AK-SYSC: A Kriging-Based Approach for Multi-failure-Mode Structural Systems Reliability Analysis

In engineering structures, many systems are subject to multiple failure modes, making the accurate and efficient analysis of system reliability a significant challenge. Among the existing estimation methods, AK-SYS and AK-SYSi approaches based on Kriging surrogate model have shown competitive performance. However, the high false identification rate associated with AK-SYS method and the omission of minimum or maximum mode information in AK-SYSi method limit both accuracy and efficiency. To enhance estimation accuracy and computational efficiency, this study proposes a novel method, referred to as AK-SYSC method, which incorporates a refined learning function Uc. AK-SYSC method addresses the limitations of AK-SYS approach by mitigating the risk of false identification of extreme modes caused by non-convergent Kriging models, which can otherwise lead to reduced estimation efficiency and inaccurate predictions of system failure probability. Furthermore, by fully utilizing the minimum or maximum value information neglected in AK-SYSi method, the proposed approach enhances the identification speed of the system limit state surface, thereby further improving computational efficiency. Three numerical examples serve to validate the performance and exactness of the proposed AK-SYSC method.