Seismic collaborative reliability evaluation of slopes using subset simulation via support vector machine

Abstract

The computational effort places a limit on the low failure probability estimate of geotechnical engineering exposed to seismic stimulation. In this research, a novel method is proposed for the seismic reliability assessment of complicated geotechnical engineering slope structures with low failure probability. First, an improved subset simulation (SS) approach is provided for the low-failure events with high-dimensional indirect cross-correlation parameters. This method expands the usefulness of SS by using the support vector machine (SVM) method. Then, in order to reduce computing effort, the seismic collaborative reliability evaluation (SCRE) framework for seismic reliability assessment based on SS and the response conditioning method (RCM) is constructed. Finally, the accuracy of SCRE is verified by comparison with the standard Monte Carlo simulation based on 2D slope. The results show that SCRE's computational effort is approximately 2% that of Monte Carlo simulation. Meanwhile, SVM does a good job of preserving the significant indirect cross-correlation that occurs in the subspace samples that were split by SS. The time-saving benefit of SCRE is further enhanced when the aforementioned situations are expanded to 3D, and the findings demonstrate that the failure probability and sliding surface of the slope are different in 2D and 3D.