Coloured Petri Nets Based Diagnosis on Causal Models Soumia Mancer and Hammadi Bennoui Computer science department, LINFI Lab. University of Biskra, Algeria mancer.soumia@gmail.com, bennoui@gmail.com Abstract. In the last decades, several approaches have been proposed in order to capture the problem of causal model-based diagnosis within Petri Nets (PNs) framework, where both the structural and behavioural analysis of the net model are exploited for reasoning. In fact, PNs are a useful tool, but most of the approaches suffer from the large size of the obtained models even for simple systems. This paper introduces a novel class of Coloured Petri Nets (CPNs) called Causal CPNs. Such a net model is motivated by representing the causal behaviour of the system to be diagnosed, as well as, simplifying the analysis methods. The diagnosis technique exploits backwardly the reachability graph of the net model. A case study is used to illustrate the usefulness of our proposal for fault diagnosis. Keywords: model-based diagnosis, causal models, coloured Petri nets, reachability graphs 1 Introduction Model-based approach as an alternative to heuristic based one, especially when the experimentations are missing, deals widely with fault diagnosis in such a manner that the examination of a given system is done on the basis of a model. It aims at explaining any observed behaviour that conflicts with the way the system is meant to behave. Among the diagnosis frameworks found in the literature, those based on causal models where the explanations would be given in terms of initial causes leading the system to a misbehaviour. In logical frameworks, a causal model-based diagnosis problem is traditionally solved through symbolic manipulations that are shown as a cumbersome task; and hence, numerous attempts to face this problem have been done. In partic- ular, Petri nets (PNs) have been used to represent the causal model, and so to exploit their analysis techniques to implement efficiently the diagnosis reason- ing mechanisms [1, 2]. For problems where the net model is large or composed of some identical parts, it is well known that Coloured PNs (CPNs) are well suited to use with respect to classical PNs. By means of data type primitives, it is possible to achieve a reduced model about the behaviour of the system under examination. The manipulation of the data values carried by tokens that