Self-tuning multirate control of binary distillation columns

Two multivariable self-tuning control algorithms based on predictive control techniques are presented and applied to some realistic simulation models of binary distillation columns. The nonlinear model of binary distillation columns subsequently used in simulation is summarized. The good agreement between the behaviour of an industrial plant and the model is highlighted by analyzing the effects on the composition products of open loop step variations in the vapour and distillate flow rates. The multirate generalized minimum variance (MGMV) algorithm is briefly presented. The algorithm is then applied to the control of an 11 plates binary distillation column model. The obtained results are presented and show the improvements obtained by means of a multirate scheme. However, the simulated plant does not exhibit the `nonminimum-phase´ behaviour typical of real columns. In these circumstances, MGMV can still ensure closed loop stability at the cost of oscillatory and sluggish transients. In order to overcome these frequently encountered difficulties, attention is drawn towards the development of a multistep-multirate algorithm obtained by suitably rearranging the generalized predictive control method. The derived algorithm multirate generalised predictive control is then used for control of a highly `nonminimum-phase´ column. The corresponding results are reported. The advantages of resorting to multirate control algorithms are clearly pointed out in this case too. Finally the paper ends with some concluding remarks and hints for future work