Predictive Maintenance of Electric Motors Using Machine Learning, IoT and Vibration Analysis

Abstract

This research presents the design and development of an IoT-based predictive maintenance system for electric motors aimed at reducing unscheduled industrial downtime and maintenance costs. The system utilizes an ESP32 microcontroller integrated with a high-precision industrial-grade accelerometer to monitor three-axis vibration signatures in real-time. Raw vibration data is acquired via the Serial Peripheral Interface (SPI) protocol and transmitted to a host workstation where a Convolutional Neural Network (CNN), developed using the MATLAB Deep Learning Toolbox, performs real-time feature extraction on RGB-mapped spectrograms generated through Short-Time Fourier Transform (STFT). The model is trained using supervised learning to distinguish between 'Normal' and 'Faulty' operational states—simulated during testing via uneven motor mounting—to provide early warning signals. Remote monitoring and instant fault notifications are facilitated through the Blynk IoT framework, which provides a live dashboard for vibration visualization and mobile alerts to ensure prompt technical intervention. Experimental results validated the system's reliability in accurately identifying anomalies and predicting potential failures, demonstrating that an affordable yet high-performance diagnostic tool can be successfully implemented for smarter industrial maintenance.