You are on page 1of 11

lnt J

R~J{k ~,teth

~,ltn 5~t & (Jcomt'lh

th~;r rot 23. No ~, pp ~ " 1'a-I'~e,

Printed In Great Britain

q)14s-ql)e,2*e~ $1.1)1~. II00 Pcrgam~m Jour~'~al'-, [.td

A Crack Tensor and Its Relation to Wave Velocity Anisotropy in Jointed Rock Masses
M. ODA*
T. YAMABE*

K. K A M E M U R A t
Crack geometry, which is closely related to the mechunical anisotropy of discontinuous materials such as rocks amt rock masses, can be concisely expressed hv a tensor F (calh, d the crack tensor), bt order to git'e a mathematical expression jor the change of longitudinal wave velocity with direction, an athtitional tensor V {s"intro~hwed. Three examples (gypsum phtster samples with art(/icial cracks, ,granites with microcracks and jointed granites) are reported ill this paper with the special emphas& oil the relation between the crack Xeontetrl" F atld the directional wave velocity V. Both tett.sor.s" are co-a.vial in the sense that the mq/or principal axis q/'F accords well with the minor one ~ffV. The anvotropic c/umge ~{1 ware relocitv with ~#rection can he associated with the anisotropic hldex measure Jor the crack geonwtr)', suk,gestin,~ a cau.s'al relation. Sm'h rehttions are u.vq/id.for e.vtractinff, itff'ormation {'ott#'rtlitlk, the crack ,;,eomctr 3' of jointed rock mos,vcs .from geophv.~'ical e.vp[oration.

INTRODUCTION Rock masses arc commonly treated as isotropic solids in conventional analyses of rock mechanics. According to the comprehensive review by Gerrard [1], however, geological mater/at is seldom isotropic. Granite, for example, is not an exception in spite of its isotropic appearance [2-5], In fact, the preferred orientation of discontinuities such as microcracks, fissures and joints is universally observed in granites, and it is believed to be one of the major sources controlling the anisotropic properties, It is clear that crack geometry (density, size and orientation of cracks) must be considered first when rock masses are treated as anisotropic solids. For the past two decades, many studies have been undertaken to show the effect of cracks on the mechanical properties of rocks and rock masses: based on continuum mechanics, elasticity moduli (static and dynamic) have been formulated by considering explicitly the presence of cracks [6-12]. Several computer models have been developed in which dominant cracks like faults are dealt with as mechanically equivalent elements (joint elements) in finite element analyses [13,14]. In laboratories, great effort has been expended to investigate the effect of the crack geometry on the elasticity and
* D e p a r t m e n t of F o u n d a t i o n Engineering. Saitama Uni,.ersit,,, 225. S h i m o - o k u b o . Urav, a. Saitama. Japan. +Taisei C o r p o r a t i o n . 1-25-1. Nishi-shinjuku. Shinjuku. Tok.,,o. Japan.
R ' . I ' . I S :~,~ \ 3S7

failure criterion of artificially jointed bodies [15, 16]. In spite of these studies, however, some difficulty still exists in characterizing in situ rock masses. This is partially because cracks are usually too complicated, especially in terms of geometry, to be discussed in general. In order to overcome the present difficulty, Oda [17-19] has introduced a tensor quantity (called the crack tensor) to handle the actual crack geometry easily. In this paper, three examples (gypsum plaster samples with artificial cracks, granites with microcracks and jointed granites) are discussed in detail with the special emphasis on the relation between the crack geometry represented by the crack tensor and the corresponding anisotropy of elastic wave velocity in cracked material. Such a relation, if it can be established, will provide useful information concerning the crack geometry from geophysical exploration.

CRACK TENSOR
Dqfinition

Here, "'crack geometry" is used to represent a concept concerned with density, size and orientation of related cracks. (I) Density O/cracks. If there are tn ~t~ cracks in a statistically homogeneous body of volume V. the crack density p is defined as,

388

ODA et al.: CRACK TENSOR AND WAVE VELOCITY IN ROCK summarized below.

~ FI(_) Fig. 1. Two unit vectors n'+' and n'-' normal to a crack.

(2) Size of cracks. For simplicity, a crack having area S is replaced_._b_y_an equivalent circle with a diameter r (i.e. r = 2 x / S i n ) . (This is not always a necessary assumption. If the shape of cracks is already known, a typical dimension can be used instead of r.) The distribution of crack sizes is then given by a density f u n c t i o n f ( r ) of diameters. It must satisfy i=f(r) dr = !
(2)

(1) Definite mathematical meaning. F is a dimensionless tensor, with non-zero components Fj ~ (i,j . . . . k = 1.2.3) only when the rank ( = the number o f subscripts of F,j k) is e~en. The components are symmetric in the sense that F,, ~ = Fj, k = F , r , . A contraction with respect to any pair of subscripts reduces its rank by 2. The zero-, second- and fourth- rank tensors, for example, are given below by using a fixed orthogonal Cartesian co-ordinate:
rip ["~ Zero-rank: F0 = -4j0 r3f(r) dr (scalar)

Second-rank:

