open access

https://hdl.handle.net/2115/100477

An improved single-loop importance sampling method based on progressive hierarchical strategy for generalized failure probability function under high-dimensional and high-reliability applications

英語

Importance Sampling Method

2026-07-01

Proceedings of the 10th International Symposium on Reliability Engineering and Risk Management (ISRERM 2026)

For high-dimensional and high-reliability aerospace applications, accurately estimating how design-parameter variations influence structural reliability levels under probabilistic inputs and fuzzy states assumption is of paramount importance, specifically, the generalized failure-probability function. Therefore, this paper proposes an improved single-loop importance-sampling method based on a progressive hierarchical technique for efficiently estimating the generalized failure probability function. The key novelties include: first, establishing an analytical expression for the unified importance-sampling density that enables rapid sample generation; second, designing a progressive hierarchical technique that adaptively updates the importance-sampling center using samples with high contributions to the generalized failure probability estimation, ensuring accurate targeting of critical high-contribution regions. By circumventing computationally expensive optimization procedures through analytical construction and progressive stratification, the proposed approach substantially improves computational efficiency in high-dimensional and high-reliability scenarios while guaranteeing estimation accuracy. Numerical example demonstrate that the proposed method outperforms existing single-loop importance-sampling methods in sampling efficiency, estimation stability, and high-dimensional adaptability.

会議発表論文

Version of Record