Sei sulla pagina 1di 18

International Journal of Rock Mechanics and Mining Sciences 36 (1999) 743760

www.elsevier.com/locate/ijrmms

Modications to the geological strength index (GSI) and their applicability to stability of slopes
H. Sonmez, R. Ulusay*
Geological Engineering Department, Applied Geology Division, Faculty of Engineering, Hacettepe University, 06532 Beytepe, Ankara, Turkey Accepted 10 May 1999

Abstract Determination of the strength of closely jointed rock masses is dicult since the size of representative specimens is too large for laboratory testing. This diculty can be overcome by using the HoekBrown failure criterion. Since its introduction in 1980, the criterion has been rened and expanded over the years, particularly due to some limitations in its application to poor quality rock masses. In the latest version, the geological strength index (GSI) was introduced into the criterion by its originators. However, the GSI classication scheme, in its existing form, leads to rough estimates of the GSI values. Another particular issue is the use of undisturbed and disturbed rock mass categories for determining the parameters in the criterion, for which clear guidelines are lacking. Furthermore, the data supporting some of these revisions, particularly the latest one, have not been published, making it dicult to judge their validity. In this study, in order to provide a more quantitative basis for evaluating GSI values, some modications are suggested by introducing easily measurable parameters with their ratings and/or intervals which dene the blockiness and surface condition of discontinuities. In addition, a method is proposed to assess the inuence of disturbance on rock mass constants due to the method of excavation. The modications to the GSI and the suggested method have been applied to slope instability case histories selected from Turkey by performing back analysis, to discuss the validity of the criterion and the methodology of parameter estimation. It was shown that the failure conditions in each case were conrmed, i.e. the analysed failure surfaces satised factors of safety of unity, when the suggested modications and disturbed rock mass condition are considered. On the basis of the results, a chart to assess the eect of disturbance in terms of method of excavation was also suggested. The back analysis of a spoil instability indicated that spoil pile materials consisting of blocky and angular rock pieces could be categorized as a disintegrated rock mass in the GSI classication and the criterion seemed to be applied to such materials. The method suggested herein must, however, be veried by additional data from slope failures before more precise guidelines can be formulated. # 1999 Elsevier Science Ltd. All rights reserved.

1. Introduction The standard method for assessing the strength of a geotechnical material is to recover representative samples and test them in the laboratory. In the case of a closely jointed rock mass it is clearly not possible to recover a sample that is large enough to represent the joint system. In order to overcome the diculties in laboratory determination of the shear strength of jointed rock masses, the empirical HoekBrown failure
* Corresponding author. Tel.:+90-29-777-62. E-mail address: resat@hacettepe.edu.tr (R. Ulusay)

criterion [1] is commonly used in conjunction with the Geomechanics Classication System [2]. This failure criterion has been rened and expanded over the years [37] as summarized in Table 1. The limitations in Bieniawski's RMR classication scheme [2] for very poor quality rock masses and for unrealistic rating adjustments for discontinuity orientation in slopes have necessitated some signicant changes in the criterion. This is probably one of the main reasons why the originators of the technique continue modifying their criterion. Recently, Hoek and Brown [69] proposed the geological strength index (GSI) based upon the visual impression on the rock mass structure. Fig. 1 shows twenty codes to identify each rock mass

0148-9062/99/$ - see front matter # 1999 Elsevier Science Ltd. All rights reserved. PII: S 0 1 4 8 - 9 0 6 2 ( 9 9 ) 0 0 0 4 3 - 1

744

H. Sonmez, R. Ulusay / International Journal of Rock Mechanics and Mining Sciences 36 (1999) 743760

Fig. 1. Characterization of rock masses on the basis of interlocking and surface condition of discontinuities: GSI classication (rearranged from Tables 3 and 4 given by Hoek and Brown [8]).

category and estimate the GSI value. It is also noted, on the basis of their recent studies on the Athens schist, Hoek et al. [9] introduced a new rock mass category into the GSI system called `foliated/laminated rock mass structure'. This new category accommodates thinly foliated, folded and predominantly sheared weak rocks of non-blocky structure. The equivalent GSI contours range from a new value of 5 up to 30 in the lower right portion of the disintegrated rock mass category. The latest version of the GSI chart [8] (Fig. 1) is sufcient for eld observations, since it is only necessary to note the letter code which identies the rock mass category. The GSI also seems a more practical parameter to estimate the strength of jointed rock masses

from eld observations when compared to the method employing rock mass classication. Because rock mass classication requires time consuming procedures and has some limitations as discussed by Sonmez et al. [10] in detail. However, due to lack of measurable and more representative parameters, and related interval limits or ratings for describing the surface conditions of the discontinuities, value of the GSI for each rock mass category appearing in Fig. 1 represents a range of values. For example, for a blocky rock with very good surface condition of discontinuity (B/VG), GSI values varying between 63 and 85 are obtained from Fig. 1. This consideration placed focus on the question ``how can a more precise GSI value be obtained from the existing chart for design?''. Hoek [11] indicates that

H. Sonmez, R. Ulusay / International Journal of Rock Mechanics and Mining Sciences 36 (1999) 743760 Table 1 Historical development of the HoekBrown criterion (rearranged from Hoek and Brown [8]) Publication Hoek and Brown [1] Coverage original criterion for heavily jointed rock masses with no nes; Mohr envelope was obtained by statistical curve tting to a number of (sn ' t ) pairs calculated by the method published by Balmer [28]. s1 ', s3 ' are major and minor eective principal stresses at failure, respectively; s1 is the tensile strength of the rock mass, m and s are material constants; sn ', t are eective normal and shear stresses, respectively original criterion for heavily jointed rock masses with no nes with a discussion on anisotropic failure and solution for the Mohr envelope by Dr. J.W. Bray as for Hoek [3] but with the addition of relationships between constants m and s and a modied form of RMR (Beniawski) [2] in which the groundwater rating was assigned a xed value of 10 and the adjustment for joint orientation was set at 0; also a distinction between disturbed and undisturbed rock masses was introduced together with means of estimating deformation modulus E modied criterion for account for the fact the heavily jointed rock masses have zero tensile strength; Balmers technique for calculating shear and normal stress pairs was utilised introduction of the generalised HoekBrown criterion, incorporating both the original criteiron for fair to very poor quality rock masses and the modied criterion for very poor quality rock masses with increasing nes content; The geological strength index, GSI, was introduced to overcome the deciencies in Bieniawski's RMR for very poor quality rock masses; the distinction between disturbed and undisturbed rock masses was dropped on the basis that disturbance is generally induced by engineering activities and should be allowed for by downgrading the value of GSI as for Hoek [6] and Hoek et al. [7], but with the addition of a chart to estimate GSI based upon the visual impression of the rock structure and the surface condition of the discontinuities indicated by joint roughness and weathering as for Hoek and Brown [8], with the addition of a new `foliated/ laminated' rock mass structure category to accommodate thinly laminated or foliated, folded and predominantly sheared weak rocks of non-blocky structure Equations

