Neural network applications in process modelling and predictive control

Abstract

Neural network techniques are investigated applied to the modelling and control of non-linear processes. The development of process models and predictive controllers using two feed-forward neural networks - the multi-layer perceptron and the radial basis function network - is described. The capabilities of these neural networks are demonstrated in two practical applications to modelling and control of a liquid level rig and a multi- variable in-line pH process. On-line results illustrate the performance of neural network predictive control schemes for set-point tracking over a wide non-linear operating range and regulation in the presence of significant disturbances.