A new performance measure approach with an adaptive step length selection method hybridized with decoupled reliability-based design optimization

Abstract

Reliability-based design optimization (RBDO) using first-order reliability methods (FORM) is widely utilized due to its simplicity and efficiency. There are two principal ingredients in the performance of RBDO approaches. One is related to the sequence of optimization and reliability loops, the other one is associated with the reliability assessment itself since it is usually computationally costly. The aim of this study is to improve the efficiency of RBDO while remaining its accuracy at a high level. To do so, sequential optimization and reliability analysis (SORA) is adopted. Moreover, the main focus of this study is on developing a fast and accurate method to conduct reliability analysis thereby, enhancing the efficiency of RBDO. The performance measure approach (PMA) of FORM is known to be very efficient for this. Thus, a new PMA approach based on an adaptive step length which is selected based on the normalized location and negative gradient vectors in two successive iterative points is presented herein. The efficiency and robustness of the proposed method are tested with six inverse reliability problems and three RBDO examples. Their performances are compared with some recently developed methods. The results confirm that the proposed method is accurate with more efficiency and competitive convergence rate.