1965 - Cook - The Failure of Rock
1965 - Cook - The Failure of Rock
Transvaal and Orange Free State Chamber of Mines, Johannesburg, South Africa
(Received 6 July 1965)
1. I N T R O D U C T I O N
GRIFFITH (1921) originally defined the low tensile strength o f brittle m a t e r i a l s in terms o f a n
inequality between the rate at which elastic energy is released a n d the rate at which surface
energy is a b s o r b e d , as a flaw or crack extends within the material. Subsequent extensions o f
Griffith's t h e o r y to c o n d i t i o n s o f compressive stress, including triaxial stresses, have usually
defined the strength o f brittle materials in terms o f a critical tensile stress c o n c e n t r a t e d n e a r
the edge o f a flaw o r c r a c k (OROWAN, [1]). U s i n g this criterion, MCCLINTOCK a n d WALSH
[2], have allowed for the closure o f Griitith cracks in c o m p r e s s i o n , a n d their results have
been shown by BRACE [3] and HOEK ( C o o k [4]) to p r o v i d e an a d e q u a t e d e s c r i p t i o n o f the
strength o f h a r d rocks.
The strength o f r o c k is usually d e t e r m i n e d by using a j a c k to a p p l y an increasing axial
l o a d to a free or confined specimen, a n d n o t i n g the stress at which the j a c k - s p e c i m e n
system becomes unstable. The system b e c o m e s u n s t a b l e at some negative slope o f the
strain-stress characteristic, d e t e r m i n e d b y the resilience o f the jack. Owing to the relatively
high resilience o f m o s t jacks, instability usually occurs at low negative slopes very close
to the p e a k o f the strain-stress characteristic o f the rock. In the case o f h a r d rocks, the
p e a k a l m o s t coincides with the t r a n s i t i o n f r o m elastic to non-elastic behaviour. Thus,
virtually no i n f o r m a t i o n concerning the non-elastic b e h a v i o u r o f h a r d rock is available.
W h i l e it is i m p o r t a n t to k n o w the strength o f rock, it is often o f equal i m p o r t a n c e , in
studies o f d y n a m i c p r o b l e m s such as r o c k b u r s t s a n d rockdrilling, to k n o w h o w energy is
a b s o r b e d in the process o f failure. The w o r k r e p o r t e d in this p a p e r considers the energy
changes associated with the extension o f Griffith cracks, and hence defines the non-elastic
389
390 N . G . w . COOK
behaviour of hard rocks. For simple cracks in tension, the results are coincident with the
Griffith locus derived by BERRY [5] in considering the kinetics of tensile and cleavage
fracture. A similar locus is obtained for closed cracks in compression, and the corres-
ponding strength criterion is found to be coincident with the M o h r - C o u l o m b criterion.
The units of measurement used in this paper are centimetres dynes and ergs, where 106
dynes/cm 2 14.5 lb/inL
2. T H E O R Y
2.1 Tension
As an introduction to this analysis, consider first the classic case of a single crack of
length 2c, in the form of a degenerate ellipse in plane strain normal to a tensile stress,
--e, applied externally to a specimen of length l, width b, and unit thickness (Fig. 1).
_ // /
ifi ,-/ /
HI ," / ./
~; / c=l'O0
Strr~ln 10 )
FIG. I. The Griffith locus defining the non-elastic behaviour in tension o f a material in which the surface
energy ~ -- 10's ergs/cm 2, and the modulus o f rigidity G = 5 x 10,1 dynes/crn 2.
Imagine that the stress is maintained constant, initially with the crack surfaces held
closed by some mechanism. If the crack surfaces are then slowly released, work, We, will
be done on the closure mechanism and additional strain energy, W.~, will be stored around
the crack.
We =- Ws = zr/2 (1 -- v) (r2/G e z ( O R O W A N [1])
where G -- modulus of rigidity
v =- Poisson's ratio.
The strain energy in the absence of the crack would be
o.2
S.E. = 4-(i + v) G (bl).
THE FAILURE OF ROCK 391
If --E is the strain associated with the externally applied stress --a in the presence of
the crack, then the total work done by the application of - - a is
Equation (1) defines linear strain-stress paths for different crack sizes and densities
prior to failure, and equation (3), which is the plane strain equivalent of BERRY'S [5] Griffith
locus, defines that region of the strain-stress plane within which equation (1) is valid.
