Harmonizing engineering geology with rock engineering for assessing rock slope stability
The Role of Engineering Geological Model
The engineering geological model of a rock slope is a comprehensive expression of the various factors which affect the slope stability, and in general, includes the following principal contents: (i) the basic geologic conditions of the slope, (ii) mechanical properties of rock mass and discontinuities, (iii) principal artificial and natural dynamic factors affecting the stability (groundwater, earthquake etc.), (iv) the developing process and characteristics of’ the rock mass deformation on the slope, and (v) the failure pattern of the slope. By combining geomorphic evidence with the above-given conditions based on detailed assessments both in field and laboratory, engineering geological model of the slope focusing the attention on primary and secondary failure modes and triggering mechanisms should be established.
Real cases and hypothetical examples showing the importance of engineering geological model in rock slope assessments are briefly discussed in the following paragraphs.
The first example is from a lignite open pit. Engineering geological conditions in this pit with movement monitoring data are illustrated in Figure 1A. In this figure, directions and magnitudes of the movement vectors are generally consistent with the bedding in the overburden and Fault 6. Shear strength of bedding in the overburden was smaller than that of the clay underlying the coal seam.
All these engineering geological information indicated that the most possible critical mode of failure should be multi-planar. A quick overburden stripping between points A and B at the toe of the slope (Figure 1A) resulted in shortening of the length of the bedding oppositely dipping to the slope and caused a movement. The uppermost benches in the alluvial soils also moved downward to fill the gap resulted from the movement of the lower benches. This movement confirmed the engineering geological model.
The second example is a simple hypothetical rock slope with the conditions illustrated in Figure 1B.
In the construction of engineering geological model for this slope, ignoring both the weak tuff layer and the fault, which may cause a bi-planar failure affecting the entire slope, the mode of failure may be anticipated to occur only throughout the highly weathered portion of the volcanic rock in the form of a circular sliding surface. Contrary to this, if the degree, depth, extent and nature of the weathering in the slope are not well identified, one of the two possible failures (circular failure) can also be missed.
The most important requirement for correct characterization is for investigators to recognize bimrocks and not approach them as stratigraphically orderly (layer-cake) rocks.Once the words “interlayered” or “interbedded” are used in boring logs, there is a tendency to imagine continuous “layer-cake” stratigraphic contacts between borings.
Otherwise, misinterpretation of the geologic data by non-geologists may result in unexpected costs and construction difficulties.
Upper sketch in Figure 2a shows the interpretation of geology based intersections by borings (terminated about 2m depth) of an assumed continuous and homogeneous sandstone bedrock surface in the Franciscan melange
(Bimrock) in US]. This model called that the slope instability was shallow, being composed of clay and boulder colluviums sliding along the contact with the underlying sandstone bedrock and it was decided to remove the shallow failed material. But during excavation, the bedrock couldn’t be found and it was recognized that the instability was actually deep-seated in sheared melange, rather than the shallow soil mass.
Bedrock was the result of the misconception by connecting straight lines (the red boundary in Figure 2A) between the interpreted soil/rock contacts as intersected by borings. The excavation was deepened below the design depth of a few meters to tens of meters. The repair finally cost more than a million dollars. Lower sketch in Figure 8b shows a more realistic model to assess bedrock conditions, in which deeper borings intersect discrete blocks in a bimrock containing blocks within a matrix.
In terms of engineering geological model, the other conditions, regarding the orientation and distribution of the blocks, that should be considered when analysing slope stability problems in bimrocksare illustrated in Figure 3.
When the bimrock includes a few blocks, it can be analysed as conventional soil (Figure 3A); failure surface is influenced by the orientation and nature of matrix shearing and it cannot be readily analysed as soil or rock (Figure 9b); blocks are oriented at high angles to the slope, which increases stability due to the increased tortuosity at the failure surface forced around the blocks (Figure 9c); and block-poor zones within generally block-rich bimrock are weaker and more likely to fail (Figure 9d).
Medley and Sanz Rehermanndeveloped a simple model to investigate the slope stability of idealized bimrocks. They compared their findings to those of Irfan and Tang], who performed stability analyses of theoretical slopes modelled bouldery soil and found that there is a good relationship between the normalized factors of safety and the volumetric block proportions, despite the significant differences in the model geometries, the orientation of blocks, the geology of the modelled materials, and analytical methods used. Although considerably more analyses should be performed to define the statistical variations, it appears that up to about 25% to 30% volumetric block proportion, the presence of blocks provides relatively little geomechanical advantage.
