Reliability analysis of embankment dam sliding stability using the sparse polynomial chaos expansion

A probabilistic stability analysis of an embankment dam using in-situ measurements is presented in the paper. Three soil properties, including the dry density ( γ d ), the effective cohesion ( C ′ ) and the friction angle ( ϕ ′ ) are simulated by means of random variables based on in-situ measurements or laboratory test results, to consider the soil variability. Concerning the deterministic calculations, two models are developed in the study. The first one is a numerical model based on the finite difference method, and the second one is an analytical model based on the limit equilibrium method. For the latter, a two-step Genetic Algorithm (GA) is proposed to locate the best noncircular critical slip surface. These two models are both used for the safety factor evaluation of the studied dam and compared in a probabilistic framework in terms of safety factor distribution and failure probability. The employed reliability method for the probabilistic analysis is the Sparse Polynomial Chaos Expansion (SPCE). The failure probability and the safety factor distribution of the considered dam under normal exploitation conditions and seismic loading conditions are presented. The influence of the random variable distribution type on the probabilistic results is studied. In addition, a PCE-based Global Sensitivity Analysis (GSA) is embedded in order to investigate the contribution of each variable to the safety factor variance.