Sampling-based adaptive design strategy for failure probability estimation
发布时间：2024-10-25
点击次数：
影响因子：8.1
DOI码：10.1016/j.ress.2023.109664
发表刊物：Reliability Engineering and System Safety
关键字：Adaptive design of experiment Failure probability Normalizing flows
摘要：Failure probability (FP) estimation problem is a crucial task in engineering. In this work we consider this problem in the situation that the underlying computer models are extremely expensive, which often arises in the practice, and in this setting, reducing the calls of computer model is of essential importance. We formulate the problem of estimating the failure probability with expensive computer models as an sequential experimental design for the limit state (i.e., the failure boundary) and propose a series of efficient adaptive design criteria to solve the design of experiment (DOE). Considering the remarkable achievements of neural networks, we aim to leverage this powerful tool for surrogate modeling and sampling purposes. In particular, the proposed method employs the deep neural network (DNN) as the surrogate of limit state function for efficiently reducing the calls of expensive computer experiment. A map from the Gaussian distribution to the posterior approximation of the limit state is learned by the normalizing flows for the ease of experimental design. Three normalizingflows-based design criteria are proposed in this work for deciding the design locations based on the different assumption of generalization error. The accuracy and performance of the proposed method is demonstrated by both theory and practical examples. The relative error of FP estimation achieved by the proposed methods is consistently below ten percent.
合写作者：Hongji Wang, Jinglai Li
第一作者：Tiexin Guo
论文类型：期刊论文
通讯作者：Hongqiao Wang
论文编号：109664
学科门类：理学
一级学科：数学
文献类型：J
卷号：241
ISSN号：0951-8320
是否译文：否
发表时间：2023-09-19
收录刊物：SCI
发布期刊链接：https://linkinghub.elsevier.com/retrieve/pii/S0951832023005781