Research On Wayside Acoustical Fault Diagnosis For Train-Wheel Bearings When The Train Is Running At A Non-uniform Velocity

The train bearing with the type of NJ(P)3226X1which is the type employed in our country, is selected to be the research object in our study. The main topic in this paper is the wayside acoustic health monitoring technique, involving two main issues: the distortion removal for the recorded signal and the fault diagnosis strategy for the train bearing.The wayside acoustic signal generated by the train bearing is obtained. The reason and characteristics of the "flow field distortion" is studied.A time domain interpolation resampling method is proposed based on the Morse acoustic theory and interpolation method. Through this method, the frequency-domain distortion as well as the time domain distortion can be both corrected, such that the distored signal obtains a maximum recovery.A data-driven Doppler effect elimination strategy is investigated in this paper. Firstly, the error of the existed formula has been pointed out. Then the construction parameters have been redefined and a "time-amplitude mapping interpolation method" is introduced to construct a Doppler-let. Secondly, based on Doppler-lets and matching pursuit method, the wayside-model parameters are identified by optimizing in the parameters set. Non-uniform operating condition is also considered in this method, so that the Doppler-shifted signal can be corrected much more accurately, especially for subway trains running at non-uniform speed.The instantaneous frequency estimation method and the signal resampling method are employed in this paper to correct the Doppler-shifted signal. The narrowband signal is obtained by band-pass filtering, and then the instantaneous frequency of the narrowband signal is calculated by Hilbert transform, and finally the Doppler-shifted signal is resampled according to the obtained instantaneous frequency curve. Both the advantages and disadvantages of this method are analyzed and summarized by comparing with another method.Two fault diagnosis strategies based on the corrected signal are introduced in this paper. One strategy is as follows:firstly the EEMD method is employed to extract the Intrinsic Mode Function (IMF) containg the fault information, then, the resonance demodulation method is employed to extract the feature. The other strategy involves the transient model analysis, the feature is extraced by matching the parametric periodic Laplace wavelets with the corrected signal.