کنفرانس بین المللی ماشین و محرکه های الکتریکی

صفحه اصلی / The 5th International Conference on Electrical Machines and Drives

Fault Detection and Classification in Induction Motors :An Explainable Convolutional Neural Networks Approach

نویسندگان :

َAli Vahidi¹ Amirata Taghizadeh² Mohammadreza Toulabi³

1- دانشگاه صنعتی خواجه نصیرالدین طوسی 2- دانشگاه صنعتی خواجه نصیرالدین طوسی 3- دانشگاه صنعتی خواجه نصیرالدین طوسی

کلمات کلیدی :

Convolutional Neural Networks،Explainable Artificial intelligence،Fault detection،Induction motors،Thermography

چکیده :

Effective fault diagnosis in induction motors is crucial for maintaining operational safety and efficiency in industrial settings. While deep learning models, particularly Convolutional Neural Networks (CNNs), have shown great promise, their inherent black box nature often hinders their adoption due to a lack of transparency and trust. This paper addresses this challenge by presenting an end-to-end, explainable diagnostic framework that leverages thermal imaging for non-invasive fault classification. We develop a tailored CNN architecture that automatically learns discriminative features from thermal images to distinguish between various motor faults. To make the model's reasoning transparent, we integrate Explainable Artificial Intelligence (XAI) through the Gradient-weighted Class Activation Mapping (Grad-CAM) technique, which generates visual heatmaps highlighting the exact image regions influencing the model's predictions. Simulation results demonstrate the framework's high effectiveness, achieving 97.3% accuracy across 11 operational conditions. Critically, the XAI visualizations confirm that the model's decisions are based on physically relevant thermal signatures, successfully identifying both concentrated hotspots and more subtle, distributed fault patterns. This combined approach provides a solution that is not only accurate but also trustworthy for industrial predictive maintenance.