Determining Fracture Parameters with Full.

field
Optical Methods
by Robert J. Sanford

ABSTRACT--In order to increase the accuracy of stressintensity-factor measurements and to obtain data on additional
parameters which may influence fracture behavior (such as
crack branching and crack curvature) a technique for fullfield fringe-pattern analysis (referred to as the local collocation
method) has been developed. This method removes the restriction of limiting the data analysis region to the near-field region
by including additional nonsingular terms in the algorithm. In
this paper the theory of the method is developed and sample
results using photoelastic, holographic and moir6 full-field
patterns are provided.

Introduction
Over the last three decades full-field optical stressanalysis methods have played an important role in experimental fracture mechanics. These methods are particularly
well suited to determining the geometric stress-intensity
factor for particular geometry/loading configurations in
situations where analytical or numerical methods fail to
provide answers. Alternatively, techniques of experimental
stress analysis can be used to verify solutions obtained by
other methods. In the area of dynamic fracture mechanics
optical techniques (particularly photoelasticity) have been
the primary tools used to study crack-arrest behavior. In
fact, one of the earliest applications of photoelasticity to
the study of fracture behavior was the recording of the
isochromatic patterns for a running crack by Wells and
Post.
The Wells and Post paper focuses on the qualitative
aspects of the fringe pattern around the crack tip. More
significantly, it demonstrates the utility of the CranzSchardin camera system s in dynamic fracture applications.
This approach has been used extensively by other
researchers in the field. 3 Irwin,' in a discussion of the
Wells and Post paper, added the quantitative analysis of
the fringe pattern needed to determine the stress-intensity
factor, He observed that the isochromatic fringes formed

Robert J. Sanford (SEM Fellow) is Professor, Mechanical Engineering

Department, University of Maryland, College Park, MD 20742.
Paper was presented at the 1988 SEM Spring Conference on Experimental
Mechanics held in Portland, OR on June 5-10.
Original manuscript submitted: July 26, 1988. Final manuscript received."
January 20, 1989.

closed loops in the near field of the crack tip and used
this observation to develop the equations necessary to
compute the stress-intensity factor from the apogee point
on a fringe (see Fig. 1). Irwin also observed that the tilt of
these fringe loops was the consequence of the constant
stress, Oox, in a direction parallel to the crack line and
demonstrated that this term must be included in the nearfield formulation of the stress equations if a correct interpretation of the isochromatic fringe distribution is to be
obtained. Analytical justifications for this additional
term in the near-field equations of fracture mechanics
were provided subsequently by various researchers.~-8
The Irwin method, also called the apogee method, for
stress-intensity-factor extraction from photoelastic patterns
was the accepted method for analysis for many years.
Kobayashi and Bradley9 modified the formulation, but,
fundamentally, the approach was unaltered. The primary
application of this approach has been dynamic fracture
behavior of transparent, birefringent materials. References
10 and 11 are typical of the results obtained with this
approach.
Stress fields in the neighborhood of cracks in threedimensional bodies have been studied extensively by Smith
et aL using the stress-freezing method (e.g., Refs. 12, 13
and 14). The approach used to extract the stress-intensity
factor in these studies differs from the apogee method
and, instead, uses fringe-order information taken along a
line perpendicular to the crack tip. By restricting data
acquisition to a radial line, the angular dependence of
the near-field equations is eliminated and the stress field
can be approximated by a function of position only. A
similar approach was used by Smith et al. ~ to determine
the stress-intensity factor from displacement information
obtained with moir~ interferometry.
Sanford and Dally ~6 developed yet another procedure
for determining stress-intensity-factor information from
photoelastic patterns. In their approach, points on fringes
in the region around the crack tip were selected at random
and the set of points matched in the least squares sense
to an expression for the maximum shearing stress based
on the modified near-field equations. This procedure,
called the over-deterministic, least-squares method, was
used to determine K,, K , and Oox. An example of the
application of this method to failure analysis of an aircraft component is presented in Ref. 17.

