Fault Feature Extraction from Wind Turbine Gearbox Based on Non-Gaussian Intensity

Complex vibration signals of wind turbine gearbox were studied so as to extract the fault features. The measured vibration signals were filtered and reconstructed by Gabor transformation, and then the meshing frequency and its harmonic components were inhibited, after which bispectral analysis was carried out to the reconstructed signals that mainly include sideband components, random components and inherent frequency components. Finally, the difference of non-Gaussian feature between normal gear vibration signals and pitting gear vibration signals was revealed, and subsequently the characteristic values of non-Gaussian intensity were extracted. Results show that gear faults may change the distribution of non-Gaussian components in vibration signals. The characteristic values of non-Gaussian intensity extracted from filtered signals are closely related to the gear pitting faults. The method is very simple in calculation, which therefore may be used in actual engineering projects.