M Echanics of Fracture: Plates and Shells With Eraeks

M echanics of fracture
VOLUME 3
Plates and shells with eraeks

M echanics of fracture
edited by GEORGE C. SIR
VOLUME 1
M ethods of analysis and solutions of eraek problems
VOLUME 2
Three-dimensional craek problems
VOLUME 3
Plates and shells with eraeks

Meehanies of fraeture 3
Plates and shells witll eraeks
A collection of stress intensity factor solutions for eraeks in plates and shelIs
Edited by
G. C. SIH
Professor of Meehanies and
Direetor of the Institute of
Fraeture and Solid Meehanies
Lehigh University
Bethlehem, Pennsylvania
NOORDHOFF INTERNATIONAL PUBLISHING

LEYDEN
© 1977 Noordhoff International Publishing
Softcover reprint of the hardcover 1st edition 1977
Adivision of A. W. Sijthoff International Publishing Company bv Leyden,

The Netherlands
All rights reserved. No part of this publication may be reproduced, stored

in a retrieval system, or transmitted, in any form or by any means, electronic,
mechanieal, photocopying, recording or otherwise, without the prior permission
of the copyright owner.
ISBN-13:978-94-0l0-1294-2 e-ISBN-13:978-94-01O-1292-8
DOI: 10.1007/978-94-010-1292-8
VAN DER LOEFF / DRUKKERS BV - ENSCHEDE

Contents
Editor's preface IX
Contributing authors XV
Introductory chapter
Strain energy density theory applied to plate bending
problems G. C. Sih XVII
Chapter 1
Interaetion of arbitrary array of eraeks in wide plates under
classical bending M. Isida
1.1 Introduction 1
1.2 Basic relations 1
1.3 Complex potentials for traction free eraeks 5
1.4 Arbitrary array of eraeks in wide plate 12
1.5 Numerieal results 23
1.6 Discussions 40
References 42
Chapter 2
Improved approximate theories of the bending and extension
of flat plates R. J. Hartranft
2.1 Introduction 45
2.2 Approximate theories by variational methods 48
2.3 Applications to crack problem s 60
2.4 Guidelines for practical appIications 81
References 82
VI Con ten ts
Chapter 3
Through eraeks in multilayered plates
R. Badaliance, G. C. Sih and E. P. Chen
3.1 Introduction 85
3.2 Minimum complementary energy applied to a layered plate 86
3.3 An approximate three-dimensional theory of multi-layered plates 89
3.4 Through crack in a layered plate 97
3.5 Stress distribution across the plate thickness 104
3.6 Discussion of numerical results 106
3.7 Appendix: Definition of constants 113
References 115
Chapter 4
Asymptotie approximations to eraek problem s in shells
E. S. Folias
4.1 Introduction 117
4.2 General theory - dassical 118
4.3 The stress field in a cracked spherical shell 122
4.4 The stress field in a cracked plate 142
4.5 The stress field in a cracked cylindrical shell 143
4.6 Approximate stress intensity factors for other shell geometries 146
4.7 Plates on elastic foundations 149
4.8 Particular solutions 155
4.9 Discussion 158
References 159
Chapter 5
Craek problems in eylindrieal and spherieal shells
F. Erdogan
5.2 Formulation of the specially orthotropic cylindrical shell problem 163
5.3 The skew-symmetric problem 167
5.4 The symmetric problem 176
5.5 Results for a specially orthotropic cylindrical shell 182
5.6 The effect of Poisson's ratio 188
Contents VII
5.7 Interaction of two eraeks 191

5.8 Further results for isotropic shells 194
References 198
Chapter 6
On eraeks in shells with shear deformation
G. C. Sih and H. C. Hagendorj
6.2 Shell theory with shear deformation 202
6.3 Symmetric loading 212
Appendix: Integrand and Kernel funetions 227
References 229
Chapter 7
Dynamie analysis of era eke d plates in bending and extension
G. C. Sih and E. P. Chen
7.2 Classical plate bending theory 232
7.3 Mindlin's theory of plate bending 239
7.4 Kane-Mindlin's equation in plate extension 247
7.5 Plates subjected to sudden loading 261
References 271
Chapter 8
A speeialized finite element approaeh for three-dimensional
eraek problems P. D. Hilton
8.2 Three-dimensional elastic calculations 274
8.3 Finite element method- background 276
8.4 Specialized elements for the crack edge 278
8.5 Applications to crack problems 282
8.6 Detaiis of the analysis 283
8.7 Results of the finite element analysis 290
8.8 Summary 297
References 297
VIII Con ten ts
Author's Index 299
Subject Index 301

Editor' s preface
This third volume of a series on Mechanies of Fraeture deals with eraeks in

plates and shelIs. It was noted in Volume 2 on three-dimensional eraek
problems that additional free surfaees can lead to substantial mathematical
complexities, often making the analysis unmanageable. The theory of plates
and shelIs forms a part of the theory of elasticity in which eertain physieal
assumptions are made on the basis that the distanee between two bounded
surfaees, either fiat or eurved, is small in eomparison with the overall dimen-
sions of the body. In modern times, the broad and frequent applieations
of plate- and shell-like struetural members have aeted as a stimulus to whieh
engineers and researchers in the field of fracture meehanies have responded
with a wide variety of solutions of teehnieal importanee. These eontributions
are covered in this book so that the reader may gain an understanding of
how analytieal treat me nt s ofplates and shells containing initial imperfeetions
in the form of eraeks are earried out.
The development of plate and shell theories has involved long standing
controversy on the eonsisteney of omitting eertain small terms and at the
same time retaining others of the same order of magnitude. This defieieney
depends on the ratio of the plate or shell thiekness, h, to other eharaeteristie
dimensions and eannot be eompletely resolved in view of the approximations
inherent in the transverse dependence of the extensional and bending stresses.
The presenee of a eraek tends to further eomplieate the situation and requires
additional cIassifieation of the basie assumptions. First of aH, the eonvention-
al subdivision of thin and thiek plates or shelIs is no longer adequate.
Suppose that a thin shell satisfies the usual requirement of (h/R)max ::; 1/20
with R being the smallest radius of eurvature of the shell. At smaH distanees
(as eompared with h)from the eraek, the shell would appear to be thiek and
the assumptions* of the cIassieal thin shell theory would lead to an inaeeu-
* Points near the crack edge in a plate or sheli, which lie on the normal to the middle
surface before deformation, cannot be assumed to remain on the normal after deformation.
This in effeet implies that shear forees eannot be disregarded.
x Editor's preface
rate description of the local stress state. In particular, the replacement of

boundary conditions on twisting moment and shearing force by one of
equivalent or Kirchhoff shear results in the satisfaction of only four boundary
conditions instead of five as required on the crack surface. A priori, the
qualitative character of the crack front stresses in a plate or shelI should be
the same as that derived from the three-dimensional theory of elasticity.
The approximations mentioned earIier should only infIuence the resuIts in
a quantitative sense through the stress intensity factors.
Another inadequacy of the cIassical approach is that the dependence of
the stresses on the transverse coordinate is assumed arbitrarily. Along the
crack front, the normal stresses in most of the interior region except in a
layer near the free surface should satisfy the plane strain condition,
(J= = V ((Jx + (Jy), in which (J= and (Jx, (Jy are the transverse and in-plane
normal stresses while v stands for the Poisson 's ratio. The stress solution in
the surface layer can be constructed separately by adopting the concept of
a boundary layer within which the stress intensity factors decrease sharply
and tend to zero * on the free surface. This desirable feature of the solution
can serve as a useful guideline in selecting the appropriate plate or shelI
theory.
In structural applications, nonalignment of loads with crack orientation
is a common practice. As arule, the stress analysis of plates or shelIs involves
more than one stress intensity factor. For example, misalignment in a plane
involves kl and k 2 while k 2 and k 3 arise simultaneously in plates or shelIs
that are twisted or sheared. Combined loadings often lead to all three stress
intensity factors k j (j = J, 2, 3). The caJculation of alIowable load in fracture
mechanics wilI depend on some combination of kl' k 2 and k 3 reaching some
critical condition. This requires the knowledge of a suitable fracture criterion
that can predict non-coplanar** crack extension. The theory of strain
* The three-dimensional theory of elasticity yields a stress solution different from 1/ v r

on the free surface where r is a radial distance. This implies the vanishing of stress intensity
factors. Since a hypothetical surface of zero thickness cannot be conceived physically, the
necessity of admitting a boundary layer in engineering applications is apparent.
** The classical energy release rate concept based on eraeks propagating in a self-similar
manner cannot be extended to the genera I ease by adding Gl, G 2 and G 3 as associated
with kl, k2 and k 3 :
G = [n(l + v)/Ej [(1 - v) (ki + kD -' qj
This is simply because eraeks do not propagate straight ahead (as this equation requires)
in eombined loading. Any attempts made to extend the energy release concept to the bent
Editor's preface XI
energy density or S-theory provides a simple means of evaluating allowable

stress in plates and shelIs subjected to complex loadings. Since this is a rela-
tively new concept, a few introductory remarks are in order.
One important feature of the S-theory, which is fundamentally different
from all existing failure theories, is that it simultaneously accounts for Sv,
energy factor associated with volume change (dilatation) and Sd, energy
factor associated with shape change (distortion). The proportion of Sv and
Sd in S = Sv + Sd is determined at locations of the stationary values of S.
Depending on the material constants, the relative minimums, Smin, occur
at locations where Sv is normally greater than Sd and the relative maximums,
Smax, correspond to locations where Sd > Sv. The basic assumption is that
brittle fracture take s place at locations of Smin (excessive dilatation) while
yielding occurs at locations of Smax (excessive distortion). In general, it is
not justified to negleet Sv against Sd or vi ee versa in a material element uniess
they differ from one another by one order or magnitude or more which is
seldom the case. The application of S-theory to analyze cracks in plates is
presented in the Introductory Chapter. For brittle fracture, the onset of
crack propagation is assumed to occur when Smin reaches a critical value, Seo
characteristic of the material. This value is then related to the load and geom-
etry of a plate or shell for determining the sub-critical or critical size of
eraeks as a function of the allowable load or vi ee versa.
The first chapter deals with the stress intensity factor solutions for a
variety of crack geometries by using the Poisson-Kirchhoff plate bending
theory. Among the problem s treated are collinear cracks, paralleI cracks,
cracks incIined at arbitrary positions, etc. The versatility of the complex
variable technique is again demonstrated for solving two-dimensional crack
problem s involving rather complicated crack geometries. Although the
theory does not satisfy the physical crack surface boundary conditions,
useful information on stress intensity factors can be assessed for complex
problem s which may not be so easily obtained from the higher order plate
theories. Chapter Two con tai ns a thorough review of existing plate theories
in connection with the mixed boundary value crack problem. Assumptions
made in each of the theories as developed from the minimum principles in
variational calculus and how they would affect the end results are discussed.
Numerieal solutions computed from integral equations for several crack
or kinked crack configuration would be unfeasible as the uncertainties encountered in the

mathematics cannot be easily resoived.
XII Editor's preface
problems show the interaetion of plate thiekness with crack len,gth. A for-
mulation of the bending and/or extension of layered plates containing a
through craek is presented in Chapter Three. Each layer can have different
elastie properties and thiekness. The three-Iayered p!ate problem with a
through eraek subjeeted to a tensile load is solved as an eXaJllple. Lamination
is found to have a signifieant infiuenee on the intensity of load transmitted
to the erack tip region. Graphieal plots of stress intensity factors for various
material and geometrie parameters are displayed.
In Chapter Four, the dassieal theory of thin shalIow shelis is applied to
analyze the stress distribution around craeks in shelis. The shalIowness
assumption justifies the projection of the cracked portion of the shell onto
a fiat plane such that the problem is formulated in terms of rectangular
eoordinates in two dimensions. Solutions based on singular integral equa-
tions are provided for craeks in shelis of different shapes such as spherieal,
cylindrieal, eonical, toroidal, etc. The presenee of eurvature in a shell gen-
erates deviation from the behavior of fiat plates, in that stretching loads
induee both extensional and bending stresses, while bending loads also
lead to both type s of stresses. Chapter Five eonsiders a method of solution
for craeked shells possessing anisotropy. The method is illustrated by solving
the problem of an orthotropie cylindrieal shelI with an axial eraek. Both
symmetric and skew-symmetrie loadings are considered. Other examples
dealing with the effeet of Poisson's ratio and interaetion of two colIinear
eraeks in shelis of an isotropie nature are also provided. Developed in Chap-
ter Six is a higher order thin shelI theory in which transverse shear deforma-
tion is induded. A tenth order system differential equation is obtained such
that satisfaction of the five natural boundary eonditions on the eraek surfaee
is made possible. This is an improvement over the dassical theory which
considers only four conditions. Solutions showing the infiuence of shell
curvature, shell thiekness, craek length, etc., on the stress intensity factor
are given and they ean differ appreeiably from those obtained by the dassical
theory eorresponding to an eighth order system of equations.
The Seventh Chapter is concerned with dynamie loading of fiat plates
with craeks where inertia effeets can no longer be negleeted. Two types of
dynamie-load sourees are treated: namely, vibratory and impact. The
dassieal and Mindlin theories are used for eraek problems in plate bending
and the Kane-Mindlin theory is applied to solve the elastodynamie plate
stretching probIems. Dynamic stress intensity faetors are defined and found
to vary as funetions of time. They tend to the statie values as time beeomes
Editor' s preface XIII
increasingly large for the case of impact loading.

Chapter Eight is devoted to numerical analysis of the three-dimensional
plate problem with eraeks. The solution is considered to be approximate in
that the equilibrium equations of elasticity are satisfied only in an average
sense over discrete regions of the plate medium. The numerical procedure
involves three-dimensional isoparametric finite elements specialized for
solving crack problem s where the crack edge singularity is assumed to be
known and embedded into the analysis. This is done by incIuding in the
crack tip elements the asymptotic solution for the displacement field as
well as the usual polynomial approximation. Values for the stress intensity
factors at nodes along the crack edge are treated as unconstrained degrees
of freedom or generalized nodal displacements to be determined through
the finite element procedure. Studies are made on the stress distribution in
a finite thickness plate with a through crack subject to tensile and bending
loads.
Much of the material in this volume is the result of research efforts
initiated in the 1960's at the California Institute of Technology and Lehigh
University. The fruitful discussions and interehange of ideas among the
researchers associated with these two institutions over the years have made
possible significant progress in the understanding of fracture behavior of
plates and shells weakened by eraeks or initial imperfections. My gratitude
goes to colleagues and students who have contributed in this field. The
many valuable hours the authors have spent in preparing their manuscripts
are deeply appreciated. I am also thankful to Mrs. Barbara DeLazaro for
neatly typing portions of the book and Mrs. Constance Weaver for willingly
providing constant secretarial assistance.
Lehigh University G. C. SIH

Bethlehem, Pennsylvania
March, 1976
Contributing authors
R. BADALIANeE
Foster-Wheeler Corporation, Livingston, New Jersey
E. P. CHEN
Institute of Fraeture and Solid Mechanies, Lehigh University, Bethlehem,
Pennsylvania
F.ERDOGAN
Pennsylvania
E.S. FOLIAS
Department of Civil Engineering, University of Utah, Salt Lake City, Utah
H. C. HAGENDORF
Fatigue and Fraeture Consultants, Hawthorne, California
R. J. HARTRANFT
Pennsylvania
P. D. HILTON
Pennsylvania
M.IsIDA
Sehool of Engineering, Kyushu University, Fukuoka 812, Japan
G. C. SIH
Pennsylvania
G. C. Sih
Introductory chapter:
Strain energy density theory applied
to plate bending problems
J. Introductory remarks
When a sl en der member is stretched gradually with consideration given

only to the principal stress in the axial direction, then failure in the global
sense is said to occur by yielding if this stress reaches the elastic limit or
yield point and by fracturing if the ultimate strength of the material is
reached. Material elements within this member, however, are in a multiaxial
stress state. Hence, yielding and/or fracture in the local sense or at a point
cannot be adequately described by considering just one of the six stress or
strain components even if the remote loading is uniaxial. The distinction
between localized and global yielding or fracture must be made before
selecting a quantity whose critical value will be assigned to determine the
load carrying capacity of the member. In principle, this quantity should
involve some combination of the stresses or strains.
The energy quantity has been frequently associated with the development
of failure theories. Two of the commonly used theories are the total energy
or Beltrami-Haigh theory and the distortion energy or Huber-Von Mises-
Hencky theory. According to the former, failure by yielding in the material
occurs when the to tal strain energy per unit volume
dW
dV
1
= ~
+ v[ 2
(lx
2 2
+ (ly + (lz - 1
V
+ V «(lx + (ly + (lz)
2
+ 2(T;y + T;z + T;x)] (1)
absorbed by the material is equal to the energy per unit volume stored in the
material upon reaching the yield point under a uniaxial stress state, i.e.,
d W/d V = (l;p/2E.The latter theory can be stated in the same way by replacing
XVIII G. C. Sih
the total strain energy density, d Wjd V, with the distortional strain energy
density:
( dW) 1 + v[
dV d = ~ (O'x - O'y)
2
+ (O'y - O'z)
2
+ (O'z - O'J 2
+ 6(T;y + T;z + T;JJ (2)
where E is the Young's modulus of elasticity and v the Poisson's ratio.

Equation (2) is obtained by decomposing d Wjd V in equation (1) into two
parts
dW (dW) (dW)
dV = dV v + dV d (3)
in which
( dW) 1 - 2v
dV v = ~(O'x + O'y + O'z)
2
(4)
represents that part of the strain energy density associated with volume
change. It has been implicitly assumed in the two foregoing theories that
the failure mode by yielding for an element in a multiaxial stress state is the
same as that in a uniaxial stress state.
The Huber-Von Mises-Hencky theory associates failure by yielding with
(dWjdV)d only while (dWjdV)v is neglected*. Except for the extreme situa-
tions of a hydrostatic stress state, where (dWjdV)d = 0, and a pure shear
stress state, where (dWjdV)v = 0, the failure of a unit volume of material
ranging from incipient yielding to brittle fracture involves the release of
both energy due to changing in volume and shape. For instance, equations
(2) and (4) indieate that for a uniaxial stress state both (dWjdV)d and
(dWjdV)v are present:
W)
( ddV 1 + V 2 (d W ) 1 - 2v 2
d = ~O'x, dV v = ~O'x (5)
Note that the distortional energy is only four times greater than the dilata-
* This assumption grew out of the test results of Bridgman [1] who showed that many
material s did not exhibit yielding when subjected to very high hydrostatic pressure. In this
case, the material elements experience volume change only with (dWjdV)d = O. It would
be inaccurate to conelude in general that the energy absorbed in changing volume has no
effeet in causing failure by yielding for other stress states as weil.
Introductory chapter XIX
tional energy if V = 0.2. This difference, bei ng less than one order of magni-
tude, cannotjustify neglecting (dWjdV)v in comparison with (d WjdV)d*.
A more general situation prevails in Figure 1 where the material element
is in a triaxial stress state. Let ro be the radius of a core region centered
z z
x x
(0) Rectangular stress (b) Principal stresses
campanents
Figure 1. Stress element outside of core region
around a convenient reference point, the origin. The element un der consid-
eration is located outside of this core region. In order to focus attention on
the location of failure, Sih [2] has defined a strain energy density factor
S = Sv + Sd such that
Sv = ro( ~~)v (6)
is the factor associated with volume change and
(7)
is the factor associated with shape change. With reference to the principal
stresses 0"1' 0"2 and 0"3 in Figure l(b), equations (4) and (6) may be written as
1-
Sv = ro ( 6E
2V) (0"1 + 0"2 + 0"3) 2 (8)
and equations (2) and (7) yield
* An accurate description of the uniaxial extension of metal bars with v between 0.2 and
0.3 should involve both changes in volume and shape.
xx G. C. Sih
(9)
For a state of plane strain where
0'3 = v(u 1 + 0'2) (10)
the ratio Sv/Sd obtained from equations (8) to (10) take s the form
Sv = (1 - 2v)(1 + v) [(0'1/0'2) + lJ2 2 (11)

Sd [(0'1/0'2) -lJ2+ [(1- v) - V(Ul/U2)J2+ [(1- v)(U 1/U2) - vJ
Figure 2 displays a plot of SV! Sd versus the ratio of the principal stresses
0' dU2 in equation (11) for v = 0.0,0.1,0.2,0.3, and 0.4. The greatest volume
change take s place when 0'1 = 0'2 corresponding to a two-dimensional
hydrostatic stress state. For most metals* with v ranging from 0.2 to 0.3,
Sv/Sd varies from 4 to 6.5. The relative magnitude of Sv and Sd shows that
both of them should be accounted for in considering the failure of material
elements.
More recentIy, Sih [3, 4] has proposed a failure theory based on the
stationary value of the strain energy density factor
reaching some critical value. Uniike that of BeItrami-Haigh, Sih's theory

possesses the additional feature of bei ng able to locate the element initiating
yielding or brittIe fracture. This is accomplished by obtaining the stationary
values of S with respeet to the angles 8 and C/J shown in Figure I(a) or I(b).
The point (8 0 , c/Jo) making S a minimum (= Smin) gives the position of the
element that initiates fracture while the location of Smax determines the
onset of yie1ding. The initiation of yielding and/or fracture usuaIIy occurs
in the vicinity of a reentrant comer, a crack front or a notch border from
which '0' the radius of the core region in Figures 1, is measured. The radius
'0 serves as a limiting distance within which the microstructural effect of
the material comes into play. In contrast to the cIassical theory of fracture
mechanics [5] which utiIizes the concept of a critical stress intensity factor
k 1c or an energy release rate quantity G1c , the strain energy density factor
theory, or S-theory, is not restricted to loading symmetric with respeet to
* Keep in mind that the present discussion holds only if yielding or fracture occurs under
smallstrain. It does not apply to rubber-like or polymer materials undergoing large defor-
mation.
Introductory chapter XXI
14.0 V=0.4
12.0
10.0
8.0
."
(J)
.......
> 0.3
(J)
o 1.0 2.0 3.0 4.0 5.0
Figure 2. Ratio of dilatational and distortional energy for a state of plane strain
the crack plane and does not require the material to possess an initial crack.
In this introductory chapter, the S-theory will be further employed in plate
bending problems involving the interaction of crack size with plate thickness.
II. Strain energy density factor theory
A rational procedure for the design of stmetural members requires that the
most likely of severaI possible failure modes of a member in service be
determined and that a quantity be assigned to predict failure. Fracture
mechanics deals with the conditions under which a member attains uncon-
trollable fracture by crack propagation. The critical value of the selected
quantity which limits the allowable load must not be sensitive to changes
XXII G. C. Sih
in the nature of loading, the type of geometry, etc., and should be a charac-
teristic of the material. The k le value in the classical theory of fracture
mechanics [5] is one such example, where k le = a.J a
for a given material
should remain constant at fracture while a, the applied stress, and a, the
half crack length, can vary individually at crack instability. Suitable experi-
ments may be designed to test a given theory by using two different loading
conditions and comparing the experimental results with those predicted by
the theory. Erdogan and Sih [6] have measured the critical tensile stress ac
in uniaxialloading andicritical shear stress re in pure shear loading for plexi-
glass plate specimenscontaining a central crack. Their measured values of
ac and re yield an average of 0.915 for re/ac and agree favorably with Sih's
prediction [2] of rclac = 0.905 for v = 0.33. The theoretieal prediction based
on the strain energy density theory was obtained by using the same critical
value of S or Se for the two different stress states. Detailed caleulations of
ac and re and additional comments will follow subsequently.
Before proceeding further, it is necessary to state the failure mode to be
treated in this discussion. In situations where the phenomenon of material
separation occurs suddenly without noticeable warning such as slow crack
growth, the fracture process is classified as brittle or simply as brittle frac-
ture*. The applieation of the S-theory to ductile fraeture is diseussed else-
where and will not be dealt with here. Within the framework of brittle
fracture where failure initiates from an element nearby the crack border,
it suffiees to compute the strain energy density S from the loeal stresses,
a x , ay, ---, r zx as shown in Figure 3. Without loss of generality, the crack
surface R bounded by an arbitrary contour e is assumed to coincide with
the xz-plane. As the point 0 travels around the crack bordeL the x-axis is
directed normal to e, the y-axis is perpendicular 10 R and the z-axis is tangent
e,
to C. The element under consideration is located at (r, w) which is related
to (x, y, z) through the relationships
* Brittle fracture includes those failures which occur without warning and give no evi-
dence of ductility or stretching prior to rapid crack propagation but does not exclusively
imply that the material must be inherently brittle in the metallurgical sense. Materials
with the same chemical composition and microstructure can behave in a brittle fashion
under one set of conditions (Le., temperature, rate of loading, mechanieal constraint, etc.)
and in a ductile fashion under other conditions. Since the continuum mechanics model
of the S-theory excludes metallurgical effects, it would not be accurate to say that the
theory is restricted to brittle material s but rather it deals with an instability phenomenon
that exhibits brittle behavior.
Introductory chapter XXIII
e rock conIour e
Figure 3. A special spherical coordinate system
x = r cos ecos 0) (13a)

y= r sin ecos 0) (13b)
z = r sin 0) (13c)
where -n/2 ::; 0) ::; n/2 and -n ::; e ::;

n. The choice of spherical coordi-
e,
nates (r, 0)) in Figure 3 yields a simple form* for the local stresses
(14a)
(14b)
(14c)
* An alternate choice of spherical coordinates (r, (), rp) such as those shown in Figures 1
lead to more complicated stress expressions [7] which are equivalent to equations (14).
An error in the asymptotic expression for one of the ellipsoidal coordinates in [7] needs
to be corrected.
XXIV G. C. Sih
kl. e e 3e
'xy = (2r eos co)l sm 2"eos 2" eos "2
k2
+(2reosco)feos2
e(1 -sm. 2esm. 23e) + 0 ("I) (14d)
'xz = - (2 e + 0(1)
k 3 co)' sin -2 (14e)
r eos 2
k3 e
'yz = (2 r eos co)' eos -2 +
T
0(1) (14f)
Substituting equations (14) into (1) and using the relation dW/dV = Sir, it
is found that
(15)
The eoeffieients aij (i, j = 1,2) depend on the material eonstants and the
angles (e, co) as given by
16J.1 eos co all = (3 - 4v - eos e) (1 + eos e) (16a)
16J.1 eos co a 12 = 2 sin e (eos e-I + 2v) (16b)
16J.1 eos co a 22 = 4(1 - v) (1 - eos 8) + (3 eos 8 -1) (1 + eos e) (l6e)
16J.1 eos co a 33 = 4 (16d)
in whieh J.1 stands for the shear modulus of elastieity.
On physieal grounds, material elements are eapable of storing energy by
experieneing volume ehange (dilatation) and shape ehange (distortion).
Exeessive dilatation may re sult in brittle fraeture while exeessive distortion
may tend to yield the material. The proportion of energy absorbed in dilata-
tion and distortion ean be determined for various material properties at
loeations of the stationary values of the strain energy density faetor. In
fraeture mechanies, interest is eentered on the eonditions that trigger eraek
instability. The assumptions made in the strain energy density theory [3, 4]
may be stated as
(1) Crack growth is directed along the !ine from the center of the sperical
core (Figure 3) to the point on the spherical surface with the minimum strain
energy density factor Smin'
(2) Growth along this direction begins when Smin reaches the maximum
critical value Sc which the material will tolerate.
Introductory chapter xxv
The values of (eo, wo) eorresponding to Smin are obtained by taking

as/ae = 0 and as/aw = O. At these loeations, the element s tend to experi-
ence more volume ehange LI V than those loeated at Smax where yielding is
assumed to take place. The path along whieh S is a relative minimum has
eoineided aeeurately with the experimentally observed trajeetory of brittle
fraeture [8, 9].
Experience indieates that failure ean seldom be associated solely with
volume ehange or shape ehange. It normally involves some of eaeh depend-
ing on the stress state. The amount of dilatation and distortion for eaeh
material element ean be determined by deeomposing equation (15) into
Sv = bllki + 2b l2 k 1 k 2 + b22k~ + b33k~ (17)

for dilatation and
Sd = cllki + 2C12klk2 + C22k~ + C33k~ (18)
for distortion. The sum of the eoeffieients bij and cij (i,j = 1,2,3) in equa-
tions (17) and (18) equal aij given by equations (16). Without going into de-
taiIs, it ean be shown that
12p eos w b ll = (1 - 2v) (1 + v) (1 + eos e) (19a)
12p eos w b l2 = 0- 2v) (1 + v) sin e (19b)
12p eos w b22 = (1 - 2v) (1 + v) (1 - eos e) (1ge)
12p eos w b 33 = 0 (19d)
and the resuIt of subtraeting equations (19) from (16) yields
16p eos w C ll = (1 + eos e) U(1 - 2v)2 +1- eos e] (20a)
16p eos w C12 = 2 sin e[eos e - tO - 2V)2] (20b)
16p eos w C22 = tO - 2V)2 (1 - eos e) + 4 - 3 sin 2 e (20e)
16p eos W C33 = 4 (20d)
Onee the loeation (eo, wo) for a given element is known, Sv and Sd ean be
computed from equations (17) and (18) for determining the effeet of dilata-
tion and distortion on the onset of fraeture or yielding.
Now, if failure is eonfined to the plane normal to the eraek front, then
w = 0 and the eore region in Figure 3 beeomes a eirele of radius r eentered
around the eraek edge. Moreover, if the load is applied such that the same
XXVI G. C. Sih
stress state prevails along the crack border, then the problem is said to be
two-dimensional. The assumption of plane strain leads to additional sim-
plification since in this case k 3 drops out of equations (14) and (15). Consider
the plane strain specimens in Figures 4(a) and 4(b) for the uniaxial tension
_ _ _ _ Te
D----+-x x
T.e - - - - - - - -
(0) Tension (k,~ 0, k2 =0)
Figure 4. Plane strain specimens
and in-plane shear of a solid containing a crack of length 2a. It will be shown
subsequently that, according to the strain energy density theory, the ratio
of the critical stresses rc/(Jc at incipient fracture is
rc = ( 3(1 - 2v) 2)1- (21)

(Jc 2(1 - v) - v
For v = 0.33, this ratio gives rc/(Jc = 0.905 which agrees weIl with the
experimentally measured value of 0.915 for cracked plexiglass plates [6] as
discussed earlier.
Uniaxial tension test. When a tensile load is applied normal to the crack
plane, only kl is nonzero and equation (15) reduces to
1
S = 16,u (3 - 4v - eos 0) (1 + eos O)ki (22)
in which kl = (J~a for a large plate. Referring to Figure 4(a), (Jc denotes
the critical value of (J at which crack propagation begins. The direction of
crack initiation is found from the condition as/ao = 0 that makes S a
relative minimum. This corresponds to 80 = 0 implying that the crack
grows in a self-similar manner and
Introductory chapter XXVII
s. = 1 - 2V k2 (23)
mm 4,u 1
At the point of eraek instability, Smin --+ Sc and equation (23) becomes
S e -_ (1 + v) (1 - 2v) /
2E (Jnja (24)
where Sc represents the fraeture toughness of the material. The measured

values of Sc for a number of metal alIoys can be found in [10]. With the
help of equations (17) and (18), Smin in equation (23) can be deeomposed into
S _ (1 + v) (1 - 2v) 2 S _ (1 - 2v)2 2
v - 6,u (J a, d - 12,u (J a (25)
The ratio
Sv _ 2(1 + v)
(26)
Sd - 1 - 2v
shows that Sv is always greater than Sd for (Jo = O. Henee, the element
assumed to trigger fraeture undergoes more volume change than shape
ehange.
The relative maximum of Sean be found in the same way. It oeeurs at
Bo = eos- 1 (1-2v) which may be inserted into equation (22) to give
(1 - V)2 2
Smax = 4,u (J a (27)
Separating equation (27) into Sd and Sv renders
S _ (1 - v) (1 - v + 4v )
2 2 S _ (1 - v 2)(1 - 2v) 2
d - 12,u (J a, v - 6,u (J a (28)
It is easily seen from
Sd 2(1 - v + 4v 2 )
(29)
Sv (1 + v)(l - 2v)
that Sd > Sv eorresponds to elements loeated at Bo = cos- 1 (1-2v) where
yielding is assumed to take place.
Pure shear test.The speeimen in Figure 4(b) is subjected to a state of in-

plane shear where kl = 0 and k z = 1:)a. With this information, equation
XXVIII G. C. Sih
(I5) can be written as

1
S = 16jl [4(1 - v) (l - eos e) + (1 + eos e) (3 cos e - 1)Jr 2 a (30)
Equation (30) may be minimized with respect to e giving

e0= cos -1 (1 - 2V)
--3- (31)
as the direction of crack initiation*, Figure 4(b). Values of eo for different

Poisson's ration are given in Table l. Putting equation (31) into (30) renders
2(1 - v) - v2 2
Smin = 12jl r a (32)
TABLE I
Fracfure angles
v 0.0 0.1 0.2 0.3 0.4 0.5
-oo 70.5° 74.5° 78.5° 82.3° 86.2° 90°
Crack instability starts when Smin = Se or

S _ (l + v) [2(1 - v) - V2 J 2
(33)
e - 6E rea
If the specimens in Figures 4(a) and 4(b) are made of the same material,
then Se in equations (24) and (33) may be eliminated to give the ratio
rc!(Jc as shown in equation (21). As in the case of tensile loading, equation
(32) may also be separated into Sv and Sd' Un der in-plane shear load, the
critical element prior to fraeture undergoes more distortion than the element
situated along the line of symmetry in Figure 4(a).
* It should be eautioned that equation (31) applies to a perfeetly sharp eraek of zero
radius of eurvature, p. Furthermore, the distanee ro at whieh Oo is measured has not been
defined. In a more refined analysis, Sih an~J(ipp [11] have shown that small values of
p and ro ean greatly affeet the ealculated and measured values of the fraeture angle 0o.
Comparison of theory and experiment on Oo without considering the effeet of p and ro
eould lead to premature eonc\usions.
Introduetory ehapter XXIX
III. Bending and twisting of cracked plates
The accurate translation of physical phenomena into mathematical terms

often presents many difficulties. Too much importance cannot be attached
to the initial stages of sifting the evidence to enable a useful analytical treat-
ment. Some of these difficulties and complexities are brought out in Volume 2
of this series on Mechanics of Fracture dealing with three-dimensional crack
probiems. The strict requirement of satisfying all the governing differential
equations and boundary conditions in the theory ofthree-dimensional theory
of elasticity cannot always be met. Assumptions and/or approximations are
frequently made to relax either the governing differential equations or
boundary conditions or both so as to make the problem manageable. Since
the discipline of fracture mechanics requires an accurate description of the
stress state near the crack tip region, it is of primary concern to preserve
the exact nature of the crack surface boundary conditions in formulating
approximate theories. More specifically, the basic character of the local
stresses in equations (14) as determined from the exact three-dimensional
theory of elasticity should be retained. Such a preference implies that ap-
proximations will be made only in a quantitative sense through stress inten-
sity factors kl, k 2 and k 3 •
Charaeter of loeal moments and shear forees. Let the thickness h of an

elastie body with a eraek be small when eompared with the other dimensions
such as crack length, width of the body, etc. Then the body is said to be a
plate. The conventional plate theories* assume that the stress variations in
terms of the thickness coordinate are known as a priori. For thin plates,
the transverse normal stress component (Jz is assumed to be zero throughout
the plate. The middle plane bisecting the plate thickness as shown in Figure 5
is usually taken as referenee. With respect to this plane z = 0, the in-plane
stresses (Jx' (Jy and r xy on the element located at (x, y, z) are assumed to vary
linearly in z, i.e.,
(34)
where Mx and My are the bending moments and M xy the twisting moment
* The theory of Hartranft and Sih [12] does not preassume the stress distribution across
the plate thickness. They determine this distribution from the plane strain condition
ahead of the crack. More detailed information can be found in Chapter 2 of this volume.
xxx G. C. Sih
per unit Iength acting on any section of the pIate paralleI to the xz and yz
pIanes. They can be obtained by integration as
MxCx, y) = f
h/2
crxCx, y, z)z dz (35a)
f
-h/2
h/2
My(x, y) = cr/x, y, z)z dz (35b)
f
-h/2
h/2
Mxy(x, y) = Tx/X, y, z)z dz (35c)
-h/2
z Crock front
~Tzx
h/2 (X,y,~-cr.
+
8 ay T xy
hl2 x
Middle plone
Figure 5. Stresses on an element ahead of crack in a thin plate
With crz aIready assumed to vanish everywhere, the free surface conditions
at z = ±h/2 can be satisfied by further requiring that
TxzCX, y, ± hJ2) = TyzCX, y, ± hJ2) = 0 (36)
It follows that the equations of equiIibrium require both T xz and T yz to vary

paraboIically in z as
T xz = ;h [1 -e:YJv x
(37a)
T yz = ;h [1 - C:YJv y
(37b)
Introductory chapter XXXI
In equations (37), Vx and Vy are the shearing forces per unit length on sec-
tions paralleI to the y and x axes and they are given by
V,,(x, y) = fh'2
'rxz(x, y, z) dz (38a)
f
-h/2
h'2
V/x, y) = 'ry.{x, y, z)dz (38b)
-h/2
Since the z-dependence of the stress components is assumed to be known,

the plate bending problem has essentially been reduced to two dimensions
and the boundary conditions will be specified in terms of the moment s and
shear forces.
For a crack whose surfaces are free from tractions, the bending moment,
twisting moment and shear force must all vanish individually. If the crack
is placed on the x-axis, then the conditions*
(39)
must prevail on the segment of the x-axis where the crack is located. To this
end, the Reissner sixth order plate bending theory [13] is used to satisfy
equations (39) and yields the desired character of the crack tip moments and
shear f orces:
Kl 0(1 . 0 . 30)
Mx = (2r)~ cos"2 - sm 2 sm 2
K2 0( 0 30)
2 2 + cos 2 cos"2 + 0(1)
•
- (2r)!" sm (40a)
My
Kl
= (2r)t cos"2
0 (1 + . 2
sm
. 30)
0 sm
2
K2 • 0 0 30
+ (2r)1- sm 2cos 2 cos"2 + 0(1) (40b)
* These eonditions eannot be eompletely satisfied by the Poisson-Kirehhoff theory of

plate bending whieh neglects the effeet of transverse shear deformation. As aresult, the
two eonditions on M xy and Vy in equation (39) are replaced by a single one on Vy
+ oMxy/ox, referred to as the equivalent or Kirehhoff shear. The eonsequences are that
the angular variations of the loeal stresses beeome dependent on the Poisson's ratio and
differ from those in equations (14) and the transverse shear stresses Txz and Tyz attain a
singularity of the order ,-t This is in eontrast to the exaet three-dimensional solution of
r-t-singularity for Txz and Tyz.
XXXII G. C. Sih
Kl . 8 8 38
M xy = (2r)t sm 2 eos 2eos"2
+ (~)t eos ~ (1 - sin %sin 3~) + 0(1) (40e)
Vx = - (~)t sin ~ + 0(1) (40d)
K3 8
Vy = (2r)t eos 2 + 0(1) (40e)
in whieh Kl and K 2 are the moment intensity faetors and K 3 the shear foree
intensity faetor. Note that the l/~ r singularity and the 8-dependenee in
equations (40) are the same as those shown in equations (14).
Moment and shear foree intensity faetors. Consider the problem of a through
eraek in a bent plate. The eraek is tiIted at an angle P with referenee to the
plane about whieh a bending moment of magnitude M is applied. This
loading as shown in Figure 6(a) is equivalent to the applieation of bending
~--~~----~~h
, r---~--L-----~-1t
r
(0) Angle crock under bending (b) Combined bending and twisting
Figure 6. Loading configuration for bending and twisting of a cracked plate
moments MI. M 2 and twisting moment M 12 in Figure 6(b) through the

transformation properties of the stresses:
MI = M sin 2 P (41a)
Introductory chapter XXXIII
M 2 =Meos 2 f3 (41b)
M 12 = M sin f3 eos f3 (41e)
For this problem, Hartranft and Sih [12] and Wang [i4] have provided the
K j (j = i, 2, 3) expressions and they are
Kl = cf>(1)M 1 via (42a)
K2 = '1'(l)M 12 via (42b)
v110
K 3 = - (1 + v)h Q(l)M 12 via (42e)
Equations (42) show that only M 1 and M 12 affeet the moment intensity
near the eraek tip while M 2 has no eontribution. The separate effeets of
M 1 , M 2 and M 12 ean be best illustrated by Figures 7(a) to 7(e) where normal
(a) Normal bending (b) Transverse bending (e) Twisting
Figure 7. Three separate problems of plate bending
bending produees Kl' parallei bending gives rise to no moment singularity or

intensity and pure twisting leads to both K 2 and K 3 • It is apparent from equa-
tions (41) and (42) that
Kl = cf>(l)M via sin2 f3 (43a)
K 2 = '1'(1) M via sin f3 eos f3 (43b)
K 3 = - (1 ~l~)h Q(l)M via sin f3 eos f3 (43e)
where the funetions cf>(l), '1'(1) and Q(l) are eomputed numerieally from
integral equations [12, 14]. Their values as a funetion of h/aJlO for different
Poisson's ratio ean be found in Figures 8 to 10.
XXXIV G. C. Sih
1.0
0.8
V = 0.5
0.3
~-
::E
0.0
......
:oo:
0.4
0.2
07-----~~------~------~------~~----~~----~~
o 0.5 1.0 1.5 2.0 2.5 3.0
h/a.fiO - Ratio of Thickness to Crack Length
Figure 8. Bending-moment-intensity factor versus ratio of thickness to crack length
OL-______ ~ ______ ~ ______- L_ _ _ _ _ _ -L~ ____ ~~ ____ ~~
o 0.5 1.0 1.5 2.0 2.5 3.0

h/av'iO - Ratio of Thickness to Crock Length
Figure 9. Twisting moment-intensity factor versus ratio of thickness to crack length

Introductory chapter xxxv
0.2
~ ~
~ 0.1
.....
~ '"
~
;:,
-+
h/a../iõ - Ratio of Thickness to Crack Length
Figure 10. Shear intensity factor versus ratio of thickness to crack Iength
IV. Direction of crack growth
In the absence of symmetry, the path of crack propagation is no longer

obvious since it may not coincide with its initial plane. The direction of
crack initiation as predicted from the strain energy density theory depends
on the intensity of the llr energy field near the crack tip. This intensity
e
measured by S varies as a function of r and through the stresses on the
element shown in Figure 5. With (lz = 0, equation (1) simplifies to
(44)
If the radius of the core region is kept sufficiently small, the n consideration
need only be given to the singular stresses. Substitution of equations (40)
into (34) and (37) yields
(lx
kl e( . 2e . 2"3e)
= (2r)+ eos 2" 1 - SID SID
XXXVI G. C. Sih
k . () ( ()
- (2r)t sm 2 2 + eos 2 eos2"
3()) + 0(1) (45a)
{1y
kl
= (2r)t eos 2
() (1 + sm. 2(),.sm 23())
k2 • () () 3()
+ (2r)t sm 2 eos 2 eos 2 + 0(1) (45b)
kl . () () 3()
L xy = (2r)t sm 2 eos 2 eos 2
-k2 eos -2()
+ (2r)t (1 . () . 3B) +
- sm - sm -
2 2
0(1) (45e)
L XZ = - (~)t sin + 0(1) I (45d)
k3 ()
L yz = (2r)t eos 2: + 0(1) (45e)
The eoeffieients k j (j = I, 2, 3) are the stress intensity faetors* whieh va ry

along the eraek front as a funetion of z, i.e.,
(46a)
k2 = 1~: 1[1(1) M Ja sin f3 eos f3 (46b)
k 3 = - 2(13 +
JI0 1- h
v)h 2 [(2Z)2] .
Q(I) M Ja sm f3 eos f3 (46e)
Inserting equations (45) into (44), the result may be east into the form
d W/d V = Sir where the strain energy density faetor S is given by
S = Allki + 2A 12 k l k 2 + A22k~ + A33k~ (47)
The quantities Aij (i,j = 1,2, 3) stand for
* Note from equations (43) and (46) that kj (j = 1,2, 3) are related to Kj (j = 1, 2, 3) as
kl = (12zjh 3) Kl, k2 = (12zjh 3) K2, k3 = (3j2h) [1 - (22jh)2]K3

Introductory chapter XXXVII
1
All = 8E (1 + eos e) [3 - v - (1 + v) eos e] (48a)
A 12 = 4~ sin e [(1 + v) eos e - (1 - v)] (48b)
A22 = 8~ [4(1 - eos e) + (1 + v) (3 eos e-I) (1 + eos e)] (48e)
(48d)
For the plate bending problem at hand, equation (47) beeomes
s=~[3<P(1)~SinPJ {(*Y F(e,p)+[1-e:)2J G(P)} (49)
with the funetions F(e, p) and G(P) being defined as

F(e, P) = B ll sin 2 P+ 2,1 B 12 sin peos P+ ,12 B22 eos2 P (50a)
5 22
G(P) = 8(1 + v) y eos P (50b)
where Bij = 8E Aij (i, j = 1, 2, 3). The parameters ,1 and y are

'1'(1) Q(l)
,1 = <P(1)' Y = <P(1) (51)
The numerieal values of ,1 and y ean be obtained from the graphs in Figures
8 to 10.
The strain energy density theory [3, 4] assumes that the eraek grows in
a direction eorresponding to e = eo whieh makes S in equation (49) a
mlmmum:
(52)
The vanishing of the first derivative of S with respeet to e gives

(1 + v) {2(1 - 3,12) sin 200 - (1 - ,1) (1 - 3,1) sin [2(0 0 + P)]
- (1 + ,1) (1 + 3,1) sin [2(0 0 - P)]} - 2(1 - v) [2(1 - ,12) sin Oo
- (1 - ,1)2 sin (Oo + 2p) - (1 + ,1)2 sin (Oo - 2P)] = 0 (53)
XXXVIII G. C. Sih
There exist two minimum values of S. One corresponds to the negative

angles eo on the tension side of the plate and the other corresponds to the
positive angles on the compression side. If the crack is always assumed to
initiate from the tension side, then only negative fracture angles will be
considered. Figure Il gives a plot of - eo in equation (53) as a function of
120·
lOO·
h la = 1.265
11= 0.0
Gl 0.1
õ>
«" 0.2
e 0.3
~
u
e 0.4
at 0.5
Gl
>
..
~
co
~
40·
Cbo
,
20·
O·
0
.
10 20· 30· 40· 50· 60· 70· 80· 90·
f3-Crack Angle
Figure 11. Variations of fracture angle with crack angle for constant h/a
the crack angle {3 for a fixed plate thickness to half crack length ratio of
h/a = 1.265. Under normal bending {3 = 90°, the predicted fracture angle is
eo = 0 which implies that the crack grows in a self-similar manner. As the
crack is tilted into the direction of bending by decreasing the angle {3, the
direction of crack initiation deviates mo re and more away from its initial
plane as the fracture angle - eo increases. The values of - eo are also seen
to be affected by the Poisson's ratio of the material. The variation appears
to be more pronounced for crack angle {3 in the middle range. Uniike the
plane theory of elasticity, the plate thickness to crack length ratio exerts an
Introductory chapter XXXIX
120°
li = 0.3
100°
80°
.!!
"""
<t 1.265
~ 1.897
.<:!
u 60° 2.530
e
u:., 3.162
.~
Õ
., 40°
z""
cbo
I
20°
0°
O' 10° 20° 30° 40° 50° 60° 70° 90°
/3- Crock Angle
Figure 12. Variations of fracture angle with crack angle for constant v
influence on the direction of crack initiation. For a fixed crack length and
angle [3, Figure 12 shows that the fracture angle - (Jo decreases with increas-
ing plate thickness. Such an influence tends to diminish for [3 cIose to O°
and 90°. Hence, changes in the hja ratio* should be accounted for when
comparing experimental data with theory.
V. Minimum strain energy density factors and

allowable bending moments
Once the fracture angles - (Jo are known, they may be inserted into equation
(49) to obtain the minimum S values or Smin. Since F«(J, [3) is the only
* Variations of fraeture angle -80 with hja also oeeurs in the stretching of plates. The
higher order theory of Hartranft and Sih [12] when applied to the mixed mode erack
problem in plane extension can prediet such an effeet.
XL G. C. Sih
funetion in equation (49) that depends on e, its minimum values Fmin (eo, [3)
are displayed graphieally in Figure 13 for hja = 1.265. The general trend of
4.0 71 = 0.0
= 0.1
= 0.2
= 0.3
= 0.4
= 0.5
1.2 '-::------'-_-'-_--'-_L------'-_-'-_--'-_"----------'-_ _
Õ 10° 20° 30° 40° 50° 60° 70° 80° 90°
f3 - Crack Angle
Figure 13. Normalized strain-energy density factor versus crack angle for constant hja
the euryes for Fmin or Smin tends to inerease with the eraek angle [3 and the
rate of this inerease beeomes greater as the Poisson's ratio is decreased. A
similar plot for F min (eo, [3) versus [3 with v = 0.3 is given in Figure 14. Here,
the ratio hja is vari ed from 0.316 to 3.162. For small values of [3, signifieant
ehange of F min are observed as hja is varied. This change diminishes as [3
approaehes 90° where all the euryes eonverge to the same value.
In order to estimate the allowable moment, Me, that a eraeked plate ean
sustain without eausing fraeture, a normalized moment quantity Mj(Eh 3 Smin)t
is defined by using equation (49):
Introductory chapter XLI
3.2
2.8
h/a = 0.316
0.632
1.265
.~ 1.897
"'c: 2.530
o'" 3.162
"0
'"
.I::!
ei
E
~
..!.
II = 0.3
f3 -Crack Angle
Figure 14. Normalized strain-energy density factor versus crack angle for constant v
Me .jhja
(Eh 3 Smin yl- = 3sinp{2cP(1)[(z/h) 2 F min(8 0 'P) + [1-(2z/h)2]G(P)}* (54)
where Me is undefined for P = O. Assuming that both v and h/a take the
respeetive values of 0.3 and 1.265, the normalized moment ean then be plotted
against P for different values of z/h, say from 0.1 to 0.5 as illustrated in
Figure 15. The eurve for z/h = 0 is outside the seale range of the graph.
For a fixed Smin, the moment M is seen to deerease monotonically with p.
The minimum value of Me oeeurs at a position where the bending moment
is applied normal to the eraek plane, i.e., p = 90°. Each value of z/h eorre-
XLII G. C. Sih
'1/ = 0.3
h/a = 1.265
6
'E
OI
E
0 4
::!E
""
OI
.!:!
Õ
E
~ 3
~'f
--"
~E
,..
"'..c:
UJ 2
::!E
O~~~~~~~~~~~~~~~~__
0° 10° 20° 30 40° 50° 60 70 ao 90°
P - Crack Anllle
Figure 15. Normalized moment versus crack angle for constant v and h/a
sponds to a layer of material in the plate with z = 0 being the middle layer
and z = 0.5h the outer or surface layer. According to the S-theory [3,4],
unstable crack propagation commences when Smin -t Sc. The critical value
Sc is a constant representing the fracture toughness of the matt"rial. When
Smin = Sc in equation (54), Mbecomes M c or the critical moment at incipient
fraeture. With this understanding, the results in Figure 15 indicate that

fracture always starts in the surface layer of the plate (zJh = 0.5) which
corresponds to the lowest critical moment M c for all crack angles {J.
With attention focused on the surface layer (zJh = 0.5) where fracture
initiates, Figures 16 and 17 study the influence of Poisson's ratio and hJa
ratio. The changes in M c as v is varied from 0.0 to 0.5 are not appreciable.
This can be seen from Figure 16. Increasing plate thickness or hJa, say with
Introductory chapter XLIII
1.4
1.2
1.0
e..
E
o
..,..
~ 0.8
~
o
E
~ 0.6
~I
-~c
~E
"'.c
UJ 0.4
:::::-
~
h/a = 1.265
0.2 z/h = 0.5
O~~~~~~~~~~~~~~~~__
0° 10° 20° 30° 40° 50° 6cf 70° 80° 90°
/3 - Crack Anllle
Figure 16. Normalized moment versus eraek angle for eonstant hja and zjh
a being constant, tends to lower the normalized moment that can be applied
to a plate without causing fracture. This tendency is cIearly shown in
Figure 17.
VI. Conelusions
When plate-like structural members are subjected to bending loads, the

in-plane stresses are required to change sign in the thickness direction and
the problem acquires a three-dimensional character. Since the through
crack geometry* prohibits any exact three-dimensional solutions. the elas-
* The major drawbaek arises from an ineomplete knowledge of the stress singularity on
the plate surface where it interseets with the erack. Further diseussions on this subject
can be found in [15, 16].
XLIV G. C. Sih
1.4
1.2
1.0
h/a = 0.316
c:
Q)
0.632
E 1.265
o
:E 0.8 1.897
2.530
3.162
0.2 V = 0.3
z/h = 0.5
f3 - Crack Angle
Figure 17. Normalized moment versus crack angle for constant v and z/h
ticity theory must be approximated by appealing to minimum principles in

the calculus ofvariations. A variety of approximations have been made and
led to various formulations known as plate theories. For instance, the most
widely used theory of Poisson-Kirchhoff in plate bending involves alteration
of the crack surface boundary conditions. Such a change has a drastic influ-
ence on the qualitative character of the stresses near the crack tip and will
seriously affect the direction of crack initiation predicted from any assumed
fracture criterion. As an example, the maximum principal stress criterion *
states that crack growth takes place in a direction normal to the plane on
which the tangentjal normal stress (Ja is a maximum or the shear stress 'ra
* This criterion yields suffieientiy accurate results only when one of the prineipal stresses
dominate.
Introductory chapter XLV
vanishes. Consider the pure twisting case shown in Figure 7(c). The 'r9 = 0
plane based on the Poisson-Kirchhoff theory can be found from Sih's work
[15] as
(J
o
= _2
cos
-1 (2(2 +V))+
5 + 3v (55)
which gives fracture angles of (Jo from - 53.2 to - 57.4 for v ranging from
0
0
0.0 to 0.5. This differs from the value of (Jo obtained from the Reissner's
theory in which the crack surface boundary conditions are satisfied. Equation
(45c) when expressed in cylindrical polar coordinates (r, (J) and 'r9 set equal
to zero gives a fracture angle*
(Jo = - cos- 1 (1) = - 70.5° (56)

The difference in (Jo as obtained from equations (55) and (56) is significant
and cannot be ignored in failure analsyis. A similar discrepancy is found if
the strain energy density factor theory is used.
Another important feature of the plate bending problem is that the mate-
rial elements in the thickness direction may not fail simultaneously as
assumed in the theory of plane elasticity. This is because the stress intensity
factors kl, k 2 and k 3 no longer remain constant along the crack front. The
location of the critical element that triggers fracture must depend on some
combination of k j (j = 1,2,3) reaching a critical value. Referring to the
cylindrical polar coordinate system (r, 0, z) in Figure 5, the position of the
critical element will be denoted by (ro, (Jo, zo), where ro is the radius of core
region, (Jo the fracture angle obtained from the condition of S reaching a
minimum and Zo corresponds to the greatest value of Smin along the crack
front. The bending of an angle crack as shown in Figure 6(a) was analyzed.
Predictions of the S-theory showed that the fracture angle (Jo not only
depended on the crack position relative to the direction of bending but also
on the Poisson's ratio of the material and the plate thickness to crack length
ratio. These factors should not be ignored in comparing experimental data
with theoretical results.
For the sake of simplicity, the in-plane stresses have been assumed to vary
linearly through the plate thickness and the transverse shear stresses are
parabolicfunctions ofz. Higherorder plate theories [12, 17] involving more
* The experimentaI resuIt of Erdogan and Sih [6) on the twisting of cracked plates gives
values of Oo cIoser to 70.5 0 than those predicted by equation (55).
XLVI G. C. Sih
generalized functions of z are available. Sih [17] has proposed to expand the
z-dependance,j'(2zJh), for the transverse shear stresses 'r xz and 'ryz in a
trigonometric series*
m
j'(2zJh) = ho - L (-it bn cos [(2nnzjh)] (57)
n=l
which satisfies the conditions that 'r xz = 'ryz = 0 on the plate surfaces
z = ±h/2, i.e.,f'(±I) = O. The constants ho and h n are given by
(58)
where a is the half crack length. The derivative of j'(2z/h) with respect to
Z or J"(2zjh) gives the variations of the in-plane stresses (lx, (lyand 'r xy
with z. A series of experiments on the bending of plexiglass plates [18] gave
data checking remarkably weIl with the predictions made by Sih [17]. The
stress intensity factor solution for this problem may also be used in conjunc-
tion with the strain energy density factor theory to predict the fracture
behavior of plates under bending andjor twisting. The theoretieal results
wiII differ but not significantly from those based on the Reissner theory.
Another shortcoming of the conventional plate theory is that (lz is neg-
lected throughout the plate and hence the plane strain condition
(59)
ahead of the crack cannot be realized. Hartranft and Sih [12] have construc-
ted a theory of plate stretching or bending in which the functionJ(2z/h) in
(60)
wiII not be assigned arbitrarily but be determined from equation (59), a
condition derived from the exact three-dimensional theory of elasticity [19].
ViIIarreal, Sih and Hartranft [20] have used a frozen stress photoelastic
technique to measure the stress variation through the thickrtess of a thick
plate containing a crack and the results were in excellent agreement with
those postulated by the theory of Hartranft and Sih.
* Reference can also be made to equation (2.82) in Chapter 2.

Introductory chapter XLVII
References
[1] Bridgman, P. W., Studies in Large Plastie Flow and Fraeture with Special Emphasis
on the Effeets of Hydrostatie Pressure, First edition, McGraw-Hill Book Co., New
York (1952).
[2] Sih, G. c., Strain energy density factor applied to mixed mode crack probIems,
International Journal of Fraeture, 10, pp. 305-321 (1974).
[3] Sih, G. c., A special theory of crack propagation: methods of analysis and solutions
of crack probIems, Meehanies of Fraeture I, edited by G. e. Sih, Noordhoff Inter-
national Publishing, Leyden, pp. 21--45 (1973).
[4] Sih, G. c., A three-dimensional strain energy density factor theory of crack propa-
gation: three-dimensional crack probIems, Meehanies of Fraeture II, edited by G. C.
Sih, Noordhoff International Publishing, Leyden, pp. 15-53 (1975).
[5] Linear Fraeture Mechanies, edited by G. e. Sih, R. P. Wei and F. Erdogan, Envo
Publishing eo., Inc. Lehigh Valley, Pennsylvania (1976).
[6] Erdogan, F. and Sih, G. c., On the crack extension in plates under plane loading
and transverse shear, Journal of Basie Engineering, 85, pp. 519-527 (1963).
[7] Sih, G. C. and eha, B. e. K, A fracture criterion for three-dimensional crack
probIems, Journal of Engineering Fraeture Mechanies, pp. 699-723 (1974).
[8] Kipp, M. E. and Sih, G. e., The strain energy density failure criterion applied to
notched elastic solids, International Journal of Solids and Struetures, 2, pp. 153-173
(1975).
[9] Sih, G. c., Discussion on some observations on Sih's strain energy density approach
for fracture prediction by I. Finnie and H. O. Weiss, International Journal of Fraeture,
10, pp. 279-283 (1974).
[10] Sih, G. C. and Macdonald, B., Fracture mechanies applied to engineering problems
-strain energy density fracture criterion, Journal of Engineering Fraeture Mechanies,
6, pp. 361-386 (1974).
[11] Sih, G. e. and Kipp, M. E., Discussion on fracture under comples stress-the angle
crack problem by J. G. Williams and P. D. Ewing, International Journal of Fraeture,
10, pp. 261-265 (1974).
[12] Hartranft, R. J. and Sih, G. c., An approximate three-dimensional theory of plates
with application to crack probIems, International Journal of Engineering Scienee, 8,
pp. 711-729 (1970).
[13] Reissner, E., On bending of elastic plates, Quarterly of Applied Mathematies, 5,
pp. 55-68 (1947).
[14] Wang, N. M., Twisting of an elastic plate containing a crack, International Journal of
Fraeture Mechanies, 6, pp. 367-378 (1970).
[15] Sih, G. c., A review of the three-dimensional stress problem for a cracked plate,
International of Fraeture Mechanies, 7, pp. 39-61 (1971).
[16] Benthem, J. P., Three-Dimensional State of Stress at the Vertex of a Quarter-Infinite
Craek in a Half-Spaee, Delft University Report No. 563 (1975).
[17] Sih, G. c., Bending of a cracked plate with an arbitrary stress distribution across
the thickness, Journal of Engineering for Industry, 92, pp. 350-356 (1970).
[18] Rubayi, N. A. and Ved, R., Photoelastic analysis of a thick square plate containing
XLVIII G. C. Sih
central crack and Ioaded by pure bending, International Journal 0/ Fraeture, 12,
pp. 435-451 (1976).
[19] Hartranft, R. J. and Sih, G. c., The use of eigenfunction expansion s in the general
solution of three-dirnensional crack problerns, Journal 0/ Mathematies and Mechanies,
19, pp. 123-138 (1969).
[20] VilJarreal, G., Sih, G. C. and Hartranft, R. J., Photoelastic investigation of a thick
plate with a transverse crack, Journal 0/ Applied Mechanies, 42, pp. 9-14 (1975).
M.Isida
1 Interaetion of arbitrary array of

eraeks in wide plates under elassieal
bending
1.1 Introduction
This paper presents a general method of analysis of arbitrary arrays of cracks

in wide plates by means of the classical theory of plate bending. This theory
is based on the Kirchhoff boundary conditions, which lead to approximate
satisfaction of the free surface conditions on the crack. In spite of this
approximation, useful stress intensity factor solutions for complicated
crack geometries can be obtained.
The analysis is similar to that in the author's previous papers on two-
dimensional [1] and antiplane shear problems [2], and is based on the concept
of the Laurent series expansions of the complex potentials proposed by
S. Moriguti [3]. In the numerical caleulations a perturbation procedure is
employed, and the crack tip stress intensity factors are given in power series
of the relative crack length. Typical problem s of practical importance are
analysed and the numerical results are discussed.
1.2 Basic relations
Consider the axes x, y in the middle plane of the plate with thickness h, and
the third axis t as shown in Figure 1.1. The moments Mx, My, M xy and
shearing forces Qx, Qy are defined as follows:
Mx = f h'2 tCTxdt, My = fh'2 tCTydt, M xy = fh'2 tTxydt (l.1a)

-h/2 -h/2 -h/2
Qx = f
h'2 Txtdt, Qy = fh'2 Tytdt (l.1b)
-h/2 -h/2
2 M.Isida
In small defleetion of tbin pIates, the bending stresses are assumed to be

proportionaI to the distance from the middIe pIane, and the first two
expressions in equations (l.1a) give
_ Mxt _ Myt
(1x - 3h3' (1y - 3h 3 (1.2)
and
Mx M~
(1B,x = (;jiZ' (1B,y = 6h (1.3)
are the extreme outer fiber stresses.