Experimental Mechanics 9 241

Nearly every full-field optical stress analysis method has

at one time or another been used to study stress fields in
the neighborhood of a crack tip. In addition to those
already cited, the work of Dudderar and Gorman 18 should
be mentioned. They examined the isopachic field near a
crack tip using holographic interferometry. Using analysis
procedures similar to those of Smith et al., they examined
the fringe order along the crack plane and obtained an
estimate of the stress-intensity factor from the inversesquare-root dependence of the isopachic fringe order
along the line.
All of the approaches described above share a common
feature, namely that the analysis procedure used to extract
the stress-intensity factor from the full-field fringe pattern
relied on the modified near-field equations of fracture
mechanics. In principle, the accuracy of these methods
should improve as the region of data acquisition is reduced
to smaller regions around the crack tip. However, it is
generally recognized and has been independently verified
by several researchers 19'=~that the plane-stress assumption
ceases to be valid in a small region at the crack tip. In this
region the stress field is three dimensional and experimental
observations are influenced by stress gradients through
the thickness. In addition, the stress field is altered by
localized crack-tip blunting. On the other hand, if this
region is excluded from the analysis, the region of data
acquisition may not lie fully within the domain in which
the inverse-square-root-r singularity dominates the fringe
pattern. It has been demonstrated that, in specimens of
finite dimensions, the singularity dominated region is
very small. 2' To overcome these problems the author and
various collaborators 22-2' have developed procedures for
analyzing fringe patterns from a variety of optical methods
in which data acquisition is not restricted to the nearfield region.
These procedures, which are an extension of the overdeterministic, least-squares method of Sanford and Dally,
retain higher order terms as variables in the analytical
expression of the field description for the type of fringe
pattern being considered. In this procedure the influence
of the singularity is separated from the higher order terms
as a consequence of the analysis. Furthermore, by purposely eliminating the very near-field, three-dimensional
region from the analysis, the problem can be fully defined

(/1 arm/aO=0

~*~ "Cm=constant

in terms of a truncated series representation of the field

equations involving a limited number of terms. Because of
the mathematical similarities between this approach and
the purely numerical method of boundary collocation, the
method is called the local collocation method. In the
following sections of this paper the mathematical formulation of the local collocation method for several popular
full-field optical fringe patterns are developed and an
example of each is presented.

Formulation of the General Equations

Methods for determining the stress-intensity factor
from full-field optical patterns which are based on the
modified near-field equations for the stresses or displacements in the neighborhood of the crack tip fall to take
advantage of the additional information contained in the
full-field pattern. This information can he used to increase
the accuracy of the analysis if the influence of the nonsingular terms in the general analytical solution for the
problem is included in the feature extraction algorithm.
In order to use this additional information, the governing
equation which describes the optical pattern is represented
by a truncated form of the general-series solution for the
problem with unknown coefficients. These coefficients,
which inchide the parameters of interest (K, Oox, etc.), are
to be determined from the analysis of the fringe pattern.
To implement this method, a set of equations is written
which relate the measured fringe orders at a number of
points in an expanded region around the crack tip to the
underlying general expression involving the unknown
coefficients. These equations are then solved in the leastsquares sense to obtain estimates of the coefficients. The
number of terms retained in the series depends on the size
of the region. Convergence of the truncated series as the
number of retained terms is increased can be used as a
solution criterion. This approach is analogous to the
boundary collocation method except that the number of
terms necessary to obtain an accurate estimate of the
stress-intensity factor is dramatically reduced since the
stress state in the region around the crack is dominated by
the lower order coefficients.
Although the general series solution for the problem
being considered can be formulated in a variety of ways,
the author has found that the generalized Westergaard
equations in complex coordinates provide an algebraically
convenient formulation. For purposes of illustration of
the local collocation method this formulation is used as
the basis for the development of the governing equations
in the following sections.
It has previously been shown that the stress state for
opening-mode crack problems can be completely described
in Westergaard notation by an Airy stress function of
the form 8
F = Re Z ( z ) + y I m 2 - ( z ) + y I m ~ ( z )
(1)

Or.

where the notations

Z-(z) =

Z(z),

Z(z)

= ~d -Z-( z ) ,

Z'(z)

dz(z)
(2)

and

Fig. I--Illustration of a typical fringe

loop near a crack tip showing t h e
coordinates of the apogee point
used to determine K from Irwin's
method

242 9

September 1989

~(z) = -~td ~(z),

dY(z) = --~tY(z),

Y ' ( z ) = - d dt Y(z)
(3)

have been employed.

