Engineering Geology 79 (2005) 230 250 www.elsevier.

com/locate/enggeo

Probabilistic analysis of rock slope stability and random properties of discontinuity parameters, Interstate Highway 40, Western North Carolina, USA
Hyuck-Jin Parka,T, Terry R. Westb, Ik Wooa
a

Department of Geoinformation Engineering, Sejong University, Gunja-dong, Gwangjin-gu, Seoul, Korea b Department of Earth and Atmospheric Sciences, Purdue University, West Lafayette, IN 47907, USA Received 26 January 2004; received in revised form 30 December 2004; accepted 4 February 2005 Available online 7 April 2005

Abstract Probabilistic analysis has been used as an effective tool to evaluate uncertainty so prevalent in variables governing rock slope stability. In this study a probabilistic analysis procedure and related algorithms were developed by extending the Monte Carlo simulation. The approach was used to analyze rock slope stability for Interstate Highway 40 (I-40), North Carolina, USA. This probabilistic approach consists of two parts: analysis of available geotechnical data to obtain random properties of discontinuity parameters; and probabilistic analysis of slope stability based on parameters with random properties. Random geometric and strength parameters for discontinuities were derived from field measurements and analysis using the statistical inference method or obtained from experience and engineering judgment of parameters. Specifically, this study shows that a certain amount of experience and engineering judgment can be utilized to determine random properties of discontinuity parameters. Probabilistic stability analysis is accomplished using statistical parameters and probability density functions for each discontinuity parameter. Then, the two requisite conditions, kinematic and kinetic instability for evaluating rock slope stability, are determined and evaluated separately, and subsequently the two probabilities are combined to provide an overall stability measure. Following the probabilistic analysis to account for variation in parameters, results of the probabilistic analyses were compared to those of a deterministic analysis, illustrating deficiencies in the latter procedure. Two geometries for the cut slopes on I-40 were evaluated, the original 758 slope and the 508 slope which has developed over the past 40 years of weathering. D 2005 Elsevier B.V. All rights reserved.
Keywords: Probabilistic analysis; Kinematic analysis; Kinetic instability; Rock slope; Monte Carlo simulation

1. Introduction Uncertainty and variability are inevitable in engineering geology studies dealing with natural materials. This prevails because rocks and soils are

T Corresponding author. Fax: +82 2 462 7537. E-mail address: hjpark@sejong.ac.kr (H.-J. Park). 0013-7952/$ - see front matter D 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.enggeo.2005.02.001

H.-J. Park et al. / Engineering Geology 79 (2005) 230250

231

inherently heterogeneous, insufficient amount of information for site conditions are available and the understanding of failure mechanisms is incomplete. Therefore, many early efforts have been made to limit or quantify uncertainty of input data and analysis results (Casagrande, 1965; Peck, 1969; Einstein and Baecher, 1983; Whitman, 1984). Slope engineering is perhaps the geotechnical subject most dominated by uncertainty since slopes are composed of natural materials (El-Ramly et al., 2002). Uncertainty in rock slope engineering may occur as scattered values for discontinuity orientations and geometries such as discontinuity trace length and spacing, and in laboratory or in situ test results. Therefore, one of the greatest challenges for rock slope stability analysis is the selection of representative values from widely scattered discontinuity data. Application of probabilistic analysis has provided an objective tool to quantify and model variability and uncertainty. In particular, the probabilistic approach to rock slope stability makes it possible to consider uncertainty and variability in geotechnical parameters of rock masses. Although probabilistic analyses have been applied to rock slope stability, a limited number of examples applied to practical cases have been fully described. Lately commercially available limit equilibrium codes (such as SWEDGE, ROCKPLANE, SLIDE, SLOPE/W) often incorporate probabilistic tools, in which variations in discontinuity properties can be assessed. In this study, an application of the probabilistic method to practical problems in rock slope stability analysis is provided, and improved procedures for the evaluation of random properties of discontinuity parameters are explained. For this purpose, a rock cut in western North Carolina provides the example where the probabilistic approach is applied to analyze slope stability. In addition, random properties of discontinuity parameters, which were measured in the field, obtained through laboratory testing and applied in the probabilistic analysis, are discussed.

2. Probabilistic analysis Slope stability analysis requires the kinematic and kinetic evaluation. In the kinematic analysis the question is whether slope failure of a rock mass is

possible based on the geometry of discontinuities and slope orientation. Combinations of discontinuity orientations and the slope face are examined to determine if specific failure modes are possible. Analysis is commonly conducted with the aid of stereographic projections of the planar features. This indicates whether kinematic instability is likely and then kinetic stability is evaluated using forces acting on the rock mass. This procedure should be carried out for the probabilistic analysis as well as the deterministic analysis. For deterministic analysis, single fixed values (typically, mean values) of representative orientation and strength parameters are determined and then the kinematic and kinetic analyses are conducted using single representative values. Therefore, the stability analysis is normally carried out with one set of geotechnical parameters. Factor of safety, based on limit equilibrium analysis, is widely used to evaluate slope stability because of its simple calculation and results. However, most input values measured in the field or obtained by laboratory tests and used subsequently to calculate a safety factor show a wide scatter across a significant range rather than being a fixed single value. Thus, each parameter should be considered as a random variable and the analysis involving different values for each parameter will result in different factors of safety. Therefore, the factor of safety itself is a random variable, depending on many input variables. However, the deterministic analysis is unable to account for variation in rock mass properties and conditions. The probabilistic analysis was developed to consider the uncertainty in parameters and results. In the probabilistic approach, the analysis carries out the analysis of random properties of the discontinuities and rock mass. Random properties of input parameters are required for probabilistic analysis and are obtained by statistical evaluation of available geological and geotechnical data. Basic statistical parameters are the mean and coefficient of variation, and the probability density function (PDF) which are obtained during this step. Subsequently, using random properties of input parameters determined previously, probability of failure is evaluated. The Monte Carlo simulation, First Order Second Moment method (FOSM) and Point Estimate Method (PEM) are commonly used, but for the current research, the

232

H.-J. Park et al. / Engineering Geology 79 (2005) 230250

Monte Carlo simulation was used to calculate probability of failure.

3. Geology of the study area The study area consists of an extensive rock cut along Interstate Highway 40 (I-40) in western North Carolina, near the Tennessee border. This area along Interstate 40 shows excellent exposures of a series of metasedimentary rocks of Late Pre-Cambrian age (Fig. 1). The area is located in the western Blue Ridge province, one of several physiographic provinces which comprise the Appalachian Highlands. The Blue Ridge structural province includes on its western boundary the Great Smoky Mountains and associated thrust faults, and on the east, the Brevard fault zone (Wiener and Merschat, 1975). This province consists of high metamorphic grade, Middle Proterozoic basement to early Paleozoic, off-shelf cover sedimentaries and Paleozoic igneous intrusives. Major rock types in this area are a gray, thin bedded to laminated feldspathic meta-sandsone and a green slate with thin interbeds of fine meta-sandstone. Bedding is

distinct and the rock is highly jointed. The I-40 site has experienced several large landslides during and after construction. An investigation for relocation of the highway concluded that wedge failures were the most common phenomena. On July 1, 1997, a large rockslide occurred in this area after heavy rainfall, when two discontinuities forming an unstable wedge, failed. More than 100,000 m3 of rock were removed during mitigation of this rockslide. In the current study a large number of discontinuity orientations and geometries were measured in the field and their random properties evaluated by the authors.

4. Random properties of discontinuity parameters In the following section random properties of geological and geotechnical parameters are determined. Information obtained from sampled data is used to make generalizations about the populations from which the samples were collected (Ang and Tang, 1975). This is an important procedure needed to obtain accurate and proper stability analysis results. However, from one study to another, the selection of

Fig. 1. Geological map of the study area.

