Professional Documents
Culture Documents
25-43 (1980)
FRANCOIS E. HEUZES
Department of Civil, Environmental and Architectural Engineering, University of Colorado, Boulder, Co.. U.S.A.
SUMMARY
A finite element model for the prediction of discrete fracture propagation in rock structures loaded in
compression is presented. The model integrates any one of three theories for mixed-mode fracture
initiation; it contains an energy balance algorithmfor predicting crack increment length, and incorporates
recent developmentsin finite element stress-intensityfactor computation.The predictionsof the model are
compared with the observed fracture response of a real rock structure. Results show that the model
accurately predicts both stable and unstable fracture progagations observed experimentally.
INTRODUCTION
At temperatures, pressures and strain rates normally encountered in rock mechanics, failure of
hard rock loaded in either tension or compression is by fracture. Fracture analysis of the
tension-loaded rock structure'-' has been facilitated by the vast body of classical fracture
mechanics literature directly applicable to this case. However, applied stresses are often
compressive rather than tensile in the geologic environment. The fracture response of structures
containing notches and cracks loaded in compression differs in a number of fundamental
respects from the response under tension loading. This paper deals specifically with compres-
sion-loaded rock structures. Because of the nature of rock materials this problem is quite
different from those usually encountered in fracture mechanics.
These differences are most conveniently illustrated with reference to the model structure
shown in Figure 1. It consists of an edge-loaded plate containing a central notch whose major
axis is oriented at an angle /3 to the major principal compressive stress. This structure was tested
in the laboratory under uniaxial compression.6 The salient features of the fracture response
were:
1. first crack growth occurs from points initially under tensile stress concentration, on the
notch. This set of two, symmetrically located cracks was labelled primary;
2. propagation of primary cracks was stable;
3. primary crack path of propagation was curvilinear;
* Portions of this paper were presented at the 18th U.S.Symposium on Rock Mechanics, Keystone, Colarado, 1977.
t Assistant Professor.
$ Associate Professor.
I
I
I
I
I
I
I
I
I
I
I
Figure 1. Angle-notched plate edge-loaded in uniaxial compression. The plate is 4 in (102 mm) square, 314 in (19 mm)
thick; a = 0.40 in (10.16 mm) and b = 0.004 in (0.102 mm)
4. after considerable primary crack propagation, a second set of two, symmetrically dis-
placed, cracks appeared. These were labelled secondary, and apparently originated from
points of initially compressive stress concentration on the notch;
5 . failure of the plate, defined as a through-going rupture, was a result of unstable secondary
crack propagation, at a load level in the range of 3 to 5 times the primary crack initiation
load.
These observations, depicted in Figure 2, were typical for model plates of Indiana limestone and
St. Cloud charcoal granodiorite, with 3O0S/3~ 9 0 "For . /3< 30", only primary cracks were
observed and plate rupture did not occur.
Observations 2 and 3 differentiate the observed fracture response of this configuration from
that usually observed in tension-loaded structures. Stable primary crack propagation indicates
that, within the assumption of linear elastic fracture mechanics (LEFM), the associated energy
release rate, G, decreases with increasing crack length for a constant load.' Consequently, under
compression, fracture initiation load is not synonymous with failure load. The curvilinear nature
of the primary crack path is a result of a variable, mixed-mode stress intensity being applied to
the incrementally advancing crack tip.
Observations 4 and 5 are particular to rock. In tests on glass, polymethylmethacrilate
(PMMA) and CR39 in the same only primary cracking was evident and
rupture did not occur. This phenomenon is shown in Figure 3 which depicts the primary crack
behaviour of a PMMA plate loaded to near its compressive yield stress. Other worker^^'^^ tried
using glass or plastic to simulate rock fracture response. But, since secondary cracking did not
occur in glass or plastic, they had to postulate the coalescence of multiple flaws to try and explain
FINITE ELEMENT MODELS FOR ROCK FRACTURE MECHANICS 27
(b)
Figure 2. Failure of two Indiana limestone angle-notched plates. Primary cracks initiate at A and propagate to B.
Secondary cracks nucleate in a region near C and propagate both toward the notch tips and the plate edges. (a) @ = 45",
(b) B = 90"
the rupture of rock loaded in compression. This hypothesis, although appealing at first, was not
validated by experimental evidence. For example, Figure 4 shows the response of a PMMA
plate containing two notches. Contrary to expectations, coalescence of primary cracks did not
occur, and the plate did not fail. In some cases, the existence of a flaw array actually retarded
primary crack growth:' and only in rare instances have primary cracks been seen to coalesce.
