F.

atmccra~~racturt Mtchmucs Vol 11, pp 817--829 0013-7944/79/0601--0g17/$02 00/0

© Pergamon Press Ltd, 1979 Printed m Great Brttmn

STRESS-INTENSITY FACTORS FOR A WIDE

RANGE OF SEMI-ELLIPTICAL SURFACE
CRACKS IN FINITE-THICKNESS PLATES
I S. RAJU and J C NEWMAN,Jr.
NASA-LangleyResearch Center, Hampton, Virgima,U.S.A

Abstract--Surface cracks are among the more commonflaws m aircraft and pressure vessel components.
Accurate stress analyses of surface-cracked componentsare needed for reliable prediction of their crack
growth rates and fracture strengths. Several calculations of stress-intensity factors for semi-elliptical
surface cracks subjected to tension have appeared m the hterature. However,someof these solutions are in
disagreement by 50-100%.
In this paper stress-intensity factors for shallow and deep semi-elliptical surface cracks in plates
subjected to tension are presented. To venfy the accuracy of the three-dimensionalfinite-element models
employed, convergencewas studied by varyingthe numberof degrees of freedom in the models from 1500
to 6900. The 6900 degrees of freedom used here were more than twice the number used in previously
reported solutmns. Also, the stress-intensity variatmns m the boundary-layerregion at the intersection of
the crack with the free surface were investigated.

NOTATION
a depth of surface crack
b half-widthof cracked plate
c half-lengthof surface crack
F stress-intensityboundary-correctionfactor
h half-lengthof cracked plate
K stress-intensityfactor (Mode I)
Q shapefactor for an elliptical crack
S applieduniform stress
t plate thickness
x, y, z Cartesiancoordinates
v Poisson's ratio
4, parametricangle of the elhpse

INTRODUCTION
SURFACECRACKSare among the more common flaws in aircraft and pressure vessel components.
Accurate stress analyses of these surface-cracked components are needed for reliable predic-
tion of their crack growth rates and fracture strengths. Exact solutions to these difficult
problems are not available; therefore, approximate methods must be used. For a semi-elliptical
surface crack in a finite-thickness plate subjected to tension (Fig. 1), Browning and Smith[I],
Kobayashi[2] and Smith and Sorensen [3] used the alternating method and Kathiresan [4] used
the finite-element method to obtain the stress-intensity factor variations along the crack front
for various crack shapes. For a deep semi-elliptical surface crack (with a l t = 0.8 and a l c = 0.2)
subjected to tension, the stress-intensity factors obtained by Smith and Sorensen[3],
Kobayashi [2] and Kathiresan [4] disagreed by 50-100%. The reasons for these discrepancies are
not well understood.
This paper presents stress-intensity factors for shallow and deep semi-elliptical surface
cracks in plates subjected to uniform tension. To test the validity of the present analysis, two
crack configurations, both embedded in a large body subjected to uniform tension, were
analyzed: (1) a circular (penny-shaped) crack and (2) an elliptical crack. These results are
compared with exact solutions from the literature [5]. To verify the accuracy of the solutions
for surface cracks in finite-thickness plates, convergence was studied by varying, from 1500 to
6900, the number of degrees of freedom in the finite-element models. The 6900 degrees of
freedom used here were more than twice the largest number used previously [4]. These models
were composed of singular elements around the crack front and isoparametric (linear strain)
elements elsewhere. Mode I elastic stress-intensity factors were calculated by using a nodal-
force method.

817
818 I.S. RAJU and J. C NEWMAN, Jr.

2h .// ! ~

-(

2 b ~ ~ / t

Fig 1. Surface crack in a plate subjected to tension.

The stress-intensity variations in a boundary-layer region at the intersection of the crack

with the free surface were also investigated for a typical semi-circular surface crack. Here the
finite-element models used were highly refined in the boundary-layer region.
The computations reported herein were conducted on a unique computer called the
STAR-100 at the NASA Langley Research Center.

