The Fault Feature Extraction Method Of Train Wheelset Bearing Based On VMD

Wheelset bearing is one of the key parts of the train running section,when its failure occurs,the normal operation of the train will be affected in the light case,but it will cause the train derailment,casualties and loss of great interests in the serious case.Therefore,it is of great significance to carry out the research on fault diagnosis of the train wheelset bearing.In this paper,Variational Mode Decomposition(VMD)method is used to extract bearing fault features.In view of the parameter sensitivity of VMD method,the number of modes and bandwidth parameters of VMD are optimized by the optimization algorithm and the correlation kurtosis index,And an improved fast spectral correlation-square fast spectral correlation index was used to improve the initial center frequency of VMD.The effectiveness of the proposed algorithms are verified by train bearing tests.The main research contents of this paper are as follows:(1)The VMD need to set parameters,which affect the result of decomposition directly,so the gray wolf algorithm is proposed to optimize the mode number and bandwidth parameters of VMD,and the correlation kurtosis index is used as the fitness function.The optimization effect was verified by bearing fault simulation signal.(2)In order to further improve the decomposition efficiency of VMD,based on the sensitivity of spectral correlation to nonstationary periodic components,an improved fast spectral correlation-square fast spectral correlation is proposed to locate the initial center frequency of VMD,and the bandwidth parameters are adjusted adaptively by the correlation kurtosis.The analysis shows that this method solves problems of the traditional VMD in the optimization of the modes number and the selection of modes,which makes the convergence speed of decomposition faster,improves the decomposition efficiency of VMD and can extract the bearing fault features quickly and accurately.(3)The train wheelset bearings are taken as the experimental object to carry out typical bearing failure experiments.The fault signals of inner ring,outer ring and compound fault signals under various working conditions are collected to verify the effectiveness of the proposed method.The research results show that the improved algorithm proposed in this paper has great advantages in computational efficiency and accuracy compared with similar methodsThe research results have certain theoretical and practical value for developing rolling bearing fault diagnosis methods and promoting the improvement of train bearing operation and maintenance technology.