Research On Fault Diagnosis Method Of Rolling Bearing Based On Signal Sparse Representation

Rolling bearing is a key component in mechanical equipment,and its operating state is related to the stability and reliability of the system.It is very important to use appropriate signal processing methods to diagnose rolling bearing faults.Rolling bearing fault diagnosis includes two major parts: fault characteristic extraction and pattern recognition.In this paper,the sparse decomposition method,a newly emerging signal processing method in recent years,is used to diagnose the faults of rolling bearings.In terms of fault feature extraction,three improvements have been made: First,the combination of sparse decomposition method and frequency domain correlation kurtosis is used to introduce frequency domain correlation kurtosis into the dictionary construction link,and the dictionary construction method is improved,using frequency domain correlation.For the value of kurtosis,the most similar atom with the fault feature is selected in the over-complete dictionary,which reduces the number of matching atoms and ensures the simplicity of the sparse representation method;second,the sparse decomposition method under the improved dictionary construction method is used for Signal denoising,the denoised signal enters the wavelet packet,EMD and EEMD to achieve secondary denoising,reconstruction and fault feature extraction,which improves the effect of extracting fault features;third,the reconstruction algorithm for sparse decomposition.The OMP algorithm has been improved,and an improved OMP algorithm-ROMP algorithm has been obtained,which improves the performance of the reconstructed signal.In the aspect of pattern recognition,the PSO algorithm is used to optimize the LS-SVM parameters,and the optimized LS-SVM is obtained,which greatly improves the accuracy of the fault classification of the rolling bearing.The main research contents of this paper are as follows:The background of the topic selection and the significance of the research are expounded,the status quo and development trend of domestic and foreign research on the fault diagnosis of rolling bearings are summarized,and the causes and failure mechanisms of rolling bearing faults are analyzed.Combining the sparse decomposition method and the frequency domain correlation kurtosis,the sparse representation method under the improved dictionary construction method is obtained.The selected atom is determined at the position of the maximum value of the frequency domain correlation kurtosis.Through simulation and experimental verification,compared with the traditional combine K-SVD with OMP sparse decomposition method,the ability to denoise,reconstruct and extract fault feature frequencies has been significantly improved.The OMP sparse reconstruction algorithm is improved,and the ROMP algorithm is obtained.After simulation and experimental verification,compared with the OMP algorithm,the sparse reconstruction performance is improved.The sparse decomposition method under the improved dictionary construction mode is used for signal denoising and filtering,and the denoised signal enters the wavelet packet,EMD and EEMD.Simulation and experimental verification have improved the ability of fault feature extraction compared with wavelet packet,EMD and EEMD methods without sparse decomposition and denoising.The LS-SVM parameters are optimized using the PSO algorithm,and the optimized parameters are input into LS-SVM for training,and the best LS-SVM model is obtained.Through simulation and experimental verification,compared with the non-optimized LS-SVM,the accuracy of rolling bearing fault classification is greatly improved.Finally,this paper uses MATLAB GUIDE to design the rolling bearing fault diagnosis platform software,which realizes the human-computer interaction.It has practical significance.There are 102 figures,1 table and 67 references in this paper.