Improving on-line fault diagnosis for discrete event systems using time

This paper presents an approach to fault diagnosis of discrete events systems modeled by Petri nets. Fault events are associated to a subset of unobservable transitions of the net. The approach is based on an extension of standard net marking, called generalized marking, which allows to efficiently estimate the explanations of the observed transitions. The algorithm proposed in this paper uses the information on the timing structure of the net to accelerate the explanations estimate. In particular, it requires the solution of linear programming problems and the computation of a tree, called timed explanation tree, introduced to compute the firing time of the minimal unobservable explanations that include a given fault transition.