State estimation and diagnosis for systems described by multiple models Application to a biological reactor.
Résumé
This communication addresses the problem of state observation of a wastewater treatment plant. As this kind of system is complex, classical observer structures are difficult to design. For that reason, we propose to describe the system by a multiple model structure which appears as a blending of linear models. In the first part of the presentation, it is explain how to design such structure and how to design a state observer. As a consequence, diagnosis and fault tolerant control may be performed using the results given by a multiple observer. Using these theoretical results, the second part of the presentation is mainly devoted to the diagnosis of a wastewater treatment process. Some numerical simulations illustrate the state and the unknown input estimation technique.