System Reliability Analysis of Soil Slopes Using an Advanced Kriging Metamodel and Quasi–Monte Carlo Simulation

Abstract

Enhancing the efficiency and accuracy of the system reliability analysis of soil slopes is of great importance in engineering design. An advanced kriging metamodel integrated with the quasi–Monte Carlo simulation (AKQMCS) is proposed for such a purpose in this paper. The proposed advanced kriging model is used as a surrogate for the limit state function (LSF) of a slope, and it is established by a sequential design of experiments, which updates the kriging model step by step, with the help of a learning function based on the entropy theory. QMCS simulation is then performed on the advanced kriging metamodel to evaluate the system failure probability of the slope. Finally, three soil slope examples, including a two-layered cohesive slope, a three-layered c-φ slope, and a single-layered sand slope, are examined to verify the capability and validity of the proposed approach. The results indicate that the proposed approach can provide a reasonable and accurate estimation of the system failure probability of slope stability with a significantly reduced number of deterministic stability analyses compared with the ordinary kriging method and the direct Monte Carlo simulation (MCS). The QMCS well outperforms the MCS in efficiency due to the low-discrepancy sequences used in the QMCS. Hence, the proposed AKQMCS provides a new and efficient tool for the system reliability analysis of soil slopes.