Reduction Rules for Diagnosability Analysis of Complex Systems Modeled by Labeled Petri Nets

Abstract

This article addresses the combinatorial explosion problem for diagnosability analysis of discrete event systems (DESs) using bounded labeled Petri nets (LPNs). Some reduction rules are given to simplify a priori the LPN model before analyzing the diagnosability. When the conditions of these reduction rules are fulfilled, some regular unobservable transitions and some specific observable transitions are suppressed. It is proven that these rules preserve the diagnosability property of the LPN system. By using reduction rules, the memory cost for diagnosability analysis is reduced. Note to Practitioners-Fault diagnosis based on discrete-event systems has been successfully used in several fields of applications, such as transportation, telecommunication, manufacturing, and so on. At the design stage of a system, the diagnosability needs to be held, which refers to the ability to determine if the system can detect the fault after its occurrence. Therefore, the diagnosability is a critical property due to its importance in terms of safety of an industrial system. In order to allow an industrial exploitation of diagnosability analysis, this article proposes reduction rules, which make it possible to reduce a priori the size of the LPN model of an industrial system.