Study On The Extraction Method Of Wind Turbine Gearbox Fault Feature

With wind turbine single unit capacity increased, the load bearing of the gearbox is also increased, coupled with the complex and changeable conditions, resulted in frequent faults of wind turbine gearbox. In this paper, through built the wind turbine gearbox fault test bench, the vibration signals and current signals of gear distributed fault and partial fault were extracted. Through analysis of vibration domain factor(peak to peak value, kurtosis) and amplitude spectrum, the preliminary fault has been judged.In order to extract gearbox fault feature further, the bispectrum analysis method was used. Bispectrum had the properties of detecting quadratic phase coupling phenomena and suppressing Gauss noise. The vibration signal of gearbox was divided into several components by its Gaussian, and the contour map and the waterfall was drawn using bispectrum, the fault characteristic frequency of distributed fault and local fault can be found respectively in the contour map. As the same fault, bispectrum peak was decreased as the strengthened of modulation signal with the fault development. Moreover bispectrum can fully shield noise and extract fault feature of gear box effectively.The diagnosis of gearbox based on vibration signal is difficult due to the vibration sensor installation inconvenience and the less relevance between fault signal and fault feature, so the idea of using the generator side current signal to detect gearbox fault was put forward. Firstly, the current signal model was established under the condtion of gearbox fault. Similar to the vibration signal, the current signal have the same modulation phenomenon, and the frequency band spacing was the fault characteristic frequency.Secondly, the changing regular of the generator side current signal while gearbox was fault was analyzed in the frequency domain. Based on this, the bispectrum picture and tangent spectrum picture of the pitting corrosion development fault(one to eight surface) and broken development fault(one to four surface) of planetary gearbox was drawn. For the pitting corrosion development fault and broken development fault, the fundamental frequency peak value of current signal was decreased with the increase of the number of fault surface. while the fault feature frequency peak value is increased. however, the fault was developed to distributed fault for the pitting corrosion of seven surface and the eight surface, the fault feature frequency peak value of current signal was increased. Therefore, gearbox fault could be extracted effectively by using the generator side current signal. The method provides a new way for the diagnosis of the actual wind farm fault diagnosis.