Equations of equiIibrium of a small pIate element are
oQx + oQy + p(x, y) =0 (1.4a)

ox oy
oMx + oMxy = Q (1.4b)

ox oy x
oMxy + oMy = Q (l.4e)

OX oy y
where p(x, y) is the density of the transverse Ioading aeti ng on the pIate
surfaee.
The moment components are reIated to the defleetion w as follows:
Mx =- D(~:~ + v~:~) (1.5a)
My =- D(~> + v~:~) (1.5b)
02 W
M xy = - D(l - v) oxoy (1.5e)
where D is the bending rigidity of the pIate defined as

Eh 3
D = 12(1 _ v2) (1.6)
and E and vare, respeetiveIy, the Young's moduIus and Poisson's ratio.
Substituting equations (1.5) into (l.4), the governing equation for the
defleetion is obtained as follows:
04W +2 04W + 04W _ p(x, y) (1.7)

ox4 ox 2oy2 oy4 - D
Arbitrary array of eraeks in wide plates under classieal bending 3
If the plate is free from transverse loading a biharmonic equation is ob-

tained:
(1.8)
Thus the displacement is a biharmonic function, and can be expressed by

two complex funetions cjJ(z) and l/1(z) as follows:
Dw = Re [zcjJ(z) + l/1(z)] (1.9)
The moments and shearing forees which are written in terms of w as

show n in equations (1.5) and (1.4) can also be expressed in terms of the
derivatives of cjJ(z) and l/1(z):
Mx + My = - 4(1 + v) Re[cjJ'(z)] (LlOa)
Mx - My - 2iM xy = - 2(1 - v) [zcjJ"(z) + 1/1" (z)] (LlOb)
Qx - iQy = - 4cjJ"(z) (LlOc)
A few comments wiII be made on the Kirchhoff boundary conditions in

the cIassical plate bending. It is sufficient, without loss of generality, to
consider the plate boundary parallei to x-axis. The physical boundary
conditions require that My, M xy and Qy take some assigned values, but
only two conditions are allowed on the plate edge as required in the cIas-
sical plate bending theory. In this connection, the quantities M xy and Qy are
replaced by single quantity referred to as the equivalent shearing force
Vy = Qy + oMxY/ox. A traction free boundary would the n require My
and Vy to vanish which in terms of the complex funetions may be written as
cjJ'(z) - flqJ'(Z) + zcjJ"(z) + l/1"(z) = an imaginary constant (1.11)

where
3+v
fl=-- (1.12)
1- v
Now, consider the IocaI stress field cIose to the crack tip. Express the stress
components obtained by the cIassicaI bending theory in the polar coordi-
nates as shown in Figure I. I, and expand them in ascending power series
of r. The first terms which are predominant in the vicinity of the crack tip
become [4]:
4 M.Isida
Figure 1.1. Bending of a plate containing a crack
kl t ( lJ SlJ) k2 t ( 9 + 7v . lJ
(1x = (2r)t 2/lh - 3 eos 2" - eosT + (2r)t 2/lh - ~SID2
+ sin S;) (1.13a)
kl t (11 + Sv lJ SlJ) kl t (. lJ . SlJ)

(1y = (2r)t 2/lh 1 _ v eos 2" + eos 2" + (2r)f 2/lh SID 2" - SID 2"
(1.13b)
kl t ( 7 + v . lJ . SlJ) k2 t (S + 3v lJ
'r xy = (2r)t 2/lh - 1 _ v SID 2- SID "2 + (2r)t 2/lh T=-; COS 2
_ eos S;) (1.13c)
f(t) ( 3lJ . 3lJ)

'rxt = (2r)t - kl eos "2 + k 2 SID "2 (1.13d)
f(t) ( . 3lJ 3lJ)

'r yt = - (2r)t kl SID "2 + k 2 eos "2 (1.13e)
where
hl _ 41 1
f(t) = 4/(3 + v) (1.14)
eoeffieients kl and kl in equations (1.13) are the stress intensity faetors in

the dassical plate bending theory. In general, the above expansion proeedure
Arbitrary array of eraeks in wide plates under dassical bending 5
may be skipped to get the stress intensity factors. They can be obtained
directly from the complex potential cj>(z) by the relation [5]
k 1 - I'k 2 -- -
12.j2(3 2
+ v) I'1m [( z - a)t 'I'"!"'( z)] (1.15)
il z--+a
1.3 Complex potentials for traetion free eraeks
Consider a crack of length 2a in the physieal plane and introduee the

mapping function
(1.16)
whieh transforms the eraek edge in z-plane and its outer region into a unit
eirele and its outer region in '-plane as shown in Figure 1.2.
o
x
20
l - plone ,- plone
Figure 1.2. Conforrnal rnapping
The inverse of equation (1.16) yields
'=~[l +(l-(:rrJ (1.17a)
el =~[1- (1-(:rrJ (1.17b)
whieh together with equation (1.16) determine the points at infinity in the
both planes.
General expressions of the eomplex potentials for traetion free eraeks wiII
6 M.Isida
be developed in series forms. Assuming that the components of the shearing

force and moment have no singularities within a region enc10sing the crack
as itis usually the case, the complex potentials 4>'(z) and tfJf/(z) must be analytic
funetions within that region as it can be seen from equations (1.10). It
follows from a weIl known theorem that they can be expressed in Laurent
series of z convergent in a region bounded by two concentric circ1es. But
here, for convenience of caleulation, Iet us begin with the following expan-
sions in ,:
oo
4>'(z) = eto + L (etm,m + PmC m)
m=l
(1.18a)
dtfJ'(z)
~
a[
=2" Yo
;, (
+ m';:l
vm + (5m':.r-m)]
Ym~ (1.18b)
where the coefficients in the righthand sides are generally complex. Now, let
(1.19)
such that dots and primes denote real and imaginary parts of the correspond-
ing quantities.
The complex funetions should aIso satisfy the single-valuedness condition
of displacement. Derive the expressions for 4>(z) and tfJ(z) by integrating
equations (1.18) and making use of the relations
(1.20)
results are then substituted into equation (1.9). The terms in the imaginary
part of log' cause multivalued displacement, and by taking their coeffi-
cients to be zero the following relations are obtained:
(1.21)
Now, the traction free relation in equation (1.11) is rewritten with the
help of the eomplex potentials in equation (1.18) as follows:
(1 - C 2{ 1X0 + m~1 (IXmC" + f3m~m) - /l{ IXo + m~1 (IXm~m + /3m~-m)}J

oo
+ (~+ ~-1) I m(lX mC"-1 - f3mC(m+l) + Yo
m=l
oo
+ I (Ym(m + 15 mC m) = iC(l _ ~-2) (1.22)
m=1
where C is a pure imaginary eonstanL Replaee ~ by I for the eraek edge, e
rearrange both sides of equation (1.22) in aseending power series of "
equate the eoeffieients of the same powers, and the following relations are
obtained:
( 1.23a)
/l(f3m+2 - 13m) = - (m + 1) (IX m + IXm+2) - Ym(m ::2: 1) (1.23b)
15 1 = /l(IX I - /31) ( 1.23e)
15m = (- iC + 1X0 - /lIXoM; - /l(IX m -2 - IXm)

+ (m - 1) (f3m-2 + 13m) (m::2: 2) (1.23d)
Equations (1.23), together with (1.21) ean be used to express 13m and 15m in
terms of IXm and Ym' Henee, the expressions in equations (1.18), (1.21) and
(1.23) give the general forms of the complex potentials in terms of ( which
are suitable for solving traetion free eraek probiems.
Finally, the re suits ean be transformed into the z-plane, by using equations
(1.17). The algebraie ealeulations are tedious and are similar to those in the
plane problem [6]. The final results of the general Laurent series expansions
for the eomplex potentials of stress free eraeks are given by
Dw = Re [zcp(z) + I/I(z)] (1.24a)
oo
cp(z) = I [(F: + jF~)z-("+1) + (L: + jL~)zn+l] (1.24b)
n=O
oo oo
I/I(z) = - D~ log z + I (D: + iD~)z -n + I (K: + iK~)zn+2 (l.24c)
n=1 n=1
where
oo
D "2n -
- "
p
L.. a 2n+2 +2(p2"K"
2p 2p + R 2p
2n L"2p ) (l.25a)
p=O
8 M.Isida
oo
F 'Zn = "azn+zp+z(QznK'
L. Zp Zp
+ SznL'
Zp Zp
) (1.25b)
p=o
oo
D'Zn+l = "Zn+Z )
L. a
P+4(pZn+1
Zp+l
K'
Zp+l + R Zn + L'
Zp+1' 2p+l ( l.25c)
p=O
oo
F 'Zn+l = ,,2n+zP+4(Qzn+'K'
L. a Zp+l Zp+l + s2p+l
zn+'L')
2p+1 ( 1.25d)
p=O
oo
"
L.
aln+zP+2(T2n K'
2p Zp
+ V Zp
2n L' )
lp (1.25e)
p=o
oo
F'2n = _"
L. a Zn+Zp+2 eU 2p K'Zp
2n + W ln L' )
2p lp ( l.25f)
p=O
oo
D'2n+ 1 -- -
,,211+2 p +4(T 2n + 1 K'
L. a Zp+ 1 2p+ I + 1/Zn+ I
r 2p+ 1
L'2p+ I ) ( 1.25g)
p=O
oo
' 1 --
F 2n+ -
,,2n+2 P +4culn + I K'
L. a 2p+ I Zp+ 1 + W 2p+
2n + I L'
I 2p+ I
) (l.25h)
p=o
The coefficients p~; to W~;! ~ in the above expressions are defined as
po =
lp
(1 _.Il!.) 2p2 + 2(2 PP+ 1) 2p + 2
( 1.26a)
{ p~n} 2p + 2[1 n. 2m + 1) (2 P + 1) p (
T~; = 2 + 2n m 37o}±1- ~l
2p p_m An -m,2m+l
1
li
- -Il1n-_L\,{Olp 2
2m + 1(2P + I)
-2
m + P- m
A n - m - , ,2m+2
J(n >= .
1 fOI
2n'
P2p) ( l.26b)
m- tf
{ Q~; }_
U 2n -
2p +
'2 2p + 1
2 L
n, p
_ nl1) A n - m,2m+l
(2 P + ( 1.26c)
2p Il m={~} P
R°
2p =(1 - .!.)~ (2 P)(2 -~)
Il 2 2p + 2 P p+1
(1.26d)
2n 2p+
R 2p = 2 2p + 2
1[( P1) 1-
(2 P)
P A n- 1 ,2
4n-
-~ m~o
' ,p 2m ( 2 P )
2m+2 p-m A n - m - 1 ,2m+2
Arbitrary array of eraeks in wide plates under classical bending 9
+ !{( _!) (2 n, _ J l _ 2m + 1)(

P)An,l + m'2;op-1 (2m+l 2p )A
n Jl Jl P Jl p-m-l n-m,2m+1
n, P
- m~l 2m
(Jl
+ 1+
6m -
Jl
1) ( 2p ) }]
p _ m A n- m,2m+l ,(n ~ 1) (1.26e)
+ 2n
2
+
n+l'p(
1 m'2;l 1 +
2m
Jl
-1)( 2p )
p - m A n- m+1 ,2m
J (1.26f)
(1.26g)
w2n = 2p +
2p 2 2P + i
1[(1 _!)Il _P_
+1
(2 P
)A
P P n, 1
1 n, p 4111 ( 2p )
+-Il m'2;o - -1
2111 + p-m A n-m,2m+l
_ 2_ 1I+1'P(1_ 2111 - 1)( 2P )A ] (1.26h)

2n + 1 m'2;l Il P -111 n-m+l,2m
{ P~;:~}
T 2n + = 1
2p + 3 [
2 2p +2 - -
1"-L1,p 2-+
2/11 + 2(2 P+ 2)
An- m-
1 1 ,2m+3
2p+ 1 Il »1=0 111 - P - 111
- 21!
1
+
n, p
1»1'2;0
(2111 + 2 - 1)
Jl +
(2PP- +l1l2) A II - m,2m+2
J (1.26i)
fQ;;:~}
") U 2n +1
2p + 3 n,p (2 P + 2)
L P _ 111 An- m,2m+2
= Jl '22p+2 m=O (1.26j)
l 2p+l
10 M.Isida
{R~;~~} 2p +
V~;~~ = 22p +2
2[ ± P+ 1)
(2
p An, 1
_ ~n-I'P2m + 1(2P + I)A

{/. m~o 2m + 3 p - m n-m-l,2m+3
1 {n.m~oP-l(2m+2-
{/. 2m + 2) ( 2p + 1)
p-m-l An- m,2m+2
-:~o
+211+1 {/.
(2m{/.+ 2 + 6m{/.+ 2) Gp_+ ~)An-m, 2m+2}] (1.26k)
{ S~;~~
W~;!!
} = 2p +
2 2p +2
2[+ 2n + 1(2P + l)A
2n+2 p n,1
+ ~ ~ 4m + 2 (2 P + l)A (1.261)
{/.m';'o2m+2 p-m n-m,2m+2
+ 2-2
n+ L
2 n+I'P(2m
m;O
1 _
-{/. ± )(2PP +l)
m
An-m+I.2m+1
]
Two values for the lower limits of the summations eorrespond to the two
quantities in the lefthand sides, and for the upper limits the smaller value
should be taken. The quantities A in equations (l.26) are defined as the
eoeffieients of the following expansion
oo
[Z - (Z2 - l)tr = L An, m Z- l 2n+m) (I Z I > 1) (1.27)
n;O
and their closed form expressions are [7]
A n-m,2m = m ( 2n ) (1.28a)
22n-n n - m
A n- m. 2m+ 1
2m +
= 22n +1(2n + 1) n -
1 (2n + 1)
nl (1.28b)
The general forms of the eomplex potentials in equations (1.24) together

with the results in equations (1.25) and (1.26) are eonvenient for solving
boundary value problems in general. In those expressions, the eraek tip
singularities do not appear explieitly, but in the final stage of the analysis
they ean be obtained in closed forms.
The method will be illustrated by a simple example eonsisting of a wide
Arbitrary array of eraeks in wide plates under c!assieal bending 11
plate with an isolated eraek subjeeted to M; = M. The doubly symmetrie

nature of the problem suggests that
oo
4J(z) = L (F;nz-(2n+l)+L;nz2n+l) (1.29a)
n=O
oo oo
l/J(z) = - D~ log z + L D;n z-2n+ L K;n z2n+2 (1.29b)
n= 1 n=O
The boundary eonditions at infinity together with equations (1.10) give
. M . M
L O = - 4(1 + v)' K O = 4(1 - v) (1.30a)
L;n = K;n = 0 (n:2: 1) (1.30b)
and the traetion free relations reduee equations (1.25) to
D;n = a 2n +2 (p~n K~ + R~n L~) (1.30e)
F;n = a 2n +2(Qgn K~ + s~n L~) (1.30d)
in whieh the eoeffieients p~n to s~n are given by equations (1.26). The
eomplex funetion 4J(z), required for ealculating the stress intensity faetor,
is transformed as follows:
oo
4J(z) = L~z + L a2n+2(Q~n K~ + s~n L~) z-(2n+l)
n=O
= L~z + n~O
~ a 2n +2 [K~.!./1 + L~.!.(.!.
2 /1
- I)JAn,l Z -(2n+ 1)
-- - Mz
4(1 + v)
+ 2(3Ma oo
+ V)n~o A
(a)2n+
-
n.l Z
1
Mz M [ 2 2
4(1 + v) + 2(3 + v)
J,.
= - z - (Z - a )'] (1.31)
Substituting equation (1.31) into (1.15), the following stress intensity faetor
is obtained:
(1.32a)
where
(1.32b)
12 M.Isida
1.4 Arbitrary array of eraeks in wide plate
Let a wide plate be subjeeted to bending at infinity. An arbitrary array of

eraeks are loeated with their eenters at points Oj (j = 1, 2, ... , N). The
axes X, Yare plaeed in the middle plane of the plate while the eorresponding
polar eoordinates are denoted by (p, [3). The loeation of the center of the
j-th eraek is (Pj' [3), its length is a j and inclination angle is IX j. Introduee
the N eoordinate systems (Xj , Y) (j = 1,2, .. . ,N) with their origins at
oj and with Xraxes along the eraeks. For eonvenienee the following dimen-
sionless eoordinates, eomplex variables and parameters are defined:
x Y .
x = d' Y = d' Z = x + lY (1.33a)
X. Y.
Xj = dJ,Yj=T,Zj=xj+iYj (1.33b)
Aj = d' rj = p.d' rjk = dP'

a· k
(j, k = 1,2, ... , N) (1.33e)
mM
mM
Figure 1.3. Arbitrary array of crack in wide plate

Arbitrary array of eraeks in wide plates under dassieal bending 13
in which d is a referenee length whieh may be arbitrarily specified. In the

numerical examples of seetion 1.5, d will be taken as the erack spaeing.
Attention is now turned to the new definition of Z whieh is different from
that in the preeeding seetion. The radius veetor between 0 j and 0 k is given by
( 1.34)
The plate is assumed to be of infinite extent and subjeeted to uniform bending

moment at infinity:
M': = 1M, M'; = mM (1.35)
Assume a displaeement funetion w of the form:
( 1.36)
where Wo is the displaeement for the uneracked plate under the same mo-
ments at infinity, and henee as
Dw o = Md 2[z<po(z) + I/Io(z)] (1.37a)

in whieh
<Po(Z) = Loz, I/Io(z) = K oz 2 (l.37b)

and
m+1 m-I
L o = - 4(1+ v)' K o = 4(1 - v) (1.37c)
The displaeement funetions wk (k = 1,2, ... , N) in equation (1,36) contain

the craek tip singularities and ean be written as:
DW k = Md 2 Re[zk<Pk(Zk) + I/Ik(Zk)] (1.38a)

such that
(1.38b)
oo
I/Ik(Zk) = - D~, k log Zk + I (D:, k + iD~, k)Z;:n (l.38e)
n=l
where the eoeffieients with dots and primes denote real and imaginary part s
14 M.Isida
of the quantities involved. They should be determined from the boundary

eonditions on the eraek. The boundary eonditions at infinity are satisfied
by W o giving rise to the applied bending moments as W k gives rise to no
stresses there. Therefore, there remains only the eonsideration of traetion
free eonditions along the edges of all eraeks.
In order to eonsider the j-th eraek, it would be eonvenient to express W
in term of the only eomplex variable Zj' From simple geometry, the following
relations hold:
Z = Z
J
.e iaj + rJ.e iPJ ( 1.39a)
Zk = zjei(aj-a kl - rjkei(PJk-ak) (1.39b)
Substituting equations (1.39) into (1.37) and (1.38), Wo and Wk (k =F j) ean

be written in terms of Zj as follows:
Dw o = Md2Re[zjcI>~(z)+ 'l'6'(Zj)J (1.40a)

with
cI>~(Z) = Loz j, 'l'~(z) = Koz} (1.40b)

and
DW k = M d 2 Re [zjcI>t(z) + 'l't(z)J (1.41a)

with
(1.41b)
(1.41e)
The eomplex potentials in equations (1.38) are reexpanded in power series

of Zj assuming that I Zj I < rjk' Henee Wo and wk are expressible in positive
power series of Zj' In this way, the total displaeement funetion W beeomes:
(1.42a)
in whieh
oo
m. j ( Zj )
.... _
-
"L. [(F·n, j + I'F')
n, j Zj-(n+ 1) + (L·n, j + I'L') n+ 1J
rl. j Zj (1. 42b)
n=O
oo oo
'l'/Z) = - D~,j log Zj + L (D~,j + iD~,)zjn + L (K~,j + iK~,)Z'j+2
n=1 n=O (1.42e)
The eoeffieients in the eomplex potentiaIs are given by
, .0"" (
Ln,j = LOLJ n + L. L. en,j '
oo
p,k F p,k
N
(P,kF')
+ Jn,j p,k (1.43a)
p=O k*j
oo N
L~,j= I L: (-J:.'fF~,k+e:"fF;,k) (1.43b)
p=O k*j
oo N
K n,' j = K oLJAOn + "L. "( ' k + bp,kD'
L. an,p,kj Dp, n, j p, k + en,
,p,kj F p,' k + dp,k ' )
n, j F p, k
p=O Hj (1.43e)
oo N
K n,' j = "L. "
L. ( - bp,kD'
n, j p, k + an,p,kD'
j p, k - dp,kF'
n, j p, k + en,p,kj F 'p, )
k (1 ' 43d)
p=O k*j
A~ are the Kroneeker deItas and the eoeffieients in the series are geometrie
eonstants whieh are defined by
o,k _ eos [(n + 2) (f3jk - aj)] bo,k _ sin [(n + 2) (f3 j k - a)]
(1.44a)
an,j - (Il + 2) (rjkt+ 2 'n,j - (n + 2) (rjkt+ 2
[p(aj - ak) + (n + p + 2) (f3jk - aj)]

( -I)P( n + p + 1) -'-----'-----:-=r-;;-:;:-,,------(p ~ 1)
n+2 (r jk t+ P+2
(1.44b)
n, J~}
eP,
{ _
dP'~
n, J
-
[ep + 2) (a j - ak) + (n + p + 4) (f3jk - a)]

(-lY ( n + p + 2) -'----'------,--(-----:-;;)n-:L+p=+,,-_ _ _ __
n +2 r jk
(1.44e)
{fep,~}
n, J
P'~
_
-
n, J
(-1 Y+ l(n + +p +1 1) -'----L------(-)-=-n

n
[ep + 2)(aj - ak) + (n + p + 2) (f3jk - a)]
jk
+O::p"+.,.-------
r
(1.44d)
The eomplex potentiaIs have been redueed to the same forms as those in
equations (1.24), and the reIations for traetion free eraeks are written in
16 M.lsida
forms given by equations (1.25). The above proeedure holds for eaeh of the
eraeks, and the relations to be used on eraek edges ean be summarized:
oo
= '\' A~n+2p+2(r.n K" .
1... J 2p 2p,]
+ R 2p
2n LO
2p, J
.) (1.45a)
p=o
oo
= '\'
1...
A~n+2p+2(Q2n KO
J 2p 2p,]
. + S2n LO
2p 2p. J
.) (1.45b)
p=o
oo
DO -
2n+l,i -
'\' A2n+2p+4(p2n+l KO
1... i
p=o
2p+l 2p+l,i + R 2p+l
2n + 1 LO
2p+l,i
) (1.45e)
oo
pO _
2n+l,i -
'\' A2n+2p+4(Q2n+l KO
1... i
p=o
2p+l 2p+l,i + S2n+l LO
2p+l 2p+l,i
) (1.45d)
oo
'
D 2n, i
= - '\'
1...
1~n+2P+2(T2nK'
AJ 2p 2p, J
. + V 2p
2n L'
2p, J
.) (1.45e)
p=o
oo
P '2n, i = _ '\' 12n+2p+2(·u2nK' .
1... A J 2p 2p,]
+ W 2p
2n L'
2p, J
.) (1.45f)
p=o
oo
'
D 2n+l,i= '\' 12n+2 p +4(T 2n + 1 K' 1/2n+1L') (1.45g)
- 1... Aj 2p+l 2p+l,j+ ~2p+l 2p+l,i
p=o
oo
'
P 2n+l,i = -
'\' A2n+2p+4(U2n+1K'
1... j
p=o
2p+l 2p+l,i + W 2p+l
2n + 1 L ' )
2p+l,i (1.45h)
where the eoeffieients p~; to w~;1! are given by equations (1.26).

The next step in the analysis is to determine the unknown eoeffieients
F~,i' P~,i' L~,i' L~,i' D~,i' D~,j, K~ ,i' K~,i (n = 0, 1, ... ; j = 1,2, ... , N) in
equations (1.43) and (l.45). To do this, a perturbation technique is applied.
For eonvenienee, the ratio of the eraek lengths is fixed by the eonstants
si (j = 1,2, ... , N). Let Aj = s). and all the unknowns are expanded in
power series of the only parameter A:
oo oo
DO (2p)
D;n,i = L
p=n+l
DO(2p)'2p
2n, i A , D;n+l,i = L
p=n+2
2n+l,j
A 2p
, (1.46a)
oo oo
D;n,i = L D'(2~)A lp
p=n+l
2n, J ' D;n+l,i= L D'2n+l
p=n+2
(2p) A 2p
i ' t
(1.46b)
oo oo
P; n, J.= L
p=n+l
pO (lp)' 2p
2n, i A P;n+l,j = L pO2n+l,i
p=n+ 2
(2p) '2p
A (1.46e)
p=/J+I
F;n+I,j =
p=n+2
L F' (2p) Jc2p
2n + I, j (1.46d)
oo oo
'
K n, j =
K AD
DLJ n + " K ' (~p112p
L.. n,) /!. ,
K'
n,]
. = "1..J K'n,j
(~p)Jc 2p (1.46e)
p=1 p=1
oo oo
L'
n,)
. = L
° ,1D
n
+ "L.. L'n,)(~p)Jc 2p L'
n. J
. = "
i...J
L' (2p)Jc 2 p
n, J (1.46f)
p=1 p=1
Substitute equations (1.46) into (1.43) and (1.45), rearrange both sides of the
resulting equations in ascending power series of Jc and equate terms of the
same powers. The final results are:
eCO) -
0, j -
L
0,
K'(O) -
0, j -
K
° (1.47a)
L'(O)
n, J
= L'(?)
n, }
= 0' n
K(?) =
,J
K'(?)
n, J
=0 (n:::O:_ 1) (1.47b)
D' (ln+2) =
ln, ) ° 0, )
s~n+2(p2nK' (~)
)
+ R °2n L'0,)(~» (1.47c)
F' (2~ + 2)
2n, )
=
°
s2n + 2(Q2n K'
)
(~)
0, )
+ S2n
° L'0,(D») ( 1.47d)
D' (2~+2) = F'
2n, J
(2~+2)
2n, J
= 0 (1.47e)
N
L'n,J
(~) = "
~
eO, ~F' (2)
n,} O,k (1.47f)
k*j
N
K'n, (~)
)
= " (aO, ~D' (2)
L.. \ n,) 0, k + e n,0, )~F'0,(2»k (1.47g)
k*j
N
L' (2) = _ " -rD,kF' (2) (1.47h)
n, J ~ J n, J 0, k
k*j
N
K,(2)= _ " (bO,k D '(2)+d D,kF'(2»
n, ) L.. n,) 0, k n,) D. k (1.47i)
k*j
D 2n,)
'(2~+2q) ="L..q-I
2n+2P+2[p2nK'(2q.-2p-2)
S) 2p 2p,)
+ R 2p
2n L'(2 q.-2 p -2)]
2p,) (1.47j)
p=D
F '(2n+2q)
2n,)
="L..q-I
S)
~n+1P+l[QlnK'(lq.-lp-l)
2p lp,)
+ slnL'(2
lp lp,)
.-1 P -2)]
Q
(1.47k)
p=O .
q-I
' (2~+2q) = _ "
D 2n,) L..
~n+2P+2[T2nK' (2q.-2p-2)
S) 2p 2p,)
+ V 2p
2n L' (2 Q.-2 p -2)]
2p,) (1.471)
p=D
q-I
F ' (2~+2Q)
2n,]
= _"
L.. S)
2n+2 p +2[U ln K' (lQ.-2p-l)
2p 2p,)
+ WlnL' (2 Q.-2 p -2)]
2p 2p,)
(1.47m)
p=O
18 M.Isida
q-Z
D '(2n+2q)
2n+l,j
_
-
" S2n+Z p +4[p2n+lK'(2 q -Z P-4)
L..- j 2p+l 2p+l,j
+ R2n+1L'(zq-ZP-4)](147
Zp+l 2p+l,j . n)
p=O
q-2
F '(2n+Zq) _
2n+l,j -
" s2n+2 p +4[Q2n+lK'(2 q -2 P-4)
L..- j 2p+l 2p+l,j
+ s2n+lL'(2q -2 P-4)]
2p+l Zp+l,j
(147 )
• P
p=O
q-2
D2n+l,j
,(2n+2q) __
-
"
L..-
2n+2P+4[T2n+lK,(2q-2P-4)
Sj Zp+l 2p+l,j
+ v2n+lL,(2q-2P-4)]
2p+l 2p+l,j
p=O (1.47q)
q-2
F,(2n+2q) _ _ "
2n+l,j -
s2n+2P+4[U2n+lK'(2q-2P-4)
L..- j 2p+l 2p+l.j
+ W2n+lL,(2q-2p-4)]
2p+l 2p+l,j
p=O (1.47r)
K' (~q)
n, J
="L..- Zq-2
N
"[aP, ~D' (2q)
L..- n, J p, k
+ bn,P' JkD'p,(2q)
k
+ cP'n, J~F'p.(2q)
k
+ dp,kF' (2 q )]
n. J P. k ( 1.47s)
k* j p=O
N Zq-2
L'~J
(~q) = "L..- "L..- [eP' ~F' (2q) + f P' ~F' (Zq)]
~J ~k ~J ~k (1.47t)
k*jp=O
K' (2q)
n.J
="L..- "L..- [- b
N Zq-2
P' ~D' (2q)
n, J p, k
+ aP'n, J~D'P,(2q)
k
- dP' ~F' (2q)
n, J p, k
+ cP,n, J~F'P,(Zq)]
k
(1.47u)
k*i p=O
N Zq-Z
L,(~q)= "
n, J L..- L..- [-fP'~F'(2q)+eP'~F'(2q)]
"
n, J p, k n, J p, k
1l.47v)
,
k*jP=O
in whieh j = 1,2, ... , N and q = 2,3, . , .,. Taking j = 1,2, ... , N, the
expansion coeffieients in equations (1.46) ean be ealeulated sueeessively to
any desired aeeuraey.
Having expressed the complex potentials lP/z j ) (j = 1,2, ... , N) in series
form, the stress intensity faetor at the erack tip A j (Zj = A) ean be found
from equation (1.15) as
12(2d)1-(3
I2
+ v)M l'1m [(. _
ZJ
1.)1- '.( .)]
AJ lP J ZJ (1.48)
1 zj"-+lj
Putting equations (1.46) into (1.42) and using the ealeuIated resuits of F; n,j
etc., lP/(z) in equation (1.48) may be written as a power series in A:
oo
lPj(z) = L (n + 1) [-(F;,j + iF~,)zj(n+2) + (L:,j + iL~,)zj]
n=O
oo q oo
= L Jc 2q L s;P[K;~~j-2P) L (2n + I)Q~;Z;(2n+2)
q=O p=O n=O
oo
- iK'(2~-2p)
2p,) "L. (2n + I)U 2p
2nZ-:-(2n+2)
)
n=O
oo q oo
+ "L... Jc 2q+1 "L...)
s~P-1[K'(2q-2.p) " (21!
2p+1,) L. + 2)Q2n+1z:-(2n+3)
2p+1 )
q=O p=O n=O
2p+1,) {(lP + 2)Z~P+1

+ L'(2q-2p) ) + ~
L.
(2n + 2)S2n+1
2p+1 Z )-:-(2n+3)}
n=O
oo
_ iK'2p+
(2q-2.p)
1,)
"L. (21! + 2)U 2p+
2n +1Z-:-(2n+3)
1 )
n=O
(2q-2.p) {(2 P + 2)Z~p+1

- iL'2p+1,) ) ~
+ L. (2n + 2)W 2p+1
2n +1Z-:-(2n+3)}J
) (1.49)
n=O
where Zj stands for
(1.50)
The infinite series containing Q~; to W~;! ~ can be summed up in c10sed

forms with the help of equations (1.26) and the relations
~ (2n + I)A Z-(2n+2) = 2m + 1 [Z _ (Z2 _ l)t]2m+1 (1 51a)

L. n-m, 2m+ 1 (Z2 _ l)t .
n=m
~
L.
(2n + l)A n-m,2m+2 Z-(2n+2) = [(2m + 2)Z _ lJ [Z _ (Z2 _ 1)t]2m+2
(Z2 _ l)t
n=m (1.51b)
together with
~ (2n + 2)A Z-(2n+3) = 2m + 2 [Z _ (Z2 _ 1)t]2m+2

~ n-m,2m+2 (Z2 _ l)t
n-m (1. 52a)
20 M.Isida
~
~
(2 +2)A
n n-m+ 1, 2m+ 1
Z-(2n+3)= [(2m+l)Z _ 1][Z_(Z2_ 1)t]2m+l
(Z2 _ l)t
n=m (1. 52b)
whieh are easily derived from equation (1.27).

Substituting the obtained expressions of tP'/z) into equation (1.48), the
stress intensity faetor at the eraek tip A jean be given as a power series whose
eoeffieients are evaluated exaetly. Computer program has been prepared
on the basis of the above formulations. It automatiealIy generates the stress
intensity faetors at all eraek tips for any given geometrie and mechanieal
parameters.
The stress intensity faetors for the j th eraek in power series form as
follows:
(1.53a)
kl , B·J = (1B Y la.F

J BJ·, F Bj = L PnC -At
n=O
(1.53b)
F~j = L
n=O
QnAn (1.53e)
s
F~j = L QnC -At
n=O
(1.53d)
where
The faetor (1 B ~ a j is kl for an isolated eraek under bending at infinity.

Therefore, F Aj , F Bj , F~j and F~j may be regarded as the magnifieation faetors,
owing to eraek interaetion effeets.
The analysis is simplified for the speeial ease of an infinite row of periodie
eraeks. Consider an infinite row of equal paralleI eraeks distributed with a
stagger angle rt. as shown in Figure 1.4. For simplieity, eonsider only three
adjaeent eraeks. Let the distanee between two adjaeent eraek eenters be the
referenee length d. The eoordinates (Xj , Y), dimensionless eoordinates
(x j ' Yj), eomplex eoordinates and parameters as shown in Figure 1.4 are
defined as
(1. 54a)
Arbitrary array of eraeks in wide plates un der classieal bending 21
(1.54b)
in whichj = - CXJ, ••• , -1,0,1, ... , oo.
XI
Xo
M
»
Figure 1.4. Periodic paralleI eraeks
Due to the periodic nature of the problem, the complex potentials

CPk(Zk) and l/!k(Zk) must have a common form for each crack. Equations (1.36)
and (1.37) are replaced by
oo
W = w(o) + L wk ( 1.55)
k= - oo
where
(1.56a)
and
(1.56b)
22 M.Isida
Moreover, wk in equation (1.55) beeomes
DWk = Md 2 Re[zkcPk(Zk) + l/!k(Zk)] (1.57)

and
oo
cPk(Zk) = L F 2n zi:(2n+l) (1.58a)
n;O
oo
l/!k(Zk) =- D~ log Zk + L D 2n zi: 2n (1.58b)
n;O
where F2n and D 2n are generally eomplex constants. Equation (1.55) permits
eonsideration of the boundary eonditions of any one of the eraeks. The
zeroth eraek will be taken for eonvenienee. Expansion of the displaeement
funetion W around the center of this eraek yields the following expression in
terms of ZO:
(1.59)
in whieh
oo
cI>o(zo) = L (F 2nZO(2n+ I) + L2nZ~n+ I) (1.60a)
n=O
oo oo
1[I0(zo) = - D~ log Zo + L D2nZ02n + L
n;1 n;O
K2nZ~n+2 (1.60b)
and the eoeffieients L 2n and K 2n are
L 1 0 ~ 2n
( 1.61a)
2n = - 4(1 + v) J n - P~o "Y 2pF 2p
K 2n = 4(1 ~ v) J~ + Jo (a~;D2p + f3i;F 2P) (1.61b)
In equations (1.61), the following eontraetions have been made:
a 2n = __1_ S e-i(2n+2)~ (1.62a)

o 2n + 2 2n+2
2n _(2n + 2p + 1) S -i(2n+2p+2)~' > 1)

a 2p - 2n + 2 2n+2p+2 e tp = ( 1. 62b)
f3 2p
2n _
-
(2n + 2p +
2n + 2
2) S 2n+2p+2
-i(2n+2p+4)~
e (1.62e)
YZp
zn = (2n + 2p +
2n+1
1) S Zn+Zp+Z e
-i(Zn+Zp+Z)~ (1.62d)
(1.63)
The Bernoulli's numbers are B zq with the following first valu es and recurrent
formula
(1.64)
B zq = 2q ~ 1[ q - t- :t e~;
1
, 1)B zm J
For this problem, equations (1.25) pertaining to the traction free conditions
become
(1.65a)
(1.65b)
oo
D'Zn = - "1.... AZn+Zp+Z(TZnK'
Zp Zp
+ VZnL'
Zp Zp
) (1.65c)
p=O
(1.65d)
The perturbation technique can now be applied to solve equations (1.61)

and (1.65) and the stress intensity factors in terms of power series in A can
then be evaluated from equation (1.48).
1.5 Numerica1 results
Fortran computer program has been prepared on the basis of the above
formulation. The values of F Aj , F Bj , F~j and F~j can be obtained auto-
matically for loading parameters I, m, dimensionless coordinates of crack
centers (r j , 13), angles ofinc1ination (Xj and crack length ratios Sj (j = 1,2, ... ,
N). The proposed method involves neither solutions of simultaneous
equations nor numerical integrations which causes deterioration of accuracy
24 M.lsida
s
Figure 1.5. Relation between N and the upper limit of S
with increasing N and S, number of eraeks and degree of the power series
of equations (1.53). Therefore N and Sean be enlarged up to the li mit of
the core capacity of the computer without loss of accuracy. Figure 1.5
shows the relation between N and the upper limit of S for the current com-
puter system at the Kyushu University. In the present scheme of numerical
cakulations, however, S is chosen to be 74 for periodic eraeks and 43 for
other cases regardless of N such that reasonable computing cost and numeri-
eal accuracy are achieved. Discussions on the accuracy of the results are
given in section 1.6. Numerical results for some typical interaction problems
will now be treated. They have been obtained by assuming the Poisson's
ratio v to be 0.3 except for the results of eollinear eraeks which are inde-
pendent of v.
Some inelined eraeks. Figure 1.6 gives the results for a pair of inclined
eraeks where the moment veetor is perpendicular to the axis of symmetry.
Solid and dashed curves correspond to F A and F H for the inner and outer
crack tips. The magnifieation faetors of kl are defined in equations (1.53).
The general trend of the curves is similar to that in the corresponding plane
problem [1, 2] except that less interaction effect is observed for the bending
Arbitrary array of eraeks in wide plates under classiea! bending 25
+--M
1.4
F.
Fe
1.2
0.8 1-- - +
FB a "4O°
0.6 F=====i===='=-=---+=====
0_50 0
0.41======::;:::===t=======---!
0 .. 600
0.2 1 . - - -- ! - - -
al70·
a _000
I
o 02 04 oo 0.8 lO
z2..
d
Figure 1.6. A pair of inclined eraeks under bending paralleI to axis of symmetry
problem for large angles of inelination. Figure 1.7 is for the ease where the
bending veetor is parallei to the symmetrie axis, showing a small interaetion
effeet for a wide range of rx. The results of uniform bending for equallength
eraeks radially distributed at equal angles are shown in Figure 1.8. For eaeh
eonfiguration, F A for the inner eraek tip is higher than F B for the outer
eraek tip. As the number of eraeks inereases, both FA and F8 tend to deerease
for the same values of a/d used in the ealculation.
Collinear eraeks. The magnifieation faetors F for eollinear eraeks under

bending at infinity are the same as those for tension, inplane shear and anti-
plane shear loads. They are defined as the eorresponding stress intensity
factors divided by those of an isolated eraek subjected to loadings at infinity.
Before diseussing the present results, available e10sed form solutions on
26 M.Isida
M M
1 r
1.0 a -'iKf
•
F:
I a - 6O"
Fo
a-7(/'
-- ...:
0.8
a-€Cf
0.6 I Q-~
0,4
I a -4(/ f·---
F.
Q-Yf
---- ------
0. 2
a . 2(j'
--- ------
0-10·
---
o 02 04 0.6 Q8 ~ 10
d
Figure 1.7. A pair of inclined eraeks under bending vertieal to axis of symmetry
__ M
+1
1.4
~
F. M 2Q l
F.
1 B~
d
1.3
M --
1.2
FA
F.
\. \
1.°0\----~0~c2~~~~0.4t,::..---:::o.J,,6,.-----,O:l.8:---0--:!1.0
d
Figure 1.8. Radially distributed eraeks under uniform bending
eollinear eraeks will be listed. The notations in the original papers are
ehanged as that the results will be eonsistent with those in the present paper.
(1) Two eraeks [8,9] (Figure 1.9)
Figure 1.9. Two eollinear eraeks
{l.66a)
(1.66b)
in whieh
m1 = 2((1 + Al _Al~) (1 _Al + A2)Y' Al = ~1, A-~

2 - d
In the speeial ease, when al = a2 = a, equations (1.66) simplify to
FA
E
= 2A(1 1_ AY!- [ (1 + ,1) 2 K(m
(m )
2)
2
- (1 - ,1)
2J (1.67a)
F =
B
(1 +2,1,1)1- [1 _
E(m 2 )]
K(m2)
(1.67b)
and
2 ',1 2
m -...; A=~
2 - 1 + ,1' d
(2) Three symmetrie eraeks [10] (Figure 1.10)
Figure 1.10 Three symmetrie eraeks

28 M.Isida
)t
where
..12 ..1 al ..1 _ a 2
m 3 = 2( (1 + ..1 2)2 _ Ai' 1 = d' 2 - d
(3) Periodie eraeks [11] (Figure 1.11)
d .1. d
Figure 1.11. Periodie eollinear eraeks
. (2
F = 1!A tan"2 '
1!A)t 2a
..1 = d (1.69)
In the foregoing expressions, K(m) and E(m) are eomp1ete elliptie integra1s
of the first and second kinds.
No other c10sed form solutions are avai1ab1e for eollinear eraeks subjeeted
to 10adings at infinity. The present method applies to any array of eraeks,
and gives the results in power series whose eoeffieients ean be evaluated
exaetIy. As an example, the F-series for periodie eollinear eraeks shown
by Figure 1.11 is given as follows:
F(A) = 1 + 0.4112..1 2 + 0.3213..1 4 + 0.2732..1 6 + 0.2413..1 8 + 0.2183..1 10
+ 0.2007..1 12 + 0.1867..1 14 + 0.1753..1 16 + 0.1657..1 18 + 0.1576..1 20
+ 0.1505..1 22 + 0.1443..1 24 + 0.1388..1 26 + 0.1339..1 28 + 0.1295..1 30
+ 0.1255..1 32 + 0.1218..1 34 + 0.1185..1 36 + 0.1154..1 38 + 0.1125..1 40
+ 0.1098..1 42 + 0.1074..1 44 + 0.1050..1 46 + 0.1029..1 48 + 0.1008..1 50
+ 0.0989..1 52 + 0.097U 54 + 0.0953..1 56 + 0.0937..1 58 + 0.0922..1 60
+ 0.0907..1 62 + 0.0893..1 64 + 0.0879..1 66 + 0.0866..1 68 + 0.0854..1 70
+ 0.0842..1 72 (1.70)
This series coincides with the expansion of the closed form solution in
equation (1.69).
11 2 ~----
102
Figure 1.12 Variation of F with crack tip stations for equal collinear eraeks (2ald = 0.5)
In what follows, the results of equally spaced eoHinear eraeks of equal size
will be examined in some detail. The values of F tend to deerease with
increasing distance from the center of the crack array. Figure 1.12 gives the
results for N ::::; 15 and 2ald = 0.5, where the solid and dashed lines eorre-
spond to odd and even numbers of eraeks respeetively.
TABLE 1.1
FA of extreme inner tips of equal eol/inear eraeks
R 2 [9] 3 [10] 4 5 7 9 Il 13 15 13,15--+ 00 oo [II]
0.1 1.0013 1.0025 1.0028 1.0031 1.0034 1.0036 1.0037 1.0038 1.0038 1.00414 1.00414
0.2 1.0057 1.0103 1.0117 1.0129 1.0140 1.0147 1.0151 1.0154 1.0156 1.01698 1.01698
0.3 1.0138 1.0241 1.0274 1.0302 1.0329 1.0344 1.0354 1.0361 1.0366 1.03983 1.03983
0.4 1.0272 1.0453 1.0517 1.0569 1:0621 1.0650 1.0669 1.0682 1.0691 1.07532 1.07533
0.5 1.04801.0766 1.0880 1.0966 1.1056 1.1106 1.1138 1.11611.1177 1.12845 1.12838
0.6 1.0804 1.1232 1.1423 1.1560 1.1707 1.1791 1.1844 1.1881 1.1909 1.20854 1.20846
0.7 1.1333 1.1956 1.2278 1.2491 1.2735 1.2873 1.2962 1.3024 1.3069 1.33626 1.33601
0.8 1.2289 1.3214 1.3782 1.4126 1.4551 1.4793 1.4949 1.5058 1.5137 1.56553 1.56497
0.9 1.4539 1.6068 1.7218 1.7860 1.8767 1.9280 1.9579 1.9811 1.9982 2.10940 2.11331
30 M.Isida
CT
I
_M T
(;)
T"
at22 t 2aj l2l1A t 2aj saja
1 1
d--+-d d--+-d
20
FA
Fe
T~ _M
..T,
18 I
CJ"
16
FA
Fe
14
12
1.0 0 02 20 1.0
0-
Figure 1.13. F of extreme inner and outer tips of equal paralleI eraeks
Table 1.1. and the solid euryes of Figure 1.13 give FA for various N and
A = 2a/d, whieh is the maximum of F at the tip of the middle eraek. Values
for N = 2, 3 and infinity in the table eorrespond to the exact solutions of
equations (1.67), (1.68) and (1.69).
Numerical comparison of the results obtained from the series in equations
(1.53) with the corresponding exact solutions will give a measure of the
aeeuraey of the present analysis. This is given in seetion 1.6.
As shown above, FA for the same value of 2a/d tends to decrease with
increasing number of eraeks. In order to be more specific, the results in
Table l.l are plotted in Figure 1.14 by taking I/N as the abscissa. Values
for odd and even N are shown by open and solid cirdes, and those for the
same values of 2ald are connected with sol id and dashed lines respectively.
Note that for N 2': 6 both lines tend to converge to almost straight lines
representing the limiting values of N = oo i.e., a periodie array of eraeks.
To check this limit numericaUy, values of N = oo are estimated by assuming
rT
2.2 I
_M
F.
_M
1.81----+------"'.,!;- I
1. 61----t---.J>..,..,..-+---+-~~
"
........ _-..."
~-
----
----
0.3 - - - --- - _
1.°0.t==~O~I =~=O~2==:::;i;:::::=--::-=
0.3
-0::l::-=--
.4
::::;;;J
I 05
"Fr
Figure 1.14. Variation of FA with I/N for constant 2a/d
straight lines passing through the values for N = 13 and 15. The results are
given in the eolumn marked '13, I 5 -+ oo' of Table 1.1, and are shown to
be remarkably cIose to the exaet ones. Therefore uneaIculated values of
FA for N larger than 15 ean be estimated by linear interporation of values
for N = 15 and infinity with excellent aeeuraey.
TABLE 1.2
FB of exlreme ouler lips of equal eollinear eraeks
~ 2 [9] 3 [l0] 4 5 7 9 11 13 15 13, 15 --> oo
0.1 1.0012 1.0015 1.0016 1.0017 1.0018 1.0018 1.0019 1.0019 1.0019 1.0020
0.2 1.0046 1.0058 1.0064 1.0067 1.0071 1.0072 1.0074 1.0074 1.0075 1.0079
0.3 1.0102 1.0130 1.0142 1.015~ 1.0158 1.0162 1.0164 1.0166 1.0168 1.0176
0.4 1.0179 1.0230 1.0253 1.0267 1.0282 1.0290 1.0294 1.0298 1.0300 1.0316
0.5 1.0280 1.0363 1.0402 1.0425 1.0449 1.0463 1.0471 1.0477 1.0481 1.0507
0.6 1.0409 1.0538 1.0600 1.0635 1.0674 1.0695 1.0708 1.0717 1.0724 1.0766
0.7 1.0579 1.0772 1.0866 1.0921 1.0982 1.1015 1.1035 1.1049 1.1059 1.1124
0.8 1.0811 1.1103 1.1249 1.1335 1.1432 1.1484 1.1516 1.1538 1.1554 1.1658
0.9 1.1174 1.1644 1.1887 1.2035 1.2206 1.2298 1.2351 1.2390 1.2418 1.2601
32 M.Isida
u-
T
_M
1.3 , - - - . ,
F. B<=~~._c~=2~a~~t=2~a=~
d:=C:::d
___ ~7
I . I I----=='f">---==±---+_~ --+-~-=--.j
___q6
-_Q_5~::=t---~==t==-
1.00
f-~_~~~9~f§~~~~~~~~~~~g
OI t 0.2 0.3 0 .4 05
Figure 1.15. Variation of FB with I/N for eonstant la/d
Values of F B , the minimums of F occurring at the extreme outer crack

tips are given in Table 1.2 and in Figure 1.13 by dashed lines. Values for
N = 2 and 3 are the exact solutions of equations (1.67) and (1.68). F B are
shown to increase with increasing N for the same A. They are replotted in
Figure 1.15 by taking I/ N as the abscissa, and are connected with al most
straight lines. Values for N = oo, where no exact solution of F B is available,
0.94 4
f~ i" 56
F
- - - - - -- \' I Y- _9_1~_ 1
1\ / / I: I
\ I I 15
\ \\ 1\ \\
\1\\1\\ \:\\\ I rl IIII

- _ N" .
092
\ \I
\ ,
\ I
\ I
0.90
\ \ \ \\!,. ~ -J I II I I I 1/
'--- "--~ /- v
I
088 I\. '\ ~ b- L /

\ \ \ ...........
r- ./""
V V /
'-.,
~~ ~~ -
0.86
l-
- eli ." ""
l ij r
Figure 1.16. Variation of F with eraek tip stations for qeual parallei eraeks (laid = 0.9)
Arbitrary array of eraeks in wide plates un der classieal bending 33
are estimated by assuming straight lines passi ng through the values of N = 13

and 15 and are given in the last eolumn of Table 1.2 as weil as in Figure 1.15
by dashed lines. The aeeuraey of those limiting values, however, may not
be so good as that in ease of F A , sinee the linearity of the euryes seems to
be disturbed in their left parts.
Parallel eraeks equally spaeed without staggering. Values of F at all the

eraek tips are smaller than unity, and they deerease monotonieally with
inereasing distanee from the center of the eraek array. This is contrary
to the eollinear eraek problem. Figure 1.16 gives an example for N = 3 to
15 and A = 0.9. F A and F B , the minimum and maximum of F occurring at
the extreme inner and outer eraek tips are tabulated in the left eolumns of
Tables 1.3 and 1.4. Table 1.3 also gives the results for periodie parallei
eraeks, whieh are in good agreement with Nisitani's previous results [12]
show n in the last column. Values in Tables 1.3 and 1.4 are plotted in Figure
1.17 against 2ald, and also shown in Figures 1.18 and l.l9 taking II N as
the abseissa. Linearity of the euryes towards larger values of N seems exeel-
JOO
FA
Fe
098
096
_M
B
094 d
id
A
092
d a a
d
OO B
__ M
oBBO
J,O
~
d
Figure 1.17. F of extreme inner and outer tips of equal paralleI eraeks
34 M.Isida
TABLE 1.3
FA of extreme inner tips of equal parallei eraeks
3 4 5 7 9 Il 13 15 13,15 --+ IX) IX) Ref. [

0.1 0.9993 0.99860.99840.99820.99800.99800.99790.99790.9978 0.9976 0.9976 0.997«
0.2 0.9971 0.9943 0.9936 0.9929 0.9923 0.9919 0.9917 0.9915 0.9914 0.9907 0.9907 0.990'
0.3 0.9937 0.9873 0.9858 0.9842 0.9829 0.9821 0.9816 0.9812 0.9810 0.9793 0.9793 0.979'
0.4 0.9889 0.9779 0.9752 0.9726 0.9702 0.9689 0.9680 0.9674 0.9670 0.9641 0.9641 0.964'
0.5 0.9831 0.9664 0.9623 0.9583 0.9547 0.9528 0.9515 0.9506 0.9499 0.9457 0.9457 0.946:
0.6 0.9764 0.9531 0.9475 0.9420 0.9371 0.9344 0.9326 0.9314 0.9305 0.9247 0.9247
0.7 0.96900.93860.93140.92420.91790.91440.9121 0.9106 0.9094 0.9019 0.9019
0.8 0.9612 0.9232 0.9143 0.9055 0.8977 0.8934 0.8906 0.8887 0.8872 0.8780 0.8779
0.9 0.953 0.908 0.897 0.887 0.877 0.872 0.869 0.867 0.865 0.8540 0.8535
TABLE 1.4
FB of extreme outer tips of equal parallei eraeks
Ä.~ 2 3 4 5 7 9 11 13 15 13, 15 --+ IX)
0.1 0.9993 0.9991 0.9990 0.9990 0.9989 0.9989 0.9989 0.9989 0.9989 0.9988
0.2 0.9971 0.9964 0.9961 0.9959 0.9958 0.9957 0.9956 0.9955 0.9955 0.9953
0.3 0.9937 0.9921 0.9914 0.9910 0.9906 0.9904 0.9902 0.9901 0.9901 0.9896
0.4 0.9889 0.9862 0.9850 0.9844 0.9836 0.9832 0.9830 0.9828 0.9827 0.9819
0.5 0.9831 0.9790 0.9772 0.9762 0.9751 0.9745 0.9742 0.9739 0.9737 0.9726
0.6 0.9764 0.9708 0.9683 0.9669 0.9654 0.9646 0.9641 0.9637 0.9635 0.9618
0.7 0.9690 0.9617 0.9585 0.9567 0.9547 0.9536 0.9530 0.9525 0.9522 0.9501
0.8 0.96120.9521 0.9481 0.9459 0.9434 0.9421 0.9413 0.9407 0.9403 0.9377
0.9 0.953 0.942 0.938 0.935 0.932 0.930 0.929 0.929 0.928 0.925
le nt. Values for N = oo are estimated by linear extrapolation of the results

for N = 13 and IS and are given in the eolumns marked '13,15 ~ oo' of
Tables 1.3 and 1.4. The estimated values of FA eoineide with the exaet ones
al most up to four figures, and the linear relationship is verified. Unealeulated
values for N > 15 ean thus be estimated by linear interpolation or extra-
polation with excellent aeeuraey.
Other parallei eraeks. Figure 1.20 gives the values of F A for the inner
eraek tips of two equal paralleI eraeks that are staggered. The staggering
effeets are inftueneed by the ratios elf and 2ald. Corresponding results
for a periodie array of paralleI eraeks are shown in Figure 1.21. In addition,
Arbitrary array of eraeks in wide p!ates un der classiea! bending 35
OI 0 .2 03 04 # 05
Figure 1.18. Variation of FA with 11 N for constant 2ald
TABLE 1.5
Variation of FA ofinner tips of two paral/el eraeks with A and stagger angle a
60°
0.1 1.0013 1.0012 1.0008 1.0003 0.9999 0.9997 0.9995 0.9993 0.9993 0.9993 0.9993
0.2 1.0057 1.0050 1.0033 1.0012 0.9993 0.9985 0.9979 0.9973 0.9971 0.9971 0.9971
0.3 1.0138 1.0121 1.0076 1.0023 0.9978 0.9961 0.9949 0.9937 0.9934 0.9935 0.9937
0.4 1.0272 1.0231 1.0134 1.0028 0.9948 0.9921 0.9903 0.9885 0.9882 0.9885 0.9889
0.5 1.0480 1.0394 1.0204 1.0020 0.9898 0.9862 0.9838 0.9817 0.9817 0.9824 0.9831
0.6 1.0804 1.0623 1.0270 0.9986 0.9824 0.9780 0.9754 0.9735 0.9740 0.9751 0.9764
0.7 1.1333 1.0929 1.0307 0.9912 0.9723 0.9677 0.9652 0.9640 0.9652 0.9670 0.9690
0.8 1.2289 1.1288 1.0271 0.9791 0.9595 0.9553 0.9534 0.9533 0.9555 0.9583 0.9612
0.9 1.4539 1.154 1.012 0.962 0.945 0.940 0.940 0.941 0.945 0.950 0.953
36 M.Isida
1.0 0 .-----=---,---------,-- - -.------ - - - r - - - - - - ,
F.
----
--- _M
B
d
092
~ '-20 -
B
_M
0
0.1 02 0 .3 0.4
-Ir 0 .5
Figure 1.19 .Variation of FB with I/N for eonstant 2a/d
1.8
_M
FA
4tT
1.6
... a
f
A
lA
•
-----M
1.2
1.0
0 80L---...J
Q2'----04
-'----Q.l..
6---Q'-
8 --
~---'I.O
Figure 1.20. F A of inner eraek tips for two paralleI eraeks staggered
TabIes 1.5 and 1.6 give their numerieaI vaIues eaIcuIated for Jc and incIination
angIe CI. = eoe 1(e/J).
A row of paraIIeI and equaI eraeks with CI. = 45° wiII now be diseussed.
TABLE 1.6
Variation of F ofperiodie parallel eraeks with A and stagger angle a
A a O° [11] 10° 20° 30° 40° 45° 50° 60° 70° 80° 90°
0.1 1.0041 1.0037 1.0026 1.0012 0.9997 0.9991 0.9986 0.9980 0.9977 0.9976 0.9976
0.2 1.0170 1.0152 1.0106 1.0046 0.9988 0.9964 0.9944 0.9919 0.9908 0.9906 0.9907
0.3 1.0398 1.0354 1.0241 1.0098 0.9968 0.9916 0.9874 0.9820 0.9798 0.9793 0.9793
0.4 1.0753 1.0661 1.0431 1.0162 0.9932 0.9843 0.9774 0.9687 0.9651 0.9642 0.9641
0.5 1.1284 1.1101 1.0676 1.0225 0.9872 0.9743 0.9645 0.9523 0.9472 0.9458 0.9457
0.6 1.2085 1.1720 1.0965 1.0271 0.9784 0.9615 0.9488 0.9334 0.9268 0.9249 0.9247
0.7 1.3360 1.2580 1.1269 1.0283 0.9663 0.9459 0.9307 0.9124 0.9046 0.9022 0.9019
0.8 1.5650 1.3735 1.1537 1.0244 0.9510 0.9276 0.9105 0.8899 0.8811 0.8784 0.8779
0.9 2.1133 1.5065 1.1706 1.0148 0.9326 0.9070 0.8885 0.8664 0.8570 0.8541 0.8535
2.0
I.e
1.6
lA
1.2 1-- - I - - - i - ----,-PS7L'----b- =
08~--L---L---L---~-~
o 02 04 06 08 ZlL 1.0 ,
d
Figure 1.21. FA of periodic paralleI eraeks staggered
Note that FB for the extreme outer tip and F A for the extreme inner tip are
the maximum and minimum of the magnification factors. Numerical results
are plotted in Figures 1.22 to 1.24 in the same way as the preceding cases,
and the linear relationship between F and I/N is confirmed.
38 M.Isida
Figure 1.22. F of extreme inner and outer eraek tips of equal parallei eraeks loeated
staggered at an angle of 45°
... -----
0.96
.. --_....
---..--,...-
__ M
090 r----r-------t-
_M
Q8BL-__- L________ ~ _____ L_ __ _ _ _L __ _ _ _ ~
o OI 0. 2 03 04 .L 05
N
Figure 1.23. Variation of FA with I/N for eonstant 2a/d

1.00
2~'0
0.2
F. -õr
----4
-Õ5-
-õ:G
091
_--51 _M
__# 09 B
__ M
____L -____L -____
*"
~L- ~ ____ ~ __ ~
o OI 02 Q.3 04 Q5
Figure 1.24. Variation of FA with I/N for constant 2a/d
The above results pertain to the magnifieation faetors of kl' Values of k 2

for skew symmetrie deformation are small as eompared with kl in all the
examples considered. Comparative large values of k 2 have been observed for
a row of parallei eraeks with et. = 45 ° as mentioned above. Their maximums
oeeur at the extreme inner eraek tips, and the eorresponding magnifieation
faetors F~ are given in Table 1.7 for various values of N and A.
TABLE 1.7
FA of extreme inner tips of equal parallei eraeks loeared with stagger angle of 45°
A. N 2 3 4 5 6 7 9 oo
0.1 0.00080.0015 0.0017 0.0019 0.0020 0.0021 0.0022 0.0025

0.2 0.0003 0.0061 0.0069 0.0076 0.0079 0.0083 0.0086 0.0100
0.3 0.0076 0.0137 0.0155 0.0171 0.0178 0.0186 0.0194 0.0224
0.4 0.01390.0243 0.0275 0.0302 0.0316 0.0328 0.0343 0.0395
0.5 0.0220 0.0376 0.0425 0.0467 0.0488 0.0507 0.0529 0.0609
0.6 0.0315 0.0528 0.0599 0.0656 0.0687 0.0713 0.0745 0.0857
0.7 0.0416 0.0689 0.0785 0.0859 0.0900 0.0934 0.0977 0.1126
0.8 0.0511 0.0847 0.0970 0.1062 0.1114 0.1157 0.1210 0.1399
0.9 0.060 0.099 0.114 0.125 0.132 0.137 0.144 0.167
40 M.Isida
1.6 Discussions
In the present analysis the magnifieation faetors F(Jc) and F'(Jc) are given as
power series of a relative eraek length Jc, and their eoeffieients are evaluated
exaetly from c10sed form expressions. Therefore, they eorrespond to the
first terms of the Maxlaurin expansions. An example has been shown by
equations (1.69) and (1.70) for the ease of a periodie array of eollinear eraeks.
The aeeuraey of the present analysis has been established in those eases
TABLE 1.8
A pair of equal eollinear eraeks
FA FB
2a (J .53) (1.67) (J .53) ( 1.67)
d S = 43 S = 43
0.1 1.001322 1.001322 1.001196 1.001196
0.2 1.005660 1.00566(, 1.004624 1.004624
0.3 1.013831 1.013831 1.010167 1.010167
0.4 1.027171 1.027171 1.017868 1.017868
0.5 1.047960 1.047960 1.027953 1.027953
0.6 1.080404 1.080404 1.040937 1.040937
0.7 1. 133262 1. 133262 1.057865 1.057865
0.8 1.228931 1.228935 1.081066 1.081067
0.9 1.4521 1.453869 1.11726 1.117412
TABLE 1.9
Three equal eollinear eraeks
FA FB Fc
2a (1.53) Sih (1.53) Sih (1.53) Sih
-
d S = 62 (1.68) S = 62 (1.68) S = 62 (1.68)
0.1 1.002518 1.002518 1.001645 1.001645 1.001503 1.001503
0.2 1.010297 1.010297 1.007017 1.007017 1.005849 1.005849
0.3 1.024070 1.024070 1.017100 1.017100 1.012959 1.012959
0.4 1.045291 1.045291 1.033530 1.033530 1.022973 1.022973
0.5 1.076630 1.076630 1.059132 1.059132 1.036311 1.036311
0.6 1.123160 1.123160 1.099151 1.099151 1.053831 1.053831
0.7 1.195578 1.195578 1.164559 1.164559 1.077235 1.077235
0.8 1.321359 1.321359 1.283484 1.283484 1.110316 1.110316
0.9 1.6066 1.606847 1.5643 1.564536 1.164394 1.164389
Arbitrary array of eraeks in wide pfates under c!assieaf bending 41
TABLE 1.10
F of periodie eollinear eraeks
2a ( 1.53) Westergaard
d S = 74 ( 1.69)
0.1 1.004145 1.004145
0.2 1.016982 1.016982
0.3 1.039830 1.039830
0.4 1.075327 1.075327
0.5 1.128379 1.128379
0.6 1.208465 1.208465
0.7 1.336005 1.336005
0.8 1.564974 1.564974
0.9 2.11317 2.113307
where the eorresponding elosed form solutions are available. Tables 1.8 to
1.10 show the eomparison of the present results with their exaet eounterparts
for a pair of equal eollinear eraeks, three equal eollinear eraeks and a periodie
array of collinear eraeks given by equation (1.67), a speeial ease of equation
(1.68) and equation (1.69), respeetively.
In Table 1.8, A and B denote the inner and outer eraek tips, and in
Table 1.9 A, B, e represent the inner, middle and outer eraek tips respeetively.
The agreement of the present results with the exaet solutions are quite
remarkable for A ;;; 0.8.
The souree of errors in the present analysis is the truneation of the infinite
series, but no theoretieal means of estimating the upper bounds of the errors
is available at the present time. However, aeeording to the author's experience
in various problems treated in this paper as weil as those in previous works
based on the perturbation technique, it appears reasonable to estimate the
upper bounds of the numerieal errors by assuming so me geometie series for
the unealculated tcrms of higher orders. Another way is to eheek the partial
sums of the obtained series in equations (1.53). An example for the peri-
odie eollinear eraeks is given in Table 1.11, showing that S may be taken
as 10, 20 and 30 in order to get reasonable results for A = 0.7, 0.8 and
0.9, respeetively. The numerieal results in the preeeding seetion have been
obtained by assuming S = 74 for periodie eraeks and S = 43 for the other
cases, and they are considered to be suffieiently aeeurate when A = 0.95.
For general arrays of eraeks, the proposed method appears to give aeeu-
rate resuits as long as ). is less than and not too elose to unity, where I. is
42 M.Isida
TABLE l.l1
Partial sums of series (1,53) for F of periodie eollinear eraeks
S 10 20 30 40 50 60 70 Eq. (1.69)
Ä
0.1 1.00414 1.00414 1.00414 1.00414 1.00414 1.00414 1.00414 1.00414
0.2 1.01698 1.01698 1.01698 1.01698 1.01698 1.01698 1.01698 1.01698
0.3 1.03983 1.03983 1.03983 1.03983 1.03983 1.03983 1.03983 1.03983
OA 1.07532 1.07533 1.07533 1.07533 1.07533 1.07533 1.07533 1.07533
0.5 1.12832 l.l2838 1.12838 1.12838 1.12838 l.l2838 l.l2838 1.12838
0.6 1.2078 1.20846 1.20846 1.20846 1.20846 1.20846 1.20846 1.20846
0.7 1.331 1.3359 1.33600 1.33601 1.33601 1.33601 1.33601 1.33601
0.8 1.53 1.562 1.5647 1.56495 1.56497 1.56497 1.56497 1.56497
0.9 2.04 2.093 2.107 2.1113 2.1127 2.1131 2.11331
the relative eraek length defined as the maximum value of (aj+a k)/ Pjk. This
limitation does not prevail in the ease of eollinear eraeks. For parallei eraeks,
however, }. may take any large values, and the present analysis no longer
applies. The diffieulty may be resolved by directly solving the simultaneous
equations instead of the perturbation preeedure.
References
[1] Isida, M., Analysis of Stress lntensity Factors for Plates Containing Random Array
of Cracks, Bulletin of the JSME, 13,59, pp. 635-642 (1970).
[2] Isida, M., Method of Laurent Series Expansion for Internai Crack Probiems, Me-
ehanies of Fraeture I, edited by G. C. Sih, Noordhoff International Publishing,
Leiden, pp. 56-130 (1973).
[3] Moriguti, S., Plane Problems of Elasticity (in Japanese), Iwanami-koza, B7-a,
Iwanami-shoten (1957).
[4] Williams, M. L., The Bending Stress Distribution at the Base of a Stationary Crack,
Journal of Applied Meellanies, 28, Trans. ASME, E, 82, pp. 78-82 (1961).
[5] Sih, G. c., Paris, P. C. and Erdogan, F., Crack-Tip Stress-Intensity Factors for
Plane Extension and Plate Bending ProbIems, Journal of Applied Mechanies, 29,
Trans. ASME, E, 83, pp. 306--312 (1962).
[6]lsida, M., On the Stress Function in the Plane Problems ofan Elastic Body Containing
a Free Elliptic Hole, Transactions of the Japan Society of Meellanieal Engineers, 21,
107, pp. 502-506 (1955).
[7] Hayashi, T., Transactions of tlle Japan Society of Meellanieal Engineers, 25, 159,
p. 1133, Ref. (8) (1959).
[8] Yokobori, T., Ohashi, M. and Ichikawa, M., The Interaction of Two coHinear
Asymmetrical Elastic Cracks, Reports of the Research Institute for Strength and
Fraeture of Materials, Tohoku University, 1,2, pp. 33-39 (1965).
Arbitrary array of eraeks in wide plates un der dassical bending 43
[9] Erdogan, F., On the stress distribution in plates with eollinear euts under arbitrary
loads, Proceedings 0/ the 4th V.S. National Congress 0/ Applied Mechanies, Berkeley,
1, pp. 547-553 (1962).
[10] Sih, G. C., Boundary Problems for I ongitudinal Shear Craeks, Proceedings 0/ the
Second Conferenee on Theoretieal and Applied Mechanies, Pergamon Press, pp.
117-130 (1964).
[11] Westergaard, H. M., Bearing Pressure and Craeks, Journal 0/ Applied Meellanies,
Trans. ASME, pp. A49-53 (1939).
[12] Nisitani, H., Journal 0/ the Japan Society 0/ Mechanieal Engineers, 71, 589, p. 209
(1968).
R. J. Hartranft
2 Improved approximate theories of

the bending and extension of
flat plates
2.1 Introduction
The analysis of thin plates for stresses and deformations is a problem

whose technical importance has grown for a century. Design refinements
and new applications to ships, planes, and spacecraft require improved
methods of analysis. As design specifications call for materials of higher
strength, the toleranee of structures to flaws is reduced. Thus the design
process must include a fracture mechanics analysis. But, although classical
plate theories are quite appropriate for ordinary problems without stress
singularities, more accurate theories are necessary to adequately model the
behavior of a plate near a crack tip.
Linear elasticity formulation. Although material and kinematical nonlinea-

rities make caution necessary, the linear theory of elasticity would be expected
to provide significant results. The particular problem of interest is that of an
infinite plate containing a crack and loaded uniformly at infinity. The
equations of elasticity are to be solved subject to boundary conditions,
which may be written with reference to Figure 2.1 as
I y I -+ oo: ryx = ryz = 0, (2.la)
I x I -+ oo: (J x = r xy = r xz = ° (2.1 b)
=°
h
I z I = 2": r zx = rzy = (Jz (2.lc)
y = 0, I x I < a: ryx = (Jy = ryz = ° (2.ld)

