Water Quality Monitoring
ESP32 sensor telemetry, historical water-quality trends and Bi-LSTM analysis in one monitoring pipeline.
2025
Sensor → API → history → analysis · IEEE ICoECIT 2026.
Sensor telemetry and temporal analysis
The water-quality framework connects ESP32 sensing, Python/Flask ingestion, MongoDB records and a React dashboard. Timestamped parameter readings make current conditions and historical changes accessible through the same pipeline. A Bi-LSTM path analyzes multivariate observed windows.
Source-defined channels cover temperature, TDS, turbidity, pH, dissolved oxygen and ammonia. Acquisition, persistence and analysis remain separate responsibilities. The work is associated with IEEE ICoECIT 2026 and Dr. Shilpa Sondkar.
Sensor to dashboard
- Sensor array
- ESP32 JSON
- Flask ingestion
- MongoDB history
- BiLSTM / React
Firmware, processing and presentation form the framework’s three connected layers.
What the framework connects
- Sensing: the ESP32 coordinates the parameter inputs and packages a measurement cycle as structured telemetry.
- Transport: Wi-Fi carries JSON observations to the application’s REST ingestion path.
- History: records retain device/time context so readings can be retrieved as meaningful sequences.
- Presentation and modeling: the dashboard visualizes parameter trends, while the forecasting workflow works over prepared multivariate windows.
Engineering decisions
ESP32 packages parameter values for network ingestion, with device and time context attached to readings. Different sensor supply/measurement characteristics belong to the acquisition layer. Structured objects reach the Flask API before MongoDB persistence, enabling retrieval of related device observations.
Sequence preparation organizes readings into windows with lagged values and rolling context. Bi-LSTM processing uses both directions within an observed window, modeling related parameter changes rather than isolated samples. Stored data supports history and visualization independently of inference.
The interface retains parameter-specific units: pH, concentrations and turbidity remain distinct signals. Threshold-oriented views and sequence analysis connect measurements to monitoring output, while the modular architecture leaves ingestion, storage and model processing inspectable.
Publication & collaboration
A Modular IoT Framework for Water Quality Monitoring with Bi-Directional LSTM Predictive Analytics
IEEE ICoECIT · 2026 · Faculty collaboration with Dr. Shilpa Sondkar ↗Professor & Head of Department of Instrumentation and Control Engineering, VIT Pune
Read the paper