Reliability analysis of seismic subsidence of loess foundation based on Kriging surrogate model

In order to evaluate the safety level of seismic subsidence of loess foundation during earthquakes, this paper proposes a reliability analysis method based on the Kriging surrogate model. The sample points of random variables are extracted based on Latin hypercube sampling (LHS). A numerical model of the loess foundation is established using FLAC^3D finite difference software, and response values corresponding to the sample point is calculated as the training sample point for establishing Kriging surrogate model. The reliability of seismic subsidence of loess foundation is calculated by Monte Carlo method, and the influence of variability of random variables on seismic subsidence reliability is analyzed. The results show that the proposed seismic subsidence reliability calculation method can reasonably consider the variability of soil parameters compared to traditional deterministic methods, making it more practical with better reliability and applicability. The variation coefficient of cohesion significantly affects the maximum reliability of seismic subsidence.