Advanced control of a polymerisation reactor

A problem common to many chemical processes is the presence of a significant pure time delay, due either to transport lag or high order dynamics. Model based control techniques, such as internal model control, (Garcia and Morari, 1982) and Smith predictors, (Smith, 1958) have been proposed for the control of such plants. For most practical applications, such as chemical, power plants, etc. Process modelling can represent a substantial task requiring much development time, (Nicholson, 1980). An alternative is to develop a suitable parametric model that fits in some manner input, output and noise data from the plant, (Ljung and Soderstrom, 1983). In particular linear parametric models form the basis of many of the popular adaptive control techniques, such as pole-placement and minimum-variance self-tuning control, (Wellstead and Zarrop, 1991) and model predictive control techniques, (Clarke et al., 1987). In this work it is proposed to utilise such model based approaches to improve the control performance of an industrial polymerisation reactor. This involves the development of a process model using system identification techniques, the simulation of the plant within the Simulink environment to allow for the design and validation of control strategies. From these studies a Smith predictor was implemented to significantly improve the polymer viscosity control. Finally a hardware platform is developed to facilitate the implementation of sophisticated algorithms such as recursive least squares, that could not be accommodated on the present DCS