2C
392 N.G.W. COOK
The Griffith locus, therefore, defines the strain-stress path along which a material fails
due to the concurrent extension of a number of identical cracks. Figure 1 shows equations
(1) and (3) in the ~--~ plane for different crack sizes with typical values assigned to
a, G, n. Once failure starts, the path followed in the E-a plane depends upon the external
characteristics of the applied stress. For example, if ~ is applied in such a way that it is
constant, irrespective of displacement, the path after failure will be parallel to the ~-axis
and, if the boundaries along which a is applied are held fixed at the instant of failure, the
path will be parallel to the a-axis. In practice most applied stresses have characteristics
lying between these two extremes.
The energy absorbed by failure, or crack extension, is proportional to the area bounded
by equation (3) and equation (1) with the appropriate initial and final crack lengths.
Therefore, the energy released at failure is proportional to the difference between this area
and the energy input per unit volume, that is, the area between the ~-axis and the path of
the externally applied stress. Unstable fracture may be defined as failure accompanied by
energy release. Paths of external stresses giving rise to failure and fracture, respectively,
are indicated by arrows 1 and 2 in Fig. 1. In the case of an external stress applied with
fixed boundaries fracture can only occur if dE/da > 0. Therefore, only materials in which
nc2a < [1/4rr(1 -- v)] are intrinsically brittle and liable to spontaneous fracture without the
external addition of energy during the process. Otherwise, it is possible to follow the path
defined by equation (3) in a quasi-static manner, during which failure will occur by stable
crack extension until the crack length is equal to the width, b.
2.2 Compression
In the second case consider a crack of length 2c, in the form of a degenerate ellipse in
plane strain inclined at some angle, 0, to the direction of a uniaxial compressive stress
applied to a specimen of length l, width b and unit thickness (Fig. 2). The crack is subject to
compressive stresses
~rc ~[1 +~ ~ cos 20] -- a cos 2 0
parallel to its surfaces. Because the crack is closed, a frictional force t~n will arise across the
crack surfaces, where ~ is the coefficient of sliding friction of these surfaces.
Imagine that the stress is maintained constant, initially with the crack surfaces held
closed by some mechanism. If the crack surfaces are then slowly released, they will slide
past one another, provided that ~- > ~an. In this process, work, We, will be done on the
closure mechanism and work, Wj, will be done against the friction forces, while additional
strain energy, Ws, will be stored around the crack when ~ = 0:
We : W8 : ~'/4 (1 - - v) 2c2/G STARR [7].
Therefore, when Izan > 0
T h e s t r a i n e n e r g y i n the a b s e n c e o f t h e c r a c k w o u l d be
oz
S.E. - - 4(1 + v) G (bl).
c~O
~tru~n x lu
FIo. 2. The Griffith loci defining the non-elastic behaviour in uniaxial and triaxial compression of a material
in which the surface energy a = 10a ergs/cm~, the modulus of rigidity G = 5 1011 dynes/cm2 and the
coeMcient of internal friction t~ = 1.0.
W . D . = ~E/2 (bI).
Now,
W . D . = S.E. + Ws + Wf/2 (see A p p e n d i x )
from which
d We (-r --/~crn)2
dc - - 7r/2 (1 - - v) G c.
394 N.G.W. COOK
2 (I - - v) G c = 4a
The latter expression is the same for states of triaxial compression and coincides with a
Mohr-Coulomb envelope in the a-~ plane.
Substituting the critical value of e from equation (6) into equation (5) yields
[d~ld~er = 1 -- 37r (1__ v2) {(sin 2 20/4)2(1+q-tzv)sinzG 0 (sin 20/2) -- 2/~z sin 4 0} neZa (8)
L ~
Equation (5) defines linear strain-stress paths prior to failure, and equation (7), the
Griffith locus, defines that region of the strain-stress plane within which equation (5) is
valid. These equations are shown in Fig. 2 for typical values of a, G, n and/z, with 0 chosen
so that the expression (~- -- tZ~n) is a maximum. Equation (6) showed that crack extension
begins when (~- /z~n)reaches a critical value. It can be shown that this expression, which
is a function of 0, has a well-defined maximum about some optimum value of 0 when the
coefficient of friction is near unity. This justifies the implicit assumption that all the active
cracks are inclined at or near the optimum value of 0. Figure 2 shows that the Griffith locus
in uniaxial compression is similar to that in tension, Fig. 1, and the same considerations
apply concerning failure and the energy released at fracture.
