Professional Documents
Culture Documents
Saliba Aloysio A Probabilistic Assasessment of Open Pit Slope Failures in The Post Closure Phase
Saliba Aloysio A Probabilistic Assasessment of Open Pit Slope Failures in The Post Closure Phase
A probabilistic assessment of
open-pit slope failures in the post-closure phase
Aloysio P. M. Saliba, TEC3 Geotecnia e Recursos Hdricos, Brazil
Fernando P. M. Saliba, TEC3 Geotecnia e Recursos Hdricos, Brazil
Jhoan S. P. Panitz, TEC3 Geotecnia e Recursos Hdricos, Brazil
Rodrigo P. de Figueiredo, Universidade Federal de Ouro Preto, Brazil
Fernanda B. S. Duarte, TEC3 Geotecnia e Recursos Hdricos, Brazil
Abstract
In the mining sector, open-pit slopes are designed according to the factor of safety values calculated for
the operational and closure phases using a deterministic approach. In this methodology, assuming a
failure mode (e.g., planar, circular) and a set of input parameter values (e.g., friction angle, cohesion,
phreatic surface level), the factor of safety is calculated in line with the proposed pit geometry (bench
heights and widths, slope angles). The open pit is divided into pit sectors of homogeneous geotechnical
properties, the geometry of which is then optimized to reach the suggested minimum factor of safety
values per engineering standards. This methodology is applied to each geometric layout of the pit during
its lifetime and closure phases; during the latter, a conservative geometry is usually adopted to allow for
the elimination of maintenance. However, this conservative geometry generally means a smaller pit slope
angle, which will lead to higher closure costs. The purpose of this paper is to examine the factor of safety
in the post-closure phase in light of a probabilistic assessment, in which each input parameter is
considered as a probability function derived from field and laboratory measurements using the Monte
Carlo method. According to this stochastic method, the probability distribution of each input parameter
is sampled 100,000 times and randomly combined with other input parameters to evaluate the slopes
factor of safety. As the function of a random variable is also a random variable, this method results in
the factor of safety probability distribution. After taking into account the minimum factor of safety values
for the post-closure phase as proposed by engineering standards, and allowing for failure risks, which
are defined as factor of safety values lower than engineering standard thresholds, the post-closure openpit geometry can be optimized using the probability distribution obtained for the slopes factor of safety.
Introduction
The design of open pits involves a variety of disciplines and requires both engineering and geology
knowledge. Rock mechanics, hydrogeology, pit surface drainage and slope stability are examples of the
different aspects to be considered.
However, pit slopes are composed of natural materials, which show variability in their geological
and geotechnical properties. The traditional design approach relies on the division of the open pit into
homogeneous zones of geotechnical properties, the geometry of which is then optimized to reach the
suggested minimum factor of safety values according to current engineering standards, professional
experience, and test results. Each pit geometry is analyzed for every phase of its lifetime, including the
closure phase, for which a conservative geometry is adopted to eliminate any future maintenance. The
factor of safety is then evaluated.
In this methodology, the factor of safety is calculated according to the proposed pit geometry
(bench heights and widths, and slope angles), assumed failure mode (e.g., planar or circular) and a set of
input parameter values (e.g., friction angle, cohesion, phreatic surface distance). In this regard, a
conservative geometry generally means a smaller pit slope angle, which will lead to higher operational
and closure costs.
An alternative approach consists of considering the variability of geomechanical properties in the
definition of pit slope geometry. In this method, a typical planar failure design method proposed by Hoek
and Bray (1977) is combined with the Monte Carlo method, and each geomechanical parameter is
defined as a random variable with a known probability distribution. The parameters of these distributions
are determined during field investigations or after analyzing drilling samples or lab test results. As a
factor of safety (FoS) is a function of these parameters according to the slope failure process considered,
their variability is incorporated into the safety factor itself. An empirical FoS probability distribution is
defined by sampling these distributions 100,000 times using the Monte Carlo method,
This distribution can subsequently be used to evaluate pit slope safety according to engineering
standards, and to consider the level of slope failure risk in the closure phase. Based on engineering
standards and risk levels, pit geometry can be proposed for the decommissioning phase, thus leading to
the optimization of closure costs.
Another positive outcome of the application of this stochastic approach is that the risk level of any
potential pit geometry can be defined, thus allowing the mining company to analyze the expected
behavior of the proposed pit geometry at any stage in the pit lifecycle.
Methodology
As defined by Hoek and Bray (1977), a plane failure occurs when a geological discontinuity, such as a
bedding plane, strikes parallel to the slope face and dips into the excavation at an angle greater than the
angle of friction, under the following assumptions:
A sliding or failure plane strikes parallel or approximately parallel (within 20 degrees) to the
face of the slope.
The failure plane daylights into the face of the slope. This condition occurs when the failure
surface dips at angle shallower than the slope face (Figure 1.a).
It is assumed that release surfaces are present so that there is no resistance to sliding at the lateral
boundaries of the failing rock mass (Figure 1.b).
The slope slice to be analyzed is taken perpendicular to the face of the slope and is assumed to
have a unit (1 m) thickness (Figure 1.c).
