Anomaly Detection in Industrial Motors using An Artificial Intelligence-Enabled Low Power Portable Device

The increasing adoption of Industry 4.0 has emphasized the need for intelligent predictive maintenance solutions to prevent industrial motor failures.This study presents a low power, edgebased anomaly detection system deployed on an Arduino Nano RP2040 Connect, leveraging a Convolutional Neural Network (CNN) for real-time fault detection.The system utilizes acoustic data from the Hitachi MIMII motor fan dataset, processed into Mel-filterbank energy (MFE) features for classification.The trained 8-layered CNN model, converted into TensorFlow Lite (TFLite) format, achieves 91.80% accuracy with an AUC score of 0.92, demonstrating high reliability in detecting anomalies.Unlike cloud-based approaches, the embedded deployment ensures low latency, reduced network dependency, and enhanced privacy.This research highlights the feasibility of deploying deep learning models on resource-constrained microcontrollers for industrial predictive maintenance, offering a cost-effective, scalable, and real-time monitoring solution.