Multi-bearing defect detection with trackside acoustic signal based on a pseudo time–frequency analysis and Dopplerlet filter

• A new method is proposed for multi-bearing defects detection.
• A novel pseudo time–frequency analysis is used to locate different bearings.
• Dopplerlet filters are used for each sound source.
• Doppler distortion is removed by a resampling method.
• Effectiveness of the technique is validated by simulated and experimental cases.

The diagnosis of train bearing defects based on the acoustic signal acquired by a trackside microphone plays a significant role in the transport system. However, the wayside acoustic signal suffers from the Doppler distortion due to the high moving speed and also contains the multi-source signals from different train bearings. This paper proposes a novel solution to overcome the two difficulties in trackside acoustic diagnosis. In the method a pseudo time–frequency analysis (PTFA) based on an improved Dopplerlet transform (IDT) is presented to acquire the time centers for different bearings. With the time centers, we design a series of Dopplerlet filters (DF) in time–frequency domain to work on the signal׳s time–frequency distribution (TFD) gained by the short time Fourier transform (STFT). Then an inverse STFT (ISTFT) is utilized to get the separated signals for each sound source which means bearing here. Later the resampling method based on certain motion parameters eliminates the Doppler Effect and finally the diagnosis can be made effectively according to the envelope spectrum of each separated signal. With the effectiveness of the technique validated by both simulated and experimental cases, the proposed wayside acoustic diagnostic scheme is expected to be available in wayside defective bearing detection.