Sunteți pe pagina 1din 7

Int. J. Rock Mech. Min. Sci. & Geomech. Abstr. Vol. 14, pp. 209-215. Pergamon Press 1977.

Printed in Great Britain

Failure of Granite under Compression


WALTER JANACH* Strength data for Westerly granite under triaxial compression lap to a confining pressure of 800 MN/m 2 (8 kbar) reveal a quadratic relationship between maximum stress and confining pressure. Neither the Coulomb nor the Griffith failure criteria are adequate for representing these data. A failure model based on the difference of elastic constants of the constituent minerals and a mechanism of structural instability yields the correct relation between maximum stress and confining pressure. The model can also explain some typical experimental observations, in particular the formation of a shear fault in the presence of confining pressure.

ning electron microscope. They make the distinction between blunt-ended cavities and sharp-ended brittle Brittle rocks submitted to a compressive state of stress cracks and suggest a number of different causes for are characterised by two remarkable features: a strong their origin. With the same technique Sprunt and pressure dependence of strength and a non-linear Brace [4] investigated the microcracking induced to volume increase prior to failure. The compressive rocks by hydrostatic pressure and temperature. The destrength increases sharply as soon as even a moderate velopment of microcracks induced by non-hydrostatic confining pressure is applied. Typically the compressive stresses in Westerly granite was studied by Tapponier strength of granite rises from 200 MN/m 2 (2 kbar) withand Brace [5] and Brace [6]. New cracks were found out confinement to about 1000 MN/m 2 (I0 kbar) at to appear first at grain boundaries and healed trans100MN/m 2 (1 kbar) confining pressure. During loadgranular cracks, whereas at higher load new transing, the rock begins to deviate from linear elastic begranular cracks also formed. At peak stress kinking of haviour at about one half of the failure stress. This biotite was observed. deviation consists of a gradual increase of volume with Directly related to the stress peak and the subsequent rising stress and is known as dilatancy. A lot of effort loss of strength is the formation of a shear fault. Jaeger has been devoted to investigating the causes and and Cook [7] state that in general the failure of brittle effects of dilatancy and its relation to rock strength, yet the understanding of brittle failure of rock under rocks under compression changes from axial splitting to shear faulting when confining pressure is applied. compression still remains incomplete. Before formulatBy means of an acoustic emission technique, Scholz [8] ing a model for the brittle failure of rock under triaxial found that in Westerly granite the microfracturing accompression, some of the present knowledge must be tivity is randomly distributed throughout the rock reviewed. specimen up to about 95~ of the peak stress. Above When brittle rocks are loaded in compression, micthis load, new microcracks are formed preferentially rocracks with a predominant direction parallel to the and tend to cluster on what eventually develops into maximum stress begin to open at about one half of a fault. Wawersik and Brace [9] investigated optically the peak stress. These open microcracks increase the polished sections of Westerly granite, which had been porosity and result in the dilatancy of the overall strucstressed to different points of the stress-strain curve. ture. The phenomenon was initially investigated by. They found that no clustering of microcracks appears Brace et al. [1], who observed a region of stable microin uniaxial compression and at confining pressures cracking at increasing load prior to failure and sugbelow about 10 MN/m 2 (100 bar). The study of quartgested a simple mechanism of crack formation. Some zite under triaxial compression by Hallbauer et al. [10] microcracks exist already before the rock is loaded in revealed the formation of a macroscopic fracture plane the laboratory. Nur and Simmons [2] suggested that in the central part of the specimen just prior to the these were formed in the early history of the rock and stress peak. With further compression this fracture resulted from the difference in thermal expansion and plane grows by stepwise joining of existing microcracks elastic constants of the constituents minerals. Removal and eventually becomes a fault. Using two optical of a thin surface layer from rock sections by ion thinmethods (slit diffraction and direct motion pictures) Liu ning allowed Sprunt and Brace I'3] the direct observaand Livanos [11] measured the bulging along incipient tion of microcavities in unstressed rock with the scanfracture zones in Westerly granite under uniaxial compression. They found that when the specimens were * Institut CERAC SA, Ch-1024 Ecublens. Switzerland. perfectly aligned, local strain inhomogeneities devel209