745

p H H H s1 s1 s3 sci 3 asci s; p ms H t Asci sn sci a2m m2 4s; st asci B ; H H H H H H H sn s3 s3 a1 d s1 ad s3 ; t sn p Hs1H H H H H H d s1 ad s3 msci a2s1 s3 s3 d s1 d s3 ; p H H H s1 s3 sci ms3 asci s; tp cot fiH H H cos fi msci a8; fi arctan1a 4hcos2 y 1; p y 90 arctan1a h3 H 1a3; h 1 16msn ssci a3m2 sci disturbed rock masses: mb/mi=exp((RMR100)/14); s = exp((RMR100)/6); undisturbed or interlocking rock masses: mb/mi=exp((RMR100)/28); s = exp((RMR100)/9); E = 10((RMR10)/40); mb, mi are for broken and intact rock, respectively H H H H H H s1 s3 sci mb s3 asci a ; sn s3 s1 H H H s3 a1 d s1 ad s3 ; p H H H H H t sn s3 d s1 ad s3 ; d s1 a H 1 amb s3 asci a1 s1 '=s 3 '+sc (ms3 '/sci+s )a; for GSI>25; mb/mi=exp((GSI100)/28); s = exp((GSI100)/9); a=0.5; for GSI < 25; s = 0; a=0.65GSI/200

Hoek [3]

Hoek and Brown [4]

Hoek et al. [5]

Hoek [6] and Hoek et al. [7]

Hoek and Brown [8] Hoek et al. [9]

as for Hoek [6]

as for Hoek [6]

although some geologists go to extraordinary lengths to try to determine an `exact' value of GSI or RMR, geology does not lend itself to such a precision and it is simply not realistic to assign a single value. He also states that for preliminary eld investigations or lowbudget projects, it may be prudent to assume larger standard deviations for the input parameters (uniaxial compressive strength of the intact rock, intact rock material constants and GSI) and they can be represented by normal distribution. Although its originators have always pointed out the criterion's approximate nature, it seems questionable to obtain a mean value represented by a normal distribution from the existing form of the GSI chart. It is also considered that it is possible to estimate dierent GSI

values for the same rock mass by dierent persons, depending on their personal experience, when the chart given in Fig. 1 is employed. From the review of the criterion, it is clear that the continuous update of the HoekBrown failure criterion has not been complemented by equal eorts to verify the same. Furthermore, the data supporting of these revisions have not been published, making it difcult to judge their validity. One important issue is use of the undisturbed and disturbed rock mass categories when determining the parameters in the criterion for which clear guidelines are lacking. Any disturbance on the rock mass due to some local factors (e.g. blasting, the presence of discrete fault zones, etc.) should be considered and, therefore, an adjustment should be

746

H. Sonmez, R. Ulusay / International Journal of Rock Mechanics and Mining Sciences 36 (1999) 743760

applied to RMR values [2]. Although the rock mass classication schemes and the HoekBrown criterion have been generally applied in tunnelling and underground mining, the back analyses of slope failures in heavily jointed rock masses, by assuming disturbed rock mass conditions showed good agreement between the estimated and the back-calculated strength parameters from the observed slope failures [10]. On the contrary, in the latest version of the criterion [7,8] average undisturbed in situ conditions are considered to estimate the GSI without application of any adjustment. Hoek and Brown [8] indicated that one of the practical problems which arose when assessing the value of GSI in the eld was related to blast damage. According to these investigators, where all the visible faces have been damaged by blasting, some attempt should be made to compensate for the lower values of GSI obtained from such faces. Also, in recently blasted faces, new discontinuity surfaces occurring due to blasting will give a GSI value which may be as much as 10 points lower than that for the undisturbed rock mass. Therefore, Hoek and Brown [8] suggest that severe blast damage can be allowed for by moving up one row in Fig. 1. This approach may be correct for the estimation of GSI from blasted rock exposures during excavation. However, moving the GSI value up one row seems to be a rough approach and also results in an increase in uniaxial compressive strength of the rock mass by more than 70%. Therefore, the reason of this assumption is still open to discussion. On the other hand, method of excavation, major planes of weakness or change in stress are treated as local features which have inuenced the rock mass at a particular location, and are not rock mass constants [1214]. Therefore, in order to compensate the inuence of such local factors, necessary adjustments should be taken into consideration. An additional practical question arising from the latest version of the Hoek and Brown's approach is ``how can the inuence of the method of excavation can be taken into account by using the existing HoekBrown's equations which consider only undisturbed rock mass, when the GSI is estimated from borehole cores or natural exposures before excavation or blasting?''. It is also noted that there is no any published case history on the back analysis of slopes in heavily jointed rock masses which conrms that the latest GSI classication yields satisfactory results when an adjustment factor is not taken into consideration. This paper is an attempt to provide a more quantitative numerical basis for evaluating GSI by introducing new parameters, and ratings, such as surface condition and structure rating. For meaningful interpretation and for providing a common basis for communication between engineers and designers, standard interval limits and ratings for the input par-

ameters are suggested. In addition, a method is proposed to assess the inuence of disturbance. The modications and the suggested method have been applied to slope instability case histories from Turkey to discuss the validity of the criterion and the methodology of parameter estimation.

