Inferential control of distillation composition using partial least squares regression

Abstract

In order to control product compositions in a multicomponent distillation column, the composition estimated from measured tray temperatures is used. In this paper, inferential models of product compositions are constructed using Partial Least Squares regression, on the basis of steady-state and time series temperature measurements. The accuracy of the estimation is greatly improved by using a dynamic model. It is also found that the use of past temperature measurements is effective for improving the performance of the inferential model. From the detailed dynamic simulation results, it is found that the cascade control system using the proposed inferential model works very well.