Engineering Froctwe Mechanics, 1970, Vol. 2, pp. I-17. Petpmon Press.

Printed inGreatBritain

ON THE FINITE ELEMENT METHOD IN LINEAR

FRACTURE MECHANICS-l-
S. K. CHAN, 1. S. TUBA and W. K. WILSON
Westinghouse Research Laboratories, Pittsburgh, Pa. 15235, U.S.A.

Ah&act-The usefulness of the finite element method for the computation of crack tip stress intensity
factors is established. Although ordinary finite element methods lack the ability to represent crack tip stress
singularity, meaningful values for crack tip stress intensity factors can be obtained by a simple process. The
results are compared not only to the results of other analytical solutions, but additional correlation is made of
two different fracture test specimen types.

NOTATION
measure of crack length. See Figs. 1,3,5 and 9.
B” thickness of fracture test specimens
strain tensor components
E,ec elastic and shear moduli
Al various functions of fI in crack tip equations
J integral defined by (6)
KI stress intensity factor for mode 1
KlC plane strain critical stress intensity factor for mode I
K: Fictitious stress intensity factor, evaluated on the basis of local results, near crack tip, by using crack
tip equations. Its limiting value at the crack tip should equal K,.
P applied load, and load at fracture
r,e,z polar coordinates; usually the origin is placed at a crack tip
&Y,Z orthogonal Cartesian coordinates
displacement components
h”,::
Y
Lamme constants of elasticity
Poissons ratio
P mass density of material
ULJ normal and shear stress components
w angular velocity at fracture.

Several quantities are defined in the figures, tables or the text.

1. INTRODUCTION
IN THE engineering application of the concepts of linear fracture mechanics to the
prediction of strength and life of cracked structures, a knowledge of the crack tip stress
intensity factor as a function of applied load and geometry of the structure is necessary.
This informatioq, combined with the experimentally determined critical stress intensity
factor and crack growth rates for the structural material make such predictions possible.
The concepts and results of linear fracture mechanics are already adopted by a great
number of engineers. The theoretical and experimental results are summarized in
several recent publications [ 1,2], therefore no attempt will be made here for the same.
It is only noted that from the practicing engineers point of view linear fracture mechan-
ics works as follows. For a given flaw (mostly cracks are assumed), and principal mode
of loading, a theoretical stress intensity factor is computed based on the laws and
principles of elasticity. For the material which is used, critical values of stress intensity
factors are determined by experiments, under the same principal mode of loading. The
tpresented at the National Symposium on Fracture Mechanics, Lehigh University, Bethlehem, Pa.,
June 17-19,1968.