2. Suggested modications for estimating GSI values Once the GSI has been estimated, the parameters which describe the rock mass strength characteristics, are calculated as follows:   GSI 100 1 mb mi exp 28 For GSI>25, i.e. rock masses of good to reasonable quality,   GSI 100 2 s exp 9 and a 0X5 3

For GSI < 25, i.e. rock masses of very poor quality, the criterion applies with s0 a 0X65 GSI 200 4 5

From the above equations it is clear that the rock mass strength parameters are sensitive to the GSI value. The lack of parameters to describe surface conditions of the discontinuities and the rock mass structure prevents to obtain a more precise value of GSI. For these reasons, the authors suggest two terms namely, `structure rating, SR' based on volumetric joint count (Jv) and `surface condition rating, SCR', estimated from the input parameters (e.g. roughness, weathering and inlling). The suggested ratings by the RMR system [2] for these parameters are selected for the purpose. According to the rating of each input parameter (Rr, Rw and Rf ) estimated from the right upper margin of the table given in Fig. 2, the total rating for surface conditions (SCR) is obtained using the following expression: SCR Rr Rw Rf 6

where Rr, Rw and Rf denote the ratings for roughness, weathering and inlling, respectively. Since the sum of

H. Sonmez, R. Ulusay / International Journal of Rock Mechanics and Mining Sciences 36 (1999) 743760

747

Fig. 2. The modied GSI classication suggested in this study.

the maximum ratings of these parameters is 18, the SCR axis in Fig. 2 is divided into 18 equal divisions. In the earlier version of the criterion (Fig. 1), Hoek et al. [7] used the terms BLOCKY/SEAMY and CRUSHED, following the terminology proposed by Terzaghi [16]. After they recognized that these terms proved to be misleading, they have been replaced, by BLOCKY/DISTURBED and DISINTEGRATED.

The authors agree with this change. On the other hand, Hoek et al. [9] proposed a new rock mass category to accommodate thinly foliated or laminated, folded and predominantly sheared weak rocks of nonblocky structure. However, Hoek [6] emphasizes that the criterion is only applicable to intact rock or to heavily jointed rock masses which can be considered homogenous and isotropic. On the contrary, the

748

H. Sonmez, R. Ulusay / International Journal of Rock Mechanics and Mining Sciences 36 (1999) 743760

strength and deformability characteristics of such rock masses are governed by the displacements along the numerous very thinly spaced presheared and slickensided foliation planes. Due to anisotropic and inhomogenous features of such rocks, introducing of this new category into the GSI scheme seems not to be realistic. The other questionable issue is that the validity of the GSI values assigned for this new category has not been conrmed yet by case studies. Therefore, only four structural categories as previously suggested by Hoek and Brown [8] are considered in this study. Block size is an extremely important indicator of a rock mass. Large blocks tend to be less deformable and develop favourable arching and interlocking in underground openings. In the case of slopes, a small block size may cause rotational slides instead of structurally controlled modes of failure. Block dimensions are determined by three rock mass parameters, namely discontinuity spacing, the number of discontinuity sets and the persistence of the discontinuities delineating potential blocks. However, in order to decrease the number of inputs, the use of a single parameter which can take into account one or two of the above mentioned parameters was considered to be more practical. Thus, volumetric joint count (Jv), which is dened as the sum of the number of joints per meter for each joint set present, is suggested to be used for the description of structure of the rock mass. Jv is estimated by one of the following expressions: Jv N1 N2 Nn FFF L1 L2 Ln 1 1 1 FFF S1 S2 Sn 7a

counted along the scanlines (Lx, Ly and Lz ) perpendicular to each other. However, in some cases it can be dicult to nd exposures along which three scanline surveys in perpendicular directions can be carried out. In such circumstances, by assuming the rock mass is homogeneous (i.e. the terms appearing in Eq. (7c) are equal to each other), Eq. (7c) can be rewritten in the following form.  Jv N L 3 7d

The intervals of Jv and related descriptions suggested by ISRM [15] were adopted for the blockiness categories to be used in the GSI classication (Table 2). Based on the intervals of Jv and corresponding descriptions for the blockiness ratings, structural rating (SR) was assigned to each category by the following procedure. 1. Using a semi-logarithmic sheet, SR and Jv are put on y and x axes, respectively. 2. While the SR axis is divided into the ratings ranging from 0 to 100, logarithmic Jv axis is divided according to the boundaries suggested for four structural categories as described in Table 2. The upper limit on the Jv axis is selected as 104 to consider pebble size. 3. Since the boundaries between the structural categories in the existing GSI table are equally divided (Fig. 1), the SR limits between the codes B-VB, VBB/D and B/D-D are selected as 75, 50 and 25, respectively. 4. By plotting the Jv values for each category as suggested in Table 2 against the boundary values of SR mentioned in item (c), the curve shown in the left margin of Fig. 2 is obtained. This curve can be used to assign a rating for SR of any rock mass using the value of Jv. It is now possible to estimate a more precise GSI value from the intersection point of SCR and SR ratings when the modied GSI chart (Fig. 2) is used.
Table 2 Descriptive terms corresponding block size and intervals of Jv suggested by ISRM [15] and by the authors of this study Descriptions by ISRM Jv (joint/m3) Descriptions for GSI (this study) Very large blocks Large blocks Medium sized blocks Small blocks Very small blocks Crushed <1 13 310 1030 3060 >60 BLOCKY (B) BLOCKY (B) VERY BLOCKY (VB) BLOCKY/DISTURBED (B/D) DISINTIGRATED (D) DISINTIGRATED (D)

