A Deep Learning System for Water Pollutant Detection Based on the SENSIPLUS Microsensor

Accurate classification of water pollutants is paramount for safeguarding the environment. This study presents an innovative approach to classifying water pollutants by integrating deep learning algorithms with effective preprocessing techniques. The Sensichips Smart Water Cable Sensor(SCW) facilitates real-time data acquisition for various water pollutants, establishing a robust foundation for comprehensive analysis. SCW utilizes the SENSIPLUS chip, which employs impedance spectroscopy to detect a variety of water-soluble pollutants via an array of sensors. Our research adopts an end-to-end sensor-to-classification framework, leveraging deep neural networks to capture temporal data dynamics. Employing a CNN model with a sliding window approach, our method demonstrates promising results, achieving an average accuracy of 95.78% across ten folds for classifying eight distinct water pollutants. The low-cost IoT-based infrastructure makes this approach scalable and accessible for deployment in water monitoring systems.