A COMPUTATIONAL APPROACH TO

ROCK FRAGMENTATION

by J. K. Dienes and L. G. Margolin, Theo-

Downloaded from http://onepetro.org/ARMAUSRMS/proceedings-pdf/ARMA80/All-ARMA80/ARMA-80-0390/2042003/arma-80-0390.pdf/1 by King Fahd University of Petroleum & Minerals user on 05 May 2024
retical Division, Los Alamos Scientific
Laboratory, University of California,
Los Alamos, New Mexico
ABSTRACT
eral, though possibly less advanced, than previous
fragmentation theories. Anisotropic effects in oil
A method for numerical modelling of rock frag- shale are associated with the bedding planes, which
mentation has been developed using a statistical ap- are visible to the naked eye, the influence of nat-
proachto estimate th6 effect of flaws on rock ural joint sets, and the effect of blasting, which
masses. Though motivated by the need to design effi- causes cracks with certain orientations to grow pref-
cient oil shale retorting techniques, the method can
erentially. In principle the effects of all these
be used to study a variety of problems concerning the
anisotropies can be accounted for by the proposed
effects of faults, joints and cracks on rock masses.
method, though in practice this may require more in-
This approach has the advantage over plasticity mod- formation about rock formations and more computer
els that it accounts for both strain-rate and size
powerthan is normally available. The approachis
effects without introducing additional parameters,
intended to represent in somedetail the micro-me-
and makes use of fundamental physical ideas concern-
chanics of rock behavior during fragmentation and
ing the behavior of cracks rather than empirical data makes use only of rock parameters that have direct
fitting.
physical significance and, hence, can often be esti-
The method is based on two theoretical results.
mated without extensive laboratory testing. In this
The first involves an explicit formula for the strain
respect we are hopeful that the methodwill have ad-
due to crack opening and crack shear resulting from
vantagesover the phenomenological(modified plasti-
stress on an ensemble of penny-shaped cracks. The
city) approaches in which a failure surface is per-
second concerns the calculation of crack opening and
mitted to vary in an empirical way, usually as a
crack stability for a penny-shaped crack under an ar-
function of strain. Such approaches tend to require
bitrary three-dimensional state of stress. The gen-
extensive laboratory testing and empirical fitting of
eral theory accounts for the effects of open cracks data. In particular, strain rate effects have to be
and closed cracks separately, with the effect of in-
included in the phenomenological approaches by intro-
terfacial friction on closed cracks playing an im-
ducing additional empirical data, whereas in statis-
portant role. A series of calculations showing the tical cracks mechanics they appear as a natural con-
effect of strain rate on stress at a fixed strain is
sequenceof the finite growth rate of cracks. In ad-
described.
dition, size effects are accounted for without addi-
INTRODUCTION AND SUMMARY tional parameters because large samples tend to con-

Fragmentation by explosives is one of the first tain larger cracks than small ones and, as a result,
have lower strength.
steps in extracting kerogen from oil shale on a com-
The idea of characterizing the behavior or rock
mercial scale. As part of a program to optimize de-
sign of oil shale retorts, we have been developing a by a superposition of the effects of a randomdistri-
bution of cracks has been explored at a number of
theoretical approach to fragmentation which will form
laboratories during the last few years. At LASL we
the basis for a computer program. Ultimately, we
have been calling this "Statistical Crack Mechanics"
hope that the program will compute detonation of the
Extensive theoretical and experimental work in this
explosive, its interaction with oil shale, and frag-
area has been done at the Stanford Research Laborato-
mentation of the rock as a result of tension, shear,
or a combination of the two. We have been particu- ry, as reported by McHugh
and Keogh(1979). Their
methods have been successful in fitting fracture data
larly interested in the effects of anisotropy, and in
for simple geometries, but the extension to more com-
this respect the approach is believed to be more gen-
plex configurations where shear and complex stress

390
states are important has not been demonstrated. A ial tests the strain often tends to increase in a

conceptually similar approach has been described by smooth fashion, probably as the result of growth and
Bamberger(1979) and applied to the analysis of con- subsequent intersection of a large number of cracks.
crete. •lis work assumes that cracks have an initial To account for this effect we have added to the dis-

opening that develops while setting up. The scatter- tribution of active cracks a distribution of "inac-

ing of waves by an ensemble of cracks has been inves- tive" cracks which are no longer capable of growing