The functions Z ( z ) and Y ( z ) are suitable Westergaard-type stress functions for the geometry/loading
being considered and are subject to the constraints that
R e Z ( z ) = 0 on the stress-free portions of the crack
faces and I m Y ( Z ) = 0 along the crack line. From eq (1),
the stresses and plane-stress displacements, u and v, can
be written as
ax = R e Z -

y l m Z ' - y I m Y ' -e 2 R e Y '

(4)

ay = R e Z +

ylmZ'

(5)

+ ylm

Y'

collocation method for this type of fringe pattern is

particularly simple and results in a set of linear equations
for the fracture parameters. The solution scheme for both
classical two beam interferometers and holographic
interferometry is identical. In either case, the fringes are
contour lines of constant sum of the in-plane normal
stresses, i.e., isopachics. A typical isopachic fringe pattern
in the region of a crack tip in Plexiglas recorded with
double-exposure holographic interferometry is shown in
Fig. 2. From eqs (4) and (5), the governing optical equation is, ideally, of the form

(6)

r~, = - y R e Z ' - y R e Y ' - I m Y

2t

= a=+ae = R e Z + R e
2

(11)

Eu = (l-v)ReZ-(l+v)ylmZ

-(l+v)ylm

Y+ 2Re

(7)

Ev =2ImZ--(l+v)yReZ

-(l+v)yReY+(l-v)Im

(8)

where E is the elastic modulus and v is Poisson's ratio.

For a single-ended, stress-free crack with the origin of
the coordinates at the crack tip the functions Z ( z ) and
Y ( z ) can be represented as
Z(Z) =

~ A , Z "-1/2
n=0

(9)

and
Y(z) =

B. z ~

where N is the fringe order, fp is an optical constant and t

is the model thickness. Equation (11) assumes that the
wavefronts for each of the two beams in the interferometer
are in perfect phase over the field o f view. In practice this
condition may be difficult to achieve and additional terms
are necessary to account for any initial pattern which may
be present. Since the success of the local collocation
method depends on matching the observed fringe pattern
to an analytical expression containing unknown coefficients
which describes the fringe pattern, any additional factors
which influence the fringe pattern must be included in the
formulation. In the simplest case, a uniform phase
difference between the reference and object wave fronts
may be present which biases all of the fringes by a constant fractional fringe order. Accordingly, eq (11) must be
modified to account for this unknown fringe shift and the
governing optical equation should be written as
Nfv2t

(10)

= R e Z + R e r + ~ No

(12)

m=O

where Ao and Bo are real constants and K = Ao~/27r and

Oo~ = 2B0. It is important to note that the functions
given by eqs (9) and (10) are the Westergaard equivalents
of the symmetric part of the Williams function ~5 and can
be used to solve exactly the same types of problems for
which the Williams solution is suitable. For other types of
opening-mode crack problems, e.g., the internal crack
problem or single-ended cracks with point or distributed
loads, alternative forms of the functions Z ( z ) and Y ( z )
must be used. Some guidelines for selecting suitable stress
functions have recently been suggested by Sanford
and Berger. ~6
Expressions similar to eqs 0)-(8) for the forward shear
mode crack problem in Westergaard notation can easily
be developed and are presented in Refs. 27 and 28. In
addition, these references a l s o contain forward shear
mode forms of the stress functions Z ( z ) and Y ( z ) for a
single-ended, stress-free crack with the origin of the
coordinates at the crack tip. These latter stress functions
are analogous to the antisymmetric part of the Williams
formulation in real variables 2~ and the earlier comments
on the equivalence of the Westergaard and Williams
representations and the restrictions on their use still apply.
The general solution for the mixed-mode plane problem
can be obtained by superposition.

