INDUSTRIAL & ENGINEERING CHEMISTRY RESEARCH, cilt.43, sa.19, ss.6102-6111, 2004 (SCI-Expanded)
An inferential control methodology, that utilizes an artificial neural network (ANN) estimator for a model predictive controller (MPC), is developed for an industrial multicomponent distillation column. In the control of product compositions by a feedback control system, because of the difficulty of on-line measurements of compositions, temperature measurements can be utilized. The selection of the temperature measurement points for the inferential control is done by the help of singular value decomposition (SVD) analysis together with column dynamics information. A moving window ANN estimator is designed to estimate the product compositions from tray temperature measurements. The composition predictions are further corrected with the actual composition data in 30-min intervals. A multi input multi output (MIMO) MPC is used with the developed ANN estimator for the dual composition control of the column. The performance of the developed control system utilizing ANN estimator is tested considering set-point tracking and disturbance rejection performances for the unconstrained and constrained cases. It is observed that the controller utilizing ANN estimator is as good as the controller utilizing direct composition values.