Comparative study evaluates machine learning and deep learning models for fault detection in induction motors, indicating significant performance improvements.
Key Points
This study aims to improve fault diagnosis accuracy in induction motors under varying load conditions using advanced machine and deep learning techniques.
Developed a multi-modal signal analysis framework using stator current, rotor speed, and flux-induced voltage signals.
Collected a fifteen-class dataset of healthy and faulty motor states at different load levels (0%, 50%, 100%).
Evaluated machine learning models, including SVM and traditional approaches, alongside deep learning models like LSTM and TCN.
LSTM model achieved a perfect classification accuracy of 100% during training and validation phases.
SVM and TCN models showed perfect prediction results during deployment on unseen test data.
Recent architectures like Transformer demonstrated strong potential; their performance can be improved with hyperparameter tuning.