Phase-Coherent Diversity Entropy for the Fault Diagnosis of Rolling Bearing

Fault diagnosis technology serves as a critical safeguard for ensuring the reliable operation of rotating machinery. In the fault diagnosis framework, the diversity entropy is a promising tool for the feature extraction and signal processing. However, the diversity entropy suffers from the incomplete feature characterization due to the conventional multiscale decomposition prioritizes low-frequency signal components while neglecting high-frequency information. To solve this problem, this paper proposed Phase-Coherent Diversity Entropy (PDE). The proposed PDE integrates a dualresolution wavelet framework combining Haar (db1) and Daubechies-4 (db4) wavelets. The db4 wavelet significantly improves frequency-domain energy concentration by incorporating smoother basis functions, effectively overcoming the problem of frequency-domain dispersion associated with the Haar wavelet. At the same time, the multi-phase method is integrated into DE, which preserves the information of the subsequences after extraction while ensuring that there is only one entropy value at each node, thereby avoiding feature redundancy. Experiment results show that the proposed PDE performs the best in bearing fault cause recognition compared with other multiscale methods.