Slope reliability analysis using Kriging-based Subset simulation

Abstract:

The Kriging method， which is an efficient and accurate metamodel， is widely used in slope reliability analysis. However， traditional methods couple the Kriging model directly with the Monte Carlo simulation method for reliability analysis， which leads to excessive memory usage or even overflow in high-dimensional slope reliability calculation with small failure probability， hence failure to find the solution. To this end， this paper proposes a Subset simulation method based on the Kriging metamodel to efficiently solve the problem of small probability slope reliability analysis. A single-layer cohesive soil slope and a practical soil slope are used to verify the effectiveness of the proposed method， and different regression models and related function models as well as the number of training samples are explored for the accuracy of the method. The results show that： （1）The proposed Subset simulation method based on the Kriging metamodel can effectively calculate the failure probability of slopes， and is more efficient than the traditional method； （2）During the construction of the Kriging model， the calculation accuracy of the model can be achieved when the number of training samples reaches10 times that of random variables. In addition， the number of additional training samples has little effect on the calculation results.