46 R. J. Hartranft
If the symmetry of the problem is exploited, and superposition is used, the

solution of the problem of Figure 2.1 is reduced to the solution of one with
the boundary conditions
I
/.~
Figure 2.1. Uniformly loaded, cracked, infinite plate
(2.2a)
x ~ OO: (1 x = Txy = Txz = 0 (2.2b)
x=O: Ux=Txy=Txz=O (2.2c)
y = 0: {
X < a: T yx = T yz ~ 0, ~(1y - (10 -
12
7J3 Moz (2.2d)
a < x. T yx - T yz - 0, u y - 0
The problem of Figure 2.1 is solved by adding the solution above to
Ux =- i x[(10 + !; MoZJ (2.3a)
Uy = 1 Y[(10 + ~; MoZJ (2.3b)
6M o
Uz = - Ev Z(1o + Eh 3 [v(x -
2 2
Z ) -
2
y ] (2.3c)
(2.3d)
(2.3e)
where E and vare Y oung's modulus and Poisson's ratio for the material.
Improved approximate theories of bending and extension of flat plates 47
In equations (2.1) through (2.3), and in subsequant work, ao represents

the constant stress at infinity for the extension of the plate and Mo is the
uniform bending couple at infinity in the plate bending problem. Equations
(2.3) comprise the linear elastic solution of the problem of Figure 2.1 when
no crack is present.
Approximations of linear elasticity. There are several ways of obtaining

approximations of the solutions ofplate probiems. There is the obvious resort
to the numerieal techniques associated with finite difference or finite element
methods. This approach may be characterized as seeking approximations
of the solution of the exact equations. This chapter is concerned with the
alternative approaches which are characterized as seeking exact solutions
of approximating equations. Most of these approaches involve manipulations
of the transverse variable, z, to obtain aset of equations in the remaining
two in-plane variabies.
The theory of generalized plane stress is an example of the technique of
averaging the equations of elasticity with respeet to the transverse variations.
The averages of the stresses, 'x=> 'yo, and a z , are the n neglected in the resuIt-
ing equations. Extensions of this method [1-3] use averages weighted by
various powers of z. These weighted averages, or moments, are specified
by aset of partial differential equations in x and y.
Equations in two-dimensions for the coefficients of series expansions of
the solution have also been obtained. In [4], expansions in both power
series and trigonometric series in the variable z are used. As indicated in [I],
the accuracy of the solution cannot be readily estimated in terms of powers
of the plate thickness, h. Expansions in powers of plate thickness have al so
been used [5, 6]. These lead to boundary layer analyses to compensate for
the fact that the boundary conditions of the problems can only be satisfied
approximately.
Variational methods have been applied as a numerieal technique for

obtaining approximate solutions and as a general technique for obtaining
approximate equations. Reissner's theory of plate bending has been derived
from the theorem of minimum complementary potential energy [7] and
from a theorem which permits both stress and displacement to be vari ed [8].
A modified theory of generalized plane stress has al so been obtained by
varjational methods [9]. The next section contains a more complete discus-
sion ofvariational methods.
48 R. J. Hartranft
2.2 Approximate theories by variational methods
There are three major variational principles available for obtaining approxi-
mate theories. In the first, a dass of displacement functions having a suitable
form is examined for the one which minimizes the potential energy. If the
dass is general enough, the exact solution may be found in this way. More
often, however, the dass of functions examined is restricted, and the mini-
mizing function is an approximation to the exact solution. The fewer the
restrictions on the dass examined, the better the approximation, and the
doser to the actual minimum value of the potential energy one arrives.
The techniques used to obtain approximate theories lead to differential
equations which the minimizing functions must satisfy. The equations for
this first principle may be regarded as approximate equilibrium equations.
The stress-displacement equations are satisfied exactly.
The second principle requires a search for the stress state, of all those
which satisfy the equilibrium equations, which minimizes the complementary
potential energy. Again, restrictions on the dass of functions considered
leads to aset of differential equations for an approximate stress field. These
equations are approximations of the stress-displacement relationships.
In the third principle [8], displaeements and stresses, without regard to
equilibrium, are examined for those which minimize a eertain functional
related to the strain energy of the body. The equations obtained approximate
the elasticity equations of equilibrium and the stress-displacement equations.
These same three principles form the basis for the stiffness, force, and
hybrid finite element methods. Whatever the application, the choice of
principle depends on what are considered to be the important features of
the problem. If a continuous displacement field is of primary interest,
minimum potential energy is appropriate. If local equilibrium is critical,
minimum complementary potential energy is indicated. The third principle
[8] would be chosen to enable a particular form to be used for both stress
and displacement. It is possible, in addition, to incorporate other features if
they can be formulated as constraints to be induded by the method of
Lagrange multi pliers [10].
Reissner's varitional principle. The third principle discussed above is usually

named after its originator [8]. In the following statement of the principle,
certain combinations of derivatives of displacement could be expressed in
Improved approximate theories of bending and extension of fiat plates 49
terms of the strains, but for clarity of thought, strain is not explieitly used.
Consider the expression,
R = Jr [Tx ' u. x + Ty ' U,y + Tz ' u,z - 2V - (T~,x + Ty,y + Tz,z

+ 2E) . uJd, - JE T~ . ud1: + hu Tn' u*d1:
T (2.4)*
The eomponents of displaeement are
(2.5)
while the stress eomponents are the eomponents of the three stress veetors,
(2.6a)
(2.6b)
(2.6e)
If n is the outward normal to the surfaee,
(2.7)
is the foree per unit area aeti ng on the surfaee. The volume oeeupied by the
body is denoted by , and its surfaee by 1:. The two parts of the surfaee,
1: T and 1: u , are those on whieh traetions, T~, and displaeements, u*, are
speeified, respeetively. The strain energy density for a linear elastie, isotropie,
homogeneous material is
(2.8)
For other materials, the expression for V should be the funetion of the stress
eomponents whose derivatives equal the strains. Finally, F is the body foree
per unit volume.
The principle states that R (equation (2.4» is an extremum** for a partic-
* Differentiation of a quantity is indicated by a eomma foJlowed by the variables with

respeet to which the derivative is to be taken.
This expression differs from that in Reissner's original paper. This has the advantage
that the treatment of boundary eonditions on displaeement is similar to that on traetions.
** An extremum eould be a minimum, but it indudes any state in which a small, first-order
change of stress and displaeement produces only a second-order ehange in R.
50 R. J. Hartranft
ular stress and displacement fieId if, and only if, the stresses and displace-
ments satisfy the equations of elasticity,
(2.9)
and
Eux,x = rr x - v(rr y + rr z), E(ux,y + uy,x) = 2(1 + v),r xy

Euy,y = rr y - v(rr x + rr z)' E(ux,z + uz,x) = 2(1 + v)r xz (2.10)
Euz,z = rr z - v(rr x + rr y), E(uy,z + UZ,y) = 2(1 + v)r yz

and the boundary conditions, u = u* on III and TlJ = T* on IT'
This is the principle which wiII be used in this chapter, for it permits one
to discuss both stresses and displacements directly. The others involve aver-
ages of one or the other, and, although the results can be equivalent, the
physical meaning is not clearo For example, in [9], the principle of minimum
complementary energy was used, and weighted averages of the displacements
were introduced. The discussion in this chapter results in the same set of
equations, but with explicit expressions for the displacements.
Reissner's theory for plate bending. The first of the improved theories was
published in the 1940's. It is restricted to plate bending. One considers the
class of displacements and stresses given by
(2.11)
rr z = - ~ [2 + 3(2zjh) - (2zjh?J
where !3x, !3y, w, Mx, My, M xy , Vx' Vyand the pressurep on thetop (z = hj2) of
the plate are functions of x and y, only. Since the z dependence is explicit,
the integration with respect to z can be performed, and the expressian (2.4)
for R becomes, for F = 0,
Improved approximate theades of bending and extension of.ffat plates 51
+ Vx(f3x + w,x) + Vy(f3y + w, y) + pw

- (Mx, x + Mxy,y - VJf3x - (Mxy,x + My,y - Vy)f3y]
- (1 _ 1 v2 )D SA dA [M
2
x +M 2y -2vMxM y +2(1 +v)M 2xy (2.12)
+ 2(1 + V)o:2(v~ + V;) + 2vo: 2p(M x + My) + Ho: 4p2]

- SST ds[ M~f3n + M~sf3s + V~w]
-k ds[Mnf3~ + Mnsf3~ + Vnw*]
where
(2.13)
The terms in the line integral over the portion, ST, of the edge on which
stresses are specified are given by
M*n = f hl2
a*zdz
n
(2.l4a)
- hl2
J
hl 2
Mn~ = '~szdz (2.14b)
f
- hl2
hl 2
V*n =
- h/2
,* dz
nz (2. 14e)
and
f3n = f3x eos eP + f3y sin eP (2.15a)
f3s = - f3x sin eP + f3y eos eP (2.15b)
The stress components, (J~, ,~S' and '~z, are shown in Figure 2.2.
Figure 2.2. Stresses and displacements on the edge of a flat plate

52 R. J. Hartranft
In the line integral over the portion, Su, where displaeements are speeified,
Mn = -!(M x + My) + -!(M x - My) eos 2ep + M xy sin 2ep (2.16a)

M ns = - -!(M x - My) sin 2ep + M xy eos 2ep (2.16b)
Vn = Vx eos ep + Vy sin ep (2.16e)

and
12 fh/2
{3~ = -h
3 u~zdz, (2.17a)
- h/2
12fh/2
{3~ = p u:zdz (2.17b)
1 - h/2
3 fh/2
w* = -21 u~[1 - (2z/h)2Jdz (2.17e)
1 - h/2
The displaeements, u~, u~, and u~, are shown in Figure 2.2.
A notable feature whieh reeurrs in plate theories is the faet that some
details of the variation of the boundary stresses and displaeements disappear.
In this ease, only the integrals (2.14) and (2.17) remain, and any z variation
of (J~, T~S' and T~z or u~, u~, and u~ is absent from R. As aresult, the theory
whieh foIlows assigns the same solution to two 10adings whieh both have
the same weighted averages as given by equations (2.14) and (2.17). The
eonneetion between the weighted averages and the unknown funetions in
equations (2.11) is given by the natural boundary eonditions of the theory.
Consider, for example, the variation [10] of the first two terms in the
expression (2.12) for R. The variations of the derivatives are eliminated as
follows by applieation of the divergenee theorem:
b SA [M x{3x.x SA [{3x,x bM x + {3y,ybMy + M xb{3x,x
+ M y{3y,yJdA =
+ M yb{3y,yJdA = SA [{3x,xbMx + {3y,ybMy - M x,xb{3x - M y,yb{3y

+ (M xb{3x),x + (M yb{3y),yJdA = SA [{3x,xbMx + {3y,ybMy - M x,xb{3x
- M y,yb{3yJdA + Ss [M xb{3x eos ep + M yb{3y sin epJds
Through similar proeedures for the variations of the other derivatives, and
by applying equations (2.15) and (2.16), the variation of equation (2.12) may
be written in the form,
+ c5My[{3y,y - (My - vM x + va 2 p)/O - v2 )D]

+ c5Mxy [{3x,y + {3y,X - 2M xy /(1 - v)D]
+ c5Vx[{3x + w,x - 2a 2 Vx/(1- v)D]
+ c5 Vy[{3y + W,y - 2a 2 Vy/O - v)D]
+ c5w[p - Vx, x - Vy, y]
+ c5{3x[Vx - Mx, x - Mxy,y]
+ c5{3y[Vy - M xy, x - My,y]}
+ JST ds{[(M n - M~)c5{3n + (M ns - M~,)c5{3s + (Vn - V~)c5w]
-[{3nc5Mn + {3sc5Mns + wc5Vn]}

-k ds{[({3n - {3~)c5Mn + ({3s - {3~)c5Mns + (w - w*)c5Vn]
- [M nc5{3n + M ns c5{3s + Vnc5w]} (2,18)
Because the eight variations, c5Mx, ... c5{3y are arbitrary in A, their coeffi-
cients in equation (2.18) must be separately zero. Thus the basic set of eight
first order, linear, partial differential equations are obtained. The boundary
integrals in (2.18) will be zero if
(2.19a)
{3n = {3~, (3s = f3~, w = w* on Su (2.19b)

for then
(2.20a)
(2.20b)
Generalization of Reissner's theory. If, in the previous discussion, the

assumed stresses and displacements of equations (2.11) are replaced by
(2.21)
and
(Jx = Mxf"(z), (Jy = MJ"(z), r xy = MxJ"(z) (2.22a)

r xz =- Vxf'(z), r yz =- VJ'(z), (Jz = pf(z) (2.22b)
54 R. J. Hartranft
where
!'(± h/2) = 0, f(h/2) = -1, f( -h/2) = °

the differentia1 equations and boundary conditions imp1ied by equation
(2.18) are reproduced. The meaning of some terms is changed. In equation
(2.12), D and arepresent
Eh 3 ~ a 2 = ~!2 (2.23)
D = 12(1 - v2 ) Il' 10 Il
where
h 3 fh/2 5hfh/2
Il = 12 U"(z)]2dz, 12 = 6" [f'(z)]2dz
-h/2 -h/2
The quantities invo1ved in the traction boundary co"ditions are again given
by equations (2.14). But the disp1acement integra1s which must be given on
the portion, Su, of the edge are altered from equations (2.17). They are
fJ: =fh/2 u:f"(z)dz, f3~ =fh/2 u:f"(z)dz,
f
-h/2 -h/2
h/2
w* = - u:!,(z)dz (2.24)
- h/2
It may be noted that if the specified stresses are of the form
a~ = Mnf"(z), Tn~ = Mnsf"(z), Tn~ = - v,,!'(z) (2.25)

the n equations (2.14) verify the expectation that
(2.26)
However, it is not necessary that the boundary stresses have the transverse
variation of equations (2.25). As pointed out in the discussion of Reissner'&
theory, two sets ofboundary stresses with the same weighted averages defined
by equations (2.14) will1ead to the same solution.
Except for notation this theory is the same as that proposed in [Il, 12].
In [11], except for so me minor differences in notation and some serious
misprints, the equations of [11, §2] are equiva1ent to the above. The symbo1s
A and B = e in [Il] are denoted by 12I1 /h 3 and 1.212 /h, respectively in this
artide. The equations in [12], after an inversion of the coordinate system,
are identical to those of this artide if f3(Z) = f(z) and kl = 12It/h\
k 2 = 1.212 /h. Reissner's theory is obtained if
Extended theories of generalized plane stress. The variationaI prineiples

ean be used to obtain two-dimensionaI, approximate theories for in-plane
deformation of plates. The equations of generalized plane stress are usually
thought of in terms of averages of equations (2.9) and (2.10). However, as
will be seen, the variationaI method leads to the same equations. The
extended theory in [9], as weIl as a eombination diseussed in [13], are also
presented from a new point of view in this seetion.
The forms chosen for approximating the displaeements and stresses, in
terms of an arbitrary funetion,j(z), are
11 3
Ux = Vx + Uxf"(z) (2.28a)
121 3
11 3
Uy = Vy + - - Uyj"(z) (2.28b)
121 3
(2.28e)
and
(lx = Sx + Sxf"(z), 'xz = - Zxf'(z)
(Jy = Sy + Syj"(z), 'yz = - Zyj'(z) (2.29)
'xy = t xy + TxJ"(z), (Jz = ZJ(z)
where f(± h/2) = !'(± h/2) = O. The two sets of unknowns, vx ' vy' sx' Sy'
t xy and Ux, Uy, UZ' Sx, Sy, Txy, Zx' Zy, Zz are funetions only of the in-plane
variabIes, x and y. If the second set are chosen to be zero, equations (2.28)
and (2.29) reduee to the form appropriate for generalized plane stress. Jf
the first set are zero, the form whieh results in the theory of [9] remains.
Exeept for rlOtation, equations (2.28) and (2.29) are those proposed in [13].
When equations (2.28) and (2.29) are substituted into equation (2.4), and
the integrations aeross the thiekness performed, R is found to be
R = JA dA{h[sxvx,x + SyVy,y + txY(vx,y + vy,J] + SxUx, x + SyUy,y

+ Txy(Ux,y + Uy,x) + ZxCUx + Uz,J + Z/U y + Uz,y)+ ZzU z}
56 R. J. Hartranft
~ SA dA{s; + S; - 2vs xsy + 2(1 + V)t;y + 2ve 2 Z z (sX + Sy)}
(1 _ 1v2 )D SA dA {Sx
2
+ Sy2 - 2vS xSy + 2(1 + v)Txy
2
+ 2wx 2 Z z (SX + Sy) + 2(1 + v)a 2(Z; + Z;) + (P2 + l)a 4 Z;}
- SA dA{h[(,sx,x + txy,y)Ux + (tXY,X + Sy,y)Uy] + (Sx, x + TXY,y - Zx)Ux
+ (Txy,x + Sy,y - Zy)Uy + (Zx, x + Zy,y - Zz)U Z}
- SST ds{h[s~un + t~sVS] + S~Un + T~.uS + Z~UZ}

+k ds{h[snv~ + tnsU~] + SnU~ + T".u~ + ZnU:} (2.30)
where
Eh 3 1 N 2 _
,U<. -
D = 12(1 _ v2 ) 13
(2.31)
where
2 fh/2 hfh/2
Il = -, [j(z)Ydz, 1 2 = -2 [f'(z)]2dz
1 -h/2 -h/2
h 3 fh/2 1 fh/2
13 = 12 [j"(z)Ydz, 14 = -h j(z)dz
-h/2 -h/2
The specified quantities in the boundary integrals of equation (2.30) are
s: = 1 -J
1
fh/2
-h/2
a:dz, t: = -h1
s
fh/2
-h/2
r:sdz
S: = 12/h
3 fh/2
a:'f"(z)dz (2.32a)
3 - h/2
T: s =
h3
121
fh/2
r:.f"(z)dz (2.32b)
f
3 - h/2
h/2
z: = - 2hj
2 -h/2
r:zf'(z)dz (2.32c)
on ST and
1 jh/2 1 fh/2
un*-
- -, *
undz, Us*-
- -h Us* dz
1 -h/2 -h/2
V n*= j h/2
-h/2
u:f "(z)dz, Vi =
fh/2
-h/2
uif"(z)dz (2.33)
V: = _Jh/2 u:f'(z)dz,
- h/2
on Su' See Figure 2.2 for the stresses and displacements on the boundary.
The remaining terms in the boundary integrals are related to the unknowns
in equations (2.28) and (2.29) by
Un = U~ cos ep + uy sin ep, Us = - Ux sin ep + vy cos ep
(2.34)
V n = Vxcosep+ Vysinep, V S = - Vxsinep+ Vycosep
and
Sn = ±(sx + Sy) + t(sx - Sy) cos 2ep + txy sin 2ep (2.35a)
tns = - t(sx - Sy) sin 2ep + t xy cos 2ep (2.35b)
Sn = t(Sx + Sy) + t(Sx - Sy) cos 2ep + Txy sin 2ep (2.35c)
Tns = - t(Sx - Sy) sin 2ep + Txy cos 2ep (2.35d)
Zn = Zx cos ep + Zy sin ep (2.35e)
Application of the usual variational methods to equation (2.30) gives
bR = 2 SA dA{hbsx[vx,x - (sx - vsy + ve 2 Z z)/EJ
+ hbsy[vy,y - (Sy - vS x + ve 2 Z z)/EJ
+ hbtxy[Vx,y + vy,x - 2(1 + v)txy/EJ

- hbVx[Sx, x + txy, yJ - hbVy[txy, x + Sy, yJ
+ bSx[Vx,x - (Sx - vSy + va 2 Z z )/{l - v2 )DJ

+ bSy[Vy,y - (Sy - vS x + va 2 Z z)/(1- v2 )D]
+ bTxy[Vx,y + Vy,x- 2Txy /(1- v)D]
+ bZx[V x + Vz,x - 2a 1 ZxI(1 - v)D] + bZy[Vy + Vz,y - 2a 1 Z y/(l- v)D]

+ bZz[V z - ve 2h(sx + sy)/E - a 2{(f31 + 1)a 2Z z + v(Sx + Sy)}/(l - v2)D]
+ bVx[Zx - Sx,x - Txy. y] + bVy[Zy - Txy, x - Sy,y]
58 R. J. Hartranft
+ e)Uz[Zz - zx,x - Zy, y]}
+ SSr ds{h[(sn - s~)bvn + (tns - t~s)bvs - (vnbs n + vsbtns )]

+ (Sn - S~)bUn + (Tns - T~s)bUs + (Zn - Z~)bUz - (UnbS n + UsbTns
+ UzbZn)} - ss.. ds{h[(vn - v~)bsn + (v s - v~)btns - (snbvn
+ tnsbv.)] + (Un - U~)bSn + (Us - U~)bTns + (U z - U~) bZn
(2.36 )
The variation bR in equation (2.36) must be zero for arbitrary variations

bsx, etc. Therefore in the area integral, each term in square brackets must
be zero everywhere in A. Thus aset of fourteen equations for the unknown
variables are obtained. One is an algebraic relationship, and the remaining
thirteen are linear, first-order, partial differential equations. The equations
of [13] are obtained if the displacement terms are eliminated and a stress
function is introduced for sx' Sy' and txy-
As was noted in [9] for the case in which Sx = Sy = t xy = V x = vy = 0,
replacing f(z) by a muItiple of f(z) does not lead to new stresses and dis-
placements in equations (2.28) and (2.29). Therefore, f(z) can be muItiplied
by a constant chosen so that 13 = 1 and 14 < O. For example, the function
suggested in [13] would be normalized to the form
j(z) = - 1 [1 - (2T)2J2
16 J15 (2.37)
and give
1 1 .J15
Il = 2f' 12 = "7' 13 = 1, 14 = - 30
(2.38)
Eh 3 2 h2 2 5 2 JJ2
D = 12(1 _ v2)' rl. = 42' fi =2' e = 30
In this connection, note that there has been a normalization in equations
(2.22) makingf(hI2) = -1. No additional requirement can be imposed.
The first five equations resuIting from equating bR (equation (2.36)) to
zero are nearly identical to those of generalized plane stress.
Ev x • x = Sx - vS y + ve 2 Z z (2.39a)
Evy, y = Sy - VS x + ve 2Z z (2.39b)
E(v x , y + vy, J = 2(1 + v)t XY (2.39c)

Sx, x + t Xl' . l' = ° (2.39d)
txl',x + Sl',y = °
The terms involving Zz are the only ones coupling these equations to the
remainder. The associated boundary conditions will require
(2.40a)
Vn = v~, V s = v~ on Su (2.40b)
Whatever the actual transverse variation of O'~, "[~s' u~, and u~, only the
weighted averages defined by equations (2.32) and (2.33) affect the solution
of the problem.
An algebraic equation also connects the two sets of unknowns.
h4
(1 - v2 )DUz = C(2[(fJ2 + 1)C(2Zz + v(Sx + Sy)] + 12/ ve 2 (sx + sl') (2.41)
3
If the initial assumptions of equations (2.29) had used Zz = 0, it would not

be necessary to satisfy equation (2.41). In such a case the two sets of un-
knowns are completely decoupIed. There would also be decoupling if
Poisson's ratio were zero (v=O), and the remains of equation (2.4i) would
be included with the equations for the second set of inknowns.
Ifthe initial assumptions included Vx = vy = sx = Sy = t XY = 0, equations
(2.39) would not be necessary. If Ux = Ul' = Uz = Sx = Sy = T XY = Zx =
= Zy = Zz = 0, then equation (2.4l) and the equations (2.42) in the next
paragraph would not have to be satisfied. In the general case both sets are
required.
The remaining differential equations are
(1 - v2 )DUx ,x = Sx - vSy + vC( 2Z z (2.42a)
(1 - v2 )DUy,y = Sy - vS x + vC( 2Z z (2.42b)
(1 - v2 )D(U x,y + Uy,x) = 2(1 + v)Txy (2.42c)
(1 - v)D(U x + Uz,x) = 2C( 2 Zx (2.42d)
(1 - v)D(Uy + Uz,y) = 2C( 2 Z y (2.42e)

(2.42f)
(2.42g)
(2.42h)
60 R. J. Hartranft
The boundary conditions for these variables involve the stress and displace-
ment averages defined in equations (2.32) and (2.33).
(2.43a)
(2.43b)
For those cases in which it is convenient to use a function of a dimension-

less variable to describe the transverse variations, the following formulas
are listed:
j(z) =j(~') = g(O = g(t z)
j'(z) = ~ g'(O,j"(z) = :2 g"(O
Il = f~l [g(O]ZdC /2 = f~l [g'(O]2d' (2.44)
/3 4
= t L1[g"(Or dc 1 = 1- f~l gmd'
The Hartranft-Sih theory of [9] consists of equations (2.42) and (2.41) with
V x = vy = Sx = Sy = t XY = 0 in equations (2.28) and (2.29).
2.3 Applications to crack problem s
The quantities of most importance in linear elastic fracture mechanics are

the stress intensity factors, which appear in the crack front stress field
singularity. Most work to date [14, 15] indicates that, except for the variation
of the stress intensity factors along the crack front, the form of the stress
field near the front is always the same. The onlyexceptions occur when
classical plate or shell theories involving a Kirchoff type boundary condition
are used [16, 17].
The improved theories considered in this chapter all result in some version
of the singular behavior,
4Gu x = kf(2r)t eos ~ [ /( - 1 + 2 sin 2 ~J (l.45a)
+ k~(2r)t sin ~ [/( + 1 + 2 cos 2 ~J + O(r)

Improved approximate theories of bending and extension of.flat plates 61
4Guy = kf(2r)t sin ~[K + 1- 2 eos 2 ~J
- kg(2r}! eos ~[K - 1 - 2 sin 2 ~J + O(r) (2.45b)
Gu z = k~(2r)t sin ~ + O(r) (2.45e)
for the displaeements, and
(lz = [ ki 0 kI.
2v (2r)! eos:2 - (2r)+ sm:2
0J + 0(1) (2.46e)
'r xz = -
k3 •
(2r)+ sm 2
e + 0(1. ) (2.46e)
k3 0
'ryz = (2r)"1 eos :2 + 0(1) (2.46f)
for the stresses. Figure 2.3 shows the eoordinate system and polar eoordi-
nates, r and e, at the eraek front. The xz-plane is the pJane of the eraek,
Figure 2.3. Coordinate system near the crack front

62 R. J. Hartranft
and the xy-plane is normal to the eraek front. For the problems of this
ehapter, the z-axis eoineides with the straight eraek front.
In equations (2.45) fot the displaeements, 2G = EI(l + v), and K = 3 - 4v
is the parameter whieh applies when the state near the eraek front is one of
plane strain. For generalized plane stress, K = (3 - v) I (1 + v). Aeeording to
[14], an exaet solution ineorporates K = 3-4v and kf = k i (i = 1,2,3). In
addition, kj = k i (i = 1,2), whieh implies that (Jz = v«(Jx+(Jy) as r-+O. The
various approximate theories fail to possess one or more of these eharac-
teristies.
Reissner's theory for plate bending. The sine or cosine transforms, as

appropriate for a problem symmetrie about the yz plane, are used to eonvert
the equations implied by expression (2.18) into ordinary differential equations
in the variable y [18]. These are satisfied by solutions of exponential form,
and when the inverse transforms are applied, an integral form of the solution
is obtained. Further applieation of the boundary eonditions,
(2.47)
leads to an integral representation in terms of one unknown function, A(s).
Thus, as in [19,20], for y ~ 0,
(1 - v)Dw = f CO
o s
1
2" A(s) {(1 - v)sy - (1 + v)}e- SY eos (sx)ds (2.48a)
(1 - v)DPx = fo
CO 1
- A(s) {[(1 - v)sy - (1 + v) - 4a 2s2]e- sy
s
+ 4a2s2me-msy} sin (sx)ds (2.48b)
(1 - v)DPy = f~ ! A(s) {[(1 - v)sy - 2 - 4a 2s2]e- sy
+ 4a2s2e-msy} eos (sx)ds (2.48e)

and
Mx = - f~ A(s) {[(1 - v) (1 - sy) + 4a 2s2]e- sy

- 4a2s2me-msy} eos (sx)ds (2.49a)
My = f~ A(s) {[3 + v - (1 - v)sy + 4a 2s2]e- sy
- 4a2s2me-mSY} eos (sx)ds (2.49b)
M xy = f~ A(s) {[2 - (1 - v)sy + 40: 2s2]e- sy

+ (2 + 40:2s2)e-msy} sin (sx)ds (2.4ge)
Vx = 2 f~ sA (s) [me- msy - e-SY] sin (sx)ds (2.49d)
Vy = 2 f~ sA(s) [e- msy - e- SY ] eos (sx)ds (2.4ge)
where (
m = m(o:s) = 1 +
l)t
0:2 S 2 (2.50)
and 0: and D are defined by equations (2.13).

The remaining boundary eonditions
fJy(x, 0) = 0, I x I> a; My(x, 0) = - M(x), Ix I < a (2.51)
express symmetry and eontinuity aeross the uneraeked portion of the xz-plane
and loading on the eraeked portion. When the stress applied to the erae k
is denoted by a~(x, 0, z), equations (2.14) and (2.51) involve
- M(x) =fh l 2 a~(x, 0, z)z dz (2.52)

- hl2
From the diseussion of seetion 2.1 and the bending portion of equations
(2.2), one finds, for
a*
y(x, 0, z) = - h123 M OZ (2.53)
the speeial form,
M(x) = Mo (2.54)
a eonstant.
The first of equations (2.51) gives the eondition
(1 - v)DfJix, 0) = - 2 J: ~ A(s) eos (sx)ds = 0, I x I> a (2.55)
Let
u(x) = 1(1 - v)DfJy(x, 0) (2.56)
Then, from inverting equation (2.55,)
-1 A(s) = - -2 Ja u(x) eos (sx)dx (2.57)

s 'Tr 0
64 R. J. HartranJt
The second of equations (2.51),
My(x, 0) = f~ t(as) A(s) cos (sx)ds = - M(x), I x I< a (2.58)

where
t(as) = 3 + v + 4a 2 s2 [1 - m(as)] (2.59)

can be integrated between zero and an arbitrary point, lxi < a, to become
f OO t(as) -1 A(s) sin (sx)ds = -

fX
M(~)d~, Ix I < a (2.60)
o s 0
In order to solve equations (2.57) and (2.60), the form
u(x) = J a
x
( 2
</>(t)tdt
t - X 2).1.> I x
2
I< a (2.61)
is chosen for the rotation in equation (2.56). This representation has the
form appropriate to a crack problem near x = a. If equation (2.61) is
substituted into equation (2.57) and the n (2.60), one additional step resuits
in a Fredholm integral equation of the second kind,
(1 + v) [</>(t) + J: F 1(t, r)</> (r)r dr] = ~ J~(t~~):~t (2.62)
where
F 1 (t, r) = J~ sg(as)J o (ts)J o (rs)ds (2.63)
and
1
g(as) = !+Vt(as) - 1 (2.64)
For
cP(~) = (1 + v) ~ ~ </>(aOIM 0 (2.65)

equation (2.62) can be written in the dimensionless form,
</>(~) + J~ F(~, '1) cP('1)d'1 = ~ ~~ (~(::~~F J: (2.66)
where
(2.67)
The non-zero stress intensity faetors for equations (2.45) (in whieh
K = (3-v) j (1 +v» and (2.46) ean be written as
(2.68)
where
6
kt = P 4>(1) Mo-.la (2.69)
is the value of kl at z = hj2. In terms of the maximum bending stress at

infinity in Figure 2.1,
(2.70)
(2.71)
The eoeffieient, 4>(1), in equations (2.69) and (2.71) is the solution of the
integral equation (2.66) evaluated at = 1. e
It is shown in [19] that the limiting value of the solution of equation (2.66)
for large aja is
4>(e) -+ -
2
M
-.le f~ (e2
M(ap)dp
2)t as
a
- -+ oo (2.72)
1c 0 0 -p a
and
-.le f~ (M(ap)dp
Mo ):2
t'
2)1 lor
c:. -- 0 (2.73)
o"-P' a
Although the theory is not expeeted to give good re suits for thiek plates,
equation (2.72) verifies the asymptotie behavior whieh is found numerieally
in Figure 2.4. It is further shown that the value of 4>(1) given by equation
(2.73) is approaehed with an infinite slope as aja -+ O.
In general, the solution depends on the parameters of the loading, M(x).
For the particular ease of the bending problem of Figure 2.1, M(x) is given
by equation (2.54). Then
!: M
fo~ (eM(ap)dp -
- p2)t - 2 0
(2.74)
The solution of equation (2.66) depends only on v and ala. The values of
66 R. J. Hartranft
10
v=O 5
f I+
=
~
12
&1
0..;10)1;
,Relssner's Theory
I
Generahzed Theory
02
OL-__~__~____~__~____~__- J
o 0.5 10 15 20 2.5 3.0
ala
Figure 2.4. Stress intensity factor for plate bending
eP(1) for this case are shown on Figure 2.4. Note the infinite slope which
occurs for IXla = 0.
The value for vanishingly thin plates
+
1
eP(l) = 3 +
v IX
v for --; = ° (2.75)
first reported in [21] should not be used. Because of the infinite slope, the
correet value of eP (1 ) for a small, but finite, thickness will be significantly
different from that of equation (2.75).
Further results in symmetric loading can be found in [19]. Twisting of a
cracked plate has also been considered [22]. All results have the proper
singularity of equations (2.45) and (2.46). The improved results, as compared
with c1assical theory, should have been expected. The Kirchoff boundary
conditions of c1assical theory are said to affect the solution only near the
edges. But it is just these edges which give the crack problem its character,
and the solution near the edges is the source of the stress intensity factors.
Generalization of Reissner's theory. The discussion of the generalized

Reissner plate bending theory leads to the conc1usion that equations (2.48)
and (2.49) apply provided that D and IX are defined by equations (2.23). In
the boundary condition of equation (2.51), the quantity
- M(x) = f
hl 2
- hl2
(J:(x, 0, z)z dz (2.76)
as required by equations (2.14), should be used.

The procedure ou tIined by equations (2.55) through (2.67) applies with

the new meanings of rt, D, and with M(x) defined by equation (2.76). But
since equations (2.52) and (2.76) are identical for the same given O"~(x, 0, z)
(or different applied stresses with the same moment), the resulting integral
equation wiII be identical to equation (2.66). Therefore, the resuIting non-zero
stress intensity factors for equations (2.46) can be expressed as
kl = kt h 2 / ,,(Z) (2.77)
6
where kt is defined in equations (2.69) and (2.71) and tP(l) is the same
function of rt ja shown in Figure 2.4.
There are several reasons for not taking the z-dependence in equation
(2.77) seriously. In the first place, it can be no more accurate than the
assumptions involved in equations (2.21) and (2.22). Furthermore, if equa-
tions (2.45) (in which K = (3 - v) j (I + v» are used, a completely different
transverse variation.
(2.78)
is obtained. The boundary condition (2.76) is identical for applied stresses

with different z-variations if the moments are the same. Thus, two stress
distributions are equivalent if they have equal moments. In this sense, equa-
tions (2.77) and (2.78) are equivalent, i.e.,
h/2 kl(Z)Z dz =fh/2 kf(z)z dz (2.79)

f
-h/2 -h/2
To obtain kt!O"b,)a (see equations (2.70) and (2.71», it is necessary to

choose a particular form for f(z). Any choice, when substituted into equa-
tions (2.23), leads to a value of rt/a. The value of the stress intensity factor
is the n determined from Figure 2.4.
A previous solution to this problem [23] used Fourier sine and cosine
series to represent the transverse variations of equations (2.21) and (2.22).
These are equivalent to the cIosed form expression for the variation of the
transverse normal stress,
1z = _ 1. _ ! sinh (2nzjh) - (2nzjh) co sh n

(2.80)
( ) 2 2 sinh n - n co sh n
where n is an arbitrary parameter chosen equal to

68 R. J. HartranJt
nh
(2.81)
2a
in [23]. As required,j(z) is -1 or zero for z = h/2 or -h/2. The variation

of the transverse shear stresses is given by
f'(z) = _ !!. eos~ (2nz/h) - eosh n (2.82)

. h smh n - n eosh n
whieh is zero at z = ±h/2. The in-plane stresses va ry as
J"(Z) = _ 2n22 sinh (2nz/h) (2.83)

h sinh n - n eosh n
For small values of n, these reduee to the variations of equations (2.11) of

Reissner's theory.
The maximum values of f'(z) and J"(z) are shown on Figure 2.5 as
funetions of n. The limiting values for large and small n are also given.
Figure 2.6 shows the normalized values of the funetions of equations (2.80),
(2.82), and (2.83) for various values of n. Note the rapid deerease of trans-
verse shear and the rapid inerease of in-plane stresses near the surfaee for
large values of n. This behavior ean be used to mode! a boundary layer
effeet at the surfaee. The integraIs of equations (2.23) ean be evaluated to
obtain
25
20
15r---__
-h !'(0)
-I
10
~ h'f'(h/2)
05
Figure 2.5. Maximum values of !'(z) andf"(z) in generalized plate bending

n3 sinh 2n - 2n
Il
12 (sinh n - n eosh n)2
(2.84)
11 2 5 4n + 2n eosh 2n - 3 sinh 2n
a2 = 10 2n 2 sinh 2n - 2n
whieh are shown on Figure 2.7.

For a partieular choice of the parameter n, equations (2.84) ean be used
in eonjunetion with Figure 2.4 to obtain kV(Jb..) a as a funetion of hja. For
n = Pno (2 .85)
where P is a eonstant and no is giyen by equation (2.81), Figure 2.8 results.

When P=Ü, the euryes eorrespond to Reissner's theory. The euryes for
P= 1 are shown in [23].
1.0
0.8
0 .6
-..
~
'" 0 .4
0 .2
0 .2 0.3 0 .4 0 .5
'1( ')- .5
Figure 2.6. Transverse varations in generalized plate bending

(a) Transverse norma I stress
0k--n~~~~.-~~-+.
o 0 .2 0.4 0 .6 0.8 1.0
1'( ' )/ f ' (O)
Figure 2.6. Transverse variations in generalized plate bending

(b) Transverse shear stresses
70 R. J. Hartranft
1.0
0 .4 0 .6 0 .6 1.0
f "( z)/ f"(h/2)
Figure 2.6. Transverse variations in generalized plate bending

(e) In-plane stresses
2.5
20
1.5
";;
~-
'-0
05
~
0
0 4 6 6 10 12
Figure 2.7. Parameters of the generalized bending theory
10
p· o
0 .8
l!;. 0 .6 II =0.3
'!!
':Ii" 0.4
n=pff
0 .2
O L
O --~----~--~
3 --~4~--4----L
6 --~
ht o
Figure 2.8. Stress intensity faetor for generalized plate bending

Hartranft-Sih theory for plate extension. This speeial version [9] of the
general theory of extended plane stress eonsists of a field of stress and dis-
plaeement given by equations (2.28) and (2.29), but with V x = vY = Sx = sY =
= t XY = O. The governing equations then beeome (2.41) and (2.42). Standard
{
methods of integral transforms lead to the solution [9]
(1 - v)DU x = rx 2 J~ 2(1 - v)pmrx2s2e-msy
- Im [(1 + iBf3) ! e-PSY]}SA(S)SinSXdS (2.86a)
(1 - v)DU y = rx 2 f: {2(I - v)f3rx2s2e-msy
- Im[(1 + iBf3)qe- PsY ]}sA(s) eos sx ds (2.86b)
(1 - v)DU z f:
= rx 2 Im[(l - iBf3) i e-PSY]A(S) eos sx ds (2.86e)
for the displaeements. The parameter m is given by equation (2.50) while

the eomplex variable
_ ( 1 - v - iBf3 ~)t
P - 1 + 1 _ v2 + 13 2 rx2s2 (2.87)
In addition,
B= e~ :~ t, q = (1 - v2 )t [1 + (1 + iBf3)rx 2 S2 ] (2.88)
The unknown funetion, A(s), is real-valued and must be determined so that
S/x, 0) = - S(x), I x I< a

(2.89)
U/x, 0) = 0, I x I> a
The other boundary eonditions,
(2.90)
are satisfied by the solution expressed by equations (2.86). This may be

verified by putting y=O into
Sx = Loo {2(1 - v)f3ma 4 s4 e -msy
(2.91a)
Im[ {- 13 _ i (~~ v2)t + (1 + iBf3)rx2S2} i e-PSY]}A(S) eos sx ds

72 R. J. Hartran!t
Sy = f; {-2(1 - v){JmIX4s4e-msy (2.91b)
+ [~2 e-PSY]}A(S) eos sx ds
{-
(1 _1 v2)f Im
Txy = J~ (1 - v){J(l + 21X2s2)1X2s2e-msy

(2.91e)
+ Im[(l + iB{J)qe-psY]1X2s2}A(s) sin sx ds
Zx = f~ {(1 - v){JmIX2s2e-msy
(2.91d)
- Im [% e-PSY]}SA(S) sin sx ds
Zy = J; {(I - v){J1X2s2e-msy - Im[qe-PSY]}sA(s) eos sx ds (2.91e)
Zz = 2 (1 - ~2)t 2 JOO Im [«1 - v2)t - i{J) IL e-PSY]A(S)eOSSXds

IX (1 - v + (J) 0 P (2.91f)
In terms of the funetion,

u(x) = !DUy(x, 0) (2.92)
whieh, by the second of equations (2.89), is zero for lxi > a,
sA(s) = - -2 2
(Jõ::'! Ja u(x) eos sx dx (2.93)
1t IX 0
In view of
S,(x, 0) = t{J1X 2 f~ S2 t(IXS)A(s) eos sx ds (2.94)
where
2 1
t(IXS) = - 4(1 - v)mlX 2s2 + (J(1 _ v2}t (2.95)
1X 2S2
and the first of equations (2.89),
JOO
o st(lXs)A(s) sin sx ds = -
2
(J1X2
JX0 S(~)d~, I x I < a (2.96)
When the funetion of equation (2.92) is represented by

a
JcfJ(t)tdt
u(x) = x (t2 _ x2)t' I x I < a (2.97)
equation (2.93) becomes
2
sA(s) = - pa2 Ja <f>(t)Jo(st)tdt
0 (2.98)
Further use of equations (2.96) and (2.98) lead to an integral equation of

the Abel type which is readily converted to the Fredholm integral equation,
tP(~) + f~ F(~, lJ)tP(rf)dlJ = ~ (2.99)
where
FG, lJ) = (~lJ)t f; Sg( ~ S)Jo(~S)Jo(lJS)dS (2.100)
1 - v2 + f32
g(as) = (1 __ v2) (f32 + 1) t(as) - 1 (2.101)
tP(~) = (1 ~ ~2:}~2 f3; 1) .j~<f>(a~)/So (2.102)
If the displacements and stresses near the crack tip are evaluated and
compared with equations (2.45) and (2.46), the non-zero stress intensity
factors are found to be
kl = kf = kif"(z) (2.103a)
(2.103b)
where
kT = tP(l)So.ja (2.104)
and
(2.105)
and the parameter in the displacement field of equations (2.45) is
(2.106)
The limiting values of equation (2.106) for zero or infinite values of f3 cor-
74 R. J. Hartranjt
respond, respectively, to the cases ofplane strain and generalized plane stress.
The parameter, 11>(1), in equation (2.104) depends on Poisson's ratio, v,
and rxJa and p. It is computed from equation (2.99) for the particular loading
S(x) = So and shown in Figure 2.9 for v = 0.30. A particular choice of the
functionj(z) and application of equations (2.31) gives p and the ratio rxJh.
1.0
I" - 0 .3
0.2
0 L-__~__~____L -_ _~_ _~_ _~

o 0 .5 1.0 1.5 2.0 2.5 3.0 )
a lo
Figure 2.9. Equivalent constant stress intensity factor in plate extension
Thus the applicable curve in Figure 2.9 can be used with an appropriate
change of scale of the abscissa to show 11> (I ) as a function of hJ a.
The function j(z) was chosen in [9] to satisfy the plane strain condition
as cIosely as possible. In order to have kl = kr
in equations (2.103), the
function must be of the form
j(z) = - Bcos (P 2:) (2.107)
Thisexpressionisincapable of satisfying the conditionsj(±hJ2) = f'(±hJ2) =

= O. Therefore, a boundary layer of thickness ehJ2 is introduced in which
the condition of plane strain is violated. Within the boundary layers, a form
is desired which matches equation (2.107) and its first and second derivatives
at the edges, z = ±(I -e)hJ2, of the boundary layer. It was considered
appropriate to relax the requirement thatf'(z) andj"(z) represent the deriva-
tives ofj(z) within the boundary layers. Thus, the three transverse variations
finally chosen were
/mproved approximate theories of bending and extension of flat plates 75
- B eos (2Pz/h), 0 :::;; z :::;; (1 - e)h/2

(2.108)
j(z) - B(l - 0 [(1 + 20 eos (1 - e)P - (Pe sin (1 - e)P],
2
h h
(1 - e) -2 -< z -< -2
(for o"z) where ( = (13-1 + 2z/h) / e,
~ PB sin (2pz/h)
1'(z) = (2.109)
~ PB(l - 0 [(1 +0 sin (1 - e)P + (Pe eos (1 - e)P + C(2]

(for 1: xz and 1:yz ) where e is chosen to satisfy equation (2.105), and
j"(z) = h42 p 2 B eos (2Pz/h), 0 < z < h/2 (2.110)
(for 0" x' o"y, and 1:Xy). When the eoeffieient B is chosen to give /3 = 1, equations
(2.108), (2.109), and (2.110) represent the transverse variations shown in
Figure 2.10. Parts (a), (b), and (e) ilIustrate the effeet of changing the
parameter, P, for a fixed boundary 1ayer thiekness, e = 0.1. The relation
13= 1/(2 + 8h/a) (2.111)

was introdueed in [9] to simulate a boundary layer thickness which becomes
eonstant for large h. Part s (d), (e), and (f) of Figure 2.10 show the effeet of
changing thiekness for P = 0.3.
10
08
~o.6
N
h/a" 10
04 (E.O I)
02
- P'f (z)
Figure 2.10. Transverse variations in plate extension

(a) Transverse normal stress
76 R. J. Hartran!t
10,-------------------_____
OL-________L -_ _ _ _ _ _ _ _L -_ _ _
o 05 10
~PftZ)

(b) Transverse shear stresses
10
08
~06
N
o4 (E Arbltrory)
0.2
O~ ________L -_ _ _ _ _ ~~Li_ _
o 05

(e) In-plane stresses
~O.6
0.2
OL-______~----~~~_
o 0.5 1.0
-P'f(z)

(d) Transverse normal stress
--
I ori---=~~~~=======
h/o =0 4
0~0------~0.~5------~10~
~ PH,)

(e) Transverse shear stresses
10
08
h/o Arbltrary
06
,
~
N
04
02
0
0

(f) In-plane stresses
Additional details of the computation of the parameters rx and f3 are given

in [9]. Equations (2.108) to (2.111) resuIt in f3 and rx/h as functions of hja
with P as a parameter as shown in Figure 2.11. The larger values of P give
smaller values of Pso that equation (2.106) is close to the appropriate form
for plane strain. In addition, the large values of P involve very little variation
of f(z) and j"(z) and a neady linear variation of I'(z) except within the
boundary layers.
Figures 2.9 and 2.11 are combined in Figure 2.12 to show c1>(l) as a func-
tion of hja for v = 0.30 and several values of P. The small values of c1>(l) for
thin plates indicate that the crack does not degrade their strength as much
78 R. J. Hartranft
r- P 1l 0 4 J a /h
2/J 08
'/;/ 1.2~
O L-____ ~ ____ ~ ______ ~ ______'

o 0 .5 10 I5 2.0
h/c
Figure 2.11. Parameters of the modified generalized plane stress

theory for plate extension
as thicker ones. The quantitative detaiis depend on the physical interpretation

of the constant So. According to equations (2.32),
h3 fh/2
So = 121 0"0 j"(z)dz (2.112)
3 - h/2
If 1(z) satisfied all of the conditions imposed on it, equation (2.112) would
result in So = o. *
However, equations (2.110) and (2.112) lead to
h4
So = 0"0 PB sin P (2.113)
121 3
10
08 04
06
08
12
11=03
OL-__ ~ ____ ~ ____ ~ __ ~ ____ ~ __ ~
o 05 10 15 2.0 25 30
h/c
Figure 2.12. Equivalent constant stress intensity factor in plate extension
* It must be noted that this seems to indicate that the full two-term representation of
equations (2.28) and (2.29) and the complete set of equations (2.39), (2.41) and (2.42)
should be used.
One aid in interpreting the results is the idea that applied stresses with
identical weighted averages as defined by equations (2.32) are equivalent.
Extend the idea to stress intensity factors, and consider the constant, kl e>
which is equivalent in the same sense to equation (2.103). It is found that
the equivalent constant stress intensity factor is
k 1e = (]J(l)(JoJa (2.114)
Thus, Figure 2.12 is a plot of k1e/(JoJa as a function of h/a.
Alternatively, the usual average,
1 fh/2
k 1a = h k1(z)dz (2.115)
- h/2
can be discussed. It is found that

Pk
k1(z) = . lpa cos (2pz/h) (2.116)
sm
where
k 1a = 4 sin 2 P k
P 2P + sin 2P le (2.117)
The curves of Figure 2. I 3 showing the dependence of the average stress

intensity factor on plate thickness are obtained by combining Figure 2. I 2
and equation (2.117).
10
12
°0~~0~.5~~~10~~~15~~2~0~~2~.5~~30
h/o
Figure 2.13. Average stress intensity factor in plate extension
Either measure, equation (2.114) or (2.1 17), of the stress intensity factor
or indeed some other such as the maximum
80 R. J. Hartranft
(2.118)
may be used to discuss the amount of reduction in load carrying capacity

produced by the crack. For definiteness, consider the average value k 1a to
be appropriate for a failure theory. That is, failure by crack growth is
presumed to occur when
(2.119)
where k c is a material constant. Then Figure 2.13 implies that as plate thick-
ness is decreased, the stress (JO! required to cause failure increases as shown
in Figure 2.14. This effect is limited by the general yielding which would
occur for highly stressed plates.
aJh
Figure 2.14. Stress at failure using the average stress intensity factor
Photoelastic analysis of thick plates. The stretching of thick cracked plates

has been subjected recently [24] to a photoelastic investigation. The results
for the transverse variation of the stress intensity factor were directly com-
pared with equation (2.116). Except near the surface, equation (2.116) fits
the experimental data weil ifthe value of the parameter P is chosen properly.
A similar study for the bending problem has also been conducted [25].
For the range of plate thicknesses studied, there is very little deviation from
the linear form of equation (2.68). In fact, the observed deviations seem to
be opposite in direction from those of equation (2.77).
2.4 Guidelines for practical applications
The various plate theories discussed in this chapter and others which may
be developed in a similar way call for the use of engineering judgement.
This is best developed by actual experience working with the theories, but
one of the ai ms of this chapter has been to show explicitly the approximations
which are involved and the reliability ofparticular results. In this concluding
section, general considerations are discussed from the viewpoint adopted for
deriving the theories.
Basic to the theories is the initial form assumed for the solution. A partic-
ular transverse variation is used and aset of differential equations is obtained.
For the generalized theories, several transverse variations can lead to the
same set of equations, for it is the integrals of the variations which dominate.
This feature makes it possible to model certain characteristics, either ob·
served experimentally or obtained from more exact analyses, with judicious
choice of the transverse variation. For example, the form chosen for extension
of a cracked plate was the one which permits the plane strain crack tip
characteristic to be included.
In a similar vein, the fact that only certain weighted averages of boundary
stresses and displacements affect the boundary value problem is an aid in
interpreting results. Much more confidenee can be placed in the averages
predicted by the theories than in the detailed transverse variations. Th e
transverse variations must be judged by comparison with experiment or by
conformity with known characteristics.
In general, the entire body of results is applicable to thin plates provided
that the initial assumptions are reasonable. However, there is the danger for
thin plates that yielding will begin at relatively low loads. On the other hand,
only the averages can be reliably used in the analysis of moderately thick
plates. But because there is less danger of yielding, the theory may actually
agree better with experiment than for thin plates.
For the crack problem, in particular, the discrepancies between the theo-
retieal form of the crack tip solution (equations (2.45) and (2.46)) and that
expected [14] are notable. Whereas the plane strain value of K is expected,
the generalized plane stress value is obtained for the plate bending theories.
In the case of extension, according to the Hartranft-Sih theory, the plane
strain value is closely approximated for some sets of transverse variations.
For one of the bending theories, different stress intensity factors were ob-
tained for the displacements than for the stresses. Finally, the transverse
82 R. J. Hartranft
normal stress is zero in the case of bending and is given by the appropriate
plane strain value for the interior portion of the crack front in the case of
extension.
References
[I] Tiffen, R., An Investigation of the Transverse Displacement Equation of Elastic

Plate Theory, Quarterly Journal 0/ Mechanics and Applied Mathematics, 14, pp. 59-74
(1961).
[2] Tiffen, R. and Lowe, P. G., An Exact Theory of GeneraJly Loaded Elastic Plates in
Terms of Moments ofthe Fundamental Equations, Prodeecings 0/ the London Mathe-
matical Society, 13, pp. 653-671 (1963).
[3] Sayer, F. P. and Calder, C. c., Plane Stress Theories, Proceedings of the Cambridge
Philosophical Society, 63, pp. 1369-1378 (1967).
[4] Green, A. E., The Elastic Equilibrium of lsotropic Plates and Cylinders, Proceedings
al the Royal Society al London, Series A, 195, pp. 533-552 (1949).
[5] Friedrichs, K. 0., Kirchoff's Boundary Condition and the Edge Effeet for Elastic
Plates, Proceedings of the Symposium on Applied Mathematics, 3, pp. 117-124 (1950).
[6] Friedrichs, K. O. and Dressler, R. F., A Boundary-Layer Theory for Elastic Plates,
Communications on Pure and Applied Mathematics, 14, pp. 1-33 (1961).
[7] Reissner, E., On Bending of Elastic Plates, Quarterly 0/ Applied Mathematics, 5,
pp. 55-68 (1947).
[8] Reissner, E., On a Variational Theorem in Elasticity, Journal 0/ Mathematics and
Physics, 29, pp. 90-95 (1950).
[9] Hartranft, R. J. and Sih, G. c., An Approximate Three-Dimensional Theory of
Plates with Application to Crack Probiems, International Journal of Engineering
Scienee, 8, pp. 711-729 (1970).
[l0] Courant, R. and Hilbert, D., Methods 0/ Mathematical Physics, 1, Ch. 4, pp. 164-274,
Interscience (1953).
[11] Goldenweiser, A. L., On Reissner's Plate Theory, Izvestia Akademii Nauk SSSR,
O.T.N. 4, pp. 102-109 (1958).
[12] Fersht, S., An Extended Reissner Thin Plate Theory, Israel Journal al Teehnology, 2,
pp. 312-317 (1964).
[13] Reissner, E., On the Calculation of Three-Dimensional Corrections for the Two-
Dimensional Theory of Plane Stress, Proceedings al the Filteellth Sellli-Annual Eastern
PI/Otoelasticity Conlerellce, pp. 23-31, Boston (1942).
[14] Hartranft, R. J. and Sih, G. c., The Use of Eigenfunction Expansions in the General
Solution of Three-Dimensional Crack ProbIems, Journal of Mathematics and Me-
cllC/llics, 19, pp. 123-138 (1969).
[15] Kassir, M. K. and Sih, G. c., Three-Dimensional Stress Distribution Around an
Elliptical Crack Under Arbitrary Loadings, Journal of Applied Mechanies, 33, pp.
601-611 (1966).
[16] Williams, M. L., The Bending Stress Distribution at the Base of a Stationary Crack,
Journal of Applied Meehanics, 28, pp. 78-82 (1961).
[l7] Sih, G. c., Paris, P. C. and Erdogan, F., Craek-Tip Stress-Intensity Faetors for
Plane Extension and Plate Bending ProbIems, Journal of Applied Meehanies, 29,
pp. 306-312 (1962).
[18] Sih, G. c., editor, Methods of Analysis and Sahttions of Craek ProbIems, Noordhoff
International Publishing, Leyden (1973).
[19] Hartranft, R. J. and Sih, G. c., Effeet of Plate Thiekness on the Bending Stress
Distribution Around Through Craeks, Journal of Mathematies and Physies, 47,
pp. 276-291 (1968).
[20] Wang, N. M., Effeets of Plate Thiekness on the Bending of an Elastie Plate Containing
a Craek, Journal of Mathematies and Physies, 47, pp. 371-390 (1968).
[21] Knowles, J. K. and Wang, N. M., On the Bending of an Elastie Plate Containing a
Craek, Journal of Mathematies and Physies, 39, pp. 223-236 (1960).
[22] Wang, N. M., Twisting of an Elastic Plate Containing a Craek, International Journal
of Fraeture Meehanies, 6, pp. 367-378 (1970).
[23] Sih, G. c., Bending of a Craeked Plate with an Arbitrary Stress Distribution Aeross
the Thiekness, Journal of Engineering for Industry, Transaetions of the ASME, 92.
pp. 350-356 (1970).
[24] Villarreal, G., Sih, G. C. and Hartranft, R. J., Photoelastie Investigation of a Thiek
Plate with a Transverse Craek, Journal of Applied Meehanies, 42, pp. 9-14 (1975).
[25] Mullinix, B. R. and Smith, C. W., Distribution of Loeal Stresses Aeross the Thiekness
of Craeked Plates under Bending Fields, International Journal of Fraeture, 10, pp.
337-352 (1974).
R. Badalianee, G. C. Sih and E. P. Chen
3 Through eraeks in multilayered

plates
3.1 Introduction
Analytical modeling of layered plates containing cracks has received very

little or no attention in the past. This is mainly because the problem is
extremely complicated since it is basicaIly three-dimensional in nature. The
presence of cracks or surfaces of discontinuity adds additional difficulties to
the development of a tractable theory of layered plates. One of the objectives
of this work is to obtain aset of governing equations that can yield effective
solutions to the mixed boundary value crack probIems. The approach of
Reissner [I] and Hartranft and Sih [2] wiIl be foIlowed in that the comple-
mentary energy functional of the assumed state of stress wiIl be minimized
with respect to the admissible variations of the stress quantities satisfying
the equilibrium equations ofthree-dimensional elasticity and traction bound-
ary conditions. The analysis is approximate as the compatibility or continuity
conditions are satisfied only in an average sense. In contrast to the conven-
tional approach adapted in the c1assical theories of plates, the stress dis-
tribution in the thickness direction of the plate will not be assumed arbi-
trarily but determined from the plane strain condition ahead of the crack
as first suggested by Sih [3].
The general formulation is valid for a plate containing any number of
layers made of different materials with different thicknesses. A bendingless
plate, however, must be constructed with even number of layers and requires
the stretching load be applied such that the system possesses material and
geometric symmetry with reference to the middle plane of the plate. As an
example, the case of a three-layer plate with a through crack subjected to
tensile loading is treated. Stress intensity factors are determined for various
values oflayer thicknesses, material con st ants, etc. Although the assumptions
86 R. Badalianee, G. C. Sih and E. P. Chen
of isotropy and homogeneity of each layer have been adopted in the devel-
opment of the theory, the analysis can be easily extended to layered plates
whose individual layers are anisotropic and nonhomogeneous.
3.2 Minimum complementary energy applied to a

layered plate
The complementary energy principle is developed for a two-material system

and the results are easily extended to the corresponding variational principle
for a system with any number of layers. Let the complementary energy, ep,
of an isotropic, homogeneous body be defined as the strain energy of that
body minus the work done on the portion of the body surface SII over which
the displacements ii i are specified, i.e.,
ep = J V tpdu - Js Tii·ds
I !l I
(3.1)
where the strain energy density is
tp = f:'J (;ijdO'ij (3.2)

Summation on the indices i,j, = x, y, z are understood.
Consider a two-material body subjected to traction boundary conditions
along its external surface as shown in Figure 3.1. For convenience, these
Reglon ®
s (I)
T
Troctlons SpeClfled on the

E)dernal Surfoces
Reglon CD
Figure 3.1. A two-phased material
conditions are assumed to prevail over the entire external surface so that
the complementary energy reduces to the strain energy and only the inter-
facial surface Sl must be considered separately. Displacement boundary
conditions can be handled in the same way and do not alter the basic devel-
opment of the analysis. The complementary energy of the two-material
Through eraeks in multilayered plates 87
system consists of the sum of the strain energies of material (l) and material
(2) as follows:
(3.3)
This funetional rp is to be minimized with respeet to the stress eomponents
whieh are eonstrained by
(1) the eonditions of equilibrium*
(J(p). = 0 (3.4)
!J. J
where
(2) the traetion boundary eonditions

(J(p)n(p)
I) J
= TI on S(rP) (3.5)
(3) and the eontinuity of traetions aeross Sl
(3.6)
where njP) are the eomponents of the outward unit normal to the surfaee
of the sub-region (p). For a two-material system, p takes the value of 1 0 r 2.
Thus, on Sl>
(3.7)
The minimization of a funetional subject to aset of constraints can be
aeeomplished with the use of Lagrange multipliers for the non-integrable
eonstraint state d in equation (3.4). For this ease, the funetional II is intro-
dueed as
II = Sv, ITId
T I V +SV2
ITId
T 2 V + Sv, ,(I)(1)d
Ai (Jij. j V +S
V2
.(2)(Jij.
Ai
(2)dj V (3.8)
where }.~1) and ),~2) are the Lagrange multipliers. Minimization of II with
respeet to admissible variations of the stress eomponents leads to aset of
differential equations and boundary eonditions. The variational proeedure
will be earried out primarily to demonstrate that it leads to eontinuity
,f f
eonditions for the displaeement eomponents aeross the material interfaee ST:
(jII = ~
alJ'1 (I)
6(Jij dv + (jlJ'2
~
(2)
(5(Jij dv
v 6(Jij V2 o(Jij
* The notation, j denotes derivative with respeet to the jth variable x, y, or Z.

