An Ultrasonic Signal Denoising Method for EMU Wheel Trackside Fault Diagnosis System Based on Improved Threshold Function

• Main Authors: Zhilin Sun, Jingui Lu
• Format: Article
• Language: English
• Published: IEEE 2021-01-01
• Series: IEEE Access
• Subjects: EMU wheel; fault diagnosis; ultrasonic data; wavelet transform; threshold denoising
• Online Access: https://ieeexplore.ieee.org/document/9467377/

In the safety protection system of the railway electric multiple unit (EMU), the safety of the running part is extremely important. The daily detection of the internal hazard defects of the wheels in the running parts relies on a professional trackside fault online diagnosis system based on the ultrasonic sensor probe array data. However, the on-line ultrasonic diagnosis of EMU wheels is usually accompanied by various interference noises. The defect echo signals collected by the sensor probe array are weak and are easily submerged by noise, which makes it impossible to perform effective defect identification. This paper proposes an improved threshold function to overcome the discontinuous shortcomings of the classical wavelet soft threshold function and hard threshold function in view of the non-stationary characteristics of the ultrasonic detection signal of the EMU wheels. This paper proposes a sine-type threshold processing function. It is characterized by adopting gradual compression processing to denoise the echo signal of the ultrasonic sensor probe array. In order to verify the validity, the continuity at the threshold is observed through the linear space vector signal, and the algorithm is simulated and tested through the three-dimensional Gaussian echo mathematical model of the ultrasonic signal and the measured ultrasonic envelope signal. Experimental results show that the improved threshold function can suppress the noise in the ultrasonic echo data, improve the signal-to-noise ratio, and retain the waveform characteristics of the defect signal, which is conducive to defect recognition.