A new method of acoustic signals separation for wayside fault diagnosis of train bearings

Abstract

For the acoustic signal acquired by a microphone is composed of a number of train bearing signals and noises, single signal of failure train bearing should be extracted to diagnose the fault type precisely in wayside fault diagnosis of train bearings. However, the phenomenon of Doppler distortion effect in the acoustic signal acquired with a microphone leads to the difficulty for signal separation. In this chapter, a new method based on Dopplerlet transform, time-frequency filtering and inverse generalized S-transform is proposed to separate different fault types of train bearing signals from the acoustic signal. Firstly, search the parameters space to find the primary functions-Dopplerlet atoms which match the original signal best by matching pursuits- based Dopplerlet transform. According to the parameters, these Dopplerlet atoms are divided into different groups corresponding to diverse acoustic sources. Through extracting the data of Dopplerlet atoms in a group and its neighborhood in time-frequency domain, the signal of corresponding train bearing can be reconstructed by the inverse transformation of GST. To diagnose the fault type of the reconstructed signal, re-sampling is carried out to remove the Doppler distortion effect in advance. After that, we can identify the fault type of reconstructed signal corresponding to a certain train bearing through the envelope spectrum. Finally, experiments with practical acoustic signals of train bearings with a defect on the outer race and the inner race are carried out, and the results verified the effectiveness of this method.