Observations 4 and 5 therefore indicate a previously unrecognized, fundamental difference in
the fracture response of real rock structures as opposed to glass and plastic models. This
difference will be explained in a subsequent section of this paper.
28 ANTHONY R. INGRAFFEA AND FRANCOIS E. HEUZE
Figure 3. Fracture response of an inclined notch in a PMMA plate: stable, primary crack propagation without plate
rupture. Load in KSI
Figure 4. (a) Fracture response on an en echelon array of notches in a PMMA plate: stable, primary crack propagation.
Although interaction effect is evident, cracks do not coalesce, and plate does not rupture. (b) Fracture response of an
array of colinear notches in a PMMA plate: stable, primary crack propagation. Cracks do not coalesce, and plate does
not rupture
FINITE ELEMENT MODELS FOR ROCK FRACTURE MECHANICS 29
The objective of this paper is to present a numerical model capable of accurate and complete
prediction of observed fracture response in a real rock plate (Figure 1).It is a finite element
model (FEM)which integrates modern theories of mixed-mode fracture mechanics, classical
energy balance concepts, and recent developments in finite element stress-intensity-factor (SIF)
computation. Theoretical developpents involved in formulating the model will first be detailed.
THEORETICAL BACKGROUND
The experimental observations cited above dictate the following requirements for a numerical
model for. fracture propagation in rock under an incrementally increasing compressive load:
1. prediction of a critical combination of mixed-mode stress-intensity factors, KI, KII
associated with the initiation of each primary crack increment;
2. prediction of the corresponding change in direction of propagation, Bo;
3. prediction of the primary crack increment length, Aa.
The first two of these requirements are met by existing theories of mixed-mode fracture
initiation.
Figure 5. Crack tip polar coordinate system, stress notation, and stress-intensity factor sign convention
by
1
rre=m
1
2
[ cos
1+. .
KIsin 6 + KII(3cos 8 - 1) .
The gemaxtheory predicts a fracture locus in the KICr-normalized KI- KIIplane given by
in which K I , is the Mode I critical SIF, or, Mode I fracture toughness. The fracture angle, 60, is
found by maximizing in the second of equations (l),which leads to
60
cos -[KI sin 60 + KII(3 cos 6 0 - 111= 0 (3)
2
Since it is expected that near the crack tip, large strain and/or inelastic material effects would
invalidate equation (l),memaxis evaluated along a contour some small distance r, from the crack
tip; r, is assumed to be a material property. The gomax theory is also applicable to fracture
initiation from notches.
In the notch case, fracture is predicted to initiate from the point of highest tangential tensile
stress, when the circumferential tangential stress a small distance, ion, from this point reaches
get. The initial fracture direction is again 60 normal to reemax. The difference in characteristic
distances, roc versus r,,, accounts for the vastly different stress gradients near the fracture
initiation points in cracks and notches.
In the S( 6)mint h e ~ r y , ’ ” ’fracture
~ initiation is again dependent on a near-tip point variable:
the strain energy density. Fracture initiates from the crack tip in a direction, BO, along which the
strain energy density at a critical distance is a minimum, when this minimum reaches a critical
material value, (aU/a V ) c .Near the crack tip
in which
S = 1K: + 2~ izKIKIr+ UZZK:I (5)
is the strain energy density intensity along a countour at a characteristic distance, r,, from the
crack tip. The coefficients, uii, are
sin 6
u 12 = - [2 cos 6 - (K - l)]
16~p
FINITE ELEMENT MODELS FOR ROCK FRACTURE MECHANICS 31
in which
p = shear modulus,
K = (3- v)/(l v ) in + plane stress,
v = Poisson’s ratio
The fracture locus predicted by the S(0),in theory is given by
-- I
'"'mar
(Ref.15) \\\
o'2
"
0 0.2 0.4 0.6 0.8 1.0
K, ' K k r
Figure 6. Comparison of four theoretical fracture initiation loci. Loci are symmetric about the K I / K I ,axis
dependent on the particular energy release rate formulation of equation (9).Different formula-
tions have been proposed (for example References 16 and 17)which would result in substantially
different fracture loci under the G(8),., criterion. Also, the curve corresponding to equation (7)
of the S(e),,. theory is for plane stress, and the Poisson's ratio is that of the Indiana limestone
plate whose fracture analysis is to be subsequently detailed.