H.-J. Park et al. / Engineering Geology 79 (2005) 230250

233

random variables can be quite different. Some authors have considered only the geometric parameters of discontinuity and groundwater conditions to be random variables, whereas others also include strength parameters as random variables. In this study, the orientation, length, spacing, persistence and strength parameters of discontinuities are considered to be random variables and their random properties are found. 4.1. Discontinuity strength parameters No detailed shear strength testing was provided for the study area. The North Carolina Department of Transportation (NCDOT, 1980) used 308 as the internal friction angle for all discontinuities in the area to calculate the factor of safety in their slope stability investigation and analysis. According to Glass (1998), this apparently was obtained using a back analysis calculation based on observations made in the field. Even though determined from a simple calculation without shear strength testing, this value seems to be reasonable. This holds true because according to Barton (1973), approximate friction angle values for siltstone, the major lithology in this study area, lies in a range between 278 and 318. However, this value includes a high level of uncertainty since it was obtained from a simple back calculation without any shear strength tests and only one value suggested for shear strength parameters. A range of friction values is preferred in the stability analysis when significant uncertainty is involved. The probability density function of the shear strength parameters, especially friction angle, can be inferred from previous research, and from an analysis performed by the authors in a nearby study area in North Carolina (Park, 1999; Park and West, 2001). Several researchers have suggested a normal distribution or truncated normal distribution for the friction angle (Mostyn and Li, 1993; Hoek, 1998; Nilsen, 2000; Pathak and Nilsen, 2004). Based on experience by the authors, this can be confirmed using a chisquare goodness-of-fit test on direct shear strength testing data. Therefore, a truncated normal distribution is considered for the density distribution of the internal friction angle for the study area. Also, the authors assumed 308 as the mean value of the distribution.

For a probability analysis, another factor is required to represent the random parameter property to delineate the dispersion of a parameter; the standard deviation or coefficient of variation. As it develops, the coefficient of variation is a fairly stable measure of variability. The coefficient of variation varies for each physical property (or geotechnical parameter) for a geological material, even within the same layer, but the coefficient of variation for the same physical properties of geological materials in many parts of the world has a value within a relatively narrow range (Rethai, 1998). Thus, we can assume reliable data for the expected standard deviation of a given physical characteristic of a geological material even before performing the laboratory tests (Harr, 1987). In research by the authors (Park, 1999; Park and West, 2001), internal friction angle data indicated a coefficient of variation of approximately a 10%. This value corresponds well to a representative coefficient of variation for parameters commonly used in civil engineering design by Harr (1987) and also agrees with the coefficients of variations of the friction angle for soil, according to Schultze (1975). Therefore the authors used a value of 10%, to calculate the standard deviation for the friction angle in this area. 4.2. Discontinuity orientation Discontinuity orientation is an important parameter affecting rock slope stability because failure type and kinematic instability are influenced mainly by this feature. The principal need is to identify the sets of preferred orientations. Orientation of these sets, and the degree of clustering within each set, has a major influence on the engineering characteristics of the rock mass. In this research, the clustering procedure, proposed by Mahtab and Yegulalp (1982), was adopted. The algorithm is based on the assumption that a discontinuity set has a significantly greater degree of clustering than would a totally random distribution of orientations. Orientation data for the study area were collected using the scanline method. Subsequently, in order to reduce sampling bias from the scanline sampling, weighting factors were applied to the orientation data. Fig. 2 shows the results of clustering in this area after applying a weighting factor. A total of 6 discontinuity sets were identified and their represen-

234

H.-J. Park et al. / Engineering Geology 79 (2005) 230250

Fig. 2. Results of clustering process of discontinuity normals on equal angle lower hemisphere projection in the study area.

tative orientations were 217/77 for J1, 183/05 for J2, 163/63 for J3, 196/56 for J4, 227/37 for J5 and 061/ 66 for J6. After performing the clustering procedure, the appropriate probability density function was determined for a discontinuity orientation distribution. Owing to its simplicity and flexibility, the Fisher distribution was selected. This distribution is based on the assumption that a population of orientation values is distributed about a btrueQ value (Fisher, 1953). This assumption is similar to the concept of discontinuity normals being distributed about some true value within a set. In view of its simplicity and flexibility, the Fisher distribution provides a valuable model to evaluate discontinuity orientation data (Priest, 1993). However, the distribution provides only an approximation for asymmetric data because it is a symmetrical distribution. Therefore, some different models have been proposed to provide better fits for asymmetric orientation data. However, these models are too complex in their parameter estimation. Furthermore, because of their complexity,

generation of random values from those asymmetric orientation distributions is difficult to accomplish and subsequently the analyses based on that probabilistic approach are difficult to perform. Hence, the Fisher distribution is commonly adopted for many probabilistic calculations and that was the case in the current study. 4.3. Discontinuity trace length Knowledge of discontinuity lengths for a rock mass is important for predicting rock behavior and analysis of rock slopes because discontinuity lengths influence the size of blocks that may be formed. Mean discontinuity length and length distribution provide important data for each joint set which are required for a probabilistic model of rock slope analysis in a jointed rock mass. However, estimation of the mean trace length is difficult because of bias errors involved in trace length measurements. Bias problems due to the scanline sampling procedure have been discussed by several authors (Baecher and

H.-J. Park et al. / Engineering Geology 79 (2005) 230250

235

Lanney, 1978; Cruden, 1977; Priest and Hudson, 1981) and several researchers have attempted to provide a procedure to provide an unbiased estimation for the mean length of joint sets. According to Mauldon et al. (2001), the approaches to circumvent or correct for sampling biases and estimate correct mean trace length of discontinuity are: (1) approach that assumes a particular form for the trace length distribution of the sampled population and (2) methods that are distribution free (Martiz, 1981). The latter approach is based on the commonly used circular scanline or window mapping and the mean trace length is obtained without consideration of the underlying trace length distribution. Therefore, it is difficult to apply the results of the second approach to the simulation procedure in this study regardless of many advantages for that approach. Regarding the first approach, Priest and Hudson (1981) showed that the corrected probability density distribution of trace lengths would have a negative exponential distribution. Moreover, they noted that if the actual trace lengths have a negative exponential distribution, the distribution and the mean of semi-trace lengths are the same as the actual trace lengths. Therefore, by measuring the semi-trace lengths of discontinuities in the field, the distribution and mean value of the actual trace lengths can be obtained. Consequently, by selecting a negative exponential distribution as the actual trace length distribution, adoption of semi-trace length measurements as a basis for estimating mean trace lengths can be supported. In addition, many studies of field measurements have showed that the negative exponential probability density distribution is suitable to represent the discontinuity trace length distribution (Wallis and King, 1980; Baecher, 1983; Kulatilake et al., 1993, 2003). Discontinuity trace length data were obtained by the authors from a field survey on road cuts along Interstate Highway 40. Approximately 300 data points were sampled with the authors collecting the semitrace length of discontinuities using the detailed scanline method. After obtaining this discontinuity length data, lengths were reclassified on the basis of discontinuity sets and the mean length for each set was evaluated. Fig. 3 shows the histograms of discontinuity trace length obtained for joint sets 1 and 2.

40

(a) Joint set 1

Occurence frequency (%)

35 30 25 20 15 10 5 0 0.4 0.8 1.2 1.6 2.0 2.4 2.8 3.2 3.6 4.0

Discontinuity trace length (m)

(b) Joint set 2

50

Occurence frequency (%)

40 30 20 10 0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

Discontinuity trace length (m)

Fig. 3. Histogram of the occurrence frequency in discontinuity trace length.

