Engineering Fracture Mechanics. 197 I, Vol. 3. pp. 7 l-96. Pergmon Press.

Printed in Great Britain

STRESS INTENSITY FACTOR FOR AN ELLIPTICAL

CRACK UNDER ARBITRARY NORMAL LOADINGt
R. C. SHAH and A. S. KOBAYASHI
Space Division, The Boeing Company, Seattle, Wash. 98 124, U.S.A. and Department of Mechanical
Engineering, University of Washington, Seattle, Wash. 98105, U.S.A., and also
Consultant, The Boeing Company, Seattle, Wash. 98124, U.S.A.

Abstract-Harmonic stress function and the stress intensity factor for an elliptical crack embedded in an
elastic solid and subjected to an arbitrary internal pressure are derived in this paper. The internal pressure
is assumed to be expressible in terms of a polynomial ofx and y as

ozz = _P(%Y)
where

The stress intensity factor for the elliptical crack subjected to this polynomial loading reduces to the solutions
given by Green and Sneddon when all coefficients A, except Aw are zero and by Kassir and Sih when all
coefficients except Alo are zero. -
The solution is then used to determine the transient stress intensity factor of an elliptical crack embedded
in a thick plate, one side of which is subjected to a sudden temperature change. Other possible applications
of this solution, such as an elliptical crack in a large beam in pure bending, are also discussed in this paper.

NOTATION
3
suitable harmonic stress function = ,20
2 & %; i+j s 3

ellipsoidal coordinates
semi-major axis of ellipse
semi-minor axis of ellipse
I_----_I yz 2”
az+s b*+s s
s(u2 + s)(b’ + s)
m tin(s) ds
I *iz?z
Poisson’s ratio
shear modulus
constants
incomplete elliptic integral of the first kind
complete elliptic integral of the first kind
incomplete elliptic integral of the second kind
complete elliptic integral of the second kind
mu,, cnul, dnu,, dcu,, cdu,, ndu,, ncu,, sdu, = Jacobian elliptic functions
k, k’ modulus and complimentary modulus of Jacobian elliptic functions respectively
0 angle in the parametric equations of ellipse
K, opening mode stress intensity factor
I moment of inertia of the beam
MO applied bending moment
f time in seconds
T temperature at any point
(I coefficient of linear expansion
E modulus of elasticity
K thermal ditfusivity.

tpresented at the Second National Symposium on Fracture Mechanics, Lehigh University, Bethlehem,
Pa., 17-19 June, 1968.
71
12 R. C. SHAH and A. S. KOBAYASHI

INTRODUCTION
A PROBLEM frequently encountered in applied fracture mechanics is the estimation of
the stress intensity factor of an elliptical crack embedded in a thick plate and subjected
to a complex stress distribution. When the stress distribution in the thick untracked
plate in the area to be occupied by the elliptical crack is a simple one such as uniform
uniaxial tension, linearly varying tension with respect to x, or simple constant shear,
then the known solutions of Green and Sneddon[l], Kassir and Sih[2], Kassir and
Sih [3], respectively can be used to determine the stress intensity factors in these prob-
lems. Also, when the surfaces of a flat elliptical crack are exposed to uniform tempera-
ture or temperature gradient, solution of Kassir and Sih [4] can be used to evaluate the
stress intensity factor. Available also are the potential functions proposed by Segedin [5]
which can be used to solve the three-dimensional elastostatic problems satisfying mixed
boundary conditions on a given plane. Using these potential functions, Segedin[S]
treated the problems of an elliptical crack in a medium in pure bending, and of a rigid
elliptical inclusion in a medium in torsion or bending. Segedin[6] also considered the
problem of thermoelastic stresses caused by uniform flow of heat perpendicular to the
plane of elliptical inclusion which acts as a conductor or an insulator. He then con-
sidered the case when the heat flow is in the plane of inclusion. When the stress dis-
tribution in the untracked thick plate is nonlinear or complex, then there exists no
solution appropriate for determining the stress intensity factor in such a plate which
is cracked internally by an elliptical crack.
In this paper, an exact solution is given for an elliptical crack embedded in an
elastic solid and subjected to complex pressure distribution expressible in terms of
a polynomial of x and y coordinates. The foundation of this paper is based on the
potential function proposed by Segedin[S]. The associated stress intensity factor for
the elliptical crack is also derived. This solution, then, can be applied to the problem
of a thick plate containing an elliptical crack if the plate thickness is sufficiently large
in comparison to the crack width (plate thickness 3 3 X crack width). This insures that
the two bounding free surfaces of the plate will have practically no effect on the stress
distribution in the vicinity of the embedded crack.
The above solution is then used to obtain the stress intensity factor of an elliptical
crack in a large beam in pure bending. Also, it is utilized to determine the transient
stress intensity factor of an elliptical crack embedded in a thick plate, one side of which
is subjected to a sudden temperature change. For this purpose, the thermal stresses
in the untracked thick plate are determined first. The thermal stresses on the crack
surface are then relieved by the solution derived in this paper and the stress intensity
factor is determined.

FUNDAMENTAL EQUATIONS
For the sake of completeness, equations of elasticity in three dimensions that can
be found in other references[3,7] are described here. In the absence of body forces
and shear stresses, 7,, and 7yZon z = 0 plane, Navier’s equations of elastic equilibrium
are satisfied by a harmonic function $ defined as follows:

V”4 = 0 (1)

where 4 is a suitable harmonic function and V* is Laplacian operator in three dimen-

sions. Cartesian components u, v and w of the displacement vector are given by
 Elliptical crack under arbitrary normal loading 73

u=z$+(l-2?)2 W

a"#
--+(l-2$)$
D=Zayaz Gb)

w=zaZch_*(*_s)2 ml
aZ2 az
where 7 is the Poisson’s ratio. The Cartesian stress components are given by

Oa)

(3b)

(3c)

(3d)

?2/2= 2Gz* (3e)

aya?

