Probabilistic finite element analyses of slope stability due to heavy rains: a comparison of homogeneous and spatially variable models

Abstract

Consequences of the climate change like melting of the permafrost and the increasing number of extreme events like heavy rains are all factors which can increase the probability of occurrence of slope instability. For geotechnical engineers, a deeper understanding of this phenomenom becomes unavoidable.

This paper aims to find the difference between two types of probabilistic slope stability analyses under a rainfall event. The first admits a homogeneous soil and takes into account uncertainty of the soil properties. The second takes into account spatial variability of the soil properties. Both analyses are performed using a Monte Carlo framework with the help of a Polynomial Chaos Expansion (PCE). The PCE is built based on 2D finite element realisations and allows an almost immmediate evaluation of new ones, which is a computing time advantage when performing the Monte Carlo method. A comparison between the probabilities of failure, e.g. the probability of the factor of safety of the slope being smaller than 1, of the two types of analyses is shown. A comparison between the probabilities of failure given different correlations lengths of the spatial variability is also presented. It is shown that the probabilities of failure, when the soil properties are variable in space, are higher than the ones where the soil is homogeneous. Indeed, with soil properties spatially variable, it is also possible to have failure planes that occur at the surface.