4.4. Discontinuity persistence 4.4.1. Traditional definition of joint persistence Discontinuity lengths determine the size of the rock blocks that form within a rock mass. Furthermore, they may also affect joint persistence, which is defined as the areal extent or size of a discontinuity along a plane (ISRM, 1978). Persistence is one of the most significant joint parameters affecting rock mass strength, but it is difficult to quantify. With reference to a joint plane (a plane through the rock mass containing a combination of discontinuities and intact rock regions), joint persistence is defined as the fraction of area that is actually discontinuous (Einstein et al., 1983). Therefore, the persistence value (K ) can be expressed in the limit form: P aDi K lim i 1 Ao Yl AD

236

H.-J. Park et al. / Engineering Geology 79 (2005) 230250

in which D is a region of the plane with area A D and a Di is the area of the i th joint in D . Einstein et al. (1983) suggested that persistence can only be roughly quantified by observing the discontinuity trace length on a rock exposure surface. This is because rock exposures are small and only two-dimensional. It is impossible in practice to measure the discontinuity area accurately in a field survey. Consequently, joint persistence can be expressed as a limit length ratio along a given line on a joint plane in terms of trace length: P K lim
i LS Y l

lSi

LS

where L S is the length of the straight-line segment S , and l Si is the length of the i th joint segment in S . 4.4.2. Importance of discontinuity persistence The reason that discontinuity persistence is important in slope stability analysis is because of its major effect on rock mass strength. The shear strength available for a rock bridge is one to two orders of magnitude greater than the shear strength available on the discontinuity. As Einstein et al. (1983) and West (1996) suggested, joint persistence can be used to estimate the strength of a rock mass against sliding along a given plane. That is, if the joint is not persistent, failure occurs through the rock bridge. This causes a significant increase in shear strength. Singh and Sun (1989) and Scavia (1990) applied a fracture mechanics concept to evaluate the stability of rock slopes which do not have a 100% persistence failure plane. Kemeny (2003) proposed a fracture mechanics model in slope stability, which is considered the time dependent degradation of rock bridge cohesion. Fracture mechanics considers rock slope failure to be a result of joint initiation and propagation. Therefore, the joint tip stress intensity factor is the governing parameter with respect to rock slope stability and the factor of safety is defined in terms of stress intensity factor (Whittaker et al., 1992). However, this approach has the limitation that the factor of safety defined as the stress intensity factor indicates crack stability, but not the overall stability of the

slope. In addition, those approaches consider the persistence as a fixed value, so it is not possible to consider the persistence as a random variable in the probabilistic analysis. The serious problem concerning persistence is that its extent is difficult to measure because direct mapping of discontinuities within a rock mass is not possible. In practice, 100% persistence is assumed. However, the possibility of a 100% persistent discontinuity on the shear planes is quite low under field conditions. In addition, as Einstein et al. (1983) suggested, because every joint in a set does not have the same value and these values are uncertain, persistence should be considered as a random variable. Therefore, a new approach is requisite for the probabilistic analysis, and random properties must be evaluated to characterize joint persistence. 4.4.3. A new concept of persistence proposed in this study Hudson and Priest (1983) recognized that two kinds of persistence could be identified: intermittent joints as in Fig. 4(a) and impersistent discontinuities as in Fig. 4(b). Intermittent joints in Fig. 4(a) require that the planes contain a patchwork of discontinuities and intact rock regions through the rock mass. As discussed previously, the previous concept of persistence implies that two or more joints occur on the same plane, so the previous persistence concept is based on the concept of intermittent joints. However, from a practical point of view, it appears that intermittent joints in Fig. 4(a) are

(a)

(b)

(c)

Fig. 4. Traces of (a) intermittent, (b) impersistent and (c) persistent joints (after Hudson and Priest, 1983).

H.-J. Park et al. / Engineering Geology 79 (2005) 230250

237

geologically unlikely (Mauldon, 1994). That is, Mauldon (1994) concluded that intermittent joints would seem to imply existence of weakness planes through the rock mass, but locally separated to form visible joints. Consequently, the intermittent discontinuity should be treated as persistent, i.e. continuous across the region of interest, at least for the purpose of mechanical analysis. To this end, a new approach is proposed in this research which accounts for persistence in spite of the limitations of measurement, based on the impersistent joint concept in Fig. 4(b). Joint persistence is described in this study as a function of the length of individual joints and the maximum sliding dimension, determined by slope geometry, joint orientation and joint dimension (Fig. 5). This approach assumes that only one joint forms the sliding surface (multiple joints do not line up end to end) and that this joint is not offset from the sliding surface, which is also proposed by Mauldon (1994). In our field survey for this study, a discontinuity is considered to occur within the same plane. To utilize this new approach, the probability that the joint length is long enough to form a block capable of sliding is evaluated. That is, using information on statistical parameters and the probability distribution of discontinuity length considered previously, a large number of individual joint length values are generated. Then each value of the generated joint length is compared to the sliding dimension and the probability that the joint length is equal to or greater than the maximum sliding length is calculated (Fig. 5). This is the probability that a fully persistent discontinuity exists. Then this value is multiplied by the probability of slope

failure with the premise that joints are fully persistent. Prock slope failure Prock slope failure j fully persistent joint Pfully persistent joint exists 3

The assumption of the fully persistent joint in rock slope stability is common in deterministic analysis as well as the probabilistic analysis. This is quite conservative approach in stability analysis. However, this approach overcomes the limitation of a conservative analysis. 4.5. Discontinuity spacing In rock slope stability analysis, spacing of discontinuity sets is part of the representation of geometric characteristics for each discontinuity set. Measurement of joint set spacing causes sampling bias since scanlines are not positioned perpendicular to discontinuities because of the limited rock face exposure (Terzaghi, 1965). Therefore, the correction of the sampling bias caused by a inclined scanline orientation is accomplished using the acute angle between the scanline orientation and the orientation of the line normal to the mean orientation for the joint set in question (Giani, 1992). Statistical parameters, i.e. mean and standard deviation of spacing for each set, are evaluated using the corrected data, and then the probability density function for discontinuity spacing was derived from these data. Although mean discontinuity spacing provides a direct measure of spacing data, several previous studies have tried to represent the distribution of measured spacing data by statistical analysis and description with the spacing data considered as a random variable. In order to determine the appropriate spacing distribution, the authors collected approximately 300 values of discontinuity data using a borehole sampling method (approximately 60% of 300 data) and the detailed scanline method (40% of 300 data) at the Interstate 26 site in northwestern North Carolina near the Tennessee border. This is because considerably more data are provided for this adjacent area than are available for the I-40 site (Park, 1999). On the basis of

Length of joint on sliding plane

Maximum sliding dimension

Fig. 5. Geometrical feature of sliding dimension and joint on sliding plane.

238

H.-J. Park et al. / Engineering Geology 79 (2005) 230250

data collected for I-26, chi-Square goodness-of-fit tests were performed for lognormal and negative exponential distributions, which are the two distribution models commonly used for spacing evaluation. This is because those theoretical distributions are bounded at zero and are skewed to the right and those characteristics are similar to the properties of the spacing distribution. Table 1 shows the results of chisquare tests for joints in the Interstate 26, Area A. Results show that both the lognormal distribution and the exponential distribution appear to be valid models for spacing at the 5% significance level. However, P because the calculated (n i e i )2 / e i value for lognormal distribution is smaller than that for the exponential distribution, the lognormal distribution is the better of the two. Table 2 for data from Interstate 26, Area B shows similar results. Therefore, the lognormal probability density distribution was used as the distribution model to represent the random property of discontinuity spacing. The literature also proposes the use of a lognormal probability distribu-

tion for discontinuity spacing. Rouleau and Gale (1985), Sen and Kazi (1984) and Kulatilake et al. (2003) suggest that the lognormal probability density distribution was appropriate, based on their goodnessof-fit tests.

