Wayside Bearing Fault Diagnosis Based On Doppler Effect Removal

All along, the failure of bearing is one of the main types of train failure and has a serious impact on driving safety of the train. Therefore, to strengthen the train bearing condition monitoring and fault diagnosis is very important for the safe operation of trains. The wayside acoustic bearing monitoring system has many advantages. However, the sound of the moving train has a Doppler distortion which make the bearing fault diagnosis difficultIn this paper, we choose the NJ(P)3226X1bearing to study. Our group design a special experimental platform for the bearing and conduct in-depth research on Doppler effect removal. Four methods of Doppler removal is proposed and provide some useful ideas for the development of the wayside acoustic bearing monitoring system.A Doppler correction method based on the instantaneous frequency estimation is proposed. We extract the instantaneous frequency by using the STFT of the signal, According to Morse theory, we can achieve the nonlinear fitting of instantaneous frequency and then establish the resampling time vector using the fitting signal. The Doppler distortion can be removed effectively by resampling.A Doppler removal method in time domain is proposed. According to the kinematic relation and Morse theory, we can get the discrete time vectors and Amplitude modulation formula. Then the amplitude modulation and Interpolation fitting is carried out for the distorted signal The results of the simulation signal and experimental signal show that the Doppler distortion is well corrected.For the sake of removing the Doppler distortion of original signal which has multiple sources, a new method based on Dopplerlet transform and re-sampling is proposed. Firstly, find the primary functions-Dopplerlet atoms in parameters space. Secondly, get the instantaneous frequency of the Dopplerlet atoms based on Morse theory and Doppler effect. Finally, establish the resampling time vector in time domain which can eliminate the Doppler distortion.In the last method, we introduced a parametric wavelet called PMDW which is used to identify the parameters of the motion model The Doppler eliminator is used to remove the Doppler effect. Then the fault frequency can be extracted through transient model analysis. In this method, there is no need to measure the parameters and it can be adapted to various types of trains.