7xz= 2Gz- a34 (3f)

aXaz2

where G is the shear modulus.

ELLIPTICAL CRACK
Consider a flat elliptical crack embedded in an elastic solid as shown in Fig. 1. The
boundary of the elliptical crack is described by

and z= 0. (4)

The crack is opened up by applying internal pressures, p(x,y), symmetrically to the

upper and lower surfaces of the elliptical crack. In view of symmet~, this problem is
equivalent to the problem of a semi-infinite solid of z Z+0 subjected to the following
boundary conditions:

a;, = -P(x*Y)
($+$ I,z=O
>
(5a)

IV=0
(f+$, l,z=O
>
(-W

‘r&.z = Tyz = 0 (2 = 0) ml
74 R. C. SHAH and A. S. KOBAYASHI

Fig. 1. An elliptical crack in an infinite solid.

together with a condition that the stresses vanish at a sufficiently large distance away
from the crack.
Using (3e) and (3f), the condition rzz = ryz = 0 on z = 0 plane is identically satisfied
and (5a) and (5b) take the following form:

aV_p(x,y) $+$< l,z=O (W

z- 2G >

a+-
dz-
0
X"
(-+L
a2 b’
l,z=O
>
(6b)

p(x,y), here, is taken to be expressible in terms of a polynomial of x and y as

p(x,y) = 5 5 Ai&+
i=O j=(l

It is assumed that the pressure distribution can be reasonably approximated by the

polynomial of (7) so that the problems of generalized pressure loading can be solved
by the analysis of this paper.

POTENTIAL FUNCTION
To solve the boundary value problem represented by (1) (6a) and (6b) for the
elliptical crack, an ellipsoidal coordinate system (h,p,v) is introduced where A, p, and v
are the roots of the equation in s,

X’
L+zz_ l= 0.
(8)
d+s+b’+s s

The Cartesian coordinate (x,y,z) of any point can be expressed in terms of (h+,u) by
 Elliptical crack under arbitrary normal loading 75

the relation:
L2*(uz-P)X*= (a~+A)(a*+~)(a”+v) (94
F(P-u2)y2= (b2+h)(b2+/L)(b2+V) C’b)
u2bZz2= A p v (9c)

where

In the plane z = 0, inside region of the elliptical crack, namely ((~‘/a’) + (y2/b2)) < 1,
isrepresented by A = 0 and the region outside the elliptical crack, i.e. ((x2/u’) + (y2/b2))
> l,isgivenbyp=O.
Segedin [5] noted that the potential function Vn) given by (lOa) is harmonic.

where

u(s) = l------
2 y2 z2
(lob)
u2+s b2+s s

Q(s) = s(u2+s)(b2+s) UOC)

and n is a positive integer. Since YCti is harmonic, YCn)and the partial derivatives of
Vti with respect to x and y coordinates satisfy the equation of equilibrium (1). Thus,
the general stress function, 4, for (1) through (3) can be represented as a polynomial
of Vn) and derivatives of Vn).

(11)

This harmonic stress function has two special properties which are useful in satisfy-
ing boundary conditions (6). The partial derivative of (11) with respect to z can be
written, on using (8), as

(124

This can also be written as

(12b)

It is obvious from (12) that on the plane z = 0 and outside of the elliptical crack,
i.e. p = 0, the normal displacement w vanishes and hence it satisfies the boundary
condition (6b). In the subsequent sections, it is shown that this harmonic function
76 R. C. SHAH and A. S. KOBAYASHI

satisfies the boundary condition (6a), namely,

(13)

where Au is a linear function of cij.

SOLUTION TO THE PROBLEM

The harmonic stress function #, as represented by (1 I), involves a double series
summed in terms of i and j. In this paper, the double summation on the series of (11)
is limited so that i+j < 3. Equation (1 l), then, becomes

(144

= 9 5 Gij iijS3 (14b)

i=O j=O

In (14), i+ j is taken less than or equal to three for two reasons.

(1) Practical considerations, such as exorbitant work (as evidenced by the Appen-
dix of this paper) involved in deriving the derivatives of the potential function Yti+j+lt
needed for ( 11) and (3) limits the number of coefficients Ati in (13). The limited number
of terms in 4 thus limits the number of terms in the pressure distribution as

P(GY) = 9 $ZAi#x’y’
i=O j=()
i+jC 3 (14c)

(2) Generally the number of unknowns Aij in (13) is limited to seven for single
precision arithmetic and eleven for the double precision arithmetic[8] so that the
pressure distribution can be fitted well (without oscillation between fitted points) by
the least square surface with a polynomial of x and y in (13).
The harmonic stress functions @ij are listed in the following equation where Cij
differs by an integral constant from cu.

@‘oo= cool/“’ = c 00
 Elliptical crack under arbitrary normal loading 77

%*=c02~=Cl12[~ (b2+OP;;gS)]1,2-4y2~ @2+;;:[)ed; )],,2]

s s s

@xl = c30~= Gl9[3 [ (a2;;)?;;; )]‘,‘-4x* [ (,2+$r)ed; s )]l,z]

(1%)
S

%I = Gl$Jg = Gy[ [ (a2+ ) ;z(J)&( )]I/”

S S

1
w(S) ds
-42 m WO
I* (a2+s)2(V+s) [Q(s)]1’2

@I2= c12a3vo=
axay2 c12x[I*mca2+S ,$~$Q( S
)I',"

- 4y2
I *a (a’+s)(~(S~~~~e(s)]l/2 1 (15)

a03= co3s= Co3Y[3 1 (1%)

Since the immediate objective of this paper is to develop an expression for the
stress intensity factor of an elliptical crack subjected to the pressure distribution in (13),
only a2@&z2 are listed in Appendix I. All of the expressions of (A2) for a2@,/az2
contain elliptic integrals, some of which are available in [9]. Other expressions were
derived and all necessary relations are listed in (A4) in Appendix II.
Each of the equations (A2a) to (A2j) contains the expression with aA/az as a factor.
The expressions containing ah/az in (A2) become singular at A = 0. It can also be seen
from (A4) that each of the elliptic integral equations (A2a) to (A2j) contains terms
which are singular at A = 0. When these singular terms at A = 0 are combined for each
equation, it is found that a2$Jaz2 is always finite at A = 0. d2Qu/az2, when evaluated
on the crack surface A = 0, yields a polynomial of x and y as shown in (16).

