Regression Based Contamination Detection for Real Time Water Quality Assessment

Authors

  • K. Sri Dhivya Krishnan, P.T.V. Bhuvaneswari

Abstract

The significance of continuous and reliable measurement of toxic contaminant level in drinking water is to maintain the strategic distance from the risk of water related diseases. Drinking water gets polluted due to urbanization, deposition of untreated effluents from industries and improper maintenance of the water bodies, resulting in the reduced fresh water supply leads to contagious diseases. Water Quality Assessment (WQA) techniques are highly computational not only for manual water sample collection and laboratory procedures, also due to the usage of numerous Water Quality Parameters (WQPs). Due to the huge demand, in exploiting cost effective and portable detection method in onsite WQA, the proposed work aims to employ the major WQPs such as pH, Total Dissolved Solids (TDS) and conductivity. Initially, different chromium (Cr) compounds were used to prepare the test samples by adding the contaminant in water from 0 to 5 mg/L in steps of 0.1 mg/L to measure the WQPs. The dataset of 918 WQPs is generated from 51 different concentrations of Cr contaminants. Then, correlation coefficient is utilized to determine the dependency between the concentration of contaminant and each WQP. Later, Regression based prediction models with different combination of WQPs are employed to estimate the level of contamination in water. It is inherent from the evaluation results that the proposed regression model using conductivity parameter is relatively better than other statistical models based on the pH and TDS parameters in contamination estimation achieving R2, adj R2, R2 (prediction) values are higher than 90%, S <0.2 and RMSE<0.39.

Keywords- chromium; contamination detection; correlation; water parameters; water quality monitoring; water quality sensors

Published

2020-12-31

Issue

Section

Articles