Professional Documents
Culture Documents
VOLUME 3
VOLUME 1
VOLUME 2
VOLUME 3
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
ISBN-13:978-94-0l0-1294-2 e-ISBN-13:978-94-01O-1292-8
DOI: 10.1007/978-94-010-1292-8
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
5.1 Introduction 161
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
Chapter 6
On eraeks in shells with shear deformation
G. C. Sih and H. C. Hagendorj
6.1 Introduction 201
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
7.1 Introduction 231
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
8.1 Introduction 273
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
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
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
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
Institute of Fraeture and Solid Mechanies, Lehigh University, Bethlehem,
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
Institute of Fraeture and Solid Mechanies, Lehigh University, Bethlehem,
Pennsylvania
P. D. HILTON
Institute of Fraeture and Solid Mechanies, Lehigh University, Bethlehem,
Pennsylvania
M.IsIDA
Sehool of Engineering, Kyushu University, Fukuoka 812, Japan
G. C. SIH
Institute of Fraeture and Solid Mechanies, Lehigh University, Bethlehem,
Pennsylvania
G. C. Sih
Introductory chapter:
Strain energy density theory applied
to plate bending problems
J. Introductory remarks
dW
dV
1
= ~
+ v[ 2
(lx
2 2
+ (ly + (lz - 1
V
+ V «(lx + (ly + (lz)
2
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
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
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
(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)
* 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)
the ratio Sv/Sd obtained from equations (8) to (10) take s the form
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
14.0 V=0.4
12.0
10.0
8.0
."
(J)
.......
> 0.3
(J)
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.
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
(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
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
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 - - - - - - - -
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
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)
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)
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)
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.
TABLE I
Fracfure angles
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
(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
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)
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
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)
Kl . 8 8 38
M xy = (2r)t sm 2 eos 2eos"2
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
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
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
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
0.2
~ ~
~ 0.1
.....
~ '"
~
;:,
-+
Figure 10. Shear intensity factor versus ratio of thickness to crack Iength
(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)
k3 ()
L yz = (2r)t eos 2: + 0(1) (45e)
(46a)
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
* Note from equations (43) and (46) that kj (j = 1,2, 3) are related to Kj (j = 1, 2, 3) as
1
All = 8E (1 + eos e) [3 - v - (1 + v) eos e] (48a)
(48d)
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)
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°
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.
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
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
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
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*
* 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.
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
1.1 Introduction
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:
and
Mx M~
(1B,x = (;jiZ' (1B,y = 6h (1.3)
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:
02 W
M xy = - D(l - v) oxoy (1.5e)
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:
(1.8)
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
kl t ( lJ SlJ) k2 t ( 9 + 7v . lJ
(1x = (2r)t 2/lh - 3 eos 2" - eosT + (2r)t 2/lh - ~SID2
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
where
hl _ 41 1
f(t) = 4/(3 + v) (1.14)
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
(1.16)
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
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
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
Arbitrary array of eraeks in wide plates under classieal bending 7
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)
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
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
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
{ 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
-:~o
+211+1 {/.
{ 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
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)
. M . M
L O = - 4(1 + v)' K O = 4(1 - v) (1.30a)
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
mM
mM
( 1.34)
( 1.36)
where Wo is the displaeement for the uneracked plate under the same mo-
ments at infinity, and henee as
(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
(1.41b)
(1.41e)
(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)
Arbitrary array of eraeks in wide plates under dassieal bending 15
, .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
n, J~}
eP,
{ _
dP'~
n, J
-
{fep,~}
n, J
P'~
_
-
n, J
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)
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)
Arbitrary array of eraeks in wide plates under classieal bending 17
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
Arbitrary array of eraeks in wide plates under classieal bending 19
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
(1.50)
~
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
~
~
(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)
F~j = L
n=O
QnAn (1.53e)
s
F~j = L QnC -At
n=O
(1.53d)
where
(1. 54a)
Arbitrary array of eraeks in wide plates un der classieal bending 21
(1.54b)
XI
Xo
M
»
Figure 1.4. Periodic paralleI eraeks
where
(1.56a)
and
(1.56b)
22 M.Isida
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)
L 1 0 ~ 2n
( 1.61a)
2n = - 4(1 + v) J n - P~o "Y 2pF 2p
f3 2p
2n _
-
(2n + 2p +
2n + 2
2) S 2n+2p+2
-i(2n+2p+4)~
e (1.62e)
Arbitrary array of eraeks in wide plates under classieal bending 23
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)
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
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.
