Scholarlink

<p>Access to safe drinking water remains a global challenge, worsened by pollution and delays in traditional lab analysis. This paper presents a low-cost (under $80), real-time Smart Water Quality Monitoring System utilizing the ESP32 microcontroller. The system measures four crucial parameters, specifically pH, TDS, temperature, and turbidity, and transmits data to a cloud backend for remote visualization on a dynamic web dashboard. A key innovation is the integration of an on-device machine learning model (TinyML) for intelligent, real-time categorization of water impurity events. A neural network, trained on a custom 6,000-point dataset and deployed using the TensorFlow Lite for Microcontrollers framework, distinguishes between ‘Normal’, ‘Rainwater Runoff’, and ‘Chemical’ impurity profiles with 99.28% accuracy. This approach enables independent anomaly detection without reliance on cloud connectivity for decision making. To address hardware constraints, an intelligent fluctuation-based logging algorithm was implemented, demonstrating a 98.2% reduction in SD card write operations during a 24-hour stability test (1,520 writes versus 86,400 in continuous mode). Combined with an automated pump control mechanism, this approach provides an autonomous, resource-efficient solution for decentralized water safety monitoring.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

An Intelligent, low-cost water quality monitoring system with on-device machine learning and cloud integration

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年2月26日

查看全文
  • Saurabh Sharma,
  • Devansh Mishra,
  • Ayush Yadav,
  • Bibek Gami,
  • E. S. Madhan

摘要

Access to safe drinking water remains a global challenge, worsened by pollution and delays in traditional lab analysis. This paper presents a low-cost (under $80), real-time Smart Water Quality Monitoring System utilizing the ESP32 microcontroller. The system measures four crucial parameters, specifically pH, TDS, temperature, and turbidity, and transmits data to a cloud backend for remote visualization on a dynamic web dashboard. A key innovation is the integration of an on-device machine learning model (TinyML) for intelligent, real-time categorization of water impurity events. A neural network, trained on a custom 6,000-point dataset and deployed using the TensorFlow Lite for Microcontrollers framework, distinguishes between ‘Normal’, ‘Rainwater Runoff’, and ‘Chemical’ impurity profiles with 99.28% accuracy. This approach enables independent anomaly detection without reliance on cloud connectivity for decision making. To address hardware constraints, an intelligent fluctuation-based logging algorithm was implemented, demonstrating a 98.2% reduction in SD card write operations during a 24-hour stability test (1,520 writes versus 86,400 in continuous mode). Combined with an automated pump control mechanism, this approach provides an autonomous, resource-efficient solution for decentralized water safety monitoring.