5. Probabilistic analysis of rock slope stability 5.1. Analysis procedure After the random properties of discontinuity parameters are defined, the probabilistic analysis is accomplished. The Monte Carlo simulation was used for the probabilistic analysis in this study. The Monte Carlo technique is frequently applied to evaluate the probability of failure of a mechanical system, in particular, when the direct integration is not practical or when the integration equation is difficult to solve (Mostyn and Li, 1993). The simulation procedure proceeds in two steps, the first being kinematic

Table 1 Chi-square test results for relative goodness-of-fit in spacing data in Interstate 26, Area A Interval Observed frequency (n i ) 0 18 5 7 11 4 1 1 1 0 0 0 2 1 0 1 0 1 0 0 0 0 0 Theoretical frequency (e i ) Exponential 00.15 0.151.00 1.001.85 1.852.70 2.703.55 3.554.40 4.405.25 5.256.10 6.106.95 6.957.80 7.808.65 8.659.50 9.5010.35 10.3511.20 11.2012.05 12.0512.90 12.9013.75 13.7514.60 14.6015.45 15.4516.30 16.3017.15 17.1518.00 N 18.00 2.5855 12.4024 9.2896 6.9580 4.7233 3.9036 2.9238 2.1900 1.6403 1.2286 0.9203 0.6893 0.5163 0.3867 0.2896 0.2169 0.1625 0.1217 0.0912 0.0683 0.0511 0.0383 0.0287 51.4260 Lognormal 0.5451 15.1551 11.5748 7.1535 4.6025 3.1028 2.1756 1.5754 1.1713 0.8904 0.6897 0.5429 0.4334 0.3502 0.2861 0.2360 0.1964 0.1647 0.1391 0.1183 0.1012 0.0870 0.0752 51.3667 (n i e i )2 / e i Exponential 2.5855 2.5263 1.9808 0.0003 8.3411 0.0024 1.2658 0.6466 0.2500 1.2286 0.9203 0.6893 4.2640 0.9727 0.2896 2.8264 0.1625 0.1217 0.0912 0.0683 0.0511 0.0383 0.0287 29.3515 Lognormal 0.5451 0.5341 3.7346 0.0033 8.8924 0.2594 0.6353 0.2101 0.0251 0.8904 0.6897 0.5429 5.6633 1.2056 0.2861 2.4732 0.1964 0.1647 0.1391 0.1183 0.1012 0.0870 0.0752 27.4725

H.-J. Park et al. / Engineering Geology 79 (2005) 230250 Table 2 Chi-square test results for relative goodness-of-fit in spacing data in Interstate 26, Area B Interval Observed frequency (n i ) 16 11 7 4 3 3 0 0 0 0 1 0 0 0 0 0 0 0 0 Theoretical frequency (e i ) Exponential 0.01.0 1.02.0 2.03.0 3.04.0 4.05.0 5.06.0 6.07.0 7.08.0 8.09.0 9.010.0 10.011.0 11.012.0 12.013.0 13.014.0 14.015.0 15.016.0 16.017.0 17.018.0 N 18.0 17.5246 10.2124 5.9513 3.4681 2.0210 1.1777 0.6863 0.4000 0.2331 0.1358 0.0792 0.0461 0.0269 0.0157 0.0091 0.0053 0.0031 0.0018 0.0011 41.9985 Lognormal 17.9265 10.9442 5.1976 2.7770 1.6230 1.0133 0.6653 0.4545 0.3207 0.2323 0.1721 0.1300 0.0998 0.0778 0.0614 0.0490 0.0395 0.0322 0.0264 41.8426 (n i e i )2 / e i Exponential 0.1326 0.0671 0.1848 0.0816 0.4742 2.8195 0.6863 0.4000 0.2331 0.1358 10.7132 0.0461 0.0269 0.0157 0.0091 0.0053 0.0031 0.0018 0.0011 16.0309

239

Lognormal 0.2070 0.0003 0.6250 0.5386 1.1683 3.9850 0.6653 0.4545 0.3207 0.2323 3.9823 0.1300 0.0998 0.0778 0.0614 0.0490 0.0395 0.0322 0.0264 12.6056

analysis, examining kinematic instability of a rock body defined by discontinuities. Based on discontinuity orientation, it is determined whether the rock body is able to move or not. If the kinematic analysis indicates that the geometric condition is potentially unstable, then the kinetic stability is assessed by the limit equilibrium method. This comprises the second step. 5.2. Evaluation for probability of slope failure To check the stability of rock slope systems, both kinematic and kinetic analyses are required to analyze the geometry and strength of discontinuities. In a complete study this should be accomplished for both probabilistic analysis as well as for deterministic analysis. However, only kinetic instability was evaluated and it was assume to be the probability of failure for the rock slope in some previous studies. Difficulty in performing kinematic analysis is considered as one reason why it is omitted. For planar failure, the kinematic analysis is relatively easy since clear criteria exist such that the dip direction of discontinuity must be within 208 of dip

direction of the slope face. However, the kinematic analysis is commonly accomplished using stereographic projections, and subsequently a calculation of probability of the kinematic instability is not easy accomplished, especially for wedge failure analysis. The stereographic projection method is not suited to conducting computational and repeated calculations used in the Monte Carlo simulation which is the typical procedure for probability analysis of wedge failure. This is because the closed form of kinematic analysis is not provided. That is, if the Monte Carlo simulation is utilized, in order to obtain the input values for each simulation, the great number of stereographic projections is required. The large number of stereographic projection is needed for each set of parameter combinations and the input values for each simulation must be measured from the stereoplot if the Monte Carlo simulation is utilized. Some years ago McMahon (1971) and Herget (1978) proposed a probabilistic kinematic analysis approach which can evaluate the probability of kinematic instability using stereographic projection. However, the procedure did not provide a closed form equation and was limited only to the

240

H.-J. Park et al. / Engineering Geology 79 (2005) 230250

planar condition. Therefore, in this study a simple equation for checking a kinematic instability for rock wedges was used (Park and West, 2001; Gu nther, 2003). X b e b aapparent aapparent tan1 tana cosbi bs 4 5

only when the block is kinematically unstable, the probability of kinetic instability is defined as Pkn=km Nf Nm 9

where N f is the number of iterations that a wedge has factor of safety less than one. Therefore, the probability of failure is Pf Pkm Pkn=km Nm Nf NT Nm 10

where e is the dip angle of the line of intersection between two discontinuities, X is the dip angles of the upper ground surface and a apparent is the apparent dip of the slope face in the dip direction of the intersection line, not in the dip direction of slope face (Park and West, 2001; Gu nther, 2003). b i and b s are the dip directions of the lines of intersection and slope face, respectively. In the current study, the probability of kinematic instability was evaluated using the apparent dip of slope face. Also, in the current study, a step-by-step procedure for evaluating the slope failure probability was used. Procedural steps were accomplished separately for kinematic stability and kinetic stability in the probabilistic analysis. That is, this procedure assesses the probability of kinematic instability, in which a number of iterations form a block or wedge that can kinematically move. Once the kinematically unstable blocks or wedges have been identified and evaluated as kinematically unstable, the kinetic probability is evaluated as a conditional probability which has a premise that the block is kinematically unstable. Therefore, the overall probability of slope failure will be Pf Pkinematically unstable Pkinetic unstablejkinematic unstable 6

6. Results of analysis 6.1. Input parameters In the probabilistic analysis, input parameters can be subdivided into two groups by their randomness: deterministic and probabilistic parameters. Deterministic parameters are those considered to be known and having a single value for all sliding blocks. In the current study, the orientation and height of the cut slope and rock density were considered to be deterministic parameters. In addition, roughness of a discontinuity was considered as a deterministic parameter. Roughness is a potentially important component of shear strength and therefore, it was measured in the field for each discontinuity using a disk clinometer. This value was added to the friction angle of the discontinuity. For the probabilistic parameters, the probability density function and the values of statistical parameters for random variables are chosen on the basis of physical properties, test results and evaluation of the measured data. In this study, joint parameters were considered to be probabilistic in nature. In all, joint orientation, geometric parameters, such as length and spacing, and shear strength parameters were considered to be probabilistic parameters. In addition, pore water pressure in the discontinuity is considered to be a random variable since the groundwater table level varies. However, those probabilistic parameters are assumed to be independent. The covariance between random parameters plays an important role in probabilistic analysis. However, research results involving the accurate evaluation of covariance between random parameters in a rock mass are limited and some

This concept was also proposed by Einstein (1996). Therefore, the probability of slope failure is Pf Pkm Pkn=km 7

The probability of kinematic instability is defined as: Pkm Nm NT 8

where N m is the number of iterations which is kinematically unstable and N T is total number of iterations. Because the kinetic analysis is performed