INTRODUCTION

210

Walter Janach

oped just prior to peak stress at 99.7?0 of the peak load and formed concentrated weak zones. These were thought to result from the interaction and coalescence of microcracks along the future fracture surfaces. The most common failure criteria used to predict the strength of rocks under triaxial compression, such as the Coulomb or Mohr criterion, are of empirical nature and do not take into account the microscopic phenomena described above. The Griffith criterion and its modifications is based on the behaviour of small pre-existing cracks. It is well suited for stress states in the intermediate region between uniaxial tension and uniaxial compression. For triaxiai compression it is usually replaced by the Coulomb criterion (Jaeger & Cook [7]). An interesting attempt to predict triaxial compressive strength from first principles was made by Peng and Johnson [12]. Their theory of faulting is based on beam-buckling theory and frictional contact among beams. It gives a good correlation of strength data for Chelmsford granite over a moderate range of confining pressure. Based on their detailed studies of microcracks in Westerly granite, Tapponier and Brace [5] suggest that the compressive strength of brittle rocks may be controlled by those mineral components which are capable of plastic slip or have elastic moduli which differ substantially from those of other, components. These critical minerals, such as biotite in Westerly granite, may not amount to more than 5-10% of the total volume. STRENGTH O F WESTERLY GRANITE U N D E R TRIAXIAL COMPRESSION One of the most extensively studied rocks, for which strength data of excellent quality exist over a wide range of confining pressures, is Westerly granite. Brace et al. [1] measured the triaxial compressive strength up to a confining pressure of 800 MN/m 2 (8 kbar) using dogbone-shaped specimens to avoid end effects. Mogi [13] made precise strength measurements with a new and simplified specimen design up to a pressure of 400 MN/m 2 (4 kbar). Figure 1 is a replot of Brace's and Mogi's strength data in at, a5 stress space, al being the maximum compression at failure and a 3 the confining pressure. As already observed by Mogi [13], the a~, a3 relationship is non-linear, ruling out any failure criterion such as the Coulomb criterion, which has a linear relation. Such a criterion could only be valid over a limited range of a3, i.e. between 0 and 50 MN/m 2 (0.5 kbar). Figure 1 contains also the curve for the extended Griffith criterion (Jaeger & Cook [7]), fitted to a uniaxial compressive strength of 230 MN/m 2 (2.3 kbar), which is equally inadequate. A surprisingly good correlation can be obtained with a quadratic relation between the principal stresses of the type a~ = aa3. Figure 1 includes a curve for the relation a~ = 104 a3. By adding a small constant to a3, one can also fit at at zero confining pressure. This property of triaxial strength for Westerly granite under compression is remarkable and to the author's knowl-

5OOO~

2'

'000/,~" I/~ 100(3r ~ "~ .~ t

.1

Griffith

t t- i

500~f f ~ 0
! I

200

400

600

8(3O

o'~, MN/m z Fig. 1. Principal stresses at failure for Westerly granite compared to Coulomb criterion, extended Griffith criterion and to quadratic relation ~ = t04 a3. Circles are measurements made by Brace et
al. [1], crosses are by Mogi 1"13].

edge has not been noticed before. In terms of predicting the strength of Westerly granite one could just use the above relation as another failure criterion with an excellent correlation with experimental data. On the other hand, the empirical quadratic relation raises some fundamental questions, which could be instrumental to a better understanding of brittle fracture of rock under compression. Two questions arise immediately: how can the complex micromechanical behaviour of granite undergoing failure produce such a simple quadratic relationship between al and a a? Is there a critical and unique micromechanism which controls the strength under triaxial compression? In order to find an answer to these questions and possibly an explanation for the empirical quadratic relationship, it is necessary to identify the critical mechanisms involved in the failure process and subsequently to describe them quantitatively. This involves deriving a theoretical failure model for the rock. The validity of such a model depends upon what assumptions and simplifications are made and upon the degree of correlation with experimental data. FAILt'RE MODEL There is ample evidence (e.g. Tapponier & Brace [5]; Hallbauer et al. [10]) that the failure process which determines the strength of brittle rocks under triaxial compression is not of a shear type. Failure has been found to be initiated by the collapse of the microstructure, which has been previously loosened up by predominantly axial microcracking. The collapse, which is due to a structural instability, is conce~'ated along a plane at a small angle to the principal compression, which in a subsequent stage becomes a fault. Once the fault is formed, shear motion takes place along it and begins to unload the surrounding rock. When a rock structure consisting of minerals with different elastic constants is submitted to a uniformly applied stress at, the internal stress and deformation