Figure 1: Geometry of slope exhibiting plane failure: (a) cross-section showing planes forming
a plane failure; (b) release surface at ends of plane failure; (c) unit thickness slide used in
stability analysis (Duncan and Christopher, 2004)
The FoS for plane failure is calculated after separating all forces acting on the slope into
components parallel and normal to the sliding planes. Details of this formula can be reviewed in the work
of Hoek and Bray (1977).
Traditional open-pit design is considered deterministic in the sense that point value estimates of all
input variables of this method are defined from field data or engineering experience. A sensitivity
analysis is usually performed to assess the expected range of variation. However, this analysis is
performed by changing one variable at a time.
The alternate method proposed in this paper is referred to as a stochastic approach, which defines
each design variable as a random variable represented by uniform distribution. A uniform distribution
occurs when each point in its domain is equally likely, which means each value has the same probability.
Priest and Brown (1983), Pine (1992), and Hoek (1998) have also proposed stochastic analyses of slope
stability, but suggest varying a pair of parameters at a time and using normal distributions.
In natural sciences, these variables are frequently assumed to be normally distributed because of
the central limit theorem, which states that any random variable will show a normal distribution provided
the sample size is large enough. However, given the variability expected from geological variables and
the small observational window that comes from field mapping and testing, or even laboratory testing,
the authors suggest the use of a uniform distribution, except in those cases in which a sufficient amount
of information is collected to support the selection of a probability distribution. This implies knowledge
of the variable variation range but not the shape of this variation.
The Monte Carlo method is used to sample each parameter distribution. In this method, a random
number is sampled from a uniform distribution bounded to zero and one. The number obtained is said to
be the cumulated probability density value (CDF) of the parameter, and its CDF is inverted to obtain the
respective parameter value. In the approach proposed here, each parameter has a uniform distribution
and this sampling procedure is repeated 100,000 times using GoldSim software (GTG, 2010).
GoldSim software uses the Latin Hypercube Sampling procedure, which ensures that all random
variables are divided into an equally likely number of strata (GTG, 2010), which means that every
variable range is divided into 100,000 sections to obtain a uniform distribution. Each variable is sampled
separately yet simultaneously, resulting in an FoS distribution produced by the addition of the
randomness in each of the input variables, which are then combined.
Deterministic
Upper bound
33
30
36
Dip angle ()
61
58
64
10
60
55
65
30
25
35
Figure 2 shows the pit slope geometry and results for the deterministic case, for which the FoS
value was calculated as 1.14. These results were produced using RocPlane software (Rocscience, 2001).
In Figure 3, one can see that the output PDF shows the form of a log-normal PDF. The FoS results
mean value was 1.20 with a standard deviation of 0.22. Figure 4 shows the cumulated probability density
function (CDF) of the calculated FoS values.
Conclusion
A stochastic method for pit slope design is presented and compared to the deterministic approach. Under
this stochastic method, many input parameters of a planar failure model are simultaneously sampled
using GoldSim software, with 100,000 realizations, which took about 15 seconds to run.
A pit slope with a 65 face angle, 10 m bench, and pit 15 m in height was designed for analysis
using both methods. Results indicated that the risk of slope failure could only be assessed using a
stochastic method, even if deterministic results were subject to sensitivity analyses. The methodology
proposed can also be extended to other slope failure modes (e.g., circular, wedge, topping).
Similar results were obtained from the average planar failure volume calculations of both methods.
However, the stochastic method also predicted the variation range of this variable, which was comparable
to the mean value itself.
The estimation of FoS values also adds the knowledge of incurred risk in the design process,
whatever the phase in the pit lifecycle. In fact, rock strength losses related to weathering and pore
pressure increases due to groundwater level recovery in the post-closure phase make the use of a
deterministic method subject to an even higher level of uncertainty when compared to other phases. In
this case, knowledge of this uncertainty is better understood using the stochastic approach proposed here.
References
Barton, N. R. and Bandis, S. (1980). Some effects of scale on the shear strengths of joints. International Journal of Rock
Mechanics, Mining Sciences and Geomechanical Abstracts, 17, pp. 69-76.
Duncan, C. and Christopher, W. (2004). Rock slope engineeringcivil and mining (4th ed). London, UK: Spons Press.
Goldsim Technology Group (GTG) (1010). GoldSim users guide. Probabilistic simulation environment (Volume 2).
Issaquah, WA: GTG.
Hoek, E. (1998). Rock engineering: Course notes. Rotterdam, Netherlands: Balkema.
Hoek, E. and Bray, J. (1977). Rock slope engineering. London: IMM.
Pine, R. J. (1992). Risk analysis design applications in mining geomechanics. Transactions of the Institution of Mining and
Metallurgy. (Section A: Mining Industry), 101, pp. A149A158.
Priest, S. D. and Brown, E. T. (1983). Probabilistic stability analysis of variable rock slopes. Transactions of the Institution of
Mining and Metallurgy (Section A: Mining Industry), 92, pp. A1A12.
Read, J. and Stacey, P. (2009). Guidelines for open pit slope design. Leiden, Netherlands: CRC Press/Balkema.
Rocscience (2001). RocPlane: Planar sliding analysis for rock slopes. Toronto, Canada: Rocscience Inc.