Metrological support for the control of wastewater composition: application of neural networks in measurement result processing methods

This article examines the metrological support for wastewater monitoring required to reduce water pollution, from assessing the initial quality of effl uents to monitoring the effectiveness of their treatment before discharge into natural waters. A priority innovative task for water-intensive chemical industries is described: comparing the results of measurements of various parameters of water composition and properties to selectively reduce the concentration of particularly hazardous toxicants. The need to improve the methodology for identifying (and subsequently removing) pollutants that most signifi cantly increase environmentally hazardous chemical and biological oxygen demand is demonstrated. A comparison of metrological analysis methods, such as predictive mathematics, traditional regression analysis, and neural networks, is conducted using the example of a study of wastewater from the sewer system of the Kemerovo Nitrogen Industry Enterprise “AZOT”. Neural networks are shown to be the most effective method, as they have been used to establish the most comprehensive and maximum number of cause-and-effect relationships, including nonlinear ones, between pollutants and chemical and biological oxygen demand, compared to other methods. Based on the results of the study, it is recommended to use neural networks to analyze the cause-and-effect relationships of measured values of the composition and properties of wastewater for metrological support of water and environmental safety of industrial wastewater disposal in nitrogen production.

metrological analysis, wastewater composition, comparability of measurement results, regression analysis, data processing, neural networks, metrological examination of physico-chemical processes