Research On Fault Diagnosis Method Of High-speed Train Axle Box Bearing Based On Scale Space Theory

With the rapid development of the rail transit industry,the number of high-speed EMUs operating on the line has increased rapidly.It is urgent to monitor the condition of high-speed EMUs,especially the axle box bearing often appears peeling,pitting and corrosion under the harsh service environment and alternating load.By installing vibration monitoring equipment at the axle box,the service performance of axle box bearings can be effectively monitored online.However,the vibration signals collected at the axle box include wheel-rail impact signal,fault bearing signal and ambient noises.Because of the complexity of the signal,the difficulty of fault diagnosis is greatly increased.Hence,it is very necessary to study a practical and feasible method to realize weak fault signal diagnosis.In this thesis,the theory of scale space is studied in depth.From the perspective of improving computational efficiency and reducing computational time,the construction of scale space,identification of resonance frequency band and online monitoring are discussed in depth.The main research work of this thesis is as follows:In this thesis,the empirical wavelet transform theory and scale space representation theory are introduced in detail.Firstly,the theory of empirical wavelet transform is introduced.It combines the adaptability of the EMD and the fast transformation characteristics of wavelet transform,and can realize the adaptive decomposition of signals.Then the construction of Gaussian kernel function,scale space plane and the determination of threshold are introduced.By using the simulated signals to analyze the characteristics of the EWT,the simulation results show that the frequency band division of the algorithm is obvious,resulting in the increase of signal components and low computational efficiency.Secondly,because the method of constructing scale space based on Gaussian kernel function has the characteristics of excessive frequency band division and low computational efficiency,a method is proposed to realize bearing fault diagnosis based on morphology by constructing scale space.Firstly,the feasibility of constructing scale space planes with triangular structure elements and flat structure elements are verified by using the simulated signals;the result proves the validity of the flat structure elements to construct the scale space plane;Finally,the algorithm is verified by using the signal of a single-axle rolling vibration test rig.Finally,this thesis proposes a bearing fault diagnosis method based on the combination of scale space and spectral kurtosis.First,in the construction of scale space plane based on morphology,the maximum number of cycles of each extreme point is calculated,which is the length of scale space curve;Then the curves are sorted according to the lengths of the curves,and the scale space curves are selected according to the number of decomposition layers to set the node division of each layer;Then,the simulated signal is used to analyze the kurtosis,correlation kurtosis to characterize the impact and periodicity of each sub-band signals,and it is proved that the correlation kurtosis can still effectively identify the bearing fault frequency band under the wheel/rail impact,so as to realize the bearing fault diagnosis.Finally,the algorithm is verified by using the data of a single-axle rolling vibration test rig.