Research On The Application Of EWT And Its Improved Methods In The Fault Diagnosis Of Rolling Bearings

With the rapid development of science and technology and the popularization of modern industrial production,rotating machinery is developing in the direction of complexity,large-scale,high-speed and automation,which puts forward higher requirements for the safe operation of equipment.Rolling bearings are one of the most commonly used components in rotating machinery,which are directly related to the smooth and reliable operation of the mechanical system.The research on rolling bearing fault diagnosis methods can effectively avoid the occurrence of production accidents,which has important safety and economic significance,as well as important academic significance and engineering value.?I?Aiming at the problem of unreasonable spectrum division result of empirical wavelet transform?EWT?,a modified fast empirical wavelet transform?MFEWT?method is proposed.The method firstly calculates and optimizes the spectral trend term by fast Fourier transform and wavelet compromise threshold function denoising,and establishes a filter band according to the trend components,and performs EWT signal decomposition;then combines the kurtosis criterion and the correlation coefficient selection principle to reconstruct the feature component,The minimum entropy deconvolution of the characteristic components to calculate the characteristic frequency;finally,through the matching of the theoretical characteristic frequency,the fault diagnosis of the rolling bearing is realized.Experimental results show that MFEWT has a good fault feature extraction effect,the method improves the performance of EWT signal decomposition and the reliability of rolling bearing fault diagnosis.?II?Based on the idea of trend item extraction in the MFEWT method,the l₁TF-EWT method is proposed for the problem of trend component extraction in the method and the over-decomposition problem in the resulting spectrum division results.This method first uses l₁trend filtering to process the signal spectrum to estimate the trend component of the spectrum,then divides the frequency spectrum according to the obtained trend component and establishes filter bands to perform empirical wavelet transform.In the vibration signal analysis experiment,l₁TF-EWT is compared with MFEWT.Results show that the trend components obtained by l₁TF-EWT can better reflect the change of the signal spectrum,and the result of feature extraction is more ideal.l₁TF-EWT further improves the performance of EWT signal decomposition and the reliability of rolling bearing fault diagnosis.?III?In order to realize intelligent fault diagnosis of rolling bearings,a rolling bearing fault diagnosis method based on the fusion of l₁TF-EWT with sample entropy and support vector machine?SVM?is proposed.The method first uses l₁TF-EWT to decompose the vibration signal of the rolling bearing and reconstruct the feature component,then calculate the sample entropy of the feature components,and input the sample entropy as the feature value to the SVM for training.Finally,the trained classifier is tested to realize the rolling bearing failure diagnosis.The experimental results show that the fault diagnosis method combining l₁TF-EWT with sample entropy and SVM has a good fault diagnosis effect.Theoretical analysis and experiments show that the rolling bearing fault diagnosis method based on improved EWT has good effect on rolling bearing fault diagnosis and has certain practical application value.It was found in the study that the selection of regularized parameters for l1 trend filtering and the application of the diagnostic method in this paper to the assessment of bearing damage and bearing life expectancy are yet to be further studied.