Multilayered neural networks for identification and control of a multivariable distillation process

This paper describes and discusses a multivariable control application of the feedforward multilayered hidden neural networks. First, an identification and control scheme of nonlinear multivariable systems is developed with reference to adaptive control. In this scheme we make use of two multilayered hidden neural networks, both trained with dynamic backpropagation algorithm and assigned respectively the identification and control tasks. The basic theoretical concepts and definitions are introduced throughout the paper. Simulation results reveal that the proposed neural networks control scheme are practically feasible. The sufficiency of the training set for identifying and controlling a nonlinear multivariable distillation process is investigated. The application results show that performances of the proposed control scheme based on the multilayered hidden neural networks are superior to conventional control strategies