~ D ~ O N A L ANALYSIS
Several attempts [6--8] have been made to develop special three.dimensional finite-elements
that account for the stress and strain singularities caused by a crack. These elements had
assumed displacement or stress distributions that simulate the square-root singularity of the
stresses and strains at the crack front. Such three-dimensional singularity elements[6, 9] were
also used herein for the analysis of finite-thickness plates containing embedded elliptical or
semi-elliptical surface cracks (see, for example, Fig. 1).

Finite-element idealization
Two types of elements (isoparametric and singular) were used in combination to model
elastic bodies with embedded elliptical cracks or semi-elliptical surface cracks. Figure 2 shows
a typical finite-element model for an embedded circular crack in a large body (h/a = b/a = 5).

S Szngularity element
I Isoparametrlc element

ith plane y
/

• r

(a) Specxmen model. (b) Element pattern around crack

front on the i th plane.

Fig. 2. Finite-element idealization for an embedded circular (penny-shaped) crack.

 Stress-intensityfactors of semi-ellipticalsurface cracks in finite-thicknessplates 819
This model idealizes one eight of the body. Various numbers of wedges were used to form the
desired configuration. Figure 2(a) shows a typical model with eight wedges. Each wedge is
composed of elements that are identical in pattern to that shown in the 4, = constant plane. The
arrangement of the elements around the crack front is shown in Fig. 2(b). The isoparametric
(linear strain) elements (denoted as I) were used everywhere except near the crack front.
Around the crack front each wedge contained eight "singularity" elements (S) in the shape of
pentahedrons. The "singularity" elements had square-root terms in their assumed displacement
distribution and, therefore, produced a singular stress field at the crack front. Details of the
formulation of these types of elements are given in Refs. [6] and [9] and are not repeated here.
The finite-element model for the embedded elliptical or semi-elliptical surface crack was
obtained from the finite-element model for the circular crack by using an elliptic trans-
formation. This transformation was needed because the stress-intensity factors must be
evaluated from either crack-opening displacements or nodal forces along the normais to the
crack front. If (x, y, z) are the Cartesian coordinates of a node in the circular-crack model and
(x', y', z') are the coordinates of that same node in the elliptical-crack model, then the
transformation is

x'=x 1+~), y'=y, z'=z for a~<lc

and (1)

x'=x, y'=y, z'=z l+x---f~) for c 1

for x and z not at the origin. Figure 3 shows how circular arcs and radial lines in the x, z plane
of the circular-crack model are transformed into ellipses and hyperbolas, respectively, in the
x', z' plane of the elliptical-crack model using eqns (1). Because eqns (1) are not valid at the
origin, a circle of very small radius, allO00, was used near the origin in the circular crack model.
The small circle maps onto an extremely narrow ellipse in the x', z' plane. The use of the small
circle avoids ill-shaped elements near the origin in the elliptical-crack model. Figure 4 shows a
typical finite-element model of a finite plate containing an elliptical crack. The transformation
reduced the bit: ratio; therefore, in order to maintain blc I> 4, additional elements were added
along the x'-axis to eliminate the influence of plate width.

-Z'

Ta a

1 ~-- a---4
x Ij c .f

(a) C x r c u l a r crack. (b) Elllptlc crack.

Fig. 3. Circleto ellipsetransformation.

Stress-intensity factor
The stress-intensity factor is a measure of the magnitude of the stresses near the crack
front. Under general loading, the stress-intensity factor depends on three basic modes of
deformation (tension and in- and out-of-plane shear). But here only tension loading was
considered and, therefore, only Mode I deformations occurred. The Mode I stress-intensity
 820 I S. RAJU and J C. NEWMAN, Jr

-z f

x ~
/
J iI
_1
I- e -I

L z ,

(a) Speclmen tnodei. (b) Element pattern on y'= 0 plane.

