Monitoring production processes of pharmaceuticals and farm chemicals with near-infrared spectroscopy coupled with the interaction moving window partial least squares method

Abstract

Online near-infrared spectroscopy is utilized to monitor pharmaceutical and farm chemical production processes. To improve model performance, interval interaction moving window partial least squares (iMWPLS) is used for wavelength selection. As a modified version of moving window partial least squares (MWPLS), iMWPLS may accurately screen out intervals and optimize the combination of the spectral intervals by iteratively accomplishing MWPLS on the selected intervals. Two online NIR datasets generated during the aspirin synthesis and dichloroacetophenone production processes are subjected to multivariate calibration with partial least squares (PLS) and wavelength selection with iMWPLS. The results show that the partial least squares models built with wavelengths selected by iMWPLS exhibit significant improvement, with low root mean squared errors of calibration and prediction, when compared with models established using both the original and preprocessed spectra, as well as models that used wavelengths selected using MWPLS. This study shows that iMWPLS has the ability to improve model performance by accurately screening spectral intervals to construct an optimized combination of intervals.