Abstract: By using the advantages of uncertainty reasoning and figurative expression of Bayesian network, a new method of modeling and assessment of multi-state system reliability based on BN is proposed to determine the nodes of BN and the multiple states of elements of system, and to give the probability of each state and then utilizing conditional probability distributing (CPD) to describe the relationship among the element states, so as to express the states of correlated nodes and build a BN model of multi-state system. The model can clearly express the multiple states of system and elements and the state probability, and also can directly calculate the system reliability on the basis of multiple state probabilities of elements, thereby carrying out qualitative analysis and quantitative assessment of multi-state system reliability. Analysis of practical examples proves the effectiveness of assessment of multi-state system reliability by using BN method.

Key words: Assessment, Bayesian networks, Modeling, Multi-state, System reliability