Multivariable inferential feedback control of distillation compositions using dynamic principal component regression models

Abstract

Multivariable inferential feedback control of distillation compositions using principal component regression (PCR) models is presented in this paper. Both static and dynamic models are studied. PCR model based software sensors are developed from process operational data, so that the top and bottom product compositions can be estimated from multitray temperature measurements. The problems of co-linearity in tray temperature measurements are addressed by using PCR. Static estimation bias and the resulting static control off-sets are eliminated through mean updating of process measurements. Application to a simulated methanol-water distillation column demonstrates the advantage of dynamic PCR model based inferential feedback control. It is shown that dynamic PCR model based inferential estimations are more robust to process operating condition variations than those based on a static PCR model.