Application of Multivariate Curve Resolution to Chemical Process Control of an Esterification Reaction Monitored by Near-Infrared Spectroscopy

Abstract

Multivariate curve resolution–alternating least squares (MCR-ALS) methodology was applied to near-infrared spectroscopy (NIR) data for the esterification reaction between glycerol and a mixture of caprylic and capric acids. Batch reaction processes were conducted either at the laboratory scale or at an industrial plant, while NIR data were obtained from samples withdrawn during the reaction processes. The process has been monitored via two typical parameters for this type of reaction, namely, the acid value (AV) and the hydroxyl value (OHV). Spectral and concentration profiles were estimated by applying soft-modeling MCR–ALS to a column-wise augmented data matrix with pure spectra of the components, and concentration values for the acid were used as a soft-equality constraint. The estimated concentration profiles have been compared with the AV and OHV values, and the estimated spectral profiles were used to predict the concentration profiles for new batches. Good results have been obtained in terms of RMSE for the prediction of AV and OHV.