Research On Feature Extraction Method Of Fault Signal Of Gearbox Bearing Of Wind Turbine

Due to the noise existing in the operating environment of the wind turbine gearbox bearing and the severe coupling vibration between different parts of the bearing,the failure characteristics of the collected vibration signals are not obvious,so it is difficult to directly complete the feature extraction and diagnosis of the wind turbine gearbox faulted bearing.For this reason,that an improved Variational Modal Decomposition method combines with the Singular Value Decomposition noise reduction technology was introduced in this paper.To accurately realize the fault feature extraction and diagnosis of wind turbine gearbox bearing vibration signal,a feature extraction model of wind turbine gearbox bearings based on Adaptive Variational Modal Decomposition-Singular Value Decomposition noise reduction was proposed.Mainly research work of the paper is as follows:(1)The impact of modal number and secondary penalty factor on VMD algorithm was studied.Thus,hybrid particle swarm algorithm was introduced to optimize the VMD parameters,then an Adaptive Variational Modal Decomposition model was constructed.It focused on selecting the minimum average envelope entropy as the fitness function in the optimization process of hybrid particle swarm algorithm and constructing the AVMD method specifically.Taking the measured vibration fault data of the high-speed shaft rolling bearing of the wind turbine gearbox as an example,the AVMD method was applied to initially detect the vibration signal.The results showed that the AVMD method can achieve signal decomposed completely without modal aliasing and under-decomposition,which laid the foundation for subsequent fault feature extraction and diagnosis of wind turbine gearbox bearing signals.(2)After the signal was decomposed by the AVMD method,several modal components were obtained.Aiming at the problem that the effective components selected by a single index was not accurate enough,this paper introduced the weighted kurtosis index to select the effective component and then reconstruct the signal.Aiming at the problem that the reconstructed signal was still mixed with some noise,the Singular Value Decomposition(SVD)noise reduction method was introduced and improved,when determining the number of effective singular values,the quadratic approximation principle method based on the average value was proposed.Therefore,this paper constructed the model for extracting fault features of wind turbine gearbox bearing signals based on AVMD-SVD.(3)Taking the bearing fault data provided by the laboratory as an example,the fault feature extraction method based on AVMD-SVD was divided into three stages and was analyzed.In the stage of AVMD method initially decomposing the vibration signal,it was compared with the traditional center frequency method.In the signal reconstruction stage,it was compared with the reconstructed signal based on the kurtosis index qualitatively and quantitatively.In the noise reduction processing stage,it was compared with AVMD and EEMD methods about the effect of feature extraction.The overall effectiveness of the AVMD-SVD method was verified from the above three stages,and the characteristic frequency of the inner ring fault was successfully extracted.Then,vibration fault data of the wind turbine gearbox high-speed shaft rolling bearing from a wind farm was taken as an example to carry out the AVMD-SVD method.The result showed that the composite fault features of outer ring-inner ring-rolling elements were successfully extracted,and the result is consistent with actual failure conditions.The result verified the feasibility and effectiveness of the AVMD-SVD method in extracting the fault features of wind turbine gearbox bearings signal.