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
Arbitrary array of eraeks in wide plates under classieal bending 27
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)
{l.66a)
(1.66b)
in whieh
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
)t
where
..12 ..1 al ..1 _ a 2
m 3 = 2( (1 + ..1 2)2 _ Ai' 1 = d' 2 - d
d .1. d
. (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)
Arbitrary array of eraeks in wide plates under dassieal bending 29
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
Arbitrary array of eraeks in wide plates under classieal bending 31
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
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
0.94 4
f~ i" 56
F
- - - - - -- \' I Y- _9_1~_ 1
1\ / / I: I
\ I I 15
\ \\ 1\ \\
\ \ \ \\!,. ~ -J I II I I I 1/
'--- "--~ /- v
I
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
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
TABLE 1.4
FB of extreme outer tips of equal parallei eraeks
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
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
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
F.
----
--- _M
B
d
092
~ '-20 -
B
_M
0
0.1 02 0 .3 0.4
-Ir 0 .5
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.
Arbitrary array of eraeks in wide plates under classieal bending 37
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
08~--L---L---L---~-~
o 02 04 06 08 ZlL 1.0 ,
d
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
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
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
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
2.1 Introduction
I x I -+ oo: (J x = r xy = r xz = ° (2.1 b)
=°
h
I z I = 2": r zx = rzy = (Jz (2.lc)
I
/.~
(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
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
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].
terms of the strains, but for clarity of thought, strain is not explieitly used.
Consider the expression,
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-
** 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
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
- (1 _ 1 v2 )D SA dA [M
2
x +M 2y -2vMxM y +2(1 +v)M 2xy (2.12)
(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
f
- hl2
hl 2
V*n =
- h/2
,* dz
nz (2. 14e)
and
f3n = f3x eos eP + f3y sin eP (2.15a)
The stress components, (J~, ,~S' and '~z, are shown in Figure 2.2.
In the line integral over the portion, Su, where displaeements are speeified,
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 =
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,
Improved approximate theories of bending and extension of fiat plates 53
where
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
(2.28e)
and
(lx = Sx + Sxf"(z), 'xz = - Zxf'(z)
(Jy = Sy + Syj"(z), 'yz = - Zyj'(z) (2.29)
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
(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
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
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
Improved approximate theories of bending and extension of flat plates 57
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
(2.36 )
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)
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
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)
/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).
(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,
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.
(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
where (
m = m(o:s) = 1 +
l)t
0:2 S 2 (2.50)
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
a*
y(x, 0, z) = - h123 M OZ (2.53)
M(x) = Mo (2.54)
a eonstant.
The first of equations (2.51) gives the eondition
Let
u(x) = 1(1 - v)DfJy(x, 0) (2.56)
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,
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
where
(2.67)
Improved approximate theories of bending and extension of.flat plates 65
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)
(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
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.
- M(x) = f
hl 2
- hl2
(J:(x, 0, z)z dz (2.76)
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)
nh
(2.81)
2a
25
20
15r---__
-h !'(0)
-I
10
~ h'f'(h/2)
05
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
0.8
0 .6
-..