Jv

7b

where S is the true spacing, N is the number of joints along a scanline, L is the length of the scanline and n is the number of joint sets. On the other hand, estimation of Jv for heavily jointed rock masses with no identiable structural pattern is extraordinarily dicult. Since the discontinuities in such rock masses do not introduce considerable dierences in their spacing in all directions, they can be assumed as homogeneous and isotropic. Therefore, expressions given in Eqs. (7a) and (7b) are not advised to be used for the determination of Jv. Instead of these, the authors suggest the following approach which is more practical in the estimation of the number of discontinuities in a rock mass with a volume of 1 m 3. Jv Nx Ny Nz Lx Ly Lz 7c

where Nx, Ny and Nz are the number of discontinuities

H. Sonmez, R. Ulusay / International Journal of Rock Mechanics and Mining Sciences 36 (1999) 743760

749

3. Validity of the HoekBrown estimates of rock mass and their impact on the assessment of stability of slopes 3.1. Theoretical background of the suggested methodology for the assessment of disturbance eect Some factors, such as method of excavation, major planes of weakness or change in stress, are treated as local features inuencing the rock mass at a particular location and are not rock mass constants. These have been discussed by Laubscher [12], Romana [13] and Kendorski et al. [14]. The greatest inuence of the method of excavation will be on the spacing of discontinuities. Depending on the blasting damage, blasted slopes may have closer discontinuity spacing than natural slopes. Therefore, in order to compensate the inuence of such local factors, necessary adjustments [2,1214] are taken into consideration in rock mass classication. Sonmez et al. [10] showed that by assuming an adjustment factor based on the method of excavation (disturbed rock mass), a good agreement was found between the estimated and the back-calculated strengths from the observed slope failures and, therefore, eect of disturbance should be taken into consideration. In order to check the validity of the equations used for the rock mass constants and to assess the eect of disturbance, four approaches with the use of modied GSI chart have been suggested and applied to failed slopes from Turkey (Fig. 3). Four cases were selected from the failures of the pit slopes in heavily jointed rock masses where joint spacing is a fraction of a meter and one is from the failures occurred in spoil piles in Turkey. The authors suggest that spoil piles composed of rock materials possess the behaviour of poorly interlocked, heavily broken rock masses with a mixture of angular and rounded rock pieces and, therefore, can be considered in disintegrated rock mass category. On the basis of this assumption, a spoil pile failure was also examined in this study. The back analysis procedure starts with the determination of the GSI value of the rock mass investigated in each case from the modied chart (Fig. 2) according to the suggested input parameters. Then the following approaches are employed in order to check the validity of the equations: . Approach 1: the shear strength parameters of the failed rock masses are estimated using the original expressions given by Eqs. (15) without application of any adjustment for the method of excavation, i.e. undisturbed rock mass condition is assumed. These parameters are then used in the back analyses for the calculation of factor of safety. . Approach 2: a disturbance (adjustment) factor of (df ) depending on the method of excavation [12,14]

is assigned in each case and multiplied by the GSI value appearing in the denominator of Eqs.(1), (2) and (5). The parameters determined in this way are then employed in the back analyses. . Approach 3: a disturbance factor (df ) is assigned in each case and multiplied by the numbers appearing in the numerator (28 and 9) of Eqs. (1) and (2) for the estimation of the rock mass parameters, mb and s. . Approach 4: in the previous form of the criterion the numbers appearing in the denominator of the equations are 28 and 14 for mb, and 6 and 9 for s for undisturbed and disturbed rock mass conditions, respectively (Table 1). In other words, it seems to be logical to conclude that the denominator of these equations vary between 14 and 28, and 6 and 9 depending on the degree of disturbance. Based on this fact, Eqs. (1) and (2) can be rewritten in the following forms: GSI 100 mb mi exp bm   bm 14 28 8

 s exp

GSI 100 bs

 bs 6 9 9

In approach 4, ve dierent values are assigned to bm (starting from 28 to 14) and to bs (starting from 9 to 6) and then the value of factor of safety (FOS) corresponding to the pairs of bm and bs for each particular case history is calculated. The results of the analyses are presented in the form of FOSbm and FOSbs curves. From these curves, bm and bs values which lead a value of factor of safety of unity are obtained for each case as depicted in Fig. 4(a). The values of bm and bs from the curves are then plotted against corresponding disturbance factors (df ) to establish a relationship between bm, bs and df (Fig. 4b), and to check the validity of the HoekBrown estimates. 3.2. Software description In this study, a computer program, HOBRSLP developed and described by Sonmez et al. [10] was employed. The program HOBRSLP was modied for this study to include the approaches described above. It can handle slope stability analysis of circular and non-circular slip surfaces for slopes involving many benches with dierent geometries, various materials and dierent groundwater conditions. It also incorporates external loading conditions.

750 H. Sonmez, R. Ulusay / International Journal of Rock Mechanics and Mining Sciences 36 (1999) 743760

Fig. 3. Location map of the back analysed case study sites and views from the investigated slope instabilities: (a) initiation of the instability in the highwall externally loaded by a spoil pile in Eskihisar strip coal mine; (b) a view from the heavily broken schist rock mass at Baskoyak barite mine; (c) a view from the jointed rock mass in Kisrakdere open pit coal mine; (d) bench failure in a closely jointed marly rock mass in Himmetoglu lignite open pit mine and (e) a view from the slope instability in a spoil pile at Eskihisar strip coal mine.

H. Sonmez, R. Ulusay / International Journal of Rock Mechanics and Mining Sciences 36 (1999) 743760

751

Fig. 5. The model with the parameters for the slope under the inuence of a symmetrical vertical triangular spoil loading in the back analysis of the failed slope in case 1.

Fig. 4. Basic concept of approach 4 to check the validity of the HoekBrown equations.

3.3. Description of the rock mass conditions and the examined slope instabilities 3.3.1. Case 1: an externally loaded highwall slope failure This case history involves the instability of a highwall in Eskihisar (Yatagan-Mugla) strip coal mine in southwestern Turkey. During a comprehensive slope stability research project by Ulusay [17], an instability of the southern part of the highwall of the ninth slice has occurred due to external loads by a temporary spoil pile (Fig. 3a). The failed slope was excavated in marl which lies above the coal seam with a thickness of 1520 m.