From this lower limit to greater than 55%, there is marked an increase in slope stability. However, more future studies are necessary, as recommended by Medley and Sanz Rehermann, perhaps by performing Monte-Carlo type simulations using 3-D geomechanical models, to understand the statistical viability of using simple analytical approaches for complex geological conditions.
Selection of the strength parametars
Reliable estimates of mechanical properties of discontinuities and rock masses are required for almost any analysis used to design rock slopes, and contributions to their selection both from engineering geology and rock mechanics are necessary. The evaluation of engineering properties, which are the quantities directly measured or empirically estimated, should consider the identified failure modes and mechanisms.
Shear strength is the main parameter for rock slope stability analysis and can be described by either linear or non-linear failure criterion. This is also valid for a differentiation between the peak and residual strength.
Based on the scale effect, which is an important issue in rock slopes, as illustrated in Figure 4A, and engineering geological conditions discussed in the previous sections, it is necessary to use the shear strength properties of either the discontinuities or of the rock mass. In the analysis of structurally controlled rock slope failures, the general trend is to use the Mohr-Coulomb and empirical Barton’s failure criteria for smooth and rough discontinuities, respectively.
If a discontinuity is filled by a soft/weak material, its shear strength is governed by the strength of the filling (Figure 4B).
Because it is extremely difficult to measure rock mass strength directly using full-scale field tests due to their high costs and practical difficulties, back-analysis of previous failures may be used for its estimation, however, it should be kept in mind that it has also some limitations.
As an alternative, the strength of rock masses usually is estimated from empirical relations and/or rock mass classification systems. However, there are some legacies inherited from these systems developed for other purposes than rock slope stability as discussed by Hack, in detail. One of the weaknesses of these systems includes the fact that they are not based on mechanics and that they combine all characteristics into a single number. Since the 1980’s the empirical Hoek-Brown failure criterion, which is based on rock mass rating and then Geological Strength Index (GSI;), is being commonly used for estimating shear strength of rock masses.
Since the Hoek-Brown criterion represents a curved failure envelope, a transition to the linear Coulomb criterion is often conducted. This is because a linear failure envelope is easier to handle both analytical and numerical methods. However, by considering that the failure envelope of a jointed rock mass, in reality, is curved, no transition to a linear envelope should be preferred. In addition, the empirical relationships based on RMR ] and Q values, are also the other alternatives to be used for the same purpose. A most recently developed new rock mass rating system, RMQR, is used to estimate the geomechanical properties of rock masses.
As an example on the selection of shear strength parameters, the complex failure surface illustrated in Figure 4 can be considered. In this figure, Zone A, the upper part of the failure surface consisting of a fault, is represented by Mohr-Coulomb criterion; for Zone B, heavily jointed central part, empirical rock mass failure criterion can be used; and for Zone C, depending on surface characteristics of the step-path discontinuities, Mohr-Coulomb or Barton’s criterion is preferred.
Determination of strength of bimrocks is one of the difficult issues in rock engineering. Mechanical properties of the matrix, the volumetric block proportion, shape and size distribution of blocks and their orientation relative to failure surfaces are the main factors affecting their overall mechanical properties.
Neglecting the contributions of blocks to overall bimrock strength, choosing instead to design on the basis of the strength of the weak matrix may result in too conservative for many bimrocks in terms of slope design. Based on the study on a physical model melange by Lindquist, when the block proportions are between about 25% and 70%, the increase in the overall mechanical properties of bimrocks are mainly related to the volumetric proportion of the blocks in the rock mass (Figure 4C).
Although some efforts have been performed to assess the strength of bimrocks based on physical models, in-situ tests and equivalent material techniques, further studies to develop more efficient methods and a database for bimrock strength are still needed.
Critical review on the methods of rock slope stability assessment
Conventional and Numerical Methods
The rock slope stability analysis may be mainly divided into two groups of methods: (i) the conventional methods (kinematic analyses based on stereographic projection technique, 2D limit equilibrium methods (LEM) and rock fall simulations), and (ii) numerical methods. The main inputs, advantages and limitations of the conventional and numerical methods, which have been discussed in literature in detail, are summarized in Table 1.
The method of kinematic analysis only helps engineers to recognize potential structurally controlled rock slope instabilities, such as planar, wedge and toppling failures and is more relevant to low height rock slopes. The LEM is used to determine the factor of safety, which is an indicator of the stability of slopes. LEMs are highly relevant to simple block failures along discontinuities or rock slopes that are behaving like soil due to their heavily fractured or weathered nature.
