Progressive Importance Sampling Approach for Reliability Analysis

Abstract

This paper introduces a progressive importance sampling (PIS) approach for reliability analysis. This new approach, an extension of the existing adaptive importance sampling (AIS) method, combines the MPP-based method, the importance sampling technique, and directional search algorithm in the standard normal variable space. Starting from the linear limit-state surface normal to the MPP, the sampling domain is expanded using a series of linear surfaces tangent to the limit-state surface to attempt to envelop the failure domain. Two examples of highly nonlinear limit state functions are presented to demonstrate the approach and the accuracy and robustness of the approach are discussed.