Research on denoising methods for rotating machinery bearing fault vibration signals based on RIME-VMD combined wavelet thresholding

Abstract

To address the challenges of noise interference and difficulty in extracting fault features from rolling bearing vibration signals, this study proposes a combined denoising method based on frost-ice optimization for adaptive VMD parameter optimization and wavelet thresholding. This method utilizes the frost-ice algorithm to optimize key VMD parameters, enabling adaptive decomposition of vibration signals. It identifies noise-dominated components by combining the energy proportion and correlation coefficient of the decomposed signals, applies wavelet soft thresholding to these components, and reconstructs the noise-reduced signal by integrating them with the retained effective components. To validate its effectiveness, simulated bearing fault experiments were conducted. Results demonstrate that this method effectively suppresses both high- and low-frequency noise. The denoised signals closely match the characteristics of clean signals. The SNR for the three operating condition fault signal types was 2.178 dB, 2.94 dB, and 3.76 dB, with RMSEs of 0.17, 0.125, and 0.026, respectively. These results outperform the other two methods, further validating that the method possesses superior universality and generalization capabilities, providing a data foundation for subsequent fault feature extraction and intelligent recognition.