There are highly persistent (about 8 m) three dominant joint sets developed parallel and/or subparallel to normal faults crossing the Tertiary deposits. The presence of cross joints, faults and at lying bedding planes results in a closely jointed rock mass. The groundwater level rises above the coal seam and tends to decline toward the marlcoal seam boundary. Thus, the failed part of the investigated slope was dry. The structure rating (SR) and surface condition rating (SCR) were estimated from the scanline survey data obtained by Ulusay [17] (Table 3). In this pit a controlled blasting is carried out with slight damage to loosen the marly overburden. For this condition, a blasting damage adjustment of 0.94 [14] was assigned. The average uniaxial compressive strength and mi of the intact marl specimens were 4.15 MPa and 9.87, respectively [17]. Based on available monitoring records by Ulusay [17] the model depicted in Fig. 5 for the slope under the inuence of a symmetrical vertical triangular spoil loading, and the failure surfaces illustrated in Fig. 6 were used in the analyses. Average values of unit weight of 13 and 16 kN/m3 were utilized for the spoil material (in-situ) and the marls, respectively. It was reported by Ulusay [17] and Sonmez et al. [10] that the back analyses, using the previous form of the HoekBrown estimates [4], which includes the RMR scheme, conrmed these failure surfaces. 3.3.2. Case 2: slope failure in a closely jointed rock mass at a barite open pit mine A comprehensive slope stability project was carried out at Baskoyak barite open pit mine, in western Anatolia between 1987 and 1988 by Ulusay and Yucel [18]. Based on the scanline surveys and a geotechnical borehole, it was reported that the schist should be regarded as comprising two types of rock mass [18]. The rst type consists of a heavily broken schist rock mass by closely spaced discontinuities and schistosity planes (Fig. 3b) and the second type is a weathered schist in dierent degrees. Due to the heavily jointed

752

H. Sonmez, R. Ulusay / International Journal of Rock Mechanics and Mining Sciences 36 (1999) 743760 smooth surfaces, (1), slightly to moderately weathered (4), soft coating < 5 mm (2) 1773 4 7 26 0.80 smooth to slickensided surfaces (1), highly weathered (1), soft coating < 5 mm (2) 15635 0 4 16 0.97 Condition of discontinuities and ratings Jv SR SCR GSIb df d slickensided surfaces (0), moderately weathered (3), soft coating < 5 mm (2) 13.3 35 5 27.5 0.97 True spacing (S1, S2, S3 for joints, Sb for bedding planes). Estimated by method of photoanalysis along x, y and z axes. b GSI determined from the modied chart in Fig. 2. d Adjustment factor for disturbance eect.
c a

Case 5 Case 4

nature of the schist, the rock mass was assumed as homogeneous and isotropic with a joint spacing of 0.04 m in all directions and, therefore, Jv of the rock mass was estimated by using Eq. (7d). The mean unit weight and uniaxial compressive strength of the heavily broken part of the schist are 22.2 kN/m3 and 5.2 MPa, respectively. The rock mass properties are tabulated in Table 3. No sign of groundwater was encountered through the geotechnical and previously drilled boreholes and on the pit benches. Thus, the pit slopes was considered as dry for stability assessments. Since the overburden material and the ore are removed by excavators without any blasting, an adjustment factor of 0.97 [14] was considered. One of three failures which occurred in closely jointed rock mass along a

Spacinga (m)

Parameters

S1=0.71, S2=0.82, S3=1.26, Sb=0.65 smooth surfaces (1), slightly weathered (5), soft coating < 5 mm (2) 6.14 63 8 43 0.94

Case 1

Sx,y,z=0.04

Case 2

Fig. 6. Slope proles, and the predicted and calculated failure surfaces employed in the back analysis of the externally loaded highwall slope in case 1.

Table 3 The parameters employed in the GSI classication for ve cases considered in this study

S1=0.75, S2=1.07, S3=0.13, Sb=0.4 smooth surfaces, (1), slightly weathered (5), soft coating < 5 mm (2) 12.5 42 8 37 0.90

Case 3

S1=0.37, S2=0.65, Sb=0.11

Sx=0.085c, Sy=0.081, Sz=0.083

H. Sonmez, R. Ulusay / International Journal of Rock Mechanics and Mining Sciences 36 (1999) 743760

753

strength of 40 MPa and mi of 9.04 has a carbonate content more than its clay content. The observed actual slip surface was in circular shape and passed through the compact marl rock mass and along the clay bed, above the coal seem. Bedding planes dip into opposite direction of the slope. Three main joint sets moderately and closely spaced, and bedding planes in the marly sequence resulted in a jointed rock mass. The rock mass characteristics and the geomechanical parameters of the clay are listed in Table 3.
Fig. 7. Slope geometries before and after the failure, and critical slip surface in closely jointed schist rock mass (case 2).

circular surface was selected for this study. The results of the back analysis [18,10] indicated that the calculated sliding surface conrms the actual failure surface delineated from the site measurements (Fig. 7). 3.3.3. Case 3: a slope instability in a coal mine at western Turkey A slope instability from the Kisrakdere open pit mine located at Soma lignite basin, western Turkey, was selected for the purpose of this study (Fig. 3c). The necessary data were collected by the authors from this pit. Fig. 8 shows the geometry of the failed slope in which a single thin coal seam with a thickness of 4.5 m is overlain by a sequence consisting of compact marl and soft clay beds about 10 m thick. The observations on the slope surfaces and available records indicated that the groundwater was below the failed marly rock mass, and the coal seam acted as an aquifer. The marly rock with a uniaxial compressive

Fig. 8. Cross-section illustrating the geometry of the failed slope and the position of the strata (case 3).

