Smart water quality monitoring: thermal variation on turbidity & TDS sensing with machine learning classification of liquid solutions

Abstract Turbidity and total dissolved solids (TDS) are the most essential factors for water quality monitoring to get an accurate assessment, which is affected by thermal variation. The research offers a thermal change sensing system for measuring these parameters to provide more accurate and stable results. The experimental research over different types of liquid samples including drinking, rain, river, tube well, coconut, calcium and mud water demonstrated a distinct style where turbidity decreased with higher temperature due to particle settling and TDS increased from enhanced solubility. The system can give accurate turbidity (NTU) and TDS (ppm) readings throughout a wide temperature range of 15 °C to 150 °C by using regulated sensors. Machine learning techniques were employed to categorise different water types based on a combination of these parameters (temperature, TDS, and turbidity). The Decision Tree approach provides the highest classification accuracy among the other algorithms like Support Vector Classifiers (SVC), Random Forests and KNN. Moreover, it has been noted that using machine learning with temperature fluctuation in sensing improves accuracy and makes it easier to distinguish between various kinds of water. Thus, using the AI-ML techniques, we can accurately estimate the types of water based on those parameters. We may also predict whether water will be suitable for drinking at a given temperature. The acceptable TDS thresholds limit of drinking water is safe up to ~80 °C for human beings. Furthermore, it was found that the TDS of river and muddy water was significantly influenced by thermal variations; however, these waters are not drinkable at any of the investigated temperatures. The discussion of the outcomes highlights the potential of this approach for real world applications.