Supervisory Control of Discrete Event Systems Modeled With Labeled Petri Nets for Diagnosability Enforcement

Abstract

This paper proposes an active fault diagnosis method to enforce the diagnosability of discrete event systems using labeled Petri nets by constructing a diagnostic supervisor. For a non-diagnosable net model, its diagnosability is viewed as a control specification and addressed by using supervisory control techniques. First, an event-based monitor and a Petri net structure, referred to as data isolation arcs, are introduced to apply control specifications for labeled Petri nets. Then, a diagnostic supervisor reachability graph is generated to estimate the current state of the diagnoser. By analyzing the diagnostic supervisor reachability graph, an integer linear programming model is formulated to design a diagnostic supervisor, which can prevent the system from entering into any indeterminate cycle only by observing the occurrence of events. Finally, by the obtained diagnostic supervisor, the resulting net model is shown to be diagnosable. Some examples are presented to demonstrate the proposed method. Note to Practitioners—Faults have a significant impact on the normal operation of a system, and timely detection and isolation are crucial to ensuring system stability and production efficiency. Nevertheless, in certain systems, the occurrence of faults cannot be determined by a finite number of observations. To address this problem, this work introduces an active fault diagnosis method in the framework of labeled Petri nets, aiming to enforce the diagnosability of a system. The diagnosability condition is treated as a control specification such that it is readily accessible for practitioners to construct a diagnostic supervisor with Petri nets by following a typical and traditional control paradigm, which ensures that the controlled system is diagnosable.