Application of multivariable generalised predictive control to the Simulink model of a paper machine

This paper presents the application of two GPC strategies, multiple model GPC and adaptive GPC, to the SIMULINK model of a paper machine. The plant features a rather large dead-time and severe filler retention variations at the wire section. In multiple model GPC an extra sensor measuring the retention level is used to select between a set of linear models identified at various fixed retention levels. At each sampling instant the selected model is employed in a normal GPC scheme. Simulation results show that the performance is much improved when compared to the fixed GPC case. The adaptive GPC method considered in this paper updates the gains of the plant transfer matrix using two 2-input-1-output least-squares estimators. Results indicate that the control is noticeably tighter than the fixed GPC case, at a price of more oscillatory responses