8'
0.75 -
a
Figure 7. Idealized G- and &curves, showing the pronounced effect of a small change in R on the computed, stable
crack increment length, Au*
Moreover, equation (9) shows that the G-curve can be computed from knowledge of the SIF
variation with Aa. R, given by the second curve, is the crack resistance to growth which, in an
ideally brittle material, is independent of crack length. For plane stress, R is given by the familiar
expression
and the units on R and G are energy per unit crack area. Consequently the R-curve is a measure
of the rate at which a unit thickness of a given material absorbs energy per unit crack extension.
Conversely, the G-curve is a measure of the rate at which the changing structural configuration
releases energy to crack extension, per unit crack extension. Integrals over Au of the G- and
R-curves thus represent opposing components of an energy balance scheme. For a fixed load, P,
the energy source is
joAn*G(P,
Aa) dAa (14)
jOAa*R
dAa
The correct crack increment length, Aa*, is obtained by solving the integral equation resulting
from equating equations (14) and (15). This mathematical manipulation is equivalent to
equating the areas ABC and CDE in Figure 7.
34 ANTHONY R. INGRAFFEA AND FRANCOIS E. HEUZE
T H E NUMERICAL MODEL
The finite element approach
Prediction of load, direction, and length corresponding to each primary crack increment
demands accurate computation of SIF's. Moreover, to investigate the mechanism of secondary
crack formation, accurate analysis of the stress state away from the primary crack tip was
required. The finite element method was selected as most suitable for these requirements.
The plate model chosen for complete, detailed fracture modelling was of Indiana limestone in
the p =45" configuration. Figures 8 show the finite element model of this structure. Iso-
parametric eight-node quadrilaterals and six-node triangles are employed away from the
propagating crack tips. Figure 8(b) is a typical near-tip detail, showing the use of +-point
singularity elements" around the crack tip. SIF's were computed from the singularity element
nodal displacements using the displacement-correlation technique. The singular term in the
appropriate element shape function expansion was equated to the corresponding, singular
analytical expression.'' Along the B = *Moo rays emanating from the crack tip, the resulting
SIF expressions are, for plane stress,
(a)
Figure 8 (a) Finite element mesh for 0 = 45". Number of elements = 78. Number of nodes = 232. Detail 1 given in
Figure 8 (b)
FINITE ELEMENT MODELS FOR ROCK FRACTURE MECHANICS 35
Scale:
\ I / S \ I
I
(b)
Figure 8 (b) Typical primary crack region detail. S denotes 1/4-point, singular triangle. Crack-tip nodal lettering as
indicated (1 in = 25.4 mm)
in which
L = length of singularity element side along the ray,
uf = crack-opening nodal displacements,
uf= crack-sliding nodal displacements.
The primes indicate that the global coordinate nodal displacements have been transformed to
the crack-tip coordinate system defined in Figure 5 . The subscripts on u’, u’ correspond to the
nodal lettering convention given in Figure 8(b).
Cracks were not constrained to follow a predetermined path. Predicted crack increment
length and direction were efficiently accommodated using the nodal grafting technique*l and
slight rezoning of the crack-tip region.
In Stage 1, Muskhelishvili’ssolution” for the stress and strain energy density fields around an
elliptical notch is used. Output of this stage for the gornax and S(8),in theories is:
1. first fracture load, PI;
2. point of initiation, pi, on the notch;
3. initial angle of propagation, yl, where y is defined in Figure 11.
Muskhelishvili’s solution gives only (2) for the G(8),., theory, which relies solely on the FEM
for predicting (1) and (3), as will be shown.
In Stage 2, the first fracture increment length is computed using the energy balance technique.
For the first critical load, P1,and first primary crack extension, Aul, an equation for G(yl,Au) is
obtained by the following algorithm:
1. the crack is extended from pi a trial length A u t , in direction yl;
2. obtain KI and KIIusing equations (16) and (17);
3. obtain G(yl, Au:) from equation (9).This is first point on the G-curve;
4. extend crack from Aui to Au:;
5. repeat (2)and (3), obtaining the next point, G(yl, A u i ) ;
6. repeat (4), (2) and (3)until n points on the curve are computed;
7. fit polynomial curve through points, G(yl, Aui) to G(yl, ha;) usingleast-squares routine;
8. solve equations (14) and (15);
9. go to Stage 3.
