Agent-based Systems for diagnosis and supervision of industrial processes

摘要

Generally, an IPCS should reconcile two contradictory properties: reacting in "real time" to the solicitations of the environment and operating out some evolved reasoning in order to meet these solicitations. This paper, proposes a diagnosis architecture using the Data Mining techniques. These techniques enable to answer different tasks of the diagnosis such as the selection of pertinent indicators, the generation of breakdown cases, the prediction and more particularly, the temporal and spatial dependencies of the IP.