Multi-sensor fusion and feature selection in ultraviolet-visible spectrometry system for predicting chemical oxygen demand

The ultraviolet-visible (UV-Vis) spectrometry system is increasingly employed in chemical oxygen demand (COD) predicting recently for its significant advantages compared with traditional standard chemical method. In this study, an investigation is undertaken to determine whether the physic-chemical parameters of samples provide a good compensation for prediction. Meanwhile, a feature selecting algorithm is employed to reduce the size of UV-Vis absorption spectroscopy provided as data input to the modeling algorithm. A high correlation of above 0.90 is obtained with the data using the stand chemical method, while less absorbance values are necessary to measure and a spectrometer with industrial wavelength resolution is adequate.