Global and local Kriging limit state approximation for time-dependent reliability-based design optimization through wrong-classification probability. (April 2021)

• Record Type: Journal Article
• Authors: Jiang, Chen; Yan, Yifang; Wang, Dapeng; Qiu, Haobo; Gao, Liang
• Is Part Of: Reliability engineering & system safety. Volume 208(2021)
• Publication Date: 2021-04
• Subjects: Time-dependent reliability-based design optimization -- Adaptive Kriging modeling -- Wrong classification probability -- False classification rate -- Estimation error of failure probability; Reliability (Engineering) -- Periodicals; System safety -- Periodicals; Industrial safety -- Periodicals; Fiabilité -- Périodiques; Sécurité des systèmes -- Périodiques; Sécurité du travail -- Périodiques; 620.00452
• DOI: 10.1016/j.ress.2021.107431
• Languages: English
• ISSNs: 0951-8320

Abstract

Highlights: The time-dependent wrong-classification probability is developed. Global and local Kriging modeling methods are used for limit state approximation. False classification rate is proposed to measure the global accuracy. Estimation error of failure probability is derived to measure the local accuracy. Abstract: Time-dependent reliability-based design optimization is an effective tool to guarantee a high reliability of the product during the full life cycle. However, the necessarily repeated probabilistic constraint evaluations bring big computational burden when this tool is applied to the complex engineering systems. To reduce the computational cost, this work employs Kriging model to approximate the limit states of time-consuming probabilistic constraints, and proposes the global and local Kriging modeling methods respectively based on the wrong-classification probability. The global one aims to reduce the wrong-classification probability in the vicinity of the whole limit states, while the local one focuses on the limits states that are potentially visited by the optimum. Based on the wrong-classification probability, two indices, i.e. false classification rate and estimation error of failure probability, are derived to measure the global and local accuracies of limit states respectively. For the global or local Kriging modeling, the approximated Kriging constraint maximizing the false classification rate or the estimation error of failure probability will be updated by the point with the maximum wrong-classification probability. Results of four case studies demonstrate the efficacy of the proposed methods.