Fig 4. Fmtte-elementideahzatlonfor a semvelhptlcalsurfacecrack

factor, K, at any point along an elliptical or semi-elliptical crack in a finite plate is taken to be

(a a, ,#) (2)

where S is the applied stress, a is the crack depth, Q is the shape factor for an ellipse and is
given by the square of the complete elliptic integral of the second kind[5]. The boundary-
correction factor, F(a/t, a/c, ~), is a function of crack depth, crack length, plate thickness and
the parametric angle of the ellipse. The present paper gives values for F as a function of alt
and 4' for a/c = 0.2, 0.4, 0.6, 1.0 and 2.0. The a/t values ranged from 0.2 to 0.8.
The stress-intensity factors from the finite-element models of embedded elliptical and
semi-elliptical surface cracks were obtained by using a nodal-force method, details of which are
given in the appendix. In this method, the nodal forces normal to the crack plane and ahead of
the crack front are used to evaluate the stress-intensity factor. In contrast to the crack-opening
displacement method[6], this method requires no prior assumption of either plane stress or
plane strain. For a surface crack in a finite plate, the state of stress varies from plane strain in
the interior of the plate to plane stress at the surface. Thus, the crack-opening displacement
method could yield erroneous stress-intensity factors.

RESULTS AND DISCUSSION

In the following sections a circular (penny-shaped) crack and an elliptical crack completely
embedded in a large body subjected to uniform tension were analyzed using the finite-element
method. The calculated stress-intensity factors for these crack configurations are compared
with the exact solutions to verify the validity of the present finite-element method.
For a semi-circular and semi-elliptical surface crack in a finite-thickness plate, convergence
of the stress-intensity factors was studied while the number of degrees of freedom in the
finite-element models ranged from about 1500 to 6900. The stress-intensity variations in a thin
boundary-layer region at the intersection of the crack with the free surface were also
investigated. The stress-intensity factor variations along the crack front for semi-circular
(a/c = I) and semi-elliptical (a/c = 0.2, 0.4, 0.6, 1.0 and 2.0) surface cracks were obtained as
functions of a/t with h/c= blc>~4. Whenever possible, these stress-intensity factors are
compared with results from the literature.

Exact solutions
In this section a comparison is made between the stress-intensity factors calculated from the
finite-element analysis and from the exact solution ["5] for an embedded circular (penny-shaped)
crack (a]c = !) and an embedded eliptical crack (ale = 0.2) in an infinite body. In the finite-
element model, h and b were taken to be large enough that the free boundary would have
 Stress-intensity factors of semi-elliptical surface cracks in finite-thickness plates 821

a negligible effect on stress intensity. The boundary correction on stress intensity for a circular
crack in a cylinder of radius b with b/a =5 is about 1%[10]. Therefore, to simulate a
large body the finite-element model was assigned the dimensions h/a = b/a = 5, along with 3078
degrees of freedom. The calculated stress-intensity factors along the crack front from the
circular crack model were about 0.4% below the exact solution.
The embedded elliptical crack (alc= 0.2) model, with ldc = blc = 4 and tla = 5, had 3348
degrees of freedom. The influence of finite boundaries on stress intensity was estimated to be
about 1%. The finite-element analysis gave stress-intensity factors along the crack front
generally within 1% of the exact solution, except in the region of sharpest curvature of the
ellipse, where the calculated values were about 3% higher than the exact solution. Further
refinement in the mesh size in this area gave more accurate stress-intensity factors.
Because the present method yielded stress-intensity factors for completely embedded
circular and elliptical cracks generally within 0.4-1% of the exact solutions, the method was
considered suitable for analyses of more complex configurations, provided that enough degrees
of freedom were used to obtain good convergence.

Approximate solutions
To verify that the finite-element meshes used for the circular and elliptical crack models
were sufficient to analyze cracked plates with free boundaries, through-the-thickness cracks
with crack length-to-width ratios ranging from 0.2 to 0.8 were analyzed under plane strain
assumptions. The meshes used here were exactly the same as those which occur on the 4' = ~r/2
plane of the circular and elliptical crack models. These meshes were then used to model the
center-crack tension specimen. For crack length-to-width ratios (c/b) ranging from 0.2 to 0.6,
the finite-element results were within 1.3% of the approximate solutions given in Ref. [10]. For
c/b = 0.8, the finite-element result was 2% below the solution given in Ref. [10]. These results
indicate that the mesh pattern along any plane has enough degrees of freedom to account for
the influence of free boundaries on the stress-intensity factor.

