Research On De-noising And Fault Identification Of Rolling Bearing Vibration Signal

Rolling bearing is the most widely used and easily damaged core support component in rotating machinery.If they break down,they will directly affect the safe and stable operation of the entire unit.In-depth study of the fault diagnosis technology of rolling bearing can provide a theoretical basis for the repair and maintenance of equipment,which has important practical significance for ensuring the safe service of equipment.Therefore,rolling bearing is selected as the research object,and then the noise interference and fault identification problems of bearing vibration signal are studied.Finally,three key problems were solved the first is to eliminate the vibration signal noise of rolling bearing effectively.The second is to extract the fault feature of rolling bearing accurately.The third is to identify the fault type of rolling bearing accurately.The main research contents are as follows:(1)Aiming at the problem of strong noise interference in the vibration signal of rolling bearing,a vibration signal denoising method combining complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN)and wavelet packet multi-threshold was proposed.The CEEMDAN algorithm was used to manage the original complex signal,the main information contained in each intrinsic mode function(IMF)component was analyzed.The IMF components dominated by noise were selected,and the wavelet packet multi-threshold method was introduced to eliminate the interference of noise on signal analysis.All the noise reduction components and the signal dominant components were fused to obtain the reconstructed signal.Simulation analysis and comparison experiment results show that this method can effectively eliminate the noise of rolling bearing vibration signal.(2)In order to extract the fault feature of rolling bearing accurately,a rolling bearing fault identification method of CEEMDAN multi-scale permutation entropy and kernel principal component analysis(KPCA)was proposed.The CEEMDAN algorithm was introduced to decompose the vibration signal,and the third-order IMF components reconstruction signal with large kurtosis value was calculated and selected.The multiscale permutation entropy value of the reconstruction signal was extracted to construct the initial feature set,and the relationship between features was analyzed.The KPCA was used to eliminate redundant information of features,and reduce the dimension of feature set,and then the low-dimensional feature set was inputted into the support vector machine(SVM)to identify bearing fault.Compared with other typical feature extraction methods,the results show that this method can precisely extract the fault feature of rolling bearing and realize the accurate identification of rolling bearing fault type.(3)Aiming at the problem of difficulty in choosing the key parameters of traditional SVM,the influence of SVM penalty parameter and kernel parameter of the identification effect was analyzed,and the parameter optimization process of grid search(GS)algorithm and PSO algorithm was studied.The advantages of two algorithms were synthesized and the two key parameters of SVM were optimized,and then a fault prediction model with better identification effect was obtained.The classification performance of the optimized model with different fault type,different fault size,and fewer training sample was analyzed.Simultaneously,the effects of different methods on the identification of different fault types of rolling bearing were compared.Experimental results show that this method can achieve accurate identification of different fault type of rolling bearing.(4)In order to further verify the effectiveness of the method proposed in this paper,the noise reduction method and the fault identification method of rolling bearing mentioned in this paper are applied to the measured bearing vibration signal.The noise reduction effect of the proposed noise reduction method on the bearing vibration signal under different working conditions is analyzed,and the results of fault identification of the measured data is obtained.Experimental results show that the methods proposed in this paper have good practical effect.