EF.M.Vd. 2.No. 1-A

2 S. K. CHAN, 1. S. TUBA and W. K. WILSON

conditions, such as temperature and irradiation, should be the same during the experi-
ment as in the actual application. If the computed theoretical stress intensity factor is
less than the critical value by a safe margin, then the flaw is acceptable. Otherwise
some sort of corrective measure must be taken to avoid failure, excessive crack growth
or hazardous operation.
It has been established that linear fracture mechanics, as described briefly above,
works very well. Of course there are still many unresolved questions which must be
investigated. In this paper only a few of these problems will be considered.
A rigorous determination of the crack tip stress intensity factor requires an exact
solution of the elasticity problem formulated for the cracked structure. In most cases
exact solutions to the actual problem are very difficult or nearly impossible to obtain.
Existing theoretical stress intensity factors are the product of highly sophisticated
mathematical analysis for idealized model configurations and loading conditions.
Considerable number of these are available in tabulated form [ 1,2], which are, of course,
very valuable for the practicing engineers. These results can be applied to a wide
variety of problems, with more or less success, depending on how well the real geometry
and loading condition agree with those used in the mathematical model.
Since in many real situations it is not possible to find suitable model representation
for which an exact solution is available, the need for a relatively straight forward
numerical method is apparent, in order to estimate stress intensity factors.
One such method will now be presented for plane structures. The extension of this
method to other types of cracked structures will also be discussed.
Faith must be gained for the validity of results by approximate methods. The
approach must be applied at first to configurations where exact or well established
solutions exist. Suitable procedure has to be developed for the computation of the
stress intensity factor and accuracy of the solution evaluated. When satisfactory results
are obtained then the method can be applied to problems where exact solutions are
impossible to obtain. Although the solution will be approximate, the magnitude of error
should be estimated.
The finite element method is suggested as the best candidate at the present time for
obtaining approximate stress intensity factors, whenever exact solutions are not
available.
Finite element methods are rapidly adopted in structural analysis. The reason is
that the method is conceptually simple, easily adoptable to high speed electronic
computations, applicable to large classes of geometries, materials and loading condi-
tions, and can be made quite accurate.
The technical details of the finite element method are described in the literature [3]
and will not be repeated in this paper.
The method given here, for the determination of crack tip stress intensity factors,
of course, contains all of the inherent advantages of the finite element stiffness method.
Various forms of mixed boundary conditions can be analyzed. Structures formed by
dissimilar metals can be handled. Body forces, such as those in rotating apparatus and
structures subjected to shock can be analyzed. In addition to thermal stresses, the
effects of plastic flow in highly stressed regions away from the vicinity of the crack tip
can be included in the analysis. Reinforced plane sheets, which are subject to plane
deformations only, can also be handled, provided that the crack tip is not in the rein-
forcement.
While the method as presented here is oriented toward plane structures, the
 Linear fracture mechanics 3

extension of the approach to other classes of structures is obvious. For example the
stress intensity factors for circumferential cracks in axisymmetrically stressed solids
of revolution can be obtained by using appropriate axisymmetric ring elements.
Extensions to mode III conditions of longitudinally cracked prismatic bars in torsion
[4], circumferentially cracked circular shafts of variable diameter subject to torsion,
and to antiplane strain problems is also possible.

2. FINITE ELEMENT PROGRAM

The program used in this study is based on the displacement method. First order
displacement functions are assumed, that is the displacements vary linearly over the
element which results in constant strains and stresses on the element. The program is
general purpose; plane-stress and plane-strain cases can be handled. Displacements
and stresses can be determined in arbitrary plane shapes by replacing the actual
geometry with an assemblage of triangular elements. The program accepts a variety
of boundary conditions and loadings, such as, prescribed displacements, boundary
tractions, body forces and temperature variations. The program is written in the Fortran
IV language and is operated on a CDC6600 Computer.

3. FINITE ELEMENT GENERATORS

Since the finite element program is general purpose, for particular geometries
additional programs are written to generate the necessary input information.
Element generators were created for four basic shapes. The configuration for the
study of mesh size effects and for comparisons with results by the collocation method
is shown on Fig. 1. Three different geometries have been used, as indicated in the
figure, by varying the crack length. The corresponding element generator is such that
all dimensions can be varied, and the element sizes away from the crack tip can be

I--“1
P

i-aY
r--l
---
1
2H

Fig. 1. Configuration for the study

of mesh size effects and for com-
parisons with results by the
collocation method.
4 S. K. CHAN, I. S. TUBA and W. K. WILSON

Fig. 2. A finite element representation of the simplified compact tension specimen. (Some
nonuniformities are present because this representation is obtained by modifications of the
mesh shown on Fig. 4.)

