Wr
z
W
z
W r
y
W
y
Wr
z
W
z
W r
2
x
W
y
W
r
2
x
W
z
W
r
2
y
W
z
W
_ _
Figure 11 Stress tensor of Example Problem 3 referenced to two different
coordinates (magnitude of all stress components in ksi).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
U
_
40:6543:08 40:6513:72 43:0813:72
1:113
2
28:95
2
10:60
2

_
ksi
2
U 350ksi
2
As expected, the second stress invariant is independent of the coordinate
system.
Third stress invariant:
xyz coordinate system:
W
1
6
r
ii
r
jj
r
kk
3r
ii
r
jk
r
jk
2r
ij
r
jk
r
ki
_ _
W r
xx
r
yy
r
zz
r
xx
r
2
yz
r
yy
r
2
xz
r
zz
r
2
xy
2r
xy
r
xz
r
yz
W 50255 5030
2
2515
2
510
2
2101530 ksi
3
W 65375ksi
3
xUyUzU coordinate system:
W
1
6
r
i
W
i
Wr
j
W
j
Wr
k
W
k
W 3r
i
W
i
Wr
j
W
k
Wr
j
W
k
W 2r
i
W
j
Wr
j
W
k
Wr
k
W
i
W
_ _
W r
x
W
x
Wr
y
W
y
Wr
z
W
z
W r
x
W
x
Wr
2
y
W
z
W
r
y
W
y
Wr
2
x
W
z
W
r
z
W
z
Wr
2
x
W
y
W
2r
x
W
y
Wr
x
W
z
Wr
y
W
z
W
W 40:6543:0813:72 40:6510:60
2
43:08
28:95
2
13:721:113
2
21:11328:95
10:60 ksi
3
W 65375ksi
3
As expected, the third stress invariant is independent of the coordinate
system.
7 PRINCIPAL STRESSES
The denition 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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
that exist in this coordinate system are called the principal stresses. In most
textbooks, principal stresses are denoted r
1
, r
2
, and r
3
. 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 p
1
, p
2
, and p
3
axes, and the principal stresses will be denoted r
p1
, r
p2
, and r
p3
.
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 xyz
coordinate systemusing the free-body diagramshown 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 r
pn
denes one axis of the principal stress coordinate system. The
direction cosines between this principal axis and the x-, y-, and z- axes are c
pnx
,
c
pny
, and c
pnz
, respectively. The surface area of triangle ABC is denoted A
ABC
.
The normal force acting over triangle ABC therefore equals (r
pn
A
ABC
). The
components of the normal force acting in the x-, y-, and z-directions equal
c
pnx
r
pn
A
ABC
, c
pny
r
pn
A
ABC
, and c
pnz
r
pn
A
ABC
, respectively.
The area of the other triangular faces are given by:
Area of triangle ABD=c
pnx
A
ABC
.
Area of triangle ACD=c
pny
A
ABC
.
Area of triangle BCD=c
pnz
A
ABC
.
Summing forces in the x-direction and equating to zero, we obtain:
c
p
n
x
r
p
n
A
ABC
r
xx
c
p
n
x
A
ABC
s
xy
c
p
n
y
A
ABC
s
xz
c
p
n
z
A
ABC
0
which can be reduced and simplied to:
r
p
n
r
xx
c
p
n
x
s
xy
c
p
n
y
s
xz
c
p
n
z
0 16a
Similarly, summing forces in the y- and z-directions results in:
s
xy
c
p
n
x
r
p
n
r
yy
c
p
n
y
s
yz
c
p
n
z
0 16b
s
xz
c
p
n
x
s
yz
c
p
n
y
r
p
n
r
zz
c
p
n
z
0 16c
Equations (16a)(16c) represent three linear homogeneous equations
which must be satised simultaneously. Since direction cosines c
p
n
x
, c
p
n
y
, and
c
p
n
z
must also satisfy Eq. (8), and therefore cannot all equal zero, the solu-
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
tion can be obtained by requiring that the determinant of the coecients
of c
p
n
x
, c
p
n
y
, and c
p
n
z
equal zero:
r
p
n
r
xx
s
xy
s
xz
s
xy
r
p
n
r
yy
s
yz
s
xz
s
yz
r
p
n
r
zz

0
Equating the determinant to zero results in the following cubic
equation:
r
3
p
n
Hr
2
p
n
Ur
p
n
W 0 17
where H, U, and W are the rst, second, and third stress invariants, respec-
tively, and have been previously listed as Eqs. (14a)(14c) and (15a)(15c).
Figure 12 Free-body diagram used to relate stress components in the xyz
coordinate system to a principal stress.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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.
Dene:
a
1
3
3U H
2
_ _
b
1
27
2H
3
9HU 27W
_ _
A

b
2

b
2
4

a
3
27
_
3

b
2

b
2
4

a
3
27
_
3

The three principal stresses [i.e., the three roots of Eq. (17)] are then given by:
r
p
1
; r
p
2
; r
p
3

A B
H
3

A B
2

3
p
A B
2
_ _

H
3

A B
2

3
p
A B
2
_ _

H
3
_

_
18
By convention, the principal stresses are numbered such that r
p1
is the
algebraically greatest principal stress, whereas r
p3
is the algebraically least.
That is, r
p1
>r
p2
>r
p3
.
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 nd 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 nding the cube root of a complex
number. The need to nd 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 nd 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 dene 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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
used as a third independent equation involving the three unknown constants
c
pnx
, c
pny
, and c
pnz
.
The process of nding principal stresses and direction cosines will be
demonstrated in the following example problem.
Example Problem 4
Given. A state of stress referenced to an xyz coordinate is known to be:
r
xx
r
xy
r
xz
r
yx
r
yy
r
yz
r
zx
r
zy
r
zz
_
_
_
_

50 10 15
10 25 30
15 30 5
_
_
_
_
ksi
Problem. Find (a) the principal stresses and (b) the direction cosines that
dene the principal stress coordinate system.
Solution. This is the same stress tensor considered in Example Problem3. As
a part of that problem, the rst, 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:
r
3
70r
2
350r 65375 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:
a
1
3
3U H
2
_ _

1
3
3350 70
2
_ _
1983
b
1
27
2H
3
9HU 27W
_ _

1
27
270
3
970350 2765375
_ _
31801
A

b
2

b
2
4

a
3
27
_
3

48134
2

48134
2
4

1983
3
27

15900 i6010
3
_
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
We nd that constant A equals the cube root of the complex number:
z 15900 i6010
Following the process described in Appendix A, the modulus and argument
of this complex number are:
r

a
2
b
2
_

15900
2
6010
2
_
16998
u tan
1
6010
15900
_ _
tan
1
0:37799 0:3614 rador2:780 rad
Since in this case:
a 15900 < 0
b 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 nd:
A

a ib
3
p
exp
lnr
3
_ _
cos
u
3
_ _
i sin
u
3
_ _ _ _
A

15900 i6010
3
_
exp
ln16998
3
_ _
cos
2:780
3
_ _
i sin
2:780
3
_ _ _ _
A 25:71 0:6005 i0:7996 f g
A 15:44 i20:56
Following an identical procedure, constant B is found to be:
B 15:44 i20:56
We now apply Eq. (18) to nd:
r
p
1
; r
p
2
; r
p
3

A B
H
3

A B
2

3
p
A B
2
_ _

H
3

A B
2

3
p
A B
2
_ _

H
3

54:21 ksi
27:72 ksi
43:51 ksi
_

_
_

_
Hence, r
p1
=54.21 ksi, r
p2
=43.51 ksi, and r
p3
=27.72 ksi.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Part (b). Determining the direction cosines. The rst two of Eqs. (16a)
(16c) and Eq. (8) will be used to form three independent equations in three
unknowns. We have:
r
p
n
r
xx
c
p
n
x
s
xy
c
p
n
y
s
xz
c
p
n
z
0
s
xy
c
p
n
x
r
p
n
r
yy
c
p
n
y
s
yz
c
p
n
z
0
c
p
n
x

2
c
p
n
y

2
c
p
n
z

2
1
Direction cosines for r
p1
. The three independent equations become:
54:21 50c
p
1
x
10c
p
1
y
15c
p
1
z
0
10c
p
1
x
54:21 25c
p
1
y
30c
p
1
z
0
c
p
1
x

2
c
p
1
y

2
c
p
1
z

2
1
Solving simultaneously, we obtain:
c
p
1
x
0:9726 c
p
1
y
0:1666 c
p
1
z
0:1620
Direction cosines for r
p2
. The three independent equations become:
43:51 50c
p
2
x
10c
p
2
y
15c
p
2
z
0
10c
p
2
x
43:51 25c
p
2
y
30c
p
2
z
0
c
p
2
x

2
c
p
2
y

2
c
p
2
z

2
1
Solving simultaneously, we obtain:
c
p
2
x
0:05466 c
p
2
y
0:8416 c
p
2
z
0:5738
Direction cosines for r
p3
. The three independent equations become:
27:72 50c
p
3
x
10c
p
3
y
15c
p
3
z
0
10c
p
3
x
27:72 25c
p
3
y
30c
p
3
z
0
c
p
3
x

2
c
p
3
y

2
c
p
3
z

2
1
Solving simultaneously, we obtain:
c
p
3
x
0:8276 c
p
3
y
0:2259 c
p
3
z
0:5138
8 PLANE STRESS
A stress tensor is dened 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 r
zz
=s
xz
=s
yz
=0, as
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
shown in Fig. 13(a). Since the three remaining nonzero stress components
(r
xx
, r
yy
, and s
xy
) all lie within the xy 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. Specically, 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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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
innitesimal 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 xy plane (i.e., r
zz
=s
xz
=s
yz
=0). This allows the
remaining components of stress to be written in the form of a column array
rather than a 3 3 array:
r
xx
s
xy
0
s
xy
r
yy
0
0 0 0
_

_
_

_ !
r
xx
r
yy
s
xy
_

_
_

_
Note that when a plane stress state is described, stress appears to be a rst-
order tensor since (apparently) only three components of stress (r
xx
, r
yy
,
ands
xy
) need be specied 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 specied in order to dene a state of
stress. When we invoke the plane stress assumption, we have simply assumed
a priori that three stress components (r
zz
, s
xz
, and s
yz
) 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:
r
x
V
x
V r
x
V
y
V r
x
V
z
V
r
y
V
x
V r
y
V
y
V r
y
V
z
V
r
z
V
x
V r
z
V
y
V r
z
V
z
V
_

_
_

_
c
x
V
x
c
x
V
y
c
x
V
z
c
y
V
x
c
y
V
y
c
y
V
z
c
z
V
x
c
z
V
y
c
z
V
z
_

_
_

_
r
xx
r
xy
r
xz
r
yx
r
yy
r
yz
r
zx
r
zy
r
zz
_

_
_

c
x
V
x
c
y
V
x
c
z
V
x
c
x
V
y
c
y
V
y
c
z
V
y
c
x
V
z
c
y
V
z
c
z
V
z
_

_
_

_repeated 12b
When transformation of a plane stress tensor is considered, it will be
assumed that the xVyVzV coordinate system is generated from the xyz
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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:
c
x
V
x
cos h
c
x
V
y
cos 90
B
h sin h
c
x
V
z
cos 90
B
0
c
y
V
x
cos 90
B
h sin h
c
y
V
y
cos h
c
y
V
z
cos 90
B
0
c
z
V
x
cos 90
B
0
c
z
V
y
cos 90
B
0
c
z
V
z
cos 0
B
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 r
zz
=
s
xz
=s
yz
=0 by assumption, then Eq. (12b) becomes:
r
x
V
x
V s
x
V
y
V s
x
V
z
V
s
y
V
x
V r
y
V
y
V s
y
V
z
V
s
z
V
x
V s
z
V
y
V r
z
V
z
V
_

_
_

_
cos h sin h 0
sin h cos h 0
0 0 1
_

_
_

_
r
xx
s
xy
0
s
yx
r
yy
0
0 0 0
_

_
_

cos h sin h 0
sin h cos h 0
0 0 1
_

_
_

_
Completing the matrix multiplication indicated results in:
r
x
V
x
V s
x
V
y
V s
x
V
z
V
s
y
V
x
V r
y
V
y
V s
y
V
z
V
s
z
V
x
V s
z
V
y
V r
z
V
z
V
_
_
_
_

cos
2
hr
xx
sin
2
hr
yy
2coshsinhs
xy
coshsinhr
xx
coshsinhr
yy
cos
2
h sin
2
hs
xy
0
coshsinhr
xx
coshsinhr
yy
cos
2
h sin
2
hs
xy
sin
2
hr
xx
cos
2
hr
yy
2coshsinhs
xy
0
0 0 0
_

_
_

_
As would be expected, the out-of-plane stresses are zero: r
zVzV
=s
xVzV
=s
yVzV
=0.
The remaining stress components are:
r
x
V
x
V cos
2
hr
xx
sin
2
hr
yy
2cos hsin hs
xy
r
y
V
y
V sin
2
hr
xx
cos
2
hr
yy
2cos hsin hs
xy
19
s
x
V
y
V cos hsin hr
xx
cos hsin hr
yy
cos
2
h sin
2
hs
xy
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Figure 14 Transformation of a plane stress element from one coordinate sys-
tem to another.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Equation (19) can be written using matrix notation as:
r
x
V
x
V
r
y
V
y
V
s
x
V
y
V
_

_
_

cos
2
h sin
2
h 2cosh sinh
sin
2
h cos
2
h 2coshsinh
coshsinh cos hsinh cos
2
h sin
2
h
_

_
_

r
xx
r
yy
s
xy
_

_
_

_
20
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]:
T
cos
2
h sin
2
h 2coshsinh
sin
2
h cos
2
h 2coshsinh
coshsinh coshsinh cos
2
h sin
2
h
_

_
_

_
21
The stress invariants [given by Eqs. (14a)(14c) or Eqs. (15a)(15c)]
are considerably simplied in the case of plane stress. Since by denition
r
zz
=s
xz
=s
yz
=0, the stress invariants become:
First stress invariant H r
xx
r
yy
Second stress invariant U r
xx
r
yy
s
2
xy
Third stress invariant W 0
22
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):
r
3
Hr
2
Ur 0 23
Obviously, one root of Eq. (23) is r=0. This root corresponds to r
zz
and for present purposes will be labeled r
p3
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 r
zz
=r
p3
=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
xy plane.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Removing the known root from Eq. (23), we have the following qua-
dratic equation:
r
2
Hr U 0 24
The two roots of this quadratic equation (that is, the two remaining princi-
pal stresses, r
p1
and r
p2
) may be found by application of the standard ap-
proach (3) and are given by:
r
p
1
; r
p
2

1
2
HF

H
2
4U
_ _ _
25
Substituting Eq. (22) into Eq. (25) and simplifying yields the following:
r
p
1
; r
p
2

r
xx
r
yy
2
F

r
xx
r
yy
2
_ _
2
s
2
xy
_
26
The angle h
p
between the x-axis and either the p
1
or p
2
axis is given by:
h
p

1
2
arctan
2s
xy
r
xx
r
yy
_ _
27
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):
r
xx
70 MPa
r
yy
15 MPa
s
xy
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 25
B
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Figure 15 Plane stress elements associated with Example Problem 5.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Equation (20) becomes:
r
x
V
x
V
r
y
V
y
V
s
x
V
y
V
_

_
_

cos
2
25
B
sin
2
25
B
2cos25
B
sin25
B

sin
2
25
B
cos
2
25
B
2cos25
B
sin25
B

cos25
B
sin25
B
cos25
B
sin25
B
cos
2
25
B
sin
2
25
B

_
_

_
70
15
50
_
_
_
_
_
_
r
x
V
x
V
r
y
V
y
V
s
x
V
y
V
_

_
_

0:8214 0:1786 0:7660


0:1786 0:8214 0:7660
0:3830 0:3830 0:6428
_

_
_

_
70
15
50
_

_
_

98:5
13:5
11:1
_

_
_

_
MPa
The rotated stress element is shown in Fig. 15(b).
Part (b). The principal stresses are found through application of Eq. (26):
r
p
1
; r
p
2

70 15
2
F

70 15
2
_ _
2
50
2

42:5 F 57:1MPa
r
p
1
99:6MPa
r
p
2
14:6MPa
The orientation of the principal stress coordinate system is given by Eq. (27):
h
p

1
2
arctan
250
70 15
_ _
31
B
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 eects. 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
p
2
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 dened at
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 innitesimally small.
In the present context, innitesimally small implies that lengths AB and
CB are small enough such that the deformed element may be treated as a
parallelogram.
Normal strain e
xx
is dened as the change in length of AB divided by the
original length of AB:
e
xx

DAB
AB
28
The change in length AB is given by:
DAB AVB AB
From the gure, it can be seen that the projection of length AVB in the x-
direction, that is, length AUB, is given by:
AWB AVBcosBAVBA 29
Figure 16 2-D element used to illustrate normal and shear strains (deforma-
tions shown are greatly exaggerated for clarity).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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:
sinBAVBAcBAVBA tanBAVBAcBAVBA cosBAVBAc1
Based on the small-angle approximation, Eq. (29) implies that AUBc
AVB, and therefore the change in length of AB is approximately given by:
DABcAUB AB AUA
Equation (28) can now be written as:
e
xx

AUA
AB
30
In an entirely analogous manner, normal strain e
yy
is dened as the
change in length of CB divided by the original length of CB:
e
yy

DCB
CB
Based on the small-angle approximation, the change in length of CB
is approximately given by:
DCB CVB CBcCUC
and therefore:
e
yy

CUC
CB
31
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
p
2
radians (i.e.,
initially BAVBC=90j). Engineering shear strain is dened as the change in
angle BABC, expressed in radians:
c
xy
DBABC BAVBA BCVBC 32
The subscripts associated with a shear strain [e.g., subscripts xy in Eq. (32)]
indicate that the shear strain represents the change in angle dened 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%.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 dierent denition of shear strain. Specically, tensoral shear
strain is dened as:
e
xy

1
2
c
xy
33
Since engineering shear strain has been dened 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 simplies 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 nal 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 10
6
meter=meter 1micrometer=meter 1 Am=m 1 Ain=in
1 10
6
radians 1 microradians 1 Arad
We must next dene 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 dene the algebraic sign of a shear strain, we rst identify the
algebraic sign of each face of the innitesimal strain element (the algebraic
sign of face was dened 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 conrm that all strains
shown in Fig. 16 are algebraically positive.
Example Problem 6
Given. The following two sets of strain components:
Set1 :
e
xx
1000 Am=m
e
yy
500 Am=m
c
xy
1500 Arad
Set2 :
e
xx
1000 Am=m
e
yy
500 Am=m
c
xy
1500 Arad
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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
dierence between the two sets of strain components is that in Set 1, c
xy
is
algebraically positive, whereas in Set 2, c
xy
is algebraically negative.
10 THE STRAIN TENSOR
A general 3-D solid body is shown in Fig. 18(a). An innitesimally small
cube isolated from an interior region of the body is shown in Fig. 18(b). The
cube is referenced to an xyz 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 dened 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 innitesimal 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 17 Strain elements associated with Example Problem 6 (not to scale).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Figure 18 Infinitesimal elements used to illustrate the strain tensor.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
symmetric second-order tensor, involving six components of strain: three
normal strains (e
xx
, e
yy
, e
zz
) and three tensoral shear strains (e
xy
, e
xz
, e
yz
). These
six strain components are dened in the same manner as those discussed in
the preceding section. Normal strains e
xx
, e
yy
, and e
zz
represent the change
in length in the x-, y-, and z-directions, respectively. Tensoral shear strains
e
xy
, e
xz
, and e
yz
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:
e
xx
e
xy
e
xz
e
yx
e
yy
e
yz
e
zx
e
zy
e
zz
_
_
_
_

e
xx
e
xy
e
xz
e
xy
e
yy
e
yz
e
xz
e
yz
e
zz
_
_
_
_
34
Alternatively, the strain tensor can be succinctly written using indicial nota-
tion as:
e
ij
; i; j x; y; or z 35
Note that if engineering shear strain is used, then Eq. (34) becomes
e
xx
e
xy
e
xz
e
xy
e
yy
e
yz
e
xz
e
yz
e
zz
_

_
_

_
e
xx
c
xy
=2 c
xz
=2
c
xy
=2 e
yy
c
yz
=2
c
xx
=2 c
yz
=2 e
zz
_

_
_

_
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 o-
diagonal positions.
In Sec. 1, it was noted that a force vector is a rst-order tensor since
only one subscript is required to describe a force tensor, F
i
. 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 xVyVzV coordinate
system (e
iVjV
) are related to the components in xyz coordinate system (e
ij
)
according to:
e
i
V
j
V c
i
V
k
c
j
V
l
e
kl
where k; l x; y; z 36a
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Alternatively, using matrix notation, the strain tensor transforms accord-
ing to:
e
i
V
j
V c
i
V
j
e
ij
c
i
V
j

T
which expands as follows:
e
x
V
x
V e
x
V
y
V e
x
V
z
V
e
y
V
x
V e
y
V
y
V e
y
V
z
V
e
z
V
x
V e
z
V
y
V e
z
V
z
V
_

_
_

_
c
x
V
x
c
x
V
y
c
x
V
z
c
y
V
x
c
y
V
y
c
y
V
z
c
z
V
x
c
z
V
y
c
z
V
z
_

_
_

_
e
xx
e
xy
e
xz
e
yx
e
yy
e
yz
e
zx
e
zy
e
zz
_

_
_

c
x
V
x
c
y
V
x
c
z
V
x
c
x
V
y
c
y
V
y
c
z
V
y
c
x
V
z
c
y
V
z
c
z
V
z
_

_
_

_ 36b
The terms c
i Vj
which appear in Eqs. (36a) and (36b) are direction cosines
and equal the cosine of the angle between the axes of the xVyVzV and xyz
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 dened as follows:
First strain invariant H
e
e
ii
37a
Second strain invariant U
e

1
2
e
ii
e
jj
e
ij
e
ij
_ _
37b
Third strain invariant W
e

1
6
e
ii
e
jj
e
kk
3e
ii
e
jk
e
jk
2e
ij
e
jk
e
ki
_ _
37c
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 H
e
e
xx
e
yy
e
zz
38a
Second strain invariant U
e
e
xx
e
yy
e
xx
e
zz
e
yy
e
zz
e
2
xy
e
2
xz
e
2
yz
_ _
38b
Third strain invariant W
e
e
xx
e
yy
e
zz
e
xx
e
2
yz
e
yy
e
2
xz
e
zz
e
2
xy
2e
xy
e
xz
e
yz
38c
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Example Problem 7
Given. A state of strain referenced to an xyz coordinate is known to be:
e
xx
e
xy
e
xz
e
yx
e
yy
e
yz
e
zx
e
zy
e
zz
_

_
_

_
1000Am=m 500Arad 250Arad
500Arad 1500Am=m 750Arad
250Arad 750Arad 2000Am=m
_

_
_

_
It is desired to express this state of strain in an xUyUzU coordinate system,
generated by:
(i) Rotation of h=20j about the original z-axis (which denes an
intermediate xVyVzV coordinate system), followed by
(ii) Rotation of b=35j about the xV-axis (which denes the nal xU
yUzU 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 xUyUzU coordinate system and
(b) Calculate the rst, second, and third invariants of the strain tensor
using both elements of the strain tensor referenced to the xyz
coordinate system, e
ij
, and elements of the strain tensor referenced
to the xUyUzU coordinate system, e
i Uj U
.
Solution
Part (a). General expressions for direction cosines relating the xyz and
xUyUzU 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:
c
xWx
cos 20
B
0:9397
c
xWy
sin 20
B
0:3420
c
xWz
0
c
yWx
cos 35
B
sin 20
B
0:2802
c
yWy
cos 35
B
cos 20
B
0:7698
c
yWz
sin 35
B
0:5736
c
zWx
sin 35
B
sin 20
B
0:1962
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
c
zWy
sin 35
B
cos 20
B
0:5390
c
zWz
cos 35
B
0:8192
Each component of the transformed strain tensor can now be found through
application of Eq. (36a) or Eq. (36b). For example, setting i V=xU, j V=xU, and
expanding Eq. (36a), strain component e
xUxU
is given by:
e
xUxU
c
xUx
c
xUx
e
xx
c
xUx
c
xUy
e
xy
c
xUx
c
xUz
e
xz
c
xUy
c
xUx
e
yx
c
xUy
c
xUy
e
yy
c
xUy
c
xUz
e
yz
c
xUz
c
xUx
e
zx
c
xUz
c
xUy
e
zy
c
xUz
c
xUz
e
zz
e
xUxU
0:93970:93971000 0:93970:3420500
0:93970250 0:34200:9397500
0:34200:34201500 0:34200750
00:9397250 00:3420750
002000
e
xUxU
1380 lm=m
Alternatively, if matrix notation is used, then Eq. (36b) becomes:
e
xUxU
e
xUyU
e
xUzU
e
yUxU
e
yUyU
e
yUzU
e
zUxU
e
zUyU
e
zUzU
_

_
_

_
0:9397 0:3420 0
0:2802 0:7698 0:5736
0:1962 0:5390 0:8192
_

_
_

1000 500 250


500 1500 750
250 750 2000
_

_
_

0:9397 0:2802 0:1962


0:3420 0:7698 0:5390
0 0:5736 0:8192
_

_
_

_
Completing the matrix multiplication indicated, there results:
e
xUxU
e
xUyU
e
xUzU
e
yUxU
e
yUyU
e
yUzU
e
zUxU
e
zUyU
e
zUzU
_

_
_

_
1380 Am=m 727 Arad 91 Arad
727 Arad 1991 Am=m 625 Arad
91 Arad 625 Arad 1129 Am=m
_

_
_

_
Notice that the value of e
xUxU
determined through matrix multiplication is
identical to that obtained using indicial notation, as expected.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Part (b). The rst, second, and third strain invariants will now be calcu-
lated using components of both e
ij
and e
iUjU
. It is expected that identical values
will be obtained since the strain invariants are independent of the coordinate
system.
First strain invariant:
xyz coordinate system:
H
e
e
ii
e
xx
e
yy
e
zz
H
e
1000 1500 2000 Am=m
H
e
4500Am=m :004500 m=m
xUyUzU coordinate system:
H
e
e
iUiU
e
xUxU
e
yUyU
e
zUzU
H
e
1380 1991 1129 Am=m
H
e
4500Am=m 0:004500 m=m
As expected, the rst strain invariant is independent of the coordinate
system.
Second strain invariant:
xyz coordinate system:
U
e

1
2
e
ii
e
jj
e
ij
e
ij
_ _
e
xx
e
yy
e
xx
e
zz
e
yy
e
zz
e
2
xy
e
2
xz
e
2
yz
_ _
U
e

_
10001500 10002000 15002000
500
2
250
2
750
2
g Am=m
2
U 5:625 10
6
Am=m
2
5:625 10
6
m=m
2
xUyUzU coordinate system:
U
e

1
2
e
iUiU
e
jUjU
e
iUjU
e
iUjU
_ _
U
e
e
xUxU
e
yUyU
e
xUxU
e
zUzU
e
yUyU
e
zUzU
e
2
xUyU
e
2
xUzU
e
2
yUzU
_ _
U
e
f13801991 13801129 19911129
727
2
91
2
625
2
g Am=m
2
U
e
5:625 10
6
Am=m
2
5:625 10
6
m=m
2
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
As expected, the second strain invariant is independent of the coordi-
nate system.
Third strain invariant:
xyz coordinate system:
W
e

1
6
e
ii
e
jj
e
kk
3e
ii
e
jk
e
jk
2e
ij
e
jk
e
ki
_ _
W
e
e
xx
e
yy
e
zz
e
xx
e
2
yz
e
yy
e
2
xz
e
zz
e
2
xy
2e
xy
e
xz
e
yz
W
e
100015002000 1000750
2
1500250
2
2000
500
2
2500250750 Am=m
3
W 2:031 10
9
Am=m
3
2:031 10
9
m=m
3
xUyUzU coordinate system:
W
e

1
6
e
iUiU
e
jUjU
e
kUkU
3e
iUiU
e
jUkU
e
jUkU
2e
iUjU
e
jUkU
e
kUiU
_ _
W
e
e
xUxU
e
yUyU
e
zUzU
e
xUxU
e
2
yUzU
e
yUyU
e
2
xUzU
e
zUzU
e
2
xUyU
2e
xUyU
e
xUzU
e
yUzU
W
e
138019911129 1380625
2
199191
2
1129
727
2
272791625 Am=m
3
W 2:031 10
9
Am=m
3
2:031 10
9
m=m
3
As expected, the third stress invariant is independent of the coordinate
system.
12 PRINCIPAL STRAINS
The denition 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 e
1
, e
2
, and e
3
.
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,
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
in this text, the axes associated with the principal strain coordinate system
will be labeled p
1
, p
2
, and p
3
axes, and the principal strains will be denoted
e
p1
, e
p2
, and e
p3
.
Since both stress and strain are second-order tensors, the principal
strains can be found using an approach analogous to that used to nd prin-
cipal stresses. Specically, it can be shown (1,2) that the principal strains
must satisfy the following three simultaneous equations:
e
p
n
e
xx
c
p
n
x
e
xy
c
p
n
y
e
xz
c
p
n
z
0 39a
e
xy
c
p
n
x
e
p
n
e
yy
c
p
n
y
e
yz
c
p
n
z
0 39b
e
xz
c
p
n
x
e
yz
c
p
n
y
e
p
n
e
zz
c
p
n
z
0 39c
Since direction cosines c
pnx
, c
pny
, and c
pnz
must also satisfy Eq. (8), and
therefore cannot all equal zero, the solution can be obtained by requiring
that the determinant of the coecients of c
pnx
, c
pny
, and c
pnz
equal zero:
e
p
n
e
xx
e
xy
e
xz
e
xy
e
p
n
e
yy
e
yz
e
xz
e
yz
e
p
n
e
zz

0
Equating the determinant to zero results in the following cubic
equation:
e
3
p
n
H
e
e
2
p
n
U
e
e
p
n
W
e
0 40
where H
e
, U
e
, and W
e
are the rst, 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:
Dene:
a
1
3
3U
e
H
2
e
_ _
b
1
27
2H
3
e
9H
e
U
e
27W
e
_ _
A

b
2

b
2
4

a
3
27
_
3

b
2

b
2
4

a
3
27
_
3

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.


The three principal strains [i.e., the three roots of Eq. (40)] are then given by:
e
p
1
; e
p
2
; e
p
3

A B
H
e
3

A B
2

3
p
A B
2
_ _

H
e
3

A B
2

3
p
A B
2
_ _

H
e
3
_

_
41
By convention, the principal strains are numbered such that e
p1
is the
algebraically greatest principal strain, whereas e
p3
is the algebraically least.
That is, e
p1
>e
p2
>e
p3
.
It is appropriate to note that many handheld calculators and computer
software packages now feature standard routines to nd 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 nding
the cube root of a complex number. The need to nd 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 nd 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 dene 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,
c
p
n
x
, c
p
n
y
, and c
p
n
z
.
The process of nding principal strains and direction cosines will be
demonstrated in the following example problem.
Example Problem 8
Given. A state of strain referenced to an xyz coordinate is known to be:
e
xx
e
xy
e
xz
e
yx
e
yy
e
yz
e
zx
e
zy
e
zz
_

_
_

_
1000Am=m 500 Arad 250 Arad
500 Arad 1500 Am=m 750 Arad
250Arad 750 Arad 2000 Am=m
_

_
_

_
Problem. Find (a) the principal strains and (b) the direction cosines that
dene the principal strain coordinate system.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Solution. This is the same strain tensor considered in Example Problem7. As
a part of that problem, the rst, second, and third strain invariants were found
to be:
H
e
0:004500 m=m
U 5:625 10
6
m=m
2
W 2:031 10
9
m=m
3
Part (a). Determining the principal strains.
In accordance with Eq. (40), the three principal strains are the roots
of the following cubic equation:
e
3
p
n
0:004500e
2
p
n
5:625 10
6
e
p
n
2:031 10
9
0
Following the standard procedure for nding the roots of a cubic equa-
tion, we have:
a
1
3
3U
e
H
2
e
_ _

1
3
35:625 10
6
0:004500
2
_ _
1:125 10
6
b
1
27
2H
3
e
9H
e
U
e
27W
e
_ _
b
1
27
20:004500
3
90:0045005:625 10
6

_
272:031 10
9
3:435 10
10
A

b
2

b
2
4

a
3
27
_
3

3:435 10
10
2

3:435 10
10

2
4

1:125 10
6

3
27

_
A

1:718 10
10
i1:524 10
10

3
_
A 594:5 10
6
i146:7 10
6

b
2

b
2
4

a
3
27
_
3

3:435 10
10
2

3:435 10
10

2
4

1:125 10
6

3
27

_
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
B

1:718 10
10
i1:524 10
10

3
_
B 594:5 10
6
i146:7 10
6

e
p
1
; e
p
2
; e
p
3

A B
H
3

A B
2

3
p
A B
2
_ _

H
3

A B
2

3
p
A B
2
_ _

H
3

2689Am=m
651Am=m
1160Am=m
_

_
_

_
Hence, e
p1
=2689 Am/m, e
p2
=1160 Am/m, and e
p3
=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:
e
p
n
e
xx
c
p
n
x
e
xy
c
p
n
y
e
xz
c
p
n
z
0
e
xy
c
p
n
x
e
p
n
e
yy
c
p
n
y
e
yz
c
p
n
z
0
c
p
n
x

2
c
p
n
y

2
c
p
n
z

2
1
Direction cosines for e
p1
. The three independent equations become:
2689 1000c
p
1
x
500c
p
1
y
250c
p
1
z
0
500c
p
1
x
2689 1500c
p
1
y
750c
p
1
z
0
c
p
1
x

2
c
p
1
y

2
c
p
1
z

2
1
Solving simultaneously, we obtain:
c
p
1
x
0:2872 c
p
1
y
0:5945 c
p
1
z
0:7511
Direction cosines for e
p2
. The three independent equations become:
1160 1000c
p
2
x
500c
p
2
y
250c
p
2
z
0
500c
p
2
x
1160 1500c
p
2
y
750c
p
2
z
0
c
p
2
x

2
c
p
2
y

2
c
p
2
z

2
1
Solving simultaneously, we obtain:
c
p
2
x
0:5960 c
p
2
y
0:5035 c
p
2
z
0:6256
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Direction cosines for e
p3
. The three independent equations become:
651 1000c
p
3
x
500c
p
3
y
250c
p
3
z
0
500c
p
3
x
651 1500c
p
3
y
750c
p
3
z
0
c
p
3
x

2
c
p
3
y

2
c
p
3
z

2
1
Solving simultaneously, we obtain:
c
p
3
x
0:7481 c
p
3
y
0:6286 c
p
3
z
0:2128
13 STRAINS WITHIN A PLANE PERPENDICULAR TO A
PRINCIPAL STRAIN DIRECTION
It has been seen that a strain tensor is dened 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: e
xz
=e
yz
=0, say (or equivalently,
c
xz
=c
yz
=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 xy
plane, e
xz
, e
yy
, and e
xy
. Two dierent circumstances are encountered in which
it is known a priori that the z-axis is a principal strain axis.
In the rst 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
e
zz
=e
xz
=e
yz
=0. Not only is the z-axis a principal strain axis in this case, but,
in addition, the principal strain equals zero: e
zz
=e
p3
=0. Since the three
remaining nonzero strain components (e
xx
, e
yy
, and e
xy
) all lie within the xy
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. Specically, 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 e
zz
is again
one of the principal strains. However, in this second case, the out-of-plane
normal strain does not, in general, equal zero: e
zz
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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 xy plane (r
xx
,
r
yy
, and/or s
xy
) can cause out-of-plane shear strains (e
xz
and/or e
yz
). 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 xy plane, e
xx
, e
yy
, and e
xy
. We will write
these strains in the form of a column array, rather than a 33 array:
e
xx
e
xy
0
e
xy
e
yy
0
0 0 e
zz
_

_
_

_
e
xx
c
xy
=2 0
c
xy
=2 e
yy
0
0 0 e
zz
_

_
_

_ !
e
xx
e
yy
e
xy
_

_
_

e
xx
e
yy
c
xy
=2
_

_
_

_
Note that e
zz
does not appear in the column array. This is not of concern in the
case of plane strain since, in this case, e
zz
=0. However, in the case of plane
stress, it is important to remember that (in general) e
zz
p 0. Although in the
following chapters we will be primarily interested in strains induced within
the xy plane, the reader is advised to remember that an out-of-plane strain
e
zz
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 simplied 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
repeated here for convenience:
e
x
V
x
V e
x
V
y
V e
x
V
z
V
e
y
V
x
V e
y
V
y
V e
y
V
z
V
e
z
V
x
V e
z
V
y
V e
z
V
z
V
_

_
_

_
c
x
V
x
c
x
V
y
c
x
V
z
c
y
V
x
c
y
V
y
c
y
V
z
c
z
V
x
c
z
V
y
c
z
V
z
_

_
_

_
e
xx
e
xy
e
xz
e
yx
e
yy
e
yz
e
zx
e
zy
e
zz
_

_
_

c
x
V
x
c
y
V
x
c
z
V
x
c
x
V
y
c
y
V
y
c
z
V
y
c
x
V
z
c
y
V
z
c
z
V
z
_

_
_

_ repeated36b
Assuming that the xVyVzV coordinate system is generated from the xyz
system by a rotation h about the z-axis, the direction cosines are:
c
x
V
x
cos h
c
x
V
y
cos 90
B
h sin h
c
x
V
z
cos 90
B
0
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
c
y
V
x
cos 90
B
h sin h
c
y
V
y
cos h
c
y
V
z
cos 90
B
0
c
z
V
x
cos 90
B
0
c
z
V
y
cos 90
B
0
c
z
V
z
cos 0
B
1
Substituting these direction cosines into Eq. (36b) and noting that by
assumption, e
xz
=e
yz
=0, we have:
e
x
V
x
V e
x
V
y
V e
x
V
z
V
e
y
V
x
V e
y
V
y
V e
y
V
z
V
e
z
V
x
V e
z
V
y
V e
z
V
z
V
_

_
_

_
cos h sin h 0
sin h cos h 0
0 0 1
_

_
_

_
e
xx
e
xy
0
e
yx
e
yy
0
0 0 e
zz
_

_
_

cos h sin h 0
sin h cos h 0
0 0 1
_

_
_

_
Completing the matrix multiplication indicated results in:
e
x
V
x
V e
x
V
y
V e
x
V
z
V
e
y
V
x
V e
y
V
y
V e
y
V
z
V
e
z
V
x
V e
z
V
y
V e
z
V
z
V
_

_
_

cos
2
he
xx
sin
2
he
yy
2cos hsinhe
xy
cos hsinhe
xx
cos hsinhe
yy
cos
2
h sin
2
he
xy
0
cos hsinhe
xx
cos hsinhe
yy
cos
2
h sin
2
he
xy
sin
2
he
xx
cos
2
he
yy
2cos hsinhe
xy
0
0 0 e
zz
_
_
_
_
As would be expected, e
xVzV
=e
yVzV
=0. The remaining strain components are:
e
x
V
x
V cos
2
he
xx
sin
2
he
yy
2coshsinhe
xy
e
y
V
y
V sin
2
he
xx
cos
2
he
yy
2coshsinhe
xy
42
e
x
V
y
V coshsinhe
xx
coshsinhe
yy
cos
2
h sin
2
he
xy
e
z
V
z
V e
zz
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 nal results, Eq. (42), to
ones which involve engineering shear strain (c
xy
). Recall from Sec. 9 that
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
e
xy

1
2
c
xy
. Hence, to convert Eq. (42), simply replace e
xy
with
1
2
c
xy
every-
where, resulting in:
e
x
V
x
V cos
2
he
xx
sin
2
he
yy
coshsinhc
xy
e
y
V
y
V sin
2
he
xx
cos
2
he
yy
coshsinhc
xy
43
c
x
V
y
V
2
coshsinhe
xx
coshsinhe
yy
cos
2
h sin
2
h
c
xy
2
e
z
V
z
V e
zz
Equation (43) relates the components of strain in two dierent coordinate
systems within a single plane and will be used extensively throughout the re-
mainder of this text.
The rst three of Eq. (43) can be written using matrix notation as:
e
x
V
x
V
e
y
V
y
V
c
x
V
y
V
2
_

_
_

cos
2
h sin
2
h 2 coshsinh
sin
2
h cos
2
h 2 coshsinh
coshsinh cosh sinh cos
2
h sin
2
h
_

_
_

e
xx
e
yy
c
xy
2
_

_
_

_
44
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 simplied when the out-of-plane z-axis is a principal axis. Since
by denition e
xz

c
xz
2
e
yz

c
yz
2
0, the strain invariants become:
First strain invariant H
e
e
xx
e
yy
e
zz
Second strain invariant U
e
e
xx
e
yy
e
xx
e
zz
e
yy
e
zz

c
2
xy
4
45
Third strain invariant W
e
e
xx
e
yy
e
zz
e
zz
c
4
xy
4
The principal strains equal the roots of the cubic equation previously
listed as Eq. (40). Substituting Eq. (45) into Eq. (40) reduces to:
e
3
p
n
e
xx
e
yy
e
zz
e
2
p
n
e
xx
e
yy
e
xx
e
zz
e
yy
e
zz

c
2
xy
4
e
p
n
e
xx
e
yy
e
zz
e
zz
c
2
xy
4
0 46
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
One root of Eq. (23) is e
pn
=e
zz
. For present purposes, this root will be
labeled e
p3
although it may not be the algebraically least principal strain. In
the case of plane strain, e
p3
=e
zz
=0.
Removing the known root from Eq. (46), we have the following qua-
dratic equation:
e
2
p
n
e
xx
e
yy
e
pn
e
xx
e
yy

c
2
xy
4
0
The two roots of this quadratic equation (that is, the two remaining prin-
cipal strains, e
p1
and e
p2
) may be found by application of the standard ap-
proach (3) and are given by:
e
p
1
; e
p
2

e
xx
e
yy
2
F

e
xx
e
yy
2
_ _
2

c
xy
2
_ _
2

47
The angle h
p
e
between the x-axis and either the p
1
or p
2
axis is given by:
h
p
e

1
2
arctan
c
xy
e
xx
e
yy
_ _
48
Example Problem 9
Given. A state of plane strain is known to consist of:
e
xx
500Am=m
e
yy
1000Am=m
c
xy
2500Arad
Problem. (a) Prepare a rough sketch (not to scale) of the deformed strain
element in the xy coordinate system; (b) determine the strain components
which correspond to an xVyV coordinate system, oriented 25j CCW from the
xy coordinate system, and prepare a rough sketch (not to scale) of the
deformed strain element in the xVyV coordinate system; and (c) determine
the principal strain components that exist within the xy plane, and prepare
a rough sketch (not to scale) of the deformed strain element in the principal
strain coordinate system.
Solution
Part (a). Asketch showing the deformed strain element (not to scale) in the
xy 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 e
xx
=500 Am/m).
The length of the element side parallel to the y-axis has decreased (cor-
responding to the compressive strain e
yy
=1000 Am/m).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Figure 19 Strain elements associated with Example Problem 9 (all deforma-
tions shown greatly exaggerated for clarity).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
The angle dened by xy axes has increased (corresponding to the
negative shear strain c
xy
=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):
e
x
V
x
V
e
y
V
y
V
c
x
V
y
V
2
_

_
_

cos
2
25
B
sin
2
25
B
2cos25
B
sin25
B

sin
2
25
B
cos
2
25
B
2cos25
B
sin25
B

cos25
B
sin25
B
cos25
B
sin25
B
cos
2
25
B
sin
2
25
B

_
_

500
1000
2500
2
_

_
_

_
Completing the matrix multiplication indicated results in:
e
x
V
x
V
e
y
V
y
V
c
x
V
y
V
2
_

_
_

725 Am=m
225 Am=m
1378 Arad
_

_
_

_
A sketch showing the deformed strain element (not to scale) in the xVyV
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 e
xVxV
=725 Am/m).
The length of the element side parallel to the yV-axis has increased (cor-
responding to the tensile strain e
yVyV
=225 Am/m).
The angle dened by the xVyV axes has increased (corresponding to the
negative shear strain c
xVyV
=2756 Arad).
Part (c). The principal strains are found through application of Eq. (47):
e
p
1
; e
p
2

500 1000
2
F

500 1000
2
_ _
2

2500
2
_ _
2

e
p
1
1208Am=m
e
p
2
1708Am=m
The orientation of the principal strain coordinate system is given by Eq. (48):
h
p
e

1
2
arctan
2500
500 1000
_ _
29:5
B
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 p
1
-axis has increased
(corresponding to the tensile principal strain e
p1
=1208 Am/m).
The length of the element side parallel to the p
2
-axis has decreased
(corresponding to the compressive principal strain e
p2
=1708
Am/m).
The angle dened 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 elds 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 specied. Strains induced in the structure will then be inferred
from these displacement elds.
In the most general case, three displacement elds are involved. Spe-
cically, these are the displacements in the x-, y-, and z-directions, typically
denoted as the u-, v-, and w-displacement elds, respectively. In general, all
three displacement elds 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 elds and the strain tensor de-
pends upon the magnitude of derivatives of displacement elds (also called
displacement gradients). If displacement gradients are arbitrarily large, then
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
the associated level of strain is said to be nite, and each component of the
strain tensor is related nonlinearly to displacement gradients as follows:
e
xx

Bu
Bx

1
2
Bu
Bx
_ _
2

Bv
Bx
_ _
2

Bw
Bx
_ _
2
_ _
e
yy

Bv
By

1
2
Bu
By
_ _
2

Bv
By
_ _
2

Bw
By
_ _
2
_ _
e
zz

Bw
Bz

1
2
Bu
Bz
_ _
2

Bv
Bz
_ _
2

Bw
Bz
_ _
2
_ _
c
xy

Bu
By

Bv
Bx

Bu
Bx
_ _
Bu
By
_ _

Bv
Bx
_ _
Bv
By
_ _

Bw
Bx
_ _
Bw
By
_ _
c
xz

Bu
Bz

Bw
Bx

Bu
Bx
_ _
Bu
Bz
_ _

Bv
Bx
_ _
Bv
Bz
_ _

Bw
Bx
_ _
Bw
Bz
_ _
c
yz

Bw
By

Bv
Bz

Bu
By
_ _
Bu
Bz
_ _

Bv
By
_ _
Bv
Bz
_ _

Bw
By
_ _
Bw
Bz
_ _
The expressions listed above dene what is known as Greens 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:
Bu
Bx
_ _
2
c0
Bv
Bx
_ _
2
c0
Bw
Bx
_ _
2
c0
Bu
Bx
_ _
Bu
By
_ _
c0; etc:
When displacement gradients are very small, the level of strain is said to be
innitesimal, and each component of the strain tensor is linearly related to
displacement gradients as follows:
e
xx

Bu
Bx
49a
e
yy

Bv
By
49b
e
zz

Bw
Bz
49c
c
xy

Bv
Bx

Bu
By
49d
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
c
xz

Bw
Bx

Bu
Bz
49e
c
yz

Bw
By

Bv
Bz
49f
For most analyses considered in this text, we will assume that strains
are innitesimal and are related to displacement elds 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 straindisplacement relationships.
As stated above, most analyses begin with the consideration of the
displacement elds induced in a structure of interest. Strain elds implied by
these displacements are then calculated in accordance with Eqs. (49a)(49f).
This process insures that strain elds are consistent with displacements. Con-
sider the opposite approach. Specically, suppose that mathematical expres-
sions for strain elds are assumed, perhaps on the basis of engineering
judgment. In this case, it is possible that the assumed strain elds correspond
to physically unrealistic displacement elds. For example, displacement elds
inferred from assumed strain elds 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 elds u(x,y,z),
v(x,y,z), and w(x,y,z). To develop the compatibility conditions, dierentiate
Eq. (49d) twice, once with respect to x and once with respect to y. We obtain:
B
2
c
xy
BxBy

B
3
u
BxBy
2

B
3
v
Bx
2
By
From Eqs. (49a) and (49b), it is easily seen that:
B
2
e
xx
By
2

B
3
u
BxBy
2
B
2
e
yy
Bx
2

B
3
v
Bx
2
By
Combining these results, we see that expressions for the strain components
e
xx
, e
yy
, and c
xy
correspond to physically reasonable displacement elds (i.e.,
are compatible) only if they satisfy:
B
2
c
xy
BxBy

B
2
e
xx
By
2

B
2
e
yy
Bx
2
50a
Equation (50a) is the rst compatibility condition. Following a similar pro-
cedure using Eqs. (49e) and (49f), we obtain:
B
2
c
yz
ByBz

B
2
e
yy
Bz
2

B
2
e
zz
By
2
50b
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
B
2
c
xz
BxBz

B
2
e
xx
Bz
2

B
2
e
zz
Bx
2
50c
These are the second and third compatibility conditions. Next, the following
expressions are obtained using Eqs. (49a), (49d), (49e), and (49f), respectively:
B
3
u
BxByBz

B
2
e
xx
ByBz
B
3
u
BxByBz

B
2
c
xy
BxBz

B
3
v
Bx
2
Bz
B
3
u
BxByBz

B
2
c
xz
BxBy

B
3
w
Bx
2
By
B
3
w
Bx
2
By

B
3
v
Bx
2
Bz

B
2
c
yz
Bx
2
Combining these four expressions, we nd that assumed expressions for
strain components e
xx
, c
xy
, c
xz
, and c
yz
, are compatible if:
2
B
2
e
xx
ByBz

B
Bx
Bc
xy
Bz

Bc
xz
By

c
yz
Bx
_ _
50d
This is the fourth compatibility condition. The nal two compatibility con-
ditions are developed using a similar process and are given by:
2
B
2
e
yy
BxBz

B
By
Bc
yz
Bx

Bc
xz
By

c
xy
Bz
_ _
50e
2
B
2
e
zz
BxBy

B
Bz
Bc
yz
Bx

Bc
xz
By

c
xy
Bz
_ _
50f
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 andstraintensors is of particular importance. Indeed, nearlyall 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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 xyz coordinate
system to the xUyUzU coordinate system, where the xUyUzU coordinate
system is generated from the xyz 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
Problem2. Program3DROTATEalso calculates the angles betweenthe xyz
and xUyUzU 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. Conrm 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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
(f ) h=45j b=60j.
(g) h=45j b=60j.
(h) h=45j b=60j.
2. Consider an xVVVyVVVzVVV coordinate system, which is generated from
an xyz coordinate system by the following three rotations:

A rotation of about the original z-axis, which denes an intermediate


xVyVzV coordinate system [see Fig. 2(a)], followed by

A rotation of b about the xV-axis, which denes an intermediate xVV


yVVzVV coordinate system [see Fig. 2(b)], followed by

Arotation of wabout the yVV-axis, which denes the nal xVVVyVVVzVVV


coordinate system.
Show that the xVVVyVVVzVVV and xyz coordinate systems are related by
the following direction cosines:
c
x
VVV
x
c
x
VVV
y
c
x
VVV
z
c
y
VVV
x
c
y
VVV
y
c
y
VVV
z
c
z
VVV
x
c
z
VVV
y
c
z
VVV
z
_
_
_
_

coswcosh sinwsinbsinh coswsinh sinwsinbcosh sinwcosb


cosbsinh cosbcosh sinb
sinwcosh coswsinbsinh sinwsinh coswsinbcosh coswcosb
_
_
_
_
3. The force vector discussed in Example Problem 1 is given by:
F 1000i

200j

600k

Using Eq. (6c), express F in a new coordinate system dened 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 conrm 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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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.
Conrm 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) r
xVxV
.
(b) r
xVyV
.
(c) r
yVyV
.
(d) r
yVzV
.
(e) r
zVzV
.
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.):
r
xx
r
xy
r
xz
r
yx
r
yy
r
yz
r
zx
r
zy
r
zz
_
_
_
_

50 10 15
10 25 30
15 30 5
_
_
_
_
ksi
7. 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 shear
stresses has been reversed.):
r
xx
r
xy
r
xz
r
yx
r
yy
r
yz
r
zx
r
zy
r
zz
_
_
_
_

50 10 15
10 25 30
15 30 5
_
_
_
_
ksi
8. Use program 3DROTATE to determine the stress invariants for the
stress tensor listed below, and compare to those determined in Example
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Problem 3. (Note: this stress tensor is similar to the one considered in
Example Problem 3 except that the algebraic sign of all stress compo-
nents has been reversed.):
r
xx
r
xy
r
xz
r
yx
r
yy
r
yz
r
zx
r
zy
r
zz
_
_
_
_

50 10 15
10 25 30
15 30 5
_
_
_
_
ksi
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.):
e
xx
e
xy
e
xz
e
yx
e
yy
e
yz
e
zx
e
zy
e
zz
_

_
_

_
1000 Am=m 500 Arad 250 Arad
500 Arad 1500 Am=m 750 Arad
250 Arad 750 Arad 2000 Am=m
_

_
_

_
10. 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 normal strain
components has been reversed.):
e
xx
e
xy
e
xz
e
yx
e
yy
e
yz
e
zx
e
zy
e
zz
_

_
_

_
1000 Am=m 500 Arad 250 Arad
500 Arad 1500 Am=m 750 Arad
250 Arad 750 Arad 2000 Am=m
_

_
_

_
11. 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 strain compo-
nents has been reversed.):
e
xx
e
xy
e
xz
e
yx
e
yy
e
yz
e
zx
e
zy
e
zz
_

_
_

_
1000 Am=m 500 Arad 250 Arad
500 Arad 1500 Am=m 750 Arad
250 Arad 750 Arad 2000 Am=m
_

_
_

_
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
REFERENCES
1. Frederick, D.; Chang, T.S. Continuum Mechanics; Scientic Publishers, Inc:
Cambridge, MA, 1972.
2. Fung, Y.C. A First Course In Continuum Mechanics; Prentice-Hall, Inc.: Engle-
wood Clis, 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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 classied as either orthotropic or
transversely isotropic materials. Sections devoted to those material properties
of primary interest to structural engineers then follow. Specically, 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 rst be dened for anisotropic
materials. These general denitions 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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
of interest. Material properties that describe the ability of a material to
conduct electricity, to transmit (or reect) visible light, to transfer heat, or to
support mechanical loading, to name but a few, have been dened. Material
properties of interest herein are those used by engineers during the design of
composite structures. Two specic examples are Youngs modulus, E and
Poissons 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 specically, 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 dierent orientations, as shown in Fig. 1.
The geometry of the three specimens is assumedtobe identical, sothat the only
dierence between specimens is the original orientation of each specimen
within the parent block. Now suppose that the axial stiness (i.e., Youngs
modulus E) is measured for each specimen. Youngs modulus measured using
specimen 1 will be denoted E
xx
(i.e., subscripts are used to indicate the original
orientation of specimen 1 within the parent block). Similarly, Youngs
modulus measured using specimens 2 and 3 will be denoted E
yy
and E
zz
,
respectively.
If the parent block consists of an isotropic material, then Youngs
modulus measured for each specimen will be identical (to within engineering
accuracies)for isotropic materials: E
xx
= E
yy
= E
zz
.
In this case, an identical value of Youngs 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 dierent Youngs
modulus will, in general, be measured for each specimenfor anisotropic
materials: E
xx
p E
yy
p E
zz
.
In this case, the value of Youngs 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 (Poissons ratio, thermal expansion coecients, 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 rst dene 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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
arranged in well-dened 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 dierent properties in dierent directions, and hence is aniso-
tropic. That is to say, if material properties are dened at a physical scale on
Figure 1 Illustration of method used to determine whether a material is iso-
tropic or anisotropic.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
the order of a grain diameter or smaller, then all metals or metal alloys must be
dened 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/mm
3
(134 10
6
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
signicant 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. Specically, it is the uniform and symmetrical orientation of the
reinforcing bers 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 bers embedded within a rela-
tively exible polymeric matrix, as shown in Fig. 2a and b. Note that the
coordinate system used to describe the plate has been labeled the 123 axes.
In this case, the 1-axis is dened to be parallel to the bers, the 2-axis is dened
to lie within the plane of the plate and is perpendicular to the bers, and the 3-
axis is dened to be normal to the plane of the plate. Note that bers are
arranged symmetrically about the 13 and 23 planes. The 123 coordinate
system will henceforth be referred to as the principal material coordinate
system. Referring to Fig. 2a, specimen 1 is machined such that the bers are
aligned with the long axis of the specimen, whereas in specimen 2, the bers
are perpendicular to the axis of the specimen. Obviously, Youngs modulus
measured for these two specimens will be quite dierent. Specically, the
modulus measured for specimen 1 will approach that of the bers, whereas the
modulus measured for specimen 2 will approach that of the polymeric matrix.
Therefore, E
11
> >E
22
, and Youngs modulus varies with direction within the
material, satisfying the denition of an anisotropic material.
The principal material coordinate system is not always aligned with the
ber 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 bers oriented in two (or more) nonorthogonal directions. The
three principal material coordinate axes lie within planes that are symmetrical
with respect to the ber array. As before, the 1- and 2-axes lie within the plane
of the plate, and the 3-axis is dened normal to the plane of the plate.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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
howthis coupling occurs in the case of a unidirectional composite is presented
in Fig. 3. Aspecimen in which the unidirectional bers are oriented at an angle
of 45j with respect to the x-axis is shown, and a small square element fromthe
gage region is isolated. Because the element is initially square and, in this
example, the bers are dened to be at an angle of 45j, bers are parallel to
diagonal AC of the element. In contrast, bers are perpendicular to diagonal
BD. This implies that the element is stier along diagonal AC than along
diagonal BD.
Figure 2 Illustration of the principal material coordinate system for thin com-
posite laminates.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Now assume that a tensile stress is applied, causing the square element
(as well as the specimen as a whole) to deform. Because the stiness 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 gure. Note
that:

The length of the square element is increased in the x-direction


(corresponding to a tensile strain, e
xx
).

The length of the square element is decreased in the y-direction


(corresponding to a compressive strain e
yy
, and associated with the
Poisson eect).

BDABis no longer p/2 rad, which indicates that a shear strain c


xy
has
been induced.
Figure 3 A 45j off-axis composite specimen used to explain the origin of cou-
pling effects.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Hence, in this example, the normal stress r
xx
has induced two normal strains
(e
xx
and e
yy
) as well as a shear strain c
xy
. Hence, coupling exists between r
xx
and c
xy
, 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 r
xx
only. In particular, note
that no shear stress exists in the xy coordinate system (s
xy
= 0). Con-
sequently, the in-plane principal stresses applied to the specimen are r
p
1
=r
xx
and r
p
2
=r
yy
=0, and the principal stress coordinate systemis dened by the
xy coordinate system. However, a shear strain does exist in the xy
coordinate system (c
xy
p 0). Consequently, the xy 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 eect hinges on the fact that the ber direction diers
fromthe direction of the applied stress r
xx
. Specically, the bers are oriented
45j away from the direction of the applied stress r
xx
. If the bers were aligned
with either the x- or y-axis, then a coupling between r
xx
and c
xy
would not
occur. We conclude that the unusual coupling eects exhibited by composites
only occur if stress and strain are referenced to a nonprincipal material
coordinate system.
Anisotropic materials are classied according to the number of planes of
symmetry dened 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 dened: the 12 plane, the
13 plane, and the 23 plane. Composites fall within one of two classications
of anisotropic behavior. Specically, 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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Although many kinds of material properties may be dened, 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 dened in the following sections. In each section, a
general denition 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 systemwill be discussed rst.
These general denitions 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
specic 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 reect
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 dierent experimental
approaches may be taken. In the rst approach, the material of interest is
subjected to a well-dened stress tensor, and components of the resulting
strain tensor are measured. In the second approach, the material of interest is
subjected to a well-dened strain tensor, and the components of the resulting
stress tensor are measured. Froman experimental standpoint, it is far easier to
impose a well-dened stress tensor than a well-dened strain tensor, and
hence the rst 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 strainspecically, 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 ve 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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
applied to a test specimen (say, r
xx
), while insuring that the remaining ve
stress components are zero (r
yy
= r
zz
= s
xy
= s
xz
= s
yz
= 0). The six com-
ponents of strain caused by r
xx
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, s
xy
), while insuring that the
remaining ve stress components are zero (r
xx
= r
yy
= r
zz
= s
xz
= s
yz
= 0).
The six components of strain caused by s
xy
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
specied 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 dicult to achieve in practice. For example, because composites are
usually produced in the form of thin platelike structures, it is very dicult to
impose well-dened 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 diculties 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 r
xx
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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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
D3410 Standard Test Method for Compressive Properties of
Polymer Matrix Composite Materials with Unsupported
Gage Section by Shear Loading
D5467 Standard Test Method for Compressive Properties of
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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
low, such that a linear relationship exists between the stress component r
xx
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 r
xx
will induce all six components of strain: e
xx
,
e
yy
, e
zz
, c
xy
, c
xz
, and c
yz
. This is not the case for isotropic materials; a uniaxial
stress r
xx
applied to an isotropic material will not induce any shear strains
(c
xy
=c
xz
=c
yz
=0); furthermore, the transverse normal strains will be
identical (e
yy
=e
zz
). 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 r
xx
is increased, the mag-
nitude of all resulting strain components is also increased. Because stress r
xx
causes six distinct components of strain for an anisotropic material, six ma-
terial properties must be dened in order to relate r
xx
to the resulting strains.
Let us rst consider material properties relating normal stress r
xx
to
normal strains e
xx
, e
yy
, and e
zz
. The relationship between r
xx
and normal
strain e
xx
is characterized by Youngs modulus E
xx
(also called the modulus
of elasticity):
E
xx
u
r
xx
e
xx
1
Youngs modulus is simply the slope of the r
xx
vs. e
xx
curve shown in
Fig. 4. In words, Youngs modulus is dened as the normal stress r
xx
divided
Figure 4 Idealized plot of the six strain components caused by the application
a uniaxial stress r
xx
.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
by the resulting normal strain e
xx
, with all other stress components equal to
zero. Subscripts xx have been used to indicate the direction in which
Youngs modulus has been measured. Because we have restricted our
attention to the linear region of the stressstrain curve, Eq. (1) is only valid
at relatively low, linear stress levels.
The relationship between the two transverse strains (e
yy
and e
zz
) and e
xx
is dened by Poissons ratio:
v
xy
u
e
yy
e
xx
v
xz
u
e
zz
e
xx
2
In words, Poissons ratio v
xy
(or v
xz
) is dened as the negative of the trans-
verse normal strain e
yy
(or e
zz
) divided by the axial normal strain e
xx
, both of
which are induced by stress r
xx
, with all other stresses equal to zero.
As before, subscripts have been used to indicate the uniaxial stress
condition under which Poissons ratio is measured. The rst subscript
indicates the direction of stress, and the second subscript indicates the
direction of transverse strain. For example, in the case of v
xy
, the rst
subscript x indicates that a uniaxial stress r
xx
has been applied, and the
second subscript y indicates that transverse normal strain e
yy
has been used to
calculate Poissons ratio.
Combining Eqs. (1) and (2), a relationship between r
xx
and transverse
strains e
yy
and e
zz
is obtained:
e
yy

v
xy
E
xx
r
xx
e
zz

v
xz
E
xx
r
xx
3
Now consider material properties relating normal strain e
xx
to shear
strains c
xy
, c
xz
, and c
yz
. Material properties relating normal strains to shear
strains were discussed by Lekhnitski [2] and are called coecients of mutual
inuence of the second kind. In this text, they will be denoted using the
symbol g, and are dened as follows:
g
xx;xy
u
c
xy
e
xx
g
xx;xz
u
c
xz
e
xx
g
xx;yz
u
c
yz
e
xx
4
In words, the coecient of mutual inuence of the second kind g
xx,xy
(or
g
xx,xz
, or g
xx,yz
) is dened as the shear strain c
xy
(or c
xz
, or c
yz
) divided by the
normal strain e
xx
, both of which are induced by normal stress r
xx
, when all
other stresses equal zero.
Subscripts have once again been used to indicate the stress condition
under which the coecient of mutual inuence of the second kind is
measured. The rst set of subscripts indicates the direction of stress, and
the second set of subscripts indicates the shear strain used to calculate the
coecient. For example, in the case of g
xx,xy
, the rst two subscripts xx
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
indicate that a normal stress r
xx
has been applied, and the second two
subscripts xy indicate that c
xy
has been used to calculate the coecient.
Combining Eqs. (1) and (4), a relationship between r
xx
and shear strain
c
xy
, c
xz
, or c
yz
is obtained:
c
xy

g
xx;xy
e
xx
r
xx
c
xz

g
xx;xz
e
xx
r
xx
c
yz

g
xx;yz
e
xx
r
xx
5
Equations (1)(5) dene 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 (s
yy
Applied)
E
yy
u
r
yy
e
yy
or e
yy

1
E
yy
r
yy
v
yx
u
e
xx
e
yy
or e
xx

v
yx
E
yy
r
yy
v
yz
u
e
zz
e
yy
or e
zz

v
yz
E
yy
r
yy
g
yy;xy
u
c
xy
e
yy
or c
xy

g
yy;xy
e
yy
r
yy
g
yy;xz
u
c
xz
e
yy
or c
xz

g
yy;xz
e
yy
r
yy
g
yy;yz
u
c
yz
e
yy
or c
yz

g
yy;yz
e
yy
r
yy
6
Properties Measured Using Specimen 3 (s
zz
Applied):
E
zz
u
r
zz
e
zz
or e
zz

1
E
zz
r
zz
v
zx
u
e
xx
e
zz
or e
xx

v
zx
E
zz
r
zz
v
zy
u
e
yy
e
zz
or e
yy

v
zy
E
zz
r
zz
g
zz;xy
u
c
xy
e
zz
or c
xy

g
zz;xy
e
zz
r
zz
g
zz;xz
u
c
xz
e
zz
or c
xz

g
zz;xz
e
zz
r
zz
g
zz;yz
u
c
yz
e
zz
or c
yz

g
zz;yz
e
zz
r
zz
7
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
2.2 Pure Shear Tests
If a pure shear stress (say, s
xy
) 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 s
xy
and the resulting strains. Strains induced at high nonlinear
stress levels, including failure stresses, will be considered in Sec. 5. Once again,
the stressstrain response of an anisotropic material diers markedly from
that of an isotropic material. Specically, for an anisotropic material, stress
s
xy
will induce all six components of strain: e
xx
, e
yy
, e
zz
, c
xy
, c
xz
, and c
yz
. If an
isotropic material is subjected to a pure shear stress s
xy
, only one strain
component is induced (c
xy
); all other strain components are zero (e
xx
=
e
yy
=e
zz
=c
xz
=c
yz
=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 xy plane) and out-of-
plane shear strains (c
xz
and c
yz
). Neither of these coupling eects occurs in
isotropic materials.
As would be expected, as the magnitude of s
xy
is increased during the
test, the magnitude of the resulting strains is also increased. Because stress s
xy
causes six distinct components of strain, six material properties must be
dened in order to relate s
xy
to the resulting strains.
Figure 5 Idealized plot of the six strain components caused by the application
a pure shear stress s
xy
.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Let us rst consider material properties relating shear stress s
xy
to shear
strains c
xy
,c
xz
, and c
yz
. The relationship between s
xy
and shear strain c
xy
is
characterized by the shear modulus G
xy
:
G
xy
u
s
xy
c
xy
8
In words, the shear modulus is dened as the shear stress s
xy
divided by
the resulting shear strain c
xy
, 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 (c
xz
, c
yz
) and c
xy
is charac-
terized by Chentsov coecients, which will be denoted using the symbol l in
this text:
l
xy;xz
u
c
xz
c
xy
l
xy;yz
u
c
yz
c
xy
9
In words, the Chentsov coecient l
xy
,
xz
(or l
xy
,
yz
) is dened as the
shear strain c
xz
(or c
yz
) divided by the shear strain c
xy
, both of which are
induced by shear stress s
xy
, with all other stresses equal to zero. The rst 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-
ecient. For example, in the case of l
xy
,
xz
the subscripts xy indicate that a
pure shear stress s
xy
has been applied, and the second two subscripts xz
indicate that c
xz
has been used to calculate the coecient.
A comparison between Eqs. (2) and (9) reveals that Chentsov coe-
cients are directly analogous to Poissons ratio. Poissons ratio is dened as a
ratio of normal strains caused by a normal stress, whereas Chentsov coef-
cients are dened as a ratio of shear strains caused by a shear stress.
Combining Eqs. (8) and (9), a relationship between s
xy
and shear strain
c
xz
or c
yz
is obtained:
c
xz

l
xy;xz
G
xy
s
xy
c
yz

l
xy;yz
G
xy
s
xy
10
Finally, consider material properties relating shear stress s
xy
to normal
strains e
xx
, e
yy
, and e
zz
. Material properties relating shear stress to normal
strains were discussed by Lekhnitski [2] and are called coecients of mutual
inuence of the rst kind. In this text, they will be denoted using the symbol
g, and are dened as follows:
g
xy;xx
u
e
xx
c
xy
g
xy;yy
u
e
yy
c
xy
g
xy;zz
u
e
zz
c
xy
11
In words, the coecient of mutual inuence of the rst kind g
xy
,
xx
(or
g
xy
,
yy
, or g
xy
,
zz
) is dened as the normal strain e
xx
(or e
yy
, or e
zz
) divided by
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
the shear strain c
xy
, both of which are induced by shear stress s
xy
, when all
other stresses equal zero. The rst set of subscripts indicates the stress
component applied, and the second set indicates the normal strain used to
calculate the coecient. For example, in the case of g
xy
,
xx
, the rst subscripts
xy indicate that shear stress s
xy
has been applied, and the second set of
subscripts xx indicates that e
xx
has been used to calculate the coecient.
Combining Eqs. (8) and (11), a relationship between s
xy
and normal
strain e
xx
, e
yy
, or e
zz
is obtained:
e
xx

g
xy;xx
G
xy
s
xy
e
yy

g
xy;yy
G
xy
s
xy
e
zz

g
xy;zz
G
xy
s
xy
12
Equations (8)(12) dene six properties measured when a pure shear
stress s
xy
is applied. Analogous material properties are dened during tests in
which pure shear s
xz
or s
yz
is applied.
Properties Measured Using Pure Shear t
xz
G
xz
u
s
xz
c
xz
or c
xz

1
G
xz
s
xz
l
xz;xy
u
c
xy
c
xz
or c
xy

l
xz;xy
G
xz
s
xz
l
xz;yz
u
c
yz
c
xz
or c
yz

l
xz;yz
G
xz
s
xz
g
xz;xx
u
e
xx
c
xz
or e
xx

g
xz;xx
G
xz
s
xz
g
xz;yy
u
e
yy
c
xz
or e
yy

g
xz;yy
G
xz
s
xz
g
xz;zz
u
e
zz
c
xz
or e
zz

g
xz;zz
G
xz
s
xz
13
Properties Measured Using Pure Shear t
yz
G
yz
u
s
yz
c
yz
or c
yz

1
G
yz
s
yz
l
yz;xy
u
c
xy
c
yz
or c
xy

l
yz;xy
G
yz
s
yz
l
yz;xz
u
c
xz
c
yz
or c
xz

l
yz;xz
G
yz
s
yz
g
yz;xx
u
e
xx
c
yz
or e
xx

g
yz;xx
G
yz
s
yz
g
yz;yy
u
e
yy
c
yz
or e
yy

g
yz;yy
G
yz
s
yz
g
yz;zz
u
e
zz
c
yz
or e
zz

g
yz;zz
G
yz
s
yz
14
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 dened by the ber direction. That is,
the 1-axis is dened parallel to the ber direction, the 2-axis is perpendicular
to the bers and lies within the plane of the composite, and the 3-axis is
perpendicular to the bers and lies out of plane. In other cases, such as the
braided composite shown in Fig. 2c and d, the 123 principal material
coordinate systemis not aligned with the ber direction, but is instead dened
by planes of symmetry associated with the ber architecture involved. For all
composite fabrics based on continuous bers and typically encountered in
practice (i.e., unidirectional, woven, or braided fabrics), the principal material
coordinate system is readily identied.
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 123 coordinate system. To simplify
our discussion, we will assume that the composite under consideration is a
unidirectional composite, and hence that the 123 axes are parallel and
perpendicular to the bers.
A stress element representing a unidirectional composite subjected to
uniaxial tensile stress r
11
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 e
11
.

The element has decreased in width in the 2- and 3-directions,


corresponding to compressive strains e
22
and e
33
, respectively.

The deformed element is a rectangular parallelepiped. That is, due to


the symmetrical distribution of bers 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 (c
12
=c
13
=c
23
=0).
Applying Eqs. (1), (2), and (4), we have:
E
11
u
r
11
e
11
15a
v
12
u
e
22
e
11
15b
v
13
u
e
33
e
11
15c
g
11;12
g
11;13
g
11;23
0 15d
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Because no shear strains are induced by r
11
, the coecients of mutual
inuence 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 coecients
of mutual inuence 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 r
22
is applied are:
E
22
u
r
22
e
22
16a
v
21
u
e
11
e
22
16b
v
23
u
e
33
e
22
16c
g
22;12
g
22;13
g
22;23
0 16d
Figure 6 Deformations induced in a unidirectional composite by uniaxial stress
r
11
(deformations are shown greatly exaggerated for clarity).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Once again, due to the symmetrical distribution of bers, stress r
22
does
not induce any shear strains, so the coecients of mutual inuence of the
second kind all equal zero.
As previously mentioned, due to the thin platelike nature of composites,
it is dicult in practice to apply a well-dened out-of-plane uniaxial stress r
33
,
or to measure the resulting normal strain induced in the out-of-plane direction
e
33
. Assuming that these practical diculties are overcome, the material
properties measured when stress r
33
is applied are:
E
33
u
r
33
e
33
17a
v
31
u
e
11
e
33
17b
v
32
u
e
22
e
33
17c
g
33;12
g
33;13
g
33;23
0 17d
For unidirectional composites, both the 2- and 3-axes are dened to be
perpendicular to the bers, and hence properties measured in the 2- and 3-
directions are typically similar in magnitude. In fact, if the distribution of
bers in the 2- and 3-directions is identical at the microlevel, then properties
measured in these directions will be equal: E
22
=E
33
, v
12
=v
13
, v
21
=v
31
, and
v
23
= v
32
. If this occurs, then the composite is classied as a transversely
isotropic material. In contrast, if the distribution of bers diers 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 classied as an orthotropic material.
Optical micrographs showing the ber distribution in the 23 plane for a
unidirectional graphitepolyimide laminate are shown in Fig. 7. Fig. 7a was
taken at a magnication of 150, and shows the ber distribution in four
adjacent plies. The ber angles are (from left to right) 0j, 45j, 90j, and 45j.
Fig. 7b was obtained for the same laminate but at higher magnication
(300), and shows ber 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 bers 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 bers in
the 2- and 3-directions diers substantially), then E
33
will dier from E
22
, and
the composite will respond as an orthotropic material.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 s
12
is
shown in Fig. 8. The deformed shape of the element is also shown. Note that:

The angle originally dened by the 12 axes has decreased, cor-


responding to a positive shear strain c
12
.
Figure 7 Optical micrographs of fibers within several plies of a [0/45/90/
45]
2s
graphitepolyimide (IM7/K3B) composite laminate. Note the resin-rich
zone between plies.
Figure 8 Deformations induced in a unidirectional composite by pure shear
stress s
12
.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

Due to the symmetrical distribution of bers 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 (c
13
= c
23
= 0).

