Research On Fault Diagnosis Of Railway Vehicle Rolling Element Bearings Based On VMD-SVM

In recent years,the rapid development of China's railway plays an irreplaceable role in the development of national economy.As an important part of the bogie of railway train,the health of rolling bearing is closely related to the safety of railway operation.Bearing failure occurs frequently because of many damage inducements,bad working conditions and discrete life of rolling bearing,which resulted in a large number of social and economic losses and damaged safety benefits.Taking rolling bearing as the research object,a fault diagnosis algorithm based on the variational mode decomposition method and support vector machine method is proposed in this thesis.The main works done are as follows:(1)The basic composition of rolling bearing is studied,and the failure mechanism,basic mode and failure vibration frequency of rolling bearing are analyzed theoretically and deduced.On this basis,the typical fault diagnosis methods of rolling bearing are analyzed,and the advantages and disadvantages of various methods are summarized.(2)This thesis studies the algorithm of VMD,and makes a comparative analysis with the classical empirical mode decomposition algorithm.On the basis of summarizing the former VMD algorithm,the existing problems of the current algorithms are analyzed and the solutions are found out.(3)A variational mode decomposition algorithm based on parameter optimization is proposed.In view of the problems existing in the current VMD algorithm,starting with the two most important parameters affecting the algorithm: mode number k and penalty factor ?,the swarm algorithm with better optimization effect,grasshopper optimization algorithm and VMD algorithm,are selected to combine,and the two parameters are selected for adaptive optimization to achieve the best decomposition effect.(4)In view of the fitness function of the optimization algorithm and the selection of the modal components after the decomposition of the variational modes,the correlated kurtosis index is proposed as the selection standard of the fitness function and the modal components of the optimization algorithm,and compared with the previous kurtosis index and the correlated coefficient index.The concept and advantages of correlated kurtosis are discussed.(5)This thesis takes the variational mode decomposition algorithm as the preprocessing method of rolling bearing fault signal.Subsequently,we use the singular value decomposition method to handle signal after preprocessing and extracts the singular value as the classification feature,uses the support vector machine method which has good effect on small sample data classification to carry out the final fault classification.(6)The verification results from the simulation signal and rolling bearing fault database of Case Western Reserve University show that the rolling bearing fault diagnosis method proposed in this paper has strong feasibility and high fault recognition rate,and has certain research significance.