* - Go
az2 h=O ab2 E(k) (164

3 =z [kr2K(k) - (l-2k2)E(k)]
a-2 h=O
(16b)

[(1+k2)E(k)-k’2K(k)] (16~)

p&p [k'2K(k)
- (l-2k')E(k)]
+&[(3-8k2++).

* E(k) -2k’Z(2++)K(k)] + -$& [kf2(2-k’)K(k) -2(kf2+P)E(k)]}

(164
78 R. C. SHAH and A. S. KOBAYASHI

a% - * [k’2(2-kZ)K(k) -2(k’2+IP)E(k)3 (16e)

a22 k=o=
(l+kZ)E(k)-k’2K(k)+ x2
%I,-, = 8Co2[ a?lPk’4 [k’Z(2 - k2)K(k) - 2(k’2+ k4)E(k)]
a5k2kr4

2(3k’- l)K(k) + 3kZ+T

>
E(k)
11 (10
a =s[ [2kf2(2k2+ l)K(k) - (2+3k’-WE(k)1
a2 h=O

+$& [(8+9k2+ 16k4-48k6)E(k) - (8+ 13k2+241P)k”K(k)]

+& [(8k4k’2+3k2k’2+2k2-8)E(k)+ (4k2k’2+5k’2+3)k’2X(k)]} (16g)

j&l =~{[2(k’Z+P)E(k)-k’z(2-k2)K(k)]+-&
a2 A=0
. [(8k4k’“+3k2k’2+2k2-8)E(k)+k’2(4k2k’2+5k’2+3)K(k)]

+&(6/P-3k’+ 16--k+$o) + (3k2kf2+ 12P-8)K(k)]] (16h)

p& =s[ [2(k’“+P)E(k) - (2-kk’)k’2K(k)] +& [(4k2kf2

az2 *=o

(3k2k’2+ 12k’2-8)K(k)+ 6k4-3k2+ 16+)E(k)]} (16i)

pJ&3
a22 )r=o
2(3k2- l)K(k) + 3k’+ = k,’
>
E(k)
1 +&

. (3k2k’2+ 12k2-8)K(k) + 6k4--3k”+ 16-$)E(k)]+&

[
. [(16k’-45k’k’2+8)k’2K(k) - (48-40k”-88k2kf2+ lSPk”)E(k)]}

(163

K(k), E(k), k and k’ are defined in (A3) in Appendix II.

Equation (12a) shows that the selected harmonic function satisfies the boundary
condition (6b). The function must also satisfy the boundary condition specified in (13)
where p(x,y) is given by (14~). Substituting (14) and (16) into (13), one obtains a set of
linear equations for undetermined constants Cii in terms of k, k’, K(k), E(k) and Aij.
This relation in matrix form is listed in the following equation.
Unknown constants Cti are determined by solving the matrix equation (17). Other
 &[kt2Kik'
C 0 0 0 -00 w
- il-2k5E(kI]
- -

0 0 0 :I( '10

- -

0 0 0 0 -01 '01

- -

’ r(34k2
XWL
--&$C2&.
0 .-i;?r‘E(k'm2k" 0 0 0 0 :2( '2c
K(k\-2ik'2+k4,Eik;
J O
. i2--bK(k1
k2 3 -
-

&4[?2(2- .k2h
0 c 0 0 '11
-11
I(kj
(kb2;kq2+k4\,
- -

0 0 :02 '02
’ (17)
- -

&I&4k2kt2+5k

0 0 0 0 GO '30
-48kqEW-(8+13k2 3\k'2K(kWL3k4k'2-:

-24k41kT2Kfk) k2kp2+2k2-8bE(k)
1 1 - -

0 0 0 ' k'2+2k2-Et\E(k1+k'2 0 321 '21

l (4k2k'2+5k'2+3)Kik'
- -

&b (8k4k'2+3k?, & [(3k2kt2+12

[
0 0 k~2+2k2-8~Efk~+(4k2. 0 kt2+?JK(k\+(dc43. 5: 5:

k2+16- *
&E(k)1
- -

0 0 0 0 0 +16-+E(kA+ 0 -02 b3

(3k2kn2+12k2-e)K(k)
I I
80 R. C. SHAH and A. S. KOBAYASHI

derivatives of Qti necessary to evaluate the remaining stress components have been
evaluated but are not listed here to keep this paper within a reasonable size. These
derivatives, on substituting in (3) show that the stresses vanish as A + 03. Hence the
solution is now complete and the displacements and stresses at any point in the medium
can be found by using (2), (3), (14b) and (17).

