A single-loop reliability analysis strategy for time-dependent problems with small failure probability

• Published in: Reliability engineering & system safety Vol. 219; p. 108230
• Main Authors: Cao, Runan, Sun, Zhili, Wang, Jian, Guo, Fanyi
• Format: Journal Article
• Language: English
• Published: Barking Elsevier Ltd 01.03.2022; Elsevier BV
• Subjects: Accuracy; Criteria; Failure analysis; Importance sampling; Kernel-density estimation; Reliability analysis; Reliability engineering; Small failure probability event; The Kriging model; Time dependence; Time-dependent reliability; The Kriging model; Kernel-density estimation; Small failure probability event; Time-dependent reliability; Importance sampling
• ISSN: 0951-8320; 1879-0836
• DOI: 10.1016/j.ress.2021.108230

•An importance sampling strategy for time-dependent reliability is proposed.•A new criterion of selecting the best training point is proposed.•A new stopping criterion is proposed. In recent years, many time-dependent reliability methods have been proposed. However, these kinds of methods are barely applied in the field of small failure probability events. This paper proposes an efficient time-dependent reliability method based on the Kriging model and the importance sampling (IS) method. The new method is a single-loop strategy, which can obtain the failure probability varying with time. For complex reliability problems, the failure region usually includes multiple sub-regions, which is challenging to deal with though the traditional IS methods (such as the first-order reliability method-based IS method (FORM-IS) and the kernel-density-estimation-based IS method (KDE-IS)). In this paper, we improve the KDE-IS method to ensure that the sample points cover all the failure sub-regions as far as possible. Aiming at selecting a sample point that can improve the accuracy of failure probability effectively to refresh the Kriging model, this paper proposes a new criterion for selecting the best training point. For accuracy, a new stopping criterion is also defined. Finally, the efficiency and accuracy of the new method are verified by four examples.