88 R. BadaUanee, G. C. Sih and E. P. Chen
-
r 1(1)~( (I»d
JVI Ai u Uij,j v -
J
V2
1(2 h ( (2»d
Ai u Uij,j V (3.9)
Application of the divergence theorem followed by a regrouping of terms

yields
(3.10)
The stresses must satisfy the symmetry condition which requires that
öU!~)
Il
= c5u~~)
l'
in VI (3.Ha)
öU!~) = öu~~)
'l II
in v2 (3.Hb)
In addition, constraint condition in equation (3.6) gives

n(·2)öu!~) = - n~l)öu!~) on Sl (3.12)
1 'l 1 'l
Under these considerations, equation (3.9) becomes
ön =f [2 ~ - (AY~ + A~I~)JdV
au!~) I, 1 l, I
f[
VI IJ
+ 2 (j1J'2
!~l -
('(2)
+ Aj,
1(2) Jd
i) V
+1
'1 A. i , j
V2 oU'l
(All) - Al 21 )njl)c5ul])ds = 0 (3.13)

s,
In order that ön = 0 for all arbitrary combinations of the six stress varia-
tions öuW in VI and on Sl and of öulP in V2 and on Sl' it is necessary that
alJ'l 1('(1) 1(1» - 0 .

au\~)-z.A.i,j+Aj,i - ,InV I (3.14a)
Il
a'P 2
+ Aj, i) -
I 1(2) 1(2) _ •
aU!~) - Z(Ai, j 0, 111 V2 (3.14b)
Il
while
(AP) - Al 2 » = 0 on Sl (3.15)
atp
is satisfied. From equation (3.2), it is recognized that - - = 6ij and hence
aUij
equations (3.14) in VI and V2 lead to
Alp) = + ul p), P = 1,2, i = 1,2,3 (3.16)
Substitution of this result into equation (3.15) renders

ull) = ul 2 ) on Sl (3.17)
The complementary energy for a nonhomogeneous body is shown to be

the sum of the strain energies of the various constituents minus the work
done over the portion of the external boundary over which displacements
are prescribed. Minimization of this energy with respeet to admissible stress
fields leads to the compatibility conditions within each material and con-
tinuity of displacements across material interfaces.
3.3 An approximate three-dimensional theory of

multi-Iayered plates
An approximate theory of layered plates wiII be developed by seeking a

stationary value for the complementary energy expression such that the
stresses satisfy the equilibrium conditions in the interior of the solid and on
that part of the boundary surface where tractions are prescribed. The
resulting system of differential equations can be solved to yield effective
solutions for the problem of eraeks in layered plates.
Consider a layered plate made of n layers with each layer bei ng of different
thickness and material properties. The layers are assumed to be perfectly
bonded and stacked in even numbers such that the system possesses material
Plone Rl!Igion
CrOS$ SKt.on of Q Symme1ric: LominOle
Figure 3.2. Geometry of multi-layered plate

symmetry with respeet to the middle plane of the laminate as shown in

Figure 3.2. In this way, stretching load produees only in-plane deformation.
The way in whieh the three-dimensional equations of elasticity is relaxed
is to assume that the stress eomponents in eaeh layer ean be approximated
as a produet of a funetion * of the thiekness variable z multiplied by a fune-
tion of the in-plane variables x and y, Le.,
(Jij(p) = f(p)()
Ij
(p) (x, y,
Z gjj
) ( no sum on I,. J,. P) (3.18)
Substitution of the stresses in equation (3.18) into the equilibrium equations
lead to the more speeifie assumption
(J~p) = fp(z)Z~P) (x, y) (3.19a)
['t~), 't~~)] = - f; (z) [Z~) (x, y), Z~P) (x, y)] (3. 19b)
[(J~>, (J~P\ 't~;] = f;'(z) [S~p) (x, y), S;P) (x, y), T;:) (x, y)] (3.1ge)
The stresses (J~P), (J~P), ... , (J~p) are assumed to vary from one layer to the next
through the funetion!p(z) and its derivatives while S~p), S;P), ... , Z;p) remain
the same for eaeh layer, i.e.,
[S~>' S~P>, ... , Z~p)] = [Sx' Sy' ... , Zz] (3.20)
Henee, equations (3.19) further reduee to
(J~p) = fp(z)Zz (x, y) (3.21a)
['t~), 't;~)] = - f;(z) [Zix, y), Z,(x, y)] (3.21b)
[(J~), (J~P), tW] = f;'(z) [Sx(x, y), S,(x, y), Tx'(x, y)] (3.21e)
where !iz) and its first derivative must match their respeetive values at the
interfaees of the adjoining layers. Making use of equations (3.21), the
complementary energy funetional beeomes
(/J = i
p= 1
2~ f [{J~2[S; + s; - 2vpSx S y
p vp
+ 2(1 + Vp)T;y] + 2f~2 (1 + vp) (Z; + Z;)
- 2vpfpf~ Zz(Sx + Sy) + f;Z;}dv - fS~) [J~(SnD~)

* This product form of the stress field solution for crack problems was first discussed by
Sih et al [4] and later by Hartranft and Sih [5].
+ Tns D(Pl)
s
-J'p Zn D(Pl]ds]
z (3.22)
The equilibrium conditions for a layered plate whose surfaces are free of
tractions with lp = I; = 0 for z = ±h/2 lead to
Zz = Zx . .>. + Zy,y (3.23a)
Zx = Sx,x + Txy,y (3.23b)

Zy = Txy,x + Sy,y (3.23c)
When referred to the normal and tangential direetions n and s, the following
relationships are to be observed:
Snn x - Tnsn y = Sxnx + Txyn y (3.24a)
Snny + Tnsn x = Txyn x + Syn y (3.24b)

(3.24c)
The quantities D~Pl, m

pl , and D~pl in equation (3.22) are the normal,
tangential, and transverse components of the displacements prescribed on

the edge S~P) of layer p as shown in Figure 3.2.
The Lagrange multipliers, Ai, are now employed to insure satisfaction of
the equilibrium equations, and thus the functional
il = rp - JA [Al(Sx,x + Txy,y - Zx)
(3.25)
is formed. The variation of il with respeet to Sx' Sy, Txy , Zx, Zy, Zz, and
liz) with the continuity conditions enforced on Ip(z) and I;(z) is
+ SybS x) + 4(1 + vp)TxybTxy] + 4J;2 (1 + vp)

x (ZxbZx + ZybZy) - 2vpJpf;,[Zz(bS x + bSy)
+ (Sx + Sy)bZ z ] + 2f;Zz 8Z z }dv - fS~Pl LJ;'(öSnU~Pl

+ 8Tns D(Pl) -J'p bZ
s n D(Pl]dsJ
z
- JA [A 1 (bS x,x + bTxy,y - bZx) (3.26)

92 R. Badaliance, G. C. Sih and E. P. Chen
+ A,z(bTxy,x + bSy,y - bZy) + A,ibZx,x + bZy,y - bZz)]dxdy
+ pt1 [2!p I}2[S; + S~ - 2v pS xS y

+ 2(1 + vp)Txy]f~'bf~' + 4(1 + vp) (Z; + Z~)fl)f~
- 2vpZ z(Sx + Sy) (fpbf~' + f;' bfp) + 2Z; fpbfp} dv
- fs~p) [(SnU~p) + TnsUs(Pl)bf;' - ZnU;P)bf~]dsJ

Note that bll consists of two parts. The first contains only variations of the
in-plane funetions Sx, Sy, T xy , Zx, Zy, and Zz; while the second includes
only variations ofJp(z) and its derivatives. The integrals containing variations
of the in-plane funetions can be integrated through the thickness and those
containing variations of Jiz) can be integrated in x and y. Thus, equation
(3.26) may be rearranged as
bll = SA {I1(SxbSx + SybSy) + 2(11 - Iz)TxybTxy
+ Iz(SxbSy + SybS x) + 13(ZxbZx + ZybZy)
+ 14 [ZzCbS x + bSy) + (Sx + Sy)bZ z] + IsZzbZ z
- A,l(bS x,x + bTxy,y - bZ x) - A,z(bTxy,x + bSy,y - bZy)
- A,3(bZx,x + bZy,y - bZz)}dxdy
- SC u (unbS n + usbTns + uzbZn)dC

(3.27)
+ ptJ::_5IXlPY;(Z)bf;(Z) + IXY:Y;(z)bf~(z)
+ 1X~)[Jiz)bf;(z) +f ;(z)bfp(z)] + 1X~>.rp(z)bfiz)
+ f31P)bf;(z) + f3~P)bf;(z)}dz
where A stands for the plate area and C u is that part of the contour on which
displacement is specified.
In equation (3.27), Un> Us and Uz are defined by
Un = PtJ::-l V~p>.r;(z)dz (3.28a)
Us = Ptf:-l V~p>.r~'(z)dz (3.28b)

Through eraeks in multifayered plates 93
(3.28c)
The quantities Ii' 12 , ••• , Is stand for
E1 ft
n p
Ii = ~ f~2(Z)dz (3.29a)
p-i p tp_1
(3.29b)
(3.29c)
(3.29d)
(3.2ge)
while lit), rx<j), ... , rx<,f) are given by

rx~p) = ~ p
f A.
[S; + S; - 2vpS x Sy + 2(1 + vp)T';y]dxdy (3.30a)
(3.30b)
rx(p)
3
=- 2f
E
P A
Z (Sx
Z
+ S y )dxdy (3.30c)
rx(p)
4
=E 1-f P A
Z Z2 dxdy (3.30d)
Finally, p~p) and p<j) are the integrals
p1 P) = Jc u (SnV~p) + Tn.V;p»dC (3.31a)

p~p) =- Jc u ZnV!p)dC (3.31b)
Integrating equation (3.27) by parts and applying the divergence theorem

give
MI = JA [(IiSx + I 2 S + I 4 Z z + A,i,JJS x
y
+ (IlSy + I 2S x + I 4 Z z + A,2,y)JSy
+ (2(11 - I 2)Txy + Al. y + A2.x)8Txy

+ (13 Z x + Al + A3.x)8Zx
+ (13Zy + A2 + A3. y)8Zy
+ (14(Sx + Sy) + I sZz + A3)8Zz]dxdy
- SCu[(un + An)8Sn +(u s + As)8Tns + (u z + A3)8Zn]dc
n
+ pi~)zz - P~)z]Jfiz)dz + I (r/tY;(z) + a<t'Y(z)
p~l
n
+ IJiP)8f;(z) 1~:-1 + I (- aipy; '(z) + aipY;(z)
p~l
- r:t.(P)f'(Z) -
3Jp
p(p)
l.z
+ P(p)8f
2 P
(z) I lp
I p -l
(3.32)
such that
In the above equations, nl and n2 are the components of the un it normal

vector n in Figure 3.2. The condition 8n = 0 leads to a system of parti al
differential equations in the variables x and y:
IlS x + I 2Sy + I 4Z z + Al. X = 0 (3.33a)
IlSy + I 2S x + I 4Z z + A2.y = 0 (3.33b)
2(11 - I 2)Txy + Al. y + A2.X = 0 (3.33c)
I 3Z x +A l +A3.X=O (3.33d)
I 3Z y + A2 + A3. y = 0 (3.33e)
I 4(Sx + Sy) + IsZ z + A3 = 0 (3.33f)
The funetions Sx, Sy, ... , Zz are constrained as
(3.34a)
(3.34b)
(3.34c)
with An = -Un> As = -Us, and A3 = -Uz on Cu- It is apparent that the
boundary conditions associated with the functions Sx' Sy, etc., can be either
of the traction or averaged displacement type. Along the contour C of the
plate, the prescribed conditions are
(3.35a)
(3.35b)
(3.35c)
The functionfiz) for the pth layer is governed by
rxiPJ~(z) + (2rx<]') - rxY!)f;(z) + rx<:J/z) = fJ~)z - fJi~)zz (3.36)
provided that,f/z),f;(z), (rxlP1;(z) + rx<]'1p + fJlP) and

[rxlP1~'(z) + (rx~p) - rx~P)f;(z) + fJl~~ - fJY!)] are continuous across the inter-
faces. The traction-free surface conditions are reflected by the requirement
that fp(z) andf;(z) must be zero on the plate surfaces.
After a considerable amount of algebra, equations (3.33) can be separated
into
(3.37a)
Zy - a6V Zy
2
= oy0 (al V A3
2
+ a 2A3) (3.37b)
a3V4A3 + a4V2A3 + a sA3 = 0 (3.37c)

where extraneous solutions are eliminated by the condition
(3.38)
The parameters al> a2 , ••• , as are defined as

{[14/(l1 + 12)] + [12 13/(1; - 1~)] - [13/(2(11 - 12 ))]}14
[1S(ll + 12 ) + 13 14 - 21~]
(3.39a)
{[14 /(/1 + 12)] + [1213/(/i - 1~)] - [13/(2(/1 - 13))]} (/1 + 12)

[1S(/l +1 2) + 1314 -21!]
(3.39b)
a3=1~-111s (3.3ge)
a 4 = - 21 4 (/1 - 12) + 1113 (3.39d)
as = - (/i - 1~) (3.3ge)
13
a6 = (3.39f)
2(/1 - 12)
14
a7 (3.39g)
[1S(/l + 12) + 14 (/3 - 21 4 )]
(/1 + 1 2)
as = (3.39h)
[1S(/l + 12) + 14 (/3 - 214 )]
The remaining unknown funetions in terms of the in-plane variables can be
expressed in terms of )'3' Zx' and Zy as
(3.40a)
(3.40b)
Sy = (/i _1 1~) {0
2A 02,1
Il 0/ - 12 0) + [11 13 - 14 (/1 - 12)] a:
oZ
- [12 1 3 + 14 (/1 - 12)] °o~x (3.40e)
The Lagrange multipliers are known through the relations

0
Al = - 1 3Z x - ox (,13) (3.41a)
0
,12 = - 1 3Z y - oy (,13) (3.41b)
The set of equations (3.37), (3.40) and the condition in equation (3.38)
together with the boundary conditions form a complete but approximate

theory of layered plates in three dimensions. One of the attractive features
of this theory is that the problem may be solved in terms of the variables
x and y while the functionJez) governing the stress distribution in the thick-
ness direction may be determined separately.
3.4 Through crack in a layered plate
In the case of a layered plate subjected to loads applied to its edges, the
formulation involves the superposition of two separate probiems. The first
considers the same layered plate and loading with the crack absent. By
removing the load on the plate boundary and considering a crack in the
plate, the second problem deals with the application oftractions to the crack
surfaces. These tractions are equal in magnitude and opposite in direction
to those found at the crack site of the first problem. The superposition of
these two problem s renders the solution to the original case of a cracked
plate under remote loading.
In what follows, attention is focused on the nontrivial problem oftractions
specified on the crack since the application of integral transforms requires
the stresses and displacements to vanish at distances far aw ay from the crack.
The crack length is also assumed to be small in comparison with the in-plane
dimensions of the plate such that the stress distribution around the crack will
not interact with the plate boundary. Hence, the dimensions of the plate
boundary may be assumed to extend to infinity. Referring to Figure 3.3,
the analysis will deal specifically with a three-layered plate containing a
crack of length 2a which penetrates through the entire plate thickness
Figure 3.3. Three layer plate containing a through crack

2(h 1 + h 2 ). A rectangular system of coordinates (x, y, z) are selected such
° °
that the crack is centered at the origin and lies in the xz-plane. The problem
possesses symmetry about the planes x = 0, y = and z = on which 'xy
°
and 'yz vanish while the displacement components Ux and Uz are zero on the
°
respective planes x = and z = 0. The conditions to be specified inside and
outside the crack at y ~ are
o/x, 0, z) = p(x, z), o~ x < a (3.42a)
u/x, 0, z) = 0, x > a (3.42b)
The equivalent conditions in terms of those funetions which depend only on

x and y are
S/x, 0) = p(x), ° ~ x <a (3.43a)
)..z(x, 0) = 0, x > a (3.43b)

°
while T xy and Zy vanish for y = and T xy, Zx and Al vanish for x = 0. The
symmetry condition about the mid-plane of the layered plate requires f(z)
to be an even function of z. The free surface conditions at z = ±h/2, where
h = h l + h 2 are satisfied by requiring f(z) and I'(z) to vanish on these
surfaces. Making use of the conditions in equations (3.43), the funetions
Zx, Zy and )'3 may be determined from equations (3.37) while Sx' Sy and T XY
follow from equations (3.40). The remaining unknown Zz is then given by
equation (3.23a).
In-plane variations of the stresses. The Fourier transform will be applied to

the space variable x for finding the funetions Sx, Sy, etc. Symmetry con-
siderations require Sx, Sy' Zy, Zz' A2 and A3 to be even in x and T xy , Zx,
and Al to be odd in x. Hence, the appropriate Fourier cosine and sine
transforms are employed as follows:
S~(s, y) = f~ Sx(x, y) eos (sx)dx, etc. (3.44a)
f:
and
T;y(s, y) = TxY(x, y) sin (sx)dx, etc. (3.44b)
Application of the Fourier sine and cosine transforms to the governing

equations (3.37) lead s to
d4A~ 2 d2A~
a 3 d y 4 + (a 4 - 2a 3 s) d y 2 + (a 3 s4 - a4s2 + a5)A~ = (3.45a) °
d 2 Z-
-dy2 -
xs ( s2 +-
1
a6
)zs _ S[al-ab
x-
d2A~
dy2 + (~
a6 - ~
a
6
S2) A~J (3.45b)
+~
a )ZCY =
al d3A~
_ ( ~ _ ~ S2) dA~
2_
d Z
_ YC _ ( S2
(3.45e)
dy
2
6 a6 dy3 a6 a6 dy
and equation (3.38) beeomes
dZ c d213c
SZ x
S
+ dy Y = - a7 --2
dy
11.
+ (a2 c
7 s - a 8 )A 3 (3.46)
A solution of equations (3.45) that satisfied the regularity eonditions at

infinity may be written as
(3.47a)
(3.47b)
Z~(S, y) = B(s)e- Q2Y + Re [P2(s)e- 1 \y] (3.47e)

in whieh the parameters ql and q2 stand for
ql = ± {S2 - 2aa43 ± {:: - (2:43) 2J +} + (3.48a)
(3.48b)
The quantities P l and P 2 in equations (3.47) are given by
P 1(S) = - 2s [a 2 - a 1(s22 - ;~)] Al(s) (3.49a)

[1 + a 6(s - ql)]
P 2(s) = - 2ql [a 2 - a l (s2 - ;~)] A 1(s) (3.49b)

[1 + a 6(s2 - q 1)]
The remaining unknowns may be expressed in terms of }.~, Z~, and Z~. For
example, the funetions T~y, S~, and A~
(3.50a)
100 R. Radalianee, G. C. Sih and E. P. Chen
- [12 13 + 14(11 - 12 )]SZ!} (3.50b)
, C( ) _ 1 ZC _ dA.~ (3.50c)
11.2 s, y = 3 Y dy
which are needed for satisfying the required boundary and symmetry eon-
ditions. First of all, the condition Z~(s, 0) = 0 yields
a2 - a l(s2 - q~)J }
B(s) = Re { 2ql [ 1 + a6(s2 _ q~) Al(s) (3.51)
and T:,(s, 0) = 0 relates the real and imaginary parts of Al (s) as
Re {2[1 + i (1- ;i ) (at ; :6(~S22 ~ q~V)JqlAl(S)} =0 (3.52)
It is eonvenient to express the above result in the form

Im[A 1(s)] = p(s)Re[A 1(s)] where
][ 2 1 3 (a 2 - 11a l )(1 + a611) - ala61~)

(1 + a6 11)2 + (a6 12)2
{ [
P(s) = - Re ql s - 2a 6
In equation (3.53), /1 and /2 are defined by
(3.54a)
_ il 2 = [(~)2
2a3
_~Jt a3
(3.54b)
The boundary eonditions in equations (3.43) may be expressed in terms of

the transformed funetions as
-2 Joo S~(s, 0) eos (sx)ds = - p(x), x <a (3.55a)

TC 0
-2 Joo A.~(s, 0) cos (sx)ds = 0, x ~ a (3.55b)

n 0
The two unknowns S~(s, 0) and A.~(s, 0) wiIl be determined in terms of a

single function R(s) defined as
(3.56)
such that
A.~(s, 0) = R(s) (3.57a)
se(s, O) = gis) R(s) (3.57b)

Y gj(s)
The functions gj(s) and gis) st and for
gj(s) = Re[qj(1 + iP)] (3.58a)
g2(S) = (li =- 1~) Re [(1 + iP) {1 q j j + 12s2 + [-qjq2(Il + 12 )1 3

+ (qi1 1 + s21 2)1 3 + (S2 - qi) (lj - 12 )1 4 ]
x [a12 ; :6(~~2 ~ q,~)j}] (3.58b)
Under these considerations, equations (3.55) become
f; R(s) cos (sx)ds = 0, x ~ a (3.59a)
J OO
o sg(s) R(s) cos (sx)ds = --u
np(x)
,x < a (3.59b)
in which
ges) = g2(S) 1 (3.60)
gl(S) sC
The constant e
is equal to m 6 /m2 where m 2 and m 6 are defined in the
Appendix and is obtained from the condition that
lim ges)
S~OO
~ 1 + o(-b-)
s
(3.61)
The system of dual integral equations (3.59) may be solved by reduction to

a standard Fredholm integral equation of the second kind. This is accom-
plished by using Copson's method [6] which involves expressing R(s) in
Jl
terms of CP( e) :
J~ p(a1])d1]
R(s) = - ea 2
o.je CPW Jo(sae) ° (e _ 1]2)4- de (3.62)
where Jo is the zero-order Bessel function of the first kind. The Fredholm
integral equation is of the form
cp(e) + f~ K(e, 1]) CP(1])d1] = .je (3.63)
whose kernel being symmetric in eand 1] is given by

K(e, 1]) = (e1])4- f~ s[ g( : ) - 1 JJ o(se) Jo(s1])ds (3.64)
Equation (3.63) wiB be approximated by aset of algebraic equations using

Simpson's rule and solved numericaBy for the case of a constant load applied
to the crack surfaces, i.e., p(x) = (Jo.
Stress intensity factor. Those terms of the stresses which are singular at
the crack front can be extracted from equation (3.62) and they are associated
with CP(l) representing the value of cp(e) evaluated at the crack tip, = 1. e
For convenience, the polar coordinates ro, rl' r2 and eo, el' e z as shown in
Figure 3.3 are introduced so that
SAx,y) = - ( a(Jo)"!: CCP(1)

r l r2
{~
r 1 r2
sin [tcel + ( 2)]
- :0 eos [eo - ye l + ( 2 )]} + 0(1) (3.65a)
a(Jo {ay.
S/x, y) = + (-----:t)
r r
CCP(1) -
r l r2
sm [1-(e l + ( 2 )]
l 2
+ : eos [eo - teel + ( 2 )]} + 0(1) (3.65b)
(3.65c)
Through eraeks in multUayered plates 103
x {~ eos [Oo - -HOl + 02 )]} + 0(1) (3.65d)
Zix,y) = auo4>(1) ~2 {[ ~1(blI2 + b 1 2 l - ::)]
xfooo .!.J
s
l (as) sin (s x) e-sYds
+ ~l (b l l2 + b2l l + !:) f~ yJ! (as) sin (sx) e-SYds} (3.65e)
el ( b l l 2
Zy(x, y) = auo4>(l) 2-
m2
+ b 2 l l + -ab2 )
6
x f~ yJ 1 (as) eos (sx) e -sy ds (3.65f)
where b l , b2 , el and m2 are given in the Appendix. Note that Zix, y) and
Zy(x, y) are nonsingular. Further, approaehing the eraek tip (a, 0), i.e,
taking the limit r -+ a, 0 -+ 0, r2 -+ 2a, O2 -+ 0, it follows from equations
(3.21) and (3.65) that the near tip stresses may be expressed as
Ux =
r:J.U e,Cz).J a
(2r l)t [eos (~l ) - ! sin (01) sin ( 3~1 )] + 0(1) (3.66a)
uY =
r:J.ue(z).J a
(2r l)!- [eos (O~ ) + t sin (01) sin ( 3~1 )] + 0(1) (3.66b)
[ t sin (01) eos (3~1 )] + 0(1)

(Xup(z).Ja
'r xy = (2r 1)t
(3.66e)
uz =
(xuT(z)~a
(2r l )t [eos (~l )] + 0(1) (3.66d)
'r xz = 'ryz = 0(1) (3.66e)

in which the parameter r:J. stands for
(x = C4>(I) (3.67)
The funetions up(z) and UT(Z) in equations (3.66) depend on the thickness
variable z and are
uiz) = uof;(z) (3.68a)
uoel (b 112 + b211)fiz)
UT(Z) = c (3.68b)
m2
104 R. Radalianee, G. C. Sih and E. P. Chen
Note that for in-plane loading, the stress components r xz and r yz remain
finite at the crack edge'l = 0 while the other stress components are singular
of the order of 1/('1)1- as '1 ~O. This behavior and the angular variations
of the stresses in el are in agreement with those found by Sih [3] for the exact
three-dimensional solution of a crack in an elastic solid. From equations
(3.66), the following stress intensity factors may be defined:
kip)(z) = IY.(Jof;(z)--!a (3.69a)
kiT)(z) = IY.c(Joe 1 (b 112 + b211)fiz)~a (3.69b)

m2
UnIike the two-dimensional solution which is independent of z, these stress
intensity factors va ry along the crack front through the funetions f(z) and
f"(z). In view of equation (3.68a), equation (3.69a) may be written as
(3.70)
which may be regarded as a measure of the normalized intensification of the

local in-plane stresses. Equation (3.67) shows that lY. can be computed from
the Fredholm integral equation for <P(I).
3.5 Stress distribution across the plate thickness
The stress distribution across each layer of the plate as governed by fiz)
and its derivatives may be determined from equation (3.36). In the immediate
vicinity of the crack front, the asymptotic stress solution in equations (3.66)
suggests an alternative approach that has also been used successfully else-
where [3, 7]. More specifically, equations (3.66) indicate that all interior
points near the crack front are in a state of plane strain, i.e.,
(3.71 )
which can be employed to obtainfp(z) in each layer. This requirement yie1ds
a differential equation for fiz):
(3.72)
where w;is equal to the asymptotic value of the function - Zz/[vp(Sx + Sy)]
expanded near the crack tip. The resuIt gives
(3.73)
Referring to Figure 3.3, the subscript p = will be attached to quantities

associated with the middle layer of thickness 2h l while the subscript p = 2
refers to the outer layers eaeh of thiekness h 2 • Solutions of equation (3.72)
that satisfy continuity of fp(z) and f;(z) across the material interfaees are
fl(z) = A eos (wlz), 0:;;;; Z :;;;; h l (3.74a)
f2(Z) = A {eos (wlh l ) eos [wiz - h l )]
- w I sin (wlh l ) sin [wiz - h l )]}, h l :;;;; z :;;;;(h l + h2) (3.74b)

w2
where A is a eonstant and w p (p = 1,2) are related to the constants b l , b 2 ,
etc., as shown in equation (3.73). Refer to the Appendix for the definitions
of the constants in equation (3.73).
Equation (3.71) or (3.72) is not expeeted to hold on the free surfaces
z = ±h(2 where f2(Z) and f; (z) must vanish sinee plane strain applies only
to the interior region of the plate. For a single layer plate, Hartranft and
Sih [2] have suggested to introduce two narrow layers of thiekness 8h(2 next
to the free surfaees. The quantity 8 = Ij2[ I + (4h(2)] as seleeted by Hartranft
and Sih [2] is referred to as the non-dimensional boundary layer* thickness.
For a three-layered plate, 8 is chosen as
8 = 2 + 16 [(~1 + h )ja] (~~)

2
(3.75)
The funetion
fiz) = YI(Z - h l - h 2)4 + Y2(Z - h l - h 2)3 + Y3(Z - h 1 - h 2)2
+Yiz - h l - h2) + Ys, [h l + h2(1 - 8)] :;;;; Z :;;;; (h l + 11 2) (3.76)

is construeted such that the coefficients Yj (j = I, 2, ... , 5) are determined
from the eontinuity of its first and second derivatives aeross the layer at
z = h 1 + h2 (1 - 8) and the traction-free conditions are satisfied by re-
quiringf2(z) = f;(z) = 0 for z = ±(hl + h 2).
* A more detailed diseussion on the theoretieal and experimental justifieations for

introducing e ean be found in the work of Villarreal, Sih and Hartranft [7].
The determination of W I in equations (3.74) is accomplished by employing

an iterative procedure in which an initial value of W I is assumed for calcu-
lating the asymptotic expressions of the funetions Sx' Sy and Zz. The result
of imposing equation (3.71) on Sx, Sy and Zz leads to a new value of W I .
This procedure is repeated until aset of discrete values of W I is found and
equation (3.72) is satisfied. The value of W z follows directly from equation
(3.73) since w~ = (vdvz)wi.
Once Ip(z) (p = I, 2) are known, the crack edge stress field as given by
equations (3.66) is completely determined. To recapitulate, the in-plane
stress components are proportional to I;(z), and the normal stress compo-
nent (Jz is equal to vp«(Jx + (Jy) except in the boundary layer while the trans-
verse shear stresses 'r xz and 'r yz are non-singular. Hence, the intensity of the
singular stress field in the vicinity of the crack edge is described by the stress
intensity factors in equations (3.69) with the values of W I and W z governing
the behavior of Ip(z) and its derivatives.
It should be mentioned that additional boundary layers of the type
discussed earlier may al so be introduced in the neighborhood of the material
interfaces. The effeet of such layers on the stress distribution has been dis-
cussed by Hilton and Sih [8] and will not be covered here.
3.6 Discussion of numerical results
Recall from equations (3.67) and (3.71) that the stress intensity factor
kiP)(z) can be computed from the Fredholm integral equation (3.63) for
<P(~) with ~ = 1. The results will depend on the geometric and material
constants hz/h l , hda, Et/Ez , VI and Vz and presented in a nondimensionless
form kip)(z)/(J p(z),j a, where (Jiz) is the crack face load distribution that
varies through the plate thickness.
Figures 3.4 to 3.7 give the values of k\P)(z) as a function of hda without
the presence ofany boundary layers. The parameter W I in/p(z) may be chosen
arbitrarily as the free-surface conditions for z = ±(hI + h z) are not satis-
fied. The symbol QI = wlh l is varied within the range 0.1 and 1.5 with
o < W I < n/2. Each of the curves with a specific value of QI or W I corre-
sponds to a given stress distribution acro~:s the layered plate. Note that the
normalized stress intensity factor increases as W I is decreased for a fixed
ratio hda. In the limit as hda becomes large, k\P)(z)/(Jp(z),ja approaches
unity for all values of W I. From equations (3.74), it can be seen that as
W I and W z (being directly proportional to W I ) decrease the stress field that
1.0
08
.0.."0 I
':: 03
05
{ 09
15
N
-t:.
""
~'
E'/E2 'OOl
0 h2/h,'05
, I2h, VI =,,! =03
I T h2
0 10 20 30
hl/ O
Figure 3.4. Normalized stress intensity factor for EI/E2 = 0.01 and h 2/h l = 0.5
O.
n,'OI
03
05
07
09
12
15
~
EI/E2= 0.1
hZ/hl I: 0.5
o 0 ...i hz
. I2h,
I T h2
o 1.0 20 30
Figure 3.5. Normalized stress intensity factor for EdE2 = 0.1 and h 2/h l = 0.5
1.0
0
n,'O.3
~N
05
07
0.9
~N
1.2
1.5
~ O.
~
EII E2'O I
h2/h, '2.0 -L h2
Q
' I2h,
1',=1'2 =0.3
I Th2
0 1.0 2.0 3.0

hl/ a
Figure 3.6. Normalized stress intensity factor for El/E2 = 0.1 and h2/hl = 2.0
10
08
n,'03
~ 05
07
N
06 09
~ I 2
15
.,.
N
-a""':.
04
~
EI/E •• O 5
h'/h, • 2 0 -'- h.
02
' T2h,
111=11'2 =0.3
I Th.
0 10 20 30
h1/a
Figure 3.7. Normalized stress intensity faetor for EI/E2 = 0.5 and h2/hl = 2.0
is governed by fiz) becomes less and le ss dependent on z. In fact, as (01

approaches zero, the crack edge stress field in equations (3.66) becomes
independent of z and reduced to the two-dimensional plane strain solution
for which k\p) = k\T) = uo.Ja.
The dependence of k\P)(z) on the change in relative layer thickness h2 /h 1
is also exhibited in Figures 3.4 to 3.7. As h2 /h 1 increases, the stress intensity
factor increases for fixed hda until it approaches the limit unity for large
h2 /h!. This is expected since increasing plate thickness results in recovering
the plane strain condition that corresponds to large values of h2 /h l •
In the absence of the boundary layers, the influence of the relative layer
stiffness Et/E2 is rather weak. No appreciable change in klP)(z) is observed
for EdE2 = 0.01, 0.1 and 0.5.
Recall that the parameter (OI can no Ionger be chosen arbitrarily if the
boundary Iayers next to the free surfaees are to be included in the analysis.
For simplicity, the case VI = V2 and henee (OI = (02 will be considered. This
suggests that (01(h l + h2 ) is al most constant and approximately equal to n/2.
NumericaI results of klP)(z) are presented graphically as function of hda for
different values of E I /E2 , h2 /h l with VI = V2 = 0.3. Figures 3.8 to 3.12
show that klP)(z) decreases as EdE2 increases while the other parameters
are heI d constant. Increasing EdE2 means that more Ioad is transmitted
to the middle Iayer. In the Iimit as E 2 goes to zero, the actual plate half
thickness is reduced from h l + h2 to h l and the stress intensity factor is
expected to decrease accordingly*. The influence of h2 /h l on this effect can
* In the present model, it is not possible to aecount eompletely for this effeet. The load
distribution permitted in this formulation for finite but small E2 is qualitatively different
than for E2 = 0, Le., for E2 finite, Pl ~ 71/2 (l(h 1 + h2» while at E2 = 0, Pl ~ 71/2 (l/hl).
9""'
'0
l:;
08
~. ~::' E 1/E 2=O I
05
20
'OO
06
b"
"-
""-
-~ 04
Figure 3.8. Variations of stress intensity factor with hIla modulus ratios and h2/hl = 0.1
(boundary layers included)
'0
a -th,
~
a
E,/E,'O I
. I2h, 0.5
0.8 2.0
/ ' Th. 10.0
QY 2
o 10 20 h'/a 30 40 50
Figure 3.9. Variations of stress intensity factor with hIla for different modulus ratios and
h2/h j = 0.5 (boundary layers included)
1.0
E'/E 2 -O I
0.5
0.8 20
100
o 1.0 20 h'la 3.0 40 50
Figure 3.10. Variations of stress intensity factor with hIla for different modulus ratios and
h2/hl = 1.0 (boundary layers included)
10
E,/EZ~O I
05
20
0.8 100
hz/h l =2 0
11'. = "2 =0.3
o 10 2.0 h'la 3.0 4.0 5.0
Figure 3.11. Variations of stress intensity faetor with Il Ila for different modulus ratios and
112/111 = 2.0 (boundary layers included)
1.0
0.8
~
a a hZ/h, -10.0
.thz
, I2h, ". ="2 = 0.3
Q
I Thz
o 1.0 2.0 h'la 3.0 4.0 5.0
Figure 3.12. Variations of stress intensity faetor with Il Ila for different modulus ratios and
11 2/11 1 = 10.0 (boundary layers ineluded)
be seen by comparing Figures 3.8 to 3.12. As h2 /h l increases, the effect of

changes of EdE2 on k\P)(z) becomes less pronounced. To understand these
interactions, reeall that at finite Ed E2 increasing h2 /h 1 causes the outer
layers to carry a larger portion of the applied load and for very large h2 /h 1
the layered plate behavior should be almost independent of the material
properties of the inside layer.
Figures 3.13 to 3.16 demonstrate the influence of the parameter h2 /h 1 on
kiP)(z). If hda is held constant, increasing h 2 /h 1 increases the total plate
thickness and k\p)(z) is observed to increase accordingly.
0.8
0.6
~
i- 0.4 h2/h l =IOO
50
20
10
0.2 05
OI
o 10 2.0 h'la 3.0 40 5.0
Figure 3.13. Relative layer thiekness effeet with boundary layers for EdE2 = 0.1
10
III =112= 0 3
08
h2/hl =IOO
50
20
1.0
05
OI
o 10 20 h'la 30 4.0 50
Figure 3.14. Relative layer thiekness effeet with boundary layers for EdEz = 0.5
10
Figure 3.15. Relative layer thickness effeet with boundary layers for EdE z = 2.0
h'/h, , 10.0
5.0
h2./hl = 2.0
1.0
0.5
0.1
10 20 h lfo 30 4.0 50
Figure 3.16. Relative layer thiekness effeet with boundary layers for EJ/E z = 10.0
Finally, Figure 3.17 gives a plot of k\P)(z)/aiz)-! a against htJa for the
special case of holding the to tal plate thickness to crack length constant
(h l + hz)/a) = 3.0). The dotted curve represents a homogeneous plate
where El = E z and k\p)(z) is not expected to vary with hl/a or hdh z once
(h l + hz)/a is fixed. The slight variation observed is associated with the
approximate nature of the constructed boundary layer thickness.
1.0
( h, t h.l 10 ' 3.0 E,/E.'0.05
0.1
III =112 = 0.3 0.5
0.8 1.0
2.0
10.0
~ 0.6
tr
"-
N ----
3~ 04
0.2 ~'"'
I
. I2h,
T h2
0 10 2.0 3.0
h./ o
Figure 3.17. Stress intensity faetor variations for eonstant normalizes plate thickness with
boundary layers
The approximate theory developed in this chapter can be easily applied

to solve problems involving the bending of layered plates. This can be
accomplished simply by replacing the symmetric functionfiz) which governs
through-the-thickness variation of the stress field due to in-plane stretching
by a skew-symmetric one for the case of bending.
3.7 Appendix: Definition of constants
The various parameters that arise in the formulation of the Iayered plate
theory are given in this Appendix. The quantities Il' 12 , etc., are integraIs
that can be computed numericaIIy and they are defined by equations (3.29).
+ (I/a6)] [(a2/al) - Il] - 12}

In
b - _ 2 ~ {[Il
(A3.1a)
1 - a6 {[ll + (1/a6)]2 +
b2 = 2 ~ 12 [(1/a6) + (~2/al?] (A3.lb)

a6 {[Il + (I/a6)] + 12}
Refer to equations (3.39) for the definitions of al' a 2, etc.
(A3.2a)
e2 = I 3 b l - 2/ 1 a 6 (A3.2b)
C3 = ~3(l2b2 -llb l ) - t (li -/~)a6 (A3.2c)
dl = 21 2 a 6 - I 3 b2 (A3.3a)
d2 = 1112a6 + Il (l2bl + llb2) (A3.3b)
d 3 = I3[t1112bl + t(li + 1~)b2 - tlz{l~ - 31~)a6] (A3.3c)
(A3.4a)
~:)
1 (A3.4b)
e2
d t (e 2 - Cl
e3 =
1
dl
[ C3 - C2d";
d2 -
C{ ~: - ( ~: Y)J (A3.4c)
m2 = 2(1 - i el) (A3.5a)
(A3.5b)
(A3.5c)
+
bz
T {
-(Il + 12)1 3
[ el 1
~ + 2a6
( Il
"2 el - e2
)
+ (li -8 I~) el
+ l~ e2 - I~] - [lill - 1 2 ) - 1113](lle2 - l2)}] (A3.5e)
ms = (Ii : I;) [ -I 112 e3 + b~ {- (I l +/2)1 3 [(e3 - 8~~ e l}2
+ 2a6
1 ( Ill z )
Il + lzf!2 + -2- el +
(li -4 I~)
+ 2:
Ill z
el
- i(l~ - 3IiI2)el] - [Ii11 - 12 ) - 1113](lze3)}
+ 2'
b2 {(Il +
2
2 3 [ 8a~1
1 )1
el +
1.
4a~
(lllei
-2- - e2
)
2 ((li - l~)
- ~ \ 4 el + 'lez - /z - 2e3) - 1 Z
S(ll -
Z
31 2 )lle l
-
(li-l~)
4
I lllz]
ez - 1 e3 + -2-
- [I4(I~ - l z) - IlI3](lle3)}] (A3.5f)
C = (A3.6)
References
[1] Reissner, E., On bending of elastic plates, Quarterly of Applied Math., 5, pp. 55-68
(1947).
[2] Hartranft, R. J. and Sih, G. e, An approximate three-dim~nsional theory of plate
with application to eraek probiems, Int. J. Engng. Sci., 8, pp. 711-729 (1970).
[3] Sih, G. e, A review of the three-dimensional stre,s problem for a eraeked plate, Int.
J. of Fracture Mechanies, 7, pp. 39-61 (1971).
[4] Sih, G. e, Williams, M. L. and Sewdlow, J. L., Three-dimensional stress distribution
near a sharp eraek in a plate of finite thiekne5s, Air Foree Materials Laboratory,
Wright-Patterson Air Foree Base, AFML- TR-66~242 (1966).
[5] Hartranft, R. J. and Sih, G. e, The use of eigenfunetion expansions in the general
solution of three-dimensional crack probiems, Journal of Mathematics and Mechanics,
19, pp. 123-138, (1969).
[6] Copson, E. T., On certain dual integral equations, Proc. GlasgolV Math. Assoe., 5,
pp. 19-24 (1961).
[7] Villarreal, G., Sih, G. e and Hartranft, R. J., Photoelastie investigation of a thick
plate with a transverse crack, Journal of Applied Meeh., 42, pp. 9-14 (1975).
[8] Hilton, P. D. and Sih, G. e, Three-dimensional analysis of laminar composites with
through eraeks, American Society of Testing and Materials, STP 593, pp. 1-35 (1975).
E. s. Folias
4 Asymptotic approximations to
crack problems in shells
4.1 Introduction
In nature, shells are the rule rather than the exception. The list of natural
shell-like structures is long, and the strength properties of some of them are
remarkable. It is logical, therefore, for man to utilize them in man-made
structures. But to do this safely, we must understand the fundamental laws
which govern the strength and displacement behavior of such structures for
they are not immune to failures, particularly in the fracture mode.
It is the intent, therefore, of this chapter to discuss a theoretieal method
which enables one to determine the stress field that exists in the neighbor-
hood of a crack and furthermore catalog the stress intensity factors for
various shell configurations and loads.
In section 2, the author gives a concise summary of the classical shell
theoryand its limitations. He the n goes on to discuss the general character
of the equations and subsequently shows that for the two simple geometries,
spherical and cylindrical, the equations reduce considerably and in the limit
the governing equations of a fiat plate are recovered.
Because the solution for a general arbitrary initial curvature presents
formidable mathematical complexities, in section 3 he chooses to display
the analytical method by specializing it to a spherical shell. In order to
preserve unity, he does this in some great length giving sufficient detaiis.
In sections 5 and 6 he gives the stress intensity factors for a cylindrical
shell with various crack orientations and for other more complicated shell
geometries.
In section 7, he examines what effect, if any, elastic foundations have on
the stress intensity factors. Such information can be of great practical value
to highway construction and the designing of storage tanks for the oil
industry.
118 E. S. Folias
4.2 General theory-c1assical
In the following, we consider bending and stretching of thin shells of revolu-

tion, as described by the traditional two-dimensional linear theory and with
the additional assumption of shallowness*. In speaking of the formulation
of two-dimensional differential equations, we mean the transition from the
exact three-dimensional elasticity problem to that of two-dimensional
approximate formulation, which is appropriate in view of the 'thinness' of
the shell. We shaII, furthermore, limit our considerations to homogeneous,
isotropic, constant thickness, shaIIow segments of shells, subjected to small
deformations and strains so that the stress-strain reIations may be established
through Hooke's law.
The basic variables in the theory of shallow sheIIs are the displacement
function w(X, y) in the direction of an axis Z and a stress function F(X, Y)
which represents the stress resultatnts tangent to the middle surface of the
shelI. Following Marguerre [2], the coupled differential equations governing
w and F, with X and Y as rectangular cartesian coordinates of the base
plane (see Figure 4.1), are given by:
q (x,y)
Figure 4.1. Initially Curved Sheet
(4.1a)
* Aeeording to Ogibalov [lI, a shell will be ealled: shallow if the least radius of eurvature
is greater by one order of magnitude than the linear dimensions, i.e., LI R ~ 0.1; and thin
ir hl R ~ 0.01.
Asymptotic approximations to crack problems in shells 119
a2F a2wo a2F a2wo

DV 4w = -q-2
aXay aXay + ax 2 avz- +
a2F a2wo
+ ay2 ax 2
(4.1b)
where V q is the biharmonie operator, E Young's modulus, h the thiekness

of the shell, D the RexuraI rigidity, q the internaI pressure and wo(X, Y) the
initiaI shape of the shell in referenee to that of a Rat plate.
The usuaI bending moment eomponents Mx, My, M xy are defined in terms
of the displaeement funetion w as
Mx = - D [;~~ + v ~~~ J (4.2a)
(4.2b)
(4.2e)
and the membrane forees in terms of the stress funetion F as

a2F
Nx = ay 2 (4.3a)
a2F
Ny (4.3b)
ax 2
a2F
N xy = - aXay (4.3e)
Finally, in view of equations (4.2) and (4.3), the bending and extensionaI
stress eomponents beeome:
(4.4a)
(4.4b)
ib)
xy = (4.4e)
and
120 E. S. Folias
a(~) = 1 a2F
h ay2 (4.5a)
= 1 a2F
ax 2 (4.5b)
a(e)
y
h
,(e) = _ 1 a2F (4.5e)

xy
Il aXay
with V bei ng Poisson's ratio.
Beeause of the eoupled nature of the differential equations (4.1), it be-
eomes apparent that there exists an interaetion between bending and streteh-
ing. That is, a bending load will generally produee both bending and
extensional stresses, and similarly a stretching load will also induee both
bending and extensional stresses. The subjeet of eventual eoneern, therefore,
is that of the simultaneous stress fields produeed in an initially eurved sheet
containing a eraek.
A theoretieal attack of the general problem for an arbitrary initial eurva-
ture presents formidable mathematieal eomplexities. However, for the two
simple geometries, spherieal and eylindrieal shells, exaet solutions ean be
obtained in an asymptotie form. On the other hand, for other more eompli-
cated shell geometries results can be obtained by a proper superposition of
these two solutions.
Spherical sheli. For a sh all ow spherieal shell the radius of eurvature

remains constant in all direetions; therefore,
(4.6)
Substituting equations (4.6) into (4.1), one reeovers Reissner's equations [3]
Eh V2w + V4 F = 0 (4.7a)
R
4 1 2 q
V W - RD V F = - D (4.7b)
Flat plate. A fiat plate represents a degenerative ease of a spherieal eap

when the radius beeomes infinite; therefore,
a2wo a2wo a2wo
aXay = ax 2 = ay2 =0 (4.8)
Asymptotic approximations to crack problems in shelis 121
Substituting equation (4.8) into (4.1), one reeovers the elassie equations for
a flat plate, i.e.,
V4 F = 0 (4.9a)
V4 w = - q (4.9b)
D
Cylindrical she/I. For a shallow eylindrieal sheli, one of the prineipal radii
of eurvatures is infinite, hile the other one is eonstant; therefore,
(4.10)
Substituting equations (4.10) into (4.1), one reeovers the equations for a
shallow eylindrieal sheli, i.e.,
Eh
(4.lla)
R
4 1 q
Vw- - (4.1lb)
RD D
Other she/! geometries. If one ehooses the eoordinate axes X and Y such
that they are paralleI to the prineipal radii of eurvature*, then
a 2wo a 2wo 1 a2wo 1
JXJY = 0; JX 2 = Rx ; Jy2 = liy (4.12)
with R x and Ry being the prineipal radii of eurvatures in the X and Y

direetions respeetively.
Substituting equations (4.12) into (4.1), one finds
Eh [ ~
Ry
a 2w
ax 2
1
+-
Rx
a2w
ay2 J+V F=O
4 (4.13a)
1[ 1 a 2F 1 2F J- - q
V4 w - aay2
ax 2 + R x - (4.13b)
D Ry D
* In general, when they are not parallei, equations (4.13) will contain additional terms of
the form (0 2 wjoXo y) and (02FjaXo Y).
122 E. S. Folias
4.3 The stress field in a cracked spherical shell
Formulation of the Problem. Consider a portion of a thin, shallow spherical

shell of constant thickness hand subjected to an internai pressure q(X, Y)
(see Figure 4.2). The material of the shell is assumed to be homogeneous and
2C-1\_ x
TOPVIEW
Y R
~:--:----....
E::=::=;=:7'~I - x
SIDE VIEW
y z
CROSS
SECTION
\
Figure 4.2. Geometrical Configurations of a Pressurized Spherical Cap
isotropic and at the apex there exists a radial cut of length 2e with respeet to
the apex. It is convenient at this point to introduce the dimensionless
coordinates
x Y
x = -, y = -, z =
Z
(4.14)
e e e
Asymptotie approximations to eraek problems in shelis 123
in view of which the coupled differential equations governing the deftection

function w(x, y) and the stress function F(x, y) with x and y as dimension-
alized rectangular coordinates of the base plane, become
Ehe 2 V 2 w + V 4 F = 0 (4.15a)
R
(4.15b)
As to boundary conditions, we require that (I) on the faces of the crack, the
normal moment, equivalent shear, and normal and tangential membrane
forees vanish, and (2) away from the crack, the appropriate loading and
support condition are satisfied.
In treating this type of problem, it is found convenient to seek the solution
into two parts, the 'undisturbed' or 'particular' solution which satisfies
equations (4.15) and the loading and support conditions but leaves residual
forees along the crack, and the 'complementary' solution which precisely
nullifies these residuals and offers no contribution far aw ay from the crack.
However, suppose that one has already found a particular solution satis-
fying equations (4.15), but that there is a residual normal moment My,
equivalent vertical shear V y, normal in-plane stress Ny, and in-plane tangen-
tial stress N xy , along the real axis lxi < I, of the form: *
= - D2 m = 0 = _ n0 = 0
M(P)
y e 0'
V(P)
y 'y
N(P)
7' N(P)
xy
(4.16)
where, for simplicity, we assume mo, no to be constants. **
Mathematieal Statement of the Problem. Assuming therefore that a partic-

ular solution has been found, we need to find two funetions of the dimension-
less coordinates (x, y), w(x, y) and F(x, y), such that they satisfy the homo-
geneous part of the differential equations (4.15) and the following boundary
conditions. At y = 0 and I x I < 1:
(4.16a)
* For particular solutions see section 4.8.

** For mo, no non-constants, see remarks after equations (4.40).
124 E. S. Folias
V/x, 0) = - c~ [~2y~
u
+ (2 - v) ~ a:~
ux uy
] = 0 (4.16b)
(4.16c)
1 a 2F
Nx/x,O) = -? axay = 0 (4.16d)
At Y = 0 and lxi > 1 we must satisfy the continuity requirements, i.e.,
lim [~nn (w+) - ~nn (W-)] = 0 (4.17a)

lyl-->O uy oy
lim
an (F+)
[""0i an] = 0
- -;;--;;(F-) (4.17b)
Iyl-O uy oy
for n = 0, 1, 2, 3. Furthermore, in order to avoid infinite stresses and
infinite displacements we require that the funetions w and F with their first
derivatives to be finite far away from the crack. These restrietions simplify
the mathematical complexities of the problem eonsiderably, and correspond
to the usual expeetations of the St. Venant Prineiple. It should be pointed
out that the boundary conditions at infinity are not geometrieally feasible.
However if the eraek is small eompared to the dimensions of the shell, the
approximation is reasonable.
Reduction of the System. Reissner [4] has shown that the homogeneous
solution to the system of equation (4.15) can be written in the following
form
(4.18)
where </J and !/J are harmonic funetions and X satisfies the same differential
equation as the deflection of a plate on an elastic foundation, i.e.,
(4.19)
(4.20)
One concludes, therefore, that the effect of the initial curvature is quali-
tatively equivalent to providing an elastic foundation for an initiaIly flat
plate, such that, as the radius of eurvature increases, the foundation modulus
beeomes weaker and weaker. This analogy has also been observed and
demonstrated experimentaIly by Seehler and Williams [5].
Method of Solution. We eonstruet next the following Fourier integral repre-
sentations with the proper behavior at infinity
w(x, Y±) = J~ {P 1 exp [ _(S2 - iA 2yt lY IJ
+ P 2 exp [ _(S2 + Oo2Y1- lY IJ + P3e-slyl} eos xs ds (4.21a)
iA 2
F(x, y±) = -
2 RDJoo
- {P 1 exp [ - (S2 - iA 2 yl: lY IJ
e 0
- P 2 exp [ - (S2 + iA 2 )t I y IJ + P 4e -s I y I} eos xs ds (4.21b)

where the P/s are arbitrary funetions of s to be determined from the boundary
eonditions, and the ± signs refer to y > 0 and y < 0 respeetively.
Assuming that one ean differentiate under the integral sign, one finds by
formally substituting equations (4.21) into the boundary eonditions In
equations (4.16) that:
·
I 1m Joo {P 1(Vos 2 - '12) -(s2-iA 2)1/2IYI
lA e + P 2 (VoS 2 + lA'12) e - (s2+iA 2)1/2IYI
Iyl-+o 0
+ Vos 2P3e-sIYI} eos (xs)ds = - mo; I x 1< 1 (4.22a)
± f~ {P l (S2 - iA 2)t (vos 2 + iA 2) + P2(S2 + iA 2)t (vos 2 - iA 2)

+ VOS 3 P 3 } eos (xs)ds = 0; I x 1< 1 (4.22b)
· iA2RDJoo {P - P
11m - 2 1e
(s2-iA 2)1/2Iyl
- 2e
-(s2+iA 2)1/2Iyl
+ P4e -slyl}
lyl-+o e 0
S2 eos (xs)ds = - no; Ix I < 1 (4.22e)

and
+ iA 2RDJoo {P 1 (S2 - iA 2yt - P 2(S2 + iA 2)t + P4 s} s sin (xs) ds = 0;

--2-
e 0 lxi < 1 (422d)
where again the ~ signs refers to y > 0 and y < 0 respeetively, and
Vo = I-v. If one, now, ehooses
VOS 3 P 3 = - {(S2 - iA 2)t (vos 2 + iA 2)P! + (S2 + iA2rl~ (vos 2 - i).2)P2}

(4.23a)
sP 4 = - {p!(S2 - iA 2)t - P 2 (S2 + iA 2)t)} (4.23b)

126 E. S. Folias
then equations (4.22b) and (4.22d) are satisfied automatiealIy and equations
(4.22a) and (4.22e) beeome, respeetively
f~ {[v os2 - i),,2 - S-I (S2 - i),,2yt (V Os2 + i),,2)]P 1
+ [V os2 + i),,2 - S-I(S2 + i),,2yt (Vos 2 - i),,2)]P 2} eos (xs) ds = - mo;

and I x I < 1 (4.24a)
i),,2RD Joo[(1 _ S-I(S2 - i),,2)t)Pl
e2 0
- (1 - S-I(S2 + i),,2)t)P2] S2 eos (x s) ds = no; I x 1< 1 (4.24b)
Furthermore, it ean easily be show n that all the eontinuity eonditions are
satisfied if one eonsiders the following two eombinations to vanish
f~
P1
(S2 - i),,2)t eos (x s) ds = 0; lxi> 1 (4.25a)
~
P2
f~ ?
(S2 + i.Jc2)! eos (xs) ds = 0; I xl> 1 (4.25b)
We have redueed, therefore, our problem to that of solving the dual integral
equations (4.24) and (4.25) for the unknown funetions Pl(s) and Pz(s).
Reduetion to Singular Integral Equations. For the determination of the

unknown funetions Pl(s) and Pz(s), we reduee the problem to aset of eoupled
singular integral equations of the Cauehy type. This ean be aeeomplished if
one lets
f~
P1
(S2 - i),,2)t eos (x s) ds = Ul(X); I xl < 1 (4.26a)
7"
f: 7 P2
(S2 + i.Jc 2r t eos (x s) ds =
whieh by Fourier inversion give

u 2(x); lxi < 1 (4.26b)
(4.27a)
(4.27b)
where the functions Ul(e) and U2(e), due to the symmetry of the problem, are
even. Next, substituting formaUy equations (4.27) into (4.24) one has after
changing the order of integration and rearranging
NY = 2iA2~D fl 1
{u (e)Lt - uie)Ln de (4.28a)
1tC -1
M=- 2D
--:rt fl -1 {ul(~)L; + u 2 (e)L!} de (4.28b)
Y
where
L *1 = -1 Joo S4 exp [ (_(S2 - iA 2)! 1Y IJ
2 ' , 2).1.
(
eos x -
J!) d
.. s s
2 0 s - lA '
- - 1Joo S3 e -s I y I eos (x - e)sds (4.29a)

2 0
L*
2 = -
lfoo S4 exp [( -2 (S2 +• 2).1.
iA 2 )!lyIJ (
eos x -
J!) d
.. s s
2 0 s + IS '
--1 Joo s3esIYleos(x-e)sds (4.29b)

2 0
L* - 2!fOO0 {S2(VOS 2 - iA 2) [-e 2 _ "2)! 1 IJ

3 - (S2 _ iA2)! exp s lA Y
- s(Vos 2 + iA 2) e-S I YI} eos (x - e)s ds (4.2ge)
* 21 Joo0 {s2(vos + iÄ. ) [ 2 . 2! J

2 2
L4 = (S2 + iA 2 )t exp -(s + lA) 1 Y 1
- s(vos 2 - iA 2) e-s I YI} eos (x - e)s ds (4.29d)
The integration in equations (4.29) may now be earried out explieitIy by

making use of the Fourier cosine transforms [6]
fo
OO e-IYlseos('s)ds = 1 Y2
p
1 (4.30a)
Jo
OO exp [ _(S2 + a 2)! 1 Y IJ eos ('s) ds = K (ap); Re a > 0
(S2 + a 2)! 0
(4.30b)
and similar results obtained by differentiating them with respeet to x and y.

In these formulas p2 = ,2 + lyl 2 , and K n denotes the modified Bessel
funetion of the third kind of order n.
128 E. S. Folias
The expressions in equations (4.29) then beeome respeetiveIy
2L! = :x { - A~2( «(2 - 31 Y 12) K o(A{3p)
_ [A 3{33(3
p3
+ 2A{3( «(2 _ 31 Y 12
pS
)J K (A{3p) + 3r._.
1 p4
8(1 Y 12}
p6
(4.31a)
where for simpIieity we have defined a 2 = i and {32 = - i. Thus, the Iimits,
as I Y I --+ 0, of Ny and My are found to be respeetiveIy
.
hm Ny
Iyl-+O
= -
2iA 2RD d
2
nc
d-
JI {uI(~)Ll - U2(~)L2} d~ (4.32a)
Jl
X-I
2D d
lim My = - --d {ul(~)L3 + UzCO L4} d~ (4.32b)
Iyl-+O n x -1
where the integraIs are understood to be of Cauehy prineipaI value and

(4.33e)
(4.33d)
Ifwe set Ny, My, in the limit as lyl --+ 0, equal to - no and - mo respeetively,
integrate with respeet to x, the n we find that they must satisfy the integral
equations
J l
-1 {U l (e)2L l - u2(02L 2} de
nn e 2
= - i).2~D x; I x I < 1 (4.34a)
f~l {u lW2L 3 + u2(e)2L4 } de = -nmox; Ix I < 1 (4.34b)
where the kernelsL 1 , L 2 , L 3 ,L4 , have singularities of the order II( = lj(x- e),
as ean easily be seen by observing their behavior for small arguments:
2L 1 = - 2(x _ e)
).
2
13 2
+ ). 4 134 (x - e)
[ 5 3y
32 - "8 - "83 In).f3 I
x -
2
eIJ
+ o ().6 (x - ~)3In). I x - ~ /) (4.35a)
2L 2
).2il(2
= - 2(x _ e) + ).
4 4
il( (x -
[ 5
0 32 - 8- - "8
3y 3
ln).il(
I-x 2
- ~ IJ
-
+ o ().6 (x - ~?ln).lx - ~}) (4.35b)
2L = _ ).2il(2(4 - VO) ).4134 ( _ 1') [5VO - 8 4 - 3v o

3 2(x _ ~) + x.. 32 + 8
(Y + In).f3l x ~ ~ I)J + O().6(x-~)3In).lx-~/) (4.35e)
2L = _ ).213 2 (4 - 1') ~5vo - 8 4 - 3vo

l
VO) 14 4( _
4 2(x _ ~) + /I. il( X .. 32 + 8
130 E. S. Folias
(4.35d)
We require, therefore, that the solutions u 1(x), u2 (x) be Hölder continuous

for so me positive Hölder indices 111 and 112 for all x in the closed interval
[-1, 1]. Thus in particular, u 1(x), uix) are to be bounded near the ends of
the crack.
However, because of the complicated nature of the kernels Li> an exact
solution for the unknown functions u 1(x) and U2(X) is extremely difficult. On
the other hand, for most practical applications the parameter A attains small
values as follows from the definition of A, namely
A= [12~~~)~2)J* (c/h) = [12(1 - v2 )]*(c/R) (R/h)t (4.36)
It is clear that A is small for large ratios of R/h and small crack lengths.
As a practical matter, if we consider crack lengths less than one tenth of the
periphery, Le., 2c < 2rrR/1O, and for R/h < 103 a corresponding upper
bound for A can be obtained, namely A < 20. Thus the range of A becomes
o < A < 20 and for most practical cases is between 0 and 2, depending
upon the size of the crack.
Solution for Small A. For the simple case A = 0, the problem reduces to
that of a fIat sheet under applied bending and stretching loads, the solution
of which has been investigated by many authors. For example, the problem
for both bending and stretching for an orthotropic plate, containing a finite
crack, was investigated by Ang and Williams [7] and a solution was obtained
by means of dual integral equations. It can easily be shown* that the du al
integral equations can be transformed to two singular integral equations of
the type (4.34) with simpler kernels. Furthermore, these are not coupled and
the solutions can easily be obtained as in § 47 of [9]. Without going into the
detaiis they are found to be of the form A(l - e)t, where A is a constant.
Similarly, the solution for an initially curved sheet must, in the limit,
check the above result and because Ul(~) and U2(~) are in particular to be
bounded near the ends of the crack, it is reasonable to assume solutions of
the form**
(4.37a)
* See Noble [8].
** In fact, one can show [10] that this is precisely the form of the asymptotic solution for
small.\.
(4.37b)
where the eoeffieients A o, AI, ... , B o, B I , ... ean be funetions of} but not
of ~.
Substituting equations (4.31) into (4.34), and making use of the reIation
f~ 1(1 - e)t (x - 0 In ,b: I\ - ~ I d~ = ~ (1 + In Jc :~2 ) x + ~x 3
(4.38)
we find by equating eoeffieients that*
{
I + 16
7IJc2
84 -- 31'0 + 4~
1'0 - 1'0
(8 -32 71'0 + 4 - 3\'0
8
Y)
2
}. 20: ) }
+In - -
16
(4.39a)
It shouId be pointed out that if eoeffieients A 0, B o of higher aeeuraey are

desired, say up to order A2n , then it is neeessary to soIve an 11 x n aIgebraie
system. In effeet, this is a method of sueeessive approximations for whieh
the question of eonvergenee is invcstigated in Referenec [10].
It thus appears that for J. < ).* the power series solutions of the form
ert I eY',
N
U\N) (0 = (1 - A,)2n(1 - (4.40a)
n=O
m= (1 -
N
uiN) e)t I B,)2n(1 - ey, (4.40b)
n=O
* For brevity, in this analysis we will restrict ourselyes to terms up to Oo. ~).
132 E. S. Folias
in the limit as N ----+ oo, will converge to the exact solutions* UI(~) and U2(~)
of the integral equations (4.34). However, since most particular solutions
will give us a non-uniform residual moment and normal membrane stress
along the crack, it is only natural to ask how the soIution changes. Suppose
for lxi < I, we expand mo and no in the form III all x 2n (even powers because
of the symmetry of the problem), then our previous method of solution will
stiil be appIicable. And as can easily be seen from equations (4.34), although
the coefficients A,,, B Il in this case may change, the character of the solution
will stiIl remain the same. Finally, because we desire to focus our attention
upon the singular stresses around the neighborhood of the crack point, we
need only to compute coefficients A o and B o.
Alternate Method of Solution. It is also possibIe to solve the coupled dual

integral equations directly by using a method which was developed by the
author some time ago [Il] and is parallei to the previous method. Thus
motivated by equations (4.25) and (4.26), one assumes the unknown func-
tions P I and P2 in the form
(4.41a)
and
p 2(S2+ i,eyt ~ Jk+l(S)
S2 ~ Bk -)k+1
( (4.41b)
k~O S
where the coefficients Ak and B k are constants to be determined.

The advantage of such a form is that equations (4.25) are automatically
satisfied and furthermore,
ul(x) = f<:l) P 1
o
2(5 -
S
2 . 2 1: ()d - ;, A
k~O
.j rr
lA) eos xs s - ~ k i+ 1r(k.3.)
+2
(1 _ .Z)k + 1:
x
(4.42a)
and
-f<:l) P z
u 2 (x) - 2 (s
2 • 2 1:
+ lA) "
eos (xs)ds
_ ~ . .jrr
- ~ Bk 2k+ Ir(k .3.) (1 - x
2)k+1:
o S k~O +2
(4.42b)
In general, the funetions U j , have some physical meaning. For exampIe, in

this case their algebraic combination represents the crack opening displace-
* This matter is discussed at same length in [10].
ment and as aresult sueh an expansion is plausible.

