Correction methods for visible near-infrared spectroscopy in water pH detection to mitigate temperature influence

Abstract

The pH value is an important indicator for determining water quality, and visible near-infrared (Vis-NIR) spectroscopy is an ideal technique for rapid detection. The Vis-NIR spectrum of water is greatly affected by temperature. When using Vis-NIR spectroscopy to detect the pH value of water, temperature interference will affect the prediction accuracy of the water pH value. In this study, several chemometric methods developed in recent years were used to correct the spectra of distilled water samples under temperature interference to improve the prediction accuracy of water pH values. Specifically, this includes the external parameter orthogonalization (EPO) algorithm and the generalized least square weighting (GLSW) algorithm. Spectra of distilled water samples at different temperatures in the 400–1049 nm range were collected, and the correction effects of the EPO algorithm and the GLSW algorithm on the spectra after temperature changes, as well as the prediction errors of water pH values, were tested and examined, compared with traditional full-temperature spectral correction models. The results show that both the EPO algorithm and the GLSW algorithm methods require only a small number of representative samples to obtain filtering matrices. By correcting the spectra of distilled water samples at all temperatures to the reference temperature level, good correction effects can be achieved. Additionally, the prediction accuracy of water pH values of the models corrected by both methods is better than that of the traditional method, and the complexity of the models is also greatly reduced.