Hybrid TinyML Sensor Detection for Microplastics

Microplastics are small plastic pieces that are now found in liquids like water and oil. Testing for them becomes difficult outside laboratories because most methods need costly equipment and controlled conditions. This study addresses this issue by examining a simple embedded sensing approach for microplastic monitoring. This proposed system is a simple, budget-friendly device that utilizes an ESP32 microcontroller, an everyday LED as a light source, and an LDR to measure changes in light for processing and potential connectivity. The entire setup is designed for cost-effective monitoring and data handling capabilities. When a liquid sample is placed in the sensing path, variations in light intensity caused by suspended particles are observed. These signals are processed directly on the device. Lightweight machine learning models are applied to identify frequently used polymer types, including Polyethylene and Polypropylene. All computation is carried out locally on the embedded system, and the output is displayed immediately without the use of cloud services or laboratory instruments. The experimental results indicate that the system can operate with low power consumption and is suitable for simple field testing. The work shows that compact TinyML-enabled optical sensing can be explored as a practical option for microplastic monitoring in resource-constrained environment.