Failure of Granite under Compression

211

~
5ti f f

Gt

Con'~lianf---...

t
Fig. 2. Distortion of an elastic column of different components illustrating origin of geometrical mismatch.

fields are inhomogeneous on the microscopic scale. This is illustrated in Fig. 2 by representing a material column which consists of stiff particles in a compliant matrix. The straight column is distorted under load. If the components are linear elastic and the deformations remain small, the lateral distortions f are proportional to the applied stress ol. expressed by f oc at. (1)

Fig. 4. Formation of a diagonal array of tipped elements.

In a competent rock such as granite, this mechanism leads to a certain degree of geometrical mismatch between the particles of constituent minerals with different elastic constants. Because the rock has tensile strength, axial cracks do not begin to open until the axial compressive stress reaches some value. In the presence of a confining pressure o3, the opening of axial cracks is prevented up to a higher stress o,. The microstructure of granite has no regular repetitive pattern as it consists of a random distribution of different mineral components. Of the many different geometrical arrangements of components in the material, some will begin to form axial cracks at an earlier stage than others. A single, isolated axial microcrack does not affect the stability of the entire rock structure. It is the interaction of suitably located cracks between each other that leads to an instability. A simple configuration of two axial microcracks, which can generate an instability, is shown in Fig. 3. The two cracks, which are roughly parallel to at, form what is going to be called a 'tipping element'. When such an element tips, an axial
q._

contraction g and a lateral expansion f are superposed on the pre-existing deformation field. It is characteristic for a tipping element that its deformations g and f impose a shear motion on the surrounding material. The structural instability arises through the formation of an array of suitably located tipping elements as shown in Fig. 4. The formation of an isolated tipped element requires more energy than the tipping of the next element of a propagating diagonal array. The reason is that the surrounding material is more distorted for an isolated element than when the element is part of a diagonal array. Therefore an element with a potential to tip will in general not tip unless it is triggered by a propagating diagonal array. This process has some similarity with a running dislocation. The point of instability of a whole structure coincides with the condition for propagation of a diagonal array of tipped elements. Of course, Fig. 4 is an over-simplification of the real situation, characterised by a random distribution of different tipping elements along the future fault plane. In order to formulate a propagation criterion for the diagonal array, the relation between the axial contraction g and the lateral expansion f of a tipping element is determined. Figure 5 shows one half of the neutral axis of an element before and after tipping. The vertical displacement g describes a curve with a horizontal tan-

I/2f

Stiff - ~ ~ / j ~

Fig. 3. Tipping element.

Fig. 5. Kinematics of neutral axis of an element during tipping.

R.M.M.S.14/4~0

212

Walter Janach

/5

o"I

work balance, which takes into account only the overall displacements at a sufficiently distant control surface. After introducing relations (3) and (4) into 15) one obtains
0.1g ~ cr3f

