Diagnostic expert systems from dynamic fault trees

A methodology for developing a diagnostic map for systems that can be analyzed via a dynamic fault tree is proposed in this paper. This paper shows how to automatically design a diagnostic decision tree from a dynamic fault tree used for reliability analysis. In particular the methodology makes use of Markov chains since they are mathematical models used for reliability analysis. We use approximate sensitivity as an intermediary step to obtain the Vesely-Fussell measure from the Markov chain. We used the Vesely-Fussell measure of importance as the corner stone of our methodology, because it provides an accurate measure of components' relevance from a diagnosis perspective. The outcome of this paper is a diagnostic decision tree, which was generated for a real dynamic system. The diagnostic decision tree produced is a map that can be used by repair and maintenance crew to diagnose a system without having previous knowledge or experience about the diagnosed system. The methodology we develop is capable of producing diagnostic decision trees that reduces the number of tests or checks required for a systems diagnosis.