Implementasi Tinyml Berbasis Esp32 Untuk Prediksi Kerusakan Motor Aerator Tambak Udang Vaname Menggunakan Analisis Vibrasi Real-Time Di Pesisir Sumatera Utara

Abstract

The aerator motor is an important component in vaname shrimp ponds because it functions to maintain dissolved oxygen levels in the pond water. Undetected aerator motor damage can cause a decrease in water quality and increase the risk of shrimp mortality. This study aims to design and implement a TinyML-based predictive maintenance system using an ESP32 microcontroller and vibration sensors to detect aerator motor damage in real-time. Vibration data was obtained using an MPU6050 sensor in several motor conditions, namely normal, damaged bearings, misalignment, unbalance, and light overload. The data was then processed using a feature extraction method and trained using an Artificial Neural Network (ANN) model based on TensorFlow Lite for Microcontroller. The results showed that the system was able to classify motor conditions with an accuracy of 95.2%, precision of 94.8%, recall of 95.1%, and F1-score of 94.9%. The TinyML implementation on ESP32 was able to work in real-time with an average inference time of 1.2 seconds and low power consumption of ±2.8 Watts. This research proves that TinyML can be an effective and inexpensive solution for predictive maintenance of aerator motors in vaname shrimp ponds in coastal areas