Reliability analysis of infinite slope using subset simulation

Abstract

It is important to incorporate the spatial variability of soil properties in slope stability analysis. Most studies use a random field to model the variability of soil properties and estimate the reliability using direct Monte Carlo simulation (MCS), which is not efficient for problems with small failure probabilities. The slope reliability can be more efficiently estimated using the subset simulation method, as it computes a small failure probability as a product of larger conditional probabilities. The approach is demonstrated using an infinite slope with undrained shear strength varying with depth. It will be shown that subset simulation requires much less samples than MCS to achieve a given accuracy. Subset simulation results are within the first order probability bounds derived from classical system reliability theory. It is found that the failure surface can be developed in any soil layer, not necessarily atthe soil-rock interface. Copyright ASCE 2009.