The length, width, and thickness of the element have not changed;
hence, all normal strain are zero (e
11
= e
22
= e
33
= 0).
Applying Eqs. (8)(10), we have:
G
12
u
s
12
c
12
18a
l
12;13
l
12;23
0 18b
g
12;11
g
12;22
g
12;33
0 18c
Because only c
12
is induced by s
12
, the Chentsov coecients as well as
the coecients of mutual inuence of the rst 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 coecients nor the coecients of
mutual inuence of the rst 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 dicult to apply well-dened out-of-plane shear stress s
13
or s
23
, or to
measure the resulting shear strains induced in the out-of-plane direction c
13
or
c
23
. Assuming that these practical diculties were overcome, the material
properties measured when stress s
13
is applied are:
G
13
u
s
13
c
13
19a
l
13;12
l
13;23
0 19b
g
13;11
g
13;22
g
13;33
0 19c
If the bers are not uniformly distributed within the 23 plane, or if the
composite is based on woven or braided fabrics, then the composite will
behave as an orthotropic material and G
12
p G
13
. If the composite is based on
a unidirectional fabric and bers are uniformly distributed, then the compo-
site is transversely isotropic and G
12
= G
13
.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Following an analogous process, material properties measured when s
23
is applied are:
G
23
u
s
23
c
23
20a
l
23;12
l
23;23
0 20b
g
23;11
g
23;22
g
23;33
0 20c
A total of 12 material properties have been dened above for ortho-
tropic or transversely isotropic composites: three Youngs moduli (E
11
, E
22
,
and E
33
), six Poissons ratios (v
12
, v
13
, v
21
, v
23
, v
31
, and v
32
), and three shear
moduli ( G
12
, G
13
, and G
23
). 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 ve 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 ve 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 ve), in practice, only four of these
properties are ordinarily required: E
11
, E
22
, v
12
, and G
12
. 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 nal comment, an often overlooked fact is that the elastic proper-
ties of composites usually dier 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 E
22
measured in tension to dier 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
dierences during a structural analysis (e.g., see Ref. 3), the bimodulus
phenomenon is a signicant complication and will not be accounted for
herein. Throughout this text, it will be assumed that in-plane elastic properties
E
11
, E
22
, v
12
, and G
12
are identical in tension and compression. The reader
should be aware that these dierences usually exist, however. If in practice the
measured response of a composite structure diers from the predicted
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
behavior, the discrepancy may well be due to dierences 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: e
xx
T
, e
yy
T
, e
zz
T
,
c
xy
T
, c
xz
T
, and c
yz
T
. 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 coecients 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 bers are oriented at an angle of 45j with respect to the x-axis. Because the
composite is initially square and, in this example, the bers are dened to be at
an angle of 45j, bers are parallel to diagonal AC and are perpendicular to
diagonal BD. Now, the coecient of thermal expansion exhibited by high-
performance bers 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).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
most polymers, it is relatively high. For example, the coecient of thermal
expansion of graphite bers is about 1 Am/mjC, 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 gure. The
fact that angle BDABhas increased reveals that a shear strain c
xy
(in this case,
a negative shear strain c
xy
) 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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 T
g
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 T
g
; 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 coecients of thermal expansion must be dened:
a
xx
u
e
T
xx
DT
a
yy
u
e
T
yy
DT
a
zz
u
e
T
zz
DT
a
xy
u
c
T
xy
DT
a
xz
u
c
T
xz
DT
a
yz
u
c
T
yz
DT
21
Because we have limited our discussion to the linear range shown in Fig.
10, the coecients of thermal expansion dened by Eq. (21) equal the slopes
of the corresponding strain vs. DT curves within the linear range. These
properties will henceforth be called linear coecients 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:
e
T
xx
c
T
xy
c
T
xz
c
T
yx
e
T
yy
c
T
yz
c
T
zx
c
T
zy
e
T
zz
_

_
_

_
DT
a
xx
a
xy
a
xz
a
xy
a
yy
a
yz
a
xz
a
yz
a
zz
_

_
_

_
22
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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
3.1 Specialization to Orthotropic and Transversely Isotropic
Composites
We will now apply these general denitions 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 123 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 123 coordinate system is shown in Fig. 11.
Because the bers 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:
a
11

e
T
11
DT
a
22

e
T
22
DT
a
33

e
T
33
DT
a
12
a
13
a
23
0 23
As before, if the bers are distributed uniformly throughout the 23
plane, then a
22
=a
33
and the composite will be transversely isotropic. If the
bers are not uniformly distributed, then a
22
pa
33
, 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).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
the distribution of bers in the 2-direction diers substantially from that in
the 3-direction.
Experimental methods of measuring thermal expansion coecients for
polymeric composites have not been standardized by organizations such as
the ASTM. In practice, the in-plane thermal expansion coecients 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 coecients 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 diuse 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
f
0% to as high as
f
5%.
From a structural point of view, the eects 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 DMwould 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
be induced: e
xx
M
, e
yy
M
, e
zz
M
, c
xy
M
, c
xz
M
, and c
yz
M
. 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 coecients 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 dened:
b
xx
u
e
M
xx
DM
b
yy
u
e
M
yy
DM
b
zz
u
e
M
zz
DM
b
xy
u
c
M
xy
DM
b
xz
u
c
M
xz
DM
b
yz
u
c
M
yz
DM
24
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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:
e
M
xx
c
M
xy
c
M
xz
c
M
yx
e
M
yy
c
M
yz
c
M
zx
c
M
zy
e
M
zz
_

_
_

_
DM
b
xx
b
xy
b
xz
b
xy
b
yy
b
yz
b
xz
b
yz
b
zz
_

_
_

_
25
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:
b
11
u
e
M
11
DM
b
22
u
e
M
22
DM
b
33
u
e
M
33
DM
b
12
b
13
b
23
0
26
If the composite is based on a unidirectional fabric and bers are
distributed uniformly throughout the 23 plane, then b
22
=b
33
and the
composite will be transversely isotropic. If the bers are not uniformly
distributed, or if the composite is based on woven or braided fabrics, then
b
22
p b
33
and the composite will be orthotropic. Recommended experimental
methods of measuring linear coecients 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 briey review the denitions of
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 e
xx
measured during a uniaxial
test are shown in Fig. 12ac. These gures are similar to Fig. 4, except that
now: (a) only one of the strain components caused by r
xx
has been plotted
(specically, only e
xx
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 stressstrain curve is nearly linear at all stress levels,
up to and including the nal 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 nal fracture. In contrast, Fig. 12b and c shows the stressstrain
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 q
xx
caused by application of a tensile
uniaxial stress j
xx
. (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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Figure 12 Continued.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
to nal fracture is (in general) far higher than that of a brittle material. The
toughness of a material is dened as the area under the entire stressstrain
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 r
p
the percent oset yield strength r
y
,
the ultimate strength r
u
, and the fracture strength r
f
. Although the axial strain
that corresponds to each of these stress levels can also be dened as a measure
of strength, typically only the percent oset yield strain e
y
and the fracture
strain e
f
are customarily used in this fashion. Many materials exhibit distinctly
dierent 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 stressstrain curves measured in tension, the superscript asso-
ciated with all measures of strength that appear in these gures includes the
character T.
The proportional limit r
pT
xx
is dened 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 dicult to precisely identify
the proportional limit based on experimental measurements. In these cases,
the stress level at which signicant nonlinear behavior begins is dened based
on the oset method. A line parallel to the initial (linear) region of the
stressstain curve is drawn on the stressstrain curve, but is oset along the
strain axis by some standard amount. The stress level at which this oset line
intersects the experimental stressstrain curve is dened as the percent oset
yield strength r
yT
xx
. A strain oset 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% oset yield strength or
the 0.2% oset yield strength, respectively. The ultimate strength r
uT
xx
is
dened as the maximum stress the material can withstand. The fracture stress
r
fT
xx
is dened as the stress that exists at nal 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 dened and only
the fracture strength (or ultimate strength) would be reported.
All of the above denitions may be applied to the data collected during a
pure shear test as well. An idealized plot of the shear strain c
xy
measured
during a pure shear test of a modestly ductile material is shown in Fig. 13. This
gure is similar to Fig. 5, except that now: (a) only one of the strain
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
components (c
xy
) caused by s
xy
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 dierent strengths in dier-
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
23 coordinate system. In the case of unidirectional composites, strengths are
measured parallel to the bers (i.e., parallel to the 1-axis) and transverse to the
bers (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
Figure 13 Idealized plot of the shear strain c
xy
caused by application of pure
shear stress s
xy
, including the shear fracture stress and strains s
xy
fP
and c
xy
fP
, and
the shear yield stress and strain s
xy
yP
and c
xy
yP
.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
bers, whereas strengths in the 2- and 3-directions are determined primarily
by the strength of the matrix. Most bers used in polymeric composites, such
as graphite or glass bers, 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 eorts 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 rst
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 r
11
vs. e
11
curve measured for most unidirectional polymeric compo-
sites resembles Fig. 12a. That is, in the ber direction, unidirectional compo-
sites exhibit nearly perfectly brittle behavior, although very modest nonlinear
behavior may begin to occur at stress levels approaching nal fracture. Due to
the limited nonlinear response, it is not appropriate to dene a yield stress in
the 1-direction. Throughout this text, it will be assumed that a uniaxial stress
in the ber direction r
11
causes failure due to fracture. That is, at failure, the
* The dierence between a thermoset and a thermoplastic polymer is discussed in Sec. 2.4 of
Chap. 1.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
specimen breaks into two or more pieces. The stress that causes fracture is
denoted r
11
fT
or r
11
fC
, depending on whether r
11
is tensile or compressive,
respectively. Similarly, the strain at failure is denoted e
11
fT
or e
22
fC
. Because very
little nonlinear behavior occurs prior to failure, it is possible to relate the
fracture stress and strain using Youngs 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 Youngs
modulus:
r
fT
11
iE
11
e
fT
11
r
fC
11
iE
11
e
fC
11

On the other hand, for most modern unidirectional polymeric compo-


sites, both the r
22
vs. e
22
and the r
33
vs. e
33
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 dened 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 (r
yT
22
; e
yT
22
; r
yT
33
; e
yT
33
or r
yC
22
; e
yC
22
; r
yC
33
; e
yC
33
; 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 Youngs modulus:
r
yT
22
iE
22
e
yT
22
r
yC
22
iE
22
e
yC
22

r
yT
33
iE
33
e
yT
33
r
yC
33
iE
33
e
yC
33

The stress and strain values present at fracture are denoted (r


22
fT
, e
22
fT
,
r
33
fT
, e
33
fT
) or r
fC
22
; e
fC
22
; r
fC
33
; e
fC
33
: Because most modern composites exhibit a
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 Youngs
modulus:
r
fT
22
< E
22
e
fT
22
r
fC
22
< E
22
e
fC
22

r
fT
33
< E
33
e
fT
33
r
fC
33
< E
33
e
fC
33

Shear stressstrain curves (i.e., a plot of s


12
vs. c
12
, s
13
vs. c
13
, or s
23
vs.
c
23
) 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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
system (e.g., in the 12 coordinate system), the shear response is insensitive to
the algebraic sign of the shear stress. That is, in the 12 coordinate system, an
identical s
12
vs. c
12
curve will be measured regardless of whether s
12
is positive
or negative.* As before, a shear failure may be dened 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
s
12
y
and c
12
y
(or s
13
y
and c
13
y
, or s
23
y
and c
23
y
), whereas the shear stress and strain at
fracture will be denoted s
12
f
and c
12
f
(or s
13
f
and c
13
f
, or s
23
f
and c
23
f
). As before,
the yield shear stress/strain may be related using the shear modulus, but the
shear fracture stress/strain may not:
s
y
12
iG
12
c
y
12
s
y
13
iG
13
c
y
13
s
y
23
iG
23
c
y
23

s
f
12
< G
12
c
f
12
s
f
13
< G
13
c
f
13
s
f
23
< G
23
c
f
23

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:
r
yT
22
r
yT
33
r
yC
22
r
yC
33
r
fT
22
r
fT
33
r
fC
22
r
fC
33
s
y
12
s
y
13
s
f
12
s
f
13
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 specically to failure of poly-
meric composites are listed in Table 2. Due to the thin platelike nature of
composites, it is dicult to apply well-dened 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 12 plane
for three common polymeric composites at roomtemperatures are included in
Table 3. These properties do not represent the properties of any specic
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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
strengths of composites are improved more or less continuously. Therefore,
the properties shown in Table 3 may not reect 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 ber angles.
As will be seen, failure predictions for a multiangle composite laminate
depend (in part) on the dierence 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
dierence 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 2 ASTM Test Standards Related to Failure of Polymeric Composites
(see also Table 1)
Designation Title
D3479 Standard Test Method for TensionTension 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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
nonreinforced bulk epoxy is commonly about 70 MPa (10 ksi), whereas from
Table 3, we see that graphiteepoxy 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 graphiteepoxy, the matrix-dominated tensile
strain at fracture (e
22
fT
) 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 [69]. Briey, two factors
lead to low matrix-dominated tensile strengths. The rst is thermal stresses
induced at the microlevel during cooldown from cure temperatures. Recall
that the thermal expansion coecient of most high-performance bers is very
low and, in fact, is often slightly negative. For example, thermal expansion
coecients of glass, Kevlar, and graphite bers are about 5, 2, and 0.5
Am/m jC, respectively (see Sec. 3 of Chap. 1). In contrast, the thermal ex-
pansion coecient 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 bers, leading to self-equilibrating
Table 3 Nominal Material Properties for Common Unidirectional Composites
Property Glass/epoxy Kevlar/epoxy Graphite/epoxy
E
11
55 GPa (8.0 Msi) 100 GPa (15 Msi) 170 GPa (25 Msi)
E
22
16 GPa (2.3 Msi) 6 GPa (0.90 Msi) 10 GPa (1.5 Msi)
v
12
0.28 0.33 0.30
G
12
7.6 GPa (1.1 Msi) 2.1 GPa (0.30 Msi) 13 GPa (1.9 Msi)
r
11
fT
1050 MPa (150 ksi) 1380 MPa (200 ksi) 1500 MPa (218 ksi)
r
11
fC
690 MPa (100 ksi) 280 MPa (40 ksi) 1200 MPa (175 ksi)
r
22
yT
45 MPa (5.8 ksi) 35 MPa (2.9 ksi) 50 MPa (7.25 ksi)
r
22
yC
120 MPa (16 ksi) 105 MPa (15 ksi) 100 MPa (14.5 ksi)
r
22
fT
55 MPa (7.0 ksi) 45 MPa (4.3 ksi) 70 MPa (10 ksi)
r
22
fC
140 MPa (20 ksi) 140 Msi (20 ksi) 130 MPa (18.8 ksi)
s
12
y
40 MPa (4.4 ksi) 40 MPa (4.0 ksi) 75 MPa (10.9 ksi)
s
12
f
70 MPa (10 ksi) 60 MPa (9 ksi) 130 MPa (22 ksi)
a
11
6.7 Am/m jC
(3.7 Ain./in. jF)
3.6 Am/m jC
(2.0 Ain./in. jF)
0.9 Am/m jC
(0.5 Ain./in. jF)
a
22
25 Am/m jC
(14 Ain./in. jF)
58 Am/m jC
(32 Ain./in. jF)
27 Am/m jC
(15 Ain./in. jF)
b
11
100 Am/m %M
(100 Ain./in. %M)
175 Am/m %M
(175 Ain./in. %M)
50 Am/m %M
(50 Ain./in. %M)
b
22
1200 Am/m %M
(1200 Ain./in. %M)
1700 Am/m %M
(1700 Ain./in. %M)
1200 Am/m %M
(1200 Ain./in. %M)
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
tensile stresses in the matrix and compressive stresses in the bers. The
second factor is the stress-concentrating eect of the bers. Because most
advanced composites are produced with a ber volume fraction of about
0.65, the tensile stresses induced within the matrix surrounding a ber are not
dictated strictly by the CTE mismatch between matrix and ber but are also
inuenced by the presence neighboring bers. Together, these two factors
give rise to thermal stresses at the microlevel that are generally tensile in the
matrix and compressive in the bers. The magnitude of the thermal stresses
induced in the ber is very low relative to the ber 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 nite element method have shown that
thermal matrix stresses can often be 5060% of the bulk matrix tensile
strength [1114]. 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-
dominated tensile strengths (i.e., r
22
yC
>r
22
yT
and r
22
fC
>r
22
fT
).
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 ber and matrix). As a typical
example, suppose a new high-performance graphite ber has recently been
developed, and properties of the ber itself have been measured. Naturally,
the structural engineer is interested in determining whether this new ber 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 ber 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 ber exists, so as to justify the time and money that will
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
be invested during the development of the composite material system based
on the new graphite ber.
As described in Sec. 6 of Chap. 1, an analysis performed at a physical
scale corresponding to the ber diameter is classied 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 bers embedded within a
polymeric matrix. The principal material coordinate system, labeled the 1
23 coordinate system, is dened by the ber direction. It is assumed that
the bers are evenly spaced, and that the matrix is perfectly bonded to the
ber.
If a force F
11
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 F
11
is related to the average stress
imposed in the 1-direction by F
11
=r
11
A, where A is the cross-sectional area
of the element. Furthermore, the sumof forces present in the matrix and bers
must equal the total applied force, which implies:
r
11
A r
f
A
f
r
m
A
m
27
where A
f
is the total cross-sectional area of the bers presented within the
element and A
m
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 ber
and matrix because the ber and matrix are assumed to be perfectly bonded.
That is:
e
f
e
m
e
11

DL
L
Stresses are assumed to be related to strains according to:
r
11
e
11
E
11
28a
r
f
e
f
E
f
e
11
E
f
28b
r
m
e
m
E
m
e
11
E
m
28c
The expressions for stresses r
f
and r
m
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 ber and matrix properties as
well as the presence of adjacent bers. 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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Figure 14 Representative composite element used to derive rule-of-mixtures
equations. (a) Composite element deformed by a load F
11
, acting parallel to the
fiber direction. (b) Composite element deformed by a load F
22
, acting perpen-
dicular to the fiber direction. (c) Composite element deformed by shear load F
12
.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
ber and matrix stresses are given by Eqs. (28b) and (28c) despite their
shortcomings. Substituting these expressions into Eq. (27) and rearranging,
we nd:
E
11
E
f
A
f
A
E
m
A
m
A
This expression allows us to predict E
11
based on properties of the
constituents (E
f
and E
m
) and the area fractions of ber and matrix (A
f
/A)
and (A
m
/A). If no voids are present, then:
A A
f
A
m
Usually, rule-of-mixtures expressions are written in terms of volume
fractions rather than area fractions. Volume fractions are given by:
V
f

A
f
A
V
m

A
m
A

A A
f
A
1 V
f

where V
f
is the volume fraction of bers and V
m
= (1V
f
) is the volume
fraction of matrix material. Consequently, the predicted value of E
11
based on
the rule of mixtures approach is given by:
E
11
E
m
V
f
E
f
E
m
29
Polymeric composites used in practice are typically produced with a
ber volume fraction V
f
of about 0.65, although it can be lower (say, V
f
=
0.30), depending on application and the manufacturing process used to
consolidate the composite. Equation (29) shows that if E
f
E
m
(which is
usually the case), then to a rst approximation, E
11
cV
f
E
f
. The value of E
11
is dictated primarily by the ber modulus E
f
and ber volume fraction V
f
. E
11
is therefore called a ber-dominated property of the composite.
Now consider the Poisson eect exhibited by the composite element
shown in Fig. 14a. As per our normal denition, the average Poisson ratio is
dened as the negative of the transverse normal strain (e
22
) divided by axial
normal strain (e
11
), both of which are caused by r
11
:
v
12

e
22
e
11
The transverse normal strain associated with the change in width of the
entire element (DW) is given by:
e
22

DW
W
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
The change in width can also be written as the sum of the change in
width of the bers present in the element DW
f
and the change in width of the
matrix present DW
m
. These are approximated as follows:
DW
f
WV
f
v
f
e
f
WV
f
v
f
e
11
DW
m
WV
m
v
m
e
m
WV
m
v
m
e
11
where v
f
and v
m
are Poisson ratios of the ber and matrix, respectively.
Hence, the transverse strain is given by:
e
22

DW
W
V
f
v
f
V
m
v
m
e
11
Applying the denition of Poissons ratio for the composite as a whole,
we have:
v
12

e
22
e
11
V
f
v
f
V
m
v
m
Noting as before that V
m
=(1V
f
), the predicted value for Poissons
ratio v
12
based on the rule of mixtures becomes:
v
12
v
m
V
f
v
m
v
f
30
Measurement of Poissons ratio of the matrix material v
m
is a straight-
forward matter. However, measuring Poissons ratio of the ber v
f
is more
dicult due to the small ber diameters involved. Experimentally measured
values of v
f
are often unavailable, even for bers widely used in practice. The
data that are available imply that both v
m
and v
f
are algebraically positive,
and also that v
m
>v
f
. Hence, Eq. (30) implies that the composite Poisson ra-
tio v
12
varies linearly with ber volume fraction V
f
, and that Poissons ratio
of the composite is less than that of the matrix (v
12
<v
m
) because usually
(v
m
v
f
)>0.
Assuming an identical ber distribution in the 12 and 13 planes, then
an identical analysis can be conducted to predict Poissons ratio v
13
, which will
result in an identical expression: v
13
=v
12
.
Next, consider prediction of the transverse modulus E
22
based on the
rule-of-mixtures approach. A composite element subjected to a force applied
perpendicular to the bers, force F
22
, is shown in Fig. 14b. This force is related
to the average stress imposed in the 2-direction by F
22
= r
22
A. We assume
that an identical and uniformstress r
22
is induced in both the ber and matrix.
Once again, this assumption is approximate at best; in reality, a triaxial state
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
of stress is induced in both the ber and matrix. Based on this assumption, the
strains induced in the ber and matrix perpendicular to the 1-axis are:
e
f

r
22
E
f
e
m

r
22
E
m
The transverse length represented by the bers present in the element
equals V
f
W, whereas the transverse length represented by the matrix equals
V
m
W. Hence, the change in width W caused by the application of r
22
is:
DW V
f
W
r
22
E
f
V
m
W
r
22
E
m
The average transverse strain caused by r
22
is:
e
22

DW
W
r
22
V
f
E
f

V
m
E
m
_ _
Youngs modulus E
22
as predicted by the rule of mixtures therefore
becomes:
E
22

r
22
e
22

1
V
f
E
f

V
m
E
m
_ _

E
f
E
m
E
m
V
f
E
f
V
m
As before, if no voids are present, then V
m
=(1V
f
), and we obtain:
E
22

E
f
E
m
E
f
V
f
E
f
E
m

31
For most polymeric composite material systems, E
f
> >E
m
. Nevertheless,
Eq. (31) shows that E
22
is dictated primarily by E
m
, and is only modestly
aected by the ber modulus E
f
. Indeed, even in the limit (i.e., as E
f
!l), the
predicted value of E
22
is only increased to:
E
22 E
f
!l

E
m
1 V
f

Because V
f
is usually about 0.65, this result shows that E
22
is still less
than three times the matrix modulus E
m
, even if the composite is produced
using a ber whose stiness is innitely high (E
f
!l). E
22
is therefore called
a matrix-dominated property of the composite.
Assuming an identical ber distribution in the 12 and 13 planes,
then an identical analysis can be conducted to predict Youngs modulus
in the 3-direction E
33
, resulting in an identical expression. Hence, E
33
=E
22
.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
As before, E
33
is dictated primarily by E
m
and is a matrix-dominated
property.
Now consider the shear modulus G
12
. An element subjected to a pure
shear force F
12
is shown in Fig. 14c. This force is related to the average shear
stress according to F
12
=s
12
A. In a rule-of-mixture analysis, it is assumed that
an identical shear stress is induced in both the bers and matrix regions. This
assumption is approximate at best. Nevertheless, on the basis of this
assumption, the shear strains induced in ber and matrix are given by:
c
f

s
12
G
f
c
m

s
12
G
m
where G
f
and G
m
are the shear moduli of the ber and matrix, respectively.
The total shear strain is given by:
c
12
V
f
c
f
V
m
c
m
V
f
s
12
G
f
_ _
V
m
s
12
G
m
_ _
The shear modulus predicted by the rule of mixtures is then:
G
12

s
12
c
12

1
V
f
G
f

V
m
G
m
_ _

G
f
G
m
G
m
V
f
G
f
V
m
Assuming no voids are present, then V
m
=(1V
f
), and we obtain:
G
12

G
f
G
m
G
f
V
f
G
f
V
m

32
Comparing Eq. (32) with Eq. (31), it is seen that the shear modulus is
related to ber and matrix properties in a manner similar to E
22
and E
33
. The
value of G
12
is dictated primarily by the shear modulus of the matrix G
m
, and
is considered a matrix-dominated property. Assuming an identical ber
distribution in the 12 and 13 planes, then G
13
=G
12
.
To summarize, the analysis presented above allows prediction of elastic
moduli E
11
, v
12
=v
13
, E
22
=E
33
, and G
12
=G
13
, based on knowledge of the
ber modulus E
f
, matrix modulus E
m
, and ber volume fraction V
f
. Although
not presented here, a rule-of-mixture approach can also be used to predict
thermal expansion coecients a
1
and a
2
, or moisture expansion coecients b
1
and b
2
(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,
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
E
11
and v
12
=v
13
are reasonably well predicted by Eqs. (29) and (30), respec-
tively. However, matrix-dominated properties E
22
=E
33
and G
12
=G
13
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 bers present in


a real composite

The triaxial state of stress induced in both matrix and ber due to the
mismatch in ber/matrix properties.

Dierences in ber distribution in the 12 and 13 planes

The adhesion (or lack thereof ) between ber and matrix

Variations in ber cross-sections from one ber to the next

The presence of voids or other defects

The anisotropic nature of many high-performance bers (e.g.,


Youngs modulus parallel and transverse to the long axis of the
ber usually diers).
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
nite 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, 913.
HOMEWORK PROBLEMS
1. An orthotropic material is known to have the following elastic properties:
E
xx
100GPa E
yy
200GPa E
zz
75GPa
v
xy
0:20 v
xz
0:25 v
yz
0:60
G
xy
60GPa G
xz
75GPa G
yz
50GPa
g
xx;xy
0:30 g
xx;xz
0:25 g
xx;yz
0:30
g
yy;xy
0:60 g
yy;xz
0:75 g
yy;yz
0:20
g
zz;xy
0:20 g
zz;xz
0:05 g
zz;yz
0:15
l
xy;xz
0:10 l
xy;yz
0:05 g
xz;yz
0:10
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
(a) What strains are induced if a uniaxial tensile stress r
xx
=300 MPa is
applied?
(b) What strains are induced if a uniaxial tensile stress r
yy
=300 MPa is
applied?
(c) What strains are induced if a uniaxial tensile stress r
zz
=300 MPa is
applied?
(d) What strains are induced if a pure shear stress s
xy
=100 MPa is
applied?
(e) What strains are induced if a pure shear stress s
xz
=100 MPa is
applied?
(f ) What strains are induced if a pure shear stress s
yz
=100 MPa is
applied?
2. An orthotropic material is known to have the following elastic properties:
E
11
100GPa E
22
200GPa E
33
75GPa
v
12
0:20 v
13
0:25 v
23
0:60
G
12
60GPa G
13
75GPa G
23
50GPa
(a) What strains are induced if a uniaxial tensile stress r
11
=300 MPa is
applied?
(b) What strains are induced if a uniaxial tensile stress r
22
=300 MPa is
applied?
(c) What strains are induced if a uniaxial tensile stress r
33
=300 MPa is
applied?
(d) What strains are induced if a pure shear stress s
12
=100 MPa is
applied?
(e) What strains are induced if a pure shear stress s
13
=100 MPa is
applied?
(f ) What strains are induced if a pure shear stress s
23
=100 MPa is
applied?
3. A tensile specimen is machined from an anisotropic material. The speci-
men is referenced to an xyz 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 xy and xz
surfaces of the specimen, as shown. A uniaxial tensile stress r
xx
=700
MPa is then applied, causing the specimen to deform as shown in Fig.
15bd. Determine the values of E
xx
, v
xy
, v
xz
, g
xx
,
xy
, g
xx,xz
, and g
xx,yz
that
correspond to these deformations.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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
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 xy
face. (d) Change in dimensions on xz face.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
4. Load vs. strain data collected during two dierent composite tensile tests
are shown in Tables 4 and 5. Use linear regression to determine the
following properties for this composite material:
(a) Determine E
11
and v
12
using the data collected using the [0]
8
specimen (Table 4).
(b) Determine E
22
using the data collected using the [90]
16
specimen (Table 5).
(c) Determine the value of v
21
for this composite material system.
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
Figure 16 Tensile specimen described in Problem 5.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
5. A thin tensile specimen is machined from a material with unknown
properties. A perfect square with dimensions 55 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 E
xx
, v
xy
,
and g
xx,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 e
xx
, e
yy
, and e
xy
, and coecients of thermal
expansion a
xx
, a
yy
, and a
xy
.
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 Coecient, 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;
Figure 17 Deformed square described in Problem 6.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 uniformchanges in moisture content. The chapter begins
with a consideration of the strains induced by stress under constant environ-
mental conditions. A generalized form of Hookes law, which relates strain
to stress for any anisotropic material, will be developed. Next, Hookes 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 rst, and will then be specialized to the case of
orthotropic and transversely isotropic materials.
Finally, the strains induced by the combined eects of stress, temper-
ature, and moisture will be discussed.
1 STRAINS INDUCED BY STRESS: ANISOTROPIC
MATERIALS
Areviewof 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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
r
ij
, where subscripts i and j take on values of x, y, and z. Alternatively, the
stress tensor can be written using matrix notation as:
r
ij

r
xx
r
xy
r
xz
r
yx
r
yy
r
yz
r
zx
r
zy
r
zz
2
4
3
5
1
Because the stress tensor is symmetrical (i.e., r
yx
=r
xy
, r
zx
=r
xz
, and
r
zy
= r
yz
), only six independent stress components appear in Eq. (1).
Similarly, strain is a symmetrical second-order tensor e
ij
and can be
written as:
e
ij

e
xx
e
xy
e
xz
e
yx
e
yy
e
yz
e
zx
e
zy
e
zz
2
6
4
3
7
5
e
xx
c
xy
=2 c
xz
=2
c
yx
=2 e
yy
c
yz
=2
c
zx
=2 c
zy
=2 e
zz

2
6
4
3
7
5 2
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,
e
xy
e
yx

c
xy
2

c
yx
2
; e
xz
e
zx

c
xz
2

c
zx
2
; e
yz
e
zy

c
yz
2

c
zy
2
3
For any elastic solid, the strain and stress tensors are related as follows
(assuming temperature and moisture content remain constant):
e
ij
S
ijkl
r
kl
4
All subscripts that appear in Eq. (4) take on values of x, y, and z.
Equation (4) is called generalized Hookes 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 S
ijkl
. Because strains
are unitless quantities, from Eq. (4), it is seen that the units of S
ijkl
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 3
4
=81 independent terms appear within the
compliance tensor. However, due to symmetry of both the strain and stress
tensors, it will be shown belowthat the compliance tensor consists of (at most)
36 material constants.
It will be very convenient to express Eq. (4) using matrix notation.
However, because S
ijkl
is a fourth-order tensor (and hence can be viewed as
having four dimensions), we cannot expand S
ijkl
as a two-dimensional
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
matrix directly. To expand Eq. (4), we must rst dene the components of
stress and strain using contracted notation, as follows:
e
xx
! e
1
r
xx
! r
1
e
yy
! e
2
r
yy
! r
2
e
zz
! e
3
r
zz
! r
3
c
yz
c
zy
! e
4
r
yz
r
zy
! r
4
c
xz
c
zx
! e
5
r
xz
r
zx
! r
5
c
xy
c
yx
! e
6
r
xy
r
yx
! r
6
5
Notice that the symmetry of the strain and stress tensors (c
yz
=c
xz
, etc.)
is embedded within the very denition of contracted notation. Also note that
the shear strain components (e
4
, e
5
, and e
6
) represent engineering shear strains,
rather than tensoral shear strains. Based on this change in notation, we can
now write Eq. (4) as:
e
j
S
ij
r
j
where i; j 16 6
In contracted notation, the strain and stress tensors are expressed with a
single subscript (i.e., e
i
and r
j
), and hence in Eq. (6), they appear to be rst-
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
S
ij
as the compliance matrix, and the use of contracted notation will be
implied.
Expanding Eq. (6), we have:
e
1
e
2
e
3
e
4
e
5
e
6
8
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
;

S
11
S
12
S
13
S
14
S
15
S
16
S
21
S
22
S
23
S
24
S
25
S
26
S
31
S
32
S
33
S
34
S
35
S
36
S
41
S
42
S
43
S
44
S
45
S
46
S
51
S
52
S
53
S
54
S
55
S
56
S
61
S
62
S
63
S
64
S
65
S
66
2
6
6
6
6
6
6
6
6
4
3
7
7
7
7
7
7
7
7
5
r
1
r
2
r
3
r
4
r
5
r
6
8
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
;
7
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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
itself be symmetrical. That is, all terms in symmetrical o-diagonal positions
must be equal:
S
21
S
12
S
31
S
13
S
32
S
23
S
41
S
14
S
42
S
24
S
43
S
34
S
51
S
15
S
52
S
25
S
53
S
35
S
54
S
45
S
61
S
16
S
62
S
26
S
63
S
36
S
64
S
46
S
65
S
56
8
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 (dened in Eq. (5)) into
Eq. (7), we have:
e
xx
e
yy
e
zz
c
yz
c
xz
c
xy
8
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
;

S
11
S
12
S
13
S
14
S
15
S
16
S
21
S
22
S
23
S
24
S
25
S
26
S
31
S
32
S
33
S
34
S
35
S
36
S
41
S
42
S
43
S
44
S
45
S
46
S
51
S
52
S
53
S
54
S
55
S
56
S
61
S
62
S
63
S
64
S
65
S
66
2
6
6
6
6
6
6
6
6
4
3
7
7
7
7
7
7
7
7
5
r
xx
r
yy
r
zz
r
yz
r
xz
r
xy
8
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
;
9
The individual constants that appear in the compliance matrix can be
easily related to the material properties dened 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 e
xx
caused by each stress component independently are given by:
r
xx
causes from Eq: 3:1 e
xx

1
E
xx
r
xx
r
yy
causes from Eq: 3:6 e
xx

m
yx
E
yy
r
yy
r
zz
causes from Eq: 3:7 e
xx

m
zx
E
zz
r
zz
s
yz
causes from Eq: 3:14 e
xx

g
yz;xx
G
yz
s
yz
s
xz
causes from Eq: 3:13 e
xx

g
xz;xx
G
xz
s
xz
s
xy
causes from Eq: 3:12 e
xx

g
xy;xx
G
xy
s
xy
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
To determine the strain e
xx
induced if all stress components act
simultaneously, simply add up the contribution to e
xx
caused by each stress
individually, to obtain:
e
xx

1
E
xx

r
xx

m
yx
E
yy

r
yy

m
zx
E
zz

r
zz

g
yz;xx
G
yz

s
yz

g
xz;xx
G
xz

s
xz

g
xy;xx
G
xy

s
xy
10a
Using an identical procedure, the remaining ve strain components
caused by an arbitrary combination of stresses are:
e
yy

m
xy
E
xx

r
xx

1
E
yy

r
yy

m
zy
E
zz

r
zz

g
xz;yy
G
yz

s
yz

g
xz;yy
G
xz

s
xz

g
xy;yy
G
xy

s
xy
10b
e
zz

m
xz
E
xx

r
xx

m
yz
E
yy

r
yy

1
E
zz

r
zz

g
yz;zz
G
yz

s
yz

g
xz;zz
G
xz

s
xz

g
xy;zz
G
xy

s
xy
10c
c
yz

g
xx;yz
E
xx

r
xx

g
yy;yz
E
yy

r
yy

g
zz;yz
E
zz

r
zz

1
G
yz

s
yz

l
xz;yz
G
xz

s
xz

l
xy;yz
G
xy

s
xy
10d
c
xz

g
xx;xz
E
xx

r
xx

g
yy;xz
E
yy

r
yy

g
zz;xz
E
zz

r
zz

l
yz;xz
G
yz

s
yz

1
G
xz

s
xz

l
xy;xz
G
xy

s
xy
10e
c
xy

g
xx;xy
E
xx

r
xx

g
yy;xy
E
yy

r
yy

g
zz;xy
E
zz

r
zz

l
yz;xy
G
yz

s
yz

l
xz;xy
G
xz

s
xz

1
G
xy

s
xy
10f
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Equation (10a)(10f) can be assembled in matrix form:
pc
e
xx
e
yy
e
zz
c
yz
c
xz
c
xy
8
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
;

1
E
xx

m
yx
E
yy

m
zx
E
zz

g
yz;xx
G
yz

g
xz;xx
G
xz

g
xy;xx
G
xy

m
xy
E
xx

1
E
yy

m
zy
E
zz

g
yz;yy
G
yz

g
xz;yy
G
xz

g
xy;yy
G
xy

m
xz
E
xx

m
yz
E
yy

1
E
zz

g
yz;zz
G
yz

g
xz;zz
G
xz

g
xy;zz
G
xy

g
xx;yz
E
xx

g
yy;yz
E
yy

g
zz;yz
E
zz

1
G
yz

l
xz;yz
G
xz

l
xy;yz
G
xy

g
xx;xz
E
xx

g
yy;xz
E
yy

g
zz;xz
E
zz

l
yz;xz
G
yz

1
G
xz

l
xy;xz
G
xy

g
xx;xy
E
xx

g
yy;xy
E
yy

g
zz;xy
E
zz

l
yz;xy
G
yz

l
xz;xy
G
xz

1
G
xy

2
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
4
3
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
5
r
xx
r
yy
r
zz
s
yz
s
xz
s
xy
8
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
;
11
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:
S
11

1
E
xx
S
22

1
E
yy
S
33

1
E
zz
S
44

1
G
yx
S
55

1
G
xz
S
66

1
G
xy
S
21
S
12

m
xy
E
xx

m
yx
E
yy
S
31
S
13

m
xz
E
xx

m
zx
E
zz
S
32
S
23

m
yz
E
yy

m
zy
E
zz
S
41
S
14

g
xx;yz
E
xx

g
yz;xx
G
yz
S
42
S
24

g
yy;yz
E
yy

g
yz;yy
G
yz
S
43
S
34

g
zz;yz
E
zz

g
yz;zz
G
yz
S
51
S
15

g
xx;xz
E
xx

g
xz;xx
G
xz
S
52
S
25

g
yy;xz
E
yy

g
xz;yy
G
xz
S
53
S
35

g
zz;xz
E
zz

g
xz;zz
G
xz
S
54
S
45

l
yz;xz
E
yz

l
xz;yz
G
xz
S
61
S
16

g
xx;xy
E
xx

g
xy;xx
G
xy
S
62
S
26

g
yy;xy
E
yy

g
xy;yy
G
xy
S
63
S
36

g
zz;xy
E
zz

g
xy;zz
G
xy
S
64
S
46

l
yz;xy
E
yz

l
xy;yz
G
xy
S
65
S
56

l
xz;xy
G
xz

l
xy;xz
G
xy
12
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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:
m
xz
E
xx

m
yx
E
yy
m
xz
E
xx

m
zx
E
zz
m
yz
E
yy

m
zy
E
zz
g
xx;yz
E
xx

g
yz;xx
G
yz
g
yy;yz
E
yy

g
yz;yy
G
yz
g
zz;yz
E
zz

g
yz;zz
G
yz 13
g
xx;xz
E
xx

g
xz;xx
G
xz
g
yy;yz
E
yy

g
yz;yy
G
yz
g
zz;xz
E
zz

g
xz;zz
G
xz
l
yz;xz
G
yz

l
xz;yz
G
xz
g
xx;xy
E
xx

g
xy;xx
G
xy
g
yy;xy
E
yy

g
xy;yy
G
xy
g
zz;xy
E
zz

g
xy;zz
G
xy
l
yz;xy
G
yz

l
xy;yz
G
xy
l
xz;xy
G
xz

l
xy;xz
G
xy
The inverse relationships are very signicant 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. Specically, 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 Hookes law in the form of strainstress
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 stressstrain form of Hookes
law. The stressstrain form of Hookes law can be obtained by simply
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
inverting previous results. For example, inverting Hookes law given by Eq.
(6), we have:
r
i
C
ij
e
j
where i; j 16 14
C
ij
is called the stiness matrix.* The stiness matrix is the mathe-
matical inverse of the compliance matrix C
ij
=S
ij
1
. In expanded form, Eq.
(14) is written as:
r
1
r
2
r
3
r
4
r
5
r
6
8
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
;

C
11
C
12
C
13
C
14
C
15
C
16
C
21
C
22
C
23
C
24
C
25
C
26
C
31
C
32
C
33
C
34
C
35
C
36
C
41
C
42
C
43
C
44
C
45
C
46
C
51
C
52
C
53
C
54
C
55
C
56
C
61
C
62
C
63
C
64
C
65
C
66
2
6
6
6
6
6
6
6
6
4
3
7
7
7
7
7
7
7
7
5
e
1
e
2
e
3
e
4
e
5
e
6
8
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
;
15
The stiness matrix is symmetrical (C
21
=C
12
, C
31
=C
13
, etc.). The units
of each stiness 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:

Coecients of mutual inuence of the second kind:


g
11;12
g
11;13
g
11;23
g
22;12
g
22;13
g
22;23
g
33;12
g
33;13
g
33;23
0
* The variable names assigned to the compliance and stiness 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 stiness matrix.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

Coecients of mutual inuence of the rst kind:


g
12;11
g
12;22
g
12;23
g
13;11
g
13;22
g
13;33
g
23;11
g
23;22
g
23;33
0

Chentsov coecients:
l
12;13
l
12;23
l
13;12
l
13;23
l
23;12
l
23;13
0
Because these coupling terms do not exist, Hookes law for orthotropic
or transversely isotropic materials is simplied considerably relative to that of
an anisotropic material. For an orthotropic material, Hookes law becomes
(compare with Eq. (11)):
e
11
e
22
e
33
c
23
c
13
c
12
8
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
;

1
E
11

m
21
E
22

m
31
E
33

0 0 0
m
21
E
11

1
E
22

m
32
E
33

0 0 0
m
13
E
11

m
23
E
22

1
E
33

0 0 0
0 0 0
1
G
23

0 0
0 0 0 0
1
G
13

0
0 0 0 0 0
1
G
12

2
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
4
3
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
5
r
11
r
22
r
33
s
23
s
13
s
12
8
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
;
Alternatively, Eq. (16) may be
16
written as:
e
11
e
22
e
33
c
23
c
13
c
12
8
>
>
>
>
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
>
>
>
>
;

S
11
S
12
S
13
0 0 0
S
12
S
22
S
23
0 0 0
S
13
S
23
S
33
0 0 0
0 0 0 S
44
0 0
0 0 0 0 S
55
0
0 0 0 0 0 S
66
2
6
6
6
6
6
6
6
6
6
6
6
6
4
3
7
7
7
7
7
7
7
7
7
7
7
7
5
r
11
r
22
r
33
s
23
s
13
s
12
8
>
>
>
>
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
>
>
>
>
;
17
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
The fact that the compliance matrix must be symmetrical (S
21
=S
12
,
etc.) has been included in Eq. (17). Each compliance termin Eq. (17) is related
to the more familiar engineering properties as follows:
S
11

1
E
11
S
22

1
E
22
S
33

1
E
33
S
44

1
G
23
S
55

1
G
13
S
66

1
G
12
S
21
S
12

m
12
E
11

m
21
E
22
S
31
S
13

m
13
E
11

m
31
E
33
S
32
S
23

m
23
E
22

m
32
E
33
18
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:
S
11
; S
22
; S
33
; S
44
; S
55
; S
66
; S
12
; S
13
; and S
23

or, equivalently, as:


E
11
; E
22
; E
33
; m
12
; m
13
; m
23
; G
12
; G
13
; and G
23

Equation (17) is the strainstress form of Hookes law suitable for use
with orthotropic materials. To obtain a stressstrain relationship, Eq. (17) is
inverted, resulting in:
r
11
r
22
r
33
s
23
s
13
s
12
8
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
;

C
11
C
12
C
13
0 0 0
C
12
C
22
C
23
0 0 0
C
13
C
23
C
33
0 0 0
0 0 0 C
44
0 0
0 0 0 0 C
55
0
0 0 0 0 0 C
66
2
6
6
6
6
6
6
6
4
3
7
7
7
7
7
7
7
5
e
11
e
22
e
33
c
23
c
13
c
12
8
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
;
19
Individual components within the stiness matrix for an orthotropic
material are related to the compliance terms as follows:
C
11

S
22
S
33
S
2
23
S
C
12

S
13
S
23
S
12
S
33
S
C
13

S
12
S
33
S
13
S
22
S
C
22

S
11
S
33
S
2
13
S
C
23

S
12
S
13
S
11
S
23
S
C
33

S
11
S
22
S
2
12
S
C
44

1
S
44
C
55

1
S
55
C
66

1
S
66
20
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
where:
S S
11
S
22
S
33
S
11
S
2
23
S
22
S
2
13
S
33
S
2
12
2S
12
S
13
S
23
Alternatively, the stiness terms may be calculated using the elastic
properties described in Sec. 2 of Chap. 3:
C
11

E
22
m
2
23
E
33
E
2
11
X
C
12

m
12
E
22
m
13
m
23
E
33
E
11
E
22
X
C
13

m
12
m
23
m
13
E
11
E
22
E
33
X
C
22

E
11
m
2
13
E
33
E
2
22
X
21
C
23

m
23
E
11
m
12
m
13
E
22
E
22
E
33
X
C
33

E
11
m
2
12
E
22
E
22
E
33
X
C
44
G
23
C
55
G
13
C
66
G
12
where:
X E
11
E
22
m
2
12
E
2
22
m
2
13
E
22
E
33
m
2
23
E
11
E
33
2m
12
m
13
m
23
E
22
E
33
Hookes law for transversely isotropic materials is simplied further-
more because in this case, E
22
=E
33
, m
12
=m
13
, m
21
=m
31
, m
23
=m
32
, and
G
12
=G
13
. Also, it can also be easily shown that for transversely isotropic
composites, G
23

E
22
21m
23

. Hence, for transversely isotropic composites, Eq.


(16) reduces to:
e
11
e
22
e
33
c
23
c
13
c
12
8
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
;

1
E
11

m
21
E
22

m
21
E
22

0 0 0
m
21
E
11

1
E
22

m
32
E
22

0 0 0
m
12
E
11

m
23
E
22

1
E
22

0 0 0
0 0 0
21 m
23

G
22

0 0
0 0 0 0
1
G
12

0
0 0 0 0 0
1
G
12

2
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
4
3
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
7
5
r
11
r
22
r
33
s
23
s
13
s
12
8
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
;
22
which may also be written as:
e
11
e
22
e
33
c
23
c
13
c
12
8
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
;

S
11
S
12
S
12
0 0 0
S
12
S
22
S
23
0 0 0
S
12
S
23
S
22
0 0 0
0 0 0 2S
22
S
23
0 0
0 0 0 0 S
66
0
0 0 0 0 0 S
66
2
6
6
6
6
6
6
6
4
3
7
7
7
7
7
7
7
5
r
11
r
22
r
33
s
23
s
13
s
12
8
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
;
23
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
It can be seen that only ve independent material constants exist for a
transversely isotropic material. The set of ve independent constants can be
viewed as:
S
11
; S
22
; S
66
; S
12
; and S
23

or, equivalently, as:


E
11
; E
22
; m
12
; m
23
; and G
12

Equation (23) is the strainstress form of Hookes law suitable for use
with transversely isotropic composites. To obtain a stressstrain relationship,
Eq. (23) is inverted, resulting in:
r
11
r
22
r
33
s
23
s
13
s
12
8
>
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
>
;

C
11
C
12
C
12
0 0 0
C
12
C
22
C
23
0 0 0
C
12
C
23
C
22
0 0 0
0 0 0 2C
22
C
23
0 0
0 0 0 0 C
66
0
0 0 0 0 0 C
66
2
6
6
6
6
6
6
6
6
6
4
3
7
7
7
7
7
7
7
7
7
5
e
11
e
22
e
33
c
23
c
13
c
12
8
>
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
>
;
24
Individual components within the stiness matrix for a transversely
isotropic composite are related to the compliance terms, as follows:
C
11

S
22
S
33
X
C
12

S
12
X
C
22

S
11
S
22
S
2
12
XS
22
S
23

C
23

S
2
12
S
11
S
23
XS
22
S
23

C
66

1
S
66
25
where:
V S
11
S
22
S
23
2S
2
12
Alternatively, the stiness terms may be calculated using the elastic
properties described in Sec. 2 of Chap. 3:
C
11

E
2
11
1 m
23

X
C
12

m
12
E
11
E
22
X
C
22

E
22
E
11
m
2
12
E
22

X1 m
23

C
23

E
22
m
23
E
11
m
2
12
E
22

X1 m
23

C
44

C
22
C
23
2

E
22
21 m
23

C
66
C
12
26
where:
X E
11
1 m
23
2m
2
12
E
22
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Example Problem 1
The properties of a composite material are known to be:
E
11
170 GPa E
22
10 GPa E
33
8 GPa
m
12
0:30 m
13
0:35 m
23
0:40
G
12
13 GPa G
13
10 GPa G
23
8 GPa
Note that nine distinct material properties have been specied, indicat-
ing that this composite material is orthotropic. Determine the strains caused
by the following state of stress:
r
11
r
22
r
33
s
23
s
13
s
12
8
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
;

350 MPa
35 MPa
15 MPa
30 MPa
10 MPa
25 MPa
8
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
;
Solution. Because the composite is orthotropic, strains are calculated using
Eq. (17), where each termwithin the compliance matrix is calculated using Eq.
(18):
S
11

1
E
11

1
170 GPa

5:88
10
12
Pa
S
22

1
E
22

1
10 GPa

100:0
10
12
GPa
S
33

1
E
33

1
8 GPa

125:0
10
12
Pa
S
44

1
G
23

1
8 GPa

125:0
10
12
Pa
S
55

1
G
13

1
10 GPa

100
10
12
Pa
S
66

1
G
12

1
13 GPa

76:9
10
12
Pa
S
21
S
12

m
12
E
11

0:030
170 GPa

1:76
10
12
Pa
S
31
S
13

m
13
E
11

0:35
170 GPa

2:06
10
12
Pa
S
32
S
23

m
23
E
22

0:40
10 GPa

40:0
10
12
Pa
In this case, Eq. (17) becomes:
e
11
e
22
e
33
c
23
c
13
c
12
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
9
>
>
>
>
>
>
=
>
>
>
>
>
>
;

5:88 1:76 2:06 0 0 0


1:76 100:0 40:0 0 0 0
2:06 40:0 125:0 0 0 0
0 0 0 125:0 0 0
0 0 0 0 100:0 0
0 0 0 0 0 76:9
2
6
6
6
6
6
6
4
3
7
7
7
7
7
7
5
1
10
12
Pa

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

350
35
15
30
10
25
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
9
>
>
>
>
>
>
=
>
>
>
>
>
>
;
10
6
Pa
1966 Am=m
2284 Am=m
246 Am=m
3750 Arad
1000 Arad
1923 Arad
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
9
>
>
>
>
>
>
=
>
>
>
>
>
>
;
Example Problem 2
The properties of a composite material are known to be:
E
11
25 Msi E
22
E
33
1:5 Msi
m
12
m
13
0:30 m
23
0:40
G
12
G
13
2:0 Msi
Note that only ve distinct material properties have been specied,
indicating that this composite material is transversely isotropic. Determine
the strains caused by the following state of stress:
r
11
r
22
r
33
s
23
s
13
s
12
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
9
>
>
>
>
>
>
=
>
>
>
>
>
>
;

50 ksi
5 ksi
2 ksi
4 ksi
1:5 ksi
3:5 ksi
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
9
>
>
>
>
>
>
=
>
>
>
>
>
>
;
Solution. Because the composite is transversely isotropic, strains are calcu-
lated using Eq. (22). Individual terms within the compliance matrix are:
S
11

1
E
11

1
25 Msi

40:0
10
9
psi
S
22
S
33

1
E
22

1
1:5 Msi

667
10
9
psi
S
44

1
G
23

21 m
23

E
22

21 0:40
1:5 Msi

1866
10
9
psi
S
55
S
66

1
G
12

1
2:0 Msi

500
10
9
psi
S
12
S
21
S
13
S
31

m
12
E
11

0:30
25 Msi

12:0
10
9
psi
S
23
S
32

m
23
E
22

0:40
1:5 Msi

266
10
9
psi
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Equation (17) becomes:
e
11
e
22
e
33
c
23
c
13
c
12
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
9
>
>
>
>
>
>
=
>
>
>
>
>
>
;

40:0 12:0 12:0 0 0 0


12:0 667 266 0 0 0
12:0 266 667 0 0 0
0 0 0 800 0 0
0 0 0 0 500 0
0 0 0 0 0 500
2
6
6
6
6
6
6
4
3
7
7
7
7
7
7
5
1
10
9
psi

50
5
2
4
1:5
3:5
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
9
>
>
>
>
>
>
=
>
>
>
>
>
>
;
10
3
psi
1966 Ain:=in:
2203 Ain:=in:
596 Ain:=in:
3200 Arad
750 Arad
1750 Arad
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
9
>
>
>
>
>
>
=
>
>
>
>
>
>
;
Example Problem 3
An orthotropic composite is subjected to a state of stress that causes the
following state of strain:
e
11
e
22
e
33
c
23
c
13
c
12
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
9
>
>
>
>
>
>
=
>
>
>
>
>
>
;

1500 Am=m
2000 Am=m
1000 Am=m
2500 Arad
500 Arad
2000 Arad
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
9
>
>
>
>
>
>
=
>
>
>
>
>
>
;
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 stiness 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 E
11
E
22
m
2
12
E
2
22
m
2
13
E
22
E
33
m
2
23
E
11
E
33
2m
12
m
13
m
23
E
22
E
33
X f17010 0:30
2
10
2
0:35
2
108 0:40
2
1708
20:300:350:40108gGPa
2
X 1457GPa
2
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
C
11

E
22
m
2
23
E
33
E
2
11
X

f10 GPa 0:40
2
8 GPag170 GPa
2
1457GPa
2
172:98 GPa
C
12

m
12
E
22
m
13
m
23
E
33
E
11
E
22
X

f0:3010 GPa 0:350:408 GPag170 GPa10 GPa


1457GPa
2
4:808 GPa
C
13

m
12
m
23
m
13
E
11
E
22
E
33
X

f0:300:40 0:35170 GPag10 GPa8 GPa


1457GPa
2
4:387 GPa
C
22

E
11
m
2
13
E
33
E
2
22
X

f170 GPa 0:35
2
8 GPag10 GPa
1457GPa
2
11:602 GPa
C
23

m
23
E
11
m
12
m
13
E
22
E
22
E
33
X

f0:40170 GPa 0:350:3510 GPag10 GPa8 GPa


1457GPa
2
3:792 GPa
C
33

E
11
m
2
12
E
22
E
22
E
33
X

f170 GPa 0:30


2
10 GPag10 GPa8 GPa
1457GPa
2
9:286 GPa
C
44
G
23
8 GPa C
55
G
13
10 GPa C
66
G
12
13 GPa
Applying Eq. (19), the stresses are:
r
11
r
22
r
33
s
23
s
13
s
12
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
9
>
>
>
>
>
>
=
>
>
>
>
>
>
;

172:98 4:808 4:387 0 0 0


4:808 11:602 3:792 0 0 0
4:387 3:792 9:286 0 0 0
0 0 0 8:0 0 0
0 0 0 0 10:0 0
0 0 0 0 0 13:0
2
6
6
6
6
6
6
4
3
7
7
7
7
7
7
5
GPa
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

1500 Am=m
2000 Am=m
1000 Am=m
2500 Arad
500 Arad
2000 Arad
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
9
>
>
>
>
>
>
=
>
>
>
>
>
>
;

245:5 MPa
19:78 MPa
10:29 MPa
20:0 MPa
5:00 MPa
26:0 MPa
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
9
>
>
>
>
>
>
=
>
>
>
>
>
>
;
Example Problem 4
A transversely isotropic composite is subjected to a state of stress that causes
the following state of strain:
e
11
e
22
e
33
c
23
c
13
c
12
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
9
>
>
>
>
>
>
=
>
>
>
>
>
>
;

1250 Ain:=in:
1000 Ain:=in:
500 Ain:=in:
2500 Arad
1000 Arad
2000 Arad
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
9
>
>
>
>
>
>
=
>
>
>
>
>
>
;
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 stiness 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 E
11
1 m
23
2m
2
12
E
22
X 25 Msi1 0:40 20:30
2
1:5 Msi 14:73 Msi
C
11

E
2
11
1 m
23

X

25 Msi
2
1 0:40
14:73 Msi
25:46 Msi
C
12

m
12
E
11
E
22
X

0:3025 Msi1:5 Msi
14:73 Msi
0:7637 Msi
C
22

E
22
E
11
m
2
12
E
22

X1 m
23


1:5 Msi

25 Msi 0:30
2
1:5 Msi

14:73 Msi1 0:40


1:809 Msi
C
23

E
22
m
23
E
11
m
2
12
E
22

X1 m
23

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

1:5 Msi

0:4025 Msi 0:30


2
1:5 Msi

14:73Msi1 0:40
0:7372 Msi
C
44

E
22
21 m
23


1:5 Msi
21 0:40
0:5357 Msi
C
66
G
12
2:0 Msi
Applying Eq. (24), the stresses are:
r
11
r
22
r
33
s
23
s
13
s
12
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
9
>
>
>
>
>
>
=
>
>
>
>
>
>
;

25:46 0:7637 0:7637 0 0 0


0:7637 1:809 0:07372 0 0 0
0:7637 0:7372 1:809 0 0 0
0 0 0 0:5357 0 0
0 0 0 0 2:0 0
0 0 0 0 0 2:0
2
6
6
6
6
6
6
4
3
7
7
7
7
7
7
5
Msi

1250 Ain:=in:
1000 Ain:=in:
500 Ain:=in:
2500 Arad
1000 Arad
2000 Arad
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
9
>
>
>
>
>
>
=
>
>
>
>
>
>
;

32:97 ksi
3:13 ksi
2:60 ksi
1:34 ksi
2:00 ksi
4:00 ksi
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
9
>
>
>
>
>
>
=
>
>
>
>
>
>
;
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 dened 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:
e
T
xx
e
T
xy
e
T
xz
e
T
yx
e
T
yy
e
T
yz
e
T
zx
e
T
zy
e
T
zz
2
6
4
3
7
5 DT
a
xx
a
xy
a
xz
a
yx
a
yy
a
yz
a
zx
a
zy
a
zz
2
6
4
3
7
5 repeated3:22
Similarly, strains caused by a change in moisture content are given by
Eq. (3.25), repeated here for convenience:
e
M
xx
c
M
xy
c
M
xz
c
M
yx
e
M
yy
c
M
yz
c
M
zx
c
M
zy
e
M
zz
2
6
4
3
7
5 DM
b
xx
b
xy
b
xz
b
xy
b
yy
b
yz
b
xz
b
yz
b
zz
2
6
4
3
7
5 repeated3:25
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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:
e
T
xx
e
T
yy
e
T
zz
c
T
yx
c
T
xz
c
T
xy
8
>
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
>
;
DT
a
xx
a
yy
a
zz
a
yz
a
xz
a
xy
8
>
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
>
;
and
e
M
xx
e
M
yy
e
M
zz
c
M
yx
c
M
xz
c
M
xy
8
>
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
>
;
DT
b
xx
b
yy
b
zz
b
yz
b
xz
b
xy
8
>
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
>
;
27
In the case of an orthotropic material a
12
=a
13
=a
23
=b
12
=b
13
=
b
23
=0, and Eq. (27) becomes:
e
T
11
e
T
22
e
T
33
c
T
23
c
T
13
c
T
12
8
>
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
>
;
DT
a
11
a
22
a
33
0
0
0
8
>
>
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
>
>
;
and
e
M
11
e
M
22
e
M
33
c
M
23
c
M
13
c
M
12
8
>
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
>
;
DT
b
11
b
22
b
33
0
0
0
8
>
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
>
;
28
In addition to these simplications, for a transversely isotropic material
with symmetry in the 23 plane, a
33
=a
22
and b
33
=b
22
. Hence, for a trans-
versely isotropic material, Eq. (28) becomes:
e
T
11
e
T
22
e
T
33
c
T
23
c
T
13
c
T
12
8
>
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
>
;
DT
a
11
a
22
a
22
0
0
0
8
>
>
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
>
>
;
and
e
M
11
e
M
22
e
M
33
c
M
23
c
M
13
c
M
12
8
>
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
>
;
DT
b
11
b
22
b
22
0
0
0
8
>
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
>
;
29
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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Sec. 2. Strains induced by a uniform change in temperature or moisture
content in the absence of stress was discussed for all three material
classications 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 eects 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 (e
ij
) is a nonlinear coupled function of these
three mechanisms:
e
ij
fr; T; M
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
dened:

Youngs modulus as the slope of linear region of the stressstrain


curve (Sec. 2 of Chap. 3)

The coecient of thermal expansion (CTE) as the slope of the linear


region of the strainDT curve (Sec. 3 of Chap. 3)

The coecient of moisture expansion (CME) as the slope of the


linear region of the strainDT 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
Youngs modulus. Similarly, an increase in moisture content often causes an
increase in CTEs and a decrease in Youngs modulus.
These coupling eects 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 Youngs 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 Youngs 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:
e
ij
e
r
ij
e
T
ij
e
M
ij
30
The superscripts j, T, and Mused in Eq. (30) indicate that the individual
components of strain are caused by the application of stress, by a uniform
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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):
e
xx
e
yy
e
zz
c
yz
c
xz
c
xy
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
9
>
>
>
>
>
>
=
>
>
>
>
>
>
;

S
11
S
12
S
13
S
14
S
15
S
16
S
21
S
22
S
23
S
24
S
25
S
26
S
31
S
32
S
33
S
34
S
35
S
36
S
41
S
42
S
43
S
44
S
45
S
46
S
51
S
52
S
53
S
54
S
55
S
56
S
61
S
62
S
63
S
64
S
65
S
66
2
6
6
6
6
6
6
4
3
7
7
7
7
7
7
5
r
xx
r
yy
r
zz
s
yz
s
xz
s
xy
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
9
>
>
>
>
>
>
=
>
>
>
>
>
>
;
DT
a
xx
a
yy
a
zz
a
yz
a
xz
a
xy
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
9
>
>
>
>
>
>
=
>
>
>
>
>
>
;
DM
b
xx
b
yy
b
zz
b
yz
b
xz
b
xy
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
9
>
>
>
>
>
>
=
>
>
>
>
>
>
;
31
Equation (31) allows the prediction of the strains induced by the
simultaneous eects 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:
r
xx
r
yy
r
zz
s
yz
s
xz
s
xy
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
9
>
>
>
>
>
>
=
>
>
>
>
>
>
;

C
11
C
12
C
13
C
14
C
15
C
16
C
21
C
22
C
23
C
24
C
25
C
26
C
31
C
32
C
33
C
34
C
35
C
36
C
41
C
42
C
43
C
44
C
45
C
46
C
51
C
52
C
53
C
54
C
55
C
56
C
61
C
62
C
63
C
64
C
65
C
66
2
6
6
6
6
6
6
4
3
7
7
7
7
7
7
5

e
xx
DTa
xx
DMb
xx
e
yy
DTa
yy
DMb
yy
e
zz
DTa
zz
DMb
zz
c
yz
DTa
yz
DMb
yz
c
xz
DTa
xz
DMb
xz
c
xy
DTa
xy
DMb
xy
8
>
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
>
;
32
where the stiness matrix C
ij
=S
ij
1
, as discussed in Sec. 1.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Following an analogous procedure, the strains induced in an ortho-
tropic material by the combined eects of stress, a uniform change in
temperature, and/or a uniform change in moisture content can be found by
superimposing Eqs. (17) and (28):
e
11
e
22
e
33
c
23
c
13
c
12
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
9
>
>
>
>
>
>
=
>
>
>
>
>
>
;

S
11
S
12
S
13
0 0 0
S
12
S
22
S
23
0 0 0
S
13
S
23
S
33
0 0 0
0 0 0 S
44
0 0
0 0 0 0 S
55
0
0 0 0 0 0 S
66
2
6
6
6
6
6
6
4
3
7
7
7
7
7
7
5
r
11
r
22
r
33
s
23
s
13
s
12
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
9
>
>
>
>
>
>
=
>
>
>
>
>
>
;
DT
a
11
a
22
a
33
0
0
0
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
9
>
>
>
>
>
>
=
>
>
>
>
>
>
;
DM
b
11
b
22
b
33
0
0
0
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
9
>
>
>
>
>
>
=
>
>
>
>
>
>
;
33
Inverting Eq. (33), we obtain:
r
11
r
22
r
33
s
23
s
13
s
12
8
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
;

C
11
C
12
C
13
0 0 0
C
12
C
22
C
23
0 0 0
C
13
C
23
C
33
0 0 0
0 0 0 C
44
0 0
0 0 0 0 C
55
0
0 0 0 0 0 C
66
2
6
6
6
6
6
6
6
6
4
3
7
7
7
7
7
7
7
7
5

e
11
DTa
11
DMb
11
e
22
DTa
22
DMb
22
e
33
DTa
33
DMb
33
c
23
c
13
c
12
8
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
;
34
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 123 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).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Finally, the strains induced in a transversely isotropic material by the
combined eects of stress, a uniformchange in temperature, and/or a uniform
change in moisture content can be found by superimposing Eqs. (23) and
(29):
e
11
e
22
e
33
c
23
c
13
c
12
8
>
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
>
;

S
11
S
12
S
12
0 0 0
S
12
S
22
S
23
0 0 0
S
12
S
23
S
22
0 0 0
0 0 0 2S
22
S
23
0 0
0 0 0 0 S
66
0
0 0 0 0 0 S
66
2
6
6
6
6
6
6
6
6
6
4
3
7
7
7
7
7
7
7
7
7
5
r
11
r
22
r
33
s
23
s
13
s
12
8
>
>
>
>
>
>
>
>
>
<
>
>
>
>
>
>
>
>
>
:
9
>
>
>
>
>
>
>
>
>
=
>
>
>
>
>
>
>
>
>
;
DT
a
11
a
22
a
22
0
0
0
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
9
>
>
>
>
>
>
=
>
>
>
>
>
>
;
DM
b
11
b
22
b
22
0
0
0
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
9
>
>
>
>
>
>
=
>
>
>
>
>
>
;
35
Inverting Eq. (35), we have:
r
11
r
22
r
33
s
23
s
13
s
12
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
9
>
>
>
>
>
>
=
>
>
>
>
>
>
;

C
11
C
12
C
13
0 0 0
C
12
C
22
C
23
0 0 0
C
12
C
23
C
22
0 0 0
0 0 0 C
22
C
23
=2 0 0
0 0 0 0 C
66
0
0 0 0 0 0 C
66
2
6
6
6
6
6
6
4
3
7
7
7
7
7
7
5

e
11
DTa
11
DMb
11
e
22
DTa
22
DMb
22
e
33
DTa
22
DMb
22
c
23
c
13
c
12
8
>
>
>
>
>
>
<
>
>
>
>
>
>
:
9
>
>
>
>
>
>
=
>
>
>
>
>
>
;
36
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 123 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).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
HOMEWORK PROBLEMS
An anisotropic material with the following properties is considered in
problems 14:
E
xx
100 GPa E
yy
200 GPa E
zz
75 GPa
v
xy
0:20 v
xz
0:25 v
yz
0:60
G
xy
60 GPa G
xz
75 GPa G
yz
50 GPa
g
xx;xy
0:30 g
xx;xz
0:25 g
xx;yz
0:30
g
yy;xy
0:60 g
yy;xz
0:75 g
yy;yz
0:20
g
zz;xy
0:20 g
zz;xz
0:05 g
zz;yz
0:15
l
xy;xz
0:10 l
xy;yz
0:05 g
xz;yz
0:10
a
xx
5 Am=m jC a
yy
10 Am=m jC a
zz
20 Am=m jC
a
xy
5 Arad=jC a
xz
15 Arad=jC a
yz
25 Arad=jC
b
xx
300 Am=m%M b
yy
60 Am=m%M b
zz
1200 Am=m%M
b
xy
150 Arad=%M b
xz
1000 Arad=%M b
yz
350 Arad=%M
1. Calculate the compliance matrix, S
ij
.
2. Calculate the stiness matrix, C
ij
=S
ij
1
. (Performthis calculation using a
suitable software package such as Maple, Matlab, Mathematica, etc.)
3. Consider the following stress tensor:
r
ij

r
xx
r
xy
r
xz
r
yx
r
yy
r
yz
r
zx
r
zy
r
zz
2
4
3
5

75 10 25
10 90 30
25 30 25
2
4
3
5
MPa
(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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
4. Consider the following strains:
e
xx
1500 Am=m e
yy
2000 Am=m e
zz
1750 Am=m
c
xy
750 Arad c
xz
500 Arad c
yz
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 eects 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 eects 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 58:
E
11
100 GPa E
22
200 GPa E
33
75 GPa
m
12
0:20 m
13
0:25 m
23
0:60
G
12
60 GPa G
13
75 GPa G
23
50 GPa
a
11
1 Am=mjC a
22
25 Am=mjC a
33
15 Am=mjC
b
11
100 Am=m%M b
22
600 Am=m%M b
33
1000 Am=m%M
5. Calculate the compliance matrix, S
ij
.
6. Calculate the stiness matrix, C
ij
.
7. Consider the following stress tensor:
r
ij

r
11
r
12
r
13
r
21
r
22
r
23
r
31
r
32
r
33
2
6
4
3
7
5
75 10 25
10 90 30
25 30 25
2
4
3
5
MPa
(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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
8. Consider the following strains:
e
11
2000 Am=m e
22
3000 Am=m e
33
1500 Am=m
c
12
750 Arad c
13
1000 Arad c
23
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 eects 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 eects 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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 formof Hookes 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
denition of the transformed, reduced form of Hookes law. The eec-
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 dened
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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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:
e
11
e
22
e
33
c
23
c
13
c
12
_

_
_

S
11
S
12
S
13
0 0 0
S
12
S
22
S
23
0 0 0
S
13
S
23
S
33
0 0 0
0 0 0 S
44
0 0
0 0 0 0 S
55
0
0 0 0 0 0 S
66
_

_
_

_
r
11
r
22
r
33
s
23
s
13
s
12
_

_
_

_
DT
a
11
a
22
a
33
0
0
0
_

_
_

_
DM
b
11
b
22
b
33
0
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 123 coordinate system. Let us now consider the
strains induced in an orthotropic material by a state of plane stress. Assuming
that r
33
=s
23
=s
13
=0, Eq. (33) of Chap. 4 becomes:
e
11
e
22
e
33
c
23
c
13
c
12
_

_
_

S
11
S
12
S
13
0 0 0
S
12
S
22
S
23
0 0 0
S
13
S
23
S
33
0 0 0
0 0 0 S
44
0 0
0 0 0 0 S
55
0
0 0 0 0 0 S
66
_

_
_

_
r
11
r
22
0
0
0
s
12
_

_
_

_
DT
a
11
a
22
a
33
0
0
0
_

_
_

_
DM
b
11
b
22
b
33
0
0
0
_

_
_

_
1
Note that Eq. (1) shows that in the case of plane stress, the out-of-plane shear
strains are always equal to zero (c
23
=c
13
=0). It is customary to write the
expressions for the remaining four strain components as follows:
e
11
e
22
c
12
_

_
_

S
11
S
12
0
S
12
S
22
0
0 0 S
66
_

_
_

_
r
11
r
22
s
12
_

_
_

_
DT
a
11
a
22
0
_

_
_

_
DM
b
11
b
22
0
_

_
_

_
2a
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
and
e
33
S
13
r
11
S
23
r
22
DTa
33
DMb
33
2b
Equations (2a) and (2b) are called the reduced forms of Hookes 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 S
66
. The denition of each com-
pliance term is not altered by the reduction fromthree to two dimensions, and
each term is still related to the more familiar engineering constants (E
11
, E
22
,
v
12
, etc.) in accordance with Eq. (18) of Chap. 4.
Inverting Eq. (2a), we obtain:
r
11
r
22
s
12
_

_
_

Q
11
Q
12
0
Q
12
Q
22
0
0 0 Q
66
_

_
_

_
e
11
DTa
11
DMb
11
e
22
DTa
22
DMb
22
c
12
_

_
_

_
3
The 33 array that appears in Eq. (3) is called the reduced stiness
matrix and equals the inverse of the reduced compliance matrix:
Q
11
Q
12
0
Q
12
Q
22
0
0 0 Q
66
_

_
_

_u
S
11
S
12
0
S
12
S
22
0
0 0 S
66
_

_
_

_
1
4
Note that we are now using a dierent symbol to denote stiness. That
is, the original 3-D stiness matrix was denoted C
ij
(as in Eq. (19) of Chap. 4,
for example), whereas the reduced stiness matrix is denoted Q
ij
. This change
in notation is required because individual members of the reduced stiness
matrix are not equal to the corresponding members in the original stiness
matrix. That is, Q
11
p C
11
, Q
12
p C
12
, Q
22
p C
22
, and Q
66
p C
66
. Relations
between Q
ij
and C
ij
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:
r
11
C
11
e
11
C
12
e
22
C
13
e
33
5a
and
r
33
C
13
e
11
C
23
e
22
C
33
e
33
5b
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
However, the reduced stiness matrix relates stress to strain under conditions
of plane stress by denition. Setting r
33
=0 and solving Eq. (5b) for e
33
, we
have:
e
33

C
13
e
11
C
23
e
22

C
33
6
The out-of-plane strain e
33
must be related to in-plane strains e
11
and e
22
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:
r
11

C
11
C
33
C
2
13
C
33
_ _
e
11

C
12
C
33
C
13
C
23
C
33
_ _
e
22
7
On the other hand, from Eq. (3), r
11
is given by (with DT=DM=0):
r
11
Q
11
e
11
Q
12
e
22
8
Comparing Eqs. (7) and (8), it is immediately apparent that:
Q
11

C
11
C
33
C
2
13
C
33
9a
Q
12

C
12
C
33
C
13
C
23
C
33
9b
Using a similar procedure, it can be shown that:
Q
22

C
22
C
33
C
2
23
C
33
9c
Q
66
C
66
9d
In essence, the denition of elements of the Q
ij
matrix diers from those of the
C
ij
matrix because the Q
ij
matrix is dened for plane stress conditions only,
whereas C
ij
can be used for any stress state.
Elements of the reduced stiness 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:
Q
11

S
22
S
11
S
22
S
2
12
Q
12
Q
21

S
12
S
11
S
22
S
2
12
Q
22

S
11
S
11
S
22
S
2
12
Q
66

1
S
66
10
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Alternatively, the elements of the reduced stiness matrix are related to
more familiar engineering constants as follows:
Q
11

E
2
11
E
11
m
2
12
E
22
Q
12
Q
21

m
12
E
11
E
22
E
11
m
2
12
E
22
Q
22

E
11
E
22
E
11
m
2
12
E
22
Q
66
G
12
11
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
r
33
=s
23
=s
13
=0. From Eq. (34) of Chap. 4, it is seen that the out-of-plane
shear strains equal zero (c
23
=c
13
=0), and the remaining four strains can be
written as:
e
11
e
22
c
12
_

_
_

S
11
S
12
0
S
12
S
22
0
0 0 S
66
_

_
_

_
r
11
r
22
s
12
_

_
_

_
DT
a
11
a
22
0
_

_
_

_
DM
b
11
b
22
0
_

_
_

_
12a
and
e
33
S
12
r
11
S
23
r
22
DTa
22
DMb
22
12b
Comparing Eq. (12a) with Eq. (2a), it is seen that the relationship
between in-plane strains (e
11
, e
22
, and c
12
) and in-plane stress components
(r
11
, r
22
, and s
12
) 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, S
13
=S
12
, a
33
=a
22
,
and b
33
=b
22
, 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 23 plane), C
33
=C
22
and C
13
=C
12
.
Inverting Eq. (12a), we obtain:
r
11
r
22
s
12
_

_
_

Q
11
Q
12
0
Q
12
Q
22
0
0 0 Q
66
_

_
_

_
e
11
DTa
11
DMb
11
e
22
DTa
22
DMb
22
c
12
_

_
_

_
Since this result is identical to Eq. (3), it is seen once again that the relationship
between in-plane strains (e
11
, e
22
, and c
12
) and in-plane stress components
(r
11
, r
22
, and s
12
) is identical for orthotropic and transversely isotropic
materials.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Example Problem 1
Determine the strains induced in a unidirectional graphiteepoxy 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:
r
11
200 MPa r
22
30 MPa s
12
50 MPa
Strains can be calculated using either Eq. (2a) or Eq. (12a). Since DT=
DM=0, we have:
e
11
e
22
c
12
_

_
_

S
11
S
12
0
S
12
S
22
0
0 0 S
66
_

_
_

_
r
11
r
22
s
12
_

_
_

_
Each termwithinthe reduced compliance matrix is calculated using Eq. (18) of
Chap. 4:
e
11
e
22
c
12
_

_
_

1
E
11
m
12
E
11
0
m
12
E
11
1
E
22
0
0 0
1
G
12
_

_
_

_
r
11
r
22
s
12
_

_
_

_
Figure 1 Unidirectional composite subjected to in-plane stresses (magnitudes
of in-plane stresses shown).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
From Table 3 of Chap. 3:
E
11
170 GPa E
22
10 GPa
m
12
0:30 G
12
13 GPa
Using these values:
e
11
e
22
c
12
_

_
_

5:88 10
12
1:76 10
12
0
1:76 10
12
100 10
12
0
0 0 76:9 10
12
_

_
_

_
1
Pa
_ _

200 10
6
30 10
6
50 10
6
_

_
_

_
Pa
Completing the matrix multiplication indicated, we obtain:
e
11
e
22
c
12
_

_
_

1230 Am=m
3350 Am=m
3850 Arad
_

_
_

_
Example Problem 2
Determine the strain induced in a unidirectional graphiteepoxy 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 dierent 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):
e
11
e
22
c
12
_

_
_

S
11
S
12
0
S
12
S
22
0
0 0 S
66
_

_
_

_
r
11
r
22
s
12
_

_
_

_
DT
a
11
a
22
0
_

_
_

_
DM
b
11
b
22
0
_

_
_

_
Numerical values for the stresses and reduced compliance matrix are given in
Example 1, and the linear thermal and moisture expansion coecients for
graphiteepoxy are (from Table 3 of Chap. 3):
a
11
0:9 Am=m=
o
C
a
22
27 Am=m=
o
C
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
b
11
150 Am=m=%M
b
22
4800 Am=m=%M
Hence, Eq. 2(a) or Eq. (12a) becomes:
e
11
e
22
c
12
_

_
_

5:88 10
12
1:76 10
12
0
1:76 10
12
100 10
12
0
0 0 76:9 10
12
_

_
_

200 10
6
30 10
6
50 10
6
_

_
_

_
155
0:9 10
6
27 10
6
0
_

_
_

_
0:5

150 10
6
4800 10
6
0
_

_
_

_
Completing the matrix multiplication indicated, we obtain:
e
11
e
22
c
12
_

_
_

1230 Am=m
3350 Am=m
3850 Arad
_

_
_

140 Am=m
4185 Am=m
0
_

_
_

75 Am=m
2400 Am=m
0
_

_
_

1445 Am=m
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 aects the state of stress, but rather aects
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 graphiteepoxy composite laminate is rmly mounted
within an innitely rigid square frame, as shown in Fig. 2. The coecients 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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Solution. According to the problem statement, the frame is innitely rigid
and is made of a material whose thermal and moisture expansion coecients
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 rmly 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:
e
11
e
22
c
12
_

_
_

0
0
0
_

_
_

_
The stresses induced can be calculated using Eq. (3):
r
11
r
22
s
12
_

_
_

Q
11
Q
12
0
Q
12
Q
22
0
0 0 Q
66
_

_
_

_
e
11
DTa
11
DMb
11
e
22
DTa
22
DMb
22
c
12
_

_
_

_
The terms within the reduced stiness matrix are calculated in accordance
with Eq. (11):
Q
11

E
2
11
E
11
m
2
12
E
22

170 GPa
2
170 GPa 0:30
2
10 GPa
170:9 GPa
Figure 2 Unidirectional composite laminate mounted within an infinitely rigid
frame.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Q
12
Q
21

m
12
E
11
E
22
E
11
m
2
12
E
22

0:30 170 GPa 10 GPa


170 GPa 0:30
2
10 GPa
3:016 GPa
Q
22

E
11
E
22
E
11
m
2
12
E
22

170 GPa 10 GPa


170 GPa 0:30
2
10 GPa
10:05 GPa
Q
66
G
12
13 GPa
Hence, in this case, Eq. (3) becomes:
r
11
r
22
s
12
_

_
_

170:9 10
9
3:016 10
9
0
3:016 10
9
10:05 10
9
0
0 0 13:0 10
9
_

_
_

0 155 0:9 10
6
_ _
0:5 150 10
6
_ _
0 155 27 10
6
_ _
0:5 4800 10
6
_ _
0
_
_
_
_
_
_

r
11
r
22
s
12
_

_
_

31:36 MPa
17:29 MPa
0
_

_
_

_
The reader may initially view this example to be somewhat unrealistic.
After all, the frame has been assumed to be innitely rigid. In reality, an
innitely rigid material (i.e., one for which E!l) does not exist. Further-
more, it is assumed that the thermal expansion coecient for the frame is zero,
which is also not valid for most real materials (the assumption that the
moisture expansion coecient is zero is true for metals and metal alloys).
Despite these unrealistic assumptions, this example problem illustrates
a common occurrence in composite laminates. Specically, most modern
composites are cured at an elevated temperature (for example, many
graphiteepoxy 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 signicant gain in
weight (most composites adsorb a maximum of 34% moisture by weight,
even if totally immersed in water), even this slight gain in moisture content
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
can nevertheless cause signicant 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 dierent coordinate systems
is shown in Fig. 3. In Fig. 3(a), the composite is referenced to the principal
material coordinate system (i.e., the 12 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) xy coordinate system. This is often called an o-axis spec-
imen since the specimen is referenced to an arbitrary xy coordinate system
rather than the principal 12 coordinate system. Suppose we wish to relate
strains and stresses referenced to the xy coordinate system. For example,
suppose we know the stresses r
xx
, r
yy
, and s
xy
, as well as the material
properties referenced to the principal material coordinate system (E
11
, E
22
,
m
12
, etc.), and wish to calculate strains e
xx
, e
yy
, and c
xy
. 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 12 coordinate system.
We can perform this calculation using a three-step process. Specically,
we can:
(a) Transform the known stresses from the xy coordinate system to
the 12 coordinate system (using Eq. 20 of Chap. 2, for example),
which will give us the stress components r
11
, r
22
, and s
12
that
correspond to the known values of r
xx
, r
yy
, and s
xy
.
(b) Apply Eq. (2a) to obtain in-plane strains e
11
, e
22
, and c
12
.
(c) Transform the calculated strains (e
11
, e
22
, and c
12
) from the 12
coordinate system back to the xy coordinate system, nally
obtaining the desired strains e
xx
, e
yy
, and c
xy
.
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 xy coordinate
system to the 12 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 formof reduced Hookes
law suitable for use in an arbitrary xy coordinate system. In eect, we will
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Figure 3 A unidirectional composite referenced to two different coordinate
systems.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
transform Hookes law from the 12 coordinate system to an arbitrary xy
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:
e
11
e
22
c
12
_

_
_

S
11
S
12
0
S
12
S
22
0
0 0 S
66
_

_
_

_
r
11
r
22
s
12
_

_
_

_
13
Equation (13) can be rewritten as:
1 0 0
0 1 0
0 0 2
_

_
_

_
e
11
e
22
c
12
=2
_

_
_

S
11
S
12
0
S
12
S
22
0
0 0 S
66
_

_
_

_
r
11
r
22
s
12
_

_
_

_
14
In Eq. (14), we have employed the so-called Reuter matrix (named
after the person who suggested this approach [1]) to, in eect, divide the shear
strain by a factor of (1/2) within the strain array. Let:
R
1 0 0
0 1 0
0 0 2
_

_
_

_ and S
S
11
S
12
0
S
12
S
22
0
0 0 S
66
_

_
_

_
so that Eq. (14) can be written in the following abbreviated form:
R
e
11
e
22
c
12
=2
_

_
_

_
S
r
11
r
22
s
12
_

_
_

_
The transformation of strains within a plane from an xy coordinate
systemto another xVyV coordinate systemwas 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:
e
11
e
22
c
12
=2
_

_
_

cos
2
h sin
2
h 2cos h sin h
sin
2
h cos
2
h 2cos h sin h
cos h sin h cos h sin h cos
2
h sin
2
h
_

_
_

e
xx
e
yy
c
xy
=2
_

_
_

_
T
e
xx
e
yy
c
xy
=2
_

_
_

_
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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:
r
11
r
22
s
12
_

_
_

cos
2
h sin
2
h 2cos h sin h
sin
2
h cos
2
h 2cos h sin h
cos h sin h cos h sin h cos
2
h sin
2
h
_

_
_

r
xx
r
yy
s
xy
_

_
_

_
T
r
xx
r
yy
s
xy
_

_
_

_
As pointed out in Chap. 2, the identical transformation matrix, [T], is used to
relate strains and stresses in the two coordinate systems:
T
cos
2
h sin
2
h 2cos h sin h
sin
2
h cos
2
h 2cos h sin h
cos h sin h cos h sin h cos
2
h sin
2
h
_

_
_

_
Inserting these transformation relationships into Eq. (14), we have:
R T
e
xx
e
yy
c
xy
=2
_

_
_

_
S T
r
xx
r
yy
s
xy
_

_
_

_
15
To simplify Eq. (15), rst 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
:
e
xx
e
yy
c
xy
=2
_

_
_

_
T
1
R
1
S T
r
xx
r
yy
s
xy
_

_
_

_
16
where:
R
1

1 0 0
0 1 0
0 0 1=2
_

_
_

_
and
T
1

cos
2
h sin
2
h 2cos h sin h
sin
2
h cos
2
h 2cos h sin h
cos h sin h cos h sin h cos
2
h sin
2
h
_

_
_

_
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
We next extract the factor of (
1

2
) from the shear strain within the strain array
using the [R]
1
matrix:
R
1
e
xx
e
yy
c
xy
_

_
_

_
T
1
R
1
S T
r
xx
r
yy
s
xy
_

_
_

_
17
Multiplying both sides of Eq. (17) by the [R] matrix, we arrive at our
nal result:
e
xx
e
yy
c
xy
_

_
_

_
R T
1
R
1
S T
r
xx
r
yy
s
xy
_

_
_

_
18
Equation (18) represents an o-axis version of Hookes Law. That is,
it relates the strains induced in the arbitrary xy coordinate system (e
xx
, e
yy
,
and c
xy
) to the stresses in the same xy coordinate system(r
xx
, r
yy
, and s
xy
) via
material properties referenced to the principal 12 coordinate system (repre-
sented by the [S] matrix) and the ber angle h (represented by the trans-
formation matrix [T]). Completing the matrix algebra indicated, we obtain:
e
xx
e
yy
c
xy
_

_
_

_
S
_
r
xx
r
yy
s
xy
_

_
_

_
19
where:
S
_
R T
1
R
1
S T
Equation (19) is known as transformed reduced Hookes law. It is called
transformed because Eq. (19) has been transformed from the 12 coor-
dinate system to the xy 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:
e
xx
e
yy
c
xy
_

_
_

S
11
S
12
S
16
S
21
S
22
S
26
S
61
S
62
S
66
_

_
_

_
r
xx
r
yy
s
xy
_

_
_

_
20
* In common parlance, the [

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


Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
where:
S
11
S
11
cos
4
h 2S
12
S
66
cos
2
h sin
2
hS
22
sin
4
h
S
12
S
21
S
12
cos
4
hsin
4
h
_ _
S
11
S
22
S
66
cos
2
h sin
2
h
S
16
S
61
2S
11
2S
12
S
66
cos
3
h sin h 2S
22
2S
12
S
66
cos h sin
3
h
S
22
S
11
sin
4
h 2S
12
S
66
cos
2
h sin
2
hS
22
cos
4
h
S
26
S
62
2S
11
2S
12
S
66
cos h sin
3
h 2S
22
2S
12
S
66
cos
3
h sin h
S
66
2 2S
11
2S
22
4S
12
S
66
cos
2
h sin
2
hS
66
cos
4
hsin
4
h
_ _
21
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 S
66
. Secondly, the [S ] matrix is
symmetric. Therefore, S
21
S
12
, S
61
S
16
, and S
62
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 o-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 12 coordinate system is referred to as an orthotropic (or transversely
isotropic) material, whereas if the same material is referenced to an arbitrary
(nonprincipal) xy 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
s
xy
will cause normal strains e
xx
and e
yy
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 Hookes law. In the 12 coordinate system, in-plane thermal
strains are given by:
e
T
11
e
T
22
c
T
12
_

_
_

_
DT
a
11
a
22
0
_

_
_

_
22
As has been previously discussed, in the 12 coordinate system, a
uniform change in temperature does not produce a shear strain, i.e., a
12
=0.
Now, thermally induced strains can be transformed from one coordinate
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
system to another in exactly the same way as mechanically induced strains are
transformed. That is, we can relate thermal strains in the 12 coordinate
system to the xy coordinate system using the transformation matrix:
e
T
11
e
T
22
c
T
12
=2
_

