Research On Fault Diagnosis Of EMU Bogie Bearing Based On Improved Kurtogram

With the development of high-speed railway industry,the requirements for EMU safety and reliability of key systems and components are increasing.Bogie bearing system as the most critical rotating component in EMUs undertakes such as bearing,guiding,damping,driving functions.,its running state is related to the safety and reliability of bogies directly.Bogie bearings are working in complex conditions which are often damaged,so there is great significance to diagnose the faults of bogie bearings of high-speed trains.In this paper,bearing of EMU bogie is taken as the research object,and the current situation of typical bearing fault diagnosis technology algorithms is summarized and analyzed.At the same time,the vibration mechanism and characteristic frequency acquisition of EMU rolling bearing are explained.Firstly,the performance of minimum entropy convolution,variational mode decomposition and fast kurtosis graph algorithm is analyzed using the simulation signals.Secondly,the test of bearing fault of EMU bogie is carried out,and the application of minimum entropy convolution,variational mode decomposition and fast kurtosis graph algorithm in bearing fault diagnosis is compared using test data,there shows the above algorithms are insufficient for bearing fault diagnosis with strong background noise.Then,in order to realize the effective diagnosis of bearing fault,an improved kurtogram is proposed in frequency domain.Witch adaptively determine the resonance frequency band in bearing fault information by calculating power spectrum amplitude kurtosis of a specific frequency band.Based on the identified resonant frequency band,the bearing vibration signal is bandpass filtered,and the frequency spectrum of the filtered demodulated signal is analyzed,to diagnose bearing fault.The algorithm overcomes noise,as well as better computational efficiency.Finally,the improved kurtogram is verified on simulation data and test-bed data,and compared with minimum entropy convolution,variational mode decomposition and fast kurtogram.The experimental results show the effectiveness and superiority in bearing fault diagnosis..