Downloaded from http://onepetro.org/ARMAUSRMS/proceedings-pdf/ARMA80/All-ARMA80/ARMA-80-0390/2042003/arma-80-0390.pdf/1 by King Fahd University of Petroleum & Minerals user on 05 May 2024
tigated by Piau (1979) and by Garbin and Knopoff because of intersections with other cracks, but
(1975). A method for calculating the reduced modulus still play a role in reducing the stiffness and
of an isotropic distribution of cracks which included strength of the rock.
the effect of tip plasticity was described by Dienes The theory has been incorporated into the SCM
(1978). Since then a more general approach was de- subroutine which produces the stress history on an
veloped which accounts for the effects of shear de- element for which the strain history and initial
formation and anisotropy of the crack distribution. stress are known. The coding is done in such a way
This paper presents the underlying theory, though that it can be used in a Lagranqian code to represent
space does not permit detailed development of all the the material behavior. A •ualitative check on the
results. method is made in the final section by comparing
In the next section we discuss the strain ener- the stress at a fixed strain for a variety of strain
gy and crack opening associated with a circular ratesfrom10 to 105sec
-1 andwefind in •ualita-
crack under an arbitrary state of stress and hypoth- tive agreement with experiments, that the stress is
esize a law relating mean strain to crack opening. about ten-fold higher at high strain rates. nuanti-
Two cases must be distinguished. In the first the tative comparisons would not be •ossible without a
normal component of traction on the crack is tensile series of very careful experiments on materials for
and the crack opens. In the second the normal com- which the flaw distribution is known.

ponent is compressive and the crack displacement in- STATISTICAL CRACK •IECHANICS
volves relative sliding of the crack faces without
In this section we develop criteria for the
opening. Different expressions for the strain ener-
stability of penny-shaped cracks and analytical ex-
gy are obtained for the two cases. This leads to
pressions for the strain resulting from an arbitrary
different stability conditions for open and closed
distribution of penny-shaped cracks. As a starting
cracks. These stability criteria are essentially
extensions of Griffith's criterion to the case of a •oint we consider the solution for a single penny-
shaped crack. The analytic solution leads to expres-
penny-shapedcrack under an arbitrary state of
stress. •,•hen the crack is unstable it is allowed to
sions for strain energy which can be used to deter-
mine crack stability. Then, by superposition we ob-
propagate at a constant speed. For oil shale, the
tain formulas for the average strain due to crack
speed should perhaps depend on crack orientation,
opening and shear. Finally, we discuss the statis-
but experimental data on this question is not avail-
tics of cracks and an expression for the number den-
able. The theory also leads to expressions for the
sity of inactive cracks is derived.
macroscopic strain due to penny shaped cracks, one
for the case of a closed crack, and the other for Theory of Penny-ShapedCracks
Keer (1966) showedhow the opening and shearing
open cracks.
The stress-strain relation we obtain can be ap- of a penny-shapedcrack in a state of biaxial loading
can be determined, and used his results to analyze
plied to any distribution of cracks, but for the pre-
the resultant release of strain energy, arriving at a
sent we assume that it is initially exponential in
crack size, as suggested by the data of McHugh and generalization of Griffith's stability criterion. In
this section we show how those results can be further
Keogh(1979). However, observation of rocks sug-
gests that intersection of cracks may play an imper-
generalized to an arbitrary, three-dimensional, state
of stress. Keer begins by analyzin• the boundary-
tant role. Our hypothesis is that if it were not
value problem for a penny-shaped crack in a biaxial
for intersections, loaded samples of rock would in-
state of stress. The normal to the crack plane is
variably be split in half by the most unstable crack
assumed to lie in the plane of the two principal
as soon as it becomes unstable. In fact, in triax-
stresses. In a coordinate system in which the crack

391
normal is taken as the z-axis, there are three com- Since shear stresses in the x-z plane do not cause

of stress, Ozz,Ozx, and Oxx,where

ponents x is displacements in the y direction• the shear strain
taken as any radial line. Keer solves for the dis- energy is
placement of the crack for each of the stresses sep-
arately,
a)
and concludes that:
the normal component,ozz , opens the crack, but
_16
1-v
Ws 3E2-v2c3[Ox2
z+ o2) yz (5)