where No is the unknown initial fringe order. For the

stress functions given by eqs (9) and (10), the modified
optical equation can be written in real, polar coordinates
as

Nfp

2t

n=0

A , r "-1/2 c o s ( n - 89

1
+ m=0
~ Bmrmcos mO + T No

(13)

Linear Algorithms
Isopachic Patterns
Although interferometric methods have not been widely
used for fracture studies, the implementation of the local

Fig. 2--Holographic interference pattern at a crack tip in

Plexiglas

Experimental Mechanics 9 243

At any point in the isopachic pattern the coordinates,

r and 0, and the fringe order, N, can be substituted into
eq (13). The result is a linear equation in the unknown
coefficients A , , Bm and No. For a large number of such
points, an over-determined set of equations is obtained
which can only be solved in the least-squares sense. In
matrix notation these equations are of the form
[NI = [S] I t ]

(14)

where [N] is the row vector of measured fringe orders,

i.e., the left side of eq (13), [S] is the matrix of coefficients
of the unknowns, and [C] is the row vector of unknown
coefficients in the modified optical equation, eq (13). The
solution of this set of equations in the least-squares sense is
[ e l = i s ' s ] -1 i s ] T I N ]

(15)

It should be noted that the unknowns Bo(= 00=/2) and

No have the same functional form and can be merged into
one unknown constant. As a consequence, the local
collocation algorithm cannot distinguish between them
and Oox cannot be uniquely determined from isopachic
patterns.

Moir~ Patterns
The displacement patterns obtained with moir~ or
moird interferometry can also be analyzed by the linear
algorithm. 2' For grating lines parallel to the Cartesian
axes the idealized optical equations are
N,, = u / fa

(16)

N v = v/fa

(17)

or

where fd is the fringe sensitivity. Following the same

procedure as previously described, the optical equations in
real, polar coordinates can be obtained by substituting the
appropriate Westergaard stress functions into eqs (7) or
(8). For those problems for which eqs (9) and (10) are
suitable stress functions the optical equations are the
following. For u-field patterns:
Co

M=

(20)

P r cos O+ Qr sin O+ R

where P , Q, and R correspond to a constant fringe field

of P fringes per unit length in the x direction, Q fringes
per unit length in the y direction and a constant shift of
magnitude R. These additional terms must be appended
to the right-hand side of eqs (18) and (19) to obtain
modified optical equations which properly describe the
observed fringe pattern.
As a practical matter, it is not necessary to determine
the absolute order of the fringes since the constant, R,
can be interpreted as a fringe-order shift factor which
includes the effect of rigid-body motion and an unknown
origin. It is necessary only that the relative order of the
fringes be maintained for the algorithm to work.
Functionally, the resulting set of modified optical equations for the displacement at each measured point is the
same as eq (14), the solution of which is given by eq 05).
An example of the application of this method for determining the stress-intensity factor and additional series
coefficients for a crack-line loaded, compact-tension
geometry from displacement data is shown in Fig. 3. This
figure shows computer-generated reconstructions of the
fringe patterns using the constants obtained with a sixterm (Ao, AI, Bo, B1, As, B2) approximation to the u
and v displacement fields plus the three rigid-body motion
terms. The white dots shown in the figure are the data
points from moir&interferometry experiments used in the
analysis. The excellent agreement between the experimental
points and the computer-generated reconstructions
demonstrates that the analysis faithfully accounted for
all of the salient features of the experiments. The use of
computer reconstructions, such as those shown in Fig. 3,
is suggested as one method for verifying the solution.

