A Fault Identification Method for Rotating Machinery Based on CEEMDAN-MDE-PCC-EWT and RCMDE

Abstract

Addressing the challenge of severe noise interference in rotating machinery, which impedes effective fault feature extraction from nonlinear and non-stationary vibration signals, this paper proposes a novel fault identification method based on secondary decomposition and Refined Composite Multiscale Dispersion Entropy (RCMDE). First, the original vibration signal is decomposed using Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) to obtain Intrinsic Mode Functions (IMFs). Mixed evaluation indices, combining Multiscale Dispersion Entropy (MDE) and Pearson Correlation Coefficient (PCC), are then constructed to separate noise-dominant from signal-dominant IMFs. Subsequently, Empirical Wavelet Transform (EWT), with parameters optimized via the Grey Wolf Optimization (GWO) algorithm, is applied for signal filtering and reconstruction. Finally, RCMDE features are extracted from the reconstructed signal and input into a Support Vector Machine (SVM) for fault classification. Experimental results demonstrate the method's excellent performance in both simulation studies and tests on public bearing datasets. Notably, for the highly noisy planetary gear fault diagnosis task, the method significantly enhances the signal-to-noise ratio (SNR) by at least 3.14 dB and achieves a fault classification accuracy of 98.61%, outperforming benchmark methods. This research provides valuable insights for rotating machinery fault diagnosis.