STRESS INTENSITY FACTOR

Kassir and Sih[3] have shown that in the vicinity of the periphery of an elliptical
crack on the plane z = 0, the ellipsoidal coordinates become

h = 2abr(a’sin* 0-t b2 cos2 O)-r/z (18a)

p=o (I=)

z~=-((a~sin~8+b~cos~~) (18~)

where r is the radial distance normal to the crack border in the plane z = 0, and 8 is
the angle in the parametric equations of ellipse

x = a cos tl and y = b sin 8. (18d)

Stress intensity factor KI is defined in terms of the normal stress, oZZ, on the plane
z=Oas

K, = 1;r [ (27rr) %*J . U9a)

Utilizing (18), it can be written as

KI= 2 “‘(a2 sin2~+b2cos2e)~~4~+li [(A)I’~~,J~+ t19b)

( >

The stress intensity factor derived for each harmonic stress function ati with the use
of (A2), (A4) and (19) is listed in (20) with the corresponding @,:

(a2 sin2 0 + b2 cos2 0)lj4 (204

alo: KI = 8c10~~os ’ ‘12(a2sin2 Bf b2 cos2 0)1/4 GObI

sol: KI = 8CO~~2sin
’ 1’2(& Sin.2 e+ 62 ~0~2 ,q) i/4
(2Oc)

azo: KI =-32c2~~cos'e "2ta2 sin2 e+ b2 cos2 ,j)1/4

(204

Q, _ K = 8C,,Gcosesine II2
11’ I ( a2 sin2 e + b* ~0~2 0)1/4 t20d
a2b2
 Elliptical crack under arbitrary normal loading 81

@02: KI = _ 32crbT SinZ ’ 2 I’* (a2 sin2 0 + b2 ~0~2 0) l/4

(200
( >

~$30: K, = - 32c30~bcos3 ’ 2 I” (a2 sin2 0 + bZ ~0~2 0) l/4

( >
cm)

@21: KI = - ’ sin
32Cz1Ga~b~”’ “* (a2 sin2 o.+ b2 cos2 0)1/d
W’h)

cp . K = _32C,,G cos 8 sin2 8 1~

12. I (a* sin* 0 + b2 cos2 6) I/4 (20i)
a2b3

112
@03: K, = - (a” sit? 0 + b2 co? 0) 1/4. VW

Some details of the derivation of the stress intensity factor are given in Appendix
III. The stress intensity factor for the total function 4 for the loading of (14~) can be
simply written by superposition of (20a)-(20j) as

4: & = s (?_)I” ca2 sin2 0 + b2 cos2 0)114( Coo + ‘10 zs o + cOl iin e _ 4c20;s2 e

+ C,, cos 8 sin 8 _ 4C02 sin2 8 - 4C,, cos3 8 - 4C,, cos2 8 sin 8
ab b’ a3 a2b

4C,, cos 8 sin2 0 - 4C,, sin3 0 .

- ab2 > (21)
b3

Coefficients Cij are already determined by the matrix equation (17) in terms ofA,.

SPECIAL CASES
(a) Constant appliedpressure, (T,, = - po, on the elliptical crack surface A = 0:
Green and Sneddon[ 11 have solved this problem. From (6a) and (16a), Coo becomes

poab2
‘O” = 8G aE(k) ’

The stress intensity factor obtained by substituting the above value of Coo in (20a)
coincides with that obtained by Irwin [lo].
(b) Linearly varying pressure, crzz= pox, on the elliptical crack surface h = 0.
From (5a), (6a) and (16b), C,, reduces to
‘ ‘

‘lo= SG[(l -2/$$;-k’YK(k)] ’

When the above value of Cl0 is substituted in (20b), the stress intensity factor Kl-f
agrees with the results of Kassir and Sih [2]$.
t K, is defined as KI = l&k,.
WZorrected for error of Kassir and Sih [2].

EFM Vol. 3 No. I- F

82 R. C. SHAH and A. S. KOBAYASHI

(c) Tunnel crack with variable pressure CT,, = -u/y, 0).

For a tunnel crack of width 2b located at-b s y 6 b and subjected to pressure uzz =
--cr(y,O), the stress intensity factor is given by Sih, Paris and Erdogan [ 111 as

(22)

For u,, = -u( y,O) = -&y, the above intensity factor becomes

K = Ao,b(mb)1’2
I
2 .
(23)

For the case of an elliptical crack with linearly varying pressure uzz = -&y, Co1
becomes, with the use of (5a), (6a) and (16c),

(24)
co’ = 8G[ (1+ k2)E(k) - k’2K(k)]

When the elliptical crack approaches the tunnel crack, i.e. when a + a, then from (A3b)

k2 + l,kr2-+ 0 (25a)
and
E(k) -+ 1, k’2K(k) --, 0. (25b)

Thus, by using (~OC), (24) and (29, the stress intensity factor for the special case of
tunnel crack reduces to the same form as given by (23).
For the loading ur+ = -poy2, the stress intensity factor for the tunnel crack, from
(22), becomes

K = pob2(&,)1/Z (26)
I
2 .

The stress intensity factor for the elliptical crack for loading of uzz = -pay” can be
obtained by solving simultaneously for Coo, C20 and Co2 from the three equations
reduced from (16a), (16d) and (16f). Limits as shown in (25) must be taken to obtain
the results corresponding to that of the tunnel crack. To avoid the painstaking process
and round-off errors in solving matrix equation (17) for unknown constants Cij, the
following values are chosen to simulate the tunnel crack.

bla = 0.05234 and a = 1.0.

The stress intensity factor thus determined numerically for this oblong elliptical
crack is K, = 0.5558~~ X lo-” which is approximately equal to KI = 0.5554~~ X lo-”
determined by (26) for the tunnel crack. The difference between the two solutions can
be attributed to the differences between an oblong elliptical crack and a true crack.
For the loading of uzz = -pay 3, the stress intensity factor for the tunnel crack
becomes
KI = #p,b”(rb)““. (27)
 Elliptical crack under arbitrary normal loading 83

Again, the same oblong elliptical crack is used to determine numerically the stress
intensity factor for this loading condition which yields K, = 0.219Op, X lo-*. This
numerically determined stress intensity factor compares favorably with KI = 0.2180~~ X
10P4determined directly for a tunnel crack from (27).
(d) Penny-shaped crack with sinusoidally varying pressure.
The stress intensity factor for the elliptical crack is compared here with that for
the penny-shaped crack subjected to sinusoidally varying pressure derived by Smith,
Kobayashi and Emery [ 121,
For the loading of cz, = -Aolr sin @= --AoIy on the crack surface, the stress inten-
sity factor for the penny-shaped crack becomes [ 121

K = 4AoI sin 13
I 3 (7r)“2
(2W

where it is assumed that a = 1. When the elliptical crack approaches the penny-shaped
crack (t, = a), then the stress intensity factor, reduced from (20~) and (24), becomes

K = 4& sin 8 a3 I”
I (28b)
3 0?T’

This is identical to (28a) when a = 1.