Downloaded from http://onepetro.org/ARMAUSRMS/proceedings-pdf/ARMA80/All-ARMA80/ARMA-80-0390/2042003/arma-80-0390.pdf/1 by King Fahd University of Petroleum & Minerals user on 05 May 2024
produces no relative shearing of the faces. and this quantity is independent of the orientation
of the x and z axes, i.e.
b) the shearstress,Ozx,resultsin a displacement
in the plane of the crack and parallel to the x-
axis, but in no crack opening. 0xz + = O•'<•.Z
•. q- 0 •Z (6)
yz
c) the crack is not perturbed by the in-plane
Combininq the expression for Ws with the strain ener-
stress, Oxx.
He then determines the strain energy release due to gy associated with tension, the total strain energy
the crack from the formula is

W=• dAzz(Uz
- +_].
Ozx(U
X u•)
, (1)
Because ozz causes no surface displacement in the x provided oZZ is tensile. If oZZ represents a com-
direction, and o
ZX,
causes no displacement in the z pressive stress, then the corresponding strain energy
direction, the associated strain energies can be su- vanishes and the strain energy consists entirely of
perposed. Keer's result for the strain energy can shear strain contributions.
also be obtained simply by combining the strain ener- If we write the magnitude squared of the trac-
gy for a penny-shaped crack in shear due to Segedin tion vector as

(1950)

--
16 1-•2 3 2 A= •ijnjoiknk (8)
Ws 3E2-• c Oxy (2)
and the normal component of traction on a crack as
and the strain energy for a penny-shaped crack in
tension originally due to Sack (1946) and later re- B : o..n.n. (9)
1,1 1 J
viewed in detail by Sneddon(1969)
then this result makes it possible to generalize the

W 8 (1-•2)C3Ozz
t : •-• 2 (3) Griffith stability criterion to

Much of the analysis presented by Keer is not essen- A< •B2/2+ C (10)
tial to his main conclusion, which can be obtained
by superposition of these energy terms. Such a pro- for open cracks, where
cedure can be used to obtain the three-dimensional

result by considering a superposition of tension and C= •E¾(2-•)/4c(1-•

2) (11)
two orthogonal shears. We take axes such that z' is
orthogonal to the crack plane and x' and y' are axes When unstable, the cracks are assumed to grow at a
in the crack plane. We can then select x and y axes uniform rate which is typically a third the speed of
oriented so that the state of stress has no shear longitudinal waves, according to data collected by
component in the x-y plane. In this system Stroh (1957). The analysis for closed cracks, for
which B < O, is more complicated. Here also we

Ozz: Oz.z. (4a) merely give the result, which is that cracks are
stable if the radius satisfies

Oxz
: Ox•y.
cos(•
+ •y.z. sin(• (4b)
c < •¾tJ(2-•)/(T + •B)(T + 3•B)(1-•) (12)
•yz= - øx'y'sin(•+ o'..
yz cos(• (4c)

392
where
T=•/A]B
2istheshear
and• isthecoefficient
of friction.

In Fig. 1 we illustrate a crack with unit nor-

mal • containedin a large volumeof measure
V.

Downloaded from http://onepetro.org/ARMAUSRMS/proceedings-pdf/ARMA80/All-ARMA80/ARMA-80-0390/2042003/arma-80-0390.pdf/1 by King Fahd University of Petroleum & Minerals user on 05 May 2024
Fig. 2 - Illustration of howcrack strain can develop
82 even when volume changes cancel.

Y Then•ll = - •22' andthe net volume

though there is an overall
change
is zero,
shear. In a coordinate
system at 45 ø to the edges of the figure
Fig. 1 - Illustration of the geometry involved in
estimating crack strain.
cf2: •1 (ell _ c22
)

This shearing component can occur as the result of

opposing openings, and must be accounted for to ob-
Wedenote the crack opening by w, and the displace-
4 tain tensor behavior of the strain. It is for this
mentof materialbythecrackis then•A wwhere
A c ,
reason that the strain tensor is written in the form
is the crack area and we assume an elliptical cross
section. Now, the z componentof crack opening is given above. In the volume V we imagine a large num-
ber of cracks of varying size and orientation, and
wn
3 andthiscanbeconsidered
to actonanareaAcn
3. the net effect is the sum over the effects of the in-
We'estimate the displacment of material across the
4 2 dividual cracks. I•e denote by m the number density
surface
S3 to be•Acwn 3, andin general
thedis- of cracks, so that
placement
of materialacrosstheit• surface
to be

vi = •4 Acw(n
i )2 5c

The associated strain is simply denotes the number of cracks per unit volume in the
range of sizes (c, c + 2c) with normals in the range
•ii = vi/Si•i =•4Acw(ni)2/V of solid angles L•. If we consider the volume V to
be small enough that the distribution of cracks is
Now, if this concept is to be useful, these strains uniform, but larae enough to contain a significant
must be components of a tensor, which can be written number of cracks, then the total strain can be ex-
as pressed as
oo