_
_

cos
2
h sin
2
h 2cos h sin h
sin
2
h cos
2
h 2cos h sin h
cos h sin h cos h sin h cos
2
h sin
2
h
_

_
_

e
T
xx
e
T
yy
c
T
xy
=2
_

_
_

_
Inverting this expression, we have:
e
T
xx
e
T
yy
c
T
xy
=2
_

_
_

_
T
1
e
T
11
e
T
22
0
_

_
_

cos
2
h sin
2
h 2cos h sin h
sin
2
h cos
2
h 2cos h sin h
cos h sin h cos h sin h cos
2
h sin
2
h
_

_
_

e
T
11
e
T
22
c
T
12
=2
_

_
_

_
23
Substituting Eq. (22) in this result, completing the matrix multiplication
indicated, and simplifying the resulting expressions, we obtain:
e
T
xx
e
T
yy
c
T
xy
_

_
_

_
DT
a
xx
a
yy
a
xy
_

_
_

_
24
where:
a
xx
a
11
cos
2
h a
22
sin
2
h
a
yy
a
11
sin
2
h a
22
cos
2
h 25
a
xy
2cos h sin h a
11
a
22

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
In Sec. 6, we will dene the properties of a unidirectional composite laminate
when referenced to an arbitrary xy coordinate system. As further discussed
there, Eq. (25) denes the eective coecients 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 c
xy
T
as well as normal
strains e
xx
T
and e
yy
T
.
In the 12 coordinate system, the in-plane strains caused by a uniform
change in moisture content are given by:
e
M
11
e
M
22
c
M
12
_

_
_

_
DM
b
11
b
22
0
_

_
_

_
26
As was the case for thermal strains, we wish to express strains induced by DM
in an arbitrary xy coordinate system. Using the identical procedure as
before, moisture strains in the xy coordinate system are given by:
e
M
xx
e
M
yy
c
M
xy
_

_
_

_
DM
b
xx
b
yy
b
xy
_

_
_

_
27
where:
b
xx
b
11
cos
2
h b
22
sin
2
h
b
yy
b
11
sin
2
h b
22
cos
2
h 28
b
xy
2cos h sin h b
11
b
22

In Sec. 6, we will dene the eective properties of a unidirectional
composite laminate when referenced to an arbitrary xy coordinate system.
As further discussed there, Eq. (28) denes the eective coecients of
moisture expansion for an anisotropic composite laminate.
We can now calculate the strains induced by the combined eects of
stress, a uniform change in temperature, and a uniform change in moisture
content. Specically, adding Eqs. (20), (24), and (27) together, we obtain:
e
xx
e
yy
c
xy
_

_
_

S
11
S
12
S
16
S
12
S
22
S
26
S
16
S
26
S
66
_

_
_

_
r
xx
r
yy
s
xy
_

_
_

_
DT
a
xx
a
yy
a
xy
_

_
_

_
DM
b
xx
b
yy
b
xy
_

_
_

_
29
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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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:
r
xx
r
yy
s
xy
_

_
_

Q
11
Q
12
Q
16
Q
12
Q
22
Q
26
Q
16
Q
26
Q
66
_

_
_

_
e
xx
DTa
xx
DMb
xx
e
yy
DTa
yy
DMb
yy
c
xy
DTa
xy
DMb
xy
_

_
_

_
30
where:
Q
11
Q
11
cos
4
h 2 Q
12
2Q
66
cos
2
h sin
2
h Q
22
sin
4
h
Q
12
Q
21
Q
12
cos
4
h sin
4
h
_ _
Q
11
Q
22
4Q
66
cos
2
h sin
2
h
Q
16
Q
61
Q
11
Q
12
2Q
66
cos
3
h sin h
Q
22
Q
12
2Q
66
cos h sin
3
h
Q
22
Q
11
sin
4
h 2 Q
12
2Q
66
cos
2
h sin
2
h Q
22
cos
4
h
Q
26
Q
62
Q
11
Q
12
2Q
66
cos h sin
3
h
Q
22
Q
12
2Q
66
cos
3
h sin h
Q
66
Q
11
Q
22
2Q
12
2Q
66
cos
2
h sin
2
h Q
66
cos
4
h sin
4
h
_ _
31
The [

Q] matrix is called the transformed, reduced stiness matrix.* This name


again reminds us that we have reduced our analysis to the 2-D plane stress
case, and that we have transformed Hookes law from the 12 coordinate
system to an arbitrary xy coordinate system. Note that the [

Q] matrix is (in
general) fully populated. This reects the anisotropic nature of unidirectional
composites when referenced to a nonprincipal material coordinate system.
Also, the [

Q] matrix is symmetric, so that



Q
21


Q
12
,

Q
61


Q
16
, and

Q
62

Q
26
, as indicated in Eq. (31).
The reader should note that the functional form of the equations that
dene the elements of the transformed reduced stiness matrix, i.e., Eq. (31),
is not identical to the functional formof the equations dening 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 S
11
for Q
11
, S
12
for
Q
12
, S
22
for Q
22
, etc. This dierence in functional form is due to the fact that
we have dened both the stiness and compliance matrices in terms of
* In common parlance, the [

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


Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
engineering shear strains, rather than tensoral shear strains. The use of Eqs.
(29) and (30) to solve simple problems involving o-axis composite laminates
is demonstrated in Example Problems 4 to 6.
Example Problem 4
Determine the strains induced in the o-axis graphiteepoxy 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 o-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:
r
xx
200 MPa r
yy
30 MPa s
xy
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
ber angle in Fig. 4 is algebraically positive: h=+30j.
Strains are calculated using Eq. (29). Since DT=DM=0, we have:
e
xx
e
yy
c
xy
_

_
_

S
11
S
12
S
16
S
12
S
22
S
26
S
16
S
26
S
66
_

_
_

_
r
xx
r
yy
s
xy
_

_
_

_
Figure 4 A 30j off-axis composite subjected to in-plane stresses (magnitudes
of in-plane stresses shown).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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:
S
11
S
11
cos
4
h 2S
12
S
66
cos
2
h sin
2
h S
22
sin
4
h
S
11
5:88 10
12
_ _
cos
4
30
B

2 1:76 10
12
_ _
76:9 10
12
_ _
cos
2
30
B
sin
2
30
B

100:0 10
12
_ _
sin
4
30
B

S
11
23:32 10
12
Pa
1
_ _
Calculating the remaining elements of the S matrix in similar fashion and
applying Eq. (29), we nd:
e
xx
e
yy
c
xy
_

_
_

23:32 10
12
4:327 10
12
33:72 10
12
4:327 10
12
70:38 10
12
47:79 10
12
33:72 10
12
47:79 10
12
101:3 10
12
_

_
_

1
Pa
_ _
200 10
6
30 10
6
50 10
6
_

_
_

_
Pa
Completing the matrix multiplication indicated, we obtain:
e
xx
e
yy
c
xy
_

_
_

6220 Am=m
1144 Am=m
10375 Arad
_

_
_

_
Example Problem 5
Determine the strains induced in the o-axis graphiteepoxy 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 dierent mechanisms that contribute
to the total strain induced in the laminate: the applied stresses, the temper-
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
ature change, and the change in moisture content. The total strains induced by
all three mechanisms can be calculated using Eq. (29):
e
xx
e
yy
c
xy
_

_
_

S
11
S
12
S
16
S
12
S
22
S
26
S
16
S
26
S
66
_

_
_

_
r
xx
r
yy
s
xy
_

_
_

_
DT
a
xx
a
yy
a
xy
_

_
_

_
DM
b
xx
b
yy
b
xy
_

_
_

_
Numerical values for the stresses and transformed reduced compliance matrix
are given in Example 4. The linear thermal and moisture expansion coef-
cients for graphiteepoxy, referenced to the xy coordinate system, are
calculated using Eqs. (25) and (28), respectively.
The thermal expansion coecients are:
a
xx
a
11
cos
2
h a
22
sin
2
h 0:9 10
6
_ _
cos
2
30j
27 10
6
_ _
sin
2
30j
a
xx
6:1 Am=m=
o
C
a
yy
a
11
sin
2
h a
22
cos
2
h 0:9 10
6
_ _
sin
2
30j
27 10
6
_ _
cos
2
30j
a
yy
20:0 Am=m=
o
C
a
xy
2 cos h sin h a
11
a
22
2 cos 30j sin 30j
0:9 10
6
27 10
6
_ _
a
xy
24:2 Arad=
o
C
The moisture expansion coecients are:
b
xx
b
11
cos
2
h b
22
sin
2
h 150 10
6
_ _
cos
2
30j
4800 10
6
_ _
sin
2
30j
b
xx
1313 Am=m=%M
b
yy
b
11
sin
2
h b
22
cos
2
h 150 10
6
_ _
sin
2
30j
4800 10
6
_ _
cos
2
30j
b
yy
3638 Am=m=%M
b
xy
2 cos h sin h b
11
b
22
2 cos 30j sin 30j
150 10
6
4800 10
6
_ _
b
xy
4027 Arad=%M
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Hence, Eq. (29) becomes:
e
xx
e
yy
c
xy
_

_
_

23:32 10
12
4:327 10
12
33:72 10
12
4:327 10
12
70:38 10
12
47:79 10
12
33:72 10
12
47:79 10
12
101:3 10
12
_

_
_

200 10
6
30 10
6
50 10
6
_

_
_

_
155
6:1 10
6
20:0 10
6
24:2 10
6
_

_
_

_
0:5

1313 10
6
3638 10
6
4027 10
6
_

_
_

_
Completing the matrix multiplication indicated, we obtain:
e
xx
e
yy
c
xy
_

_
_

6220 Am=m
1144 Am=m
10375 Arad
_

_
_

946 Am=m
3100 Am=m
3751 Arad
_

_
_

657 Am=m
1819 Am=m
2014 Arad
_

_
_

5931 Am=m
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 aects the state of stress, but rather aects
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 o-axis graphiteepoxy composite laminate is rmly mounted within
an innitely rigid square frame, as shown in Fig. 5. The coecients 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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 innitely rigid frame with zero
thermal expansion coecients. Since the frame is innitely rigid and is
made of a material whose thermal and moisture expansion coecients equal
zero, the frame will retain its original shape. Since the o-axis composite is
rmly 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:
e
xx
e
yy
c
xy
_

_
_

0
0
0
_

_
_

_
The stresses induced can be calculated using Eq. (30):
r
xx
r
yy
s
xy
_

_
_

Q
11

Q
12

Q
16

Q
12

Q
22

Q
26

Q
16

Q
26

Q
66
_

_
_

_
e
xx
DTa
xx
DMb
xx
e
yy
DTa
yy
DMb
yy
c
xy
DTa
xy
DMb
xy
_

_
_

_
Recall that the elements of the reduced stiness matrix, the [ Q] matrix, were
calculated as a part of Example Problem 3. Each term within the transformed
Figure 5 Off-axis composite laminate mounted within an infinitely rigid frame.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
reduced stiness matrix, the [

Q] matrix, can therefore be calculated via a


straightforward application of Eq. (31). Calculation of

Q
11
, for example,
proceeds as follows:

Q
11
Q
11
cos
4
h 2 Q
12
2Q
66
cos
2
h sin
2
h Q
22
sin
4
h

Q
11
170:9 10
9
_ _
cos
4
30j
2 3:016 10
9
2 13:0 10
9
_ _ _ _
cos
2
30j sin
2
30j
10:05 10
9
_ _
sin
4
30j

Q
11
107:6 10
9
Pa
Calculating the remaining elements of the [

Q] matrix in similar fashion, and


using the thermal and moisture expansion coecients referenced to the xy
coordinate system (calculated as a part of Example Problem 5), we nd:
r
xx
r
yy
s
xy
_

_
_

107:6 10
9
26:1 10
9
48:1 10
9
26:1 10
9
27:2 10
9
21:5 10
9
48:1 10
9
21:5 10
9
36:0 10
9
_

_
_

0 155 6:1 10
6
_ _
0:5 1313 10
6
_ _
0 155 20:0 10
6
_ _
0:5 3638 10
6
_ _
0 155 24:2 10
6
_ _
0:5 4027 10
6
_ _
_

_
_

_
r
xx
r
yy
s
xy
_

_
_

19:0 MPa
5:04 MPa
21:1 MPa
_

_
_

_
3 CALCULATING TRANSFORMED PROPERTIES USING
MATERIAL INVARIANTS
The stresses and strains in a unidirectional composite referenced to an
arbitrary xy 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 stiness matrix, [

Q], and individual elements of the [

Q]
matrix are calculated in accordance with Eq. (31).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Now, both Eqs. (21) and (31) involve trigonometric functions raised to a
power (i.e., sin
4
h, cos
4
h, cos h sin
3
h, etc.). These equations can be simplied
somewhat through the use of the following trigonometric identities:
sin
4
h
1
8
3 4 cos 2h cos 4h
cos
4
h
1
8
3 4 cos 2h cos 4h
cos h sin
3
h
1
8
2 sin 2h sin 4h
cos
2
h sin
2
h
1
8
1 cos 4h
cos
3
h sin h
1
8
2 sin 2h sin 4h
32
For example, substituting these identities into Eq. (21) and simplifying
result in:
S
11
U
S
1
U
S
2
cos 2h U
S
3
cos
4
h
S
12
S
21
U
S
4
U
S
3
cos 4h
S
16
S
61
U
S
2
sin 2h 2U
S
3
sin 4h 33
S
22
U
S
1
U
S
2
cos 2h U
S
3
cos 4h
S
26
S
62
U
S
2
sin 2h 2U
S
3
sin 4h
S
66
U
S
5
4U
S
3
cos 4h
The terms U
i
S
which appear in Eq. (33) are called compliance invariants and
are dened as follows:
U
S
1

1
8
3S
11
3S
22
2S
12
S
66

U
S
2

1
2
S
11
S
22

U
S
3

1
8
S
11
S
22
2S
12
S
66
34
U
S
4

1
8
S
11
S
22
6S
12
S
66

U
S
5

1
2
S
11
S
22
2S
12
S
66

The superscript S is used to indicate that these quantities are calculated using
members of the reduced compliance matrix [S]. They are called compliance
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
invariants because they dene 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:
Q
11
U
Q
1
U
Q
2
cos 2h U
Q
3
cos 4h
Q
12
Q
21
U
Q
4
U
Q
3
cos 4h
Q
16
Q
61

1
2
U
Q
2
sin 2h U
Q
3
sin 4h 35
Q
22
U
Q
1
U
Q
2
cos 2h U
Q
3
cos 4h
Q
26
Q
62

1
2
U
Q
2
sin 2h U
Q
3
sin 4h
Q
66
U
Q
5
U
Q
3
cos 4h
where the stiness invariants, U
i
Q
, are dened as:
U
Q
1

1
8
3Q
11
3Q
22
2Q
12
4Q
66

U
Q
2

1
2
Q
11
Q
22

U
Q
3

1
8
Q
11
Q
22
2Q
12
4Q
66
36
U
Q
4

1
8
Q
11
Q
22
6Q
12
4Q
66

U
Q
5

1
8
Q
11
Q
22
2Q
12
Q
66

The superscript Q is used to indicate that these quantities are calculated using
members of the reduced stiness matrix [ Q]. The reader should note that the
functional forms of the stiness invariants dened in Eq. (36) are not identical
to those of the compliance invariants dened in Eq. (34). The dierence 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 stiness
invariants simplies the calculation of the terms within the [Q] matrix. At this
point, this simplication may seem to be a trivial matter since, in practice,
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 stiness invariants leads to a
convenient method of transforming the stiness 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 graphiteepoxy 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:
S
11
S
12
0
S
12
S
22
0
0 0 S
66
_
_
_
_

5:88 10
12
1:76 10
12
0
1:76 10
12
100:0 10
12
0
0 0 76:9 10
12
_
_
_
_
1
Pa
_ _
The compliance invariants may be calculated using these values and
Eq. (34):
U
S
1

1
8
3S
11
3S
22
2S
12
S
66

1
8
_
3 5:88 10
12
_ _
3 100:0 10
12
_ _
2 1:76 10
12
_ _
76:9 10
12
_
48:9 10
12
U
S
2

1
2
S
11
S
22

1
2
5:88 10
12
100:0 10
12
_
47:1 10
12
U
S
3

1
8
S
11
S
22
2S
12
S
66

1
8
_
5:88 10
12
100:0 10
12
2 1:76 10
12
_ _
76:9 10
12
_
4:06 10
12
U
S
4

1
8
S
11
S
22
6S
12
S
66

1
8
_
5:88 10
12
100:0 10
12
6 1:76 10
12
_ _
76:9 10
12
_
2:296 10
12
U
S
5

1
2
S
11
S
22
2S
12
S
66

1
2
_
5:88 10
12
100:0 10
12
2 1:76 10
12
_ _
76:9 10
12
_
93:17 10
12
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Using the rst equation of Eq. (33) and setting h=30j:
S
11
U
S
1
U
S
2
cos 2h U
S
3
cos 4h 48:9 10
12
_ _
47:1 10
12
_ _
cos 60j 4:06 10
12
_ _
cos 120j
23:3 10
12
Pa
1
Similarly, using the second equation of Eq. (33) and setting h=30j:
S
12
S
21
U
S
4
U
S
3
cos 4h
2:296 10
12
_ _
4:06 10
12
_ _
cos 120j
4:33 10
12
Pa
1
The additional terms within the [S] matrix are calculated using the rest of
Eq. (33). A summary of our results is:
S
23:3 10
12
4:33 10
12
33:7 10
12
4:33 10
12
70:4 10
12
47:8 10
12
33:7 10
12
47:8 10
12
101:3 10
12
_

_
_

_
1
Pa
_ _
Note that the [S] matrix is identical to that calculated in Example Problem 4.
4 EFFECTIVE ELASTIC PROPERTIES OF A UNIDIRECTIONAL
COMPOSITE LAMINATE
The denitions of common engineering material properties were reviewed in
Chap. 3. In this section, these concepts will be used to dene the eective
properties of a unidirectional composite laminates referenced to an arbitrary
xy coordinate system.
4.1 Effective Properties Relating Stress to Strain
Let us rst consider the elastic properties measured during uniaxial tests.
Consider the unidirectional composite laminate subjected to a uniaxial stress
r
xx
, 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 denition, r
yy
=s
xy
=0, Eq.
(29) becomes for this case:
e
xx
e
yy
c
xy
_

_
_

S
11
S
12
S
16
S
12
S
22
S
26
S
16
S
26
S
66
_

_
_

_
r
xx
0
0
_

_
_

_
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
In-plane strains caused by uniaxial stress r
xx
are therefore given by:
e
xx
S
11
r
xx
37a
e
yy
S
12
r
xx
37b
c
xy
S
16
r
xx
; 37c
In Sec. 2 of Chap. 3, Youngs modulus was dened as the normal stress
r
xx
divided by the resulting normal strain e
xx
, with all other stress components
equal zero. Applying this denition to the unidirectional laminate shown in
Fig. 6, Youngs modulus in the x-direction is given by:
E
xx

r
xx
e
xx

r
xx
S
11
r
xx

1
S
11
38a
Inserting the relation for Q
11
listed in Eq. (22), we have:
E
xx

1
S
11
cos
4
h 2S
12
S
66
cos
2
h sin
2
h S
22
sin
4
h
38b
Since each of the compliance terms (S
11
, S
12
, etc.) can also be related to the
more familiar engineering constants using Eq. (18) of Chap. 4, Youngs
modulus can also be written as:
E
xx

1
cos
4
h
E
11

1
G
12

2m
12
E
11
_ _
cos
2
h sin
2
h
sin
4
h
E
22
38c
Figure 6 Unidirectional composite laminate subjected to uniaxial stress r
xx
.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
In Sec. 2 of Chap. 3, Poissons ratio v
xy
was dened as the negative of
the transverse normal strain e
yy
divided by the axial normal strain e
xx
, both of
which are induced by stress r
xx
, with all other stresses equal zero. Poissons
ratio for the unidirectional laminate shown in Fig. 6 is therefore given by:
v
xy

e
yy
e
xx

S
12
S
11
39a
Using Eq. (22), this can be written as:
v
xy

S
12
cos
4
h sin
4
h
_ _
S
11
S
22
S
66
cos
2
h sin
2
h
_ _
S
11
cos
4
h 2S
12
S
66
cos
2
h sin
2
h S
22
sin
4
h
39b
or equivalently, using Eq. (18) of Chap. 4:
v
xy

v
12
E
11
cos
4
h sin
4
h
_ _

1
E
11

1
E
22

1
G
12
_ _
cos
2
h sin
2
h
cos
4
h
E
11

1
G
12

2v
12
E
11
_ _
cos
2
h sin
2
h
sin
4
h
E
22
39c
InSec. 2 of Chap. 3, the coecient of mutual inuence of the secondkind
g
xx,xy
was denedas the shear strainc
xy
dividedby the normal straine
xx
, both
of which are induced by normal stress r
xx
, when all other stresses equal zero.
For a unidirectional composite laminate, g
xx,xy
is therefore given by:
g
xx;xy

c
xy
e
xx

S
16
S
11
40a
which may be written as:
g
xx;xy

2S
11
2S
12
S
66
cos
3
h sin h 2S
22
2S
12
S
66
cos h sin
3
h
S
11
cos
4
h 2S
12
S
66
cos
2
h sin
2
h S
22
sin
4
h
40b
or equivalently:
g
xx;xy

2
E
11

2v
12
E
11

1
G
12
_ _
cos
3
h sin h
2
E
22

2v
12
E
11

1
G
12
_ _
cos h sin
2
h
cos
4
h
E
11

1
G
12

2v
12
E
11
_ _
cos
2
h sin
2
h
sin
4
h
E
22
40c
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
An identical procedure can be employed to dene the properties
measured during a uniaxial test in which only r
yy
is applied. In this case,
Eq. (29) becomes:
e
xx
e
yy
c
xy
_

_
_

S
11
S
12
S
16
S
12
S
22
S
26
S
16
S
26
S
66
_

_
_

_
0
r
yy
0
_

_
_

_
In-plane strains are:
e
xx
S
12
r
yy
41a
e
yy
S
22
r
yy
41b
c
xy
S
26
r
yy
41c
These strains can be used to dene the Youngs modulus E
yy
, Poissons ratio
v
yx
, and coecient of mutual inuence of the second kind g
yy,xy
:
E
yy

1
S
22
42a
v
yx

S
12
S
22
42b
g
yy;xy

S
26
S
22
42c
If desired, these relations can be expanded in terms of compliances referenced
to the 12 coordinate system using Eq. (22) or written in terms of measured
engineering properties using Eq. (18) of Chap. 4.
Next, consider the eective material properties measured during a pure
shear test. A composite laminate subjected to pure shear stress s
xy
is shown in
Fig. 7. Assuming that DT=DM=0, Eq. (29) becomes:
e
xx
e
yy
c
xy
_

_
_

S
11
S
12
S
16
S
12
S
22
S
26
S
16
S
26
S
66
_

_
_

_
0
0
s
xy
_

_
_

_
Hence, the strains caused by a pure shear stress are given by:
e
xx
S
16
s
xy
43a
e
yy
S
26
s
xy
43b
c
xy
S
66
s
xy
43c
In Sec. 2 of Chap. 3, the shear modulus was dened as the shear stress
s
xy
divided by the resulting shear strain c
xy
, with all other stress components
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
equal zero. Applying this denition to the laminate shown in Fig. 7, the shear
modulus referenced to the xy coordinate axes is given by:
G
xy

s
xy
c
xy

1
S
66
44
As before, this expression can be expanded in terms of compliances
referenced to the 12 coordinate system using Eq. (21) or written in terms of
measured engineering properties using Eq. (18) of Chap. 4.
The coecient of mutual inuence of the rst kind g
xy,xx
(or g
xy,yy
) was
dened as the normal strain e
xx
(or e
yy
) divided by the shear strain c
xy
, both
of which are induced by shear stress s
xy
, when all other stresses equal zero.
For a unidirectional composite laminate, the coecient of mutual inuence of
the rst kind g
xy,xx
is therefore given by:
g
xy;xx

e
xx
c
xy

S
16
S
66
45a
while g
xy,yy
is given by:
g
xy;yy

e
yy
c
xy

S
26
S
66
45b
Chentsov coecients were dened in Sec. 2 of Chap. 3 as the shear
strain c
xz
(or c
yz
) divided by the shear strain c
xy
, both of which are induced by
Figure 7 Unidirectional composite subjected to pure shear stress s
xy
.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
shear stress s
xy
, 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 xy plane (s
xy
) and out-of-plane shear strains (c
xz
or c
yz
). Consequently,
Chentsov coecients 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 coecients of thermal expansion
are measured by determining the strains induced by a uniform change in
temperature and forming the following ratios:
a
xx

e
T
xx
DT
a
yy

e
T
yy
DT
a
xy

c
T
xy
DT
46
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 r
xx
=r
yy
=s
xy
=DM=0, Eq. (29) becomes for this
case:
e
xx
e
yy
c
xy
_

_
_

_
DT
a
xx
a
yy
a
xy
_

_
_

_
47
Hence, the thermal expansion coecients for a unidirectional laminate are
given by Eq. (25), repeated here for convenience:
a
xx
a
11
cos
2
h a
22
sin
2
h
a
yy
a
11
sin
2
h a
22
cos
2
h
a
xy
2 cos h sin h a
11
a
22

Similarly, the linear coecient of moisture expansion is measured by
determining the strains induced by a uniform change in moisture content and
forming the following ratios:
b
xx

e
M
xx
DM
b
yy

e
M
yy
DM
b
xy

c
M
xy
DM
48
The superscript Mis 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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Eq. (29). Assuming that r
xx
=r
yy
=s
xy
=DT=0, Eq. (29) becomes for this
case:
e
xx
e
yy
c
xy
_

_
_

_
DM
b
xx
b
yy
b
xy
_

_
_

_
49
Hence, the moisture expansion coecients for a unidirectional laminate are
given by Eq. (28), repeated here for convenience:
b
xx
b
11
cos
2
h b
22
sin
2
h
b
yy
b
11
sin
2
h b
22
cos
2
h
b
xy
2 cos h sin h b
11
b
22

Example Problem 8
Plot the eective properties listed below for a unidirectional hj graphite
epoxy laminate, for all ber angles ranging 0jVh V90j:
(a) Eective Youngs moduli, E
xx
and E
yy
.
(b) Eective Poissons ratio, v
xy
and v
yx
.
(c) Eective shear modulus G
xy
.
(d) Coecients of mutual inuence of the rst kind, g
xy,xx
and g
xy,yy
.
(e) Coecients of mutual inuence of the second kind, g
xx,xy
and g
yy,xy
.
(f) Coecients of thermal expansion, a
xx
, a
yy
, and a
xy
.
(g) Coecients of moisture expansion, b
xx
, b
yy
, and b
xy
.
Use the material properties listed in Table 3 of Chap. 3.
Solution. Plots of the eective elastic properties for unidirectional hj graph-
iteepoxy laminates are presented in Figs. 814.
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 bers are often very brittle, whereas
modern polymeric matrices are fairly ductile. Consequently, most ber-
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 12
plane is provided in Appendix B.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Figure 8 A plot of the effective Youngs moduli E
xx
and E
yy
for unidirectional
graphiteepoxy laminates and fiber angles ranging over 0jVh V90j.
Figure 9 A plot of the effective Poisson ratios v
xy
and v
yx
for unidirectional
graphiteepoxy laminates and fiber angles ranging over 0jVh V90j.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Figure 10 A plot of the effective shear modulus G
xy
for unidirectional
graphiteepoxy laminates and fiber angles ranging over 0jVh V90j.
Figure 11 A plot of the effective coefficients of mutual influence of the first
kind g
xy,xx
and g
xy,yy
for unidirectional graphiteepoxy laminates and fiber
angles ranging over 0jVh V90j.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Figure 12 A plot of the effective coefficients of mutual influence of the second
kind g
xx,xy
and g
yy,xy
for unidirectional graphiteepoxy laminates and fiber
angles ranging over 0jVh V90j.
Figure 13 A plot of the effective coefficients of thermal expansion a
xx
, a
yy
, and
a
xy
for unidirectional graphiteepoxy laminates and fiber angles ranging over
0jVh V90j.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 ber direction, failure is assumed to
involve fracture, whereas transverse to the bers, failure is assumed to
involved yielding, dened on the basis of a % strain oset. 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:

Fracture stress in the 1-direction: r


11
fT
, r
11
fC
.

Yield stress in the 2- and 3-directions: r


22
yT
, r
22
yC
, r
33
yT
, r
33
yC
, s
12
y
, s
13
y
,
s
23
y
.
If the composite is transversely isotropic, then the number of independent
material strengths involved is reduced, since in this case:
r
yT
22
r
yT
33
r
yC
22
r
yC
33
s
y
12
s
y
13
Recall that all of these strengths may vary with temperature.
Figure 14 A plot of the effective coefficients of moisture expansion b
xx
, b
yy
,
and b
xy
for unidirectional graphiteepoxy laminates and fiber angles ranging
over 0jVh V90j.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 eects 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 eects
between individual stress components, consider the two dierent tests of
unidirectional composite shown in Fig. 15. A composite subjected to uniaxial
stress r
11
is shown in Fig. 15(a). This is, of course, the very state of stress used
during the measurement of the fundamental material strength r
11
fT
(as
Figure 15 An illustration of what is meant by coupling effects of stress on
the failure phenomenon.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 r
11
is
increased to r
11
fT
, by denition. 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 (r
22
and s
12
) are applied prior to the appli-
cation of r
11
. It is assumed that the combination of r
22
and s
12
does not cause
failure prior to the application of r
11
. While maintaining r
22
and s
12
at
constant values, stress r
11
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 r
22
and/or s
12
change the value of r
11
at which failure occurs?
Coupling eects refer to the fact that the application of r
22
and/or s
12
often
does alter the value of r
11
necessary to cause failure. If the test represented by
Fig. 8(b) is conducted and r
11
p r
11
fT
at failure, then a coupling eect has oc-
curred, i.e., the presence of r
22
and s
12
has changed the value of r
11
necessary
to cause failure. Conversely, if the test depicted in Fig. 15(b) is performed and
r
11
=r
11
fT
at failure, then no coupling has occurred. Experimental measure-
ments have shown that the coupling phenomenon is much more signicant
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 eects are pronounced for a given material or not. For
metals, the general trend is that coupling eects 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 ber direction but ductile transverse to the ber. It is generally accepted
that coupling eects 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 TsaiHill failure
criterion, and the TsaiWu failure criterion. As will be seen, the maximum
stress criterion does not account for coupling eects, whereas potential
coupling eects are accounted for in the TsaiHill and TsaiWu 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 satised:
1*r
fC
11
< r
11
< r
fT
11
50
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
and
1*r
yC
22
< r
22
< r
yT
22
and
1*r
yC
33
< r
33
< r
yT
33
and
js
12
j < s
y
12
and
js
13
j < s
y
13
and
js
23
j < s
y
23
According to the maximum stress failure criterion, failure (i.e., fracture in the
ber direction or yield-like behavior transverse tothe ber) is predictedstrictly
on the basis of individual stress components. Thus, failure is assumed to be
independent of any coupling eects between individual stress components.
In the case of plane stress (r
33
=s
13
=s
23
=0), the maximum stress
failure criterion reduces to the following ve inequalities:
1*r
fC
11
< r
11
< r
fT
11
and
1*r
yC
22
< r
22
< r
yT
22
and
s
12
j j < s
y
12
51
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 TsaiHill 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 modied version of the
* The von Mises yield criterion is also mathematically equivalent to the octahedral shear
stress and distortional energy yield criteria.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 TsaiHill failure criterion or as the quadratic
failure criterion. For general 3-D states of stress, the TsaiHill criterion
predicts that failure of an orthotropic composite will not occur if the following
inequality is satised:
r
11

2
r
fT
11
_ _
2

r
22

2
r
yT
22
_ _
2

r
33

2
r
yT
33
_ _
2

s
23

2
s
y
23
_ _
2

s
13

2
s
y
13
_ _
2

s
12

2
s
y
12
_ _
2
r
11
r
22
1
r
fT
11
_ _
2

1
r
yT
22
_ _
2

1
r
yT
33
_ _
2
_

_
_

_
r
11
r
33
1
r
fT
11
_ _
2

1
r
yT
22
_ _
2

1
r
yT
33
_ _
2
_

_
_

_
r
22
r
33
1
r
fT
11
_ _
2

1
r
yT
22
_ _
2

1
r
yT
33
_ _
2
_

_
_

_ < 1
52
In the case of plane stress conditions (r
33
=s
13
=s
23
=0), the TsaiHill
criterion reduces to:
r
11

2
r
fT
11
_ _
2

r
22

2
r
yT
22
_ _
2

s
12

2
s
y
12
_ _
2
r
11
r
22
1
r
fT
11
_ _
2

1
r
yT
22
_ _
2

1
r
yT
33
_ _
2
_

_
_

_ < 1 53
It is interesting to note that according to the TsaiHill failure criterion,
failure of orthotropic composites is sensitive to the out-of-plane strength term
(r
33
yT
), even for plane stress conditions.
If the composite is transversely isotropic (that is, if r
33
yT
=r
22
yT
), then Eq.
(53) reduces to:
r
11

2
r
fT
11
_ _
2

r
22

2
r
yT
22
_ _
2

s
12

2
s
y
12
_ _
2

r
11
r
22
r
fT
11
_ _
2
< 1 54
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
A potential advantage of the TsaiHill failure criterion is that coupling
eects between individual stress components are accounted for, in contrast
with the maximum stress failure criterion. On the other hand, most com-
posites exhibit signicantly dierent failure strengths in tension and com-
pression (as indicated in Table 3 of Chap. 3), and so a shortcoming of the
TsaiHill failure criterion is that it does not directly account for these dif-
ferences. That is, an implicit assumption of the TsaiHill criterion (as well as
the original von Mises criterion) is that failure strengths in tension and
compression have equal magnitudes. Hence, only tensile strengths r
11
fT
, r
22
yT
,
and r
33
yT
appear in Eqs. (53) and (54). Dierences in tensile and compressive
strengths can be accounted for articially in the TsaiHill 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 r
11
,
r
22
, and s
12
, and also that r
11
is tensile but r
22
is compressive. In such a case,
the dierences in tensile/compressive strengths can be accounted for by using
the tensile strength in the 1-direction, r
11
fT
, but the compressive strength in
the 2-direction, r
22
yC
.
While the TsaiHill criterion can be modied in this way to account for
dierences in tensile and compressive strengths, it would be ideal if a failure
criterion was available that accounts for both coupling eects and dierences
in tensile and compressive strengths automatically. One such criterion is
the TsaiWu criterion, described in the next paragraph.
5.3 The TsaiWu 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-Dstates of
stress, the TsaiWu criterion predicts that failure will not occur if the
following inequality is satised:
X
1
r
11
X
2
r
22
X
3
r
33
X
11
r
2
11
X
22
r
2
22
X
33
r
2
33
X
44
s
2
23
X
55
s
2
13
X
66
s
2
12
2X
12
r
11
r
22
2X
13
r
11
r
33
2X
23
r
22
r
33
< 1
55
Most of the constants that appear in this inequality (i.e., X
1
, X
2
, X
3
, X
11
, etc.)
can be determined based on fundamental strength measurements (i.e., r
11
fT
,
r
11
fC
, r
22
yT
, r
22
yC
, etc.). First, consider a uniaxial strength measurement in which
only stress r
11
is applied (that is, a test in which r
11
p 0, r
22
=r
33
=s
23
=s
13
=s
12
=0). Stress r
11
is increased monotonically from zero until
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
failure occurs. If r
11
is tensile, then at the moment of failure, r
11
=r
11
fT
, and the
TsaiWu criterion reduces to:
X
1
r
fT
11
X
11
r
fT
11
_ _
2
1
Conversely, if r
11
is compressive, then at the moment of failure, r
11
=r
11
fC
(where the measured compressive strength, r
11
fC
, is treated as an algebraically
positive number), and the TsaiWu criterion reduces to:
X
1
r
fC
11
X
11
r
fC
11
_ _
2
1
Solving for X
1
and X
11
, we nd:
X
1

1
r
fT
11

1
r
fC
11
X
11

1
r
fT
11
r
fC
11
56
Similarly, if failure is measured during two uniaxial stress tests in which only
r
22
is applied, we nd:
X
2

1
r
yT
22

1
r
yC
22
X
22

1
r
yT
22
r
yC
22
57
Using measurements obtained during two tests in which only r
33
is applied:
X
3

1
r
yT
33

1
r
yC
33
X
33

1
r
yT
33
r
yC
33
58
Three additional constants are determined using measured shear strengths:
X
44

1
s
y
23
_ _
2
X
55

1
s
y
13
_ _
2
X
66

1
s
y
12
_ _
2
59
Only three coecients remain to be determined, X
12
, X
13
, and X
23
. Several
methods of determining these coecients 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, X
12
can be determined by conducting
a biaxial test in which r
11
=r
22
=r and r
33
=s
23
=s
13
=s
12
=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 r
11
=r
22
=r
y
and
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
r
33
=s
23
=s
13
=s
12
=0. Substituting these values into the TsaiWu criterion
and solving for X
12
results in:
X
12

1
2 r
y

2
1 X
1
X
2
r
y
X
11
X
22
r
y

2
_ _
60
At least conceptually, X
13
and X
23
can also be determined in a similar manner.
Two additional biaxial tests to failure would be required, where in one test,
r
11
=r
33
=r, and in the second test, r
22
=r
33
=r. These data would then
allow the calculation of X
13
and X
23
, respectively. In practice, however, these
tests would be very dicult to perform. Since composites are usually quite
thin, it is especially dicult to apply well-dened out-of-plane stress compo-
nents (i.e., r
33
, s
13
, or s
23
). Hence, in most instances, determining X
13
or X
23
in
this manner is impractical.
A second approach is to assume that X
12
, X
13
, and X
23
can be calculated
as follows:
X
12

1
2

X
11
X
22
_

1
2

r
fT
11
r
fC
11
r
yT
22
r
yC
22
_
X
13

1
2

X
11
X
33
_

1
2

r
fT
11
r
fC
11
r
yT
33
r
yC
33
_
X
23

1
2

X
22
X
33
_

1
2

r
yT
22
r
yC
22
r
yT
33
r
yC
33
_
61
The basis of this approach is that if Eq. (61) is enforced and isotropic
strengths are assumed (i.e., if r
11
fT
=r
11
fC
=r
22
yT
=r
22
yC
=r
33
yT
=r
33
yC
=r
y
, and s
y
12
s
y
13
s
y
23
r
y
=

3
p
), then the TsaiWu 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 X
12
,
X
13
, and X
23
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 12 plane. In this case, the
TsaiWu criterion reduces to:
X
1
r
11
X
2
r
22
X
11
r
2
11
X
22
r
2
22
X
66
s
2
12
2X
12
r
11
r
22
< 1 62
Hence, in the plane stress case, six constants are involved, ve of which can
be calculated using readily available strength data (r
11
fT
, r
11
fC
, r
22
yT
, etc.). Only
one problematic coecient remains, X
12
. This term can be determined using
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
an o-axis specimen (which is, in eect, a biaxial test). For example, suppose a
uniaxial stress r
xx
is applied to a unidirectional composite specimen in which
the bers are oriented at h=45j with respect to the direction of loading.
Under these conditions, the stresses in the 12 coordinate system are easily
calculated:
r
11
r
22
s
12
_
_
_
_
_
_

cos
2
45j sin
2
45j 2cos 45j sin 45j
sin
2
45j cos
2
45j 2cos 45j sin 45j
cos 45j sin 45j cos 45j sin 45j cos
2
45j sin
2
45j
_

_
_

_
r
xx
0
0
_

_
_

_
or
r
11
r
22
s
12
_

_
_

r
xx
cos
2
45j
r
xx
sin
2
45j
r
xx
cos 45j sin 45j
_

_
_

r
xx
=2
r
xx
=2
r
xx
=2
_

_
_

_
The strength of the 45j o-axis specimen is measured by increasing
stress r
xx
until failure occurs. Let us assume that failure occurs due to yielding
and denote the stress level at which failure occurs as r
xx
=r
xx
y
. At failure, the
ply stresses are r
11
=r
22
=s
12
=r
xx
y
/2. Substituting these stresses into Eq.
(62) and solving for X
12
, we nd:
X
12

1
2 r
y
xx

2
4 r
y
xx
2 X
1
X
2
r
y
xx
X
11
X
22
X
66

_ _ _
63
While this example has been based on a 45j o-axis specimen, a similar
approach can be used with any hj o-axis specimen.
From an analytical standpoint, the TsaiWu failure criterion is an
improvement over the other two failure criteria considered. First, unlike the
maximum stress failure criterion, the coupling eects between individual
stress components are accounted for in the TsaiWu criterion. Second, unlike
the TsaiHill criterion, dierences in tensile and compressive strengths are
automatically and naturally accounted for via the X
1
, X
11
, X
2
, X
22
, X
3
, and
X
33
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 r
xx
, r
yy
, and s
xy
are referenced to an arbi-
trary xy coordinate system. There are, of course, an innite number of
dierent combinations of r
xx
, r
yy
, and s
xy
that (collectively) dene a state of
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
plane stress. For illustrative purposes, two simple stress states will be con-
sidered: rst, a state of uniaxial stress (i.e., r
xx
p 0, r
yy
=r
xy
=0), and second,
a state of pure shear (s
xy
p 0, r
xx
=r
yy
=0).
Numerical results for a unidirectional graphiteepoxy composite will be
used to facilitate these comparisons. The following failure strengths are taken
from Table 3 of Chap. 3 and are typical for graphiteepoxy at room temper-
ature:
r
fT
11
1500 MPa r
yT
22
50 MPa s
y
12
75 MPa
r
fC
11
1200 MPa r
yC
22
100 MPa
6.1 Uniaxial Stress
An o-axis composite ply subjected to a uniaxial stress r
xx
has been
previously shown in Fig. 6. The stresses induced in the 12 coordinate system
by stress r
xx
can be determined using Eq. (20) of Chap. 2:
r
11
r
22
s
12
_

_
_

cos
2
h sin
2
h 2 cos h sin h
sin
2
h cos
2
h 2 cos h sin h
cos h sin h cos h sin h cos
2
h sin
2
h
_

_
_

_
r
xx
0
0
_

_
_

_
or equivalently:
r
11
r
xx
cos
2
h
r
22
r
xx
sin
2
h 64
s
12
r
xx
cos h sin h
6.1.1 Maximum Stress Criterion
Substituting Eq. (64) into Eq. (50), we obtain:

r
fC
11
cos
2
h
< r
xx
<
r
fT
11
cos
2
h
and
r
yC
22
sin
2
h
< r
xx
<
r
yT
22
sin
2
h
and
r
xx
j j <
s
y
12
coshsin h
65
According to the maximum stress criterion, failure will not occur if
these ve inequalities are satised. The predicted tensile and compressive
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
failure strengths, r
xx
fT
and r
xx
fC
, respectively, for a hj o-axis graphiteepoxy
laminate are therefore the smallest values returned by the following expres-
sions:*
Tensile strength:
r
fT
xx

r
fT
11
cos
2
h

1500 MPa
cos
2
h
66a
r
fT
xx

r
yT
22
sin
2
h

50 MPa
sin
2
h
66b
r
fT
xx

s
y
12
cos h sin h

75 MPa
cos h sin h

66c
Compressive strength:
r
fC
xx

r
fC
11
cos
2
h

1200 MPa
cos
2
h
67a
r
fC
xx

r
yC
22
sin
2
h

100 MPa
sin
2
h
67b
r
fC
xx

s
y
12
cos h sin h

75 MPa
cos h sin h

67c
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 bers or yielding of the matrix.
Equations (66a)(66c) and (67a) (67b) (67c) were used to create the failure
envelope for a unidirectional graphiteepoxy 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. Specically, Fig. 16 is valid only if:
r
xx
p 0
r
yy
r
zz
s
xy
s
xz
s
yz
0
As will be seen later, failure envelopes for other states of stress dier
substantially from Fig. 16.
* As before, compressive strength is treated as an algebraically positive number.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

The mode of failure depends on whether r


xx
is tensile or compressive,
and also on the ber angle h:
If r
xx
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 r
xx
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 ber 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
Figure 16 Failure envelope for a unidirectional graphiteepoxy laminate
subjected to a uniaxial stress r
xx
, based on the maximum stress criterion.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
ber failure. In general, ber failure will only occur when the composite is
tested under carefully controlled laboratory conditions in which the uniaxial
stress is aligned with the ber direction to within a fewdegrees. Also, note that
the ber angle at which a change from shear failure to matrix failure occurs
diers in tension and compression.
6.1.2 TsaiHill Criterion
According to the TsaiHill 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 TsaiHill failure criterion predicts
that failure will not occur if the following inequality is satised:
r
xx
<
cos
2
h cos
2
h sin
2
h
_
r
fT
11
_ _
2

sin
4
h
r
yT
22
_ _
2

cos
2
h sin
2
h
s
f
12
_ _
2
_

_
_

_
1=2
68
As previously noted, the TsaiHill criterion does not automatically account
for dierences in tensile and compressive strengths. A failure envelope for a
unidirectional graphiteepoxy composite will be generated using tensile or
compressive strengths, as appropriate. Thus, the tensile strength predicted by
the TsaiHill criterion is:
r
fT
xx

cos
2
h cos
2
h sin
2
h
_
r
fT
11
_ _
2

sin
4
h
r
yT
22
_ _
2

cos
2
h sin
2
h
s
f
12
_ _
2
_

_
_

_
1=2
Similarly, the compressive strength predicted by the TsaiHill criterion is:
r
fC
xx

cos
2
h cos
2
h sin
2
h
_
r
fC
11
_ _
2

sin
4
h
r
yC
22
_ _
2

cos
2
h sin
2
h
s
f
12
_ _
2
_

_
_

_
1=2
Substituting the strength values that have been assumed for graphiteepoxy,
we have:
r
fT
xx

cos
2
h cos
2
h sin
2
h
_
1500 MPa
2

sin
4
h
50 MPa
2

cos
2
h sin
2
h
75 MPa
2
_ _
1=2
69a
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
r
fC
xx

cos
2
h cos
2
h sin
2
h
_
1200 MPa
2

sin
4
h
100 MPa
2

cos
2
h sin
2
h
75 MPa
2
_ _
1=2
69b
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 TsaiHill criterion
but for other states of stress dier substantially from Fig. 17.
6.1.3 TsaiWu Criterion
The TsaiWu criterion for the case of plane stress is given by Eq. (62).
Substituting Eq. (64) into Eq. (62), we obtain:
r
f
xx
_ _
2
X
11
cos
4
h X
22
sin
4
h cos
2
h sin
2
h X
66
2X
12

_
r
f
xx
X
1
cos
2
h X
2
sin
2
h
_
1 0
70
Figure 17 Failure envelope for a unidirectional graphiteepoxy laminate
subjected to a uniaxial stress r
xx
, based on the TsaiHill criterion.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
The constants that appear in Eq. (70) are calculated using the strength
properties that have been assumed for graphiteepoxy and Eqs. (56)(59),
as appropriate:
X
1

1
r
fT
11

1
r
fC
11

1
1500 MPa

1
1200 MPa

10
6
6000 Pa
X
11

1
r
fT
11
r
fC
11

1
1500 MPa 1200 MPa

10
15
1800 Pa
2
X
2

1
r
yT
22

1
r
yC
22

1
50 MPa

1
100 MPa

10
6
100 Pa
X
22

1
r
yT
22
r
yC
22

1
50 MPa 100 MPa

10
15
5 Pa
2
X
66

1
s
y
12
_ _
2

1
75 MPa
_ _
2

10
12
5625 Pa
2
As previously discussed, there is no widely accepted technique used to
calculate X
12
. For present purposes, X
12
will be calculated in accordance with
Eq. (61):
X
12

1
2

X
11
X
22
_

1
2

10
15
1800 Pa
2
_ _
10
15
5 Pa
2
_ _

10
p
600 Pa
2
_ _
10
15
_ _
Substituting these values into Eq. (70), we obtain:
r
f
xx
_ _
2
_
10
15
_ _
cos
4
h
1800 Pa
2

10
15
_ _
sin
4
h
5 Pa
2
cos
2
h sin
2
h

10
12
5625 Pa
2

10
15
_ _
10
p
300 Pa
2
_ __
r
f
xx
10
6
cos
2
h
6000 Pa

10
6
sin
2
h
100 Pa
_ _
1 0
71
Equation (71) is a second-order polynomial in the unknown failure stress,
r
xx
f
. For any given ber angle h, there will be two roots to this equation. The
predicted tensile strength equals the algebraically positive root, whereas the
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
predicted compressive strength equals the negative root. For example, for a
ber angle h=30j, Eq. (71) becomes:
r
f
xx
_ _
2 4417 10
20
Pa
2
_ _
r
f
xx
2375 10
12
Pa
_ _
1 0
The two roots of this expression are found to be (125.910
6
Pa, 179.710
6
Pa). Hence, the strengths predicted by the TsaiWu criterion for a 30j
graphiteepoxy specimen are:
r
fT
xx
125:9 MPa
r
fC
xx
179:7 MPa
A failure envelope based on the TsaiWu criterion for a unidirectional
graphiteepoxy laminate subjected to a uniaxial state of stress is shown in
Fig. 18. This gure is analogous to those obtained using the maximum stress
criterion and the TsaiHill criterion (Figs. 16 and 17, respectively). As before,
Figure 18 Failure envelope for a unidirectional graphiteepoxy laminate
subjected to a uniaxial stress r
xx
, based on the TsaiWu criterion.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
it is important to realize that this failure envelope is valid for a uniaxial state of
stress only. Failure envelopes based on the TsaiWu criterion but for other
states of stress dier 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
viewof just the rst quadrant is presented in Fig. 20. It is apparent that similar
predictions are obtained on the basis of all three criteria, although a
signicant numerical dierence occurs at low ber angles, near the region
at which the failure mode shifts from ber 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 dier substantially. One
stress state that exhibits this eect 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 r
xx
, obtained using the maximum
stress, TsaiHill, and TsaiWu failure criteria.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 o-axis composite ply
subjected to a pure shear stress state is shown in Fig. 21. The stresses induced
in the 12 coordinate system can be determined using Eq. (20) of Chap. 2:
r
11
r
22
s
12
_

_
_

cos
2
h sin
2
h 2 cos h sin h
sin
2
h cos
2
h 2 cos h sin h
cos h sin h cos h sin h cos
2
h sin
2
h
_

_
_

_
0
0
s
xy
_

_
_

_
or equivalently:
r
11
2s
xy
cos h sin h
Figure 20 Comparison of the failure envelopes (first quadrant only) for a
unidirectional graphiteepoxy laminate subjected to a uniaxial stress r
xx
,
obtained using the maximum stress, TsaiHill, and TsaiWu failure criteria.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
r
22
2s
xy
cos h sin h
s
12
s
xy
cos
2
h sin
2
h
_
72
6.2.1 Maximum Stress Criterion
Substituting Eq. (72) into Eq. (51), we obtain:
r
fC
11
2cos h sin h
< s
xy
<
r
fT
11
2cos h sin h
73a
(and)
r
yC
22
2cos h sin h
< s
xy
<
r
yT
22
2cos h sin h
73b
(and)
s
xy

<
s
y
12
cos
2
h sin
2
h
73c
Equations (73a)(73c) will now be used to generate a failure envelope for a
graphiteepoxy 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
Figure 21 Unidirectional composite subjected to pure shear stress s
xy
.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
shear stress s
xy
will induce a tensile value r
11
and a compressive value
for r
22
. 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 s
xy
will induce a compressive stress r
11
but tensile stress r
22
.
Of course, if a negative shear stress s
xy
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 dier in tension and compression.
With these observations in mind, the following equations may be used
to generate a failure envelope for a graphiteepoxy laminate subjected to pure
shear, based on the strength properties previously listed:
Positive shear strengths:

For 0j<h<90j:
s
P
xy

r
fT
11
2 cos h sin h

1500 MPa
2 cos h sin h
74a
s
P
xy

r
yC
22
2 cos h sin h

100 MPa
2 cos h sin h
74b
s
P
xy

s
y
12
cos
2
h sin
2
h

75 MPa
cos
2
h sin
2
h
74c

For 90j<h<0j:
s
P
xy

r
fC
11
2 cos h sin h

1200 MPa
2 cos h sin h
74d
s
P
xy

r
yT
22
2 cos h sin h

50 MPa
2 cos h sin h
74e
s
P
xy

s
y
12
cos
2
h sin
2
h

75 MPa
cos
2
h sin
2
h
74f
Negative shear strengths:

For 0j<h<90j:
s
N
xy

r
fC
11
2 cos h sin h

1200 MPa
2 cos h sin h
75a
s
N
xy

r
yT
22
2 cos h sin h

50 MPa
2 cos h sin h
75b
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
s
N
xy

s
y
12
cos
2
h sin
2
h

75 MPa
cos
2
h sin
2
h
75c

For 90j<h<0j:
s
N
xy

r
fT
11
2 cos h sin h

1500 MPa
2 cos h sin h
75d
s
N
xy

r
yC
22
2 cos h sin h

100 MPa
2 cos h sin h
75e
s
N
xy

s
y
12
cos
2
h sin
2
h

75 MPa
cos
2
h sin
2
h
75f
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 dier sub-
stantially (for example, compare Figs. 16 and 22, both of which are
based on the maximum stress failure criterion).
Figure 22 Failure envelope for a unidirectional graphiteepoxy laminate
subjected to a pure shear stress s
xy
, based on the maximum stress criterion.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

None of the curves shown in Fig. 22 is associated with ber 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 graphiteepoxy laminate sub-
jectedto a pure shear stress state will never occur due to ber failure.

The shear strength of an o-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 s


xy
is positive or
negative as well as on ber angle h.
If s
xy
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.
Matrix failure is predicted for 26.6j<h<63.4j.
Shear failure is predicted for 63.4j<h<90j.
If s
xy
is negative, then:
Shear failure is predicted for 90j<h<63.4j.
Matrix failure is predicted for 63.4j<h<26.6j.
Shear failure is predicted for 26.6j<h<16.8j.
Matrix failure is predicted for 16.8j<h<73.2j.
Shear failure is predicted for 73.2j<h<90j.
6.2.2 TsaiHill Criterion
Substituting Eq. (72) into Eq. (54), we obtain:
s
2
xy
4 cos
2
h sin
2
h
2
r
fT
11
_ _
2

1
r
yT
22
_ _
2

1
2 s
y
12
_ _
2
_

_
_

_
cos
4
h sin
4
h
s
y
12
_ _
2
_

_
_

_
< 1
77
Equating the left-hand side to unity and solving for s
xy
:
s
xy

1
4 cos
2
h sin
2
h
2
r
fT
11

2

1
r
yT
22

2

1
2 s
y
12

2
_ _

cos
4
h sin
4
h
s
y
12
_ _
2
_

_
_

_
1=2
78
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Recall that the TsaiHill criterion does not automatically account for dier-
ences in tensile and compressive stresses. Therefore, to predict shear strengths
using Eq. (78), the failure strengths used must be selected according to
whether r
11
and r
22
are positive or negative.
Positive shear strengths:

For 90j<h<0j, both r


11
and r
22
are positive; therefore, use
r
11
fT
and r
22
yT
.

For 0j<h<90j, r
11
is positive and r
22
is negative; therefore,
use r
11
fT
and r
22
yC
.
Negative shear strengths:

For 90j<h<0j, both r


11
and r
22
are negative; therefore, use
r
11
fC
and r
22
yC
.

For 0j<h<90j, r
11
is negative and r
22
is positive; therefore,
use r
11
fC
and r
22
yT
.
A failure envelope based on these failure strengths and Eq. (78) is shown in
Fig. 23.
Figure 23 Failure envelope for a unidirectional graphiteepoxy laminate sub-
jected to a pure shear stress s
xy
, based on the TsaiHill criterion.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
6.2.3 TsaiWu Criterion
Substituting Eq. (72) into Eq. (62), we obtain:
s
2
xy
X
66
cos
4
h sin
4
h
_
2 cos
2
h sin
2
h 2X
11
2X
22
79
_ _
4X
12
X
66
__
2s
xy
cos h sin h X
1
X
2
< 1
Numerical values for constants X
1
, X
2
, X
11
, etc., were calculated in Sec. 6.1.3
(based on strengths assumed for graphiteepoxy). Substituting these values in
the left-hand side of Eq. (79) and equating to unity, we obtain:
s
2
xy
10
12
5625Pa
2
cos
4
h sin
4
h
_
cos
2
h sin
2
h
_

67 10
15
150Pa
2

10
p _ _
10
15
_ _
75Pa
2
_ __
s
xy
cos h sin h
61 10
6
3000
_ _
1
80
Equation (80) is a second-order polynomial in the unknown shear failure
stress, s
xy
f
. For a given ber angle, h, there are two roots to this equation. The
predicted positive shear strength, s
xy
fP
, equals the algebraically positive root,
whereas the predicted negative shear strength, s
xy
fN
, equals the negative root.
For example, for a ber angle h=30j, Eq. (80) becomes:
s
f
xy
_ _
2
202:8 10
18
Pa
2
_ _
s
f
xy
88:05 10
10
Pa
_ _
1 0
The two roots of this expression are found to be (95.22 10
6
Pa, 51.8 10
6
Pa). Hence, the shear strengths predicted by the TsaiWu criterion for a 30j
graphiteepoxy laminate are:
s
fP
xy
95:2 MPa
s
fN
xy
51:8 MPa
A failure envelope based on the TsaiWu criterion for a unidirectional
graphiteepoxy laminate subjected to a pure shear stress state is shown in
Fig. 24. This gure is analogous to those obtained using the maximum stress
criterion and the TsaiHill 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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
view of just the rst quadrant is presented in Fig. 26. The dierence 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 TsaiHill and TsaiWu criteria are similar,
although some dierence exists. The maximum stress criterion predicts local
maximums in shear strength near ber 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 24 Failure envelope for a unidirectional graphiteepoxy laminate sub-
jected to a pure shear stress s
xy
, based on the TsaiWu criterion.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 dene these values using a consistent set of units. For example,
program UNIDIR requires the user to input elastic moduli, thermal expan-
sion coecients, and moisture expansion coecients 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 graphiteepoxy:
E
11
170 10
9
Pa E
22
10 10
9
Pa v
12
0:30
G
12
13 10
9
Pa
a
11
0:9 10
6
m=m
o
C a
22
27:0 10
6
m=m
o
C
b
11
150:0 10
6
m=m%M b
22
4800 10
6
m=m%M
Figure 25 Comparison of the failure envelopes for a unidirectional graphite
epoxy laminate subjected to pure shear stress s
xy
, obtained using the Maximum
Stress, TsaiHill, and TsaiWu failure criteria.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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
r
xx
=20010
6
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 e
xx
=200010
6
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 graphiteepoxy material
system:
E
11
25:0 10
6
psi E
22
1:5 10
6
psi v
12
0:30
G
12
1:9 10
6
psi
a
11
0:5 10
6
in:=in:
o
F a
22
15 10
6
in:=in:
o
F
b
11
150:0 10
6
in:=in:%M b
22
4800 10
6
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 r
xx
=30,000
psi. If, instead, the analysis requires the user to input numerical values for
Figure 26 Comparison of the failure envelopes (first quadrant only) for a
unidirectional graphiteepoxy laminate subjected to a uniaxial stress r
xx
,
obtained using the maximum stress, TsaiHill, and TsaiWu failure criteria.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
strains, then strains must be input in in./in. (not in Ain./in.). A typical value
would be e
xx
=200010
6
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 dierent types of analyses may be performed. The program may
be used in either of the following:

Calculating total strains (e


xx
, e
yy
, c
xy
) caused by a specied com-
bination of stresses (r
xx
, r
yy
, s
xy
), 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 (r
xx
, r
yy
, s
xy
) caused by a specied combination
of total strains (e
xx
, e
yy
, c
xy
), a uniformtemperature 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 eective properties of a unidirectional
composite based on the denitions described in Sec. 4. An implicit assumption
in these calculations is that all material properties (E
11
, a
11
, b
11
, 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, TsaiHill, or TsaiWu
failure criteria introduced in Sec. 5. Two dierent 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 specied ber angle h.

Generating a data le 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 le (named Envelop.txt) that contains the
stress(es) predicted to cause failure of a unidirectional composite as a function
of ber angle, based on the particular failure criterion specied by the user. A
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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. ber angle. For example, any of the failure envelopes
presented in Sec. 6 may be easily recreated in this way. As before, an implicit
assumption is that failure strengths (r
11
fT
, a
22
yT
, 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 bers 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,
rst 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 stiness matrix for the materials listed below, rst
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:
r
11
350 MPa
r
22
40 MPa
s
12
60 MPa
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Determine the resulting strains (e
11
, e
22
, and c
12
), rst 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):
e
11
1250 Ain:=in:
e
22
2000 Ain:=in:
c
12
0
Determine the stresses (r
11
, r
22
, and s
12
), rst by hand calculation and
then using program UNIDIR. Use the material properties listed in
Table 3 of Chap. 3.
7. Repeat Problem 6 for a unidirectional Kevlar/epoxy composite lami-
nate.
8. Repeat Problem 6 for a unidirectional graphite/epoxy composite lami-
nate.
9. A square unidirectional glass/epoxy composite laminate with dimen-
sions 11 m is clamped between four innitely 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 coecient of the rigid walls
is zero, so the rigid walls do not expand or contract.
(a) Calculate the stresses (r
11
, r
22
, s
12
) induced by this change in
temperature, rst by hand calculation and then using
program UNIDIR.
(b) Predict whether the composite will fail based on the maxi-
mum stress failure criterion.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
10. Repeat Problem 9 for a unidirectional Kevlar/epoxy composite lami-
nate.
11. Repeat Problem 9 for a unidirectional graphite/epoxy composite lami-
nate.
12. A 11 m square unidirectional glass/epoxy composite laminate is
placed within a cavity dened 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 ll 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 (r
11
,
r
22
, s
12
) caused by the change in moisture content.
13. Repeat Problem 12 for a unidirectional Kevlar/epoxy composite lami-
nate.
14. Repeat Problem 12 for a unidirectional graphite/epoxy composite lami-
nate.
15. A perfectly square unidirectional glass/epoxy composite laminate is
mounted in a frame consisting of four innitely rigid frame members,
Figure 27 Clamped composite laminate considered in Problems 9, 10, and 11.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
as shown in Fig. 29(a). The frame members are pinned at each corner.
Since the laminate is perfectly square, the angle dened 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 (r
11
,
r
22
, s
12
) are induced in the panel?
16. Repeat Problem 15 for a unidirectional Kevlar/epoxy composite lami-
nate.
17. Repeat Problem 15 for a unidirectional graphite/epoxy composite lami-
nate.
18. Create a plot of the following eective properties for a unidirectional
glass/epoxy composite, with ber angles ranging from 90j to +90j.
(a) E
xx
and E
yy
.
(b) v
xy
and v
yx
.
Figure 28 A composite laminate placed within a cavity defined by four rigid
walls (considered in Problems 12, 13, and 14).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
(c) G
xy
.
(d) g
xy,xx
and g
xy,yy
.
(e) g
xx,xy
and g
yy,xy
.
(f ) a
xx
, a
yy
, and a
xy
.
(g) b
xx
, b
yy
, and b
xy
.
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.
19. Repeat Problem 18 for a unidirectional Kevlar/epoxy composite lami-
nate.
Figure 29 Unidirectional composite laminate considered in Problems 15, 16,
and 17.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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, TsaiHill, and TsaiWu failure cri-
teria. Compare these results with your calculations obtained in part
(a).
21. Repeat Problem 20 for a unidirectional Kevlar/epoxy composite lami-
nate.
22. Repeat Problem 20 for a unidirectional graphite/epoxy composite lami-
nate.
23. On the same graph, plot failure envelopes for a unidirectional glass/
epoxy composite laminate for the following two conditions:
(a) Unidirectional stress: r
xx
, r
yy
=s
xy
=0
(b) Biaxial normal stress: r
yy
=r
xx
/10, s
xy
=0
Use the TsaiHill failure criterion. Suggested solution procedure: use
program UNIFAIL (twice) to generate data les corresponding to the
specied loading conditions, and then plot these data les.
24. Repeat Problem 23 for a unidirectional Kevlar/epoxy composite lami-
nate.
25. Repeat Problem 23 for a unidirectional graphite/epoxy composite lami-
nate.
REFERENCES
1. Reuter, R.C. Concise property transformation relations for an anisotropic lami-
na. J. Compos. Mater., Vol. 5, April 1971, 270272.
2. Hill, R. The Mathematical Theory of Plasticity; New York, Oxford University
Press, 1998.
3. Tsai, S.W. Strength theories of lamentary structures. In: Fundamental Aspects
of Fiber Reinforced Plastic Composites; Schwartz, R.T., Schwartz, H.S., eds;
Wiley Interscience: New York, 1968; 311.
4. Tsai, S.W.; Wu, E.M. A general theory of strength for anisotropic materials. J.
Compos. Mater., Vol. 5, January 1971, 5880.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 xyz coordinate systemis dened as indicated. Note that
the origin of the xyz 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 dened
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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 xy 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 xy plane, these load components are
often referred to as in-plane loads.
First, consider load components N
xx
, N
yy
, and N
xy
. 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 rst subscript indicates the face of the plate a given load acts upon,
whereas the second subscript indicates the line of action of the load. Apositive
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 N
xx
and N
yy
(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 N
xy
and N
yx
(Fig.
2b) is not as immediately apparent, but application of the sign convention just
Figure 1 A thin plate with in-plane dimensions a and b and thickness t.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Figure 2 Schematic of allowable plate loadings: (a) in-plane normal forces N
xx
and N
yy
, (b) in-plane shear forces N
xy
and N
yx
, (c) bending moment M
xx
, (d)
bending moment M
yy
, (e) in-plane torques M
xy
and M
yx
.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
described will conrm 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, jN
xy
j =
jN
yx
j. These requirements are also analogous to those of shear stresses acting
on an innitesimal stress element, as discussed in Sec. 2.5. It is emphasized
that N
xx
, N
yy
, and N
xy
are all dened as distributed loads, expressed in units of
force/plate length (such as N/m or lbf/in.).
The remaining loads shown in Fig. 2 (M
xx
, M
yy
, and M
xy
) are bending
moments (or torques) distributed along the edge of the plate. Load compo-
nents M
xx
and M
yy
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 M
xx
and M
yy
may seempuzzling at rst because, for example, M
xx
represents a bending moment acting about the y-axis. However, M
xx
is
directly related to the distribution of r
xx
through the thickness of the plate,
as will be shown below. Because M
xx
arises due to the distribution of r
xx
(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 dened 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 M
xx
tends to cause tensile stresses in the positive z-face (i.e., the
lower face) of the plate. Hence, M
xx
is positive as drawn. A similar
observation holds for M
yy
, as shown in Fig. 2d.
Finally, loads M
xy
and M
yx
are dened as uniformly distributed in-
plane moments (or torques) acting along neighboring edges of the plate, as
shown in Fig. 2d. It will be shown belowthat M
xy
and M
yx
are directly related
to in-plane shear stresses s
xy
and s
yx
, respectively. Because M
xy
and M
yx
arise
due to the distribution of s
xy
and s
yx
, it can be shown that for statical
equilibrium to be maintained, jM
xy
j = jM
yx
j. An algebraically positive
distributed torque is dened 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 forcelength
(such as Nm or lbfin). Because M
xx
, M
yy
, and M
xy
all represent uniformly
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
distributed moments acting along the plate edge, they are all expressed in units
of forcelength/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 denitions 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 dened 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
beamis 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 beamis 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 specied in terms of distributed loads, as described
in preceding paragraphs. For example, in SI units, N
xx
is expressed in terms of
Newtons per meter, whereas M
xx
is expressed in terms of Newton meter per
meter. In contrast, in fundamental beam theory, point loads are often
specied. 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 N
xx
and M
xx
, when applied to a beam, have
also been shown in Fig. 3. They have been denoted as N
b
xx
and M
b
xx
, where the
superscript b is used to denote that these loads are dened in the sense
traditionally used in beam theory and therefore have dierent units than the
Figure 3 Aprismatic beamwithrectangular cross-section (compare withFig. 1).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
distributed loads used in plate theory. Because the width of the beam is b, the
two load denitions are related according to N
b
xx
=bN
xx
and M
b
xx
=bM
xx
. As
mentioned above, the sign convention used to dene 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 diers 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 viewof a plate loaded only by distributed load N
xx
and moment M
xx
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
N
xx
and distributed moment M
xx
on the left-hand side, and the resulting
internal stress r
xx
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 r
xx
through the thickness of the plate must be exactly
balanced by the distributed load N
xx
. Let the free-body diagramhave 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 dF
xx
associated with the
stress acting over this thin strip is dF
xx
=dN
xx
1=r
xx
dA=r
xx
dz. We can now
relate the total distributed force N
xx
acting on the left-hand side of the free-
body diagram to the distribution of r
xx
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:
N
xx

_
t=2
t=2
dN
xx

_
t=2
t=2
r
xx
dz 1a
In an entirely equivalent manner, we can relate distributed forces N
yy
and N
xy
to stresses r
yy
and s
xy
, respectively:
N
yy

_
t=2
t=2
r
yy
dz 1b
N
xy

_
t=2
t=2
s
xy
dz 1c
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Now consider moment M
xx
. As before, the plate is assumed to be in
statical equilibrium, and hence moments acting about the y-axis must sum to
zero: SM
y
=0. Again consider an incremental strip of height dz, which is
located a distance z fromthe midsurface. The incremental distributed moment
dM
xx
contributed by N
xx
acting over the incremental strip is dM
xx
=dN
xx
z=
r
xx
zdz. 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:
M
xx

_
t=2
t=2
r
xx
zdz 2a
In an entirely equivalent manner, we can relate moments M
yy
and M
xy
to
stresses r
yy
and s
xy
, respectively:
M
yy

_
t=2
t=2
r
yy
zdz 2b
M
xy

_
t=2
t=2
s
xy
zdz 2c
Figure 4 Edge view of a thin plate subjected to loads N
xx
and M
xx
.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 N
xx
, N
yy
, and N
xy
are commonly called
stress resultants, and moments M
xx
, M
yy
, and M
xy
are commonly called
moment resultants.
2 PLATE DEFORMATIONS: THE KIRCHHOFF HYPOTHESIS
Let us now consider the deformation of a thin, at plate. Fig. 5 represents a
(magnied) viewof 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 gure, and the positive z-direction is downward, which is
consistent with the original denition 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 gure. If the at 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 gure. We will base our analysis on the Kirchho
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 bo
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 Kirchho hypothesis, line bo
Figure 5 Initial and deformed positions of a flat plate.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
d has been drawn perpendicular to the midplane in both the initial and
deformed positions. In eect, we have assumed that out-of-plane shear strains
are zero (c
xz
=c
xz
=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 z
c
from
the midsurface. Point c is shown in Fig. 5, and lies along line bod. Denote
the displacement of point o in the x-direction and z-direction as distances
u
o
and w
o
, respectively. From the gure, it can be seen that the distance point
c has moved in the x-direction, u
c
, is approximately given by:
u
c
iu
o
z
c
sin a 3
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 z
c
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 4
Based on this assumption, Eq. (3) can be written as:
u
c
iu
o
z
c
a 5
Now, fromFig. 5, it can be seen that tan a=dw/dx. Applying the small-angle
approximation once more, we can say that tan a ia idw=dx. Substituting
this result in Eq. (5), we obtain:
u
c
u
o
z
c
dw
dx
6
To summarize, we have expressed the displacement in the x-direction of
arbitrary point c (which we have called distance u
c
) as a function of the
displacement in the x-direction of a point on the plate midsurface (distance
u
o
), the position of point c with respect to the midsurface (length z
c
), 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 v
c
. Using a procedure that is entirely equivalent to that just
described, it can be shown that:
v
c
v
o
z
c
dw
dy
7
Equations (6) and (7) represent the displacements of point c in the x-
direction and y-direction, respectively, and followdirectly fromthe Kirchho
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
hypothesis. We can now determine the innitesimal in-plane strains at point
c, in accordance with Eq. (49) of Chap. 2:
e
xx

Bu
Bx
e
yy

Bv
By
c
xy

Bv
Bx

Bu
By
8
Substituting Eqs. (6) and (7) into Eq. (8), we obtain the following expressions
for the strains induced at point c:
e
c
xx

Bu
o
Bx
z
c
B
2
w
Bx
2
e
c
yy

Bv
o
By
z
c
B
2
w
By
2
9
c
c
xy

Bu
o
By

Bv
o
Bx
2z
c
B
2
w
BxBy
Let:
e
o
xx

Bu
o
Bx
10a
e
o
yy

Bv
o
By
10b
c
o
xy

Bu
o
By

Bv
o
Bx
10c
j
xx

B
2
w
Bx
2
10d
j
yy

B
2
w
By
2
10e
j
xy
2
B
2
w
BxBy
10f
where e
o
xx
; e
o
yy
; and c
o
xy
are the in-plane strains that exist at the midplane of the
plate. The terms j
xx
, j
yy
, and j
xy
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 dene curvatures varies fromone author
to the next. The sign convention used throughout this text and dened by Eqs.
(10a10f ) is most commonly used in the study of composite plates. However,
in many textbooks devoted to beam theory, curvatures are dened using the
opposite sign convention. For example, in beamtheory curvature, j
xx
is often
dened as j
xx
=+B
2
w/Bx
2
, rather than j
xx
=B
2
w/Bx
2
as indicated above.
Also, in some textbooks devoted to plate theory, j
xy
is dened as j
xy
=+B
2
w/
BxBy, rather than as j
xy
=2B
2
w/BxBy as indicated in Eq. (10f ). These
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
unfortunate deviations from one author to the next have developed over
many years, and a universal agreement on algebraic signs or even the
fundamental denition of j
xy
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. (10a10f)
in Eq. (9), we obtain:
e
xx
e
o
xx
zj
xx
e
yy
e
o
yy
zj
yy
11
c
xy
c
o
xy
zj
xy
Equation (11) can be conveniently written in matrix form as:
e
xx
e
yy
c
xy
_

_
_

e
o
xx
e
o
yy
c
o
xy
_

_
_

_
z
j
xx
j
yy
j
xy
_

_
_

_
12
Equation (12) is the primary result we require for present purposes from
classical thin-plate theory. It allows us to calculate the innitesimal in-plane
strains (e
xx
,e
yy
,c
xy
) induced at any position z through the thickness of the
plate, based on the midplane strains (e
o
xx
; e
o
yy
; c
o
xy
) and midplane curvatures
(j
xx
,j
yy
,j
xy
). Note that this result is based strictly on the Kirchho hypoth-
esis. We have made no assumptions regarding the mechanism(s) that caused
the at 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 t=2 0:5 mm : e
xx
250 Am=m; e
yy
1500 Am=m; c
xy
1000 Arad
at z t=2 0:5 mm : e
xx
250 Am=m; e
yy
1100 Am=m; c
xy
800 Arad
What midplane strains and curvatures are induced in the plate?
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Solution. To solve this problem, we simply apply Eq. (12) to both surfaces of
the plate. For example, using the measured strains for e
xx
, we have:
at z t=2 0:0005 m : e
xx
250 Am=m e
o
xx
0:0005j
xx
at z t=2 0:0005 m : e
xx
250 Am=m e
o
xx
0:0005j
xx
Solving simultaneously, we nd:
e
o
xx
0 Am=m; j
xx
0:50 rad=m
Using a similar approach utilizing the measured values for e
yy
and c
xy
, we
nd:
e
o
yy
1300 Am=m; j
yy
0:40 rad=m
c
o
xy
900 Arad; j
xy
0:20 rad=m
3 PRINCIPAL CURVATURES
In Sec. 2, we invoked the Kirchho 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 Kirchho hypothesis has ultimately allowed us to calculate the in-
plane strains referenced to the xy coordinate system (e
xx
, e
yy
, and c
xy
)
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 (e
o
xx
, e
o
yy
, and c
o
xy
) and midplane curvatures (j
xx
, j
yy
, and j
xy
).
In this section, we will consider a special case. Specically, we will
consider a state of deformation in which the midplane strains are zero: e
o
xx

e
o
yy
c
o
xy
0. In this special case, Eq. (12) becomes:
e
xx
e
yy
c
xy
_

_
_

_
z
j
xx
j
yy
j
xy
_

_
_

_
13
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 (e
o
xx
e
o
yy
c
o
xy
0) and
the midplane of the plate is called the neutral surface.
Equation (13) gives the in-plane strains referenced to the xy coordinate
system. Referring to Fig. 6, suppose we wish to express these strains relative to
a newxVyVcoordinate system, obtained by rotating through angle a about the
z-axis. As noted in Sec. 2, the Kirchho hypothesis implies that c
xz
=c
yz
=0.
Therefore, the z-axis is a principal strain axis. Consequently, we can rotate in-
plane strains from the xy coordinate system to the new xVyV coordinate
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
system using Eq. (44) of Chap. 2 (developed in Sec. 13 of Chap. 2), repeated
here for convenience:
e
x Vx V
e
y Vy V
c
xVy V
2
_

_
_

cos
2
a sin
2
a 2cosa sina
sin
2
a cos
2
a 2cosa sina
cosa sina cosa sina cos
2
a sin
2
a
_

_
_

_
e
xx
e
yy
c
xy
2
_

_
_

_
2:44
Substituting Eq. (13) into Eq. (44) of Chap. 2, we can write:
e
x VxV
e
y Vy V
c
xVy V
_
_
_
_
_
_
z
j
xVxV
j
y Vy V
j
x Vy V
_
_
_
_
_
_
14
where:
j
xVxV
j
y Vy V
j
xVy V
2
_

_
_

cos
2
a sin
2
a 2cosa sina
sin
2
a cos
2
a 2cosa sina
cosa sina cosa sina cos
2
a sin
2
a
_

_
_

_
j
xx
j
yy
j
xy
2
_

_
_

_
15
Note that midplane curvatures in the xVyV coordinate system are related to
curvatures in the xy 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
Figure 6 In-plane coordinate system xVyV, obtained by rotating through angle
a about the z-axis.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 systemcan also be determined using Eqs. (47) and (48) of Chap. 2,
respectively, repeated here for convenience:
e
p
1
; e
p
2

e
xx
e
yy
2
F

e
xx
e
yy
2
_ _
2

c
xy
2
_ _
2

2:47
h
p
e

1
2
arctan
c
xy
e
xx
e
yy
_ _
2:48
Substituting Eq. (13) into Eq. (47) of Chap. 2, we nd that for pure bending,
the in-plane principal strains are given by:
e
p
1
; e
p
2
z
j
xx
j
yy

2
F

j
xx
j
yy

2
_ _
2

j
xy
2
_ _
2

_
_
_
_
16
Substituting Eq. (13) into Eq. (48) of Chap. 2, we nd that the orientation of
the principal strain coordinate system is given by:
h
p
e

1
2
arctan
j
xy
j
xx
j
yy
_ _
17
Noting that j
xx
, j
yy
, and j
xy
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 simplied expression for the principal strains is obtained by writing
Eq. (16) as:
e
p1
j
p
1
18
e
p2
j
p
2
where j
p
1
and j
p
2
are called principal curvatures and are given by:
j
p
1
; j
p
2
z
j
xx
j
yy

2
F

j
xx
j
yy

2
_ _
2

j
xy
2
_ _
2

_
_
_
_
19
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 systemin which the principal curvatures exist. Because
shear strain is zero in the principal strains coordinate system, j
p
1
p
2
=0 as well.
A physical interpretation of the preceding results can be obtained
through sketches of deformed strain elements parallel to the xy plane, as
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
was done in Sec. 2.13 (in particular, refer to Sample Problem2.9). Athin plate
reference to an xyz coordinate systemis shown in Fig. 7a. Arectangular 3D
element cut out of this plate by two pairs of planes parallel to the xz and yz
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 xy coordinate system, can be
calculated using Eq. (13).
Consider as representative examples three 2Dstrain elements parallel to
the xy plane and located at positions dened by:

z=t/2 (element abcd, shown in Fig. 7b)

z=0 (element efgh, shown in Fig. 7c)

z=+t/2 (element ijkl, 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 e
xx
and e
yy
, 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).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
the element changes from p/2 rad (i.e., 90j), in accordance with the algebraic
sign of c
xy
.
First, consider strain element ijkl, 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):
e
xx
j
z t=2
tj
xx
=2
e
yy
j
z t=2
tj
yy
=2
c
xy
j
z t=2
tj
xy
=2
Assume for illustrative purposes that all curvatures are positive (j
xx
,j
yy
,
j
xy
>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 e
xx
and e
yy
are positive, and angle
jil has decreased because c
xy
is positive.
Now consider strain element efgh, 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 efgh is not deformed, as
shown in Fig. 7c.
Finally, consider strain element abcd, located at z=t/2 (Fig. 7b).
Using Eq. (13), the strains induced at this position are:
e
xx
j
z t=2
tj
xx
=2
e
yy
j
z t=2
tj
yy
=2
c
xy
j
z t=2
tj
xy
=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. Asketch 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 e
xx
and e
yy
are negative, and angle bad
has increased because c
xy
is negative.
The deformed 2D strains elements shown in Fig. 7bd are assembled to
create a sketch of the entire 3D element in Fig. 8. Note that, in accordance
with the Kirchho hypothesis, the four line segments that dene the vertical
edges of the element (line segments aei, bfj, cgk, and dhl) remain
straight lines after deformation. However, the transverse planes are no longer
plane after deformation. For example, plane bjhc has been twisted during
deformation of the plate. Inspection of Figs. 7bd and 8 reveals that trans-
verse planes do not remain plane after deformation due to curvature j
xy
. That
is, if j
xy
p 0, shear strain c
xy
varies with through-thickness position z, in
accordance with Eq. (13). It is this through-thickness variation in c
xy
that
leads to twisting of the transverse planes. For this reason, j
xy
is known as the
twist curvature.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
We will nowrepeat this process for a rectangular 3Delement 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. 9bd. 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 (j
p
1
,j
p
2
>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. 9bd are assembled to create a sketch of the deformed 3D element in
Fig. 10. As before, the four line segments that dene the vertical edges of the
element (line segments aei, bfj, cgk, and dhl) 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 systemin which the
principal curvatures exist, the transverse planes of the strain element simply
rotate about the neutral surface.
Asummary 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 fromone coordinate systemto another using the same process as that
Figure 8 A 3D strain element assembled from the 2D deformed elements
shown in Fig. 7bd (deformations shown greatly exaggerated for clarity).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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).
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).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
used to transformthe strain or stress tensors. For a thin plate governed by the
Kirchho hypothesis, midplane curvatures transform according to Eq. (15).
In general, three midplane curvatures are induced in a thin plate: j
xx
, j
yy
, and
j
xy
. Curvature j
xy
is called the twist curvature because it represents a twisting
of a plane transverse to the midplane of the plate. The principal curvatures j
p
1
and j
p
1
are the maximumand minimumcurvatures, 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 e
o
xx
,e
o
yy
,c
o
xy
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 magnied edge view of a thin composite laminate that contains n plies is
shown in Fig. 11. The gure is similar to the edge viewof a thin plate shown in
Fig. 4, except that nowthe ply interface positions are shown. The thickness of
ply k will be denoted t
k
. The origin of the xz 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 S
n
k 1
t
k
.
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 z
0
(i.e., z
0
u t/2). Note that z
0
is always an
algebraically negative number. The coordinate position of the interface
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
between plies 1 and 2 is denoted z
1
, and z
1
=z
0
+t
1
. Similarly, the coordinate
position of the interface between plies 2 and 3 is denoted z
2
, and z
2
=z
1
+t
2
,
etc. For an n-ply laminate, the outermost surface of the laminate in the
positive z-direction is will be labeled z
n
; obviously, z
n
=+t/2. Note that in all
cases, z
n
is an algebraically positive number. Also note that the total thickness
of the laminate equals (z
n
z
0
), and the thickness of an individual ply k is
t
k
=(z
k
z
k1
). For example, the thickness of ply 2 is t
2
=(z
2
z
1
).
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 bers, or may consist of a
woven or braided fabric. In these latter two cases, there are two or more ber
directions present within each ply, although the orientation of the principal
material coordinate systemis always evident due to the symmetrical pattern of
the ber architecture. For simplicity in the following discussion, it will be
assumed that all plies are composed of unidirectional bers. In this case, the
angle between the principal material coordinate system and the x-axis is
equivalent to the angle between the bers and the x-axis. Hence, in the
discussion to follow, we will simply refer to the ber 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
Figure 11 An edge view of an n-ply laminate showing ply interface positions.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
because the phrase ber angle is more concise than the phrase principal
material coordinate system angle.
To describe the stacking sequence of a laminate, we list ber angles
within square brackets [ and ]. The ber angle (in degrees) of ply 1 is
listed rst, followed by the ber angle of ply 2, ply 3, etc. Each ber angle will
be separated by a slash /. For example, a four-ply laminate consisting of
plies with ber 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 ber 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 ber 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 ber angles, which allows us to abbreviate the
listing of ply ber 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 ber 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 ber angles are symmetrical about the
laminate midplane. This laminate is denoted [0/F45/90]
s
. The subscript s
indicates that the four ber 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.
Anine-ply laminate containing ber 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 ber 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: rst to indicate that ber 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.
Anal example is the 10-ply laminate shown in Fig. 12e. In this case, the
ber 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
ber pattern listed within the parentheses occurs twice. Note that this
laminate is similar but not identical to that shown in Fig. 12d.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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
.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Sample Problem 2
A(0/F30/90)
2
/45/20
s
laminate is fabricated using a graphiteepoxy 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 ber angles that appear in the laminate is
as follows:
0j; 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
midplane
The laminate contains a total of 19 plies, the ber 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:
Note that the total laminate thickness equals the dierence between z
19
and z
0
, as expected: t=z
19
z
0
=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 ber angles of each ply)