~
'" 0 .4
0 .2
0 .2 0.3 0 .4 0 .5
'1( ')- .5
0k--n~~~~.-~~-+.
o 0 .2 0.4 0 .6 0.8 1.0
1'( ' )/ f ' (O)
1.0
0 .4 0 .6 0 .6 1.0
f "( z)/ f"(h/2)
2.5
20
1.5
";;
~-
'-0
05
~
0
0 4 6 6 10 12
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
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 z f:
= rx 2 Im[(l - iBf3) i e-PSY]A(S) eos sx ds (2.86e)
_ ( 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)
{-
(1 _1 v2)f Im
sA(s) = - -2 2
(Jõ::'! Ja u(x) eos sx dx (2.93)
1t IX 0
In view of
where
2 1
t(IXS) = - 4(1 - v)mlX 2s2 + (J(1 _ v2}t (2.95)
1X 2S2
JOO
o st(lXs)A(s) sin sx ds = -
2
(J1X2
JX0 S(~)d~, I x I < a (2.96)
2
sA(s) = - pa2 Ja <f>(t)Jo(st)tdt
0 (2.98)
where
1 - v2 + f32
g(as) = (1 __ v2) (f32 + 1) t(as) - 1 (2.101)
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)
(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
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
h h
(1 - e) -2 -< z -< -2
~ PB sin (2pz/h)
1'(z) = (2.109)
(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
08
~o.6
N
h/a" 10
04 (E.O I)
02
- P'f (z)
10,-------------------_____
OL-________L -_ _ _ _ _ _ _ _L -_ _ _
o 05 10
~PftZ)
10
08
~06
N
o4 (E Arbltrory)
0.2
O~ ________L -_ _ _ _ _ ~~Li_ _
o 05
~O.6
0.2
OL-______~----~~~_
o 0.5 1.0
-P'f(z)
I ori---=~~~~=======
h/o =0 4
0~0------~0.~5------~10~
~ PH,)
10
08
h/o Arbltrary
06
,
~
N
04
02
0
0
r- P 1l 0 4 J a /h
2/J 08
'/;/ 1.2~
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
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.
Improved approximate theories of bending and extension of flat plates 79
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
k 1a = 4 sin 2 P k
P 2P + sin 2P le (2.117)
12
°0~~0~.5~~~10~~~15~~2~0~~2~.5~~30
h/o
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)
aJh
Figure 2.14. Stress at failure using the average stress intensity factor
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
[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
3.1 Introduction
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.
ep = J V tpdu - Js Tii·ds
I !l I
(3.1)
where the strain energy density is
Reglon ®
s (I)
T
Reglon CD
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
(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
-
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)
(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)
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
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
Through eraeks in multilayered plates 89
Plone Rl!Igion
(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
+ 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)
(3.28c)
E1 ft
n p
Ii = ~ f~2(Z)dz (3.29a)
p-i p tp_1
(3.29b)
(3.29c)
(3.29d)
(3.2ge)
(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)
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
94 R. Badaliance, G. C. Sih and E. P. Chen
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
I 3Z x +A l +A3.X=O (3.33d)
I 3Z y + A2 + A3. y = 0 (3.33e)
(3.34a)
Through eraeks in multilayered plates 95
(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
(3.37a)
Zy - a6V Zy
2
= oy0 (al V A3
2
+ a 2A3) (3.37b)
(3.38)
a3=1~-111s (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
0
,12 = - 1 3Z y - oy (,13) (3.41b)
The set of equations (3.37), (3.40) and the condition in equation (3.38)
Through eraeks in multilayered plates 97
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
° °
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)
f:
and
T;y(s, y) = TxY(x, y) sin (sx)dx, etc. (3.44b)
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)
(3.47a)
(3.47b)
(3.48b)
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
, 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)
(3.54a)
_ il 2 = [(~)2
2a3
_~Jt a3
(3.54b)
such that
A.~(s, 0) = R(s) (3.57a)
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)
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
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
a(Jo {ay.
