A Novel Quantile-Based Sequential Optimization and Reliability Assessment Method for Safety Life Analysis

Xia Jiang,Zhenzhou Lu
DOI: https://doi.org/10.1016/j.ress.2023.109810
IF: 7.247
2024-01-01
Reliability Engineering & System Safety
Abstract:Safety life analysis under random uncertainty is an effective tool for guiding the maintenance and design of structures. However, existing methods for safety life analysis employ a nested solution strategy by dichotomy, with inner time-dependent reliability analysis, resulting in high computational costs. This paper proposes a quantile-based sequential optimization and reliability assessment (QSORA) method to overcome the shortcoming of nested solution strategy in safety life analysis. In QSORA, the probability constraint is first equivalently transformed into a quantile constraint corresponding to the target time-dependent failure probability (TDFP). Then, the coupling relationship between the search for safety life and the quantile estimation is then eliminated, and the safety life is solved iteratively through a series of cycles. Each cycle contains two independent executions: an equivalent deterministic optimization and quantile estimation of extreme performance function corresponding to the target TDFP. Finally, a sampling-based method combined with a double-loop Kriging model is proposed to efficiently estimate the quantile, and it is embedded into QSORA to improve the efficiency of safety life analysis. Two learning functions are adopted to ensure the accuracy of double-loop Kriging model. Numerical and engineering examples verify the efficiency and accuracy of the proposed method for safety life analysis.
What problem does this paper attempt to address?