..+1/2
"

NuEfa =

Nonlinear Algorithms (Photoelastic Patterns)

[(1 - u) cos(n + 89

A,, ~
n

-(1

+ v) (n + 89 sin 0 sin ( n - 89

~ B rm+----~x
m=0 " m + 1 [2 cos(m + 1)0)

- (1 + v) (m + 1) sin 0 sin(m0)]

(18)

and for wfield patterns:

r
rn+ll2
N , Efd = n=O
~" A , ~
[2 sin(n + 89

- (1 + v) (n + 89 sin 0 cos(n - 89
+

As was the case for isopachic patterns, these equations

must be modified to include additional effects which result
from the practical aspects of performing the experiment.
In particular, the displacement equations assume that the
loads are applied in such a way as to produce complete
symmetry about the crack line and the original stays
fixed. In practice the fixed point is usually at one of the
load points and the crack tip undergoes both a rigid-body
translation and rotation. This nonstrain related fringe
field can be expressed as

t " = (Tmax)2

(21)

r m+t

m=0 B m ~

[(1 - v) sin(m + 1)0

- (1 + v) (m + 1) sin 0 cos(toO)]

244 9 September 1989

The most widely used optical technique for studying

fracture behavior is the photoelastic method. The local
collocation method can also be applied to the analysis of
photoelastic fringe patterns. The solution scheme results
in a set of nonlinear equations to be solved. The approach
is somewhat different from that previously described;
however, the simplicity of a photoelastic experiment
compensates for the additional complexity in the analysis.
As in the previous cases, the starting point for the
formulation is the governing optical equation which, for
photoelasticity, can be expressed as:

(19)

where fo is the photoelastic constant for the specimen

material. In terms of Westergaard stress functions, the
optical equation becomes

(---fr
t " = (--q~

)2 + r~y~ = D ~ + T 2

(22)

the fringe value is Nx. Accordingly

Ok = Ok(An,B.,)

(28)

where

Tk = T k ( A . , B . )

D = ~ y2- a. = y l m Z ' + y l m

Y'-ReY

(23)

Taking a Taylor's series expansion of eq (27) and retaining only the linear terms yields

and
g k ,+, =

T = z~ = - y R e Z ' - y R e

Y'-ImY

( n - 89 A . r "-1/2 sin 0 s i n ( n - 3/2)0

n=O

+ ~

B~r~[msinOsin(mO)+cos(mO)]

(25)

m=O

T=

+ ...

Ogk

Ogk

--gk = "-~o AAo + - - ~

9A

za.'t, + . . .

+ ~Ogk ABo + ~d3gk AB . . . . .

(26)

Substitution of eqs (25) and (26) into eq (22) provides

the desired optical equation to be solved in the leastsquares sense. Clearly, eq (22) is nonlinear in the unknown
constants A . and Bin, and the solution scheme of the
previous section does not apply.
There are a variety of methods available to solve nonlinear, algebraic systems of equations in the least-squares
sense. An iterative procedure based on the NewtonRaphson method can be used to solve the optical equations if eq (22) is cast in a modified form. Let

(29)

where superscript i denotes the iterative step and AAo,

AA, . . . . .
ABo, AB, . . . . are corrections to the previous
estimates of the unknowns Ao, AI . . . . Bo, B, . . . . .
respectively.
Note from eq (27) that the desired result IS
" g ki+1 = 0
yields a system of linear equations (one for each point k)
in the correction terms A A . , AB,. of the form

(n- 89 A.r ~-l/2sinOcos(n-3/2)O

~ B~r ~ [m sin 0 cos(m0) + sin(m0)]

m=O

ag~

Og~
Og~
+'~o
ABo + ~ A B ~

n=O
-

(24)

Unlike the other experimental methods, the optical equation for photoelasticity normally does not need to be
modified to include additional experimental effects in
order to completely describe the fringe pattern.
For the stress functions given by eqs (9) and (10), eqs
(23) and (24), in real variables, become

D =

ag~

(30)

For simplicity, the iteration counter i has been eliminated.

