Development of a Portable Water Quality Monitoring System Integrating pH, TDS Sensors, and CNN-Based Visual Analysis

Abstract

Access to safe, clean drinking water is still a global challenge. Currently, over 2.2 billion people are using water sources that are not safe for consumption. Traditional water quality testing is labor intensive and time-consuming, limiting access. This study focuses on the design and development of a smart, portable prototype to assess tap water quality in real-time and to recommend filters/filtration methods. Using pH and TDS sensors along with a Convolutional Neural Network (CNN) on a Raspberry Pi (RP), the prototype classifies water samples into clean, colored, oily, saline, and turbid categories. The water sample’s sensor readings and images are used to make suggestions about the suitable methods for filtration through an intuitive  user interface. The classification accuracy of the CNN model is 75.53% on the test dataset and the sensors proved to be highly reliable against commercial meters. The entire inference pipeline takes less than three seconds, including image capture, classification, and recommendation. The proposed solution is a viable and cost-effective solution for monitoring water quality in both domestic and industrial settings; providing real-time support for decision-making regarding filtration methods while improving public health outcomes. Future work could include increasing CNN dataset diversity, incorporating additional sensors, adaptive learning mechanisms from user feedback, and adding cloud capabilities in the proposed system.