Due to the common acceptance of the safety factor approach as the main criterion of slope stability, LEM has been used more often and seems still to remain the most commonly adopted method. In the case of rockfalls, it is generally impossible to secure all blocks, and therefore, consideration is given to the design of protective measures around structures endangered by the falling blocks. Rockfall protection works, therefore, largely involves the determination of travel paths and trajectories of unstable blocks that have detached from a rock slope face.
For the purpose, 2D and 3D computer-based programs, called “rockfall simulators”, analyse the trajectory of falling blocks and also help to determine remedial measures. However, 3D simulations are more reliable and realistic than 2D simulations in case of complex morphology and then trajectories not parallel to each other along the slope. In addition, unexpected bounces, due to local obstacles and not intersected by 2D sections, can be seen only with 3D simulations.
This is because at such scales stability is affected by the strength and deformation properties of both intact rock and discontinuities, the geometry and distribution of discontinuities throughout the rock mass and complexities related to material anisotropy, non-linear behaviour, in-situ stresses and the presence of coupled processes such as pressures,
seismic loading etc. These types of failure cannot be modelled by means of conventional methods of analysis.
The numerical modelling techniques have been developed to generate a range of possible solutions for such rock slope stability problems. Numerical methods can be considered in three groups, namely continuum modelling, discontinuum modelling and hybrid modelling. Continuum modelling is applicable to rock slopes composed of intact rock, weak rocks and heavily jointed rocks and include the finite element (FEM) and finite difference (FDM) methods.
Discontinuum modelling is suitable for rock slopes of which stability is controlled by discontinuity behaviour and is referred to as the discrete element method (DEM). Most recently hybrid approaches, including the coupling of continuum and discontinuum approaches, which often fails to realistically simulate the progressive failure of rock slopes, particularly the dynamics of kinetic release accompanying complex internal distortion, dilation and fracture, are increasingly being adopted rock slope stability analysis as an advanced numerical modelling (Figure 5).
Methods for Earthquake-Induced Slope Failures
The use of empirical bounds by Keefer and an equation proposed by Aydan for the maximum distance of disrupted and coherent earthquake-triggered slope instabilities are the approaches for preliminary assessments.
Three basic inertial methods: (i) pseudo-static analysis, (ii)Newmark’s displacement method, and (iii) dynamic analysis, for assessing the earthquake-induced rock slope failures are available in the literature.
The greatest difficulty with the first method, based on a static LEM analysis treating earthquake motion as a static force, involves conservative results, the selection of an appropriate seismic coefficient and the value of an acceptable safety factor and no information on the magnitude of displacement. Newmark’s method is the extension of the simple pseudo-static method by directly considering the accelerogram of the sliding mass within the slope to determine permanent displacements. However, it involves many assumptions and limitations.
Dynamic analysis typically incorporates an FDM (e.g. FLAC3D code; DEM approximation (e.g. 3DEC code; or a FEM approximation, which compute slope stresses and strains using earthquake accelerograms as input and obtain permanent deformation of the slope. The estimation of the travel distance of natural slopes upon dynamic failure is also of great importance. Although this issue is well known and some methods are available, the present numerical methods are still insufficient to model earthquake-induced post-failure motions.
By considering that the best laboratory is the nature itself, back-analysis of previous slope failures is an attractive method to estimate operative shear strength along the sliding surface at the time of failure. For back-analysis, the failure mode should be well established and accurate information on the failure geometry, groundwater conditions and other factors, which are considered to have contributed to the failure, should be available. Often LEM is used to back-calculate the strength, assuming safety factor is equal to unity. In practice, LEM is used to estimate cohesion (c) and friction angle by assuming one of these parameters and back-calculating the other for a safety factor of unity.
The most efficient solution, in the back-analysis of failure surfaces of which shear strength is expressed by Mohr-Coulomb linear failure criterion, is to analyse more than one cross-section obtained from the instability and then to estimate the operative shear strength parameters from the intersection of the c-f curves satisfying a factor of safety of unity, called multiple solutions, or to compare these intersections with the range of peak and residual strengths of the sliding surface experimentally determined in laboratory or field, if available (Figure 6A).
In the case of failures along rough discontinuities or through closely jointed rock masses, which obey to non-linear failure criteria (i.e. different c and f values are operative along the sliding surface depending on normal stress), the above-mentioned approach is not valid.
As shown in Figure 12b-c, in such cases, shear stresses estimated from different points selected along the sliding surface under different normal stresses depending on the suitable failure criterion (Hoek-Brown criterion for the heavily jointed rock mass in Figure 6B and Barton’s criterion for the rough discontinuity surfaces in Fig. 6C) are plotted onto the failure envelope of the sliding surface and if there is a good match between the failure envelope and shear stress plots, it can be concluded that the failure geometry analysed represents the actual failure and the failure criterion used is correctly chosen.