Fij=--~

PfoI

r3ninjE(n,r)d~dr r 3ninjn, nt E(n, r) d ~ dr ( i , j , k , l = 1,2,3)


(6)

Fourth-rank: Fijkt= --~

where r,, is the maximum size of diameters. (3) Orientation of cracks. Orientation of a crack ts indicated by two unit vectors, n I+~and d-~. normal to the major principal plane (Fig. 1). (Note that d +~ is parallel, but opposite, to n'-'. Hereafter. n stands for both n ~+' and n~-L) A density function E(n, r) is used to represent the statistical distribution of n's. It also satisfies

f 'I ' n .E ( n , r ) d f ~ d r = l ,

(3)

where Q is solid angle corresponding to the entire surface of a unit sphere. Here. E(n, r) is symmetric in the sense that E(n, r) = E( - n . r), and it is written as E(n)f(r) if n and r are statistically independent. A tensor F (called the crack tensor) has been introduced to give a mathematical definition for the crack geometry in such a manner that all these elements are included [17]; F= 4 J0

where n~ is a direction cosine of n with respect to the reference axis x,. (2) Close relation to crack geometry. On the assumption that crack aperture increases in proportion to crack size, Oda has proved that the scalar F0 is equivalent to the porosity associated with cracks [17]. Since F0 is a symmetric second-rank tensor just like a stress tensor, one can always find three orthogonai principal axes even though non-orthogonal crack sets are concerned. In order to show that the crack tensor is closely related to the crack geometry, the concept o f isotropy is examined here. Two-dimensional cracks are illustrated in squares o f size a (depth T) in Fig. 2 whose abscissa and ordinate are parallel to the reference axes xl and x, respectively. Let 0 ~k~be an inclination angle of n ~k~(i.e. a unit vector

r~nn...nE(n,r)d~dr

(4)

where stands for tensor product and the number of n designates the rank of the tensor. If all information concerning individual cracks is available, the integral form of equation (4) can be rewritten by the following additive form;

I--11
(1) Sample (A)
(4) Saml:~e (D)
O

V = n ,w, (r'k*)3n(k) n (k) "'" n Ik)

(5) /\ / /,

where the superscript (k) denotes a k th crack among r # r~ cracks. (If cracks are idealized by squares of side length r rather than circles, then n/4 is omitted from equations (4) and (5). Note also that major cracks are only important in the calculation of F because each crack contributes to it in proportion to r 3. In other words, relatively smaller cracks can be disregarded without introducing any serious error.)

I
(2) Samole (B)
[ .

,. a

(5) Samole (E)


x2

\
T ~ X1

Characteristics of the crack tensor


The reasons that the crack tensor has been accepted as a representative measure of the crack geometry are

(31 Saml~le (C)

Fig. 2. Five crack models with isotropie nature m the sense of the second-rank crack tensor.

ODA

et al.:

C R A C K T E N S O R A N D WAVE VELOCITY IN R O C K

389

normal to a k th crack) to the x~-axis, and r and T conditions are satisfied: be the crack length and the crack depth respectively. Note that, for such two-dimensional cases, the direction r--.0 cosines n~ and n~ are given by cos0 ~e~ and sin0~kL respectively. The two-dimensional version of the crack ~o "-'* I. tensor, for example F~:_,. is given by Here, E(O) equals l/(2n) since each n has an equal chance to be oriented at any solid angle. Accordingly, F, iz2=~.TkEl= T(r4k')ZcosZO~lsin20 ~k~. (7) the crack system is an ideal one having the following isotropic tensor: Sample (A): sixteen cracks with the same crack length Fgel= 1 of r/a = 0.25 are inserted. Half of them are oriented at 0 = 0 , and the remainder at 0 = 90L The crack tensor is given in matrix form. 0 = 5

F~'~= I
\&, &:,/

oo)
"

,,

L,

o/

Sample (B): sample (B) is exactly the same as sample (A) except that all cracks are rotated by 45 '~ clockwise. The crack tensor is

All samples (A) to (E) are characterized by the same second rank isotropic tensor. (The slight deviation from isotropy is ignored in sample (D).) This may seem strange in view of the quite different appearance of these samples. The difference among them appears clearly, however, if higher rank tensors are considered. An isotropic tensor of fourth-rank is given in general form by

F~0"~= I
F(B) =

F,jkt = ;'.5,jakl + ~L(a,k6; + 6a6jk) + V(a,k6#-- 6a6,~)


0

(13)

'0.5

'J

o5)
0.25 0.25 0

Fire tlkl

'0.25 0.25

o.i5).

where 5,j is Kronecker's delta, and 2, it and v are constants. Since any subscripts in F,j~q are interchangeable, 2 equals Iz and v is zero: (9) Ft,ii
= F2222 =

3Fll22 = 32
(14)

FI222 = Flit2 = 0.