S/x, y) = + (-----:t)
r r
CCP(1) -
r l r2
sm [1-(e l + ( 2 )]
l 2
(3.65c)
Through eraeks in multUayered plates 103
xfooo .!.J
s
l (as) sin (s x) e-sYds
el ( b l l 2
Zy(x, y) = auo4>(l) 2-
m2
+ b 2 l l + -ab2 )
6
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)
uz =
(xuT(z)~a
(2r l )t [eos (~l )] + 0(1) (3.66d)
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:
(3.70)
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
Through eraeks in multifayered plates 105
(3.73)
The funetion
fiz) = YI(Z - h l - h 2)4 + Y2(Z - h l - h 2)3 + Y3(Z - h 1 - h 2)2
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
Through eraeks in multilayered plates 107
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
Figure 3.6. Normalized stress intensity factor for El/E2 = 0.1 and h2/hl = 2.0
108 R. Badalianee, G. C. Sih and E. P. Chen
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
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
Figure 3.10. Variations of stress intensity factor with hIla for different modulus ratios and
h2/hl = 1.0 (boundary layers included)
110 R. Badaliance, G. C. Sih and E. P. Chen
10
E,/EZ~O I
05
20
0.8 100
hz/h l =2 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
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)
0.8
0.6
~
i- 0.4 h2/h l =IOO
50
20
10
0.2 05
OI
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
112 R. Badaliance, G. C. Sih and E. P. Chen
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 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).
e2 = I 3 b l - 2/ 1 a 6 (A3.2b)
dl = 21 2 a 6 - I 3 b2 (A3.3a)
(A3.4a)
~:)
1 (A3.4b)
e2
d t (e 2 - Cl
e3 =
1
dl
[ C3 - C2d";
d2 -
C{ ~: - ( ~: Y)J (A3.4c)
(A3.5b)
(A3.5c)
114 R. BadaUanee, G. C. Sih and E. P. Chen
+
bz
T {
-(Il + 12)1 3
[ el 1
~ + 2a6
( Il
"2 el - e2
)
+ (li -8 I~) el
+ 2a6
1 ( Ill z )
Il + lzf!2 + -2- el +
(li -4 I~)
+ 2:
Ill z
el
+ 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-
C = (A3.6)
Through eraeks in multilayered plates 115
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
q (x,y)
(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
(4.2b)
(4.2e)
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
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)
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)
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
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
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)
(4.16a)
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)
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
Asymptotic approximations to crack problems in shells 125
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
iA 2
F(x, y±) = -
2 RDJoo
- {P 1 exp [ - (S2 - iA 2 yl: lY IJ
e 0
· iA2RDJoo {P - P
11m - 2 1e
(s2-iA 2)1/2Iyl
- 2e
-(s2+iA 2)1/2Iyl
+ P4e -slyl}
lyl-+o e 0
then equations (4.22b) and (4.22d) are satisfied automatiealIy and equations
(4.22a) and (4.22e) beeome, respeetively
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).
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 =
(4.27a)
(4.27b)
Asymptotic approximations to crack problems in shelis 127
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 '
L*
2 = -
lfoo S4 exp [( -2 (S2 +• 2).1.
iA 2 )!lyIJ (
eos x -
J!) d
.. s s
2 0 s + IS '
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)
_ [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
(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)
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
-
(4.35d)
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.\.
Asymptotic approximations to crack problems in shelis 131
(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
(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)
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.
(4.41a)
and
p 2(S2+ i,eyt ~ Jk+l(S)
S2 ~ Bk -)k+1
( (4.41b)
k~O S
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)
and
S2 ~
J ' )
(S)k+ eos (xs)ds =
1
n'
0'
Ix I < 1 (4.43b)
(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)
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
(4.47)
then
(4.48a)
(4.48b)
in which L1 is given by
Fo, 0(,1.; i) = + i~2 I 1(e- itt / 4 Aj ~2)Kl (e- itt / 4 Aj ~2) (4.50a)
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.
Asymptotic approximations to crack problems in shells 135
(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:
k(e)
(J~e) = (2~)f eos (Om [1 + sin (Oj2) sin (30j2] (4.53b)
k(e)
(J~~ = (2:)t eos (012) sin (012) eos (3012) (4.53e)
k~b) [
(Jy(b) = (2r)+ K eos (5012) + 111 + 5v
_ v eos (012)
J (4.54b)
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)
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
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
In-plane Shear Load. If on the other hand, the residual loads, i.e., equation
(4.16), are of the form
then the solution ean be eonstrueted in a similar manner [10] and the results
are
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)
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)
Asymptotic approximations to crack problems in shells 139
with
(4.64)
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)
k (b l _ {_ ). 4Rn O 2Jc4
1 - 2(1 _ V )C
[1 +
32
7v
+
~~ (
8
1 ).)]
