Research On Sound Characteristics And Fault Diagnosis Method Of Axle Box Bearing Of High-speed Train

Axle box bearing is one of the most important components of the bogie of rail transit train,which can transform the rotary motion of the wheelset into the translation along the rail direction,and also bear the load of the car body.Under the coupling effect of long-time highspeed rotation and the heavy load,the axle box bearing is prone to failure because of its complex operating environment.Once the failure occurs,the safety of the train will be greatly affected.Therefore,it is very important to monitor the health of axle box bearing.At present,vibration data,temperature data and sound data are often used to monitor the health status of axle box bearing.This dissertation is mainly based on the sound data to study the fault diagnosis algorithm of axle box bearing.The main research work is as follows:(1)The vibration and noise mechanism of axle box bearing is analyzed,the calculation formula of fault frequency of axle box bearing and the basic characteristic frequency under various fault conditions are listed.Through the decomposition of bearing structure,the vibration caused by different parts of bearing and the effect on the superposition of sound pressure are analyzed.The numerical simulation model of bearing sound signal is established,which can extract the sound data of bearing under different fault conditions,and establish the simulation model database.(2)Based on the original high-speed train axle box bearing test-bed of the research group,the test-bed has been rebuilt and reconstructed,which greatly improves the use and replacement efficiency of the fault axle box bearing.The test-bed can add different vertical static loads,adjust the bearing speed and excitation frequency in real time,and construct complex and changeable test conditions.Finally,a sound database of axle box bearing fault is established.(3)A method of fault feature extraction of sound data of axle box bearing based on spectrum whiteness and energy operator is proposed.Firstly,the noise of the sound signal is reduced by AR model filter,and then the signal is whitened in the frequency domain to eliminate the influence of discrete frequency.Combined with its average energy value,the whitened signal is constructed.Finally,the characteristic parameters of the data are extracted by the energy operator demodulation,and the fault location is determined by comparing with the theoretical eigenvalue.(4)Aiming at the noise interference frequency in the spectrum whitening energy operator demodulation algorithm,a method combining deconvolution filtering and cross-correlation spectrum is proposed to extract the bearing sound data characteristic parameters.Firstly,the original data is denoised by deconvolution filter,which improves the signal-to-noise ratio,the Hilbert envelope and energy operator of the filtered signal are calculated,and then the crosscorrelation analysis of the two demodulation signals is carried out,and the characteristic parameters of the sound data are found from the spectrum diagram.The validity of the diagnosis method is verified by using the model simulation data and the experimental data collected.