•ij = •4Acwni
nj/V cij d•i dc•,c 3o•nI 3
To illustrate the need for a tensor formulation con- where we have written the crack opening as cc, with
sider the effect of two perpendicular cracks one of c independent of crack orientation or size. Now, the
which is opening with volume displacement Av and the crack opening is qiven by Sneddon(1969) as
other closing with volume displacement - •v so that
2 1-¾
the net effect cancels, as illustrated in Fig. 2. u = -•a..n.n.c (14)
z T• U •J I j

393
where o..n.n. is the componentof stress perpendicu- in the coordinate
system
withn3 = l, n1 = n2 = 0
lar to the crack surface, assumed tensile. Then

2 1-•
m-
• t•
0..
1j
n.n.
1 j
(15) TST
i s 2 + 023
i = ø13 2 ' (21)

Downloaded from http://onepetro.org/ARMAUSRMS/proceedings-pdf/ARMA80/All-ARMA80/ARMA-80-0390/2042003/arma-80-0390.pdf/1 by King Fahd University of Petroleum & Minerals user on 05 May 2024
and the integral can be written and this is just the quantity whose square root ap-

pearsin the preceeding

expression
for ut. Sincethe
•o.
lj =•oOk•
• d•nkn•n
i j •-•
2•lc3dc (16) stress axes are not usually aligned with the crack
axes, for our purpose the more general expression
where

T•T•
• 1 =0ijnj oiknk- ( njnk)2 (22)
•o : 8(1-u)/3• (17)
is more useful, and we will find it convenient to
andthe superscripton •.•J denotesthe contribution write

from crack opening.

In addition to the strains resulting from crack D:•i ls i (23)
opening, there is a significant contribution from
shearing in the plane of the cracks. In a coordinate for the magnitude of shearing stress. The direction
system in which the x and y axes are selected so that of•tangential traction we denote by t.1 with
oxy = 0 the tangential displacement of the crack sur-
face due to the shearing stress oxz is given by t.1 : TS./D
1
(24)
Segedin (1950) as
To analyze the effect of crack shear on the far-
4(1-•)
Ux: •(2-•) øxz
•-- •/c
2- x2- y2 field strain we considered the sum of two displace-
ments whose net result is a shear strain without vol-
and there is an analogous expression for u . Then ume change. In Figure 3 the strain in the n-t plane
Y
the tangential displacement due to the combined is shownwith the vertical displacement given by v
shears is and the horizontal displacement by u.

ut- s(2-u)tJ •c2-x

-y
The direction of displacement in the plane of
the crack is the direction of the resultant shear.
This direction is given by subtracting out from the
total traction on the crack surface

T.1 = o..n. (18)

1J J Fig. 3 - Illustration of the macroscopic shear due to
tangential displacement of the crack surface.
the normal componentof traction
Denoting
byLt the (tangential)
crackdisplacement,
we may write
Tinink = oi.n.n.n.
J J 1 K ' (lg)
The resulting vector can be written
u = •t sine , v: •t cosO

The vertical displacement is, in the same spirit as

T•• = c•ij .nj - ni ojk n.n
j k
(20) before,

and it is readily shownthat this is perpendicular to

the normal. It is also straightforward to show that, 4 (•c2 sine)
T (•t cose)

394
for a single crack. This may be written in terms of
the componentsof the tangential and normal unit
Bijkg bijkg ac
vectors as
with

3•c2ntt3n3 bijkg= nink6j•

+ njnk6ig
- 2ninjnkn
• (29)

Downloaded from http://onepetro.org/ARMAUSRMS/proceedings-pdf/ARMA80/All-ARMA80/ARMA-80-0390/2042003/arma-80-0390.pdf/1 by King Fahd University of Petroleum & Minerals user on 05 May 2024
wheret 3 = sine, n3 = cose, and the vertical strain Heremoandms denotethe densitiesof openandshear
then takes the form (i.e., closed) cracks. Assumingsuperposition to
hold, we may write the total strain as
4 2
•33 = • • c At t 3 n3/g3A
3
•ij = (Aijk%
+ Bijkg+ Cijk%)økg (30)
It is natural to generalize this by expressing the
strain tensor as withCijk•thecompliance
of thematrix
material
in
the absence of cracks. In fact, in SCMwe use a rate
form of this relation, but the physical interpreta-
•s.
lj =•2•c2/•t(tinj
+njti)/g•A
•3 tion is virtually the same.

