Algorithms for Bayesian network modeling and reliability inference of complex multistate systems with common cause failure

In constructing the Bayesian network (BN) reliability model, too many components will make the memory storage requirements of the conditional probability table (CPT) exceed the computer random access memory (RAM), especially for the complex multistate system with common cause failure (CCF). However, the existing methods cannot solve the BN modeling's large memory storage requirements problem of the complex multistate system with CCF. Thus, this paper proposes a BN block to process the nodes with CCF, converting CPT to a super multistate node's joint probability table, based on which a multistate BN compression modeling algorithm under CCF is proposed to reduce the memory storage requirements of BN reliability modeling. By deriving the intermediate inference factor constructing rules, this paper proposes a multistate BN compression inference algorithm under CCF to perform the compressed BN reliability inference. Finally, two engineering cases validate the proposed algorithms' performance. The results show that the proposed algorithms can significantly decrease the BN modeling's memory storage requirements and accurately analyze the reliability of the complex multistate system with CCF.

Automated summary

This paper proposes a BN block to process nodes with common cause failure (CCF) and presents an algorithm to reduce the memory storage requirements of BN reliability modeling. By deriving the intermediate inference factor constructing rules, a multistate BN compression inference algorithm under CCF is proposed. Two engineering cases are used to validate the performance of the proposed algorithms. The results show that the algorithms can significantly decrease the memory storage requirements of BN modeling and accurately analyze the reliability of complex multistate systems with CCF.