Y+ n 4 +
where
(4.68)
(i) for no i= 0, mo = 0:
Asymptotic approximations to crack problems in shelis 141
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
o8 ~
06
04
02
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
Asymptotic approximations to crack problems in shells 143
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
~--~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
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)
Figure 4.8
+ [ 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+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)
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
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
+ 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.
Rl = R- e tan E
R2 = R + e tan E
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)
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)
Asymptotic approximations to crack problems in shells 149
Example 3: Toroidal shell (see Figure 4.10). For an axial crack in the outer
surface
(4.80)
k (e)
1 ~
-(e)
(J V C
I{l + 645rr /\, 2
1
1 - 64
1 2
rr /\, 3 },. "AI. 3 <1 (4.81)
k (e)
I
~
_(J
-(e) I
yC
{I + 64 5rr"A22 rr"lel2}.
+ 64 ,
1
/\'1.2<
1 (4.82)
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)
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
x TOP VIEW
-I
SIDE VIEW
Figure 4.11
Asymptotic approximations to crack problems in shelis 151
(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
On the other hand, for large values of the parameter A referenee [28] gives
[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.
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
(4.89)
Asymptotie approximations to eraek problems in shelis 153
E* JH
(1 + :~) D 0 1/1 2 (z) dz
E~ JH {'/"(Z)}2 dz
(1 - v~2)D 0 'I'
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
beeome
1.0
0.8
r::::'
;:. 0.6
+
r<l
:::::
Ib"
.........
'- 0.4
(1....
0.2
CASE I., 8«r<l, lI=I/3
k=O
0.8
r-;::::"1
,
;::. 0.6
+
r<")
Ib'"
'--'
....... 0.4
0....
0.2
CASEli., r=k8<1, V=I/3
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.
(4.95)
(i(e) = (q oRI4h) (3 + cos 21X), (i(h) = 0, and iCe) = (q oRI4h) sin 21X (4.98)
(4.101)
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
(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
Asymptotic approximations to crack problems in shells 159
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
5.1 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
(5.2a)
(5.2b)
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
VI
4
=
(iPOX~ + ~
(E 2 /E I )2 OX~
(2
)
=
4
V2 (5.8)
(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
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
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)
. 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)
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
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)
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:
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)
J:
Y-+ +0 0 a 0 1 D 2
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:
(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)
(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)
(5.30b)
Crack problems in cylindrical and spherical shells 171
(5.30e)
(5.32)
f:
it is seen that
=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
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
kzz(x, t) = - a 22
t
+fro [F
0
22 (C(,O) - a22]sin C(t - x) - sin C(t]dC(. (5.38b)
(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:
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]
'"
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,
as m e .
(jxy = ha2 l(E 2 I1D 1 )2
i ~
f (-1)11-1 ( All - ----cz-e B
VJ - 2
n
)
and using the results in [24] to evaluate the integrals, we obtain the leading
term in (5.45) as follows:
174 F. Erdogan
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
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
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.S2)
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)
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.
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 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 !
(5.60e)
After some manipulations it ean be shown that the eontinuity eonditions
are satisfied if
JOO
o
•
_11_
A
IX
(mIR I - mlR l ) eos IXX dlX = 0, (lxi> 1). (5.61b)
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
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)
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
(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.
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).
Crack problems in cylindrical and spherical shelis 179
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]
(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)
where (r, 8) are the polar coordinates at the crack tip defined by
Defining the membrane component of the stress intensity factor in the shell
by
km =
s
i(E 2 hD 1 a)t [eo H (1) - H
ha2 e2 1 2'
(1)J (5.75)
(5.76)
where
(5.77)
the funetions Hj are obtained from equations (5.65) after the normalization
given by equation (5.77).
