Research On The Key Techniques Of The Wayside Acoustic Defective Train Bearing Detector System

The defect of the roller bearing is the dominant type of fault for a train, which lead to serious accidents and significant costs for the rail transport industry. Consequently, to develop the techniques of condition monitoring and fault diagnosis for the train bearings is indispensable. The wayside acoustic defective bearing detector system has attracted widespread attention recently for its particular advantages in fault diagnosis for the train bearings. However, some characteristics of the wayside acoustic signal, such as steep distortion, strong noises and multi-sources make the construction of that system difficult.The train bearings with the type of NJ(P)3226X1which is the dominated type in China, is selected to be the research object in our study. A test bench was designed for this type of bearings to achieve the acoustic signal. Then the core of this paper is to study the methods to overcome some of the technical difficulties which include the removal of Doppler Shift and the filtering of the noises in the wayside acoustic defective bearing detector system. Furthermore, some improvements and verification have been discussed. The whole work is aimed to provide some basic theories and ideas to the development of the system eventually.Firstly the analysis of the train bearings was discussed. The modal analysis based on ANSYS for all the components of the train bearing had been carried out. The vibration of each component was excited by the hammer with an accelerometer. The respondent acoustic signals were acquired by the microphone and the natural frequencies were obtained, which provides the reference data for further research in fault diagnosis of the bearing. Meanwhile, under the consideration of the rotation speed and load, the test bench was designed. The artificial cracks had been set by the wire-electrode cutting machine on the outer race, inner race and the roller respectively to acquire the acoustic signal of those three typical types of the bearing fault.A new Doppler Shift removal method based on the instantaneous frequencies (IFs) evaluation for the train bearings is proposed. Specifically, the IF estimation based on short time Fourier transform (STFT) is applied to attain the IF vector. According to the acoustic theory of the moving source, the data fitting is then carried out to achieve the fitting IFs with which the re-sampling sequence can be established as the re-sampling vector in time domain. To demonstrate the effectiveness of this method, a simulation with three adjacent frequencies and an experiment with practical acoustic signals of the train bearings with a crack on the outer race and the inner race had been carried out.The Doppler Shift removal error based on the method above had been analyzed. Then from the perspective of improving the precision of IF estimation, the method to estimate the IFs based on the Short time discrete spectrum correction was proposed. Based on this method, the criterion of full-period sampling was employed to increase the precision of the IF estimation and the using of variable window length was implemented to restrain the affection of the Gauss noise.A novel de-noising method has been proposed based on Doppler Shift. With the crazy climber detection algorithm, the IFs with relatively high level could be extracted. According to the acoustic theory of the moving source, the information of the position of these acoustic sources is revealed. Finally, the variable digital filtering was employed to remove the noises from the other positions of the train. A simulation on the signal with several frequency components and the experiment with the acoustic signal of the faulted bearing mixed with several noises with high amplitude demonstrated the validity of this method.Considering the drawbacks of the proposed method above, the filtering method based on the Doppler Shift removal with the noisy frequencies and with the objective frequencies were proposed respectively under the promise that interleaves of the IFs does not appeared in the signal. After the pre-processing of distinguish the signal frequencies from the noises, the Doppler Shift removal was carried out with the parameters of the noisy frequencies or the objective frequencies. Then a filter with certain frequency band could be used to remove the noises. Both of the methods could decrease the calculating time.