Multivariable Adaptive Control of a Binary Distillation Column

Abstract In this paper we consider the adaptive control of a binary distillation column. The considered process is multivariable and non minimum phase. The two products to be separated are water and methanol. The control inputs are the reflux flow rate at the top of the column and the heating power at the bottom of the column. The controlled outputs are the top product composition and flow rate. This distillation column is non linear (nearly bilinear) and the cross couplings are fairly high. For these reasons, we use the adaptive control technictues in order to try to cope with the non linearities. The linear process model is non minimum phase so we use an indirect adaptive control scheme. The underlying linear control scheme is L.Q. type. We aim at following a decoupled reference model while minimizing the quadratic criterion. In a first step, multivariable recursive identification of a binary distillation is carried out. Some simulation results on the identified model are presented.