Sample (C): sixteen crack are inserted at 0 = 0 , 4 5 , 90 ~ and 135:, four cracks for each orientation. The crack tensor is

F~oC~= 1
'J 0.5

0,2 a )
__,#,,r'~c/=/0.~50.3750 0.125 . (10) Sample (D): orientation of cracks is chosen as random as possible to make an isotropic sample. The crack tensor is not exactly isotropic, but it becomes F~)DI= I =( 0.488 \-0.09 - 0 . 0 9 "] 0.512J 0.109 0.403 --0.056 -0.039\ --0.056). 0.109/

Based on equation (14), samples (A) and (B) are no longer isotropic while sample (C) is still isotropic. However, the isotropy of sample (C) disappears if the sixth-rank tensor is introduced. (3) General index measure for crack anisotropy. Let F~, F2 and /73 be the principal values of F v, and consider a space whose axes are taken as F~, F2 and F3 (Fig. 3). In this space, any second-rank tensor is expressed by a vector O P with the components of the corresponding principal values. If O P is on the diagonal (for three-dimensional cases, & = F , = F 3 ; and for twodimensional cases, F~ = F_,), the second-rank tensor is
F3

,,otr#$ic T p
B\

,,:,=F2

F'D'

0.378 --,1ktV~D)= 0.109 --0.039

(11)
(a) Three-dimension (b) Two-dimension

Sample (E): an infinite number of tiny cracks are inserted completely at random such that the following

Fig. 3. Three- and two-dimensional fabric spaces to show the isotropic and anisotropic components of the crack tensor.

390

ODA et al.: CRACK TENSOR AND WAVE VELOCITY IN ROCK (iii) open cracks remain open during the application of stress:
Eii ~ -

isotropic. This means that decomposing the vector OP into two vectors is essential: i.e. OP = OA + OB. The length IOAI of the vector OA is proportional to the first invariant of F~/. That is. for three-dimensional cases.

(1-

V) 5,k6j`

~5,j6~,
v

+ ~ , (6,tFjk + 6j, Fj, + 6ikFu + 6~,Fjt) ak, (18)


in which the summation convention is used, and E and v are Young's modulus and Poisson's ratio for the elastic matrix, respectively. [If F,j is an isotropic tensor, then equation (18) is reduced to Walsh's solution (7).] The first two terms in the right-hand side of equation (18) correspond to the elastic deformation of the matrix, while the last one is the correction term due to the presence of cracks. Here, i/D equals 8/(3 hE) for pennyshaped cracks and rc/(2E) for elliptical cracks stressed under plane stress. When a~ is only a non-zero stress component (uniaxial stress in the x~-direction), equation (! 8) becomes O'u = g ' = E

IOA[=

(F,

+ F2 + 1:3)= -.~ Fo

and for two-dimensional cases,

IOA]=-~(Ft + F,)=-~Fo.

(15)

The vector OB, which is placed on the deviatoric plane. has the following magnitude: for three-dimensional cases.

IOnl =

(F~F,j)

i ~

"=

~-~

[(F,

F:)" (16)

+ (F2- F3)'- + (Fj - F,):]'::


and for two-dimensional cases.
Ion I = - F,)

c,,

(EFIt)/D + 1"

(19)

where F~ is the deviatoric tensor of F o. In the sequel, the following measure A (r~ is employed as a representative index of anisotropy; for three-dimensional cases. A'") [OBI [(F'-F:)"+(F:-EO2+(F3-F')2]:" =[~= F, + F , + F3

A more general formulation has been given by Cowin [20]. Furthermore, according to a recent study by Oda [t9], the second-rank crack tensor plays an important role when a permeability tensor for jointed rock masses is considered. (In the sequel, we are mainly concerned with the second-rank tensor because of the important role in the hydro-mechanical analyses of jointed rock masses. No difficulty exists, if necessary, to introduce higher rank tensors.) SUPERSONIC WAVE VELOCITY TESTS ON GYPSUM PLASTER SAMPLES

and for two-dimensional cases, A F~= FI -- F., (17)

F,+F,'
Two points are important: (1) A er) alone gives a false impression about the anisotropy of crack geometry. Note always that the combination of F0 and A eF) is extremely important. For example, a sample with one crack has the same anisotropic index as a second sample with many parallel cracks, even though the later is more anisotropic than the former. (2) In the two-dimensional cases which will appear in the sequel, A (~) ranges from zero for isotropic crack systems to unity for anisotropic systems with completely parallel alignment of cracks.

Tested samples
Supersonic wave velocity tests were performed on gypsum plaster samples having artificial two-dimensional cracks. Samples were made as follows: water-gypsum mixture (2:3 by weight) was poured into a cylindrical mould with a circular cross section of 200 mm dia. Strips of greased paper were inserted into the water-gypsum mixture such that a given crack system was made. After the gypsum plaster was well hardened, it was taken out of the mould, and was trimmed to make an equilateral polygon. A generator of supersonic waves was tightly pressed at the diagonal position of a receiver (Fig. 4).
generator

(4) Close relation with the mechanical and hydrological properties. In the analyses of the mechanical behaviour
of rock masses, engineers commonly deal with tensor quantities such as stress tr~ and strain qj. It can be said, therefore, that if the concept of crack geometry is quantified by a tensor, it is easily taken into account in the analyses. For example, consider elasticity for discontinuous rock masses. Oda et al. [12] has formulated an elastic constitutive equation on the following assumptions: (i) each crack is simplified as a penny-shaped open crack of diameter r. (ii) there is no elastic interaction among cracks, and

receiver Fig. 4. Supersonicwave velocitytests on gypsumplaster sampleswith artificial cracks.