The material properties and thickness of each ply.


z
0
=t/2=1.1875 mm z
1
=z
0
+t
1
=1.0625 mm z
2
=z
1
+t
2
=0.9375 mm
z
3
=z
2
+t
3
=0.8125 mm z
4
=z
3
+t
4
=0.6875 mm z
5
=z
4
+t
5
=0.5625 mm
z
6
=z
5
+t
6
=0.4375 mm z
7
=z
6
+t
7
=0.3125 mm z
8
=z
7
+t
8
=0.1875 mm
z
9
=z
8
+t
9
=0.0625 mm z
10
=z
9
+t
10
=0.0625 mm z
11
=z
10
+t
11
=0.1875 mm
z
12
=z
11
+t
12
=0.3125 mm z
13
=z
12
+t
13
=0.4375 mm z
14
=z
13
+t
14
=0.5625 mm
z
15
=z
14
+t
15
=0.6875 mm z
16
=z
15
+t
16
=0.8125 mm z
17
=z
16
+t
17
=0.9375 mm
z
18
=z
17
+t
18
=1.0625 mm z
19
=z
18
+t
19
=1.1875 mm
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Note that the plies are not necessarily all of the same material type. For
example, some plies within a laminate may be of graphiteepoxy whereas
others may be glassepoxy.
Once the stacking sequence and thickness of each ply have been
specied, the total laminate thickness and interface positions throughout
the laminate may be determined, as previously illustrated in Fig. 11. Also, the
transformed reduced stiness matrix 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
(e
o
xx
; e
o
yy
; c
o
xy
) and (j
xx
,j
yy
,j
xy
), respectively, in accordance with the Kirchho
hypothesis (Eq. (12)). For example, the strains induced in a distance z
k
from
the laminate midplane are:
e
xx
e
yy
c
xy
_

_
_

zz
k

e
o
xx
e
o
yy
c
o
xy
_

_
_

_
z
k
j
xx
j
yy
j
xy
_

_
_

_
Note that these strains are referenced to the xy coordinate system. If desired,
these strains can be rotated fromthe xy coordinate systemto the local 12
coordinate systemfor each ply (dened by the ply ber angle) using Eq. (44) of
Chap. 2:
e
11
e
22
c
12
=2
_
_
_
_
_
_j
z z
k
T
k
e
xx
e
yy
c
xy
=2
_
_
_
_
_
_j
zz
k

cos
2
h sin
2
h 2cosh sinh
sin
2
h cos
2
h 2cosh sinh
cosh sinh cosh sinh cos
2
h sin
2
h
_

_
_

_
k
e
xx
e
yy
c
xy
=2
_

_
_

_j
z z
k
Ply Stresses: Once ply strains are determined, ply stresses are calculated
using Hookes law, as discussed in Sec. 5.2. For example, the stresses induced
at a distance z
k
from the laminate midplane are calculated using Eq. (30) of
Chap. 5:
r
xx
r
yy
s
xy
_

_
_

zz
k

Q
11
Q
12
Q
16
Q
12
Q
22
Q
26
Q
16
Q
26
Q
66
_

_
_

zz
k
e
xx
DTa
xx
DMb
xx
e
yy
DTa
yy
DMb
yy
c
xy
DTa
xy
DMb
xy
_

_
_

z z
k
Note that the material properties used in this calculation (specically, Q, a
ij
,
and b
ij
) are properties of the ply that exists at position z
k
, and in particular are
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
functions of the ply ber angle h. Because ber angle generally varies fromone
ply to the next, these material properties also vary from one ply to the next.
If desired, stresses can be rotated from the xy coordinate system to the
local 12 coordinate system for each ply using Eq. (20) of Chap. 2:
r
11
r
22
s
12
_
_
_
_
_
_j
zz
k
T
zz
k
r
xx
r
yy
s
xy
_
_
_
_
_
_j
zz
k
A numerical example that illustrates these calculations is presented in
the following Sample Problem.
Sample Problem 3
Assume that the panel considered in Sample Problem1 is actually an eight-ply
[0/30/90/30]
s
graphiteepoxy laminate. Assume that the laminate was
initially at 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. FromSample Problem1, the midplane strains and curvatures are:
e
o
xx
0 Am=m j
xx
0:50 rad=m
e
o
yy
1300 Am=m j
yy
0:40 rad=m
c
o
xy
900 Arad j
xy
0:20 rad=m
To determine ply interface positions, rst note that the total laminate
thickness is:
t 8 plies0:125 mm 1:0 mm 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:
z
0
=t/2=(0.001 m)=0.000500 m
z
1
=z
0
+t
1
=0.000500 m+0.000125 m=0.000375 m
z
2
=z
1
+t
2
=0.000375 m+0.000125 m=0.000250 m
z
3
=z
2
+t
3
=0.000250 m+0.000125 m=0.000125 m
z
4
=z
3
+t
4
=.000125 m+0.000125 m=0.000000 m
z
5
=z
4
+t
5
=0.000000 m+0.000125 m=0.000125 m
z
6
=z
5
+t
6
=0.000125 m+0.000125 m=0.000250 m
z
7
=z
6
+t
7
=0.000250 m+0.000125 m=0.000375 m
z
8
=z
7
+t
7
=0.000375 m+0.000125 m=0.000500 m
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 z
o
=0.000500 m)
are:
e
xx
e
yy
c
xy
_

_
_

_j
zz
0

e
o
xx
e
o
yy
c
o
xy
_

_
_

_
z
0
j
xx
j
yy
j
xy
_

_
_

0
1300 10
6
m=m
900 10
6
m=m
_

_
_

_
0:000500 m
0:50 rad=m
0:40 rad=m
0:20 rad=m
_

_
_

_
e
xx
e
yy
c
xy
_
_
_
_
_
_
j
zz
0

250 Am=m
1500 Am=m
1000 Arad
_

_
_

_
Similarly, strains present at the interface between plies 1 and 2 (i.e., strains
present at z
1
=0.000375 m) are:
e
xx
e
yy
c
xy
_

_
_

_j
zz
1

e
o
xx
e
o
yy
c
o
xy
_

_
_

_
z
1
j
xx
j
yy
j
xy
_

_
_

0
1300 10
6
m=m
900 10
6
m=m
_

_
_

_
0:000375 m
0:50 rad=m
0:40 rad=m
0:20 rad=m
_

_
_

_
e
xx
e
yy
c
xy
_
_
_
_
_
_j
zz
1

188 Am=m
1450 Am=m
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 (e
xx
, e
yy
, and c
xy
)
are predicted to be linearly distributed through the plate thickness. This linear
distribution is a direct consequence of the Kirchho 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 specied in Sample Problem1. 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 xy coordinate system. As will be seen, knowledge of ply strains
referenced to the local 12 coordinate system(dened by the ber angle within
each ply) is often required. Transformation of the strain tensor from one
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
coordinate system to another was reviewed in Chap. 2 and, in particular,
strains can be rotated from the xy coordinate system to the 12 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 h
1
=0j, the xy and 12 coordinate systems are coincident,
and therefore the description of the strain tensor is identical in both coor-
dinate systems. This can be conrmed through application of Eq. (44) of
Chap. 2:
Top interface:
e
11
e
22
c
12
=2
_

_
_

_
j
ply 1
z z
0

cos
2
h
1
sin
2
h
1
2 cos h
1
sin h
1
sin
2
h
1
cos
2
h
1
2 cos h
1
sin h
1
cos h
1
sin h
1
cos h
1
sin h
1
cos
2
h
1
sin
2
h
1
_

_
_

_
e
xx
e
yy
c
xy
=2
_

_
_

_
j
ply 1
z z
0
e
11
e
22
c
12
=2
_

_
_

_
j
ply 1
z z
0

1 0 0
0 1 0
0 0 1
_

_
_

_
250 Am=m
1500 Am=m
1000 Arad=2
_

_
_

_
j
ply1
z z
0
e
11
e
22
c
12
=2
_

_
_

_
j
ply 1
z z
0

250 Am=m
1500 Am=m
1000 Arad
_

_
_

_
j
ply 1
z z
0
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) e
xx
(Am/m) e
yy
(Am/m) c
xy
(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 xy coordinate system.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Figure 13 Through-thickness strain plots dictated by the midplane strains and
curvatures discussed in Sample Problem 1. Strains are referenced to the xy
coordinate system.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Bottom interface:
e
11
e
22
c
12
=2
_

_
_

_
j
ply 1
z z
1

cos
2
h
1
sin
2
h
1
2cos h
1
sin h
1
sin
2
h
1
cos
2
h
1
2cos h
1
sin h
1
cos h
1
sin h
1
cos h
1
sin h
1
cos
2
h
1
sin
2
h
1
_

_
_

_
e
xx
e
yy
c
xy
=2
_

_
_

_
j
ply 1
z z
1
e
11
e
22
c
12
=2
_

_
_

_
j
ply 1
z z
1

1 0 0
0 1 0
0 0 1
_

_
_

_
188 Am=m
1450 Am=m
975 Arad=2
_

_
_

_
j
ply 1
z z
1
e
11
e
22
c
12
_

_
_

_
j
ply 1
z z
1

188 Am=m
1450 Am=m
975 Arad
_

_
_

_
j
ply 1
z z
1
Ply 2. In this case, h
2
=30j and consequently the description of strain in the
xy and 12 coordinate systems diers substantially. Applying Eq. (44) of
Chap. 2, we have:
Top interface:
e
11
e
22
c
12
=2
_

_
_

_
j
ply 2
z z
1

cos
2
h
2
sin
2
h
2
2 cos h
2
sin h
2
sin
2
h
2
cos
2
h
2
2 cos h
2
sin h
2
cos h
2
sin h
2
cos h
2
sin h
2
cos
2
h
2
sin
2
h
2
_

_
_

_
e
xx
e
yy
c
xy
=2
_

_
_

_
j
ply 2
z z
1
e
11
e
22
c
12
=2
_

_
_

_
j
ply 2
z z
1

cos
2
30j sin
2
30j 2 cos30j sin30j
sin
2
30j cos
2
30j 2 cos30j sin30j
cos30j sin30j cos30j sin30j cos
2
30j sin
2
30j
_

_
_

188 Am=m
1450 Am=m
975 Arad=2
_

_
_

_
j
ply 2
z z
1
e
11
e
22
c
12
=2
_

_
_

_
j
ply 2
z z
1

0:750 0:250 0:866


0:250 0:750 0:866
0:433 0:433 0:500
_

_
_

_
188 Am=m
1450 Am=m
975 Arad=2
_

_
_

_
j
ply 2
z z
1
e
11
e
22
c
12
_

_
_

_
j
ply 2
z z
1

200 Am=m
1463 Am=m
931 Arad
_

_
_

_
j
ply 2
z z
1
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Bottom interface:
e
11
e
22
c
12
=2
_

_
_

_
j
ply 2
z z
2

cos
2
h
2
sin
2
h
2
2 cos h
2
sin h
2
sin
2
h
2
cos
2
h
2
2 cos h
2
sin h
2
cos h
2
sin h
2
cos h
2
sin h
2
cos
2
h
2
sin
2
h
2
_

_
_

_
e
xx
e
yy
c
xy
=2
_

_
_

_
j
ply 2
z z
2
e
11
e
22
c
12
=2
_

_
_

ply 2
z z
2

cos
2
30j sin
2
30j 2 cos30j sin30j
sin
2
30j cos
2
30j 2 cos30j sin30j
cos30j sin30j cos30j sin30j cos
2
30j sin
2
30j
_

_
_

125 Am=m
1400 Am=m
950 Arad=2
_

_
_

_
j
ply 2
z z
2
e
11
e
22
c
12
=2
_

_
_

_
j
ply 2
z z
2

0:750 0:250 0:866


0:250 0:750 0:866
0:433 0:433 0:500
_

_
_

_
125 Am=m
1400 Am=m
950 Arad=2
_

_
_

_
j
ply 2
z z
2
e
11
e
22
c
12
_

_
_

_
j
ply 2
z z
2

155 Am=m
1430 Am=m
846 Arad
_

_
_

_
j
ply 2
z z
2
Ply strains referenced to the local 12 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 12 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 stiness matrix for each ply. Using graphiteepoxy material
properties fromTable 2 of Chap. 3 and Eqs. (11) and (31) of Chap. 5, we nd:
For 0j plies:
Q
_
0j plies

170:9 10
9
3:016 10
9
0
3:016 10
9
10:05 10
9
0
0 0 13:00 10
9
_

_
_

_
Pa
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
For 30j plies:
Q
_
30j plies

107:6 10
9
26:06 10
9
48:3 10
9
26:06 10
9
27:22 10
9
21:52 10
9
48:3 10
9
21:52 10
9
36:05 10
9
_

_
_

_
Pa
For 90j plies:
Q
_
90j plies

10:05 10
9
3:016 10
9
0
3:016 10
9
170:9 10
9
0
0 0 13:00 10
9
_

_
_

_
Pa
For 30j plies:
Q
_
30j plies

107:6 10
9
26:06 10
9
48:3 10
9
26:06 10
9
27:22 10
9
21:52 10
9
48:3 10
9
21:52 10
9
36:05 10
9
_

_
_

_
Pa
Table 2 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
Ply number z-coordinate (mm) e
11
(Am/m) e
22
(Am/m) c
12
(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 12 coordinate system local to individual plies.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Figure 14 Through-thickness strain plots dictated by the midplane strains and
curvatures discussed in Sample Problem 1. Strains are referenced to the 12
coordinate system.
Stresses present at the outer surface of ply 1 (i.e., strains present at
z
0
=0.000500 m) can now be calculated:
r
xx
r
yy
s
xy
_

_
_

_
j
ply 1
z z
0

Q
11
Q
12
Q
16
Q
12
Q
22
Q
26
Q
16
Q
26
Q
66
_

_
_

_
j
ply 1
z z
0
e
xx
e
yy
c
xy
_

_
_

_
j
z z
0
r
xx
r
yy
s
xy
_

_
_

_
j
ply 1
z z
0

170:9 10
9
3:016 10
9
0
3:016 10
9
10:05 10
9
0
0 0 13:00 10
9
_

_
_

_
250 Am=m
1500 Am=m
1000 Arad
_

_
_

_
r
xx
r
yy
s
xy
_

_
_

_
j
ply 1
z z
0

38:2 MPa
14:3 MPa
13 MPa
_

_
_

_
To calculate stresses at the interface between plies 1 and 2 (i.e., at z
1
=
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 =
z
1
()
, whereas ply 2 begins at z=z
1
(+)
. Hence, the stresses within ply 1 at z =
z
1
()
are:
r
xx
r
yy
s
xy
_

_
_

_
j
ply 1
z z
1

Q
11
Q
12
Q
16
Q
12
Q
22
Q
26
Q
16
Q
26
Q
66
_

_
_

_
j
ply 1
z z
1
e
xx
e
yy
c
xy
_

_
_

_
j
z z
1
r
xx
r
yy
s
xy
_

_
_

_
j
ply 1
z z
1

170:9 10
9
3:016 10
9
0
3:016 10
9
10:05 10
9
0
0 0 13:00 10
9
_

_
_

_
188 Am=m
1450 Am=m
975 Arad
_

_
_

_
r
xx
r
yy
s
xy
_

_
_

_
j
ply 1
z z
1

27:8 MPa
14:0 MPa
12:7 MPa
_

_
_

_
The stresses within ply 2 (a 30j ply) at z=z
1
(+)
are:
r
xx
r
yy
s
xy
_

_
_

_
j
ply 2
z z
1

Q
11
Q
12
Q
16
Q
12
Q
22
Q
26
Q
16
Q
26
Q
66
_

_
_

_
j
ply 2
z z
1
e
xx
e
yy
c
xy
_

_
_

_
j
z z
1
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
r
xx
r
yy
s
xy
_

_
_

_
j
ply 2
z z
1

107:6 10
9
26:06 10
9
48:3 10
9
26:06 10
9
27:22 10
9
21:52 10
9
48:3 10
9
21:52 10
9
36:05 10
9
_

_
_

_
188 Am=m
1450 Am=m
975 Arad
_

_
_

_
r
xx
r
yy
s
xy
_

_
_

_
j
ply 2
z z
1

29:5 MPa
13:6 MPa
13:0 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 xy 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 ber angle (in general)
changes from one ply to the next. Indeed, in this example problem, the same
ber angle occurs in only two adjacent plies (namely, plies 4 and 5, both of
which have a ber 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
number
z-coordinate
(mm)
r
xx
(MPa)
r
yy
(MPa)
s
xy
(MPa)
Q
(MPa)
A
(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 xy coordinate system.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 xy coordinate
system.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 subjectedtoelastic stress/strainlevels, and(b) whenthe Q
_
matrix
does not vary from one ply to the next (i.e., for unidirectional laminates in
which the ber angle does not vary from one ply to the next).
Knowledge of ply stresses referenced to the local 12 coordinate system
(dened by the ber 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 xy coordinate
system to the 12 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=z
1
=0.375 mm) proceeds as follows:
r
11
r
22
s
12
_
_
_
_
_
_

ply 2
z z
1

cos
2
h
2
sin
2
h
2
2 cos h
2
sin h
2
sin
2
h
2
cos
2
h
2
2 cos h
2
sin h
2
cos h
2
sin h
2
cos h
2
sin h
2
cos
2
h
2
sin
2
h
2
_

_
_

_
r
xx
r
yy
s
xy
_
_
_
_
_
_

ply 2
z z
1
r
11
r
22
s
12
_
_
_
_
_
_

ply 2
z z
1

cos
2
30j sin
2
30j 2 cos30j sin30j
sin
2
30j cos
2
30j 2 cos30j sin30j
cos30j sin30j cos30j sin30j cos
2
30j sin
2
30j
_

_
_

29:3 MPa
13:6 MPa
13:0 MPa
_
_
_
_
_
_

ply 2
z z
1
r
11
r
22
s
12
_
_
_
_
_
_

ply 2
z z
1

0:750 0:250 0:866


0:250 0:750 0:866
0:433 0:433 0:500
_
_
_
_
29:3 MPa
13:6 MPa
13:0 MPa
_
_
_
_
_
_

ply 2
z z
1
r
11
r
22
s
12
_
_
_
_
_
_

ply 2
z z
1

29:8 MPa
14:1 MPa
12:1 MPa
_
_
_
_
_
_

ply 2
z z
1
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Ply interface stresses referenced to local 12 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 conrm that the ply stresses referenced
to the xy coordinate system and listed in Table 3 are equivalent to the ply
stresses referenced to the 12 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 Q r
xx
r
yy
Second stress invariant A r
xx
r
yy
s
2
xy
Third stress invariant C 0
repeated 2:22
For plane stress conditions, the third stress invariant always equals zero, and
so Ccannot be used to evaluate whether two plane stress states are equivalent.
The rst and second stress invariants, Q and A, respectively, have been
calculated using the ply stress components referenced to both the xy and 12
Table 4 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
number
z-coordinate
(mm)
r
11
(MPa)
r
22
(MPa)
s
12
(MPa)
Q
(MPa)
A
(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 12 coordinate system.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Figure 16 Through-thickness stress plots predicted for a [0/30/90/30]
s
graphiteepoxy laminate subjected to the midplane strains and curvatures
discussed in Sample Problem 1. Stresses are referenced to the 12 coordinate
system.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
coordinate systems. Values calculated for Qand Aare 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
dierent 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 Kirchho hypothesis was invoked, which
allows us to calculate the in-plane strains induced at any location through the
thickness of a thinplate. Inthis section, we will combine the material presented
in Secs. 1 and 2, as well as certain material presented in Chap. 5. This will lead
tothe ability torelate stress andmoment resultants tothe 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 rst 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 r
xx
inducedina thincomposite laminate are relatedtostress
resultant N
xx
in accordance with Eq. (1a), repeated here for convenience:
N
xx

_
t=2
t=2
r
xx
dz repeated 1a
The composite laminate consists of n plies, and the ber 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:
r
xx
r
yy
s
xy
_

_
_

_
k

Q
11
Q
12
Q
16
Q
21
Q
22
Q
26
Q
61
Q
62
Q
66
_

_
_

_
k
e
xx
e
yy
c
xy
_

_
_

_
k
20
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
The subscript k in Eq. (20) indicates that the stresses, transformed reduced
stiness 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 r
xx
induced in ply k is:
r
xx

k
Q
11
e
xx
Q
12
e
yy
Q
16
c
xy
_ _
k
Substituting this relationship into Eq. (1a), we have:
N
xx

_
t=2
t=2
Q
11
e
xx
Q
12
e
yy
Q
16
c
xy
_ _
k
dz 21
The strains induced in ply k can be related to the midplane strains and
curvatures via the Kirchho hypothesis, in accordance with Eq. (11) or Eq.
(12). Substituting Eqs. (11) and (12) into Eq. (21), we obtain:
N
xx

_
t=2
t=2
Q
11
e
o
xx
Q
12
e
o
yy
Q
16
c
o
xy
zQ
11
j
xx
zQ
12
j
yy
zQ
16
j
xy
_ _
dz
22
We cannot integrate Eq. (22) directly because the integrand is a discontinuous
function of z. That is, the transformed reduced stiness terms Q
11
, Q
12
, and
Q
16
are all directly related to the ply material properties and ber angle h (see
Eq. (31) of Chap. 5). Because the ply material and/or ber angle may change
fromone ply to the next, the transformed reduced stiness 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 fromunder the integral sign.
Equation (15) can therefore be broken into six individual integrals:
N
xx
e
o
xx
_
t=2
t=2
Q
11
_ _
k
dz e
o
yy
_
t=2
t=2
Q
12
_ _
k
dz
c
o
xy
_
t=2
t=2
Q
16
_ _
k
dz j
xx
_
t=2
t=2
Q
11
_ _
k
dz
j
yy
_
t=2
t=2
z Q
12
_ _
k
dz j
xy
z Q
16
_ _
k
dz
23
Because the transformed stiness terms are constant over each ply thickness,
each of the six integrals in Eq. (23) can be evaluated in a piecewise fashion:
N
xx
e
o
xx
Q
11
_ _
1
_
z
1
z
0
dz Q
11
_ _
2
_
z
2
z
1
dz Q
11
_ _
3
_
z
3
z
2
dz
: : :
_
Q
11
_ _
n1
_
z
n1
z
n2
dz Q
11
_ _
n
_
z
n
z
n1
dz
_
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
e
o
yy
Q
12
_ _
1
_
z
1
z
0
dz Q
12
_ _
2
_
z
2
z
1
dz Q
12
_ _
3
_
z
3
z
2
dz
: : :
_
Q
12
_ _
n1
_
z
n1
z
n2
dz Q
12
_ _
n
_
z
n
z
n1
dz
_
c
o
xy
Q
16
_ _
1
_
z
1
z
0
dz Q
16
_ _
2
_
z
2
z
1
dz Q
16
_ _
3
_
z
3
z
2
dz
: : :
_
Q
16
_ _
n1
_
z
n1
z
n2
dz Q
16
_ _
n
_
z
n
z
n1
dz
_
j
xx
Q
11
_ _
1
_
z
1
z
0
zdz Q
11
_ _
2
_
z
2
z
1
zdz Q
11
_ _
3
_
z
3
z
2
zdz
: : :
_
Q
11
_ _
n1
_
z
n1
z
n2
zdz Q
11
_ _
n
_
z
n
z
n1
zdz
_
j
yy
Q
12
_ _
1
_
z
1
z
0
zdz Q
12
_ _
2
_
z
2
z
1
zdz Q
12
_ _
3
_
z
3
z
2
zdz
: : :
_
Q
12
_ _
n1
_
z
n1
z
n2
zdz Q
12
_ _
n
_
z
n
z
n1
zdz
_
j
xy
Q
16
_ _
1
_
z
1
z
0
zdz Q
16
_ _
2
_
z
2
z
1
zdz Q
16
_ _
3
_
z
3
z
2
zdz
: : :
_
Q
16
_ _
n1
_
z
n1
z
n2
zdz Q
16
_ _
n
_
z
n
z
n1
zdz
_
24
Although Eq. (24) may appear daunting at rst, 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
k
z
k1
dz z
k
z
k1

or
_
z
k
z
k1
zdz
1
2
z
2
k
z
2
k1

Hence, evaluating all integrals that appear in Eq. (24), we obtain:


N
xx
e
o
xx
Q
11
_ _
1
z
1
z
0
Q
11
_ _
2
z
2
z
1
Q
11
_ _
3
z
3
z
2

: : :
_
Q
11
_ _
n
z
n
z
n1

_
e
o
yy
Q
12
_ _
1
z
1
z
0
Q
12
_ _
2
z
2
z
1
Q
12
_ _
3
z
3
z
2

: : :
_
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Q
12
_ _
n
z
n
z
n1

_
c
o
xy
Q
16
_ _
1
z
1
z
0
Q
16
_ _
2
z
2
z
1
Q
16
_ _
3
z
3
z
2

: : :
_
Q
16
_ _
n
z
n
z
n1

_

1
2
j
xx
Q
11
_ _
1
z
2
1
z
2
0
_
Q
11
_ _
2
z
2
2
z
2
1
_
Q
11
_ _
3
z
2
3
z
2
2
_
_

: : :
Q
11
_ _
n
z
2
n
z
2
n1
_
_

1
2
j
yy
Q
12
_ _
1
z
2
1
z
2
0
_
Q
12
_ _
2
z
2
2
z
2
1
_
Q
12
_ _
3
z
2
3
z
2
2
_
_

: : :
Q
12
_ _
n
z
2
n
z
2
n1
_
_

1
2
j
xy
Q
16
_ _
1
z
2
1
z
2
0
_
Q
16
_ _
2
z
2
2
z
2
1
_
Q
16
_ _
3
z
2
3
z
2
2
_
_
. . . Q
16
_ _
n
z
2
n
z
2
n1
_
_
25
Equation (25) can be simplied substantially by dening the following terms:
A
11
Q
11
_ _
1
z
1
z
0
Q
11
_ _
2
z
2
z
1
Q
11
_ _
3
z
3
z
2

: : :
_
Q
11
_ _
n
z
n
z
n1

_
A
12
Q
12
_ _
1
z
1
z
0
Q
12
_ _
2
z
2
z
1
Q
12
_ _
3
z
3
z
2

: : :
_
Q
12
_ _
n
z
n
z
n1

_
A
16
Q
16
_ _
1
z
1
z
0
Q
16
_ _
2
z
2
z
1
Q
16
_ _
3
z
3
z
2

: : :
_
Q
16
_ _
n
z
n
z
n1

_
B
11

1
2
Q
11
_ _
1
z
2
1
z
2
0
_
Q
11
_ _
2
z
2
2
z
2
1
_
Q
11
_ _
3
z
2
3
z
2
2
_

: : :
_
Q
11
_ _
n
z
2
n
z
2
n1
_
_
B
12

1
2
Q
12
_ _
1
z
2
1
z
2
0
_
Q
12
_ _
2
z
2
2
z
2
1
_
Q
12
_ _
3
z
2
3
z
2
2
_

: : :
_
Q
12
_ _
n
z
2
n
z
2
n1
_
_
B
16

1
2
Q
16
_ _
1
z
2
1
z
2
0
_
Q
16
_ _
2
z
2
2
z
2
1
_
Q
16
_ _
3
z
2
3
z
2
2
_

: : :
_
Q
16
_ _
n
z
2
n
z
2
n1
_
_
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
With these denitions, Eq. (25), becomes:
N
xx
A
11
e
o
xx
A
12
e
o
yy
A
16
c
o
xy
B
11
j
xx
B
12
j
yy
B
16
j
xy
26a
Following an entirely analogous procedure for stress resultants N
yy
and N
xy
,
it can be shown that:
N
yy
A
21
e
o
xx
A
22
e
o
yy
A
26
c
o
xy
B
21
j
xx
B
22
j
yy
B
26
j
xy
26b
N
xy
A
61
e
o
xx
A
62
e
o
yy
A
66
c
o
xy
B
61
j
xx
B
62
j
yy
B
66
j
xy
26c
where:
A
ij

n
k1
Q
ij
_ _
k
z
k
z
k1
27a
B
ij

1
2

n
k1
Q
ij
_ _
k
z
2
k
z
2
k1
_ _
27b
and i, j =1, 2, or 6. Because subscripts i and j may take on one of three values,
both A
ij
and B
ij
can be written as 3 3 matrices. Also, recall that the
transformed reduced stiness matrix is symmetrical (see Eq. (31) of Chap. 5).
Hence, both A
ij
and B
ij
are also symmetrical:
A
ij

A
11
A
12
A
16
A
21
A
22
A
26
A
61
A
62
A
66
_

_
_

_
A
11
A
12
A
16
A
12
A
22
A
26
A
16
A
26
A
66
_

_
_

_
B
ij

B
11
B
12
B
16
B
21
B
22
B
26
B
61
B
62
B
66
_

_
_

_
B
11
B
12
B
16
B
12
B
22
B
26
B
16
B
26
B
66
_

_
_

_
Equation (26a) (26b) (26c) can be written in matrix form as follows:
N
xx
N
yy
N
xy
_

_
_

A
11
A
12
A
16
A
12
A
22
A
26
A
16
A
26
A
66
B
11
B
12
B
16
B
12
B
22
B
26
B
16
B
26
B
66
_

_
_

_
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

_
28
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 A
ij
and B
ij
matrices. The values of each term within the
A
ij
and B
ij
matrices depend on the material properties and ber angle of each
ply (i.e., they depend on terms within the Q
ij
matrix) as well as the stacking
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
sequence (i.e., the distance z
k
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 r
xx
induced in a thin composite laminate are related to moment
resultant M
xx
in accordance with Eq. (2a), repeated here for convenience:
M
xx

_
t=2
t=2
r
xx
zdz repeated 2a
Substituting the expression for r
xx
from Eq. (20), we obtain:
M
xx

_
t=2
t=2
Q
11
e
xx
Q
12
e
yy
Q
16
c
xy
_ _
k
zdz 29
Each strain that appears in Eq. (29) can be related to the midplane strains and
curvatures via the Kirchho hypothesis. Hence, substituting either Eq. (11) or
Eq. (12), we have:
M
xx

_
t=2
t=2
zQ
11
e
o
xx
zQ
12
e
o
yy
zQ
16
c
o
xy
z
2
Q
11
j
xx
z
2
Q
12
j
yy
z
2
Q
16
j
xy
_ _
dz
30
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:
M
xx
e
o
xx
Q
11
_ _
1
_
z
1
z
0
zdz Q
11
_ _
2
_
z
2
z
1
zdz Q
11
_ _
3
_
z
3
z
2
zdz
: : : :
_
Q
11
_ _
n1
_
z
n1
z
n2
zdz Q
11
_ _
n
_
z
n
z
n1
zdz
_
e
o
yy
Q
12
_ _
1
_
z
1
z
0
zdz Q
12
_ _
2
_
z
2
z
1
zdz Q
12
_ _
3
_
z
3
z
2
zdz
: : : :
_
Q
12
_ _
n1
_
z
n1
z
n2
zdz Q
12
_ _
n
_
z
n
z
n1
zdz
_
c
o
xy
Q
16
_ _
1
_
z
1
z
0
zdz Q
16
_ _
2
_
z
2
z
1
zdz Q
16
_ _
3
_
z
3
z
2
zdz
: : : :
_
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Q
16
_ _
n1
_
z
n1
z
n2
zdz Q
16
_ _
n
_
z
n
z
n1
zdz
_
j
xx
Q
11
_ _
1
_
z
1
z
0
z
2
dz Q
11
_ _
2
_
z
2
z
1
z
2
dz Q
11
_ _
3
_
z
3
z
2
z
2
dz
: : : :
_
Q
11
_ _
n1
_
z
n1
z
n2
z
2
dz Q
11
_ _
n
_
z
n
z
n1
z
2
dz
_
j
yy
Q
12
_ _
1
_
z
1
z
0
z
2
dz Q
12
_ _
2
_
z
2
z
1
z
2
dz Q
12
_ _
3
_
z
3
z
2
z
2
dz
: : : :
_
Q
12
_ _
n1
_
z
n1
z
n2
z
2
dz Q
12
_ _
n
_
z
n
z
n1
z
2
dz
_
j
xy
Q
16
_ _
1
_
z
1
z
0
z
2
dz Q
16
_ _
2
_
z
2
z
1
z
2
dz Q
16
_ _
3
_
z
3
z
2
z
2
dz
: : : :
_
Q
16
_ _
n1
_
z
n1
z
n2
z
2
dz Q
16
_ _
n
_
z
n
z
n1
z
2
dz
_
31
The piecewise integrals that appear in Eq. (31) are of one of the following two
forms, both of which are easily evaluated:
_
z
k
z
k1
zdz
1
2
z
2
k
z
2
k1
_ _
or
_
z
k
z
k1
z
2
dz
1
3
z
3
k
z
3
k1
_ _
Hence, evaluating all integrals, we obtain:
M
xx

1
2
e
o
xx
Q
11
_ _
1
z
2
1
z
2
0
_
Q
11
_ _
2
z
2
2
z
2
1
_
Q
11
_ _
3
z
2
3
z
2
2
_
_

: : : :
Q
11
_ _
n
z
2
n
z
2
n1
_
_

1
2
e
o
yy
Q
12
_ _
1
z
2
1
z
2
0
_
Q
12
_ _
2
z
2
2
z
2
1
_
Q
12
_ _
3
z
2
3
z
2
2
_
_

: : : :
Q
12
_ _
n
z
2
n
z
2
n1
_
_

1
2
c
o
xy
Q
16
_ _
1
z
2
1
z
2
0
_
Q
16
_ _
2
z
2
2
z
2
1
_
Q
16
_ _
3
z
2
3
z
2
2
_
_

: : : :
Q
16
_ _
n
z
2
n
z
2
n1
_
_
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

1
3
j
xx
Q
11
_ _
1
z
3
1
z
3
0
_
Q
11
_ _
2
z
3
2
z
3
1
_
Q
11
_ _
3
z
3
3
z
3
2
_
_

: : : :
Q
11
_ _
n
z
3
n
z
3
n1
_
_

1
3
j
yy
Q
12
_ _
1
z
3
1
z
3
0
_
Q
12
_ _
2
z
3
2
z
3
1
_
Q
12
_ _
3
z
3
3
z
3
2
_
_

: : : :
Q
12
_ _
n
z
3
n
z
3
n1
_
_

1
3
j
xy
Q
16
_ _
1
z
3
1
z
3
0
_
Q
16
_ _
2
z
3
2
z
3
1
_
Q
16
_ _
3
z
3
3
z
3
2
_
_

: : : :
Q
16
_ _
n
z
3
n
z
3
n1
_
_
32
The rst three quantities on the right-hand side of the equality sign involve the
previously dened terms B
11
, B
12
, and B
16
. We now dene three new terms,
associated with the last three quantities:
D
11

1
3
Q
11
_ _
1
z
3
1
z
3
0
_
Q
11
_ _
2
z
3
2
z
3
1
_
Q
11
_ _
3
z
3
3
z
3
2
_

: : : :
_
Q
11
_ _
n
z
3
n
z
3
n1
_
_
D
12

1
3
Q
12
_ _
1
z
3
1
z
3
0
_
Q
12
_ _
2
z
3
2
z
3
1
_
Q
12
_ _
3
z
3
3
z
3
2
_

: : : :
_
Q
12
_ _
n
z
3
n
z
3
n1
_
_
D
16

1
3
Q
16
_ _
1
z
3
1
z
3
0
_
Q
16
_ _
2
z
3
2
z
3
1
_
Q
16
_ _
3
z
3
3
z
3
2
_

: : : :
_
Q
16
_ _
n
z
3
n
z
3
n1
_
_
Hence, Eq. (32) can be written in the following simplied form:
M
xx
B
11
e
o
xx
B
12
e
o
yy
B
16
c
o
xy
D
11
j
xx
D
12
j
yy
D
16
j
xy
33a
Following an entirely equivalent procedure for M
yy
and M
xy
, it can be shown
that:
M
yy
B
21
e
o
xx
B
12
e
o
yy
B
26
c
o
xy
D
21
j
xx
D
22
j
yy
D
26
j
xy
33b
M
xy
B
61
e
o
xx
B
62
e
o
yy
B
66
c
o
xy
D
61
j
xx
D
62
j
yy
D
66
j
xy
33c
The B
ij
terms that appear in Eqs. (33a) (33b) (33c) have been previously
encountered and are given by Eq. (27b). The new terms D
ij
are given by:
D
ij

1
3

n
k 1
Q
ij
_ _
k
z
3
k
z
3
k1
_ _
34
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
The D
ij
terms can be written as a symmetrical 3 3 matrix:
D
ij

D
11
D
12
D
16
D
21
D
22
D
26
D
61
D
62
D
66
_

_
_

_
D
11
D
12
D
16
D
12
D
22
D
26
D
16
D
26
D
66
_

_
_

_
Equations (33) can be written in matrix form as follows:
M
xx
M
yy
M
xy
_

_
_

B
11
B
12
B
16
B
12
B
22
B
26
B
16
B
26
B
66
D
11
D
12
D
16
D
12
D
22
D
26
D
16
D
26
D
66
_

_
_

_
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

_
35
Equation (35) relates the moment resultants applied to a composite
laminate to the resulting midplane strains and curvatures via the B
ij
and D
ij
matrices. The value of each termwithin the B
ij
and D
ij
matrices depends on the
material properties and ber angle of each ply (i.e., they depend on terms
within the Q
ij
matrix) as well as the stacking sequence (i.e., the distance z
k
of
each ply fromthe 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:
N
xx
N
yy
N
xy
M
xx
M
yy
M
xy
_

_
_

A
11
A
12
A
16
B
11
B
12
B
16
A
12
A
22
A
26
B
12
B
22
B
26
A
16
A
26
A
66
B
16
B
26
B
66
B
11
B
12
B
16
D
11
D
12
D
16
B
12
B
22
B
26
D
12
D
22
D
26
B
16
B
26
B
66
D
16
D
26
D
66
_

_
_

_
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

_
36
Equation (36) will sometimes be written in abbreviated form as:
N
M
_ _

A B
B D
_ _
e
o
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., A
12
=A
21
, B
12
=B
21
, D
12
=D
21
, etc.), the entire ABDmatrix
is also symmetrical.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
It should be noted that the above results are applicable to thin laminates
fabricated using any combination of ply materials. Because the A
ij
, B
ij
, and D
ij
matrices are each calculated based on a summation over all plies, and
individual ply properties (represented by the Q
ij
matrix) are embedded within
these summations, both ply material type and ber 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 dierent prepreg material
systems) are calculated using (Eq. (27a), (27b), and (34).
The A
ij
matrix relates in-plane stress resultants to in-plane midplane
strains. For this reason, the A
ij
terms are called extensional stinesses.
Similarly, the D
ij
matrix relates moment resultants to midplane curvatures,
and elements within the D
ij
matrix are therefore called bending stiness. The
B
ij
matrix relates in-plane stress resultants to midplane curvatures, and also
relates moment resultants to the in-plane midplane strains. The B
ij
terms are
called coupling stinesses. For an isotropic plate, the coupling stinesses 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 Hookes 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:
e
o
j
_ _

A B
B D
_ _
1
N
M
_ _
37
In this text, the inverse of the ABD matrix will be called the abd matrix:
a b
b d
_ _

A B
B D
_ _
1
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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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:
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

a
11
a
12
a
16
b
11
b
12
b
16
a
12
a
22
a
26
b
21
b
22
b
26
a
16
a
26
a
66
b
61
b
62
b
66
b
11
b
21
b
61
d
11
d
12
d
16
b
12
b
22
b
62
d
12
d
22
d
26
b
16
b
26
b
66
d
16
d
26
d
66
_

_
_

_
N
xx
N
yy
N
xy
M
xx
M
yy
M
xy
_

_
_

_
38
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, a
ij
and d
ij
, respectively, are also symmetrical. However, the
3 3 matrix that appears in the upper right-hand quadrant is not symmetrical
(b
12
p b
21
, b
16
p b
61
, and b
26
p b
62
). 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. Aside 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: t
1
=t
2
=t
3
=0.125 mm.
Figure 17 Side viewof the [30/0/90]
T
laminate consideredinSample Problem4.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Note that because an odd number of plies are used, the origin of the xyz
coordinate system exists at the midplane of ply 2. The ply interface
coordinates can be calculated as:
z
0
t=2 0:375mm =2 0:1875mm 0:0001875m
z
1
z
0
t
1
0:1875mm 0:125mm 0:0625mm 0:0000625m
z
2
z
1
t
2
0:0625mm 0:125mm 0:0625mm 0:0000625m
z
3
z
2
t
3
0:0625mm 0:125mm 0:1875mm 0:0001875m
We will also require the transformed reduced stiness 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):
Q
_
30
j
ply

Q
11
Q
12
Q
16
Q
12
Q
22
Q
26
Q
16
Q
26
Q
66
_

_
_

107:6 10
9
26:06 10
9
48:13 10
9
26:06 10
9
27:22 10
9
21:52 10
9
48:13 10
9
21:52 10
9
36:05 10
9
_

_
_

_
Pa
For ply 2 (the 0j ply):
Q
_
0
B
ply

Q
11
Q
12
Q
16
Q
12
Q
22
Q
26
Q
16
Q
26
Q
66
_

_
_

107:9 10
9
3:016 10
9
0
3:016 10
9
10:05 10
9
0
0 0 13:00 10
9
_

_
_

_
Pa
*
The Q
_
matrix for a 30j graphite-epoxy ply was calculated as a part of Example Problem 5.6
of Chap. 5.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
For ply 3 (the 90j ply):
Q
_
90
B
ply

Q
11
Q
12
Q
16
Q
12
Q
22
Q
26
Q
16
Q
26
Q
66
_

_
_

10:05 10
9
3:016 10
9
0
3:016 10
9
170:9 10
9
0
0 0 13:00 10
9
_

_
_

_
Pa
We can now calculate each member of the A
ij
, B
ij
, and D
ij
matrices, in
accordance with (Eq. (27a), (27b), and (34), respectively.
.
Using Eq. (27a), element A
11
is calculated as follows:
A
11

3
k 1
Q
11
_ _
k
z
k
z
k1

A
11
Q
11
_ _
1
z
1
z
0
Q
11
_ _
2
z
2
z
1
Q
11
_ _
3
z
3
z
2

A
11
107:6 10
9
_ _
:0000625 0:0001875 170:9 10
9
_ _
0:0000625 0:0000625 10:05 10
9
_ _
0:0001875 0:0000625
A
11
36:07 10
6
Pa m
The remaining elements of the A
ij
matrix are found in similar fashion:
A
ij

36:07 4:012 6:016
4:012 26:02 2:690
6:016 2:690 7:756
_

_
_

_ 10
6
Pa m
.
Using Eq. (27b), element B
11
is calculated as follows:
B
11

1
2

3
k 1
Q
11
_ _
k
z
2
k
z
2
k1
_ _
B
11

1
2
Q
11
_ _
1
z
2
1
z
2
0
_ _
Q
11
_ _
2
z
2
2
z
2
1
_ _
Q
11
_ _
3
z
2
3
z
2
2
_ _
_ _
B
11

1
2
107:6 10
9
_ _
:0000625
2
0:0001875
2
_ _ _
170:9 10
9
_ _
0:0000625
2
0:0000625
2
_ _
10:05 10
9
_ _
0:0001875
2
0:0000625
2
_ __
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
B
11
1:524 10
3
Pa m
2
The remaining elements of the B
ij
matrix are found in similar fashion:
B
ij

1:524 0:3601 0:7521
0:3601 2:245 0:3362
0:7521 0:3362 0:3601
_

_
_

_ 10
3
Pa m
2
_ _
In passing, in this example, it appears that B
12
is numerically equal to B
66
. This
is not true, in general. In this problem, the apparent numerical equivalence is
due to the fact that only four signicant digits have been used. Nevertheless,
for laminates produced using a single material system, it is often (but not
always) the case that B
12
cB
66
. 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 B
12
and B
66
are directly related to Q
12
and Q
66
,
respectively, their values are often nearly identical. Also, in Sec. 6.2, it will be
seen that all elements within the B
ij
matrix are zero for symmetrical laminates.
Hence, for symmetrical laminates, these two terms are, in fact, numerically
equal, that is, B
12
=B
66
=0 for symmetrical laminates.
.
Using Eq. (34), element D
11
is calculated as follows:
D
11

1
3

3
k 1
Q
11
_ _
k
z
3
k
z
3
k 1
_ _
D
11

1
3
Q
11
_ _
1
z
3
1
z
3
0
_ _
Q
11
_ _
2
z
3
2
z
3
1
_ _
Q
11
_ _
3
z
3
3
z
3
2
_ _
_ _
D
11

1
3
107:6 10
9
_ _
:0000625
3
0:0001875
3
_ _ _
170:9 10
9
_ _
0:0000625
3
0:0000625
3
_ _
10:05 10
9
_ _
0:0001875
3
0:0000625
3
_ __
D
11
0:2767Pa m
3
The remaining elements of the D
ij
matrix are found in similar fashion:
D
ij

0:2767 0:0620 0:1018
0:0620 2:513 0:0455
0:1018 0:0455 0:1059
_

_
_

_
Pa m
3
_ _
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
The [ABD] matrix can now be assembled:
ABD
36:07 10
6
4:012 10
6
6:016 10
6
1524 360:1 752:1
4:012 10
6
26:02 10
6
2:690 10
6
360:1 2245 336:2
6:016 10
6
2:690 10
6
7:756 10
6
752:1 336:2 360:1
1524 360:1 752:1 0:2767 0:0620 0:1018
360:1 2245 336:2 0:0620 2:513 0:0455
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:
abd
3:757 10
8
1:964 10
9
1:038 10
8
1:440 10
4
3:905 10
6
8:513 10
5
1:964 10
9
1:037 10
7
4:234 10
8
1:866 10
5
6:361 10
4
4:268 10
4
1:038 10
8
4:234 10
8
2:004 10
7
3:661 10
4
3:251 10
4
1:851 10
5
1:440 10
4
1:866 10
5
3:661 10
4
7:064 3:122 10
2
4:572
3:905 10
6
6:361 10
4
3:251 10
4
3:122 10
2
6:429 3:620
8:513 10
5
4:268 10
4
1:851 10
5
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:
N
xx
50 kN=m N
yy
10 kN=m N
xy
0 N=m
M
xx
1 N m=m M
yy
1 N m=m M
xy
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 xy coordinate system
(c) Ply stresses relative to the xy 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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
and curvature may be obtained through application of Eq. (38), which
becomes:
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

3:757 10
8
1:964 10
9
1:038 10
8
1:440 10
4
3:905 10
6
8:513 10
5
1:964 10
9
1:037 10
7
4:234 10
8
1:866 10
5
6:361 10
4
4:628 10
4
1:038 10
8
4:234 10
8
2:004 10
7
3:661 10
4
3:251 10
4
1:851 10
5
1:440 10
4
1:866 10
5
3:661 10
4
7:064 3:122 10
2
4:572
3:905 10
6
6:361 10
4
3:251 10
4
3:122 10
2
6:429 3:620
8:513 10
5
4:628 10
4
1:851 10
5
4:572 3:620 17:41
_

_
_

50 10
3
10 10
3
0
1
1
0
_

_
_

_
Completing this matrix multiplication, we obtain:
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

2039Am=m
518Am=m
55Arad
14:48m
1
0:096m
1
1:323m
1
_

_
_

_
(b) Ply strains relative to the xy 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 z
0
= 0.0001875 m) are:
e
xx
e
yy
c
xy
_

_
_

_j
z z
0

e
o
xx
e
o
yy
c
o
xy
_

_
_

_
z
0
j
xx
j
yy
j
xy
_
_
_
_
_
_

2038 10
6
m=m
518 10
6
m=m
55 10
6
m=m
_

_
_

_
0:0001875m
14:48rad=m
0:096rad=m
1:328rad=m
_

_
_

_
e
xx
e
yy
c
xy
_
_
_
_
_
_
j
z z
0

677Am=m
536Am=m
194Arad
_

_
_

_
Strains calculated at the remaining ply interface positions are summarized in
Table 5.
(c) Ply stresses relative to the xy coordinate system. The Q
_
matrix for all
plies was calculated as a part of Example Problem 4. Ply stresses may now be
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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=z
0
) are:
r
xx
r
yy
s
xy
_

_
_

_
j
ply 1
z z
0

Q
11
Q
12
Q
16
Q
12
Q
22
Q
26
Q
16
Q
26
Q
66
_

_
_

_
j
ply 1
z z
0
e
xx
e
yy
c
xy
_
_
_
_
_
_
j
z z
0
r
xx
r
yy
s
xy
_

_
_

_
j
ply 1
z z
0

107:6 10
9
26:06 10
9
48:13 10
9
26:06 10
9
27:22 10
9
21:52 10
9
48:13 10
9
21:52 10
9
36:05 10
9
_

_
_

_
677 10
6
536 10
6
194 10
6
_

_
_

_
r
xx
r
yy
s
xy
_

_
_

_
j
ply 1
z z
0

77:5MPa
28:1MPa
37:1MPa
_

_
_

_
Stresses calculated at remaining ply interface positions are summarized
in Table 6.
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
z-coordinate (mm) e
xx
(Am/m) e
yy
(Am/m) c
xy
(Arad)
0.1875 677 536 194
0.0625 1133 524 28
0.0625 2943 512 137
0.1875 4753 500 303
Strains are referenced to the xy coordinate system.
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) r
xx
(MPa) r
yy
(MPa) s
xy
(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
Stresses are referenced to the xy coordinate system.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
6.2 Including Changes in Environmental Conditions
Recall that we simplied the analysis leading up to Eq. (36) by assuming that
DT=DM=0. We will nowconsider howto 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:
r
xx
r
yy
s
xy
_

_
_

_
k

Q
11
Q
12
Q
16
Q
21
Q
22
Q
26
Q
61
Q
62
Q
66
_

_
_

_
k
e
xx
DTa
xx
DMb
xx
e
yy
DTa
yy
DMb
yy
c
xy
DTa
xy
DMb
xy
_

_
_

_
k
repeated 5:30
Stress r
xx
in ply k is given by:
r
xx
Q
11
e
xx
DTa
xx
DMb
xx
f g Q
12
e
yy
DTa
yy
DMb
xx
_ _
39
Q
16
c
xy
DTa
xy
DMb
xy
_ _
Stress resultant N
xx
is related to r
xx
via Eq. (1a). Substituting Eq. (39) into Eq.
(1a), we have:
N
xx

_
t=2
t=2
Q
11
e
xx
Q
12
e
yy
Q
16
c
xy
_ _
k
dz
DT
_
t=2
t=2
Q
11
a
xx
Q
12
a
yy
Q
16
c
xy
_ _
k
dz 40
DM
_
t=2
t=2
Q
11
b
xx
Q
12
b
yy
Q
16
b
xy
_ _
k
dz
The rst 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:
DT
_
t=2
t=2
Q
11
a
xx
Q
12
a
yy
Q
16
c
xy
_ _
k
dz
DT

n
k 1
Q
11
a
xx
Q
12
a
yy
Q
16
a
xy
_
k
z
k
z
k1

_ _
This quantity is called a thermal stress resultant, and will be denoted N
xx
T
.
That is,
N
T
xx
DT

n
k 1
Q
11
a
xx
Q
12
a
yy
Q
16
a
xy
_
k
z
k
z
k1

_ _
41a
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Similarly, the third integral in Eq. (40) can be evaluated to give the moisture
stress resultant, denoted N
M
xx
:
N
M
xx
DM

n
k 1
Q
11
b
xx
Q
12
b
yy
Q
16
b
xy
_
k
z
k
z
k1

_ _
42a
Hence, after evaluating all integrals, Eq. (40) may be written as:
N
xx
A
11
e
o
xx
A
12
e
o
yy
A
16
c
o
xy
B
11
j
xx
B
12
j
yy
B
16
j
xy
N
T
xx
N
M
xx
43a
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 (N
T
xx
and N
M
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), ve additional
thermal stress/moment resultants and ve additional moisture stress/moment
resultants will be identied, as follows:
N
T
yy
u DT

n
k 1
Q
12
a
xx
Q
22
a
yy
Q
26
a
xy
_
k
z
k
z
k1

_ _
41b
N
T
xy
u DT

n
k 1
Q
16
a
xx
Q
26
a
yy
Q
66
a
xy
_
k
z
k
z
k1

_ _
41c
M
T
xx
u
DT
2

n
k 1
Q
11
a
xx
Q
12
a
yy
Q
16
a
xy
_
k
z
2
k
z
2
k1
_ _ _
41d
M
T
yy
u
DT
2

n
k 1
Q
12
a
xx
Q
22
a
yy
Q
26
a
xy
_
k
z
2
k
z
2
k1
_ _ _
41e
M
T
xy
u
DT
2

n
k 1
Q
16
a
xx
Q
26
a
yy
Q
66
a
xy
_
k
z
2
k
z
2
k1
_ _ _
41f
N
M
yy
u DM

n
k 1
Q
12
b
xx
Q
22
b
yy
Q
26
b
xy
_
k
z
k
z
k1

_ _
42b
N
M
xy
u DM

n
k 1
Q
16
b
xx
Q
26
b
yy
Q
66
b
xy
_
k
z
k
z
k1

_ _
42c
M
M
xx
u
DM
2

n
k 1
Q
11
b
xx
Q
12
b
yy
Q
16
b
xy
_
k
z
2
k
z
2
k1
_
_ _
42d
M
M
yy
u
DM
2

n
k 1
Q
12
b
xx
Q
22
b
yy
Q
26
b
xy
_
k
z
2
k
z
2
k1
_
_ _
42e
M
M
xy
u
DM
2

n
k 1
Q
16
b
xx
Q
26
b
yy
Q
66
b
xy
_
k
z
2
k
z
2
k1
_
_ _
42f
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 formsimilar
to Eq. (36):
N
xx
N
yy
N
xy
M
xx
M
yy
M
xy
_

_
_

A
11
A
12
A
16
B
11
B
12
B
16
A
12
A
22
A
26
B
12
B
22
B
26
A
16
A
26
A
66
B
16
B
26
B
66
B
11
B
12
B
16
D
11
D
12
D
16
B
12
B
22
B
26
D
12
D
22
D
26
B
16
B
26
B
66
D
16
D
26
D
66
_

_
_

_
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

N
T
xx
N
T
yy
N
T
xy
M
T
xx
M
T
yy
M
T
xy
_

_
_

N
M
xx
N
M
yy
N
M
xy
M
M
xx
M
M
yy
M
M
xy
_

_
_

_
44
Equation (44) will sometimes be abbreviated as:
N
M
_ _

A B
B D
_ _
e
o
j
_ _

N
T
M
T
_ _

N
M
M
M
_ _
Equation (44) is comparable to Eq. (36), except we have now included the
eects due to a change in temperature and/or moisture content. Equation (44)
can be viewed as Hookes 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:
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

a
11
a
12
a
16
b
11
b
12
b
16
a
12
a
22
a
26
b
21
b
22
b
26
a
16
a
26
a
66
b
61
b
62
b
66
b
11
b
21
b
61
d
11
d
12
d
16
b
12
b
22
b
62
d
12
d
22
d
26
b
16
b
26
b
66
d
16
d
26
d
66
_

_
_

_
N
xx
N
T
xx
N
M
xx
N
yy
N
T
yy
N
M
yy
N
xy
N
T
xy
N
M
xy
M
xx
M
T
xx
M
M
xx
M
yy
M
T
yy
M
M
yy
M
xy
M
T
xy
M
M
xy
_

_
_

_
45
where, as before:
a b
b d
_ _

A B
B D
_ _
1
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
A subtlety embedded fact within the preceding discussion is that most
composites are subjected to a signicant 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 120jCor 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
uniformchange in temperature of DT=100jCor 155jCduring 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 roomtemperature 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 signicant diculty 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:
(a) Midplane strains and curvatures
(b) Ply strains relative to the xy coordinate system
(c) Ply stresses relative to the xy 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 2050jC below the nal cure
temperature. Nevertheless, this complication will be ignored in this text; it will be assumed that
the nal cure temperature denes the stress-free temperature.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 = (20175) = 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:
N
xx
N
yy
N
xy
M
xx
M
yy
M
xy
_

_
_

N
M
xx
N
M
yy
N
M
xy
M
M
xx
M
M
yy
M
M
xy
_

_
_

0
0
0
0
0
0
_

_
_

_
The eective thermal expansion coecients for each ply are calculated using
Eq. (25) of Chap. 5, repeated here for convenience:
a
xx
a
11
cos
2
h a
22
sin
2
h
a
yy
a
11
sin
2
h a
22
cos
2
h
a
xy
2cos h sin h a
11
a
22

5:25
From Table 3 of Chap. 3, the thermal expansion coecients for graphite-
epoxy (relative to the 12 coordinate system) are a
11
= 0.9 Am/m jC and
a
22
= 27Am/m jC. Therefore:
For ply 1 (the 30j ply):
a
1
xx
0:9Am=m
B
C cos
2
30
B
27Am=m
B
C sin
2
30
B

6:08Am=m
B
C
a
1
yy
0:9Am=m
B
C sin
2
30
B
27Am=m
B
C cos
2
30
B

20:0Am=m
B
C
a
1
xy
2cos 30 sin 30 0:9 27 Am=m
B
C 24:2Arad=
B
C
For ply 2 (the 0j ply):
a
2
xx
0:9Am=m
B
C cos
2
0
B
27Am=m
B
C sin
2
0
B

0:9Am=m
B
C
a
2
yy
0:9Am=m
B
C sin
2
0
B
27Am=m
B
C cos
2
0
B

27:0Am=m
B
C
a
2
xy
2cos 0
B
sin 0
B
0:9 27 Am=m
B
C 0Arad=
B
C
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
For ply 3 (the 90j ply):
a
3
xx
0:9Am=m
B
C cos
2
90
B
27Am=m
B
C sin
2
90
B

27:0Am=m
B
C
a
3
yy
0:9Am=m
B
C sin
2
90
B
27Am=m
B
C cos
2
90
B

0:9Am=m
B
C
a
3
xy
2cos 90
B
sin 90
B
0:9 27 Am=m
B
C 0Arad=
B
C
Both the ply interface positions as well as the Q
ij
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:
N
T
xx
u DT

n
k 1
Q
11
a
xx
Q
12
a
yy
Q
16
a
xy
_
k
z
k
z
k1

_ _
N
T
xx
DT
__
Q
11
a
xx
Q
12
a
yy
Q
16
a
xy
_
1
z
1
z
0

_

_
Q
11
a
xx
Q
12
a
yy
Q
16
a
xy
_
2
z
2
z
1

_
Q
11
a
xx
Q
12
a
yy
Q
16
a
xy
_
3
z
3
z
2

_ __
N
T
xx
155
_
107:6 10
9
_ _
6:08 10
6
_ _
26:06 10
9
_ _
20:0 10
6
_ _ _ _
48:13 10
9
_ _
24:2 10
6
_ __
0:0625 0:1875 10
3
_
170:9 10
9
_ _
0:9 10
6
_ _
3:016 10
9
_ _
27:0 10
6
_ _ _ _
0 0
_
0:0625 0:0625 10
3
_
10:05 10
9
_ _
27 10
6
_ _ _ _
3:016 10
9
_ _
0:9 10
6
_ _
0 0
_
0:1875 0:0625 10
3
_
_
N
T
xx
4060N=m
The remaining thermal stress and moment resultants are calculated in similar
fashion, eventually resulting in:
N
T
xx
N
T
yy
N
T
xy
M
T
xx
M
T
yy
M
T
xy
_

_
_

4060N=m
7360N=m
2860N=m
0:62N m=m
0:62N m=m
0:36N m=m
_

_
_

_
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
We can now calculate midplane strains and curvature using Eq. (45), which
becomes:*
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

3:757 10
8
1:964 10
9
1:038 10
8
1:440 10
4
3:905 10
6
8:513 10
5
1:964 10
9
1:037 10
7
4:234 10
8
1:866 10
5
6:361 10
4
4:628 10
4
1:038 10
8
4:234 10
8
2:004 10
7
3:661 10
4
3:251 10
4
1:851 10
5
1:440 10
4
1:866 10
5
3:661 10
4
7:064 3:122 10
2
4:572
3:905 10
6
6:361 10
4
3:251 10
4
3:122 10
2
6:429 3:620
8:513 10
5
4:628 10
4
1:851 10
5
4:572 3:620 17:41
_

_
_

4060
7360
2860
0:62
0:62
0:36
_

_
_

_
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

285Am=m
1424Am=m
908Arad
2:16m
1
10:9m
1
9:4m
1
_

_
_

_
(b) Ply strains relative to the xy coordinate system. Ply strains may nowbe
calculated using Eq. (12). For example, strains present at the outer surface of
ply 1 (i.e., strains present at z
o
=0.0001875 m) are:
e
xx
e
yy
c
xy
_

_
_

z z
0

e
o
xx
e
o
yy
c
o
xy
_

_
_

_
z
0
j
xx
j
yy
j
xy
_

_
_

285 10
6
m=m
1424 10
6
m=m
908 10
6
m=m
_

_
_

_
0:0001875m
2:16rad=m
10:9rad=m
9:4rad=m
_

_
_

_
*The [abd] matrix for a [30/0/90]
T
graphite-epoxy laminate was calculated in Sample Problem 3.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
e
xx
e
yy
c
xy
_
_
_
_
_
_

z z
0

120Am=m
3468Am=m
2672Am=m
_
_
_
_
_
_
Strains calculated at the remaining ply interface positions are summa-
rized in Table 7.
(c) Ply stresses relative to the xy 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=z
0
):
r
xx
r
yy
s
xy
_

_
_

ply 1
z z
0

Q
11
Q
12
Q
16
Q
12
Q
22
Q
26
Q
16
Q
26
Q
66
_

_
_

ply 1
z z
0
e
xx
DTa
xx
e
yy
DTa
yy
c
xy
DTa
xy
_

_
_

z z
0
r
xx
r
yy
s
xy
_

_
_

ply 1
zz
0

107:6 10
9
26:06 10
9
48:13 10
9
26:06 10
9
27:22 10
9
21:52 10
9
48:13 10
9
21:52 10
9
36:05 10
9
_

_
_

120 1556:08 10
6
3468 15520:0 10
6
2672 15524:2 10
6
_

_
_

_
r
xx
r
yy
s
xy
_
_
_
_
_
_

ply 1
z z
0

53MPa
5:2MPa
4:8MPa
_

_
_

_
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
z-coordinate (mm) e
xx
(Am/m) e
yy
(Am/m) c
xy
(Arad)
0.1875 120 3468 2672
0.0625 150 2100 1500
0.0625 420 750 320
0.1875 690 620 860
Strains are referenced to the xy coordinate system.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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:
(a) Midplane strains and curvatures
(b) Ply strains relative to the xy coordinate system
(c) Ply stresses relative to the xy 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 modied due to the
slow diusion 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 eective moisture expansion coecients
for each ply are calculated using Eq. (28) of Chap. 5, repeated here for
convenience:
b
xx
b
11
cos
2
h b
22
sin
2
h
b
yy
b
11
sin
2
h b
22
cos
2
h 5:28
b
xy
2coshsinhb
11
b
22

Table 8 Ply Interface Stresses in a [30/0/90]


T
Graphite-Epoxy
Laminate Caused by a Cooldown from 175jC to 20jC
Ply number z-coordinate (mm) r
xx
(MPa) r
yy
(MPa) s
xy
(MPa)
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
Stresses are referenced to the xy coordinate system.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
From Table 3 of Chap. 3, the moisture expansion coecients for graphite-
epoxy (relative to the 12 coordinate system) are b
11
=150 Am/m %M and
b
22
=4800 Am/m %M. Therefore:
For ply 1 (the 30j ply):
b
1
xx
150Am=m %M cos
2
30
B
4800Am=m %M sin
2
30
B

1310Am=m %M
b
1
yy
150Am=m %Msin
2
30
B
4800Am=m %Mcos
2
30
B

3640Am=m %M
b
1
xy
2cos30sin30 150 4800Am=m %M 4030Arad=%M
For ply 2 (the 0j ply):
b
2
xx
150Am=m %M cos
2
0
B
4800Am=m %M sin
2
0
B

150Am=m %M
b
2
yy
150Am=m %M sin
2
0
B
4800Am=m %M cos
2
0
B

4800Am=m %M
b
2
xy
2cos0
B
sin0
B
150 4800Am=m %M 0Arad=%M
For ply 3 (the 90j ply):
b
3
xx
150Am=m %M cos
2
90
B
4800Am=m %M sin
2
90
B

4800Am=m %M
b
3
yy
150Am=m %M sin
2
90
B
4800Am=m %M cos
2
90
B

150Am=m %M
b
3
xy
2cos90
B
sin90
B
150 4800 Am=m %M 0Arad=%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:
N
M
xx
DM

3
k 1
Q
11
b
xx
Q
12
b
yy
Q
16
b
xy
_
k
z
k
z
k1

_ _
N
M
xx
DM Q
11
b
xx
Q
12
b
yy
Q
16
b
xy
_
1
z
1
z
0

_ _ _
Q
11
b
xx
Q
12
b
yy
Q
16
b
xy
_
2
z
2
z
1

_ _
Q
11
b
xx
Q
12
b
yy
Q
16
b
xy
_
3
z
3
z
2

_ __
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
N
M
xx
0:5 107:6 10
9
1312 10
6

_
26:06 10
9
3638 10
6
48:13 10
9
4027 10
6

0:0625 0:1875 10
3
170:9 10
9
150 10
6

3:016 10
9
4800 10
6
000:0625 0:0625
10
3
10:05 10
9
4800 10
6
3:016 10
9

150 10
6
000:1875 0:0625 10
3

_
N
T
xx
8190 N=M
The remaining thermal stress and moment resultants are calculated in similar
fashion, eventually resulting in:
N
T
xx
N
T
yy
N
T
xy
M
T
xx
M
T
yy
M
T
xy
_

_
_

8190 N=m
8460 N=m
233 N=m
0:05 N m=m
0:05 N m=m
0:03 N m=m
_

_
_

_
We can now calculate midplane strains and curvatures using Eq. (45), which
becomes:
*
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

3:757 10
8
1:964 10
9
1:038 10
8
1:440 10
4
3:905 10
6
8:153 10
5
1:964 10
9
1:037 10
7
4:234 10
8
1:866 10
5
6:361 10
4
4:628 10
4
1:038 10
8
4:234 10
8
2:004 10
7
3:661 10
4
3:251 10
4
1:851 10
5
1:440 10
4
1:866 10
5
3:661 10
4
7:064 3:122 10
2
4:572
3:905 10
6
6:361 10
4
3:251 10
4
3:122 10
2
6:429 3:620
8:513 10
5
4:628 10
4
1:851 10
5
4:572 3:620 17:41
_

_
_

4060 8190
7360 8460
2860 233
0:62 0:05
0:62 0:05
0:36 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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

18Am=m
509Am=m
420Arad
1:0m
1
5:0m
1
4:4m
1
_

_
_

_
(b) Ply strains relative to the xy 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 z
0
=0.0001875 m) are:
e
xx
e
yy
c
xy
_

_
_

zz
0

e
o
xx
e
o
yy
c
o
xy
_

_
_

_
z
0
j
xx
j
yy
j
xy
_

_
_

18 10
6
m=m
509 10
6
m=m
420 10
6
m=m
_

_
_

_
0:0001875m

1:0rad=m
5:0rad=m
4:4rad=m
_

_
_

_
e
xx
e
yy
c
xy
_

_
_

zz
0

206Am=m
1450Am=m
1240Am=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%
z-coordinate (mm) e
xx
(Am/m) e
yy
(Am/m) c
xy
(Arad)
0.1875 206 1450 1240
0.0625 80 820 690
0.0625 44 190 150
0.1875 170 440 400
Strains are referenced to the xy coordinate system.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
(c) Ply stresses relative to the xy coordinate system. Ply stresses may nowbe
calculated using Eq. (30) of Chap. 5. The stresses present at the outer surface
of ply 1 are (i.e., at z=z
0
):
r
xx
r
yy
s
xy
_

_
_

ply 1
z z
0

Q
11
Q
12
Q
16
Q
12
Q
22
Q
26
Q
16
Q
26
Q
66
_

_
_

ply 1
z z
0
e
xx
DTa
xx
DMb
xx
e
yy
DTa
yy
DMb
yy
c
xy
DTa
xy
DMb
xy
_

_
_

z z
0
r
xx
r
yy
s
xy
_
_
_
_
_
_

ply 1
z z
0

107:6 10
9
26:06 10
9
48:13 10
9
26:06 10
9
27:22 10
9
21:52 10
9
48:13 10
9
21:52 10
9
36:05 10
9
_

_
_

206 1556:08 0:51312 10


6
1450 15520:0 0:53638 10
6
1240 15524:2 0:54027 10
6
_

_
_

_
r
xx
r
yy
s
xy
_

_
_

ply 1
z z
0

25MPa
2:4MPa
2:2MPa
_

_
_

_
Stresses calculated at the remaining plies and ply interface positions are
summarized in Table 10.
Acomparison of the results obtained in Example Problems 6 and 7 leads
to the following observation: The initial ply stresses and strains caused by
cooldown fromcure temperatures to roomtemperatures 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 number z-coordinate (mm) r
xx
(MPa) r
yy
(MPa) s
xy
(MPa)
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 xy coordinate system.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
by the subsequent adsorption of moisture. Although the interaction between
temperature and moisture eects 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 eects 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
simplied through proper selection of the laminate stacking sequence. Before
these simplications are discussed, however, it is illustrative to consider the
simplest case of all-specically, 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:
E
11
E
22
E
m
12
m
21
m
G
12
G
a
11
a
22
a
b
11
b
22
b
Also recall that for isotropic materials, only two of the elastic moduli are
independent. That is:
G
E
21 m
If these interrelations between material properties are enforced, then Eq. (44)
reduces to:
N
xx
N
yy
N
xy
M
xx
M
yy
M
xy
_

_
_

A
11
A
12
0 0 0 0
A
12
A
11
0 0 0 0
0 0
A
11
A
12
2
_ _
0 0 0
0 0 0 D
11
D
12
0
0 0 0 D
12
D
11
0
0 0 0 0 0
D
11
D
12
2
_ _
_

_
_

_
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

N
T
N
T
0
0
0
0
_

_
_

N
M
N
M
0
0
0
0
_

_
_

_
(46)
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
where:
A
11

Et
1 m
2
A
12
mA
11

mEt
1 m
2
D
11
D
22

Et
3
121 m
2

D
12
mD
11

mEt
3
121 m
2

D
66

D
11
D
12

2

Et
3
241 m
N
T
DT
Eta
1 m
_ _
N
M
DM
Etb
1 m
_ _
The constant D
11
is often called the exural rigidity of an isotropic plate.
Taking the inverse of Eq. (46), we nd:
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

a
11
a
12
0 0 0 0
a
12
a
11
0 0 0 0
0 0 2 a
11
a
12
0 0 0
0 0 0 d
11
d
12
0
0 0 0 d
12
d
11
0
0 0 0 0 0 2 d
11
d
12

_

_
_

_
N
xx
N
T
N
M
N
yy
N
T
N
M
N
xy
M
xx
M
yy
M
xy
_

_
_

_
47
where:
a
11

1
A
11
1 m
2


1
Et
a
12
ma
11
m
A
11
1 m
2


m
Et
d
11
d
22

1
D
11
1 m
2


12
Et
3
d
12
md
11

m
D
11
1 m
2


12m
Et
3
d
66
2 d
11
d
12

24 1 m
Et
3
Comparing Eqs. (44) and (45) with Eqs. (46) and (47), it is apparent that
multiangle composite laminates may exhibit unusual coupling eects, 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
N
xx
will (in general) induce a midplane shear strain cj
xy
and curvatures j
xx
,
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
j
yy
, and j
xy
, due to the presence of the a
16
, b
11
, b
12
, and b
16
terms, respectively.
Physically, these means that a uniformin-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 (N
ij
T
and M
ij
T
, respectively) develop
during cooldown. Equation (45) shows that the thermal stress resultants N
ij
T
and M
ij
T
will cause curvatures j
xx
, j
yy
, and j
xy
to develop upon cooldown.
Physically, this means that (in general) a composite laminate that is at at the
elevated cure temperature will bend and warp as it is cooled to room
temperature. Coupling eects due to moisture stress and moment resultants,
N
ij
T
and M
ij
T
, 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 N
ij
T
=M
ij
T
=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 eects greatly complicate the design of composite
structures, it is of interest to determine whether these coupling eects can be
reduced or eliminated. It will be seen that it is indeed possible to reduce or
eliminate many of these coupling eects through proper selection of the
laminate stacking sequence. Common stacking sequences used to eliminate
coupling eects 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 ber angle. Several examples of
symmetrical stacking sequences have been previously shown in Fig. 12. For
a symmetrical n-ply laminate, the material and ber angle used in ply 1 is
identical to that used in ply n, the material and ber 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-
cations to Eqs. (44) and (45). Specically, for a symmetrical laminate:
All coupling stinesses equal zero (B
ij
=0).
All thermal moment resultants equal zero (M
ij
T
=0).
All moisture moment resultants equal zero (M
ij
M
=0).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
To demonstrate that coupling stinesses are zero for a symmetrical
laminate, consider the coupling stiness B
11
. From Eq. (27b), B
11
is given by:
B
11

1
2

n
k1
Q
11
_ _
k
z
2
k
z
2
k1
_ _
In expanded form, B
11
is given by:
B
11

1
2
Q
11
_ _
1
z
2
1
z
2
0
_
Q
11
_ _
2
z
2
2
z
2
1
_
Q
11
_ _
3
z
2
3
z
2
2
_
_

: : :
Q
11
_ _
n2
z
2
n2
z
2
n3
_
Q
11
_ _
n1
z
2
n1
z
2
n2
_
48
Q
11
_ _
n
z
2
n
z
2
n1
_
g
Because the laminate is assumed to be symmetrical, it must be that:
Q
11
_ _
1
Q
11
_ _
n
Q
11
_ _
2
Q
11
_ _
n1
49a
Q
11
_ _
3
Q
11
_ _
n2

etc:
Also, due to symmetry, the ply interface positions are located symmetrically
about the midplane, and hence (recalling that z
0
<0 and z
n
>0):
z
0
z
n
z
1
z
n1
z
2
z
n2
49b

etc:
Together, the relations listed as Eqs. (49a) and (49b) imply that for any
symmetrical laminate:
Q
11
_ _
1
z
2
1
z
2
0
_
Q
11
_ _
n
z
2
n
z
2
n1
_
Q
11
_ _
2
z
2
2
z
2
1
_
Q
11
_ _
n1
z
2
n1
z
2
n2
_
50
Q
11
_ _
3
z
2
3
z
2
2
_
Q
11
_ _
n2
z
2
n2
z
2
n3
_

etc:
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Substituting Eq. (50) into Eq. (48), it is seen that B
11
=0. Similar results may
be demonstrated for all other coupling stinesses, and hence B
ij
= 0 for any
symmetrical laminate, as stated.
To demonstrate that thermal moment resultants are zero for a symmet-
rical laminate, consider the thermal moment resultant M
T
xx
. From Eq. (42d),
M
T
xx
is given by:
M
T
xx

DT
2

n
k 1
Q
11
a
xx
Q
12
a
yy
Q
16
a
xy
_
k
z
2
k
z
2
k1
_ _ _
In expanded form, M
T
xx
is given by:
M
T
xx

DT
2
_
Q
11
a
xx
Q
12
a
yy
Q
16
a
xy
_
1
z
2
1
z
2
0
_
Q
11
a
xx
Q
12
a
yy
Q
16
a
xy
_
2
z
2
2
z
2
1
_
Q
11
a
xx
Q
12
a
yy
Q
16
a
xy
_
3
z
2
3
z
2
2
_
. . . . . . 51
Q
11
a
xx
Q
12
a
yy
Q
16
a
xy
_
n2
z
2
n2
z
2
n3
_
Q
11
a
xx
Q
12
a
yy
Q
16
a
xy
_
n1
z
2
n1
z
2
n2
_
Q
11
a
xx
Q
12
a
yy
Q
16
a
xy
_
n
z
2
n
z
2
n1
_
_
Because the laminate is assumed to be symmetrical, it must be that:
Q
11
a
xx
Q
12
a
yy
Q
16
a
xy
_
1
z
2
1
z
2
0
_
Q
11
a
xx
Q
12
a
yy
Q
16
a
xy
_
n
z
2
n
z
2
n1
_
Q
11
a
xx
Q
12
a
yy
Q
16
a
xy
_
2
z
2
2
z
2
1
_
Q
11
a
xx
Q
12
a
yy
Q
16
a
xy
_
n1
z
2
n1
z
2
n2
_
Q
11
a
xx
Q
12
a
yy
Q
16
a
xy
_
3
z
2
3
z
2
2
_
Q
11
a
xx
Q
12
a
yy
Q
16
a
xy
_
n2
z
2
n2
z
2
n3
_
Substituting Eq. (52) into Eq. (51), it is seen that M
xx
T
=0 for a symmetrical
laminate. Similar results may be demonstrated for M
yy
T
and M
xy
T
, 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.
(52)
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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:
N
xx
N
yy
N
xy
M
xx
M
yy
M
xy
_

_
_

A
11
A
12
A
16
0 0 0
A
12
A
22
A
26
0 0 0
A
16
A
26
A
66
0 0 0
0 0 0 D
11
D
12
D
16
0 0 0 D
12
D
22
D
26
0 0 0 D
16
D
26
D
66
_

_
_

_
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

N
T
xx
N
T
yy
N
T
xy
0
0
0
_

_
_

N
M
xx
N
M
yy
N
M
xy
0
0
0
_

_
_

_
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

a
11
a
12
a
16
0 0 0
a
12
a
22
a
26
0 0 0
a
16
a
26
a
66
0 0 0
0 0 0 d
11
d
12
d
16
0 0 0 d
12
d
22
d
26
0 0 0 d
16
d
26
d
66
_

_
_

_
N
xx
N
T
xx
N
M
xx
N
yy
N
T
yy
N
M
yy
N
xy
N
T
xy
N
M
xy
M
xx
M
yy
M
xy
_

_
_

_
53b
Due to these dramatical simplications, 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 stinesses and thermal and moisture moment
resultants.
It is noted in passing that unidirectional composite laminates [h]
n
are
symmetrical and hence B
ij
=M
ij
T
=M
ij
M
=0 for unidirectional laminates. In
addition, for the case of [0j]
n
or [90j]
n
laminates, A
16
=A
26
=D
16
=D
26
=N
T
xy
=N
M
xy
0.
7.2 Cross-Ply Laminates
Composite laminates that contain plies with ber 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, Q
16
=Q
26
0. From(Eq. (27a) (27b) and (34) it is seen that
A
16
, A
26
, B
16
, B
26
, D
16
, and D
26
all involve a summation of terms involving Q
16
and Q
26
, and hence all of these terms also equal zero for cross-ply laminates.
Furthermore, fromEq. (25) of Chap. 5, it is seen that a
xy
=0 for any 0j or 90j
ply, and fromEq. (38) of Chap. 5 that b
xy
=0 for any 0j or 90j ply. FromEqs.
(42c), (42d), (42e), and (42f), it is seen that N
T
xy
M
T
xy
N
M
xy
M
M
xy
0 for
(53a)
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
cross-ply laminates (because Q
16
Q
26
a
xy
b
xy
0 . For cross-ply
laminates then, Eqs. (44) and (45) reduce to:
N
xx
N
yy
N
xy
M
xx
M
yy
M
xy
_

_
_

A
11
A
12
0 B
11
B
12
0
A
12
A
22
0 B
12
B
22
0
0 0 A
66
0 0 B
66
B
11
B
12
0 D
11
D
12
0
B
12
B
22
0 D
12
D
22
0
0 0 B
66
0 0 D
66
_

_
_

_
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

N
T
xx
N
T
yy
0
M
T
xx
M
T
yy
0
_

_
_

N
M
xx
N
M
yy
0
M
M
xx
M
M
yy
0
_

_
_

_
54a
and
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

a
11
a
12
0 b
11
b
12
0
a
12
a
22
0 b
21
b
22
0
0 0 a
66
0 0 b
66
b
11
b
21
0 d
11
d
12
0
b
12
b
22
0 d
12
d
22
0
0 0 b
66
0 0 d
66
_

_
_

_
N
xx
N
T
xx
N
M
xx
N
yy
N
T
yy
N
M
yy
N
xy
M
xx
M
T
xx
M
M
xx
M
yy
M
T
yy
M
M
yy
M
xy
_

_
_

_
54b
If a cross-ply laminate is also symmetrical, then all remaining coupling
stinesses equal zero, as well as all remaining thermal and moisture moment
resultants. Hence, for symmetrical cross-ply laminates, Eqs. (54a) and (54b)
are further simplied to:
N
xx
N
yy
N
xy
M
xx
M
yy
M
xy
_

_
_

A
11
A
12
0 0 0 0
A
12
A
22
0 0 0 0
0 0 A
66
0 0 0
0 0 0 D
11
D
12
0
0 0 0 D
12
D
22
0
0 0 0 0 0 D
66
_

_
_

_
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

N
T
xx
N
T
yy
0
0
0
0
_

_
_

N
M
xx
N
M
yy
0
0
0
0
_

_
_

_
55a
(and)
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

a
11
a
12
0 0 0 0
a
12
a
22
0 0 0 0
0 0 a
66
0 0 0
0 0 0 d
11
d
12
0
0 0 0 d
12
d
22
0
0 0 0 0 0 d
66
_

_
_

_
N
xx
N
T
xx
N
M
xx
N
yy
N
T
yy
N
M
yy
N
xy
M
xx
M
yy
M
xy
_

_
_

_
55b
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Note from Eqs. (55a) and (55b) that symmetrical cross-ply laminates do not
exhibit any coupling stinesses. That is:
A
16
A
26
D
16
D
26
B
ij
0
or, equivalently,
a
16
a
26
d
16
d
26
b
ij
0
Laminates that do not possess these coupling stinesses 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 ber angle h, there exists a
second ply whose ber 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 Q
16
and Q
26
terms for these
balanced plies will always be of equal magnitude but opposite algebraic sign
(i.e., Q
16
j
h
Q
16
j
h
and Q
26
j
h
Q
26
j
h
. Consequently, from Eq. (27a),
A
16
=A
26
=0 for a balanced laminate. Further, from Eqs. (25) and (28) of
Chap. 5, it is seen that a
xy
and b
xy
for the two balanced plies will always be of
equal magnitude but opposite algebraic sign (i.e., a
xy
j
h
a
xy
j
h
and b
xy
j
h
b
xy
j
h
. Consequently, from Eqs. (42c) and (43), N
T
xy
=N
M
xy
=0 for a
balanced laminate. Equations (44) and (45) are therefore simplied to:
N
xx
N
yy
N
xy
M
xx
M
yy
M
xy
_

_
_

A
11
A
12
0 B
11
B
12
B
16
A
12
A
22
0 B
12
B
22
B
26
0 0 A
66
B
16
B
26
B
66
B
11
B
12
B
16
D
11
D
12
D
16
B
12
B
22
B
26
D
12
D
22
D
26
B
16
B
26
B
66
D
16
D
26
D
66
_

_
_

_
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

N
T
xx
N
T
yy
0
M
T
xx
M
T
yy
M
T
xy
_

_
_

N
M
xx
N
M
yy
0
M
M
xx
M
M
yy
M
M
xy
_

_
_

_
56a
and
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

a
11
a
12
0 b
11
b
12
b
16
a
12
a
22
0 b
21
b
22
b
26
0 0 a
66
b
61
b
62
b
66
b
11
b
12
b
61
d
11
d
12
d
16
b
12
b
22
b
62
d
12
d
22
d
26
b
16
b
26
b
66
d
16
d
26
d
66
_

_
_

_
N
xx
N
T
xx
N
M
xx
N
yy
N
T
yy
N
M
yy
N
xy
M
xx
M
T
xx
M
M
xx
M
yy
M
T
yy
M
M
yy
M
xy
M
T
xy
M
M
xy
_

_
_

_
56b
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
If a balanced laminate is also symmetrical, then Eqs. (56a) and (56b) are
simplied to:
N
xx
N
yy
N
xy
M
xx
M
yy
M
xy
_

_
_

A
11
A
12
0 0 0 0
A
12
A
22
0 0 0 0
0 0 A
66
0 0 0
0 0 0 D
11
D
12
D
16
0 0 0 D
12
D
22
D
26
0 0 0 D
16
D
26
D
66
_

_
_

_
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

N
T
xx
N
T
yy
0
0
0
0
_

_
_

N
M
xx
N
M
yy
0
0
0
0
_

_
_

_
57a
and
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

a
11
a
12
0 0 0 0
a
12
a
22
0 0 0 0
0 0 a
66
0 0 0
0 0 0 d
11
d
12
d
16
0 0 0 d
12
d
22
d
26
0 0 0 d
16
d
26
d
66
_

_
_

_
N
xx
N
T
xx
N
M
xx
N
yy
N
T
yy
N
M
yy
N
xy
M
xx
M
yy
M
xy
_

_
_

_
57b
7.4 Balanced Angle-Ply Laminates
The denition 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 simplication that
occurs.
All plies in an angle-ply laminate have a ber angle of the same
magnitude. That is, all plies within an angle-ply laminate have a ber 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 ber angle+h may dier from
the number of plies with ber angle h. However, a balanced angle-ply
laminate must have an equal number of plies with ber angle+h and h, so as
to satisfy the preceding denition 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 ber angle. For
example, a [35/65/-35/-65]
T
laminate is balanced (although not symmetrical),
and A
16
=A
26
=0 for this laminate. In contrast, a balanced angle-ply laminate
may involve only one distinct angle. A stacking sequence of either [35/-35/
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
35/-35]
T
or [65/65/-65/-65]
T
result in balanced angle-ply laminates, for
example.
The following simplications occur for a balanced angle-ply laminate:
B
11
B
22
B
66
M
T
xx
M
T
yy
M
T
xx
M
T
yy
0
In addition, the simplications that exist for any balanced laminate
(A
16
=A
26
=N
xy
T
=N
xy
M
=0) also occur for a balanced angle-ply laminate.
Equations (44) and (45) are therefore simplied to:
N
xx
N
yy
N
xy
M
xx
M
yy
M
xy
_

_
_

A
11
A
12
0 0 0 B
16
A
12
A
22
0 0 0 B
26
0 0 A
66
B
16
B
26
0
0 0 B
16
D
11
D
12
D
16
0 0 B
26
D
12
D
22
D
26
B
16
B
26
0 D
16
D
26
D
66
_

_
_

_
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

N
T
xx
N
T
yy
0
0
0
M
T
xy
_

_
_

N
M
xx
N
M
yy
0
0
0
M
M
xy
_

_
_

_
58a
and
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

a
11
a
12
0 0 0 b
16
a
12
a
22
0 0 0 b
26
0 0 a
66
b
16
b
62
b
66
0 0 b
61
d
11
d
12
d
16
0 0 b
62
d
12
d
22
d
26
b
16
b
26
0 d
16
d
26
d
66
_

_
_

_
N
xx
N
T
xx
N
M
xx
N
yy
N
T
yy
N
M
yy
N
xy
M
xx
M
yy
M
xy
M
T
xy
M
M
xy
_

_
_

_
58b
If a balanced angle-ply laminate is also symmetrical, then Eqs. (58a) and (58b)
are simplied still further to:
N
xx
N
yy
N
xy
M
xx
M
yy
M
xy
_

_
_

A
11
A
12
0 0 0 0
A
12
A
22
0 0 0 0
0 0 A
66
0 0 0
0 0 0 D
11
D
12
D
16
0 0 0 D
12
D
22
D
26
0 0 0 D
16
D
26
D
66
_

_
_

_
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

N
T
xx
N
T
yy
0
0
0
0
_

_
_

N
M
xx
N
M
yy
0
0
0
0
_

_
_

_
59a
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
and
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

a
11
a
12
0 0 0 0
a
12
a
22
0 0 0 0
0 0 a
66
0 0 0
0 0 0 d
11
d
12
d
16
0 0 0 d
12
d
22
d
26
0 0 0 d
16
d
26
d
66
_

_
_

_
N
xx
N
T
xx
N
M
xx
N
yy
N
T
yy
N
M
yy
N
xy
M
xx
M
yy
M
xy
_

_
_

_
59b
7.5 Quasi-Isotropic Laminates
A quasi-isotropic laminate is one that satises the following conditions:

Three or more distinct ber angles must be present within a laminate.