Equation (30) can be expressed in matrix form as

[g] = [c] [z~]

(31)

where
- g,
[g] =

AAo
A =

AAN

(32)

ABo
-- gL

gk

( N k f u ~2 = 0
D2k+ T~k-- . 2t "

(27)

where the subscript k denotes the value of the function

evaluated at a point in the fringe field (rk, Ok) at which

Z~BM
and

Fig. 3--Computer
reconstruction of the u
and v moir~ patterns
containing rigid-body
motion. The white dots
denote the locations of
data points from the
experimental pattern

Experimental Mechanics 9 245

Og,
OAo

Og,
OAn

Og,
OBo

Og, q
" " " OBu
(33)

[c] =

OgL
OAo

OgL
OA~

Og~
OBo

]
OgL
OBM

where L is the total number of data points used in the

least-squares algorithm, L > M + N + 2, and N and M
are the upper limits of the truncated series approximations to Z(z) and Y(z).
At first glance it might appear that obtaining the
elements of matrix [c] involves evaluating a large number
of partial derivatives; however, this is not the case9 The
functional form of each column of [c] is identical, only
the coordinates used to evaluate the row elements are
changed. Furthermore, the partial derivatives in each of
the column elements in [c] can be rewritten as

Accordingly, an iterative procedure must be used to

obtain the best-fit set of coefficients. The steps in the
procedure are the following.
(a) From the experimental fringe pattern, select a
sufficiently large number of data points with coordinates (rk, Ok, Nk), selected so as to characterize
the features of the fringe pattern 9
(b) Assume initial values for the unknown coefficients
Ao, AI . . . . . An, Bo . . . . . BM (usually estimates for
only the first few leading terms are required, the remainder are initially set equal to zero).
(c) Compute the elements of [ g] and [c].
(d) Solve eq (35).
(e) Revise the estimates of the unknowns, i.e.,
A~*1 = A ~ + A A o

A~+' : A ~ + AA~
Bi*l
= Bio + ABo
o

Ogk = 2Dk ODk

OTk
OA,
~
+ 2Tk--OA,
or

i+1

~M

(34)

= B M -t" A B M

[Jgk = 2Dk ODk

Tk
OB---~
~ + 2T0OBm

(f) Repeat steps (c) through (e) until the elements of [A]
become acceptably small.

Since the functions D k and Tk are linear functions of the

differentiation variables ( A , , Bin), the partial derivatives
in eq (34) can be obtained from eqs (25) or (26) by
inspection 9 Finally, the expressions Dk and Tk, which
need to be evaluated in eq (34), are the same as those in
eq (27) from which the row elements of matrix [g] are
also obtained.
From the above discussion, it is clear that obtaining the
numerical values of the matrix elements necessary to form
eq (31) is a straightforward algebraic exercise which can
be implemented recursively in a variety of computer
languages. Similarly, since the functional form of eq (31)
is identical to eq (14), the solution of eq (31) in the leastsquares sense is given as

A n example of the application of this approach applied

to the determination of the stress-intensity factor for a
three-point bend specimen is shown in Fig. 4 (only a portion of the fringe pattern in the region of interest is
shown). Note that the load was applied through an autocalibration disk to eliminate errors due to imprecise measurement of the applied load and the photoelastic fringe
constant. Figure 5 shows the portion of the experimental
fringe pattern used for data acquisition, the data set
extracted from this region and the computer-generated
reconstruction of the fringe pattern using a six-coefficient
approximation to the maximum shear stress field9 Although the equations are nonlinear and an iterative
procedure was used to solve for the coefficients, the
solution converged rapidly. A comparison of the results
from the local collocation method for determining the
stress-intensity factor with the results from the boundarycollocation method for same problem is shown in Fig. 6.
Several additional examples of the application of this
algorithm for the analysis of dynamic isochromatic fringe
patterns from running cracks are presented in Ref. 27.
This reference also describes some practical matters
related to the implementation of the method including:
selection of points, number of terms to be retained,
convergence criteria, etc.

[zX] = [c~c1-1 [c] r [g]

(35)

Unlike the linear algorithms in which the solution of the

matrix equation gives the desired series coefficients
directly, the solution of eq (35) results in correction
terms for prior estimates of the coefficients A , and B=.