Convergence
In the previous section the mesh pattern along any 4' = constant plane was found to be
sufficient to account for the influence of free boundaries even for very deep cracks. In this
section, the mesh pattern in the angular direction, 4,, is studied. Figures 5 and 6 show the results
of a convergence study on the stress-intensity factors for a semi-circular surface crack and a
semi-elliptical surface crack, respectively, in a finite-thickness plate. The larger numbers of
degrees of freedom are associated with smaller wedges in the 4,-direction and a more accurate
representation of the crack shape.

1.2

S ~ .8
Degrees of freedom

•1500

Q 4317 .
.4 I

0 6195 t
2a

I I I I
0 .25 .5 .75 i
2~

Fig. 5. Convergence of stress-intensity factors for a deep semi-circular surface crack (O = f2/4; a/t = 0.8).
822 I S. RAJU and J. C. NEWMAN,Jr
2.0

1.6

1.2

D e g r e e s of f r e e d o m
.8

O 2727 , ._
[3 4 7 9 7 ~ ~= _,
Q 6867 2c -I

I ,,. i I
0 21
. 5 .5 .75 1
2~

Fig. 6. Convergenceof stress-intensityfactorsfor a deepsemi-ellipticalsurfacecrack(Q = 1.104;alt = 0.8,

alc= 0.2).

For the semi-circular crack (Fig. 5), hla = bla = 5 and alt = 0.8. This configuration was
chosen because the close proximity of the back surface to the crack front was expected to
cause difficulty in achieving convergence. The models were composed of either 2, 4, 8 or 12
equal wedges. The number of degrees of freedom ranged from 1500 to 6195. The two finest
models (4317 and 6195 degrees of freedom) gave stress-intensity factors within about I% of
each other.
Figure 6 shows the convergence study for a semi-elliptical surface crack (ale = 0.2). The h/c
and blc ratios were equal to four and, again, alt was chosen as 0.8. The number of degrees of
freedom ranged from 1692 to 6867. The eight-wedge model with 4797 degrees of freedom gave
results within about I% of those from the finest model.
The results shown in Figs. 5 and 6 demonstrate rapid convergence of the solution and
indicate that the eight-wedge model may be used to obtain stress-intensity factors as a function
of alt and ale. However, further consideration of the stress-intensity variations in the
bounding-layer region at the intersection of the crack with the free surface must be given.

Boundary.layer eHect
Hartranft and Sih[ll] proposed that the stress-intensity factors in a very thin "boundary
layer" near the free surface drop off rapidly and equal zero at the free surface. To investigate
the boundary-layer effect, a semi-circular surface crack in a large body was considered. Three
different finite-element models with 8, I0 and 14 wedges were analyzed. The eight wedge model,
shown in Fig. 2(a), had eight equal wedges. The other models had nonuniform wedges and were
obtained by refining the eight-wedge model near the free surface. The smallest wedge angle for
the ten- and fourteen-wedge models were ~-/32 and ~-/128, respectively. The stress-intensity
factors obtained f~om the three models are shown in Fig. 7. The maximum stress-intensity
factor obtained from the eight-wedge model at the free surface was slightly less (1.4%) than the
peak value obtained from the fourteen-wedge model. These results also show that the stress
intensities near the free surface were affected by the refinements and drop off rapidly as
proposed by Hartranft and Sih. However, the stress-intensity distributions in the interior
(~b > ~r/16) were unaffected by the refinements. These results strongly suggest that the stress
intensifies in the interior are insensitive to whatever is happening in the boundary layer. An
estimate for the size of the boundary layer was obtained from Ref. [I I], assuming that the crack
was a through crack of length, 2a, and this estimate is also shown in Fig. 7.
In view of the highly localized boundary-layer effect, the eight-wedge model was used
subsequently to obtain stress-intensity factors as a function of ale and alt. The stress-intensity
factor obtained at the free surface with the eight-wedge models were interpreted as an average
stress-intensity factor near the free surface.
 Stress-intensityfactorsof semi-ellipticalsurfacecracksin finite-thicknessplates 823

1.25

e 8 Wedge model
i. 20 ® 10 Wedge model
4 Wedge m o d e l

1.15

I.i0
K

sN
1.05
Estlmated

2a
I I I I
0 .25 .5 .75 1

ff

Fig. 7. Effects of mesh refinementnear the free surface on the distribution of stress intensity for
semi-circularsurfacecrack(Q = 'n'2/4; alt = 0.2).