It foIlows that equations (4.24) take the forms
Jof~ {[VtsSz4 -=- i~~~~z - s(vos z + iA 2 )JAk

+ [ VOS +iAZS
4
.,Z)'
2
Z· zJ }
J k + 1(S) ()d
- s(vos - lA ) Bk ()k+l eos xs s
(s + lA
Z
2 S
= - mo; I x 1< 1 (4.43a)
and
k~O J~{[(S2 _S;A )t - SJA 2 k -[(S2 ;:A2 )t - SJB k}
S2 ~
J ' )
(S)k+ eos (xs)ds =
1
n'
0'
Ix I < 1 (4.43b)
Multiplying, subsequently, both sides by
(1 - x 2 )j+t
2j ru + !)Jrr
and integrating with respeet to x from °to 1 one finds
oo
L {AkGk,/A.; i) + BkGkjA; -i)} = - moHj ; j = 0,1,2,... (4.44a)
k=O
and
oo
L {AkFkjA; i) -
k=O
BkFkjA; - i)} = noH j ; j = 0,1,2, ... (4.44b)
where for simplieity we have made the following definitions
J j + 1(S) d
(S)j+l s
( 4.45a)
J j + 1 (s) ds
(S)j+ 1
(4.45b)
and
1
(4.46)
2j + 1 (j + I)!
134 E. S. Folias
Equations (4.44) now represent two infinite systems of algebraic equations

which are to be solved for the unknown coefficients* Ak and B k • Such
systems have been studied extensively and the questions of existence and
uniqueness of the solution are discussed in reference [13].
As apraetieal matter now, if one addresses himself to the major contribu-
tion of the solution, which comes primarily from the terms with k = j = 0,
and uses the following integral approximation
(4.47)
then
(4.48a)
(4.48b)
in which L1 is given by
LI ~ Fo, oP·; -i) Go,oP.; i) + Fo, 0(,1.; i) Go,o(A; -i) (4.49)

Furthermore, in view of the approximation (4.47) ,one has
Fo, 0(,1.; i) = + i~2 I 1(e- itt / 4 Aj ~2)Kl (e- itt / 4 Aj ~2) (4.50a)
Go,o(A; +i) = voFo,o(A; +i) + i~2 {l- 2I 1(e- itt / 4 A/~2)

K 1 (e- itt / 4 AI ~2)} - iA 2 (4.50b)
It is clear now from the above, that the general expressions for the coef-
ficients A o and B o are complicated series expressions involving the modified
Bessel funetions In and Kn. As complicated as they may seem, the use of an
electronic computer makes the work a routine.
* In the field of fraeture mechanies it is only neeessary to eompute the first eoefficients
Ao and Bo for only the first term of the series in equations (4.41) lead to the weil known
[12]I/vr stress singular behavior ahead of the eraek tip.
Determination of w and F. In view of equations (4.27), (4.37) to (4.39) and

the relation
f~ S-Il Ji as) eos (x s) ds

= {.J1t(2a)-1l [reu + 1)r 1 (a 2 - x2 t- t ; 0 < x < a ;Re Il> -1
0; a < x < oo; Re Il > -1
(4.51)
(4. 52a)
(4.52b)
(4.52e)
(4.52d)
Therefore a substitution of the above relations into equations (4.21) wilI
determine the bending defleetion w and membrane stress funetion F.
Furthermore, the eorresponding integrals wilI converge and the differentia-
tions under the integral sign are al so justified at lea st for y =I- O. The values
of the derivatives at y = 0 and lxi < 1 ean be obtained by a proper limiting
process.
The Stress Field. Without going into the detail s, the stress distribution
around the eraek tip for a symmetriealloading* is found to be:
Extensional stresses: through the thickness

k(e)
(J~e) = (2~}t eos (Oj2) [1 - sin (Oj2) sin (30j2)] (4.53a)
k(e)
(J~e) = (2~)f eos (Om [1 + sin (Oj2) sin (30j2] (4.53b)
* The antisymmetric Joading case will be discussed Jater.

136 E. S. Folias
k(e)
(J~~ = (2:)t eos (012) sin (012) eos (3012) (4.53e)
Bending stresses: on the 'tension side' of the shell

k(b)
(Jt;) = - (2~)"f K[3 eos (012) + eos (5012)] (4.54a)
k~b) [
(Jy(b) = (2r)+ K eos (5012) + 111 + 5v
_ v eos (012)
J (4.54b)
(Jt;J = (~~~+ K[sin (5012) + G: :) sin (O/2)J (4.54e)
where v is Poisson's ratio and (r, 0) are the polar eoordinates around the
eraek tip. In general, the stress intensity faetors kie ) and k\b) are funetions
of eraek size, geometry of the shell, material properties and loading charac-
teristics. In this ease, they are related to the coefficients Aa and Ba by the
expressions
(e) A4 RD
kl = 2hc 4 (Aa + Ba)V c (4.55a)
(b) iEhA 2 I
kl = 4(1 _ V2)C2 (Aa - Ba) (3 + v)~ e (4.55b)
Thus, in view of equations (4.39), the stress intensity factors
k(e) = ij(e) JC{l + 3n A2} + ij(b) I/Jc2) (1 1- v2rt{~ + ~(Jc + In Jc)}

1 32 ~ 3 32 8 4
+ 0().4In Jc) (4.56)
and
k(b) _ _ -(e)
1 - (J
).2(Jt»)C
(1-v 2yt(3+v)
{I +32 + 1+ 3V(
7v
8
I
y+ n 4
3.)}
_ ij(b) ~C{l + 1 + 3v nJc 2} + 0(Jc 4 1n Jc) (4.57)
3 +V 32
are obtained.
The reader should be cautioned for equations (4.56) and (4.57) represent
exaet asymptotic expansion s up to O(Jc 2) terrns. Consequently, they are good
approximations for 0 ::s; )~ < 1. For larger values of )~, one must also indude
Asymptotie approximations to craek problems in shells 137
higher order terms in order to guarantee convergence. This can be done by

either of the two methods discussed previously and the aid of a computer.
For example, the numerieal solution of the system of equation (4.44) leads
to anumerieal result that may weil be approximated within a 5% error by
the simple relation *
kle) ~ a(e) .Je (1 + 0.466 A2yt (4.58)
which is valid for all values of A.
In view of the above, one may conjecture that in an initially curved sheet,
1. the stresses are proportional to (ejr)!,
2. the stresses have the same angular distribution as that of a fiat plate,
3. the stress intensity factors are funetions of the shell geometry and, in
the limit, we recover the fiat plate,
4. the stresses inelude interaction terms for bending and streetching.
A typical term for a spherical shell is
ashell
- - ~ 1
aplate
+ -,
RJ
e2
(
al + e)
b l In (Rl )!
J
+ 0 (If11 ) (4.59)
where the expression inside the parenthesis is a positive quantity. One con-
eludes, therefore, that a spherical initial curvature, in reference to that of a
fiat sheet, is to increase the stresses in the neighborhood of the crack tip and,
as aresult, reduce its resistance to fracture initiation.
It should be emphasized that elassical bending theory has been used in
deducing the foregoiog results. Henee it is inhereot that only the Kirehoff
equivalent shear free condition is satisfied along the crack [15], and not the
vanishing of both individual shearing stresses. While outside the local region
the stress distribution should be aecurate, one might expect the same type of
discrepancy to exist near the crack point as that found by Knowles and
Wang [16] in eomparing Kirehoff and Reissner bending results for a fiat
plate. In this case the order of the stress singularity remained unchanged but
the angular distribution around the crack changed so as to precisely be the
same as that due to solely extensionalloading.
Recently, Sih and Hagendorf [17] investigated this matter further by
deriving an improved theory of shallow shelis which incorporates the effeet
of a transverse shear deformation. As expected, their results** showed that
* For the predietion of failures in pressurized vesseis [14] the contributions of k\b) are
negligible in comparison to those of k\e).
** See chapter 6.
138 E. S. Folias
elassie theory eannot adequately prediet the exaet angular dependenee of

the bending stresses in the vieinity of a eraek. However, in general, these
bending stresses are so small when eompared to the extensional stresses that
ean be negleeted. On the other hand, for very long eraeks sueh eontributions
beeome signifieant and eonsequentIy may no longer be negleeted. Unfortu-
nately, in sueh eases bulging effeets beeome extremely important and any
theory, whether elassie or shear, is inadequate.
In-plane Shear Load. If on the other hand, the residual loads, i.e., equation
(4.16), are of the form
M lP ) = 0 V(P) = 0 N(P) = 0 N(P) = -~ (4.60)

y 'Y 'y 'xy e2
then the solution ean be eonstrueted in a similar manner [10] and the results
are
Extensional stresses: through the thiekness:

keel
(J~e) = - (2;)} sin (0(2) [2 + eos «(J(2) eos (30/2)] (4.61a)
k(e)
(J~e) = (2;)+ sin (8(2) eos (0/2) eos (38m (4.61b)
k(e)
(J~~ = (2;)"l: eos (8/2) [1 - sin (Oj2) sin (38(2)] (4.61e)
Bending stresses: on the 'tension side' of the sheli:
(J~b) = (~;;! {sin (50;2) - (91~ vV) sin

7 «(Jm] (4.62a)
ktb)
(J~b) = (2~)l: K[sin (0/2) - sin (58(2)J (4.62b)
(Jlb)
xy
ktb)
= ~
(2r)2 1- v
5)
-+- eos (Oj2) - eos (50/2)
K[(3V ] (4.62e)
where the stress intensity faetors kie ) and k~b) are given by
(4.63a)
ktb) = _ -(e) ,j3(3 + v) A2{7 + V- 4(5 - V)y _ 5 - v In A

2 T 2(I-v2r~ 16(1-v) 4(1-v) 4
+ O(A 2 In A)} (4.63b)
with
(4.64)
In the limit, as J. ~ 0 one reeovers preeisely the results of referenees [18]

and [I5]. Here again the results are eorreet up to 0(A4 ) terms and as aresult
they are only good approximations for 0 ~ A < 1.
Effect of Transverse Vibrations. In addition to the usual external applied

loads, pressure vesseis are frequently also subjeeted to vibrations. Conse-
quently, an investigation was earried out in order to assess analytieally what
effeet, if any, do vibrations have on the meehanism of fraeture. The analysis
has shown [19] that in general, transverse vibrations reduee the stress
intensity faetof. However, when the foreing frequeney w approaehes the
natural frequeney of the uneraeked sheli, the stress intensity faetor increases
without bound. This phenomenon, eoupled with the usual I/~r singular
behavior, eauses the pressure vessel to fail at nominal values even lower than
the yield stress.
Thus, without going into the mathematieal details [19], the stress intensity
faetors for a residualload * of the form
lvl~P) = - ~ mo eos (wt + 4», V~P) = 0, (4.65a)

C
N(P)
y
no eos (wt + ,+"
= - c2 A.) N(P)
xy
= 0, (4.65b)
are
case (i) A4 ~ 0:
k\e) = ,j c{ :C02+ (3':o~R[ 372 + ~(y + In ~)J+ O(A 41n A)} eos (wt + 4»
(4.66a)
* See equation (4.16).

140 E. S. Folias
k (b l _ {_ ). 4Rn O 2Jc4
1 - 2(1 _ V )C
[1 +
32
7v
+
~~ (
8
1 ).)]
Y+ n 4 +
+ 2~o:/~~~4 + O().4 1n )')} eos (OJt + </J) (4.66b)
case (ii) ).4 < 0:
klel = JC{~~ [1 + ~; ).2] + R(;: v) [372+ ~(y + In ~)]

+ O().41n)')} eos (OJt + </J) (4.67a)
k (b l _{_ ).4Rno JC[I+7V 1+3V( I ~)]

1 - 2(1 _ V 2 )C 4 32 + 8 Y+ n 4 +
moEhJc [ 1T).2 1 + 3V]
_ V2)C4 1 + 32 3 + v
4 }
+ 2(1 + O()' In).) eos (OJt + </J) (4.67b)
where
(4.68)
(JJthe foreing frequeney and p the density of the material.

From these results, the following speciallimiting cases of praetieal interest
ean be examined:
1. If (JJ ~ 0 and R i= oo, the stresses of a non-vibrating eraeked spherieal
shell are reeovered and eoineide with those obtained in [10].
2. If (JJ i= 0 and R ~ oo, we reeover the vibrating eraeked plate expressions
in [20].
3. If (JJ = 0 and R ~ oo, the stresses of a flat sheet are reeovered and
eoincide with those obtained previously for bending [15] and extension
[18].
4. If)' ~ 0, i.e., when the foreing frequency reaches the natural frequeney
(EI p}t (lI R) of the uneraeked sheli, the extensional stress intensity fac-
tor becomes infinite.
As apraetieal matter, it is of some value to compare the dynamie with the
static stress along the line of eraek prolongation. For e = I in., h = 0.1 in.,
R = 32.6 in., v = 1/3, E = 16 X 106 psi, p = 0.315 1bfjin. 3 •
(i) for no i= 0, mo = 0:
[0.67(1 + 0.29A 2 ) - 1.24A 4 ( 0.25 + 0.13 In :~)]

(J Ydynanl.ic eos (wt + </J); A4 < 0
O'Ystatic
[0.67 - 1.24A 2 (0.25 + 0.13 In :~ )]eos(wt + </J);
A4 > 0
(ii) for no = 0, mo =I 0:
[ 0.87(1 + 0.18A 2 ) + 0.14 ( 0.43 + 0.19 In A162 ) ]
O'YdynamiC = eos (wt + </J); A4 < 0
~~)] eos (wt + </J); A

O'Ystatic
[0.87 + 0.14(0.43 + 0.19 In 4 >0
where
A4 = 2.1 X 10- 5 w 2 - 1
The plots of the ratio
I = 0"Ydynande
0"Yst. tle eos (wt + <jJ)
for various values of w are given in Figures 4.3 and 4.4.
10~~----------------------~
08
06
04
02
100 200 300 400 500 600 700