3.3.4. Case 4: a bench failure in a coal mine Himmetoglu open pit coal mine, operated by the Turkish Coal Enterprises (TKI), is located in northwest Anatolia and produces low caloric value of coal. A local bench failure occurred in 1998 in the eastern slope, excavated in heavily jointed marly rock mass, as a result of steepening of the slope (Fig. 3d). On the basis of the scanline surveys [19], the parameters of discontinuities given in Table 3 were obtained and a GSI value of 27.5 for the rock mass was determined. Detailed instability plan and cross-section of the failed bench are shown in Fig. 9. The visible part of the failure surface was in circular form. Detailed hydrogeological investigations [19] indicated that the slope was dry. Since the overburden material was removed by excavators without blasting, an adjustment factor of 0.97 was considered in this case. The back analysis of the failure surface (surface 1 in Fig. 9) showed that circular failure did not appear as a realistic mechanism for this instability with a factor of safety considerably greater than unity. The position of the oor strata dipping towards the excavation and the visible upper part of the sliding surface indicated the possibility of another mode of failure by combination of a planar sliding surface along the weak oor strata and a circular failure surface through the rock mass (Fig. 9; failure surface 2). The back analysis of the multiplanar failures along both bedding planes and the faults in this pit indicated that the residual shear strength parameters of the weak and slickensided bedding planes were cr=1.4 kPa and f=128 [19]. By employing these parameters, rock mass properties of the marls and Janbu's method of analysis [20] for this combined failure surface, a back analysis was performed. The analysis which yielded a factor of safety of unity indicated that a combined failure surface was the realistic mode of failure for this instability. After the removal of the failed material, the combined failure surface clearly appeared and conrmed the surface labeled 2 in Fig. 9 as the real failure surface. Therefore, the parameters given in Table 3 and the predicted mode of failure were employed in this study for further assessments.

754

H. Sonmez, R. Ulusay / International Journal of Rock Mechanics and Mining Sciences 36 (1999) 743760

Fig. 9. Plan of the shallow-seated bench failure in a heavily jointed rock mass (a) and cross-section of the stability (b) deduced from surveying (case 4).

3.3.5. Case 5: spoil pile instability in a strip coal mine Spoil piles in the Eskihisar mine, suer from numerous problems. Engineering geological characterization of the waste material and stability of the piles in this mine were investigated by Ulusay et al. [2123], who reported deep-seated rotational instabilities covering agricultural areas, shallow-seated rotational failures both sides of the haul road and bi-planar wedge failures along the operating slices. The authors of the present article consider that spoil pile materials can be dened as poorly interlocked and heavily broken rock

masses with a mixture of angular and rounded rock pieces, unless they do not contain high proportion of nes as a result of hauling, dumping and subsequent deformation. This approach indicates that the categorization of spoil piles mainly consisting of rock material as an disintegrated rock mass and the use of the HoekBrown criterion to estimate the shear strength parameters of such materials seems to be logical. Therefore, a selected spoil instability from the Eskihisar mine was also employed for this investigation. For this purpose, an instability occurred along the haul road (Fig. 3e) was backanalysed. The spoil pile consisted of marly rock. The cross-sections prepared from the instability plan by Ulusay et al. [21,22] revealed that the failure did not involve the foundation material. The curvature of the exposed sliding surface and slightly curved escarpments in plan conrmed a rotational type of failure has occurred (Fig. 10). No water table or seepage was encountered in the pile. Inplace unit weight determination indicated a mean value of 14 kN/m3 for the spoil material [21]. The main question in the case of spoil materials is ``how can the structural rating in conjunction with Jv be estimated?''. It was considered that determination of the grain size distribution of the fragments in the spoil, (i.e. measurement of the distance between the boundaries of the rock fragments along the selected directions and calculation of the average value), can be used as a practical and economic method to estimate Jv. Although sieving the excavated or blasted rock through screens is the most direct method to determine rock fragmentation, the cost of this method is high. Photoanalysis is one of the more recent and well established methods [2426]. In this study the method of photoanalysis has been employed. A test site near the investigated spoil pile was selected for the fragmentation analysis. Using a shovel truck, a small sized pile was dumped and photographs of this pile from its two sides were taken by a 35 mm camera. Since there were no established gures in the literature on the minimum number of photographs that were needed to accurately sample a given volume of rock fragments, analysis of two randomly selected photographs was considered to be sucient for the purpose of this study. Attention was paid to keep the camera perpendicular to the surface of which photograph was taken. In each photograph a scaling object (a circular plate) and a reference area (a wooden mesh) were used (Fig. 11). Outlines of the rock fragments were then digitized into the computer. Along the x and y axis (Fig. 11), all fragments larger than approximately 2 cm were automatically dimensioned. In order to measure the dimensions along the third axis (z ), the same process was applied on the photograph taken in perpendicular direction to the previous one. The information obtained was used in the statistical

H. Sonmez, R. Ulusay / International Journal of Rock Mechanics and Mining Sciences 36 (1999) 743760

755

Fig. 10. (a) Plan of the shallow-seated spoil instability along the haul road and (b) cross-sections of the spoil pile showing the failure surfaces and pile geometries (case 5; after Ulusay et al. [21]).

756

H. Sonmez, R. Ulusay / International Journal of Rock Mechanics and Mining Sciences 36 (1999) 743760

Table 4 The results of the back analysis of the failed slopes based on dierent approaches to assess the eect of disturbance. The values in parentheses indicate the average values for three sections df b Case No. Calculated factor of safety (FOS) approach 1 approach 2 approach 3 approach 4: at limiting equilibrium conditiona bm Case 1: Section Section Section Case 2: Case 3: Case 4: Case 5: Section Section Section Section
b a

bs 7.05 (6.98) 6.90 (6.98) 6.98 (6.98) 7.34 6.68 7.35 6 6 6

1-1'/1 1-1'/2 2-2'

0.94

1-1' 2-2' 3-3' 4-4'

0.97 0.90 0.97 0.80

1.48 1.48 1.45 1.70 1.41 1.32 2.71 2.64 2.64 2.69

1.44 1.43 1.40 1.62 1.34 1.23 1.93 1.80 1.84 1.84

1.42 1.41 1.39 1.59 1.28 1.19 2.10 2.03 2.04 2.07

18.9 (18.55) 18.2 (18.55) 18.55 (18.55) 20.28 17.15 20.3 14 14 14

Adjustment factor for disturbance eect. Obtained from Fig. 12.

