Investigating the Influence of Conditional Simulation on Small-Probability Failure Events Using Subset Simulation

Abstract

Spatial variation of soil strength parameters is a dominating uncertainty in slope stability analysis. This uncertainty can be accounted for in a stochastic description, based on a global geostatistical characterization of the soil strength parameters, which leads to a wide range of possible slope responses, of which only a small proportion typically concern slope failure. This paper investigates the effect of including additional data to reduce the range of possible scenarios in the stability analysis of slopes in spatially variable soils. Subset simulation, which is a technique to focus the random sampling of Monte Carlo analyses in the region of interest, is used here to address the already small probabilities of slope failure. The reduction in the range of possible scenarios with respect to both the probability of failure and modes of failure is then investigated. The strong reduction in the range of expected failure modes through conditional simulation demonstrates the uncertainty reduction and the relationship between the distribution of weaker zones in the slope and the development of sliding surfaces.