The number of distinct ber angles will be denoted m, and hence if a
laminate is quasi-isotropic, then m z 3.

The m distinct ber angles must appear at equal increments of (180/


m) degrees.

An equal number of plies must be present at each of the m distinct


ber angles.
It can be shown (see Ref. 2) that if the three conditions are met, then
members of the A
ij
matrix for the laminate are related as follows:
A
11
A
22
A
66

1
2
A
11
A
12
A
16
A
26
0
Now, these same relations between extensional stiness 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
laminate, N
T
xy
M
T
xy
N
M
xy
M
M
xy
0 . In this case, Eqs. (44) and (45)
reduce to:
N
xx
N
yy
N
xy
M
xx
M
yy
M
xy
_

_
_

A
11
A
12
0 B
11
B
12
B
16
A
12
A
22
0 B
12
B
22
B
26
0 0
A
11
A
12
2
_ _
B
16
B
26
B
66
B
11
B
12
B
16
D
11
D
12
D
16
B
12
B
22
B
26
D
12
D
22
D
26
B
16
B
26
B
66
D
16
D
26
D
66
_

_
_

_
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

N
T
xx
N
T
yy
0
M
T
xx
M
T
yy
0
_

_
_

N
M
xx
N
M
yy
0
M
M
xx
M
M
yy
0
_

_
_

_
60a
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

a
11
a
12
0 b
11
b
12
b
16
a
12
a
22
0 b
21
b
22
b
26
0 0 2 a
11
a
12
b
61
b
62
b
66
b
11
b
21
b
61
d
11
d
12
d
16
b
12
b
22
b
62
d
12
d
22
d
26
b
16
b
26
b
66
d
16
d
26
d
66
_

_
_

_
N
xx
N
T
xx
N
M
xx
N
yy
N
T
yy
N
M
yy
N
xy
M
xx
M
T
xx
M
M
xx
M
yy
M
T
yy
M
M
yy
M
xy
_

_
_

_
60b
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 ber 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 stinesses for these laminates are quasi-isotropic,
they still exhibit coupling stinesses (i.e., B
ij
p 0); furthermore, the bending
stinesses are not isotropic (e.g., D
11
p D
22
).
If a quasi-isotropic laminate is also symmetrical (e.g., [0/45/90/45]
s
),
then the coupling stinesses B
ij
=0, and all remaining thermal moment
resultants and moisture moment resultants equal zero. Hence, Eqs. (60a)
and (60b) are simplied still further:
N
xx
N
yy
N
xy
M
xx
M
yy
M
xy
_

_
_

A
11
A
12
0 0 0 0
A
12
A
11
0 0 0 0
0 0
A
11
A
12
2
_ _
0 0 0
0 0 0 D
11
D
12
D
16
0 0 0 D
12
D
22
D
26
0 0 0 D
16
D
26
D
66
_

_
_

_
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

N
T
xx
N
T
yy
0
0
0
0
_

_
_

N
M
xx
N
M
yy
0
0
0
0
_

_
_

_
61a
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

a
11
a
12
0 0 0 0
a
12
a
11
0 0 0 0
0 0 2 a
11
a
12
0 0 0
0 0 0 d
11
d
12
d
16
0 0 0 d
12
d
22
d
26
0 0 0 d
16
d
26
d
66
_

_
_

_
N
xx
N
T
xx
N
M
xx
N
yy
N
T
yy
N
M
yy
N
xy
M
xx
M
yy
M
xy
_

_
_

_
61b
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Again, note that only the in-plane extensional stinesses are isotropic; the
bending stinesses 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, D
16
,D
26
,d
16
,d
26
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 dierent
analysis requirements are commonly encountered. In the rst 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 dierent calculation path is used in these two dierent
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:
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
1. Dene the problem:
(a) Specify the number of dierent ply materials to be used in the
laminate.
(b) Specify the elastic properties for each ply material used (i.e.,
specify E
11
, E
22
, v
12
, G
12
, a
11
, a
22
, b
11
, and b
22
for each ply
material).
(c) Specify the laminate description (i.e., specify the number of plies
n and the ply material and ber angle for each ply).
(d) Specify the mechanical and environmental loads applied to the
laminate (specify N
xx
, N
yy
, N
xy
, M
xx
, M
yy
, M
xy
, DT, and DM).
2. Calculate the ABD matrix:
(a) Calculate the reduced stiness matrix Q
ij
for each material used
in the laminate (using Eq. (11) of Chap. 5).
(b) Calculate the transformed reduced stiness matrix Q
ij
for each
ply based on the appropriate reduced stiness matrix and ply
ber angle (using Eq. (31) of Chap. 5).
(c) Calculate the A
ij
, B
ij
, and D
ij
matrices, using (Eq. (27a), (27b),
and (34), respectively.
(d) Assemble the ABD matrix.
3. Calculate the inverse of the ABD matrix: abd=ABD
1
.
4. Calculate the thermal and moisture stress and moment resultants:
(a) Calculate the eective thermal and moisture expansion coef-
cients 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 xy 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 z
0
, z
1
, z
2
, z
3
, . . ., z
n
). Strains are also often transformed to
the local 12 coordinate system, dened by the ber angle within
each ply.
(b) Calculate the ply stresses in the xy 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 z
0
, z
1
, z
2
, z
3
, . . ., z
n
). Stresses
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
are also often transformed to the local 12 coordinate system,
dened by the ber 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. Dene the problem:
(a) Specify the number of dierent ply materials to be used in the
laminate.
(b) Specify the elastic properties for each ply material used (i.e.,
specify E
11
, E
22
, v
12
, G
12
, a
11
, a
22
, b
11
, and b
22
for each ply
material).
(c) Specify the laminate description (i.e., specify the number of plies n
and the ply material and ber angle for each ply).
(d) Specify the midplane strains and curvatures and environmental
loads experienced by the laminate (specify e
xx
j , e
yy
j , c
xy
j , j
xx
, j
yy
,
j
xy
, DT, and DM).
2. Calculate the ABD matrix:
(a) Calculate the reduced stiness matrix Q
ij
for each material used
in the laminate (using Eq. (11) of Chap. 5).
(b) Calculate the transformed reduced stiness matrix Q
ij
for each
ply based on the appropriate reduced stiness matrix and ply
ber angle (using Eq. (31) of Chap. 5).
(c) Calculate the A
ij
, B
ij
, and D
ij
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 eective thermal and moisture expansion coef-
cients for each ply, using Eqs. (25) and (28) of Chap. 5,
respectively.
(b) Calculate the thermal and moisture stress and moment resultants
using Eqs. (42a42f) (43), respectively.
4. Calculate the applied stress and moment resultants, using Eq. (44).
5. For each ply:
(a) Calculate the ply strains in the xy 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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
positions z
0
, z
1
, z
2
, z
3
, . . . , z
n
). Strains are also often transformed
to the local 12 coordinate system, dened by the ber angle
within each ply.
(b) Calculate the ply stresses in the xy 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 z
0
, z
1
, z
2
, z
3
, . . . , z
n
). Stresses
are also often transformed to the local 12 coordinate system,
dened by the ber angle within each ply.
9 EFFECTIVE PROPERTIES OF A COMPOSITE LAMINATE
The denitions of common engineering material properties were reviewed in
Chap. 3. In this section, these concepts will be used to dene the eective
properties of multiangle composite laminates. Before we begin this discus-
sion, it is pertinent to note that most of the properties dened in Chap. 3 do
not take into account the peculiar coupling eects exhibited by general
composite laminates. As a typical example, consider the denition of the
thermal expansion coecient. As discussed in Sec. 3.3, the thermal expansion
coecient is dened as the ratio of a thermal strain divided by the temperature
change that caused that strain. Six dierent thermal expansion coecients
may be dened for an anisotropic material (see Eq. (21) of Chap. 3). For
example, if a uniform temperature change (DT) causes a normal strain in the
x-direction (e
xx
T
), then the thermal expansion coecient associated with the x-
direction is:
a
xx

e
T
xx
DT
However, this denition does not anticipate the out-of-plane coupling eects
exhibited by general composite laminates. That is, for a general composite
laminate, a temperature change DT causes both midplane strain e
xx
j and
midplane curvature j
xx
, and consequently e
xx
T
varies through the laminate
thickness. In this case then, what strain should be used to calculate a
xx
?
A reasonable approach is to dene a
xx
based on the thermal strain
induced at the laminate midplane, a
xx
=e
xx
j /DT, regardless of whether or not
a curvature is induced as well. Asecond thermal property can then be dened
as the ratio of midplane curvature to temperature change (j
xx
/DT).
Although several new properties representing unusual coupling eects
can be dened in this manner, this topic will not be pursued here. Recall that
coupling eects occur in composite laminates with arbitrary stacking sequen-
ces because B
ij
p 0. Symmetrical laminates are almost always used in practice
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
so as to insure that B
ij
= 0. Consequently, these couplings are rarely
encountered in practice. The complications due to out-of-plane coupling
eects present in general laminates will be ignored in this section. The eective
properties for symmetrical composite laminates are discussed in the following
subsections.
9.1 Effective Properties Relating Stress to Strain
The standard denitions of material properties used to relate stress to strain
were discussed in Sec. 3.2. Three types of properties are measured during
uniaxial tests: Youngs modulus, Poissons ratios, and coecients of mutual
inuence 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 Youngs modulus measured during a uniaxial
test is identical to that measured in pure bending. However, for composite
materials, the properties measured during uniaxial tests dier substantially
from those measured during pure bending tests. For example, Youngs mod-
ulus of a composite as measured during a uniaxial test is substantially dierent
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 exural properties.
Extensional Properties
Consider a symmetrical composite laminate subjected to uniaxial loading
N
xx
, 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 denition N
yy
=N
xy
=
M
yy
=M
yy
=M
xy
=0, Eq. (45) becomes (for symmetrical laminates):
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

a
11
a
12
a
12
0 0 0
a
12
a
22
a
26
0 0 0
a
16
a
26
a
66
0 0 0
0 0 0 d
11
d
12
d
16
0 0 0 d
12
d
22
d
26
0 0 0 d
16
d
26
d
66
_

_
_

_
N
xx
0
0
0
0
0
_

_
_

_
The midplane strains caused by uniaxial loading are therefore given by:
e
o
xx
a
11
N
xx
62a
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
e
o
yy
a
12
N
xx
62b
c
o
xy
a
16
N
xx
62c
Recalling that N
xx
is dened as a constant load/unit plate length (with units of
N/m or lbf/in.), the eective (or nominal) normal stress r
xx
applied to the
laminate is given by r
xx
=N
xx
/t, where t is the total laminate thickness.
In Sec. 3.2, Youngs modulus was dened as the normal stress r
xx
divided by the resulting normal strain e
xx
, with all other stress components
equal to zero.
Applying the standard denition to the laminate shown in Fig. 18, the
eective extensional Youngs modulus in the x-direction is given by:
E
ex
xx

r
xx
e
o
xx

N
xx
=t
a
11
N
xx


1
ta
11
63
The superscript ex is used to denote that this property is Youngs modulus
measured in extension.
In Sec. 3.2, Poissons ratio v
xy
was dened as the negative of the
transverse normal strain e
yy
divided by the axial normal strain e
xx
, both of
which are induced by stress r
xx
, with all other stresses equal to zero. The
eective Poissons ratio in extension for the laminate shown in Fig. 18 is given
by:
m
ex
xy

e
o
yy
e
o
xx

a
12
N
xx
a
11
N
xx

a
12
a
11
64
Figure 18 A symmetrical composite laminate subjected to uniaxial load N
xx
.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Once again, a superscript ex has been used to indicate that this Poissons
ratio is measured in extension.
The coecient of mutual inuence of the second kind g
xx,xy
was
dened as the shear strain c
xy
divided by the axial normal strain e
xx
, both
of which are induced by normal stress r
xx
, when all other stresses equal
zero. For a composite laminate, the eective coecient of mutual inuence
of the second kind g
ex
xx;xy
is therefore given by:
g
ex
xx;xy

c
o
xy
e
o
xx

a
16
N
xx
a
11
N
xx

a
16
a
11
65
An identical procedure can be employed to dene properties measured
during a uniaxial test in which only r
yy
=N
yy
/t is applied. In this case, Eq.
(45) becomes:
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

a
11
a
12
a
12
0 0 0
a
12
a
22
a
26
0 0 0
a
16
a
26
a
66
0 0 0
0 0 0 d
11
d
12
d
16
0 0 0 d
12
d
22
d
26
0 0 0 d
16
d
26
d
66
_

_
_

_
0
N
yy
0
0
0
0
_

_
_

_
Midplane strains induced are therefore:
e
o
xx
a
12
N
yy
66a
e
o
yy
a
22
N
yy
66b
c
o
xy
a
26
N
yy
66c
These strains can be used to dene the eective extensional Youngs mod-
ulus, E
ex
yy
, Poissons ratio, m
ex
yx
, and coecient of mutual inuence of the sec-
ond kind, g
yy,xy
ex
:
E
ex
yy

1
ta
22
67a
m
ex
yx

a
12
a
22
67b
g
ex
yy;xy

a
26
a
22
67c
Next, consider the eective material properties measured during a pure
shear test. Asymmetrical composite laminate subjected to pure shear loading
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
N
xy
is shown in Fig. 19. Because the load N
xy
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:
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

a
11
a
12
a
12
0 0 0
a
12
a
22
a
26
0 0 0
a
16
a
26
a
66
0 0 0
0 0 0 d
11
d
12
d
16
0 0 0 d
12
d
22
d
26
0 0 0 d
16
d
26
d
66
_

_
_

_
0
0
N
xy
0
0
0
_

_
_

_
Hence, the midplane strains caused by pure shear loading are given by:
e
o
xx
a
16
N
xy
68a
e
o
yy
a
26
N
xy
68b
c
o
xy
a
66
N
xy
68c
The eective (or nominal) shear stress s
xy
applied to the laminate is given by
s
xy
=N
xy
/t, where t is the total laminate thickness.
In Sec. 3.2, the shear modulus was dened as the shear stress s
xy
divided by the resulting shear strain c
xy
, with all other stress components
Figure 19 A symmetrical composite laminate subjected to shear load N
xy
.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
equal to zero. Applying this denition to the laminate shown in Fig. 19, the
eective shear modulus referenced to the xy coordinate axes is given by:
G
ex
xy

s
xy
c
o
xy

N
xy
=t
_ _
a
66
N
xy
_ _

1
ta
66
69
The coecient of mutual inuence of the rst kind g
xy,xx
(or g
xy,yy
) was
dened as the normal strain e
xx
(or e
yy
) divided by the shear strain c
xy
, both
of which are induced by shear stress s
xy
, when all other stresses equal zero.
For a composite laminate, the eective coecient of mutual inuence of the
rst kind g
ex
xy;xx
is therefore given by:
g
ex
xy;xx

e
o
xx
c
o
xy

a
16
N
xx
a
66
N
xy

a
16
a
66
70a
whereas g
ex
xy;yy
is given by:
g
ex
xy;yy

e
o
yy
c
o
xy

a
26
a
66
70b
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 (c
xz
j and c
yz
j ) are equal
to zero. Consequently, Chentsov coecients, which were dened in Sec. 3.2,
are always equal to zero for thin composite laminates.
The eective properties of a laminate obey the inverse relations fol-
lowed by any anisotropic plate, which are dened in Eq. (13) of Chap. 4. In
particular:
m
ex
xy
E
ex
xx

m
ex
yx
E
ex
yy
g
ex
xx;xy
E
ex
xx

g
ex
xy;xx
G
ex
xy
g
ex
yy;xy
E
ex
yy

g
ex
xy;yy
G
ex
xy
Cross-ply, balanced, and balanced angle-ply laminates were discussed in
Secs. 7.2, 7.3, and 7.4, respectively. Recall that a
16
=a
26
=0 for these types of
stacking sequences. Therefore, coecients of mutual inuence of the rst
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-
cussed in Sec. 7.4), it will be found that the eective in-plane moduli E
ex
xx
=
E
ex
yy
, g
ex
xy
=g
ex
yx
, and G
ex
xy
are related in the same manner as for an isotropic
plate:
G
ex
xy

E
ex
xx
2 1 m
ex
xy
_ _
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Flexural Properties
All of the eective 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 eective 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 eective exural properties. Eective exural properties do
not follow from the fundamental denitions of material properties reviewed
in Sec. 3.2. Rather, eective exural properties represent the response of a
structure subjected to bending, and are dened in analogy to those exhibited
by an isotropic plate. As discussed in Sec. 7, the bending stinesses (i.e., the D
ij
matrix) for an isotropic plate are given by:
D
11
D
22

Et
3
12 1 m
2

D
12
mD
11

mEt
3
12 1 m
2

D
66

D
11
D
12

2

Et
3
24 1 m
Also, for an isotropic plate, the inverse of the D
ij
matrix (i.e., the d
ij
matrix) is
given by:
d
11
d
22

1
D
11
1 m
2


12
Et
3
d
12
md
11

m
D
11
1 m
2


12m
Et
3
d
66
2 d
11
d
12

24 1 m
Et
3
Thus, for isotropic plates, both the D
ij
and d
ij
matrices can be calculated
directly frommaterial properties Eand v and the plate thickness t. Given these
results, one might anticipate that the D
ij
and d
ij
matrices for a composite
laminate could be calculated in a similar manner, using the total laminate
thickness t and eective extensional properties E
ex
xx
, E
ex
yy
, g
ex
xy
, and g
ex
yx
. This is
not the case. The bending stinesses of a composite laminate (i.e., the D
ij
matrix) cannot, in general, be calculated on the basis of eective extensional
properties. For example, in general:
D
11
p
E
ex
xx
t
3
12 1 m
ex
xy
_ _
2
_ _
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
for a composite laminate. In essence, this cannot be done because the exural
properties of a laminate are dictated, in part, by the laminate stacking
sequence. In contrast, the eective extensional properties are independent of
the stacking sequence. Hence, eective exural properties cannot be directly
related to eective extensional properties.
As stated above, eective exural properties are dened for a state of
pure bending. Note from the above discussion that for an isotropic plate:
d
11

12
Et
3
d
12
md
11
The eective Youngs moduli and Poisson ratios exhibited by a composite
laminate in exure, denoted E
xx

, E
yy

, g
xy

, and g
yx

, can therefore be dened as


follows:
E
fl
xx

12
d
11
t
3
E
fl
yy

12
d
22
t
3
m
fl
xy

d
12
d
11
m
fl
yx

d
12
d
22
Although the eective exural Youngs moduli and Poisson ratios dier from
the corresponding eective extensional properties, direct substitution will
shown that the exural properties obey the inverse relations:
m
fl
xy
E
fl
xx

m
fl
yx
E
fl
yy
It is also interesting to note that d
11
p d
22
for a quasi-isotropic laminate.
Consequently, E
fl
xx
p E
fl
yy
; m
fl
xy
p m
fl
yx
; even for a quasi-isotropic laminate.
The eective exural coecient of mutual inuence of the second kind
g
fl
xx;xy
can be dened as the midplane shear strain c
xx
j divided by the midplane
normal strain e
xx
j , both of which are induced by moment resultant M
xx
, when
all other stress resultants equal zero. For a composite laminate, the eective
exural coecient of mutual inuence of the second kind g
xx,xy

is therefore
given by:
g
fl
xx;xy

c
o
xy
e
o
xx

d
16
M
xx
d
11
M
xx

d
16
d
11
An identical procedure can be employed to dene the eective exural
coecient of mutual inuence of the second kind g
yy,xy

, a property measured
during a test in which only M
yy
is applied:
g
fl
yy;xy

c
o
xy
e
o
yy

d
26
M
yy
d
22
M
yy

d
26
d
22
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
It is mentioned in passing that exural coecients of mutual inuence of the
second kind are not encountered during study of isotropic plates because
d
16
=d
26
=0 for isotropic plates. The eective exural 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 coecients of thermal expansion are measured
by determining the strains induced by a uniform change in temperature, and
forming the following ratios:
a
xx

e
T
xx
DT
a
yy

e
T
yy
DT
a
xy

c
T
xy
DT
71
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 N
xx
N
yy
N
xy
M
xx
M
yy
M
xy
0):
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

a
11
a
12
a
12
0 0 0
a
12
a
22
a
26
0 0 0
a
16
a
26
a
66
0 0 0
0 0 0 d
11
d
12
d
16
0 0 0 d
12
d
22
d
26
0 0 0 d
16
d
26
d
66
_

_
_

_
N
T
xx
N
T
yy
N
T
xy
0
0
0
_

_
_

_
72
Substituting the midplane strains indicated by Eq. (72) into Eq. (71), the
eective linear thermal expansion coecients for a general laminate are:
a
xx

1
DT
a
11
N
T
xx
a
12
N
T
yy
a
16
N
T
xy
_ _
a
yy

1
DT
a
12
N
T
xx
a
22
N
T
yy
a
26
N
T
xy
_ _
73a
a
xy

1
DT
a
16
N
T
xx
a
26
N
T
yy
a
66
N
T
xy
_ _
It is noted in passing that these results are for a symmetrical laminate and can
therefore be inverted to give:
N
T
xx
DT A
11
a
xx
A
12
a
yy
A
16
a
xy
_
N
T
yy
DT A
12
a
xx
A
22
a
yy
A
26
a
xy
_
73b
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
N
T
xy
DT A
16
a
xx
A
26
a
yy
A
66
a
xy
_
The eective linear coecient of moisture expansion is measured by
determining the strains induced by a uniformchange in moisture content, and
forming the following ratios:
b
xx

e
M
xx
DM
b
yy

e
M
yy
DM
b
xy

c
M
xy
DM
74
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 calculatedaccording to
Eq. (45), which becomes (for a symmetrical laminate and for DT=N
xx
=N
yy
=N
xy
=M
xx
=M
yy
=M
xy
=0):
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

a
11
a
12
a
12
0 0 0
a
12
a
22
a
26
0 0 0
a
16
a
26
a
66
0 0 0
0 0 0 d
11
d
12
d
16
0 0 0 d
12
d
22
d
26
0 0 0 d
16
d
26
d
66
_

_
_

_
N
M
xx
N
M
yy
N
M
xy
0
0
0
_

_
_

_
75
Substituting the midplane strains indicated by Eq. (75) into Eq. (74), the
eective linear moisture expansion coecients for a general laminate are:
b
xx

1
DM
a
11
N
M
xx
a
12
N
M
yy
a
16
N
M
xy
_ _
b
yy

1
DM
a
12
N
M
xx
a
22
N
M
yy
a
26
N
M
xy
_ _
76a
b
xy

1
DM
a
16
N
M
xx
a
26
N
M
yy
a
66
N
M
xy
_ _
Because these results are for a symmetrical laminate, they can be inverted to
give:
N
M
xx
DM A
11
b
xx
A
12
b
yy
A
16
b
xy
_
N
M
yy
DM A
12
b
xx
A
22
b
yy
A
26
b
xy
_
76b
N
M
xy
DM A
16
b
xx
A
26
b
yy
A
66
b
xy
_
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Sample Problem 8
Determine the eective extensional and exural moduli, thermal expansion
coecients, and moisture expansion coecients 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, eective 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:
ABD
72:2 10
6
8:02 10
6
12:0 10
6
0 0 0
8:02 10
6
52:0 10
6
5:38 10
6
0 0 0
12:0 10
6
5:38 10
6
15:5 10
6
0 0 0
0 0 0 4:23 0:676 1:19
0 0 0 0:676 0:988 0:532
0 0 0 1:19 0:532 1:03
_

_
_

_
Because the laminate is symmetrical, all elements of the B
ij
matrix are zero, as
expected. We obtain the [abd] by inverting the [ABD] numerically:
abd
16:0 10
9
1:23 10
9
12:0 10
9
0 0 0
1:23 10
9
20:0 10
9
6:0 10
9
0 0 0
12:0 10
9
6:0 10
9
75:8 10
9
0 0 0
0 0 0 3:51 10
1
2:92 10
2
3:92 10
1
0 0 0 2:92 10
2
1:408 6:96 10
1
0 0 0 3:92 10
1
6:96 10
1
1:79
_

_
_

_
The eective extensional moduli of the laminate can now be calculated using
Eqs. (63)(70b):
E
ex
xx

1
ta
11

1
0:000750 16:0 10
9

83:3GPa
E
ex
yy

1
ta
22

1
0:000750 20 10
9

66:7GPa
G
ex
xy

1
ta
66

1
0:000750 75:8 10
9

17:6GPa
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
m
ex
xy

a
12
a
11

1:23 10
9
_ _
16:0 10
9
0:077
m
ex
yx

a
12
a
22

1:23 10
9
_ _
20 10
9
0:061
g
ex
xx;xy

a
16
a
11

12 10
9
16 10
9
0:75
g
ex
yy;xy

a
26
a
22

6:0 10
9
20 10
9
0:30
g
ex
xy;xx

a
16
a
66

12:0 10
9
75:8 10
9
0:16
g
ex
xy;yy

a
26
a
66

6:0 10
9
75:8 10
9
0:079
Eective exural properties are found to be:
E
fl
xx

12
d
11
t
3

12
0:351 0:000750
3
81:0GPa
E
fl
yy

12
d
22
t
3

12
1:408 0:000750
3
20:2GPa
m
fl
xy

d
12
d
11

2:92 10
2
0:351
0:083
m
fl
yx

d
12
d
22

2:92 10
2
1:408
0:021
g
fl
xx;xy

d
16
d
11

0:392
0:351
1:12
g
fl
yy;xy

d
26
d
22

0:696
1:408
0:49
Note that the values of extensional properties are quite dierent from anal-
ogous exural properties.
The thermal stress resultants associated with a given change in temper-
ature must be determined in order to calculate the eective thermal expansion
coecients. Numerically speaking, any change in temperature can be used,
but for present purposes, a unit change in temperature will be assumed (i.e.,
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
DT=1). Using (42a(42f), the thermal stress resultants associated with DT=1
are:
N
T
xx
j
DT1
52:3N=m N
T
yy
j
DT1
94:9N=m N
T
xy
j
DT1
36:9N=m
The eective thermal expansion coecient a
xx
can now be calculated in
accordance with Eq. (73a):
a
xx

1
DT
a
11
N
T
xx
a
12
N
T
yy
a
16
N
T
xy
_ _
a
xx

1
1
16:0 10
9
52:3 1:23 10
9
94:9
_
12:0 10
9
36:9
a
xx
1:16Am=m
B
C
Using an equivalent procedure:
a
yy
2:06Am=m
B
C a
xy
4:00Arad=
o
C
Finally, moisture stress resultants associated with a given change in moisture
content must be determined in order to calculate the eective moisture
expansion coecients. 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:
N
M
xx
j
DM1
32; 800N=m N
M
yy
j
DM1
33; 800N=m N
M
xy
j
DM1
930N=m
Applying Eqs. (76a) and (76b), we nd:
b
xx
494Am=m%M b
yy
643Am=m%M b
xy
667Arad%M
10 TRANSFORMATION OF THE ABD MATRIX
The A
ij
, B
ij
, and D
ij
matrices all involve a summation over the thickness of the
laminate in accordance with (Eq. (27a) (27b) and (34), repeated here for
convenience:
A
ij

n
k1
Q
ij
_ _
k
z
k
z
k1
repeated 27a
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
B
ij

1
2

n
k1
Q
ij
_ _
k
z
2
k
z
2
k1
_ _
repeated 27b
D
ij

1
3

n
k1
Q
ij
_ _
k
z
3
k
z
3
k1
_ _
repeated 34
As discussed in Sec. 5.2, Q
ij
may be calculated using Eq. (31) of Chap. 5, which
involves use of the Q
ij
matrix and various trigonometrical functions raised to
a power (e.g., cos
4
h, cos
2
h sin
2
h, sin
4
h, etc.). Alternatively, as discussed in
Sec. 5.3, Q
ij
may be calculated using Eq. (35) of Chap. 5, which involves the
use of material invariants U
i
Q
and trigonometrical functions, whose argu-
ments involve ber 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. Specically, 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 transformthe
ABD matrix from one coordinate system to another.
To aid in our development, dene the following geometry factors,
which are related to the ber angles and ply interface positions:
V
A
0

n
k 1
z
k
z
k1
t V
A
1

n
k 1
cos 2h
k
z
k
z
k1

V
A
2

n
k 1
sin 2h
k
z
k
z
k1
V
A
3

n
k 1
cos 4h
k
z
k
z
k1

V
A
4

n
k 1
sin 4h
k
z
k
z
k1
t
77
V
B
0

1
2

n
k 1
z
2
k
z
2
k1
_ _
0 V
B
1

1
2

n
k 1
cos 2h
k
z
2
k
z
2
k1
_ _
V
B
2

1
2

n
k 1
sin 2h
k
z
2
k
z
2
k1
_ _
V
B
3

1
2

n
k 1
cos 4h
k
z
2
k
z
2
k1
_ _
V
B
4

1
2

n
k 1
sin 4h
k
z
2
k
z
2
k1
_ _
78
V
D
0

1
3

n
k 1
z
3
k
z
3
k1
_ _

t
3
12
V
D
1

1
3

n
k 1
cos 2h
k
z
3
k
z
3
k1
_ _
V
D
2

1
3

n
k 1
sin 2h
k
z
3
k
z
3
k1
_ _
V
D
3

1
3

n
k 1
cos 4h
k
z
3
k
z
3
k1
_ _
79
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
V
D
4

1
3

n
k 1
sin 4h
k
z
3
k
z
3
k1
_ _
Next, consider the steps necessary to calculate the A
11
term. In this case, Eq.
(27a) becomes:
A
11

n
k 1
Q
ij
_ _
k
z
k
z
k1
80
Using the invariant formulation the expression for Q
11
listed in Eq. (35) of
Chap. 5 may be substituted, resulting in:
A
11

n
k 1
U
Q
1
U
Q
2
cos 2h U
Q
3
cos 4h
_ _
k
z
k
z
k1
81
where U
1
Q
, U
2
Q
, and U
3
Q
are the stiness invariants dened in Eq. (36) of
Chap. 5. If all plies within the laminate are composed of the same material,
then stiness invariants are constant for the laminate and only ber angle h
varies from one ply to the next. Therefore, Eq. (81) can be rewritten as:
A
11
U
Q
1

n
k 1
z
k
z
k1
U
Q
2

n
k 1
cos 2h
k
z
k
z
k1

82
U
Q
3

n
k 1
cos 4h
k
z
k
z
k1

The geometry factors V
0
A
, V
1
A
, and V
3
A
appear in Eq. (82), and hence the
equation may be written as:
A
11
U
Q
1
V
A
0
U
Q
2
V
A
1
U
Q
3
V
A
3
83a
Following an identical procedure, the remaining elements of the A
ij
matrix are
given by:
A
22
U
Q
1
V
A
0
U
Q
2
V
A
1
U
Q
3
V
A
3
83b
A
12
A
21
U
Q
4
V
A
0
U
Q
3
V
A
3
83c
A
66
U
Q
5
V
A
0
U
Q
3
V
A
3
83d
A
16
A
61

1
2
U
Q
2
V
A
2
U
Q
3
V
A
4
83e
A
26
A
62

1
2
U
Q
2
V
A
2
U
Q
3
V
A
4
83f
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Analogous procedures can be used to calculate members of the B
ij
and D
ij
matrices. It can be shown that elements of the B
ij
matrix are given by:
B
11
U
Q
1
V
B
0
U
Q
2
V
B
1
U
Q
3
V
B
3
84a
B
22
U
Q
1
V
B
0
U
Q
2
V
B
1
U
Q
3
V
B
3
84b
B
12
B
21
U
Q
4
V
B
0
U
Q
3
V
B
3
84c
B
66
U
Q
5
V
B
0
U
Q
3
V
B
3
84d
B
16
B
61

1
2
U
Q
2
V
B
2
U
Q
3
V
B
4
84e
B
26
B
62

1
2
U
Q
2
V
B
2
U
Q
3
V
B
4
84f
Members of the D
ij
matrix are given by:
D
11
U
Q
1
V
D
0
U
Q
2
V
D
1
U
Q
3
V
D
3
85a
D
22
U
Q
1
V
D
0
U
Q
2
V
D
1
U
Q
3
V
D
3
85b
D
12
D
21
U
Q
4
V
D
0
U
Q
3
V
D
3
85c
D
66
U
Q
5
V
D
0
U
Q
3
V
D
3
85d
D
16
D
61

1
2
U
Q
2
V
D
2
U
Q
V
D
4
85e
D
26
D
62

1
2
U
Q
2
V
D
2
U
Q
3
V
D
4
85f
Let us now consider transformation of the ABD matrix from one
coordinate system to another. A multiangle composite laminate referenced
to an xy 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 ber
angles referenced to the xy coordinate system. Now suppose that a dierent
coordinate system is of interest, the xVyV coordinate system, orientated h
degrees counterclockwise from the original xy coordinate system. The
transformed stiness matrices referenced to the xVyV coordinate system will
be labeled A
U
ij
, B
U
ij
, and D
U
ij
.
Inspection of Fig. 20 shows that a ply with ber angle h
k
relative to the
original x-axis will form an angle (h
k
/) relative to the xV-axis. Element A
11
may therefore be calculated using Eqs. (83a)(83f) by substituting (h
k
/) for
h
k
. Hence:
A
11
U
Q
1

n
k 1
z
k
z
k1
U
Q
2

n
k 1
cos 2 h
k
/ z
k
z
k1
86
U
Q
3

n
k 1
cos 4 h
k
/ z
k
z
k1

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Recalling the general trigonometrical identity:
cos a b cos a cos b sin a sin b
Equation (86) can be written as:
A
11
U
Q
1
V
A
0
U
Q
2
V
A
1
cos 2/ V
A
2
sin 2/
_ _
87a
U
Q
3
V
A
3
cos 4/ V
A
4
sin 4/
_ _
The geometry factors V
i
A
have been dened in Eq. (77). A similar procedure
can be applied to all remaining elements of the A
ij
matrix as well as the B
ij
and D
ij
matrices:
A
22
U
Q
1
V
A
0
U
Q
2
V
A
1
cos 2/ V
A
2
sin 2/
_ _
87b
U
Q
3
V
A
3
cos 4/ V
A
4
sin 4/
_ _
A
12
A
21
U
Q
4
V
A
0
U
Q
3
V
A
3
cos 4/ V
A
sin 4/
_ _
87c
A
66
U
Q
5
V
A
0
U
Q
3
V
A
3
cos 4/ V
A
4
sin 4/
_ _
87d
A
16
A
61

1
2
U
Q
2
V
A
2
cos 2/ V
A
1
sin 2/
_ _
87e
U
Q
3
V
A
4
cos 4/ V
A
3
sin 4/
_ _
A
26
A
62

1
2
U
Q
2
V
A
2
cos 2/ V
A
1
sin 2/
_ _
87f
U
Q
3
V
A
4
cos 4/ V
A
3
sin 4/
_ _
Figure 20 A general composite laminate, showing the original xy
coordinate system and the new xVyV coordinate system.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
B
11
U
Q
1
V
B
0
U
Q
2
V
B
1
cos 2/ V
B
2
sin 2/
_ _
88a
U
Q
3
V
B
3
cos 4/ V
B
4
sin 4/
_ _
B
22
U
Q
1
V
B
0
U
Q
2
V
B
1
cos 2/ V
B
2
sin 2/
_ _
88b
U
Q
3
V
B
3
cos 4/ V
B
4
sin 4/
_ _
B
12
B
21
U
Q
V
B
0
U
Q
3
V
B
3
cos 4/ V
B
4
sin 4/
_ _
88c
B
66
U
Q
5
V
B
U
Q
3
V
B
3
cos 4/ V
B
4
sin 4/
_ _
88d
B
16
B
61

1
2
U
Q
2
V
B
2
cos 2/ V
B
1
sin 2/
_ _
88e
U
Q
3
V
B
4
cos 4/ V
B
3
sin 4/
_ _
B
26
B
62

1
2
U
Q
2
V
B
2
cos 2/ V
B
1
sin 2/
_ _
88f
U
Q
3
V
B
4
cos 4/ V
B
3
sin 4/
_ _
D
11
U
Q
1
V
D
0
U
Q
2
V
D
1
cos 2/ V
D
2
sin 2/
_ _
89a
U
Q
3
V
D
3
cos 4/ V
D
4
sin 4/
_ _
D
22
U
Q
V
D
0
U
Q
2
V
D
1
cos 2/ V
D
2
sin 2/
_ _
89b
U
Q
3
V
D
3
cos 4/ V
D
4
sin 4/
_ _
D
12
D
21
U
Q
4
V
D
U
Q
3
V
D
3
cos 4/ V
D
4
sin 4/
_ _
89c
D
66
U
Q
5
V
D
0
U
Q
3
V
D
3
cos 4/ V
D
4
sin 4/
_ _
89d
D
16
D
61

1
2
U
Q
2
V
D
2
cos 2/ V
D
1
sin 2/
_ _
89e
U
Q
3
V
D
4
cos 4/ V
D
3
sin 4/
_ _
D
26
D
62

1
2
U
Q
2
V
D
2
cos 2/ V
D
1
sin 2/
_ _
89f
U
Q
3
V
D
4
cos 4/ V
D
3
sin 4/
_ _
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 ABDmatrix for a hybrid composite
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 dierent analysis paths. One analysis path corresponds to the case in
which the loads applied to the laminate are specied, whereas the second
corresponds to the case in which midplane strains and curvatures applied to
the laminate are specied. In either case, the user must provide various
numerical values required during the calculations performed. The user must
dene these values using a consistent set of units. For example, the user must
input elastic moduli, thermal expansion coecients, and moisture expansion
coecients 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:
E
11
170 10
9
Pa E
22
10 10
9
Pa m
12
0:30 G
12
13 10
9
Pa
a
11
0:9 10
6
m=m
B
C a
11
27:0 10
6
m=m
B
C
b
11
150:0 10
6
m=m %M b
22
4800 10
6
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 N
xx
=15010
3
N/m and M
xx
=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
ej
xx
=200010
6
m/m=0.002000 m/m and j
xx
=0.5 m
1
. 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).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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:
E
11
25:0 10
6
psi E
22
1:5 10
6
psi m
12
0:30 G
12
1:9 10
6
psi
a
11
0:5 10
6
in=in
B
F a
11
15 10
6
in=in
B
F
b
11
150:0 10
6
in=in %M b
22
4800 10
6
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 N
xx
=1000 lbf/in. and M
xx
=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 e
xx
j =2000 10
6
in./
in. =0.002000 in./in. andj
xx
=0.01 in
1
. All temperatures wouldbe inputted
in degrees Fahrenheit. Ply thicknesses must be inputted in inches. A typical
value would be t
k
=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:
e
xx
2000Ain:=in: e
yy
500Ain:=in: c
xy
1000Arad
Similarly, the strains measured by strain gage rosette 2 are:
e
xx
3000Ain:=in: e
yy
2000Ain:=in: c
xy
1000Arad
Determine by hand calculation:
(a) Midplane strains and curvatures induced in the laminate
(b) Strains (e
xx
,e
yy
,c
xy
) induced at the interface between plies 1
and 2.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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:
e
xx
500Ain:=in: e
yy
1000Ain:=in: c
xy
750Arad
Similarly, the strains measured by rosette 2 are:
e
xx
1000in:=in: e
yy
2000Ain:=in: c
xy
750Arad
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:
(a) Midplane strains and curvatures induced in the laminate
(b) Strains (e
xx
,e
yy
,c
xy
) induced at the interface between plies 1
and 2
(c) Stresses (r
xx
,r
yy
,s
xy
) induced in ply 1, at the interface between
plies 1 and 2
(d) Stresses (r
xx
,r
yy
,s
xy
) induced in ply 2, at the interface between
plies 1 and 2.
Figure 21 Edge view of the composite laminate described in Problem 1.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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:
e
xx
500Ain:=in: e
yy
1000Ain:=in: c
xy
750Arad
Similarly, the strains measured by rosette 2 are:
e
xx
1000Ain:=in: e
yy
2000Ain:=in: c
xy
750Arad
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:
(a) Midplane strains and curvatures induced in the laminate
(b) Strains (e
xx
,e
yy
,c
xy
) induced at the interface between plies 1
and 2
(c) Stresses (r
xx
,r
yy
,s
xy
) induced in ply 1, at the interface between
plies 1 and 2
(d) Stresses (r
xx
,r
yy
,s
xy
) 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 150jFin 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:
e
o
xx
1500Ain:=in:
e
o
yy
2200Ain:=in:
c
o
xy
1000Arad
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
j
xx
j
yy
j
xy
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):
(a) N
T
xx
(b) N
T
yy
(c) N
T
xy
(d) M
T
xx
(e) M
T
yy
(f) M
T
xy
:
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
1418. Assuming a ply thickness of 0.125 mmand using properties listed
in Table 3 of Chap. 3 (SI units), the [Q] matrix for ply 2 is:
10:76E10 2:606E10 4:813E10
2:606E10 2:722E10 2:152E10
4:813E10 2:152E10 3:605E10
_

_
_

_
The [ABD] matrix for this laminate is:
9:906E7 1:454E7 0 0 0 0
1:454E7 5:885E7 0 0 0 0
0 0 2:452E7 0 0 0
0 0 0 11:89 1:032 0:7521
0 0 0 1:032 1:628 0:3362
0 0 0 0:7521 0:3362 1:864
_

_
_

_
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
The [abd] matrix for this laminate is:
1:048E 8 0:2558E 8 0 0 0 0
0:2558E 8 1:763E 8 0 0 0 0
0 0 4:078E 8 0 0 0
0 0 0 9:029E 2 5:160E 2 2:713E 2
0 0 0 5:160E 2 0:6674 9:958E 2
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:
N
T
xx
8607N=m
N
T
yy
21; 825N=m
N
T
xy
M
T
xx
M
T
yy
M
T
xy
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:
N
M
xx
6092N=m
N
M
yy
6098N=m
N
M
xy
M
M
xx
M
M
yy
M
M
xy
0
14. A [0/F30/90]s graphite-epoxy laminate is cured at 175jC and then
cooled to roomtemperatures (20jC). Use hand calculation to determine
the following:
(a) Midplane strains and curvatures induced during cooldown
(b) Strains (e
xx
,e
yy
,c
xy
) induced in ply 2 during cooldown
(c) Stresses (r
xx
,r
yy
,s
xy
) induced in ply 2 during cooldown
(d) Stresses (r
11
,r
22
,s
12
) 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%
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
(b) Strains (e
xx
,e
yy
,c
xy
) induced in ply 2 after moisture content
increased to 0.5%
(c) Stresses (r
xx
,r
yy
,s
xy
) induced in ply 2 after moisture content
increased to 0.5%
(d) Stresses (r
11
,r
22
,s
12
) 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:
N
xx
N
yy
N
xy
10; 000N=m M
xx
M
yy
M
xy
0
Ignoring thermal and moisture eects, use hand calculation to determine
the following:
(a) Midplane strains and curvatures
(b) Strains (e
xx
, e
yy
, c
xy
) induced in ply 2
(c) Stresses (r
xx
, r
yy
, s
xy
) induced in ply 2
(d) Stresses (r
11
, r
22
, s
12
) induced in ply 2.
17. A[0/F30/90]
s
graphite-epoxy laminate is cured at 175jCand 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:
N
xx
N
yy
N
xy
10; 000N=m M
xx
M
yy
M
xy
0
Use hand calculation to determine the following:
(a) Midplane strains and curvatures
(b) Strains (e
xx
,e
yy
,c
xy
) induced in ply 2
(c) Stresses (r
xx
,r
yy
,s
xy
) induced in ply 2
(d) Stresses (r
11
,r
22
,s
12
) induced in ply 2.
18. A10 10 in.
2
[0/F30/90]
s
graphite-epoxy laminate is supported between
three innitely rigid walls and rollers, as shown in Fig. 22. A load
N
xx
=7500 N/m is applied to the plate. Ignoring thermal eects,
moisture eects, and the possibility of buckling, use hand calculation
to determine N
xx
, N
yy
, N
xy
, e
xx
j , e
yy
j , and c
xx
j .
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:
N
xx
30kN=m N
yy
7kN=m N
xy
0
N
xx
10Nm=m M
yy
M
xy
0
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 e
xx
, e
yy
, and c
xy
.
(c) Prepare plots similar to Fig. 14, showing the through-thickness
variation of strains e
11
, e
22
, and c
12
.
(d) Prepare plots similar to Fig. 15, showing the through-thickness
variation of stresses r
xx
, r
yy
, and s
xy
.
(e) Prepare plots similar to Fig. 16, showing the through-thickness
variation of stresses r
11
, r
22
, and s
12
.
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
roomtemperatures (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:
N
xx
30kN=m N
yy
7kN=m N
xy
0
N
xx
10Nm=m M
yy
M
xy
0
Moisture content remains constant at 0%. Using properties listed in
Table 3 of Chap. 3 and assuming that the graphite-epoxy, glass-epoxy,
Figure 22 A [0/F30/90]
s
graphite-epoxy laminate described in Problem 18.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 e
xx
, e
yy
, and c
xy
.
(c) Prepare plots similar to Fig. 14, showing the through-thickness
variation of strains e
11
, e
22
, and c
12
.
(d) Prepare plots similar to Fig. 15, showing the through-thickness
variation of stresses r
xx
, r
yy
, and s
xy
.
(e) Prepare plots similar to Fig. 16, showing the through-thickness
variation of stresses r
11
, r
22
, and s
12
.
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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 eorts 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 eort. 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 diculties 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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
1 PRELIMINARY DISCUSSION
1.1 Yielding and Fracture of Isotropic Metals vs.
Unidirectional Polymeric Composites
Let us begin our discussion of the diculties 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 ber and the polymeric matrix),
and the atomic and/or molecular structure of these two constituents diers
substantially. Furthermore, the molecular structure of the polymeric matrix
in the immediate vicinity of the ber/matrix interface usually diers from the
molecular structure at positions removed from the ber. That is, the molec-
ular structure developed during polymerization of a polymer in bulk diers
from that developed in the region immediately adjacent to the surface of the
ber. The region near the ber is often referred to as the interphase, rather
than the interface, and the mechanical properties exhibited by the polymer
in the interphase dier from bulk properties. Hence, from a continuum-
mechanics point of view, three more or less distinct materials can be
dened: the ber, the matrix, and the bermatrix interphase.
Consider the case of graphiteepoxy. As discussed in Sec. 3 of Chap. 1,
graphite is a highly ordered crystalline structure, and the high strength and the
high stiness exhibited by a graphite ber at the macroscale are achieved by
aligning the basal planes of the graphite crystal with the axis of the ber. The
macroscopical stressstrain response of an individual graphite ber is almost
perfectly linear up to nal 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 stressstrain 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 stressstrain 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 stressstrain response of the interphase is
dicult to measure (or even dene), but is likely somewhere between that of
the ber and matrix. The stress necessary to cause dislocations or other
imperfections to develop or move within the crystalline lattice of the graphite
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
ber are far higher than those necessary to cause molecular segments to slide
in the epoxy matrix. The initial nonlinear deformations exhibited by a graph-
iteepoxy 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 formmicrocracks, often located (at least initially) at
grain boundaries. As loading is further increased (and/or if loading uctuates
with time, as in fatigue loading), these microcracks growin 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 nal 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
ber or the epoxy matrix, the imperfections that lead to the formation of a
crack in these two mediums are wholly dierent. In addition, the stress levels
necessary to cause crack growth in a graphite ber 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 ber. The orientation of matrix crack(s) that forms in a unidirectional
laminate is invariably related to the ber direction (e.g., cracks form either
parallel or perpendicular to the bers) and is independent of the principal
stress and principal stress coordinate system when dened 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 bermatrix interface. Matrix cracks gen-
erally occur in planes either parallel or perpendicular to the ber
direction.

Fibermatrix debonding. In this case, the crack has formed in the


interphase region, and a (nonplanar) crack extends around the peri-
phery of the ber.

Fiber cracks. These are cracks that occur in the ber itself. Fiber
cracks almost always occur in a plane perpendicular to the axis of the
ber, and extend across the entire width of the ber.
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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 bers 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, polymermatrix 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 eects 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 eects are not pronounced) with an acceptable degree of accuracy.
Predicting the failure of multiangle composite laminates is far more dicult
than for unidirectional composites, however. One diculty is that a single
macroscopical crack or aw 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 signicance 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].
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
contrast to the case of a unidirectional composite (e.g., a [90]
n
laminate), in
which the nal fracture is ultimately governed by a single (or, at most, a few)
crack or aw. 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 nal fracture of a
multiangle laminate does not occur until hundreds or even thousands of
cracks and other aws have formed.
Asecond 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 diering Poisson ratios or coecients of
mutual inuence.
Thirdly, pre-existing thermal and/or moisture stresses occur in multi-
angle laminates due to a mismatch in eective thermal expansion and
moisture expansion coecients 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 nal fracture of the laminate
as a whole.
The many diculties 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 N
xx
(or, equivalently, a tensile eective stress r
xx
N
xx
=t) is applied and steadily
increased until the nal laminate fracture occurs.
An axial strain e
xx
is, of course, induced as r
xx
is increased from zero.
The initial slope of the r
xx
vs: e
xx
represents the eective Youngs modulus of
the laminate,

E
xx
, and can be predicted using classical lamination theory
(CLT), as discussed earlier. As r
xx
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 formin one
(or more) ply. For a [0/45/90/45]
s
laminate subjected to uniaxial tensile
loading, the rst plies to yield and then crack are almost always the 90j plies;
that is, the 90j plies yield when r
xx
N
xx
=t has been increased to a critical
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
level. The dierence 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
eective stress level necessary to cause yielding of the 90j plies is called the
rst-ply failure stress. Hence, cracks begin to form in the 90j plies at load
levels above the rst-ply failure stress. Note that if the laminate were a
unidirectional [90]
n
laminate, then nal 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
Figure 1 An idealized stressstrain plot for a [0/45/90/45]
s
laminate,
showing the evolution of internal damage.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
the 90j plies can be predicted on the basis of so-called shear lag models. A
number of researchers have developed shear lag models. Adetailed discussion
of such models will not be presented here, but the interested reader should
refer to Refs. 25 for representative example analyses.
Once the 90j plies have cracked/fractured, they can no longer contrib-
ute fully to the eective stiness of the laminate. Hence, it would be expected
that the slope of the r
xx
vs: e
xx
curve (i.e., the eective Youngs modulus E
xx
)
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 gure).* The failure of
the 90j plies has little impact on the eective Youngs modulus because the
stiness of these plies in the x-direction, relative to the stiness of the F45j
and 0j plies, is very low even before failure occurs.
As the eective stress level is further increased, cracks eventually begin
to formwithin the F45j plies. Once again, many cracks formwithin 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
signicant damage has yet occurred within the 0j plies. Hence, the laminate as
a whole can support still higher eective stress levels. An additional decrease
in slope (i.e., a decrease in apparent Youngs 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 xy plane. As the eective 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 xy 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 eective 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 xy plane.
The nal laminate fracture is precipitated by ber failures within the 0j
plies. The eective stress level at which nal fracture occurs is often called the
last-ply failure stress. At the nal 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 eective Poisson ratio v
xy
occurs due to transverse cracking of the 90j plies [4].
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
delamination that occurred at lower stress levels, as well as the large amount
of energy release associated with ber failure.
Two other forms of damage encountered in composites are bermatrix
debonding and ber microbuckling. In the case of bermatrix debonding, a
crack forms around the periphery of a ber, so that the load can no longer be
transferred fromthe matrix to the ber. In the case of ber microbuckling, the
bers within a ply that experiences compressive stresses in the ber direction
buckle. This reduces the compressive stiness exhibited by the ply and
ultimately leads to failure of the bers 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 signicant internal damage at load levels far belowthat required to cause
nal fracture. It should now be clear that damage refers to many failure
events including matrix cracks, delaminations, bermatrix debonding, ber
microbuckling, ber cracks, etc. Hence, a fundamental question arises during
the design of composite structures: What is failure and what is a safe
eective stress level? There is no single answer to this question, and the
denition 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. Astructure 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 eective 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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
a pressurized gas (the body of a re extinguisher, perhaps) to that of a pres-
surized airplane fuselage. In the case of a re 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 dicult 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 dierent
loading conditions during takeo, 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 dicult 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 dened. First, the laminate failure stress may be
dened as the eective stress level required to initiate a ply failure of any kind.
Ply failure is meant to imply either yielding or ber fractures in one or more
plies. The laminate failure stress or load dened in this manner is called the
rst-ply failure stress or rst ply failure load. This is a conservative design
approach (in some cases overly conservative) because for most multiangle
laminates, it will be found that rst-ply failure is initiated due to yielding of
one or more plies rather than ber 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 ber
fractures. Hence, the ultimate load-carrying capacity of most multiangle
laminates is well above the rst-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 rst-ply failure approach may be appropriate.
The second limiting philosophy is to dene the laminate failure stress as
the eective stress level that causes nal catastrophic fracture; in this case, the
laminate failure stress is dened as the last-ply failure stress. Obviously, this
latter denition 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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 N
xx
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-
ned 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 ber angle is assumed to vary from one ply to the next for the
laminate shown in Fig. 2. In other words, the eective material properties of
each ply (in particular, the eective Poissons ratio v
xy
and the eective
coecient of mutual inuence of the second kind g
xx,xy
) varies fromone 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 (r
xx
, r
yy
, and s
xy
)
k
. Hence, even for a
Figure 2 A thin multiangle (and/or multimaterial) composite laminate subjected
to a uniaxial load N
xx
.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
simple tensile specimen, a uniaxial load in the x-direction (N
xx
) is predicted to
cause stresses r
yy
and s
xy
, as well as stress r
xx
.
Referring again to Fig. 2, note that although it may be possible for r
yy
and s
xy
to exist in the central regions of the laminate, neither r
yy
nor s
xy
can
possibly exist at the free edge because neither N
yy
nor N
xy
is applied to the
laminate. Therefore, even if r
yy
and s
xy
are induced in the central regions of
the laminate, as y !Fb, it must be that r
yy
!0 and s
xy
!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 N
xx
(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 diers from that used
in ply 2. Furthermore, assume that E
1
11
E
2
11
; E
1
22
E
2
22
; G
1
12
G
2
12
, but
that v
1
12
> v
2
12
. In this hypothetical case then, a mismatch in Poissons ratio
exists between plies, but all other material properties are exactly identical. In
addition, the coecient of mutual inuence of the second kind equals zero
for all three laminates because the 12 axes are coincident with the xy axis for
all three plies. We assume that the thickness of each ply is identical (say,
t
1
=t
2
=t
3
=t
p
), so that the total laminate thickness is t =3t
p
.
Based strictly on physical reasoning, it is clear that all plies will
experience a tensile axial stress and strain (r
xx
and
xx
) because the overall
loading applied to the laminate (N
xx
) has been assumed to be tensile. If the
three plies are rmly 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 eect. 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 Poissons ratio of the material used in these plies is
greater than of ply 2: v
1
12
> v
2
12
. 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: |e
yy
|
plies 1 and 3
>|e
yy
|
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 byanequal amount: je
yy
j
plies 1 and 3
je
yy
j
ply 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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
that they would otherwise, and hence it is expected that a transverse tensile
stress will be induced in plies 1 and 3.
CLT calculations conrm 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:
r
xx
j
plies 1 and 3

3E
11
E
22
v
2
12
2v
2
12
v
1
12
h i n o
N
xx
t
p
9E
11
E
22
v
1
12
n o
2 v
2
12
n o h i
2

1a
r
yy
j
plies 1 and 3

h
v
1
12
v
2
12
i
E
22
N
xx
t
p
9E
11
E
22
v
1
12
n o
2 v
2
12
n o h i
2

1b
s
xy
j
plies 1 and 3
0 1c
r
xx
j
ply 2

3E
11
E
22
v
1
12
2v
2
12
v
1
12
h i n o
N
xx
t
p
9E
11
E
22
v
1
12
n o
2 v
2
12
n o h i
2

2a
r
yy
j
ply 2

2 v
1
12
v
2
12
h i
E
22
N
xx
t
p
9E
11
E
22
v
1
12
n o
2 v
2
12
n o h i
2

2r
yy
j
plies 1 and 3
2b
s
xy
j
ply 2
0 2c
It is interesting to note that Eqs. (1a) and (2a) show that the axial stress
r
xx
induced in plies 1 and 3 is not precisely equal to that induced in ply 2
because v
1
12
p v
2
12
. Of greater immediate interest, however, is that Eqs. (1b)
and (2b) predict that a transverse stress r
yy
is induced in all three plies, again
because v
1
12
p v
2
12
. Hence, a load in the x-direction (N
xx
) 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 Poissons ratio. That is, if
v
1
12
v
2
12
then from Eqs. (1b) and (2b), the transverse stress in all three plies
becomes r
yy
=0.
* Because this is a relatively simple laminate, Eqs. (1) and (2) can be conrmed by hand
calculations.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
The physical reasoning described above is simple and compelling;
furthermore, the expected algebraic sign of the transverse stresses expected
in each ply is conrmed by CLT calculations, lending credence to the CLT
approach. Nevertheless, as pointed out earlier, a transverse stress r
yy
cannot
possibly exist at the free edge of the laminate because no transverse load N
yy
is
applied to the laminate.
Afree-body diagramof an individual ply can nowbe used to understand
why free-edge stresses develop and are three-dimensional. A cross-section
within the yz 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 F
yy
(associated with
stress r
yy
) 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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
equilibrating force F
yy
cannot be present on the right side of the free-body
diagrambecause no load is applied to the free edge of the laminate. Therefore,
a shear force V
zy
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 equilibriumin the
y-direction (SF
y
=0). The fact that a shear force V
zy
is present implies, of
course, that a shear stress s
yz
= s
zy
is present near the free edge. Note, how-
ever, that it must be that V
zy
!0 as y !Fb because no shear force can exist at
the free edge; V
zy
can only be nonzero at interior regions of ply 3. Also, the
normal force acting on the left side F
yy
is acting through the centroid of the
free-body diagram, whereas the shear force on the right side V
zy
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 SM
x
= 0) is if a force acting normal to the xy plane ( F
zz
) also
develops. The fact that a normal force F
zz
is present implies, of course, that a
normal stress r
zz
is present. Numerical and analytical solutions (discussed
below) indicate that F
zz
(and hence r
zz
) 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 N
xx
, internal force F
zz
(or, equiva-
lently, stress r
zz
) must be self-equilibrating: m
b
y 0
r
zz
dxdy 0. This implies
that r
zz
must undergo a change in algebraic sign at regions near the free edge.
If, for example, r
zz
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 r
zz
and s
zy
are
caused by a uniaxial tensile loading N
xx
. The magnitude and distribution of
r
zz
and s
zy
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 coecient of mutual inuence of
the second kind g
xx,xy
=0, and therefore according to CLT, a uniaxial load
N
xx
does not cause an in-plane shear stress (s
xy
) to be induced at interior
regions in any ply. However, for more general laminates, g
xx,xy
p 0, and a
uniaxial loading N
xx
will cause both r
yy
and s
xy
at interior regions of the
laminate. In these cases, the state of stress near a free edge involves all six
components of stress (r
xx
, r
yy
, r
zz
, s
yz
, s
xz
, and s
xy
). 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 inuence on the delamination failure of
multiangle composite laminates, and have therefore been the topic of a great
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
many research papers. The existence of free-edge strains in multiangle
composite laminates has been well documented experimentally [58]. Meth-
ods used to predict free-edge stresses may be roughly grouped into numerical
solutions [814] and approximate analytical solutions [1521]. With the
exception of Ref. 9 (summarized below), all of the numerical solutions
referenced here were based on the nite-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 rst numerical study of free-edge
stresses in 1970. Their analysis was based on the method of nite dierences.
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 e
xx
. The following
properties, which were typical of transversely isotropic graphiteepoxy
material systems at the time, were assumed in their study:
E
11
138GPa20Msi
E
22
E
33
14:5GPa2:1Msi
G
12
G
23
G
13
5:86GPa 0:85Msi
v
12
v
23
v
31
0:21
The thickness of each ply was denoted h
o
. In this text, the total laminate
thickness has been denoted t; thus, in the Pipes and Pagano study, the total
laminate thickness was t = 4h
o
. The width of the laminate was denoted 2b.
Laminates with the three dierent 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/h
o
=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 =h

o
(i.e., just within the +45j ply), is presented in Fig. 5.
In this gure, 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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
the 45j plies in a [45/45]
s
laminate subjected to a uniformaxial strain e
xx
are
as follows:
r
xx
e
xx
20:4GPa2:96Msi
r
yy
0
s
xy
e
xx
7:93GPa1:15Msi
Figure 5 Ply stresses calculated for a graphiteepoxy [45/45]
s
laminate, at
the through-thickness position z = h
0
+
. (From Ref. 9, with permission from
Sage Publications Ltd.)
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Because CLT is based on the assumption of plane stress, for a CLT
calculation, the remaining stress components are zero by assumption:
r
zz
=s
xz
=s
yz
=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 nite-dierence 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 r


xx
/e
xx
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 nite value of about 15.9 GPa (2.3 Msi) at the free edge.

A normalized transverse stress r


yy
/e
xx
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 r


zz
/e
xx
is predicted to occur in
regions very near the free edge; r
zz
/e
xx
is essentially zero for y/
b<0.70. As y/b !1.0, r
zz
/e
xx
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 s
xy
/e
xx
increases very slightly fromthe 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 s


xz
/e
zz
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 s
xz
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 s


yz
/e
zz
is almost
nonexistent throughout the entire width of the laminate. Values of
s
yz
/e
zz
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 specic case of a [45/45]
s
graphite
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
epoxy laminate subjected to a remote uniform axial strain e
xx
, based on the
ply properties listed above. Dierent stress distributions would be predicted if
a dierent stacking sequence were involved, if the laminate were subject to
dierent loading conditions, and/or if dierent 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 s
xz
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 nite-element technique. Several numerical analyses devoted to free-edge
stresses and based on the use of the nite-element technique appeared in the
literature following the Pipes and Pagano study (Refs. 1113) are representa-
tive examples. Although predictions based on nite-element methods were in
general agreement with the original nite-dierence study by Pipes and
Pagano, there was growing consensus that a singular stress eld exists at
the free edge [1618]. That is, as nite-element analyses became more rened,
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 r
zz
and s
xz
) did
not converge to a stable value as mesh densities were increased near the free
edge is indicative of a singular stress eld, 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 conrmed the
singular nature of free-edge stress elds 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. 19,20,2325)
have been based on the use of Lekhnitskis 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:
r
ij
fy; zr
d

f y; z
r
d
3
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
where f ( y,z) represents a generalized function that depends on laminate
stacking sequence, material properties, and applied loading, and r represents
the distance fromthe 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 dierence in ber 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/r
d
!land,
in accordance with Eq. (3), the stresses r
ij
are predicted to be innitely 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 dierence in ber 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 nite-dierence analysis (listed above). A plot of d as a function
of ber 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 ber 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 s
xz
near the free edge, but substantially alters the predicted distribution of the
out-of-plane normal stress r
zz
. In fact, in the original Pipes and Pagano nite-
dierence solution, r
zz
was predicted to be tensile at the free edge, whereas in
the Wang and Choi analysis, r
zz
was predicted to be compressive. In both
analyses, the shear stress s
xz
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 r
zz
is generally less than s
xz
(except at the
free edge, where both are singular) and exists over a narrower region, r
zz
is
believed to play a central role in delamination failure. If r
zz
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 r
zz
is compressive,
it will tend to hold plies together, inhibiting delamination.
Once the singular nature of free-edge stresses had been claried on the
basis of analytical studies, it became possible to dene special singular
elements for use in nite-element studies of free-edge stresses. The stress
singularity represented by Eq. (3) is embedded within the shape function used
to dene these singular elements, following an approach suggested by Stern
[27]. For example, Nailadi et al. [15] used singular elements in a nite-element
study of free-edge stresses, and compared their predictions to displacements
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
measured at the free-edge using Moire interferometry. These authors also
considered a [45/45]
s
graphiteepoxy laminate subjected to axial tensile
strain, although the material properties used in this study diered slightly
from those in the earlier analyses of Pipes and Pagano and Wang and Choi.
They concluded that stress component r
zz
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 r
zz
is particularly
sensitive to inclusion of the stress singularity in the problem formulation.
Figure 6 Strength of the free-edge stress singularity in [Fh]
s
graphiteepoxy
laminates. (From Ref. 20, with permission from ASME.)
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Three important points should be emphasized. First, the distribution of
free-edge stresses depends on material property and stacking sequence.
Because the number of dierent combinations of materials and stacking
sequences is innitely 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 linearelastic material behavior and,
furthermore, all analyses assume that a distinct interface exists between plies.
Neither of these assumptions is rigorously valid. Adistinct 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 nite. 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, ber microbuckling, etc.) occurs. It is the integrated
eect of all of these factors that governs the nal 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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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. CLTcan be used to predict the rst-ply
failure of a multiangle laminate with a reasonable degree of accuracy. Recall
that, in this text, rst-ply failure is meant to imply the initiation of either
yielding or ber 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 rst-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 rst-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 rst-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 rst ply failure stress for the specic 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 dierent stacking sequences and/or to more general thermo-
mechanical loading conditions. Although several dierent solution paths may
be followed, one approach is illustrated in Fig. 7. The ow diagram shown in
this gure 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 (N
xx
, N
yy
, N
xy
, M
xx
, M
yy
, M
xy
). It is
further assumed that the relative magnitude of each load component is
known. That is, if N
xx
=1, then it is assumed that the magnitudes of the
remaining loads can be related to N
xx
(i.e., N
yy
=k
1
|N
xx
|, N
xy
=k
2
|N
xx
|,
M
xx
=k
3
|N
xx
|, M
yy
=k
4
|N
xx
|, M
xy
=k
5
|N
xx
|). The relative magnitudes of each
load component (i.e., constants k
i
) are referred to as unit loads in Fig. 7.
The calculation steps shown in Fig. 7 can be summarized as follows:

Step 1. The rst step is to dene the problem. Ply properties,


laminate stacking sequence, unit loads applied to the laminate, stress-
free temperature (often assumed to be the cure temperature), service
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Figure 7 A flow diagram illustrating calculation of the first-ply failure load
using CLT.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
temperature, moisture content, and the failure criterion to be
employed must all be specied.

Step 2. The [ABD] matrix for the laminate is determined. The re-
duced stiness matrix Q
ij
for each material used in the laminate is
calculated (using Eq. (11) of Chap. 5), and then the transformed
reduced stiness matrix Q
ij
for each ply is determined (using Eq. (31)
of Chap. 5). This allows calculation of the A
ij
, B
ij
, and D
ij
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 specied DTand DMare 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 xy 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 xy coordinate system using Eq. (12) of Chap. 6, and
calculation of the ply stresses in the xy 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, TsaiHill, or TsaiWu failure cri-
terion), then ply stresses are rotated to the local 12 coordinate
system dened 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 12 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 identied. The unit loads are then increased,
and steps 4 and 5 are repeated iteratively, until the rst 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 rst-ply failure condition has
been identied.
The use of CLTto predict rst-ply failure according to the owdiagram
shown in Fig. 7 is illustrated in the following Example Problem.
Example Problem 1
Predict the rst-ply failure load for a [0/30/60]
s
graphiteepoxy laminate
subjected to a uniaxial tensile load N
xx
as well as a change in temperature
corresponding to cooldown fromcure temperatures (175jC) to roomtemper-
atures (20jC). Assume no change in moisture content, and that the laminate is
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
stress-free at the cure temperature.* Base the prediction on the maximum
stress failure criterion, use properties for graphiteepoxy 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:
N
yy
N
xy
M
xx
M
yy
M
xy
DM 0
The following properties for graphiteepoxy are taken from Table 3 of
Chap. 3:
E
11
170GPa E
22
10GPa v
12
0:30 G
12
13GPa
r
f T
11
1500MPa r
yT
22
50MPa s
y
12
75MPa
r
fC
11
1200MPa r
yC
22
100MPa
a
11
0:9Am=m
B
C a
22
27Am=m
B
C
Employing CLTcalculations 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:
r
11
55:54MPa r
22
28:36MPa s
12
22:83MPa
For the 30j plies (plies 2 and 5), a DT=155jC will cause:
r
11
29:59MPa r
22
24:79MPa s
12
0MPa
For the 60j plies (plies 3 and 4), a DT=155jC will cause:
r
11
55:54MPa r
22
28:36MPa s
12
22:83MPa
The additional stresses caused by the external load N
xx
are superimposed with
the initial thermal strains. A CLT analysis gives the following predicted
stresses:
For the 0j plies (plies 1 and 6):
r
11
2:750 10
3
N
xx
55:54MPa
r
22
0:5193 10
2
N
xx
28:36MPa
s
12
0:1748 10
3
N
xx
22:83MPa
For the 30j plies (plies 2 and 5):
r
11
1:112 10
3
N
xx
29:59MPa
r
22
0:1206 10
3
N
xx
24:79MPa
s
12
0:2646 10
3
N
xx
* 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 nal cure temperature denes the
stress-free temperature.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
For the 60j plies (plies 3 and 4):
r
11
0:2098 10
3
N
xx
55:54MPa
r
22
0:1760 10
3
N
xx
28:36MPa
s
12
0:8985 10
2
N
xx
22:83MPa
These results can be used to determine the stresses induced by the
combined eects of DT = 155jC and any value of N
xx
, assuming a linear
elastic response. For example, if N
xx
= 100 kN/m, then the stress induced in
the ber direction in plies 1 and 6 is: r
11
= (2.75010
3
)(10010
3
)55.54
MPa=219.5 MPa. Because the maximum stress failure criterion is used in
this sample problem, the value of N
xx
necessary to cause stresses to reach
critical levels in any ply can be determined directly using the specied failure
stresses:
For the 0j plies (plies 1 and 6):
For ber failure : N
fail
xx

1500 10
6
55:54 10
6
2:75 10
3
565:7 10
3
N=m
For matrix failure : N
fail
xx

50 10
6
28:36 10
6
0:5193 10
2
416:7 10
3
N=m
For shear failure : N
fail
xx

75 10
6
22:83 10
6
0:1748 10
3
559:7 10
3
N=m
For the 30j plies (plies 2 and 5):
For ber failure : N
fail
xx

1500 10
6
29:59 10
6
1:112 10
3
1322 10
3
N=m
For matrix failure : N
fail
xx

50 10
6
24:79 10
6
0:1206 10
3
209:0 10
3
N=m
For shear failure : N
fail
xx

75 10
6
0:2646 10
3
283:4 10
3
N=m
For the 60j plies (plies 3 and 4):
For ber failure : N
fail
xx

1200 10
6
55:54 10
6
0:2098 10
3
5455 10
3
N=m
For matrix failure : N
fail
xx

50 10
6
28:36 10
6
0:1760 10
3
123:0 10
3
N=m
For shear failure : N
fail
xx

75 10
6
22:83 10
6
0:8985 10
2
580:6 10
3
N=m

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.