made larger or smaller. A corresponding typical mesh layout is shown in Fig. 2. In the
crack tip area the possibility of independent refinement in mesh size is incorporated.
On Fig. 2, around the crack tip two families of geometrically similar element groups
are superimposed, one on top of the other, but smaller in size. Any number of element
size reductions following this pattern can be carried out automatically. All necessary
numbering of the elements and of the nodal points are carried out also automatically.
Results of production plotting routines indicate all of these numbers, but they are
omitted from Fig. 2, for the sake of clarity. For the same reason only one element
size reduction is shown. In the actual computations the ratio of the smallest area used
to the smallest area shown was &. The smallest elements used are difficult to see with
the naked eye.
The second basic shape corresponds to the compact tension fracture test specimen,
as is shown on Fig. 3. A corresponding typical element outline is shown on Fig. 4.
The automatic mesh size reduction feature is included not only in these element
generators but in all others with cracks.
The third basic shape is the case of a crack of length 2a in an infinitely large flat
plate. No separate figures are included here for this problem, because the insert in Fig.
9 is descriptive enough.
The final basic shape is shown on Fig. 5, for the rotating fracture test specimens and
a finite element setup is given on Fig. 6.

4. COMPUTATIONAL METHODS FOR STRESS INTENSITY FACTORS

Once the numerical solution has been obtained for a particular finite element
representation, crack tip stress intensity factors can be estimated by the use of estab-
lished crack tip relations. There are many possible methods which can be used. The
 Linear fracture mechanics

- 2H
---.

Fig. 3. Compact tension specimens used in

analysis. Dimensional ratios are taken from actual
test specimens.

Fig. 4. A finite element representation of the compact tension specimen.

three specific methods considered here are: (1) displacement method, (2) stress method,
and (3) line integral (energy) method. The major emphasis has been placed on the
displacement method due to its relative simplicity and ease of interpretation. All three
methods are conceptually straight forward and will be described below.
The displacement method involves a correlation of the finite element nodal point
displacements with the well known crack tip displacement equations. The plane strain
displacement equations are:
llf = $$ [ r/27r]1’2h(6,v) (1)
6 s. K. CHAN, I. S. TUBA and W. K. WILSON
where

fi(O, v) = cos (O/2) [ 1 - 2 v + sin2 (t?/2)]

f,(& V) = sin (192) 12 - 2 v - co? (O/2)]

and the crack tip stress intensity KI is a function of the geometry of the body containing
the crack and of the applied loading conditions. A corresponding set of equations
can be written for plane stress conditions. By substituting a nodal point displacement
Us?at some point (r,O) near the crack tip into (1) a quantity KT can be calculated.

K,* = [27r/rl”2 Gu;/[fr(&v)]. (2)

From plots of K,* as a function of r for fixed values of 8 and a particular displacement
component, estimates of KI can be made. If the substituted displacements were the
exact theoretical values then the value of the K: obtained as r approaches zero would
be the exact value of K,. Since the finite element displacements are rather inaccurate
at an infinitesimal distance from the crack tip, this limiting process is not useable.
Instead a tangent extrapolation of the K,* curve is used to estimate KI. With suitable
refinement of element size the KF curves obtained from element displacements

Geometry +o 5
K,
7 o-2244 0.1027
8 0.3037 0 -1027
9 0.3740 0.1027

Fig. 5. Rotating fracture test specimens.

 Linear fracture mechanics

Fig. 6. A finite element representation of the rotating test

specimen.

rapidly approach a constant slope with increasing distance (r) from the crack ‘tip.
The intercept of the tangent to the constant slope portion of the curve with the KT
axis is used as the KI estimate. As will be shown when specific examples are considered
this method gives very good estimates of K, and as the exact displacements are ap-
proached by element displacements the exact value of KI is approached. The most
accurate estimates are obtained from the KT curve corresponding to the u displacement
on the crack surface (v,).
K* = mY2~ tr,.
I (3)
4( 1 - v’) r1’2

The stress method for determining crack tip stress intensity factors is similar to
the displacement method. Now the nodal point stresses are correlated with the well
known crack tip stress equations. For plane strain conditions the equations are

Uij=-fij(6)
KI
(27r r)“2
where
f,,(O) = cos (O/2) [ 1 - sin (O/2) sin (38/2)]
f,,(e) = cos (e/2) [ 1 + sin (e/2) sin (3e/2)1
f,,(e) =fUz = sin (e/2) cos (e/2) cos (3812).