.--/xy

Taking into account that the axial contraction g and the lateral expansion f obey the quadratic kinematical relation (2) and furthermore that the lateral expansion f is proportional to a~ (1), equation (5) finally reduces to a~ -s_ 0"3, (6) which represents the criterion for the propagation of a diagonal array of tipped elements. Once the array of tipped elements has propagated across the entire rock specimen, it becomes a region of weakened material. The reason for this is that a global shear motion can now take place simultaneously along the plane of the array while the more intact regions on both sides of this plane do not need to deform. Consequently the axial stress a~ begins to decrease as soon as the array goes across the entire specimen. Therefore the propagation criterion (6) is at the same time the criterion for the maximum axial stress al that the rock can support at a given confining pressure a3. In other words equation (6) is a failure criterion for triaxial compression. Equation (6) is incomplete as the constant is unknown and the uniaxial compressive strength becomes zero. Predicting the constant would require a quantitative analysis of the structure of granite. Here the constant is empirically established so as to give the best fit to the available data. Because the opening of axial cracks must overcome the tensile strength of the rock besides doing expansion work against the confining pressure, it is evident that a small term should be added to 0"3 . Figure 7 is a comparison of Brace's and Mogi's failure data for Westerly granite with the relation tr 2 = 1.05" 10'*(6 3 -I- 5) (MN/m2).
30OC - -

Fig. 6. Displacements &x and Ay (grossly exaggerated) of a control surface S, resulting from the local displacements g and f of a tipping element.

gent at the origin. This curve can be described by a power series in f, the linear term being zero because of the horizontal tangent. Keeping only one term by truncating the power series, one obtains the kinematical relation for a tipping element as
g ~2 f 2 . (2) To a first approximation the axial displacement 9 is therefore proportional to the square of the lateral displacement f It is interesting to note that this kind of relation is also fundamental in buckling theory. But here the tipping of an element is not a buckling instability as in the model suggested by Peng and Johnson [12]. The criterion for propagation of a diagonal array of tipped elements is going to be based on the balance between the work done by al through the contraction 0 and the expansion work against the confining pressure a3 through f A control surface S surrounding an element as shown in Fig. 6 is slightly deformed as a result of tipping. It is assumed that the axial displacements &x of the control surface depend only on g and the lateral displacements Ay only on f As the mean stress of the material inside the control surface remains unchanged during tipping, the displaced volumes of the control surface in axial and lateral direction (for unit thickness) are equal to the volumes displaced by the tipping element through 9 and f Therefore the work I4'1 done by 0.1 on the material inside the control surface through the axial contraction is proportional to alg: Wl = a t l s A x ds oc aiO

(7)

25O(

(3) against o"3


(4)

2000

and accordingly becomes

the expansion

work

W 3 = t r 3 $ s A y d s oc t y 3 f

IO(30

At this point the hypothesis is made that propagation of a diagonal array of tipped elements will occur whenever the contraction work W1 equals the expansion work W3, w~ = w3. (5) The redistribution of inhomogeneous strain energy around the tipping element is not considered in this

/
I
200

500

400 % , MN/m 2

600

800

Fig. 7. Principal stresses at failure for Westerly granite compared to relation at= = 1.05-104 (a 3 + 5). Circles are measurements made by Brace et al. [1], crosses by Mogi [13].

Failure of Granite under Compression Equation (7) predicts a uniaxial compressive strength of 229 MN/m 2 (2.29 kbar), which corresponds well to measured values. The additive constant is 5 M N / m z (50 bar), which is about one half of the tensile strength. This is not unreasonable, because the local tensile resistance drops as soon as a crack opens. DISCUSSION Despite the good correlation with measured strength data, the derived failure model is only a tentative description of the rock failure process, providing a qualitative and plausible explanation of the observed quadratic relationship. The model is also incomplete because it is set up for a plane state of stress and does not take the second principal stress tr2 into account. This should really be considered even when tr2 and tr3 are equal. Moreover the derivation is not at all rigorous and is based on numerous assumptions and a hypothesis. The real behaviour of granite undergoing failure is by far more complicated than the simple mechanism of a tipping element. Figures 8 and 9 show a typical shear fault in a failed cylinder of Swedish granite. From Fig. 9 it appears that there is a gradual increase of open microcracks towards the fault. The overall failure process is the statistical average over all the different microprocesses. In reality the tipping of elements and the propagation of the array of already tipped elements are rather gradual processes which instead of being sharp and discrete are dispersed in time and space.

