Copula-Based Matrix System Reliability for Reliability Assessment of General Systems

Abstract

This paper proposes a novel system reliability analysis method as an alternative to matrix-based system reliability (MSR). Traditional MSR requires the fitting of a Dunnett-Sobel (DS) class correlation matrix to the general correlation coefficient matrix to account for the correlation between the components in the presence of a common source of hazard. This issue significantly limits the applicability of the MSR, especially when a large number of common source random variables (CSRVs) are involved. The novel method proposed in this research, called copula-based matrix system reliability (CMSR) eliminates the need for DS class correlation fitting, which results in a faster, generally applicable, and more accurate MSR. The efficiency and accuracy of the proposed method are evaluated through several work examples.