Application of VMD and dynamic wavelet noise reduction techniques in rolling bearing fault diagnosis

Abstract

This study proposes a denoising technique based on VMD and dynamic wavelet to address the difficulty of extracting rolling bearing fault vibration signal features in a noisy background. Firstly, the vibration signal is decomposed using the VMD method and the extracted IMFs are filtered. Then, dynamic wavelet are used to denoise the confused IMFs, and the IMFs are reconstructed to achieve joint denoising. Finally, a deep learning method is applied to adaptively extract the features of the noise-reduced signals for fault diagnosis. The effectiveness of the method is confirmed through experimental verification, demonstrating its ability to remove noise information from the signal and improve the presentation of fault characteristics, leading to improved accuracy in fault diagnosis.