H.-J. Park et al. / Engineering Geology 79 (2005) 230250

241

researchers propose that the assumption of independence is conservative (Cherubini et al., 1983; Li and White, 1987). Parameters used in this study are listed in Table 3. Input data for slope geometry are included in Table 3(a) and the input for discontinuity properties is given in Table 3(b). A point to note here is the length of joint set 5. All other discontinuity lengths for discontinuity sets were determined in the field, except for J5 which was identified as a bedding plane. In many cases, bedding plane lengths are assumed to be infinite, but in this study, in order to show that the length of bedding was much greater than that of other joint sets, it was assigned a value of 60 m. 6.2. Results for planar failure Table 4 shows the results of the deterministic and probabilistic analyses of planar failure for each joint set. In the deterministic analysis, mean values of each random variable are used and the factor of safety is calculated for each set. In contrast, random properties of random variable are considered and the probability of failure is evaluated using the simulation procedure. In order to compare the results between the deterministic and probabilistic analysis, the same performance function suggested by Hoek and Bray (1981) was utilized. In Fig. 6, the histogram of safety factor is the result of calculation using the repeated simulation procedure. Based on results of the deterministic analysis, joint sets 4 and 5 were found to be unstable having a factor of safety of 0.29 and 0.69, respectively. That is, kinematic analyses for J4 and J5
Table 3 Input values for I-40 area (a) Input for slope geometry Orientation of slope (dip direction/dip) 210/75 (b) Input for discontinuity properties Set I.D Mean orientation (dip direction/dip) Fisher constant Mean friction angle (deg) S.D of friction angle Mean length (m) Roughness (deg) Mean spacing (m)

Table 4 Comparison of results for the deterministic analysis and the probabilistic analysis at cut slope angle of 758 Joint Factor Probability of failure set I.D. of safety Kinematic Kinetic Total J1 J2 J3 J4 J5 J6 Stable Stable Stable 0.29 0.69 Stable 0.345 0 0 0.621 0.759 0 0.018 0 0 0.014 0.671 0 Average volume of possible block (m3)

0.006 5.3 0 0 0.0087 18.9 0.509 42.6 0

indicate they are kinematically unstable and subsequently, factors of safety for those joint sets are computed and found to be less than 1. In Table 4, the word dstableT indicates that the kinematic analysis found the set to be kinematically stable and subsequently, the kinetic analysis was not conducted. Therefore, all joint sets except J4 and J5 are kinematically stable. However, the probabilistic analysis for the planar failure shows different results. According to the probability of planar failure in Table 4, the analysis indicates that joint sets J1, J4 and J5 have a possibility of failure. Especially for joint set 1, the probability of kinematic instability calculated is 34.5% and the probability of kinetic instability is 1.8% despite the fact that J1 was found to be stable in the deterministic analysis. For the deterministic analysis, the mean orientation of J1 does not show a possibility of kinematic instability because the mean dip angle for J1 is greater than the slope cut angle and the join will not daylight. However, when the variation in orienta-

Height of slope (m) 34

Unit weight of rocks (t/m3) 2.56

J1 217/77 42 30 3 1 1 1.6

J2 183/5 53 30 3 1 1 3.5

J3 163/63 29 30 3 1 1 0.43

J4 196/56 119 30 3 0.5 1 4.5

J5 227/37 36 30 3 60 0 1.0

J6 061/66 106 30 3 1 1 2.7

PDF Fisher Normal Exponential Lognormal

242

H.-J. Park et al. / Engineering Geology 79 (2005) 230250

(a) Set 1 at 75 degree slope angle

0.40 0.35 0.30 0.40 0.35 0.30

(b) Set 4 at 75 degree slope angle

Frequency

Frequency
0.5 1.0 1.5 2.0 2.5 3.0

0.25 0.20 0.15 0.10 0.05 0.00

0.25 0.20 0.15 0.10 0.05 0.00 0.5 1.0 1.5 2.0 2.5 3.0

Factor of safety

Factor of safety

(c) Set 5 at 75 degree slope angle

0.25 0.20 0.25 0.20

(d) Set 4 at 50 degree slope angle

Frequency

0.15 0.10 0.05 0.00

Frequency
0.5 1.0 1.5 2.0 2.5 3.0

0.15 0.10 0.05 0.00

0.5

1.0

1.5

2.0

2.5

3.0

Factor of safety
0.25 0.20

Factor of safety

(e) Set 5 at 50 degree slope angle

Frequency

0.15 0.10 0.05 0.00

0.5

1.0

1.5

2.0

2.5

3.0

Factor of safety
Fig. 6. Histogram of factor of safety calculated in probabilistic analysis for planar failure.

tion is considered, many of scattered orientations have the possibility of kinematic instability even though the mean orientation does not yield a kinematically unstable condition. Consequently, the deterministic analysis based on a fixed representative orientation of discontinuities fails to show the possibility of kinematic instability.

The probability of kinetic instability for J4 of 1.4% is much lower than the probability of kinetic instability for J5, 67.1%. In addition, the probability of slope failure for J4, 0.9% is lower than that for J5, 50.9%. Therefore, based on probabilistic analysis, J5 represents a greater risk and more a dangerous condition. Discontinuity trace length is a possible

H.-J. Park et al. / Engineering Geology 79 (2005) 230250

243

reason. In this study, discontinuity length was considered to be a form of persistence. This is used when the probability of the kinetic instability is evaluated. Therefore, J5, a bedding plane with a mean length of 60 m, has a higher kinetic instability probability than J1 whose mean length is only 0.5 m. Regarding a specific aspect of probabilistic analysis, the mean volume of blocks for each case of possible failure, is evaluated using discontinuity orientation data and cut slope geometry. That is, if each block whose dimension is calculated by randomly selected discontinuity parameter is kinematically and kinetically unstable, the volumes of each block are calculated using the Hoek and Bray (1981) equation. Then the mean volume of possible blocks is evaluated. For this, the daylight point (where possible failure surface meets the slope face) is randomly selected from slope face. Mean volumes of possible blocks are 5.3 m3 for J1, 18.9 m3 for J4 and 42.6 m3 for J5. Therefore, joint set 5, which is a bedding plane, exhibits a high possibility of plane failure and the size of the failure block will also be great. In addition, J4, a plane indicated as a failure possibility by NCDOT (1980), shows in the current study, a maximum 19.6% failure probability and 18.9 m3 volumes for block size. However, since the cut slopes were excavated more than 40 years ago and several large and many small slides occurred in this area, the slope cut has become much flatter. Based on observation of the authors the cut slope angle is now approximately 508 rather than the original 758. Using this slope angle, the factor of safety and probability of failure were evaluated again and the results compared to the previous calculations. Table 5 includes results of both deterministic and probabilistic analyses for the 508 slope angle. A
Table 5 Comparison of results for the deterministic analysis and the probabilistic analysis at cut slope angle of 508 Joint Factor Probability of failure set I.D. of safety Kinematic Kinetic Total J1 J2 J3 J4 J5 J6 Stable Stable Stable Stable 0.61 Stable 0 0 0 0.053 0.663 0 0 0 0 0.014 0.669 0 Average volume of possible block (m3)

deterministic analysis of the current slope shows joint set 4 is now stable following slope flattening. Representative orientation data for joint set 4 are no longer kinematically unstable because the dip of the discontinuity is greater than the 508 slope dip angle. Therefore, without further investigation of the shear strength, joint 4 is designated to be stable because the joint will not daylight kinematically. For joint 5, however, the safety factor is not much different from the previous value, 0.69. Hence, reducing the slope angle does not significantly change the factor of safety for joint set 5. Considering the results of probabilistic analysis, joint set 1 is stable for a 508 cut slope. That is, both probabilities of kinematic and kinetic instability are zero. In addition, the probabilities of kinematic instability for J4 and J5 are reduced from 62.1% and 75.9% to 5.3% and 66.3%, respectively, but the probabilities of kinetic instability in both joint sets are unchanged. However, the total probability of slope failure is reduced somewhat because of the multiplication effect. As the slope angle is reduced, volumes of possible rock blocks are also reduced. For joint set 4, the mean volume is reduced from 18.9 m3 to 3.2 m3 and the mean volume for joint set 5 is reduced from 42.6 m3 to 19.1 m3. Therefore, the risk of a large slope failure is significantly reduced. However, in practice, the planar failure on joint set 5 will occur only when the lateral extent of potential failure mass is isolated by a lateral release surface, which is a prerequisite for planar failure to occur (Hoek and Bray, 1981). 6.3. Results for wedge failure Relative to wedge failure, for only two joint set combinations (J2 and J3 and J2 and J6), do the results of the deterministic analysis for kinematic stability agree with those of the probabilistic analysis (Table 6; Fig. 7). They are kinematically stable based on the deterministic analysis and have zero probability of kinematically instability in the probabilistic analysis. However, some joint combinations indicated as kinematically stable by the deterministic analysis are shown to be kinematic unstable using probabilistic analysis. Those combinations (J1 and J2, J1 and J5, J2 and J4, J2 and J5, J3 and J6, J5 and J6) have small probabilities of kinematic instability, ranging from