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
b _ ~ ~ a Z Jz 3 G 12 Vc .58
(Jx/r, 8, Z) - Ab (2ra)t h 12D 1 [
? sm 2
where
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
where X, Y, Z are the aetual reetangular eoordinates and the aetual geometrie
angle 0 in the shell is given by
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
Crack problems in cylindrical and spherical shelis 183
TABLE 5.1
Elastic constants of the orthotropic materials
Titanium Graphite
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),
Figure 5.2. Membrane eomponent of the stress intensity faetor ratio Am for a pressurized
Titanium and for an isotropie (v = 1/3) eylinder
(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
Crack problems in cylindrical and spherical shelis 185
-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
TABLE 5.2
The efj'eet %rthotropy on the stress intensity ratios (al y/(Rh) = 1.66)
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
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)
Crack problems in cylindrical and spherical shelis 191
TABLE 5.3
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
-~-_ .. ~~---
- - - - ._-------- -----
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)
Crack problems in cylindrical and spherical shelis 193
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.
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
* 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
of Fraeture Mechanies, 1, pp. 104-113 (1965).
[4] Folias, E. S., A finite line crack in a pressurized spherical sheli, International Journal
of Fraeture Mechanies, 1, pp. 20-46 (1965).
[5] Folias, E. S., A circumferential crack in a pressurized cylinder, International Journal
of Fraeture Mechanies, 3, pp. 1-12 (1967).
[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).
Crack problems in cylindrical and spherical shelis 199
[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].
* 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
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.
ix, r
[17] :
SECTION A\ - A\
(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
On eraeks in shelis with shear deformation 205
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
oN xy oN"y Qy
ox + ay + R
=0 (6.5b)
oQx
+
oQy N xx + N yy = 0 (6.5c)
ox ay R
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)
206 G. C. Sih and H. C. Hagendorf
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
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):
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)
(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)
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)
208 G. C. Sih and H. C. Hagendorf
6(1 - v)
{ -a{ly
ay + vax-
a{lx} (6.12b)
(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)
ux = Rl (~epx - l/J),
U u y = 1R oep
oy (6.15)
and
(6.16)
1
V4 l/J - V 2 l/J 0 (6.17b)
k2
* 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
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)
h2 h2R 2 1 + V eo
12/ 1 ,Jc4 = 12(1 _ v2)1 2 ' e = ~ T; (6.19)
(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)
On eraeks in shells with shear deformation 211
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)
N xy = N = 2/1h { -
02cjJ
- - 1- ot/!}
- (6.24e)
yx R oxoy 2 oy
Similar expressions are obtained for the moments
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)
Qx = ph { oep} oep
OX - 'I' , Qy = /1h oy (6.26)
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.
212 G. C. Sih and H. C. Hagendorf
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 '
(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)
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
214 G. C. Sih and H. C. Hagendorf
+ 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)
- 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*
e + (kS)2
+{ 2(1 _ elex p(-sly/)-[1+(ks)2]sexp(-Ply/)
1
+ 2(1 _ e) [2 - 3e + (1 - 2e) (kS)2]SPF
[{-
(6.37a)
+ 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
F(lyl;s)=-1 [ ::-exp(-eeIY/)+
1
2 ee
1
-exp(-eeIY/)
ee
J (6.39a)
f~ ~ Al(s) eos(xs)ds = o}
°
(6.41a)
J
(J Ixl>a
OO - Ais)eos(xs)ds =
o s
and
Al(s) = 2Jäs
Ehi
Ja r;
-V t 4J1 (t)J oCst)dt (6.42a)
f: Jt
5 0
where JO is the zero order Bessel function of the first kind and
1 + y
15 = 1 + -1--8
-y
(6.43)
(6.44b)
it is found that the funetions cPj(~) (j = 1,2) are solutions to the following
system of eoupled FredhoIm integraI equations
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
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)
218 G. C. Sih and H. C. Hagendorf
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)
(lxx == ~~;~1 eos (Oj2) [1 - sin (0/ 2 ) sin (30j2)] + 0(1) (6.50a)
k2J )Z! eos (8j2) sin (8j2) eos (38j2) + 0(1) (6.50c)
( r '
The stress intensity factor kl (z) in equations (6.50) depends on the thickness
coordinate z and is defined as
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)
4
EXTENSION OF TOP L AYER 2 · hl2l., O.3)
005
0'0
I!:o
,z I
,
N
l 0.15
.5
o 02 04 06 oe '0 ,2
SHELL THICKNESS/CRACK LE..:lTH. hl20
40 r05
,
POl5
-
I ~ 30
~' 020
; -2
( :::::. 025
: ~ 0 .30
~ i'"
--:.-
o 2.5 3.0
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.