O D A et al.:

C R A C K T E N S O R A N D W A V E V E L O C I T Y IN R O C K

3'-)1

//
[
/ f

/
/

'\/

//

(a) Sample (I)

(b)

/ 2

/
E
~- X I

. . . .

E0

(c)
Fig. 5. Result of supersonic wave velocity test on sample (I) with an almost isotropic, second-rank crack tensor.

Longitudinal wave velocity was measured at eight directions for each sample. Actual cracks are commonly filled with weathered products having much less rigidity than the solid matrix. Accordingly, the greased paper can be a likely model since it is highly compressible as compared with the hardened gypsum plaster. Uniaxial compression tests on

these samples have shown that the cracks of the greased paper behave like elliptical open cracks [12]. Three samples (I), (ll) and (IIl), selected from nine experiments, are shown in Figs 5--7. The tensors F~,]~, U,] u and FUm are calculated usin~ the reference axes 2q and x, which are parallel to the major and minor principal axes respectively, as follows:

(b)

x2

E . . . . 0 E-o

(c)
Fig, 6 Result of supersonic wave velocity test on sample (11) with between (I) and [Ill). a second-rank crack tensor intermediate

392

ODA

et al.:

CRACK

TENSOR

AND

WAVE VELOCITY

IN R O C K

1.o

(a)

Sample (~)

x2

X1

,!

.0

E Eo

(c)
Fig. 7. Result o f supersonic wave velocity test on s a m p l e (111) with a n a n i s o t r o p i c c r a c k tensor.

for sample (I),

__(0oO 08)
for sample (II), 'J and sample (III), Ft,,,) q
=

(0:4

0)
0.14

(20)

is also in harmony with the anisotropic crack tensor. Sample (II) is intermediate. Local axes x~ and x~ are set by rotating the principal axes x, and x2 counterclockwise at ~. If samples (I), (1!) and (Iit) are uniaxially compressed parallel t o the new axis x; (i.e. a;, ~ 0; a~2 = a;., =0), then equation (19) becomes E' E ! rc (F, cos2~ + F2 sin2~) (21)

(0"088 00) where r~/(2E) is used in the place of I/D. This is because the cracks of the greased paper are expected to behave like ellipses Using the values of F~ and F., of equation (20) in equation (21), the change of E'/E with directions can be expressed as a function of a (seeFigs 5c, 6c and 7c). In the case of samples (II) and (IH), E'/E varies from the maximum to the minimum in the similar way to that of (v/v0)2. In the case of sample (I), on the other hand, both E'/E and (v/v0) 2 are almost independent of ~. Now, consider a mathematical expression for the directional change of the squared velocity ratio. For this purpose, assume that the squared velocity ratio in a direction m (unit vector) is approximated by
(--~
V "

where F~ and F, are the principal values in the principal directions.

Experimental results
The squared longitudinal wave velocities v 2 (normalized by the squared longitudinal wave velocity o 2 for the crack-free sample) at eight diagonal positions are shown in the polar diagrams of Figs 5b, 6b and 7b. (The reasons that the squared velocity ratio (v/v0)2 is selected rather than the velocity ratio (v/v0) are: (1) for an isotropic elastic solid, v 2 is more directly related to the elastic constants than v itself, and (2) on the basis of the observed relation between (v/r0) 2 and RQD, Deere [21] has suggested that (v/v0) 2 can be a representative measure for the Rock Quality Designation.) For sample (I) having the nearly isotropic crack tensor, the change of the squared velocity ratio with directions is quite small In other words, it behaves like an elastically isotropic body. For sample (Ill), on the other hand, the change is remarkable for the maximum in the x,direction to the minimum in the x -direction. This result

\ t'0 /

_~ Vi/~lim /

(22)

where V~ is a non-dimensional tensor to be determined. In fact, V~ has been calculated by using Kanatani's least square error method [22];

4~(v'k'y
v,J=~tk~,\v0 / j

1
~ 0

:'~' ulk) 2 t
,

where N is the number of the measured directions.

ODA et al.: CRACK TENSOR AND WAVE VELOCITY IN ROCK

3'93

2.0P gypsum plaster samples siteA Vo siteC 1.0


o

\--.

siteB

0 0.1

I 0.5

I 1.0

5.0

10.0

I 50.0

vo
Fig. 8. Relation between the isotropic index 1:,~for the squared longitudinal wave velocity ratio (r 'r,): and the isotropic index F~ for the crack geometry.

