失效模式重要性高效分析的序列元模型 重要抽样法

Reliability issues in complex systems with multiple failure modes are addressed through failure mode importance analysis. An efficient algorithm for failure mode importance analysis is proposed based on sequential metamodel-based importance sampling. The relationship between key factors of the probability classification function for constructing importance sampling probability density functions in multi-failure-mode reliability analysis and surrogate models of all failure modes is investigated. A surrogate model update strategy is developed, enabling successful application of metamodel-based importance sampling to estimate reliability in multi-failure-mode scenarios by identifying the most critical failure mode. The relationship between failure mode importance indices, multi-failure-mode reliability and single-failure-mode reliability is analyzed. A sequential metamodel-based importance sampling algorithm is designed to share training samples between multi-failure-mode and single-failure-mode reliability analyses. By sequentially updating surrogate models for each failure mode, the importance indices are efficiently evaluated. Numerical examples and finite-element-based turbine blade are analysed to validate the proposed method. Results confirm its efficiency and accuracy in evaluating failure mode importance indices, demonstrating superior performance compared to existing approaches.