Adaptive surrogate-assisted sampling pool reduction strategy for low failure probability estimation

To improve the efficiency and accuracy of rare-event reliability analysis of complex structures, an advanced adaptive kriging-based candidate sample reduction (AK-CSR) method is proposed by integrating the CSR strategy and the advanced AK method. Through the improved first-order reliability method and the updated kriging model (KM), the accurate most probable failure point can be obtained with KM updating. By domain constraint and distance constraint functions, the CSR strategy can ceaselessly find desired samples to update the KM. The proposed method was verified using three numerical examples and two engineering examples. The results demonstrated that the AK-CSR method can be used to perform rare-event reliability analysis of complex structures and improve computational efficiency while maintaining good accuracy. Moreover, this study offers a useful insight into reliability-based design optimisation of complex structures and enriches the field of structural reliability theory.