Research On Fault Diagnosis Of Rolling Bearing Based On Full Vector-empirical Wavelet Transform

Bearing vibration has always been a research hotspot in the machinery industry.In recent years,with the rapid development of China's manufacturing industry,the rotation speed and load of rolling bearings are also increasing.Using vibration signals of rolling bearings for fault diagnosis research is conducive to ensuring the normal operation of mechanical equipment.However,the energy of the rolling bearing fault signal is weak,and the fault characteristic information is more susceptible to noise in the working conditions,making it difficult to effectively achieve signal-to-noise separation.Empirical wavelet transform is a new type of signal decomposition method,which can accurately decompose fault signals from complex signals.The vibration signal collected by the traditional single sensor cannot fully reflect the true state of the bearing state,which will affect the accuracy of the fault diagnosis.The full vector spectrum technology can integrate the two-channel signals and effectively avoid information leakage.This paper takes rolling bearings as the research object,comprehensively studies the advantages of the empirical wavelet transform and the full vector spectrum technology,extracts the fault characteristics of the rolling bearing vibration signals,and combines the limit learning machine for fault diagnosis.The main research contents are as follows:(1)Firstly,the theory and performance of the EWT method are studied,and the simulation signal is used to verify that the EWT method has stronger signal analysis capabilities than EEMD and ITD.Aiming at the problem of insufficient decomposition of weak fault signals by EWT,combined with the advantages of KICA,an EWT-KICA combined rolling bearing fault feature extraction method is proposed.First,EWT is used to decompose the bearing fault signal,the component results obtained by the decomposition are screened and reconstructed according to the correlation coefficient criterion,and then KICA is used to demix the reconstructed signal to achieve the separation of signal and noise.Experimental results show that this method can enhance EWT's ability to extract fault features.(2)Secondly,the research found that the traditional EWT method may cause over-decomposition or under-decomposition of the signal due to the spectrum segmentation problem.This paper proposes an EWT spectrum segmentation optimization method for the maximum envelope,and at the same time,in order to solve the problem of missing information caused by single-channel signals,Introduce the full vector spectrum information fusion technology,and propose an improved EWT-KICA full vector fusion fault feature extraction method.It is verified through experiments that the method not only achieves the purpose of signal-to-noise separation,but also can extract the fault features of rolling bearings more comprehensively and accurately.(3)Finally,the principle of the ELM method is introduced,and the characteristic factors of the main vibration vector and the time-domain signal after full vector fusion are calculated respectively,and they are composed into feature vectors.An ELM bearing fault diagnosis method based on full vector empirical wavelet transform is proposed.The method uses the improved EWT-KICA full vector information fusion method to extract the fault features of the rolling bearing,then calculates the characteristic parameters of the main vibration vector and the time domain signal and constructs the feature vector for identification,finally inputs the feature vector into the ELM Classifier and classify test samples.Experiments show that this method can accurately identify the normal rolling bearing,outer ring fault,cage fault and cage rolling outer ring composite fault.