Research On Fault Feature Extraction Of Wind Turbine Drive Trains Based On Spectral Kurtosis

Wind turbine has the characteristics of complicated load and poor operating environment,so the failure rate and maintenance costs are high.Among them,the problem of the transmission chain of the wind turbine is particularly prominent.According to the characteristics of wind turbine drive system,the new theory and method of fault feature extraction is studied,which is important to ensure the healthy operation of wind turbines.In the diagnosis and analysis of the actual vibration signal of the wind turbine,the classical vibration signal processing methods can't capture the accurate information which reflects the fault characteristic,which affects the diagnosis and analysis of the wind turbine fault.In this paper,the method of vibration signal analysis and fault feature extraction based on spectral kurtosis is studied,which combines the minimum entropy deconvolution(MED)and the maximum correlation kurtosis deconvolution(MCKD)to analyze the vibration signal.And extract the fault feature information from the vibration signal under strong background noise.It is verified by applying the measured signal of wind turbine transmission chain.The main contents and conclusions are as follows:(1)Study on the spectral kurtosis theory,two kinds of traditional spectral kurtosis algorithms are introduced,compared analyzed the wind turbine simulation signal and measured signal,Study shows characteristics and shortcomings of spectral kurtosis.(2)In the strong noisy environment,the spectral kurtosis is difficult to extract the fault feature.Discussed the AR-MED and spectral kurtosis,MCKD and spectral kurtosis combination of weak fault feature to extract part of the method respectively.Firstly,the signal is filtered and pretreated,the noise component is suppressed,and the impact characteristic is improved.Then,the kurtosis of the signal after noise reduction is calculated,and the characteristic frequency of the fault is extracted by envelope demodulation.The method is tested with the measured vibration signals of the wind turbine,and the results show that the early fault feature extraction ability is improved.(3)According to the vibration data of generator bearings from normal to irregular,the data are screened at equal time interval,and the fault feature is extracted by using the above research methods,analyze the trend of the amplitude of fault characteristic.Compared with the change trend of the time-domain eigenvalue,it reflects the practical value of the research methods in the early fault diagnosis of the equipment.