Semi-circular surface crack in a finite-thickness plate

Figure 8 shows and Table 1 presents the stress-intensity factors for a semi-circular surface
crack in a finite-thickness plate as a function of the parametric angle, 4,, and the crack
depth-to-plate thickness ratio, aJt. Near the intersection of the crack and the free surface, the
stress-intensity factor increases more rapidly with aJt than at the deepest point (4, = ~r/2). For
each value of a/t, the stress-intensity factor calculated from the eight-wedge model is largest at
the free surface.
For a semi-circular surface crack with a/t = 0.55, the stress-intensity factors calculated by
Browning and Smith[l], who used the alternating method, were about 1-3% below the present
results for various values of 4'.

Semi-elliptical surface crack in a finite-thickness plate

The stress-intensity factors for a semi-elliptical surface crack (a/c = 0.2, 0.4, 0.6 or 2.0) in a
finite-thickness plate as a function of the parametric angle, 4,, and the crack depth-to-plate
thickness ratio, alt, are given in Table 1.
1.6
a/t
.8

1.2

S ~ .8

.4
~:: J,
2a

I I I I
0 .25 .5 .75 1
2~

Fig. 8. Distribution of stress-intensity factors along crack front for a semi-circular surface crack (Q =
lr214).
824 1 S. RAJU and J C NEWMAN, Jr

Table 1 Boundary correction factors, F, for semi-elliptical

surface cracks subjected to tension (~,=03: F =
K/SVlrra/Ol)

air
ale 2~ 02 04 0.6 0.8

0 0617 0724 0899 I 190

0.125 0,650 0 775 0 953 I 217
0 25 0 754 0 883 I 080 1 345
0 375 0 882 1 009 I 237 I 504
02 05 0 990 I 122 1 384 1.657
0.625 1 072 I 222 1 501 I 759
0 75 I 128 1,297 1.581 1.824
0 875 1,161 1 344 1 627 1.846
I0 I 173 1.359 I 642 1 851

0 0 767 0.896 1.080 I 318

0 125 0 781 0 902 1 075 I 285
025 0842 0946 1 113 1 297
0 375 0 923 I 010 1 179 1 327
04 0,5 0 998 I 075 1 247 1 374
0.625 1 058 1 136 I 302 1.408
0.75 1 103 1.184 1.341 1.437
0.875 I 129 1 214 1 363 1.446
I0 1 138 I 225 I 370 1 447

0 0.916 1.015 1 172 I 353

0.125 0.919 1,004 I 149 1 304
0 25 0.942 1 009 1 142 1.265
0 375 0 982 1.033 1 160 1 240
06 0.5 1 024 1 062 1.182 1 243
0 625 1.059 I 093 I 202 1.245
075 I 087 1,121 1 218 1 260
0.875 I 104 1 139 1.227 1.264
1.0 I 110 1 145 I 230 1 264

0 1 174 1 229 1 355 1,464

0 125 1.145 I 206 1 321 410
0 25 1.105 1 157 I 256 314
0.375 I 082 1 126 1.214 234
10 05 1067 I 104 1 181 193
0 625 I 058 1 088 I 153 150
0 75 1.053 1.075 1 129 134
0 875 1.050 1.066 1.113 1.118
10 1 049 I 062 1 107 I 112

0 0 821 0 848 0 866 0 876

0.125 0.794 0.818 0.833 0 839
0.25 0 740 0 759 0.771 0.775
0,375 0.692 0 708 0.716 0.717
20 05 0646 0659 0664 0,661
0 625 0,599 0,609 0,610 0,607
0.75 0.552 0.560 0.560 0,554
0875 0512 0519 0.519 0.513
1.0 0 495 0 501 0 501 0 4%

