Research on threshold denoising method for correlation coefficient spectrum of vibration signal

• Published in: 2015 12th IEEE International Conference on Electronic Measurement & Instruments (ICEMI) Vol. 1; pp. 359 - 363
• Main Authors: Gu Yuhai, Ma Chao, Han Qiushi, Xu Xiaoli
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.07.2015
• Subjects: autocorrelation; Correlation; Correlation coefficient; Feature extraction; FFT; Noise reduction; Root mean square; Signal to noise ratio; threshold method; vibration signal; Vibrations
• DOI: 10.1109/ICEMI.2015.7494201

Vibration signal is the main signal used for fault diagnosis of large mechanical and electrical equipment. Because of the bad work condition, the vibration signal contains complex noises and its SNR(signal-to-noise ratio) is very low, so it is difficult to extract the fault feature in vibration signal. According to the characteristic of vibration signal, an denoising method is provided to eliminate the broadband random noise by using autocorrelation coefficient spectrum threshold. The acquisition of the signal is calculated by autocorrelation function firstly, and then the autocorrelation signal of amplitude spectrum is computed with FFT. After that the special frequency points are selected from amplitude spectrum of autocorrelation signal, which meet the threshold requirements. The FFT result of original signal is filtered according to the selected points before, and the results satisfying the conditions are reserved and the rest is cleared to zero. The denoised signal is reconstructed by IFFT with the selected FFT value of original signal. To verify the effectiveness of this denoising method, the simulation signal and the measured signal is used for testing, and the results show that this denoising method is s simple but a good denoising effect is obtained.