A Noise Reduction Method for Acoustic Emission Signals of Dry Gas Seals Based on Frequency Subspace Dictionary Learning
Aiming at the problems of low signal-to-noise ratio and susceptibility to background noise in-terference in dry gas seal acoustic emission signals,a denoising method based on frequency domain sub-space dictionary learning was proposed.Firstly,ob-tain the mutual relationship between each frequency band based on adjacent relevant information of the time-frequency distribution of the acoustic emission signal.Based on this,the boundary of the frequency domain division was determined,and the correspond-ing empirical wavelet family was constructed.The sparse reconstruction of the acoustic emission signal was carried out in each subspace using the time shift invariant dictionary learning algorithm.On this ba-sis,the kurtosis index of the reconstructed signal was used to weight each component.The experimental re-sults showed that the proposed algorithm improves the signal kurtosis index from 48.43 to 185.93,achie-ving noise reduction of acoustic emission signals and enhancement of collision and wear characteristics dur-ing dry gas sealing start-up process.
dry gas sealacoustic emissionempiri-cal wavelet transformsparse dictionary learningsig-nal noise reduction
国家科技期刊平台
NSTL
万方数据
