A MONITORING APPROACH FOR DISCRETE EVENT SYSTEMS BASED ON A TIME PETRI NET MODEL Mohamed Ghazel ∗ Armand Toguyéni ∗ Michel Bigand ∗∗ ∗ Laboratoire d’Automatique, Génie Informatique et Signal ∗∗ Équipe de Recherche en Génie Industriel École Centrale de Lille, BP 48, 59651 Villeneuve d’Ascq, France Abstract: We develop in this paper a monitoring approach for Discrete Event Systems (DES) starting from a time Petri net model representing the a priori known behavior of such a system. The originality of our approach lies in the combination made of the concept of event observability with the exploitation of the temporal constraints on these events in order to refine the result of the monitoring process. Copyright c 2005 IFAC Keywords: Monitoring, diagnosis, time Petri nets, discret event system, time. 1. INTRODUCTION In this paper, we propose a monitoring approach to be applied to Discrete Event Systems (DES) for which one knows the behavior a priori. This behavior is represented by a Time Petri Net model (TPN hereafter). In addition, the events which can occur are of two types: observable and un- observable. Our objective here is to develop a method which allows filling up this partial ob- servability on the system in order to track online its state and to identify the events which occur. That will mainly enable to monitor the system by discerning online possible failures. The monitoring approach that we propose uses first the observ- able events to estimate the states the system can assume. We exploit thereafter the temporal con- straints on the events in order to refine the results of the estimation. The paper is organized as follows: In section 2, we present the Time Petri Net formalism. A rep- resentation of the state of such model is also proposed. Section 3 is dedicated to the discussion of the Enumerative Approach, a reachability anal- ysis method for TPNs, and in the fourth section we present our monitoring approach. Finally, we conclude the paper and we present the prospects of this work in the last section. 2. TIME PETRI NET: PRESENTATION - ANALYSIS 2.1 Definition Let T ⊂ Q + be a temporal field, a Time Petri Net (Merlin, 1974) on T is a 6-tuple N =<P,T,B,F,M 0 ,SIM > such that: • N =<P,T,B,F,M 0 > is a marked Petri Net (PN), (B = backward and F = forward ), • SIM : T → T × T ∞ is the Static Interval Mapping, which associates to each transition in T its static firing interval, with rational bounds of firing (as T ⊂ Q). We say here static firing interval, because when studying the dynamics of the TPN, these intervals will then evolve and one would then speak about