Nodal point stresses u;“Jin the vicinity of the crack tip can be substituted into (4) and
values of K; can be calculated

(3
8 S. K. CHAN, I. S. TUBA and W. K. WILSON

From plots of K: as a function of r for a fixed 8 and a particular stress component,

estimates of K, can be made. If the exact theoretical stresses were substituted into (5)
then the intercept of the curve with the KJ axis at r = 0 would be the exact value of K,.
Due to the inability of the finite element method to represent the stress singularity
conditions at the crack tip, the KF curve for I greater than zero must again be extra-
polated back to r = 0. The extrapolated value of K: at r = 0 is the estimated K,. Good
KI estimates by the stress method are obtained from the KT curve corresponding to
the ‘T, stress on the 6 = 0 plane.
Rice has shown [5] that the value of the line integral

J= Wdy-T. 2d.r
>

where I is an arbitrary contour surrounding the crack tip as shown in Fig. 7, is propor-
tional to the square of the crack tip stress intensity factor. For plane strain conditions
the following relation is given by Rice:

(7)

In (6) W is defined as the strain energy density, T is the traction vector defined accord-
ing to the outward normal along I, Ti = ofjnj, and u is the displacement vector. The
line integral is evaluated in a counterclockwise sense starting from the lower crack

Fig. 7. Crack tip coordinates and typical

contour r.

surface and continuing along the path I to the upper flat surface and ds is an element
of arc length along I. Rice’s path independent integral is applicable to linear and non-
linear elastic materials. For linear elastic conditions under consideration

W(e) =+h(~?ii)~+p?ijeij. (8)

By numerically evaluating the integral of (6) for the finite element solution over a path
surrounding the crack tip, an estimate of the crack tip stress intensity factor can be
made by use of (7).
In the methods just presented and in the specific applications to follow, only mode
I crack tip loading conditions are considered. For the plane problems under considera-
tion the displacement and the stress methods as described can be extended to obtain
 Linear fracture mechanics 9
mode II stress intensity factors, KI,, or combinations of K, and KI,. For a mixed
mode condition K, and KI, estimates are made in the manner described above. To
uncouple the mixed mode conditions, K, estimates are made from the KT curve
constructed from the u displacement on the crack surface (8 = V) by the displacement
method, or from uy on the 8 = 0 plane by the stress method. Similarly the KI, compon-
ent can most effectively be obtained from Ki curves constructed from the u dis-
placement on the crack surface by the displacement method:

(2?7)“% U,
KZ=4(*-yo) @)1/2

or from r,, on the 8 = 0 plane by the stress method:

K; = (27~)“~~~~.

The value of the path independent line integral J (6) is related to the sum of the
squares of the stress intensity components:

5. INFLUENCE OF ELEMENT SIZE

The influence of element size on estimated K, appears to be approximately the
same for all three methods considered. Since the displacement method is the simplest
to work with and interpret it will be the method considered in detail here.
The effect of relative element size on the KT curve as calculated by the displace-
ment method is shown in Fig. 8. Here KT calculated from the u displacement on the
crack surface is shown as a function of r for a number of different mesh reductions for a
specific proportion (a/W = O-5 of the simplified CT geometry (Fig. 1). The finite ele-
ment curves are compared with the KT curve calculated from displacements obtained
by a boundary collocation solution for the same geometry and loading conditions. The
Williams’ stress function[6-81 was used in the collocation process. The collocation
curve shown in Fig. 8 is considered to be accurate within O-50 per cent and is used as
the basis for convergence considerations of the finite element method. Two independent
phases of element size reduction are considered. The first phase is concerned with the
size of elements very close to the crack tip. The size of elements in this area will be
referred to as the inner element size. The second phase of reduction is concerned with
the elements away from the crack tip and near the outer bounds of the geometry. Over
this area the term outer element size will be used. Element size will be considered in
terms of element area. The inner and outer element sizes for the cases shown in Fig. 8
are listed in Table 1. As indicated, cases l-4 have the same outer element sizes but
significantly different reductions in crack tip element sizes. The relative size of ele-
ments for case 1 is shown in Fig. 2. The inner element sizes are decreased in going
from case 1 to case 4 by a factor of 256. Cases 4 and 5 have the same inner element
size but the outer elements for case 5 are approximately one fourth of those of case 4.
From a study of Fig. 8 the following observations can be made. All of the finite
element curves approach a constant slope as r/W increases. The higher the degree of
inner element reduction, the more rapid the curve approaches a constant slope. For
 S. K. CHAN, I. S. TUBA and W. K. WILSON

