An information reuse-based method for reliability updating

Pei-Pei Li,Yi Zhang,Yan-Gang Zhao,Zhao Zhao,Enjian Cai

DOI: https://doi.org/10.1016/j.ress.2023.109536

2023-08-08

Abstract:Reliability updating of structures is computationally demanding, especially when the likelihood function involves inverse analysis and needs to be updated multiple times. In this study, a novel efficient approach for updating the failure probability based on the method of moments and information reuse is proposed. Instead of updating the failure probability using newly obtained posterior distributions, the proposed method transforms the estimate of posterior failure probability into a computation of the reliability index based on the posterior moments of the limit-state function at each update. When calculating these posterior moments, the terms containing the likelihood function are separated from the limit-state function analysis and then the bivariate dimension-reduction method (BDRM) in combination with the point estimate method (PEM) [1,2] is used to conduct the calculation. In this way, the results evaluated in the first updating time can be reused to update the probability of failure in the subsequent updates, avoiding the need to reconduct limit-state function analysis and inverse analysis using the posterior samples, thus improving computational efficiency. Four numerical examples are presented to validate the efficiency and accuracy of the proposed method, and the results indicate that the proposed method has obvious computational advantages for multiple reliability updating problems without loss of accuracy.

engineering, industrial,operations research & management science