Analysis and diagnosis with Petri net models

Mon, 05/02/2007 (All day)
El curso se impartirá los días:
  • 5 de febrero de 17 a 21 horas en el seminario 22 del edificio Ada Byron
  • 6 de febrero de 17 a 20 horas en el seminario 22 del edificio Ada Byron
  • 7 de febrero de 17 a 20 horas en el seminario 23 del edificio Ada Byron
The first point of this course is to explain performance evaluation in the context of reliability analysis and fault detection and diagnosis. Both studies are concerned with safe exploitation and control of industrial systems. Reliability analysis is usefull to prevent faulty behaviours and to decrease the frequency of faults. Fault detection and diagnosis is usefull to protect system and to decrease the gravity of faults.

The second point is a short introduction about Petri nets and there use for the modelling of discrete event systems (DES), continuous systems (CS) and hybrid dynamical systems (HDS). The modelling of fault events and of faulty behaviours is introduced and discussed.

The third point concerns reliability analysis with Petri nets. The idea is to give the basic notions that are required to start a simulation with stochastic Petri nets and to give a correct interpretation of the results. The exponential distribution of delays is introduced and Petri net variables and reliability indicators are linked and discussed.

The last and most important point is to discuss the use of Petri net for fault diagnosis of DES and HDS. The diagnosability problem is first presented. Structural analysis of Petri nets is then introduced in the context of fault diagnosis: directed paths and causality relationships are defined and characterized, influence areas of fault events and dependance areas of observable variables are also introduced. The idea is to present and compare two different point of view concerning the diagnosis : the behavioural approach based on a partial extansion of the reachability graph, and the structural approach based on event detectability and directed paths.