Summary

Fig. 4--Typical photoelastic fringe pattern for a threepoint bend specimen at a l W = 0.6. Circle denotes the
region of data acquisitiOn used in the local collocation
analysis

246 9 September 1989

The local collocation method provides a general

procedure for analyzing full-field optical fringe patterns 9
The method does not depend on unique features of the
fringe pattern, as in the apogee method, nor is it confined
to the near-field region around the crack tip, as previous
methods were. The method is also general in the sense
that it can be used with a variety o f stress functions and
does not depend on the single-ended crack-tip expansion
for the stress-field equations 9 By extension, the methods
described can be applied to mixed-mode loading by adding
to the optical equations the additional mode II terms for

the stresses. The method has proven to provide accurate

answers with a small number of unknown coefficients. In
part this increase in accuracy is the result of the high
degree of data redundancy used in the method (typically
six to ten data points are used for each unknown retained
in the analysis).
Although the methods described in this paper are
somewhat more difficult to implement than previous
methods, the use of modern desk-top computers and
digitizing equipment makes the methods practical. The
increase in accuracy and the ability to infer additional
information about the fracture parameters from the nonsingular terms more than compensates for the additional
time it takes to implement the algorithms. It is not difficult
to envision how new image-processing equipment and
feature-extraction algorithms could be incorporated into
the experimental analysis in order to fully automate the
procedure and produce results while the experiment is
in progress.

Acknowledgments
The development of the local collocation method
described in this paper has evolved over a period of ten
years under the sponsorship from a variety of sources,
including U.S. Nuclear Regulatory Commission, Oak
Ridge National Laboratory, Sandia Laboratories, Office
of Naval Research and the National Science Foundation.
Additional financial support was also received from the
Computer Science Center of the University of Maryland.

Fig. 5--Sequence of events for local collocation of the

fringe pattern of Fig. 4

16

PHOTOELASTIC RESULTS
BOUNDARY COLLOCATION
(Srowley, i976)

14

12

IC

The author wishes to express his appreciation to Dr. G.R.

Irwin for his advice and encouragement. Finally, the
contributions of Dr. R. Chona over these last ten years
are gratefully acknowledged.