analyses. By putting the mean fragment sizes of 0.085, 0.081 and 0.083 m calculated for x, y and z axes, respectively, into Eq. (7c), a Jv value of 1773 which represents a disintegrated rock with a very low structure rating of 4 was obtained. A surface condition rating of 7 and a GSI value of 26 were estimated. 3.4. Back analysis of the selected slope instabilities The results obtained from the back analysis of these slope failures are evaluated by following the steps summarized below. . Step 1: using the GSI value for each case (Table 3), the HoekBrown constants were calculated from the original equations (Eqs.(1), (2) and (5)) which do not consider the disturbance eect or an adjustment factor. The factors of safety, based on these parameters tabulated in the rst column of Table 4, are considerably greater than unity and indicate that failure can not occur through these slopes. This situation focuses the attention on the fact that the negligence of an adjustment factor (df ) yields unrealistic assessments for stability of slopes, and therefore, a modication based on df value seems to be necessary for HoekBrown equations. . Step 2: using the same model, the analyses including an adjustment factor for each case (Table 3) were performed. However, in this step, the adjustment factor was multiplied by the GSI value to consider the disturbance eect on the rock mass, as applied in the RMR scheme. Thus, Eqs.(1), (2) and (5) were rewritten in the following forms:

mb mi  s exp

GSI df 100 exp 28 GSI df 100 9 GSI df 200 

 10a

10b

a 0X65

10c

The results of the back analysis of the failed slopes yielded factors of safety considerably greater than unity suggesting that the slopes were stable (Table 4). This approach indicated that the above modication did not satisfy the failure condition. . Step 3: in this step, the eect of disturbance was assigned onto the numbers appearing in the numerators of the Eqs. (1) and (2) as follows:   GSI 100 11a mb mi exp 28 df GSI 100 s exp 9 df   11b

The results of the analysis indicated that the values of the factor of safety calculated for each case (Table 4) were still greater than unity. . Step 4: in this step, considering the results obtained from step 3, ve dierent values were assigned to bm and bs in Eqs. (8) and (9), ranging between 14 and 28, 6 and 9, respectively. Then the values of factor of safety corresponding to selected values of bm and bs for each particular case were calculated to esti-

H. Sonmez, R. Ulusay / International Journal of Rock Mechanics and Mining Sciences 36 (1999) 743760

757

Fig. 11. A view from the spoil material taken for image processing, and the scaling factor (case 5).

mate the pairs of bmbs satisfying the limit equilibrium condition. Since the disintegrated material forming the spoil piles were the mixture of excavated, blasted, hauled and dumped overburden, a lower bound adjustment factor of 0.8 was assigned for this material. Besides, the values of 14 and 6, which were suggested as the lower bounds in the original HoekBrown equations [4], were employed. Therefore, for dierent values of bm and bs for the spoil pile instabilities mentioned in case 5 the trial and error method was not used. The back analysis of these instabilities yielded values of factor of safety equal to unity, indicating that the suggested approach seemed to be satisfactory. The results of the back analysis are presented in FOSbm and FOSbs forms (Fig. 12) to obtain the pairs of bm and bs satisfying limiting equilibrium condition. Considering that an adjustment factor (df ) of 1 corresponds to bm and bs values of 28 and 9, respectively, for undisturbed rock masses and, similarly, values of 14 (bm)and 6 (bs) correspond to a df value of 0.8 for highly disturbed rock mass and using the combinations of df bm and df bs which lead a factor of safety of unity, the plots given in Fig. 13 are established. It is now possible to estimate bm and bs values for closely jointed rock masses, depending on the disturbance eect, when the following expressions derived from the curve shown in Fig. 13 are used.  bm 3X14 ln df df 3401 df  28 12

 bs 0X67 ln

df df 3401 df

 9 13

The results of the back analysis of the slope instabilities in closely jointed rock masses and in the spoil piles with high proportion of disturbed rock pieces indicated that the disturbance eect due to the inuence of the method of excavation could not be ignored. In other words, the equations of the criterion based on the undisturbed rock mass condition did not work well if an adjustment factor was not considered. For this purpose, it is advised that, in the estimation of the rock mass constants, determination of the values of bm and bs of any particular rock mass determined from the curves (Fig. 13) or from Eqs. (12) and (13) for a given df value seems to be better. The rock mass constants then should be estimated by using Eqs. (8) and (9) proposed in this study. 4. Conclusions Due to the limitations in the RMR classication scheme, particularly for very poor quality rock masses, the geological strength index (GSI) has been introduced into the HoekBrown failure criterion. However, there are no published case histories on the back analysis of slopes or underground openings in heavily jointed rock masses which conrm that the current GSI methodology yields satisfactory results. In this study, an attempt has been made to provide a more quantitative numerical basis for evaluating the GSI and to suggest quantities which make more sense

758

H. Sonmez, R. Ulusay / International Journal of Rock Mechanics and Mining Sciences 36 (1999) 743760

Fig. 12. Variation in bm and bs values with factor of safety (FOS) for the case studies from the slope failures in closely jointed rock masses.

H. Sonmez, R. Ulusay / International Journal of Rock Mechanics and Mining Sciences 36 (1999) 743760

759

masses and even for rock spoil pile materials. However, the authors hope that the application of the suggested modications onto various failure case histories both from surface and underground excavations may lead to provide a better tool for more precise guidelines and to check validity of the equations employed by the non-linear failure criterion.

