A New Assessment Method for System Reliability Based on Dynamic Fault Tree

Abstract

According to the deficiency of traditional Markov chain approach in dynamic fault tree analysis, a new modular method for system reliability analysis is proposed. This paper focuses on dividing the fault tree of system into independent subtrees using a linear-time algorithm, and the processing method for different subtrees: Binary decision diagram solution for static subtrees and Bayesian Network solution for dynamic subtrees, respectively. In addition, an approach is also provided for mapping some dynamic logic gates into discrete-time Bayesian network. At last, the modular method has been applied to assess the reliability of a satellite key device and the results have shown that the proposed method can overcome the state explosion problem, and is useful for assessing the reliability of large and complex systems.