Bearing Fault Detection Using Continuous Wavelet Transform and VGG۱۶-Based Transfer Learning

This paper proposes a method for bearing fault detection using transfer learning and Continuous Wavelet Transform (CWT). Vibration signals are transformed into time-frequency images using CWT, enabling the application of pre-trained convolutional neural networks (CNNs) such as VGG۱۶, ResNet, and Inception for fault classification. The dataset, collected from simulated faults in rotating machinery, provides diverse fault scenarios for evaluation. By leveraging transfer learning, models pre-trained on large datasets are adapted to the fault detection task, overcoming data limitations. Performance metrics, including accuracy, precision, recall, and F۱ score, were compared across the models. Results show that VGG۱۶ outperforms ResNet and Inception in all metrics, demonstrating its effectiveness in fault classification. ResNet showed moderate performance, while Inception struggled with lower accuracy. This study highlights the potential of combining transfer learning with CWT for reliable and efficient fault detection, offering a valuable approach for predictive maintenance in industrial settings.

Keywords:

Fault Detection , Transfer Learning , Continuous Wavelet Transform