Proc. 18th Int. Conf. Near Infrared Spectrosc., pp. 23–29 (2019)
Fast real-time monitoring of water quality can help facilitate reuse of industrial process water. Near-infrared spectroscopy is a well-established process monitoring tool within the pharma and food industry. Utilising near-infrared spectroscopy as one of the pieces in the puzzle for optimising water reuse is therefore attractive. Partial least squares regression models were computed on 286 near-infrared spectra of mono salt solutions of KCl, K2SO4, KNO3, CaCl2, CaSO4 and Ca(NO3)2. Concentrations ranging from 0 ppm to 1000 ppm (w/w) in steps of 100 ppm were measured at 10, 25 and 40 °C. Analysis showed that the concentration of salt could be predicted independently from temperature resulting in a root mean squared error of cross validation, RMSECV, of 186 ppm and an R2 of 0.67. A global temperature model and an individual model on K2SO4 at 25 °C cross validated by leave-one-concentration-out resulted in RMSECV values of 181 ppm and 115 ppm and R2 values of 0.68 and 0.87. The limit of detection and limit of quantification for K2SO4 was estimated to be 140 ppm and 400 ppm.