Wiener model based nonlinear predictive control of a pH neutralisation process

A case-study evaluation of a Wiener model based nonlinear predictive control method is presented. The approach makes use of Wiener model identification and does not require first-principles modelling. It was tested on a simulated model of a pH neutralisation process that was reconstructed from the literature. The obtained results are better than those resulting from the original physical model based nonlinear control and also of an artificial neural network (ANN) based approach. The performance is excellent also in the case of a considerable plant-to-model mismatch. There is a clear relation to the underlying linear model based method in a form of gain scheduling, so that the properties of the nonlinear control system can be analysed from the comprehensible linear control aspect. The method combines advantages of linear model based predictive control and gain scheduling while retaining a moderate level of computational complexity, thus it can be applied as the first next step in cases where performance of linear control is unsatisfactory due to process nonlinearity. Copyright (C) 1998 IFAC.