Important Sampling for Structural Reliability Based on Radial Basis Function Neural Network

摘要

The methods of the structural reliability mainly involve analytical approximate reliability index or numerical simulation, which using the finite element solver is time-consuming and large computation. Important sampling (IS) for structural reliability analysis based on radial basis functions neural network (RBFNN) is proposed in the paper, in which trained RBFNN can model the implicit function between the structure response and input random variables. And limit state function of structure is simulated with RBFNN model applied to calculate the design point. The results show that the RBFNN can simulate the limit state functions of structures. Besides, calculation procedure based on finite element solver for structural analysis is greatly reduced and the efficiency in structural reliability evaluation is improved.