From [ 121, the stress intensity factors for the loadings prescribed on the crack sur-
faces, as indicated in the parentheses, are given by the following equations.

(29a)

K = 16 0cos28
I 1;($+,2 (r.,, = -Pox2+PoYZ) Wb)

K = 32p. cos 38
I
35(n)“2 (PZZ= --P& + 3PoxY2) (29~)

K = 32p. sin 30
I 35(,)1/Z (uzz = -3w2Y+PoY3). W-W

The stress intensity factor for the penny-shaped crack for the above loadings can be
obtained from those of the elliptical crack by solving the matrix equation (17). How-
ever, extensive rearrangement of (17) by taking the proper limits when k + 0 is
necessary in order to solve this equation for b/a = 1. The following values are chosen
to simulate the penny-shaped crack in order to avoid this limiting process and round-
off errors in solving (17).

b/a = O-982 and a = 1-O.

The stress intensity factors for this nearly penny-shaped crack are determined numeri-
cally for the above loadings. They are given for the particular values of 8 in the following
equations.
 R. C. SHAH and A. S. KOBAYASHI

K,ls=450 = 0*588p, (uzr = -2poxy) (30a)

&I,=,~ = 0*602p, ( oZZ= -p,,Y + p,,y”) (30b)

&I o=oo = 0*522p, (ozz = -p($ + 3poXy”) (3Oc)

K,le=3,,0 = 0*501p, (uZZ= -3p,,xy”+p0y3) (30d)

These values of stress intensity factor compare favorably with those determined
from (29) and the range of the deviations from those of (29) is + 1.2 to -2.9 per cent.

APPLICATIONS
(a) Elliptical crack in a large beam in pure bending
Consider a large beam with an embedded elliptical crack and subjected to pure
bending. The crack is located such that the line perpendicular to it and passing through
its center is parallel to the neutral axis of the beam and is at a distance c from the neutral
axis as shown in Fig. 2. The crack is small compared to the size of the beam cross-
section and is located such that its center is at least 3(b) and 3(a) away from the four

Fig. 2. Elliptical crack in a large beam in pure bending.

free surfaces respectively. This insures that the bounding free surfaces of the beam will
have practically no effect on the stress distribution in the vicinity of the embedded
crack. The stress distribution in the crack-free beam at the location of the crack is

urz = --
Mot & (31)
1 I

where M,, is the applied bending moment to the beam and 1 is the moment of inertia of
the beam. Since the crack surfaces are stress free, the stress intensity factor can be
obtained by superimposing the solution for a crack with --mZr prescribed on its surfaces.
The resulting stress intensity factor is
 Elliptical crack under arbitrary normal loading 85

1’ (32)
KI = 7 1” (b2 ~082 0 + Q,Z sin2 0) l/4
k2b sin 8
(1+k2)E(k)-k’2K(k)

The first term in the above expression is the stress intensity factor resulting from the
average stress over the crack surface (M,,c/Z) and the second term is the stress intensity
factor due to linearly varying stress (M,yll). The second term represents the error
which would occur if the first term alone were used as an approximation. If the crack
is far away from the neutral axis, i.e., b/c is small, the error due to the assumption of a
constant stress on the crack is small. On the other hand, if the crack is very close to
the neutral axis, this error can be significant.
The above derivation of the stress intensity factor is valid only when no part of the
two crack surfaces interferes with each other upon loading. This, then, from (32)
requires that

(33)

which is identical to the result derived by Segedin [5].

(b) Elliptical crack in a plate subjected to a thermal shock

The solution of (14~) is used to determine the stress intensity factor for an elliptical
crack embedded in a thick plate subjected to thermal shock. Consider a plate of thick-
ness L, one side of which is subjected to a sudden temperature I/, as shown in Fig. 3.

T=O@ t=O

T=O T=V

Fig. 3. Plate subjected to thermal

shock.

The ambient temperature of plate is 0. Assuming the other dimensions of plate much
larger than the thickness L, temperature variation, T(g), occurs only through the
thickness of the plate. This can be, then, represented as

a2T 1 CAT
~=Kdt
(344

with the boundary conditions as

86 R. C. SHAH and A. S. KOBAYASHI

T=O (5 = 0) (34b)

T=V (5=L) (34c)

T=O (t = 0) (34d)

where 5 is the coordinate in the thickness direction, K is the thermal diffusivity of the
material and t is time.
The temperature distribution T([,t) can be represented from 1131as

T([,t) = V L+Z 2 (-1)“sinSL exp(-F)]. (35)

L rTTn=, n
The thermal stress distribution for a plate with traction-free edges and surfaces and
having the temperature variation only through the thickness can be represented as [ 141
L/2

TdSfF T5 d5 (36)
J -L/2
I
where coordinate axes x and z are perpendicular to axis 4 as shown in Fig. 3. t = 5 - L/2,
Q is the coefficient of linear expansion and E is the modulus of elasticity.
The stress distribution in the untracked plate can now be determined by evaluating
numerically (35) and (36). The stress intensity factor of the elliptical crack embedded
in this plate can be determined by relieving the normal stress existing in the corre-
sponding crack area of the untracked plate.
As a numerical example, a one-inch thick plate of 4335 steel is subjected to a sudden
temperature rise of 1OWF. The following are the properties of 4335 steel:

cx= 6.6 X lop6 in./in./“F

K = 0.01496 in.‘/sec

E = 29 X lo6 psi
7 = 0.32
oys = yield strength in tension = 210 ksi
oUzl= ultimate tensile strength = 240 ksi.