Conditions of triaxial stress can be analysed in the same way, assuming that the crack
is parallel to the direction of the intermediate compressive stress and using the relations
f f l -}- 0"3 ~1 - - o'3
~n-- 2 2 cos 20
2.3 Micro-seismicity
Many rocks are heterogeneous aggregates of different parts which may differ in their
mechanical properties. Assume that the major heterogeneities can be considered as elastic
inclusions in an approximately homogeneous elastic surround. When such a rock is com-
pressed, the stress in the inclusion will be different from that in the surround, and the
inclusion may fail in a stable or unstable fashion before the rock as a whole fails, depending
upon the relative mechanical properties of the inclusion and the surround. Unstable
fracture of these inclusions would give rise to micro-seismic activity which is often observed
in rock samples under compression. In this part these considerations are analysed for the
case of rock in uniaxial compression.
MUSKELISHVILI [9] has analysed the effects of circular elastic inclusions in an elastic
surround. He has shown that, for uniaxial compression in plane strain, the ratio between
the stress in the inclusion and that in the surround is given by
3M0 (9)
cr/~ = 2M0 q- M
ba2Gn
~r = ~ d + ~3 (lO)
where
a = 2(1 -+- v)
32 {(sinz 20/4) + t~ sin 2 0 (sin 20/2) -- 2/~z sin a 0 }
b--
~r(1 -- v) {(sin 20/2) -- tz sin 2 0}4
and the locus can be expressed in terms of the effective Young's modulus, M, and the
stress at which failure starts, ~ as
cr aGo~
M = --~er = a4 "}- aba2G2n" (11)
Considering the Griffith locus in the M-or plane, as illustrated by Fig. 3(c), and bearing
in mind that M and cr can only decrease during fracture, it will be seen that the criterion
for unstable fracture of the inclusion can be expressed as
If the inequality (13) is not satisfied but equation (14) and inequality (15) are satisfied,
failure of the inclusion will be stable.
The left- and right-hand sides of inequality (13) are plotted in Fig. 3(a), against the
13- u.s~ ."~-~L..
t," ~5~-
12 ' ~ *~,E~,
10
" //i
2 3 4 5 6
Indusion stress do dvn{zS/cm ~ x 10 -a
M-s /
M:I
2
,, -. -.
' ~o 3t
Fro. 3(a). The conditions for unstable inclusion fracture in uniaxial compression of a heterogeneous material
in which the surface energy a = 10 8 ergs/cm ~, the modulus of rigidity G = 5 1011 dynes/cm 2, the coefficient
of internal friction tz = 1, and the difference between the inclusions and the surround lies only in the crack
density n.
(b). The relations between inclusion stress and surround stress.
(c). The Griffith loci of surrounds with different crack densities.
THE FAILURE OF ROCK 397
inclusion stress, a0, at which failure starts, for two inclusion crack densities and surround
moduli using typical values of a, G, n and/z. These values are assumed to be the same for
both the surround and the inclusions, so that only the crack density varies. Figure 3(a)
refers only to the start of inclusion failure, as equations (9) and (11) were used to eliminate
M0 and o from the right-hand side of inequality (13). As soon as unstable fracture of the
inclusion starts the actual stress on the inclusion and the stress defining its locus will no
longer be the same and the equality between ~0 in equations (9) and (11) ceases to hold.
The surround stress, ~r, corresponding to the inclusion stress, cro, at which failure starts,
is shown in Fig. 3(b), making use of equations (9) and (11). The Griffith loci for surrounds
with different crack densities are shown in Fig. 3(c). This can be used together with Fig. 3(b)
to determine whether the strength requirements, equations (14) and (15), are satisfied, that
is, whether M and ~ of the surround are above the hatching.
