Application of near-infrared spectroscopy in batch process control

While batch processes are gaining ever increasing importance in the manufacturing industries, control of the product quality remains to be a serious challenge. To improve overall process understanding and control, new analytical techniques, such as Near-Infrared (NIR) Spectroscopy, are starting to be employed in industry. Currently, these techniques are primarily used for process monitoring purposes and have not yet been explicitly included in feedback control systems. This paper investigates the ability of three different control systems to adequately control a simulated batch reactor using the NIR spectra as feedback information such that the product meets quality specifications. The particular problem considered in this paper is adequate representation of the NIR spectrum using a single variable that is then controlled by employing Model Predictive Controller (MPC). It is shown that the resulting controller performances are highly variable if the controlled variable is chosen by selecting a single peak in the NIR spectrum to represent that variable. On the other hand, by using Principal Component Analysis (PCA) to extract information from all of the wavenumbers and represent it using a single composite variable, which is then controlled, it is shown that the process can be adequately regulated.