Figure 9 shows the stress-intensity factors for a semi-elliptical surface crack (a/c = 0.2) as a
function of the parametric angle and the crack depth-to-plate thickness ratio. For each value of
a/t, the maximum stress-intensity factor occurs at the deepest point (~ = ¢r12). Also, the
maximum stress-intensity factor is larger for larger values of a[t.
Figure 10 shows the stress-intensity distributions for various semi-elliptical surface cracks
as a function of alc with alt = 0.8. For smaller values of a/c, the normalized stress-intensity
factor at ~k = ~r12is larger. The maximum normalized stress-intensity factor occurred at ~b = ¢r[2
for a/c <~0.4 but occurred at ~b = 0 for a/c >I 1.0.
Figures 11 and 12 shows stress-intensity factors obtained by several investigators for a
semi-elliptical surface crack in a finite-thickness plate. Figure 11 shows the results for a surface
 Stress-intensity factors of semi-elliptical surface cracks in finite-thickness plates 825
2,0

1.6

1.2,

I
.8

.4

i i I i
.25 .5 .75 1
2#

Fig. 9. Distribution of stress-intensity factors along crack front for a semi-elliptical surface crack (Q =
I 104", alc = 0.2).

2.0 a/c
.2

1.6

.4 A

1.2
K

.8

.4

f:= 2¢ q
* I i t
.25 .5 .75 1
2_¢
;r

Fig. 10. Distributionof stress-intensityfactors alongcrack front for semi-ellipticalsurfacecracks (air =

0.8).

crack with a/c = 0.2 and alt = 0.8. Smith and Sorenson[3] and Kobayashi[2] used the alternat-
ing method and Kathiresan[4] used the finite-element method to obtain stress-intensity factor
variations along the crack front. These three solutions disagree by as much as 50-100%. The
reasons for these discrepancies are not well understood. The present results, shown as solid
symbols, are considerably higher than the previous solutions[2-4]. The results from Sm/th and
Sorensen [3] are generally closer to the present results, though 10-25% lower.
Figure 12 shows a comparison of the maximum stress-intensity factors obtained by several
investigators for a semi-elliptical surface crack as a function of a/t. The maximum stress-
intensity factors occurred at 4, = 7r/2. The present results are shown as solid symbols. The open
symbols show the results from Smith and Sorensen[3], Kobayashi[2, 12] and Rice and
Levy[13]. The results from Rice nad Levy, obtained from a line-spring mo.del, are about 3.5%
below the present results over an air range from 0.2 to 0.6. For a/t >0.6, the Rice-Levy
solution shows a reduction in stress intensity. The dash-dot curve shows the results of an
approximate equation proposed by Newman[14] for a wide range of a/c and a/t ratios.
826 I S RAJU and J. C NEWMAN, Jr.
2.0

1.6

1.2
K

.8

Present results
.4 O Smlth a n d S o r e n s e n {3]
[3 Kobayash~ [2]
Kathlresan [4]
i | | |
.25 5 .75 1

Fig. 11. Comparison of stress-intensity factors for a deep semi-elliptical surface crack (Q = 1.104; a/t = 0.8,
alc = 0.2).

2.2
• Present results /
R i c e a n d Levy [13]
2.0 O Smlth and S o r e n s e n [3] /
/A Kobayashx [2] /,
Q Kobayash~. [12] ~
1.8
----- Newman [1~1 / . /

g 1.6
X/f/.="- i" /
1.4

1.2

I.C

I i i i !
.2 .4 .6 .8 1
a
t

Fig. 12. Comparison of the maximum stress-intensity factor for a semi-elliptical surface crack as a function
of a/t (4 = ~r/2; Q = 1.104; a/c = 0.2).

Newman's equation is within _+5% of the present results over an sit range from 0.2 to 0.8.
Newman's equation gives a good engineering estimate for the maximum stress-intensity factor.

CONCLUDING REMARKS
A three-dimensional finite-element elastic stress analysis was used to calculate stress-
intensity factor variations along the crack front for completely embedded elliptical cracks in
large bodies and for semi-elliptical surface cracks (crack depth-to-half crack length ratios were 0.2
to 2.0) in finite-thickness plates. Three-dimensional singularity elements were used at the
crack front. A nodal force method which requires no prior assumption of either plane stress or
plane strain was used to evaluate the stress-intensity factors along the crack front.
Completely embedded circular and elliptical cracks were analyzed to verify the accuracy of
the finite-element analysis. The stress-intensity factors for these cracks were generally about
0.4--1% below the exact solutions. However, for the elliptical crack the calculated stress-
intensity factors in the region of sharpest curvature of the ellipse were about 3% higher than the
 Strcss-mtensity factors of semi-elliptical surface cracks in finite-thicknessplates 827