The transient temperature distribution, through the thickness, as determined by (35),

is shown in Fig. 4 for various values of time. The transient thermal stress (T,, through
the thickness for t = 2 set is shown in Fig. 5(a). Figure 5(b) shows the stress distribu-
tion ozz between 5 = 0.55 in. to 5 = 0.75 in. for various values of time. From Fig. 4, it is
evident that the steady state condition for temperature distribution for this material
is reached at t = 60 set and accordingly the stress crz vanishes (Fig. 5(b)).
Since the stress uzz is maximum between 5 = 0.55 in. and [ = 0.75 in. for most
values of time, the center of the elliptical crack is assumed to be located at 5 = 0.65 in.
with the crack geometry of a = 1.0 in. and b = 0.1 in. The stress oZL between 5 =
0.55 in. and 5 = 0.75 in. is relieved by fitting the negative of u,, with the following
polynomial of y by the method of least squares.

u,, = A60+Ao,Y+A6zYZ+A63Y3. (37)

 Elliptical crack under arbitrary normal loading 87

0’
u‘
5
2
lli
4
E

0.7 0.8 0.9 1.0

DISTANCE, 5 (INCHES)

Fig. 4. Temperature distribution in 4335 steel plate.

6.0-

Fig. S(a). Thermal stress wzz in the plate.

88 R. C. SHAH and A. S. KOBAYASHI

01 I I I

0.55 0.60 0.65 0.70 0 ‘5

DISTANCE, l (INCH)

Fig. 5(b). Thermal stress ‘T,, in the plate.

The maximum deviation between the stress calculated from (36) and (37) is less than
half of one per cent. These four coefficients obviously differ for each thermal stress
distribution and, therefore, differ for each time increment. The values of these four
coefficients Aij is substituted in matrix equation (17) for each time increment with all
other Ati taken as zero. After solving the matrix equation (17) for Cti, the resultant
stress intensity factors determined from (21) are shown in Fig. 6 for various values of
time. The stress intensity factor is maximum at one of the two ends of the minor
diameter (4 = 0.55 in. and 8 = -90” or 5 = 0.75 in. and 8 = 90’) of the ellipse. The
distribution of the stress intensity factor along the crack periphery is nearly symmetric
about the major axis. The stress intensity factor which is generally maximum at 4 =
0.75 in. (0 = 90”) is plotted against time in Fig. 7. As expected, the stress intensity
factor at the minor axis (5 = 0.75 in., 13= 900) reaches a maximum value of 268 ksi V’%.
a short time (time = 2 set) after the application of the sudden temperature rise of
100°F. It is interesting to note that the fracture toughness KIc of 4335 steel is 50
ksi fi Thus, thermal stress alone due to a sudden temperature rise of 100°F could
account for approximately 5 per cent of this value.

CONCLUSION
(1) The harmonic stress function and stress intensity factor for an elliptical crack
embedded in an elastic solid and subjected to an internal pressure loading of err =
-P(x,Y) where P(X,Y> =A~~+A~(~~~+A~IY+A~&+A~,~Y+A~~Y*+A~~+A~x*Y+
A12xy2 + Ao3y3 are derived. This solution can be applied to the thick plate containing
an elliptical crack if the plate thickness is at least three times greater than the crack
width.
(2) The results derived in this paper for the stress intensity factor of an elliptical
crack agree well with those of the tunnel crack and the penny-shaped crack.
 89

Go III -70 -50

I / 30, , -10
, 1 , , , ] j , ) , ,

0 10 30 50 m 90
ANGLE 0 IN DEGREES
Fig. 6. Transient stress intensity factor distribution with 8.

3.

TIME IN SECONDS

Fig. 7. Transient stress intensity factor at 0 = 90”.

90 R. C. SHAH and A. S. KOBAYASHI

(3) Stress intensity factor for an elliptical crack in a large beam in pure bending is
given.
(41 The derived results are used to determine the transient stress intensity factor
for an elliptical crack in a thick plate, one side of which is subjected to a sudden
temperature rise.

Acknowled~emenrs-This investigation was accomplished by the authors as a part of the independent

fracture mechanics research which is being uerformed bv the Space Division of The Boeing Company. The
authors wish to thank Mr. C. F. Tiffany of-the Space Division, The Boeing Company, for his encouragement
and general counsel.

REFERENCES
[ 1J A. E. Green and I. N. Sneddon, The distribution of stress in the neighborhood of a flat elliptical crack
in an elastic solid. Proc. Camb. Phil. Sot. 46, 159- 163 (1950).
[2] M. K. Kassir and G. C. Sih, Geometric discontinuities in elastostatics. J. Math. Mech. 16, 927-948
(1967).
[3] M. K. Kassir and G. C. Sih, Three-dimensional stress distribution around an elliptical crack under
arbitrary loadings. J. appl. Mech. 60 l-6 11 (1966).
[4] M. K. Kassir and G. C. Sih, Three-dimensional thermoelastic problems of planes of discontinuities
or cracks in solids. Ofice of Naval Res. Contr. No. 610(06), Task Order NR064-476, Tech. Rep.
No. 2 (1966).
[5] C. M. Segedin, Some three-dimensional mixed boundary value problems in elasticity. Univ. of Wash.,
Dept. ofAeronautics andAstronautics Rep. No. 67-3 (1967).
[6] C. M. Segedin, Thermoelastic stresses due to uniform heat flow past rigid elliptic inclusions. L’niu. of
Wash., Dept. ofAeronautics andAstronautics, Rep. No. 67-5 (1967).
171 I. N. Sneddon and D. S. Berry, The classical theory of elasticity. Handb. Phys. 6, Springer, Berlin
(1958).
WI J. F. Price and R. H. Simonsen, Various methods and computer routines for approximation, curve
fitting, and interpolation. Boeing scienr. Res. Lab., Dl-82-015 1 Revised.
[91 P. F. Bvrd and M. D. Friedman. Handbook qf Elliptical Integrals for Engineers and Physicists.
Springer; Berlin (1954).
[lOI G. R. Irwin, Crack extension force for a part-through crack in a plate. J. appl. Mech. 65 l-654 (1962).
[ill G. C. Sih, P. C. Paris and F. Erdogan, Crack-tip stress intensity factors for plane extension and plate
bending problems. J. uppl. Mech. 306-3 12 (1962).
1121F. W. Smith, A. S. Kobayashi and A. F. Emery, Stress intensity factor for penny-shaped cracks, part I,
infinite solid. J. appl. Mech. 946-95 1 (1967).
u31 H. S. Carslaw and J. C. Jaeger, Conduction of Heat in Solids. Clarendon Press, Oxford (1947).
[I41 B. A. Boley and J. C. Weiner, Theory of Thermal Stresses. Wiley, New York (1960).

