A CPN-Approach for DistributedAbductive Reasoning : Application to Causal Model-Based Diagnosis

Abstract

This thesis deals with fault diagnosis of distributed systems from a model-based view where Coloured Petri Nets are used to describe the systembehaviour. The systems concerned here are those comprising different interactingsubsystems. Coloured Behavioural Petri Nets are defined as a particular CPNintended for the description of a system’s causal behaviour, where each transitionis labelled with a matrix describing explicitly its firing ways. The use of suchmatrices helps in tackling the problem of complexity during backward analysis,and gives rise to a very specific technique based on reachability of CBPNs calledCW-analysis. CBPNs together with the CW-analysis are used to develop a dis-tributed model-based diagnosis approach. The diagnostic system is defined as setof diagnostic agents where each is assigned to diagnose a subsystem. Accordingly,the system model consists of a set of place-bordered CBPNs, whereas CW-analysisis exploited to implement a local diagnosis scheme. Once local diagnoses are ob-tained by the different agents, a cooperation process should be initiated to ensureglobal consistency of such diagnoses