References
1. Wells, A. and Post, D., "'The Dynamic Stress Distribution
Surrounding a Running Crack--A Photoelastic Analysis, "" Proc. SESA,
16 (1), 69-92 (1958).
2. Cranz, C. and Schardin, H., "'Kinematographic avor ru hendem
film und mit extrem hoher bildfrequenz, "" Zeit, f. Phys., 56, 147 (1929).
3. Dally, d.W., "'Dynamic Photoelastic Studies o f Fracture, "" William
M. Murray Lecture, EXPERIMENTALMECHANICS, 19 (10), 349-362
(1979).
4. Irwin, G.R., Discussion o f Ref. l, Proc. SESA, 16 (1), 93-96 (1958).
5. Sih, G.C., "'On the Westergaard Method o f Crack Analysis, "" Int.
J. Fract. Mech., 2, 628 (1966).
6. Eftls, J. and Liebowitz, H., "'On the Modified Westergaard
Equations f o r Certain Crack Problems, "" Int. J. Fract, Mech., 8 (4),
383-392 (1972).
7. Larsson, S.G. and Carlsson, A.J., "'Influence o f Non-singular
Stress Terms and Specimen Geometry on Small-scale Yielding at Crack
Tips in Elastic-plastic Materials, "" J. Mech. Phys. Solids, 21, 263-277
(1973).
8. Sanford, R.J., ",4 Critical Re-examination o f the Westergaard
Method Solving Opening Mode Crack Problems, "" Mech. Res. Comm.,
6, 289-294 (I979).
9. Bradley, W.B. and Kobayashi, A.S., "'An Investigation o f
Propagating Cracks by Dynamic Photoelasticity,'" EXPERIMENTAL
MECHANICS,l0 (3), 106-113 (March 1970).
I0. Kobayashi, T. and Dally, J.W., "'The Relation Between Crack
Velocity and the Stress Intensity Factor in Birefringent Polymers," A S T M
STP 627, 257-273 (1977).
11. Dally, J.W. and Kobayashi, T., "'Crack Arrest in Duplex Specimens, "" Int. J. Solids, Struct., 14, 121-129 (1978).
12. Smith, C.W., "Use o f Three-dimensional Photoelasticity in Fracture Mechanics, "' EXPERIMENTALMECHANICS,13 (12), 539-544 (1973).
13. Schroedl, M.A., McGowan, J.J. and Smith, C. W., "'Determination
o f Stress Intensity Factors from Photoelastic Data with Application to
Surface Flaw Problems, "" EXPERIMENTALMECHANICS,14 (10), 392-399
(1974).
14. Smith, C. W., "'Photoelasticity in Fracture Mechanics, "' EXPERIMENTALMECHANICS,20 (10, 390-396 (1980).
15. Smith, C.W., Post, D., Hiatt, G. and Nicoletto, G., "Displacement
Measurements Around Cracks in Three-dimensional Problems by a Hybrid
Experimental Technique,'" EXPERIMENTALMECHANICS, 23 (1), 15-20
(1983).
16. Sanford, R.J. and Dally, J.W., "'A General Method for Determining Mixed-Mode Stress Intensity Factors from lsochromatic Fringe
Patterns, "" Eng. Fract. Mech., 11, 621-633 (1979).
17. Parks, V.J., Sanford, R.J. and Dally, J.W., "'Optical Fracture
Mechanics and Stress Analysis o f a Turbine Engine Fan Disk, "" CASE
STUDIES O N S T R U C T U R A L I N T E G R I T Y A N D RELIABILITY, Proc.
A r m y Syrup. on Solid Mech., 113-131 (Sept. 1978).
18. Dudderar, T.D. and Gorman, H.J., "'The Determination o f Mode I
Stress-intensity Factors by Holographic Interferometry, "" EXPERIMENTAL
MECHANICS,13 (4), 145-149 (1973).
19. Schroedl, M.A. and Sin#h, C. IV., ",4 Study o f Near and Far Field
Effects in Photoelastic Stress Intensity Determination, "" Eng. Fract.
Mech., 7, 341-355 (1975).
20. Rosakls, A.J. and Ravi-Chandar, K., "'On Crack Tip Stress State."
A n Experimental Evaluation o f Three-dimensional Effects, "" Int. J. Solids
and Struct., 22 (2), 121-134 (1986).
21. Chona, R., Irwin, G.R. and Sanford, R.J., "Influence o f Specimen
Size and Shape on the Singularity Dominated Zone, "" A S T M 791, I3-I23
(1983).
22. Sanford, R.J., "'Application o f the Least Squares Method to
Photoelastic Analysis, "" EXPERIMENTALMECHANICS,20, 192-197 (1980).
23. Sanford, R.J. and Chona, R., "'Photoelastic Calibration o f the
Short Bar Chevron-notch Specimen," A S T M ST,~ 855, 81-97 (1984).
24. Barker, D.B., Sanford, R.d. and Chona, R., "'Determining K and
Related Stress-Field Parameters from Displacement Fields," EXPERI-

MENTALMECHANICS,

4
2
C

O.I

0!2

0.3

0!4

O,S

0.6

0.7

01.8

o/w

Fig. 6--Comparison of the dimensionless shape

function for the three-point-bend geometry from
boundary and local collocation

0.9

25

(I2), 399-407 (1985).

25. Williams, M.L., "'On the Stress State at the Base o f a Stationary
Crack, "" J. AppL Mech., 24, 109-114 (March 1957).
26. Sanford, R.J. and Berger, J.R., "'The Numerical Solution o f
Opening Mode Finite Body Fracture Problems Using Generalized Westergaard Functions, "" Eng. Fract. Mech., in press.
27. Chona, R. and Sanford, R.J., "'Analyzing Crack Tip Isochromatic
Fringe Patterns," VI Int. Cong. on Exp. Mech. (1988).
28. Chona, R., "'The Stress Field Surrounding the Tip o f a Crack
Propagating in a Finite Body, "" Phl) Dis&, Univ. o f Maryland, College
Park, M D 1987.

Experimental Mechanics 9 247