APPENDIX I
Evaluation of partial derivatives of @ii
a%p/az” is required to find the stress grz at any point on z = 0 plane. These deriva-
tives of I#Jare listed here with the needed relation

aA_ 2z(a2+h)(b2+h)
(Ala)
az - (A-/A)@--v)

a2Qoo
-- _ m ds 2c,z ah
az*
_2c
00
I* m(~)l”’ +A[Q(A)]“‘z
WW

a%
10 = -_2C& (A2b)
a22 s(az+s~~~(s)~“2-A(az+A)~~(A)~1~z~ >
 Elliptical crack under arbitrary normal loading 91

a”@20
--
_
az2
_4c
20
IfpS(a’+S;;Q(S),112-3xz
f*ffi
s(a’+s)~;Q(s)]“’
-y2
[Q&3,*-32.
f s’(02+s~~Q(s)~“2+A(~
fAm*=
I
WW

a'@ ds
-_-U = -2c19y WeI
a2 [Q(s)]:~“- [Q(k)]$ 1

a2@ =
02
a22
-4c,,
(IAm ~(b~+s;;Q(s),"~-~~ I
hm [Q ps”) ,312 - 3y2
I Am s(b’+s)~;Q(s),“’

-322 (A2f)
I Ams2(b”fsPse(s)l”2+h(hi+h:~;ZQ(X)11/2~

s(a’+s)~~Q(~)]‘l~-~~’ fAms(a’+$Q(r)]“”

-3y2
fkm (cz~+s;Q(s)]~~~-~~~ fkm s~(u~+s;;Q(s)]“~

2X22 ah
dz (A23
+h(a2+A)3[Q(h)]112

a2@,,_
-
a22
- --4czly [Q $]3/2 -3x2 f Am (a'+s;Q(~)]~~~-~' f f (b"+s;Q(s)]""

m
ds 2X2Z
-322 (A2h)
f ,, dQW2+ (~~~+h)[Q(h),~/~s

a2Q12
-- _ _4c
a2
x
Ifim f10 fAm
m
12 [QF),3”-x2
(cz~+s)~Q(s)]“‘~-~~~ (62+~;Q(~)]312

f s[QW1312+
ds 2y2z
-322 (A2i)
h (b2+A)[Q(x)]3/2$

fhm
a2@
-_-B = -4c,,y 3
a22 (I Am ~(b~+s):;Q(s)]~~~-~~~ (b2+~;Q(~)]312

-5y2
fkm fAm
s(b’+s)~;Q(~)]~~~-~~~ ~‘(bz+~;l’s[Q(~)]”

2y2z ai
Wj 1
+h(b2+h)3[Q(h)]1~2dz

These elliptical integrals are evaluated in terms of Jacobian elliptic functions in Appen-
dix II.
92 R. C. SHAH and A. S. KOBAYASHI

APPENDIX II
Evaluation of elliptic integrals
Various elliptic integrals are encountered in the text and Appendix I. Some ot
these elliptic integrals appear in parts in [91. For continuity of development, however,
all elliptic integrals are listed in the order in which they appear in the text and they
are evaluated in terms of incomplete elliptic integrals of the first and second kinds,
and Jacobian elliptic functions snu Ir cnul and dnu,. The Jacobian elliptic function SW,
is related to A by
a2
sr?u, = - (A3a)
a2 + A

where the modulus k and complementary modulus k’ are defined as

k2 = -,
a2 - b2 k’2 = $. (A3b)
a2

The incomplete elliptic integral of the first kind, ul, in (A3a) takes all real values be-
tween 0 and K(k). The following standard notations of Jacobian elliptic functions are
used.
tnuI = %; dcu, = ---L dnu
cnu , cdu, = cnul
cm1 1 dnu,
(A3c)
1 1 mu,
ndu, = -* , ncu, = --;sdu,=-
dnu, 1 dnu,

E(u,) = j-;’ d n2u du = incomplete elliptic integral of the second kind. (A3d)
m
ds
= & [-E(u,) + tnu, dnu,] (A44
J ,, s[Qb)l"

=~[(1-2k2)E(uI)-k’2uI+k2tnu,dnul] (A4b)
IAms(a2+sysQ(s)]li2
CC
=-&-J-(l+k’)E(u,)+k” u, + k2snu, (dcu, + cdu,)] (A4c)
J h s(b’+s;;Q(s),“’

2
3a,1Pk,z [(2+3k2-8k4)E(u,)-2(2k2+l)k’2u,
I Ams(a’+s)~~Q(s)]1~2=P
+3Pdnu, tnu, - k”k”snu, cnul dnu,] (A4d)

=& [-2(k4+k’Z)E(u1)+ (2-k2)k’2u,+k2snu,(cdu,+k2dcu,)]

I : [Q&r,z
(A4e)

m
=&[(8k2k”+5k2-3)E(u,)+ (3k’2-k2)k’2u,
s A s’(a’+s~;Q(s)]lil
+k2tnu, dnuI(kr2nc’u, - 5k” - 2)] (A4f)
 Elliptical crack under arbitrary normal loading 93

2 3k2+2-- 1W (E(u,)
I
,p s(bZ+s)$Q(sp =- 3a7k4k14K kf2 >
-Pk2snu1 cdu,)