W
Figure 4.3. Ratio of Dynamic and Static Stresses vs w for mo = 0

142 E. S. Folias
o8 ~
06
04
02
100 200 300 400 500
Figure 4.4. Ratia of Dynamic and Static Stresses vs w for no = 0
4.4 The stress field in a cracked plate
The problem of a flat plate containing a finite crack has been investigated
by many authors for various types of loadings and the results are reported in
other chapters of this volume. The solution, however, for an infinite plate
(see Figure 4.5) may al so be obtained from that of a spherical cap by simply
~~~~~~~~~'--x
11111111111
j q(x,y)
z
Figure 4.5. Cracked Plate Subjected to a Lateral Load q
letting R ~ oo or A ~ O. Thus the stress field around the crack tip is given
by equations (4.53) and (4.54) where the stress intensity factors now are
k~e) = ij(e) ~ e (4.69a)

k~b) = _ je
ij(b) .. (4.69b)
4.5 The stress field in a cracked cylindrical shell
For a cylindrical shell, one of the principal radii of curvature is infinite and
the other constant. It appears therefore that this geometric simplicity leads
to rather straightforward analytical solutions. However, the fact that the
curvature varies between zero and a constant as one considers different an-
gular positions - say around the point of a crack which is aligned paralleI
to the cylinder axis-more than obviates the initial geometric simplification
and therefore increases the mathematical complexities considerably. For
this reason, Sechler and Williams [5] suggested an approximate equation,
based upon the behavior of a beam on an e1astic foundation, and were abI e
I l
L..-2_C-j_:....~_I-~~X__-,"';j
TOPVIEW
~h SIDE VIEW
Figure 4.6. Geometry and Coordinates of an Axially Cracked Cylindrical Shell Un der
Uniform Axial Extension N x and Interrrai Pressure qo
144 E. S. Folias
EI-x .; .l 2r TOP VIEW
~--~Y----------~~
SIDE VIEW
Figure 4.7. Geometry and Coordinates of a Peripherally Cracked Cylindrical Shell Under
Uniform Axial Extension N x and Internai Pressure qo
to obtain a reasonable agreement with the experimental results. Subsequently,

using the method of section 2, the author investigated this problem in a more
sophisticated manner and the detaiis for an axial and a peripheral crack
(see Figures 4.6 and 4.7) can be found in references [21] and [22], respectively.
Again, omitting the mathematical details, the stresses around the crack
tip are given also by equations (4.53) and (4.54), where the stress intensity
factors are: *
* It should be emphasized that these results are only valid for small A, and that for large A,
one must consider more terms of the asymptotic expansions. Using the method described
previously on the alternate method of solution, the stress intensity factors have been
determined numerically for v = -! and a(b) = 0 and may weil be approximated within a
6% error and for all A, by the simple relations:
kie ) =yc (1 + 0.317 A,2)'
and
k~e) = yc (1 + 0.05 A,2)! .
(i) for an axial crack (see Figure 6)
k(e) _ -(e) /
1 - (J Y C
{I + 51tA 2 }
64 + (J
-(b) (1 - v2 yt A2 ..jc
..j3(3 + V)
5 + 37v 1 + 5v ( A)} -4
{ 96(1- V) + 16(1- V)' Y + In S + O(A In .1); ;. < 1 (4.70a)
eb) __ -(e) / ..j3A 2 {5 + 37v

1 - (J YC(l_vZ)l 96 +
1 + 5v
16
(1A+ I
I1
~)}
S
_ -(b) / {I _ 1 + 2v + 5v 2 1tA 2 O( 14 I < 1

+
} 1. 1
(J yC (3+v) (l-v) 64 AnA), A (4.70b)
(ii) for a peripheral crack (see Figure 4.7)
k (e) -(e) / {I 1tA 2 } -(b) (1 + V2 )!;,2 ..jC

1 = (J Y C + 64 + (J ..j3 (3+ V)
(I + v)
{ 32(1 _ V)
(1 + v) ( , 4,
+ 16(1 _ V) A + In"8 + 0(;' In A);
A)} A < 2.5 (4.71a)
ktb) __ _(e)..j3 .1 2 ..jC{!

1 - (J (1 _ v2)1
+V
32 +
1+
16
v(Y + I n8
~)} _ (i(b)
3 +V
+ 2v + v2 1t;' 2}
{ 1 - (35 + 4
V)(1 _ v) 64 + 0(.1 In .1); ;. < 2.5 (4.71 b)
(iii) for an arbitrary orientation crack* (see Figure 4.S)
Figure 4.8
k(e) _ -(e) / {I (5 cos 2 rx + sin 2 rx)1t;.2} -(b) (1 - v2 )tA2 ..jc

1 - (J Y C + 64 + (J ..j3 (3 + V)
146 E. S. Folias
{[9~(~ ~7~) + 1~(7 ~Vv) (Y + In Ae~s a)] eos 2 a

+ [32~t~ v) + l;(t~ v) (Y + In A s~n a)] sin 2 a} + O(A 4
111 A); A< 1
(4.72a)
ktb) = _ -(e) .j3 A2 .jC{[5 + 37v 1 + 5v ( I A eos a)] eos 2 a

1 (1 (1 _ v2r~ 96 + 16 Y + 11 8
+ [ 1~+
+ v 16""" A sin
1 + v ( Y + In-
8-
a)] sm. 2}a
(1 + 2v + 5v 2 ) eos 2 a + (5 + 2v + v2 ) sin 2 a
-(1
-rbl /
yC
{I -
(3 + v)(1 - v) n6~}
+ O(A 4 In A); A < 1 (4.72b)
and*
kie ) = ile) .jC {I + .j5~~\in 2a}
-rbl (I - V 2 )"!-A 2 [ 5 + 37v 1 + 5v ( In A eos a)]t
+ r .jc .j3(3 + v) 96(1 - v) + 16(1 - v) Y + 8
(l+v) (1 +v) ( +1 ASina)]\in2a+ O(A 4 111 A)·

[ 32(I-v)+16(1-v) A< 1 (4.73a)
Y n 8 '
(b) -(e) .j3A 2 .jc {[5 + 37v 1 + 5v ( I A eos a)Jt

k2 = r (1 _ v2)t 96 + 16 Y+ n 8
[l~V+l~V(Y+ln ASi;a)T}sin2a
-rbl / { ( 5v o2 - 12vo + 8)"t(v 2 + 2v + 5)+ nA 2 . 2}
+ r yC 1+ (4 _ vo)vo 64 sm a
+ O(A4 In A); A < 1 (4.73b)
4.6 Approximate stress intensity factors for other

shell geometries
Beeause the eomplementary or perturbed solution presents eontributions
only in the immediate vieinity of the eraek tip, one may eonsider-at least
* The reader should note that the angular distribution here is given by equations (4.61)
and (4.62).
Ioeally - the prineipaI radii or eurvatures eonstant. Thus, assuming that the
eraek is paralleI to one of the prineipaI axes, say aIong the x-axis, one may
hypothesize that the stress intensity faetors depend primariIy on the eurva-
tures that one observes as he traveIs aIong and perpendieuIar to the eraek
proIongation. Consequently, one may estimate the stress intensity faetors by
a proper superposition of the resuIts of an axiaI and a peripheraI crack in a
eyIindrieaI sheI!. In partieuIar, for (i(b) = 0
k(e) ~ (i(e) ..JC {l + 7rA x 2+ 57rA y 2} + O(A4 In A)' A< 1 (4.74a)

1 64 64 '
k(b) '" -(e)

1 - (J
..J3
(1 _ v
..Jc
2)"!"
{5 + 37v A 2 1 + 5v 1 2( I
96 x + 16 Ay Y + n
b:)
8
+ 1 ~ v A/ + 1 ~ v A/(Y + In ~x)} + O(A 4 1nA); A< 1 (4.74b)
In order to eheek the vaIidity of such a superposition we will consider as

our first exampIe* a spherieaI eap the stress intensity factors for whieh we
know exaetly.
Example 1: Sphere. For this shell the eurvature is constant in all direc-
tions; therefore, in view of equations (4.70a) and (4.7Ia), one has
k(e) '" -(e)..J

1 - (J C
{I + 7rA
64
2
+
57rA
64
2} = -(e)
(J
{I + 37rA
32
2}.' A< 1 (4.75)
whieh is identieaI to equation (4.56). SimiIarly,
k (b) -(e) ..J3 A2 ..Jc {5 + 37v 1 + 5v ( I A)

1 ~ -(J (1 _ v2 }t 96 + 96 Y + n 8
+ 13~ v + 1~ v (Y + ln~)}
_ _ -(e)
- (J
..J3 A2 ..Jc
(1 _ v2 )!
{-O.l + 32
5v
+
1 + 3v (
8
1 ~)}.
Y+ n 4 '
A< 1 (4.76)
which agrees fairly well with equation (4.57). One may conclude, therefore,
that such a hypothesis may not be unreasonabIe.
* In the following examples, we have assumed u(b) = o.

148 E. S. Folias
Rl = R- e tan E
R2 = R + e tan E
Figure 4.9. Conical Circular Shell
Example 2: Circular conieal shell (see Figure 4.9). In this case, one curva-
ture is infinite, the other finite; therefore,
(i) for an axial erack:
k~e) =::. ij(e) .JC{l + ~;Ai}; Al < 1 (4.77a)
k~e) =::. ij(e) .JC {l + ~: A~}; A2 < 1 (4.77b)
(ii) for a peripheral erack:
k(e)
1-
'" ij(e) IC{1 + 643'
'v
~A2}. A3 < 1 (4.78)
where
2
2 2 ]~ C
Al = [12(1 - v ) 2 -(R---c-ta-n-e~)h~
2
2
A2 =[12(1 - 2 ]~
v) 2
C
(-R-+-c-t-an~e)~h
A~ =[12(1 (4.79)
Figure 4.10. Toroidal Shell
Example 3: Toroidal shell (see Figure 4.10). For an axial crack in the outer
surface
(4.80)
for an axial crack in the inner surface
k (e)
1 ~
-(e)
(J V C
I{l + 645rr /\, 2
1
1 - 64
1 2
rr /\, 3 },. "AI. 3 <1 (4.81)
for a peripheral crack in the outer surface
k (e)
I
~
_(J
-(e) I
yC
{I + 64 5rr"A22 rr"lel2}.
+ 64 ,
1
/\'1.2<
1 (4.82)
and for a peripheral crack in the inner surface
k (e)
1 ~
-(e)
(J V
I'{l
e - 5rr'2
64 le 3
rr 1 2 } .
+ 64 /\, 1 , 1
/lI. 3 < 1 (4.83)
4.7 Plates on elastic foundations
Analyses of plates resting on foundations usually fall into two groups. The
first group follows the weil known theory of Winkler and Zimmerman [23]
in which the elastic foundation is considered as a system of separate uncon-
nected springs. Such a hypothesis simplifies considerably the analysis of
150 E. S. Folias
structures on elastic foundations and lead s frequently to incorreet results.

The second group follows the theory in which one describes the physical
properties of the natural foundation more accurately by the hypothesis that
the foundation is an elastic isotropic semi-infinite space [24]. Here again,
such a hypothesis leads to cumbersome calculations and therefore the method
becomes impractical.
Recently a new theory based on Vlasov's general variational method [25]
has been proposed [26]. This theory considers the elastic foundation as a
single or double layer model whose properties are described by two or more
generalized elastic characteristics. The advantage of this theory is that it is
more accurate than the theory of Winkler and Zimmermann and simpler
than the theory of the elastic semi-infinite space.
Winkler-Zimmermann Foundation. The characteristics of the fracture of

plates resting on a Winkler-Zimmermann foundation have been investigated
and the results are reported in references [27] and [28]. In this case, the
governing differential equation for the displacement function w(x, y), with
x and yas dimensionless coordinates (see Figure 4.11), is given in the c1assical
theory by
x TOP VIEW
-I
SIDE VIEW
Figure 4.11
(4.84)
where
(4.85)
and k the elastie foundation spring eonstant. In referenee [27], the author,
using the method deseribed in seetion 4.3, was able to obtain an asymptotie
expansion of the solution for small values of the parameter A. Thus without
going into the details, the stresses at the surfaee z = h(2e are given by equa-
tions (4.54) where the bending stress intensity faetor now is given by
ktb) = ij(b) .jC{l

1
+
(3
3 + 2v
+
+ 3v 2 7L.1. 2
v) (I - v) 32
+[ 3 + 6v
(3 +
+ 15v 2
v) (1 - v)
(.
y + In ±)
+ (3 +
2
3v -1
v) (1 _ v)
J A4}-1
128 + O(A 6
In A); A< 2 (4.86)
On the other hand, for large values of the parameter A referenee [28] gives
k~b) = ij(b).jC ~A + O(r 3/2 ); A> 4 (4.87)
For example, along the eraek prolongation and for v = 1(3,
[1 + ~ ~22J ;
.jc 1
A< 1
(2r)!
5 32
(4.88)
2>4
Sinee the behavior of the stress intensity faetor at the two extremes is
known, one may eonstruet a eurve with the proper asymptotes. Such a plot
is given in Figure 4.12. One concludes, therefore, that the general effeet of
an elastie foundation is to deerease the stresses in magnitude by a faetor
whieh depends on the type of foundation, the eraek Iength, and the material
properties.
Single Layer Foundation. Following [26], the differential equation governing

152 E. S. Folias
the displacement function is w(X, Y) of a plate resting on single layered

elastic foundation (see Figure 4.13).
1.0
~
0.8 '. ,,
"" ...
...
1
b'~
~0.6
Ib..<l
. . . r- ...
... - r--- r---
0.4
0.2
-
o 2 3 4 5 6 7 8 9
STRESS INTENSITY FACTOR VS A
Figure 4.12
.AL
r
I
-T'
L
-: -:: TOP VIEW
h
I
SIDE VIEW
Figure 4.13. A Cracked Plate on a Single Layered Foundation
(4.89)
with the quantities r*, b* and m* as constants defined by*
E* JH
(1 + :~) D 0 1/1 2 (z) dz
E~ JH {'/"(Z)}2 dz
(1 - v~2)D 0 'I'
(lE.!: + ~fH 1/1 2(z) dZ) ~ (4.90)

g g 0 D
and
E*s
1 - v~
v*
s
=
r-=v:Vs (4.91)
D the f1exural rigidity of the plate, 'lp and Ys the specific weights of the plate
and elastie foundation and g the gravitational acceleration.
The problem of a finite crack, of length 2e, in the plate has been investi-
gated and the results are reported in referenee [29]. Thus without going
through the detail s, the stress intensity factor, in view of the definitions
r = cr*, b = cch, k = rjb, (4.92)
beeome
(i) for b < r < 1
kib ) = õ'bv'c(12 - fr 2 ) {12 + {- ~{v(2 - v) (5 - 12y + 121n 2) +

- 2(3 - 4y + 41n 2) + 16v(1 - 2y + 21n 2)] +
1'2
+ }(3v + 1)(1 - v)(1 - In 2 + 21n r)} (1 _ v)(3 + v) +
+ {ö34 [v(2 - v)(5 - 12y + 121n 2)(3 - 4y + 41n 2) - 8v(1 - 2y + 21n 2)]
* It is assumed that no horizontal displaeements oeeur in the elastic foundation and that
the vertieal displaeement is given by a single funetion If/(z). From referenee [26] a typieal
funetion is
where r* is a eoeffieient determining the variation with depth of the displaeement.

154 E. S. Folias
- !z(3v + 1)(1 - v)(1- In2 + 2Inr)} (1- v~(3 +

4 v) + ...
}-1
(4.93)
Notice that <5 does not appear in equation (4.93) for it is negIigible. A
pIot of this is given in Figure 4.14.
1.0
0.8
r::::'
;:. 0.6
+
r<l
:::::
Ib"
.........
'- 0.4
(1....
0.2
CASE I., 8«r<l, lI=I/3
o 0.2 0.4 0.6 0.8 1.0
Figure 4.14. Stress Coefficient Versus r
(ii) for r = k<5 < 1, where k is a real constant
k~b) = ~b.jC { 1 + 8(e _ 1)+ In k 2

1 [k 2
+ (e -
- (e - 1)+
1)+J
(1 + v) + t(3v + 1) (1 - v) (2k 4 - 1) (i2
(1 - v)(3 + v)
1
16(1 _ v) (3 + v) [v(2 - v)(5 - 12y + 12In 2) + (3 - 4y + 4In 2) +
- 8v(1 - 2y + 2In 2) - 2(1 + 3v) (1 - v) (1 - 2In 2)]k 2 <5 2 +
lk 2 ~ 2 3k 2 <5 2 3 [
+4 (In u)<5 +4 _ 3k2<52 32(1 _ v) (3 + v) v(2 - v)
(5 - 12y + 121n 2) + (3 - 4y + 41n 2) - 8v(1 - 2y + 21n 2) - 2(1 + 3v)

1 e(j4 3 k 2 + (k 4 - 1)t
(z- - 21n 2)] 4 _ 3k 2(j2 + 16(e -1)t In k 2 _ (k4 _ 1)+
(1 + v) + t(3v + 1) (1 - v) (2e - 1) k 2 (j4 3k4(j4

(1 + v) (3 + v) 4 - 3k 2(j2 + 32 - 24k 2(j2
-1
(4.94)
In (j + ... }
k=O
0.8
r-;::::"1
,
;::. 0.6
+
r<")
Ib'"
'--'
....... 0.4
0....
0.2
CASEli., r=k8<1, V=I/3
o 0.2 0.4 0.6 0.8 1.0

8
Figure 4.15. Stress Coefficient Versus .5
A plot is given in Figure 4.15. Furthermore, in the limit as k ~ 0 one

recovers the results for a Winkler and Zimmermann foundation.
4.8 Partieular solutions

In general, the actual stress fields will depend upon the contributions of the
particular solutions reflecting the magnitude and distribution of the applied
load. On the other hand, the singular part of the solution, that is the terms
producing infinite elastic stresses at the crack tip, will depend upon the local
156 E. S. Folias
stresses existing along the locus of the crack before it is cut, which of course
are precisely the stresses which must be removed by the particular solutions
described above in order to obtain the stress-free edges as required physically.
Figure 4.16. Pressurized Spherical Cap With Fixed Ends
Clamped Spherieal Sheli. Consider a clamped segment of a shallow spherical

shell of base radius R o and containing at the apex a finite radial crack of
length 2e (see Figure 4.16). The shell is subjected to a uniform internaI pres-
sure qo with radial extension Nr = (qol2)R, and because it is clamped we
require that the displacement and slope vanish at R = R o. For this problem
the residual 'applied bending' and 'applied stretching' loads at the crack
are:*
(4.95)
Closed Cylindrieal Tank. Consider a shallow cylindrical shell containing a

crack of length 2e. The shell is subjected to a uniform internaI pressure qo
with an axial extension N" = (q oR/2), My = 0, far away from crack. For
this problem, if the crack is paralleI to the axis of the cylinder, then
q(e) = (qoR/h), q(b) = 0 and iCe) = 0 (4.96)
If the crack is perpendicular to the axis of the cylinder, then
q(e) = q oR/2h, q(b) = 0 and iCe) = 0 (4.97)
In the event that the crack makes an angle IX with the axis of the cylinder,
then
* For more detaiis see reference [30].

(i(e) = (q oRI4h) (3 + cos 21X), (i(h) = 0, and iCe) = (q oRI4h) sin 21X (4.98)
Infinite Plate. Consider an infinite thin plate containing a crack of length

2e. At infinity the plate is subjected to a uniform extensionalload (Joo and
in-plane shear load Too, then
(4.99)
Reetangular Strip on a Spring Foundation. Consider a rectangular strip,

infinitely long in the x-direction and of finite width b in the y-direction.
Furthermore, let the strip be subjected to a constant moment Mo and zero
shear at y = ±b, and simultaneously subject to a uniform normal loading
qo. Then*
(i(e) = iCe) = °
and
6M o cos (Abl.)2) sinh (Abi .)2) + sin (Abl.)2) cosh (Abl.)2)
h2e2 cosh (Abl.)2) sinh (Abl.)2) + sin (Abl.)2) cos (Abl.)2)
(4.100)
where
(4.101)
Plate on a Single Layered Foundation. Consider an infinite elastic plate

which rests on a single-Iayered elastic foundation and contains a finite,
through the thickness, crack oflength 2e. The plate is subjected to two equal
concentrated lateral static loads of intensity P O with corresponding points
of application (0, L, -h) and (0, -L, -h) (see Figure 4.13). Furthermore, it
will be assumed that L> > e. Then**
-(e) _
(J - Oan
d -Ch) _
(J -
_ 6D P Oe4 {(I +- v)12 [ 1-
h2e2 2nD(A; _ A~) (x2 )!
{(A+K 1 [A+(x 2 + 12)1-J - LK 1 [L(x 2 + 12)!J} +
+ 1: 2++ VI~2 {A!K o[A+(x 2 + 12)!J _ A~Ko[L(x2 + 12)1-J }} (4.102)
* For more detaiis see reference [28].

** See reference [29].
158 E. S. Folias
where
I = L (4.103)
e
Now, since we have al ready assumed that 1 « I, it is easy to see that the
above bending moment (alo ng the crack) is approximately a constant, i.e.,
- Dmolc z . Alternately, as an engineering approximation, one may think of
the quantity ( - DmolcZ) as an upper bound, or lower bound, or even a me an
value of the precise bending moment along the crack in order to obtain an
estimate of the stresses in the vicinity of the crack.
4.9 Discussion
From the above analysis it becomes evident that in an initiaIly curved sheet
the stresses near the crack tip possess the usuall/.}r singular behavior which
is characteristic to two-dimensionaI Iinear elastic crack probIems. Further-
more, the angular distribution around the crack tip is precisely the same as
that of a fiat pIate and that the initial curvature appears only in the stress
intensity factors and it appears in such a way that in the limit as R -+ oo
one recovers the fiat sheet behavior.
A typical term is of the form
Ushell ~ 1 + {~ + aZ + !!1..ln ~ +.!2 In -_c--}t .~

Upl ate Rl Rz Rl (R1h) Rz (Rzh) h
+0 (}i' }~) (4.104)
Thus the general effeet of a positive (negative) initial curvature, in reference

to that of a fiat sheet, is to increase (decrease) the stresses in the neighborhood
of the crack point and reduce (increase) its resistance to fracture initiation.
For a cyIindrical sheIl with an axial crack, for example, equation (4.178)
reads
(4.105)
which correlates fiat sheet behavior with that of initiaIly curved specimens.
In a similar manner, the general effeet of an elastic foundation is to
decrease the magnitude of the stress intensity factor in the neighborhood of
the crack tip and as aresult prevent further fracture. This decrease clearly
depends on the values of the pararneters which characterize the elastic
foundation.
References
[I] Ogibalov, P. M., Dynamies and strength 0/ shelis, translated from Russian, published
for NASA and NSF by the Israel Program for Scientific Translations.
[2] Marguerre, K., Zur theorie der gekrummten platte grosser formanderung, Proe. 5th
Int. Congr. Appi. Meeh. pp. 93-101 (1938).
[3] Reissner, E., On some problems in shell theory, Structural Mechanies, Proceedings
o/the First Symposillm on Naval Strlletllral Mechanies, pp. 11-14, pp. 74-113 (1958).
[4] Reissner, E., A note on membrane and bending stresses in spherical shelis, Soe.
Industr. Appi. Math. 4, pp. 230-240 (1956).
[5] Sechler, E. E. and Williams, M. L., The critical crack length in pressurized monocoque
cylinders, Final Report, GALCIT 96, Calif. Inst. Tech., September 1959. See also:
M. L. Williams, Proe. Craek Propagation Symp., Cranfield, 1, pp. 130 (1961).
[6] Erdelyi, A et al., Tables of integral transforms, Bateman Manuseript Project, McGraw-
Hill, N. Y. (1954).
[7] Ang, D. D. and Williams, M. L., Combined stresses in an orthotropic plate having
a finite craek, J. Appi. Meeh., pp. 372-378 (1961).
[8] Noble, B., The approximate solution of dual integrai equations by Variational
Methods, Proceedings Edinburgh Mathematieal Society, 11, pp. 113-126 (1958-59).
[9] Muskheleshvili, N. 1., Singular integral equations, English translation by J. R. M.
Radok, P. Noordhoff, Ltd., The Netherlands (1953).
[10] Folias, E. S., The stresses in a spherieal shell containing a crack, ARL 64-23, Aerospace
Research Laboratories (1964).
[11] Folias, E. S., The effeet o/initial curvature on eraekedfiat sheets, UTEC DO 68-070,
University of Utah (1968).
[12] Sneddon, I. N. and Lowengrub, M., Craek problems in the classical theory 0/ elasticity,
John Wiley & Sons, Inc., New York (1969).
[13] Kantorovieh, L. V. and Krylov, V. 1., Approximate methods 0/ higher analysis, P.
Noordhoff, The Netherlands (1964).
[14] Folias, E. S., On the predictions of eatastrophie failures in pressurized vesseis, Pros-
peets 0/ Fraeture Mechanies, edited by G. C. Sih, H. C. van Elst and D. Broek,
Noordhoff International Publishing (1974).
[15] Williams, M. L., The bending stress distribution at the base of a stationary crack,
J. Appi. Meeh., pp. 78-82 (1961).
[16] Knowles, J. K. and Wang, N. M., On the bending of an elastic plate containing a
eraek, J. Math. and Phys. 39, pp. 223-236 (1960).
[17] Sih, G. C. and Hagendorf, H. C., A new theory of spherical shelis with eraeks, Thin
Shell Struetures, edited by Y. C. Fung and E. E. Seehler, Prentice-Hall, pp. 519-545
(1974).
160 E. S. Folias
[18] Williams, M. L., On the stress distribution at the base of a stationary crack, J. Appi.
Meeh. (1957).
[19] Do, S. H. and Folias, E. S., On the steady-state transverse vibrations of a cracked
spherical shell, International Journal of Fraeture Mechanies, 7, No. 1, pp. 23-37 (1971).
[20] Folias, E. S., On the steady-state transverse vibrations of a cracked plate, Engineering
Fraeture Meehanies Journal, Vol. 1, No. 2, pp. 363-368 (1968).
[21] Folias, E. S., An axial crack in a pressurized cylindrical sheli, International Journal
of Fraeture Mechanies, 1,2, pp. 104-113 (1965), or The stresses in a cylindrical shell
containing an axial crack, ARL-64-174, Aerospace Research Laboratories, United
States Air Foree, Dayton, Ohio, pp. 1-42 (1964).
[22] Folias, E. c., A circumferential crack in a pressurized cylindrical sheli, International
Journal ofFraeture Mechanies, 3, 1, pp. 1-11 (1967).
[23] Zimmermann, H., Die bereehnung des eisenbahn oberbaues, Berlin: W. Ernst U. Sohn
(1888).
[24] Boussinesq, J., Application des potentiels Il !'etude de l'equilibre et du mouvement des
solides etastiques, Paris: Gauthier-Villars (1888).
[25] Vlasov, V. Z., General theory of shelis and its application in engineering, Moskva
Leningrad, Gostekhizdat (1949).
[26] Vlasov, V. Z. and Leont'ev, U. N., Beams, plates and shelis on elastie foundations,
NASA Technical Translation No. TTF-357 (Aec. No. TT65-50135).
[27] Folias, E. S., On a plate supported by an elastic foundation and containing a finite
crack, International Journal of Fraeture Mechanies, 6, 3 pp. 257-263 (1970).
[28] Ang, D. D., Folias, E. S. and Williams, M. L., The bending stress in a cracked plate
on an elastiefoundation, J. Appi. Mech., 30, pp. 245-251 (1963).
[29] Lin, Si-Tsai, and Folias, E. S., On the fracture of highway pavement, International
Journal of Fraeture, 11, pp. 93-106 (1975).
[30] Folias, E. S., A finite line crack in a pressurized spherical sheli, International Journal
of Fracture Mechanies, 1, 1, pp. 20-46 (1966).
F. Erdogan
5 Crack problems in cylindrical

and spherical shelis
5.1 Introduction
ParticularIy within the past decade the so-called linear fracture mechanics
has established itself as a highly satisfactory working tool in studying the
phenomena of brittle fracture and fatigue crack propagation in structural
solids. The technique appears to be most effective when 'plane strain'
conditions prevail along the existing crack front. It has also been shown that
the stress intensity factor, which is the basic element in the linear fracture
mechanics, is the most appropriate correlation parameter in fatigue crack
propagation studies of relatively thin-walled plates un der membrane loading
where the crack is a through crack, and 'generalized plane stress' conditions
are assumed to exist. The plane assumption here, of course, is an approxi-
mation in which the three-dimensional effects resulting from the intersection
of the crack plane with the stress-free surfaces of the plate are neglected. *
From the view point of practical applications this boundary layer or thick-
ness effect does not appear to be very significant. Therefore, one may be
justified in using standard plate or shell theories for studying the fracture
problem s in thin-walled structures provided the plane of the crack is per-
pendicular to the surface of the sheet.
With the assumption oflinearity, it is known that the relevant information
in crack problem s may be obtained from a local perturbation problem in
which the only externalloads are the crack surface tractions. In 'thin-walled'
structures this would mean that after solving the plate or shell problem under
given external loads by ignoring the crack, the stress intensity factors may
* See Chapter 2 of this volume for the effeet of plate thickness and related approximate
teehniques.
162 F. Erdogan
be found by using the equal and opposite of the membrane and bending
resultants at the location of the crack as the crack surface tractions. Since at
the present time linear problems are the only tractable crack prohIerns, the
geometry of the particular thin-walled structure must then be such that
locally, small deformation plate or linearized shallow shell theories are
applicable. At first sight it may appear that in such cases it is sufficient to
approximate the structure locally by a Rat plate. However, recent studies
have shown that local shell curvatures may have a rather considerable
efIect on the stress intensity factors. Hence, in thin-walled curved structures
the crack problem must be considered in conjunction with a shell rather
than a plate theory.
Because of the peculiarity of the crack problem s in shells, there are analyt-
ical limitations regarding the type of problem s which can be solved by the
existing techniques. Aside from the considerations regarding the linearity of
the problem, the two major limitations arise from the geometry and material
behavior of the shell. The geometrical factors inc1ude the relative size of
the crack with respect to the radii of curvature and dimensions of the sheli,
spatial variation of the curvatures, and the shape and orientation of the crack.
The material factors are primarily the anisotropy and nonhomogeneity. In
addition to linear elasticity, in the existing solutions [e.g., 1-13] it is assumed
that the shell is 'infinitely' large, the curvatures are constant (i.e., the shell
is a circular cylinder or a sphere), the crack is along a principal plane of
curvature, and the material is isotropic and homogeneous. If these assump-
tions are disregarded, mathematically the problem does not seem to be
tractable. Further remarks will be made in this chapter regarding this point.
If the material is isotropic and homogeneous, in applications one could
obtain approximate solutions with an acceptable accuracy by approximating
local shell-crack geometry with an ideal shell which has a solution, namely
a spherical shell with a meridional crack, a cylindrical shell with an axial
crack, or a cylindrical shell with a circumferential crack.
From a practical view point the assumption ofhomogeneity of the material
in shelis does not seem to be a critical restriction. Even in thin-walled struc-
tures made of composites one may easily as sume that the gross behavior of
the material is homogeneous. However, in practice a mild or st rong aniso-
tropy in shells appears to be a rule rather than an exception. Most metallic
shelis are manufactured through rolling or extrusion process, and hence,
are generally mildly anisotropic. Shells which are made of composites such
as fiberglass, boron-epoxy, graphite-epoxy, etc., are of course, strongly
Crack problems in cylindrical and spherical shelis 163
anisotropie. Sinee the treatment of anisotropie or, even orthotropie shelis is

not traetable, it is therefore desirable to have a technique for approximately
evaluating the effeet of material anisotropy on the critical fraeture parameter,
namely, the stress intensity faetor.
This ehapter deseribes a method of solution for the speeialIy orthotropie
shelIs containing a eraek. The method is deseribed by considering symmetrie
and skew-symmetrie problems in eylindrieal shelIs with an axial eraek (for
details see [14-16]). Its extension to the other two ideal geometries seems
to be straightforward. Most of the numerieal results given in the ehapter,
whieh indudes the effeet of Poisson's ratio and interaetion of two eraeks,
is, however, on the isotropie shelIs. The analysis and the results given in
this ehapter are based on an 8th order linearized shalIow shelI theory in
whieh Kirehhoff assumption is made with regard to the transverse shear and
the twisting moment on the eraek surfaee. Sinee there are five traetion
eomponents on the boundary, to satisfy all the boundary eonditions individ-
ually a 10th order theory should be used. * Also, sinee any bending theory is
neeessarily approximate, one would expeet that the sheIl thiekness wilI have
a slight effeet on the membrane eomponent and a eonsiderable effeet on the
bending eomponent of the stress intensity faetor. **
5.2 Formulation of the specially orthotropic cy-

lindrical shell problem
The linearized bending theory of anisotropic shallow shells dates back to a

paper by Ambartsumyan [18] and the detailed treatment of the subjeet may
be found in [19-21]. Referring to Figure 5.1, let an infinitely long orthotropie
Figure 5.1. Geometry of a cylindrical shell with an axial crack
* See Chapter 6 of this volume.

** See Chapter 2 of this volume. See also [17] for cracked plate under bending.
164 F. Erdogan
eireular eylindrieal shell of elastie eonstants El' VI' E 2 , V2' G J2 , thiekness

h, and mean radius R eontain an axial through eraek of length 2a. Assume
that through a proper superposition the problem has been redueed to aloeal
perturbation problem in whieh the eraek surfaee traetions are the only exter-
naI loads. Defining the nondimensional orthogonal eoordinates in the
tangent plane by
Xl = X/a , YI = Y/a (5.1)
The differential equations for the orthotropie eylindrieal shell may be written
as
(5.2a)
(5.2b)
where Xl and X2 are the prineipal direetions of orthotropy take n respeetively

along axial and eireumferential direetions, F is a stress funetion, w is the
displaeement eomponent normal to the shell surfaee, and the operators vi
and vi
are defined by
04 E2 84
vi oxi + 2 [ V2 + 2(1 - VI V2) E::
G 12 ] 04
ox~ox~ + EI ~,
X2
(5.3a)
04 84
vi =
ox I4 + 2 ( 2G
E2
12
-V2
) ~
84
XI X 2
E2
+ E I 8x1· (5.3b)
The notation for the orthotropie elastie eonstants are defined by the follow-
ing stress strain relations:
1
ell = El (0"11 - V I 0"22), (5.4a)
1
e22 = E (0"22 - V 2 0"11), (5.4b)
2
1
el2 = 2G 12 0"12' (5.4e)
~ ~ (5.4d)
El E2
Crack prob/ems in cylindrical and spherica/ shelis 165
The stress and moment resultants are related to F and w through the follow-
ing expressions:
1 jPF
NII = 2a 7)2' (5.5a)
Xl
1 a2F
N zz = 2 ~, (5.5b)
a Xl
1 a2F
N I2 = -
az axlax z
, (5.5c.)
D I ( azw aZw)
MII
-az -;;---y
(iX I + Vz ax~ , (5.5d)
= -
Dz ( al\V c w)
2
M 22
-az ~
(iXl + vI::JT
(XI '
(5.5e)
VI = QI +
aM 12
CX 2
D I [a~
-3
a
3
w
dX I
+ ( Vz +
3
---w-;-
h G 12 )
3
c \V
CXICX~
J
,
( 5.5f)
V2 = Qz +
aM 12
cX I
D
Q3
2 [a 3w
aX~ + (
VI + -
3 IZ )
h G-
3D z
3
a w
-;;-Y;;--
dXldxl
J
, ( 5.5g)
where
Dk = Ek h3 j12(1- VI V2), (k = 1,2). (5.6)

The membrane and bending stress components are obtained from relations
of the form
(5.7)
In solving the problem, for example by expressing the function F and w in

terms of appropriate Fourier integrals, (2) may be reduced to a system of two
fourth order linear ordinary differential equations. The eharaeteristie fune-
tion of this system will be an 8th degree polynomial the coefficients of which
will be funetions of the transform variable. For the problem to be analytie-
ally traetable, it is essential that the roots of the characteristic equation be
obtainable in closed form. For anisotropic shells in general and for ortho-
tropic shelis in particular this does not seem to be possible. In order to
express the roots in closed form the operators vi and must be properlyvi
faetorized. From equations (5.3) it may be seen that these operators can
indeed be factorized and expressed in the following form
166 F. Erdogan
VI
4
=
(iPOX~ + ~
(E 2 /E I )2 OX~
(2
)
=
4
V2 (5.8)
if the elastic constants satisfy the following conditions:
(5.9a)
E2
( 2G - V2
) (Et/E », = 1. (5.9b)
2
12
Now, by direet substitution it may easily be shown that the conditions (9)
are satisfied provided the elastic constants are related by
G _ (E I E 2 Y!:
(5.10)
12 - 2[1 + (V I V2 Y!T
Considering also the relation in equation (5.4d), this means that the sheet
material has only three independent constants. Such a material is said to be
'specially orthotropic'. The analysis and certain results given in this chapter
will the n be valid only for those sheet materials in which the measured value
of G12 and that calculated from equation (5.10) in terms of measured Ei and
Vi' (i = 1,2) are in reasonably good agreement.
If the variables are changed once more as follows
Xl = X, (E 1jE 2)t x2 = y, (5.11)
the operators v1 and Vi become

4
VI = V2 =
4 (0Ox22 + oy2
02 )2 =V 4
. (5.12)
With equation (5.12), equations (2) become identical to the differential

equations for isotropic shelis in which E and D = Eh 3 j[12(1 - v2 )] are
replaced by E 2 and DJ> respectively, i.e.,
02
Ox 2 F(x, y) = 0, (5.13a)
02
ox 2 lI'(x, y) = O. (5.13b)
Let the stress and moment resultants on y = 0, - a < X < a obtained from
the solution of the shell under given externalloads by ignoring the crack be
Ny{X, 0) = N~(X) = no(x) (5.14a)

Crack problems in cylindrical and ~pherical shelis 167
N Xy(X, 0) = N~y(X) = to(x) (5.14b)
My(X, 0) = M~(X) = l11 o(X) (5.14c)
Vy(X, 0) = V~(X) = vo(x) (5.14d)
Considering the perturbation problem and referring to (5.1), (5.5), and (5.11),
the system of differential equations (5.13) must then be solved under the
following boundary conditions specified on the crack surface:
. D Z( Z aZw OZw)
!tm 2 e ~+ VI ~ = moex), (5.15a)
y-±o a oy ox
. 1 aZF
!tm -:::z ~ = - no(x), (5.15b)
y-±o a dx
lim e aZF
y-+O 2;--;::-= to(x), (5.15c)
- a uxoy
where
(5.16)
It is now clear that by properly decomposing the input funetions given by

equations (5.15) into eve n and odd components, the solution of the general
problem may be expressed as the sum ofa 'symmetric' and a 'skew-symmetric'
solution. In the following two sections the solutions of these problems will
be presented in so me detail.
5.3 The skew-symmetric problem
From apraetieal view point the important skew-symmetric problem is that

having the following crack surface tractions:
111 0(X) = 0, no(x) = 0, to(x) = t o( -x), vo(x) = -v o( -x), (-I < x < 1)
(5.17)
Outside the crack, the antisymmetry of the problem and the conditions of
continuity require that
MiX, 0) = 0, N/X, 0) = ° (5.18a)

168 F. Erdogan
. an
hm -;-nw(x, y) =
. an
Ilm -;-nw(x, y), (Il = 0,1,2,3), (S.l8b)
y-++O uy y-+-O uy
lim
y-++O
:nn F(x, y) =
uy
lim
y-+-O
:nn F(x, y),
uy
(n = 0, 1,2, 3), I x I > 1 (S.18e)
Sinee the external loads in equations (5.17) are self-equilibrating 10eaI

traetions, the funetions F and W satisfy the reguIarity eonditions at x = =f oo
and henee may be expressed in terms of Fourier integraIs. Thus using the
symmetry eonsiderations, after some routine manipuIations the solution of
equations (5.13) may be expressed as
w(x,y) = sgn(y) fLOO
o
4
Q/!X)emJ I y I sin !xx d!X (S.19a)
fL
1
OO 4
F(x,y) = sgn(y) KjQj(!X)e mJ I v I sin !xx d!X, (S.19b)
o 1
where the funetions Qi!X), (j = I, ... , 4) are unknown and

Kl = K 2 = - i(E 2hD I)t, K 3 = K 4 = i(E 2hD 1)!
mi = - (!X 2 + i 1A.!X)t, m2 =- (!X 2 - i 1 A.!X)t
m3 = - (!X 2 + i 2 A.!X)!, m4 =- (!X 2 - i 2 A.!X)!

4
.
'l=e"i/4.
,12 =e -"i/4,4
,II. =
12(E 2lE 1) (1 a
-VIV2)~J2 (5.20)
Rl
Substituting from equations (5.19) into equations (5. ISa), (S.I5b), (5.17a),
(5.17b), and (5.18a) it is found that
1-
2
Q3 = [!X(Vi2A.C2 c ) + lJ (Ql + Q2) -
2" "2i (Ql - Q2) (S.21a)
2 1-
Q4 = - [!X(Vi2A.C2 c ) - lJ (Ql + Q2) +
2"
i
2 (Ql - Q2) (S.21b)
The two remaining equations to determine Qj (j = I, ... ,4) are obtained

from the mixed boundary eonditions in equations (5. ISe), (S.ISd), (S.I8e)
and (5.18d). Sinee W and F are odd funetions of y, equations (5.18e) and
(S.l8d) are automatieally satisfied for n = I and n = 3. Using equations
(5.21) and (5.19) it may be shown that the eonditions in equations (S.I8e)
and (S.18d) for n = 0 and n = 2 will be satisfied if
Crack problen1S in cylindrical and spherical shelis 169
f; (Ql + Q2) sin rt.x drt. = 0 (5.22a)
f: (Ql + Q2)rt. 2 sin rt.x drt. = 0 (5.22b)
J OO
o
Vl 2 2
e (Ql + Q2)rt. 2 sin rt.x drt. + J: i 1 A(Ql - Q2)rt. sin rt.x drt. = 0,
(I x I > 1) (5.22e)
Here equations (5.22a) and (5.22b) refer to the conditions that w and
vanish on y = 0, I x I > I. Since analytically equation (5.22b)
(jZW/ cy 2
follows from equation (5.22a), equations (5.22) is actually equivalent to
only two independent conditions. For dimensional consistency these con-
ditions will be selected as follows:
f: (Ql + Q2)rt. 2 sin rt.x drt. = 0 (5.23a)
f: i 1 A(Ql - Q2)rt. sin rt.x drt. = 0, Ix I > 1 (5.23b)
With the selection of equation (5.23) as the conditions for i x I > I it should
be noted that a single-valuedness condition i.e., Il' = 0 for y = 0, I x I > I
stiil remains to be satisfied. This condition will be necessary to obtain a
unique solution for the resulting integral equations.
Substituting from equations (5.19) into equation (5.15c) and (5.15d), and
again, for dimensional consistency, integrating equation (5.15d), it is found
that
lim [ - e2 JOC!
Y-+ +0 a 0
i
1
KjmjQje m1Y rt. eos rt.x drt.] = - to (x), (5.24a)
lim JX {- D;fOC! i [e 3m; - rt. 2em j (v l + h33G12)JQjemlY sin rt.x drt.}dX
J:
Y-+ +0 0 a 0 1 D 2
= - vo(x) dx, (I x I < 1) (5.24b)
With equations (5.21), (5.23) and (5.24) give a system of dual integral equa-
tions to determine QI and Q2' Define now the following auxiliary functions:
ul(x) = J: iIArt.(QI - Q2) sin rt.x drt. (5.25a)
(5.25b)
N ote that U 1 and u 2 are related to the second derivati yes of w and F and hence
170 F. Erdogan
are expected to have the same type of singularity as Nij and Mij at the crack
tips (x = =F I, y = 0). From equations (5.25) and (5.23) it follows that
(5.26a)
QI + Qz = -2z
na
Jl
0
uz(t) sin at dt (5.26b)
Substituting now from equations (5.26) and (5.21) into equations (5.24) and
observing that U I and U 2 are odd functions, the following integral equations
are obtained to determine U I and U 2 :
lim -
y~+o
1
n
JI-1
Iz
I
hi;Cx, t, y)u/t)dt =!,(x), (i = 1,2, I x I< I), (5.27)
where
iaZto(x) a 3 fX
!I(X) e(EzhD I}!:' f2(X) = D2 0 vo(x) dx (5.28)
Il 1j(x, t, y) = J: F 1j(a, y) sin a(t - x)det, (j = 1,2) (5.29a)
F ll (!X,y) = l[_ :z,

2 I,A
/I, - !X/l2 - :2,
12 A
11 3 - !X11 4 ] (5.29b)
(2a 2c vI - e2 )
F 12 (a, y) = 2"1 [ - .
1 1 ),11 1 - etl12 + i 2 ),e 2 + 12.A) 11 3
+ Cet(V\; e
2) + et )11 4 ] (5.29c)
h 2j (x, t,a) = f: i
F 2 et , y) esin a(t - x) - sin ett]det, (j = 1,2) (5.30a)
F 21 (et,y) = ~ [(e 2- V, - 1l;~~2)( i~: 111 - i:~ 11 3)
(5.30b)
Crack problems in cylindrical and spherical shells 171
(5.30e)
em1Y e m2Y emlY em2Y

11 1 =----, n2=-+--'
mi m2 111 1 /11 2
emJY e m4Y e mJY e,"4Y

n3 =-----, n4 = - +-, (5.3\ )
111 3 111 4 111 3 111 4
By examining the asymptotie behavior of the integrands for large and

small values of rt. in equations (5.29) and (5.30) for the kernels hij' it may be
seen that some of the integrals are uniformly eonvergent. I n these integrals
the li mit ean be put under the integral sign and the resulting kernels are
simple Fredholm kernels. In the expression sinexp (p = t - x or p = t) as
p ~ 0 it may also be seen that the remaining integrals beeome divergent,
meaning that the kernels contain singularities. These singular parts of the
kernels may be separated in a standard way by adding and subtraeting the
asymptotie value of the integrand for large values of C/.. For example, noting
that for large ex
(5.32)
f:
it is seen that
ex 2 fl l Cex, y) sin ap dex
= ilA.J: e-aYsin ap dex + f: [ex 2 fl l (ex, y) - iIAe- ay ] sin exp dex
=ilA
p
2 p
+Y
2+ Joo0 [(e-111,- - - -me x
mIY
e
m2Y
2
) 2 J
-i I Ae- aY sinrt.pdex (5.33)
where the last integral is uniformly eonvergent for all p and y ~ 0, and henee,
the limit y = 0 ean be put under the integral sign, whereas the integrated
term gives a Cauchy type singularity I/p. Similarly
J~ exnzCrt., y) sin Cip dex

=
p
2 2P
+Y
2+ Joo
0
[(elillY
/11,
e m1Y )
--+-ex-2e- ay
111 2
J sinrt.pdex. ( 5.34)
Thus, after separating the singular part s of the kernels and going to limit,
equation (5.27) becomes
Jl 2
-1
dt
Iaijuj(t)--
, t - X
+
Jl Ikij(x,t)u/t)dt
2
-1 1
= n/;(x) (lxi < 1), (5.35)
172 F. Erdogan
where
a 22 = - 3v + e + (1 + ~i )(v
I Z I + ~~12), (5.36)
and the Fredholm kernels k ij(x, t) are given by
klj(x, t) = f: [FliC(,O) - alJ sin C(t - x)dC(, (j = 1,2), (5.37)
and
k 21 (x, t) = f: F 21 (C(, 0) [sin C((t - x) - sin C(t]da, (5.38a)
kzz(x, t) = - a 22
t
+fro [F
0
22 (C(,O) - a22]sin C(t - x) - sin C(t]dC(. (5.38b)
Since U I and U z are related to the second derivatives of F and w, the

elements of the fundamental matrix of the singular integral equations (5.35)
will be (I - xz) - t and the solution will be of the form
(5.39)
where the functions Gj are bounded in ( - 1 ~ x ~ 1). Thus the index of the
system in equation (5.35) is K = I, and hence theoretically the solution is
not unique and will contain two arbitrary constants [22]. These constants
are determined by using the single-valuedness condition mentioned earlier,
namely that w(x,O) = °
for I x I > I. Referring to equations (5.19), (5.21),
° °
(5.22) and (5.25) it may be shown that 2u z(x) = - (oz/ox 2 )w(x,0). Since
uz(x) = for I x I > 1, it then follows that for w(x,O) = for
I x I > 1 uix) must satisfy the following conditions:
f~l uz{x) dx = 0, f~l dx f~l U2(t) dt = 0. (5.40)
The unknown funetions Gl and G2 defined by equation (5.39) may be ob-

tained from equations (5.35) and (5.40) in a straightforward way by using
the technique outIined in [23].
To examine the asymptotic behavior of the stresses around the crack tips
let us as sume that the bounded funetions Gl and Gz are expressed in terms
of the following infinite series:
Craek problems in eylindrieal and spherieal shells 173
oo oo
GI(.x) = I A Il T 21l - 1(X), G 2(x) = I B Il T 21l - I(X), (5.41)
where T k(X) is the Chebyshev polynomial of the first kind and the symmetry
property of ulx) = - ul-x) (j = 1,2) has been used. From equations
(5.26), (5.39) and (5.41) by using the relation [24]
J~T21l+1(X)(1- x 2 )-+sinIXxdlX = (_1)" ~ J 211 +1(1X), (n=0,1,2 ... )
it may be shown that (5.42)

oo
i 1AIX(QI - Q2) = Le _1)"-1 A nJ 2n - 1(1X), (5.43a)
1
'"
1X 2(QI + Q2) = I( _1)"-1 B nJ 2n - 1(1X). (5.43b)
1
The expressions for the stresses may the n be obtained by substituting from
equations (5.43), (5.21), (5.19) into (5.5) and (5.7). For example,
(j~y= _,~ uXUY

~2~ Ja
= ,e 2
Ja
fooIKjQ/IX)rnpem/coSIXXdlX,(y;;:::O).
0 1
(5.44)
At (y = 0, x = ±I), the integrals in (5.44) are divergent, meaning that the
stresses will have a singularity at the crack tips. Noting that the integrand
in equation (5.44) is integrable around IX = 0 and is bounded and continuous
elsewhere in the domain, the divergent behavior of the integral must be due
to the asymptotic behavior of the integrand for large values of IX. Thus,
substituting from equation (5.43) and (5.21), equation (5.44) may be expressed
as m e .
(jxy = ha2 l(E 2 I1D 1 )2
i ~
f (-1)11-1 ( All - ----cz-e B
VJ - 2
n
)
x J~ J 2n - 1(1X) [- I + IXY + 0(1X- 1)]e- ay cos IXX dlX (y 20). (5.45)
Noting that for large values of IX [24]
J2n_l(IX)~(-1)n-1JI(IX)~ (~lXr (_1)n-l X

X [cos (IX - ~7r) + o(1X- 1)J (n = 1,2, .. .), (5.46)
and using the results in [24] to evaluate the integrals, we obtain the leading
term in (5.45) as follows:
174 F. Erdogan
CT~/r, (J) = h~2 (E 2hD t )! ~ ( - An + Vt ; ; e2 En)
X 4(;r)! (3 i +
eos eos 5~) + D(r!), (5.47)
where (r, (J) are the polar eoordinates measured from the eraek tip,
(x = 1, Y = 0), r 2 = (x - 1)2 + y2, tan (J = y/(x - 1).
For example, if
to(x) = No, vo(x) = 0,

defining the following normalized funetions (see equation (5.28»;
G;(x) = Gix)/uo, (j = 1,2), (5.48a)

aZN o )"zRNoe
U o = e(E 2 hD j )! = h(EtE z)!' (5.48b)
equation (5.46) may be expressed as
CTmxy =
(No.Ja)[.~(
-h- I f - an + vt-e
eZ
Z
1
bn) ] 4(2ra)f ( 3 eos "2(J 5(J)
+ eos 2
(5.49)
Observing that the stress intensity faetor in a fiat plate under uniform shear
stress Nolh and that in a shell are defined by
kp = Noatlh, k';' = !im [2(x - l)a]tCTx;(x, 0), (5.50)

x->t
from equation (5.49) the membrane eomponent of the stress intensity faetor
ratio for the shell is found to be
(5.51)
Further, noting that Tn(l) = I, (n = 0, 1,2, ... ) from equation (5.48) and
Crack problems in cylindrical and spherical shells 17S
(S.SO) it follows that
(S.S2)
The remaining membrane stresses may be obtained in a similar way. Thus,

for small values of r the membrane stresses in the shell may be expressed as
a~x(r, 0) =
Cmk p
4e(2rayf (7 sin ~ + sin S;) + O(r+), (S.S3a)
a~k, 0) =
Cmk p
4(2ra)t
( . 0
-SIn 2 + sin S~) + 0(1"1), (S.S3b)
a~y(r, 0) =
Cmk p
4(2rayf (3 eos ~ + SO)
eosT + 0(/"').
-" (S.S3e)
Defining now the bending component of the stress intensity factor by
k: = lim [2(x - l)aJ~ a;/x, 0,11) = Cbk p , (S.S4)

x-->!
in a similar way the asymptotic expressions for the bending stresses around
the erack tip may be obtained as follows:
(S.SSa)
(S.SSb)
176 F. Erdogan
where the bending component of the stress intensity ratio is found to be*:
e
b
= _ {3[1 -
1+
):]}+
(V I V 2
(V I V2 )2
(2 + VI -
e2
e2 ) G!(I). (5.56)
Thus, onee the singular integral equations (5.35) are solved after normaliza-
tion deseribed by equations (5.48), the stress intensity faetors ean be obtained
without further analysis. The analysis in this section remains valid for the
isotropie shell with EI = E 2 = E, VI = V2 = V, G 12 = G, and e = (EdE2 )!
=1.
5.4 The symmetric problem
Consider now the symmetrie problem in whieh the only external loads are
the following eraek surfaee traetions (see equations (5.14»:
mo(x) = moe -x), no(x) = no( -x), to(x) =0
vo(x) = 0, (-1 < x < 1) (5.57)
In addition to the boundary eonditions speeified by equations (5.14), (5.15),

and (5.57) on the eraek surfaee (-1 < x < 1, Y = 0), outside the eraek
(I x I > 1, Y = 0) the symmetry and continuity considerations require that
(18e, d) and the following eonditions be satisfied:
Nxy(X, 0) = nx/x, 0) = 0, Vy(X, 0) = v/x, 0) = 0 (I x I> 1) (5.58)
In this ease, using again the Fourier transforms, the solution of equation
(5.13) may be expressed as
w(x, y) = JL
OO
o
4
Rkt.)e IllJ I y I eos ax da, (5.59a)
JL
1
OO 4
F(x, y) = KjRj(a)e mJ I y I eos ax da, (5.59b)
o !
where K j and m j are defined by equations (5.20). Substituting from equations

(5.59) into (5.15), the homogeneous eonditions in equations (5.57e), (5.57d),
and (5.58) give the following two algebraie relations:
m3R3 = (i:~:2 +~) (miR! + +I

m2 R2 ) (m 2 R 2 - miRI), (5.60a)
* See [16) for details.

Crack problems in cylindrical and spherical shells 177
(5.60e)
After some manipulations it ean be shown that the eontinuity eonditions
are satisfied if
f~ (mlR I + mlRl)eoslXxdlX = 0, (5.61a)
JOO
o
•
_11_
A
IX
(mIR I - mlR l ) eos IXX dlX = 0, (lxi> 1). (5.61b)
The remaining boundary eonditions in equations (5.57a), (5.57b) with (5.14),

(5.15), and (5.59) may be expressed as
lim
y-> +0
fX [- -\fOC!
0 a 0
I
I
(clm; - V11X1)R je mJY eos IXX dlXJdX
= - D~
1 JXol11o(x)dx, (5.62a)
(5.62b)
With equations (5.6), (5.61) and (5.62) give a system of dual integral equations
to determine Rl and Rl' In equation (5.62) the integral equations are written
in integrated form to make them dimensionally eonsistent with equations
(5.61) (i.e., the quantities whieh appear in equations (5.61) and (5.62) now
represent the first derivatives of F and IV). Defining
f~ (mIR I + 1111R l ) eos IXX dlX = vl(x), (5.63a)
J OO
o
-
ilA
IX
(mIR I - mlR l ) eos IX dlX = vl(x), 0 .:s; x < oo (5.63b)
and using equation (5.60), and the symmetry eonditions vix) = Vj ( -x),
(J = 1,2), equations (5.61) and (5.62) may be redueed to:
lim
y-4 + 0 Jr
~fl ±
-1 1
gij(x, t, y)vit)dt = p;(x), (i = 1,2), (I x I < 1) (5.64)
Following a proeedure similar to that of the previous seetion to separate the

178 F. Erdogan
singular kernels and going to limit, equation (5.64) may be put into the
following standard form:
JI t
-t t
2 [
_'J_
t-x
J
h·. + lij(x, t) vit)dt = npi(X), (i = 1,2, I x I < 1) (5.65)
where
Pl(X) = i (E:':D t ) J: no(x) dx, P2(X) = ~: f: moCx) dx; (5.66)
(5.67)
[(!:.- + ~ + 2) -
111(x, t) = Joo
o mt m2
- (~+
m
~
m
+ 3 4
2)J sin a(t - x) da, (5.68a)
ldx, t) = J OO [
0
a2
i 1'1.
x sin a(t - x) da, (5.68b)
12t(x, t) = J OO [
o
2 (a
(Vt - e) --
mt
+-
a + -a
m2
+a
m3
- +)
4
m4
(5.68c)
Here m j and constants it, i 2, and Jc are defined by equations (5.20) and Co is
given by equation (5.60c).
From equations (5.63) and (5.59) it may be seen that physically the
quantities VI and V 2 eorrespond to the first derivatives of F and w. Therefore,
the elements of the fundamental matrix of the system of singular integral
equations (5.65) will be (I - x 2 )+ and the solution will be of the following
form:
(5.69)
where H j is bounded in' - 1 ::; x ::; 1. Thus the index of the system in equa-
tions (5.65) is K = - I, and the re are no additional conditions (other than
the consisteney conditions of the singular integral equations) necessary for
a unique solution [22]. Note, again, that equations (5.65) are complex and
are equivalent to four real integral equations which may be solved numeri-
eally in a simple way by using the technique outlined in [23].
To examine the stress state around the crack tips let the functions H j be
expressed in terms of the following infinite series:
oo oo
Hl(x) = L A n U 2ix), Hz(x) = L B n U 2n(x), (5.70)
o 0
where U k(X) is the Chebyshev polynomial of the second kind and the
symmetry property Vj(x) = vi -x) has been used. By using the relation [24]
t U 2it)(1 - t 2 )!eoso:tdt = (-1)" ~ (2n + 1) ~ J Zn + 1(0:),

(n = 0, 1,2, ... ), (5.71)
from equations (5.63), (5.69), and (5.70) it follows that
oo 1
mlR l + mzR z = L:(-I)"(2n + I)An-JZn+l(O:), (5.72a)
o 0:
(5.72b)
Substituting into the stress expressions from equations (5.72), (5.60), and
(5.59) and omitting the detaiis it may now be shown that*
m
O"yy(r, 0) =
i(EzhD1)t ~
4ha 2 · 7t
(2/1 + 1) ? (co An - Bn
)
X (2!)! (5 eos ~ - eos 5~) + O(r!) (5.73)
* See [16] for the evaluation of the related integrals.

180 F. Erdogan
where (r, 8) are the polar coordinates at the crack tip defined by
r2 = (x - 1)2 + y2, tan e = y/(x - 1).
Defining the membrane component of the stress intensity factor in the shell
by
k: = lim [2(x - l)a]t (J"~(x, 0), (5.74)

X"" 1
and observing that U2n(l) = 2n + 1, k~ is found to be
km =
s
i(E 2 hD 1 a)t [eo H (1) - H
ha2 e2 1 2'
(1)J (5.75)
For example, if no(x) = No = constant, mo(x) = 0, and the corresponding

plate stress intensity faetor is defined by kp = (No.Ja)/h, the membrane
component of the stress intensity factor ratio beeomes
(5.76)
where
(5.77)
the funetions Hj are obtained from equations (5.65) after the normalization
given by equation (5.77).
The reamining stress eomponents may be obtained in a similar way. The

asymptotic stress state in the neighborhood of the crack tip may the n be
expressed as
1( 5 cos -20
-
4
- eos -2 58) + O(r'), Jc
(5.78a)
(ixx
m (
r,
8)
= Am
kp
(2raY~
2
-e
4
( 3 cos -8
2
+ eos 2 58) + 0(r , 2 ), (5.78b)
4"e ( sm"2
. 8 . 2
- sm 50) + O( r',') (5.78c)
Craek problems in eylindrieal and spherieal shells 181
VoV
c/ eos 258] + O(r , T
), (5.79a)
b _
(Jxx(r, 8, Z) - Ab (2ra)t
~_ 2Z
Jz
[(
8+ "2
5v
- 8v ez + 3vzv )
c c eos "28
Vc z 58] + 0(r -'- (5.79b)

- 4e z (1 - vze ) eos"2 2
),
b _ ~ ~ a Z Jz 3 G 12 Vc .58
(Jx/r, 8, Z) - Ab (2ra)t h 12D 1 [
? sm 2
- (8 + ;~) sin ~] + O(rt ), (5.7ge)
where
Vo = ez - VI' e = (Et/Ez)"ic, Vc = e z -(VI (5.80)
and the bending component of the stress intensity ratio is found to be
b.
Ab = ks/k 1. [ ]* b
p = klIm 2(x - l)a - (jyiX, 0, h)
px-l
-_2_
1z [12(1-
1 V1Vz)]t[(e oz _ 2)( e z _ VI ) + 2Co ]H*(I)
1 . (5.81)
e -V1Vz e
By Ietting El = E 2 = E, VI = v2 = V, G12 = G, and e = the results of

this seetion too reduee to the soIution of isotropie eylindrieal shelI.
The 'bulging' of the sheIl in the neighborhood of the eraek, i.e., w may be
directly evaluated in terms of the solution given in this seetion. Also, in the
present as weIl in the two other ideal sheIl geometries (that is, in the eylin-
drieaI sheIl with an axial or a eireumferentiaI erack, and in the spherical sheIl
with a meridionai craek) it ean be shown that the auxiliary funetions defined
to reduee the problem to singular integral equations are directly related to
the eraek surfaee displaeements. For the isotropie shelIs these displaeements
are abtained and presented for variaus values of the shelI parameter ). in [8].
The 8-dependenee in the asymptotie stress expressions given by equations
(5.53), (5.55), (5.78), and (5.79) is identieal to the expressions for isotropie
shelIs. However, note that the dimensionless eoordinates r, 8, x, and y in the
speeialIy orthotropie shelIs are defined by
r 2 = (x - 1)2 + y2, tan 8 = y/(x - 1), x = X/a, y = e Y/a (5.82)
182 F. Erdogan
where X, Y, Z are the aetual reetangular eoordinates and the aetual geometrie
angle 0 in the shell is given by
tan0=Y/(X-a)= Y 1)= el tane. (5.83)

e(x -
Therefore, beeause of equation (5.83), the angular variation of the asymptotie
stresses in the speeially orthotropie shelIs is different and a good deal more
eomplieated than that in isotropie shelis.
The analysis given in this and the previous seetions indieates that, sinee
the roots of the eharaeteristie equation mj' (j = I, ... ,4), shown in equa-
tions (5.19) and (5.20) are funetions of the transform variable ex, (0 < ex < oo)
mathematieally the problem would have been intraetable if mj(ex) were not
evaluated in cIosed form. This is essential for extraeting the singular parts
of the kernels of the resulting integral equations as weil as for studying and
obtaining the eorreet singular behavior of the solution. The analysis also
shows that this critical aspeet of the problem relating to the singular nature
of the integral equations and their solution is entirely dependent on the
asymptotie behavior of eertain funetions for large values of ex (see, for exam-
ple, equations (5.29) to (5.34)). The variable ex appears in these funetions
explieitly as weil as through mj(ex). In the equation whieh determines mj' the
eoeffieients of the eharaeteristie funetion, whieh is an 8th degree polynomial,
are funetions of ex. Therefore, for the problems in whieh the roots miex)
eannot be expressed in cIosed form, it appears that ifthe asymptotie solution
of the eharaeteristie equation giving mj(ex) for large values of ex ean be
obtained eorreetly in cIosed form, then the singular parts of the kernels ean
be separated and the singular nature of the solution ean be studied. Further-
more, by also evaluating m/ex) for small values of rJ. in cIosed form and for
intermediate diserete values of ex numerieally, at least in prineiple, it is possible
to evaluate the Fredholm kernels in the integral equations numerieally and,
at the co st of a rather high eomputational effort, to obtain a meaningful
approximate solution.
5.5 Results for a speeially arthatropie eylindrieal

shell
In order to give an idea about the effeet the material orthotropy may have
on the stress intensity faetors in a eylindrieal shell containing a longitudinal
through eraek, in this seetion some numerieal results on eylinders made of
three different materials wiII be presented. These are an isotropic cylinder, a

titanium cyIinder which is mildly orthotropic, and a graphite cylinder which
is strongly orthotropic. The measured elastic constants of the orthotropic
materials are shown in Table 5.1. The table also shows the 'average shear
TABLE 5.1
Elastic constants of the orthotropic materials
Titanium Graphite
El(psi) 1.507 x 10 7 1.5 X 106

E2(psi) 2.08 x 10 7 40 X 10 6
VI 0.1966 0.0075
V2 0.2714 0.2000
GI2 6.78 x 106 4.0 X 10 6
Gav • 7.15 X 106 3.73 X 106
moduIus' caIcuIated from (see equation (5.10))

G _ (E 1 E2 )t
(5.84)
av. - 2[1 + (V 1 V2 )t]'
where EI is the moduius in the axial direction and the notation is given by
equations (5.4). If the measured shear modulus G12 were equal to the caIcu-
lated modulus Gav ., then the material would be speciaIly orthotropic and the
analysis given in the previous sections would be valid without any approxi-
mations. The table indicates that these two values are sufficiently elose so
that the speciaI orthotropy assumption may be used to study the effect of
material orthotropy on the stress intensity factors.
Figures 5.2 to 5.5 show the results for a pressurized sheIl with an axial
crack. The membrane and bending components of the stress intensity factor
ratio Am and Ab shown in the figures are defined by equations (5.74), (5.76)
and (5.81). For the pressurized sheIl the corresponding fiat plate stress
intensity factor is
k PoR (5.85)
P = -h- v(a),
where Po is the internal pressure and the dimensions R, h, aare shown in

Figure 5.1. Generally the results in cylindrical as weIl as spherical shelis are
presented in terms of the dimensionless 'shell parameter' A defined by
184 F. Erdogan
Figure 5.2. Membrane eomponent of the stress intensity faetor ratio Am for a pressurized
Titanium and for an isotropie (v = 1/3) eylinder
Jc = [(12(1 - v 2 )]t al ~(Rh) (5.86)

in isotropie shells, and
(5.87)
in orthotropie sheIls (see equation (5.20)). It is seen that the parameter Jc
in the speeiaIly orthotropie sheIls depends on two elastie eonstants and,
therefore, is not an appropriate eorrelation eoeffieient to be used for the
purpose of eomparing the results in two different sheIls with the same
geometry and different materials. Thus, in Figures 5.2 to 5.5 a purely geo-
metrieal parameter, namely aIJ(Rh) is used as the independent variable.
Also, from the analysis given in the previous seetions it is clear that the
dependenee of the results on elastie constants is not through Jc only. Henee,
the orthotropie results shown in the figures are for the speeifie material
eonstants given in Table 5.1. Similarly, for the isotropie shelis the Poisson's
ratio v appears in the analysis through Jc as weil as elsewherc. The isotropie
sheIl shown in Figures 5.2 to 5.5 under the designation (E 1IE 2 ) = I are thus
-0.8
Figure 5.3. Bending component of the stress intensity factor ratio Ab for a pressurized
Titanium and for an isotropic cylinder
5
Am
1.0
0.0375
5 10
c/./Rh
Figure 5.4. Membrane component of the stress intensity factor ratio Am for a pressurized
Graphite and for an isotropic cylinder
186 F. Erdogan
-I
Figure 5.5. Bending component of the stress intensity factor ratio Ab for a pressurized
Graphite and for an isotropic cylinder
obtained for one value of vonly, namely v = 1/3. The effeet of von the stress
intensity faetors in isotropie shelis is diseussed in the following seetion. In
eaeh figure there are two sets of orthotropie results whieh eorrespond to the
alignment of the stiff direction of the material in the axial or the eireumferen-
tial direction of the eylinders.
The results indieate that in the speeially orthotropie shelis the stress
intensity faetors are strongly dependent on the modulus ratio E t /E 2 , and
generally they inerease with deereasing Ell E 2' Et being the modulus in axial
direction. This does not, of course, neeessarily mean a reduc:ion in the
resistanet. of the shell to erack propagation as the shell beeomes stiffer in
circumferential direction. Any material, particularly a composite, whieh is
not isotropic in elastic properties, would not be expected to be isotropic in
Craek problems in eylindrical and spherical shells 187
3
Cm----T
Figure 5.6. Membrane and bending components of the stress intensity factor ratio, Cm
and Cb for a specially orthotropic (Titanium) and for an isotropic (v = 1/3) cylinder
under torsion
its resistance to crack propagation. In each case the load-bearing strength

of the structure wouId, of course, be decided by the ratio of the stress
intensity factor or whatever the measure of the severity of the externalloads
and the crack geometry to the corresponding strength parameter of the
material.
Figure 5.6 shows the results for a cylinder with an axial crack under skew
symmetric loading. Here it is assumed that the cyli nder is un der torsion and
aw ay from the crack region the uniform shear N xy = No is the only nonzero
stress component aeti ng on the shell. Thus, the corresponding ftat plate stress
188 F. Erdogan
intensity factor is a mode II component given by kp = No~(a)lh. The

membrane and bending stress intensity factor ratios Cm and Ch shown in the
figure are defined by equations (5.50), (5.51), and (5.54). In this example too
a/~(Rh) rather than the shell parameter Jc is used as the independent variable
and for the isotropic case (designated by (EdE2) = I) it is again assumed
that v = 1/3. Figure 6 shows the same trend as Figures 5.2-5.5, namely, the
stress intensity factors increase with decreasing Et/E2 • This appears to be
primarily due to the multiplicative factor (Ed E 2)* in the expression of the
shell parameter Jc given by equation (5.85). In fact for a quick estimate of
the stress intensity ratios in skew-symmetric as weil as in symmetric problems
for the specially orthotropic shells the isotropic results may be sufficient
provided }o is calculated from equation (5.87).
Table 5.2, which show s the results for onlyone value of the variable
a/~(Rh) = 1.66, gives so me idea about the relative effect of material ortho-
tropy. Here the results for graphite, titanium, and an isotropic material
(v = 1/3) are compared. In this case too, the strong influence of material
orthotropy is apparent.
TABLE 5.2
The efj'eet %rthotropy on the stress intensity ratios (al y/(Rh) = 1.66)
Isotropie Titanium Graphite

Material
Ei/ E 2 1.0 1.381 0.724 26.667 0.0375

,\ 3.0 2.811 3.304 1.359 7.018
C III 1.942 1.880 2.044 1.340 4.045
Cb 0.199 0.158 0.239 0.019 1.241
5.6 The effect of Poisson's ratio
As indicated in the previous section, in the isotropic shelIs the Poisson's

ratio v appears in the analysis explicitly as weil as through }. defined by
equation (5.86). This means that the stress intensity factors are functions of
two independent variabies, namely v and a/~(Rh). However, since in most
metallic structural materials v is in the neighborhood of 1/3 and since v
affects the results partly if not mostly through Jo, in practice the tendency
has been to present the resuIts by using only Jo as the independent variable
for a fixed Poisson's ratio, v = 1/3. To justify this or to throw some light on
the approximation involved, the effect of v for some selected values of A or
a/~(Rh) has to be studied.
(a) No # 0, Mo = 0
2 Am (A =3)
0.5
Am
Am (A= I)
~----~----~------~----~----~O
o 0.1 0.2
1/
0.3 0.4 0.5
Figure 5.7. The effect of Poisson's ratio on the stress intensity factors in a pressurized
isotropic cyIinder with an axial crack
(b)Na=O,Ma#O
0.1 f==:~~~~:;;:::===~=:;;:::::J 1.0

Ab (}"=3)
L -_ _ _ _ ~ ____ ~ ______ ~ ____ ~ ____ __JO

o 0.1 0.2 0.3 OA 0.5
1/
Figure 5.8. The effect of Poisson's ratio on the stress intensity factors in an axially cracked
cylindrical shell under uniform bending, M yy = Mo
Figures 5.7 to 5.9 show some resuIts for a cyIindricaI shell with an axiaI
crack. Figure 5.7 show the variation of the symmetric stress intensity factor
ratios Am and Ab for A = I and ). = 3 in a pressurized cyIinder where
190 F. Erdogan
(5.88)
In this case the effect of v on the main stress intensity component Am appears
to be negligible. Figure 5.8 shows some symmetric results for the shell under
cylindrical bending only in which M yy = - Mo is the only nonzero crack
surface loading. For this loading the corresponding flat plate stress intensity
factor is defined by
(5.89)
and Am and Ab are again given by equations (5.88). In this case too the varia-
tion of the main stress intensity component Ab with v for the values of )0 = 1
and Jo = 3 does not appear to be significant. Even though there is a consider-
abI e relative change in Am as v goes from zero to 0.5, it should be observed
that the absolute value of Am itself is rather small.
An example for the skew-symmetric problem is shown in Figure 5.9. Here
it is assumed that a cylinder containing an axial crack is under torsion and
)0 = 5. The related stress intensity factors are defined by equations (5.50),
(5.51), and (5.54). For this Jo value, the effect of v again appears to be
negligible.
It should be noted that in Figures 5.7 to 5.9 Jc is used as a constant para-
4 . 0 , - - - - - - - - - - - - - , 1.6
lA
3.0 1.2
1.0
Cm 2.0 0.8 Cb
0.6
1.0 OA
0.2
0
o 0.1 0.2 0.3 0.4 0.5
v
Figure 5.9. The effeet of Poisson's ratio on the stress intensity faetors in an axially eraeked
isotropic eylinder under torsion (A = 5)
meter. Sinee }. is a deereasing funetion of v, this would eompensate some of

the increases in the stress intensity faetor ratios observed for inereasing v.
A somewhat more meaningful result would be obtained by eomparing the
stress intensity faetor ratios for different Poisson's ratios and a fixed geometrie
parameter aIJ(Rh). A very limited such eomparison for the symmetrie
problem is shown in Table 5.3 whieh leads to the same general eonclusion
that the effeet of v on the stress intensity faetors is not very signifieant.
TABLE 5.3
The effect of Poisson's ratio
No ep 0, Mo = 0 No = 0, Mo ep 0
Rh v Am Am
0.3 0.5 2.157 0.353 0.097 0.810
1/3 2.163 0.364 0.078 0.865
-~-_ .. ~~---
- - - - ._-------- -----
2.63 0.15 2.066 0.352 0.057 0.912

1/3 2.074 0.370 0.076 0.873
2.6 o 2.045 0.326 0.043 0.932
1/3 2.059 0.372 0.076 0.875
5.7 Interaetion of two eraeks
In plane problem s it is known that if the medium eontains more than one
eraek, depending on the relative distanee between the eraeks, there eould be
a strong interaetion between the respeetive stress fields and the stress inten-
sity faetors eould be highly affeeted. In order to give some idea about the
effeet of interaeting stress fields on the stress intensity faetors in shells, in
this seetion the results of a simple problem for a pressurized eylindrieal shell
containing two axial eraeks are presented. From the formulation and the
solution of the eraek problem in shelis given in Seetion 2-4 of this ehapter
it is clear that there is no major diffieulty in formulating the problem and in
deriving the governing system of singular integral equations if the shell
contains, instead of a single eraek, aset of eollinear eraeks. Therefore,
there is no need to present further analytieal details.
In the example under eonsideration the two eraeks are assumed to be equal
in length. The partieular crack geometry and dimensions are shown by the
192 F. Erdogan
'4
~ '--I~-F""'~-1-+l:-G:I!5õ2~=>~ - ~ x
y
A=3
:2
o
'0 1.0
Figure 5.10. The membrane eomponents of the stress intensity faetor ratio in a pressurized
isotropie eylinder (v = 1/3) with two eollinear axial eraeks. A:" for the inner eraek tip,
A~ for the outer eraek tip
insert in Figure 5.10, and the resuits are shown in Figures 5.10 and 5. 11.
The stress intensity factor ratios Am and Ab are again defined by equations
(5.88). The superscripts i and 0 on Am and Ab refer to the inner and outer
crack tips, respectively. The figures show the resuits for J. = I, 2, 3 where J.
is defined by equation (5.86) and v again is assumed to be 1/3. For the pur-
pose of comparison, Figure 5.10 also shows the stress intensity factor ratios
for the fiat plate with the same crack geometry (i.e., for A. = 0) evaluated
from [25]
(5.90a)
0.8
0.4 A= 3
o/e
-0.2
Figure 5.11. The bending eomponents of the stress intensity faetor ratio in a pressurized
eylinder with two axial eraeks. A; for the inner eraek tip, Ag for the outer crack tip
AD _~ __ bi[1 - E(m)/K(m)]
(5.90b)
m- kp - (b t - atHbt(b t + a )/2]+'
j
where
(5.91 )
and K(m) and E(m) are the eomplete elliptie integrals of the first and the
second kind, respeetively.
For alc = 0 the two eraeks are far apart, there is no interaetion, and the
resuIts eorrespond to that of a single eraek in a pressurized eylinder. On
the other hand as alc ...... I, i.e., as the length of the net ligament between the
two eraeks approaehes zero, as expeeted, the stress intensity faetor at the
inner eraek tip goes to infinity and that at the outer tip approaehes the
value obtained for a single eraek of length 4a. However, for alc > 0.4 and
). 2: 2 the results show a somewhat unexpeeted behavior. In f1at plates A~ is
always greater than A; whereas in shelis the results show that for eertain
194 F. Erdogan
ranges of a/c and .Jc it is possible to have A~ < A~. This behavior seems to be
even more pronouneed for the bending stress intensity faetors shown in
Figure 5.11. A parti al explanation of this phenomenon may be found in the
distribution of the displaeement eomponent w(x, y) normal to the shell sur-
faee. In a pressurized isotropie shell containing a single eraek of length 2a,
evaluating w in the plane of the eraek, i.e., for x > 0, Y = 0, one obtains,
for example for .Jc = 2, the resuIt shown in Figure 5.12 (where, in the nota-
0.8
Figure 5.12. Disp!acement component w norma! to the thell surface in the plane of the
crack for a pressurized cylinder with an axial crack, x = X/a
tion of Figures 5.1 and 5.10, w > °inward). The normalization faetor whieh
appears in the figure is given by
POR 2.Jc2
cw = 2Eh' (5.92)
and the eoordinate x is normalized with respeet to a. The figure shows that,
although around the eraek there is an outward bulging in the shell, further
along the x axis w ehanges sign and there is a zone of depression. When the
distanee e is small enough for the stress and displaeement fields of the two
eraeks to interaet, for a eertain range of e this 'depression' may eause a
reduetion in the stress intensity faetors.
5.8 Further results for isotropic shelIs
This seetion presents a summary of the ealeulated results for the three
51- 1"5~ Q
,.,!:l;>;-
~
..,
e
<:J-
~
41- ~Y/ ::!
Ab
'"
«e"
t-. l' S'
~
S'
-
t}
;::;.
Am 31- ,'5/ / // -.......... !:l
-.
!:l
;:::
!:l..
{J
:::-
~
..,;::;.
21- // _,c:tC"~ ("l~
!:l
.....
:::-
'"
~
.....
t:::;-
~\ ~"C~
""q,. 4 C,,""
8 10
>..
Figure 5.13. Membrane component of the stress intensity factor ratio Am in symmetri-
cally-Ioaded shelis: N yy = No # 0, M yy = 0
Figure 5.14. Bending component of the stress intensity factor ratio Ab in symmetri-
'-D
cally-Ioaded shelis: N yy = No # 0, M yy = 0 vl
-
1.2 ..-
\0
0\
0.12 Ab
1.1
1.01<
Am
0.9
0.8
0.7
0.6
4 6 8
A
Figure 5.15. Membrane component of the stress intensity factor ratio Am in shelJs sym- ~
metrically-loaded in bending: M yy = Mo =F 0, Nyy = 0
~
Figure 5.16. Bending component of the stress intensity factor ratio Ab in shelJs sym- ~
metrically-loaded in bending: M yy = Mo =F 0, Nyy = 0 §
4 Q
:::,
~
7 '1:::1
~
t::r-
~
3 ~
6
'"
S·
~
g:
Cm 5 f}
Cb 2 :::,
§
-"'.
4 :::...
{3
;:,..
....
3 :::,
"'"".
~
-
1"":;-
-
2
o 2 4 6
A
II < -=r=,
o 2 4 6 8 10
A
Figure 5.17. Membrane component of the stress intensity factor ratio Crn in shells under
uniform skew-symmetric membrane loading: N.cy = No
Figure 5.18. Bending component of the stress intensity factor ratio C b in shelis under ......
1.0
uniform skew-symmetric membrane loading: N,.y = No -....)
198 F. Erdogan
idealized crack geometries, namely a cylindrical shell with an axial crack,

that with a circumferential crack, and a spherical shell with a meridionaI
crack. The loading condition is assumed to be homoengeous and either sym-
metric or skew-symmetric. The Poisson's ratio of 1/3 is assumed in all
caleulations. The technique used to solve the related shell problem s is sim-
ilar to that described in Sections 5.3 and 5.4 of this chapter [6-9].
Figures 5.13 and 5.14 show, respectively, the membrane and bending,
stress intensity factor ratios Am and Ab for the three shell geometries. In this
symmetric case the only nonzero crack surface traction is assumed to be
N yy = - No = constant, x bei ng the coordinate along the crack. * The
results for the symmetric problem in which M yy = - Mo = constant is the
only nonzero crack surface load are given in Figures 5.15 and 5.16. Finally,
Figures 5.17 and 5.18 show the skew-symmetric results for the three crack
geometries where N xy = - No = constant is the nonzero crack surface
traction. In presenting these results Jc is defined by equation (5.86), Am and
Ab are defined by equations (5.88) (with kp as given by equations (5.89) for
Figures 5.15 and 5.16), and Cm and C b are defined by equations (5.50),
(5.51), and (5.54).
* It should be noted that there was anumerieal error in Ab for the eylindrieal shell with
a circumferential crack given in [7].
References
[1] Copley, L. G. and Sanders, J. L., Jr., A longitudinal crack in a eylindrical shell under
intemal pressure, International Journal of Fraeture Mechanies, 5, pp. 117-131 (1969).
[2] Duncan, M. E. and Sanders, J. L., Jr., The effect of circumferential stiffener on the
stress in a pressurized cylindrical shell with a longitudinal crack, International Journal
of Fraeture Mechanies, 5, pp. 133-155 (1969).
[3] Folias, E. S., An axial crack in a pressurized cylindrical sheli, International Journal
[4] Folias, E. S., A finite line crack in a pressurized spherical sheli, International Journal
