DEVELOPMENT OF AN ARTIFICIAL NEURAL NETWORK BASED APPLICATION FOR PREDICTION OF EFFLUENT WASTEWATER QUALITY.

Abstract

Water quality monitoring is undertaken to improve water quality, responsible for water borne diseases and deaths. Over five years to 2017, water-borne diseases–cholera, diarrhea, typhoid, and viral hepatitis–caused 10,738 deaths ( diarrheal deaths account for 60%). Water quality data has the following characteristics: Missing values, censored values that lie below the least count of measuring devices, outlying values, and uncertain distributions. Therefore, it is necessary to use an appropriate statistical method to draw valid conclusions and aid policy development. Here, data from 4 Sewage treatment plants are used to find the relation between input parameters and output variable(For. E.g., pH, Chemical oxygen demand, temperature, Biochemical oxygen demand, total suspended solids, and discharge). Various basic methods of trend discovery and regression are used to determine the correlation between different parameters like input pH and out pH. The aim is to incorporate Artificial Neural Networks to train models and predict the output parameters with the help of input parameters. The central objective is to build an ANN model-based application that can use plant operation and influent characteristics data to model treatment plant behaviour without going through the rigor of establishing a mechanical model. A proposed extension of the current modelling framework is to use the knowledge from this project to predict the concentrations of Emerging contaminants in the water.