An Importance Sampling Based Approach for Reliability Analysis

In this paper, an importance sampling based approach for reliability analysis is proposed. The fundamental of this approach is to bias the realization of random variables around the most probable point (MPP) such that the number of simulations can be reduced significantly. Compared to the basic Monte Carlo simulation (MCS), the proposed approach requires less computational effort since it only evaluates the system performance functions at the reduced probability space. Two comparison experiments are conducted at the end of the paper. One is used to demonstrate the proposed method improves the efficiency comparing with basic MCS without losing accuracy. The second one is used to illustrate the proposed method generates more accurate results than that of FORM (first order reliability method).