exact solution. The numbers of degrees of freedom in the embedded crack models were about
3000. A convergence study on stress-intensity factors for semi-elliptical surface cracks in
finite-thickness plates showed that convergence was achieved for both the semi-circular and the
semi-elliptical surface crack with about 4500 degrees of freedom.
For the semi-circular surface crack the maximum stress-intensity factor occurred near the
intersection of the crack with the free surface. On the other hand, for the semi-elliptical surface
crack (crack depth-to-crack half length ratio of 0.2), the maximum stress-intensity factor
occurred at the deepest point. For both the semi-circular and semi-elliptical surface crack the
stress-intensity factors were larger for larger values of crack depth-to-plate thickness ratio.
For the semi-circular surface crack, the stress-intensity factors calculated by Browning and
Smith [1] using the alternating method, agreed generally within about 3% with the present results.
However, for semi-eUiptical surface cracks (crack depth-to-plate thickness ratio of 0.2) Smith
and Sorensen[3] using the alternating method gave stress-intensity factors in considerable
disagreement (10-25%) with the present results. For semi-elliptical surface cracks the results
from Rice and Levy[13] for crack depth-to-plate thickness ratios less than or equal to 0.6 and
an approximate equation proposed by Newman[14] were in good agreement with the present
results.
The stress-intensity factors obtained herein should be useful in correlating fatigue crack
growth rates as well as fracture toughness calculations for the surface-crack configurations
considered.

REFERENCES
[I] W. M. Browning and F. W. Smith, An analysis for complex thrce-dimensional crack problems. Developments in
Theoretical and Applied Mechanics, Vol. 8, Proc. 8th SECTAM Conf. (1976).
[2] A. S. Kobayashi, Surface flaws in plates in bending. Proc. 12th Annual Meeting of the Soc. of Engng Sci., Austin,
Texas (1975).
13] F. W. Smith and D. R. Sorensen, Mixed mode stress intensity factors for semi-elliptic~ surface cracks. NASA
CR-134684 (1974).
[4] K. Kathircsan, Three-dimensionallinear elastic fracture mechanicsanalysis by a displacement hybrid finite element
model. Ph.D. Thesis, Georgia Institute of Technology(1976).
[5] A. E. Green and I. N. Sneddon, The distribution of stress in the neighborhoodof a flat elliptical crack in an elastic
solid. Proc. CambridgePhil. Soc. 46 (1959).
[6] D. M. Tracey, Finite element for three-dimensionalelastic crack analysis. NucL Engr. and Design 25 (1974).
[7] R. S. Barsoum,On the use of isoperametric finite elements in linear fracture mechanics. Int. J. Num. Meth. Engr. 10(I)
25-37 (1976),
[8] S. Atluri and K. Kathiresan, An assumed displacement hybrid finite element model for three.dimensional linear
fracture mechanicsanalysis. Proc. 12th Annual Meetingof the Soc. Engr. Science, Universityof Texas, Austin, Texas
(1975).
[9] I. S. Raju and J. C. Newman, Jr., Three-dimensional finite-element analysis of finite-thickness fracture specimens.
NASA TN D-8414 (1977).
li0] H. Tada, P. C. Paris and G. R. Irwin, The Stress Analysis of Cracks Handbook, p. 28.1. Del. Research Corp., C.
(1973).
[11] R. J. Hartranft and G. C. Sih, An approximate three-dimansionaltheory of plates with application to crack problems.
Int. J. Eng. ScL S(8), 711-729(1970).
[12] A. S. Kobayashi, Crack opening displacement in a surface flawed plate subjected to tension or plate bending.
Presented at the 2nd Int. Conf. Mech. Behavioro[ Materials, Boston, MA (1976).
[13] J. R. Rice and N. Levy, The part-through surface crack in an elastic plate. Trans. ASMEJ. AppL Mech. PaperNo.
71-APM-20(1970).
[14] J. C. Newman, Jr., Fracture analysis of surface- and through-crackedsheets and plates. Engng Fracture Mech. 4,
667--689(1973).

