Application of Metaheulistic Method to Reliability Analysis for Lifeline Network Involves Multiple Failure Modes
Ken ISHIBASHI, Hitoshi FURUTA, Yasutoshi NOMURA,Koichiro NAKATSU and Kyosuke TAKAHASH
Abstract:The reliability analysis is necessary to maintain the safety of structure. In order to evaluate the soundness of structure quantitatively, the calculation of failure probability is one of useful measures. In the reliability analysis for large-scale structures, the enhancement of calculation accuracy and the estimation of factors for the failure require the failure probability calculated by considering the various failure modes. However, the sampling method like Monte Carlo simulation is difficult to calculate the failure probability and estimate multiple failure modes efficiently. This is because the search of various failure modes increases the calculation cost significantly. Therefore, in order to overcome the trade-off, this paper attempts to propose an efficient method by applying Metaheuristic methods. Metaheuristic methods are characterized by the high search ability with keeping the diversity among solution candidates. Firstly, the search of failure modes is formulated as the combinatorial optimization. In this optimization, the metaheuristic method searches failure modes in consideration of the occurrence probability of minimal cut set. In this way, various failure modes that have high occurrence probability are obtained efficiently. Secondly, Probabilistic Network Evaluation Techniques (PNET) is applied to calculate the failure probability of network in consideration of the correlation among obtained failure modes. Through these processes, the proposed method can calculate the accurate failure probability efficiently. Several numerical experiments are presented to demonstrate the applicability of the proposed method for the reliability analysis of large-scale networks. Key Words:Reliability Analysis, Lifeline network, Failure modes, Metaheuristic, Minimul cut set, Probabilistic network evaluation techniques, Markov chain monte carlo