Weighted low-discrepancy samplings: A novel method for slope system reliability analysis

Abstract

This paper presents an efficient method for system reliability analysis of geotechnical examples based on the low-discrepancy sequences (LDS) sampling technique in combination with probabilistic weighting of the samples, i.e., the weighted low-discrepancy samplings (WLDS). In the framework of quasi-Monte Carlo simulation (QMCS), the Sobol’s quasi-random LDS are employed to generate the basic random variable samples that can significantly improve the efficiency of probabilistic weighting for the probability of failure calculation. In particular, the most probable system failure points (MPPs) can be directly captured according to the probabilistic weighting calculation. Via Nataf transformation, the WLDS computational framework incorporating correlated random variables in soils is also illustrated. Focusing on geotechnical applications, the reliability analyses of example soil slopes are presented to verify the performance of the proposed method. It is found that the proposed WLDS method not only can accurately obtain the probability of system failure but also can capture the multi-MPPs defining different earth slope failure modes in a cost-effective manner. The results in this study demonstrate that the proposed WLDS method is a feasible and promising method for performing the system reliability analysis of complex geotechnical earth structures.