Reliability - based slope stability analysis and associated severity of failure of tailings dams

Abstract

Slope stability analysis is crucial to the design of every new tailings storage facility, and while deterministic methods such as limit-equilibrium analysis are the most widely used method of analysis, the safety factors produced by them are ultimately difficult to comprehend for non-engineers and are often applied unknowingly to conditions involving widely varying degrees of uncertainty. There is, therefore, an increasing trend towards the determination of probabilities of failure in addition to safety factors to assess slope stability. The most common method of statistical analysis within the engineering field has historically been the Monte-Carlo method due to its conceptual simplicity, and reliability-based analysis has up until recently had little practical usage due to its computational and conceptual barriers. This dissertation compares pre-programmed probability methods available in “off-the-shelf” slope stability packages to various reliability-based analysis methods. Each of these methods is applied to several embankments/ waste facilities, specifically selected and configured to investigate a wide range of safety factors, probabilities of failure and the effect of material uncertainty. Based on the results of the analysis, and the discussions in the literature review, a suitable reliabilitybased analysis method is selected to be implemented in the remaining analyses and experiments. Consequence of failure is not often thoroughly assessed in early stages of design and is usually only quantified by the engineering judgement of experienced practitioners. This is generally done using only the critical failure surface returned in the limit-equilibrium analysis, and while larger failure surfaces may indeed have higher safety factors, their associated probability of failure may be unacceptable for that specific size of failure. Furthermore, the prescribed limits of acceptable probabilities of failure do not take into account the complex behaviour of tailings (such as the potential for liquefaction and flow to occur) and are, additionally, presented against the potential number of fatalities caused by failure which is considered inappropriate for the assessment of consequence of failure of tailings storage facilities. This dissertation therefore also introduces the concept of measuring ‘severity’ of failure and ‘slope performance’, which takes into account the effect material uncertainty has on the probability of failure of slip surfaces smaller and larger than the critical slip surface, and limits of acceptable, intermediate and unacceptable probabilities of failure are proposed in order to provide a completely objective assessment of whether the performance of a facility is acceptable.