Competing Failure Analysis in Sequence-Dependent Systems

Abstract

Many real-world fault-tolerant systems exhibit sequence-dependent failure behavior, which takes place when the system failure depends not only on combinations of its component failure events but also on the sequence in which the component failures occur. Such a behavior has been modeled using the priority-AND (pAND) gate in the dynamic fault tree, which requires extensions of the traditional static fault tree analysis methods for system reliability evaluation. Further complicating the system reliability analysis is the competing failure behavior due to function dependence existing among the system components. Specifically, time-domain competitions exist between the occurrence of propagated failures originating from dependent components and the occurrence of trigger component failure. Different occurrence orders can lead to dramatically different system statuses. Moreover, the sequence-dependent behavior and the competing failure behavior can be correlated, and thus should be handled together for accurate system reliability modeling and analysis. This paper proposes a combinatorial reliability analysis method that can address both sequence-dependent and competing failures as well as their correlations. The proposed method has no limitation on the types of time-to-failure distributions for system components. Advantages and applications of the proposed method are illustrated through a detailed case study on an example of a computer network system.