AMRBF-SS: Subset Simulation with Active Learning and Multiple Kernels Radial Basis Function for Small Failure Probability Prediction

Abstract

Combining surrogate models with simulation methods is an effective way to address failure probability problems involving time-consuming computational models in structural reliability analysis. The radial basis function (RBF) has been widely used in the context of uncertainty quantification owing to its flexibility, nonlinearity, computational efficiency, and ability to handle high-dimensional data. Multiple RBF kernel functions are integrated in this study with subset simulation (SS) to formulate the proposed AMRBF-SS method to efficiently solve the problems of small failures. The local uncertainty of the prediction is estimated and integrated into to formulate an active learning function. Various numerical and practical examples are considered to verify the accuracy and efficiency of the proposed method. The results show that the proposed AMRBF-SS method provides an effective and efficient technique that can solve high-dimensional small-probability problems with similar accuracy levels but fewer evaluation times than other existing methods.