As before, we must integrate over the population of Statistics of Cracks

cracks and orientations, as well as the volume. Then It is often assumed that the number density of
cracks is exponential in character. This assumption
is attractive for three reasons; it is supported by
laboratory and field statistics, especially the SRI
Now, using Segedin's result for shearing displace-
group (McHugh and Keogh,1979); it leads to very
ment, which we write as
convenient growth laws (Dienes, 1978); and it is
possible to makeheuristic statistical arguments
4(1-•) D (24)
At : •(2-•) • c leading to this kind of statistics, as shownby
Dienes, 1979. In the current theory we extend this
and Eqs. 20, 23 and 24 this becomes concept by assumingthat for each orientation the
cracks are exponentially distributed. If we assume
lj • d•l(njnkOik
+ ni no.
kjk that all cracksgrowat a rate • then it is straight-
co
forward to show that

- 2ninjnknCøkC)•o
dcc3
a?-!• ac
(25)

where
a_E•
at
+ • ac
an_ am
at
(31)

where the right hand side represents the rate at

•s : 81-• 1
3 2-• • '
(26) which active cracks are transformed into inactive
cracks. For constant•, andunderthe assumption
Here the superscript s is used to emphasize that this that
is the strain due to shearing of cracks. It is in-
terestingto notethat •.11 : O, confirming
that crack •l = kn (32)
shearing does not contribute to overall dilatation,
as we know on physical grounds that it should not. the initial value problem is solved by
These results can be expressed in a compact no-
tation if we write
n=r•00
e-(c-•t)/E-kt (33)

Aijk•= •ø•F,•
'n.n.nkn
• • • d•l[
.
dcc ac 3amø
(27) where E is the initial
valid only for c > •t:
mean radius. This solution
if this doesnot hold, • = O.
is

and
Dienes (1978) provides an estimate for k for an iso-
tropic distribution, and Dienes(1979) describesthe

395
underlying theory for computing k in the case of ties of active and inactive cracks change with time,
cracks bedded in a plane. and show that at late times the distribution of in-

However, observations of tile exposed surfaces of active cracks approaches a definite limit. It is
rocks shows that cracks are often stopped by other useful to consider the ground plane of the independ-
cracks in what we may loosely call a T configuration. ent variables c and t illustrated in Fig. 4.
In other instances the cracks may cross to form an X

Downloaded from http://onepetro.org/ARMAUSRMS/proceedings-pdf/ARMA80/All-ARMA80/ARMA-80-0390/2042003/arma-80-0390.pdf/1 by King Fahd University of Petroleum & Minerals user on 05 May 2024
configuration. In this analysis, we do not discuss
the mechanics of pairs of growing cracks, which con-
stitutes a very difficult problem, but merely consi-
c-(•t)O
Solutions
basedon•
/:•;t
der the problem from a geometric and statistical initialcrack •
point of view.
to that in the kinetic
Tile situation is somewhatanalogous
theory of gases. There one
density
/ .
/ c-•t(0
considers collisions of spheres and determines the / Solutions
basedon
frequency of collision, hence the mean free path, by
simply arguments. From them many useful
are obtained without any knowledge of the detailed
results
nucleation
of
cracks
mechanics of collision. Here, we consider the ex- / ,•,•.Nucleation
t
tent to which cracks can grow before intersection
from a statistical point of view. The arõunlents are
rough, but results concerning crack intersections Fig. 4 - Illustration of the ground plane (c-t) showing
the two regions into which cracks are divided.
are necessary if we are to characterize rubblization
processes by numerical simulation, for without them If initially the distribution is exponential,
it would be necessary to let cracks grow without at later timesit is exponential
onlyfor • > ct.
bound in the analysis, and this would lead to unreal- For • < ct, •: O, since eventhe smallest cracks
istically large numbers of fines when a great many will havegrownto a size •t at time t. Thus,the
cracks intersect in a given volume. distributionof inactive cracksis givenfor t < c/•
In the course of deformation, it appears likely by
that a mass of rock will contain some cracks that
.t Nk
grow and somethat don't.
inactive cracks, respectively.
These, we call active and
Active cracks are
•t=Jokl•
dt=__o_
(e•t
_1)e-c/E (34)
those that are unstable. Inactive cracks are those andfor t > c/• by
that are stable because they have intersected several
other cracks and their
stress concentration.
edges are no longer regions of
The number density of inactive
m=
jo
'c/• •o
k kc/•
e-C/E
k;• dt : • (e- - ). (35)