I I

I I
0 045 0.1 0.15

Fig. 8. Effect of mesh size on KT for geometry 2.

Table 1. Element sizes for cases

shown in Fig. 8

Element size?-A/a*

Case Inner ( 1O-6) Outer ( 10m2)

1 312 2
2 78 2
3 20 2
4 1.20 2
5 1.20 1

tA - Element area, a -Crack-

length.

fixed outer element size, the sequential reduction in inner element size results in a
convergence of the curves to a curve which is a straight line over a major part of its
length with a sudden drop near rl W = 0. For fixed inner element size a reduction in the
outer element size moves the constant slope portion of the curve closer to and more
parallel with the collocation curve. It can be concluded, that the finer the outer element
size is, the closer the actual curve is to the exact near but not at the crack tip. By
sequential reduction of the inner element size the shape of the curve near the tip will
deviate less and less from the curve of the exact solution which is a straight line.
The best estimate of the stress intensity factor which can be obtained from any
finite element K,* curve is obtained by extrapolating the straight portion of the curve
 Linear fracture mechanics II

back to the vertical axis. The result obtained by the extrapolation of curve 5 is 4 per
cent lower than the collocation value. On Fig. 8 since only part of the scale is used, the
deviations are greatly exaggerated.
The curves of Fig. 8 apply to a case in which the untracked section of the crack
plane is subjected to a very high degree of bending. It is also beneficial to consider the
case of a crack in a uniform stress field which represents the opposite extreme. In
Fig. 9 the theoretical curve K,* obtained from Westergaards[9] exact solution is

.8-
.

I.4 -

I I I I I I 1 I I I I I I
0 O-1 0.2 0.3 0.4 0.5 0.6
L
a
Fig. 9. Variation of KF with r/a for a finite crack in an infinite medium.

compared to a finite element curve for a refined element representation. The com-
parison is essentially similar to that of Fig. 8, except that the slopes of the two curves
are negative. The estimated stress intensity factor obtained from a linear back extra-
polation of the straight portion of the finite element curve results in a value of 5.5
per cent lower than the exact value.
A plot of KT values obtained from the stress method for the same geometry and
loading condition is shown on Fig. 10. The element refinement is the same as that for
case 5 in Table 1. The points indicate values calculated from nodal point stresses
ry on the 8 = 0 plane. Whereas a smooth curve could be drawn through the KT-
displacement calculated points as shown in Fig. 8, such is not the case for the nodal
stress calculated values. There is much more scatter with these points than with the
corresponding displacement points. The courser the element size, the greater the scatter
that is observed for the stress calculated KT values. Therefore, the stress method KI
is estimated by fitting the points for small values of r/W with a straight line and taking
the vertical axis intercept as KI. Such an operation applied to the points of Fig. 10
12 S. K. CHAN, I. S. TUBA and W. K. WILSON

I I

8
P
< 0.
m
.-
Y
6

.
4

: i
0

Fig. 10. Kf plot for stesssmethod(case4).

