امير صادقي

Electric motors play a crucial role in various industries as one of the key components ensuring system functionality. Failures and malfunctions in these motors can lead to production downtime, financial losses, and even safety hazards. Early and accurate fault detection in electric motors is of great importance, as it helps prevent unexpected failures and their negative consequences. In this research, transfer learning has been employed for fault diagnosis in electric motors. This method leverages pre-trained models to solve a new problem by utilizing existing knowledge. Specifically, deep pre-trained neural networks have been used for extracting features from motor current signals, followed by classification to diagnose faults. In the first part of this study, fault detection in the bearing of a three-phase induction motor was examined using motor current signal analysis in an experimental setup. The performance of AlexNet, EfficientNetB0, and ResNet50 transfer learning algorithms was eva‎luated. The results demonstrated that EfficientNetB0 outperformed the other two models, providing higher accuracy in fault detection. In the second part, the current signal of apron station needle machines was analyzed to detect mechanical faults. This analysis was conducted using AlexNet and EfficientNetB0 transfer learning models, and the results again showed that EfficientNetB0 outperformed AlexNet. Furthermore, various transfer learning scenarios were examined in both cases, and their impact on model accuracy and efficiency was analyzed. The findings indicate that transfer learning can be effectively used for fault diagnosis in electric motors, achieving high accuracy and low error rates. Compared to traditional fault detection methods, this approach offers advantages such as higher accuracy, faster processing speed, and reduced dependence on large training datasets. This research highlights that transfer learning serves as a powerful tool for diagnosing faults in electric motors and industrial machinery. It can be utilized in various industries to enhance the reliability and efficiency of electrical equipment.

يادگيري انتقالي , تشخيص خطا , موتورهاي الكتريكي , شبكه عصبي عميق

Transfer learning , Fault diagnosis , Electric motors , Deep neural networks

Amir Sadeghi

Ahmad Mirabadi