A systematic determination approach of a models' base for uncertain systems: experimental validation

The increasing complexity of industrial processes and the higher performance requirements, make the multimodel approach necessary. This approach is often confronted to the determination of the useful models' base problem. An approach for a systematic determination of a models' base for the representation of uncertain linear systems is suggested in this article. The application of this approach requires two main steps. We are firstly interested in the classification of the data by using Kohonen self-adapting artificial neural network and in the determination of the necessary number of models. Secondly, we exploit the obtained data relative to the clusters for the structural and parametric identification of different base-models. An experimental validation of the suggested modelling approach is carried out on an olive oil esterification reactor. The obtained results are satisfactory and show a very good precision relatively to the case in which the classical modelling, based on a unique model, is adopted.