results in an estimate which is 4.5percent lower than the collocation value. For courser
element sizes the increase in scatter reduces confidence in this method. The reason
for the scatter is of course that in the method constant stresses are computed for
individual elements, and nodal point stresses are obtained by an averaging process.
The line integral method was also applied to this configuration. The contour
integration path was taken over the outer boundary of the geometry and the integral
was evaluated numerically. The strain energy densities were calculated from nodal
point stresses and the nodal point forces were used as the surface tractions. The stress
intensity factor estimated by this method for element size reduction of case 5 fromTable
1 is 3.5 per cent less than the collocation value. The advantage of this method is that
no extrapolations are required, only one number is calculated. The disadvantage
is that for a single element representation it is difficult to estimate the degree of error
in the estimated value, whereas for the other two methods a study of the KI* plot
gives some indication of the degree of accuracy.
As these specific cases demonstrate, estimates of K, values within engineering
accuracy can be made by this approach. The higher the degree of accuracy required,
the larger the number of elements needed. The primary limitation on the number of
elements used is the available computer storage capacity. The stability of the total
stiffness matrix appears to be relatively independent of the number of elements used
or of the variation in element size over the structure.

6. RESULTS FOR COMPACT TENSION SPECIMENS

Compact tension fracture test specimens are the latest in an evolution series, in
which they were preceded by the Manjoine and the WOL (wedge opening load)
variety. Since the development of these specimens is well documented[lO, 111,
discussion here will be limited to the comparison of some recent experimental data to
corresponding solutions by the finite element method. Greenberg, Wessel, and Pryle
[121 have tested a Cr-MO-V turbine rotor steel. The actual dimensional ratios from
three of their test specimens are shown on Fig. 3. These same dimensions were used
also in the analysis. A finite element representation is shown on Fig. 4. The ratio of
 Linear fracture mechanics 13

the smallest element used to the smallest element shown is &. The results from the
finite element method for the stress intensity factors are compared to collocation
results on Fig. 11. Actual dimensions and fracture loads were used in the calculations.
During the test procedure, the change (2~3 in a gage length (215~ on Fig. 3) is plotted
as a function of the applied load. The actual values at fracture of the ratio vu/P, often
referred to as compliance, are compared to the corresponding computed values in
Table 2. The agreement is gratifying.

0 Finite Element
P-

18 -

16 -

14 -

rz 12-

Y_I
*n
10 -

8-

81 . .
a
W

Fig. 11. Comparison of results of collocation and finite

element solutions.

7. RESULTS FOR ROTATING TEST SPECIMENS

Rotating tests are also used to evaluate fracture properties of materials. Specimen
preparation, test procedures and results can be found in [ 13]-[ 161 and they will not be
reviewed here. The geometry considered is shown on Fig. 5. Once the test is run,
the present practice is to evaluate KIc from the following expression

Krc= *,G (12)

14 S. K. CHAN, I. S. TUBA and W. K. WILSON

where
3-2~
go = 8(1 Pw2JGlZ (13)

which is the stress at the center of a solid rotor of radius &, and the other quantities
are defined in the nomenclature.
A triangular finite element representation of the rotating test specimen is shown on
Fig. 6. In the actual computation at the crack tip much smaller elements were used
than those shown on the figure. The ratio of the area of the smallest element used to
the smallest element shown was &.
For the prediction of KI the displacement method described in part 4 was used. For
geometry 8 the variation of K.7 with r/R0 is shown on Fig. 12. Of course, from the

Table 2. Compact tension specimen results

Experimentalt Finite element

c Crack Stress
5E Fracture opening Compliance Compliance intensity
8 IoadP 2vr h/P VYIP KlC
0 (lb) (10e2in.) (lO+in./lb) (lo-“inJIb) (ksi G)

5 5820 I.27 IO-9 10‘04 41.3

6 5630 1.07 9,s 9.79 39. I
7 5940 1.16 9.8 9.69 40.1

tRef. [12].

Fig. 12. Variation of KT with r/R, for geometry 8.