This algorithm assumes the crack to progagate at a fixed angle, yl, until energy supply at P1, y1 is
exhausted. At this point, the fracture initiator is no longer a notch, but a crack at angle yl. To
check whether continued propagation at a new angle but under the same load, P1,is possible,
Stage 3 is entered.
In Stage 3, load and direction for initiation from a crack tip are computed. This stage is
entered after length-stability of the previous fracture increment at angle yi-l is reached, if such
stability is possible. Here the algorithm is:
1. obtain K1 and KIIat the tip of previous crack increment;
2. substitute these SIF’s into equation (l),equation (4)or equation (9);
3. obtain Bo, equation (3), equation (8) or equation (12), and corresponding new crack
direction, yi ;
4. substitute 80 into equation (2), equation (7) or equation (11). Obtain stationary values of
governing parameter;
5. compare stationary value with critical stationary value. If smaller than critical, increase
load until larger. When or if larger than critical, increment initiates in new direction, yi,
obtained in (3);
6. go to Stage 4.
In Stage 4, the length of an increment extending from a previous increment’s tip is computed by:
1. obtaining first point on G ( y i ,Au) curve by substituting critical combination of SIF‘s ((1)of
Stage 3) and corresponding direction, Bo ((3) of Stage 3), into equation (9);
2. extending crack a trial length, Au:, from tip of previous stable increment in new direction
yi ((3) of Stage 3);
3. repeating (2) through (9) of Stage 2 to obtain correct increment length, Aui.
Stages 1through 4 are the complete analysis strategy for the gemax and S(t9),in theories. It has
been previously argued that, since their governing parameters are point variables measured
internally, these theories are applicable to fracture initiation from both notches and cracks. No
such claim has been made for the G(B),, theory. That is, the availability of sufficient energy to
sustain initiation from a notch is a necessary condition. This global condition has nor been shown
to imply the suficienr condition of local material failure. Moreover, the G(B),., theory, as
FINITE ELEMENT MODELS FOR ROCK FRACTURE MECHANICS 37
presently formulated, is expressible only in terms of SIF's which do not exist at a non-stress
singular point of initiation on a notch. The problem may be approached, however, from a
different point of view. In this investigation, it was assumed that, at the point of initiation on the
notch, there exists a crack of vanishinglysmall length oriented normal to the notch surface. That
such a crack could exist a priori or grow due to the high tensile tangential stress at this point is not
beyond imagination, especially in a rock structure. An algorithm for predicting the first fracture
increment load using the G(O),, theory is then:
1. extend a crack of length Aal from point pi normal to the notch surface;
2. compute KI and K I Ifor a unit load at the tip of this crack increment, equations (16) and
(17);
3. extend a second, normal increment from pi such that Aa2>Aal;
4. repeat (2);
5. repeat (3) and (2) until n increments Aal < Aa2 . . . Aa, and corresponding SIF's are
known;
6 . obtain G(7,ha,) using equation (9) for each increment tip;
7. fit a polynomial equation through these points using a least-squares routine;
8. compare the extrapolated value, G * ( y , 0), with G,and obtain the first fracture initiation
load from
P: = G,/G*(r, 0 )
9. go to Stage 4.
An examplebf the procedure from Stage 4, step 1to Stage 3, step 1 is given in Figures 9 and
10. Figure 9 was used to compute the length of the third, primary crack increment according to
0.07[ ,"E&;
0.06
0.04
I
I
0.02 - 1
0.0I - I
V A a , ~0.014"
I 1 I 1
Ok d.004 0.012 0.020
Aa3 ( i n )
Figure 9. Example summary of Stage 4 computations. Energy release rate, G,and energy absorptionrate, R,curvesfor
third primary crack increment, S(e),, theory (1 in = 25.4 mm, 1 Ib = 4.45 N)
38 ANTHONY R. INGRAFFEA AND FRANCOIS E. HEUZE
KII
(psi $in)
700
1 00 FEM
om EQ 10
500 1
1 I
300
I00
Aa, ~0.014" I
n I 1
"0 0.0I 0.02
Aa3 (in)
Figure 10. Example computation (Stage 3, step 1) of SIF's at tip of third primary crack increment, S(S),,. theory.