Using Fig. 3 to consider any case which allows prior unstable fracture of the inclusion,
it will be seen that, as the surround stress is increased inequality (13) becomes more gross
and successively stronger inclusions will fracture. At the same time the surround stress
approaches the Griffith locus at which the surround will fail. Thus, micro-seismic activity
in a rock containing many heterogeneous inclusions will tend to increase as the stress on
the rock is increased until ultimate failure of the rock as a whole occurs.
3. E X P E R I M E N T
3.1 Description
A Griffith locus to describe the non-elastic behaviour of hard rock was developed, based
on the assumption that a number of cracks, all of identical size, extend simultaneously.
The flaws in rock almost certainly consist of cracks of different sizes, which would begin to
extend at different stresses. In a completely homogeneous material the first crack to extend
might grow into a major fracture. The heterogeneous nature o f rock makes it quite likely
that this crack will extend only a short distance, before it is stopped by a region of greater
resistance to crack propagation. The crack will begin to extend again only at some higher
stress, after other cracks have already started to extend. Thus, it may be reasonable to
suppose that the non-elastic behaviour of a heterogeneous material, such as rock, can be
described by a synthesis of Griffith loci according to the distribution in the rock of cracks
of different sizes. The effect of such a synthesis would be to alter the transition between
elastic and non-elastic behaviour from a discontinuous to a continuous change. As the
Griffith locus arises from crack extension, the slope of the synthesized transition can be
expected to depend upon crack extension.
A series of uniaxial compression tests were conducted on rectangular prisms of Tennessee
marble, a relatively homogeneous hard rock, and St. Cloud granite, a relatively hetero-
geneous hard rock. These specimens were machined in a surface grinder to nominal dimen-
sions of 3.75 x 3.75 x 15 cm or 3.75 x 5.00 x 15 cm, care being taken to ensure that their
ends were parallel to within 0.001 cm. They were loaded parallel to their long axes, using
the device illustrated in Fig. 4, by means of an hydraulic compression testing machine. An
attempt to reduce end-effects to a minimum was made by placing a hard steel plate, having
the same cross-section as the prism, at each end of the rock specimen, and lubricating the
interface between this plate and the hard steel loading platten with molybdenum disulphide.
The object of these experiments was to study the relations between the following:
(i) The resilience of the loading system and the violence of fracture.
(ii) The slope of the strain-stress curve and the extension of cracks within the rock.
398 y.G.W. COOK
Steel tube
FIo. 4. A device for determining the strain-stress characteristics of rock specimens and for increasing the
stiffness of a compression testing machine.
3.2 Results
Theory suggests that energy, in addition to that stored as strain energy in the rock
specimen, is necessary to produce failure of the specimen. Provided that the loading is
done in a quasi-static manner, this additional energy can only derive from the energy
stored by the resilience of the testing machine. Accordingly, it should be possible to reduce
the violence of fracture by increasing the stiffness of the testing machine and, in the extreme,
prevent violent fracture altogether. Figure 5 shows the strain-stress characteristics obtained
by loading to fracture two similar specimens of Tennessee marble in a flexible and stiff
machine, respectively. The machine resilience is represented in this figure by the lines with
negative slopes --k, corresponding to resiliences of 2.7 x 1011 dynes/cm, and 13.5 x 1011
dynes/cm, respectively. It can be seen that fracture of the specimen occurred as soon as the
slope of the strain-stress curve (da/dE)/k > 1. The fracture which occurred during the test
with the flexible configuration was violent, and the resulting shock from the testing machine
could be felt extensively in the laboratory building. The fracture which occurred during the
test with the stiff configuration was almost non-violent and hardly produced a noticeable
shock. This suggests that almost all the energy stored by the resilience of the stiff machine
was absorbed in non-elastic deformation of the specimen, so that its Griffith locus must
have been close to the machine characteristic. Unfortunately it was not possible to increase
the stiffness of the machine further to any significant extent. A complete Griffith locus,
theoretically obtainable with an infinitely stiff machine, could not, therefore, be obtained.
