Contributions to the multimodel approach : systematic determionation of a models base and validities estimation

Abstract EN

The multimodal approach is a research subject developed for modeling, analysis and control of complex systems.

This approach supposes the definition of a set of local model.

Each local model can describe the system’s behavior in a given region.

Different strategies exist in the literature dealing with the determination of these local model.

In this paper we propose new method for the systematic determination of the local model's base.

This method is based on the classification of a set of identification data related to the considered system.

The number of the local models, their orders and their parameters are generated automatically without any prior knowledge about the full operating range of the process.

The good precision, that has demonstrated this new method in simulating, is confirmed by an experimental validation using real processes.

As well as the local models determination, the multimodal representation is still confronted to the models selection problems which my use either a simple commutation or a fusion technique.

We present, also, in this paper a fusion strategy based on a fuzzy supervisor which evaluates the pertinence of each model of the base using a computation of both the residues and their variation.

This strategy is experimentally validated.