Prioritized experience replay-based adaptive hybrid method for aerospace structural reliability analysis

Reliability assessment of aerospace structures is crucial for ensuring operational safety and preventing catastrophic failures. However, conventional reliability methods face significant challenges in adapting to diverse scenarios and capturing complex failure mechanisms under extreme conditions. To address the dual challenges of single-model dependency and modeling inaccuracy in such scenarios, the Prioritized Experience Replay-based Adaptive Hybrid Method (PER-AHM) is proposed by integrating PER-based hybrid optimization framework (PER-HOF) and adaptive modeling method (AMM). Specifically, PER-HOF is developed to efficiently obtain optimal parameters of surrogate models, by combining prioritized experience pool, dynamic update mechanism and trial switching strategy; AMM is designed to select the most suitable model for specific problems, via considering multiple metrics. The significance of the PER-AHM lies in its ability to select models in a modular manner, share solving experience and enhance the accuracy and efficiency of aerospace structural reliability analysis. To illustrate the efficiency of the developed method, two examples are used, including reliability analysis of two-dimensional nonlinear system and aeroengine high-pressure turbine radial deformation, which demonstrate that PER-AHM outperforms traditional methods in prediction accuracy and reliability assessment precision. This research provides an accuracy and efficient solution for aerospace structural reliability analysis.