IoT integrated real-time monitoring and AI-based forecast system for optimising tea withering process

Abstract

Tea withering is one of the most important steps in tea processing that directly affects the final quality of tea. Traditional withering procedures are based on manual visual observation and controlled environmental conditions, which often lead to moisture variations, energy consumption, and unstable tea quality. This study circumvents these limitations by proposing a new IoT-based monitoring and AI-based prediction combined system for optimising the tea withering process in real time. The objectives of this research are to develop a system for monitoring environmental conditions and tea leaf characteristics in real time, and to construct and train an AI model that can accurately forecast the optimal withering time. A system was developed using ESP32 microcontrollers, DHT11 and DS18B20 temperature sensors, HX711 load cell modules, and OLED displays. A Blynk-based web dashboard was deployed for remote visualisation, and a Flask server was used to deploy the AI prediction model. Over 300 withering cycles accumulated over two years were used to train time-series forecasting and regression models. Pilot results attest that real-time monitoring enables precise heating and fan control, and AI-driven prediction significantly improves the accuracy of estimating withering time based on initial moisture content, atmospheric temperature, and tray conditions as primary parameters. Pilot implementation confirmed that real-time monitoring maintained a 5–10 °C tray-to-air temperature difference, while the predictive model accurately forecasted withering duration between 13.4 - 18.3 hours. The combined system reduced operator intervention and energy wastage, offering a practical pathway toward automation in tea manufacturing. The system demonstrated scalability for factory-level applications, showing stable performance during continuous trials across two years and more than 300 withering cycles. Automated control of blowers and heaters ensured timely termination at 60% moisture, conserving energy and preventing over- or under-withering. These findings highlight the system’s potential to enhance sustainability and consistency in tea processing industries. This study demonstrates that the integration of IoT and AI provides a practical pathway for automation in tea manufacturing. The findings have significant implications for enhancing energy efficiency, improving product quality, and promoting sustainability in tea processing industries.