Reliability analysis of complex uncertainty multi-state system based on Bayesian network

In modern engineering, multi-state system (MSS) is a kind of system that represents a capability allowing for more than two performance states in a system besides perfect functionality and complete fault [21]. Compared with the two-state system, MSS can define the components states of a system, and express the effect of the changes of component performance on system performance more flexibly and precisely. In the 1970s, Barlow and Wu [2] first proposed the concept of MSS and gradually established the related theory. Then, the reliability theory of MSS has been widely concerned by scholars. And the following reliability analysis methods for MSS have been developed: the extended Boolean model method [22, 26], random process theory [1, 14, 18], Monte-Carlo simulation method [23, 25], function model method [8, 16, 31], Bayesian network method [13, 29], and so on. The uncertainty, which is caused by the insufficient information about internal structures, the scarcity of historical data and the changeability of operation environment, is one of the most crucial problems in MSS reliability analysis. Therefore, it is very difficult to define and obtain the component state performances and state probabilities. Meantime, the boundaries among component fault states fail to define and obtain with precision. So the traditional probability-based method is no longer applicable. However, non-probabilistic methods, such as evidence theory [7], grey system theory [33], probability-box [27], and fuzzy theory [15, 30], have been proposed and developed for reliability analysis of complex uncertainty MSS. WANG H, DUAN F, MA J. Reliability analysis of complex uncertainty multi-state system based on Bayesian network. Eksploatacja i Niezawodnosc – Maintenance and Reliability 2019; 21 (3): 419–429, http://dx.doi.org/10.17531/ein.2019.3.8.