Diagnosability Enforcement in Discrete Event Systems Using Labeled Petri Nets

This research solves the problem of diagnosability enforcement in discrete event systems modeled by labeled Petri nets. Since fault diagnosis relies on system observations, the detection of the occurrence of faults can be interfered with by deadlocks. Thus, most documented approaches in the literature assume that a plant is deadlock-free. In this paper, given a plant that may contain deadlocks, a revised topology is constructed to verify the diagnosability of the original plant. It is shown that a plant is diagnosable if there does not exist an elementary faulty path in the extended diagnoser derived from its simplified model. Further, it is shown that the diagnosability of a plant may remain unchanged if the initial tokens of certain places increase. An online control policy is designed to enforce the diagnosability without computing the (partial) state space of the plant. Finally, a manufacturing example is presented to illustrate the proposed method.