213

Fig. 9. Detail of Fig. 8, showing part of the shear fault and axial microcracks in the surrounding rock, On the other hand the model is consistent with and can help to explain some of the observed phenomena. The geometrical mismatch of particles with different elastic constants produces an inhomogeneous elastic field, which is superposed on the average stresses ~t and ~3. At a certain stage this leads to the opening of microcracks in all those locations where the tensile strength is exceeded. This occurs most often in the direction of the minimum compression tr3 and explains the predominantly axial direction of dilating microcracks. Isolated microcracks with a random distribution throughout the material do not affect the overall stability of the structure and produce dilatancy well before the peak stress is attained. The stress peak is a result of an interaction process between existing microcracks. It is characteristic for this interaction process (in the model the propagation of a diagonal array of tipped elements) that it is preferential and localised along suitably oriented planes. The affected material is loosened and becomes weaker. Any further deformation is therefore concentrated in this zone of weakness and the load which the entire structure can support begins to decrease. Under normal testing conditions the entire system including the loading machine becomes unstable at this point and leads to a violent fracture of the specimen. The preferential formation of a diagonal array of tipped elements which subsequently develops into a shear fault can also be explained by an overall energy argument. Because the tipping of elements requires expansion work against the confining pressure tr a, the

Fig. 8. Shear fault in a specimenof Bohus granite failed under triaxial compression at trt = 430 MN/m2 (4.3kbar) and tra = 20 MN/m2 (0.2 kbar).

Zt4

Walter Janach

under shock load. one calculates a Hugoniot ctastic limit of 45 kbar. This successful predictio,a of strength at very high strain rate using a model based on static data is an indication that the compressive strength of granite is independent or only weakly dependent on strain rate. This is in agreement with strength data collected by Brace and Jones [16] at different strain rates for many rocks, which show that the strength increases by only about 15, for a strain rate change of I000. The much higher strain rate sensitivity of strength observed in split-Hopkinson bar experiments with rock is an apparent effect, which was shown by Janach [17] to be due to bulking of the failing material and the resulting additional radial inertia forces. The tipping element failure model based on the difference of elastic constants of the constituent minerals can correlated failure data for Westerly granite ranging from uniaxial compressive strength to the Hugoniot elastic limit. Whether the model applies also to other brittle rocks or even to brittle fracture under compression in general, is unknown at this stage. If a brittle material consisting of only one component which is isotropically elastic has internal defects in form of preexisting microcracks, it will also develop an inhomogeneous stress and deformation field when loaded in compression. It is suggested that these defects could Fig. 10. Part of a specimen of Bohus granite failed under uniaxial play a role similar to the difference of elastic constants compression at a~ = 2iX)MN'm2 t2 kbar}, showing columns formed in the derived model by also leading to a geometrical by axial splitting. mismatch of the microstructure and forming tipping elements, The influence of the intermediate principal stress az concentration of such elements along a diagonal plane minimises the total expansion work. Once the diagonal on the failure mechanism when it is not equal to a 3 array is fully developed, it allows the unloading of the has not been investigated. This would include the interstill intact material through a Iocalised shear motion esting region of stress states encountered in extension along what becomes the fault plane. Fault formation, experiments, where the compressive stresses at and a: as typically shown in Fig. 8 for Buhus granite, is there- are equal while tr 3 can vary from small tension to fore the process which minimises the expansion work moderate compression. Future work should include a quantitative elastic required for unloading the main compression. This reasoning does not apply when a3 is very small or zero analysis of a simplified structure with different elastic as in uniaxial compression. In this case the material constants, possibly supplemented by model experiseparates into long columns through large scale axial ments, in order to verify the formation and instability splitting as shown in Fig. 10, The columns most prob- of tipping elements. Other rocks and brittle materials ably fail through a buckling instability. It is remarkable would need to be tested in a wide range of confining that there is no discontinuity of the measured failure pressures in order to investigate whether the quadratic stress at the transition from axial splitting to faulting relation between the principal compressions 0-1 and a 3 and that both types of failure are equally well described is also valid and to what extent it might be characteristic for brittle fracture under compression in general. by the quadratic relationship between al and a 3. It is interesting to use equation (7J to extrapolate to failure stress states at higher confining pressures than Receired 14 January 1977. the 800 M N / m 2 (8 kbar} used by Brace e t al. [1]. By calculating the point of intersection of the curve REFERENCES 0"7 = 1.05" 104 {O" 4- 5} with the straight line a 1 = rr~ 3 11 - vljv corresponding to uniaxial strain one can preI. Brace W. F.. Paulding B. W. & Scholz C. Ditatancy in the fracdict the Hugoniot elastic limit of Westerly granite. The ture of crystalline rocks. J. oeophys. Res. 71, 3939-3953 (1966). 2. Nur A. & Simmons G. The origin of small cracks in igneous Hugoniot elastic limit is the strength at uniaxial strain rocks. Int. J. Rock Mech. Min. Sci 7, 307-314 {1970}. under shock load. It is measured as the amplitude of 3. Sprunt E. S. & Brace W. F. Direct observation of microactivities the elastic precursor of a strong shock wave. For an in crystalline rocks. Int. d. Rock Mech. Min. Sci. & Geomech. Ahst. 11, 139-150 (1974). unspecified granite, Grine [14] found this amplitude to 4. Sprunt E. S. & Brace W. F. Some permanent structural changes be about 40kbar. Using a Poisson ratio v = 0.30 as in rocks due to pressure and temperature. Proc. 3rd Int. Conq. suggested by Walsh and Brace [15] for Westerly granite Rock Mech., Denrer, 2A. 524-529 11974).