 Linear fracture mechanics I5

nature of the approximations inherent to the method, it is known that the exact value of
the stress intensity factor is 4-5 per cent higher than those predicted here.
The resulting factor for various crack lengths are given in Fig. 13. On the same
figure two different comp~sons are made. The empirical formula, (12), used by
Sankey[l6] crosses over the line predicted by the finite element solution. Another
approximate method has been investigated recently by Williams and Isherwood[l7]
for plane stress problems with finite geometry. Modifying their method to plane strain
conditions, the corresponding results are also shown.

N
3i
0
u- ,” Empirical Formula
3
Q Defined by Eq. 12

:I , G~metries7,8and9 i

o-3 o-4
a
RO

Fig. 13. Stress intensity factors for various a/R,.

The rotating test specimens were machined from the same forging as the compact
tension specimens discussed in the previous secticn. It is possible, therefore, to com-
pare stress intensity factors for the @-No-V steel as obtained by the two different
types of test specimens. The comparison is made in Table 3. It is interesting to note that
regardless of the computational method used for obtaining the critical stress intensity
factors, the values are consistently higher for the rotating test specimen, than for the
compact tension specimens. If K,, is truly a material property, then the critical stress
intensity factors should be the same. The values for compact tension specimen are
approximately 15 per cent lower than those for the rotating case. For engineering pur-
poses, this is satisfactory. The comparison of this type must be carried out on large
number of test results before one should seek reasons for discrepancies either in the
basic theory or in the experimental process. That, of course, is beyond the objectives
of this study.

8. POTENTIAL ~PRO~~N~ AND EX~NSIONS

The specific approach presented here is considered to be only the initial step in the
use of the finite element methods to obtain crack tip stress intensity factors. This
general approach can be refined in a number of ways. Higher order elements such as
linear strain elements can be used in place of the constant strain elements considered
16 S. K. CHAN, I. S. TUBA and W. K. WILSON

Table 3. Comparison of test results for a Cr-MO-V turbine

rotor steel

Fracture specimen type

Compact tension Rotating

(at O”F) (at - 11°F)

Geometry 4 5 6 8

Load at
fracture 5820 lbt 5630 lbt 5940 lbt 10,800 rev/mint
Finite
2 element 41.3 39.1 40.1 41.0
$ Collocation 42.6 40.35 41.4
2 Equation ( 12) 45.4
x Williams-
g Isherwood
‘- i method 47.8

tRef. [12].
*Ref. [16].

here. Also combinations of elements of various orders with the higher order elements
closer to the crack tip is possible. The use of a special crack tip element which includes
the crack tip singularity condition as its assumed displacement pattern is also very
promising. The potential of these variations is now under consideration by the writers.

9. CONCLUSIONS
It has been demonstrated in this paper that the finite element method, supplemented
with special computational procedures, can be used to find crack tip stress intensity
factors in various shapes under different types of loading conditions. The accuracy
of the prediction is satisfactory without the use of excessive computer time, and of
course, can be improved, within the limitations of the computer. Although the method is
most signiticant for geometries and loading conditions where no exact solutions exist,
it seems desirable to analyze both the structures containing the cracks, and the test
specimens used to evaluate the material properties by the same method. The inherent
approximations this way would become compensated for automatically.

Acknowledgements-The writers are greatly indebted to Mr. S. E. Gabrielse and Dr. C. Visser for the
development of the basic computer program, to Mrs. V. Ho and Mr. D. Wei for programming the element
generators, and to Mr. G. 0. Sankey and Mr. E. T. Wessel for freely providing the results of their experi-
mental work. All named persons are with the Westinghouse Research and Development Center.

