Decentralized Approach to Diagnose Manufacturing Systems

Manufacturing systems are an example of discrete event systems (DES). They are modular and informationally as well as geographically decentralized. Thus, a centralized diagnoser is not adapted for them. This paper presents a decentralized diagnosis approach to perform the diagnosis of manufacturing systems. This approach is based on a Boolean modelling and event-state-based local diagnosers. The use of a Boolean model is useful to obtain an abstracted model easier to exploit especially in the case of complex systems where the number of states is large. Local diagnosers combine event and state based models to infer the fault´s occurrence using event sequences and state conditions characterized by sensors´ readings and commands issued by the controller. A co-diagnosability notion is defined in order to verify that the diagnosis performance of the local diagnosers is equivalent to the one of the centralized diagnoser