Multi-models approach for the diagnosis of Discrete Events Systems

International audienceThis paper presents the first work of a multimodels approach to realize the fault diagnosis of Discrete Event Systems (DES). This approach integrates different information sources to construct two separated models: control and plant models. The goal is to enrich the image about process functioning in using different representation tools adapted to each information source. This image is enough rich to realize the three diagnosis functions: detection, localization and identification. This approach is based on a Boolean modelling of each process actuator with its according sensors, a temporal model based on a fuzzy expectation of event arrivals and a set of if… then rules. This approach is decentralized, thus both the complexity and the model manual construction task are reduced. An example of manufacturing systems is used to illustrate the functionality of this approach. Copyright© 2005 IFA