Research On Noise Reduction And Feature Extraction Of Dry Gas Seal Vibration Signal

With the rapid development of modern industry,dry gas seal is widely used in many fields,especially in nuclear power,aviation,navigation and military industry.The characteristic information of vibration signal during the operation of dry gas seal is weak and disturbed by strong external noise,which is difficult to truly reflect the tribological characteristics of dry gas seal.This will cause great difficulties for the research directions of dry gas seal fault diagnosis,condition monitoring and life prediction under extreme working conditions such as high speed and high pressure in the future.In view of the above problems and the difficulty of separating vibration signals from noise,a noise reduction method based on the combination of adaptive ensemble empirical modal analysis(CEEMDAN)and independent component analysis(ICA)theory is proposed.On this basis,S-transform is used to extract the features of the noise reduced vibration signals.Firstly,the noise reduction model of vibration signal is established by using the principles of CEEMDAN,ICA and fuzzy entropy.The acceleration signal during the operation of dry gas seal is collected through experiment,and the acceleration signal is CEEMDAN decomposed to obtain multiple intrinsic mode function(IMF)components.Then the corresponding independent component components are obtained through ICA transformation and the fuzzy entropy is calculated.The components with unqualified fuzzy entropy are set to zero,and the qualified components are reconstructed to obtain the signal after noise reduction.At the same time,the vibration signal is simulated by the simulation signal,and the noise reduction method proposed in this paper is verified.The root mean square error and signal-to-noise ratio are introduced as the evaluation indexes of the noise reduction results.The results show that the noise reduction method proposed in this paper is feasible and reasonable.Secondly,the orthogonal test is carried out on the dry gas seal test-bed from 6pressures and 6 groups of rotating speeds,and the acceleration signal during the operation of dry gas seal is collected through the experiment.The noise reduction method proposed in this paper is used for noise reduction,and the noise reduction methods in relevant literature are compared.The experimental results show that the signal-to-noise of CEEMDAN + ICA method is about 3.8 d B higher than that of EEMD+ ICA method,and the root mean square error is reduced by 5%;In terms of noise component selection,the signal-to-noise of CEEMDAN + ICA method is about 9.5 d B higher than CEEMDAN+WT method,and the root mean square error is reduced by66%,indicating that the noise reduction model based on CEEMDAN+ICA has certain reliability.Then,in order to further verify the advantages of S-transform in feature extraction,the bearing experimental data set of Western Reserve University is selected for vibration feature extraction.By comparing the performance of short-time Fourier transform and wavelet transform in feature extraction,the results show that Stransform has high time-frequency aggregation and high precision in feature extraction,so S-transform is selected to extract features in this paper.Finally,according to the relevant parameters in the noise reduction model,the effects of three parameters on the noise reduction results are discussed in CEEMDAN analysis;In the part of fuzzy entropy,the influence of phase space dimension and similarity tolerance on the noise reduction results is discussed,and finally the best parameter value in the noise reduction model is obtained.According to the noise reduction method proposed in this paper,the noise reduction analysis of the vibration signal collected by the test-bed is carried out,and the vibration signal after noise reduction is obtained.In feature extraction,the vibration signal after noise reduction is analyzed and studied by S-transform.By comparing the frequency band range of vibration signal obtained by power spectral density analysis and CEEMDAN analysis,they have high similarity and can be judged as vibration fundamental frequency signal.Through the acceleration test signal of dry gas seal,it is proved that the method in this paper is more effective than other traditional noise reduction methods,and provides a new way for the research of gas seal fault diagnosis and condition monitoring.