If the slope of the strain-stress characteristic is due to crack extension this should be
detectable as a change in the tensile strength of the rock, which is theoretically sensitive
to crack size. Accordingly, a set of carefully matched specimens of Tennessee marble were
THE FAILURE OF ROCK 399
subjected to different complete load cycles, identified by the slope, d(r/dE, of the strain-
stress characteristic immediately before the load was released. Four cylindrical tensile test
specimens, 2.5 cm in diameter by 5.0 cm in length, with their axes perpendicular to the
long axis of the original prism, were afterwards diamond-drilled from each member of
this set. These cylinders were glued to steel end-caps, and their section was reduced by
'7
o 15 -k2
&o ~ . " -\
! s ~ _ _ \\\\\
0.5 1.0 1.5 2.0 2.5 3.0
Stroin x 10 3
FIG. 5. Strain-stress characteristics for Tennessee mazble obtained during compression tests to fracture in a
flexibleand stifftesting machine.
5 per cent over threequarters of their length. The end-caps were screwed onto long steel
rods 0.6 cm in diameter, to avoid applying bending loads to the specimens in the tensile
testing machine. Most of these tensile specimens broke across the reduced section with a
nearly plane cleavage to the axis. The tensile strengths of these specimens are plotted
in Fig. 6 against the slope of the strain-stress characteristic at the end of each cycle. Two
groups of tensile strength were obtained for each compression specimen. In every case, the
tensile specimens belonging to the stronger group came from the ends of the compression
specimens, which can be attributed to two factors. First, friction forces between the end-
plates and the compression specimen probably provided some measure of constraint at the
ends of the specimens, especially during non-elastic deformation. Second, the rigid end-
plates applied a constant strain condition across the end of the compression specimen,
so that if a weak region began to fail the stress on it dropped, whereas more freedom for
readjustment existed towards the centre of the specimen, where a weak region would
continue to fail. Nevertheless, a distinct linear relation appears to exist between the slope
of the strain-stress curve and the residual tensile strength, which can be attributed to crack
extension during the original loading. It is also interesting to note that while the tensile
strength of the specimens changed by an order of magnitude they were all sufficiently
strong to bear similar maximum loads.
In Fig. 7 the slope of the strain-stress curve of a Tennessee marble specimen is plotted
against the applied axial compressive stress for a complete stress cycle, followed by a
stress increasing to fracture. This figure emphasizes how the slope of the strain-stress
characteristic changes with applied stress, and shows how the slope changed suddenly
400 N. 6. w. c o o k
when the load was reversed at the end o f the first cycle. The friction forces across the crack
reverse when the direction of loading is reversed and are sufficient to prevent sliding for a
small load change. The slope at this point probably, therefore, represents the true Young's
modulus o f Tennessee marble, provided that no open cracks or pores are present at this
stress (WALSH [10]).
~gr, 7
C)
/
u~ .
5/
f-'
5 j/
o:J 5m" j
J
~' E n d cyhnders
Centre cylinders
d
dynes/Zorn ~ x 1 0 -'t
FIG. 6. The tensile strengths of specimens taken from Tennessee marble subjected to different load cycles
identified by d~r/dE at the end of each cycle.
10-
~p 5-
>(
~E
0
d dynes x 10 -8
cm 2
-5-
-10-
The strain-stress characteristics to fracture for Tennessee marble and St. Cloud granite
are shown in Figs. 8 and 9, respectively, together with records o f micro-seismic activity
drawn along the same strain base. These show that in the case o f Tennessee marble, which
is relatively homogeneous, the micro-seismic activity was limited to a single release at a
point, marked by a dip in the strain-stress characteristic, where an early macroscopic
fracture occurred. In the case o f St. Cloud granite, which is relatively heterogeneous,
increasing micro-seismic activity preceded ultimate fracture.
THE FAILURE OF ROCK 401
or
1 Stroin x2 10 3 :3
o
>~
.ta
J I i t
i i L i i i
0 1 2 3
Stroin x 10 s
FIG. 8. The strain-stress characteristics and the envelope of micro-seismic activity for Tennessee marble
loaded to fracture in compression.
4. C O N C L U S I O N
A Griffith locus describing the non-elastic behaviour of hard rock is derived according
to a particular model of crack extension. The general basis of the theory is that elastic
potential energy is converted to surface energy during the failure of rock, by an inefficient
conversion process in which energy is dissipated by friction forces. Even if the details of the
model prove to be incorrect the analysis is based upon such general energetic principles
that the concept of a Griffith locus to describe non-elastic behaviour will probably still be
valid.
