System reliability analysis of landslides involving sliding surfaces using intelligent multiple response surfaces

Abstract

The shear strength reduction method provides an effective tool of numerical analysis for landslides reliability analysis. However, it ignores the failure probability of the secondary failure surfaces and requires huge computational cost. To avoid these common criticisms, an intelligent multiple response surfaces method for system reliability using multiple response-surface method (MRSM) and least-squares support vector machine (LSSVM) is presented to evaluate the stability of complex multistage historic landslides with multiple sliding surfaces. Deterministic analysis of each sliding surface is first performed using the finite element method of sliding surface stress analysis, which is applied to obtain the safety factors of different sliding surfaces from the stress fields generated by finite element simulations. The LSSVM model with excellent nonlinear fitting ability is then employed to construct the multiple response-surface method (MRSM) of the sliding surfaces and a genetic algorithm (GA) is adopted to optimize the LSSVM. This proposed methodology is finally applied to investigate the probability of system failure of the Zhenggang landslide in southwestern China. The results indicate that the proposed approach can reduce the computational cost of finite element analysis in direct Monte Carlo simulation (MCS) by proper training using a limited of samples, and the calculation accuracy meets the engineering requirements of complex multistage historic landslides.