An adaptive method based on PC-Kriging for system reliability analysis of truss structures

Abstract

In practice, a truss consists of a large number of members which makes it a complex system. This leads to difficulties to estimate the system reliability due to computational costs. An adaptive method is thereby proposed to deal with this issue. It constructs a global metamodel to quickly estimate the rough reliability index of a truss. According to the estimated reliability index, the differential evolution algorithm is performed to generate more samples located in an expanded domain so that more representative failure modes can be identified. Combined with AK-SYSi, local metamodels of representative failure mods are built, and updated through active learning. When the convergence criterion is satisfied, the results of system reliability analysis can be obtained. Eventually, two examples of truss structures are studied to illustrate the superiority of the proposed method in balancing accuracy and efficiency. The results indicate that the proposed method makes a good balance between accuracy and efficiency when it is applied to analyze the system reliability of the truss.