Abstract

Clean drinking water is fundamental for the welfare of humankind, and although many efforts are being made in the direction of automated water quality monitoring, the process of sampling sensors manually and sending them to a lab where their content undergoes chemical analysis is still predominantly aperiodic [6], [8], thus resulting in contamination of the water being discovered too late. Recent research has confirmed that inexpensive IoT devices can convey real-time data concerning the water's pH, turbidity, total dissolved solids (TDS), and temperature to the cloud [3], whereas ML has proven capable of classifying water quality as either good or bad, calculating the Water Quality Index (WQI), and detecting anomalies [2], [4]. We conducted a review of the literature consisting of eight relevant studies and found out that there are three recurring trends: the majority of IoT systems lack a prediction feature [3], [6], most ML studies utilize training data for their algorithms [7], and a consistent Internet connection is required for most systems [1], [4]. We propose a layered system consisting of Sensing Layer, Edge and Communication Layer, Cloud Layer, Intelligence Layer, and Application and Alert Layer using ESP32 microcontroller as the main element, with the methodology for data preprocessing, WQI calculation, ML classification, anomaly detection, and alert delivery being described in detail, including a representative example scenario. Since the research is currently at the conceptual stage, no implementation or performance evaluation is carried out, and an evaluation approach is provided for further research.

Keywords

How to Cite This Article

Rights