The squared velocity ratios at eight directions are substituted in equation (23) to calculate the tensor 1~;~ first. Using it, the change of the squared velocity ratio with directions is mathematically expressed by means of equation (22). As shown by the broken lines in Figs 5b-7b, the mathematical expression, based on equations (22) and (23), is precise enough to envelop the experimental results. It can be said, therefore, that the tensor 1~ is a reasonable index to measure the anisotropic change of longitudinal wave velocity. On the analogy of the crack tensor, the following indexes can be defined: for three-dimensional cases, V()= V~+ V : + V~ A~. ~ = [(I/'1- V 2 ) 2 + ( V , V3)2+(V,

and Laurentian), together with detailed microscopic fabric data on microcracks, fluid inclusion planes, biotite cleavage planes and quartz optic axes. From the three granites all of which are anisotropic, Barre granite is discussed here in detail. Douglass and Voight have concluded that the microcracks in quartz particles give the d o m i n a n t effect on the mechanical anisotropy of the granite, especially at low stress levels. In this paper, therefore, the microcracks

,o[
I

0.9

K,

"

VI):] 12

(24a)

siteC

and for two-dimensional cases,


V o = VI + I'~ A v

07

AI n _ I,:I - VVo

06 (24b)
0.5 i

where V0 is the isotropic c o m p o n e n t of V o while A ~4 r~ is the anisotropic one. Referring to the two tensors F,j and V,:, the directional change of longitudinal wave velocity can be quantitatively discussed. Data taken from the nine tests are plotted as open circles in the relation between V0 and F0 of Fig. 8 and also in the relation between A ~ and A m of Fig. 9. Note that the increase of F o is associated with the decrease of t%. and the anisotropy index of wave velocity have a strong tendency to proportionally increase with the anisotropy index of crack geometry. ANISOTROPY OF GRANITES DUE TO MICROCRACKS Douglass and Voight [3] reported Young's moduli and unconfined compressive strengths of granites, sampled from three North American localities (Barre, Stanstead

siteB
0.4

siteA

0.3

Barre granite

//plaster samples

/yDSum
o

0.2 ~

;/-

/ /

0.5
A F

10

Fig. 9. Relation between the anisotropic index ,4~ l'or the squared longitudinal wave velocity ralio and the anisotropic index A '~' for the crack geometry.

394

ODA et al.:

C R A C K TENSOR A N D WAVE VELOCITY IN ROCK

/
o,

/
X3

\
t

/
la'm2

.IF) X3 Or ~2 )

\o,
F)
"5 "/

7
,

Xt

X*

(a) Poles to microfractures:510 8%.6~, 4% and 2% per 1%

(b) =x~Fi,principal "xiF31 Fit

o x (E~ principal a xl~ I ell~soicl

area (Douglass 8 Voight [3])

(Douglass 8 Voight [3])

Fig. 10, Sehmidt's equal area nets showing (a) poles to mlcrocracks and (b) principal axes for Barre granite [3].

are only considered, without referring to other fabric elements. The density of poles normal to the microcracks is reproduced in Schmidt's equal area net of Fig. 10a in which the reference axes x~. x, and x 3 are perpendicular to the three quarry planes named rift, hardway and grain respectively. If it can be assumed that orientation of microcracks is independent of the size. then the secondrank crack tensor becomes F,, = F0 f nin, E(n) df~ - F~IN,, (25)

where N,, is a second-rank tensor depending 0nly on the directional distribution of n. To calculate the crack tensor F,j of the granite, first read the densities of the microcracks from Fig. 10a. and next calculate Ngj by inserting the readings into the additive form of equation (25): 0.454 F,, = Ff, 0.013 -0.004 0.013 0.315 0.017 --0.004\ 0.017/. 0.231/ (26)

It is clearly seen, on Schmidt's net of Fig. 10b. that the corresponding three principal axes x] rl (major), x~r' (intermediate) and x; r~ (minor) accord well with the normals to the quarry planes. Thill et al. [4] reported the anisotropic, longitudinal wave velocity of Barre granite. They measured the velocities in various directions using a spherical sample of the granite (70mmdia). In Fig. !1. the result is reproduced by equivalent lines of the squared velocity ratio (r/ro)'-. In this case. the reference velocity r0 is set to 5.90 km/sec. It corresponds to the longitudinal wave velocity of the granite under a hydrostatic pressure of about 70 MPa. This is because most microcracks are expected to close under such a high pressure. Three orthogonal principal axes are clearly detected, by inspection. in the wave velocity profile: i.e. the maximum velocity ratio in the x~r~-direction and the minimum in the x'~r~-direction. Using Fig. il. the corresponding three-dimensional tensor V~j can be calculated, whereas

additional information is required to calculate F,j. In spite of the lack of data. however, the anisotropy index A ~r~ can still be determined through equation (26), and the result is given by a black circle in the relation between A tvt and A ~rl of Fig. 9. Douglass and Voight [3] carried out nine uniaxial compression tests on Barre granite to investigate the change of Young's modulus E' with directions. Following Rodrigues's method [2]. the change of E' was approximated by an ellipsoid-type quadratic law with six coefficients of the quadratic. Using the least squares method, the most probable values for the six coefficients were estimated from the actual measurements of E' at nine directions. The three characteristic roots of the quadratic were used to define three principal values E1 (minor), E2 (intermediate) and E3 (major) which were associated with the three principal axes x~C~,-,,_"cE~ ,,,,~"'4~,3"~t~ of the ellipsoid respectively. For the purpose of finding the principal axes and the corresponding principal values, Rodrigues's method differs from the present one based on equations (22) and (23). It is worthy of note. however, that both methods yield similar results especially when extremely anisotropic cases are excluded. The following indexes are tentatively accepted

,_.

k._.oo p

o. ,

X1

Fig. I 1 Change of the squared wave velocity ratio with directions in Barre granite 14].