0 0 0 0.0007 3.2 0.444 19.1 0

244

H.-J. Park et al. / Engineering Geology 79 (2005) 230250

(a) Combination of joint sets 1 and 2

0.05 0.04 0.30 0.25

(b) Combination of joint sets 1 and 3

Frequency

0.03 0.02 0.01 0.00

Frequency
1 2 3 4 5

0.20 0.15 0.10 0.05 0.00 1 2 3 4 5

Factor of safety

Factor of safety

(c) Combination of joint sets 1 and 4

0.05 0.04 0.12 0.10

(d) Combination of joint sets 1 and 5

Frequency

0.03 0.02 0.01 0.00

Frequency
1 2 3 4 5

0.08 0.06 0.04 0.02 0.00 1 2 3 4 5

Factor of safety

Factor of safety

(e) Combination of joint sets 1 and 6

0.07 0.07 0.06 0.06 0.05 0.04 0.03 0.02 0.01 0.00 1 2 3 4 5

(f) Combination of joint sets 2 and 4

Frequency

Frequency

0.05 0.04 0.03 0.02 0.01 0.00 1 2 3 4 5

Factor of safety

Factor of safety

(g) Combination of joint sets 2 and 5

0.05 0.04 0.20 0.18 0.16 0.14 0.12 0.10 0.08 0.06 0.04 0.02 0.00 1 2 3 4 5 0.00

(h) Combination of joint sets 3 and 4

Frequency

0.03 0.02 0.01

Frequency

Factor of safety

Factor of safety

Fig. 7. Histogram of factor of safety calculated in probabilistic analysis for wedge failure at 758 slope angle.

H.-J. Park et al. / Engineering Geology 79 (2005) 230250

245

(i) Combination of joint sets 3 and 5

0.12 0.10 0.16 0.14 0.12

(j) Combination of joint sets 3 and 6

Frequency

Frequency
1 2 3 4 5

0.08 0.06 0.04 0.02 0.00

0.10 0.08 0.06 0.04 0.02 0.00 1 2 3 4 5

Factor of safety

Factor of safety

(k) Combination of joint sets 4 and 5

0.10 0.08 0.18 0.16 0.14 0.12 0.10 0.08 0.06 0.04 0.02 0.00 1 2 3 4 5 0.00

(l) Combination of joint sets 4 and 6

Frequency

0.06 0.04 0.02

Frequency

Factor of safety

Factor of safety

(m) Combination of joint sets 5 and 6

0.08 0.07 0.06

Frequency

0.05 0.04 0.03 0.02 0.01 0.00 1 2 3 4 5

Factor of safety
Fig. 7 (continued ).

0.1% to 3.7%. By contrast, the joint set combinations shown to be kinematically unstable in the deterministic analysis show high probabilities of kinematic instability in the probabilistic analysis. That is, unstable joint combinations (J1 and J3, J1 and J4, J1 and J6, J3 and J4, J3 and J5, J4 and J5, J4 and J6) show high failure probabilities ranging from 18.9% to 93.5%.

For the kinetic analysis, a sliding mode for wedge failure is determined and a factor of safety value is calculated using a deterministic analysis. However, in the probabilistic approach, four different kinetic probabilities can be evaluated for four different sliding modes because the scattered orientations of discontinuities can produce these different sliding modes for combinations of the two discontinuities. For example,

246

H.-J. Park et al. / Engineering Geology 79 (2005) 230250

for the J3 and J5 combination in Table 6, the deterministic analysis shows a factor of safety of 0.36 for the wedge sliding with contact on plane 2. However, based on the probabilistic analysis, there is 56.1% possibility of sliding with contact on plane 2. Also, there is 19.1% possibility of sliding without contact (that is, the contact is lost on both planes since water pressures on both planes is greater than the normal force) on both planes. In fact, the combinations of J2 and J5, J4 and J5 and J5 and J6 which are stable in the deterministic analysis have three or four different sliding modes with high probabilities of failure. In Table 6, refer to the sliding mode where the factor of safety for each joint set combination is evaluated using a deterministic analysis. The sliding mode with the highest probability of kinetic instability among those four different sliding modes shows the lowest factor of safety. For example, in the J3 and J5 combination, the deterministic analysis has a factor of safety of 0.36 which is a sliding mode in contact on plane 2, and this sliding mode has the highest probability of kinetic instability ( P f = 56.1%) of all the other sliding modes. Comparing these probabilities with the average volume of the failure block, the combinations of J1 and J3, J1 and J5, J3 and J4, J4 and J5 and J5 and J6 indicate the possibility of small volume blocks, that is, 0.011 m3, 0.095 m3, 6.86 m3, 1.72 m3 and 0.041 m3, respectively. However, for the J3 and J5 combination,

with a high probability of failure, 55.8%, the average block volume is approximately 82.38 m3. These results agree with the failure history for slopes in this area. According to Glass (1998), several small slides were reported each year and the large rockslide that occurred on July 1, 1997 was a wedge failure formed by J3 and J5. This finding for the site is verified by the results of the current research, showing high probabilities of failure for several small slides and for a large wedge slide. When the slope angle is assumed to be 508, the results of a deterministic analysis are much different from the previous deterministic results (Table 7; Fig. 8). Only the J4 and J5 combination is found to be kinematically unstable and the factor of safety for the sliding mode, with contact on plane 2, equals 0.33. The other combinations show kinematically stable conditions. However, based on probabilistic analysis, a total of eight combinations show kinematic instability, even though the probabilities are reduced significantly from the previous probabilities on the 758 cut slope. As can be observed in Table 7, the probabilities of total kinetic instability (that is, the sum of four probabilities of kinetic instability based on a sliding mode) for J2 and J5, J3 and J5 and J4 and J5 combinations are still high, 68.3%, 75.5% and 69.8%, respectively. As the results in Table 7 show, we observe that J5 plays an important role in kinetic

Table 6 Results of wedge failure for the deterministic analysis and the probabilistic analysis in I-40 area at 758 cut slope Set no. 1 Set no. 2 Factor of safety Stable FS1 = 0 FS3 = 0.32 Stable FS3 = 0.43 Stable Stable Stable Stable FS3 = 0.09 FS3 = 0.36 Stable FS3 = 0.33 FS1 = 0 Stable Probability of failure Kinematic 0.001 0.4 0.374 0.037 0.189 0 0.002 0.002 0 0.935 0.738 0.015 0.471 0.309 0.009 No contact 0.025 0.014 0.004 0.038 0.005 0 0 0.338 0 0.012 0.191 0.023 0.046 0.011 0.554 Plane 1 0 0 0 0 0 0 0 0.005 0 0 0 0 0.001 0 0.012 Plane 2 0.036 0.0125 0.0074 0.244 0.0067 0 0.0045 0.331 0 0.013 0.561 0 0.514 0 0.193 Both planes 0 0 0 0 0 0 0 0.0029 0 0 0.0037 0 0.0014 0 0 Total probability of failure 0 0.011 0.004 0.010 0.002 0 0 0.001 0 0.023 0.558 0.0004 0.265 0.0034 0.0068 Average volume of possible wedge (m3) 0.006 0.011 0.004 0.095 0.002 0 5.39 15.35 0 6.86 82.38 0.13 1.72 0.019 0.041

