A Model-Following Control Scheme Using ArtificialNeural Networks for Multivariable Systems with Unknown Models

A model-following control scheme for multivariable systems with unknown mathematical models is proposed. A neurocontroller is designed with artificial neural network to implement the model-following control law. For the controlled system to follow the reference model, an algorithm for optimal learning control is developed in this paper, by which the numerical solution of the optimal model-following control law is obtained. The numerical solution is employed as samples in training the neurocontroller. As neural networks can learn to approximate functions, the neurocontroller is trained to generate the optimal model-following control law. The proposed model following control scheme is applied to an actual multivariable system, a paper machine. The simulation results show that the scheme described in this paper is effective.