OD-\ et al.: CRACK TENSOR AND WAVE VELOCITY IN ROCK

3q5

time, that .4 ,L, proportionally increases with the increase of A 'e' for each stress level. 06Laurent~an Barre Star,stead

SEISMIC TESTS ON JOINTED GRANITES Three sites [site (A), site (B) and site iC)]. located near Nakatsugawa, Central Japan, were selected for investigation because they seemed homogeneous in terms of the distribution of joints. According to the geological sur;ey around the tested sites (fresh to moderately weathered granite), most joints were classified, depending on the orientations, into one of the three major sets: i.e. two vertical joint sets with EW and NS strikes and one horizontal set. In site (By, the vertical joints were commonly filled with thin films of brownish material derived from weathering product, while the}' were tightly closed without tiny filling materials in site (A). Site (C) is intermediate, The longitudinal wave velocity for two collected samples was 4.5 km sec. The conventional method for the measurement of longitudinal wave velocity of soils was followed except that transducers were set in place with gypsum plaster. [Of course, in addition to the cases for soils, special care may be required to obtain reproducible wave velocities for jointed rock masses. Since special attention to error sources was not paid. the relative (not absolute) values of wave velocities tire only considcrcd here.} Ninc transducers with frequency characteristics ranging from 20 to 200 Hz were positioned along a straight scanlinc at a distance less than 1 m. To determine thc change of wave velocity ~ith directions in horizontal sections, four scanlines were selected for site (A), six for site (By and two for site (C). An elastic wave was generated by striking the surface with a hammer as hard as possible. After repeating the procedure several times, the average longitudinal wave velocity for a particular direction was determined by making a regression anabsis of the distance vs travel time. The squared wave ~elocities [normalized by (4.5 km sec):] tire shown bv solid lines in the polar diagrams of Figs 13 and 14. These measured values arc st, bstituted in equation (23) to calculate the corresponding tensors I",i in the same way its for the gypsum plaster samples. Using the tensors, the corresponding mathematical expressions for the change of the squared wa~e velocities tire obtained as shown by broken lines in Figs 13 and 14. Note the fairly good agreement between the broken and solid lines in these figures. Joint traces, which were all visible on the horizontal exposures in sites (A), (BY and (C), were carefully mapped. [No information was recorded for joint traces shorter than a cut-off length (about 0.2 m) so that the trace length distribution was truncated. The error introduced by the truncation, which is important in conventional scanline sur,eys, has only a negligible effect on the final result in the crack tensor analyses.] In order to facilitate the mapping, a square ~ith a dimension of I I m ~as superimposed on the exposures as a scale. Two vertical and one horizontal joint sets form a threedimensional crack system which should be characterized

05-

AE 04/
~ /

/
J
~ncrease of stress level level
I

level

I1U

i
ol o~

0.2

03 A F,

0.4

0.5

Fig. 12. Dcpcndcncc of the a n i s o t r o p y index ,-I "~ ' l-or Y o u n g ' s m o d u l u s on the anisotropy index ,-I '*' for the crack gcometr',.

to measure the isotropic and anisotropic components: E, = El + E, + E,, 4 : ' = l~,, [(E, - E: ) : + (E_, - / 7 , ) : + ( E ~ - E. )-']'-" (27) The uniaxial stress-strain curves reported by Douglass and Voight are not linear so that the principal Youn,,'s moduli Et, E, and E~ are not constant, depending significantly on the applied stress levels. At each stress level, a set of the principal moduli and the corresponding principal axes tire determined with successive numbers from 1 to 8. The principal axes x(il:~, .v',~-' and .v~~' tire plotted as O , A and [] respectively on Schmidt's net of Fig. 10b. It is clearly seen that the principal directions tire changeable, dcpending on the stress levels. In fact, the axes .v~r) and .v~, ~:' are rotated about zr/2 around the axis .v~" tit a stress level between 25.3 MPa and 33.7 MPa. In order to explain the rotation, as suggested by Douglass and Voight, other fabric elements must be taken into account: e.g. quartz optic axes. It can be said, however, that the principal axes accord well with the principal axes ,v~, *' (i = I, 2, 3) of the microcracks if our interest is limited below the stress of 25.3 MPa. This may be because the microcracks tire responsible for the elastic behaviour of the granite at the low stress level. By treating the data on Stanstead and Laurentian granite in the same way as Barre granite, A :' (anisotropy index for Y o u n g ' s modulus) and .4 'r~ (anisotropy index for microcracks) were calculated to prepare Fig. 12. Note that the , 4 ' t o ' ,4 '*~ relation seriously depends on the stress level, decreasing A 'L, with increasing the stress level from 6.7 to 59.2 MPa. This fact is mainly caused by the closure of the microcracks which makes the crack tensor isotropic. It is clearly seen, tit the same
I

396
N

ODA

et al.:

