Fast Analysis of Structural Reliability Using Gaussian Process Classification Based Dynamic Response Surface Method

The performance function of large-scale complicated engineering structure is always highly nonlinear and implicit, and its reliability needs to be evaluated through a time-consuming Finite Element method (FEM). A new method, Gaussian process classification (GPC) dynamic response surface based on Monte Carlo Simulation (MCS) was proposed. Small training samples were created using FEM and Markov chain. Then, the most probable point (MPP) is predicted quickly using MCS without any extra FEM analysis. Furthermore, an iterative algorithm is presented to reduce the errors of GPC by using information of MPP to improve the reconstructing precision constantly. Then, Monte Carlo method combined with GPC surface is applied to get the probability of failure. Several examples results demonstrate the efficiency and robustness of the proposed method, compared with the results of common reliability methods.