222 G. C. Sih and H. C. Hagendorf
----<: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
0 .9 0_01
o.o'!' 0.10
hIA-o.l:i
0.8
l!:
t: 0.7
~ 0.6
~
>- 0 4 MY
"~ , ,
,..
t- 0 .3
M.
õi
Z
~ 0. 2
~
'"~ OI
'"
t-
<Il
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
10
09
TOP L~AYER
• , ,
M. •
LAYER
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
On eraeks in shells with shear deformation 225
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].
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
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
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.
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)
l+v . 3+v
a 21 (s) = - -l-- e - I:Fo(AS) + -1-- (l - 2e)k 2 s2 {I - FoUs)}
-v -v
1
0 22
+ k s2) + k s(1 + kl s2)1
(s) = -2 e (1
2
eo - e
leading to
and the parameters k, A, e, etc., can be found in equations (6.19) and (6.20).
(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)
On eraeks in shells with shear deformation 229
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
7.1 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.
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
(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:
(7.5)
[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)
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)
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).
w(s) (x,y,t) = 21
n
Joo-oo {Al(s) exp [- Y(S2 + cx rt ] 2
(7.15)
+ ~~ '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:
f:
whose kernels LI and L 2 are given by the expressions
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
Dynamic analysis of cracked plates in bending and extension 237
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)
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)
(7.26)
in whieh Kl1) represents the amplitude of the moment intensity faetor and is
given by
(7.27)
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·
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
Dynamic analysis of cracked plates in bending and extension 239
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.
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.
My = D( °t; + °t: )
v exp( -iwt) (7.30b)
(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
_ n2/1h (I/! y
+ Ol/!z)
oy = _ p/z3 w 21/! y (7.32b)
I/! x
= (a -
2
1) Oepl + (a _
OX I
1) Oep2 + Oep3
OX 0Y
(7.33a)
2 (CX.)2 J'=12
a·=--.::.1.. (7.34)
J 1 - V CX3' ,
The wave numbers CXj (j = 1,2,3) stand for
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)
242 G. C. Sih and E. P. Chen
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(i)(X
xy \ , ° °
, t) = (7.39c)
(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.
Dynamic analysis of cracked plates in bending and extension 243
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 .
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)
244 G. C. Sih and E. P. Chen
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:
(7.50b)
These equations ean be solved numerieally for the intensity of the moment
distribution around the eraek tip.
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)
Dynamic analysis of cracked plates in bending and extension 245
(7.53b)
(7.53e)
°
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
(7.56)
There is a phase angle between Kl and the input wave whieh ean be computed
from
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 ~
Dynamic analysis of cracked plates in bending and extension 247
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.
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
248 G. C. Sih and E. P. Chen
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.
y
Middle
plane
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
Dynamic analysis of cracked plates in bending and extension 249
Dx = (7.60a)
oep _ 01/1
Dy = oy ox (7.60b)
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)
2
jj j = 2fJ n )2
1( h [ (0: + fJ) ( W)2
õ5 - . J
1 - ( - 1YH , j = 1, 2 (7.63a)
The frequency
_ nC t
w= - (7.65)
Il
f
of length defined as
hl 2
(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
_ JlIlZ ~ . ocjJj
Ry - 6 L... eJ ~ (7.70b)
j=1 dy
[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)
Dynamic analysis of cracked plates in bending and extension 251
(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)
. 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)
(7.76d)
. 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
Dynamic analysis of cracked plates in bending and extension 253
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)
(7.86)
The satisfaetion ofequations (7.8Ia) and (7.8Ib) requires that A(s) be deter-
mined from the dual integral equations
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:
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)
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)
(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
Ixl<a (7.96b)
and g2(S) is given by
(7.99)
where 0/ 3 and 0/ 4 can be obtained by solving the Fredholm integral equa-
tions
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)
i(2s 2 - 6;) [ ~
C 3(s) = - 2SfJ3 (;l(S) + Cis)] (7.103e)
J:
The remaining conditions in equations (7.83) yield
J-roOO
g3(s)CCs)exp(-isx)ds =
2niR..