REFERENCES
[1] C. F. Tiffany and J. N. Masters, Fracture toughness testing and its applications. ASTM Spec. Tech.
Publ. No. 38 1, pp. 249-277 (1965).
[2] E. T. Wessel, W. G. Clark and W. K. Wilson, Engineering methods for the design and selection of
materials against fracture. U.S. Army Tank and Automotive Center Rep., Contract No. DA-30-069-
AMC602(T) (1%6).
[3] 0. C. Zienkiewicz and Y. K. Cheung, The Finite Element Method in Structural and Continuum
Mechanics. McGraw-Hill, New York (1%7).
[4] G. Sih, Strength of stress singularities at crack tips for flexural and torsional problems. Trans. ASME J.
appl. Mech. (1%3).
[S] J. R. Rice, A path independent integral and the approximate analysis of strain concentration by notches
and cracks. J. uppl. Mech. 35,379 (I 968).
 Linear fracture mechanics 17

t61M. L. Williams, On the stress distribution at the base of a stationarv crack. J. aoD1.
. . Mech. 24. 109-l 14
(1957).
171 W. K. Wilson, Analytic determination of stress intensity factors for the manjoine brittle fracture test
specimen. AEC Res. Devel. Rep. No. WERL-0029-3 (1965).
[81 B. Gross, E. Roberts, Jr. and J. E. Srawley, Elastic displacement for various edge-cracked plate speci-
mens. NASA Tech. Note No. D-4232 (1967).
191 H. M. Westergaard, Bearing pressures and cracks. Trans. ASME .I. appf. Mech. (1939).
[lOI M. J. Manjoine, Biaxial brittle fracture tests. Trans. ASME 293-298 (1965).
r111 E. T. Wessel, State of the art of the WOL specimen for K,,- .- fracture toughness testing. Enme Frac-
fure Mech. 1,77-103 (1968).
r121 H. D. Greenberg, E. T. Wessel and W. H. Pryle, Fracture toughness of turbine generator rotor forgings.
Presented at 2nd National Symp. Fracture Mech., Lehigh University (1968).
1131 D. H. Winne and B. N. Wundt, Application of the Griffith-Irwin theory of crack propagation to the
burstinn behavior of disks. includina analvtical and exnerimental studies. Trans. ASME 80 (1958).
1141 G. 0. Sankey, Spin fracture testsof NiMoV rotor sheets in the brittle fracture range. Pioc. Am. Sot.
Test. Mater. 60.721-732 (1960).
USI G. D. Cooper, G. 0. Sankey and E. T. Wessel, Discussion of paper results of bursting tests of alloy
steel disks and their application to design against brittle fracture. Symp. Large Rotor Forgings, ASTM
A. Mtg, Purdue University (1965).
tl61 G. 0. Sankey, Spin tests to determine brittle fracture under plane strain. Presented at SESA Spring
Mtg, Albany (1968).
[171 J. G. Williams and D. P. Isherwood, Calculation of the strain-energy release rates of cracked plates by
an approximate method. J. Strain Analysis 3, 17-22 (1968).

(Received 29 May 1968)

R&xIIu~-O~ Ctablit I’utilite de la methode de l’element fini pour les calculs des facteurs d’intensid de la
tension de rupture en bout. Bien qu’habituellement les m&odes d’elements finis ne peuvent pas representer
la singularit de la force de rupture en bout, il est possible d’obtenir par un pro&C simple des valeurs signi-
ficatives de ces facteurs d’intensite. Les resultats sont compares B ceux des autres solutions analytiques, et
une correlation supplCmentaire est effect&e avec types specimens differents d’essai de rupture.

Zusammenfassang-Die Zweckmhsigkeit der Methode der endlichen Elemente fur die Errechnung von
Spanmmgsfaktoren an Rissspitzen wird dargelegt. Obzwar die gewohnlichen Methoden der endlichen Ele-
mente nicht imstande sind die Singularitiit der Rissspitzenspannung vorzustellen, kiinnen mittels eines
einfachen Verfahrens doch sinnvolle Werte fur Spannungsfaktoren an den Rissspitzen erhalten werden.
Die Ergebnisse werden nicht nur mit den Ergebnissen anderer analytischer Losungen verglichen, sondem es
wird such eine zusatzliche Korrelation von zwei verschiedenen Typen von Bruchprobekorpem vorgenommen.