+2(3k2 - l)u, + 3~k’~t~u~ nd3u,

+(3P+ l)k2snu1 mu, ndSul

1 Wk)
=y&j-p [(2k2k’2+5k”+3)E(u,) - (k2+3)k’2u,

+k2k’4tnu, nc2ul ndu, + (k2krz - kt!- 3)k%zu, cdu,

+ ( 3kz - 7) k2snu, dcu,] (A4h)

ds 2
‘xi

J
h s(a2+s)3[Q(s)]“2
=. ,5agpk,2 [(8+9k2+ 16~-48~)~(ff~) - (24kI+ 13k2+8)

* kf2u1 + 15Pdnu, mu, - (3kzsn2u, + 9k2 t 4)

- k2kr2mul emI dnul] (A4i)

+3)k’*u, + 3k2snu, cdu, + 3k%u, dnu, f k2k’4snu1 cnu, dnuJ

(A4j 1

=& [(16kilkr2+8kl-4k2-2)E(u,) + (2+5k2

s’(a’+s;~Q(s),”

-8k4)k’2u, + (k’2m2u1 - 8k12 - 2)IPtnu, dnu,

+k”k’4snu, mu, dnu,] (A4k)

Jkm (b'+s;Q(s)]""
=A [(6k“-3k2+ 16-8/k’2)E(u,)

+ (3k2kt2 + 12k” -8)u, + 3k4k12tnu, nd3u,

-3k+nu, drzu, + (3kl+ l)k2snu, cdu, nd2ut

- ( 13 + 3k2 - 8/k’2) k2snu, cdul] (A411

ds
m
2
I
h sEQ!(s)l”‘”
=~[(8-2kfz-3k2~~-8k2k’4~~(u~)+2(2k2~z-kt~-2)

* kr2u1 + PkF4sn5u1 tn3u, sdu, + ( 2k2 - 6) k4 tnu, dnu,

+ (3k’2 - 4)snu, cdul + (2k2 + 1 + k2sn2u1) k’4snuI cnu, dnu,]

(A4m)
94 R. C. SHAH and A. S. KOBAYASHI
m
2
I ,5a9k8k,8 { (40kZk12-48k2 + 88k-‘kf2- 15k“k’4)E(u,)

+ (l&Y - 451i4kf2+ 8k2)k’2u, + 15k’6k%n3u, mu, sd5u,

+ Psnu, cdul [ (3k2 - 115k2kr2- 58k4kt2 + 45)

+ k’2nd2u, (k2+.S) ( 13k2-9+ 3kZkf2sd2u,) + 15k’4nd2u,]

- 15k2k’6snu, cnu, dnu,) (A4n)

+ ( 11k’2 - 8 - k’4) k2u, + k’4k%n3u1 tn3ul sd3u,

- 6k4k’“tnu, nd3u, + 6k9nul dnu,
+ k?mu, cdu, nd2u, (6Pk’” - 7k2 - 1)
+ k%zu, cdu, (18 + 16k2- 16/k’2)
- kzkr4snu, cm, dnu,] (ASP)

APPENDIX III
~~r~~ati~n of stress intensity factur
The derivation of the stress intensity factors is lengthy and hence some of the
intermediate steps in the derivation are given. In the following derivations only those
terms of CT,,at p = 0 are considered which involve singularity of A at A = 0. From (3~)
and (19b), stress intensity factor can be written as

221I*=0
liz(a2 sin2 0 + b2 cos2 8)“* F% Xij2 aZ
i * (A5)

For the sake of brevity, 2G(r/ab)1’2 (a2 sin2 8 + b2 co9 6)1/4 is substituted by A in
the following expressions. On using (Al) through (A5) stress intensity factors can be
written as follows:

@oo:
K, = 2ACoo lii h112

_ 4AGoo (A6a)
ab
dnu, tnu,
@IO: KI = 4AClo lim
A--t0
X”~X
a3b2 -a2b2(h-:)(h-vv) [h~~,]l’z),;o
= 4AC,, cos 8
(A6b)
a2b
dnu, tnu,
@OX: Ki = 4AC,,%lim h1’2y
h-+0 ab4 -a2b2(A-y)(h-v) [h~~)]1’2}p_,
= 4AC,, sin B WC)
ab2
 Elliptical crack under arbitrary normal loading

( 1_3X2_L _ z2
Q20: K, = 8ACZo Fz A”2{ “;;;;“I a2 b2 b2cn2uI >

+aw::F(w [h(Z;)311’21p=0

= _ l6AC20 COS’ 8
(A64
a3b

dnu, tnu,
Q11: KI = 4AC,, lim
A-t0
X1’2xy
a3b4 -a2b2(h-p) (r”v) [Q(h)]1/2 p=.

= 4AC,, cos 8 sin 13

VW
a2b2

&?: KI = 8AC,, hi A”2

- Z2
a2dn2ulcn2u, > +a2b2(?~~(k-v) [h$~k)3~2]wZ,,

l&4 Co2 sin2 8

b460

= - l6AC30 cos3 8
a4b
Wig)

-- 2x2/M
1
b2z2 sn6ultn2ul
a4 dn2ul +a2b2(h-p)(h-v)(a2+h)[Q(h)]1i2

= - 16AC,, cos2 8 sin 8 Wh)

a3b2

-- b2z2 sn6u,tn2u1
a4 dn2ul 1 2Y2P
+a2b2(A-p)(h-v)(b2+h)[Q(A)]1~2 I Ir=o

= - l6AC12 cos 8 sin2 8 (A6i)

a2b3
96 R. C. SHAH and A. S. KORAYASHI

1
5y2sn8u,
_~_
3z2sn6ultn2u, 2Y2P
a2dn2u, a2 +a2b2(h--)(A--)

= _ 16AC,, sin3 8
ab4

(Receiurd 25 Septrmber 1968)