_ _J exp(-lxc)j eos Yj), lxi <a
~
(7.104b)
Dynamic analysis of cracked plates in bending and extension 257
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
J:
with L s and L 6 being the kernels given by
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
258 G. C. Sih and E. P. Chen
+ --,
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
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)
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)
~
~- 0.8
0.4
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
Figure 7.9. Thickness mode stress intensity faetor versus wave number for input P-wave
at norma! incidenee
260 G. C. Sih and E. P. Chen
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°
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
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)
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(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)
(7.121)
Dynamic analysis of cracked plates in bending and extension 263
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)
(7.128)
1.0
a/h= 1.0
0.8
~
0.6
Q.
-
~
*-
090 0.4
0.2
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 + ...
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)
1.2
a/h = 1.0
0.8
~ .
::E
0.6
-
~
-,.:-
0.4
0.2
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.
(7. 132a)
(7.132b)
(7.132c)
(7.133a)
(7.133b)
Dynamic analysis of cracked plates in bending and extension 267
(7. 133e)
t/I~(x,y,p) = -2 Joo [sel - ai) Bi(s,p) exp( -fJiy) + sel - ai) Bi(s,p)
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)
(7. 137b)
(7. 137c)
B* ()
3 S, P =
sB*(s,
2
p)
2
(1 -
) (* - a *)
val 2 (7. 140e)
al - a2
'J
.::.
n
Joo fi(s, p) B*(s, p) eos(sx)ds =
0
MO
p
D , °< x<a (7.141b)
!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
A solution that satisfies equations (7.141) can be written in the form [16]
such that epi can be computed from a standard Fredholm integral equation
of the second kind:
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
Figure 7.13. Solution of Fredholm integral equation in the Laplace trall5form domain for
a Mindiin plate
270 G. C. Sih and E. P. Chen
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
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
1.0
1I=0.25
0.8
~ 0.6
~.
-...
~ 0.4
~-
0.2
Figure 7.14. Normalized dynamic moment intensity factor versus time for a Mindiin
plate under impact
Dynamic analysis of cracked plates in bending and extension 271
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).
272 G. C. Sih and E. P. Chen
8.1 Introduction
Plane Normal to
Crack Front at A
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)
-r xy =
kj(z)
- - sm -
(2r}t
. 8
2
8
eos - eos -
2 2
38
+ 0(1) (8.ld)
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)
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)
Uz = O(r) as r -+ 0 (8.3e)
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.
----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
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
A specializedjinite element approachjor three-dimensional crack problems 279
(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
(8.12)
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
(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
A specializedjinite element approachjor three-dimensional crack problems 281
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,
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
Jl Jl IJl
The element stiffness matrix then becomes
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.,
A specializedjinite element approachjor three-dimensional crack problems 283
3
kj(s) = I Nj(s)k j
1
where
Nt = ses - 1)/2
N";.=(L-s 2 )
N~ = ses + 1)/2
/
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.
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
A specializedjinite element approachJor three-dimensional crack problems 285
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
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
0.3
.1.. ~=2
h
0.2
t=1 ta =0.15
1I=0.3
0.1
Midplane af Plate
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-
A specializedjinite element approachJor three-dimensional crack problems 287
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
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
A specializedjinite element approachjor three-dimensional crack problems 289
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
°
-0 . 1 0 ,1 0 .2 0.3 0.4 0 ,5 0 .6 0 .7 0 .8
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.
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
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
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.
A specializedjinite element approachJor three-dimensional crack problems 293
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
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
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
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
A specializedjinite element approachJor three-dimensional crack problems 297
8.8 Summary
References
[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
Subject index