Quality Control of Industrial Detergents through Infra-Red Spectroscopy Measurements Coupled with Partial Least Square Regression

Abstract

Quality control of industrial products has become essential in modern industry as it aims to satisfy customer demands. Therefore it requires fast and simple procedures in order to ensure efficient on-line process monitoring and detect abnormal deviation from certain product specifications. In this work, commercial detergent quality control was performed by means of (i) Fourier Transform Infra-Red (FT-IR) spectroscopy and(ii) Partial Least Square Regression (PLS-R) that allows for prediction from multivariate spectra. Sodium hydroxide and non-ionic surfactant concentrations were considered for the calibration PLS-R model. Results demonstrated excellent predictive performance of the PLS-R model. In addition, its robustness was evaluated by mimicking a fault in the process, in this case a deviation of anionic surfactant concentration. It was found that a Q_x statistic can be introduced with the purpose to assess whether sodium hydroxide and non-ionic surfactant concentration are correctly estimated in presence of external interference.