Application of cyclic coherence function to bearing fault detection in a wind turbine generator under electromagnetic vibration

In a wind turbine generator, there is an intrinsic electromagnetic vibration originated from an alternating magnetic field acting on a low stiffness stator, which modulates vibration signals of the generator and impedes fault feature extraction of bearings. When defects arise in a bearing, the statistics of the vibration signal are periodic and this phenomenon is described as cyclostationarity. Correspondingly, cyclostationary analysis enables finding the degree of cyclostationarity representing potential fault modulation information. In this paper, the electromagnetic vibration acting as a disturbance source for fault feature extraction is deduced. Additionally, the spectral correlation density and cyclic coherence function used for vibration analysis are estimated. A real 2 MW wind turbine generator with a faulty bearing was tested and the vibration signals were analyzed separately using conventional demodulation analysis, cyclic coherence function, complex wavelet transform and spectral kurtosis. The analysis results have demonstrated that the cyclic coherence function can detect the fault feature of inner race successfully, while the feature is concealed by intensive electromagnetic vibration in the other three methods. The disassembled bearing of the wind turbine generator illustrates the effectiveness of the analysis result, and precautionary measures for protecting bearings in generators are suggested.