Reliability Modeling of Redundant Systems Considering CCF Based on DBN

Abstract

Common cause failure (CCF) has become a hot topic in reliability and availability analysis of redundant systems. Aiming at the shortage of $$\beta $$ -factor model in distinguishing three or more failures, multiple error shock theory is put forward. Having the problem in determining minimal cut sets and structure functions for dynamic fault tree (DFT), dynamic Bayesian network (DBN) is applied to transit DFT events to corresponding nodes and express causal relations between the nodes. On the base of explicit modeling for CCF, DBN models for hot spare (HSP) gate, cold spare (CSP) gate and warm spare (WSP) gate are established considering CCF processes. For a HSP gate, all failure processes are listed in the stress event layer for each component. For a CSP gate and a WSP gate, CCF node and intermediate nodes are introduced to express causal relations. At last, a control unit and its improved type are taken as examples. The DBN models are built by referring to corresponding DFT structures when taking CCFs into consideration. From the results, it is obvious that the modern unit has a relative higher reliability and availability. And it is less easily influenced by uncertainty such as failure rates and coverage factor through sensitivity analysis.