Fault diagnosis in labelled Petri nets: A Fourier–Motzkin based approach

Abstract

We propose techniques for fault diagnosis in discrete-event systems modelled by labelled Petri nets, where fault events are modelled as unobservable transitions. The proposed approach combines an offline and an online algorithm. The offline algorithm constructs a diagnoser in the form of sets of inequalities that capture the legal, normal and faulty behaviour. To implement the offline algorithm, we adopt the Fourier–Motzkin method for elimination of variables from these sets of inequalities. Upon observing an event, the diagnoser is used to determine whether a fault occurred or might have occurred. The occurrence of a fault can be verified by checking the observed sequence against the sets of inequalities. This approach has the advantage that the tradeoff between the size of the diagnoser and the time for computing the diagnosis is achieved. In addition, fault diagnosis in both bounded and unbounded Petri nets can be addressed.