Failure of Granite under Compression


5. Tapp~nnicr P. & Brace W. F. Dc~ch~pmcnt of stress-induced microcracks in ~,'cstcrl.~ granite. I,t. J. Rock Mech. Min. Sci. & (h,ometh. Ah.~tr. 13. 103 112 119761. 6. Brace W. F. Direct observation of dilatant voids in rock. The Effects of Voids on Material Deformation (Edited by Cowin S. C. & Carroll M. M,) AMD Vol. 16. pp. 1-12. ASME. New York ( 1976~. 7. Jaeger J. C. & Cook N. G. W. Fundamentals of Rock Mechanics. pp. 86, 95, 147. Methuen, London (1969). 8. Scholz C. H. Microfracturing and the inelastic deformation of rock in compression. J. oeophys. Res. 73, 1417--1432 (1968). 9. Wawersik W. R. & Brace W. F. Post-failure behaviour of a granite and diabase. Rock Mech. 3, 61-85 (1971). 10. Hallbauer D K., Wagner H. & Cook N. G. W. Some observations concerning the microscopic and mechanical behaviour of quartzite specimens in stiff, triaxial compression tests. Int. J. Rock Mech. Min. Sci. Geomech. Ahso'. il), 713--726 (1973). I I. Liu H. P. & Livanos A. C. R. Dilatancy and precursory bulging

215

12. 13. 14. 15. 16. 17.

along incipient fracture zones in uniaxially compressed Westerly grat~,~ J. gophys~ Res, gl, M95-3510 (1976). Peii~S? k Johnson A. M. Crack growth and faulting in cylindrical Specimens of Chelmsford granite. Int. J. Rock Mech. Min. Sci. 9, 37-86 (19721. Mogi K. Some precise measurements of fracture strength of rocks under uniform compressive stress. Rock Mech. Enang Geol. 4, 41-55 (1966). Grine D. R. Equations of state granite and salt. Stanford Res. Inst. Final Rept~ Proj. PGD-3244 (1961). Walsh J. B. & Brace W. F. Elasticity of rock in uniaxial strain. Int. J. Rock Mech. Min. Sci. 9, 7-15 (1972). Brace W. F. & Jones A. H. Comparison of uniaxial deformation in shock and static loading of three rocks. J. geophys. Res, 76, 4913-4921 (1971). Janach W. The role of bulking in brittle failure of rocks under rapid compression. Int. J. Rock Mech. Min. Sci. & Geomech. Absir. 13, 177-186 (1976).

S-ar putea să vă placă și