Spectral Decomposition Method for Water Chemical Oxygen Demand Measurement Based on UV-Vis Absorption Spectroscopy

Abstract

We introduce a spectral decomposition method by modifying the GMM (Gaussian mixture module) algorithm. The first key point is to circumvent the process of random data generation for less error; the second key point is to replace the Gaussian model with a bi-Gaussian model to overcome the limitations of conventional symmetric Gaussian approximations. Comparative analysis with the Levenberg–Marquardt algorithm and standard GMM approaches demonstrates the superior accuracy of our method, as evidenced by the minimization of spectral reconstruction errors across all tested wavelength regimes. The effectiveness of this decomposition method for chemical oxygen demand (COD) detection was assessed via a series of experiments with real samples of sewage from various plants. The analysis showed less root mean squared error (RMSE) value by using the B-band of the benzene ring after the procedure of decomposition in contrast to the methods of peak searching and fixed wavelengths. The proposed method can improve the environmental suitability of COD detection effectively.