Persistence of relevance: A missing issue in imperfect coverage models

Abstract

In traditional imperfect fault coverage models, simply coverage models (CMs), the coverage (including identification and isolation) is typically limited to the faulty components regardless of their relevance. The relevance is typically defined in the context of perfect fault coverage, and a component is called irrelevant if its (covered) failure does not affect the system state, otherwise it is relevant. Although it is generally assumed that all components are initially relevant in these models, such an assumption does not consider the fact that an initially relevant component could become irrelevant afterwards due to the failures of other components, and we call it a non-persistent component. A system with only persistent components is called persistent, otherwise it is called non-persistent. For a non-persistent system, it is important to cover (identify and isolate) the non-persistent components in time whenever they become irrelevant, such that their future uncovered failures will not affect the system anymore. This paper formalizes the concept of persistence and analyzes its impact on reliability of the systems subjected to imperfect fault coverage. It is demonstrated that with the coverage of irrelevant components (in addition to the faulty components), the reliability of a non-persistent system can be (significantly) improved without increasing redundancy.