(1 psi Jin = 1.1 kNm-''*)
the S(0),i,, theory. Figure 10 received the resulting Au3 and computes the corresponding S I F s
at this new primary crack tip position. With these SIF's, step 2 of Stage 3 is entered and the
algorithm proceeds.
These analysis procedures require input of a number of material properties. Detailed
descriptions of experimental techniques and results are given in Reference 6, and a summary of
those results is presented in Tables I, I1 and 111.
ion, in r,, in in
ueC,'
in (mm) in (mm) psi (MN/m*)
E, in G,,1*2in
psi (MN/m2) Ib-in/in2 (N-mm/mm')
PRIMARY CRACK
Figure 11. Comparison of predicted and observed primary crack paths. Indiana limestone, 6 =45". Final load,
PI,,= 5000 psi (34.5 MPa) (1 in = 25.4 mm)
AB'C' = C'D'E' in Figure 7)about 20 per cent longer than that predicted using the full R value.
Since, by equation (13), a 5 per cent error in R corresponds to only about a 2 per cent error in
KI,,very accurate estimates of fracture toughness are required for accurate prediction of stable
crack growth.
Figure 12. FEM computed principal tensile stresses in region of Indiana limestone, /3 = 45" plate. End of eighth
primary crack increment, Pe = 3200 psi (22.08 m a ) . S element stresses not shown for clarity.
of the u2compressive principal stress parallel to the initiation flaw is neglected, the applicable
fracture initiation formula isz3
It is assumed that a, is the critical radius of a 'penny-shaped' flaw, such as a grain boundary,
embedded in the thickness of the plate, and oriented normal to ul.When equation 17 is solved
simultaneously with the relationship between KI, and a, experimentally derived for Indiana
limestone,6 the resulting critical flaw radius is a, = 0-03 in (0.787 mm). This radius is in good
agreement with the observedz4 half-grain size in Indiana limestone of 0.010 in-0.040 in
(0.254 mm-1.016 mm).
Once nucleated in this manner, the secondary crack grows along the corridor of tensile stress,
both toward the notch tip and the plate edge, until it reaches a globally unstable length. At this
point, through fracture of the plate occurs. Photographic proof obtained from plates on which
testing was halted just prior to rupture shows that secondaliy cracks do nucleate in the plate
interior. They do not start from a point of high compressive stress concentration on the notch, as
might appear from casual inspection of failed specimens.
42 ANTHONY R. INGRAFFEA AND FRANCOIS E. HEUZE
DISCUSSION
The scenario of primary/secondary crack progagation observed and successfully modelled
strongly suggests the following: vertical splitting typical in the failure of uniaxially loaded rock
specimens is the result of unstable, secondary crack propagation, rather than of coalescence of
stable primary cracks. Unstable secondary cracking from a single flaw can lead to specimen
rupture. It is now clear why secondary cracking is not observed in glass or plastic: the intrinsic
critical flaw is not present in the nucleation zone.
It is now reasonable to imagine that the so-called ‘inclinedshear fractures’ typically observed
in triaxial tests are the expression of unstable, secondary cracks whose direction of propagation
is governed by the ratio of principal applied stresses. The natural sequel to this investigation
must be the performance and modelling of biaxial compression tests on the structural configura-
tion of Figure 1.
CONCLUSIONS
Comparison of the predictions of the new finite element model with experimental observations
shows that:
1. the model duplicates the stable and curvilinear nature of observed primary crack growth;
2. primary crack trajectories predicted by the gemax and G(O),, theories are in good
agreement with each other and with observed trajectories;
3. primary crack trajectory predicted by the S(t9),i, theory agrees only qualitatively with the
observations;
4. predictions of primary crack length were consistently short. Error analysis indicates that
this is due to the high sensitivity of the length prediction algorithm to slight mis-estimation
of fracture toughness;
5 . the model predicts the formation of observed secondary cracks;
6. secondary cracks nucleate in the interior of a rock structure owing to stress redistributions
resulting from primary crack growth;
7. the model predicts unstable propagation of secondary cracks resulting in structural failure,
as observed.
REFERENCES
1. R. A. Schmidt, ‘Fracture toughness testing of limestone’, Exp. Mech. 16 (9,161-167 (1976).
2. R. A. Schmidt, ‘Fracture mechanics of oil shale-unconfined fracture toughness, stress corrosion cracking, and
tension test results’, Proc. 18rh US.Symp.on Rock Mechanics, pp. 2A2-1,2A2-6 (1977).
