An effective approach for fault diagnosis of Discrete-Event Systems modeled as safe labeled Petri nets

Abstract

In this paper, a fault diagnosis method for Discrete-Event Systems (DES) modeled as safe labeled Petri nets (LPN) is proposed. We assume that some transitions of the Petri net are unobservable, including the fault transitions. The diagnosis method consists of two steps: (i) in the first step, online fault detection is carried out; and (ii) then, by using the observed sequence of events and the labeled Petri net system model, the fault candidates are isolated. The online fault detection is based on the construction of an LPN from the fault-free system behavior model, called observable behavior Petri net (OBPN), whose transitions are all observable, and whose generated language is guaranteed to be equal to the observable language of the fault-free system model when some conditions are satisfied. It is also shown that the OBPN can be implemented for fault detection instead of implementing the reachability graph, or even part of it, which leads to a fast fault detection method without requiring the use of a large amount of memory. Two case studies are presented to illustrate the proposed method.