J1 J1 J1 J1 J1 J2 J2 J2 J2 J3 J3 J3 J4 J4 J5

J2 J3 J4 J5 J6 J3 J4 J5 J6 J4 J5 J6 J5 J6 J6

H.-J. Park et al. / Engineering Geology 79 (2005) 230250

247

(a) Combination of joint sets 1 and 3

0.35 0.30 0.25

(b) Combination of joint sets 1 and 5

0.07 0.06 0.05

Frequency

Frequency
1 2 3 4 5

0.20 0.15 0.10 0.05 0.00

0.04 0.03 0.02 0.01 0.00

1 2 3 4 5

Factor of safety

Factor of safety

(c) Combination of joint sets 2 and 5

0.16 0.14 0.12

(d) Combination of joint sets 3 and 4

0.10 0.08

Frequency

Frequency

0.10 0.08 0.06 0.04 0.02 0.00 1 2 3 4 5

0.06 0.04 0.02 0.00

Factor of safety

Factor of safety

(e) Combination of joint sets 3 and 5

0.14 0.12 0.10 0.18 0.16 0.14 0.12 0.10 0.08 0.06 0.04 0.02 0.00 1 2 3 4 5 0.02 0.00

(f) Combination of joint sets 4 and 5

Frequency

0.08 0.06 0.04

Frequency

Factor of safety

Factor of safety

Fig. 8. Histogram of factor of safety calculated in probabilistic analysis for wedge failure at 508 slope angle.

instability. However, there is an interesting conclusion: as the slope angle is reduced, the probabilities of kinetic failure when contact on both planes is lost actually increase. This indicates that the effective normal forces acting on both planes are reduced and these values become smaller than the pore water pressures on the sliding planes as the slope angle is reduced (Duzgun et al., 2003). However, probabilities

of total kinetic instability are the same as the probabilities for the 758 cut slope. That is, the decrease in slope angle does not affect the distribution and values of factor of safety, as shown previously for planar failure. This can be confirmed by noting that the factor for safety for the combination of J4 and J5 at the 758 slope angle in Table 6, 0.33, is the same as the value for the 508 slope angle in Table 7.

248

H.-J. Park et al. / Engineering Geology 79 (2005) 230250

Table 7 Results of wedge failure for the deterministic analysis and the probabilistic analysis in I-40 area at 508 cut slope Set no. 1 Set no. 2 Factor of safety Stable Stable Stable Stable Stable Stable Stable Stable Stable Stable Stable Stable FS3 = 0.33 Stable Stable Probability of failure Kinematic 0 0.003 0.003 0.004 0 0 0.001 0.001 0 0.190 0.683 0 0.332 0 0 No contact 0 0.015 0.005 0.17 0 0 0.003 0.460 0 0.017 0.434 0 0.231 0 0 Plane 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Plane 2 0 0.012 0.011 0.22 0 0 0.004 0.220 0 0.008 0.320 0 0.467 0 0 Both planes 0 0 0 0 0 0 0 0.002 0 0 0.001 0 0 0 0 Total probability of failure 0 0 0 0.0014 0 0 0 0.0006 0 0.0047 0.481 0 0.232 0 0 Average volume of possible wedge (m3) 0 0.005 0.001 0.002 0 0 0.118 0.441 0 0.314 3.78 0 0.0381 0 0

J1 J1 J1 J1 J1 J2 J2 J2 J2 J3 J3 J3 J4 J4 J5

J2 J3 J4 J5 J6 J3 J4 J5 J6 J4 J5 J6 J5 J6 J6

Regarding the average volume of a possible block for the 508 slope angle, volumes are reduced significantly from those of the 758 slope, as expected. Among seven combinations which are analyzed as stable by the deterministic but unstable in the probabilistic analysis in Table 7, six joint combinations have mean volume less than 0.6 m3. Many combinations become stable and the mean volumes of possible wedge are reduce as slope angle decreases. In particular, the volume for the J3 and J5 combination is reduced from 82.39 m3 to 3.78 m3, showing that the probability of a large volume failure was significantly reduced after the cut slope angle was flattened by slope processes.

7. Summary and conclusions Rock slope stability is highly dependent on discontinuity characteristics, and the random properties of each parameter have an important effect in the probabilistic analysis. Therefore, random properties for geometric and strength parameters of discontinuities play a critical role in the probabilistic analysis. In this study, discontinuity parameters including orientation, length and spacing were measured in the field and their random properties determined on the basis of physical considerations and goodness-of-fit testing. In addition, the new concept of persistence, which can be utilized effectively in the probabilistic

approach, was proposed. The proposed approach simply uses joint length data rather than the persistence value. This is significant because field determination of persistence is not possible on a practical basis. Therefore, the proposed approach expresses the probability that the joint length is equal to or greater than the maximum sliding length, which is multiplied by the probability of failure of the rock slope, the latter being evaluated assuming a fully persistent joint. To evaluate rock slope stability, both kinematic and kinetic conditions were examined. Both conditions were evaluated simultaneously. This is because examination of kinetic condition is conducted only after the kinematic failure is indicated. Then the probabilities of the two conditions were calculated separately. They were combined to evaluate the overall probability of slope failure. The Monte Carlo simulation technique was utilized to analyze the possibility of failure for planar and wedge features in the study area. This probabilistic analysis was applied to a study area on I-40 in western North Carolina. Comparisons between the deterministic analysis and probabilistic analysis showed that results of the probabilistic method yields significantly different results from those of the deterministic analysis. In some cases the deterministic analysis, based on a fixed value for discontinuity and slope parameters, did not indicate a slope failure condition whereas the

H.-J. Park et al. / Engineering Geology 79 (2005) 230250

249

probabilistic method did. Variations in discontinuity orientation are one cause for this difference, as discontinuity orientations have a wide scatter within the same set. This illustrated by J1 in Table 4 and the J3 and J5 combination of Table 7. In this study, the study area has experienced several small slope failures and a few large slope failures. With the details of the slope failure history, the probability of slope failure can be updated. According to Bayesian approach, the probability of failure can be updated by new information and subjective judgment. In the approach, the probability of failure is treated as a random variable and the probability of failure is updated using all available information, both theoretical and experimental (Powell and Pine, 1996). Therefore, the probability of slope failure in this area can be updated based on detailed failure history and the authors will focus on this subject in a further study.

Acknowledgements This research is partly supported by Korea Ministry of Science and Technology (Project No. M1-030200-0063). The authors would like to express their gratitude to the anonymous reviewers for their valuable comments.

References
Ang, A.H.S., Tang, W.H., 1975. Probability Concepts in Engineering Planning and Design, vol. 1. Wiley, New York. Baecher, G.B., 1983. Statistical analysis of rock mass fracturing. J. Math. Geol. 15 (2), 329 347. Baecher, G.B., Lanney, N.A., 1978. Trace length biases in joint surveys. Proceedings of 19th US Symposium on Rock Mechanics, pp. 56 65. Barton, N.R., 1973. Review of a new shear strength criteria for rock joints. Eng. Geol. 7, 287 332. Casagrande, A., 1965. Role of the bCalculated RiskQ in earthwork and foundation engineering. J. Soil Mech. Found. Div. 91 (4), 1 40. Cherubini, C., Cotecchia, V., Renna, G., Schiraldi, B., 1983. The use of bivariate probability density functions in Monte Carlo simulation of slope stability in soils. Proc. 4th Int. Conf. on the Application of Statistics and Probability to Soil and Structural Engineering, pp. 1401 1411. Cruden, D.M., 1977. Describing the size of discontinuities. Int. J. Rock Mech. Min. Sci. Geomech. Abstr. 14, 133 137.

