A New Stochastic Response Surface Method in Spatial Variability Slope Stability Analysis

Abstract

Spatial variability is becoming more and more common in limit equilibrium slope stability analysis. With the recent advances in the abilities for limit equilibrium slope stability analysis to handle complicated models, it is more important than ever to have a fast and accurate method to perform slope stability analysis in spatially varying soils. The use of response surfaces is widely adopted in the literature to deliver efficient stochastic analyses. However, their implementation is usually deeply correlated with the choice of a field generation algorithm. This makes the response surface method inflexible since it sometimes requires the field generation algorithm to be completely rewritten. Finely discretized spatially varying random fields can have many random variables which make the response surface difficult to determine. To solve this, the authors have proposed a modified response surface guided approach to slope reliability analysis in spatially varying soils which is independent of the choice of field generation algorithm. This approach was evaluated using a two-dimensional problem, and validated by comparing its accuracy and speed against the traditional Monte Carlo simulation.