CRACK TENSOR AND WAVE VELOCITY IN ROCK

by a three-dimensional crack tensor F, (i,j = t. 2.3). It is worthy of note, however, that the longitudinal wave velocities in the horizontal directions are mainly retarded by the two vertical joint sets. For simplicity, therefore, the two joint sets are only taken into account without paying any respect to the horizontal one. Using the horizontal trace maps, the corresponding crack tensors are calculated. For example, for sites (A) and (B), UA,= /' 6.33 '~ \ --0.38 --0.38']. 4.00j" (28)

s (a) Change of squared velocity ratio

F~'=("

12.2-3.00" / - 3.00 5.76/'

X2 X2,X1

F, V"

(b) Joint trace m a p in site (A)

Fig. 13. Joint trace map of site (A) and the change of the squared wave velocity ratio with directions.
N

W:

o(o)

(a) Change of squared velocity ratio

X2

X2F'X~V,

,F,
x,V

(b) Joint trace map in site(B)


Fig. 14. Joint trace map of site (B) and the change o f the squared wave velocity ratio with directions.

il

The principal axes associated with these tensors are shown by x~r~ and xf,r) in Figs 13 and 14. It should be pointed out that the field data reported here are so limited that further study is required to reach any general conclusions. However. the following trends can provide a useful guide for further investigation: (i) In the jointed granites, the minor axis x~, r) of F~ tends to accord with the major axis x~v~ of Vie Such co-axiality is observed not only in the gypsum plaster samples but also in granites (ii) The anisotropy of Vi, is associated with the anisotropy of F~. In order to make this point clear, the relation between A "~ and A ~ i s shown by three squares in Fig. 9. [In site (C), as pointed out before, only two squared velocity ratios were measured. One is parallel to one vertical joint set with NS strike which only appears in this site, and one more is perpendicular to it. In the calculation of A'V~ for site (C), the two values of the squared velocity ratios are considered as the corresponding principal ones. unlike the other two sites.] Direct comparison of the data with those of the gypsum plaster samples, of course, is meaningless because F0 differs so much. It can still be said that the A~V~~ A t~ relation for the jointed gramtes is consistent with that for the gypsum plaster samples. Note the strong anisotropy of wave velocity in jointed gramtes. (iii) In Fig. 8. three squares give the relation between Iio and F0 for the jointed granites. It may safely be said that this relation depends on the rigidity &joints. If the rigidity is high enough to be comparable with the solid matrix, the increase of Fo do not cause any reduction of 1Io, as shown by the dotted line in Fig. 8. On the other hand. if the rigidity is low as in a case where joints are filled with soft weathered products, the Vo-F o relation may give a lower bound given by the broken line in Fig. 8. Data for site (A) are plotted close to the upper (dotted) line, while data for site (B) are near to the lower (broken) line. The result is quite consistent with the observation o f joints in these sites.

tm

CONCLUDING REMARKS Crack geometry, which is related to the anisotropy of cracked bodies, has been expressed in general form

ODA et al,:

CRACK TENSOR AND WAVE VELOCITY IN ROCK

~"

