Abstract

This paper investigates the possible use of artificial neural networks (ANN), more precisely multi-layer perceptrons (MLPs), for the nonlinear modelling and predictive control of a milk pasteurisation plant. Model predictive control (MPC) schemes require the development of a predictive model. Using data gathered from an industrial milk plant, a nonlinear multi-step ahead neural network predictor model (NNM) was established. A neural predictive controller (NPC) was then designed on the same basis for the control of milk pasteurisation temperature. Simulation results are presented and conclusions are drawn.