Adaptive neural predictive control for industrial multivariable processes

Abstract

An adaptive neural predictive control for industrial multivariable processes is presented. Neural network models together with linear dynamic models are used to approximate weakly coupled industrial multivariable processes with time delay. The predictive control law is derived on the basis of the minimization of a generalized predictive performance criterion. A real-time adaptive control algorithm, including the recursive least-squares estimation and neural predictors of back-propagation neural networks, is proposed and then successfully applied to achieve the performance specifications for multivariable heating process of a plastic injection moulding machine. Both simulations and experimental results are used to show the effectiveness of the proposed method for time-delay multivariable processes with set-point changes, load disturbances and significant plant uncertainties.