cracks, denoted by m(c,•), represents the numberof The reduced modulus theory requires the mean
cracks per unit volume per sterradian whose radii ex- cube of crack sizes. The mean cube for inactive
ceed c. Thus, m(c,•)A• represents the numberof cracks is defined by
cracks per unit volume whose radii exceed c and whose
normals lie
• + •),
in the range given by (e, e + &e), (½,
where e and • denote polar coordinates. For
• : •oc3mdc (36)
brevity, we may write the solid angle as •œ = and we readily find
sine •e • and we use • to denote symbolically the
direction
9.
represented in polar coordinates by 0 and
If E dependson crack orientation, then the as-
•=Nøk
!6(E4
-E
4)e-kt[•3(E
•

c•, -a) + (37)

sumption that the total distribution

•) is exponential no longer holds in general.
(integrated over
A pro-
+3•2(E2-•2)+6c(E3-a3
cedure for determining • whenthe distribution is where we have set

isotropic is described by Dienes (1979). In the re- ^

mainder of this section, we examine how the densi- C : •t , a : •/k : • m/2•

2 E2 N o
(38)

396
F will, in general, depend on orientation.
The distribution of inactive cracks is important
because all the active cracks eventually become inac-
tive. The distribution of inactive cracks approaches
a definite limit with mean size that is given by

Downloaded from http://onepetro.org/ARMAUSRMS/proceedings-pdf/ARMA80/All-ARMA80/ARMA-80-0390/2042003/arma-80-0390.pdf/1 by King Fahd University of Petroleum & Minerals user on 05 May 2024
•=•/k
+•:E12 3•No+•
•2E3 (39) 5 5 8
STRAIN RATE EFFECT

Strain rate effects in rocks appear to be very

important. For example, Grady and Kipp (1979) pre-
sent data showing an increase in fracture stress by
an order of magnitude when the strain rate increases
from
lOs-1 to 104s-1 In thissection
wereport
on
the result of varying the strain rate input to SCM
using the theory outlined in the preceeding sections.
Though in general the strain rate input to SCMmay
Fig. 5 - The 9 regions on the unit sphere into which
vary in an arbitrary way, in the calculations re-
the crack normals may fall. The shaded
ported here it is taken to be constant and similar to regions are equivalent to the opposite, un-
shaded, region.
that experienced by a triaxial test specimen when
there is no change in sample density, i.e. Each segment has a similar segment on the oppo-
site side, but since a 180ø rotation of a penny-

dll = - 2D , d22= d33= D shaped crack has no effect, it is only necessary to

consider points on half the sphere. All segments on
with the other components of strain vanishing. SCM the surface have the same area. Three segments are
makes use of the physics described in %he preceeding spherical caps lying symmetrically with respect to
section, though the detailed formulation is somewhat the coordinate axes. The remaining 6 segments are
complicated because of the variety of logical paths, bounded by great circles passing through one axis
the need to invert the stress-strain relations to ob- and bisecting the other two. Though the angular re-
tain stress from the given strain rate, and the need solution is not very fine, our first indications are
to express the constitutive relations described in that it is good enough to give stresses within a few
the preceeding section in rate form. percent of the limiting (infinite resolution) case.
In computing the surface integrals required in For these calculations we assume an isotropic
the previous section and shown in Eqs. 27 and 28 it distribution of cracks with a mean radius of 1 cm, a

is convenient to divide the unit sphere upon which fracture velocity of 400 m/s and an initial crack
the ends of the crack normals lie into 9 segments, density of 0.4• cracks/cm
3. Thematerial,typical
as illustrated in Fig. 5. of oil shale,is takento havea density
of 2 g/cm
3,
a shear modulus of 10 GPA and a Poisson's ratio of
0.2. Thesurface
energy
wastaken
as1J/m
2 andthe
coefficient of friction and cohesion as O. In spite
of these patently round and somewhatarbitrary num-
bers, the results of these calculations, shown in
Fig. 6, are in fair general agreement with the data
compiled Grady and Kipp. These stresses are taken at
a strain of 1%, since there is no definite failure
point, bu• the assumption that failure always occurs
at the same strain is not inconsistent with the re-

sults of laboratory data.

