Estimation of COD from UV-Vis Spectrometer Exploiting Machine Learning in Leather Industries Wastewater

Abstract

We present a method for the analysis of wastewater in the context of the leather industry. In this context, the determination of the Chemical Oxygen Demand parameter is essential for the determination of the degree of water pollution. Conventional methods for measuring it require time-consuming laboratory analysis, sample preparation and the usage of toxic chemicals. The proposed method is based on machine learning and soft sensing, employing nonspecific sensors to derive the quality indicators of wastewater. In particular, we leverage ultraviolet and visible spectroscopy measurements, that provide wastewater absorbance, that is the quantity of light absorbed by a solution, to estimate Chemical Oxygen Demand. We stress that, after deployment, our approach does not require any (timeexpensive) laboratory analyses, and hence it can be used to implement systems of real-time monitoring of wastewater in a leather production context.

Machine Learning for the Estimation of COD from UV-Vis Spectrometer in Leather Industries Wastewater

“…In this paper, we propose an automatic data analysis approach for the analysis of wastewater. The proposed method, built over our preliminary work [14], leverages soft sensing and machine learning, allowing the determination of a water quality indicator using nonspecific sensors. In particular, the method can determine the COD by exploiting an optical sensor (a spectrophotometer), using the ultraviolet and visible (UV-Vis) wavelengths.…”