Acknowledgements
Fig. 13. Relationship between bm and bs and adjustment (disturbance) factor df derived from the back analysis of the investigated slope instabilities.

than that of the RMR System when used for the estimation of rock mass strength. For the purpose, two terms, `structure rating, SR', and `surface condition rating, SCR', have been introduced into the existing GSI classication scheme. In order to assign the ratings to these terms, the use of some easily measurable input parameters such as, roughness, weathering, inlling and volumetric joint count have been suggested. According to the selected rating intervals, the GSI chart has been modied to estimate more precise values of GSI. Five well documented slope instability examples have been given to illustrate the application of the proposed method in practical geotechnical engineering. The application of the suggested method and the approaches indicated that the use of GSI value determined from the suggested modied chart and consideration of disturbance eect conrmed the limit equilibrium condition for the failed slopes. The other issues concluded in the study were that the spoil pile materials consisting of angular and rounded rock pieces with low proportion of nes could be categorized as disintegrated rock masses in the GSI classication and it seemed possible to estimate their shear strength parameters from the modied Hoek Brown equations presented herein. Some engineering geologists feel that the visual descriptions upon which the GSI system is based are preferable to the numbers of Bieniawski's classication. On the other hand, some engineers may be unhappy with the largely descriptive nature of the GSI system and the comments oered by the authors. However, the attempt by the authors is to address the discussion of the GSI chart and its perceived deciencies to improve the GSI and to suggest a method for practitioners. A better understanding of the mechanics of jointed rock mass behaviour is a problem of major signicance in geotechnical engineering. The authors believe that the HoekBrown failure criterion provides a good estimate for the shear strength of closely jointed rock

The authors would like to thank Professor Dr. Hasan GERCEK from Zonguldak Karaelmas University, Turkey for his valuable comments on the revised manuscript.

References
[1] Hoek E, Brown ET. Underground excavations in rock. London: Inst. Min. Metall. Stephen Austin and Sons, 1980. [2] Bieniawski ZT. Engineering rock mass classications. John Wiley and Sons, 1989. [3] Hoek E. Strength of jointed rock masses, 1983 Rankine Lecture. Geotechnique 1983;33(3):187223. [4] Hoek E, Brown ET. The HoekBrown failure criterion: a 1988 update. In: Jurran JC, editor. Rock Engineering for Underground Excavations, Proc. 15th Canadian Rock Mech. Symp. University of Toronto, 1988. p. 318. [5] Hoek E, Wood D, Shah S. A modied HoekBrown criterion for jointed rock masses. In: Hudson JA, editor. Proc. Eurock '92. Thomas Telford, 1992. p. 20913. [6] Hoek E. Strength of rock and rock masses. ISRM News J 1994;2(2):416. [7] Hoek E, Kaiser PK, Bawden WF. Support of underground excavations in hard rock. Rotterdam: A.A. Balkema, 1995. [8] Hoek E, Brown ET. Practical estimates of rock mass strength. Int J Rock Mech Min Sci 1997;34(8):116586. [9] Hoek E, Marinos P, Benissi M. Applicability of the goelogical strength index (GSI) classication for very weak and sheared rock masses: the case of the Athens schist formation. Bull Eng Geol Environ 1998;57:15160. [10] Sonmez H, Ulusay R, Gokceoglu C. A practical procedure for back analysis of slope failures in closely jointed rock masses. IntJ Rock Mech Min Sci 1997;35(2):21933. [11] Hoek E. Reliability of the HoekBrown estimates of rock mass properties and their impact on design. Int J Rock Mech Min Sci 1998;35(1):638. [12] Laubscher DH. A geomechanics classication system for the rating of rock mass in mine design. J South Afr Inst Miner Metall 1990;90(10):25773. [13] Romana MA. Geomechanical classication for slopes: slope mass rating. In: Hudson JA, editor. Comprehensive rock engineering, Vol. 3. London: Pergamon Press, 1993. p. 57599 [ch. 22]. [14] Kendorski FS, Cumming RA, Bieniawski ZT, Skinner EH. Rock mass classication for block caving mine drift support. In: Proc. 5th Int. Cong. Rock Mech. ISRM., Melbourne, 1983. p. B51B63. [15] ISRM (International Society for Rock Mechanics). In: Brown ET, editor. ISRM suggested methods: rock characterization, testing and monitoring. London: Pergamon Press, 1981. [16] Terzaghi K. Rock defects and loads on tunnel supports. In:

760

H. Sonmez, R. Ulusay / International Journal of Rock Mechanics and Mining Sciences 36 (1999) 743760 Proctor RV, White TL, editors. Rock tunnelling with stell suppots, vol. 1. Youngstown, OH: Commercial Shearing and Stamping Company, 1946. p. 1799. Ulusay R. Geotechnical evaluations and deterministic design cosiderations for pitwall slopes at Eskihisar (Yatagan-Mugla) strip coal mine. Unpublished PhD dissertation, Middle East Technical University, Ankara, Turkey, 1991. Ulusay R, Yucel Z. An example for the stability of slopes excavated in weak rocks: Baskoyak Barite Open Pit. Earthsciences (Bull of Earth Sciences Application and Research Center of Hacettepe University) 1989;15(2):1527 [in Turkish]. Ulusay R, Ekmekci M, Gokceoglu C, Sonmez H, Tuncay E, Erdogan S. Slope stability investigation for Himmetoglu lignite open pit mine. Hacettepe University Report, Project No.: 970058, 1998. [In Turkish]. Janbu N. Slope stability computations. In: Hirscheld RC, Paulos SJ, editors. Embankment dam engineering: Cassagrande volume. New York: Wiley, 1973. p. 4787. Ulusay R, Arikan F, Yoleri MF, Caglan D. Engineering geological characterization of coal mine waste material and an evaluation in the context of back-analysis of spoil pile instabilities in a strip mine, SW Turkey. Eng Geol 1995;40:77101. Ulusay R, Yoleri MF, Caglan D, Arikan F. Design evaluations for spoil piles at a strip coal mine considering safety of the haul road. Int J Surf Min Recl Environ 1995;9:13340. Ulusay R, Caglan D, Arikan F, Yoleri MF. Characteristics of biplanar wedge spoil pile instabilities and methods to improve stability. Can Geotech J 1996;33(1):5879. Franklin JA, Mearz NH, Bennett CP. Rock mass characterization using photoanalysis. Int J Min Geol Eng 1988;6:97112. Singh A, Scoble M, Lizotte Y, Crowther G. Characterization of underground rock fragmentation. Geotech Geol Eng 1991;9:93 107. Goktan RM, Ayday CA, Zengin M. Case study on the comparison of measured and predicted mean fragment size from a large-scale blast. In: Fujii T, editor. Proc 8th Rock Mech Congress, vol. 1. A.A. Balkema, 1995. p. 779.

[17]

[22]

[18]

[23]

[24] [25]

[19]

[20] [21]

[26]

Potrebbero piacerti anche