dentification des systèmes non linéaires par réseaux de neurones artificiels et validation par application d’une commande proportionnelle intégrale (PI)

Abstract

Abstract : Identification of nonlinear systems using artificial neural networks provides an efficient solution for large classes of nonlinear systems because these networks have the ability of learning, approximation and generalization. This work includes the exploitation of the artificial neural networks to the identification of nonlinear systems which are a tank model and a continuous stirred tank reactor CSTR model with cooling jacket. We have also applied classical proportional integral PI control method on the real system and its neural model to control the liquid level in the tank and the molar concentration of a reagent in the continuous stirred tank reactor CSTR. The objective of this work is to show the effectiveness of artificial neural networks to identify nonlinear systems.