AI powered health monitoring for electrical machines using machine learning
Loading...
Date
2025
Journal Title
Journal ISSN
Volume Title
Publisher
UMT Lahore
Abstract
This project develops and demonstrates a low-cost AI powered health monitoring system for a single-phase capacitor-start induction motor using an Aurdino UNO microcontroller and sensors (ACS712 for current, ZMPT101B for voltage, DS18B20 for winding temperature and SW-420 for vibration pulses). The firmware acquires synchronized measurements in short sliding windows, computes electrical, thermal and vibration features, and applies a simple, interpretable random Foresto classify the motor state as healthy or faulty in real time. Sensor data and health status are streamed to a Google Sheet for historical logging and displayed on a Blynk mobile dashboard for remote visibility and alerts. Controlled tests with induced overload, restricted cooling and imbalance scenarios show that the system detects abnormal conditions within a few seconds, maintains stable readings during normal operation. The results confirm that an explainable edge AI approach can provide timely, reliable condition monitoring for motors without expensive industrial hardware, laying a practical foundation for future multi-class diagnostics and predictive maintenance enhancements.