REFERENCES 1. Gerrard C. M. Background to .14athemattcal ~[o&'tin~ m Gc,)mechanics {Edited by Gudehus G.), Chap. 2. pp. 33-120 Wile>. New York (1977). 2. Rodrigues F. P. Anisotropy of granite. Proc. l~'t Com,,r. I.SR).[. Vol. 1, pp. 721-731. Lisbon (1966t. t r e a t e d as an a n i s o t r o p i c , elastic solid. T h r e e e x a m p l e s 3. Douglass P. M. and Voight B. Anisotropy of granite: a rellcction ( g y p s u m plaster s a m p l e s w i t h artificial cracks, g r a n i t e s of microscopic fabric. G&~technique 19, 376 398 (19691. with m i c r o c r a c k s a n d j o i n t e d granites) are r e p o r t e d h e r e 4. Thill R. E.. Bur T. R. and Steckley R. C. Velocity anisotrophy in w i t h special e m p h a s i s on the r e l a t i o n b e t w e e n the c r a c k dry and saturated rock sphere and its rdation to rock Pabric. b~r J. Rock Mech. Min. Sci. & Geomech. Ahstr. 10, 537~ 557 (19":31. g e o m e t r y and the a n i s o t r o p i c w a v e velocity, w i t h the 5. Friedman M. and Bur T. R. Investigations ol" the relations among following conclusions. residual strain, fabric, fracture and ultrasonic attenuation and (1) T h e c r a c k t e n s o r F,j has the f o l l o w i n g utility in velocity in rock. Int. J. Rock Mech. 3,[in. Sci. & Geomech. .4h~tr. It, 221-234 (1974). p r a c t i c a l r o c k m e c h a n i c s : it is a n o n - d i m e n s i o n a l t e n s o r 6. Wa[sh J. B. The effects of cracks on the compressibility of rock. c a r r y i n g a clear i m a g e related to the c r a c k g e o m e t r y . J. geophys. Res. 70(2), 381-389 (1965a). E v e n if a r o c k m a s s is intersected by v e r y c o m p l i c a t e d 7. Walsh J. B. The effect of cracks on the uniaxial elastic compression of rocks. J. geophys. Res. 70(2), 399 411 (1%5b). j o i n t sets, the t e n s o r m a k e s it possible to q u a n t i f y the 8. Budiansky B. and O'Connell R. J. Elastic moduli of a cracked c o r r e s p o n d i n g c r a c k g e o m e t r y in a c o n c i s e m a n n e r . solid, btt. J. Solids Struct. 12, 81 97 ([976). A l s o , the a n i s o t r o p y o f the c r a c k g e o m e t r y is g e n e r a l l y 9. Piau M. Attenuation of plane compressional wa'~e by a random distribution of thin circular cracks, hit. J. Em,,ng Sci. 17, 151 167 e x p r e s s e d in t e r m s o f the i n v a r i a n t s o f the tensor. (1979). (2) T h e c h a n g e o f s q u a r e d l o n g i t u d i n a l w a v e v e l o c i t y 10. Kachanov M. A micro-crack model of rock elasticity. (Part I: with d i r e c t i o n s in a n i s o t r o p i c r o c k masses c a n be Frictional sliding on micro-cracks.) Mech. Mater. l, 19 27 (19S2). I1. Horii H. and Nemat-Nasser S. Overall moduli of solids ~ith e x p r e s s e d by i n t r o d u c i n g an a d d i t i o n a l t e n s o r V,. U s i n g microcracks: Load-induced anisotropy. J. Mech. Ptn~. Solids the tensor, the a n i s o t r o p i c c h a r a c t e r i s t i c o f w a v e v e l o c i t y 31(2), 155 171 (1983). can be q u a n t i t a t i v e l y related to the a n i s o t r o p i c g e o m e t r y 12. Oda M., Suzuki K. and Macshibu T. Elastic compliance for rock-like materials. Soils F~hls 24(3), 27 40 (19S41. o f cracks. 13. Goodman R. E., Taylor R. L. and Brekkc T. L. Model for the (3) T h e present e x p e r i m e n t a l s t u d y yields s o m e intermechanics of jointed rock. J. Sod Mech. I=dtl~' Dit. ..Ira. Soc ('it, esting o b s e r v a t i o n s on the r e l a t i o n b e t w e e n k;, a n d V,,: Engrs 94(S~,13), 637-659 (1968). b o t h tensors are c o - a x i a l in the sense that the m a j o r axis 14. Zienkiewicz O. C. and Dullage C. Analysis o( non-linear problems in rock mechanics with particuIar reference to jointed rock ssstem. o f F~j a c c o r d s with the m i n o r axis o f V,~. T h e r e l a t i o n s Proc. 2nd Congr, on Rock Mcchanics, Vol. 3, pp. 50l 509 ([97(I). b e t w e e n F,~ and f/i,, which are e x p r e s s e d in t e r m s o f the 15. Brown E. T. Strength of models of rock v.,ilh intermittent joints. J. Soil Moth. Fdns Dir. Am. Soc. cir. I'n,zr~ 96(S~,!6), 1935 1949 i n v a r i a n t s , can be used to e x t r a c t useful i n f o r m a t i o n (1970). c o n c e r n i n g the g e o m e t r i c a l a n d m e c h a n i c a l p r o p e r t i e s o f 16. Einstein H. H. and llirschfcld R. ('. Model studies ou mcchamcs geological discontinuities from geophysical exploration. of jointed rock. J. Soil Mech. Fc~hr~ Dir ..Ira. Soc ,'il. [SnW'~ 99(SM3), 229-248 (1973). [7. Oda M. Fabric tensor for discontinuous geoh)gicaI materials..';,il,~ F&r~" 22(4), 96 I08 (1982). Acknowledgements--We would like to thank Dr T. Maiguma for instructing us in the detailed procedure of wave velocity measurements 18, Odu M. Similarity rule of crack geometry in statistically homt~geneous rock masses. Mech. Matcr. 3, li t) 129 (19S4) in situ. We would also like to thank Messrs Y. Hatsuyama, N. Hara, M. Koono and H. Naka~,ama for helping us with the field measure19. Oda M. Permeability tensor for discontinuous rock masses, (i,;otechnique 35(4), 483 495 (1985). ment. 20. Cowin S. C. The relationship between the elastich,, teus~r and the fabric tensor. Mech. Mater. 4, I 1[ (1985). 21. Deere D. V. Geological consideration. Ro~k Ale~ham~' in Engineering Practice (Edited by Stagg and Zienkic~icz), ('hap. 1, pp. 1-20. Wiley, New York (196S). 22. Kanatani K. Distribution of directional dat:.t alld fabric tei>ors Received 25 February 1986. Int. J. Engng Sci. 22(2), 149-164 (1984)

(called the crack tensor) by t a k i n g i n t o a c c o u n t the density, size and o r i e n t a t i o n o f cracks. It has been s h o w n t h r o u g h the p r e s e n t study t h a t the c r a c k t e n s o r c o n c e p t p r o v i d e s a p o w e r f u l tool w h e n a j o i n t e d r o c k m a s s is

You might also like