The rst-ply failure load equals the lowest value of N
xx
that causes the stress in
any ply to reach a critical level. Examination of these results shows that rst-
ply failure is predicted to occur when N
xx
=123.0 kN/m due to a matrix
yielding failure in the 60j plies (plies 3 and 4). Equivalently, the eective stress
at rst-ply failure is:
r
xx

N
xx
t

123:0kN=m
60:125mm
164MPac24ksi:
3.2 Predicting Last-Ply Failure
Last-ply failure calculated using CLTis based on the so-called ply-discount
scheme. A ow diagram showing typical steps in the analysis is presented in
Fig. 8. The process begins with calculations similar to those described for a
rst-ply failure analysis. The rst step is to dene the problemto 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 rst-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 suered a matrix crack;
therefore, the stinesses of a failed ply are reduced in a discrete step. As an
example of typical values, the stinesses of a failed ply could be related to the
intact values according to:
E
failed
11
E
11
v
failed
12
v
12
E
failed
22
0:3E
22
G
failed
12
0:3G
12
In this example, the ber-dominated properties of the failed ply (E
11
failed
and
v
12
failed
) are not reduced at all from those of an intact ply. In contrast, the
matrix-dominated properties (E
22
failed
and G
12
failed
) are reduced to 30% of those
of an intact ply. These estimates reect the fact that matrix fractures occur
well before ber fractures and, in fact, a signicant number of ber fractures
do not ordinarily occur until the last-ply failure load is reached. Matrix-
dominated stinesses are not reduced to zero because even a cracked ply will
contribute to some extent to the overall laminate stiness.
Although a discrete reduction in ply stiness has been the traditional
approach, it is by no means clear that this technique can be used to model the
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Figure 8 A flow diagram illustrating calculation of the last-ply failure load using
CLT.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 signicant 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, rst-ply
failure is dened as the load at which ply yielding rst occurs, and the stiness
of yielded plies should be reduced in a smooth manner to represent this
phenomenon, rather than by a discrete reduction in ply stiness.
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
signicantly in error. As already discussed, signicant errors in prediction
may occur at load levels beyond the rst-ply failure load because the ductile
nature of modern composite material systems is often not properly modeled.
Asecond 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 Kirchho
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 Kirchho hypothesis upon which CLTis based.
Once signicant 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 viewsuch analyses with a healthy level
of skepticism.
4 LAMINATE FIRST-PLY FAILURE ENVELOPES
Methods usedtodetermine the rst-ply failure loadfor a laminate subjectedto
a specic single combination of stress and moment resultants were discussed
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
in Sec. 1 of Chap. 3. Obviously, there are an innite number of combinations
of stress and moment resultants that will cause rst-ply failure. Conceptually,
those combinations of stress and moment resultants that (collectively) cause
rst-ply failure dene a laminate failure surface. It is dicult 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
N
xx
and N
yy
that cause rst-ply failure can be calculated using the approach
described in Sec. 1 of Chap. 3, while assuming that the remaining four
resultants are zero (N
yy
=M
xx
=M
yy
=M
xy
=0). A plot of the (N
xx
,N
yy
)
combinations calculated under these assumptions is called a rst-ply failure
envelope. A typical rst-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 gure. In one case, thermal and moisture
stresses and strains are ignored (i.e., this curve was generatedby 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
graphiteepoxy
laminate based on the maximum stress failure criterion and combinations of
N
xx
and N
yy
(N
xy
= M
xx
= M
yy
= M
xy
= 0).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
cooldown are included in the analysis (this curve was generated by assuming
that DT=155jC, DM=0). This curve illustrates the deleterious eects of
thermal stresses and strains on rst-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 rst-ply
failure loads. Finally, the third curve includes both thermal eects (based on
DT=155jC) as well as a change inmoisture 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 rst-ply
failure loads for the analysis that include both thermal and moisture eects are
slightly higher than the analysis that includes only thermal eects.
It should be kept in mind that the rst-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 rst 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
graphiteepoxy laminate based on any combi-
nation of (N
xx
,N
xy
), (N
xx
,M
xx
), and (M
xx
, M
yy
) are shown in Figures 1012,
respectively. As before, three curves are shown and are based on the
maximum stress failure criterion: one in which temperature and moisture
eects 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 eects are included.
Failure envelopes of distinctly dierent appearances may be obtained if
a dierent failure criterion is used. For example, failure envelopes obtained
using the maximum stress, TsaiHill, and TsaiWu failure criteria for
combinations of M
xx
and M
yy
are compared in Fig. 13. For this loading
condition, the TsaiWu criterion gives the most conservative rst-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
TsaiHill or TsaiWu criterion would provide the best prediction, whereas for
brittle composites the maximum stress criterion may be best.
* The failure envelope based on the TsaiWu criterion was generated by calculating the
coupling term X
12
using Eq. (61) of Chap. 5. A signicantly dierent failure envelope may be
obtained if X
12
were determined on the basis of a biaxial test or by some other means (see Sec.
5.3 of Chap. 5).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 rst-ply failure analysis, as described in Sec. 1 of Chap. 3


and illustrated in Fig. 7; or

Generate rst-ply failure envelopes as discussed in Sec. 4 and


illustrated in Figs 913.
The program prompts the user to input all information necessary to
perform the analysis. Properties of up to ve dierent materials may be
dened. 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 bers (r
f T
11
; r
f C
11
) or
Figure 10 First-ply failure envelope for a [0/45/90/45]
s
graphiteepoxy lam-
inate based on the maximum stress failure criterion and combinations of N
xx
and N
xy
(N
yy
=M
xx
=M
yy
=M
xy
=0).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
yielding of the matrix (r
yT
22
; r
yC
22
; s
y
12
). The user species the number of plies
within the laminate and material type and ber angle for each ply. The user
also selects from the maximum stress, TsaiHill, or TsaiWu failure crite-
rion.*
If a rst-ply failure analysis for a particular combination of stress and
moment resultants is to be performed, the programprompts 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 specied by the user (i.e., any two of the six stress resultants
N
xx
, N
yy
, N
xy
, M
xx
, M
yy
, M
xy
). Note that programLAMFAIL itself does not
create a failure envelope. Rather, the program creates a le (named Envel-
op.txt) that contains the stress resultant pairs predicted to cause rst-ply
failure of a composite laminate specied by the user. A failure envelope may
* If the TsaiWu criterion is selected, then the coupling strength term X
12
is calculated ac-
cording to Eq. (61) of Chap. 5.
Figure 11 First-ply failure envelope for a [0/45/90/45]
s
graphiteepoxy
laminate based on the maximum stress failure criterion and combinations of N
xx
and M
xx
(N
yy
= N
xy
= M
yy
= M
xy
= 0).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
then be created using a second software package to import the data
generated by program LAMFAIL. The rst few lines of the le 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
graphiteexpoxy 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 N
xx
and N
yy
. The rst few lines of the le represent header information. Notice
that the le contains the ply number predicted to fail and the failure type (i.e.,
ber, matrix, or shear), in addition to the combinations of N
xx
and N
yy
predicted to cause failure. Table 1 shows 12 such combinations of N
xx
and
N
yy
. The complete le Envelop.txt contains a total of 2001 pairs. Thus, a rst-
ply failure envelope may be created by importing and plotting the N
xx
and N
yy
pairs.
File Envelop.txt diers slightly if either the TsaiHill or TsaiWu
criterion is selected for use. Specically, if either of these criteria is selected,
then le 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 le
Figure 12 First-ply failure envelope for a [0/45/90/45]
s
graphiteepoxy
laminate based on combinations of M
xx
and M
yy
(N
xx
= N
yy
= N
xy
= M
xy
= 0).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Envelop.txt created during an analysis for a [0/45/90/45]
s
graphiteepoxy
laminate in which the TsaiHill 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 graphiteepoxy prepreg
systemwith 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., N
yy
=N
xx
/2 and N
xy
=M
xx
=M
yy
=M
xy
=0). Use the
Figure 13 First-ply failure envelope for a [0/45/90/45]
s
graphiteepoxy lam-
inate for combinations of M
xx
and M
yy
, based on the maximum stress, TsaiHill,
and TsaiWu criteria (N
xx
=N
yy
=N
xy
=M
xy
=0, DT=155jC, DM=1%). Fig-
ure 13 A [0/F30]
s
laminate loaded in biaxial tension, as described in Problem 1.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
following process to determine the tensile loads N
xx
and N
yy
necessary to
cause rst-ply failure, according to the maximum stress failure criteria:
(a) Use program CLT to determine the stresses (r
11
, r
22
, s
12
)
induced in each ply by unit loads N
xx
=1, N
yy
=1/2 (i.e.,
k
1
=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 1 First Few Lines of the File Envelop.txt, Created by Program LAMFAIL
During an Analysis of a [0/45/90/45]
s
GraphiteEpoxy 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 N
xx
and N
yy
were considered. The le also contains the number of the ply predicted to fail and
the failure type (i.e., ber, matrix, or shear) for each combination of N
xx
and N
yy
predicted
to cause failure.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
(c) Calculate the increase in unit loads that will cause rst-ply ply
failure.
(d) Conrm the results of part (c) using program LAMFAIL.
2. Repeat Problem 1 for a room temperature cure glassepoxy system
3. Repeat Problem 1 for a room temperature cure Kevlarepoxy system.
4. Repeat Problem 1 for the following load condition:
N
yy
N
xx
=2 N
xy
5N
xx
M
xx
M
yy
M
xz
0
5. Repeat Problem 1, except assume that the laminate is produced using a
graphiteepoxy system cured at 175jC (350jF).
6. Repeat Problem 1, except assume that the laminate is produced using a
glassepoxy system cured at 175jC (350jF).
Table 2 First Few Lines of the File Envelop.txt, Created by Program LAMFAIL
During an Analysis of a [0/45/90/45]
s
GraphiteEpoxy Laminate
**Program LAMFAIL***
Laminate failure predictions based on the TsaiHill 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 TsaiHill failure criterion had been selected for use, and stress resultants N
xx
and N
yy
were considered. The le also contains the number of the ply predicted to fail for each
combination of N
xx
and N
yy
predicted to cause failure.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
7. Repeat Problem 1, except assume that the laminate is produced using a
Kevlarepoxy system cured at 175jC (350jF).
8. Three dierent rst-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) N
xx
=327 kN/m, N
yy
=0 kN/m (this point lies on the
DT=DM=0 curve).
(b) N
xx
=79.4 kN/m, N
yy
=0 kN/m (this point lies on the DT=
155jC, DM=0 curve).
(c) N
xx
=159 kN/m, N
yy
=0 kN/m (this point lies on the
DT=155jC, DM=1% curve).
9. Three dierent rst-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) N
xx
=0 kN/m, N
yy
=32 7kN/m (this point lies on the
DT=DM=0 curve).
(b) N
xx
=0 kN/m, N
yy
=79.4 kN/m (this point lies on the
DT=155jC, DM=0 curve).
Figure 14 A [0/F30]
s
laminate loaded in biaxial tension, as described in
Problem 1.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
(c) N
xx
=0 kN/m, N
yy
=159 kN/m (this point lies on the
DT=155jC, DM=1% curve).
10. Three dierent rst-ply failure envelopes for a [0/45/9045]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) N
xx
=115.2 kN/m, N
yy
=212.1 kN/m (this point lies on the
DT= DM=0 curve).
(b) N
xx
=205.2 kN/m, N
yy
=91.6 kN/m (this point lies on the
DT= 155jC, DM=0 curve).
(c) N
xx
=172.7 kN/m, N
yy
=154.6 kN/m (this point lies on the
DT=155jC, DM=1% curve).
11. Use program LAMFAIL to create rst-ply envelopes for a [0/F45]s
graphiteepoxy laminate subjected to dierent combinations of N
xx
and
M
yy
, based on the maximum stress, TsaiHill, and TsaiWu failure
criteria. Assume that the laminate was cured at 175jCand will be used at
roomtemperatures (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-bre/epoxy composites under dry, sat-
urated and immersed fatigue conditions. Compos. Sci. Technol. 1999, 59 (14),
21152123.
4. Surgeon, M.; Vanswijgenhoven, E.; Wevers, M.; van der Biest, O. Transverse
cracking and Poissons ratio reduction in cross-ply carbon bre-reinforced
polymers. J. Mater. Sci. 1999, 34 (22), 55135517.
5. Joe, R.; Varna, J. Analytical modeling of stiness reduction in symmetric and
balanced laminates due to cracks in 90j layers. Compos. Sci. Technol. 1999, 59
(11), 16411652.
6. Pipes, R.B.; Daniel, I.M. Moire analysis of the interlaminar shear edge eect in
laminated composites. J. Compos. Mater. 1971, 5, 255.
7. Oplinger, D.W.; Parker, B.S.; Chiang, F.-P. Edge-eect studies in ber-re-
inforced laminates. Exp. Mech. 1974, 14 (9), 347354.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
8. Czarnek, R.; Post, D.; Herakovich, C.T. Edge eects in composites by Moire
interferometry. Exp. Tech. 1983, 7 (1), 1821.
9. Herakovich, C.T.; Post, D.; Buczek, M.B.; Czarnek, R. Free edge strain con-
centrations in real composite laminates: experimentaltheoretical correlation.
J. Appl. Mech. 1985, 52, 787793.
10. Pipes, R.B.; Pagano, N.J. Interlaminar stresses in composite laminates under
uniform axial extension. J. Compos. Mater. 1970, 4, 538548.
11. Rybicki, E.F. Approximate three-dimensional solutions for symmetric laminates
under in-plane loading. J. Compos. Mater. 1971, 5, 354360.
12. Wang, A.S.D.; Crossman, F.W. Some new results on edge eects in symmetric
composite laminates. J. Compos. Mater. 1977, 11, 92106.
13. Hsu, P.W.; Herakovich, C.T. Edge eects in angle-ply composite laminates.
J. Compos. Mater. 1977, 11, 422428.
14. Herakovich, C.T. On the relationship between engineering properties and de-
lamination of composite materials. J. Compos. Mater. 1981, 15, 336348.
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 13, pp. 5962.
16. Puppo, A.H.; Evensen, H.A. Interlaminar shear in laminated composites under
generalized plane stress. J. Compos. Mater. 1970, 4, 204220.
17. Pagano, N.J.; Pipes, R.B. The inuence of stacking sequence on laminate
strength. J. Compos. Mater. 1971, 5, 5055.
18. Pagano, N.J. On the calculation of interlaminar normal stress in composite
laminate. J. Compos. Mater. 1974, 8, 6581.
19. Pagano, N.J. Free-edge stress elds in composite laminates. Int. J. Solids Struct.
1978, 14, 401406.
20. Wang, S.S.; Choi, I. Boundary-layer eects in composite laminates: Part 1. Free-
edge stress singularities. J. Appl. Mech. 1982, 49, 541548.
21. Wang, S.S.; Choi, I. Boundary-layer eects 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, 561569.
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 ecient method for the calculation of in-
terlaminar stresses in composite materials. J. Appl. Mech. 1986, 53, 744750.
25. Kassapoglou, C.; Lagace, P.A. Closed form solutions for the interlaminar stress
eld in angle-ply and cross-ply laminates. J. Compos. Mater. 1987, 21, 292308.
26. Becker, W. Closed-form solutions for the free-edge eect in cross-ply laminates.
Compos. Struct. 1993, 26 (12), 3945.
27. Stern, M. Families of consistent conforming elements with singular derivative
elds. Int. J. Numer. Methods Eng. 1979, 14, 409421.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 eective axial and exural 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 eective rigidities are determined.
1 PRELIMINARY DISCUSSION
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 dened 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 deections 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. (15). In virtually all of these
references, it is assumed that the beamis composed of an isotropic material or
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
at most is a beamin which dierent layers of isotropic materials are bonded or
bolted together to formthe 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. (15), 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 classication, the stresses, strains, and deections
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 articial
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 deections induced are due to the combined eects 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 eective 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 signicant free-edge stresses occur has a width
approximately equal to the laminate thickness. Now, by denition, 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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Consequently, free-edge eects may signicantly inuence 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 eects of free-edge stresses will be ignored. The error
in predicted response caused by neglecting free-edge stresses is dicult 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 beamstiness, but may dominate beamfailure
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 dene the
beamaxis and allowthe 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 (N
xx
, N
yy
, and N
xy
) and three moment resultants (M
xx
, M
yy
, and
M
xy
). For present purposes, we will only consider two of these loads: stress
resultant N
xx
and moment resultant M
xx
. That is, to compare CLT to iso-
tropic beam theory, we will specify that N
yy
= N
xy
= M
yy
= M
xy
= 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 N
xx
is described in units of N/m (or lbf/in.), and moment resultant
M
xx
is described in units of Nm/m(or lbf in./in.). In contrast, in beamtheory,
point or concentrated loads are usually specied. In beam theory, an
axial load (corresponding to N
xx
) is described in units of N(or lbf ), whereas a
bending moment (corresponding to M
xx
) is described in units of N m (or lbf
in.). In this textbook, the superscript b is used to dierentiate between loads
as dened in beamtheory vs. loads as dened in CLT. Thus, an axial load and
bending moment applied to a beam will be denoted N
xx
b
and M
xx
b
, respec-
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
tively. Since the width of the beam is b, the two load denitions are related
according to N
xx
= N
xx
b
/b and M
xx
= M
xx
b
/b.
We are now ready to compare CLT with beamtheory. 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:
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

a
11
a
12
a
16
b
11
b
12
b
16
a
12
a
22
a
26
b
21
b
22
b
26
a
16
a
26
a
66
b
61
b
62
b
66
b
11
b
21
b
61
d
11
d
12
d
16
b
12
b
22
b
62
d
12
d
22
d
26
b
16
b
26
b
66
d
16
d
26
d
66
_

_
_

_
N
xx
N
T
xx
N
M
xx
N
yy
N
T
yy
N
M
yy
N
xy
N
T
xy
N
M
xy
M
xx
M
T
xx
M
M
xx
M
yy
M
T
yy
M
M
yy
M
xy
M
T
xy
M
M
xy
_

_
_

_
Since our current objective is to compare CLT to isotropic beam theory, we
will simplify our discussion by ignoring the eects of changes in temperature
and moisture content (i.e., let N
ij
T
=M
ij
T
=N
ij
M
=M
ij
M
=0). For an isotropic
beam or plate, substantial simplications occur in the [abd] matrix, as was
already discussed in Sec. 7 of Chap. 6. Noting that we have already specied
that N
yy
=N
xy
=M
yy
=M
xy
=0, for an isotropic beam, Eq. (45) of Chap. 6
reduces to:
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

a
11
a
12
0 0 0 0
a
12
a
11
0 0 0 0
0 0 2a
11
a
12
0 0 0
0 0 0 d
11
d
12
0
0 0 0 d
12
d
11
0
0 0 0 0 0 2d
11
d
12

_
_

_
N
b
xx
=b
0
0
M
b
xx
=b
0
0
_

_
_

_
where for an isotropic material, the elements of the [abd] matrix reduce to (see
Sec. 7 of Chap. 6):
a
11

1
Et
a
12

v
Et
d
11

12
Et
3
d
12

12v
Et
3
d
66
2d
11
d
12

241 v
Et
3
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Let us rst consider the midplane strains induced by N
xx
b
alone (that is, let
M
xx
b
= 0). From the above result, the axial midplane strain induced by N
xx
b
(only) is given by:
e
o
xx
a
11

N
b
xx
b

1
Et
_ _
N
b
xx
b
Since the cross-sectional area of the rectangular beam is A = (t)(b), we can
write this result as:
e
o
xx

N
b
xx
AE

r
xx
E
1a
The quantity r
xx
=N
xx
b
/Ais simply the uniaxial stress induced in a prismatic
isotropic beamsubjected to an axial load. Hence, CLT reduces to the uniaxial
form of Hookes 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:
e
o
yy
a
12

N
b
xx
b

v
Et
_ _
N
b
xx
b

vN
b
xx
AE

vr
xx
E
ve
o
xx
1b
Hence, for uniform axial loading, the transverse strain e
yy
o
is related to the
axial strain e
xx
o
via Poissons 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):
c
o
xy
0 1c
It is mentioned in passing that the stress tensor induced in the beam is
uniaxial (i.e., r
xx
p 0, r
yy
=r
zz
=s
xy
=s
xz
=s
yz
=0). That is, the principal
stresses are r
p1
= r
xx
, r
p2
= r
p3
= 0, and the principal stress coordinate sys-
tem is coincident with the xyz coordinate system. Similarly, since shear
strain is zero, the principal strain coordinate system is coincident with the
xyz coordinate system and e
p
1
= e
xx
o
, e
p
2
= e
yy
o
. The principal stress and
principal strain coordinate systems are therefore coincident, which is always
the case for isotropic materials.
Now consider the midplane strains and curvatures induced by M
xx
b
alone (that is, let N
xx
b
=0). In this case, CLTpredicts that all midplane strains
are zero (e
xx
o
= e
yy
o
= c
xy
o
= 0). Therefore, the midplane represents the neutral
surface, as expected for an isotropic beam subjected to pure bending. The
midplane curvatures are:
j
xx
d
11
M
b
xx
b
_ _
j
yy
d
12
M
b
xx
b
_ _
j
xy
0
The fact that the twist curvature j
xy
is zero implies that the xz and yz planes
are the principal planes of curvature, as expected for a prismatic isotropic
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
beam with symmetric rectangular cross section. Substituting the reduced
forms for d
11
and d
12
for an isotropic material, we nd:
j
xx

12
Et
3
M
b
xx
b
_ _
j
yy

12v
Et
3
M
b
xx
b
_ _
The moment of inertia for a rectangular cross section of width b and height t is
I = bt
3
/12. Therefore the above expressions can be written as:
j
xx

M
b
xx
EI
2a
j
yy

vM
b
xx
EI
2b
Equations (2a) and (2b) are the well-known moment-curvature equations for
isotropic beams. The product EI is known as the exural rigidity of an
isotropic beam. Together, Eqs. (2a) and (2b) imply that j
yy
= vj
xx
, which
shows that anticlastic bending of an isotropic beam with symmetric rectan-
gular cross section is predicted by CLT and, further, that j
yy
is related to j
xx
via Poissons ratio, as expected.
The through-thickness variation in normal strains e
xx
and e
yy
is given by:
e
xx
e
o
xx
zj
xx
z
M
b
xx
EI
_ _
e
yy
e
o
yy
zj
yy
z
vM
b
xx
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):
r
xx
r
yy
s
xy
_

_
_

E
1 v
2

vE
1 v
2

0
vE
1 v
2

E
1 v
2

0
0 0
E
21 v
_

_
_

_
e
xx
e
yy
c
xy
_

_
_

E
1 v
2

vE
1 v
2

0
vE
1 v
2

E
1 v
2

0
0 0
E
21 v
_

_
_

_
zM
b
xx
EI
zvM
b
xx
EI
0
_

_
_

_
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Upon completing the matrix multiplication indicated above, we nd:
r
xx
r
yy
s
xy
_

_
_

M
b
xx
z
I
0
0
_

_
_

_
Hence, the CLT analysis predicts that r
yy
and s
xy
are both zero, and that the
axial stress r
xx
induced in an isotropic beam by M
xx
b
(only) is given by the
familiar exure formula:
r
xx

M
b
xx
z
I
3
As before, since only an axial stress is induced, the stress tensor is
everywhere uniaxial and principal stresses are r
p1
= r
xx
, r
p2
= r
p3
= 0. The
principal stress coordinate system is coincident with the xyz coordinate
system. Similarly, since shear strain is zero, the principal strain coordinate
systemis coincident with the xyz coordinate systemand e
p1
=e
xx
o
, e
p2
=e
yy
o
.
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
subjected to an axial load N
xx
b
and bending moment M
xx
b
, both of which are
constant along the length of the beam. If N
xx
b
=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. Specically, traditional
beam theory allows one to calculate the shear stresses induced in a beam.
Recall that shear stresses s
xz
are given by the so-called shear formula (15):
s
xz

VQ
Ib
where V is the shear force present at a specied beam cross section, Q is the
rst moment of an area about the neutral axis (for a rectangular beam
Q = b[(t
2
/4)z
2
]/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 s
xz
is due to the fundamental
assumption that the bending moment M
xx
b
is constant along the length of
the beam. That is, recall fromearlier studies [see, for example, Refs. (15)] that
the shear force V is related to the bending moment M
xx
b
according to:
V
dM
b
xx
dx
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Since M
xx
b
is assumed constant in a CLT analysis, V=dM
xx
b
/dx =0. There-
fore a CLT analysis implies that shear stress s
xz
= 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 dierent. In particular, for composite beams, the exure formula [Eq.
(3)] is only valid if the beam is oriented in a specic 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
Fig. 1. An externally applied axial load N
xx
b
, bending moment M
xx
b
, and
transverse distributed load q(x) are also shown. Both the beam cross section
and applied loads are referenced to an xyz 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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
y- and z-axes are orthogonal to the long axis of the beam. Bending moment
M
xx
b
and transverse load q(x) act within the xz plane, and hence the xz
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 dierent 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 eective axial rigidity
exhibited by a composite beam is identical in either orientation. As will be
seen, however, the eective exural rigidity of a rectangular composite beam
will dier substantially depending on whether the beamis 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 exural 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 at 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-
ned as the xz 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 ange laminate (or ange laminates). Thus, for example, a
composite I-beam has one web laminate and two ange laminates, whereas a
composite box-beam has two ange 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 ange 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 eects.
That is, if a beam is produced using nonsymmetric web or ange laminates,
then substantial thermal and/or moisture moment resultants (M
ij
T
and M
ij
M
)
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, M
ij
T
=M
ij
M
=0, and hence this complication is eliminated.
Secondly, we require that all beamcross sections possess both geometric
and material symmetry about the plane of loading (the xz plane). Having
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
already stipulated that all laminates are symmetric, then the requirement of
geometric and material symmetry about the xz plane is achieved automati-
cally for solid rectangular composite beams and in these cases requires no
further discussion. However, symmetry about the xz 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 xz plane is automatically satised 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
ber angles that are symmetric about the xz plane, as well as being symmetric
about the local web laminate midplanes.
Note that symmetry about the xz plane places no restrictions on the
ange laminates. Thus, the top and bottomanges in an I-, hat-, or box-beam
are allowed to have a dierent stacking sequence and thickness, although they
must both be symmetric about their respective midplanes (the stacking
sequences and widths of the two bottom anges in a hat-beam must be
identical, however). The beamcross section may therefore be nonsymmetrical
about the xy plane.
In practice, it can be problematic to produce thin-walled composite
beams that feature the symmetries just described. Diculties 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 diculties 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 rst curing the ange 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 ange laminates are completely
independent. Since any stacking sequence can be used in either laminate, it
is easy to produce a T-beamwith the required symmetric stacking sequence in
both web and ange laminates in this manner and to insure symmetry about
the xz plane. However, a distinct disadvantage is that the web and ange
laminates are joined solely by the adhesive bondno continuous bers cross
the junction between web and ange. 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 ange
laminate. Additional plies, which span the width of the ange, are added to
complete the ange laminate. During layup, an internal v-shaped cavity is
formed near the webange junction. This cavity is lled with some ller
material (often pre-impregnated unidirectional tow) prior to curing the
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
T-beam. This second approach of producing a T-beam has the distinct ad-
vantage of providing continuous bers across the junction between the web
and ange. However, the stacking sequences used in the web and ange
laminates are no longer independent. Since the ange laminate is required to
be symmetric (at least for the analysis presented in this section), the ber
angles used in the web and inner portions of the ange laminates must be
repeated on the outer surface of the ange. 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 ange laminate,
Figure 2 Methods to produce a composite T-beamusing 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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
unless only 0j or 90j plies are used in the web. This diculty may not be
immediately obvious, but is illustrated in Fig. 3. The gure 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 ange
results in a single ange ply in which the ber angle is +hj on one side of the
angeweb junction, but hj on the other side. Hence, it is not possible to add
additional plies across the entire width of the ange to produce both a sym-
metric laminate, unless web ber 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 sti-
ness. Asimilar diculty is encountered when producing a beamwith two web
laminates such as a hat- or box-beam. A single ply that becomes part of both
web laminates and both ange 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 xz plane since a web ply with a ber angle of +hj on one side of the xz
plane becomes a ply with a ber angle of hj on the other side of the xz 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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
As before, the only way this can be avoided (if the beam is produced using
unidirectional pre-preg tape) is to use ber angles of h =0j or 90j in the web,
which may lead to web laminates with unacceptably low shear stiness.
These diculties 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 ber directions, oriented
symmetrically about the warp direction. Since each individual woven or
braided ply features a symmetric Fhj ber pattern, plies extending from the
web into the ange (as in a T-beam) or completely around the circumference
of the cross section (as in a box-beam) retain the identical Fhj ber pattern in
both web and ange 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 ange 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 xy plane.
Suppose, for example, that an I-beam is produced in which the (symmetric)
top ange is produced using a very dierent stacking sequence than the
(symmetric) bottom ange. This implies that the eective thermal expansion
coecient a
xx
of the top ange may be very dierent than that of the bottom
ange. Consequently, if the beam experiences a uniform change in temper-
Figure 4 Illustration of why unidirectional fabrics or pre-pregs cannot be used
to produce web laminates that are symmetric about the xz plane.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Figure 5 The use of woven or braided fabrics or pre-preg avoids the difficulties
shown in Figs. 3 and 4.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
ature, then the top ange will tend to expand or contract at a dierent rate
than the bottom ange, causing the beam to warp. The symmetric web
laminate will tend to restrict such warping, but nevertheless, this eect may
be signicant depending on details of a given beamdesign. This possibility will
not be addressed in this text, although the theory presented can be easily
modied to account for this eect.
4 EFFECTIVE AXIAL RIGIDITY OF RECTANGULAR
COMPOSITE BEAMS
A rectangular composite beam subjected to axial load N
xx
b
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
beamto pure axial loading. As mentioned earlier, signicant free-edge stresses
may be present (depending on the stacking sequence and material systemused
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
is N
xx
b
, the midplane strains and curvatures induced in a symmetric compo-
site beam can be predicted using Eq. (45) of Chap. 6, which becomes (for
N
yy
=N
xy
=M
xx
=M
yy
=M
xy
=b
ij
=M
ij
T
=M
ij
M
=0):
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

a
11
a
12
a
16
0 0 0
a
12
a
22
a
26
0 0 0
a
16
a
26
a
66
0 0 0
0 0 0 d
11
d
12
d
16
0 0 0 d
12
d
22
d
26
0 0 0 d
16
d
26
d
66
_

_
_

_
N
b
xx
=b N
T
xx
N
M
xx
N
T
yy
N
M
yy
N
T
xy
N
M
xy
0
0
0
_

_
_

_
4
Note that thermal and moisture stress resultants are (in general) present in
the symmetric composite beam, even though the only externally applied
load is N
xx
b
. 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
application of an external axial load N
xx
b
. From Eq. (4), it is easy to see
that the incremental change in midplane strains caused by N
xx
b
(only) is
given by:
e
o
xx

a
11
b
N
b
xx
e
o
yy

a
12
b
N
b
xx
c
o
xy

a
16
b
N
b
xx
5
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
The eective extensional Youngs modulus, Poissons ratio, and coecient of
mutual inuence of the second kind for a composite laminate were all dened
as follows in Sec. 9.1.1 of Chap. 6:
E
ex
xx

1
ta
11
v
ex
xy

a
12
a
11
g
ex
xx;xy

a
16
a
11
Substituting these eective properties into Eq. (5), we nd:
e
o
xx

N
b
xx
tbE
ex
xx

N
b
xx
AE
ex
xx

r
xx
E
ex
xx
6a
e
o
yy
v
ex
xy
N
b
xx
AE
ex
xx
_ _
v
ex
xy
r
xx
E
ex
xx
_ _
v
ex
xy
e
o
xx
6b
c
o
xy
g
ex
xx;xy
N
b
xx
AE
ex
xx
_ _
g
ex
xx;xy
r
xx
E
ex
xx
_ _
g
ex
xx;xy
e
o
xx
6c
From Eq. 6(a), we see that the eective axial rigidity of a rectangular com-
posite beam can be written as:
AE
ex
xx
_ _
tb E
ex
xx
7a
Figure 6 A composite beam with cross-section bt, subjected to axial load
N
xx
b
.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
where A is the cross-sectional area of the beam and E
xx
ex
is the eective
extensional Youngs modulus of the laminate (dened in Sec. 9.1.1 of Chap.
6). Alternatively, the eective axial rigidity may be written as:
AE
ex
xx
_ _

b
a
11
_ _
7b
Although the beam shown in Fig. 6 is oriented such that ply interfaces are
orthogonal to the xz plane, the analysis is also applicable to a beamoriented
such that ply interfaces are parallel to the xz plane, as shown in Fig. 1(b).
Hence, Eqs. (7a) and (7b) give the eective 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 eective stress tensor is uniaxial (i.e, r
xx
p 0; r
yy
r
zz
s
xy
s
xz
s
yz
0. That is to say, the eective principal stresses are r
p
1
=r
xx
;
r
p
2
=r
p
3
0, and the eective principal stress coordinate system is coinci-
dent with the xyz coordinate system. However, for a composite beam, an
axial stress j
xx
will (in general) cause a shear strain c
xy
o
due to the presence of
a
16
(or equivalently, due to the presence of D
xx,xy
ex
). Therefore, the principal
strain coordinate system is not, in general, coincident with the eective prin-
cipal stress coordinate system.
Recall from Sec. 7 of Chap. 6 that a
16
=0 for the following commonly
used stacking sequence:

Cross-ply.

Balanced.

Balanced angle-ply.

Quasi-isotropic.
If an axially loaded symmetric composite beamis produced using any of
these stacking sequences, then shear strain c
xy
o
is zero and the eective
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 exural
rigidity of a beam is dened in pure bending. In this section, the eective
exural rigidities of composite beams with rectangular cross sections will be
determined. In order to do so, we must consider two dierent cases. In the rst
case, we consider a beam oriented such that ply interfaces are orthogonal to
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
the plane of loading, as previously shown in Fig. 1(a). In the second case, we
consider a beamfor 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,
signicant free-edge stresses may be present (depending on the stacking
Figure 7 A composite beam with ply interfaces orthogonal to the xz plane,
subjected to pure bending. (a) A composite beam subjected to pure bending.
(b) Cutaway view of composite beam at cross-section AA.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
sequence and material systemused to produce the beam) but will be neglected
in the following discussion.
Since we have assumed that the beam is symmetric and subjected to
M
xx
b
only, the midplane strains and curvatures can be predicted using Eq.
(45) of Chap. 6, which becomes (for b
ij
N
xx
N
yy
N
xy
M
yy
M
xy

M
T
ij
M
M
ij
0 and M
xx
=M
xx
b
/b):
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

a
11
a
12
a
16
0 0 0
a
12
a
22
a
26
0 0 0
a
16
a
26
a
66
0 0 0
0 0 0 d
11
d
12
d
16
0 0 0 d
12
d
22
d
26
0 0 0 d
16
d
26
d
66
_

_
_

_
N
T
xx
N
M
xx
N
T
yy
N
M
yy
N
T
xy
N
M
xy
M
b
xx
=b
0
0
_

_
_

_
8
Thermal and moisture stress resultants N
ij
T
and N
ij
M
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 inuence the initial elastic response of the beam,
however. From Eq. (8), it is easy to see that midplane strains are not changed
upon the application of M
xx
b
since b
ij
=0. Hence, the midplane of the beam
cross section represents the neutral surface in the sense that application of
M
xx
b
does not contribute to (or alter) preexisting midplane strains. As noted in
Sec. 2, the neutral surface for an isotropic beamwith rectangular cross section
is also coincident with the midplane.
From Eq. (8), the midplane curvatures are given by:
j
xx
d
11
M
b
xx
b
_ _
j
yy
d
12
M
b
xx
b
_ _
j
xy
d
16
M
b
xx
b
_ _
9
The fact that the twist curvature j
xy
p 0 (in general) shows that the xz and
yz planes are not the principal planes of curvature. This is in direct contrast
to an isotropic beamsince for a prismatic isotropic beamwith symmetric rect-
angular cross section, the xz and yz planes represent the principal planes of
curvature.
The eective exural Youngs modulus, Poissons ratio, and coecient
of mutual inuence of the second kind for a composite laminate were all
dened as follows in Sec. 9.1 of Chap. 6:
E
xx

12
d
11
t
3
v
xy

d
12
d
11
g
xx;xy

d
16
d
11


Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Substituting these eective properties into Eq. (9), we nd:
j
xx

12
E
xx
t
3
M
b
xx
b
_ _
j
yy

12v
xy
E
xx
t
3
M
b
xx
b
_ _
j
xy

12g
xx;xy
E
xx
t
3
M
b
xx
b
_ _
The moment of inertia for a rectangular cross section of width b and height t
is I = bt
3
/12. Therefore the above expressions can be written as:
j
xx

M
b
xx
E
xx
I
10a
j
yy

v
xy
M
b
xx
E
xx
I
10b
j
xy

g
xx;xy
M
b
xx
E
xx
I
10c
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
xx

I represents the eective exural rigidity for a composite beam


in this orientation. That is, the eective exural rigidity of a rectangular
composite beam with ply orientation orthogonal to the plane of loading can
be written as:
IE
xx
_ _

bt
3
E
xx
12

b
d
11
_ _
11
where E
xx

is the eective exural Youngs modulus of the laminate (dened


in Sec. 9.1.2 of Chap. 6).
Equations (10a) and (10b) also imply j
yy
=v
xy

j
xx
, 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
(10c) imply j
xy
=g
xx,xy
fl
j
xx
, which shows that a twist curvature is induced in a
composite laminate, unless g
xx,xy
fl
= 0. A midplane twist curvature j
xy
does
not occur for isotropic beams subjected to pure bending. Recalling that g
xx,xy

=d
16
/d
11
, it is seen that g
xx,xy

is zero only for specially orthotropic laminates


(i.e., if d
16
=0). This rarely occurs for composite beams used in practice. The
only common stacking sequences that lead to a specially orthotropic beamare
unidirectional [0]
n
or [90]
n
stacks or symmetric cross-ply [0/90]
ns
stacks. The
rst two stacking sequences are essentially never used in practice because they
possess signicant strength and stiness in one direction only. Cross-ply
beams may be used on occasion but suer from very low shear stiness.
Beams produced using any other stacking sequence, for example, a symmetric

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.


quasi-isotropic stacking sequence, will exhibit a twist curvature when sub-
jected to a pure bending.
Based on the above denitions, we see that through-thickness variation
in strains e
xx
, e
yy
, and c
xy
induced by pure bending is given by:
e
xx
z zj
xx
z
M
b
xx
E
xx
I
_ _
12a
e
yy
z zj
yy
z
v
xy
M
b
xx
E
xx
I
_ _
12b
c
xy
z zj
xy
z
g
xx;xy
M
b
xx
E
xx
I
_ _
12c
Strains are predicted to vary as linear functions of z, as expected from results
in earlier chapters. Note, however, that ply stresses r
xx
, r
yy
, and s
xy
are not
linear functions of z. It is not possible to develop a exure formula for use
with composite beams with ply orientation orthogonal to the plane of
loading. Rather, stresses must be calculated using Hookes 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 ber
angles are referenced to the +xV-axes and are measured positive from the
+xV-axis towards the +yV-axis, in accordance with the right-hand rule.
Note that the bending moment is referenced to the xyz coordinate
system, whereas the ply stacking sequence is referenced to a new xVyVzV
coordinate system. As is apparent from Fig. 8, the +zV- and +yV-axes are
coincident, as are the +x- and +xV-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
potentially confusing change. For example, the bending moment M
xx
b
shown

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.


in Fig. 8 causes a very dierent state of stress in the composite beamthan does
the bending moment shown in Fig. 7 (which is also denoted M
xx
b
). This
dierence 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
properties are symmetric about the xz plane. Since the moment M
xx
b
is
constant along the length of the beam, all beamcross sections must deformin
an identical manner. Under these conditions, the long axis of the beam must
deform into a circular arc with a radius of curvature r
xx
, as shown in Fig. 9.
Hence, beam cross sections that are initially plane and perpendicular to the
axis of the beammust remain plane and perpendicular to the deformed axis of
the beam following loading. These are precisely the same conditions encoun-
Figure 8 A composite beam with ply interfaces parallel to the xz plane,
subjected to pure bending. (a) A composite beam subjected to pure bending.
(b) Cutaway view of composite beam at cross-section A-A.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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:
e
xx
zj
xx
13
where curvature j
xx
= 1/r
xx
. 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 j
xx
represents bending in the xz plane (the plane of loading) and is
not comparable to the curvature j
xx
discussed previously. The curvature pre-
viously discussed occurs in the xVzV plane and would now be labeled j
xVxV
.
Since we have assumed both a symmetric stacking sequence and
symmetric rectangular cross section, the axial strain induced at any position
z will be uniformacross the width of the beam. Hence, the axial strain given by
Eq. (13) represents the strain induced at the midplane of the laminate:
e
o
xx
e
o
xVxV
zj
xx
14
We also note that for the loading condition shown in Fig. 8:
N
yVyV
N
xVyV
M
xVxV
M
y VyV
M
xVyV
0:
Therefore, Eq. (45) of Chap. 6 becomes:
e
o
xVxV
e
o
yVyV
c
o
xVyV
j
xVxV
j
yVyV
j
xVy V
_

_
_

a
11
a
12
a
16
0 0 0
a
12
a
22
a
26
0 0 0
a
16
a
26
a
66
0 0 0
0 0 0 d
11
d
12
d
16
0 0 0 d
12
d
22
d
26
0 0 0 d
16
d
26
d
66
_

_
_

_
N
xVxV
N
T
xVxV
N
M
xVxV
N
T
yVyV
N
M
yVyV
N
T
xVyV
N
M
xVyV
0
0
0
_

_
_

_
Our current objective is to evaluate the eective exural 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 N
xVxV
=N
xx
is:
e
o
xVxV
e
o
xx
a
11
N
xVxV
a
11
N
xx
Combining this result with Eq. (14), we nd:
N
xx
z
j
xx
a
11
_ _
15
Now consider a free-body diagramshowing the internal distribution of stress
resultant N
xx
at arbitrary cross-section AA, as shown in Fig. 10. Since the
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
beam is in static equilibrium, the externally applied moment M
o
b
must be
balanced by the distribution of internal stress resultant N
xx
. Recalling that the
units of N
xx
are force-per-length (e.g., N/m), the incremental axial force
acting over an innitesimal strip of height dz, located at an arbitrary distance
z from the neutral axis, is given by:
dN
b
xx
N
xx
dz
Since this force acts at distance z from the neutral axis, the incremental
internal moment associated with this force is:
dM
b
xx
dN
b
xx
z N
xx
zdz
The total moment is obtained by integrating over the height of the beam:
M
b
xx
m
b=2
b=2
N
xx
zdz m
b=2
b=2
j
xx
a
11
_ _
z
2
dz
Neither j
xx
nor a
11
is a function of z, so they can be removed fromthe integral
sign. Completing the integration indicated, we nd:
M
b
xx

j
xx
b
3
12a
11
The eective extensional Youngs modulus is given by Eq. (63) of Chap. 6,
repeated here for convenience:
E
ex
xx

1
ta
11
Figure 10 Free-body diagram used to relate external bending moment M
o
b
to
internal stress resultant N
xVxV
.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Our expression for M
xx
b
can therefore be written as:
M
b
xx

j
xx
E
ex
xx
tb
3
12
Noting that the area moment of inertia about the y-axis for the rectangular
cross section is I = tb
3
/12, we can rearrange this result to nd:
j
xx

M
b
xx
E
ex
xx
I
16
Equation (16) is a moment-curvature equation for a composite beamwith
ply orientation parallel to the plane of loading. The product E
xx
ex
I represents
the eective exural rigidity of a composite beam in this orientation. That is,
the eective exural rigidity of a rectangular composite beam with ply
orientation parallel to the plane of loading can be written as:
IE
ex
xx
_ _

tb
3
E
ex
xx
12

b
3
12a
11
_ _
17
where E
xx
ex
is the eective extensional modulus of the laminate (dened in
Sec. 9.1.1 of Chap. 6). Equation (17) should be compared with our earlier
expression for the eective exural rigidity of a composite beam with plies
orthogonal to the plane of loading Eq. (11). It is seen that the exural rigidity
of a composite beamwith plies parallel to the plane of loading is dominated by
extensional stinesses (i.e., E
ex
xx
or a
11
). In contrast, the exural rigidity of a
beam with plies orthogonal to the plane of loading is dominated by exural
stinesses (i.e., E
xx

or d
11
).
Anticlastic bending will occur for the composite beamshown in Fig. 10.
Normal strains e
yy
=e
z Vz V
will vary linearly with distance fromthe neutral axis.
Assuming a positive bending moment, then in regions below the neutral axis
(i.e., for z = yV >0), the transverse normal strains will be compressive,
whereas in regions above the neutral axis (for z =yV<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 eective (or nominal) normal stress is given by
r
xVxV
=r
xx
=N
xx
/t. Substituting this denition into Eq. (15) and rearranging,
we have:
j
xx

r
xx
ta
11
z

r
xx
zE
ex
xx
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Combining this result with Eq. (16) and solving for eective stress, we
have:
r
xx

M
b
xx
z
I
18
Equation (18) is analogous to the exure formula for isotropic beams [Eq. (3)]
and can be used to determine the eective stress at any position z. As would be
expected, the eective stress r
xx
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
eective stress r
xx
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 eective axial and exural 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 eective
rigidities is similar for all cross sections. Acomposite box-beamwill 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 ange laminates may dier. The
thickness of the top and bottom ange laminates is labeled t
tf
and t
bf
,
respectively. The width of both anges is labeled b. The two identical web
laminates have a thickness of t
w
and height h. Thus, the overall width and
depth of the beam cross section are (b) and (h+t
tf
+t
bf
), respectively.
To dene the axial rigidity, we must determine the axial strains e
xx
induced when the beam is subjected to an axial load N
xx
b
(only). Further,
N
xx
b
must be applied such that uniform axial strains are induced; that is, the
beammust not bend. This is equivalent to saying that the line of action of N
xx
b
must pass through the centroid of the beam cross section. It is known a priori
that the centroid lies along within the xz plane since we have required
symmetry about this plane. However, we have not required symmetry about
the xy plane. Hence, the position of the centroid relative to either outer
surface of the beam cross section (that is, coordinates z
t
and z
b
shown in Fig.
11) is initially unknown. Note that since the z-axis is positive downwards, in
all cases, z
t
<0 and z
b
>0. Our rst objective is to determine either distance z
t
or z
b
. In the following derivation, we will determine distance z
b
. Note that
once z
b
is known, z
t
is also known since z
t
= z
b
(h+t
bf
+t
tf
).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Sketches of a composite box-beam subjected to axial load N
xx
b
(only)
are shown in Fig. 12. In this gure, it is assumed that the line of action of N
xx
b
does pass through the centroid. Consequently, a uniform axial strain e
xx
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:
N
xx
j
w

e
xx
a
w
11
t
w
E
ex
xx w
e
xx
j 19a
N
xx
j
bf

e
xx
a
bf
11
t
bf
E
ex
xx bf
e
xx
j 19b
N
xx
j
tf

e
xx
a
tf
11
t
tf
E
ex
xx tf
e
xx
j 19c
Symbols w, bf, and tf have been used to denote variables associated
with the web laminate, bottom ange laminate, and top ange laminate,
respectively. For example, E
xx
ex
j
w
represents the eective extensional modulus
of the web laminates.
Figure 11 Cross section of a composite box-beam. The positive x-axis is out
of the plane of the figure.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
The stress resultants given by Eqs. (19a)(19c) are indicated in the free-
body diagramshown 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 (SM
y
=0),
we obtain:
2 N
xx
j
w
h z
b
t
bf
h=2 N
xx
j
bf
b z
b
t
bf
=2 N
xx
j
tf
b
z
b
t
bf
h t
tf
=2 0
Substituting [Eqs. (19a), (19b), (19c)] and solving for distance z
b
, we nd:
z
b

1
2
2A
w
E
ex
xx

w
h 2t
bf
t
bf
A
bf
E
ex
xx

bf
A
tf
E
ex
xx

tf
2t
b
2h t
tf

2A
w
E
ex
xx

w
A
bf
E
ex
xx

bf
A
tf
E
ex
xx

tf
_

_
_

_20a
Figure 12 Side view of a composite box-beam subjected to axial load N
xx
b
. The
line-of-action of N
xx
b
is assumed to pass through the centroid, located distance
z
b
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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
where
A
w
= t
w
h = cross-sectional area of the web laminate.
A
bf
= t
bf
b = cross-sectional area of the bottom ange laminate.
A
tf
= t
tf
b = cross-sectional area of the top ange laminate.
Noting that the eective extensional Youngs moduli can be written as
E
ex
xx
A
w
1=t
w
a
w
11
, E
ex
xx
A
bf
1=t
bf
a
bf
11
, and E
ex
xx
A
tf
1=t
tf
a
tf
11
, this result can also
be expressed as:
z
b

1
2
_
2ha
bf
11
a
tf
11
h 2t
bf
ba
w
11
t
bf
a
tf
11
a
bf
11
2t
bf
2h t
tf

_
2ha
bf
11
a
tf
11
ba
w
11
a
bf
11
a
tf
11

_
20b
Equations (20a) and (20b) gives the distance from the centroid to the bottom
surface of the composite box-beam. If the bottomand top ange laminates are
identical (i.e., if A
bf
=A
tf
, t
bf
=t
tf
, and E
ex
xx
A
bf
E
ex
xx
A
tf
), then Eqs. (20a) and
(20b) reduces to z
b
=t
bf
+h/2, as would be expected.
Let us now determine the eective axial rigidity of the composite box-
beam. Since a uniformaxial strain (e
xx
) is induced, the stress resultants present
in each segment of the beam to the axial strain are given by:
N
xx
j
w

e
xx
a
w
11
t
w
E
ex
xx

w
e
xx
N
xx

bf

e
xx
a
bf
11
t
bf
E
ex
xx

bf
e
xx
N
xx

tf

e
xx
a
tf
11
t
tf
E
ex
xx

tf
e
xx
21
Also, the total load applied to the beamas a whole (N
xx
b
) must equal the
sum of the forces acting in each beam segment:
N
b
xx
2 N
xx
w
j h N
xx
bf
j b N
xx
tf
j b
Dene the eective stress applied to the beamas the force applied to the beam
divided by the area of the beam cross section:
r
beam

N
b
xx
A
beam

2N
xx

w
h N
xx

bf
b N
xx

tf
b
2A
w
A
bf
A
tf
22
Substituting Eq. (21) into Eq. (22), we nd:
r
beam
e
xx
2A
w
E
ex
xx

w
A
bf
E
ex
xx

bf
A
tf
E
ex
xx

tf
2A
w
A
bf
A
tf
_
_
_
_
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
We can now dene the eective Youngs modulus of the beam as:
E
beam
xx

2A
w
E
ex
xx

w
A
bf
E
ex
xx

bf
A
tf
E
ex
xx

tf
2A
w
A
bf
A
tf
23
The eective axial rigidity of the composite box-beam is simply the eective
beam Youngs modulus multiplied by the cross-sectional area:
AE
beam
xx
2A
w
E
ex
xx

w
A
bf
E
ex
xx

bf
A
tf
E
ex
xx

tf
24a
Alternatively, the eective axial rigidity of the box-beam can be written as:
AE
beam
xx

2h
a
w
11

b
a
bf
11

b
a
tf
11
24b
If the bottom and top ange laminates are identical (i.e., if A
bf
=A
tf
, t
bf
=t
tf
,
and E
xx
ex
j
bf
=E
xx
ex
j
tf
), then Eqs. (24a) and (24b) reduce to:
AE
beam
2 A
w
E
ex
xx

w
A
bf
E
ex
xx

bf
_ _
2
h
a
w
11

b
a
bf
11
_ _
Next, consider the eective exural rigidity of a composite box-beam.
Recall that the exural rigidity of a beam is dened 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 xz
plane, the long x-axis of the beam must deform into a circular arc with a
radius of curvature r
xx
, as shown. Hence, cross sections of the box-beam that
are initially plane and perpendicular to the axis of the beammust 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:
e
xx
zj
xx
25
where curvature j
xx
=1/r
xx
. We assume that the axial strain induced at any
position z is uniformacross the width of the web and ange laminates; that is,
we assume that e
xx
is not a function of y.
Our rst objective is to determine the location of the neutral axis. That
is, we wish to determine either distance z
t
or z
b
, previously shown in Fig. 11. In
the following derivation, we will determine distance z
b
, 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 z
t
can easily be
developed.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Figure 13 Side view of a composite box-beam subjected to pure bending. (a)
Composite box-beam deformed into a circular arc of radius r
xx
. (b) Distribution
of internal stress resultants N
xx
A
w
, N
xx
A
bf
, and N
xx
A
tf
.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
First consider the web laminates. We note that for the web laminate, the
ply interfaces are parallel to the xz plane. Also, we have assumed that the
strain e
xx
is not a function of y. Hence, Eq. (25) indicates that the midplane
strains induced in the web laminates vary with z:
e
o
xx

w
zj
xx
26
Since the web laminates are symmetric, the stress resultant induced at any
point in the webs, N
xx

w
, is given by:
e
o
xx

w
a
w
11
N
xx

w
27
Equating Eqs. (26) and (27) and solving for N
xx

w
, we have:
N
xx

w
z
j
xx
a
w
11
_ _
zt
w
E
ex
xx

w
j
xx
This result shows that N
xx

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:
dN
w
xx
N
xx

w
dz
Hence, the total axial load that supported the web, N
b
xx

w
, is given by:
N
b
xx

m
z
b
t
bf
z
b
ht
bf
N
xx

w
dz m
z
b
t
bf
z
b
ht
bf
t
w
E
ex
xx

w
j
xx
_ _
zdz
Evaluating this integral, we nd:
N
b
xx

t
w
E
ex
xx

w
j
xx
2
z
b
t
bf

2
z
b
h t
bf

2
_ _
28
Now consider the bottom ange laminate. Referring to Fig. 11, note that the
midplane of the bottom ange is located at z = z
b
t
bf
/2. Substituting this
value into Eq. (25), the strain induced at the midplane of the bottom ange
laminate is given by:
e
o
xx

bf
z
b

t
bf
2
_ _
j
xx
29
Since the bottom ange laminate is symmetric, the stress resultant induced in
the bottom ange, N
xx

bf
, is:
e
o
xx

bf
a
bf
11
N
xx

bf

N
xx

bf
t
bf
E
ex
xx

bf
30
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Equating Eqs. (29) and (30) and solving for N
xx

bf
, we nd:
N
xx

bf
z
b

t
bf
2
_ _
t
bf
E
ex
xx

bf
j
xx
31
This result gives the value of N
xx

bf
induced at the midplane of the bottom
laminate, shown in Fig. 13(b). The total axial force supported by the bottom
ange, N
b
xx

bf
, is obtained by multiplying stress resultant N
xx

bf
by the width of
the bottom ange, b:
N
b
xx

bf
bN
xx

bf
z
b

t
bf
2
_ _
A
bf
E
ex
xx

bf
j
xx
32
Note that the cross-sectional area of the bottom ange, A
bf
=bt
bf
, has been
used in Eq. (32).
Finally, we must determine the total axial force supported by the top
ange. FromFig. 11, we see that the midplane of the top ange laminate exists
at z=(z
b
t
bf
ht
tf
/2). Based on this value, expressions analogous to Eqs. (28)
(32) can be obtained for the top ange laminate. We nd that the total axial
load supported by the top ange is given by:
N
b
xx

tf
z
b
t
bf
h
t
tf
2
_ _
A
tf
E
ex
xx

tf
j
xx
33
All equations necessary to determine the position of the neutral axis
have nowbeen developed. The box-beamis modeled as four rectangular lam-
inates (two web laminates, the bottom ange laminate, and the top ange
laminate) and is in pure bending. Hence, the axial loads present in all four
laminates must sum to zero:
2N
b
xx

w
N
b
xx

bf
N
b
xx

tf
0
Substituting Eqs. (28), (32), and (33) and solving for z
b
, we obtain:
z
b

1
2
2A
w
E
ex
xx

w
h 2t
bf
t
bf
A
bf
E
ex
xx

bf
A
tf
E
ex
xx

tf
2t
b
2h t
tf

2A
w
E
ex
xx

w
A
bf
E
ex
xx

bf
A
tf
E
ex
xx

tf
_

_
_

_ 34
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 identied, the
eective exural rigidity of the composite box-beamcan be determined. First,
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
consider the moment supported by a web laminate. The incremental moment
associated with the force dN
b
xx

w
located distance z from the neutral axis is
given by:
dM
b
xx

w
dN
b
xx

w
_ _
z t
w
E
ex
xx

w
j
xx
_ _
z
2
dz
The total moment supported by a web laminate is then given by:
M
b
xx

w
m
z
b
t
bf
z
b
ht
bf
t
w
E
ex
xx

w
j
xx
_ _
z
2
dz
Evaluating this integral, we nd that the moment supported by a web
laminate is given by:
M
b
xx

t
w
E
ex
xx

w
j
xx
3
_
_
_
_
z
b
t
bf

3
z
b
h t
bf

3
_ _
35
Now consider the moment supported by the bottom ange laminate.
Since the midplane of the bottom ange laminate is not located at the neutral
axis of the overall beamcross section, the ange 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
ange is identical to that experienced by the beamas a whole: j
xx

bf
=j
xx
. The
moment resultant associated with this curvature is:
M
xx

bf
j
xx
=d
bf
11
t
3
bf
E
xx

bf
j
xx
=12
This moment resultant contributes to the total moment supported by the
bottomange. In addition, the stress resultant induced N
xx
j
bf
also contributes
to the moment supported by the bottomange and may be of equal or greater
importance. The bottom ange laminate is located at z = (z
b
t
bf
/2). Sub-
stituting this coordinate into Eq. (25), we nd that the midplane strain
induced in the bottom ange laminate is
e
o
xx

bf
z
b
t
bf
=2j
xx
This midplane strain is related to stress resultant N
xx

bf
according to:
e
o
xx

bf
a
bf
11
N
xx

bf

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.


Equating and solving for N
xx

bf
, we nd:
N
xx

bf
z
b
t
bf
=2
j
xx
a
bf
11
_ _
z
b
t
bf
=2 t
bf
E
ex
xx

bf
j
xx
_ _
This stress resultant acts at a distance (z
b
t
bf
/2) from the neutral axis, and so
the moment contributed by N
xx

bf
is obtained by multiplying by this distance.
Both N
xx

bf
and M
xx

bf
are resultants (i.e., loads per unit length) and must be
multiplied by the width of the ange to determine the bending moment. Based
on the above discussion, the total moment supported by the bottom ange
laminate is:
M
b
xx

bf
z
b
t
bf
=2
2
bt
bf
E
ex
xx

bf
_ _

bt
bf
E
xx

bf
12
_
_
_
_
j
xx
This expression can be simplied by noting that the area of the bottom ange
is A
bf
=bt
bf
and, further, that the area moment of inertia of the bottomange
(taken about the local midplane of the bottom ange) is I
bf
=bt
bf
3
/12:
M
b
xx

bf
z
b
t
bf
=2
2
A
bf
E
ex
xx

bf
_ _
I
bf
E
xx

bf
_ _
j
xx
36
In passing, it might be anticipated that Eq. (36) could be further simplied by
expressing the area moment of inertia of the bottomange 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 bottomange laminate
about the neutral axis of the beam cross section is given by:
I
bf

na
z
b
t
bf
=2
2
A
bf
I
bf
Comparing this expression with Eq. (36), it is seen that the parallel axis
theorem cannot be used to advantage because the eective extensional
modulus of the bottom ange, E
ex
xx

bf
, diers from the eective exural
modulus of the bottom ange, E
xx

bf
.
Finally, we consider the moment supported by the top ange. Following
an equivalent process, we conclude that the moment support by the top ange
laminate is given by:
M
b
xx

tf
z
b
t
bf
h t
tf
=2
2
A
tf
E
ex
xx

tf
_ _
I
tf
E
xx

tf
_ _
j
xx
37

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.


The total moment applied to the beam, M
o
b
, must equal the sum of the
moment components supported by the two web laminates, the bottom ange
laminate, and the top ange laminate:
M
b
o
2M
b
xx

w
M
b
xx

bf
M
b
xx

tf
Also, by denition, the eective exural rigidity of the box-beam is related to
M
o
b
according to:
EI
M
b
o
j
xx
Substituting Eqs. (35)(37), the eective exural rigidity of a composite box-
beam is given by:
EI
2t
w
E
ex
xx

w
3
z
b
t
bf

3
z
b
h t
bf

3
_ _

_
z
b
t
bf
=2
2
A
bf
E
ex
xx

bf
38
z
b
t
bf
h t
tf
=2
2
A
tf
E
ex
xx

tf
I
bf
E
xx

bf
I
tf
E
xx

tf
_
If a box-beam is considered for which the bottom and top ange laminates
are identical (i.e., if A
bf
= A
tf
, t
bf
= t
tf
, and E
ex
xx

bf
= E
ex
xx

tf
), then Eq. (38)
reduces to:
EI 2I
w
E
ex
xx

w
2
t
bf
h
2
A
bf
E
ex
xx

bf
4
I
bf
E
xx

bf
_
_
_
_
where I
w
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. Specically, both the ange 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 xz
plane. This implies that the two web laminates must be produced using an
identical symmetric stacking sequence, such that the set of ber angles
represented by the two web laminates is symmetric about the xz plane.
The top and bottom ange laminates must also be symmetric, but they need
not be identical. Therefore the beam cross section need not be symmetric
about the yz plane. Having dened a beamcross section in this way, then the

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.


location of the centroid (which also denes the location of the neutral axis in
pure bending) is given by Eqs. (20a) and (20b) or Eq. (34). The eective axial
rigidity of the box-beam is given by Eqs. (24a) and (24b), and the eective
exural rigidity is given by Eq. (38). As will be discussed in following sections,
once the eective axial and exural rigidities of the composite box-beamhave
been determined, then deection 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 13.
Example Problem 1
Three box-beams are constructed using unidirectional graphiteepoxy pre-
preg tape with the material properties listed in Table 3 of Chap. 3 and ply
thickness of 0.125 mm. The rst box-beamis shown in Fig. 14(a). In this case,
the web and ange 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 ange. Hence, both web laminates and
the bottom ange 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 ange 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 eective axial and exural
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 anges are produced using an 8-ply stacking
sequence, laminate thicknesses are:
t
w
t
bf
t
tf
80:125 mm 1 mm
The length of the bottom and top ange laminates equals the overall width
of the beam (30 mm). The length of the web laminates is:
h 50 mm t
bf
t
tf
50mm 21mm 48mm
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Table 1 Centroid Location of Some Thin-Walled Composite Beams with
Symmetric Web and Flange Laminates
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Table 2 Effective Axial Rigidities of Some Thin-Walled Composite Beams
with Symmetric Web and Flange Laminates
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Table 3 Effective Flexural Rigidities of Some Thin-Walled Composite Beams
with Symmetric Web and Flange Laminates
We can now calculate the area properties of the web and ange laminates:
A
w
ht
w
48 mm1 mm 48 10
6
m
2
A
bf
A
tf
bt
bf
30 mm1 mm 30 10
6
m
2
I
bf
I
tf

bt
bf

3
12

30 mm1 mm
3
12
2:5 10
12
m
4
The eective moduli for the web and ange laminates are calculated using the
process described in Sec. 9.1 of Chap. 6 and are found to be:
E
ex
xx

w
E
ex
xx

bf
E
ex
xx

tf
90:4 Gpa
E
xx

bf
E
xx

tf
120 Gpa
FromTable 1, for a composite box-beam, the location of the centroid is given
by:
z
b

1
2
2A
w
E
ex
xx

w
h 2t
bf
t
bf
A
bf
E
ex
xx

bf
A
tf
E
ex
xx

tf
2t
bf
2h t
tf

2A
w
E
ex
xx

w
A
bf
E
ex
xx

bf
A
tf
E
ex
xx

tf
_

_
_

_
Substituting all known values and completing the indicated calculations, we
nd:
z
b
25 mm
This result is as anticipated since the bottom and top ange laminates are
identical and the total depth of the beam is 50 mm.
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-beamwith additional plies added to both top
and bottom flange laminates.

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
From Table 2, for a composite box-beam, the eective axial rigidity is:
AE 2A
w
E
ex
xx

w
A
bf
E
ex
xx

bf
A
tf
E
ex
xx

tf
Substituting all known values and completing the indicated calculations,
we nd:
AE 14:110
6
N
From Table 3, for a composite box-beam, the eective exural rigidity
is:
IE
2t
w
E
ex
xx

w
3
z
b
t
bf

3
z
b
h t
bf

3
_ _

_
z
b
t
bf
=2
2
A
bf
E
bf
xx
z
b
t
bf
h t
tf
=2
2
A
tf
E
ex
xx

tf
I
bf
E
xx

bf
I
tf
E
xx

tf
_
Substituting all known values and completing the indicated calculations,
we nd:
IE
_ _
4:92 10
3
N m
2
Part (b). The web and bottom anges are produced using the same 8-ply
stacking sequence considered in Part (a), but now the top ange is a 20-ply
laminate. Adjusting our previous calculations, we nd:
t
w
t
bf
80:125 mm 1 mm
t
tf
200:125 mm 2:5 mm
h 50 mm t
bf
t
tf
50 mm 1 mm 2:5 mm 46:5 mm
A
w
ht
w
46:5 mm1 mm 46:5 10
6
m
2
A
bf
bt
bf
30 mm1 mm 30 10
6
m
2
A
tf
bt
tf
30 mm2:5 mm 75 10
6
m
2
I
bf

bt
bf

3
12

30 mm1 mm
3
12
2:5 10
12
m
4
I
tf

bt
tf

3
12

30 mm2:5 mm
3
12
39:1 10
12
m
4
E
ex
xx

w
E
ex
xx

bf
90:4 GPa

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
E
xx

bf
120 GPa
E
ex
xx

tf
83:3 GPa
E
xx

tf
94:0 GPa
Based on the appropriate expression taken from Table 1, we nd that the
centroid is located at:
z
b
29:4 mm
Note that the increased thickness of the top ange laminate has had the eect
of moving the centroid 4.4 mm towards the top ange. The new axial and
exural rigidities are:
AE 17:4 10
6
N
IE 6:35 10
3
N m
2
The increased thickness of the top ange laminate has increased the eective
axial and exural 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 anges consist of a 20-ply laminate. We nd:
t
w
80:125 mm 1 mm
t
tf
t
bf
200:125 mm 2:5 mm
h 50 mm t
bf
t
tf
50 mm 2:5 mm 2:5 mm 45 mm
A
w
ht
w
45 mm1 mm 45 10
6
m
2
A
bf
A
tf
bt
bf
30 mm2:5 mm 75 10
6
m
2
I
bf
I
tf

bt
bf

3
12

30 mm2:5 mm
3
12
39:1 10
12
m
4
E
ex
xx

w
90:4 GPa
E
ex
xx

bf
E
ex
xx

tf
83:3 GPa
E
xx

bf
E
xx

tf
94:0 GPa
Employing the appropriate expressions from Tables 13, we nd:
z
b
25:0 mm

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.


AE 20:6 10
6
N
IE 8:43 10
3
N m
2
Since the top and bottom anges 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 eective rigidities. Compared to the values calculated in Part (a),
the increased thickness of the bottom and top ange laminates has increased
the eective axial and exural rigidities by 46% and 71%, respectively.
7 STATICALLY DETERMINATE AND INDETERMINATE
AXIALLY LOADED COMPOSITE BEAMS
Deections 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
deections.
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 N
b
B
, N
b
C
, and N
b
D
. In addition, the eective axial rigidity changes dis-
cretely at points B and C. In practice, a discrete change in eective 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 (SF
x
= 0):
N
b
xx

AB
N
b
B
N
b
C
N
b
D
N
b
xx

BC
N
b
C
N
b
D
N
b
xx

CD
N
b
D
The change in length of a beam caused by axial loading is called the
elongation, u
o
. The elongation induced within each segment of the beam
shown in Fig. 15 is given by:
u
o

AB

N
b
xx

AB
L
AB
AE
xx

AB
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
u
o

BC

N
b
xx

BC
L
BC
AE
xx

BC
u
o

CD

N
b
xx

CD
L
CD
AE
xx

CD
where L
AB
, L
BC
, and L
CD
represent the length of beamsegments AB, BC, and
CD, respectively. The elongation of the bar as a whole is simply the sumof the
elongation induced over each segment:
u
o

i
N
b
xx

i
L
i
AE
xx

i
39
A statically indeterminate composite beam is shown in Fig. 16. In this
case, the beamis clamped at both ends and subjected to two known axial loads
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 R
xx
= N
B
b
+N
C
b
+N
D
b
. (c) Free-body
diagram used to determine internal load in segment BC equals N
xx
b
= N
C
b
+N
D
b
.
(d) Free-body diagram used to determine internal load in segment CD equals
N
xx
b
= N
D
b
.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
N
B
b
and N
C
b
. An unknown reaction force is induced at each end of the bar,
R
xx
b
j
A
and R
xx
b
j
D
. Note that both reaction forces have been assumed to act in
the +x-direction in Fig. 16. Enforcing the appropriate equation of equilib-
rium (SF
x
= 0), we nd:
R
b
xx

A
N
b
B
N
b
C
R
b
xx

D
0 40
Since R
xx
b
j
A
and R
xx
b
j
D
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
beammust be zero. By requiring that the total elongation is zero, we develop a
second independent equation, known as an equation of compatibility. Treating
Figure 16 A statically indeterminate composite beam subjected to two known
intermediate loads. (a) Original structure. (b) Free-body diagram.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
reaction force R
xx
b
j
D
as a redundant force, applying Eq. (39), and equating the
total elongation to zero, we nd:
u
o

AD

N
b
B
N
b
C
R
b
xx

D
_ _
L
AB
AE
xx

AB

N
b
C
R
b
xx

D
_ _
L
BC
AE
xx

BC

R
b
xx

D
_ _
L
CD
AE
xx

CD
0
41
Equations (40) and (41) represent two independent equations in two un-
knowns, R
xx
b
j
A
and R
xx
b
j
D
. 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 graphiteepoxy 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
beamhas 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 ange laminates are produced using a 20-ply [(0/90)
2s
/ F45]
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
ange laminates are produced using an 8-ply [0/90]
2s
stacking sequence. The
length of each beam segment is as follows:
L
AB
250 mm
L
BC
500 mm
L
CD
250 mm
Thus, the total length of the beam is 1.0 m. Finally, the applied loads are:
N
b
B
15 kN
N
b
C
10 kN
N
b
D
5 kN
Determine the total beam elongation produced by this loading condition.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Solution. The eective axial rigidities provided by the beam cross sections
involved were determined as a part of Example Problem 1 and found to be:
AE
xx

AB
AE
xx

BC
20:6 10
6
N
AE
xx

CD
14:1 10
6
N
The elongation can be determined through the application of Eq. (39):
u
o

i
N
b
xx

i
L
i
AE
xx

i
u
o

N
b
xx

AB
L
AB
AE
xx

AB

N
b
xx

BC
L
BC
AE
xx

BC

N
b
xx

CD
L
CD
AE
xx

CD
u
o

30000N0:25m
20:6 10
6
N

15000N0:50m
20:6 10
6
N

5000N0:25m
14:1 10
6
N
u
o
0:817 mm
Example Problem 3
Assume that the composite beam shown in Fig. 16 is a graphiteepoxy 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
beamhas 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 ange laminates are produced using a 20-ply [(0/90)
2s
/ F45]
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
ange laminates are produced using an 8-ply [0/90]
2s
stacking sequence. The
length of each beam segment is as follows:
L
AB
250 mm
L
BC
500 mm
L
CD
250 mm
Thus, the total length of the beam is 1.0 m. Finally, loads N
B
b
= 15 kN and
N
C
b
= 10 kN. Determine the reaction forces R
xx
b

A
and R
xx
b

D
.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Solution. The reaction forces are determined through solving Eqs. (40) and
(41). The eective axial rigidities provided by the beamcross sections involved
were determined as a part of Example Problem 1 and found to be:
AE
xx

AB
AE
xx

BC
20:6 10
6
N
AE
xx

CD
14:1 10
6
N
Using the values provided, Eq. (41) becomes:
25 kN R
b
xx

D
_ _
0:25 m
20:6 10
6
N

10 kN R
b
xx

D
_ _
0:5 m
20:6 10
6
N

R
b
xx

D
_ _
0:25 m
14:1 10
6
N
0
Solving for the redundant force:
R
b
xx

D
10:1 kN
The fact that R
xx
b

D
is negative implies that it acts in the direction opposite to
that shown in Fig. 16(b).
The unknown reaction force R
xx
b

A
can now be determined through the
application of Eq. (40):
R
b
xx

A
N
b
B
N
b
C
R
b
xx

D
R
b
xx

A
15 kN 10 kN 10:1 kN 0
R
b
xx

A
14:9 kN
Reaction R
xx
b

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
Deections 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. Adistributed load, q(x), is a load distributed over a length of the beam
and is specied 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 specic cross section. Apoint load is usually specied in units
of either Newtons or pounds-force (Nor lbf ). Finally, a concentrated moment
(also called a couple), M
b
o
, is a moment applied at a particular cross section
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
and is specied in units of Newton-meters or pounds-force-in. (Nmor lbf in.).
As in earlier sections, a superscript b has been used to dierentiate between
a concentrated moment and a moment resultant, M
xx
. 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 deections,
we must determine these internal forces. In order to do so, the manner in
which the beam is supported must rst be dened. The three most common
beamsupport conditions are illustrated in Fig. 18. Apinned support is one that
restrains the beam from deection in the x- and z-directions, but does not
prevent rotation. In contrast, a roller support restrains the beam from
deection in the z-direction, but does not restrain deections in the x-
direction nor does it restrict rotation. Finally, a clamped (or xed) support
does not allow the beam to rotate or deect 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. Acantilever beamis shown in
Fig. 19(b) and is dened as a beamclamped 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 17 Illustration of external transverse loads that lead to beam bending.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
third common conguration is an overhanging beam, shown in Fig. 19(c). This
condition is similar to a simply supported beamexcept that the roller support
is at an interior cross section, such that one end of the beam is free.
Having specied 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 xz plane
(the plane of loading), three equations of equilibrium are satised automati-
cally: SF
y
=SM
x
=SM
z
= 0. In this section, we do not consider any axial
loads; therefore SF
x
=0 is also satised automatically. Consequently, only
two equations of equilibrium are available for use. Specically, 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:

F
z
0

M
y
0
Beamproblems may nowbe 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 19 Common combinations of beam support conditions. (a) Simply sup-
port beam. (b) Cantilevered beam. (c) Overhanging beam.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
minate and the reaction forces are determined through the application of the
equations of equilibrium and consideration of beam deections. An example
of a statically indeterminate beamis shown in Fig. 21. In this case, the beamis
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
z
(A)
, R
z
(B)
, and M
o
b
), as indicated in the
accompanying free-body diagram. The three unknown reaction forces cannot
Figure 20 Free-body diagrams for common statically determinate beams. (a)
Simply supported beam. (b) Cantilevered beam. (c) Overhanging beam.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
be determined solely on the basis of equilibrium since only two equations of
equilibrium are available for use. The dierence 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 rst degree. One way to determine reaction
force in this case is to enforce the requirement that the beamz-deection 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 beamof interest, either through application of the equations
of equilibrium or through an appropriate combination of the equations of
equilibriumand compatibility. The reaction forces and the applied transverse
loads represent external forces applied to the beam, and these induce internal
forces within the beam. Internal forces may consist of a shear force V
xx
b
and
bending moment M
xx
b
. In general, V
xz
b
and M
xx
b
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 bendingin these earlier
sections, the internal bending moment M
xx
b
did not vary along the length of
the beam. Further, in pure bending, no shear force exists: V
xz
b
= 0. We now
wish to determine transverse beam deections under more general loading
conditions in which both V
xz
b
and M
xx
b
exist. This is often called a state of
nonuniform bending. Rigorously speaking, both V
xz
b
and M
xx
b
contribute to
transverse beam deections. However, the deections due to V
xz
b
are less
signicant than the deections due to M
xx
b
and can usually be ignored. That is,
we will assume that the contribution of shear force V
xz
b
to the total deection
of composite beams is negligibly small. In essence, we assume Secs. 5 and 6
Figure 21 Free-body diagram for a beam statically indeterminate to the first
degree.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
remain valid for the case of nonuniformbending, 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:
j
xx

M
b
xx
IE
xx
42
The eective exural 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, (IE
xx
) was developed in Sec. 6. Also, for a state of pure bending,
the midplane curvature j
xx
is related to transverse beamdeections according
to:
j
xx

d
2
w
dx
2
43
where w=w(x) represents the z-deection of the composite beammidplane.
*
Combining Eqs. (42) and (43), we nd:
d
2
w
dx
2

M
b
xx
IE
xx
44
Most textbooks devoted to mechanics of materials (Refs. [15] for example)
describe several techniques to determine transverse beam deections. 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 Castiglionos Theorem
(which is a technique based on energy methods). Two techniques will be briey
discussed here: direct integration of the governing equations and the method
of superposition.
Direct integration of the governing dierential 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
moment as a function of x: M
b
xx
= M
b
xx
(x). Hence, the beam deection 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 dierential equation with constant coecients. Consequently, the
* For pure bending, Eq. (43) is exact and is derived in any text devoted to mechanics of
materials, including Refs. [15] for example.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
deection of a beam subjected to several dierent load components acting
simultaneously can be found by superimposing (i.e., adding together) the
deections caused by each load component acting separately. In practice then,
tables of beam deections caused by simple loading conditions and for
commonly encountered beam support conditions are generated, and the
deections caused by a combination of two or more loads are obtained by
adding together the deection caused by each load component. Tables for
deections 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 deections. The deections of a statically
determinate composite beamwill be obtained by direct integration of Eq. (44)
in Example Problem 4. The method of superposition is then applied to a
statically determinate composite beamsubjected to a combination of loads in
Example Problem5. Finally, the method of superposition is used to determine
the reaction forces and deections for a statically indeterminate beam in
Example Problem 6.
Example Problem 4
A cantilevered graphiteepoxy box-beam subjected to a linearly increasing
distributed load qx q
o
=Lx is shown in Fig. 22(a):
(a) Obtain analytical expressions giving the reaction forces and
transverse beam deections.
(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 beamdeections if q
o
=200 N/m.
Solutions
Part (a). A free-body diagramof the beam is shown in Fig. 22(b). Since the
beam is clamped, two reaction forces may exist at the left-hand end: R
z
and
M
b
o
. The rst 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 R
z
and M
b
o
, we rst replace the distributed load q(x) with
the statically equivalent force q
o
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 (SF
z
=
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
SM
y
= 0), we obtain analytical expressions giving the reaction forces at the
clamped end:
R
z

q
o
L
2
M
b
o

q
o
L
2
3
a
Note that if q
o
>0, then the bending moment M
b
o
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 nd:
V
b
xz
x
q
o
2L
L
2
x
2

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) = (q
o
/L)x. (b) Free-body diagram used to determined the unknown reaction
forces (distributed load q
o
).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
A so-called shear force diagram is created by plotting V
b
xz
(x), as shown in
Fig. 24. Similarly, summing moments about an axis passing through the left-
hand side of the free-body diagramshown in Fig. 23 and parallel to the y-axis,
we nd:
M
b
xx
x
q
o
6L
2L
3
3L
2
x x
3

A so-called bending moment diagram is created by plotting M


b
xx
(x), as shown
in Fig. 24. Substituting the above expression for the bending moment M
b
xx
(x)
into Eq. (44), we nd:
d
2
w
dx
2

q
o
6LIE
xx

2L
3
3L
2
x x
3

Integrating once results in:


dw
dx

q
o
6LIE
xx

2L
3
x
3L
2
x
2
2

x
4
4
_ _
C
1
where C
1
is a constant of integration. Since the beamis clamped at the left end,
the slope must equal zero there: dw/dx =0 at x =0. Enforcing this boundary
condition, we nd that the constant of integration must equal zero:
C
1
0
Performing a second integration and simplifying, we obtain:
w
q
o
x
2
120LIE
xx

20L
3
10L
2
x x
3
C
2
Figure 23 Free-body diagram used to determine internal shear and bending
moment, V
xz
b
and M
xx
b
, respectively, acting at an arbitrary cross section located
at position x.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
where C
2
is a second constant of integration. Since the beam is clamped
at the left end, the deection must also equal zero there: w = 0 at x = 0.
Enforcing this second boundary condition, we nd that the constant of inte-
gration must again equal zero: C
2
= 0 Hence, we nd that beam deections
are given by:
w
q
o
x
2
120LIE
xx

20L
3
10L
2
x x
3
b
Note that this result is included in the list of solutions tabulated in Appendix
C, Table C.1.
Part (b). The eective exural rigidity of the box-beam shown in Fig.
14(c) was determined as a part of Example Problem 1 and was found to be:
IE
xx
8:43 10
3
N m
2
Figure 24 Shear and bending moment diagrams for the cantilevered beam
considered in Example Problem 4.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Substituting this and the other specied numerical values (L =1 m, q
o
=200
N/m) into expression (b) above, we nd that the transverse deections are
predicted to be (in millimeters):
wx 0:1977x
2
20L
3
10L
2
x x
3
_ _
A plot of these beam deections is shown in Fig. 25. Note that both the de-
ection and slope are zero at the clamped end, as dictated by this boundary
condition. A maximum deection of 2.17 mm is predicted to occur at the
free end.
Example Problem 5
Suppose the cantilevered graphiteepoxy beam considered in Example Prob-
lem4 is subjected to both a bending moment M
o
b
applied at the free end and a
linearly increasing distributed load qx q
o
=L x, as shown in Fig. 26.
Perform the following:
(a) Use the method of superposition to obtain an analytical expression
giving the predicted deection w(z).
(b) Plot numerical values of the predicted deections, assuming the
following beam length and loading:
L 1 m
q
o
200 N=m
M
b
o
40 N m
Figure 25 Deflections predicted for the graphiteepoxy composite box-beam
considered in Example Problem 4.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Solution
Part (a). Deections induced by a concentrated moment M
o
b
and distrib-
uted load q(x) when acting separately are included in Table C.1. The deec-
tions due to M
o
b
acting alone are given by:
wx

M
b
o

M
b
o
x
2
2IE
xx

The deections due to the distributed load are given by:


wx

q x

q
o
x
2
120LIE
xx

20L
3
10L
2
x x
3

Applying the principle of superposition, the beam deections caused by both


load components acting simultaneously are simply the sum of these two
solutions:
wx wx

M
b
o
wx

qx

M
b
o
x
2
2IE
xx

_ _

q
o
x
2
120LIE
xx

20L
3
10L
2
x x
3

_ _
Part (b). The distributed load q(x) will tend to deect the beamdownwards
(i.e., in the positive z-direction), whereas M
b
o
will tend to deect the beam
upwards (in the negative z-direction).
Using the values specied for beam dimensions, material properties,
and loads, deections are given (in millimeters) by:
wx 0:1977x
2
20 10x x
3
2:372x
2
which can be rearranged as:
wx 1:582x
2
1:977x
3
0:1977x
5
Figure 26 Cantilevered graphiteepoxy beamconsideredinExample Problem5.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
A plot of predicted deections is shown in Fig. 27. As before, both the
deection and slope are zero at the left end, as required by the clamped
boundary condition. A locally maximum positive (downwards) deection of
0.160 mm occurs at an axial position of x = 0.563 m. The globally maximum
deection occurs at the free end, where a negative (upward) deection 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
deections if a = 3L/4 = 0.75 m and P = 1000 N.
Solution
Part (a). The beam shown in Fig. 21 is statically indeterminate to the rst
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
the reaction forces (R
A
z
, R
B
z
, or M
b
o
) can be treated as the redundant force.
For present purposes, R
B
z
will be treated as the redundant.
This problem can be solved based on the method of superposition, as
summarized in Fig. 28. The rst step is to determine the beamdeections that
would occur if the redundant force was removed. Removal of redundant
reaction force R
B
z
implies that the roller support at the right end is removed,
as shown in Fig. 28(b). The process of obtaining beam deections when the
redundant force is removed is often called the reduced problem. FromAppen-
Figure 27 Deflections predicted for the beamconsidered in Example Problem5.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
dix C Table C.1 we nd that beam deections over the range a V x V L
associated with the reduced problem are given by:
wx

reduced

Pa
2
6IE
xx

3x a
The next step is to determine beam deections that would be caused if
only the redundant force was applied, as shown in Fig. 28(c). FromTable C.1,
we nd that the beam deections associated with the redundant force R
B
z
are given by:
wx

R
B
z

R
B
z
x
2
6IE
xx

3L x
This expression is valid for any axial position 0 V x V L.
We now superimpose the deections associated with the reduced and
redundant problems. In this case, we have, for a V x V L:
wx wx

reduced
wx

R
B
z

1
6IE
xx

Pa
2
3x a R
B
z
x
2
3L x
_ _
a
We require that the total beam deection equals zero at the right end
(at x=L) since the beam is supported by a roller at that point. Hence:
wx L
1
6IE
xx

