Neuro Control for Multicomponent Distillation Column

Abstract The control problem of a multicomponent nonideal distillation column is proposed by using a dynamic neural network approach. The holdup, liquid and vapor flow rates are assumed to be time-vaying (nonideality). The technique proposed in this paper is based on two central notions: a dynamic neural identifier and a neuro-controller for output trajectory tracking. The first one guarantees boundness of the state estimation error with a small enough tolerance level. The tracked trajectory is generated by a nonlinear reference model, and we derive a control law to minimize the trajectory tracking error. The controller structure which we propose is composed of two parts: the neuro-identifier and the local optimal controller. Numerical simulations, concerning a 5 components distillation column with 15 trays, illustrate the effectiveness of the suggested approach.