(Received 18 July 1978;receivedfor publication 4 September 1978)

APPENDIX
Determination of stress-intensityfactors using the nodal-force method
The nodal-forcemethod,developedin Reg.[9] for evaluatingstress-intensity factors is presented herein. In this method,
the nodal forces normal to the crack plane and ahead of the crack front are used to evaluate the stress-intensity factor. In
contrast to the crack-openingdisplacement method[6],this method requires no prior assumption of either plane stress or
plane strain.
Consider an elliptic surface crack idealized by several wedges as shown in Fig. 13. Each wedge was composed of
various number of elements. The nodal forces (F~)normal to the y' -- 0 plane along the junction (dashed curve) between
wedges i and i + 1 are assumed to be contributed from the normal stresses acting over one-half of the wedges on either
side of the junction (shaded area). For clarity, the shaded area has been enlarged (R 4 a). The dashed curves are
hyperbolas normal to the crack front. Nodal forces along these hyperbolas are needed to evaluate the stress-intensity
factors.
828 I, S RAJU and J C. NEWMAN, Jr

__ I

Fig. 13. Nodal forces on y'= 0 plane and at the junction between wedges i and ~+ 1

The normal stress cry at any point in the shaded area is assumed to be given by

~v = ~ + A; r112+ A2 t~12+
v ~.zlrr)
(3)
where r is the distance normal to the crack front. In general, K and A', are functions of the coordinate along the crack
front. However, in the present analysis K and A', were assumed to be constant over the shaded area. This assumption is
jusfiflndwhen the wedges are thin or the actual K variations across a wedge are small.
The total force acting in the y'-directionover the shaded area is given by

F, = t. crv dr (4)

where to is the average width of the shaded area (t~ = (t,+ t,+,)/2).SubsUtuUng eqn (3) into eqn (4) gives

K
Fv = ta( (-~--~)
2VR + A,R"2+ A2RSI2+. ) (5)

where A, represents constants. Using only the firstterm on the righthand side of eqn (5), K is replaced by the apparent
stress-intensityfactor K~ and is evaluated as

Fv = v[z~r;
~ x/(R)2to. (6)

1 . 2 .C

1.2(
~/128

1.1-c ~
7r/8

- -K a P I. i( n'/4 /

1.0.c

2a

I i I I
.16
,.

.04 .08 .12

R
a

Fig. 14. Apparent stress-intensity factors for semt-circular surface crack (a/t = 0.2) as a function of Rla
and the parametric angle, 4,
 Stress-intensity factors of semt-elliptical surface cracks in finite-thickness plates 829

The force Fy was obtained from the finite-element analysis and was determined as follows. For example, let nodes 1-4 in
Fig. 13 be the nodes lying on the junction between wedges t and i+ I. Node 4 is a distance R from the crack front. The
nodal forces F, (/= I-4) are obtained from the finite-element analysis. Thus, the total force acting over a distance R is

~=~+~+~+~ (7)
where F~, the nodal force at node 4, is calculated using the elements that are connected to node 4 and are reside the shaded
area. For various values of R, the number of nodal forces contributing to Fy would also vary.
The total force Fy obtained from the finite-element analysis is substituted into eqn (6), and the apparent stress-intensity
factor is evaluated. The procedure is repeated for various values of R. Equating eqns (5) and (6) gives

K~p = K + BIR + B2R 2 + (g)

where B, represents constants. From eqn (8), a plot of K o vs R is linear for small values of R and the intercept at R = 0, at
the crack front, gives the stress-intensity factor at a particular location along the crack front. The procedure is repeated at
various locations along the crack front.
Figure 14 shows the apparent stress-intensity factor for a typical semi-circular surface crack obtained by the
nodal-force method plotted against the distance R/a for various values of the parametric angle, 4'. If the Kop value nearest
the crack front (R = 0) is neglected, the relationship between K,~ and R/a are nearly linear, as expected. Therefore, the
Koo value closest to the crack front was neglected. For each value of 4,, a stratght line was fitted (by the method of least
squares) to the Kap results to evaluate the stress-intensity factor The intercept of these lines at Rla = 0 gives the
stress-intensity factor.