[5] Folias, E. S., A circumferential crack in a pressurized cylinder, International Journal
[6] Erdogan, F. and Kibler, J. J., Cylindrical and spherical shelis with eraeks, International
Journal of Fraeture Mechanies, 5, pp. 229-237 (1969).
[7] Erdogan, F. and Ratwani, M., Fatigue and fracture of cylindrical shelis containing
a circumferential crack, International Journal of Fraeture Mechanies, 6, pp. 379-392
(1970).
[8] Erdogan, F. and Ratwani, M., Plastieity and eraek opening displaeement in shelis,
International Journal 0/ Fraeture Mechanies, 8, pp. 413-426 (1972).
[9] Erdogan, F. and Ratwani, M., A cireumferential eraek in a eylindrieal shell under
torsion, International Journal 0/ Fraeture Mechanies, 8, pp. 87-95 (1972).
[10] Murthy, M. Y. Y., Rao, K. P., and Rao, A. K., Stresses around an axial eraek in a
pressurized eylindrieal sheli, International Journal 0/ Fraeture Meehanies, 8, pp.
287-297 (1972).
[11] Murthy, M. Y. Y., Rao, K. P., and Rao, A. K., On the stress problem oflargeelliptieal
eutouts and eraeks in cireular eylindrieal shelis, International Journal 0/ Solids
Struetures, 10, pp. 1243-1269 (1974).
[12] Erdogan, F. and Ratwani, M., A note on the interferenee of two eolIinear eraeks in
a eylindrieal sheli, International Journal 0/ Fraeture, 10, pp. 463-465 (1974).
[13] Erdogan, F. and Ratwani, M., Fraeture of eylindrieal and spherieal shelIs containing
a eraek, Nuclear Engineering and Design, 20, pp. 265-286 (1972).
[14] Yueeoglu, U. and Erdogan, F., A eylindrieal shell with an axial eraek under skew-
symmetrie loading, International Journal 0/ Solids Struetures, 9, pp. 347-362 (1973).
[15] Erdogan, F., Ratwani, M., and Yueeoglu, U., On the effeet of orthotropy in a eraeked
eylindrical sheli, International Journal 0/ Fraeture, 10, pp. 369-374 (1974).
[16] Yueeoglu, U., Ph. D. Dissertation, Lehigh University (1971).
[17] Hartranft, R. J. and Sih, G. C., Effeet of plate thiekness on the bending stress distri-
bution around through eraeks, Journal 0/ Mathematies and Physies, 47, pp. 276-291
(1969).
[18] Ambartsumyan, S. A., On the theory of anisotropie shallow shelis, NACA Tech.
Memo. 1424 (1956).
[19] Ambartsumyan, S. A., Theory of anisotropie shallow shelis, NASA Tech. Transl.
F-118 (1964).
[20] Flügge, W. and Conrad, D. A., Singular solutions in the theory 0/ shallow shelis,
Tech. Report No. 101/ Stanford University (1956).
[21] Apeland, K., Analysis of anisotropie shallow shelis, Acta Teehnica Scandinavica, 22
(1963).
[22] Muskhelishvili, N. 1., Singular Integral Equations, P. Noordhoff, Groningen (1953).
[23] Erdogan, F., Gupta, G. D., and Cook, T. S., Numerieal solution ofsingular integral
equations, in Methods of Analysis and Solutions 0/ Crack Prohlems, G. C. Sih, ed.,
Noordhoff International Publishing, Leyden, The Netherlands (1973).
[24] Gradshteyn, I. S. and Ryzhic, J. M., Tahles o/Integrals, Series and Produets, Academie
Press (1965).
[25] Erdogan, F., On the stress distribution in plates with eollinear euts under arbitrary
loads, Proc. 4th U.S. National Congress Applied Meehanics, I, pp. 547-553 (1962).
G. C. Sih and H. C. Hagendorf
6 On eraeks in shelis
with shear deformation
6.1 Introduction
Although mueh work has been done on the stress analysis of eraeks in
initially flat plates [I], theoretieal treatment of eraeks in initially eurved
plates or shells has reeeived attention only in reeent years. The presenee of
eurvature in a shell generates deviation from behavior of flat plates, in that
stretching loads will induee both extensional and bending stresses, while
bending loads will al so lead to both types of stresses. In the neighborhood of
geometrie defeets such as man-made flaws or eraeks, the stresses are redistrib-
uted and ean lead to a reduetion of the load earrying eapaeity of shell strue-
tures.
One of the simplest shell geometries is that of a spherieal shell whose
eurvature radius R is everywhere eonstant. The first investigation on the
stresses in a spherieal shell containing a eraek was made by Ang, Folias and
Williams [2], who associated the shell problem with that of an initially flat
plate resting on an elastie foundation. The equivalenee of the two problem s
was made by identifying the foundation modulus with Eh/R 2 where E is the
Young's modulus and h the shell thiekness. The general eharaeter of the
eraek tip stress field in a shell was studied by Sih and Setzer [3] who pointed
out that the funetional reiationships of the loeal (extension-bending) stresses
are identieal with those obtained by superimposing the individual extensional
and bending stresses of an initially flat plate. They further emphasized the
eoupling effeets of extension and bending through the intensity of the loeal
stress fieid. A separate treatment of the spherieal shell problem was later
given by Folias [4] who utilized singular integral equations with Cauehy
type kernels, as devised by Knowles and Wang [5], for solving eraek problems
involving flat plates. The formulation in [4] relies on eertain approximations
202 G. C. Sih and H. C. Hagendorf
in the kernels for small values of the eurvature parameter. Other methods for
solving shell problems with eraeks have been presented by Erdogan and
Kibler [6], and Sih and Dobreff [7]. Their results covered a signifieantly
wider range of the eurvature parameter. Problems of eraeks in eylindrieal
shelIs have also been treated by Folias [8,9], Copley and Sanders [10], and
Erdogan and Ratwani [Il].
One of the shorteomings inherent in all of the foregoing work s, whieh are
based on the classieal shallow shell theory of Reissner [12], is that the
boundary eonditions on the eraek surfaees are only satisfied approximately
in the Kirehhoff sense. The drawbaek gives rise to a differenee in the angular
distribution of the extensional and bending stress fields near the eraek tip*.
This fundamental diffieulty was overeome by Sih and Hagendorf [13]. They
developed a tenth-order system of shell equations whieh aeeounted for the
effeet of transverse shear. The new results satisfy five individuaI boundary
eonditions on a free edge of the shell and aequired a three dimensional
eharaeter in the eraek front stresses resembling those found by Sih [I] for
a through eraek in a thiek plate.
This ehapter will be eoneerned with a tenth-order system of shell equations
applied to soIve shell problems with eraeks. Fourier transform is employed
redueing eaeh problem to the solution of a system of eoupled Fredholm
integral equations of the second kind. Both symmetrie and skew-symmetrie
loadings are considered and the resulting stress soIutions ean differ substan-
tially from those based on the classieal theory of Reissner [12].
6.2 Shell theory with shear deformation
The fundamental equations of the three-dimensional theory of elasticity

involve overwhelming diffieulties when applied to soIve the shell problem
unIess simplifying assumptions are invoked. A reasonable deseription of the
behavior of elastie shelis ean be made by assuming that the shell thiekness is
everywhere at least one order of magnitude smaller than the radius of
eurvature of the referenee surfaee. On this basis, approximate theories of
thin elastie shelis may be developed and they generally ean differ from one
another essentially in the formulation of appropriate stress-strain relations.
A theory of thin elastie isotropie shelis whieh aeeounts for transverse
* Since the extensional and bending stress intensity factors can no longer be conveniently
combined, special caution should be exercised when the results are used to analyze failure.
On eraeks in shelis with shear deformation 203
shear deformation and normal stress was derived by Naghdi [14]. He

obtained the stress-strain relations by applying Reissner's [15] variational
theorem of elasticity. Kalnins [16] has also proposed a shear deformation
theory of shells. In order to obtain a system of shell equations, which are
amenable to effective solutions of eraeks or cavities in shells, Sih and Hagen-
dorf [13] developed a theory that satisfies five physical boundary conditions
on an edge of the shell. This general theory of elastic shells is based on the
following three assumptions:
(1) The displacement vector is linear in the normal coordinate curve to
the middle surface,
(2) the distance between points on a normal to the undeformed middle
surface do not change during deformation, and
(3) the stresses are replaced by a system of stress resultants and stress
couples.
The stress-strain relations are deduced for an elastic orthotropic material
satisfying a generalized Hooke's law that is consistent with the restriction
imposed on the kinematics of deformation by assumption (2). The resulting
system of equations are invariant with respeet to rigid-body motion.
Shallow shell equations. A shell is said to be shallow (i.e., slightly curved) if

the middle surface is sufficiently smooth and all points on this surface are
sufficiently close to some plane. In a theory of shallow shells, the material
particles on the middl~ surface are located by the Cartesian coordinates of
their projection on the X Y-plane, i.e.,
x= X, Y= y, Z =f(x,y) (6.1)
in which x and y are the curvilinear coordinates andf(x, y) is the equation
of the middle surface of the shell. The points in the shell are described by
(x, y, z) with reference to the undeformed middle surface, where z is the
distance measured along the normal to the middle surface and is such that
-h12 ::; z ::; h12. Here, h stands for the thickness of the shell as shown in
Figure 6.1.
The theory of shallow shells is based on three simplifying assumptions
ix, r
[17] :
(1) Negleet. the squar~s of the derivatives and their products in

companson to umty. Y
SECTION A\ - A\
Figure 6.1. Spherical Cap with a Crack
(2) Negleet the transverse shear resultants Qx, Qy in the first two equations
of equilibrium.
(3) Negleet the tangential displaeements ux ' uy in the expressions for the
transverse shearing strains Yxz' yyz·
In what foIIows, only the first assumption wiII be ineorporated:
( Of)2 ( Of)2
ox < < 1, ay < < 1,
of of
ox ay < < 1 (6.2)
Let the referenee surfaee be the middle surfaee of a shaIIow spherieal shelJ
of eonstant thiekness, h. Then the prineipal radii of eurvature are R x = Ry =
= R. The sheII strain-displaeement equations are
OU x U.
I: = - +~ (6.3a)
.\x OX R
oU y Uz
Gyy -
ay+ -
R
(6.3b)
oU x y oU
Yxy
ay +ox- (6.3c)
ou z {3
Ux+ x
Yxz -y (6.3d)
ox
oU z u y {3
Yyz -/i..+y (6.3e)
=
ay
in which ux ' uy are the displacements tangential to the x and y coordinate
lines and U z is the displacement in the direction of the normal. The quantity
{3x is the angle through which the normal n rotates in the direction of the
x-coordinate line. A corresponding interpretation holds for {3y. The curvature
relations can be written as
o{3x o{3y o{3x o{3y

K xx = ax' K yy = ay' K xy = ay + --;;; (6.4)
It follows that the equations of equilibrium which express the vanishing of

the foree stress-resultants at each point of the middle surface are
oN xx oN yx Qx
ox + ay + R
=0 (6.5a)
oN xy oN"y Qy
ox + ay + R
=0 (6.5b)
oQx
+
oQy N xx + N yy = 0 (6.5c)
ox ay R
Similarly, the equilibrium equations which express the vanishing of the

moment stress-resultants at every point of the middle surface take the form s
oMxx oM yx _ Qx = 0
ox + ay (6.6a)
oMxy + oM yy _ Q = 0
ax ay y
(6.6b)
1
N xy - N yx + R (Mxy - M yx ) = 0 (6.6c)
For a homogeneous and isotropic elastic medium, the stress resultants are
expressed in terms of the shell strains as
(6.7a)
(6.7b)
N xy = N yx = Jlhyxy (6.7c)
and the shear resultants are given by
Qx = Jlhyxz' Qy = Jlhyyz (6.8)
where v is Poisson's ratio and Jl the shear modulus of elasticity. The moment
stress-resultants M xx ' M yy are considered positive when they produce positive
stresses on the part of the shell above the middle surface (Figures 6.2):
Figure 6.2a. Notations for Stress Resultants and Shear Forees
Figures 6.2b. Notations for Bending and Twisting Moments

ph 3
Mxx 6(1 _ v) (K xx + VK yy ) (6.9a)
ph 3
6(1 v) (K yy + VK<J (6.9b)
lih 3
M xy = M yx = U K xy (6.9c)
Equations (6.7) to (6.9) complete the system of equations which represent

the mathematical mode! used for this shaIIow shell theory. Jt is obvious from
the above relations that the last equation of equilibrium in equations (6.6)
is satisfied identicaIIy and may be discarded from further consideration. A
point to be made here is that since the present theory admits transverse shear
strains Yxz and yyz' the transverse shear forces Qx and Qy can be determined
in the same way as the other stress resultants. The actual stresses are express-
i ble in terms of the stress resultants as follows:
(1 + = Z)a
R xx
1
-Il N xx + {;3
12Z
M.u (6. tOa)
(1 + ~ = li )ayy
1
N yy + V
12Z
M yy (6.10b)
(1 + ~ )r = T1 XY
1N·xy
12Z
+ 71 Mxy (6.lOc)
(1 + ~ )r = xz
1
-
h
Qx (6.lOd)
(1 + ~ )r = h yZ
1
Qy
(6.10e)
The foregoing relations may be further rearranged to express the stress

resultants N w N yy ' etc., directly in terms of the displacements:
N = 2ph { ou x + ou v 1 + v }
v -" + - - u . (6.1ta)
xx 1 - v ox oy R -
N = 2ph { oU y
+ v ou x 1 + v } (6.11 b)
~+-R-uz
yy I - v oy
N XY = N yx = plI {OU
ayx + oU
oxy } (6.llc)
The moments M xx , M yy , etc., are

M xx = Jlh 3 {a{lx + v a{ly} (6.12a)
6(1 - v) ax ay
6(1 - v)
{ -a{ly
ay + vax-
a{lx} (6.12b)
_ _ Jlh 3 {a{lx a{ly}

M xy - M yx - 12 ay + ax (6.12c)
and the transverse shear resultants become*
Q4 = Jlh { ~; - i + {lx}, Qy = Jlh { ~; - ~ + {ly} (6.13)
Substituting the appropriate quantities in equations (6.11) to (6.13) into

equations (6.5) and (6.6) render
(6.14a)
(6. 14b)
v u + a{l,,- + a{ly _ !
2
z ax ay R
(3 + v){ auaxx + auayy} _
1- v
4
R2
( ~)u
1- v Z
=0 (6. 14c)
a {lx + 1 -
2 v a {lx + 1 + v a {ly _ 6(1
2 2 v) {au + {l _ x}
z U
ax 2 2 ay2 2 axay h2 ax x R
=0 (6.14d)
* If Naghdi's [14] general shell equations were used, then the transverse shear resultants
in equations (6.13) should be multiplied by the factor 5/6. The other expressions of N xx ,
N yy , etc., remain the same.
On eraeks in shells with shear deformation 209
_6(:.-1---.,.-------'--v) {OU z + f3 _ uY } = 0
h2 oy Y R
(6. 14e)
These coupled partial differential equations constitute the governing equa-

tions for the equilibrium of the middle surface of a shallow spherical shell
in terms of the five displacement measures.
Introduction of stress and displacement functions. The system of equations

(6.14) can be uncoupled by introducing the funetions ep, l/J and cfJ, lJ' such that
ux = Rl (~epx - l/J),
U u y = 1R oep
oy (6.15)
and
(6.16)
After a considerable amount of algebra, the following system of equations*

governing the non-symmetric deformation of a shallow spherical shell are
obtained:
1
vu +
6 z J,.4
V 2 uz 0 (6.17a)
1
V4 l/J - V 2 l/J 0 (6.17b)
k2
V 4 epo = 0 (6. 17c)
* If all of the three simplifying assumptions stated earlier were used, then the application
of equations (6.15) and (6.16) willlead to a different system of uncoupled equations given
by
176uz - [2(1 + v)/R2]W 4 uz + [12(1 - vZ)/hZRZ]I7Zuz = °

I7 zlf1 = 0, 17 zQJo = 0
I7 z'P-(12/h 2)'P = 0
with the provision that
alfl { Uz + 2[3(1 -hRV2)] I) 17 Z [2[3(1

hR h[l(1 + V)JI] }
-a;- -- 2
_ vZ)]} I 17
Z- -
R 3 1- v
- -- Uz
provided that the normal displacement Uz and the functions 1/1, epo satisfy the
conditions
(6.18a)
2 2e 4 4
V epo = 1 _ 2/Uz + Jc V Uz) (6.18b)
in which k, Jc and e stand for
h2 h2R 2 1 + V eo
12/ 1 ,Jc4 = 12(1 _ v2)1 2 ' e = ~ T; (6.19)
with eo bei ng h2 /12R 2 • The quantities lj (j = I, 2, 3) are given by

2 3 + V
Il = 1 + eo, 12 = 1 + -1--
- v
eo, 13 = 1 + - - e o
1- v
(6.20)
The remaining functions in equations (6.15) and (6.16) can be determined

from
(6.21a)
(6.21c)
(6.21d)
in which
4
14 = 1 + l--eo (6.22)
-v
The appearance of the functions epo, rfJ o in the expressions for ep, rfJ respec-
tively result from integrations after manipulating the equilibrium equations
(6.14). A cIose examination of the equilibrium conditions leads to equations
(6.18) and the expressian for rfJ o in equations (6.21). Upon defining a new
function
ep
X= Uz - rfJ - R2 (6.23)
and making use of the transformations in eq uations (6.15) and (6.16), the
stress resultants in terms of the normal displaeement U z and the auxiliary
funetions CjJ, t/!, ep, 'I'
and X are found:
h
2/1 { [PCjJ 02cjJ ot/! }
N xx ="
1- v R
-
ox2 + v oy2 + (1 + v)u z - --a; (6.24a)
2/1 h { 02cjJ 02cjJ ot/! }

N yy = oy2 + v ox 2 + (1 + v)u z - v~ (6.24b)
1 - v R
N xy = N = 2/1h { -
02cjJ
- - 1- ot/!}
- (6.24e)
yx R oxoy 2 oy
Similar expressions are obtained for the moments
M xx = - 6(l/1~ v) {:~; + v ~:~ + :x t)}

('I' + (6.25a)
v -
ox
o
('P+t)}
(6.25b)
M xy = M yx = - -6-
ph 3
{02 X
oxoy
+l 0
2 oy ('I' + :2)} (6.25e)
and shear resultants
Qx = ph { oep} oep
OX - 'I' , Qy = /1h oy (6.26)
The set of equations (6.15) to (6.26) forms a eomplete system of equations

for a theory of shallow spherieal shelis. The positive eonventions for the
stress resultants N xx ' N xy ' N yy , shear forees Qx, Qy, and bending moments
M xx , M xy , M yy are shown in Figures 6.2a and 6.2b.
A quantitative estimate of the error inherent in this theory of shallow
spherieal shelis resulting from the approximation introdueed by equations
(6.2) is given by the inequality
L
-R -< ( +8J)t
1 2J ' J =
[(Of)2
1 + Tx max
Jt - 1 (6.27)
where L is the diameter of the plane outline of the shallow spherieal eap
shown in Figure 6.1, and J < < 1 denotes the maximum deviation in
approximating the Lame parameters by unity [17]. For example, a deviation
of 5 percent limits R ~ 1.66 L.
6.3 Symmetric loading
Consider a portion of a spherical shell of constant thickness h containing a

through crack at the apex of the shell. The shallowness condition requires
that the crack oflength 2a in Figure 6.1 is small in comparison with the radius
of curvature Rand hence the problem can be solved in the projected XZ-
plane. The solution to this problem may be considered as the sum of two
separate probIems. The first concerns with loads applied along the edge of an
uncracked shell. The membrane and moment stresses will be different from
zero along the prospective crack site: y = 0, I x I :::; a. The second problem
deals with a cracked shell with tractions specified on the crack surfaces.
These tractions are equal and opposite to those found along y = 0, I x I :::; a
in the first problem. Since the solution for the uncracked shell is straight-
forward, attention will be focused only on the solution for a cracked shell
subjected to surface tractions.
Let the shell in Figure 6.1 be loaded symmetrically with respeet to the
X-axis along which the line crack is located. From the symmetry conditions,
the in-plane shear resultant, transverse shear and twisting moment vanish
everywhere along the line y = 0, i.e.,
Nxy(x, 0) = Qix, 0) = Mxy(x, 0) = 0 for all x (6.28)
The boundary conditions on the crack surfaces are
Nyy(x, 0) = N(x); Myy(x,O) = M(x), I x 1< a (6.29)

and the continuity requirements are given by
lim f3y= lim uy=O, lxi> a (6.30)

lyl~O lyl~O
where u/x, y, z) = uy(x, y) + zf3llC, y). Imposed on the solution are al so

the regularity conditions that the stresses and displacements must be bound-
ed as (x 2 + y2yl: --+ oo. This implies that the funetions uz ' 1jJ, epo and their
derivatives must be bounded at infinity. Although these conditions at infinity
are not geometrically feasible in the shell problem, they are satisfied within
the approximation introduced in the shallow shell theory by equations (6.2).
Integral representations. The integral solutions to the shallow shell equa-

tions (6.17) and (6.18) can be obtained by means of the Fourier sine and
cosine transforms depending upon whether the transformed function is odd
On eraeks in shelts with shear deformation 213
or even in x. Without going into the details, the integral representations for
tfJ amd ePo that possess the appropriate symmetrieal behavior with respeet
UZ '
to x and y and satisfy the regularity eonditions at infinity are

uz(x,y) = f~ [(1- 210) Al exp(-slyl) + As exp(-rtlyl)
+ A3 exp ( - ii I y I)] eos (xs) ds (6.3Ia)
tfJ(x, y) = f~ [A 4 exp (-s I y I) + A 2 exp (- 131 y 1)] sin (x s) ds (6.3Ib)
ePo(x,y) = f~ [A 6 + lylA 7 ] exp(-slyl) eos (xs)ds (6.3Ie)
For abbreviation, the funetions rt and 13 stand for

rt = rt(s) = [52 + (iIA 2)]t, 13 = 13(5) = [52 + (llk 2 )]t
in whieh i = ~ - I and li is the eomplex eonjugate of rt. The funetions
A j = A is) (j = 1,2, ... , 7) are not independent and are related to eaeh
other as
Al' A 2 are real, A 3 = As,
(6.32)
It is eonvenient to define two additional funetions A and B such that
(6.33)
The satisfaction of the symmetry conditions in equations (6.24) to (6.26)
leads to the following expressions for A and B in terms of A 1(s) and A 2 (s)
13 (6.34a)
A = -(1 - 2e)sA 1 - 2sk 2 A 2
(6.34b)
and A 6 is found in the same way:
s2A 6 = -{l - e + 2[(1 - 10)(1 - 210) - (e/eo)]k 2s2}A I

13(10 + k 2 s 2 )
+ 210 0 (1 _ e) A 2 (6.35)
Making use of the above relations, the membrane stress resultants, transverse
shears, and bending moments may be expresses in terms of only two un-
knowns A 1 (s) and A 2 (s). Sinee the resulting expressions are lengthy, only
those quantities which are pertinent to the specification of symmetry and

boundary conditions in equations (6.28) and (6.29) will be listed. The str~ss
resultants N yy and N xy are
N yy = 2~h f:[{[ - G~: + sl Y I)e + ~ ~ :(1 - 2e)(kS)2J
exp(- sl Y I) -[e + ~~ :(1- 2e)(kS)2}F - [[e(l- e)]t
+ 11 -+ 2e J}
{e + (kS)2
eo (1 + V)(AS)2 sG A1(s) + 2e o(1-e) P exp (-s 1 Y I)
1 [3 - v - (5 - v)e
+ [1 + (kS)2]S exp ( - P 1Y I) - 2(1 _ e) 1_ v (6.36a)
14 (1 - 2e) 2J 1 [e 1 +v- (1 + 3v)e

+ eo (ks) sPF - 2[e(1 - e)]t (kS)2 + 1- v
(1 - 2e) (ksf J SPG} Az(s) ] cos (x s) ds
N xy = 2~h f~[{[ - es 1 Y 1 +~ ~ ~ (1 - 2e)(kS)2] exp( -s 1 Y I)
- 31 + v (1 _ 2e)(ks)2
- v
F*- ~ -+ 2eeo (1 + V)(AS)2 G*} A1(s)
e + (kS)2
+ { 2e (1 _ e) exp( -s 1Y I)
o + [t + (kS)2] exp( -P 1Y I)
_ 1- [_2_ + ~ (kS)2J
2e
2(1 - e) I-veo
F*
- 2[ e~I-_2:)]f [~ ~ : - (kS)2 JG*} PAz(s) ] sin (x s) ds (6.36b)
and the moment expressions M yy and M xy take the forms
M yy = 2/lhf:H - [(1 + 12~ov) e + eeoslyl

J
+ (1 - 2e) (kS)2 exp( - sl Y I) + [e + (1 - 2e)(ks)2]sF
+ [[e(1 - eHi- - (1 - v) e (2S)2 ] sG }A 1(S)
e + (kS)2
+{ 2(1 _ elex p(-sly/)-[1+(ks)2]sexp(-Ply/)
1
+ 2(1 _ e) [2 - 3e + (1 - 2e) (kS)2]SPF
+ 2[e(1 ~ e)]t [(k:)2 +e- (1 - 2e)(kS)2] spG }AiS) ]eOS(XS)dS
[{-
(6.37a)
M xy = 2p.h f~ [eeo s I y I + (1 - 2e) (ks)2]exp( -s lY/)
+ (1 - 2e) (kS)2F* - (1 - v) e (2S)2G*}A 1(s)

e + (ks)2 2
+ { 2(1 _ e)exp(-sly/)-[!+(ks) ]exp(-Ply/)
+ 2(1
1 - _ 2ee) [1 + (kS)2]F* - ! (1 - v) (2S)2G* } PA 2 (s) ] sin (xs)ds
(6.37b)
The transverse shear resultant is
Qy = 2p.h f~ [e exp (-s lY/) - eF* -[e(1 - e)]tG*]sAt(S)
+ ~ [exP ( -PI yl) - F* - e~ e~ (k2~2 t + 1)G* }PAiS) }
eos (xs)ds (6.38)
In equations (6.36) to (6.38), the funetions F and G stand for
F(lyl;s)=-1 [ ::-exp(-eeIY/)+
1
2 ee
1
-exp(-eeIY/)
ee
J (6.39a)
G(lyl;s) = 2i 1 [1 ~ exp(-eely/) - 1exp(-eelyl)J

a (6.39b)
while F* and G* are given by
F*(I Y I;s) =k [exp (- äi y I) + exp( -ee I y I)] (6.40a)

216 G. C. Sih and H. C. Hagendorj
G*(lyl;s) = ;i [exp(-iXIYI)-exp(-alyl)] (6.40b)
Similar expressions for the remaining stress resultants, bending moments

and transverse shear may al so be abtained.
Fredholm integral equations. The crack problem will now be reduced to a

system of dual integral equations which can be salved numerically. By
substituting the appropriate stress resultants, shear resultant and moments
into equations (6.28) and (6.29), it can be easily shown that the unknowns
Al(s) and A 2 (s) are governed by
f~ ~ Al(s) eos(xs)ds = o}
°
(6.41a)
J
(J Ixl>a
OO - Ais)eos(xs)ds =
o s
and
J: [alles) Al(s) + a I2 (s) PAis)] eos (xs)ds = M(x) }

R~tx) Ix I < a (6.41b)
f~ [a 21 (s) AI(s) + anCs) fiAz(s)] eos (xs)ds =
2J1h
The funetions aij(s) (i,j = 1,2) depend on the variaus physical parameters
are given by equations (A 6.1) in the Appendix. Equations (6.41) ean be
satisfied by the following representations of Aj(s) (j = 1,2):
Al(s) = 2Jäs
Ehi
Ja r;
-V t 4J1 (t)J oCst)dt (6.42a)
f: Jt
5 0
Ais) = 4Ef;s 4J2(t)Jo(st)dt (6.42b)
where JO is the zero order Bessel function of the first kind and
1 + y
15 = 1 + -1--8
-y
(6.43)
The newly introdueed funetions 4Jit) (j = I, 2) are assumed continuous

over the intervaI [0, a] and required to satisfy the condition
!im [t-t 4Jit)] = 0, j = 1,2
1-+0
On eraeks in shelis with shear defarmatian 217
A further ehange of the unknown funetions by expressing cPi~) in terms of

tPjm for j = I, 2 aeeording to
_ cPlm = RcP1m + (h/6)cP 2(Ü (6.44a)

1 + 80
(6.44b)
it is found that the funetions cPj(~) (j = 1,2) are solutions to the following
system of eoupled FredhoIm integraI equations
tP 1(0 + f: [KlI(~,I]) cJ)IC11) +~ ~ Kd~'11)cP2(I])Jd11

2~~ f~
-
- n 0 (eN(ax)
- x )! d x
2
(6.45a)
tP 2 (0 + f~ [~ K21(~,I])cPl(11) + K22(~'11)tP2(11)Jdl]
- 2~~ f~(6/h)M(aX) d (6.45b)
- n 0 (e - x 2 )! x
where 0 < ~ s 1 and the symmetrie kernels are given by
Kij(~, 1]) = (~I])! f~ sCij(sla)J o( ~s)J o(I]s)ds, 0 < ~ sl; 0 < I] s 1 (6.46)
The quantities cij (i,j = 1,2) are given in equations (A 6.3) of the Appendix.
Equations (6.45) ean be solved numeriealIy for tPi~) (J = 1,2) onee the
stress resuItant N(x) and bending moment M(x) on the eraek surfaees are
speeified.
Stress intensity factor. The stress intensity faetor may be extraeted from
the singular eontributions of the stress resultants and moments given by the
improper integrals in equations (6.36) and (6.37). These integraIs are eon-
vergent everywhere exeept at the eraek tips x = ±a and y = 0 at whieh
points they beeome unbounded. To obtain the singuIar terms, equations
(6.42) wiIl first be integrated by parts giving
( 6.47a)
4a
P(s)A 2 (s) = Eh
and then the results ean be inserted into equations (6.36) and (6.37). A
solution near the eraek may be obtained by retaining only the first term in
the above expressions for A I and A 2 , and expanding the integrands for large
values of the argument s.
The near field solution is obtained by retaining only the first term in the
above expressions for AI(s) and A 2 (s). Let rand 8 denote the polar eoordi-
nates measured from the line of expeeted craek extension, the positive X-axis
in Figure 6.l. For symmetrie loading ease, the asymptotie expansions* for
the stress resultants in the neighborhood of the eraek tip are found to be
(Sih and Hagendorf [13])
N xx = - c.PI(l) (a/2r}t eos (8/2) [1 - sin (8/2) sin (38/2)] + ... (6.48a)
N yy = - c.PI(l) (a/2r)! eos (8/2) [1 + sin (8/2) sin (38/2)] + ... (6.48b)
N xy = - c.PI(l) (a/2r)! eos (8/2) sin (8/2) eos (38/2) + ... (6.48e)
and the moments are
M xx = - ~ c.P 2 (l) (a/2r)! eos (8/2) [1 - sin (8/2) sin (38/2)] + ...
(6.49a)
!!.. c.P 2 (l) (a/2r)! eos (8/2) [1 + sin (8/2) sin (38/2)] + ...
6
(6.49b)
M xy = - ~ c.P 2(l) (a/2r)! eos (8/2) sin (8/2) eos (38/2) + '" (6.4ge)
in whieh the nonsingular terms in r have been negleeted.

An important point to be made here is that the angular distribution of
the stress resultants and moments around the eraek tip are identieal. Equa-
tions (6.48) and (6.49) differ only in the eoeffieients or amplitude of the
singular stress field determined by c.PI(l) and c.P 2(1). This feature of the solu-
tion permits the membrane and moment stress fields to be eombined in a
natural f(l~hion by equations (6.10) whereas in the classieal shell theory
* At this point, the quantities Ba and B are neglected in camparisan to unity.

On eraeks in shelis with shear defarmatian 219
[4, 6, 7] the corresponding expressions for Mxx> M yy and M xy are incompatible

with those for N xx , N yy and N xy in that the O-dependence for the stress re-
sultants and moments are entirely different. It is obviously more meaningful
to consider the combined stress field:
(lxx == ~~;~1 eos (Oj2) [1 - sin (0/ 2 ) sin (30j2)] + 0(1) (6.50a)
~~;~1 eos (8j2) [1 + sin (Oj2) sin (30j2)] + 0(1) (6.50b)
k2J )Z! eos (8j2) sin (8j2) eos (38j2) + 0(1) (6.50c)
( r '
'x: = 'yo = 0(1) (6.50d)
The stress intensity factor kl (z) in equations (6.50) depends on the thickness
coordinate z and is defined as
k (z) = _ [<P 1 (1) + (2Z j h)<P i l)] .Ja (6.51)

1 1 + (zjR) h
It is seen that the effects of shell curvature enter into the local stress field
only through the stress intensity factor k 1 (z). Furthermore, it is c1ear that
in the shell problem, both the extensional and bending effects are always
coupled regardless of the nature of the loading. The numerical results for
three specific cases wiII be discussed subsequently.
Extensianal laad: N(x) = No. Let the crack surfaces be opened out by
uniform extensionalloads of constant magnitude No. The loading situation
is illustrated in Figure 6.3. Of interest are the stress intensity factors at the
top and bottom surface layers z = ±h/2 of the she11. Equation (6.51) wiII
be rewritten for the extensional case as
(6.52)
where ",je)(l) are related to <Pj(l) (j = 1,2) by
<Pi~) = - No"'je) (~) for ~= 1

The numerical values of k\e) at the top layer of the sheII are shown graphi-
cally in Figure 6.3 as a function of h/2a for a Poisson's ratio of v = 0.3 and
4
EXTENSION OF TOP L AYER 2 · hl2l., O.3)
005
0'0
I!:o
,z I
,
N
l 0.15
" II,' 020

, 025
.5
o 02 04 06 oe '0 ,2
SHELL THICKNESS/CRACK LE..:lTH. hl20
Figure 6.3. Variations of ExtensionaJly-Loaded Stress-Intensity Factor at Ouler Layer

with Shell Thickness to Crack Length Ratio
40 r05
EXTENS'O OF SOTT LAYER
,
POl5
-
I ~ 30
~' 020
; -2
( :::::. 025
: ~ 0 .30
~ i'"
--:.-
o 2.5 3.0
o 02 04 06 o.e '0 1.,

SHELL TH'CKNESS/CRACK LENGTH. hl20
Figure 6.4. Variations of ExtensionaJly-Loaded Stress-Intensity Factor at Inner Layer

with Shell Thickness to Crack Length Ratio
h/R ranging from 0.01 to 0.30 inclusive. These euryes depiet the effeet of
inereasing radius of eurvature, R. Note that for a given value of h/2a, the
norma!ized stress intensity faetor hk(t)(h/2)/ No-J a decreases with inereasing
radius of eurvature. This effeet is more pronouneed as R is increased. The
dotted euryes in Figure 6.3 represent !ines of eonstant R/2a, and henee
show the thiekness effeet. A similar set of euryes for k\e) at the bottom layer
of the sh ell z = -h/2 are given in Figure 6.4. For large values of R/2a, say
3, k~e)( - h/2) does not ehange appreeiably with the shelI thiekness to eraek
length ratio, h/2a, i.e., with inereasing thiekness. A eomparison of the stress
intensity faetors k\e) for z = ±h/2 is made in Figure 6.5 for v = 1/3 and a
~
z EXTENSION FOR hlR~O 2. "~1/3
§ 16 --TOP LAYER-~~
~
~
$....:-
~
14
a:
o
>-
>I BOTTOM LAYER-
e: 12
>-
>-
iii
~ 10
>-
~
<f)
~ ~_ _L -_ _~_ _~_ _~_ _-L___~
a: 0 02 04 06 08 10 12 14
I;;
SHELL THICKNESS/CRACK LENGTH. h/2a
Figure 6.5. Com paris on of Extensionally-Loaded Results at Outer and [nner Layer ofShell
partieular value of h/R = 0.2. For the given loading eondition, the k\e)
value at the top layer is slightly higher than that at the bottom layer. The
point of interseetion eorresponds to the ease of a short eraek whieh falls
outside the validity of the present theory of thin shelIs. Figure 6.6 illustrates
the quantitative differenee in the values of k\e) at the inner layer of the shell
z = - h/2 between the proposed theory and the classieal shell theory*
[6, 7]. The proposed theory prediets larger values of k\e) and the differenee
beeomes more pronouneed for longer eraeks, i.e., as h/2a ~ O.
Bending laad: M(x) = Mo. When the loading on the eraek surfaees are of
the bending type, say equal and opposite uniform moments with magnitude
* This comparison can only be made in the region ahead of the crack, i.e., for 0 = 0, as
the extensional and bending effects in the classical theory cannot be combined for other
values of O.
EXTENSION OF BOTTOM LAYER

FOR h/R'O I, .-113
30
----<:LASSICAL THEORY
./
/--SHEAR THEORY
o 02 04 06 08 10
SHELL THICKNESS/CRACK LENGTH, h/20
Figure 6.6. Classical Theory versus Shear Theory for Stretching Load
Mo, the results are quite different. The functions eP/I) in equation (6.51) are
equal to
while the stress intensity factor for z = ±h/2 beeomes
k(bl (+ h/2) = =+= [r/I~bl (1) ± r/I\bl (1)J 6M o .Ja (6.53)

1 - 1 ± (h/2R) h
Under bending, one side of the eraek tends to close as it is subjeeted to

compression. The maximum value of k~bl will always oeeur on the tension
side, Henee, kibl (h/2) and Wl ( - h/2) at the outer and inner surfaces of the
shell will correspond to couples -Mo and Mo, respectively. With the tension
side at z = h/2, Figure 6.7 gives a plot of h 2k\b l (h/2)/6Mo.Ja versus h/2a for
v = 0.3 and h/R = 0.01,0.05, etc. Uniike the re suits for extensionalloading,
the magnitude of k\bl increases monotonically with h/2a for a fixed value
of h/R. This behavior was also observed in the fiat plate theory by Hartranft
and Sih [l8]. The dotted euryes in Figure 6.7 shows the variations of k\bl
along the lines of R/2a = eonstant as the thiekness of the shell is ehanged.
Figure 6.8 gives the results for the situation where the tension side is reversed
to the inner surface of the sheli, z = -h/2. By fixi ng h/2a < 0.8, kibl
increases with decreasing values of h/R. The opposite effect is seen when
0 .9 0_01
o.o'!' 0.10
hIA-o.l:i
0.8
l!:
t: 0.7
~ 0.6
BENOING OF TOP LAYER Z ' h/21.'0.31

ri
0
~
>- 0 4 MY
"~ , ,
,..
t- 0 .3
M.
õi
Z
~ 0. 2
~
'"~ OI
'"
t-
<Il
0 0 .2 0.4 0 .6 0 .8 10 1.2 14 1.6 18 2 .0 2 .2 2,4

SHELL THICKNESS I CRACK LENGTH h/2a
Figure 6.7. Stress-Intensity Factor at Outer Layer Induced by Bending Laad
1.0
0.9
0.8
l!:
.r~ 0.7
~
e-,,-
BENOING OF BOTTOM LAYER Z··hI2l. ,0 31
"'"
","
~
0
t-
::.",.. 0 .4
M.
0-
o;
Z
"';:::
0-
0.3
<Il
<Il
"''">-
<Il 0.2
o 0.5 1.0 1.5 2.0 2.5

SHELL THICK NESS ICRACK LENGTH. h/2a
Figure 6.8. Stress-Intensity Factar at Inner Layer Induced by Bending Laad

10
09
TOP L~AYER
• , ,
M. •
LAYER
6ENDING fOR "IR~Ol5 . • - 0.3
o 05 '0 15 20 25
SH ELL THICKNESS/CRACK LENGTH. "/20
Figure 6.9. Results at Outer and Inner Layer of Shell Subjected to Bending Load
h/2a is approximately greater than 0.8. At this ratio of h/2a the shelI is rela-
tively thick in comparison with the crack length 2a and the assumption of
thin shelI is violated.
Figure 6.9 makes a comparison of k\bl for z = ±h/2 with h/R = 0.15 and
v = 0.3. The two curves interseet at h/2a ~ 0.48 meaning that
k\bl(h/2) > k\b 1( -h/2) for h/2a < 0.48 and k\bl(h(2) < k\bl( -h(2) for
~
..
u
It
0 .6
>-
0- ~
~
iii
z
"'
~
BENOING Of TOP LAYER fOR , • 1/ 3
zl
:;;"'
a:
o ~-r0~2'--0td4--~0~6--'0~.8'-~1.0~~1~2--~I~4--~16~~1.8
S>iELL THICKNESS ICRACK. "/20
Figure 6.10. Comparison of Classical and Shear Deformation Theory for Bending Load
h/2a > 0.48. It should be bome in mind that these results eorrespond to two
different loading eonditions, namelyapplied moments ± Mo, i.e., opposite
in sign.
A dramatie departure of the present results from those obtained by the
classieal theory is shown in Figure 6.10 where the trend of the variations of
kiti with h/2a is, in faet, opposite. The differenee beeomes more and more
signifieant as the ratio h/2a decreases. The results based on the classieal
theory is known to be approximate sinee the boundary eonditions on the
eraek surfaees are satisfied only in the Kirehhoff sense. A more detailed
diseussion on the approximate nature of the classieal shell theory has been
given by Sih and Dobreff [7].
Combined loading: 6Mo/h 2 = No/h = (J. In general, both extensional and

bending loads may be present and the total normal traetions applied to the
eraek surfaees eonsists of NyiX,o) and that due to the eouple MyiX, 0).
Consider the speeial ease where the eraek surfaees are not permitted to eome
into eontaet. This eondition ean be satisfied by taking
0 .05
n/R . O OI
.
3.5 COMBINEO EXTENSION AND &NOING (TOP L AYER I
1010
01 02 03 O. 05 06 0 .7 0 .8 0.9
SHELL THICKfilESS / CRACK LENGTH , hl2a
Figure 6.11. Results at Outer Layer for Combined Extensional and Bending Loads
226 G. C. Sih and H.C. Hagendorf
where both No and Mo are constants. According to the above equation,

there arises two possible conditions of combine d loading which are shown in
Figures 6.11 and 6.12.
In Figure 6.11, the maximum tension oeeurs at the outer layer of the sheli,
z = h/2 with the combined stress intensity factor kl (h/2). For convenience,
let the crack surfaees be subjeeted to the combined load such that
3.0 COMBINEO EXTENSION AND SENDING (SOnOM L AYERI

FOR 6M o /h2 • No/h lt 0'"
0.75
,, ''
~2 . 5
0:
...
o
...~
I:
enz
'"~ 2.0
<Il
<Il
~
<Il
"
j
o 1.5
o 0 .2 0 .4 0.6 0 .8 tO 12 1.4
SHELL THICKNESS J CRACK LENGTH. hJ20
Figure 6.12. Results at Inner Layer for Combine d Extensional and Bending Loads
I-No /hl = I ±6Moh2 1· A plot of k l (h /2)!(Jy'a versus hJ2a is shown in

Figure 6.11 and shows that kl (h/2) is redueed sharply with decreasing erack
length. eUryeS for hjR = 0.01 to OJO inclusive are given. By following the
dotted curve for Rj2a = 2, it can be seen that an increase in the shell
thickness from h = 0.04a to h = 0.2a corresponds to a reduction in the
stress intensity factor of approximately 45 percent. If the direction of the
applied couple is reversed, then the side of maximum tension switches to the
inner layer of the shell at z = - hi2. The numerical results for kl ( - h/2)
are show n graphically in Figure 6.12. In the range 1.5 5 Rj2a 5 2.5,
k l ( -h/2) tirst decreases reaehing a minimum at h ~ 0.6a and then increases
as hJ2a is raised.
On cracks in shelis with shear deformation 227
Appendix: Integrand and Kernel Functions
The quantities 0;/.5) in the duaI integral equations (6.41) are complicated
funetions of the geometric parameters of the cracked spherical sh ell and
their expressions are listed below. In addition, Ci/S) (j = 1, 2) in equations
(6.46) are given separately in the Appendix so that unnecessary complications
in the presentation can be avoided.
A6.1 lntegrand functions

The integrand funetions aij(s) (i,j = 1,2) in the dual integraI equations
(6.41) are given by
a 11 (s) = - ( 1 + 1 : v I:o} - (1 - 21:)k 2 s 2
+ {I: + (1 - 21:)k 2 s 2 }F oUs)

+ {[I:(I - I:)J± - (l - v)e),2 s2}G o(AS) (A6.1a)
a12 (s) = -1 I: + es 2
- k s(I +'k2 S2)"-2
2 1 - I:
1
+ 2(1 _ 1:) {2 - 38 + (l - 21:)k 2 s 2 }F o(AS)
+ 2[s(1 1_ s)J± k 2 s 2 + I: - (l - 2e)k 2s 2J GoCAS)

[I: (A6.lb)
l+v . 3+v
a 21 (s) = - -l-- e - I:Fo(AS) + -1-- (l - 2e)k 2 s2 {I - FoUs)}
-v -v
- {[e(l - e)J± + ~ ~ ~: (1 + V)}c2 S 2 } Go()'s) (A6.1c)
1
0 22
+ k s2) + k s(1 + kl s2)1
(s) = -2 e (1
2
eo - e
_ 1 {3 - v - (5 - v)e + 14 (1 - 28) k 2 sl }Fo(AS)

2(1 - 8) 1- v eo
1 {e 1 + v - (1 + 3v)e
2[8(1 - 8)]+ k 2 s2 + 1- v
(1 - 28)k l S2}G O(AS) (A6.ld)

in which Fo, Go are defined
Fo (AS) = sF(O; s), GO(AS) = sG(O; s)
leading to
Fo{s) = {[S2(1 + /}t + /]/2(1 + /)}t (A6.2a)
Go(s) = s/{[2(1 + S4)] [(I + S4)! + S2]}t (A6.2b)
and the parameters k, A, e, etc., can be found in equations (6.19) and (6.20).
A6.2 Integrand Kernelfunction

The functions ciis) (i, j = 1,2) in the kernel Kij of equation (6.46) stand for
c l1 (s) = 12eo [{(I + k 2s2yl: + ks}]-Z - -8I2 [eo + k 2s2] [I - FO{AS)]

+v -v
(A6.3a)
3- v 1
+ 1 _ v2 [1 - Fo{AS)] - (I _ V2)A2S2 Go{AS)
(I - V»)4 S 2 I
+ k2 · [I - FO(AS)] - ~ {I - 2A 2s2Go{As)}
.1 2
- k 2 Go{As) (A6.3b)
c2is) = -2
1 [{{I + k 2s2)t + kS}]-2
+ V
(A6.3c)
References
[1] Sih, G. c., A review of the three-dimensional stress problem for a eraeked plate,
International Journal 0/ Fraeture Meehanies, 7, pp. 39-61 (1971).
[2] Ang, D. D., Folias, E. S. and Williams, M. L., The bending stress in a eraeked plate
on an elastie foundation, Journal 0/ Applied Meehanies, 30, pp. 245-251 (1964).
[3] Sih, G. C. and Setzer, D. E., Diseussion of The bending stress in a eraeked plate on
an elastie foundation, Journal 0/ Applied Meehanies, 31, pp. 365-367 (1964).
[4] Folias, E. S., The stresses in a eraeked spherieal sheli, Journal 0/ Mathematies and
Physies, 44, pp. 164-176 (1965).
[5] Knowles, J. K. and Wang, N. M., On the bending of an elastie plate containing a
eraek, Journal 0/ Mathematies and Physies, 39, pp. 223-236 (1960).
[6] Erdogan, F. and Kibler, J. J., Cylindrieal and spherieal shelis with eraeks, International
Journal 0/ Fraeture Meehanies, 5, pp. 229-237 (1969).
[7] Sih, G. C. and Dobreff, P. S., Craek-like imperfeetions in a spherieal sheli, Glasgow
Mathematieal Journal, 12, pp. 65-88 (1971).
[8] Folias, E. S., An axial eraek in a pressurized eylindrieal sheli, International Journal
0/ Fraeture Meehanies, 1, pp. 104-113 (1965).
[9] Folias, E. S., A cireumferential eraek in a pressurized eylindrieal shell, International
Journal 0/ Fraeture Meehanies, 3, pp. 1-12 (1967).
[10] Copley, L. G. and Sanders, J. L., A longitudinal eraek in a eylindrical shell under
internaI pressure, International Journal 0/ Fraeture Meehanies, 5, pp. 113-131 (1969).
[11] Erdogan, F. and Ratwani, M., A eireumferential eraek in a eylindrieal shell under
torsion, International Journal 0/ Fraeture Meehanies, 8, pp. 87-95 (1972).
[12] Reissner, E., On some prob!ems in shell theory, Struetural Meehanies, Proceedings
First Symposium on Naval Struetural Maehenies, pp. 74-113 (1958).
[13] Sih, G. C. and Hagendorf, H. c., A new theory of spherieal shelis with eraeks,
Thin-Shell Struetures: Theory, Experiment, and Design, Edited by Fung, Y. C. and
Seehler, E. E., Prentice-Hall, New Jersey, pp. 519-545 (1974).
[14] Naghdi, P. M., On the theory of th in elastie shelis, Quarterly Applied Mathematies,
54, pp. 369-380 (1957).
[15] Reissner, E., On a variational theorem in elasticity, Journal 0/ Mathematies and Phys-
ies, 29, pp. 90--95 (1950).
[16] Kainins, A., On the derivation of a general theory of elastie shelis, Indian Journal 0/
Mathematies, 9, pp. 381-425 (1967).
[17] Novozhilov, Y. Y., Thin Shell Theory, Second edition, Noordhoff, Groningen, Nether-
lands (1964).
[18] Hartranft, R. J. and Sih, G. c., Effeet of plate thickness on the bending stress distri-
butian around through eraeks, Journal 0/ Mathematies and Physies, 47, pp. 276-291
(1968).
G. C. Sih and E. P. Chen
7 Dynamic analysis of cracked plates

in bending and extension
7.1 Introduction
The present chapter deals with dynamic loading of fiat plates containing a
through crack where inertia effects can no longer be neglected. Two types
of dynamic-Ioad sources are considered: namely, vibratory and impact.
When the applied load is time dependent, waves travel through the plate
and result in a complex stress pattem upon striking the crack modeled as a
plane of discontinuity in the material. Because of the complexities encoun-
tered in the treatment of elastodynamic crack problems in three dimensions,
simplifying assumptions are made to relax the system of equations in the
theory of elasticity. The classical approach in the development of plate
theories [1-3] is to assume that the stress distribution in the thickness
direction of the plate is known as a priori. For the elastostatic crack problem,
Hartranft and Sih [4] suggested to determine the stress variations through
the plate thickness from the plane strain condition. Although their method
al so applies to elastodynamic crack probIems, it will not be discussed here.
One of the most commonly used plate bending theories is identified with
the Kirchhoff boundary conditions [I]. In this theory, only two boundary
conditions are applied for obtaining the plate defiection that satisfies a
fourth order differential equation whereas physieal reasoning requires three
eonditions. This inconsisteney has been removed in the Mindlin's [2] theory
of plate bending which aeeounts for the rotatory inertia and shear effeets
and henee involves the satisfaetjon of three boundary conditions on the
eraek surfaee. In plane extension, Kane and Mindlin [3] have developed a
higher order plate theory that eouples the extensional motion and the thiek-
ness mode vibration. The dynamic stress or moment intensity faetor solutions
are obtained for these higher order plate theories and are eompared with
232 G. C. Sih and E. P. Chen
those found from the classical thin plate theory of Kirchhoff and the plane
theory of elasticity.
7.2 Classical plate bending theory
In the application of the classicaI plate bending theory, the crack length and
input wave length are assumed to be large in comparison with the plate
thickness h. The input waves are generated by a combination of bending
moments applied to the plate edge causing the plate to vibrate in the trans-
verse direction. As mentioned earlier, since the classical plate bending theory
fails to account for all the physical boundary conditions on the crack surface,
the stress distribution in the immediate neighborhood of the crack will
naturally be affected. A discussion of this will follow.
Lagrange equation. Let the coordinate axes x and y be chosen such that they
are in the middle plane of the plate and the z-axis is perpendicular to this
plane as illustrated in Figure 7.1. It is assumed that the norrnaIs of the xy-
plane before bending remain as norrnais of the sam e planc after bending
while the middle or xy-plane is unstrained during loading. Under these
considerations, the rectangular displacement components u" !lvand u=
become
Middle
plane
Figure 7.1. Flexural waves impinging on a cracked plate

Dynamic analysis of crackE'd plates in bending and extension 233
(7.1)
in which w represents the defiection of the middle plane of the plate and it
depends on the variables x, y and t. Referring to Figure 7.1, the bending
and twisting moments per unit length, Mx, My andMxy may be expressed in
terms of w(x, y, t) by making use of the stress-strain relations and collecting
the moments produced by (Jx, (Jy and 'xy acting on any section of the plate
paralleI to the xz and yz planes. The resuIts are
( 02 W W)
02-
Mx = D ox 2 + voy2 (7.2a)
( 02 W 02 W )
My = D oy2 + V ox 2 (7.2b)
(Jzw
M xy = (1 - v) D ~
xoy
(7.2e)
in which
Eh 3
D= (7.3)
12(1 - v 2 )
is ealled the flexural rigidity of the plate. The Young's modulus and Poisson's
ratio are denoted by E and v. Similarly, the vertieal shearing forees Qx and
Qy acting on the plate edge due to the shearing stresses 'zx and 'zy may also
be obtained:
Qx = D (J~ (~:~ + ~:~) (7.4a)
o (OZw (iZw) (7.4b)

Qy = D ay OX Z + oy2
By neglecting the rotatory inertia which IS of order h 3 , the equation of

motion yields
(7.5)
where V4 is the biharmonie operator in two dimensions and p is the mass

density of the material.
[neident and seattered waues. Let a plate with a eraek of length 2a as shown
in Figure 7.1 be exeited by propagating f1exural waves whieh when eneoun-
tered by the eraek are refleeted and refraeted, that is, seattered. The souree
that emits the ineident waves eorresponds to moments applied symmetrie-
ally about the eraek plane y = O. The expression for the ineident wave may
be written as
w(i)(x, y, t) = W o eos (xy sin y) exp [ - i(xx eos y + wt)] (7.6)
in whieh W o is the amplitude of the input wave and w is the eireular frequeney.
The angle of incidenee y lies between the limits - Jr and Jr and is measured
from the positive x-axis. In equation (7.6), the parameter
(7.7)
is determined such that w(i) satisfies equation (7.5). The frequeney w in

equation (7.7) eorresponds to the Rayleigh-Lamb frequeney [5] for long
waves, a limitation of the classieal plate bending theory.
The total wave field in the plate ean be separated into two parts, one
associated with ineident wave and the other with seattered wave:
w(x, y, t) = W(i) (x, y, t) + W(S) (x, y, t) (7.8)
In the same way, the plate defleetion, moments and shear forees may also
be divided into two parts. For example,
lvI ix, y, t) = M~i) (x, y, t) + M~S) (x, y, t), etc. (7.9)

From equation (7.6), all the quantities associated with the ineident waves
ean be easily derived. Henee, the f1exure problem is primarily eoneerned with
finding the seattered field.
For a stress-free eraek in plate bending, the three quantities MY' M xy and
Qy should vanish individually for I x I < a and y = O. The classieal theory,
however, involves only two eonditions. This is done by eombining M xy and
Qy into a single one, Vy, referred to as the equivalent shear foree [I], i.e.,
My(x,o,t) = Vy(x,o,t) = 0, lxi < a (7.10)

The eonditions to be speeified on the eraek for the seattered field beeome
M (S)
y = - MUl.
y' y = -
V(S) 0
V(i)=
y' Ix I < a; y =0 (7.11)
Outside the eraek, the problem is symmetrie about the x-axis and henee
Dynamic analysis of cracked plates in bending and extension 235
u(S)
Y = v(s)Y = 0, I x I' -
> a·
, y =0 (7.12)
The quantities M~i) and V~i) in equations (7.11) are given by
M~i)(X, 0, t) = - Dw ocx 2 (sin 2 y + vcos 2 y) exp[ -iCcxx cosy + cot)](7.13a)
V~i)(x, 0,
OM(i)
t) = [ ~ + Q~i) J = 0 (7.13b)
Ox y=o
in which M~~ and Q~i) may be computed from equations (7.2c), (7.4b) and
(7.6).
Fredholm integral equations. Application of Fourier transform to the

vari abi e x in equation (7.5) leads to an ordinary differential equation in the
variable y whose time dependent solution for the seattered wave field takes
the form
w(s) (x,y,t) = 21
n
Joo-oo {Al(s) exp [- Y(S2 + cx rt ] 2
+ A 2(s) exp[ - Y(S2 - c( 2 )t]}exp [ - i(sx + cot)]ds, y ;::: 0 (7.14)
which satisfies the regularity condition that w(s) ~ 0 as y ~ oo. It follows

from equation (7.14) that the conditions in equations (7.11) and (7.12) are
satisfied if Al(s) and A 2 (s) are expressed in terms of a single function A(s):
(7.15)
such that A(s) satisfies the dual integral equations
-1 Joo A(s) exp (- isx)ds = 0, I x I ;::: a (7.16a)

2n - oo
_I_JOO fl(S)A(s) exp (-isx)ds = ~o (sin 2y+ vcos 2y)exp(-icxxcosy),

2n -oo
lxi < a (7.16b)

in which
The magnitude of the maximum moment generated at the incident wave

front is given by Mo = Dwocx2 which is assumed to remain finite as co ~ o.
Equations (7.16) may be solved by a method deseribed in [6]. Without going

into detail s, a solution is obtained:
= (1 _2:)~t: v)D (sin y+ veos2y{f~ ~~ <1\(~)Jo(sa~)d~

t
A(s) 2
+ ~~ 'l'1(~)Jl(Sa~)d~J (7.17)
with Jo and J I being the first kind Bessel funetions or order zero and one.
The funetions <P I (~) and 'l' 1 (~) may be ealeulated numerieally from the
Fredholm integral equations of the second kind:
<Jil(~) + J~ L 1G,I1) <Ji t (l1)dl1 = ~~ Jo(lY.a~ eosy) (7.18a)
'l'1(~) +f~ L2(~,I1) 'l'1(I1)dlJ = ~~ Jl(lY.a~ eosy) (7.18b)
f:
whose kernels LI and L 2 are given by the expressions
Ll(~' 11) = (~IJ)t s[F 1(sja) - 1] Jo(~s) Jo(lJs)ds (7.19a)
L2(~' 11) = (~I1)t j: s[ F j (sja) - 1] J j (~s) Jt (lJs)ds (7.19b)
Note that equations (7.19) is symmetrie in the variables ~ and 11 and Fj(s)
stands for
2 11(s)
F 1(5) = (1 _ v)(3 + v) -5- (7.20)
Onee <P j (~) and 'l' j (~) are known, the moment s and shear forees in the craek-
ed plate ean be determined.
Moment intensity factor. The moment and shear foree distribution around
the eraek ean be determined from the asymptotie solution expressed in terms
of aset of polar eoordinates rand () referred to the eraek tip as shown in
Figure 7.1 with
2
r = (x - a)2 + y2, () = tan- 1 (-y-)
x-a
(7.21)
Near the craek tip x = a and y = 0, the seattered field solution bei ng
singular dominates while the incident field contributes at distanees away
from the eraek. Inserting A(s) in equation (7.17) into the appropriate moment
expressions, it ean be shown that as r ~ 0
Mx
Kl
= Mx(s) = (2r)+ (1 -+ v)v [1"2 . e . 3e"2 - 2:9J + ()
3 sm sm eos 01 (7.22a)
My = M y (s) Kl [1 (1 - v) . e . 30
= (2r)t "2 + v 3 sm 2" - 2:eJ +
sm cos
0(1) (7.22b)
M xy = M xy
(s) Kl ( 1 ) [2 sm
= (2r)t 3 + v
1 - -v sm
. 2 + -2 . e ecos 2"
30J + 0 ( 1)
(7.22c)
The eoefficient Kl in equations (7.22) is defined as the moment intensity

factor
(7.23)
Since the stresses (Fx, (Fy and T xy are related to the moments Mx, My and M xy
by the relations
12z 12z 12z
(Fx = V Mx, (Fy = fi3 My, Txy = V M xy (7.24)
the stress intensity faetor kl that varies linearly with z in the thickness
direction of the plate may be obtained by multiplying equation (7.23) by
the factor 12z/h 3 • Note that equations (7.22) for Mx. My and M xy possess
the same inverse square-root of r singularity as in plane elasticity. The
angular distribution of the moments, however, is a function of the Poisson's
ratio of the material. A physicalIy unrealistic feature of the classical plate
bending theory is that the shear forces
Kl (
Qx = Qx = (2r) 3/2 3 + v eos
(s) 2) (3e)
T + ... (7.25a)
Qy -_ Q(") _ ~ (~) •
y - (2r)3/2 3 + v sm
(3e)
2 + ... (7.25b)
have a singularity of order r - 3/2 as r ~ O. This higher order singularity arises

from the failure of satisfying all three boundary conditions on the erack
surfaee and can be removed by a theory in which shear deformation and
rotatory inertia are included. In subsequent work, it will be shown that Qx
and Qy actually remain finite as the crack tip r = 0 is approached for pure
bending loads.
The numerical results can be best displayed by expressing equation (7.23)
as
(7.26)
in whieh Kl1) represents the amplitude of the moment intensity faetor and is
given by
(7.27)
The ph ase angle 6 1 between Kl and ineident wave in equation (7.26) is
6 = ~ tan- l {Iill[4i I(l)] - Re['PI(l)]}

(7.28)
I 2rc Re[4i I (1)] + Im['P I (1)]
with r = 2rr/w being the period of oseillation. In the li mit as w -.0, the
statie value of Kl = Mo.../a is reeovered [7]. The numerieal values of epl
and 'PIin equation (7.27) are obtained from the Fredholm integral equa-
tions (7.18) for v = 0.25 and f1exural waves at normal incidenee, i.e.,
y = 90°. Figure 7.2 shows a plot of I KlI) I versus frequeney w normalized
11=0.25. r=go·
o 0.1 0.2 0.3

W/W.
Figure 7.2. Normalized moment intensity factor versus frequency for different a/h values
against the eut-off frequeney of the plate, Wo = rrcz/h, where C 2 is the shear
wave veloeity of the material. The dynamie moment intensity faetor is found
to be smaller than the static case and decrease in magnitude as the frequency
is increased. This effeet becomes more and more pronouneed as the crack
length to plate thickness ratio increases.
7.3 Mindlin's theory of plate bending
Since the c1assical plate theory is not expected to be valid for sharp transients
or for the frequencies of modes of higher order, Mindiin [2] advanced a
higher order theory which includes shear deformation as weil as rotatory
inertia. The same difference prevails between the Bernoulli-Eulear beam and
the Timoshenko beam. For crack probiems, it is essential that the three
physical boundary conditions of vanishing bending moment. twisting
moment, and transverse shear force are satisfied individually on the crack
surface.
Governing equations of flexural motions. In the Mindlin's theory [2] of

flexural motions of plates, three types of flexural waves are considered:
slow flexural, fast flexural and thickness shear waves. In general, any one of
these waves propagating toward a crack gives rise to reflected waves of all
three types. If the cracked plate is set into steady-state motion by the pro pa-
gating flexural waves, the rectangular components of the displacement vector
may assume the forms
ux(x, y, z, t) = zljJx(x,y) exp( - iwt) (7.29a)
uix, y, z, t) = zljJix, y) exp( - iwt) (7.29b)
uz(x, y, z, t) = ljJ/x, y) exp( - iwt) (7.29c)
In equations (7.29), w is the circular frequency of the harmonic wave and z
is the thickness coordinate. The normal displacement of the plate is ljJ z and
the rotations of the normals about the x- and y-axes are denoted by ljJ x and
ljJy as shown in Figure 7.3. From equation (7.29c), it is c1ear that the normal
strain in the thickness direction has been negleeted.
With reference to the rectangular coordinate system x, y and z of Figure
7.3, the bending and twisting moments can be expressed in terms of IjJx, ljJy
and IjJz as
M = D(OljJx + v OljJy) exp(-iwt) (7.30a)

x OX oy
My = D( °t; + °t: )
v exp( -iwt) (7.30b)
Mxy = e; v) D( °tx + °o~ ) exp( - iwt) (7.30e)
Figure 7.3. Displacement functions in the shear deformation theory of Mindlin
where D = Eh 3 /12(1 - v2 ) is the flexural rigidity of the plate. The shear

forees Qx and Qy per unit length of the plate are given by
Qx = K 2 fJ.h (~~z + I/Jx) exp( - iwt) (7.31a)
(7.31b)
The shear eoeffieient K 2 assumes the value n 2 /12 so that the frequeney of
the first anti-symmetrie mode of thiekness-shear vibration matches that of
the exaet three-dimensional theory. Making use of equations (7.30) and
(7.31), the three equations of motion beeome
6D[(1 - v)V 2 l/Jx + (1 + v) o~ (°tx + ~~y ) J

- n 2 fJ.h (I/Jx + °tz) = - ph 3 w 2 I/Jx (7.32a)
6D[(1 - V)V21/Jy + (1 + v) o~ (ot x

+ °o~ )J
_ n2/1h (I/! y
+ Ol/!z)
oy = _ p/z3 w 21/! y (7.32b)
n2/1h (VZI/! z + oI/! x + ol/!y) = _ 12phw 2./•• (7.32c)

OX oy '1'_
The Laplacian operator V2 in two dimensions is a2 /ax 2 + aZ/ay2.

It is convenient to introduce three displacement potentials epl' epz and ep3
such that they are related to l/I x' l/I)I and l/I z as follows:
I/! x
= (a -
2
1) Oepl + (a _
OX I
1) Oep2 + Oep3
OX 0Y
(7.33a)
./. ( Oepl Oep2 Oep3

'I'y = a2 - 1) ay + (al - 1) ay - OX (1.33b)
l/Iz = epl + epz (7.33c)

in which
2 (CX.)2 J'=12
a·=--.::.1.. (7.34)
J 1 - V CX3' ,
The wave numbers CXj (j = 1,2,3) stand for
where the cut-offfrequency is W o = nCz/h and cz = C/1/p)t is the shear wave

velocity. The rotatory inertia and transverse shear effects are associated with
R and S as given by
h2 12D
R = 12' S = n 2 ph (7.36)
Substituting equations (7.33) into (7.32), the potentials epl' epz and ep3
that generate slow flexural, fast flexural and thickness shear waves satisfy
differential equations of the Helmholtz type:
(V 2 + CX])epix, y) = 0, j = 1,2,3 (7.37)
Equations (7.35) show that the three wave numbers aj (j = I, 2, 3) are

dependent on the frequeneies; thus, all three flexural waves are dispersive.
In addition, al is real for all frequeneies, whereas al and a3 may be real or
imaginary depending upon whether OJ is greater or smaller than OJo.
Ineident flexural waves. Suppose that the plate with a eraek of length 2a
in Figure 7.1 is set into motion by a slow flexural wave (OJ < OJ o)* arriving
from distanees far away from the eraek. At low frequeneies, it is the sl ow
flexural waves that are of interest sinee the fast flexural and thiekness shear
waves attenuate gradually from the souree of loading. The form of the input
flexural wave that results in symmetrie bending with referenee to the line
eraek ean be written as
cpli) = 4>0{ exp [ - ia l(x + y)j ~2J + exp [ - ia l(x - y)j ~2J} exp ( - iOJt)
(7.38a)
(7.38b)
with 4>0 being the amplitude of the input wave. Along the axis of symmetry
y = 0, the moments of the ineident field are
M~)(x, 0, t) = (al - 1) Ma(l + v) exp [ -i (~~ + OJt)] (7.39a)
M~i)(x, 0, t) = (al - 1) M o(1 + v) exp [ - i (~~ + OJt)] (7.39b)
M(i)(X
xy \ , ° °
, t) = (7.39c)
The ineident moment Mo = 4> ort.i D is assumed to remain finite as OJ --+ 0.

Similarly, the ineident shear forees are
(7.40a)
(7.40b)
The eomplete solution of the waves as diffraeted by the eraek is obtained
by adding the ineident and seattered waves, i.e.,
(7.41)
* The case of fast flexural wave w > Wo may be treated by the same method.
Likewise, the plate displaeements, moments and shears ean al so be found by

superposing the ineident and seattered part s and the results are obvious. For
a traetion-free craek, the quantities My, M yx and Qy must each vanish for
I x I < a and y = O. This implies that
M(S)
y
= _ M(i) M(s)
Y' yx
= _ MU) =0
rx'
Q(s)
y
= _ Q(i)
y
=0 (7.42)
for I x I < a and y = 0 while
./,(S)
"v y
= Q(s)
y
= M(s)
yx'
= 0 I x I > a·, y = 0 (7.43)
must hold outside the crack. There remains the determination of the seat-
tere d wave field.
Solution of duaf integraf equation. With the aid of Fourier transform, a

solution of equation (7.37) is
cjJjs\x, y, t) = 1 fOCi Bis) exp { - [f3jY + i(sx + wt)]}ds, Y ~ 0

-2 (7.44)
n -0Ci
where the funetions

- (s2 -CLj2)t -- - I C.(L2j - S 2)t·
13 j - ,j-- 1, 2, 3 (7.45)
represent the branch cuts. Equation (7.44) satisfies the conditions of vanish-
ing displaeements, moments and shear forees for the seattered waves suf-
fieiently far aw ay from the erack.
Let the unknowns Bis) (j = 1, 2, 3) in equations (7.44) be related to a
single funetion B(s) as
BJo(s) = (-l)j (1 - 2
( CL
l -
2)13 j CL~ B()
V)S2 -
CL2
s , j
0 = 12
, (7.46a)
(7.46b)
The eonditjons in equations (7.42) and (7.43) ean then be satisfied if B(s) is
governed by the following system of dual integral equations
- 1 fOCi B(s) exp( - isx)ds = 0, I x I ~ a (7.47a)

2n -0Ci
1 fOCi .
2n -0Ci sfis) B(s) exp( -lsx)ds =
2( a2 - l)M 0 (
(1 _ v)D exp -
iCL 1
.,}2
X) ,I x 1< a
(7.47b)
in whieh/is) is known:
_ 2 {_ (a2 - 1) [(1 - V)S2 - aD2
f2(S) - (1 - v2 )s (ai - ex~)J11
(al - 1) [(1 - V)S2 - a~Y 2(1 - V)S2 P3}
(7.48)
+ (exf - a~)J12 - ex~
By following a proeedure outlined in [6], a solution of equations (7.47) is
obtained:
B(s) = - 21t(~: -=- v~~Moa2 [f: v'e eP2(e) Jo(sae)de
+ f: v'e P2(e) Jl(Sae)de] (7.49)
The funetions eP 2 and P 2 ean be ealculated from the Fredholm integral

equations
l
ePie)- oL J
3 (e,r,)ePi11)d11 = v'eJo
(av'2ae)
1
(7.50)
(7.50b)
whose symmetric kernels are given by
L 3 (e, 11) = (e11)!- f~ s[!2(sja) + 1JJ o(se) J o(S11)ds (7.51a)
Lie, 11) = (e11)t f~ s[f2(sja) + 1JJ l(Se) J l(S11)ds (7.5ib)
These equations ean be solved numerieally for the intensity of the moment
distribution around the eraek tip.
Crack Iront moment distribution. The asymptotic behavior of the moment

distribution near the eraek ean be found by introdueing the polar eoordi-
nates rand 0 eentered at x = a and y = 0 such that
x = a + reosO, y = rsinO (7.52)
Keeping in mind that the moments of the ineident waves are nonsingular,
it is only neeessary to eonsider the seattered field for small r:
(7.53a)
(7.53b)
(7.53e)
The dynamie moment intensity faetor is
Kl = (1 - az) (1 + v) Mo.J a [<Pil) + i'l' il)] exp( - iwt) (7.54)
It is worthwhile noting that the funetional dependenee of Mx, My and M xy

e
on is identieal with that found by Sih and Loeber [8] for plane extensional
waves impinging on the eraek. In this eonneetion, the plate may be viewed
as an assembly of thin sheets bonded together and eaeh of the sheets is
behaving loeally at the eraek as if it were being stretehed and eompressed
°
by oseillating membrane forees. Note al so that the bending loads yield
finite values of Qx and Qy at r = whieh is in eontrast to the singular
solution in the classieal plate bending theory.
An alternative way of expressing Kl in equation (7.54) is
Kl = Mo.Ja I K\Zl I exp[ - iw(t - bz)] (7.55)

in whieh K\2) is given by
(7.56)
There is a phase angle between Kl and the input wave whieh ean be computed
from
5: _' -1 {Im[<pz(l)] + Re['l'i l )]}

Uz - - tan (7.57)
2n Re[<Pz(I)] - Im['l'z(l)]
Reeall that the input waves in equation (7.38a) eorrespond to
M(i)
y
= _ (1 + v)JVI 0 eos (iJ(lX)
2 ' M(i)
yx
= Q(i)
y
= ° (7.58)
with M~i) being the only non-vanishing moment. Henee, in the limit as
iJ(1~ 0, a state of pure bending with an applied moment of magnitude
(l + v)Mo prevails in the region where the stresses are undisturbed by the
eraek. Moreover, it ean be shown that equation (7.55) reduees to the statie
result of Hartranft and Sih [9] based on the Reissner theory of plate bending,
if h/JIO is replaeed by h/n.
Displayed in Figure 7.4 is the variations of the moment intensity faetor
1.0 ~
c\
1.2 11 = 0.25
0.8
;;;
-:.:: ;;;
-:.::-
~
OA 0
o 0.1 0.2 0.3
WIW o
~
;:r:
o 0.1 0.2 0.3 I:l
~
WIW o I:l..
M
Figure 7.4. Moment inteosity factor as a function of frequency for various ratios of a/h ~
in Mindlin's theory
Q
Figure 7.5. Effect of Poisson's ratio on moment intensity factor for a/h = 1 ~
0.20
v =0.25
0.16
f... 0.12
"-
OI
<xl
0.08
0.04
W/W.
Figure 7.6. Phase angle versus frequency for a/h = 1.0,0.5 and 0.25
I K\Z) I with w/w o for different ratio of a/h. All the curves are below the
static value of 1.25 and, I K\2) I decreases more rapidly with frequency when
a/h is increased. This general trend is similar to that of the classical plate
theory and the corresponding problem of moment concentration for a
circular hole [10] in a Mindlin plate. The effect of Poisson's ratio on the
moment intensity factor is shown in Figure 7.5 for a/h = 1.0. A larger value
of v tends to increase I K\Z) I for a given frequency. Finally, the phase angle
bz is plotted against w for v = 0.25 and a/h = 1.0, 0.5 and 0.25. For fixed
values of w and a, the phase angle between Kl and the incident waves is
seen to increase as the plate thickness is decreased as show n in Figure 7.6.
7.4 Kane-Mindlin's equation in plate extension
Sih [II] has discussed the two-dimensional problem ofplane waves impinging
on a through crack. The dynamic equations of plane elasticity are applicable
to the extensional vibration of plates provided the wave lengths are large in
comparison with the thickness of the plate. In terms of frequencies, the plane
elasticity equations yield good results for frequencies which are lower than
the frequency of the first mode of thickness vibration of the plate.
In order to overcome some of the foregoing shortcomings of the plane
elasticity equations, the higher order extensional vibration theory of Kane
and Mindlin [3] will be employed. This theory is also two-dimensional being
reduced from the three-dimensional theory of elasticity. However, the reduc-
tion is not carried out as far as to arrive at the elastodynamic equations in
plane elasticity and does account for the coupling between extensional
motions and the first mode of thickness vibration.
Governing equations. Consider a plate bounded by planes z = ±h/2 and

a right handed cartesian coordinate system (x, y, z) with the xy-plane lying
in the middle plane of the plate as shown in Figure 7.7. The components of
displacement are assumed to be approximated by
y
Middle
plane
Figure 7.7. Extensional waves striking a through crack
ux(x, y, t) = vix, y) exp( - iwt) (7.59a)
uy(x, y, t) = v/x, y) exp( - iwt) (7.59b)
uz(x, y, t) = 2hz vz(x, y) exp( - iwt) (7.59c)
for incident waves that vary sinusoidaIly in time with circular frequency w.
In terms of the potentials of the dilatation and rotation in the plane of the
plate, the displacement funetions Dx , Dy and Dz in equations (7.59) may be

expressed as
Dx = (7.60a)
oep _ 01/1
Dy = oy ox (7.60b)
vz=elepl +e2epZ (7.60c)

where
ej =
ep =
A~2
epl + ep2
(A + 2/1)
and
[)~ - (~ rJj =
The dilatationaI wave velocity is C l = [(A + 2,u)jp]t with). and ,u bei ng the
1,2 (7.61 )
Lame constant and p the mass density of the material. The three independent
displacement potentials epl' ep2 and 1/1 satisfy the equations
(V 2 + jj~)epj = 0, j = 1,2 (7.62a)
(V 2 + jj~)1/I = 0 (7.62b)
as derived from the equations of motion. The numbers jj j (j = 1, 2, 3) stand

for
2
jj j = 2fJ n )2
1( h [ (0: + fJ) ( W)2
õ5 - . J
1 - ( - 1YH , j = 1, 2 (7.63a)
jj~ = (wjc 2)2 (7.63b)

The quantities 0:, fJ and H are given by
(A + 2/1)2 A + 2p
0: = 4p(A + /I)' fJ = 4(A + /I) (7.64a)
The frequency
_ nC t
w= - (7.65)
Il
makes the Kane-Mindlin equations applicable to much higher frequencies

than those accommodated by the dynamic equations of plane elasticity.
Examination of bj (j = 1, 2, 3) in equations (7.63) shows that bi and b;

are always larger than zero but b~ is greater than zero for w/ (jj > 1 and less
than zero for wj(jj < 1. At the eut-off frequeney w = (jj, b~ is identieally
zero.
Referring to Figure 7.7, N x, Ny and N xy are the in-plane forees per unit
f
of length defined as
hl 2
(Nx, Ny, N xy) = (o"x, o"y' 'xy)dz (7.66)

- hl2
whieh in terms of the displaeement potentials <PI' <Pz and !/J may be written as
(7.67a)
(7.67b)
(7.67e)
The foree N= is h times the average transverse stress 0"= and is given by
7rJlh ~ z-
jL:, <PJ2b j - (1 - v)b 3 J
Z
Nz = 2~v (7.68)
f
The shear forees R x and Ry are obtained from
hlZ
(R x, Ry) = ('ZX' 'zy)z dz (7.69)

- hlZ
whieh in turn yields

Jllzz Z o<p j
R x =-6- I
j= I
er-;;;-:
uX
(7.70a)
_ JlIlZ ~ . ocjJj
Ry - 6 L... eJ ~ (7.70b)
j=1 dy
These quantities play a role analogous to that of transverse shear forees

Qx and Qy in the eorresponding plate bending theory.
[neident wave field. Let the eraek of length 2a in Figure 7.7 be engulfed by
a plane ineident wave. The resulting wave field ean be separated into two
parts: an ;neident wave field and a seattered wave field, i.e.,
(7.7la)
(7.71b)
The potentials <p ji) (j = 1, 2) and '" (i) are specified while the potentials
<p ~s) (j = I, 2) and '" (s) for the seattered waves must be found by solving
equations (7.62). Three types of input waves will be considered.
Suppose that the incident wave is a plane harmonic compressional wave
making an angle 1'1 with the x-axis along which the crack lies, i.e.,
<p~i) = cJJ t exp { - i [6 1 (x eos 1'1 + Y sin 1'1) + (Ut]} (7.72a)
(7.72b)
in whieh cJJ 1 represents the amplitude of the ineident P-wave. From equations
(7.67) and (7.68) the in-plane and transverse normal forees per unit length
are obtained:
N~i) = J1h<p~i) (26~ sin 2 1'1 - c5~) (7.73a)
N~i) = J1h<p~i) (2c5i eos 2 1'1 - b~) (7.73b)
N(i)
xy
= - 2j1hJ,(i)
'f' 1
15 12 sin"I 1 eos"I I (7.73e)
N(i) = 7rJ1~ rJ..(i) [215 2_ il _ v)b 2 ] (7.73d)

z 2~3v'Vl 1\ 3
while the ineident shear forees are given by

. 12
R(i) - _ ~ rJ..(;) e c5 eos y (7.74a)
x - 6 'VIII 1
. 12
R(i)
y
= _ ~
6 'rJ..(i)
VIII
e c5 sin"II (7.74b)
In the ease of an ineident wave with amplitude cJJ 2 impinging on the eraek
at an angle 1'2 with the x-axis, the potentials are
<p~) = cJJ 2 exp {- i [bix eos 1'2 + Y sin 1'2) + wt]} (7.75a)
<pli ) = ",(i) = 0 (7.75b)
The eorresponding in-plane and transverse normal forees beeome

N~i) = J1h<p~i) (2bi sin 2 1'2 - b~) (7.76a)
N;i) = j1h<p~i) (2bi eos 2 1'2 - b~) (7.76b)
N(i)
xy
rJ..(j) 15 22 sin"12 eos"12
= _ 2j1h 'V2 (7.76e)
(7.76d)
Similarly, the transverse shear forees are

. h2
R(i)
x
= - ~ 6 ,/..(i)
'1-'2
e2 82 eos Y2 (7.77a)
. h2
R(i)= _
y
L6 ,/..(i)
'1-'2
e2 82 sin Y2 (7.77b)
When ro/ö5 < 1, 4J~) goes to zero sinee in this ease 82 is imaginary. This
implies that the ineident wave attenuates near infinity.
If the ineident wave is associated with 83 and makes angle Y3 with the
x-axis, then the following representation prevails:
t/1(i) = t/13 exp {- i [b 3(x eos Y3 + Y sin Y3) + rot]} (7.78a)
4Jii) = 4J~) = 0 (7.78b)
with t/1 3 being the wave amplitude. Equations (7.78) give rise only to the
in-plane forees
N~i) =- 2{tht/1(i) 8; sin Y3 eos Y3 (7.79a)
N~i) = 2{tht/1(i) 8; sin Y3 eos Y3 (7.79b)
N(i)
xy
= r"h,I,(i)
'I'
832 eos 2y 3 (7.7ge)
while N(i)
x'
R(i)
x
and R(i)are
y
all zero.
Boundary conditions on the crack. The eraek in Figure 7.7 lying along the
x-axis from x = -a to x = a is assumed to be free of surfaee traetions.
This means that the forees per unit Iength Ny, N xy and Ry must vanish on
I x I < a and y = 0, i.e.,
N(i)
y.
+ N(s)
y
=0 (7.80a)
N(i)
xy
+ N{s)
xy
=0 (7.80b)
R(i)
y
+ R(s)
y
=0 (7.80e)
The problem ean thus be formulated in three parts in terms of the symmetry
eonditions of equations (7.80).
The symmetrie portion of the seattered wave problem refers to equation
(7.80a) such that

N~S) (x, 0) = - N~i) (x, 0), I x I< a (7.8Ia)
v~S) (x, 0) = 0, I x I~ a (7.8Ib)
N(s)
xy'
(x 0) = R\s)
y'
(x 0) = 0 , for all x (7.8le)
The skew-symmetrie seattered wave field is determined by the eonditions
N~) (x, 0) =- N~; (x, 0), I x I< a (7.82a)
v~s) (x, 0) = 0, I x I~ a (7.82b)
N(s)
Y'
(x 0) = R(s)
y'
(x 0) = 0 , for all x (7.82e)
Equation (7 .80e) represents the speeifieation of the negative transverse shear

foree R;i) on the eraek for the seattered wave problem i.e.,
R~S) (x, 0) = - R~i) (x, 0), I x I< a (7.83a)
v~s) (x, 0) = 0, I x I ~ a (7.83b)
N(s)
Y'
(x 0) = N(s)
xy'
(x 0) = 0 , for all x (7.83e)
This is analogous to the problem of anti-plane shear in the theory of elas-
tieity.
The quantities N~i), N~~ and R~) in the above equations are those given
in equations (7.73), (7.74), (7.76), (7.77) and (7.79). Three different formula-
tions referred individually to equations (7.81), (7.82) and (7.83) will now be
treated.
In-plane normal force specijied. With the aid of Fourier transforms and
inversion theorem, equations (7.62) ean be solved:
qyjS) (x, y) = 1
-2 Joo A/s) exp [ -(fJjY + isx)]ds. (j = 1,2); Y ~ 0 (7.84a)
n: -oo
(x, y) = 1 Joo Ais) exp [ -(fJ3Y + isx)]ds, Y ~ 0
-2 (7.84b)
n: -oo
ljt(S)
The braneh euts in the problem are deseribed by
{Jj = (S2 - c5W- = - i(c5] - S2)!, j = 1,2,3 (7.85)

The three unknowns A/s) (j = 1,2,3) may be expressed in terms of a single
funetion A(s) as given by
(7.86)
The satisfaetion ofequations (7.8Ia) and (7.8Ib) requires that A(s) be deter-
mined from the dual integral equations
J: oo A(s) exp( - isx)ds = 0, I x I ~a (7.87a)
J OO
-oo
nP.
gl(S) A(s) exp( - isx)ds = Il [1- (C~jCl)2] exp( - ixc5j eos y), I x I < a
(7.87b)
where j = 1, 2, 3 and gl (s) is a known funetion:
In equation (7.87b), P j (j = 1,2, 3) are defined as

P j = IlcPJc5~ - 2c5] eos 2 yJ,j = 1,2 (7.89a)
P3 = - 21l1jJ3c5~ sin Y3 eos Y3 (7.89b)
It ean be shown that
A(s) = n~~a2 [1 -(~:rJ-l J~ ~~[tpl(~)JO(sa~)+ tp2(~)

J 1 (sa~)]d~ (7.90)
satisfies the dual integral equations (7.87) provided that tp 1 and tp 2 are
determined from the following system of Fredholm integral equations
tplG) - f~ Lj(~,17) tpl(17)d17 = - ~~ Jo(c5ja~ eos Yj) (7.91a)
tp2G) - J: Li~, 17) tpz(17)d17 = - ~~ Jl(c5ja~ eos Yj) (7.91b)

The kernels L j and L 2 whieh are symmetrie in ~ and 17 take the forms
Lj(~, 17) = (~17)t J~ s[Gj(sja) + l]Jo(s~) Jo(s17)ds (7.92a)
Li~, 17) = G17)t f~ s[Gj(sja) + I]Jj(s~) Jj(s17)ds (7.92b)

in whieh G j is given by
Dynamie analysis of eraeked plates in bending and extension 255
In-plane shear foree speeijied. A solution of equations (7.62) that pertains

to the conditions specified in equations (7.82) is
ep)S) (x, Y) = 21