The most important aspects of this concept are that amounts of energy, large in relation
to the stored elastic strain energy, are necessarily dissipated during the failure of rock,
and that failed rock retains strength and cohesion during extensive non-elastic deformation.
These factors are of vital importance in understanding any process concerning rock failure
such as rockdrilling or rockbursts. The results imply that in drilling it is necessary to
supply more energy to fracture and destroy the cohesion of the rock than to stress the rock
to the point of failure. Any solution to the problem of rockbursts will also probably depend
upon the fact that the vast amounts of energy which must be released by underground
excavation (Cook [I 1]) can be dissipated by a process of stable rock failure.
402 N.G.w. cook
'
1 2 3 4 5 6
StrC~ln 10 ~
m
2
u
0 1 2 3 4 5 6 7
Stvom x 10 3
FIG. 9. The strain-stress characteristics and the envelope of micro-seismic activity for St. Cloud granite
loaded to fracture in compression.
Acknowledgements--Most of the work reported in this paper was done at the School of Mineral and Metal-
lurgical Engineering of the University of Minnesota during 1964, with funds provided by the American
Petroleum Institute and the National Science Foundation. I am pleased to acknowledge the contribution
which Mr. S. CROUCH made to the experimental work.
REFERENCES
1. OROWAN E. Fracture and Strength of Solids: The Physics Society (London). Rep. Progr. Phys. 12,
185-232 (1949).
2. McCLINTOCK F. A. and WALSH J. B. Friction on Griffith Cracks in Rock under Pressure. Int. Congr.
Appl. Mech., Berkeley (1962).
3. BRACE W. F. An Extension of the Griffith Theory of Fracture to Rocks. J. geophys. Res. 65, 3477-3480
(1960).
4. Cook N. G. W. A Note on Rockbursts considered as a Problem of Stability. S.A.I.M.M. 65, (8) 437
(1965).
5. BERRY J. P. Some Kinetic Considerations of the Griffith Criterion for Fracture--1. Equations of
Motion at Constant Force. J. Mech. Phys. Solids 8, 194-206 (1960).
6. SALAMON M. D. G. Elastic Analysis of Displacements and Stresses Induced by Mining of Seam or
Reef Deposits--IV. Inclined Reef S.A.LM.M. 65, 319-338 (1964).
7. STARR A. T. Slip in a Crystal and Rupture in a Solid due to Shear. Proc. Cam& phil. Soc. 24, 489-500
(1928).
8. VON KARMAN T. Festigkeitsversucht unter albreitigem druk. Z. Vet. deutsch. Ing. 55, 1749-1757 (1911).
9. MUSKELISnVIU N. I. Some Basic Problems of the Mathematical Theory of Elasticity, translated by
J. R. M. Rodak, Noordhoff, Croningen, Netherlands, pp. 218-220 (1953).
10. WALSH J. B. The Effect of Cracks on the Compressibility of Rock. J. geophys. Res. 70, 399-411 (1965).
11. COOK N. G. W. The Seismic Location of Rockbursts. Rock Mechanics, Proceedings of the Fifth Rock
Mechanics Symposium, Pergamon Press (1963).
THE FAILURE OF ROCK 403
A P P E N D I X
Let~/be the mean shear displacement of the crack surfaces relative to one another per unit change in the
shear stresses across these surfaces. Refer to Fig. 10 where the shear stress across the crack surfaces is
plotted against the mean relative displacement of these surfaces. The work done as the closure mechanism
is released, and hence also the additional strain energy is
W = tic 72 if t~a. = 0
W = ~ 72 - - 3 c ~2~2
We = dc (7 -- t, an) 2.
dc may be found by equating W for the condition when/*an = 0 to STARR'S [7] solution of that problem
Thus,
,r 2
de7 3 = ~/4(1-- v ) ~ e 2
As a is increased from zero, in the presence of the crack, the surfaces displace linearly from 0 to ti(7 -- ~an)
and the work done against the friction forces is W~/2.
tD
~t
o a(r-~.a.) ~r
Mean displacemc~nt of crack
Fla. 10. Shear stress and displacement of a crack surface.