Descamps and Yankey, who discussed the limitations in the back-analysis in detail, indicate that back-analysis is reliable only when the model and all assumptions are reasonable and accurate representations of the real system. Therefore, in practice, both LEM and numerical modelling tools, and if possible, movement monitoring data should be used together to generate a more accurate solution in back-analysis of the failed rock slopes.
Role of Movement Monitoring
Slope failures, such as occurred in theVaiont Dam, result in loss of life and damage to or destruction of engineering structures. In open-pit mines, where access is generally restricted, slope failures may cause loss of life or not, but they interrupt the mining operation and cause damage to property. Therefore, threats due to gradual deformations in slopes should be well managed and for this purpose slope monitoring is necessary.
Slope monitoring, which forms an integral part of slope management, is mainly conducted to detect potentially unstable ground, to assess the performance of slope design and to apply suitable remedial measures. If this procedure is properly carried out, the risk can be eliminated or reduced. For long years slope monitoring were practiced through visual observations including manual inspection and mapping of tension cracks along the slope face. Although these methods are useful, they are time-consuming and have limited accuracy.
In addition, some basic instrumentations such as borehole extensometer, inclinometer etc., which are installed in boreholes, are commonly used as subsurface monitoring techniques for slopes particularly in civil engineering practice and it seems that their use will also keep their popularity in the future.
At the beginning of the 21st century, the depth of open pits considerably increased reaching up to more than 1000 m and this progress resulted in an increased frequency of large pit slope failures. Advance warning of these slope instability problems became more important, and therefore, the importance of slope monitoring in open pit mining increased.
This situation necessitated the development of various new slope monitoring systems, which allow continuous monitoring for the routine inspection of the rock and their deformation. The recent technologies used to monitor slopes are Automated Total Station Networks, Light Detection and Ranging (LIDAR) scanning, Slope Stability Radar (SSR), Global Positioning System (GPS), Time Domain Reflectometry (TDR), high-resolution micro-seismic monitoring. Particularly in the last decade, active monitoring of pit slopes with radar, which is suited for predicting movements on the larger surface area, become standard practice. In recent years, microseismic monitoring technique also gained considerable attention for the early forecast of slope failures and in estimating deformations with considerable accuracy.
From the literature, it appears that although the microseismic monitoring method in the monitoring of natural and cut slopes for civil excavations is not regarded as having reach maturity, while its popularity and use in the monitoring of pit slopes are increasing since 2002.
Researches over the last two decades have led to the development of a system for slope monitoring based on the concept of measuring Acoustic Emission (AE). AE rates generated by active waveguides are proportional to the velocity of slope movement, and can, therefore, be used to detect changes in rates of movement in response to destabilizing and stabilizing effects, such as rainfall and remediation, respectively. However, for its more common use, further studies seem necessary.
Due to the more complex nature of rock masses when compared to soils, rock slope engineering is a complex field that requires rock engineers and engineering geologists to work together for developing a site-specific model.
Engineering geology provides direct inputs to rock engineering design approaches and harmonization of an accurate engineering geological model, which includes geology, structure, degree of deterioration, groundwater conditions and triggering static and dynamic factors, with material behaviour and rock engineering methodologies essential for rock slope stability assessments.
A number of studies are mainly concentrated on small and moderate size slope instabilities. Depending on the increase in the depth of slopes, due to complex failure mechanism in which the slip surface of composite shape can be formed, the current trend is integrating conventional and numerical methods together in conjunction with the data from conventional and advanced movement monitoring techniques.
It is also noted that particularly experiences on failure mechanisms associated with large-scale slopes in hard, brittle and jointed rock masses are limited and new approaches for the estimation of the strength of large-scale rock masses are needed.
Compilation of rock mass strength data from the back-analysis results particularly of large scale slope failures may form a very useful database for the assessment of slopes which have not subjected to any failure. Development of more powerful numerical methods which can better model the propagation of failure along discontinuities and through intact rock is one of the expectations.
Further studies are also necessary for a better understanding of characterization, strength and mechanical behaviour of bimrocksand faulted-fractured zones in the assessment of rock slope stability. In addition, the consideration of failures of natural rock slopes has received very little attention in earthquake engineering, and therefore, much attention should be given to the possibility of such slope failures and to assessment methodologies in terms of rock engineering.
Resat Ulusay; DGT BiH Geotehnika