n
Joo -co
Bj(s) exp [ -([3jY + isx)Jds, (j = 1,2); Y 2 ° (7.94a)
1
l/1(s) (x, Y) = 2n Joo -co
.
B 3 (s) exp [ -(f33Y + isx)Jds, Y 2 ° (7.94b)
From equation (7.82c), the three unknown B/s) (j = I, 2, 3) can be related

to B(s) by the following relationships
(7.95b)
Equations (7.82a) and (7.82b) are the n satisfied by requiring that B(s) be
J:
the solution of the dual integral equations
co B(s) exp( - isx)ds = 0, I x I 2 a (7.96a)
f oo_ co g2(S) E(s) exp( - isx)ds = 2nQj

p[(I-(e 2(e 1)2J
exp( ixJ eos}')
- j j ,
Ixl<a (7.96b)
and g2(S) is given by
g2(S) = [2[31 [3zCe 1 -

([32e 2 -
e2)~2 + (2s 2
[31 el)(53
- :~)2J
2[33(53
[1 _(~)2 Jel
-I (7.97
The amplitude of the input wave is embedded in Qj (j = I, 2, 3):
Qj = prpj(5~ sin 2Yj' j = 1,2 (7.98a)

Q3 = - pl/1 3(5~ eos 2Y3 (7.98b)
Solving equations (7.96), an integral representation of B(s) is found:

nQ j a 2
B(s) = 2p
(7.99)
where 0/ 3 and 0/ 4 can be obtained by solving the Fredholm integral equa-
tions
lJ'i~) - J~ L3(~,I1) lJ'ir,)dl1 = - .J~ Jo(bja~ eos y) (7.100a)
IJ'4(0 - f~ L4(~' 11) 0/il1)dl1 = - .J~ J l(bja~ eos y) (7.100b)
The kernel funetions L 3 and L 4 stand for
L3(~,I1) = (~'1)t J~ s[G (sja) + 1] Jo(s~) Jo(s'1)ds

2 (7.101a)
Li~,I1) = (~'1)t f~ s[Gisja) + 1] J1(S~) J l(sl1)ds (7.IOIb)
with Gis) = gis)js.

Transverse shear force speeijied. Referring to the conditions specified in
equations (7.83), the following potentials are taken as solutions to equations
(7.62) :
4>jS) (x, Y) = 21 Joo eis) exp [ -(fJjY + isx)]ds, (j = 1,2); Y ;;::: 0 (7.102a)
n -ro
1 fro C 3(s) exp [-(fJ3Y
(x, y) = -2 - + isx)]ds, Y ;;::: 0 (7.102b)
n -ro
tjJ(s)
As before, the symmetry conditions given by equation (7.83c) reduce the

number ofunknowns from C/s) (j = 1,2,3) to CCs):
( s) = (2S2 - t5~)2 - 4fJ2fJ3S2 CCs) (7.103a)

(el - e2) (2s - t5e) + 4fJie 2 fJl - e l fJ2)s
el 2 2 2
e s - _ (2s 2 - 6~)2 - 4fJd3 3 s 2 es

2( ) - (el - e2) (2s 2 - 6;)2 + 4fJ3(e2fJl _ e l fJ2)s2 () (7.103b)
i(2s 2 - 6;) [ ~
C 3(s) = - 2SfJ3 (;l(S) + Cis)] (7.103e)
J:
The remaining conditions in equations (7.83) yield
oo CCs) exp( - isx)ds = 0, I x I ;;::: a (7.104a)
J-roOO
g3(s)CCs)exp(-isx)ds =
2niR..
_ _J exp(-lxc)j eos Yj), lxi <a
~
(7.104b)
in whieh g 3(S) is given;
( ) (Plel - P2e2) (2s 2 - b~)2 + 4PlP2P3(e 2 - e l )s2

(7.105)
g3 s = - (el - e2)(2s 2 - bD2 + 4P3(P l e2 - P2el)S2
In equation (7 .104b), R j (j = I, 2) takes the form
R j = JlcjJj ejbj sin Yj' j = 1,2 (7.106)
Jl
The system of dual integral equations (7.104) may be solved to give
inR.a 2
C(s) = _ J_ .J~ [Ps(~) Jo(sa~) + P6(~) Jl(sa~)]d~ (7.107)
Jl 0
such that PS and P 6 are to be found from
Ps(~) - f~ Ls(~,rJ) Ps(rJ)drJ = - .J~ Jo(bja~ eOSYj) (7.108a)
P6(~) - f~ L6(~,rJ) P6(rJ)drJ = - .J~ Jl(bja~ eos "I) (7.108b)
J:
with L s and L 6 being the kernels given by
Ls(~, rJ) = (~rJ}t s[G 3 (sja) + I] Jo(s~) Jo(srJ)ds (7.109a)
L6(~' rJ) = (~rJ)t J: s[ G3 (sja) + 1] J 1 (s~) J 1(srJ)ds (7.109b)

The funetion G 3(s) = g3(S)jS is given sinee gis) has al ready been defined by
equation (7.105).
Singular stresses. In terms of the loeal polar eoordinates rand () measured
from the eraek edge as shown in Figure 7.7, the normal forees N x , Ny, etc.,
and shear forees R x , Ry, etc., are unbounded as r ~ O. The nature of this
singularity ean be found from the appropriate integral solutions that repre-
sent N x, Ny, . .. , and R x, Ry, .... By earrying the integration for large
values of the variable of integration s, it is found that near the eraek edge
the following results prevail;
23(})
k () (1
eos 2 . () .
h- l Nx = _1_
(2r)!
- sm 2 sm
~ sin
(2r)!
2() (2 + eos 2() eos
3(}) + 0(1)
2" (7. 11 Oa)
h- l N y =
k
_1_ eos - () (1 + sm-
. () sm. -3() )
(2r)! 2 2 2
+ --,
kl . 0
Sln -2 eos
o eos T30
"2 + 0(1) (7.l10b)
(2r»
h-l N = kl O. o .
30
xy (2r)1- eos "2 sm -2 sm -2
+ kl, eos !!.. (1 _ sin !!.. sin 30) + 0(1) (7.1l0e)

(2r)" 2 2 2
h- l
N z = 2v
( re )[ kl
2 ~3
0
(2r)1- eos 2" - k2 . 0
(2r)1- sm"2
J+ 0(1) (7.1l0d)
The transverse shear forees beeome
6h- 2 R =
x
~
(2r)t
0
sin -2 + 0(1) (7.11la)
1 k3 ()
6h- Ry = (2r)1- eos "2 + 0(1) (7.111b)
Equations (7.110) and (7.111) show that the eraek front stress field in the
Kane-MindIin theory is three-dimensionaI in charaeter. In faet, the funetional
dependence of the six stress eomponents on rand () is the Slme as that
found by Sih [12] from the three-dimensional theory of elasticity. The only
exception is that equations (7.102) do not yield the exact plane strain con-
dition sinee
(7.112)
However, the discrepancy is not serious as the factor 1t/2~3 does not deviate
appreeiable from unity. The interesting aspeet of this theory is that there is
a eoupling between extensional motions and thiekness shear vibration as
evidenced by the existenee of all three stress intensity faetors
kl = [P l (1) - iPil)] Pj~a,j = 1k,2,3 (7.113a)
kl = [P 3 (!) - iPil)] Qj~a,j = 1,2,3 (7. 113b)
k 3 = [P 6(1) + jp 5(1)] R j ~ a, j = 1,2 (7.1l3e)
The funetions Pl' P l , . . • , P 6 in equations (7.l13) are eaIeuIated numeri-

ealIy from equations (7.91), (7.100) and (7.108) and evaIuated at the eraek
tip where the dimensionless variable ~ = 1.
The numerical values of the stress inetnsity factors in equations (7.113)

have been obtained [13] for v = 0.25 and various value, of a/h and wa(c 2 •
1.6
~
~- 0.8
0.4
o 0.4 0.8 1.2 1.6
Figure 7.8. Extensiona! stress intensity faetor versus wave number for input P-wave at
norma! incidenee
0.15 v = 0.25
>; = 90°
0.10
~a,-
.,
......
.><
2.0
o 2.0 4.0 6.0 8.0 10.0
Figure 7.9. Thickness mode stress intensity faetor versus wave number for input P-wave
at norma! incidenee
Figure 7.8 shows a pIot of the normaIized stress intensity faetor kt/Pl'!at t
in equation (7.113a) versus the normalized wave number OJa/c 2 whieh varies
from 0.0 to 1.6. The magnitude of the stress at the front of the input P-wave
is P I = 11<1>1c5~ for normal incidenee, i.e., YI = 90°. It is seen that the kl
values inerease at first reaehing a maximum and then deerease. The peak
value of kl increases as a/h is raised from 0.5 to 1.0 and then it start s to
decIine with the plane strain solution as a lower limit. The eoupling between
the extensionaI and thiekness motion gives rise to the stress intensity factor
k 3 in equation (7.113c) whieh when normalized with respeet to Pl> i.e.,
I k 3 /PI Ja I is plotted in Figure 7.9 as a funetion of OJa/c 2 . Note that the
amplitude of k 3 is mueh lower than that of kl shown in Figure 7.8. The
euryes have the same general trend of inereasing to a peak and then deeays
in amplitude. The peak decreases in magnitude as a/h is increased. This
shows that the thiekness eoupling effeet beeomes less and less important as
the plate thiekness is decreased. The oeeurrenee of k 3 alone eorresponds to
anti-pIane shear motion [14] exeited by input SH-waves.
If rf;3 is speeified, while <1>1 = <1>2 = 0, then k z in equation (7.113b) is the
only nonzero stress intensity faetor. At normaI incidenee, Y3 = 90°, equa-
tion (7.98b) gives Q3 = 11rf;3c5~ for the magnitude of the stress at the input
1.2
~
0 '"
..... 0.8
... N
0.4 v =0.25
>;= 90°
o 0.4 0.8 1.2 1.6
Figure 7.10. In-plane shear stress intensity faetor versus wave number for input shear
wave at normal incidenee
Dynamie analysis of eraeked plates in bending and extension 261
wave front. Plotting I k 2/Q3.ja I as a function of roa/e2' the results are

displayed graphically in Figure 7.10. Again, the curves increase in amplitude
for small values of a/h and then decrease as a/h is increased with the plane
strain solution as the li mit of an infinitely thick plate.
7.5 Plates subjected to sudden loading

Sudden loading is also referred to as shock or impact where the interval of
time involved in applying the load is very short. Common examples are
collision, an explosion, or some other mechanieal means. Since impact is
unavoidably accompanied by vibration, the duration of impact should be
compared with the fundamental period of natural vibration of the system
being considered. Generally speaking, if the time required to apply load, i.e.,
increase the load from zero to its maximum value, is less than half the fun-
damental natural period of the system, then the loading is regarded as im-
pact. In other words, the dynamic effects are negligible only if the time of
loading* is many times greater than the fundamental natural period.
The results obtained in this section on the transient response of cracked
plates apply to waves impinging on the crack as caused by the sudden
application of load and the case of sudden appearance of crack in a plate
under impact.
Classieal plate under impaet. The classical plate bending theory is used to
solve the problem of a cracked plate under impact. The source of impact is
the sudden appearance of uniform bending moment on the surface of a
crack of length 2a lying on the x-axis. Because of symmetry with reference
to the x-axis, it suffices to consider only the region of the plate where y > 0.
Hence the following mixed boundary conditions
M,(x,O,t) = -MoH(t), Vy(x,O,t) =0, lxi <a (7. 114a)
uy(x, 0, t) = V,(x,O, t) = 0, I x I ~ a (7. 114b)
are imposed on the solution of the differential equation (7.5). In equation

(7.114a), H(t) is the Heaviside unit step function. The initial conditions of
this problem is assumed to be zero while the displacements at distances far
away from the crack are required to vanish, i.e.,
* Periodically repeated loading has been discussed earlier.

Iim [uix, y, t), u,(x, y, t), uz(x, y, t)] = 0 (7.115)

(;x2 + r')"" oo
The formulation is similar to that of section 7.2 except that the problem
will now be solved in the Laplace transform domain. Let the Laplace trans-
form pair be wrltten as
g*(p) = J~ g(t) exp( - pt)dt (7.116a)
g(t) = 21 .
nl
f Br
g*(p) exp(pt)dt (7. 116b)
where the integral in equation (7.1l6b) is taken over the Bromwich path.
Applying equation (7.116a) to equation (7.5), the deftection w* in the trans-
form domain depends only on the space variables x and y:
(7.117)
In addition, if the Fourier cosine transform is employed on the variable x,

equation (7.117) reduces to an ordinary differential equation which can be
readily solved. With the aid of the Fourler inversion theorem, a solution in
terms of p is found:
w*(x,y,p) = ~ Joo {At(s,p) exp[ _ (S2 + ~2)t]
n 0
+A;(s,p)exp[ - (S2 - iX 2 )t]} cos(sx)ds, y ~ 0 (7.118)

The parameter iX is given by
-4 ph 2
0: = - - P (7.119)
D
Solving for the appropriate moment and equivalent shear expressions in
the transform domain and making use of equation 0.118), the conditions
in equations (7.114) lead to a pair of dual integral equations
-2 Joo A*(s,p) cos(sx)ds = 0, x ~ a (7. 120a)

n 0
-2 Joo f!(s, p) A *(s, p) cos(sx)ds = MD O, x<a (7. 120b)

n 0 p
in whichf1 is known, i.e.,
* _ (S2 _ iX2 )t[(1 _ V)S2 + ii 2]2 _ (S2 + ii2)t [(1 _ V)S2 _ ii 2]2
fl(S,P) - 20:2(S4 _ 0:4)t
(7.121)
Onee A*(s,p) in equations (7.120) is found, At and Ai in equation (7.118)

may be obtained from
-(1 - V)S2 + ,;x2
At(s, p) (S2 + ,;x2)t
2!2 A*(s,p) (7.122)
(1 - V)S2 + ,;x2
Ai(s, p) (S2 _ ,;x2)f
Following the proeedure for solving duaI integraI equations as outlined in

[6], it is found that
A*(s,p) = (1 _ :)~;: v)Dp J: ~~ ep!(~,p) Jo(sa~)d~ (7.123)
such that ept satisfies the Fredholm integraI equation
ept(~, p) + f~ Lt(~, t/, p) ep!(~,p)dt/ = ~ ~ (7.124)
whose kerneI being symmetrie in ~ and t/ is
Lt(~,t/,p) = (~t/)t f~ s[F!(sja,p) - 1] Jo(s~) Jo(st/)ds (7.125)
The funetion Ft takes the form

*() 2ft(s, p)
F 1 S, P = (1 _ v) (3 + v)s (7.126)
It is apparent from equations (7.123) and (7.118) that ept determines the
expressions of moments and shear forees in the Laplaee transform domain.
Referred to a system of IoeaI polar eoordinates r and e measured from the
eraek tip, the moments are
M*-
(2r)'
(1 -+ v) [1. .
K!(p) - - -smesm -3e - eos
x- - - ,
3 v 2 2 ~ J+ 0(1) (7. 127a)
M*=
y
Kt(p)
(2r)+
[I2" (1 +- v). . 23e
3 v sme sm - eos
J
2"e + 0(1) (7.127b)
*
M xy =
K!(p) (
(2r)+ 3 +1 ) [
v
.
2 sm
- v sm
"20 + -1 2 -
. e eos T30] + 0 ( 1)
(7. 127e)
The Laplaee transform of the moment intensity faetor is

(7.128)
1.0
a/h= 1.0
0.8
~
0.6
Q.
-
~
*-
090 0.4
0.2
o 1.0 2.0 3.0 4.0 5.0
c 2 /ap
Figure 7.11. Solution of Fredholm integral equation in the Laplace transform domain for
a Classical plate
NumerieaI values of ePT versus e2/ap is shown in Figure 7.11 for a/h = I and
2. Note that ePT rises sharply at first and then IeveIs off. The r- and 8-
dependenee of the moments for transient response in equations (7.127) is the
same as that in equations (7.22) for the steady-state ease. This eonclusion
was first made by Sih [15] who diseussed the eharaeter of the transient
stresses around a eraek under anti-plane strain. The transverse shear forees
Ki(p) ( 2 ) 38 (7.129a)
Q~ = (2r)! 3 + v eos '2 + ...
Q*y = Ki(p) ( 2 ) sl'n 38 + (7. 129b)

(2r)! 3+v' 2 ' ..
are again found to be singular of the order r- 3 / 2 as in equations (7.25).
Henee, the nature of dynamie Ioading affeets only the intensity of the moment
field in the neighborhood of the eraek tip and does not influenee the fune-
tional dependenee of the moments on rand O.
Now, let the inverse Laplaee transform of the funetion <PT< I, p)jp be
denoted by N\(t). Then, the moment intensity faetor as a funetion of time
beeomes
(7.130)
A plot of the normalized moment intensity faetor K\(t)jMoJa against

c2/tla is given in Figure 7.12 for a/h = I and 2. The dynamie moment
1.2
a/h = 1.0
0.8
~ .
::E
0.6
-
~
-,.:-
0.4
0.2
o 1.0 2.0 3.0 4.0 5.0
Figure 7.12. Normalized dynamic moment intensity factor versus time for a Classical
plate under impact
intensity faetor is seen to increase with time and approaeh the static value.
Deereasing plate thiekness tends to lower the K\ (t) value for small time t.
Sudden bending of a Mindlin plate. The effeet of bending on the transient

response of a cracked plate using the Mindlin theory [2] is considered. In

this theory, the rectangular components of the displacement vector are given
by
ux(X, y, z, t) = zl/l x(x, y, t) (7.131a)
uy(x, y, z, t) = zl/l/x, y, t) (7.131b)
uz(x, y, z, t) = I/Iz(x, y, t) (7.131c)

Refer to Figure 7.2 for the physical meaning of I/Ix' I/Iy and 1/1=. The expres-
sions for the moments and shear forees are the same as those shown in
equations (7.30) and (7.31) except that the exponential factor, exp( - iwt),
should be removed and I/Ix, I/Iy and I/Iz are to be understood as functions ef
x, y and t. Keeping this in mind, the equations ofmotion ofthree-dimensional
elasticity theory may be written in terms of the displacement functions:
(7. 132a)
(7.132b)
(7.132c)
Applying the Laplace transform to eq uations (7.132) gives
(7.133a)
(7.133b)
(7. 133e)
where p is the Laplaee transform variable. The reduetion of the preeeding

equations to aset of homogeneous ordinary differential equations neeessi-
tates the usage of the Fourier cosine and sine transforms. In the ease of
symmetrie bending with respeet to the eraek plane, t/ly and t/I= are eve n in x
and t/lx is odd in X. That is
t/lxCx,y,t) = - t/lx(-x,y,t) (7. I 34a)
t/ly(x,y,t) = t/ly(-x,y,t) (7.134b)
t/lz(x, y, t) = t/lz( -x, y, t) (7.1 34e)
It foIlows that the Fourier sine transform is applied to equation (7.133a)

and the Fourier cosine transform to equations (7.133b) and (7.133e). The
resulting eouplet ordinary differential equation ean then be solved together
with the Fourier inversion theorem to render
t/I~(x,y,p) = -2 Joo [sel - ai) Bi(s,p) exp( -fJiy) + sel - ai) Bi(s,p)
11: 0
X exp( - fJiY) - fJ~ B~(s, p) exp( - fJ~y)] sin(sx)ds, y ~ 0 (7. 135a)
t/I~(x,y,p) = -2 Joo [(l-aDfJiBi(s,p)exp(-pfy)

11: 0
+ (1 - ai) Pi Bi(s, p) (- f3!y) - sB!(s, p) exp( - P!y)] cos(sx)ds, y ~ 0

(7. 135b)
t/I~(x, y, p) = -2 Joo [Bi(s, p) exp( - pfy) + Bj,(s, p) exp( - f3!y)] eos(sx)ds,
11: 0
Y~0 (7.135c)
The parameters fJ j and aj are defined as
fJj = (S2 + (5)+, j = 1,2,3 (7.1 36a)
a*
J
= __!!.i
2
1- V
(-)2
ä3 '
j = 1 2
,
(7.136b)
and the numbers fi j (j = I, 2, 3) are given in terms of the variable p:

(7. 137b)
(7. 137c)
in which R and S have al ready been defined by equation (7.36) and W o is

the cut-off frequency nC2/h.
The unknowns Bj (j = 1, 2, 3) in equations (7.135) are to be determined
from the following boundary conditions
M/x, 0, t) = - MoH(t); Qy(x, 0, t) = MxY(x, 0, t) = 0, 0::; I x I < a
(7.138)
on the crack surface and the symmetry conditions

ljJ/x, 0, t) = 0; Qy(x, 0, t) = Mx/x, 0, t) = 0, I x I ;?: a (7.139)
on the x-axis outside the crack. Expressing Bj (j = 1 ,2, 3) in terms of a

single unknown B* given by
B*( ) _ B*(s, p) [(1 - V)S2 + ä~]2 (7.140a)
I S, P - f3i(a~ _ a;) (1 - V)S2 + a~
B*(s, p) [ 2 -2]
Bi(s, p) = B *( 2 2) (1 - v)s + a 2 (7. 140b)
2 al - a2
B* ()
3 S, P =
sB*(s,
2
p)
2
(1 -
) (* - a *)
val 2 (7. 140e)
al - a2
equations (7.138) and (7.139) reduee to the satisfaetion of a pair of dual

integrations:
-2 Joo B*(s, p) eos(sx)ds = 0, x;?: a (7.141a)

n 0
'J
.::.
n
Joo fi(s, p) B*(s, p) eos(sx)ds =
0
MO
p
D , °< x<a (7.141b)
with li being a known funetion:
!i(s, p) =
1 {I -{3* ai [
--2--:::Z (1 - v)s 2 + -2]2
al -
1-Bi aT [(1 - v)s
2
al - a2 I
+ ä~Y - (1 - v)2(ai - a s2 n {3!} (7.142)

Dynamic analysis of cracked plates in bending and extensiol1 269
A solution that satisfies equations (7.141) can be written in the form [16]
B*(s,p) = PD~~~2V2)J~ J~epH~,p)Jo(sa~)d~ (7.143)
such that epi can be computed from a standard Fredholm integral equation
of the second kind:
epi(~,p) + J~ q(~,IJ,p) epi(lJ,p)dlJ = J~ (7.144)

The kernel function in equation (7.144) is
Li(~,IJ,p) = (~IJ)1: J~ s[Fi(s/a,p) - lJ Jo(s~)Jo(slJ)ds (7.145)

where Fi is defined as
F*( ) _ 2fi(s, p)
2 s,p - (I _ v2)s (7.146)
For v = 0.25, equation (7.144) has been solved numericaIIy by Embley and
Sih [16] and the results for ep! plotted against c 2/pa are displayed graphicaIIy
in Figure 7.13 for a/h = I and 2.
1.0
li =0.25
0.8
a/h = 1.0
~-----
0.6
*oflo
OI
0.2
o 1.0 2.0 3.0 4.0 5.0
Figure 7.13. Solution of Fredholm integral equation in the Laplace trall5form domain for
a Mindiin plate
To obtain the solution near the crack tip, the Laplaee transform of the
moments are expressed in terms of infinite integrals whieh are eonvergent
everywhere except at the singular points eorresponding to the crack tips,
x = ±a. At these points the moments are unbounded and their behavior
ean be found by expanding the integrands for large values of the argument s.
By retaining only the highest order term in s, using the well-known Bessel
integral identities and applying the Laplaee inversion theorem, Embley and
Sih [16] have found that
Mx = Kt(t) eos -9[1 .9 (7.l47a)

2 sin 3%J 0(1)
- SlD- +
(2r)1- 2
Kt(t)
My = (2r)1-
eos -9[1
2
.9
+SlD-
2 sin 3%J + 0(1) (7.147b)
Kt(t) 9 . 9 39
M xy = eos - SlD- eos - +0(1) (7.147c)
(2r)1- 2 2 2
and the dynamic moment intensity faetor is
Kt(t) = Mo.Ja N 2(t) (7.148)

in whieh N 2 (t) is the inverse Laplace transform of I/>!(l, p)/p. The trans-
verse shear forees Qx and Qy are non-singular everywhere which is expeeted
1.0
1I=0.25
0.8
~ 0.6
~.
-...
~ 0.4
~-
0.2
0 1.0 2.0 3.0 4.0 5.0

c2 t1a
Figure 7.14. Normalized dynamic moment intensity factor versus time for a Mindiin
plate under impact
sinee the applied moment is symmetrie with respeet to the eraek and does
not give rise to anti-plane shear deformation of the eraek surfaee.
Figure 7.14 shows a plot of Kl(t)/Mo~a versus the normalized time
C2t/a for v = 0.25. The euryes eorresponding to a/h = I and 2 rise rapidly
within a short time interval reaehing their maximum and then begin to
oseillate. For a fixed eraek length, the Kt(t)-faetor for the thieker plate
reaches its maximum at a later time. In both cases, the dynamie moment
intensity is lower than the static limit.
References
[I] Kirchhoff, G., Über das Gleichgewicht und die Bewegung einer elastisehen Seheibe,
Journal reine u. angew. MallI., 40, pp. 51-88 (1850).
[2] Mindiin, R. D., Influence of rotatory inertia and shear on flexural motions of isotropie
elastie plates, Journal of Applied MeelIanies, 18, pp. 31-38 (1951).
[3] Kane, T. R. and Mindiin, R. D., High-frequency extensional vibrations of plates,
Journal of Applied MeelIanies, 23, pp. 277-283 (1956).
[4] Hartranft, R. J. and Sih, G. c., An approximate three-dimensional theory of plates
with applieation to crack probiems, Imernational Journal of Engineering Seienee, 8,
pp. 711 729 (1970).
[5] Achenbach, J. D., Wave propagation in elastie solids, North-Holland/American
Elsevier, Holland (1973).
[6] Copson, E. T., On certain dual integral equations, Proceedings of Glasgow Mathe-
matieal Association, 5, pp. 19-24 (1961).
[7] Sih, G. c., Paris, P. C. and Erdogan, F., Crack tip stress intensity faetors for plane
extension and plate bending probiems, Journal of Applied Mechanies, 29, pp. 306-312
(1962).
[8] Sih, G. C. and Loeber, J. F., Wave propagation in an clastic solid with a line of
discontinuity or finite crack, Quarterly of Applied Mathematies, 27, 2, pp. 193-213
(1969).
[9] Hartranft, R. J. and Sih, G. C., Effeet of plate thickness on the bending stress distri-
bution around through eraeks, Journal of Mathematies and Physies, 47, 3, pp. 276-291
(1968).
[10] Pao, Y. H., Dynamic stress concentration in an elastic plate, Journal of Applied
Mechanies, 29, pp. 299-305 (1962).
[11] Sih, G. c., Propagation of elastic waves around a crack, Proceedings of Third Con-
ference on Dimensioning, Budapest, pp. 577-588 (1968).
[12] Sih, G. c., Review of three-dimensional stress state in a eracked plate, International
Journal of Fraeture Mechanies, 7, I, pp. 39-61 (1971).
[13] Macdonald, B. D., The coupUng al in-plane and thiekness motions of a craeked plate,
Ph. D. thesis, Lehigh University, Bethlehem (1970).
(14] Loebef, J. F. and Sih, G. c., Diffraction of antiplane shear waves bya finite crack,
Journal of the Acoustical Society of America, 44, I, pp. 90-98 (1968).
[15] Sih, G. C, Some eIastodynamie probIems of eraeks, International Journal of Fracture

Mechanies, 1, 1, pp. 51-68 (1968).
[16] Embley, G. T. and Sih, G. C, Sudden appearanee of a eraek in a bent plate, Inter-
national Journal of Solids and Structures, 9, pp. 1349-1359 (1973).
Peter D. Hilton
8 A specialized finite element

approach for three-dimensional
crack problems
8.1 Introduction
The purpose of fracture mechanics is to predict the conditions of failure for

structural components containing flaws. Accurate stress analysis is required
to determine the stress field in the immediate vicinity of these existing flaws.
Typical examples of such flaws in structural members include surface cracks,
interior cracks, and through cracks. In each of these cases the near stress
field is generally three-dimensional in character. No analytic solution proce-
dures are available for considering the relevent problems and most previous
numerical analyses are based on the two-dimensional models of plane stress,
plane strain, or plate bending theories. Therefore, it is appropriate to attempt
to develop numerical stress analysis procedures applicable to the three-
dimensional crack problem s of interest. The finite element method is a logical
procedure to consider in this regard because of its ability to treat, in a
standard manner, a large range of structural geometries and boundary
conditions.
This chapter gives a brief review of three-dimensional finite element
procedures followed by a presentation of a specialized approach for finite
element applications to crack probIems. Example problem s involving a
through crack in a finite thickness plate subjected to symmetric loadings are
given to demonstrate this approach. The through crack geometry has been
chosen as the primary three-dimensional example for two reasons: first thi s
crack geometry is standard for test specimens and often observed in struc-
tural components. Second, literally hundreds of research papers have been
written concerning this crack geometry in the past two decades. The vast
majority of these papers employed two-dimensional models of plane stress
and/or strain to treat the through crack problem for tensile loading and
274 Peter D. Hilton
plating bending theories to analyze the effects of bending moments. There-

fore, it is appropriate to study the influence of three-dimensional effects on
this problem and to determine the range of parameters over which the
two-dimensional models yield adequate results for failure prediction.
8.2 Three-dimensional elastie ealeulations
For purposes of stress analysis a crack is generally modeled as a surface of

displacement discontinuity across which no forces are transferred. Using this
sharp crack edge model of the geometry in conjunction with an elastic,
homogeneous, isotropic model for material response, Hartranft and Sih [2]
have proven that the plane strain singular solution is asymptotically correct
at the crack edge except possibly at corners, i.e., at the points where the
crack edge penetrates the body surface.
The singular portion of the asymptotic solution along the crack edge is
given below for symmetric (mode I) crack probiems. A local cylindrical
coordinate system whose axis is tangent to the crack edge is used, Figure 8.1.
The rectangular (local coordinates) components of the stress tensor are
Portian of Crack Front
Plane Normal to
Crack Front at A
Figure 8.1. Local coordinate system at the crack front
k1(z)
(2r)+
cos sin ~ sin 3:) + 0(1) (8.1a)
ay = k1(z! cos
(2r)2
-29 (1 + sin -29 sin 329) + 0(1) (8.1b)
az = v(ax + ay) + 0(1) (8.1c)

A specializedjinite element approachJor three-dimensional crack problems 275
-r xy =
kj(z)
- - sm -
(2r}t
. 8
2
8
eos - eos -
2 2
38
+ 0(1) (8.ld)
T yz = 0(1), T xz = 0(1) as r -+ 0 ( 8.1 e)

The eorresponding strain eomponents are
kj(z) (I + v) 8 [ .
ex = - - , - - e o s - (1 - 21') - sm
(2r)2 E 2
~ sin 3:J+0(1)
= kj(z)sx(r, 8) + 0(1) (8.2a)
ey =
kj(z) (1
(2r}!: -e
+ 1')
eos
e[
2" (1 - 21') + sin"2 sin
8 38J
2 + 0(1)
= k z)B ir, 8) + 0(1)
j ( (8.2b)
ez = 0(1) (8.2e)
kj(z) 2(1 + v). e 8 38

'Yxy = (2r)+ - E - sm 2 eos "2 eos 2" + 0(1)
= k (z)Yxy{r,8) + 0(1)
j (8.2d)
'Yyz = 0(1), Yxz = 0(1) as r -+ 0 (8.2e)
and the displaeement eomponents reIative to the eraek edge are
Ux = (I + 1') ,[ 0 30
~ k 1 (z)(2r)' (5 - 8v) eos"2 - eos"2 + O(r)
J
= kj(z)üx(r,O) + O(r) (8.3a)
uy = (1~v) kj(Z)(2r)~[(7 - 81') sin ~ - sin 3:J + 0(1')
= kj(z)üy(r, 0) + O(r) (8.3b)
Uz = O(r) as r -+ 0 (8.3e)
where E is Young's modulus and v is Poisson's ratio.

The amplitude, kj(z), of this singular solution is not determined by the
asymptotie analysis refereneed above, but, depends on the global geometry
and loading eonditions for the partieular eomponent under eonsideration.
For the speeial ease of plane strain, kl (z) is eonstant along the eraek front,
i.e., k j (z) = kl' and is known as the stress intensity faetor. It ean serve as
276 Peter D. Hilton
a parameter to represent the effeet of load transmission to the eraek tip

region. Henee, regardIess of the eriterion of fraeture, it is useful and neees-
sary to ealeulate the stress intensity faetor for a speeimen and/or struetural
eomponent containing a eraek in terms of geometry and loading. This need
has led to eonsiderable interest and effort in the development of numerieal
proeedures for analysis of eraek probIems, partieularly for three-dimensional
problems where most of the analytieal methods become unmanageable. The
finite element method has reeeived particular attention beeause of its abiIity
to treat, in a standard manner, eomplicated geometries and boundary
conditions.
8.3 Finite element method-background
The finite element displaeement method [3] is based on the variational

prineiple ofminimum potential energy, employing a modified Ritz procedure
to obtain approximate solutions. The domain under eonsideration is divided
into subregions ealled elements. The displaeement field within eaeh element
is then approximated in terms of discrete (nodal) displaeement values using
interpolating polynominals known as displaeem~nt shape funetions. Using
this assumed form for the displacement field, the strain field, stress field, and
finally the strain energy of an element ean be expressed in terms of the
unknown nodal displaeement values. The potential energy for the body is
the n ealeulated as the sum of the strain energies of the elements minus the
work done by applied forees.
Minimization of this potential energy funetional with respeet to the nodal
displacement values results in aset of governing algebraie equations whieh
is to be solved for these diserete quantities. The stress and strain fields within
an element are then expressible in terms of these nodal displaeement values.
Two-dimensional finite element pro(;edures for eraek problem s have been
the subjeet of a number of recent review articIes [4, 5]. It has been reeognized
that the character of the crack tip singularity must be introdueed a priori
into the finite element method to obtain aeeurate values for the stress
intensity faetor (or faetors for nonsymmetrie situations). Several proeedures
have been developed for this purpose. The first, presented by W. K. Wilson
[6], involved the formulation of a eireular eraek tip element whose displace-
ment shape funetion was based on the asymptotic solution for the near tip
displacement field, equations (8.3) with kl (z) = kl; i.e., with the stress
intensity factor treated as a generalized nodal displaeement eomponent
A specializedjinite element approachjor three-dimensional crack problems 277
(degree of freedom). The approach was extended to elastic-plastic problems

through the introduction of the plastic singularity by Hilton and Hutchinson
[7]. Tracey [8] later developed an alternative approach using a sequenee of
pie-shaped elements to surround the crack tip with displaeement shape
functions that contain the square root dependenee of the near tip displaee-
ment field on distanee from the crack tip. Pian [9] developed a hybrid erack
tip element eonsistent with his standard elements. Stiil other approaches
have been presented. The one which is of interest for the present work was
due to Benzley [I]. This approaeh employs a reetangular grid pattern with
special treatment for the elements adjacent to the crack tip, i.e., their dis-
plaeement field indudes the asymptotie field, equations (8.3), as weil as the
standard terrns. The author is not aware of any analogous finite element
treatments for plating bending problems involving eraeks.
The only previous three-dimensional treatment of craek problems by
speeialized finite element proeedures of whieh the author is aware is due to
Traeey [10]. Mr. Traeey has extended his two-dimensional approaeh men-
tioned above to three-dimensions and treated both through eraek and
surfaee eraek (semi and quarter eirde erack edges were considered) probIems.
The Traeey approaeh enforees a square root singularity in the in-plane
strain eomponents and does not allow for non-singular eontributions to
these components in the immediate vieinity of the eraek edge. An alternative
approaeh employing Traeey's element goemetry with an extension of
Benzley's displaeement shape funetions to three-dimensions will be pre-
sented here. The new approaeh is be1ieved to be a signifieant improvement
to the work of Traeey beeause it allows for nonsingular eontributions to the
strain eomponents along the eraek edge. These contributions, while not
partieularly important for two-dimensional probIems, allow for the large
distortional strains with small (or zero) stress intensity faetor believed to
exist in the vieinity of the interseetion of the eraek edge with the body
surfaee.
The finite element approaeh to three-dimensional eraek problem s to be
deseribed here is applieable to eurved (smooth) eraek fronts; however, the
examples considered are for a straight eraek front norrnaI to the plate surfaee.
The straight crack front geometry is implicitly assumed in all two-dimen-
sional analyses and also it is employed in manufacture oftest speeimens. On
the other hand, test speeimens, after suffieient eraek growth, are observed
to contain a eurved (thumb nail shaped) crack front. Thus the straight crack
front assumed here is not believed to be the natural crack geornetry. One
278 Peter D. Hilton
aim of the present work is to explain the growth of the initial manufactured
crack front into the th umb nail shape observed in laboratory specimens.
8.4 Specialized elements for the crack edge
The finite element displacement method incorporating isoparametric e1ements

[3] is the basis of the approach to be presented. In general, an element is
enclosed by six surfaces with nodal points chosen along the element edges
(a five surface element will be introduced later in the text). A coordinate
transformation is employed to map the element into a cube, Figure 8.2. The
inverse mapping function can be expressed in the form
y
----f'=-IR-=::If---- P
.....0 , - - - - - - - x
z
Figure 8.2. Mapping of curvilinear finite elements
where i = 1, the number ofnodes associated with the element, and (Xi' Yh Zi)
are the coordinates of the ith node. The funetions N;(p, s, t) associated with
the mapping are polynomial funetions with the property
1 for p, s, t = Ph Si. tiethe point in the mapped shape which

{ corresponds to Xi' Yi' Zi)
Ni(p, s, t) =
ofor p,s,t = Pj,Sj,tjj =f. i (8.5)
The same funetions N;(p, s, t) are used in the approximation for the dis-
placement field within the element, i.e., the interior displacement component s
are expressed in terms of the corresponding nodal components as
(8.6)
It is now appropriate to mentian that these funetions are known as 'dis-
placement shape functions'.
The strain components are abtained in terms of the nodal displacement
components by differentiating equation (8.6), i.e.,
" aN·
s", = '-- ~ U"'i'
i uX
y",y = "'--i ~
oN U"'i + '-- -;-
uY
j
i
oN UYi
uX
" j
(8.7a)
Sy
" TaN·
= '-- U yt , (8.7b)
UYj
aN. aN·
Yzx = Li TuX U Zi + Li TuZ U Xi (8.7c)
where
oN j OX ay OZ aN;
ap ap ap ap OX
oN j OX ay oz oN j
(8.8)
as =
as as as ay
aN; ox oy oz oN;
at at at at az
is inverted to abtain the needed partial derivatives. This ealeulatian is
expressed formallyas
s = [B] <5 (8.9)
where
(8.10)
n = number of nodes per element
The stress components are related to the strain components through the
generalized Hooke's law
280 Peter D. Hiltan
0" = [D] e (8.11)
The strain energy in an element can then be expressed as
(8.12)
where [k] is the element stiffnes matrix given by
[k] = f~J~J~l [BY[D][B] detJ dp ds dt (8.13)
and where detJ is the determinate of the Jacobian for the transformations
given in equations (8.8).
The potential energy of the body, calculated as the sum of the element
strain energy contributions minus the work done by applied forces, is
minimized with respect to nodal displacement components to obtain aset
of Iinear algebraic equations for the determination of their values.
Elements which have an edge that contacts the crack edge are treated
speciaIIy. The approximate form for their displacement field is chosen so as
to incIude the crack edge singularity, equations (8.3), with unknown ampli-
tude kl (z) pIus the standard nonsinguIar terms for the corresponding isopara-
metric element. The variation of the stress intensity factor along the crack
edge is approximated in a piece-wise fashion consistent with the finite
element procedure. Letting s be the variable in the mapped space correspond-
ing to the crack edge, the stress intensity factor is assumed in the form
k 1(s) = L N~(s)ki
i
where the summation is for values of i from one to the number of nodes
along the edge. The dispIacement field in an element adjacent to the crack
edge is the n approximated in the local coordinate system shown in Figure
8.1 as
Ux = L Nb, s, t) [u x , - UXi L kjNj(s)] + ux (r,8) L kjNj(s) (8.14a)

i j j
Uy = L Ni(p, s, t) [u y, - u L kjNj(s)] + uk, 8) L kjNj(s)

i
Yi
j j
(8.l4b)
(8.14c)
where ux(r, 8) and uir, 0) are given by equations (8.3) and UXi' uy , are the
values of these functions at the ith node of the element.
The strain components are obtained by differentiating equations (8.14) as
ex = I Ni,x [U x, - ii xi I N~kj] +I N~exkj (8.15a)

i j j
ey = Ii Ni,y [U Yi - iiYi I
j
NjkJ +I
j
Njeykj (8.15b)
(8.15c)
x I
j
N~kJ +I
j
NHxyk j (8.15d)
Yyz = Ii
Ni,. [U Yi - iiYI I NjkJ
j
+I
i
Ni,y Uz,
+ I N'L (ii y, - I Ni iiy)k j (S.15e)

j i
Yzx = Ii Ni,z [U z, - ii zi I NjkJ

j
+I
i
Ni,z Uz,
+I Nj,z (ii xi - I Ni iix)kj (8.15f)

j i
The last summations in the equations for ex , ey, Yxy contain the strain sin-
gularity while all other terms remain finite at the crack tip.
Note that the nodal values of the stress intensity factor are treated as
generalized nodal displacement components, i.e., let
where n is the number of nodes along the crack front and m is the number
of nodes of the elements adjacent to the crack front. Then the strains can
be expressed in the form
e = [B*] ~*
and the element stiffness matrix for elements adjacent to the crack front is
given by
[k] = f~J~lJ~l [B*Y [D][B*] detJ dp ds dt (8.16)

It needs to be pointed out that in cases where the local coordinate system
(x, y, z) along the crack edge differs from the global coordinate system
(X, Y, Z), a coordinate transform [3] for the nodal displacement vector b
is required in the form
~ X.y,z =[A.]~ X,Y,Z (8.17)
282 Peter D. Hilton
Jl Jl IJl
The element stiffness matrix then becomes
[k] = -1 -1 -1 [AY [B*f [D][B][A] detJ dp ds dt (8.18)
8.5 Applications to crack problems
The results to be presented are based on use of 20 node isoparametric

elements with nodes at each element corner and an intermediate node along
each edge. This element yields a quadratic approximation in the mapped
space for the displacement field along each edge with some additional cubic
terms in the interior. Specialized element s with rectangular cross section
(the logical extension of Benzley's element) and with pie-shaped cross section
(similar to Tracey's element) have been tested. The pie-shaped configuration
was found preferable because it enabled more accurate modeling of circum-
ferential variations for the displacement field in the vicinity of the crack edge.
Typical grid patterns are shown in Figures 8.3 and 8.4.
Pie Shaped Element
Figure 8.3. First octant for center cracked plate with two layer grid pattem
The variation of the stress intensity factor along the portion of the crack
edge corresponding to a particular element was taken as quadratic in the
mapped space for consistency with the element chosen, i.e.,
Figure 8.4. Refined (four layer) grid pattem
3
kj(s) = I Nj(s)k j
1
where
Nt = ses - 1)/2
N";.=(L-s 2 )
N~ = ses + 1)/2
8.6 Detaiis of the analysis
The specialized finite element procedure described above was employed to

consider the problem of a through crack in a plate subjected to (a) tensile
and (b) bending loads. The in-plane dimensions of the plate were kept con-
stant at the values shown in Figure 8.5 and the loading was maintained at
either (Jo = 1 psi for the tension case or (Jo = 2y/h for the bending problem.
The influence of plate thickness was studied by considering two values for
h at 1.0 inch and 0.6 inch.
Note that the geometry chosen for consideration makes both the tension
and bending problem s symmetric about the xz and yz planes. Further, the
tension case is symmetric about the xy plane while the bending problem is
skew-symmetric about the same xy plane.
In this analysis it is assumed that no contact exists between the crack
faces; yet, the results for the bending case will indicate a negative stress
284 Peter D. Hilton
/
y
f}
/
V=0.3
e E = 1.0 psi
/-; L~ta-lb
x
1(---_----;
b
a = 2 in. .E..=
a 2
b = 4 in. ~ = I
e = 2 in. lL
2a = 025
. or 0.15
h = 1.0 in. or 0.6 in.
Figure 8.5. Plate geometry
intensity factor and material overlap in the compressive side of the plate.
This inconsistency is avoided by considering plates subjected to simultane-
ous tension and bending with sufficient tension applied to prevent the
overlap. Three-dimensional finite element caleulations for crack elosure
problem s associated with bending loads can be carried out using an iterative
procedure to determine the region of crack face contact. While this approach
is not conceptually difficult its application to three-dimensional crack prob-
lems requires excessive computation time and is therefore postponed in the
hope of determining methods which will substantially reduce computational
costs. The linear character of elasticity solutions allows for the use of the
superposition principle to add tensile and bending solutions in appropriate
proportions such that the result satisfies the consistency argument eve n
though the bending solutian itself is not meaningful. This is the appraach
taken here.
Finite element analyses were carried out for two grid patterns: the first
shown in Figure 8.3 contains 24 elements and 161 nodes: The second grid
is a refinement of the first developed by dividing each of the two layers of

elements in half in the thickness direction to form a four layer grid with
48 elements and 279 nodes, Figure 8.4.
One of the objectives of this work is to determine the variation of the
stress intensity factor through the specimen thickness and the variations of
the near field stress components and energies in the thickness direction and
with distance from the crack edge. Graphs illustrating these variations are
given in Figures 8.6-8.11 for the tensile case and Figures 8.12-8.14 for the
bending problem.
Surfoce of Plote
0.4
.E.
o =2
.fo = I
o
0.3
v=0.3
z
11
0.2
0.1
Midplone of Plote
o 0.5 1.0 1.5 2.0 2.5
Figure 8.6. Variation of stress intensity factor along crack front, tensile loading
Notice that for the case of tensile loading the stress intensity factor is
al most constant along most of the crack edge but decreases sharply near the
plate surface. The slope or rate of decrease is increased as the grid pattern
is refined; however, the numerical values for the stress intensity factor at the
free surface are far from zero. Considerable research efforts concerning this
behavior of the stress intensity in the vicinity of the free surface have been
reported by Hartranft and Sih [2, 11, 12]. Their results indicate that a thin
286 Peter D. Hilton
Surface of Plate
0.4 4 Layer Grid

1L=0.25{
2a 2 Layer Gri d ;a =0.25
0.3
.1.. ~=2
h
0.2
t=1 ta =0.15
1I=0.3
0.1
Midplane af Plate
o 0.5 1.0 1.5 2.0 2.5
Figure 8.7. Influence of procedure on the variation of the stress intensity factor along the
crack front, tensile loading
boundary Iayer is expected at the free surface across which the stress inten-
sity factor varies very rapidIy and that it should approach zero at the free
surface. Their caleuIations further clearly indicate that a numericaI proce-
dure, as used here, without special treatment of the boundary layer region
is not adequate to model these rapid variations and should not be expected
to result in zero stress intensity at the free surface. On the other hand, the
finite element procedure used here enables one to enforce kl Ch12) = O. That
condition has been imposed on some of the runs reported here and com-
parisons with and without enforcing kl (hI2) = 0 are shown in Figures
8.7-8.10 for the tensile loading situation and in Figures 8.12-8.14 for the
bending case. Notice that, to the accuracy of the resuits obtained and re-
ported in graphical form by the finite element procedure, the interior fields
are not substantially altered by the introduction of the surface constraint,
k l(h!2) = O.
It is appropriate to make some estimates of the accuracy of the results
presented here. For that purpose, a two-dimensional finite element calcula-
0.5
4 Loyer Grid} ..!L=o 25
--...."""'--~-2Loyer Grid 20 .
0.4
..h..=015
20 .
2. Loyer Grid
0.3 JL=2
o
z ~=I
Il o
V=0.3
0.2 .h.
20 .
=0 15
2 Loyer Grid
;0=0.25
0.1
4 Loyer Grid
i" =0.08 .!..
Midplone of Plote o =0.02
0 2 4 6 8 10 12
rr/a;,
Figure 8.8. Variation of the in-p!ane norma! stress through the p!ate thickness
tion (plane strain) was performed using the NSRDC [13] twelve node iso-
parametric element program with a circular singularity element at the crack
tip. Results from that program with appropriate grid patterns are believed
to be accurate to 2%. The value of the stress intensity factor obtained from
the two-dimensional finite element analysis is kl = 1.96 (5 oJ a. This result
is within 10% of the interior model values obtained for the stress intensity
factor using the three-dimensional finite element solution procedure for the
case of tensile loading. It is interesting to note that the three-dimensional
results for the stress intensity in the interior of the plate are actuaUy larger
in magnitude than that obtained from the two-dimensional analysis. As the
finite element method generally underestimates the primary unknowns (in
this case, the nodal values for the stress intensity and for the displacement
components), this observation is an indication that the numerical resuits
for the stress intensity factor obtained from the three-dimensional analysis
are at least as accurate as is implied by the 10% maximum deviation from
the two-dimensional predictions. Tracey [10] al so observed that comparable
288 Peter D. Hilton
2 Loyer Grid
4 Loyer Grid
}.!!....:o
20 .
25 )
l
Il ..!!.: 2
o
0.2 .fo: I
o
li :0,3
h
20: 0 .15
~:O
20 .
15
2 Loyer Grid Midplane of Plale
-I o 2 3 4 5
Figure 8.9, Variation of transverse normal stress at 8 = 0 through the plate thickness,
tensile loading
grid patterns yielded more aeeurate results for three-dimensional analysis

of eraek problems than for eorresponding two-dimensional ealeulations.
Comparative results required to estimate the aeeuraey of the bending
solutions are more diffieult to obtain. Solutions for an infinite plate eontain-
ing a through eraek and subjected to a uniform far field bending moment
distribution have been obtained using dassieal plate bending theory in 1962
by Sih et al [14]. The solution yields kl = (12zl h3 )M o.J a. Hartranft and Sih
[15] solved the same problem employing the Reissner plate bending theory
in 1968. The latter solution results in a thickness effeet predietion, i.e.,
12z
kl = F(h la) h3 Mo .Ja
where the function F(hla) has been determined numerieally and presented
a graphieal form. The author is not aware of analogous solutions for fi nite
size plates whieh would prediet the inftuenee of the in-plane dimensions
(bla and ela) on the stress intensity amplitude as weil as distribution. A crude
Surfoce of Plole
~=2
o
.f..: I
o
ZI = 0 .3
0 .3
z f =0 .08
"il .h...= 015
2.0 '
0 ,2 2 Loyer Grid
i =0,08 i-=0 .02

h ..!L=0. 15
0 ,1 20 =0 .25 , 20
4 Loye r Grid \ 2 Loye r Gr id
\
\
Midplane of Plole
\
°
-0 . 1 0 ,1 0 .2 0.3 0.4 0 ,5 0 .6 0 .7 0 .8
Figure 8.10. Degree of plane strain constraint at 0 = 0, tensile loading
estimate of this effect can be obtained from the plane strain solution discussed
aboveo That solution indicates that the stress intensity factor for the in-plane
geometry considered here is approximately 1.96 times the corresponding
value for an infinite plate. Jf a similar influence of in-plane dimensions is
anticipated for the bending problem, the Sih et al and Hartranft - Sih
solutions given above should be muItiplied by 1.96 before comparison with
the finite element resuits reported here. The normalized slopes for the stress
intensity factor against distance from the mid-plane [(k I /rJo.J a)/(2z/h)]
reported by Hartranft and Sih and corrected for the in-plane plate dimensions
as discussed above are 1.1 7 and 1.37 for h = 0.6 and 1.0 respectively.
Three dimensional finite element predictions for the transverse variation
of the stress intensity factor in the bending case are reported in Figure 8. I 2
for the two and four layer grid patterns employed. The two layer grid pattem
ealeulations have been performed for both the cases of kl treated as a free
parameter at the plate surface and kl constrained to be zero at that surface.
(Results are shown in Figures 8.12-8.13 for 1l/2a = 0.25). The four Iayer grid
290 Peter D. Hilton
Surfaee of Plote
0.5
0.4
ta
i=OoOB
=0.25 4 Loyer Grid
003
*=2
..l.
h
t=1
002 ZI=003
Ool
Midplone of Plote
o 2 3 4
r 4 jJ. dW
Õ 1-211 (1': dV
Figure S.lla. Variation of strain energy density minimum «(} = 0) across plate thickness,
tensile loading, Il = 005a
pattem ealeulations have only been earried out for the constrained situation.
Notice that for the bending problem, unIike the tensile ease, the four layer
grid pattem ealeulation yields larger values for the stress intensity faetor in
the plate interior than those predieted by the two layer grid results. This is
an indieation that the finer grid pattem is neeessary to obtain aeeurate
numerieal results for the plate bending problem. These fine grid ealeulations
result in normalized slopes for the stress intensity faetor (ktfuo../a)j(2zjh)
in the specimen interior of 1.23 and 1.25 for h = 0.6 and 1.0, respeetively.
The trend of inereasing normalized slope for the stress intensity faetor
predicted by Hartranft and Sih is confirmed by the present results. Further,
the magnitudes for this slope, reported here, are in reasonable agreement
with those given by the Hartranft - Sih ealeulations.
8.7 Results of the finite element analysis
The finite element solution was obtained for the geometry shown in Figure
A specializedjinite element approaehJor three-dimensional craek problems 291
0.5r-------------~~--~-----, __--------------------------
0.4
.h. t=0.08
20 =0 . 15
0.3
~~2
z o
h %= I
0.2 ZI = 0.3
0.1
Midplone of Plole
o 2 3 4
r 4 fL dW
a 1-211 2dV
0;;
Figure 8.11 b. Variation of strain energy density minimum (0 = 0) across plate thickness,
tensile loading, h = 0.3a
8.5. From dimensional considerations, there are three basic geometrical

parameters: the ratios of plate width, plate length, and plate thickness to
crack length. These ratios, bla, ela, hl2a, take the respective values, 2.0, 1.0,
0.25 in one case, and 2.0, 1.0, 0.15 in another. Because of the linear nature
of the elastic analysis, the stress state may be normalized by the applied
stress, 0"0 which is uniform for the tensile case.
The basic result, i.e., the variation of the computed stress intensity factor
along the crack front from midplane to the surfaee, is shown for the case of
tensile loading in Figure 8.6. For the two plate thicknesses used the values
for the stress intensity factor are fairly cIose along the center portion of the
crack front, but deviate significantly near the surfaces. It is expected that a
wider range of plate thicknesses would produce larger differences throughout
the plate. The zero value of the stress intensity factor at the surfaee of the
plate has been imposed on the problem. The effeet of aHowing the finite
element method to determine the value at the surface, and the accuracy of the
results can be judged by referring to Figure 8.7.
292 Peter D. Hilton
1!...=0.25
20
0.4 2 Loyer Grid
0.3
z
h
h
0.2 20 =0.25}
h 4 Loyer Grid
20 =0.15
0.1
Midplone of Plote
o 0.25 0.5 0.75 1.0
kl/crov'õ
Figure 8.12. Variation of stress intensity factor across plate thickness for bending case
The variations in Figure 8.7 of the stress intensity faetor show the effeets
of inereasing the number of elements modeIing the plate. For h/2a = 0.25,
there are two euryes whieh both result in kl = 0 at the surfaees. Comparison
of them wiII reveal that they differ only sIightIy near the surfaee of the plate.
The eurve eorresponding to four layers (48 elements) gives higher values
than that for two layers (24 elements) of elements. This indieates a more
rapid deerease of stress intensity faetor in the boundary layer for the more
aeeurate solution.
A eomparison of the euryes in Figure 8.7 for h/2a = 0.25 whieh result
from the four layer grid pattem may be made. For one of the euryes, the
value of kl at the surfaee was determined from the finite element solution
in the same way as at all other points. For the other, this value was chosen
to be zero. The euryes may be seen to be indistinguishable for z/h < 0.3, and
quite distinet at the surfaees. Nevertheless, the deerease shown at the surfaee
by the eurve in whieh kl at the surfaee was left free reinforees the suspieion
[2, 16, 17] that in the boundary layer, the stress singularity is relatively le ss
important.
Surfoce of Plote
0.5r---------------------~~~~====~~=_--~~---
0.4
0.3
z
11
0.2
0.1
Midplone of Plote
o 2 3 5 6 7 8
Figure 8.13. Variation of Uy/uo across plate thickness for bending problem
To further investigate the condition of the neighborhood of the crack front,

stresses were computed. These stresses indude the nonsingular as weIl as
the singular parts. For the positions considered, the nonsingular portion is
seen to be considerable. The normal stress component, (ly, aeti ng on the
plane directly ahead of the crack front is shown in Figure 8.8. The two
families of curves are for two distances ahead of the crack front. The larger
stresses act at the doser points a distance of O.02a, and the smaller at O.08a.
The decrease in stress as the surface is approached is primarily due to the
decrease in the singular terms (compare Figures 8.6 and 8.7). However,
since the singular terms are zero at the surface for all but two curves (shown
dashed), the surface values of (ly must be due to nonsingular effects. At the
larger distance from the crack, the singular part is relatively small.
The transverse normal stress, (lz> acting on planes paralleI to the plate
surfaces is shown in Figure 8.9 for the same thicknesses and distances from
the crack front used for (ly" Again, as for Figure 8.8, the dashed curves were
obtained without constraining kl to be zero at the surfaces. This stress
should be zero at the surface, but because of the nature of finite element
294 Peter D. Hilton
Surface af Plate
0 . 5 1 - - -.........~~====--L----7---------
0.4
0.3
lL = 0.25
z 2a } 4 Layer Grid
Il
~a=0.15
0,2
0.1
Q 0.5 1.0 1.5

r 4 fL dW
il 1-211 2' dV
(To
Figure 8.14. Transverse variatian of normalized strain energy density for bending case
methods imposing displacement fields, it is not. The surface values are low
compared to other values, and so the small discrepancy is tolerable. The
element interfaces for the four layer grid pattem occur at 2z/h = 0.2, 0.4,
and 0.45. Rapid changes in magnitude observed for (Jz/(Jo at these points
(Figure 8.9) are numerieal and, as such, indicate the order of inaccuracies
associated with the finite element caleulations employing this grid pattern.
The singular portion of the crack tip stress state is known [2] to satisfy
the plane strain condition
(Jz = v((Jx + (Jy)

Figure 8.10 ilIustrates the effect of the nonsingular terms on the validity of
this condition for points on the plane directIy ahead of the crack. The ratio
of (Jz to the plane strain value is plotted as a function of distance from the
midplane of the plate for the various parameters considered. The cIoser to
the crack front that the ratio is computed, the cIoser to the plane strain
value of one it becomes. In addition, for a given distance from the crack
front, the deviation from the plane strain value increases as the surface is
approached. Thus, as the model of reference [17] proposed, the stress sin-
gularity may be taken to be a state of plane strain except in a boundary
layer near the surface. Notice, finally, that the thinner plate deviates signifi-
cantly more from plane strain than the thicker plate.
The plane strain value of the strain energy density for an infinite, uniformly
stressed, cracked body has been discussed in [18]. The minimum value of
the limit
is attained in the plane directly ahead of the plane of the crack. The fracture
criterion proposed in [18]* is equivalent to several others in cases such as
this in which only Mode I loading is present.
In this three-dimensional case, the strain energy density, d W/d V, was
computed for several values of distance, r, from the crack front and several
positions through the thickness. The results are normalized by the plane
strain value given above. As shown in Figure 8.11, this value,
r 4 .J!... dW
a 1 - 2v (T~ dV
is large at midplane, reflecting the large stress intensity factors there. The
values decrease as the surface is approached but, because of the nonsingulaJ
terms included, not to zero. The points closer to the crack front show the
sharpest decrease at the surface. For values of r close to zero, the curves ir
Figure 8.11 approach the variation of the normalized stress intensity factor
squared as obtained from Figures 8.6 and 8.7, i.e.,
r
a
4
1 - 2v
dW
dV
+k 2
(Toa
as r-+O
The finite element solutions for bending were obtained using the same
plate geometries (Figure 8.5) and grid patterns (Figures 8.3 and 8.4) as
considered in the tensile loading case. The tractions applied to the surfaces
y = ± e for this case are
* Fracture occurs in the direction of minimum S at a load which causes Smin to reach a
critical value, Sor, characteristic of the material.
296 Peter D. Hilton
(J = 2 (JOZ
h
with (Jo = 1 psi
leading to a bending moment per unit length of
Mo = (Joh 2
6
Numerical results for the variations of the stress intensity factor across
the tensile portion of the plate thickness are given in Figure 8.12. Calcula-
tions were performed for two values of the normalized plate thickness,
h/2a = 0.15 and 0.25. For the thicker plate computations were carried out
with the stress intensity at the plate surface unconstrained and with the zero
values of the stress intensity factor imposed at the plate surface. Unlike the
tensile case, the results for the stress intensity factor with no surface con-
straint imposed show little tendency to drop off at the free surface. In this
case the tendency of the numerical results for the stress intensity factor to
approach zero at the free surface is opposed by the behavior interior to the
plate for which the stress intensity factor is expected to increase with distance
from the mid-plane.
Notice further that, uniike the tensile loading case, refinement of the grid
pattem in the transverse direction for the bending problem results in signifi-
cantly improved predictions for the stress intensity factor in the plate
interior. This observation indicates that the four layer grid pattem is need ed
to obtain accurate solutions for the plate bending problem considered here.
In comparing the results presented in Figure 8.12 for the fine grid calcula-
tions with the stress intensity factor constrained at the plate surface, it is of
interest to note that the relative 'boundary layer' thickness increases as plate
thickness decreases. Similar observations can be made for the tensile loading
case. They indicate a more dominate influence of the plate surfaces on the
interior solution for thinner plates and, as such, suggest that two-dimensional
models become less applicable as plate thickness is decreased.
Figure 8.13 contains curves for the variation of the normalized stress
component ((Jy/(Jo) normal to the crack plane along the line r = .Ola directly
ahead of and paralleI to the crack edge. These curves al so show the influence
of the conditions imposed at the plate surface and the thickness effect on the
near field transverse distributions.
Figure 8.14 contains plots of the normalized strain energy density directly
ahead of the crack edge (r = .Ola) as a function of the transverse variable

2z/h.
8.8 Summary
A three-dimensional finite element procedure has been developed for treat-

ing crack probIems. Results were obtained for problems involving a plate
containing a through crack and subjected to tensile or bending loads.
Transverse variations for the stress intensity factor and stress components
in the vicinity of the crack edge were reported. These results indicate the
infiuence of plate thickness and demonstrate the Iimitations of previously
reported solutions based on the two-dimensional models of plane stress.
plane strain, and plate bending theories.
References
[1] S. E. Benzley, Representation of Singularities with Isoparametric Finite Elements,

Int. J. Nam. Meth. Eng., Vol. 8, pp. 537-545 (1974).
[2] R. J. Hartranft and G. C. Sih, The Use of Eigenfunction Expansions in the General
Solution of Three-Dimensional Crack Problems,!. Math. Meeh., Vol. 19, pp. 711-729
(1970).
[310. C. Zienkiewicz, The Finite Element Method in Engineering Scienee, McGraw-Hill,
London (1971).
[4] P. D. Hilton and G. C. Sih, Application of the Finite Element Method to the Cal-
culations of Stress Intensity Factors, Methods of Analysis and Solations of Craek
Probiems, pp. 426-483, (G. C. Sih, editor), Noordhoff International Publishing,
Leyden, The Netherlands (1973).
[5] J. R. Rice and D. M. Tracey, Computational Fracture Mechanies, Namerieal and
Compater Methods in Straetaral Mechanies, (S. J. Fenves, editor), Academic Press,
New York, pp. 585-623 (1973).
[6] W. K. Wilson, Combined Made Fraetare Mechanies, Ph. D. Dissertation, University
of Pittsburgh (1969).
[7] P. D. Hilton and J. W. Hutchinson, Plastic Intensity Factors for Crack Probiems,
Eng. Fraet. Meeh., Vol. 3, pp. 435-451 (1971).
[8] D. M. Tracey, Finite Elements for Determination of Crack Tip Stress Intensity
Factors, Eng. Fraet. Meeh., Vol. 3, pp. 255-265 (1971).
[9] T. H. Pian, P. Tong and C. H. Luk, Elastic Crack Analysis by a Finite Element
Hybrid Method, Prac 3rd Int. Conl on Matrix Methods in Straet. Meeh., Wright-
Patterson Air Force Base, Ohio, (November 1971).
[10] D. M. Tracey, Finite Elements for Three-dimensional Elastic Crack Analysis,
Naelear Eng. and Design, Vol. 26, pp. 282-290 (1973).
298 Peter D. Hilton
[11] R. J. Hartranft and G. C. Sih, Alternating Method Applied to Edge and Surfaee
Craek Problerns, Methods of Analysis and Softttions of Craek Probiems, pp. 179-238,
(G. C. Sih, editor), Noordhoff International Publishing, Leyden, The Netherlands
(1973).
[12] G. C. Sih, Three-Dirnensional Stress State in a Craeked Plate, Int. J. Fraet., Vol. 7,
pp. 39-61 (1971).
[13] P. D. Hiiton, L. N. Gifford Jr. and O. Lornaeky, Finite Element Fraeture Meehanies
Analysis of Two-Dirnensional and Axisyrnrnetrie Elastie and Elastic-Plastie Craeked
Struetures, NSRDC Report 4493, (November 1974).
[14] G. C. Sih, P. C. Paris, and F. Erdogan, Craek Tip Stress Intensity Faetors for Plane
Criterion and Plate Bending Problerns, Transactions, Am. Soe. Meellanieal Engineers,
J. Appi. Meeh., Vol. 29, pp. 307-312 (1962).
[15] R. J. Hartranft and G. C. Sih, Effeet of Plate Thiekness on the Bending Stress Distri-
bution Around Through Craeks, J. o{Maths. and Plrysies, Vol. 47, No. 3, pp. 276-291
(1908).
[16] G. Villarreal, G. C. Sih and R. J. Hartranft, Photoelastie Investigation of a Thiek
Plate with a Transverse Craek, J. Appi. Meeil., Vol. 42, pp. 9-14 (1975).
[17] R. J. Hartranft and G. C. Sih, An Approxirnate Three-Dirnensional Theory of Plates
with Applieations to Craek Problerns, Int. J. Engr. Sci., Vol. 8, pp. 711-729 (1970).
[18] G. C. Sih, A Speeial Theory of Craek Propagation, Metllods of Analysis and Solutions
of Craek Probiems, pp. XXI-XLV, (G. C. Sih, editor), Noordhoff International
Publishing, Leyden, The Netherlands (1973).
299
Author' s index
Aehenbaeh, J. D. 234,271 144, 150, 151, 153, 156, 157, 159, 160,
Ambartsumyan, S. A. 163, 199 162, 198, 201, 219, 229
Ang, D. D. 130, 150, 151, 157, 159, 160, Friedrichs, K. O. 47 [5] [6]
201,229
Apeland, K. 163, 199 Gifford, L. N., Jr. 287,298
Goldenweiser, A. L. 54,82
Badalianee, R. 85 Gradshteyn, I. S. 173, 179, 199
Benthem, J. P. XLIII, XLVII Green, A. E. 47,82
Benzley, S. E. 277,282,297 Gupta, G. D. 173, 179, 199
Boussinesq, J. 150, 160
Bridgman, P. W. XVIII, XLVII Hagendorf, H. C. 137,159,201,202,203,
218,229
Calder, C. C. 47,82 Hartranft, R. J. XXIX, XXXIII, XXXIX,
Cha, B. C. K. XXIII, XLVII XLV, XLVI, XLVII, XLVIII, 45, 47,
Chen, E. P. 85,231 50, 55, 58, 60, 62, 65, 66, 71, 74, 75,
Conrad, D. A. 163, 199 77, 80, 81, 82, 83, 85, 90, 104, 105,
Cook, T. S. 172, 179, 199 115, 163, 199, 222, 229, 231, 245, 271,
Copley, L. G. 162, 198,202,229 274,285,288,289,290,292,294,297,298
Copson, E. T. 102, 115, 236, 244, 263, Hayashi, T. 10,42
271 Hilbert, D. 48, 52, 82
Courant, R. 48, 52, 82 Hilton, P. D. 106,115,273,276,287,297,
298
Do, S. H. 139, 160 Hutehinson, J. W. 277,297
Dobreff, P. S. 202, 219, 221, 225, 229
Dressler, R. F. 47,82 Ichikawa, M. 27,42
Duncan, M. E. 162, 198 Isida, M. 1,7,24,42
Embley, G. T. 269,270,272
Erdelyi, A. 127,135, 159 Kainins, A. 203, 229
Erdogan, F. XX, XXII, XXVI, XLV, Kane, T. R. 231,247,248,249,258,271
XLVII, 5, 27, 43, 60, 83,161,162,163, Kantorovieh, L. V. 134, 159
172, 179, 181, 192, 198, 199, 202, 219, Kassir, M. K. 60, 82
221, 229, 238, 271, 288, 298 Kibler, J. J. 162,198,202,219,221,229
Kipp, M. E. XXV, XXVIII, XLVII
Fersht, S. W. 54, 82 Kirehhoff, G. 231,234,271
Flügge, W. 163, 199 Knowles, J. K. 66,83,137,159,201,229
Folias, E. S. 117,131,132,137,139,140, Krylov, V. I. 134, 159
300 Author's index
Leont'ev, U. N. 150,153, 160 Sayer, F. P. 47, 82

Lin, S. T. 153, 157, 160 Sechler, E. E. 125, 143, 159
Loeber, J. F. 245,260,271 Setzer, D. E. 201,229
Lomacky, O. 287,298 Sih, G. C. XVII, XIX, XX, XXII, XXIII,
Lowe, P. G. 47,82 XXIV, XXV, XXVI, XXVII, XXVIII,
Lowenbrug, M. 134, 159 XXIX, XXXI, XXXIII, XXXVII,
Luk, C. H. 277,297 XXXIX, XLII, XLIII, XLV, XLVI,
XLVII, XLVIII, 5, 27, 42,43, 47, 50, 55,
MacdonaId, B. XXVII, XL VII, 259, 271 58,60,62,65,66,67,68,69,71,74,75,
Marguerre, K. ll8, 159 77,80,81,82,83,85,90, 104, 105, 106,
Mindiin, R. D. 231, 239, 240, 247, 248, ll5, 137, 159, 163, 199, 201, 202, 203,
249,258, 265, 271 218, 219, 221, 223,225, 229,231,238,
Moriguti, S. 1,42 245, 247, 258, 260,264, 269, 271, 272,
Mullinix, B. R. 80,83 274, 276, 285, 288, 289, 290, 292, 294,
Murthy, M. V. V. 162, 198 295, 297, 298
Muskhelishvili, N. I. 130, 159, 172, 179, Smith, C. W. 80, 83
199 Sneddon, I. N. 134, 159
Swedlow, J. L. 90, ll5
Naghdi, P. M. 203,208,229
Nisitani, H. 33,43 Tiffen, R. 47,82
Noble, B. 130, 159 Tong, P. 277, 297
Novozhilov, V. V. 203, 2ll, 229 Tracey, D. M. 277,282,287,297
Ved, R. XLVI, XLVII

Ogibalov, P. M. ll8, 159
Villarreal, G. XLVI, XLVIII, 80, 83,
Ohashi, M. 27, 42
104, 105, ll5, 292, 298
Pao, Y. H. 247,271 Vlasov, V. Z. 150, 151, 160
Paris, P. C. 5,42,60,83, 238, 271, 288,
Wang, N. M. XXXIII, XLVII, 62, 66,
298 83, 137, 159, 201, 229
Pian, T. H. 277,297
Wei, R. P. XX, XXII, XLVII
Westergaard, H. M. 28, 43
Rao, A. K. 162, 199 Williams, M. L. 3, 42, 60, 82, ll5, 125,
Rao, K. P. 162, 199 130, 137, 138, 139, 140, 143, 150, 151,
Ratwani, M. 162,163,181,198,199,202, 157, 159, 160,201,229
229 Wilson, W. K. 276,297
Reissner, E. XXXI,47, 48, 55, 58, 82, 85, Winkler, E. 150
ll5, 120, 124, 159,202,203, 229
Rice, J. R. 276, 297 Yokobori, T. 27,42
Rubayi, N. A. XLVI Yuceoglu, U. 163, 176, 179, 199
Ryzhik, I. M. 173, 179, 199
Zienkiewicz, O. C. 276; 278, 281, 297
Sanders, J. L. 162, 198,202,229 Zimmermann, H. 150, 160
301
Subject index
Anisotropie shells 163, 164 vibration 139, 231

Asymptotic methods 117, 130, 131
Elastie foundations l49,201
Beltrami-Haigh theory XVII Extension ofplates 55,85,247,273
Bending of plates
dassical theory l, 232 Finite element methods 47, 276
dynamic loadings 237,245,257,258, Fracture angles XXVIII, XXXVII,
270 XXXVIII, XXXIX, XL V
finite element solution 288
shear deformation theory XXIX, 62 Hartranft-Sih theory 60, 85, 231, 295
variational methods 48,88 Huber-Von Mises-Hencky theorem
Boundary conditions XVII, XVIII
approximate 52,54,95,212,234
exact 45, 50, 283 Impact loads 261
Kirchhoff type 1, 3, 60, 123, 137, Integral transform method 62, 98, 125,
163, 202,231 168, 176,212,213,235
Boundarylayers 47,68,74, LOS, 106,296
Kane-Mindlin's theory 247
Chebyshev polynomials l72, l73, l79 Kirchhoffs theory 4,232
Classieal theory of bending
see Kirchhoff' s theory Lagrange multipliers 87
Complementary energy 48, 86 Laurent series 6, 7
Complex variable method 3 Layered plates 85
Composite plates
see layered plates Mapping function 5
Crack front stresses and moments Mindlin's theory 239
see stress and moment fields Nonhomogeneous media 85
Cracks Numerical accuracy 41, 13l, 286, 287
see stress intensity factors
Critical moments XLI, XLII, XLIII, Orthotropie shells 163, 182
XLIV
Planestrain XLVI,62, 74,107,108,288,
Differential equations 294,295
see plate bending, plate extension and Plane stress 47, 55
shell equations Plate bending equations
Dynamic loads dassieal 2,3, 120, 15l, 233
impact 26l shear deformation 52, 239
302 Subject index
Plate extension equations parallel eraeks 20, 21, 32 to 39

dynamic 248,249 radial eraeks 20, 21, 26
generalized plane stress 58 shear load 168
laminates 94 three dimensional 285,286,292
modified plane stress 58, 59 twisting XXXIV, XXXVI
Poisson's ratio 188,246,247 uniform bending XXVI, XXIV, 46,
Potential energy 48 65,66,67,70
vibration 139,237, 238, 245,246,247
Reissner's theory 50,53,62 Stress intensity faetors in plate extension
dynamie 259,260
Shallow shelis 118, 202, 203 laminates 97, 109 to 112
Shear deformation 137, 202 symmetrie extension 46, 73, 74, 78,
Shell equations 79
anisotropie 164 three dimensional 285, 286
das sie al 118 Stress intensity faetors in shelis
eylindrical 121 arbitrary shell 147
shear deformation 205, 209 eonieal 148
spherieal 120 eylindrieal 145,146,174,180 to 193,
Singular integral equations 126, 172, 178 195 to 197
Spherieal shelis 122, 203, 204 spherieal 137, 195 to 197, 198,219 to
Strain energy density faetor 226
critical value XXIV, XXV, XXVI, toroidal 149
XXVIII Stress and moment fie1ds for
dilatation XIX, XXV arbitrary eraek front XXIII, XXIV
distortion XXV dassical shel1 135, 138
maximum value XXVII dynamie loading 237,245,257,270
minimum value XX, XXIV, XXVII layered plate 103
normalized valm! XL, XLI plane eraek XXXVI
plate bending XXVI, XXXVII Reissner shell XXXI
three dimensional XXIV, 294, 295 shell with shear deformation 217,
Strain energy density funetion XVII, 218
XVIII, XXXV, 49, 86, 87, 295
Stress funetions 118, 164, 209 Toroidal shelis 149
Stress intensity faetors in plate bending Twisting of plates XXIX
angle eraek XXXII
eollinear eraeks 20, 21, 25, 27, 28, Variational methods 48, 52, 53, 57, 58,
29,30,31,32 88,92,203
elastic foundation 151, 152, 153, Vibration 139,231
154, 155
impaet 264,265,269,270 Waves 234, 250, 251
nonparallei eraeks 20,21,24,25,26

M Echanics of Fracture: Plates and Shells With Eraeks

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

M Echanics of Fracture: Plates and Shells With Eraeks

Uploaded by

Copyright:

Available Formats

M echanics of fracture

Plates and shells with eraeks

M ethods of analysis and solutions of eraek problems

Three-dimensional craek problems

Plates and shells with eraeks

Plates and shells witll eraeks

NOORDHOFF INTERNATIONAL PUBLISHING

Adivision of A. W. Sijthoff International Publishing Company bv Leyden,

All rights reserved. No part of this publication may be reproduced, stored

VAN DER LOEFF / DRUKKERS BV - ENSCHEDE

5.7 Interaction of two eraeks 191

Author's Index 299

Subject Index 301

This third volume of a series on Mechanies of Fraeture deals with eraeks in

rate description of the local stress state. In particular, the replacement of

* The three-dimensional theory of elasticity yields a stress solution different from 1/ v r

energy density or S-theory provides a simple means of evaluating allowable

or kinked crack configuration would be unfeasible as the uncertainties encountered in the

increasingly large for the case of impact loading.

Lehigh University G. C. SIH

When a sl en der member is stretched gradually with consideration given

+ 2(T;y + T;z + T;x)] (1)

+ 6(T;y + T;z + T;JJ (2)

where E is the Young's modulus of elasticity and v the Poisson's ratio.

Figure 1. Stress element outside of core region

Sv = ro( ~~)v (6)

is the factor associated with volume change and

and equations (2) and (7) yield

For a state of plane strain where

0'3 = v(u 1 + 0'2) (10)

Sv = (1 - 2v)(1 + v) [(0'1/0'2) + lJ2 2 (11)

reaching some critical value. Uniike that of BeItrami-Haigh, Sih's theory

o 1.0 2.0 3.0 4.0 5.0

II. Strain energy density factor theory

Figure 3. A special spherical coordinate system

x = r cos ecos 0) (13a)

where -n/2 ::; 0) ::; n/2 and -n ::; e ::;

The values of (eo, wo) eorresponding to Smin are obtained by taking

Sv = bllki + 2b l2 k 1 k 2 + b22k~ + b33k~ (17)

16p eos w C ll = (1 + eos e) U(1 - 2v)2 +1- eos e] (20a)

16p eos w C12 = 2 sin e[eos e - tO - 2V)2] (20b)

16p eos w C22 = tO - 2V)2 (1 - eos e) + 4 - 3 sin 2 e (20e)

16p eos W C33 = 4 (20d)

(0) Tension (k,~ 0, k2 =0)

Figure 4. Plane strain specimens

rc = ( 3(1 - 2v) 2)1- (21)

where Sc represents the fraeture toughness of the material. The measured

Separating equation (27) into Sd and Sv renders

It is easily seen from

Pure shear test.The speeimen in Figure 4(b) is subjected to a state of in-

(I5) can be written as

Equation (30) may be minimized with respect to e giving

as the direction of crack initiation*, Figure 4(b). Values of eo for different

v 0.0 0.1 0.2 0.3 0.4 0.5

-oo 70.5° 74.5° 78.5° 82.3° 86.2° 90°

Crack instability starts when Smin = Se or

III. Bending and twisting of cracked plates

The accurate translation of physical phenomena into mathematical terms

Charaeter of loeal moments and shear forees. Let the thickness h of an

Figure 5. Stresses on an element ahead of crack in a thin plate

It follows that the equations of equiIibrium require both T xz and T yz to vary

Since the z-dependence of the stress components is assumed to be known,

* These eonditions eannot be eompletely satisfied by the Poisson-Kirehhoff theory of

OL- ~ ~ __- L_ _ _ _ _ _ -L~