An approach for reliability-based sensitivity analysis based on saddlepoint approximation

Reliability-based sensitivity analysis quantifies the effects of input random variables on model outputs. It is an integral part of reliability-based design and robust optimal design. An efficient reliability sensitivity method is developed in this article by combining the stochastic perturbation technique and the moment-based saddlepoint approximation. The main advantage of the presented method is that only the first three moments of the structural response are needed and the cumulant generating functions of input random variables are not strictly required. Several examples are illustrated, and the analysis procedure shows that the method is convenient to be applied to parameter sensitivity analysis of structural reliability.