System Reliability Analysis using Multivariate Statistical Modeling

초록

Reliability analysis is the evaluation of the probability to satisfy a design constraint. Conventional reliability analysis methods that have been developed assume that all responses are independent and calculates the reliability for each response. This can lead to a significant error in the reliability of the entire system. Therefore, system reliability analysis is required to consider the joint probability of the responses and accurately evaluate the reliability of the system. The calculation of the system reliability usually entails multi-dimensional integration that is extremely difficult and numerically expensive. To resolve this issue, several system reliability analysis methods have been developed. Although Monte Carlo simulation method has been widely used, it still requires a high computational cost [1]. Ditlevsen’s second-order upper bound combined with MPP-based dimension reduction method and system reliability analysis using the first-order reliability method assuming independence between failure modes have also been proposed [2-3]. However, these methods neglect the joint probability of failure of some responses or assumes that the responses are independent. This study proposes a system reliability analysis method using a multivariate statistical model. The main advantage of this method is that the system reliability is directly evaluated by a multivariate statistical model without any assumptions or approximations.