3. A. S. Abou-Sayed and E. R. Simonson, ‘Fracture toughness, KIc,of triaxially-loaded Indiana limestone’,Prm. 18rh
US.Symp. on Rock Mechanics, pp. 2A3-1,2A3-8 (1977).
4. S. H.Advani, L. Z. Shuck, H. Y. Chang and H.V. Gangarao. ‘Analytical and experimental investigations on
induced fracturing of reservoir rock’, ASME, Petr. Div. Preprint No. 76-PET-8 (1976).
5 . M . P. Hardy, ‘Fracture mechanics applied to rock’, Ph.D. dissertation, University of Minnesota (1973).
6. A. R. Ingraffea. ‘Discrete fracture propagation in rock: laboratory tests and finite element analysis’, Ph.D.
dissertation, University of Colorado (1977).
7. J. Glucklich, ‘Oncrack stability in some fracture tests’, Engng Fruct. Mech. 3, 333-344 (1971).
8. B. Cotterell, ‘Brittle fracture in compression’, Int. I.Fmr. Mech. 8, 195-207 (1972).
9. E. G. Bombalakis, ‘Photoelastic study of initial stages of brittle fracture in compression’, Tectonophysics, 6,
461473 (1968).
10. E. Hoek and Z. T.Bieniawski, ‘Brittle fracture propagation in rock under compression’, Inr. J. Fracr. Mech. 1,
139-155 (1965).
11. F. Erdogan and G. C. Sih, ‘Onthe crack extension in plates under in plane loading and transverse shear’, I.Busic
Engrg, ASME, 85, 519-527 (1963).
12. J. G . Williams and P. D. Ewing, ‘Fracture under complex stress-the angled crack problem’, Int. J. Frucr. Mech. 8,
441-446 (1972).
FINITE ELEMENT MODELS FOR ROCK FRACTURE MECHANICS 43
13. G. C. Sih, ‘Strain-energy-density factor applied to mixed-mode crack problems’, Inr. J. Frucr. Mech. 10,305-321
(1974).
14. M. E. Kipp and G.C. Sih, ‘The strain energy density criterion applied to notched elastic solids’, Inr. J. Solids Srrucr.
11,153-173 (1975).
15. M. A. Hussain, S. L. Pu and J. H. Underwood, ‘Strain energy release rate for a crack under combined Mode I and
Mode 11’, Fracture Analysis, ASTMSTP 560, 2.28 (1974).
16. T. K. Hellen and W.S. Blackburn, ‘The calculation of stress intensity factors for combined tensile and shear
loading’, Inr. J. Frucr. Mech. 11,605-617 (1975).
17. M. L.Vanderglas and R. J. Pick, ‘Onvirtual crack extension methods for combined tensile and shear loading’, Proc.
Int. Conf. on Frucr., 4rh. Waterloo, 3, 501-505 (1977).
18. H. Miyamoto, S. Fukuoa and K. Kageyama, ‘Finite element analysis of crack propagation under compression’, Proc.
Inr. Conf. on Frucr., 4rh. Waterloo, 3,491-499 (1977).
19. R. S. Barsoum, ‘Triangular quarter-point elements as elastic and perfectly-plastic crack tip elements’, Inr. J. Nurn.
Merh. Engng. 11, 85-98 (1977).
20. C . F. Shih, H. G. Delorenzi and M. D. German, ‘Crack extension modeling with singular quadratic isoparametric
elements’, Inr. J. Fracr. 12,647-651 (1976).
21. A. R. Ingraffea, ‘Nodal grafting for crack propagation studies’, Inr. J Num. Merh. Engng, 11, 1185-1187 (1977).
22. N. I. Muskhelishvili, Some Basic Problems of rhe Marhemaricul Theory OfElasticiry, P. Noordhoff, Groningin. 1963.
23. 1. N. Sneddon, ‘The distribution of stress in the neighborhood of a crack in an elastic solid’, Proc. Roy. SOC.London,
Series A, 187,229 (1946).
24. A. R. Ingraffea and R. A. Schmidt, ‘Experimental verification of a fracture mechanics model for tensile strength
prediction of Indiana limestone’, Proc. 19rh US . Symp. on Rock Mechanics, Mackay School of Mines, University of
Nevada, 247-253 (1978).