Research On Compound Fault Feature Extraction For Wind Turbine Drive Train

Condition monitoring and fault diagnosis are significant for the security operation of wind turbines.Suffering from harsh operational environment and impact load,compound faults(e.g.gear-bearing fault,multiple gears fault)arise frequently in the wind turbine drive train with complex structure.When compound faults emerge,the weak fault is usually concealed by obvious fault or intensive noise,which can not easily be found by rountine inspection.However,the weak fault is significant incentive of obvious fault or the failure of drive train,and its fault characteristic need be timely extracted.Therefore,it is necessary to study failure mechanism,identify the weak fault information of multi-component and master the healthy state.These measures are benefical to making preventive maintenance schedule and avoiding catastrophic results,which can promote the development of wind energy with high efficiency and relibility.In this dissertation,in order to improve the accuracy of fault detection,some advanced compound fault characteristic extraction methods are studied based on the structure of wind turbine drive train,and the modulation model of faulty components in drive train.The main contents are listed as following:(1)The structures of double-fed and direct drive wind turbines are respectively analyzed,and the fault characteristic frequencies of the critical components in wind turbine gearboxes are computed.The modulation principle of faulty part in wind turbine drive train and the feature of corresponding power spectrum density are researched.A novel modulation model for distributed fault of planetary gear in wind turbine gearbox is presented.According to the vibration data from a renewable energy corporation,the fault mechanisms and fault representations of six typical faults of wind turbine drive train are analyzed,providing the foundation for future research.(2)Considering the compact structure of wind turbine gearbox,multi-components with intensive coupling,wide vibration frequency band,and large difference of vibration energy at different gear stages,the cepstrum based method is studied to identify the neighbor fault frequencies,and the complex wavelet based fault characteristic method is proposed to match different fault components and realize envelope demodulation simultaneously.A method combining cepstrum editing with complex wavelet transform is proposed to detect the weak bearing fault information hidden by intensive vibration energy of faulty gears.Comparing the conventional envelope demodulation analysis in detecting the weak bearing fault,the proposed method combining cepstrum editing with complex wavelet transform is superior due to its ability of eliminating obvious periodic components and multi-scale demodulation.(3)Massive vibration signals of wind turbine drive train are being generated each day.For these signals,conventional demodulation analysis need artificially design bandpass filters,and wavelet analysis need select wavelet basis function by hand,which bring inaccurate diagnosis results and low efficiency.The empirical mode decomposition for adaptive fault characteristic extraction is reviewed.The Fourier spectrum in empirical wavelet transform can be divided into different parts based on local maxima segmentation and scale space segmentation.And the empirical law is introduced to realize parameterless in empirical wavelet transform.Empirical wavelet transform is utilized to analyze the vibration signal with compound gear faults,and detect multiple rotating frequencies denoting ordinary gear fault and meshing frequency of planetary gear denoting defects in planetary stage,which demonstrates the effectiveness of proposed vibration model of faulty planetry stage.A margin factor based method is proposed to sort the decomposed empirical modes,and the empirical mode with maximum margin factor is regarded as the critical component including fault information.The analysis of faulty bearing in wind turbine generator verifies the effectiveness of the proposed method.(4)The vibration signals of bearing in wind turbine generator are easily polluted by the electromagnetic disturbance,unbalance and misalignment between high speed shaft of gearbox and generator shaft.In light of this,the characteristic of cyclostationary theory is studied,and a fault characteristic extraction method based on cyclic coherence fuction and cyclic ratio coefficient is proposed.The faulty bearing signal under intensive electromagnetic vibration is analyzed by complex wavelet transform,spectral kurtosis and cyclic coherence function.Among of these,cyclic coherence function can match second cyclostationary characteristic of faulty bearing in wind turbine generator,and the fault characteristic can be evidenced by cyclic ratio coefficient.Another faulty bearing signal under intensive rotating frequency and its vibration signal six months earlier are analyzed by cyclic coherence function,demonstrating the effectiveness of cyclic coherence in detecting severe fault and incipient fault.