Abstract
Traditional most probable point (MPP) reliability analysis using sensitivity information to find the MPP is difficult in vehicle reliability-based design optimization. In this paper, an adaptive surrogate model using Bayesian metric developed in previous work is used to represent the true performance functions and replace the true limit state function. The score function with reweighting scheme is exploited to compute the sensitivities of probabilistic responses with respect to the design variables, which are the mean values of the random variables. Numerical results indicate that the proposed methods can produce the best surrogate model and estimate the sensitivities of probabilistic responses accurately. The proposed methodology is demonstrated by a vehicle reliability-based design optimization problem with full frontal and offset frontal impacts.
Keywords Surrogate Model, Reliability-based Design Optimization, Score Function, Reweighting Scheme