Pa
2
3L a R
B
z
L
2
3L L
_ _
0
which reduces to:
2R
B
z
L
3
Pa
2
a 3L b
Figure 28 Summary of the method of superposition applied to the indetermi-
nate beam considered in Example Problem 4.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Expression (b) represents an equation of compatibility. Only one such equa-
tion is necessary in this problem since the beam is indeterminate to the rst
degree.
Next, we enforce the equations of equilibrium, SF
z
= 0 and SM
y
= 0,
resulting in (respectively):
R
A
z
R
B
z
P c
R
B
z
L M
b
o
Pa d
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:
0 2L
3
0
1 1 0
0 L 1
_

_
_

_
R
A
z
R
B
z
M
b
o
_

_
_

Pa
2
a 3L
P
Pa
_

_
_

_
Solving this system of equations results in:
R
A
z

P
2L
3
a
3
2L
3
3a
2
L e
R
B
z

Pa
2
2L
3
3L a f
M
b
o

Pa
2L
2
a
2
2L
2
3aL g
Part (b). Based on the specied numerical values, the reaction forces are:
R
A
z

1000N
21m
3
0:75m
3
21m
3
30:75m
2
1m
_ _
367N
R
B
z

1000N0:75m
2
21m
3
31m 0:75m 633N
M
b
o

1000N0:75m
21m
2
0:75m
2
21m
2
30:75m1m
_ _
117Nm
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Note that the calculated values for R
A
z
, R
B
z
, and M
b
o
are all algebraically
negative, which indicates that these reaction forces are acting in a sense
opposite to that shown in Fig. 21.
To plot beamdeections, note that expression (a), developed in the rst
part of this example problem, gives the deection over the range a V x V L.
Hence, we need an additional expression for beam deections that is valid
over 0 Vx Va. From Table C.1, we nd that beamdeections over the range
0 V x V a associated with the reduced problem are given by:
wx

reduced

Px
2
6IE
xx

3a x
As before, from Table C.1, we nd that the beam deections associated
with the redundant force R
z
(B)
are given by:
wx

R
B
z

R
B
z
x
2
6IE
xx

3L x
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 deection, valid for
0 V x V a:
wx wx

reduced
wx

R
B
z

1
6IE
xx

Px
2
3a x R
B
z
x
2
3L x
_ _
h
Figure 29 Deflections predicted for the beamconsidered in Example Problem6.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Recall from Example Problem 1 that the eective exural rigidity for this
beam is ( IE
xx
) = 8.4310
3
N m
2
. Substituting this value (and the other
numerical values specic to this problem) into expressions (a) and (h) and
plotting over the appropriate ranges, results in Fig. 29. As dictated by the
specied boundary conditions, the beam deection is zero at either end of the
beam, and the slope of the beam is zero at the left end. A maximum positive
(downward) deection of 0.94 mm is predicted to occur at the axial location
x = 0.64 m.
9 COMPUTER PROGRAM BEAM
The computer programBEAMhas beendevelopedtosupplement 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,
eective axial rigidity, and eective exural rigidity of a composite beam
with either a rectangular cross section or with any of the thin-walled cross
sections shown in Tables 13. The program prompts the user to input all
information necessary to perform these calculations. Properties of up to ve
dierent materials may be dened. 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 graphiteepoxy:
E
11
170 10
9
Pa E
22
10 10
9
Pa v
12
0:30
G
12
13 10
9
Pa
Since 1 Pa = 1 N/m
2
, all lengths must be input in meters. For example, ply
thicknesses must be input in meters (not millimeters). A typical value would
be t
k
=0.000125 m(corresponding to a ply thickness of 0.125 mm). Similarly,
if an I-beam that involves a 50-mm-wide ange laminate was under consid-
eration, then the width of the ange 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 graphiteepoxy material system:
E
11
25:0 10
6
psi E
22
1:5 10
6
psi v
12
0:30
G
12
1:9 10
6
psi
All lengths would be input in inches. A typical ply thickness would be
t
k
= 0.005 in., and for an I-beam, a typical ange width would be 2.0 in.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Having determined the eective axial and exural rigidities of a
composite beam of interest, then beam deections 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 [(0
2
/F30/F45/F60)
3
/90
2
]
s
composite beam with rectangular cross
section is shown in Fig. 30. Assume this laminate is produced using a
graphiteepoxy 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 eective axial
rigidity of the beam, (ii) the eective exural rigidity of the
beam when ply interfaces are orthogonal to the plane of
loading [Fig. 30(a)], and (iii) the eective exural 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 eective axial and
exural 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.
3. Repeat Problem 1, except assume that the beam is produced using a
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 graphiteepoxy 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 graphiteepoxy 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 ange laminate, web laminate, and bottom ange laminate are
[(0
2
/90
2
/0
2
)
2
/45
2
/0/45
2
]s, [0
2
/90
2
/0
2
]
2s
, and [(0
2
/90
2
/0
2
)
2
/45
2
/0/45
2
]
s
,
respectively (note that Fig. 31 implies that the web plies extend into the
top and bottom ange laminates).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
(a) Determine the centroid location, eective axial rigidity, and
eective exural rigidity of the beam using hand calculation
and appropriate expressions from Tables 13. Determine the
eective elastic moduli involved using program CLT.
(b) Use the program BEAM to calculate the centroid location,
eective axial rigidity, and eective exural rigidity of the
beam, and compare these results with your hand calculations
obtained in step (a).
Figure 30 Composite beam with rectangular cross section described in
Problem 1 (not all ply interfaces shown).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
6. Repeat Problem 5, except assume that the beam is produced using a
glass/epoxy pre-preg with a thickness of 0.150 mm.
7. Repeat Problem 5, except assume that the beam is produced using a
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 ange laminates are [(0
2
/90
2
/0
2
)
2
/45
2
/0/45
2
]
s
and [0
2
/90
2
/
0
2
]
2s
, respectively.
9. Refer to Example Problem4, part (b). Suppose the maximumdeection
of the cantilevered beam must be reduced by a factor of 2. That is, the
maximum deection must be reduced to 1.085 mm (or less). Deections
will be reduced by adding 0j plies to the top and bottomange laminates
of the box-beam. The stacking sequence of the ange laminates will then
be of the type [(0/90)
2s
/F45/0
n
]
s
. Determine the value of n necessary to
reduce deections to the desired level.
10. Refer to Example Problem 5, what maximum deection occurs if a
negative bending moment is applied, i.e., if M
o
b
= 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.
Figure 31 Composite I-beam described in Problem 5 (not all ply interfaces
shown).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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
deections induced in a thin plate by varying loads we must rst 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 simplication
of the governing equations. This limitation is reasonable, as most composites
used in practice are symmetric.
1 PRELIMINARY DISCUSSION
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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 expose 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 dierent 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. Specically, 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
(r
zz
=s
xz
=s
yz
=0), and

The Kirchho 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
N
xx
, N
yy
, N
xy
, M
xx
, M
yy
, M
xy
) 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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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.
Specically, 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 denition 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. Asketch
showing stress resultants N
xx
acting on opposite edges of a thin laminate is
presented in Fig. 1a. Because N
xx
is a load applied to the two plate edges that
are parallel to the y-axis, N
xx
cannot vary as a function of x. Therefore along
the plate edge the externally applied resultant N
xx
is either a constant or, at
most, a function of y only:
N
xx
N
xx
y:
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 N
xx
( y) applied to the negative laminate edge (i.e., the edge whose
outward normal points in the negative x-direction) and the resultant
N
xx
( y) applied to the positive laminate edge. Therefore the load applied
to the negative x-edge will be labeled N
xx
(x)
( y), whereas the load applied to
the positive x-edge will be labeled N
xx
(+x)
( y). Although not shown in Fig. 1a,
stress resultant N
xy
and moment resultants M
xx
and M
xy
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
negative x-edge:
Resultants acting on the
positive x-edge:
N
xy
(x)
( y) N
xy
(+x)
( y)
M
xx
(x)
( y) M
xx
(+x)
( y)
M
xy
(x)
( y) M
xy
(+x)
( y)
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 N
yy
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 N
yx
, M
yy
,
Figure 1 Illustration of varying stress resultants acting on opposite edges of a
thin plate. (a) Variation of stress resultants N
xx
(+)
(y). (b) Variation of stress
resultants N
yy
(+)
(x).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
and M
yx
also act on these plate edges. As before, we must distinguish between
loads applied to the negative and positive y-edge:
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 N
xy
=N
yx
. We
can no longer make this assumption. For example, the variation of the shear
stress resultant over the nite length of the positive x-face of the laminate,
N
xy
(+x)
( y), is independent of the variation of the shear stress resultant applied
over the nite length of the positive y-face, N
yx
(+y)
(x). Similarly, the variation
of the twisting moment applied to the positive x-face, M
xy
(+x)
( y), is now
independent of the twisting moment applied to the positive y-face, M
yx
(+y)
(x).
Figure 2 Illustration of varying transverse load q(x,y) acting over the surface of
a thin plate and perpendicular to the xy plane and the resulting out-of-plane
shear stress resultants V
xz
(Fx)
( y) and V
yz
(Fy)
(x).
Resultants acting on the
negative y-edge:
Resultants acting on the
positive y-edge:
N
yy
(y)
(x) N
yy
(+y)
(x)
N
yx
(y)
(x) N
yx
(+y)
(x)
M
yy
(y)
(x) M
yy
(+y)
(x)
M
yx
(y)
(x) M
yx
(+y)
(x)
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Exceptions to this statement occur at the four corners of the laminate. For
example, at the corner (x=a, y=b):
N
x
xy
b N
y
yx
a and M
x
xy
b M
y
yx
a
A distributed transverse force acting over the surface of the laminate,
q(x,y), is showninFig. 2. The transverse force q(x,y) has units of force per area,
such as N/m
2
=Pascals or lbf/in
2
=psi. Because q(x,y) is a distributed force
acting in the z-direction, shear stress resultants V
xz
( y) and/or V
yz
(x) must be
present along one or more edges of the laminate, so as to maintain static
equilibrium (i.e., to maintain AF
z
=0). V
xz
(Fx)
( y) and V
yz
(Fy)
(x) are given by:
V
Fx
xz

_
t=2
t=2
s
Fx
xz
dz V
Fy
yz

_
t=2
t=2
s
Fy
yz
dz 1
The reader should immediately object to the inclusion of q(x,y), V
xz
(Fx)
( y),
and V
Fy
yz
x in our analysis. After all, they violate our assumption of plane
stress. That is, the presence of q(x,y) implies that stress r
zz
will be induced in
the plate, and furthermore Eq. (1) implies that shear stresses s
xz
and s
yz
will
also be induced. These objections are entirely valid. If a transverse load q(x,y)
is applied then a 3-Dstate 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: r
zz
, s
xz
, s
yz
<<r
xx
, r
yy
, s
xy
.
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 ve 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 specied along the edges of a rectangular laminate. Specied 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
specied boundary conditions. Rather, in many instances midplane displace-
ments along the plate edges are known. Specied midplane displacements
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 species that a beam end is clamped, one has
specied something about the displacement eld in the beam, rather than
something about the applied loading.
Similar displacement boundary conditions can also be dened for thin
plates. Hence to dene a thin-plate problem of interest we must specify the
boundary conditions along each edge of the plate. The specied 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 probleminvolving
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. Diculties in obtaining exact
solutions occur for either isotropic plates or anisotropic composite plates,
and arise from two dierent sources. First, the boundary conditions encoun-
Figure 3 A thin plate subjected to stress resultants and transverse loading that
vary along the length and width of the plate.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
tered in a problemmay preclude obtaining an exact solution. For example, an
exact solution can be found for a at rectangular isotropic plate subjected to
transverse loading if all four edges are simply supported, but an exact solution
cannot be found for an identical at rectangular isotropic plate subjected to
transverse loading if all four edges are clamped. Because boundary conditions
are independent of material properties, the diculties in obtaining exact
solutions for certain boundary conditions are encountered for either isotropic
or anisotropic composite plates.
The second source of diculty has to do with the anisotropic nature of
composites and is not encountered during the study of isotropic plates.
Specically, exact solutions cannot be found for laminates that exhibit the
following couplings terms:

in-plane normal-shear coupling (i.e., laminates for which A


16
,
A
26
p 0),

bending-twisting coupling (i.e., laminates for which D


16
, D
26
p 0),
and/or

coupling between in-plane loads and out-of-plane twisting (i.e.,


laminates for which B
16
, B
26
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 B
ij
, M
ij
T
, M
ij
M
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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 rst 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 dierential equations involving the internal stress and moment
resultants, transverse loading q(x,y), and the out-of-plane displacement eld,
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 elds, u
o
(x,y), v
o
(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 at prior to application
of external edge loads or displacement. In particular, the laminate is assumed
to be symmetric and hence B
ij
=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 (M
ij
T
=M
ij
M
=0). An innitesimal
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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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
innitesimal 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 innitesimal 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 innitesimal
length and width of the element. For example, the stress resultant N
xx
* acting
on the negative x-face of the element is assumed to vary slightly over distance
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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 innitesimal element equate to zero:

F
x

F
y

F
x

M
x

M
y

M
z
0:
Let us rst sumforces 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 gure. The element has been deected out of the
original xy plane by the transverse loading. A view of the xy plane is shown
in Fig. 5a, while a view of the xz plane is shown in Fig. 5b. In the deected
condition the left and right side of the element form angles a and (a+(Ba/Bx)
dx), respectively, with the original xy plane. Consider the x-directed force
associated with N
xx
* , acting on the left side of the element. Recall that the units
of a stress resultant are (force/length), so to convert N
xx
* 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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
multiply it by the distance over which it acts, namely, by distance dy. Fig. 5b
shows that the line-of-action of N
xx
* is not precisely parallel to the x-axis, due
to the out-of-plane deection. The component of the force associated with
N
xx
* that is parallel to the x-axis is therefore given by:
N*
xx
dy cosa
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: cosa
c
1. With this
assumption the component of the force associated with N
xx
* that is parallel
to the x-axis reduces to:
N*
xx
dy
In eect, invoking the small-angle approximation implies that we will ignore
the fact that N
xx
* 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:


N
xx
*
BN
xx
*
Bx
dx
_ _
dy

x-directed force acting on the negative y-face:


N
xy
*dx

x-directed force acting on the positive y-face:


N
xy
*
BN
xy
*
By
dy
_ _
dx
Figure 6 Free-body diagram of an infinitesimal element, showing only those
resultants that contribute toward forces in the y-direction: (a) xy plane; (b) yz
plane.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Summing the force components acting on all four sides and equating to zero
(SF
x
=0):
N
xx
* dy N
xx
*
BN
xx
*
Bx
dx
_ _
dy N
xy
*
_ _
dx N
xy
*
BN
xy
*
By
dy
_ _
dx 0
Simplifying, we obtain:
BN
xx
*
Bx

BN
xy
*
By
0 2a
Eq. (2a) is the rst equation of equilibrium. It shows that if N
xx
* 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 xy plane
of the element is shown in Fig. 6a while the yz plane is shown in Fig. 6b.
Summing forces in the y-direction, invoking the small angle approximation,
and equating to zero (SF
y
=0), we nd:
N
yy
*
_
dx N
yy
*
BN
yy
*
By
dy
_ _
dx N
xy
*
_
dy N
xy
*
BN
xy
*
Bx
dx
_ _
dy 0
which simplies to:
BN
yy
*
By

BN
xy
*
Bx
0 2b
Eq. (2b) is the second equation of equilibrium.
Let us now consider forces parallel to the z-axis. Due to out-of-plane
deection 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 N
xx
* that is parallel to the z-axis is given by:
N
xx
* dy sina
The small angle approximation (where a is expressed in radians) is:
sina
c
a
c
Bw
Bx
Hence, the z-directed component of the force associated with N
xx
* is:
N
xx
* dy
Bw
Bx
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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:


N
xx
*
BN
xx
*
Bx
dx
_ _
Bw
Bx

B
2
w
Bx
2
dx
_ _
dy

z-directed force acting on the negative y-face:


N
xy
*
_ _
dx
Bw
Bx

z-directed force acting on the positive y-face:


N
xy
*
BN
xy
*
By
dy
_ _
Bw
Bx

B
2
w
BxBy
dy
_ _
dx
Adding these results, the resultants shown in Fig. 5 result in a component of
force in the z-direction given by:
N*
xx

Bw
Bx
dy N*
xx

BN*
xx
Bx
dx
_ _
Bw
Bx

B
2
w
Bx
2
dx
_ _
dy
N*
xy
_ _
Bw
Bx
dx N*
xy

BN*
xy
By
dy
_ _
Bw
Bx

B
2
w
BxBy
dy
_ _
dx
When the algebra indicated is completed, two higher-order terms appear:
BN*
xx
Bx
_ _
B
2
w
Bx
2
_ _
dx
2
dy and
BN*
xy
By
_ _
B
2
w
BxBy
_ _
dx dy
2
Neglecting these higher-order terms, the above expression simplies to:
N*
xx
B
2
w
Bx
2

BN*
xx
Bx
Bw
Bx
N*
xy
B
2
w
BxBy

BN*
xy
By
Bw
Bx
_ _
dxdy a
Following an identical procedure, the stress resultants shown in Fig. 6
represent a force in the z-direction given by:
N*
yy
B
2
w
By
2

BN*
yy
By
Bw
By
N*
xy
B
2
w
BxBy

BN*
xy
Bx
Bw
By
_ _
dxdy b
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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:
V
xz
* dy V
xz
*
BV
xz
*
Bx
dx
_ _
dy V
yz
*
_
dx V
yz
*
BV
yz
*
By
dy
_ _
dx
q x; y dxdy
which simplies to:
BV
xz
*
Bx

BV
yz
*
By
q x; y
_ _
dxdy c
We are now ready to sum all forces in the z-direction. Adding expressions (a),
(b), and (c), equating the resulting sum to zero (SF
z
=0), and rearranging,
there results:
N
xx
*
B
2
w
Bx
2
N
yy
*
B
2
w
By
2
2N
xy
*
B
2
w
BxBy

BN
xx
*
Bx

BN
xy
*
By
_ _
Bw
Bx

BN
yy
*
By

BN
xy
*
Bx
_ _
Bw
By

BV
xz
*
Bx

BV
yz
*
By
q x; y 0
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:
N
xx
*
B
2
w
Bx
2
N
yy
*
B
2
w
By
2
2N
xy
*
B
2
w
BxBy

BV
xz
*
Bx

BV
yz
*
By
q x; y 0 2c
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-
gramshowing the resultants that contribute to the moment about the x-axis is
shown in Fig. 7a. Summing moments and equating to zero (SM
x
=0), we
have:
M
yy
*
_
dx M
yy
*
BM
yy
*
By
dy
_ _
dx M
xy
*
_
dy M
xy
*
BM
xy
*
Bx
dx
_ _
dy
V
yz
*
BV
yz
*
By
dy
_ _
dxdy V
xz
*
BV
xz
*
Bx
dx
_ _
dy
dy
2
_ _
V
xz
* dy
dy
2
_ _
q x; y dxdy
dy
2
_ _
0
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
which simplies to:

BM
yy
*
By
dxdy
BM
xy
*
Bx
dxdy V
yz
* dxdy
BV
yz
*
By
dx dy
2

BV
xz
*
Bx
dx
2
_ _
dy
2

q x; y
2
dx dy
2
0
The last three terms in this equation contain the higher-order factor
(dy)
2
. Hence dropping the last three terms and simplifying, we obtain:
V
yz
*
BM
xy
*
Bx

BM
yy
*
By
3a
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 (SM
y
= 0), we have:
M
xx
* dy M
xx
*
BM
xx
*
Bx
dx
_ _
dy M
xy
*
_
dx M
xy
*
BM
xy
*
By
dy
_ _
dx
V
xz
*
BV
xz
*
Bx
dx
_ _
dy dx V
yz
*
BV
yz
*
By
dy
_ _
dx
dx
2
_ _
V
yz
*
_
dx
dy
2
_ _
q x; y dxdy
dx
2
_ _
0
Upon completing the algebra indicated, a negligible higher-order term
appears (in this case dx
2
). Neglecting all terms that include this factor and
simplifying, we obtain:
V
xz
*
BM
xx
*
Bx

BM
xy
*
By
3b
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
(SM
z
=0), we obtain:
N
xx
*dy
dy
2
_ _
N
xx
*
BN
xx
*
Bx
dx
_ _
dy
_ _
dy
2
_ _
N
xy
*
BN
xy
*
Bx
dx
_ _
dy
_ _
dx N
xy
*
BN
xy
*
By
dy
_ _
dx
_ _
dy
N
yy
*
BN
yy
*
By
dy
_ _
dx
_ _
dx
2
_ _
N
yy
* dx
_ _
dx
2
_ _
0
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Simplifying:

BN
xx
*
Bx
dxdy
2
2

BN
xy
*
Bx
dx
2
dy
BN
xy
*
By
dxdy
2
N
xy
* dx
2
dy dxdy
2
_ _

BN
yy
*
By
dx
2
dy
2
0
All terms in this equation contain a higher-order term (either dx
2
or dy
2
).
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 ve useful equations through a
summation of forces and moments. The rst 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 fth 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, V
xz
* and V
yz
* , to the moment resultants M
xx
* , M
yy
* and M
xy
* , and
that the shear stress resultants also appear in Eq. (2c). This means that only
three of the ve 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:
N
xx
*
B
2
w
Bx
2
N
yy
*
B
2
w
By
2
2N
xy
*
B
2
w
BxBy

B
2
M
xx
*
Bx
2

B
2
M
yy
*
By
2
2
B
2
M
xy
*
BxBy
4
q x; y 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 SF
i
=0 and SM
i
=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:
BN
xx
*
Bx

BN
xy
*
By
0 5a
BN
yy
*
By

BN
xy
*
Bx
0 5b
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
N
xx
*
B
2
w
Bx
2
N
yy
*
B
2
w
By
2
2N
xy
*
B
2
w
BxBy

B
2
M
xx
*
Bx
2

B
2
M
yy
*
By
2
2
B
2
M
xy
*
BxBy
5c
q x; y 0
These results have been developed strictly on the basis of static
equilibrium. Note that we have not specied 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, dened by the slope of the deected 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 simplied for cases in
which the loads applied to the plate result in a maximum out-of-plane
deection less than about half the laminate thickness: w(x,y)j
max
< 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 (N
xx
* ,
N
yy
* , and N
xy
* ) can be neglected. If the in-plane resultants are ignored during
the summation of forces in the z-direction [setting (N
xx
* dy)Bw/Bx=0, for
example] and the preceding derivation is repeated, the equations of equili-
brium simplify as follows:
BN
xx
*
Bx

BN
xy
*
By
0 6a
BN
yy
*
By

BN
xy
*
Bx
0 6b
B
2
M
xx
*
Bx
2

B
2
M
yy
*
By
2
2
B
2
M
xy
*
BxBy
q x; y 0 6c
Eqs. (6a)(6c) are valid if the maximumout-of-plane displacement is less than
about half the plate thickness: w(x,y)j
max
< 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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
relationship is summarized by Eq. (44) in Chapter 6. We have since limited our
discussion to symmetric laminates (B
ij
=M
T
ij
=M
ij
M
) and are now using the
superscript * to denote internal stress and moment resultants. Conse-
quently, Eq. (44) is modied slightly to become:
N*
M*
_ _

A 0
0 D
_ _
e
o
j
_ _

N
T
0
_ _

N
M
0
_ _
7
In this equation:

N* and M* represent internal stress and moment resultants


associated with externally applied forces,

N
T
represents thermal stress resultants associated with a uniform
through-thickness change in temperature, DT, and

N
M
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, N
T
and N
M
should be
viewed as constants, i.e., they are not functions of x, y, or z.
Expanding the rst of the six equations represented by Eq. (7), we have:
N
xx
* A
11
e
o
xx
A
12
e
o
yy
A
16
c
o
xy
N
T
xx
N
M
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 nite strains and innitesimal strains was
made at that point. Basically, strains can be considered to be innitesimal
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 deected 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 innitesimal strains. As we will see, this assumption will
ultimately lead to the conclusion that in-plane displacement elds u
o
(x,y) and
v
o
(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 elds and/or
in-plane forces will change, reecting 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 eect,
the assumption of innitesimal strains has eliminated (in a mathematical
sense) the coupling between transverse loads and in-plane displacements/
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 innitesimal and can
therefore be expressed in terms of midplane displacement elds in accordance
with Eqs. 49(a)(49f) of Chapter 2:
e
o
xx

Bu
o
Bx
e
o
yy

Bv
o
By
c
o
xy

Bu
o
By

Bv
o
Bx
j
xx

B
2
w
Bx
2
j
yy

B
2
w
By
2
j
xy
2
B
2
w
BxBy
Therefore, the internal stress resultant N
xx
* can be written:
N
xx
* A
11
Bu
o
Bx
A
12
Bv
o
By
A
16
Bu
o
By

Bv
o
Bx
_ _
N
T
xx
N
M
xx
Given this result, the quantity
BN
xx
*
Bx
_ _
is easily obtained:
BN
xx
*
Bx
A
11
B
2
u
o
Bx
2
A
12
B
2
v
o
BxBy
A
16
B
2
u
o
BxBy

B
2
v
o
Bx
2
_ _
Note that this quantity appears as a term in the rst equation of equilibrium,
Eq. (5a). Also note that neither the thermal nor moisture stress resultant is
involved in this result, because as previously noted N
T
and N
M
are constants
and therefore BN
T
xx
=Bx BN
M
xx
=Bx 0. Following an analogous procedure,
the second term in Eq. (5a) is found to be:
BN
xy
*
By
A
16
B
2
u
o
BxBy
A
26
B
2
v
o
By
2
A
66
B
2
u
o
By
2

B
2
v
o
BxBy
_ _
Adding these two results, Eq. (5a) can be written in terms of elements of the
[ABD] matrix and midplane displacements as follows:
A
11
B
2
u
o
Bx
2
A
12
A
66

B
2
v
o
BxBy
2A
16
B
2
u
o
BxBy
A
16
B
2
v
o
Bx
2
A
26
B
2
v
o
By
2
A
66
B
2
u
o
By
2
0
8a
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Following an identical procedure, the second and third equations of equilib-
rium [Eqs. (5b) and (5c)] can be written:
A
16
B
2
u
o
Bx
2
A
12
A
66

B
2
u
o
BxBy
2A
26
B
2
v
o
BxBy
A
26
B
2
u
o
By
2
A
22
B
2
v
o
By
2
A
66
B
2
v
o
Bx
2
0
8b
A
11
Bu
o
Bx
A
12
Bv
o
By
A
16
Bu
o
By

Bv
o
Bx
_ _
N
T
xx
N
M
xx
_ _
B
2
w
Bx
2
_ _
A
12
Bu
o
Bx
A
22
Bv
o
By
A
26
Bu
o
By

Bv
o
Bx
_ _
N
T
yy
N
M
yy
_ _
B
2
w
By
2
_ _
8c
2 A
16
Bu
o
Bx
A
26
Bv
o
By
A
66
Bu
o
By

Bv
o
Bx
_ _
N
T
xy
N
M
xy
_ _
B
2
w
BxBy
_ _
D
11
B
4
w
Bx
4
4D
16
B
4
w
Bx
3
By
4D
26
B
4
w
BxBy
3
2 D
12
2D
66

B
4
w
Bx
2
By
2
D
22
B
4
w
By
4
q x; y 0
Eqs. (8a)(8c) represent the equations of equilibrium for a symmetric
laminate in terms of the [ABD] matrix, transverse loading, and midplane
displacement elds. 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)j
max
< t/2] then the equations of equilibrium given
by Eqs. (6a)(6c) are applicable, which become:
A
11
B
2
u
o
Bx
2
A
12
A
66

B
2
v
o
BxBy
2A
16
B
2
u
o
BxBy
A
16
B
2
v
o
Bx
2
A
26
B
2
v
o
By
2
A
66
B
2
u
o
By
2
0
9a
A
16
B
2
u
o
Bx
2
A
12
A
66

B
2
u
o
BxBy
2A
26
B
2
v
o
BxBy
A
26
B
2
u
o
By
2
A
22
B
2
v
o
By
2
A
66
B
2
v
o
Bx
2
0
9b
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
D
11
B
4
w
Bx
4
4D
16
B
4
w
Bx
3
By
4D
26
B
4
w
BxBy
3
2 D
12
2D
66

B
4
w
Bx
2
By
2
D
22
B
4
w
By
4
q x; y 0
9c
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 ab, 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 ve 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 innitesimal 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 simplied 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 elds, summarized as Eqs. (8a)(8c) and Eqs. (9a)
(9c).
In this section we will formally dene the boundary conditions of the
plate. That is, we wish to precisely dene 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 specied 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 u
o
(x,y), v
o
(x,y), and w(x,y), respectively, as in earlier chapters.
Hence, two categories of boundary conditions can be dened. We can either
specify components of the edge displacements, or we can specify components
of the edge loads. Boundary conditions involving specied displacements are
called geometric (or kinematic) boundary conditions, while boundary con-
ditions involving specied 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 u
o
(+x)
( y),
v
o
(+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 specied only one of several possible gra-
dients in the three displacement elds. Initially, one might suspect that, for
reasons of mathematical symmetry perhaps, other displacement eld gra-
dients should also be specied. For example, perhaps Bu
o
(+x)
( y)/By or
Bv
o
(+x)
( y)/Bz should be specied along x=a as well. These other gradients
need not be considered for two reasons. First, in thin plate theory the
boundary conditions are dened for the plate edge at the midplane. By
denition the midplane has a thickness of zero. This fundamental denition
precludes any midplane boundary conditions with a z-dependency; i.e., the
midplane displacement eld gradients Bu
o
(+x)
(y)/Bz, Bv
o
(+x)
(y)/Bz, or
Bw
(+x)
( y)/Bz are undened. This leaves us with possible gradients in the x-
or y-directions. With the exception of (Bw
(+x)
( y)/Bx), a specied gradient in
geometric boundary conditions along a plate edge in the x- or y-directions
results in an over-specied problem. For example, having specied displace-
ments in the x-direction along the edge x=a, namely, u
o
(+x)
( y), the value
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
of Bu
o
(+x)
( y)/By is also specied 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 u
o
(x,y), v
o
(x,y), and w(x,y), the geometric boundary conditions
that may be specied along each of the four edges of the plate are:
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 ve external stress and moment
resultants may be specied along this edge: N
xx
(+x)
( y), N
xy
(+x)
( y), V
xz
(+x)
( y), M
xx
(+x)
( y), and M
xy
(+x)
( y). Values for N
xx
(+x)
( y), N
xy
(+x)
( y), and M
xx
(+x)
( y) can be specied independently, without consideration of any other stress
or moment results. However, the remaining two resultants, V
xz
(+x)
( y) and
M
xy
(+x)
( y), are not independent and must be considered together. The in-
terdependence between V
xz
(+x)
( y) and M
xy
(+x)
( y) was rst noted by Kirchho
in 1850 and occurs because (as far as static equilibrium is concerned) the
twisting moment M
xy
(+x)
( 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
M
xy
(+x)
( y)dy, whereas the element on the left is subjected to a twisting moment
[M
x
xy
(y)+(dM
x
xy
y/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
For edge x=a: For edge x=0:
u
0
a; y u
x
o
y u
0
0; y u
x
o
y
v
0
a; y v
x
o
y v
0
0; y v
x
o
y
Bwa; y
Bx

Bw
x
y
Bx
Bw0; y
Bx

Bw
x
y
Bx
wa; y w
x
y w0; y w
x
y
For edge y=b: For edge y=0:
u
0
x; b u
y
o
x u
0
x; 0 u
y
o
x
v
0
x; b v
y
o
x v
0
x; 0 v
y
o
x
Bwx; b
By

Bw
y
x
By
Bwx; 0
By

Bw
y
x
By
wx; b w
y
x wx; 0 w
y
x
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
magnitude M
xy
(+x)
( y), acting in the directions shown and separated by
distance dy. Similarly, the element on the left is subjected to two vertical
shear resultants of magnitude [M
x
xy
(y)+(dM
x
xy
( 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 innitesimal elements,
we obtain {[M
x
xy
( y)+(dM
x
xy
( y)/dy)dy]M
x
xy
( y)}=(dM
x
xy
( y)/dy)dy.
Hence the variation of the twist moment M
xy
(+x)
( y) along the edge x=a is
statically equivalent to a vertical shear resultant of magnitude dM
x
xy
( y)/dy.
Adding this statically equivalent vertical shear resultant component to the
external vertical shear that is actually applied to the edge, V
xz
(+x)
( y), we dene
the Kirchho shear resultant acting along the edge x=a:
V
xK
xz
y V
x
xz
y
dM
x
xy
y
dy
10
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).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
The superscript +xK is used to denote the Kirchho shear resultant
(sometimes called the eective shear resultant) acting along the positive
x-edge.
Now, the equations of equilibrium (developed in Sec. 2) must be
satised at all points within the laminate, including the laminate edge. There-
fore the distribution of the external vertical shear resultant, V
xz
(+x)
( y), must
satisfy Eq. (3b). Along the edge x=a this equation becomes:
V
x
xz

BM
x
xx
Bx

BM
x
xy
By
Recalling that all boundary conditions along the positive x-edge are either
constant or are functions of y only, it must be that dM
x
xx
/dx=0. Therefore
along the edge x=a:
V
x
xz

BM
x
xy
By
11
Eq. (11) shows that the variation of the externally applied shear V
xz
(+x)
( y)
along the edge x=a is intimately related to the variation of the externally
applied twist moment M
xy
(+x)
( y) along this same edge, and that the functional
form of these two resultants cannot be specied independently.* Suppose, for
example, that the laminate is subjected to a vertical shear resultant that varies
linearly over the edge x=a. That is, suppose V
xz
(+x)
( y)=Ay+B, where Aand
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.
Furthermore, the twist moment must vary according to dM
x
xy
( y)/dy=
Ay+B, which implies that M
x
xy
y A=2 y
2
By C , where C is a
constant of integration.
As a second case of interest, suppose that V
xy
(+x)
is constant, i.e., assume
A=0 and hence that V
xz
(+x)
( y)=B. In this case the rate of change in the twist
moment must equal dM
x
xy
( y)/dy=B; otherwise, Eq. (11) is not satised 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
constant twist moment M
xy
(+x)
=B is applied, then the vertical shear must
equal zero: dM
x
xy
( y)/dy=V
x
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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Substituting Eq. (11) into Eq. (10), we nd that the eective Kirchho
shear resultant acting along the edge x=a can be written in two equivalent
ways:
V
xK
xz
y V
x
xz
y
dM
x
xy
y
dy
or
V
xK
xz
y 2
dM
x
xy
y
dy
12
Note that the Kirchho shear resultant diers from the applied vertical shear
only if there is variation in the twisting moment. That is, V
xz
(+xK)
( y) diers
from V
xz
(+x)
( y) only if dM
x
xy
( y)/dy p 0. In eect, the Kirchho shear
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 N
xx
* (x,y), N
yy
* (x,y), N
xy
* (x,y), N
yx
* (x,y),
M
xx
* (x,y), M
yy
* (x,y), M
xy
* (x,y), M
yx
* (x,y), V
xz
* (x,y), and V
yz
* (x,y), the static
boundary conditions that may be specied along the edges x=a are as
follows:
For edge x=a:
N
xx
*a; y N
x
xx
y
N
xy
*a; y N
x
xy
y
M
xx
*a; y M
x
xx
y
BM
xx
*a; y
Bx
2
BM
xy
*a; y
By
V
x
xz
y
dM
x
xy
y
dy
2
dM
x
xy
y
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:
N
xx
*0; y N
x
xx
y
N
xy
*0; y N
x
xy
y
M
xx
*0; y M
x
xx
y
BM
xx
*0; y
Bx
2
BM
xy
*0; y
By
V
x
xz
y
dM
x
xy
y
dy
2
dM
x
xy
y
dy
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
For edge y=b:
N
yy
*x; b N
y
yy
x
N
yx
*x; b N
y
yx
x
M
yy
*x; b M
y
yy
x
BM
yy
*x; b
By
2
BM
yx
*x; b
Bx
V
y
yz
x
dM
y
yx
x
dx
2
dM
y
yx
x
dy
For edge y=0:
N
yy
*x; 0 N
y
yy
x
N
yx
*x; 0 N
y
yx
x
M
yy
*x; 0 M
y
yy
x
BM
yy
*x; 0
By
2
BM
yx
*x; 0
Bx
V
y
yz
x
dM
y
yx
x
dx
2
dM
y
yx
x
dy
3.3 Combinations of Geometric and Static Boundary
Conditions
Two fundamental categories of boundary conditions were dened in the
preceding subsections: Geometric Boundary Conditions and Static Boundary
Conditions. If either a geometric or static boundary condition is specied 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 u
o
(a,y) =0 and
u
o
(a,y)=1 mm, respectively. Examples of homogeneous and inhomogeneous
static conditions along the edge x=a are N
xx
(a,y)=0 and N
xx
(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 u
o
(a,y)=0. Note that
this requirement is an example of a homogeneous geometric boundary
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
condition. In such a case the geometric boundary condition is specied and is
therefore known (i.e., u
o
(a,y)=0), whereas the corresponding static boun-
dary condition is unknown. In fact, the question becomes, what stress
resultant N
xx
(+x)
( y) must exist along the edge x=a, so as to maintain the
boundary condition u
o
(a,y)=0?
Conversely, suppose in a dierent 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
xx
(+x)
( 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 u
o
(a,y) exist along the edge x=a,
given that the edge loading is known to be N
xx
(+x)
( 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 specied for each edge of a rectangular plate
are summarized below:
The following conditions can be specied for the edge x=a:
Geometric Condition Static Condition
u
o
a; y u
x
o
y OR N
xx
*a; y N
x
xx
y
v
o
a; y v
x
o
y OR N
xy
*a; y N
x
xy
y
Bwa; y
Bx

Bw
x
y
Bx
OR M
xx
*a; y M
x
xx
y
wa; y w
x
y OR
BM
xx
*a; y
Bx
2
BM
xy
*a; y
By
V
x
xz
y
dM
x
xy
y
dy
2
dM
x
xy
y
dy
The following conditions can be specied for the edge x=0:
Geometric Condition Static Condition
u
o
0; y u
x
o
y OR N
xx
*0; y N
x
xx
y
v
o
0; y v
x
o
y OR N
xy
*0; y N
x
xy
y
Bw0; y
Bx

Bw
x
y
Bx
OR M
xx
*0; y M
x
xx
y
w0; y w
x
y OR
BM
xx
*0; y
Bx
2
BM
xy
*0; y
By
V
x
xz
y
dM
x
xy
y
dy
2
dM
x
xy
y
dy
The following conditions can be specied for the edge y=b:
Geometric Condition Static Condition
u
o
x; b u
y
o
x OR N
yx
*x; b N
y
yx
x
v
o
x; b v
y
o
x OR N
yy
*x; b N
y
yy
x
Bwx; b
By

Bw
y
x
By
OR M
yy
*x; b M
y
yy
x
wx; b w
y
x OR
BM
yy
*x; b
By
2
BM
yx
*x; b
Bx
V
y
yz
x
dM
y
yx
x
dx
2
dM
y
yx
x
dx
The following conditions can be specied for the edge y=0:
Geometric Condition Static Condition
u
o
x; 0 u
y
o
x OR N
yx
*x; 0 N
y
yx
x
v
o
x; 0 v
y
o
x OR N
yy
*x; 0 N
y
yy
x
Bwx; 0
By

Bw
y
x
By
OR M
yy
*x; 0 M
y
yy
x
wx; 0 w
y
x OR
BM
yy
*x; 0
By
2
BM
yx
*x; 0
Bx
V
y
yz
x
dM
y
yx
x
dx
2
dM
y
yx
x
dx
The number of dierent combinations that may be dened is enormous.
Four conditions must be specied to dene the boundary conditions along
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
an edge. Because there are two possibilities in each case, a total of 2
4
=16
possible combinations exist along each edge. Because there are four edges, a
total of 16
4
=65,536 dierent combinations of boundary conditions can be
dened for a thin rectangular plate. However, some combinations are
encountered so frequently that they have been given special names. Speci-
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
Afree edge is dened 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 satised:
For x=a:
N
xx
*a; y 0
N
xy
*a; y 0
M
xx
*a; y 0
BM
xx
*a; y
Bx
2
BM
xy
*a; y
By
0
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 elds. Recalling that
we have limited our discussion to symmetric laminates (B
ij
=M
ij
T
=M
ij
M
=0)
the boundary conditions for a free edge may also be written:
For x=a:
A
11
Bu
o
a; y
Bx
A
12
Bv
o
a; y
By
A
16
Bu
o
a; y
By

Bv
o
a; y
Bx
_ _
N
T
xx
N
M
xx
0
A
16
Bu
o
a; y
Bx
A
26
Bv
o
a; y
By
A
66
Bu
o
a; y
By

Bv
o
a; y
Bx
_ _
N
T
xy
N
M
xy
0
D
11
B
2
wa; y
Bx
2
D
12
B
2
wa; y
By
2
2D
16
B
2
wa; y
BxBy
0
D
11
B
3
wa; y
Bx
3
4D
16
B
3
wa; y
Bx
2
By
D
12
4D
66

B
3
wa; y
BxBy
2
2D
26
B
3
wa; y
By
3
0
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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
deection 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 M
xx
(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 dened in beam theory only two requirements are
involved (w=M
x
=0), whereas in thin plate theory four conditions must be
specied along each plate edge. Hence for thin plates four distinct combina-
tions of geometric and static boundary conditions may be classied 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 dened along the edge x=a as follows:
For x=a:
S1 : wa; y 0 M
xx
*a; y 0 u
o
a; y u
x
o
y v
o
a; y v
x
o
y
S2 : wa; y 0 M
xx
*a; y 0 N
xx
*a; y N
x
xx
y v
o
a; y v
x
o
y
S3 : wa; y 0 M
xx
*a; y 0 u
o
a; y u
x
o
y N
xy
*a; y N
x
xy
y
S4 : wa; y 0 M
xx
*a; y 0 N
xx
*a; y N
x
xx
y N
xy
*a; y N
x
xy
y
Note that a simple support is by denition a mixture of geometric and
static boundary conditions. Also, depending on the values specied for
[u
x
o
( y) or N

xx
( y)] and [v
x
o
( y) or N

xy
( 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 elds.
For example, the four conditions that (collectively) dene a simple support
of type S4 for a symmetric laminate along the edge x=a can be written:
S4 simple support, for x=a:
wa; y 0
D
11
B
2
wa; y
Bx
2
D
12
B
2
wa; y
By
2
2D
16
B
2
wa; y
BxBy
0
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
A
11
Bu
o
a; y
Bx
A
12
Bv
o
a; y
By
A
16
Bu
o
a; y
By

Bv
o
a; y
Bx
_ _
N
T
xx
N
M
xx
N
x
xx
A
16
Bu
o
a; y
Bx
A
26
Bv
o
a; y
By
A
66
Bu
o
a; y
By

Bv
o
a; y
Bx
_ _
N
T
xy
N
M
xy
N
x
xy
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 xed edge) is one in which the out-of-plane
deection 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 satised: 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 dene a clamped boundary condition, whereas in the case of thin
plates four conditions are required. Hence, four distinct combinations can be
classied as a clamped edge. The possible clamped boundary conditions
are often numbered C1 through C4 and are dened as follows:
For x=a:
C1 : wa; y 0
Bwa; y
Bx
0 u
o
a; y u
x
o
y v
o
a; y v
x
o
y
C2 : wa; y 0
Bwa; y
Bx
0 N
xx
*a; y N

xx
y v
o
a; y v
x
o
y
C3 : wa; y 0
Bwa; y
Bx
0 u
o
a; y u
x
o
y N
xy
*a; y N

xy
y
C4 : wa; y 0
Bwa; y
Bx
0 N
xx
*a; y N

xx
y N
xy
*a; y N

xy
y
A clamped boundary may or may not involve a mixture of geometric and
static conditions, depending on the values specied for [u
o
(+x)
( y) or N
xx
(+)
( y)]
and [v
o
(+x)
( y) or N
xy
(+)
( y)]. Also, these conditions may be either homoge-
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
neous or inhomogeneous. In those cases in which the internal stress or
moment resultants are required to take on specic 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 elds, 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 xy 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:
qx; y q
o
sin
px
a
_ _
sin
py
b
_ _
13
where q
o
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 innite 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:
qx; y

l
m1

l
n1
q
mn
sin
mpx
a
_ _
sin
npy
b
_ _
14
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 coecients q
mn
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 q
mn
, multiply Eq. (14) by the factor sin(nVpy/b)dy, and integrate from0 to b:
_
b
0
qx; ysin
nVpy
b
_ _
dy
_
b
0

l
m1

l
n1
q
mn
sin
mpx
a
_ _
sin
npx
b
_ _
_ _
sin
nVpy
b
_ _
dy
15
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
It can be shown that:
_
b
0
sin
npy
b
_ _
sin
nVpy
b
_ _
dy 0; if n p nV
_
b
0
sin
npy
b
_ _
sin
nVpy
b
_ _
dy
b
2
; if n nV
Consequently, upon performing the integration on the right side of Eq. (15),
only one term for the range 0<n<lremains, specically, the term for which
n=nV. Hence Eq. (15) becomes:
_
b
0
qx; ysin
nVpy
b
_ _
dy
b
2

l
m1
q
mnV
sin
mpx
a
_ _
16
We nowrepeat 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:
_
a
0
_
b
0
qx; ysin
mVpx
a
_ _
sin
nVpy
b
_ _
dxdy

_
a
0
b
2

l
m 1
q
mnV
sin
mpx
a
_ _
_ _
sin
mVpx
a
_ _
dx
ab
4
q
mVn V
Solving this result for the coecient q
mVnV
, we nd:
q
mVnV

4
ab
_
a
0
_
b
0
qx; ysin
mVpx
a
_ _
sin
nVpy
b
_ _
dxdy 17
Eq. (17) allows calculation of the constant coecients, 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 innite number of terms (as
m,n ! l). Of course, it is impossible to use an innite number of terms;
in practice a nite 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:
qx; y q
o
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
where q
o
is the magnitude of the uniformly distributed load. Substituting
q(x,y) = q
o
into Eq. (17), we nd the coecients in the Fourier series ex-
pansion associated with specied value of m and n to be:
q
mn

4q
o
ab
_
a
0
_
b
0
sin
mpx
a
_ _
sin
npy
b
_ _
dxdy
16q
o
p
2
mn
; m; n odd integers
q
mn

4q
o
ab
_
a
0
_
b
0
sin
mpx
a
_ _
sin
npy
b
_ _
dxdy 0; m; n even integers
Hence a constant transverse load q(x,y)=q
o
can be written:
qx; y
16q
o
p
2

l
m 1

l
n 1
1
mn
sin
mpx
a
_ _
sin
npy
b
_ _
m; n 1; 3; 5; . . . 18
How well Eq. (18) describes a constant transverse loading depends on the
number of terms used. Anormalized 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 a
i
b
i
,
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 dened as a force, with units of
Newtons or pounds-force, for example. Eq. (17) becomes in this case:
q
mn

4P
aba
i
b
i
_
na
i
=2
na
i
=2
_
gb
i
=2
gb
i
=2
sin
mpx
a
_ _
sin
npy
b
_ _
dxdy
Evaluating this integral we nd:
q
mn

16P
p
2
mna
i
b
i
sin
mpn
a
_ _
sin
npg
b
_ _
sin
mpa
i
2a
_ _
sin
npb
i
2b
_ _
19
Together, Eq. (14) and (19) dene the Fourier series expansion of a force P,
uniformly distributed over the internal region a
i
b
i
. To illustrate this series
expansion, consider the following specic example. Assume that the interior
region is centered in the middle of the plate, i.e., let n=a/2, g=b/2. Further,
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 a
i
b
i
. The
interior region is centered at the point x=n, y=g.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
assume that the length and width of the interior region equals one half the
length and width of the plate, i.e., let a
i
=a/2, b
i
=b/2. For this specic
example then, Eq. (19) becomes:
q
mn

64P
p
2
mnab
sin

mp
2

sin

np
2

sin

mp
4

sin

np
4

; m; n odd integers
q
mn
0; m; n even integers
Note that coecients 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 nd:
qx; y
64P
p
2
ab

l
m1

l
n1
1
mn
sin

mp
2

sin

np
2

sin

mp
4

sin

np
4

_ _
sin

mpx
a

sin

mpy
b

; m; n odd integers
20
As before, how well Eq. (20) describes a load P uniformly distributed over the
interior region depends on the number of terms used. Anormalized 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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 coecients for this case can be obtained from
Eq. (19) by allowing a
i
!0, b
i
!0, i.e., the internal area over which load Pacts
is allowed to shrink to a point. It can be shown that in the limit Eq. (19)
becomes:
q
mn

4P
ab
sin

mpn
a

sin

mpg
b

21
Together, Eqs. (14) and (21) dene the Fourier series expansion of a
concentrated force P applied at the point x=n and y=g. As a specic
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:
q
mn

4P
ab
1
mn=21
; m; n odd integers
0; m; n even integers
_
_
_
Combining this result with Eq. (14), we nd that the Fourier series expansion
for a concentrated load applied at the center of the plate is given by:
qx; y
4P
ab

l
m 1

l
n 1
sin

mpx
a

sin

npy
b

1
mn=21
22
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), 134140.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 deections caused by the combined eects 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 specied boundary con-
ditions often cannot be found. Sometimes, the diculty in obtaining an exact
solution is because of the boundary conditions involved; diculties of this
sort are encountered for both isotropic and anisotropic plates. In other cases,
the diculties are strictly because of the anisotropic nature of composites.
527
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
We have previously restricted our discussion to symmetric stacking
sequences, and so B
ij
M
T
ij
M
M
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 A
16
A
26
D
16
D
26
N
T
xy
N
M
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 stinesses and thermal/moisture resultants and can
therefore be classied 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 denition of a specially orthotropic
laminate is inconsistently used in the literature, and is sometimes relaxed
from the denition just described. Specically, the denition 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
stinesses D
16
and D
26
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
A
16
A
26
N
T
xy
N
M
xy
= 0, but place no restriction on the values of D
16
or D
26
. 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 A
16
A
26

N
T
xy
N
M
xy
= 0. Hence both symmetric balanced laminates and symmetric
angle-ply laminates (see Sec. 7 of Chap. 6) are specially orthotropic under this
relaxed denition. 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 stinesses of the laminate plays an
important role. Therefore we must maintain the more restrictive denition
and stipulate that a specially orthotropic laminate must have A
16
A
26

D
16
D
26
N
T
xy
N
M
xy
= 0.
For a specially orthotropic laminate, the equations of equilibrium for
a symmetric laminate [i.e., Eq. (8) of Chap. 9] are simplied further still,
and become:
A
11
@
2
u
o
@x
2
A
12
A
66

@
2
v
o
@x@y
A
66
@
2
u
o
@y
2
0 1a
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
A
12
A
66

@
2
u
o
@x@y
A
22
@
2
v
o
@y
2
A
66
@
2
u
o
@x
2
0 1b
A
11
@u
o
@x
A
12
@v
o
@y
N
T
xx
N
M
xx
_ _
@
2
w
@x
2
_ _
1c
A
12
@u
o
@x
A
22
@v
o
@y
N
T
yy
N
M
yy
_ _
@
2
w
@y
2
_ _
2 A
66
@u
o
@y

@v
o
@x
_ _ _ _
@
2
w
@x@y
_ _
D
11
@
4
w
@x
4
2D
12
2D
66

@
4
w
@x
2
@y
2
D
22
@
4
w
@y
4
qx; y 0
In the following sections, we will rst obtain solutions for plate
deections 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 rst 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. 15 for a discussion of this method.
Anal 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 deections induced in specially
orthotropic composite plates that are simply supported along all four edges.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
As discussed in Sec. 3 of Chap. 9, four types of simple supports may be dened
for a thin plate. For the edge x=a, these are:
S1: wa; y 0 M
*
xx
a; y 0 u
o
a; y u
x
o
y v
o
a; y v
x
o
y
S2: wa; y 0 M
*
xx
a; y 0 N
*
xx
a; y N
x
xx
y v
o
a; y v
x
o
y
S3: wa; y 0 M
*
xx
a; y 0 u
o
a; y u
x
o
y N
*
xy
a; y N
x
xy
y
S4: wa; y 0 M
*
xx
a; y 0 N
*
xx
a; y N
x
xx
y N
*
xy
a; y N
x
xy
y
Hence the denition of a simply supported edge requires that either in-plane
displacements, in-plane stress resultants, or some combination thereof must
be specied 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 roomtemperature, 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 N
xx
, N
yy
, and/or N
xy
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 mentionedin Section 6.2 of Chap. 6, in practice the stress- andstrain-free temperature is often
2050jC 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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
that assumed during the development of CLT in Chap. 6. Therefore we can
develop general expressions giving the in-plane displacement elds caused by
any combination of N
ij
, 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:
e
o
xx
e
o
yy
c
o
xy
j
xx
j
yy
j
xy
_

_
_

a
11
a
12
0 0 0 0
a
12
a
22
0 0 0 0
0 0 a
66
0 0 0
0 0 0 d
11
d
12
0
0 0 0 d
12
d
22
0
0 0 0 0 0 d
66
_

_
_

_
N
xx
N
T
xx
N
M
xx
N
yy
N
T
yy
N
M
yy
N
xy
0
0
0
_

_
_

_
As in earlier chapters, we assume innitesimal strain levels. Therefore mid-
plane strains are related to midplane displacements according to Eq. (10) of
Chap. 6:
e
o
xx

@u
o
@x
e
o
yy

@v
o
@y
c
o
xy

@u
o
@y

@v
o
@x
Consequently, midplane displacement elds are given by:
@u
o
@x
a
11
N
xx
N
T
xx
N
M
xx
a
12
N
yy
N
T
yy
N
M
yy
2a
@v
o
@y
a
12
N
xx
N
T
xx
N
M
xx
a
22
N
yy
N
T
yy
N
M
yy
2b
@u
o
@y

@v
o
@x
a
66
N
xy
2c
Integrating Eq. (2a) with respect to x, we nd:
u
o
x; y a
11
N
xx
N
T
xx
N
M
xx
a
12
N
yy
N
T
yy
N
M
yy

_ _
x f
1
y k
1
Where f
1
( y) is an unknown function of y (only) and k
1
is an unknown constant
of integration. Similarly, integrating Eq. (2b) with respect to y we nd:
v
o
x; y a
12
N
xx
N
T
xx
N
M
xx
a
22
N
yy
N
T
yy
N
M
yy

_ _
yf
2
x k
2
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
where f
2
(x) is an unknown function of x (only) and k
2
is a second unknown
constant. Without a loss in generality, we assume that midplane displace-
ments are zero at the origin (i.e., let u
o
= v
o
= 0 at x = y = 0), and con-
sequently we conclude k
1
= k
2
= 0. Substituting the above expressions for
u
o
(x,y) and v
o
(x,y) into Eq. (2c), we nd:
@u
o
@y

@v
o
@x

@f
1
@y

@f
2
@x
a
66
N
xy
Because the terms that appear on the right side of the equality (i.e., a
66
and
N
xy
) are known constants, it follows that f
1
and f
2
must be at most linear
functions of y and x, respectively:
f
1
y k
3
y
f
2
x k
4
x
Hence we can write:
k
3
k
4
a
66
N
xy
Constants k
3
and k
4
can take on any value as long as they sum to the product
(a
66
N
xy
). To determine particular values convenient for our use, we now
require that the innitesimal rotation vector in the xy plane, x
xy
(which
represents rigid body motion of the plate), is zero. The innitesimal rotation
vector is given by [see Ref. 3]:
x
xy

1
2
@u
o
@y

@v
o
@x
_ _
Requiring that x
xy
=0 leads to:
k
3
k
4

1
2
a
66
N
xy
Combining the preceding results, we conclude that in-plane midplane dis-
placement elds 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:
u
o
x; y a
11
N
xx
N
T
xx
N
M
xx
a
12
N
yy
N
T
yy
N
M
yy

_ _
x
1
2
a
66
N
xy
y
v
o
x; y
1
2
a
66
N
xy
x a
12
N
xx
N
T
xx
N
M
xx

_
a
22
N
yy
N
T
yy
N
M
yy
y
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
These expressions can be simplied through the use of the eective thermal
expansion coecients of the laminate, dened by Eq. (73a) of Chap. 6, and the
eective moisture expansion coecients, dened by Eq. (76a) of Chap. 6. For
a specially orthotropic laminate, these become:
a
xx

1
DT
a
11
N
T
xx
a
12
N
T
yy
_ _
a
yy

1
DT
a
12
N
T
xx
a
22
N
T
yy
_ _
a
xy
0
b
xx

1
DM
a
11
N
M
xx
a
12
N
M
yy
_ _
b
yy

1
DM
a
12
N
M
xx
a
22
N
M
yy
_ _
b
xy
0
Hence we see that the midplane displacement elds can be written as:
u
o
x; y a
11
N
xx
a
12
N
yy
DTa
xx
DMb
xx
_
x
1
2
a
66
N
xy
y 3a
v
o
x; y
1
2
a
66
N
xy
x a
12
N
xx
a
22
N
yy
DTa
yy
DMb
yy
_
y 3b
Note that the displacement elds are independent of the transverse load
q(x,y). As pointed out in Sec. 2 of Chap. 9, u
o
(x,y) and v
o
(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
c0]. 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
u
o
(x,y) and v
o
(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 elds
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 xy plane
according to:
qx; y q
o
sin
px
a
_ _
sin
py
b
_ _
4
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 dene DT
c
as the change in temperature from cure to room
temperature:
DT
c
T
RT
T
C
where T
RT
= room temperature and T
C
= 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 N
xx
=N
yy
=N
xy
=0:
u
c
o
x; y a
xx
DT
c
x 5a
v
c
o
x; y a
yy
DT
c
_ _
y 5b
These displacements are induced before assembly of the simply supported
plate. Note that these in-plane displacement elds satisfy the rst two
equations of equilibrium, Eqs. (1a) and (1b). A type S1 simple support will
simply maintain these displacements during subsequent loading and/or
Figure 1 Thin rectangular plate of thickness t and in-plane dimensions a b,
subjected to a transverse load q(x,y) = q
o
{sin[(px)/a]}{sin[(py)/b]}. All four
edges of the plate are subject to simple supports of type S1.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
temperature changes. Therefore type S1 boundary conditions for all four
edges of the plate become:
For x 0 For x a :
w0; y 0 wa; y 0
M
*
xx
0; y D
11
@
2
w
@x
2
D
12
@
2
w
@y
2
0 M
*
xx
a; y D
11
@
2
w
@x
2
D
12
@
2
w
@y
2
0
u
o
0; y u
x
o
y 0 u
o
a; y u
x
o
y a
xx
DT
c
a
v
o
0; y v
x
o
y a
yy
DT
c
_ _
y v
o
a; y v
x
o
y a
yy
DT
c
_ _
y
For y 0 For y b :
wx; 0 0 wx; b 0
M
*
yy
x; 0 D
11
@
2
w
@x
2
D
12
@
2
w
@y
2
0 M
*
yy
x; b D
11
@
2
w
@x
2
D
12
@
2
w
@y
2
0
u
o
x; 0 u
y
o
x a
xx
DT
c
x u
o
x; b u
x
o
x a
xx
DT
c
x
v
o
x; 0 v
y
o
x 0 v
o
x; b v
x
o
y a
yy
DT
c
_ _
b
We wish to determine the out-of-plane displacement eld w(x,y) that
satises 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 eld is
given by:
wx; y c sin
px
a
_ _
sin
py
b
_ _
6
where c is an unknown constant. Substituting this assumed form into the
boundary conditions will reveal that they are identically satised 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);
(b) write the thermal stress resultants N
xx
T
and N
yy
T
in terms of eec-
tive thermal expansion coecients [using Eq. (73b) of Chap. 6]; and
then
(c) solve the resulting expression for constant c.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Following this process we nd:
c
q
o
p
4
D
11
a
4

2
a
2
b
2
D
12
2D
66

D
22
b
4

DT
c
DT
p
2
1
a
2
A
11
a
xx
A
12
a
yy
_

1
b
2
A
12
a
xx
A
22
a
yy
_
_ _ _ _
7a
Notice that the temperature change as dened in earlier chapters (DT) ap-
pears in Eq. (7a). That is:
DT current temperature cure temperature
Also note that if the temperature is not changed following assembly in the
type S1 simple supports, then DT=DT
c
, and the eects 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:
c
q
o
R
4
b
4
p
4
D
11
2R
2
D
12
2D
66
R
4
D
22

DT
c
DTa
2
p
2
a
xx
A
11
R
2
A
12
_ _
a
yy
A
12
R
2
A
22
_ _ _ _
_ _
7b
The predicted out-of-plane deection is obtained by combining either
Eq. (7a) or (7b) with Eq. (6). Using Eq. (7b) for example, we have:
wx; y
q
o
R
4
b
4
sinpx=asinpy=b
p
4
D
11
2R
2
D
12
2D
66
R
4
D
22

DT
c
DTa
2
p
2
a
xx
A
11
R
2
A
12
_ _
a
yy
A
12
R
2
A
22
_ _ _ _
_ _
8
Equations (5a), (5b), and (8) give the predicted displacement eld 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 roomtemperature. 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 elds are given by Eqs. (5a),
(5b), and (8). A typical application of this solution is discussed in Sample
Problem 1.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 [0
2
/90)
2
]
s
graphiteepoxy laminate is cured at 175jC and then cooled to
room temperature (20jC). After cooling, the at laminate is trimmed to in-
plane dimensions of 300 150 mmand mounted in an assembly that provides
type S1 simple supports along all four edges. The x axis is dened 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
graphiteepoxy 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 graphiteepoxy and the specied stacking sequence, the [ABD]
matrix is:
ABD
1:76 10
6
4:52 10
6
0 0 0 0
4:52 10
6
95:6 10
6
0 0 0 0
0 0 19:5 10
6
0 0 0
0 0 0 40:1 0:848 0
0 0 0 0:848 10:8 0
0 0 0 0 0 3:66
_

_
_

_
where the units of A
ij
are Pa m and the units of D
ij
are Pa m
3
.
We also require the eective thermal expansion coecients. Based on
the properties listed in Table 3 of Chap. 3 for graphiteepoxy and the specied
stacking sequence, these are:
a
xx
0:29 lm=m jC a
yy
2:44 lm=m jC
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Because the laminate is specially orthotropic, the eective shear thermal
expansion coecient is zero (a
xy
= 0). The cooldown from the cure temper-
ature to room temperature is:
DT
c
20jC 175jC 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 eective thermal expansion coecients 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:
wx; y
324
p
4
278
88:6DT
c
DT
p
2
_ _
sin
px
0:30
_ _
sin
py
0:15
_ _
meters
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 graphiteepoxy plate
considered in Sample Problem 1 as a function of temperature.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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
= 20jC175jC = 155jC, and tem-
peratures ranging from50jC<T<20jCare to be considered, the quantity
(DT
c
DT) ranges from 70jC z(DT
c
DT) z0jC in this problem. A plot
of maximumout-of-plane displacement as a function of temperature is shown
in Fig. 2. At room temperature (20jC), a maximum deection of 12 mm is
predicted. As wouldbe expected, the plate becomes stier 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 deection of 3.7 mm is
predicted.
4 SPECIALLY ORTHOTROPIC LAMINATES SUBJECT
TO SIMPLE SUPPORTS OF TYPE S4
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,
subjected to a transverse load q(x,y) = q
o
sin(
px
a
)sin(
py
b
). All four edges of the
plate are subject to simple supports of type S4 (compare with Figure 1).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
elds along the plate edges are not required to take on any specied 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: N
xy
=
N
yx
= 0. The plate is also subjected to a uniform change in temperature, DT,
and a transverse loading that varies over the xy plane according to:
qx; y q
o
sin
px
a
_ _
sin
py
b
_ _
9
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 dene 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:
N
*
xx
0; y N
*
xx
a; y N
xx
N
*
yy
x; 0 N
*
yy
x; b N
yy
N
*
xy
0; y N
*
xy
a; y N
*
yx
x; 0 N
*
yx
x; b 0
Because we have limited consideration to symmetric specially orthotropic
laminates, A
16
A
26
D
16
D
26
B
ij
N
T
xy
N
M
xy
M
T
ij
M
M
ij
=0.
Therefore the boundary conditions can be written as:
For x = 0,a:
w0; y wa; y 0 10a
M
*
xx
0; y M
o
xx
a; y D
11
@
2
w
@x
2
D
12
@
2
w
@y
2
0 10b
N
*
xx
0; y N
*
xx
a; y A
11
@u
o
@x
A
12
@v
o
@y
N
T
xx
N
xx
10c
N
*
xy
0; y N
*
xy
a; y A
66
@u
o
@y

@v
o
@x
_ _
0 10d
For y = 0,b:
wx; 0 wx; b 0 11a
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
M
*
yy
x; 0 M
*
yy
x; b D
12
@
2
w
@x
2
D
22
@
2
w
@y
2
0 11b
N
*
yy
x; 0 N
*
yy
x; b A
12
@u
o
@x
A
22
@v
o
@y
N
T
yy
N
yy
11c
N
*
yx
x; 0 N
*
yx
x; b A
66
@u
o
@y

@v
o
@x
_ _
0 11d
Using Eqs. (3a) and (3b) for the assumed conditions (i.e., N
xy
=DM=0), the
in-plane displacements elds are:
u
o
x; y a
11
N
xx
a
12
N
yy
DTa
xx
_
x 12a
v
o
x; y a
12
N
xx
a
22
N
yy
DTa
yy
_
y 12b
Let us conrm that these equations satisfy the appropriate boundary con-
ditions. Substituting Eqs. (12a) and (12b) into boundary condition (10c) and
rearranging, we nd that the following expression must be satised:
N
xx
A
11
a
11
A
12
a
12
N
yy
A
11
a
12
A
12
a
22

DTA
11
a
xx
A
12
a
yy
N
T
xx
N
xx
Because the laminate is specially orthotropic,
A
A
11
A
12
0
A
12
A
22
0
0 0 A
66
_
_
_
_
a A
1

A
22
A
11
A
22
A
2
12

A
12
A
11
A
22
A
2
12

0
A
12
A
11
A
22
A
2
12

A
11
A
11
A
22
A
2
12

0
0 0
1
A
66
_

_
_

_
Therefore by direct substitution we nd:
A
11
a
11
A
12
a
12
1
A
11
a
12
A
12
a
22
0
Also, from Eq. (73b) of Chap. 6, we nd (for A
16
= 0):
N
T
xx
DTA
11
a
xx
A
12
a
yy

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.


Hence the boundary condition represented by Eq. (10c) is satised identically
by the in-plane displacement elds. A similar process can be used to conrm
that the boundary condition given by Eq. (11c) is also satised. Both Eqs.
(10d) and (11d) are satised as well, because @u
o
/@y = @v
o
/@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 eld is given by:
wx; y c sin
px
a
_ _
sin
py
b
_ _
13
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 satised 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 nd:
c
q
o
p
4
1
a
4
D
11

2
a
2
b
2
D
12
2D
66

1
b
4
D
22

1
p
2
a
2
N
xx

1
p
2
b
2
N
yy
_ _
14a
Using the denition of the plate aspect ratio, R = a/b, this result can also be
written as:
c
q
o
R
4
b
4
p
4
D
11
2R
2
D
12
2D
66
R
4
D
22

a
2
p
2
N
xx
N
yy
R
2
_ _
_ _ 14b
The predicted out-of-plane deection is obtained by combining either Eq.
(14a) or (14b) with Eq. (13). Using Eq. (14b) for example, we have:
wx; y
q
o
R
4
b
4
sinpx=asinpy=b
p
4
D
11
2R
2
D
12
2D
66
R
4
D
22

a
2
p
2
N
xx
N
yy
R
2
_ _
_ _ 15
Equations (12a), (12b), and (15) give the predicted displacement elds in-
duced in the plate and represent the solution to this problem.
To summarize, we have found the displacement elds 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 N
xx
and N
yy
. A typical application of this
solution is discussed in Sample Problem 2. It should be kept in mind that we
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
have not considered the possibility of buckling in this section. The solution we
have obtained is not valid if N
xx
and/or N
yy
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 [(0
2
/90)
2
]
s
graphiteepoxy laminate is cured at 175jC and then cooled to
room temperature (20jC). After cooling, the at laminate is trimmed to in-
plane dimensions of 300150 mm and mounted in an assembly that provides
type S4 simple supports along all four edges. The x axis is dened 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 (i.e., N
xx
= N
yy
) 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
dened by y = 0.075 m, if in-plane loads N
xx
= N
yy
= 50 kN/m
are applied.
(b) Plot the maximum out-of-plane displacement as a function of in-
plane loads, over the range 0<(N
xx
= N
yy
)<70 kN/m.
(c) Compare the maximum out-of-plane displacement caused by the
specied 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 N
xx
= N
yy
= 0
Use the properties listed for graphiteepoxy in Table 3 of Chap. 3, and assume
each ply has a thickness of 0.125 mm.
Solution. A [(0
2
/90)
2
]
s
graphiteepoxy 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 stinesses, dimensions, and
the specied transverse loading, Eq. (15) becomes:
wx; y
324
p
4
278
0:090
p
2
N
xx
4N
yy
_ _
_ _
_

_
_

_
sin
px
0:3
_ _
sin
py
0:15
_ _
meters
This expression can be used to calculate the out-of-plane displacement
induced at any point (x,y) over the surface of the plate.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Part (a). Using N
xx
N
yy
= 50 kN/m and y = 0.075 m, out-of-plane dis-
placements are:
wx; y 1:30 10
3
sin
px
0:3
_ _
meters
A plot of these displacements over the length of the plate (0<x<0.3m is
showninFig. 4(a). Displacements are zeroat the edges denedby x=0, 0.3 m,
as dictated by the specied 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:
wj
max

324
p
4
278
0:090
p
2
N
xx
4N
yy
_ _
_ _
_

_
_

_
meters
Aplot of maximumout-of-plane displacement as a functionof in-plane tensile
loads is shown in Fig. 4b. As would be intuitively expected, the plate is
stiened 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
maximumdeection of 12 mmoccurs when N
xx
=N
yy
=0, whereas the max-
imum deection is reduced to 0.96 mm when N
xx
= N
yy
= 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 xed 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 specied
rather than in-plane displacements.
Referring to the results presented in these two sample problems, we
nd that identical deections are predicted, despite the dierences in
boundary conditions. That is, a maximum deection of 12 mm is predicted
at room temperature for type S1 condition, and an identical 12 mm
deection is predicted if N
xx
= N
yy
= 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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Figure 4 Out-of-plane displacements induced in the graphiteepoxy plate
considered in Sample Problem 2. (a) Out-of-plane displacements induced along
centerline y = 0.075 m when N
xx
= N
yy
= 50 kN/m, (b) Maximum out-of-plane
displacements as a function of in-plane loads N
xx
= N
yy
.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
for type S4 conditions is not predicted because we have based our analysis
on innitesimal strains. The consequences of the innitesimal 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
elds u
o
(x,y) and v
o
(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 dierent boundary conditions are iden-
tical because we have considered the case in which N
xx
= N
yy
= 0. If N
xx
and/or N
yy
p 0, then the transverse displacements for a plate supported by
type S4 simple supports is quite dierent 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) = q
o
{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).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 N
xx
and known displacements v
o
. In
contrast, the twoedges y =0,b are subject tosimple supports of type S1, which
means that we require these edges to maintain known in-plane displacements
u
o
and v
o
. The plate is also subjected to a uniform change in temperature, DT,
and a transverse loading that varies over the xy plane according to:
qx; y q
o
sin
px
a
_ _
py
b
_ _
16
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 N
xx
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 v
o
(x,y) induced during cooldown. Using Eq. (3b), this can be written
(with N
xx
= N
yy
= N
xy
= DM = 0):
v
o
x; y DT
c
a
yy
y 17
We can use this expression to specify known values of v
o
along all four edges
of the plate. Because a known stress resultant N
xx
is applied to the two edges
x = 0 and x = a, all quantities necessary to dene the type S2 simple support
along these two edges are known. To dene the type S1 simple supports along
the edges y = 0,b, we must specify known values of u
o
. Toward that end,
equate Eqs. (3b) and (17):
1
2
a
66
N
xy
x a
12
N
xx
a
22
N
yy
DT a
yy
_
y DT
c
a
yy
y
This relation must be satised along the edges y =0,b. Substituting y =0, it is
seen that N
xy
must vanish (N
xy
= 0). Substituting y = b, and solving for N
yy
,
we obtain:
N
yy

a
yy
DT
c
DT a
12
N
xx
a
22
18
Equation (18) gives the stress resultant N
yy
that must be provided by the
simple supports along the y = 0,b so as to maintain the stipulated value of v
o
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 N
xx
is applied (i.e.,
if N
xx
= 0), then N
yy
= 0, as would be expected.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
We can now calculate the displacement u
o
implied by these conditions,
using Eq. (3b):
u
o
x; y a
11
N
xx
a
12
a
yy
DT
c
DT a
12
N
xx
a
22
_ _
DT a
xx
_ _
x
This expression can be written as:
u
o
x; y N
xx
a
11

a
2
12
a
22
_ _

a
12
a
22
a
yy
DT
c
DT a
xx
DT
_ _
x 19
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
w0; y wa; y 0 20a
M
*
xx
0; y M
*
xx
a; y D
11
@
2
w
@x
2
D
12
@
2
w
@y
2
0 20b
N
*
xx
0; y N
*
xx
a; y A
11
@u
o
@x
A
12
@v
o
@y
N
T
xx
N
xx
20c
v
o
0; y v
o
a; y DT
c
a
yy
y 20d
For y = 0,b:
wx; 0 wx; b 0 21a
M
*
xx
x; 0 M
*
yy
x; b D
12
@
2
w
@x
2
D
22
@
2
w
@y
2
0 21b
u
o
x; 0 u
o
x; b
N
xx
a
11

a
2
12
a
22
_ _

a
12
a
22
a
yy
DT
c
DT a
xx
DT
_ _
x 21c
v
o
x; 0 0 v
o
x; b DT
c
a
yy
b 21d
Now consider out-of-plane displacements w(x,y). Guided by the func-
tional formof the transverse pressure [i.e., Eq. (16)], we once again assume the
out-of-plane displacement eld is given by:
wx; yc sin
px
a
_ _
sin
py
b
_ _
22
where c is an unknown constant. Substituting this assumed form into
boundary conditions Eqs. (20a), (20b), (21a), and (21b) will reveal that they
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
are identically satised 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 nd:
c
q
o
p
4
D
11
a
4

2D
12
2D
66

a
2
b
2

D
22
b
4

N
xx
p
2
1
a
2

A
12
A
11
b
2
_ _

a
yy
A
12
A
22
A
2
12

p
2
b
2
A
11
DT
c
DT
_ _
23
Using the denition of the plate aspect ratio, R = a/b, this result can also be
written as:
c
q
o
R
4
b
4
p
4
D
11
2R
2
D
12
2D
66
R
4
D
22

R
4
b
2
p
2
N
xx
1
R
2

A
12
A
11
_ _

a
yy
A
12
A
22
A
2
12

A
11
DT
c
DT
_ _ _ _
24
The predicted out-of-plane deection is obtained by combining either Eq.
(23) or (24) with Eq. (22). Using Eq. (24), for example, we have:
wx; y
q
o
R
4
b
4
sinpx=asinpy=b
p
4
D
11
2R
2
D
12
2D
66
R
4
D
22

R
4
b
2
p
2
N
xx
1
R
2

A
12
A
11
_ _

a
yy
A
12
A
22
A
2
12

A
11
DT
c
DT
_ _ _ _
25
Equations (17), (19), and (25) give the predicted displacement elds induced
in the plate and represent the solution to this problem.
To summarize, we have found the displacement elds 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. Atypical 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
N
xx
is compressive or if the change in temperature DT tends to cause the
laminate to expand.
Sample Problem 3A
[(0
2
/90)
2
]
s
graphiteepoxy laminate is cured at 175jC and then cooled to
room temperature (20jC). After cooling, the at laminate is trimmed to in-
plane dimensions of 300 150 mmand mounted in an assembly that provides
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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
dened parallel to the 300-mm edge (i.e., a = 0.3 m; b = 0.15 m). The
laminate is then subjected to a uniformin-plane tensile loading N
xx
along x =
0,a and a transverse pressure given by q(x,y) = 40 {sin[(px)/a]}{sin[(py)/b]}
(kPa).
(a ) Plot the maximumout-of-plane displacement as a function of tensile
load over the range 0<N
xx
< 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 N
xx
= 50 kN/m.
Use the properties listed for graphiteepoxy in Table 3 of Chap. 3, and assume
each ply has a thickness of 0.125 mm.
Solution. A [(0
2
/90)
2
]
s
graphiteepoxy laminate was also considered in
Sample Problem 1, and numerical values for the [ABD] matrix and eective
thermal expansion coecients 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 stinesses, dimensions, and the specied transverse loading, Eq. (25)
becomes:
wx; y
324
p
4
278
0:099
p
2
N
xx

82:5
p
2
DT
c
DT
_ _
_

_
_

_
sin
px
0:3
_ _
sin
py
0:15
_ _
meters
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:
wj
max

324
p
4
278
0:099
p
2
N
xx
_ _
meters
A plot of maximum out-of-plane displacement as a function of N
xx
is shown
in Fig. 6(a). As would be intuitively expected, the plate is stiened as N
xx
is
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Figure 6 Out-of-plane displacements induced in the graphiteepoxy plate
considered in Sample Problem 3. (a) Maximum out-of-plane displacements at
room temperature as a function of in-plane load N
xx
. (b) Maximum out-of-plane
displacements as a function of temperature (N
xx
= 50 kN/m).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
increased. A maximum deection of 12 mm occurs when N
xx
= 0, whereas
the maximum deection is reduced to 3.4 mm when N
xx
= 70 kN/m.
Part (b). As before, the maximumout-of-plane displacement occurs at x =
a/2 = 0.15 m and y = b/2 = 0.075. Because a constant in-plane tensile load
N
xx
= 50 kN/m is applied, the maximum out-of-plane displacement is given
by:
wj
max

324
p
4
781
82:5
p
2
DT
c
DT
_ _
meters
Aplot of maximumout-of-plane displacement as a function of temperature is
shown in Fig. 6(b). At roomtemperature (20jC), a maximumdeection of 4.3
mm is predicted. As would be expected, the plate becomes stier as the
temperature is decreased. Out-of-plane displacements are decreased because
of the in-plane tensile load N
yy
that develops as temperature is decreased. At
the lowest temperature considered (50jC), a maximumdeection 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,
uniformin-plane loads N
xx
and N
yy
, a uniformtemperature change DT, and a
sinusoidal transverse loading given by:
qx; y q
o
sin
px
a
_ _
sin
py
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,
repeated here for convenience:
qx; y

l
m1

l
n1
q
mn
sin
mpx
a
_ _
sin
npy
b
_ _
9:14
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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
dierences are that the transverse load is given by Eq. 14 of Chap. 9 and the
transverse deections w(x,y) are assumed to be of the form:
wx; y

l
m1

l
n1
c
mn
sin
mpx
a
_ _
sin
npy
b
_ _
26
where c
mn
are unknown constants to be determined using the equations of
equilibrium. The assumed form and solution for in-plane displacements
u
o
(x,y) and v
o
(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 coecients (following the same
procedures as used in Sec. 4), we obtain:
c
mn

q
mn
a
4
b
4
p
4
D
11
mb
4
2D
12
2D
66
mnab
2
D
22
na
4

ab
p
_ _
2
N
xx
mb
2
N
yy
na
2
_ _
_ _
27a
Using the denition of the plate aspect ratio, R=a/b, this result can also be
written as:
c
mn

q
mn
R
4
b
4
p
4
D
11
m
4
2D
12
2D
66
mnR
2
D
22
nR
4

a
2
p
2
N
xx
m
2
N
yy
nR
2
_ _
_ _
27b
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 N
xx
and N
yy
, a
uniformtemperature change DT, and homogenous simple supports of type S4
along all four edges are given by:
u
o
x; y
A
22
N
xx
A
12
N
yy
A
11
A
22
A
2
12
_ _
a
xx
DT
_ _
x 28a
v
o
x; y
A
11
N
yy
A
12
N
xx
A
11
A
22
A
2
12
_ _
a
yy
DT
_ _
y 28b
wx; y
R
4
b
4
p
4

l
m1

l
n1
q
mn
sinmpx=asinnpy=b
D
11
m
4
2D
12
2D
66
mnR
2
D
22
nR
4

a
2
p
2
N
xx
m
2
N
yy
nR
2
_ _
_ _
28c
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
The method of using a double-Fourier series expansion to represent the
transverse loading and out-of-plane displacements was rst 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 [15] 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
A [(0
2
/90)
2
]
s
graphiteepoxy laminate is cured at 175jC and then cooled to
room temperature (20jC). After cooling, the at laminate is trimmed to in-
plane dimensions of 300150 mm and mounted in an assembly that provides
type S4 simple supports along all four edges. The x axis is dened 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 N
xx
= N
yy
= 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 dened by y = 0.075 m. Use the
properties listed for graphiteepoxy in Table 3 of Chap. 3, and assume each
ply has a thickness of 0.125 mm.
Solution. A [(0
2
/90)
2
]
s
graphiteepoxy 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.
The double-Fourier series expansion of a uniform transverse loading
was discussed in Sec. 4 of Chap. 9. The coecients in the Fourier series
expansion were found to be:
q
mn

16q
o
p
2
mn
; m; n odd integers
Combining these coecients with Eq. (28c) allows prediction of out-of-plane
displacements. A plot of these displacements along the centerline of the plate
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
dened by y =b/2 =0.075 mis 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-
termexpansion (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 at symmetric laminate subjected to
constant and uniformboundary edge loads N
x
xx
N
x
xx
=N
xx
and N
y
yy
=
N

yy
= N
yy
(shear resultants are assumed zero: N
xy
= N
yx
= 0). As we have
seen in preceding sections, coupling stinesses B
ij
= 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.
Figure 7 Out-of-plane displacements induced in the graphiteepoxy plate con-
sidered in Sample Problem 4, along centerline y = 0.075 m.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
This statement is always true if both N
xx
and N
yy
are tensile; a at
symmetric laminate subjected to tensile edge loads N
xx
and/or N
yy
will remain
at regardless of the magnitude of load (unless, of course, the loads are high
enough to cause fracture). However, suppose that either N
xx
or N
yy
(or both)
are compressive. If the magnitude of the compressive load(s) is (are) relatively
low, then the initially at laminate remains at; that is, only in-plane
displacements u
o
(x,y) and v
o
(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-Dview
(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
stiness of the laminate, plate aspect ratio, applied loading, and the boundary
conditions. Aparticular 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 at rectangular
specially orthotropic laminate subjected to simple supports of type S4 and
uniformedge loads N
x
xx
N
x
xx
N
xx
and N
y
yy
N
y
yy
=N
yy
, 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, N
T
xx
and N
T
yy
, 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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
to buckling, the internal stress resultants at all points within the laminate are
uniform and given by:
N
*
xx
x; y A
11
@u
o
@x
A
12
@v
o
@y
N
T
xx
N
x
xx
N
x
xx
N
xx
29a
N
*
yy
x; y A
12
@u
o
@x
A
22
@v
o
@y
N
T
yy
N
y
yy
N
y
yy
N
yy
29b
N
*
xy
x; y A
66
@u
o
@y