Duzgun, H.S.B., Yucemen, M.S., Karpuz, C., 2003. A methodology for reliability based design of rock slopes. Rock Mech. Rock Eng. 36 (2), 95 120. Einstein, H.H., 1996. Risk and risk analysis in rock engineering. Tunn. Undergr. Space Technol. 11 (2), 141 155. Einstein, H.H., Baecher, G.B., 1983. Probabilistic and statistical methods in engineering geology; specific methods and examplesPart 1: exploration. Rock Mech. Rock Eng. 16, 39 72. Einstein, H.H., Veneziano, D., Baecher, G.B., OReilly, K.J., 1983. The effect of discontinuity persistence on rock sloe stability. Int. J. Rock Mech. Sci. Geomech. Abstr. 20, 227 236. El-Ramly, H., Morgenstern, N.R., Cruden, D.M., 2002. Probabilistic slope stability analysis for practice. Can. Geotech. J. 39, 665 683. Fisher, R.A., 1953. Dispersion on a sphere. Proc. R. Soc. Lond., A 217, 295 305. Giani, G.P., 1992. Rock Slope Stability Analysis. A.A. Balkema. Glass, F.R., 1998. A large wedge failure along Interstate 40 at North CarolinaTennessee State line. Proceedings of 48th Highway Geology Symposium. Arizona Department of Transportation, Prescott, Arizona, pp. 65 75. Gu nther, A., 2003. SLOPEMAP: programs for automated mapping of geomaterical and kinematical properties of hard rock hill slopes. Comp. Geosci. 29, 865 875. Harr, M.E., 1987. Reliability Based on Design in Civil Engineering. McGraw-Hill, New York. Herget, G., 1978. Analysis of discontinuity orientation for a probabilistic slope stability design. Proceedings of 19th U.S. Symposium on Rock Mechanics, Reno, Nevada. University of Nevada, pp. 42 50. Hoek, E.T., 1998. Factor of safety and ptrobability of failure (Chpater 8). Course notes, Internet edition, http://www.rockeng. utoronto.ca/hoekcorner.htm. Hoek, E.T., Bray, J.W., 1981. Rock Slope Engineering. Institute of Mining and Metallurgy. Hudson, J.A., Priest, S.D., 1983. Discontinuity frequency in rock masses. Int. J. Rock Mech. Sci. Geomech. Abstr. 20, 73 89. ISRM, 1978. Suggested methods for the quantitative description of discontinuities in rock masses. Int. J. Rock Mech. Sci. Geomech. Abstr. 15, 319 368. Kemeny, J., 2003. The time reduction of sliding cohesion due to rock bridge along discontinuities: a fracture mechanics approach. Rock Mech. Rock Eng. 36 (1), 27 38. Kulatilake, P.H.S.W., Wathugala, D.N., Stephansson, O., 1993. Joint network modeling with a validation exercise in Strip mine, Sweden. Int. J. Rock Mech. Sci. Geomech. Abstr. 30, 503 526. Kulatilake, P.H.S.W., Um, J., Wang, M., Escandon, R.F., Varvaiz, J., 2003. Stochastic fracture geometry modeling in 3-D including validations for a part of Arrowhead East Tunnel, California, USA. Eng. Geol. 70, 131 155. Li, K.S., White, W., 1987. Probabilistic approaches to slope design. Research Report, vol. 20. Dept. of Civil Engineering, Australian Defense Force Academy, Canberra, Australia, p. 4. Mahtab, M.A., Yegulalp, T.M., 1982. A rejection criterion for definition of clusters in orientation data. Proceedings of 22nd Symposium on Rock Mechanics. American Institute of

250

H.-J. Park et al. / Engineering Geology 79 (2005) 230250 Priest, S.D., Hudson, J.A., 1981. Estimation of discontinuity spacing and trace length using scanline surveys. Int. J. Rock Mech. Min Sci. Geomech. Abstr. 18, 183 197. Rethai, L., 1998. Probabilistic Solutions in Geotechnics. Elsevier, Amsterdam. Rouleau, A., Gale, J.E., 1985. Statistical characterization of the fracture system in the Strip Granite, Sweden. Int. J. Rock Mech. Min. Sci. Geomech. Abstr. 22 (6), 353 367. Scavia, C., 1990. Fracture mechanics approach to stability analysis of rock slopes. Eng. Fract. Mech. 35, 899 910. Schultze, E., 1975. The general significance of statistics for the civil engineer. Proceedings of the 2nd International Conference on Application of Statistics and Probability in Soil and Structural Engineering, Aachen. Sen, Z., Kazi, A., 1984. Discontinuity spacing and RQD estimates from finite length scanline. Int. J. Rock Mech. Min. Sci. Geomech. Abstr. 21 (4), 203 212. Singh, R.N., Sun, G.X., 1989. Fracture mechanics approach to slope stability analysis. Int. Symp. on Surface Mining. University of Nottingtton, England, pp. 93 97. Terzaghi, R.D., 1965. Source of error in joint surveys. Geotechnique 15, 287 304. Wallis, P.F., King, M.S., 1980. Discontinuity spacings in a crystalline rock. Int. J. Rock Mech. Min. Sci. Geomech. Abstr. 17, 63 67. West, T.R., 1996. The effects of positive pore pressure on sliding and toppling of rock blocks with some considerations of intact rock effects. Environ. Eng. Geosci. 2, 339 353. Whitman, R.V., 1984. Evaluating calculated risk in geotechnical engineering. J. Geotech. Eng., ASCE 110 (2), 145 186. Whittaker, B.N., Singh, R.N., Sun, G., 1992. Rock Fracture Mechanics; Principles, Design and Application. Elsevier. Wiener, L.S., Merschat, C.E., 1975. Field Guidebook to the Geology of the Central Blue Ridge of North Carolina and Spruce Pine Mining District. Association of American State Geologists.

Mining Metallurgical and Petroleum Engineers, New York, pp. 116 123. Martiz, J.S., 1981. Distribution-free Statistical Methods. Chapman and Hall, London. Mauldon, M., 1994. Intersection probabilities of impersistent joints. Int. J. Rock Mech. Sci. Geomech. Abstr. 31 (2), 107 115. Mauldon, M., Dunne, W.M., Rohrbaugh, M.B., 2001. Circular scalines and circular windows: new tools for characterizing the geometry of fracture traces. J. Struct. Geol. 23, 247 258. McMahon, B.K., 1971. Statistical methods for the design of rock slopes. 1st AustralianNew Zealand Conference on Geomechanics, pp. 314 321. Mostyn, G.R., Li, K.S., 1993. Probabilistic slope analysisstate of play. Proceedings of Conference on Probabilistic Methods in Geotechnical Engineering. A.A. Balkema, Canberra, Australia, pp. 89 109. Nilsen, B., 2000. New trend in rock slope stability analysis. Bull. Eng. Geol. Environ. 58, 173 178. NCDOT, 1980. I-40:Slope Stability Study Final Report, vols. 1,2,3. North Carolina Department of Transportation. Park, H.J., 1999. Risk analysis of rock slope stability and stochastic properties of discontinuity parameters in western North Carolina. PhD thesis. Purdue University. Park, H.J., West, T.R., 2001. Development of a probabilistic approach for rock wedge failure. Eng. Geol. 59, 233 251. Pathak, S., Nilsen, B., 2004. Probabilistic rock slope stability analysis for Himalayan condition. Bull. Eng. Geol. Environ. 63, 25 32. Peck, R.B., 1969. Advantages of limitations of the observational method in applied soil mechanics: 9th Rankie Lecture. Geotechniques 19 (2), 171 187. Powell, N., Pine, R.J., 1996. Bayesian approach to slope stability assessment by updating probability of failure treated as a random variable. Trans. Inst. Min. Metall. 105, A31 A36. Priest, S.D., 1993. Discontinuity Analysis For Rock Engineering. Chapman and Hall, New York.