An impact feature extracting method based on S-transformation spectrum threshold denoising

Yuan-jing GUO,Yan-ding WEI,Xiao-jun ZHOU,Fang TANG

DOI: https://doi.org/10.13465/j.cnki.jvs.2014.21.008

2014-01-01

Abstract:Extracting impact features from a rotating machinery fault vibration signal is significant to some fault diagnoses.Based on the fact that S-tansformation is suitable for processing signals with impact features,an effective impact feature extracting method with the threshold denoising for S-transformation spectrum was proposed.Firstly,the time-frequency spectrum of a time-domain signal was obtained using S-transformation.As the spectrum coefficients were complex numbers,the spectrum threshold denoising was performed according to the magnitude of coefficients.During denoising,both the hard and the soft threshold functions were used,respectively.Most importantly,the optimal threshold estimation was obtained between zero and the maximum modulus of the coefficients with the step iterative algorithm,and evaluated with the introduced modified risk function.Finally,the inverse S-transformation was carried out on the denoised time-frequency spectrum to reconstruct the time-domain impact features.The processing results of simulated signals and rolling bearing fault vibration signals showed that using the acquired optimal threshold,S-transformation spectrum threshold denoising method can be used to successfully extract impact features from noise mixed signals,so as to realize the related fault dignosis.