@v
o
@x
_ _
0 29c
It is emphasized that Eqs. (29a), (29b), and (19c) are valid prior to buckling.
After buckling has occurred and signicant out-of-plane displacements have
Figure 8 A typical mode [2,1] buckling mode, caused in a rectangular plate by
a uniaxial compressive loading N
xx
.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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:
wx; y 0 30a
M
*
xx
0; y D
11
@
2
w0; y
@x
2
D
12
@
2
w0; y
@y
2
0 30b
Along x = a:
wa; y 0 31a
M
*
xx
a; y D
11
@
2
wa; y
@x
2
D
12
@
2
wa; y
@y
2
0 31b
Along y = 0:
wx; 0 0 32a
M
*
yy
x; 0 D
12
@
2
wx; 0
@x
2
D
22
@
2
wx; 0
@y
2
0 32b
Along y = b:
wx; b 0 33a
M
*
yy
x; b D
12
@
2
wx; b
@x
2
D
22
@
2
wx; b
@y
2
0 33b
The equation of equilibrium governing out-of-plane displacements, Eq. (1c),
is [with q(x,y) = DM = 0]:
A
11
@u
o
@x
A
12
@v
o
@y
N
T
xx
_ _
@
2
w
@x
2
_ _
34
A
12
@u
o
@x
A
22
@v
o
@y
N
T
yy
_ _
@
2
w
@y
2
_ _
2 A
66
@u
o
@y

@v
o
@x
_ _ _ _
@
2
w
@x@y
_ _
D
11
@
4
w
@x
4
2D
12
2D
66

@
4
w
@x
2
@y
2
D
22
@
4
w
@y
4
0
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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
buckling phenomenon is approached. Hence substituting Eqs. (29a), (29b),
and (29c) into Eq. (34), we have:
D
11
@
4
w
@x
4
2D
12
2D
66

@
4
w
@x
2
@y
2
D
22
@
4
w
@y
4
N
xx
@
2
w
@x
2
_ _
N
yy
@
2
w
@y
2
_ _
35
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:
wx; y c sin
mpx
a
_ _
sin
npy
b
_ _
36
where mand n are positive integers and c is an unknown constant representing
the maximum out-of-plane deection. It is easy to show that Eq. (36) satises
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 deections is satised. Sub-
stituting Eq. (36) into Eq. (35) and simplifying, we nd that in order for
equilibrium to be maintained, it is required that:
D
11
mp
a
_ _
4
2D
12
2D
66

mp
a
_ _
2
np
b
_ _
2
D
22
np
b
_ _
4
N
xx
mp
a
_ _
2
N
yy
np
n
_ _
2
37
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 eld,
i.e., m and n dene how w(x,y) varies with x and y, and consequently dene
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 N
xx
and N
yy
, it is
convenient to rearrange Eq. (37) for three dierent loading conditions. As
explained above, N
xx
and/or N
yy
must be compressive to cause buckling.
Assume for the moment that a constant transverse tension N
yy
z0 is applied,
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
in which N
xx
case must be compressive to cause buckling. Solving Eq. (37) for
N
xx
, we nd:
N
xx

p
2
m
2
a
2
b
2
_
D
11
mb
4
2D
12
2D
66
mnab
2
38
D
22
na
4
N
yy
na
2
b
p
_ _
2
_
Using the denition of the plate aspect ratio, R = a/b, this result can also be
written as:
N
xx

p
2
ma
2
_
D
11
m
4
2D
12
2D
66
mnR
2
39
D
22
nR
4
N
yy
naR
p
_ _
2
_
Hence, given some value for N
yy
z0 and assumed integer values for m and n,
either Eq. (38) or (39) can be used to calculate a corresponding value for N
xx
.
Because m and n can be any combination of positive integers, there are an
innite number of values for N
xx
that satisfy Eq. (38) or (39). The critical
buckling load, denoted N
xx
c
, corresponds to the particular combination of m
and n that leads to the value of N
xx
with lowest magnitude. The combination
of m and n that correspond to this lowest load dene 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 N
yy
z 0, all variables that
appear on the right side of the equality sign are algebraically positive. Hence
N
xx
c
must be algebraically negative, i.e., the critical buckling load N
xx
c
is
predicted to be compressive, as would be expected. Secondly, the minimum
magnitude of N
xx
c
will always correspond to n = 1 because N
yy
z 0. Finally,
note that if the constant transverse load N
yy
is increased, then the magnitude
of N
xx
c
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 N
xx
z 0 is applied, and that a compressive load in the y direction
causes buckling. Solving Eq. (37) for N
yy
, we nd:
N
yy

p
2
n
2
a
4
b
2
_
D
11
mb
4
2D
12
2D
66
mnab
2
40
D
22
na
4
N
xx
mab
p
_ _
2
_
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Using the plate aspect ratio, R = a/b, this result can be written as:
N
yy

p
2
nbR
2

2
_
D
11
m
4
2D
12
2D
66
mnR
2
D
22
nR
4
N
xx
ma
p
_ _
2
_
41
As before, m and n can be any combination of positive integers, and therefore
there are an innite number of values for N
yy
that satisfy Eq. (40) or (41). The
critical buckling load N
yy
c
corresponds to the particular combination of m
and n that leads to the value of N
yy
with lowest magnitude, and the
combination of m and n that correspond to this load dene the predicted
buckling mode. Because it has been assumed that N
xx
z0, m = 1 in all cases,
and if the constant tensile load N
xx
is increased, the magnitude of the critical
buckling load N
yy
c
will increase.
Finally, consider a third loading condition in which buckling is caused
by the simultaneous increase in both N
xx
and N
yy
. Further, assume the two
loads are linearly related. That is, assume N
yy
= kN
xx
, where k is a known
constant. Substituting this relation into Eq. (37) and solving for N
xx
, we nd:
N
xx

p
2
ab
2
D
11
mb
4
2D
12
2D
66
mnab
2
D
22
na
4
_ _
mb
2
kna
2
_ _
42
As before, m and n can be any combination of positive integers, and therefore
there are an innite number of values for N
xx
that satisfy Eq. (42). The critical
buckling condition, dened by the two simultaneous loads, N
xx
c
and N
yy
c
=
kN
xx
c
, corresponds to the particular combination of m and n that leads to the
value of N
xx
(and N
yy
) with lowest magnitude. The combination of m and n
that correspond to this load condition denes the predicted buckling mode.
Note that if k < 0, then N
xx
and N
yy
are of opposite algebraic signs. This
implies that there are two distinct buckling load conditions, one in which
(N
xx
< 0, N
yy
>0), and a second in which (N
xx
>0, N
yy
<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 N
xx
, by N
yy
, or by some combination thereof.
This is not the case. Other forms of loading may cause buckling, e.g., a shear
load N
xy
. In this chapter, only buckling caused by N
xx
and/or N
yy
is con-
sidered. The reader interested in buckling caused by other types of loading is
referred to Refs. 4 and 5.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 nite 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 stiness. 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
A [(0
2
/90)
2
]
s
graphiteepoxy laminate is cured at 175jC and then cooled to
room temperature (20jC). After cooling, the at laminate is trimmed to in-
plane dimensions of 300 150 mmand mounted in an assembly that provides
type S4 simple supports along all four edges. The x axis is dened parallel to
the 300 mm edge (i.e., a = 0.3 m; b = 0.15 m).
(a) Predict the critical buckling load N
xx
c
and mode for this laminate, if
0 V N
yy
V 400 kN/m.
(b) Predict the critical buckling load N
yy
c
and mode for this laminate, if
0 V N
xx
V 400 kN/m.
Use the properties listed for graphiteepoxy in Table 3 of Chap. 3, and assume
each ply has a thickness of 0.125 mm.
Solution. A [(0
2
/90)
2
]
s
graphiteepoxy 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.
Part (a). It is noted that n = 1, because N
yy
z 0. Equation (39) becomes in
this case:
N
xx

p
2
0:09m
2
40:1m
4
65:34m
2
172:8 N
yy
0:6
p
_ _
2
_ _
A plot of the critical buckling load for 0 V N
yy
V 400 kN/m is presented in
Fig. 9(a). As expected, N
xx
c
is increased as N
yy
is increased. A change in
buckling mode also occurs as N
yy
is increased. The plate buckles in mode
[2,1] over the range 0 V N
yy
< 35 kN/m, in mode [3,1] over the range 35 kN/
m V N
yy
< 150 kN/m, and in mode [4,1] over the range 150 kN/m V N
yy
< 400 kN/m. These buckling modes are illustrated in Fig. 9(b,c,d),
respectively.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Part (b). It is noted that m=1, because N
xx
z 0. Equation (41) becomes in
this case:
N
yy

p
2
0:36n
2
40:1 65:34n
2
172:8n
4
N
xx
0:09
p
2
_ _
A plot of the critical buckling load for 0 V N
xx
V 400 kN/m is presented in
Fig. 10(a). As expected, N
yy
c
is increased as N
xx
is increased. A change in
Figure 9 Buckling response of the [(0
2
/90)
2
]
s
graphiteepoxy plate considered
in Sample Problem 5.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
buckling mode also occurs as N
xx
is increased. The plate buckles in mode [1,1]
over the range 0 VN
xx
<72 kN/m, and in mode [1,2] over the range 72 kN/m
V N
xx
< 400 kN/m. These buckling modes are illustrated in Fig. 10(b,c),
respectively.
Note that the magnitudes of N
yy
c
calculated in part (b) are far lower than
those calculated for N
xx
c
in part (a). This pronounced dierence is largely
because of the stacking sequence involved. For the [(0
2
/90)
2
]
s
laminate under
Figure 10 Buckling response of the [(0
2
/90)
2
]
s
graphiteepoxy plate consid-
ered in Sample Problem 5.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
consideration, 8 of 12 plies are 0j plies (i.e., plies with bers parallel to the x
axis). Hence the resistance to buckling because of a compressive load N
xx
is
far higher than resistance to buckling because of a compressive N
yy
.
Sample Problem 6
A structure is being designed that will involve a [(0
2
/90)
2
]
s
graphiteepoxy
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 N
xx
(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.
Solution. A [(0
2
/90)
2
]
s
graphiteepoxy 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.
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 [(0
2
/90)
2
]
s
graphiteepoxy panel considered in Sample Problem 6.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
ratio therefore varies over 1 VRV5. Because transverse loading is zero
(N
yy
= 0), n = 1 and Eq. (39) becomes:
N
xx

p
2
ma
2
40:1m
4
16:34mR
2
10:8R
4
_ _
Aplot of the predicted critical buckling load over the specied 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
square plate), buckling is predicted to occur at N
xx
c
= 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 N
xx
and/or N
yy
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 at symmetric laminate was subjected to uniform
boundary edge loads N
x
xx
N
x
xx
and N
y
yy
= N
y
yy
. Because we specied
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 specied
edge loads.
We now wish to consider buckling caused by a dierent 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 conguration. 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):
u
c
o
x; y a
xx
DT
c
x
v
c
o
x; y a
yy
DT
c
y
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
where a
xx
and a
yy
are the eective thermal expansion coecients 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 0 : For x a :
w0; y 0 wa; y 0
M
*
xx
0; y 0 M
*
xx
a; y 0
u
o
0; y u
x
o
y0 u
o
a; y u
x
o
y a
xx
DT
c
a
v
o
0; y v
x
o
y a
yy
DT
c
y v
o
0; y v
x
o
y a
yy
DT
c
y
For y 0 : For y b :
wx; 0 0 wx; b 0
M
*
yy
x; 0 0 M
*
yy
x; b 0
u
o
x; 0 u
y
o
x a
xx
DT
c
x u
o
x; b u
y
o
x a
xx
DT
c
a
v
o
x; 0 v
y
o
x 0 v
o
x; b v
y
o
y a
yy
DT
c
b
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
satised, repeated here for convenience:
D
11
mp
a
_ _
4
2D
12
2D
66

mp
a
_ _
2
np
b
_ _
2
D
22
np
b
_ _
4
N
xx
mp
a
_ _
2
N
yy
np
b
_ _
2
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
In the present case, loads N
xx
and N
yy
are caused by a deviation from room
temperature, and are given by:
N
xx
DT
c
DTA
11
a
xx
A
12
a
yy
43a
N
yy
DT
c
DTA
12
a
xx
A
22
a
yy
43b
Substituting Eqs. (43a) and (43b) into Eq. (37), solving for the temperature
dierence, and using the denition of the plate aspect ratio, R = a/b, we
obtain:
DT
c
DT
p
2
b
2
R
4
D
11
m
4
2D
12
2D
66
mnR
2
D
22
nR
4
n
2
A
12
a
xx
A
22
a
yy

m
R
_ _
2
A
11
a
xx
A
12
a
yy

_
_

_ 44a
It is convenient to express the temperature dierences involved as:
DT
c
DT T
RT
T
where T
RT
=room temperature (i.e., the temperature at which the laminate
was mounted in the simple support xture); T = current temperature.With
this change in notation we have:
T
p
2
b
2
R
4
D
11
m
4
2D
12
2D
66
mnR
2
D
22
nR
4
n
2
A
12
a
xx
A
22
a
yy

m
R
_ _
2
A
11
a
xx
A
12
a
yy

_
_

_T
RT
44b
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, T
RT
. 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 innite number of values of temperature T that will satisfy Eq. (44b). The
temperature at which thermal buckling will occur, denoted T
bk
, 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 dene 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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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:
T
bk

p
2
b
2
R
4
D
11
2D
12
2D
66
R
2
D
22
R
4
A
12
a
xx
A
22
a
yy

1
R
2
A
11
a
xx
A
22
a
yy

_
_

_ T
RT
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:
T
bk

p
2
D
22
b
2
A
12
a
xx
A
22
a
yy

T
RT
Sample Problem 7
Two [0
2
/90)
2
]
s
graphiteepoxy 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 graphiteepoxy in Table 3 of
Chap. 3, and assume each ply has a thickness of 0.125 mm.
Solution. A [(0
2
/90)
2
]
s
graphiteepoxy laminate was also considered in
Sample Problem 1, and numerical values for the [ABD] matrix and eective
thermal expansion coecients are listed there. The aspect ratios involved in
this problem are:
R 300 mm=150 mm 2
and
R 3000 mm=150 mm 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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 deections
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 ve dierent materials may be
dened. Numerical values must be dened 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 graphiteepoxy:
E
11
170 10
9
Pa E
22
10 10
9
Pa v
12
0:30 G
12
13 10
9
Pa
Because 1 Pa = 1 N/m
2
, all other lengths must be input in meters. For ex-
ample, ply thickness must be input in meters (not millimeters). Atypical value
would be t
k
=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 graphiteepoxy material system:
E
11
25:010
6
psi E
22
1:510
6
psi v
12
0:30 G
12
1:910
6
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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 deections caused by the combined
eects 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 PRELIMINARY DISCUSSION
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 A
16
= A
26
= D
16
= D
26
=
N
xy
T
= N
xy
M
= 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 D
16
,D
26
p 0 for this stacking sequence.
571
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 innite 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 nite 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
signicant gures.
The distinction between an exact and approximate solution does not
refer to the need to use a nite number of terms during a numerical
evaluation. Rather, the distinction refers to the rigor with which the boundary
conditions are satised. 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 classied 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 deection 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 satised exactly. For example, in the solutions for simply
supported specially orthotropic plates discussed in Chap. 10 all geometric and
static boundary conditions are satised exactly by the solutions presented. In
contrast, in an approximate solution, only the geometric boundary conditions
are specied and enforced directly. Static boundary conditions are not en-
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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. Briey stated, the principle of minimum
potential energy states that for all possible displacement elds that satisfy
given boundary conditions, the displacement eld 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
equilibriumand 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, suce 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 elds induced in the body are known. During a typical
structural analysis, the elastic properties and the external forces are known,
but the displacement elds are not (in fact, the objective during a structural
analysis is often to determine the displacement elds induced by some
specied loading). During application of the Ritz method it is assumed, in
eect, that the functional form of the displacement elds are known. For
general thin-plate problems the midplane displacement elds are assumed to
be of the form:
u
o
x; y

M
1
m 1

N
1
n 1
a
mn
U
mn
x; y 1a
v
o
x; y

M
2
m 1

N
2
n 1
b
mn
V
mn
x; y 1b
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
wx; y

M
3
m 1

N
3
n 1
c
mn
W
mn
x; y 1c
where a
mn
, b
mn
, and c
mn
are unknown constants; and U
mn
, V
mn
, and W
mn
are
known functions that vary over x and y. In general, the number of terms used
to describe each displacement eld may dier (e.g., M
1
does not necessarily
equal M
2
or M
3
). Also, the number of terms used to describe the variation in x
and y may dier (e.g., M
1
does not necessarily equal N
1
).
Having assumed the functional form for the displacement elds as
represented by Eqs. (1a)(1c), then the potential energy P of an elastic plate
subjected to specied external loading can be calculated. Note that the values
of constants a
mn
, b
mn
, and c
mn
in Eqs. (1a)(1c) eectively dene the
magnitudes of the displacement elds. Hence, as constants a
mn
, b
mn
and/or
c
mn
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 eld actually adopted by a solid body is one in which the
potential energy is a minimum. This condition is therefore dened by the
following criteria:
BP
Ba
mn
0
m 1; 2; . . . ; M
1
n 1; 2; . . . ; N
1
_
2a
BP
Bb
mn
0
m 1; 2; . . . ; M
2
n 1; 2; . . . ; N
2
_
2b
BP
Bc
mn
0
m 1; 2; . . . ; M
3
n 1; 2; . . . ; N
3
_
2c
Equations (2a)(2c) lead to (M
1
N
1
)+(M
2
N
2
)+(M
3
N
3
) equations
that must be satised simultaneously. Hence, by solving these equations for
constants a
mn
, b
mn
, and c
mn
, the magnitudes of the displacement elds 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 elds actually
induced in the structure.
It is seen therefore that solutions obtained using the Ritz method are
based on the functions U
mn
, V
mn
, and W
mn
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 dierentiable 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 U
mn
(x,y),
V
mn
(x,y), and W
mn
(x,y) are selected to satisfy the prevailing geometric
boundary conditions, but the static boundary conditions are not considered
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
during this selection. Hence, these functions may, or may not, satisfy the pre-
vailing static boundary conditions. If the functions U
mn
(x,y), V
mn
(x,y), and
W
mn
(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 U
mn
, V
mn
, and W
mn
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 U
mn
, V
mn
, and W
mn
that satisfy both geometric and static boundary conditions will be illustrated
for a simply supported plate. Functions U
mn
, V
mn
, and W
mn
are usually
selected to be separable functions of x and y. That is, function W
mn
, for
example, is usually selected to be of the form:
W
mn
X
m
xY
n
y
For simply supported plates, it is common to assume that X
m
and Y
n
are
sinusoidal functions of x and y:
X
m
x sin
_
mpx
a
_
Y
n
y sin
_
npy
b
_
3
where a and b are the length and width of a rectangular composite plate,
respectively. Thus, based on this selection, Eq. (1c) becomes:
wx; y

M
3
m 1

N
3
n 1
c
mn
sin
_
mpx
a
_
sin
_
npy
b
_
4
These functional forms for X
m
and Y
n
are appropriate for simply supported
plates because they satisfy the geometric boundary conditions, regardless of
the magnitudes of c
mn
. 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 satised along edges y=0,b):
D
11
B
2
w
Bx
2
D
12
B
2
w
By
2
2D
16
B
2
w
BxBy
M
*
xx
0
Upon substituting the assumed displacement eld [Eq. (4)] into this static
boundary condition, the rst two terms lead to:
D
11
B
2
w
Bx
2
D
11

M
3
m 1

N
3
n 1
c
mn
mp
a
_ _
2
sin
mpx
a
_ _
sin
npy
b
_ _
_ _
0;
for x 0; a
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
D
12
B
2
w
By
2
D
12

M
3
m 1

N
3
n 1
c
mn
_
np
b
_
2
sin
_
mpx
a
_
sin
_
npy
b
_
_ _
0;
for x 0; a
However, the third term leads to:
2D
16
B
2
w
BxBy
2D
16

M
3
m 1

N
3
n 1
c
mn
_
mp
a
__
np
b
_
cos
_
mpx
a
_
cos
_
npy
b
_
_ _
Evaluating this third term along x=0, a:
2D
16
B
2
w
BxBy

2p
2
D
16
ab

M
3
m 1

N
3
n 1
mn c
mn
cos
_
npy
b
_
; for x 0; a
Hence, the static boundary condition associated with simply supported
edges x = 0,a is satised only if:

M
3
m 1

N
3
n 1
mnc
mn
cos
npy
b
_ _
0; for 0 Vy Vb 5
If only a single term is used to describe the displacement eld (i.e., if M
3
=
N
3
=1), then the static boundary condition is clearly not satised even ap-
proximately because:
c
11
cos
py
b
_ _
p 0; for all 0VyVb
However, if the number of terms used is increased (i.e., as M
3
and/or N
3
are
increased), then Eq. (5) may be satised more and more exactly, through
proper selection of the values of constants c
mn
. For example, if four terms are
used (i.e., if M
3
=N
3
=2), then the static boundary condition along edges
x=0,a is satised if:
c
11
cos
py
b
_ _
2c
12
cos
2py
b
_ _
2c
21
cos
py
b
_ _
4c
22
cos
2py
b
_ _
0;
for all 0 Vy Vb
It is now possible to satisfy the static boundary condition along edges x = 0,
a exactly by setting c
21
= c
11
/2 and c
22
= c
12
/2. Of course, selecting
constants that satisfy these requirements may not lead to the displacement
eld that represents the state of minimum potential energy because constants
c
11
, c
12
, c
21
, and c
22
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 eld, it is possible to satisfy the static
boundary conditions along all four edges more and more exactly.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
In general then, the validity of a solution based on the Ritz approach is
increased as the number of terms is increased. Assuming functions X
m
and Y
n
are selected to satisfy the geometric boundary conditions, then a Ritz analysis
will converge toward the exact solution as M
3
and/or N
3
are increased. It is
also appropriate to note that the diculty in obtaining an exact solution is due
to the D
16
term. If D
16
= 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 satised exactly by the assumed sinusoidal
form for X
m
and Y
n
.
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 specic 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 elds u
o
(x,y) and v
o
(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 u
o
(x,y) or v
o
(x,y) in terms of the
double series as listed as Eqs. (1a,b) nor to determine the magnitudes of in-
plane displacement elds using Eqs. (2a,b). For the problems considered
herein, only w(x,y) is unknown. The out-of-plane displacement eld 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 X
m
and Y
n
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 M
3
!M and N
3
!N; thus, w(x,y) will be written:
wx; y

M
m 1

N
n 1
c
mn
sin
_
mpx
a
_
sin
_
npy
b
_
6
Although not discussed in this text, solutions are also available based on
alternate (nonsinusoidal) function forms for X
m
and Y
n
. Alternate forms
include polynomials in x and y:
X
m
x x
2
ax
2
x
m1
Y
n
y y
2
ay
2
y
n1
forms involving other trigonometric functions:
X
m
x 1 cos
2mpx
a
_ _ _ _
Y
n
y 1 cos
2npy
b
_ _ _ _
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
or forms involving so-called beam functions:
X
m
x c
m
cos
k
m
x
a
_ _
c
m
cosh
k
m
x
a
_ _
sin
k
m
x
a
_ _
sinh
k
m
x
a
_ _
Y
n
x c
n
cos
k
n
x
b
_ _
c
n
cosh
k
n
x
b
_ _
sin
k
n
x
b
_ _
sinh
k
n
x
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
elds induced in a symmetric laminate for specic 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 roomtemperature. The laminate is then trimmed to the desired nal
rectangular dimensions a b. Finally, the laminate is subjected to constant
and uniformly distributed stress resultants (N
xx
, N
yy
, and/or N
xy
) along all
four edges and/or a further change in temperature. Although we will not
consider the possibility of a uniformchange 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:
N
*
xx
0; y N
*
xx
a; y N
xx
N
*
yy
x; 0 N
*
yy
x; b N
yy
N
*
xy
0; y N
*
xy
a; y N
*
yx
x; 0 N
*
yx
x; b N
xy
*
As mentioned in Sec. 6.2 of Chap. 6, in practice, the stress- and strain-free temperature is often
2050jC below the nal 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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
For a symmetric laminate subjected to uniform in-plane stress resultants and
a change in temperature, Eq. (45) of Chap. 6 becomes:
e
0
xx
e
0
yy
c
0
xy
j
xx
j
yy
j
xy
_

_
_

a
11
a
12
a
16
0 0 0
a
12
a
22
a
26
0 0 0
a
16
a
26
a
66
0 0 0
0 0 0 d
11
d
12
d
16
0 0 0 d
12
d
22
d
26
0 0 0 d
16
d
26
d
66
_

_
_

_
N
xx
N
T
xx
N
yy
N
T
yy
N
xy
N
T
xy
0
0
0
_

_
_

_
As in earlier chapters, we assume innitesimal strain levels, and therefore
midplane strains are related to midplane displacements according to Eq. (10)
of Chap. 6:
e
0
xx

Bu
o
Bx
e
0
yy

Bv
o
By
c
0
xy

Bu
o
By

Bv
o
Bx
Consequently, midplane displacement elds are related to the stress and ther-
mal resultants as follows:
Bu
o
Bx
a
11
N
xx
N
T
xx
a
12
N
yy
N
T
yy
a
16
N
xy
N
T
xy
7a
Bv
o
By
a
12
N
xx
N
T
xx
a
22
N
yy
N
T
yy
a
26
N
xy
N
T
xy
7b
Bu
o
By

Bv
o
Bx
a
16
N
xx
N
T
xx
a
26
N
yy
N
T
yy
a
66
N
xy
N
T
xy
7c
Integrating Eq. (7a) with respect to x, we nd:
u
o
x; y a
11
N
xx
N
T
xx
_ _
a
12
N
yy
N
T
yy
_ _
a
16
N
xt
N
T
xy
_ _ _ _
x
f
1
y k
1
where f
1
( y) is an unknown function of y (only); and k
1
is an unknown constant
of integration. Similarly, integrating Eq. (7b) with respect to y, we nd:
v
o
x; y a
12
N
xx
N
T
xx
_ _
a
22
N
yy
N
T
yy
_ _
a
26
N
xy
N
T
xy
_ _ _ _
y
f
2
x k
2
where f
2
(x) is an unknown function of x (only); and k
2
is a second unknown
constant. Without a loss ingenerality, we assume that midplane displacements
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
are zero at the origin (i.e., at x = y = 0) and, consequently, k
1
= k
2
= 0.
Substituting these expressions for u
o
(x,y) and v
o
(x,y) into Eq. (7c), we nd:
Bu
o
By

Bv
o
Bx

Bf
1
By

Bf
2
Bx
a
16
N
xx
N
T
xx
a
26
N
yy
N
T
yy

a
66
N
xy
N
T
xy

Because all terms on the right side of the equality are known constants, it
follows that f
1
and f
2
must be at most linear function of y and x, respectively:
f
1
y k
3
y
f
2
x k
4
x
Hence, we can write:
k
3
k
4
a
16
N
xx
N
T
xx
a
26
N
yy
N
T
yy
a
66
N
xy
N
T
xy

Because all quantities that appear in this relation are constants, k


3
and k
4
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 innitesimal rotation vector in the xy plane x
xy
(which
represents rigid body motion of the plate) is zero. The innitesimal rotation
vector is given by (3):
x
xy

1
2
Bu
o
By

Bv
o
Bx
_ _
Requiring that x
xy
=0 leads to:
k
3
k
4

1
2
a
16
N
xx
N
T
xx
_ _
a
26
N
yy
N
T
yy
_ _
a
66
N
xy
N
T
xy
_ _ _ _
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:
u
o
x; y a
11
N
xx
N
T
xx
_ _
a
12
N
yy
N
T
yy
_ _
a
16
N
xy
N
T
xy
_ _ _ _
x

1
2
a
16
N
xx
N
T
xx
_ _
a
26
N
yy
N
T
yy
_ _
a
66
N
xy
N
T
xy
_ _ _ _
y
v
o
x; y
1
2
a
16
N
xx
N
T
xx
_ _
a
26
N
yy
N
T
yy
_ _
a
66
N
xy
N
T
xy
_ _ _ _
x
a
12
N
xx
N
T
xx
_ _
a
22
N
yy
N
T
yy
_ _
a
26
N
xy
N
T
xy
_ _ _ _
y
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
These expressions can be simplied through the use of the eective thermal
expansion coecients of the laminate, dened by Eqs. (73a) of Chap. 6 and
repeated here for convenience:
a
xx

1
DT
a
11
N
T
xx
a
12
N
T
yy
a
16
N
T
xy
_ _
a
yy

1
DT
a
12
N
T
xx
a
22
N
T
yy
a
26
N
T
xy
_ _
6:73a
a
xy

1
DT
a
16
N
T
xx
a
26
N
T
yy
a
66
N
T
xy
_ _
Hence, we see that the midplane displacement elds can be written:
u
o
x; y a
11
N
xx
a
12
N
yy
a
16
N
xy
DTa
xx
_ _
x

1
2
a
16
N
xx
a
26
N
yy
a
66
N
xy
DTa
xy
_ _
y
v
o
x; y
1
2
a
16
N
xx
a
26
N
yy
a
66
N
xy
DTa
xy
_ _
x
a
12
N
xx
a
22
N
yy
a
26
N
xy
DTa
yy
_ _
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 elds
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
dened as a simple support. The distinction between the dierent types of
simple supports has to do with the boundary condition assumed for the in-
plane displacement elds. For example, to dene a simple support of type S1,
one species known values of in-plane displacements, whereas to dene a
simple support of type S4, one species known values of in-plane stress
resultants. For the problems considered here, we are able to calculate the
midplane displacement eld induced by a specied 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 uniformbut rather varied along the plate edges, then
it would be more dicult (and, in most cases, impossible) to determine
associated in-plane displacement elds.
Because we have limited discussion to cases in which stress resultants
applied to the edges are constant and uniform, the midplane displacement
7
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
elds 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.
Specically, 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 N
xx
, N
yy
, and/or N
xy
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
denition 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
dened using units of force/area. The product ( q)(dx)(dy) represents the
transverse force provided by q(x,y) acting over an innitesimal element of area
(dx)(dy). The distance through which this force travels equals the out-of-plane
deection 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:
W
__
qx; ywx; ydxdy 8
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:
U
1
2
_ _ _
r
xx
e
xx
r
yy
e
yy
r
zz
e
zz
s
yz
c
yz
s
xz
c
xz
s
xy
c
xy
dxdydz
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 (r
zz
=s
yz
=s
xz
=0),
we can immediately discard terms involving these stress components:
U
1
2
_ _ _
r
xx
e
xx
r
yy
e
yy
s
xy
c
xy
dxdydz 9
Next, the stresses in ply k of the laminate are given by Eqs. (30) of Chap. 5,
which become (for DM = 0):
r
xx
r
yy
r
xy
_
_
_
_
_
_
k

Q
11
Q
12
Q
16
Q
12
Q
22
Q
26
Q
16
Q
26
Q
66
_
_
_
_
k
e
xx
DTa
xx
e
yy
DTa
yy
c
xy
DTa
xy
_
_
_
_
_
_
k
Substituting these expressions for ply stresses into Eq. (9) and rearranging,
we nd:
U
1
2
__ _
Q
k
11
e
2
xx
2Q
k
12
e
xx
e
yy
2Q
k
16
e
xx
c
xy
2Q
k
26
e
yy
c
xy
Q
k
22
e
2
yy
Q
k
66
c
2
xy
_
DT a
k
xx
Q
k
11
a
k
yy
Q
k
12
a
k
xy
Q
k
16
_ _
e
xx
10
DT a
k
xx
Q
k
12
a
k
yy
Q
k
22
a
k
xy
Q
k
26
_ _
e
yy
DT a
k
xx
Q
k
16
a
k
yy
Q
k
26
a
k
xy
Q
k
66
_ _
c
xy
_
dxdydz
We now invoke the Kirchho 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:
e
xx
e
0
xx
zj
xx
e
yy
e
0
yy
zj
yy
c
xy
c
0
xy
zj
xy
Substitution of Eqs. (12) of Chap. 6 into Eq. (10) results in:
U
1
2
_ _ _
Q
11
e
0
xx
_ _
2
2ze
0
xx
j
xx
z
2
j
2
xx
_ _ _
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
2Q
12
e
0
xx
e
0
yy
ze
0
xx
j
yy
ze
0
yy
j
xx
z
2
j
xx
j
yy
_ _
2Q
16
e
0
xx
c
0
xy
ze
0
xx
j
xy
zc
0
xy
j
xx
z
2
j
xx
j
xy
_ _
2Q
26
e
0
yy
c
0
xy
ze
0
yy
j
xy
zc
0
xy
j
yy
z
2
j
yy
j
xy
_ _
Q
22
_
e
0
yy
_ _
2
2ze
0
yy
j
yy
z
2
j
2
yy
_
11
Q
66
_
c
0
xy
_ _
2
2zc
0
xy
j
xy
z
2
j
2
xy
_
DT
_
a
xx
Q
11
a
yy
Q
12
a
xy
Q
16
_
e
0
xx
zj
xx
_ _
DT
_
a
xx
Q
12
a
yy
Q
22
a
xy
Q
26
_
e
0
yy
zj
yy
_ _
DT
_
a
xx
Q
16
a
yy
Q
26
a
xy
Q
66
_
c
0
xy
zj
xy
_ __
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 specic examples are:
_
t=2
t=2
Q
11
dz; which after integration becomes A
11
_
t=2
t=2
Q
11
zdz; which after integration becomes B
11
_
t=2
t=2
Q
11
z
2
dz; which after integration becomes D
11
_
t=2
t=2
DT a
xx
Q
11
a
yy
Q
12
a
xy
Q
16
_ _
dz;
which after integration becomes N
T
xx
_
t=2
t=2
DT a
xx
Q
11
a
yy
Q
12
a
xy
Q
16
_ _
zdz;
which after integration becomes M
T
xx
Hence, after integration Eq. (11) can be written:
U
1
2
__
A
11
e
0
xx
_ _
2
2A
12
e
0
xx
e
0
yy
A
22
e
0
yy
_ _
2
2 A
16
e
0
xx
A
26
e
0
yy
_ _
c
0
xy
_
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
A
66
c
2
xy
2B
11
e
0
xx
j
xx
2B
12
e
0
xx
j
yy
e
0
yy
j
xx
_ _
2B
16
e
0
xx
j
xy
c
0
xy
j
xx
_ _
2B
26
e
0
yy
j
xy
c
0
xy
j
yy
_ _
2B
22
e
0
yy
j
yy
2B
66
c
0
xy
j
xy
D
11
j
2
xx
2D
12
j
xx
j
yy
12
2 D
16
j
xx
D
26
j
yy
_ _
j
xy
D
22
j
2
yy
D
66
j
2
xy
N
T
xx
e
0
xx
N
T
yy
e
0
yy
N
T
xy
c
0
xy
M
T
xx
j
xx
M
T
yy
j
yy
M
T
xy
j
xy
_
dxdy
Because we have limited discussion to symmetric laminates, B
ij
= M
xx
T
=
M
yy
T
= M
xy
T
= 0 in all cases considered. Equations (12) therefore simplies
to:
U
1
2
_ _ _
A
11
e
0
xx
_ _
2
2A
12
e
0
xx
e
0
yy
A
22
e
0
yy
_ _
2
2 A
16
e
0
xx
A
26
e
0
yy
_ _
c
0
xy
A
66
c
2
xy
D
11
j
2
xx
2D
12
j
xx
j
yy
2
_
D
16
j
xx
D
26
j
yy
_
j
xy
13
D
22
j
2
yy
D
66
j
2
xy
N
T
xx
e
0
xx
N
T
yy
e
0
yy
N
T
xy
c
0
xy
_
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 N
ij
are applied to the
laminate rst, before the application of any other load(s) that cause bending.
Our rst step is therefore to calculate the strain energy within a laminate
subjected to N
ij
only even if other loads are involved in the problem under
consideration. We will label this component of strain energy U
I
. Calculation
of U
I
is straightforward because we have limited our discussion to symmetric
laminates. That is, for a symmetric laminate, stress resultants N
ij
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 B
ij
p 0. If this were the case, then N
ij
would also contribute
to curvatures; furthermore, both M
ij
and q(x,y) would contribute to
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
midplane strains. These coupling eects would greatly complicate our
analysis. Because symmetry has been assumed, the laminate remains at
following application of N
ij
. Therefore, the only loads that contribute to
midplane strains are stress resultants N
ij
. The strain energy associated with N
ij
only (i.e., the strain energy component U
I
) can therefore be obtained fromEq.
(13) simply by setting j
xx
=j
yy
= j
xy
=0. Ultimately our expression for U
I
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 j
xx
, j
yy
, and j
xy
. We
will label this component of the strain energy U
II
. Because the laminate is
symmetric, midplane curvatures are caused by the combined eects of the
transverse load q(x,y) and/or bending moment resultants M
ij
, but are
independent of stress resultants N
ij
. Strain energy component U
II
can be
obtained from Eq. (13) simply by setting e
0
xx
=e
0
yy
=e
0
xy
=0. Ultimately, our
expression for U
II
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 U
I
associated with stress resultants N
ij
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 U
III
. We assume that the stress resultants N
ij
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
U
III
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)=q
o
, will dier from the mathematical form of W if the transverse
load varies sinusoidally: q(x,y)= q
o
sin(px/a)sin(py/b).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 = U
I
+U
II
+U
III
W.
3.1 Evaluation of Strain Energy Component U
I
Strain energy component U
I
is obtained from Eq. (11) by setting j
xx
=j
yy
=
j
xy
=0:
U
I

1
2
__
A
11
e
0
xx
_ _
2
2A
12
e
0
xx
e
0
yy
A
22
e
0
yy
_ _
2
2 A
16
e
0
xx
A
22
e
0
yy
_ _
c
0
xy
_
A
66
c
2
xy
N
T
xx
e
0
xx
N
T
yy
e
0
yy
N
T
xy
c
0
xy
_
dxdy
As in earlier chapters, we assume innitesimal strain levels and therefore
midplane strains are related to midplane displacements according to Eq. (10)
of Chap. 6:
e
0
xx

Bu
o
Bx
e
0
yy

Bv
o
By
c
0
xy

Bu
o
By

Bv
o
Bx
With this substitution, we have:
U
I

1
2
_ _
A
11
Bu
o
Bx
_ _
2
2A
12
Bu
o
Bx
_ _
Bv
o
By
_ _
A
22
Bv
o
By
_ _
2
_
2 A
16
Bu
o
Bx
A
26
Bv
o
By
_ _
Bu
o
By

Bv
o
Bx
_ _
A
66
Bu
o
By

Bv
o
Bx
_ _
2
N
T
xx
Bu
o
Bx
_ _
N
T
yy
Bv
o
By
_ _
N
T
xy
Bu
o
By

Bv
o
Bx
_ _
_
dxdy
Equation (14) gives the strain energy component U
I
for any in-plane
displacement elds u
o
(x,y) and v
o
(x,y). We will now integrate this expression
using the displacement elds 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:
C
1
a
11
N
xx
a
12
N
yy
a
16
N
xy
DTa
xx
_ _
C
2
a
12
N
xx
a
22
N
yy
a
26
N
xy
DTa
yy
_ _
15
C
3

1
2
a
16
N
xx
a
26
N
yy
a
66
N
xy
DTa
xy
_ _
14
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
With this change in notation, Eqs. (7) can be written:
u
o
x; y C
1
x C
3
y
v
o
x; y C
3
x C
2
y
Substituting these expressions into Eq. (14), we have:
U
I

1
2
_
b
0
_
a
0
A
11
C
1

2
2A
12
C
1
C
2
A
22
C
2

2
_
2A
16
C
1
A
26
C
2
2C
3
A
66
2C
3

2
N
T
xx
C
1

N
T
yy
C
2
N
T
xy
2C
3

_
dxdy
Evaluating this denite integral, we nd:
U
I
A
11
C
1

2
2A
12
C
1
C
2
A
22
C
2

2
_
4 A
16
C
1
C
3
A
26
C
2
C
3
A
66
C
2
3
_ _
N
T
xx
C
1
N
T
yy
C
2
2N
T
xy
C
3
_
ab
Substituting Eqs. (73b) of Chap. 6, which give the thermal stress resultants
in terms of elements of the [A] matrix, and Eq. (5), we nd that U
I
can be
written:
U
I
C
1
N
xx
C
2
N
yy
2C
3
N
xy
_
ab 16
3.2 Evaluation of Strain Energy Component U
II
As previously discussed, for symmetric laminates, strain energy component
U
II
can be obtained from Eq. (13) by setting e
0
xx
e
0
yy
e
0
xy
=0:
U
II

1
2
__
D
11
j
2
xx
2D
12
j
xx
j
yy
2D
16
j
xx
D
26
j
yy
j
xy
D
22
j
2
yy
_
D
66
j
2
xy
_
dxdy
Midplane curvatures are related to out-of-plane displacements according to
Eq. (10) of Chap. 6:
j
xx

B
2
w
Bx
2
j
yy

B
2
w
By
2
j
xy
2
B
2
w
BxBy
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
With this substitution, we have:
U
II

1
2
_ _
D
11
B
2
w
Bx
2
_ _
2
2D
12
B
2
w
Bx
2
_ _
B
2
w
By
2
_ _
_
4 D
16
B
2
w
Bx
2
_ _
D
26
B
2
w
By
2
_ _ _ _
B
2
w
BxBy
D
22
B
2
w
By
2
_ _
2
4D
66
B
2
w
BxBy
_ _
2
_
dxdy
17
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:
wx; y

M
m 1

N
n 1
c
mn
sin
_
mpx
a
_
sin
_
npy
b
_
Because Eq. (6) will be used in all problems considered, we will integrate
Eq. (17) based on this displacement eld. The following derivatives appear
in Eq. (17):
B
2
w
Bx
2

M
m 1

M
n 1
c
mn
_
mp
a
_
2
sin
_
mpx
a
_
sin
_
npy
b
_
B
2
w
By
2

M
m 1

N
n 1
c
mn
_
np
b
_
2
sin
_
mpx
a
_
sin
_
npy
b
_
B
2
w
BxBy

M
m 1

N
n 1
c
mn
_
mp
a
__
np
b
_
cos
_
mpx
a
_
cos
_
npy
b
_
Consider the rst term under the integral sign in Eq. (17). Upon substituting
the expression for (B
2
w/Bx
2
) listed above, this term becomes:
1
2
_
b
0
_
a
0
D
11
B
2
w
Bx
2
_ _
2
dxdy

1
2
_
b
0
_
a
0
D
11

M
m 1

N
n 1
c
mn
_
mp
a
_
2
sin
_
mpx
a
_
sin
_
npy
b
_
_ _

M
i 1

N
j 1
c
ij
ip
a
_ _
2
sin
ipx
a
_ _
sin
jpy
b
_ _
_ _
dxdy
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
The evaluation of this integral is greatly simplied by noting the following
identities:
_
a
0
sin
_
mpx
a
_
sin
ipx
a
_ _
dx 0; if m p i
_
b
0
sin
_
npy
b
_
sin
ipy
b
_ _
dy 0; if n p j
Hence, during integration of this term, we need only consider:
1
2
_
b
0
_
a
0
D
11
B
2
w
Bx
2
_ _
2
dxdy

1
2
_
b
0
_
a
0
D
11

M
m 1

M
n 1
c
2
mn
_
mp
a
_
4
sin
2
_
mpx
a
_
sin
2
_
npy
b
_
_ _
dxdy
After integration and evaluation, this term becomes:
1
2
_
b
0
_
a
0
D
11
B
2
w
Bx
2
_ _
2
dxdy
p
4
b
8a
3
D
11

M
m 1

N
n 1
m
4
c
2
mn
The following terms also appear in Eq. (17) and are evaluated in a similar
manner:
1
2
_
b
0
_
a
0
2D
12
B
2
w
Bx
2
_ _
B
2
w
By
2
_ _
dxdy
p
4
4ab
D
12

M
m 1

N
n 1
m
2
n
2
c
2
mn
1
2
_
b
0
_
a
0
D
22
B
2
w
By
2
_ _
2
dxdy
p
4
a
8b
3
D
22

M
m 1

N
n 1
n
4
c
2
mn
1
2
_
b
0
_
a
0
4D
66
B
2
w
BxBy
_ _
2
dxdy
p
4
2ab
D
66

M
m 1

N
n 1
m
2
n
2
c
2
mn
The remaining terms in Eq. (17) involve D
16
and D
26
. Upon substituting the
appropriate derivatives, the rst of these becomes:
1
2
_
b
0
_
a
0
4D
16
B
2
w
Bx
2
_ _
B
2
w
BxBy
_ _
dxdy
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

1
2
_
b
0
_
a
0
4D
16

M
m 1

N
n 1
c
mn
_
mp
a
_
2
sin
_
mpx
a
_
sin
_
npy
b
_
_ _

M
i 1

N
j 1
c
ij
ip
a
_ _
jp
b
_ _
cos
ipx
a
_ _
cos
jpy
b
_ _
_ _
dxdy
This integration does not simplify as readily as those considered above. We
make use of the following identities:
_
a
0
sin
_
mpx
a
_
cos
ipx
a
_ _
dx
0; if m i is even
2ma
pm
2
i
2

; if m i is odd
_

_
_
b
0
sin
_
npx
b
_
cos
jpx
b
_ _
dy
0; if n j is even
2nb
pn
2
j
2

; if n j is odd
_

_
On the basis of these identities, we nd after integration and evaluation:
1
2
_
b
0
_
a
0
4D
16
B
2
w
Bx
2
_ _
B
2
w
BxBy
_ _
dxdy

2p
2
a
2
D
16

M
m 1

N
n 1

M
i 1

N
j 1
c
mn
c
ij
m
2
nij
m
2
i
2
n
2
j
2

_ _
MINJ
where:
MI 1
m
1
i
1
NJ 1
n
1
j
1
Notice that:
MI
0; if m i is even
2; if m i is odd
_
_
_
NJ
0; if n j is even
2; if n j is odd
_
_
_
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
The nal term that appears in Eq. (17) is evaluated in a similar manner and
becomes:
1
2
_
b
0
_
a
0
4D
26
B
2
w
By
2
_ _
B
2
w
BxBy
_ _
dxdy

2p
2
b
2
D
26

M
m 1

N
n 1

M
i 1

N
j 1
c
mn
c
ij
mn
3
ij
m
2
i
2
n
2
j
2

_ _
MINJ
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:
U
II

M
m 1

N
n 1
p
4
8
c
2
mn
bm
4
a
3
D
11

2m
2
n
2
ab
D
12
2D
66

an
4
b
3
D
22
_ _ _
2p
2
mnc
mn

M
i 1

N
j 1
c
ij
ij
m
2
i
2
n
2
j
2

_ _ _

m
2
a
2
D
16

n
2
b
2
D
26
_ _
MINJ
__
3.3 Evaluation of Strain Energy Component U
Strain energy component U
III
represents the change in strain energy associ-
ated with stress resultants N
ij
caused by bending. We assume that the stress
resultants N
ij
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 e
0
xx
may be determined via Fig. 1.
The gure shows an element of length dx, which represents an innitesimal
element of the midplane that has already been deformed by stress resultants N
ij
during step 1, as previously discussed. The length dx is further increased to
length dxV if the plate is deected out of plane. From Fig. 1, we see that:
dxV dx
2

Bw
Bx
_ _
2
dx
2
_ _
1=2
dx 1
Bw
Bx
_ _
2
_ _
1=2
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 n
1=2
1
1
2
n
1
2
_ _
1
4
_ _
n
2

1
2
_ _
1
4
_ _
3
6
_ _
n
3
. . .
18
III
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
We adopt this general formula for our use by retaining only the rst two terms
and letting n = (Bw/Bx)
2
. Hence:
1
Bw
Bx
_ _
2
_ _
1=2
c
1
1
2
Bw
Bx
_ _
2
With this approximation, the new length of the element is given by:
dxV dx 1
1
2
Bw
Bx
_ _
2
_ _
The change in-plane strain e
0
xx
caused by out-of-plane displacement w(x,y)
is labeled e
0;b
xx
. Based on the above, e
0;b
xx
is given by:
e
0;b
xx

dxV dx
dx

1
2
Bw
Bx
_ _
2
19a
Using a similar approach, the change in in-plane strains e
0
yy
and c
0
xy
caused by
out-of-plane displacement w(x,y) are given by:
e
0;b
yy

1
2
Bw
By
_ _
2
19b
Figure 1 Sketch used to determine change in in-plane strain caused by out-of-
plane deflections.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
c
0;b
xy

Bw
Bx
_ _
Bw
By
_ _
19c
We can now calculate the strain energy associated with the change in in-plane
strains caused by bending. We assume in-plane stress resultants N
ij
remain
constant as bending develops. Consider, for example, the force represented
by stress resultant N
xx
acting over an innitesimal element with length and
width dx and dy. The x-directed force is (N
xx
dy), and the distance through
which this force moves as bending develops equals e
0;b
xx
dx. The incremental
strain energy associated with this force and caused by bending is therefore
dU
III
=(N
xx
dy)(e
0;b
xx
dx). Analogous expressions hold for stress resultants N
yy
and N
xy
. Hence, strain energy U
III
is given by:
U
III

__
N
xx
e
0;b
xx
N
yy
e
0;b
yy
N
xy
c
0;b
xy
_ _
dxdy 20
Equation (20) represents a general expression for U
III
. 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 N
xx
, N
yy
, and
N
xy
in terms of displacement elds, which will ultimately allow us to specify
geometric boundary conditions that represent known values of N
xx
, N
yy
, and
N
xy
. From Eq. (44) of Chap. 6, we can write (for B
ij
= DM = 0)
N
xx
A
11
e
0
xx
A
12
e
0
yy
A
16
c
0
xy
N
T
xx
A
11
Bu
o
Bx
_ _
A
12
Bv
o
By
_ _
A
16
Bu
o
By

Bv
o
Bx
_ _
N
T
xx
Similarly,
N
yy
A
12
Bu
o
Bx
_ _
A
22
Bv
o
By
_ _
A
26
Bu
o
By

Bv
o
Bx
_ _
N
T
yy
N
xy
A
16
Bu
o
Bx
_ _
A
26
Bv
o
By
_ _
A
66
Bu
o
By

Bv
o
Bx
_ _
N
T
xy
Substituting these expressions as well as Eqs. (19a,b,c) into Eq. (20), we
obtain:
U
III

1
2
_ _
A
11
Bu
o
Bx
_ _
A
12
Bv
o
By
_ _
A
16
Bu
o
By

Bv
o
Bx
_ _
N
T
xx
_ _ _
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

Bw
Bx
_ _
2
A
12
Bu
o
Bx
_ _
A
22
Bv
o
By
_ _
A
26
Bu
o
By

Bv
o
Bx
_ _
N
T
yy
_ _

Bw
By
_ _
2
2 A
16
Bu
o
Bx
_ _
A
26
Bv
o
By
_ _
A
66
Bu
o
By

Bv
o
Bx
_ _
N
T
xy
_ _

Bw
Bx
_ _
Bw
By
_ __
dxdy 21
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 N
ij
and/or a temperature change DT.
For these conditions, the in-plane displacement elds 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:
U
III

1
2
_
b
0
_
a
0
_
A
11
C
1
A
12
C
2
2A
16
C
3
N
T
xx
_ _

M
m 1

N
n 1
c
mn
_
mp
a
_
cos
_
mpx
a
_
sin
_
npy
b
_
_ _
2
A
12
C
1
A
22
C
2
2A
26
C
3
N
T
yy
_ _

M
m 1

N
n 1
c
mn
_
np
b
_
sin
_
mpx
a
_
cos
_
npy
b
_
_ _
2
2 A
16
C
1
A
26
C
2
2A
66
C
3
N
T
xy
_ _

M
m 1

N
n 1
c
mn
_
mp
a
_
cos
_
mpx
a
_
sin
_
npy
b
_
_ _

M
i 1

N
j 1
c
ij
jp
b
_ _
sin
ipx
a
_ _
cos
jpy
b
_ _
_ __
dxdy
Integration of this expression is simplied through the use of the trigono-
metric identities listed in preceding section. We obtain:
U
III

M
m 1

N
n 1
p
2
8
c
2
mn
m
2
b
a
A
11
C
1
A
12
C
2
2A
16
C
3
N
T
xx
_ _
_ _
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

n
2
a
b
A
12
C
1
A
22
C
2
2A
26
C
3
N
T
yy
_ _
_
2m
2
nc
mn
A
16
C
1
A
26
C
2
2A
66
C
3
N
T
xy
_ _

M
i 1

N
j 1
c
ij
j
m
2
i
2
n
2
j
2

_ _
MINJ
_ __
This result can be further simplied 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 nally nd:
U
III

M
m 1

N
n 1
p
2
8
c
2
mn
m
2
b
a
N
xx

n
2
a
b
N
yy
_ _
2m
2
nc
mn
N
xy
_

M
i 1

N
j 1
c
ij
j
m
2
i
2
n
2
j
2

_ _
MINJ
_ _ _
22
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
Greens strain tensor, mentioned in Sec. 14 of Chap. 2. Recall that Greens
strain tensor represents a denition of nite strains, which must be ac-
counted for during analyses involving large displacement gradients. For ex-
ample, nite strain e
xx
is dened as:
e
xx

Bu
Bx

1
2
Bu
Bx
_ _
2

Bv
Bx
_ _
2

Bw
Bx
_ _
2
_ _
Because we have incorporated the term e
0;b
xx
= 1/2(Bw/Bx)
2
during our
calculation of U
III
, it would be natural to conclude that our analysis is valid
for nite strain levels. This conclusion would be incorrect. Our analysis is
based on innitesimal strains, despite the inclusion of Eq. (18) during cal-
culation of U
III
. To perform an analysis based on energy methods that ac-
counts for nite strain levels, we would need to develop new expressions
comparable to Eqs. (16), (18), and (20) (i.e., our current expressions for U
I
,
U
II
, and U
III
, 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 nite 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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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:
W
_ _
_
qx; y
_

M
m 1

N
n 1
c
mn
sin
_
mpx
a
_
sin
_
npx
b
_
_ _
dxdy 23
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 deections of simply supported symmetric
composite panels will be predicted on the basis of a Ritz analysis. The gen-
eral approach is to rst obtain an expression for the total potential energy
P of the plate, given by:
P U
I
U
II
U
III
W
In general, P is a function of the elastic properties of the plate, plate di-
mensions, and midplane displacement elds u
o
(x,y), v
o
(x,y), and w(x,y).
For the problems considered herein, in-plane displacement elds u
o
(x,y) and
v
o
(x,y) are known, while the out-of-plane displacement eld w(x,y) is un-
known. The out-of-plane displacement eld is assumed to be of the form:
wx; y

M
m 1

N
n 1
c
mn
sin
_
mpx
a
_
sin
_
npy
b
_
The magnitude of out-of-plane deections are obtained by applying the prin-
ciple of minimum potential energy, which requires:
BP
Bc
mn
0
m 1; 2; . . . ; M
n 1; 2; . . . ; N
_
This process leads to (M N) equations that must be satised simulta-
neously. Hence, by solving these equations for constants c
mn
, 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 U
I
, U
II
, and U
III
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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
4.1 Deflections Due to a Uniform Transverse Load
Consider a composite plate subjected to a constant and uniform transverse
load, q(x,y)=q
o
. In this case, Eq. (23) becomes:
W q
o
_
b
0
_
a
0

M
m 1

N
n 1
c
mn
sin
_
mpx
a
_
sin
_
npx
b
_
_ _
dxdy
During integration of this expression, we note the following:
_
a
0
sin
_
mpx
a
_
dx
0; if m is even
2a
mp
; if m is odd
_

_
_
b
0
sin
_
npx
b
_
dy
0; if n is even
2b
np
; if n is odd
_

_
Hence, the work done by a uniform transverse load can be written:
W

M
m 1

N
n 1
abq
o
c
mn
p
2
mn
_
1
m
1
__
1
n
1
_
_ _
24
The total potential energy can now be obtained by combining Eqs. (16), (18),
(22), and (24):
P C
1
N
xx
C
2
N
yy
2C
3
N
xy
_
ab
_ _

f

M
m 1

N
n 1
p
4
8
c
2
mn
bm
4
a
3
D
11

2m
2
n
2
ab
D
12
2D
66

an
4
b
3
D
22
_ _ _
2p
2
mnc
mn

M
i 1

N
j 1
c
ij
ij
m
2
i
2
n
2
j
2

_ _
m
2
a
2
D
16

n
2
b
2
D
26
_ _
MI NJ
_ __
g
25

f

M
m 1

N
n 1
p
2
8
c
2
mn
m
2
b
a
N
xx

n
2
a
b
N
yy
_ _ _
2m
2
nc
mn
N
xy

M
i 1

N
j 1
c
ij
j
m
2
i
2
n
2
j
2

_ _
MI NJ
_ __
g

M
m 1

N
n 1
abq
o
c
mn
p
2
mn
1
m
1
_ _
1
n
1
_ _ _ _
_ _
The four individual energy components U
I
, U
II
, U
III
and W are shown within
the large braces in Eq. (25). This expression is unwieldy so a change in
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
26
notation will be made to facilitate inspection of the mathematical structure of
P. Toward that end, dene the following constants:
F
1

p
4
b
8a
3
D
11
F
2

2p
4
8ab
D
12
2D
66
F
3

p
4
a
8b
3
D
22
F
4

p
2
b
8a
N
xx
F
5

p
2
a
8b
N
yy
F
6

2p
2
a
2
D
16
F
7

2p
2
b
2
D
26
F
8
2N
xy
F
9

abq
o
p
2
Note that these terms are all known constants for a given laminate and
loading condition. On the basis of these denitions, Eq. (25) can be rear-
ranged as follows:
P U
I

M
m 1

N
n 1
f
c
2
mn
F
1
m
4
F
2
m
2
n
2
F
3
n
4
F
4
m
2
F
5
n
2
_ _
c
mn
F
9
mn
1
m
1
_ _
1
n
1
_ _ _ _
c
mn

M
i 1

N
j 1
c
ij
m
2
i
2
n
2
j
2

_
_
F
6
m
3
nij F
7
mn
3
ij F
8
m
2
nj
_ _
MINJ
_
_g
To further explore the Ritz method, we must now expand the expression for
P, based on some specied 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 formof Pbased 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 nd:
P U
I
c
2
11
F
1
F
2
F
3
F
4
F
5

c
2
12
F
1
4F
2
16F
3
F
4
4F
5

c
2
21
16F
1
4F
2
F
3
4F
4
F
5

4c
2
22
4F
1
4F
2
4F
3
F
4
F
5

27
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

40
9
c
11
c
22
2F
6
2F
7
F
8

40
9
c
12
c
21
2F
6
2F
7
F
8
4c
11
F
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 coecients c
mn
.
Second, the term involving F
9
is a linear function of c
mn
. Referring to the
denitions of F
1
through F
9
listed above, it is seen that F
9
is the only term
related to the transverse load q
o
. Hence, that portion of the total expression
for P which is related to the transverse loading is a linear function of the
unknown coecients c
mn
. 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 c
mn
, and terms involving
the transverse load are always linear functions of c
mn
.
Proceeding with the Ritz analysis, we now apply the principle of min-
imumpotential energy. That is, we wish to identify the particular values of c
mn
dictated by:
BP
Bc
mn
0
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 c
11
and equating to zero, we have:
BP
Bc
11
2F
1
F
2
F
3
F
4
F
5
c
11

40
9
2F
6
2F
7
F
8
c
22
4F
9
0
Three additional equations are also formed (BP/Bc
12
=BP/Bc
21
=BP/
Bc
22
=0). The four equations can be represented using matrix notation as
shown in Fig. 2(a). These equations may be easily solved for coecients c
mn
by
multiplying both sides of the equation by the inverse of the [4 4] array, as
shown in Fig. 2(b). Once coecients c
mn
have been determined, the out-of-
plane deections can be calculated using Eq. (6) and the problem is solved.
Referring to the denitions of F
1
through F
9
listed above, it is seen that
normal stress resultants N
xx
and N
yy
appear only in terms F
4
and F
5
, re-
spectively, while the shear stress resultant N
xy
appears only in term F
8
.
Inspection of the [4 4] array shown on the left side of the equality in Fig. 2(a)
reveals that F
4
and F
5
appear only along the main diagonal of the matrix,
whereas F
8
appears only in o-diagonal positions. This pattern always oc-
curs, regardless of the value of the value of M and N or the nature of the
transverse loadingthe normal stress resultants appear only along the main
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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/Bc
mn
= 0, for M = N = 2. (b) Solving the set of simultaneous equations shown in part (a).
C
o
p
y
r
i
g
h
t


2
0
0
4

b
y

M
a
r
c
e
l

D
e
k
k
e
r
,

I
n
c
.

A
l
l

R
i
g
h
t
s

R
e
s
e
r
v
e
d
.
diagonal, while the shear stress resultant appears only in o-diagonal posi-
tions. In summary, Fig. 2 represents the solution for a simply supported
symmetric composite panel subjected to a uniform transverse pressure q
o
,
a uniform in-plane stress resultants N
xx
, N
yy
, and N
xy
, and a uniform change
in temperature, DT, based on M = N = 2. Depending on details of a specic
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 specied values of M and N, to perform the required partial dieren-
tiation, to determine the inverse of the resulting [M N] matrix, and to
complete the nal matrix multiplication that gives the coecients c
mn
. 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
room temperature (20jC). After cooling, the at 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 x axis is dened
parallel to the 300-mmedge (i.e., a =0.3 m, b =0.15 m). The laminate is then
subjected to a uniformtransverse load q(x,y) =30 kPa. No in-plane loads are
applied (i.e., N
xx
= N
yy
= N
xy
= 0). Determine the maximum out-of-plane
displacement based on a Ritz analysis and plot the out-of-plane displacement
eld. 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:
ABD
145:2 10
6
35:3 10
6
0 0 0 0
35:3 10
6
64:8 10
6
0 0 0 0
0 0 50:2 10
6
0 0 0
0 0 0 22:3 7:97 2:20
0 0 0 7:97 14:3 2:20
0 0 0 2:20 2:20 10:8
_

_
_

_
where the units of A
ij
are Pa m and the units of D
ij
are Pa m
3
. Notice that
neither D
16
nor D
26
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
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 eld 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
D
16
, D
26
p 0. However, for this problem, the magnitudes of D
16
and D
26
(relative to D
11
and D
22
) are very small. Specically, for the laminate
considered in this problem D
16
/D
11
= D
26
D
11
= 0.0986 and D
16
/D
22
=
D
26
D
22
= 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 D
16
and D
26
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 D
16
and D
26
.
Figure 3 Convergence of predicted plate deflections based on a Ritz analysis
as Mand N are increased from 1 to 10 (i.e., as the number of terms is increased
from 1 to 100).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 N
xx
= N
yy
. 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 specied range in N
xx
and N
yy
, and 100
terms were used in the displacement eld 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 N
xx
= N
yy
= 0, a maximum out-of-plane displacement of
8.03 mm is predicted. In contrast, if tensile stress resultants N
xx
= N
yy
= 100
kN/m are applied, the maximum out-of-plane displacement is reduced to
0.71 mm.
Figure 4 A contour plot of out-of-plane displacements for the [(F45/0)
2
]
s
laminate considered in Sample Problem 1.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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., N
xx
=
N
yy
= N
xy
= 0). Determine the maximum out-of-plane displacement based
on a Ritz analysis and plot the out-of-plane displacement eld. 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 ber angle of 25j has been selected for consideration in this prob-
lem because it results in high relative values of D
16
and D
26
, resulting in an
interesting distortion of the predicted out-of-plane displacement eld. Spe-
cically, for this laminate:
D
16
=D
11
0:370 D
16
=D
22
2:21
D
26
=D
11
0:126 D
26
=D
22
0:755
Figure 5 Predicted maximum deflections of the plate considered in Sample
Problem 2 (M=N=10 in all cases).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
These relative values of D
16
and D
26
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 maximumdisplacement of 16.1 mmis 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 eld 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:
qx; y q
o
sin
px
a
_ _
sin
py
b
_ _
Figure 6 A contour plot of out-of-plane displacements for the [25]
12
laminate
considered in Sample Problem 3.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
In this case, the work done by the transverse load, calculated using Eq. (23),
is given by:
W q
o
_
b
0
_
a
0
sin
px
a
_ _
sin
py
b
_ _ _ _

M
m1

N
n1
c
mn
sin
mpx
a
_ _
sin
npx
b
_ _
_ _
dxdy
During integration of this expression, we note the following:
_
a
0
sin
px
a
_ _
sin
mpx
a
_ _
dx
0; if m p 1
a
2
; if m 1
_
_
_
_
b
0
sin
py
b
_ _
sin
npx
b
_ _
dy
0; if n p 1
b
2
; if n 1
_

_
Hence, the work done by a sinusoidal transverse load is simply:
W
q
o
abc
11
4
The total potential energy is P=U
I
+U
II
+U
III
+W. Our earlier expressions
for U
I
, U
II
, and U
III
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:
P C
1
N
xx
C
2
N
yy
2C
3
N
xy
_ _
ab
_ _

f

M
m 1

N
n 1
p
4
8
c
2
mn
bm
4
a
3
D
11

2m
2
n
2
ab
D
12
2D
66

an
4
b
3
D
22
_ _ _
2p
2
mnc
mn

M
i 1

N
j 1
c
ij
ij
m
2
i
2
n
2
j
2

_ _ _
28

m
2
a
2
D
16

n
2
b
2
D
26
_ _
MI NJ
__
g

f

M
m 1

N
n 1
p
2
8
c
2
mn
m
2
b
a
N
xx

n
2
a
b
N
xx
_ _ _
2m
2
nc
mn
N
xy

M
i 1

N
j 1
c
ij
j
m
2
i
2
n
2
j
2

_ _
MI NJ
_ __
g

q
o
abc
11
4
_ _
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
As before, the next step is to apply the principle of minimum potential
energy:
BP
Bc
mn
0
m 1; 2; . . . ; M
n 1; 2; . . . ; N
_
This process leads to (M N) equations (similar in form to those shown in
Fig. 2) that must be satised simultaneously. Hence, by solving these equa-
tions for constants c
mn
, 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 a
i
b
i
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:
qx; y
16P
p
2
a
i
b
i

l
m 1

l
n 1
1
mn
sin
_
mpn
a
_
sin
_
npg
b
_
sin
_
mpa
i
2a
_ _ _
sin
npb
i
2b
_ __
sin
_
mpx
a
_
sin
_
npy
b
__
Substituting the above into the expression representing the work done by
transverse loads W [Eq. (23)] and integrating, it will be found:
W
4abP
p
2
a
i
b
i

M
m1

N
n1
c
mn
mn
sin
mpn
a
_ _
sin
npg
b
_ _
sin
mpa
i
2a
_ _
sin
npb
i
2b
_ _ _ _
The total potential energy is P=U
I
+U
II
+U
III
+W. Our earlier expressions
for U
I
, U
II
, and U
III
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:
P C
1
N
xx
C
2
N
yy
2C
3
N
xy
_
ab
_ _

M
m 1

N
n1
p
4
8
c
2
mn
bm
4
a
3
D
11

2m
2
n
2
ab
D
12
2D
66

an
4
b
3
D
22
_ _ _
2p
2
mnc
mn

M
i 1

N
j 1
c
ij
ij
m
2
i
2
n
2
j
2

_ _ _

m
2
a
2
D
16

n
2
b
2
D
26
_ _
MI NJ
__
g f
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

f

M
m 1

N
n 1
p
2
8
c
2
mn
m
2
b
a
N
xx

n
2
a
b
N
yy
_ _ _
2m
2
nc
mn
N
xy

M
i 1

N
j 1
c
ij
j
m
2
i
2
n
2
j
2

_ _
MI NJ
_ __
g

4abP
p
2
a
i
b
i

M
m 1

N
n 1
c
mn
mn
sin
mpn
a
_ _
sin
npg
b
_ _
sin
mpa
i
2a
_ _
sin
npb
i
2b
_ _ _ _
_ _
As before, the next step is to apply the principle of minimum potential energy:
BP
Bc
mn
0
m 1; 2; . . . ; M
n 1; 2; . . . ; N
_
This process leads to (M N) equations (similar in form to those shown
in Fig. 2) that must be satised simultaneously. Hence, by solving these
equations for constants c
mn
, the out-of-plane displacements caused by a
transverse force P uniformly distributed over an interior rectangular region
of dimensions a
i
b
i
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 Papplied at x =n and y =g can
be obtained by allowing a
i
! 0, b
i
! 0 in Eq. (29). In the limit, we obtain:
W P

M
m 1

N
n 1
c
mn
sin
mpn
a
_ _
sin
npg
b
_ _ _ _
The total potential energy is:
P C
1
N
xx
C
2
N
yy
2C
3
N
xy
_
ab
_ _

M
m 1

N
n 1
p
4
8
c
2
mn
bm
4
a
3
D
11

2m
2
n
2
ab
D
12
2D
66

an
4
b
3
D
22
_ _ _
2p
2
mnc
mn

M
i 1

N
j 1
c
ij
ij
m
2
i
2
n
2
j
2

_ _ _

m
2
a
2
D
16

n
2
b
2
D
26
_ _
MI NJ
__
g
29
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

f

M
m 1

N
n 1
p
2
8
c
2
mn
m
2
b
a
N
xx

n
2
a
b
N
yy
_ _ _
2m
2
nc
mn
N
xy

M
i 1

N
j 1
c
ij
j
m
2
i
2
n
2
j
2

_ _
MI NJ
_ _ _
g
P

M
m 1

N
n 1
c
mn
sin
mpn
a
_ _
sin
npg
b
_ _ _ _
_ _
30
We apply the principle of minimum potential energy:
BP
Bc
mn
0
m 1; 2; . . . ; M
n 1; 2; . . . ; N
_
This process leads to (M N) equations (similar in form to those shown in
Fig. 2) that must be satised simultaneously. Hence, by solving these equa-
tions for constants c
mn
, 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
TO SIMPLE SUPPORTS OF TYPE S4
In this section, buckling of type S4 simply supported symmetric composite
panels will be considered. Abrief 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 N
xx
, N
yy
,
and/or N
xy
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 B
ij
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 at plate subjected to in-
plane loading is given by:
P U
I
U
II
U
III
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 U
I
, U
II
, U
III
for a composite panel have been
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
presented as Eqs. (14), (17), and (21), respectively. Inspection of these equa-
tions reveals that U
I
is a function of in-plane displacements u
o
(x,y) and
v
o
(x,y), that U
II
is a function of out-of-plane displacements w(x,y), and that
U
III
is a function of u
o
, v
o
, and w. If an initially at plate is subjected to
relatively low levels of in-plane loads, then only in-plane displacements oc-
cur; that is, the plate remains at because at low load levels, the at con-
guration corresponds to the state of minimum potential energy. In this case,
U
I
is the only nonzero strain energy component: U
II
= U
III
= 0 because
w(x,y) = 0. However, as the magnitude of in-plane loading is increased, then
the conguration that corresponds to the state of minimum potential energy
may no longer be at but rather bent; the conguration 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 U
I
, U
II
, U
III
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:
P U
I

M
m 1

N
n 1
_
c
2
mn
F
1
m
4
F
2
m
2
n
2
F
3
n
4
F
4
m
2
F
5
n
2
_ _
c
mn

M
i 1

N
j 1
c
ij
m
2
i
2
n
2
j
2

_
F
6
m
3
nij F
7
mn
3
ij
_
_
F
8
m
2
nj
_
MI NJ
_
__
31
The constants F
1
through F
8
were dened in Sec. 4, in conjunction with Eq.
(26). Note that one constant, namely, F
9
, 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
Bc
mn
0
m 1; 2; . . . ; M
n 1; 2; . . . ; N
_
This process leads to (M N) equations that must be satised 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 F
9
= 0.
Referring to Fig. 2(a), we see that in the present case, all terms on the right
side of the equality are zero.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 N
xx
that will cause buckling, given some constant values for N
yy
and N
xy
.
Alternatively, we may wish to determine the value of N
xy
that will cause
buckling, given some constant values for N
xx
and N
yy
. A third possibility is a
loading situation in which all three resultants are increased proportionately
(i.e., N
yy
=k
1
N
xx
and N
xy
=k
2
N
xx
, where k
1
and k
2
are known constants). In
this case, we are interested in buckling caused by a simultaneous increase in
N
xx
, N
yy
, and N
xy
.
For illustrative purposes, let us assume that we are interested in
buckling caused by uniaxial loading, i.e., for the loading condition N
xx
p 0,
N
yy
= N
xy
= 0. In this case, we have:
F
5

p
2
a
8b
N
yy
0 F
8
2N
xy
0
Also, F
9
=0, because no transverse load is applied. For this situation, the set
of equations shown in Fig. 2(a) reduce to:
2F
1
F
2
F
3
F
4
0 0
40
9
2F
6
2F
7

0 2F
1
4F
2
16F
3
F
4

40
9
2F
6
2F
7
0
0
40
9
2F
6
2F
7
216F
1
4F
2
F
3
4F
4
0
40
9
2F
6
2F
7
0 0 84F
1
4F
2
4F
3
F
4

_
_

c
11
c
12
c
21
c
22
_

_
_

0
0
0
0
_

_
_

_
The unknown buckling load is contained in the term F
4
= p
2
bN
xx
/8a. We
can therefore rearrange the above expression by bringing all terms involving
F
4
to the right side of the equality:
2F
1
F
2
F
3
0 0
40
9
2F
6
2F
7

0 2F
1
4F
2
16F
3

40
9
2F
6
2F
7
0
0
40
9
2F
6
2F
7
216F
1
4F
2
F
3
0
40
9
2F
6
2F
7
0 0 84F
1
4F
2
4F
3

_
_

_
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

c
11
c
12
c
21
c
22
_

_
_

_
F
4
2 0 0 0
0 2 0 0
0 0 8 0
0 0 0 8
_

_
_

_
c
11
c
12
c
21
c
22
_

_
_

_
32
N
xx
p
2
b
4a
0 0 0
0
p
2
b
a
0 0
0 0
p
2
b
a
0
_

_
_

_
c
11
c
12
c
21
c
22
_

_
_

_
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 N
xx
, and the eigen-
vector corresponds to the column matrix containing coecients c
ij
. Because
we have assumed M = N = 2, there are four values of N
xx
that will satisfy
Eq. (32). The eigenvalue with lowest magnitude represents the critical
buckling load N
xx
c
, and the coecients c
ij
that correspond to this eigenvalue
represent the critical buckling mode because wx; y S
M
m 1
S
N
n 1
c
mn
sin
(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 N
xx
p 0, N
yy
=N
xy
=0, and M=N=2. As a second example, consider
buckling caused by a pure shear load. That is, assume N
xy
p 0, N
xx
=N
yy
=0.
We now have F
4
= F
5
= F
9
= 0. Following a process identical to that
described above, we arrive at the following generalized eigenvalue problem:
2F
1
F
2
F
3
0 0
40
9
2F
6
2F
7

0 2F
1
4F
2
16F
3

40
9
2F
6
2F
7
0
0
40
9
2F
6
2F
7
F
8
216F
1
4F
2
F
3
0
40
9
2F
6
2F
7
0 0 84F
1
4F
2
4F
3

_
_

c
11
c
12
c
21
c
22
_

_
_

_
N
xy
0 0 0
80
9
0 0
80
9
0
0
80
9
0 0

80
9
0 0 0
_

_
_

_
c
11
c
12
c
21
c
22
_

_
_

_
33
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Once again, there will be four eigenvalue/eigenvector pairs that satisfy Eq.
(33) because we have assumed M = N = 2. The eigenvalue with lowest
magnitude represents the critical buckling shear load N
xy
c
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 ve
dierent materials may be dened. 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:
E
11
170 10
9
Pa E
22
10 10
9
Pa v
12
0:30
G
12
13 10
9
Pa
Because 1 Pa=1 N/m
2
, all lengths must be input in meters. For example, ply
thicknesses must be input in meters (not millimeters). A typical value would
be t
k
=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:
E
11
25:0 10
6
psi E
22
1:5 10
6
psi v
12
0:30
G
12
1:9 10
6
psi
All lengths would be input in inches. A typical ply thickness might be
t
k
=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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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
Clis, 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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Appendix A
Finding the Cube-Root of a Complex Number
The process of nding 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 DeMoivres 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 DeMoivres theorem to nd the cube root of a complex
number.
Any complex number z can be written as:
z h iv
where:
i u

i
p
h the real part of z
v the imaginary part of z
The modulus (r) and argument (u) of a complex number are dened as:
r

h
2
v
2
p
A:1
617
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
u tan
1
v=h A:2
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 denitions, DeMoivres theorem can be used to
show that the cube root of a complex number z = h+iv is given by:

z
3
p

h iv
3
p
exp
ln r
3

cos
u
3
h i
isin
u
3
h i n o
A: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 tan


1
(v/h) corresponds to
two dierent arguments (that is, two dierent 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 rst quadrant of the complex plane:
0 Vu
p
2
radians
In contrast, consider a dierent complex number zV, where zV=1
*
z=
hiv. 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).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
(A.1)] is identical. However, the argument of zV, angle uV say, is an angle in the
third quadrant of the complex plane:
p rad V uV V
3p
2
radians
Hence, when calculating the argument of a complex number using Eq. (A.2),
one must insure that the angle returned by the tan
1
( ) function corresponds
to the proper quadrant, as determined by the algebraic signs of h and v.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Appendix B
Experimental Methods Used to Measure
In-Plane Properties E
11
, E
22
, v
12
, and G
12
An abbreviated discussion of tests used to measure tensile in-plane proper-
ties E
11
, E
22
, v
12
, and G
12
is provided in this Appendix. The fact that this
discussion is limited to tensile test methods is signicant and should be
noted by the reader. First, since E
11
, E
22
, and v
12
often dier in tension and
compression, in practice both tensile and compressive values should be
measured. Specimen geometries used during compression testing generally
dier 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 lament wind-
ing, for example. Third, material properties can also be inferred from ex-
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. (13) 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. 14. As indicated, tensile specimens are at and straight-sided
with rectangular cross section. Very often adhesively bonded end tabs are
used. The [0]
n
specimen, the o-axis [h]
n
specimen, and the [90]
n
specimen
621
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
shown in Figs. 13 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 1 A [0]
n
tensile specimen, showing typical dimensions and adhesively
bonded end tabs.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
specimen. The bevel angle is usually about 57j, although angles as high as
30j are used on occasion. Use of a shallowbevel 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 dened as the tab-to-tab distance) to specimen width.
Figure 2 A [h]
n
tensile specimen, showing typical dimensions and adhesively
bonded end tabs.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
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 3 A [90]
n
tensile specimen, showing typical dimensions and adhesively
bonded end tabs.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Strains induced during a tensile test are measured using either bonded
resistance strain gages or extensometers. Properties E
11
and v
12
are measured
using a [0]
n
specimen while E
22
is measured using a [90]
n
specimen. It is
possible to measure v
21
as well, using a [90]
n
specimen. However, for most
advanced unidirectional composites the numerical value of v
21
(the so-called
minor Poisson ratio) is very small and is at least an order of magnitude
smaller than the major Poisson ratio v
12
. For example, for graphite/epoxy
the major Poisson ratio v
12
c
0.3 whereas the minor Poisson ratio v
12
c0.01.
The small value of v
21
can lead to a relatively high measurement error.
Figure 4 A [F45]
ns
tensile specimen, showing typical dimensions.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Consequently, v
21
is rarely measured in practice. If the value of v
21
is required
it is calculated using the inverse relation:
v
21
v
12
E
22
E
11
Several techniques by which the shear modulus G
12
can be measured
using a tensile specimen are available. One approach is based on an o-axis
[h]
n
specimen. Referring to Eq. (38c) in Chapter 5, the axial modulus E
xx
of
such a specimen is given by:
E
xx

1
cos
4
h
E
11

1
G
12

2v
12
E
11

cos
2
hsin
2
h
sin
4
h
E
22
Solving this equation for G
12
, we obtain:
G
12

E
xx
E
11
E
22
cos
2
hsin
2
h
E
11
E
22
E
xx
E
22
cos
4
h 2v
12
E
22
cos
2
hsin
2
h E
11
sin
4
h
B:1
Thus if E
xx
is measured for a [h]
n
specimen (where h is known), and assuming
that E
11
, v
12
, and E
22
have also been measured and are known, then the shear
modulus G
12
can be calculated using Eq. (B.1). For example, if h = 45j (the
most common case) then Eq. (B.1) reduces to:
G
12

E
xx
E
11
E
22
4E
11
E
22
E
xx
E
22
2v
12
E
22
E
11

B:2
An alternate method is based on the use of a [F45]
ns
specimen. In this
case the axial and transverse strains (e
xx
and e
yy
) induced by a uniaxial tensile
stress (r
xx
) are measured, usually using biaxial strains gages. For this stacking
sequence it can be shown (4) that the shear stress (s
12
) and shear strain (c
12
)
induced in the +45j plies are given by:
s
12
A
45j
r
xx
=2
c
12
A
45j
e
xx
e
yy

In contrast, the shear stress and strain induced in the 45j plies are given
by:
s
12
A
45j
r
xx
=2
c
12
A
45j
e
xx
e
yy

Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.


Hence for the [F45]
ns
stacking sequence the shear modulus is given by:
G
12

s
12
c
12

r
xx
2e
xx
e
yy

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.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Appendix C
Tables of Beam Deflections and Slopes
Tables of typical beam deections and slopes can be found on the following
pages.
629
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Table 1 Deection and Slopes of Cantilever Beams
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Table 1 Continued
Source: Ref. 1.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Table 2 Deections and Slopes of Simply Supported Beams
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Table 2 Continued
Source: Ref. 1.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

Вам также может понравиться