Diagnose online de sistemas a eventos discretos modelados por Redes de Petri rotuladas acíclicas

Abstract

The objective of this thesis is to present a method for online failure diagnosis for discrete event systems (DES) modeled by acyclic labeled Petri nets in which different transitions can be labeled by the same event. Initially, it is proposed an online diagnoser that makes its decision about the occurrence of the failure by storing sequences of observed events and verifying, after the occurrence of an observable event, if two sets of inequalities are satisfied: the first set accounts for the normal or fault-free behavior whereas the second set accounts for the faulty behavior of the system. In addition, we consider the problem of diagnosability verification in which sets of inequalities, when satisfied, allow to decide if the language generated by a Petri net is diagnosable. The method adopted for online diagnosis has the advantage over those found in the literature since its depends only on the verification if a vector of variables satisfies an inequality system, being self-contained within the formalism the Petri net. However, the proposed method is restricted to Petri nets that do not present failure and normal sequences sharing the same number of occurrences of the observable events. A modified diagnoser is proposed to encompass all the acyclic labeled Petri nets, in which more than two sets must be checked for system behavior definition being the number of sets dependent on the reachability tree of the Petri net. A manufacturing system is used to illustrate the methods proposed here.