A fast model-based diagnosis engine

In this paper we present a novel fast model-based diagnosis engine. Our novel engine is based on a two-step approach to diagnosis, i.e., off-line system analysis and on-line diagnosis. The efficiency of our novel method results from the fact that, by performing a detailed analysis of the target system, it drastically reduces the amount of computation needed for diagnosis. In particular, our new algorithm relies on the concept and use of minimal set of ARRs to achieve a much better efficiency in the diagnosis process. Our novel diagnosis engine is based on our two recent results. First, it uses our recently developed method for generation of the complete set of ARRs. Second, it uses the minimal set of ARRs; as we have recently shown that for any given number of faults, i.e., single, double, triple, etc., there is a corresponding minimal set of ARRs which is usually significantly smaller than the complete set of ARRs. We present and discuss the performance of our diagnosis engine by its application to several examples. We show that, even by using a non-exoneration assumption, we achieve a much better efficiency over the GDE as well as full ARR-based approaches for model-based diagnosis.