397
 io,ooo I F I
' I Dieres, J. K., "A Statistical Theory of Fragmen-
tation", 19th U.S. Symposium on Rock Mechanics,
Stateline, Nev., May 1-3, 1978.
5,000
Dieres, J. K., "On the Inference of Crack Statis-
2,000
•--
tics from Observations on an Outcropping", 20th
1,000 --
U.S. Symposiumon Rock Mechanics, Austin, Texas,
June 4-6, 1979.

Downloaded from http://onepetro.org/ARMAUSRMS/proceedings-pdf/ARMA80/All-ARMA80/ARMA-80-0390/2042003/arma-80-0390.pdf/1 by King Fahd University of Petroleum & Minerals user on 05 May 2024
Garbin, H. D. and Knopoff, L., "The Shear Modulus
of a Material Permeated by a RandomDistribution
of Free Circular Cracks", Quart. Appl. Math. Vol.
33, No. 3, 296-300, Oct. 1975.

Grady, D. E. and Kipp, M. E., "Oil Shale Fracture

and Fragmentation at Higher Rates of Loading",
20th U.S. Symposiumon Rock Mechanics, Austin,

50
•/ Texas, June 4-6, 1979.

Keer, L. M., "A Note on Shear and Combined Load-

ing for a Penny-ShapedCrack," J. Mech. Physics
o 04 0.8 12 2.0 24 28
Solids, 1•4, 1, (1966).
Stress, K•lob(3rs
McHugh, S. L. and Keough, D. D., "An Analysis of
the Effect of Input Stress Pulse Shape on the
Fracture of Devonian Shale", 20th U.S. Symposium
on Rock Mechanics, June 4-6, 1979, Austin, Texas.
Fig. 6 - The relation between strain rate and the
stress at 1% strain as determined by SCM. Piau, M., "Attenuation of a Plane Compressional
Wave by a RandomDistribution of Thin Circular
The asymptotic stresses at high and low strain Cracks", Int. J. Engng. Sci, Vel. 17, pp. 151-
rates are a simple consequenceof the assumptions. 167, 1979.
At very low strain rates the cracks appear to grow
Sack, R. A., "Extension of Griffith's Theory of
rapidly, and change from active to inactive in a Rupture to Three Dimensions", Proc. Phys. Soc. 58
(1946).
relatively short time. On the other hand, at high
strain rates there is not enough time for the cracks Segedin, C. M., "Note on a Penny-ShapedCrack Un-
der Shear", Proc. Camb. Phil. Sec. 47, 396
to grow significantly. As a result the stress is
(1950).
simply that predicted by elastic theory, which is in-
dependent of strain rate. Each family ef cracks is Sheddon, I. N. and Lowergrub, M., Crack Problems
in the Classical Theory of Elasticity, Wiley,
allowed to grow independently of the others, depend- 1969).
ing on its stability.
Stroh, A. N., "A Theory of the Fracture of
REFERENCES Metals" in Advancesin Physics Vol VI, 1957
Taylor and Francis, LTD, London.
Bamberger, Y., Cannard, G. and Marigo, J.-J.,
"Microfissuration du Beton et Propagation d'Ondes
tr•s•ores,", to appear
in theproceedings
of
ne bo//oque International du C.N.R.S. n 295,
EuromechColloquium 115, Villard de Lans, 19-22
June, 1979.

398
Figure 1 Figure 2

S3
I
I

Downloaded from http://onepetro.org/ARMAUSRMS/proceedings-pdf/ARMA80/All-ARMA80/ARMA-80-0390/2042003/arma-80-0390.pdf/1 by King Fahd University of Petroleum & Minerals user on 05 May 2024
I

S2

Y
I _j

Fiqure 4

Fi qure 3

c-6t)O
Solutions
basedon
///•:6t
initial
crack
density
• c-•t<0
J Solutions
basedon

• nucleation
of
cracks
/ •,..Nucleation
t

Fi qure 5 Fiqure 6

IO,OOO
[ 1 I I I
5,000

2,000
I,OOO
I

• 200

100
• • ß
20,

100 0.4 0.8 12 116 2J4 2.8

Stress• Kilobars

399