Fault Diagnosis of Wind Turbine Bearings Based on CEEMDAN-GWO-KELM

To solve the problem of fault signals of wind turbine bearings being weak, not easy to extract, and difficult to identify, this paper proposes a fault diagnosis method for fan bearings based on Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) and Grey Wolf Algorithm Optimization Kernel Extreme Learning Machine (GWO-KELM). First, eliminating the interference of noise on the collected vibration signal should be conducted, in which the wavelet threshold denoising approach is used in order to reduce the noise interference with the vibration signal. Next, CEEMDAN is used to decompose the signal after a denoising operation to obtain the multi-group intrinsic mode function (IMF), and the feature vector is selected by combining the correlation coefficients to eliminate the spurious feature components. Finally, the fuzzy entropy for the chosen IMF component is input into the GWO-KELM model as a feature vector for defect detection. After diagnosing the Case Western Reserve University (CWRU) dataset by the method presented in this research, it is found that the method can identify 99.42% of the various bearing states. When compared to existing combination approaches, the proposed method is shown to be more efficient for diagnosing wind turbine bearing faults