Application Research Of Variational Mode Decomposition In Fault Diagnosis Of Rolling Bearings

As a common component in rotating machinery and equipment,rolling bearing will directly affect the performance of the whole machine.When the rolling bearing fails and cannot be checked in time,it may cause the whole machine to be paralyzed,resulting in huge economic losses,and in serious cases may cause casualties.Therefore,whether from the perspective of safety or economy,it has important practical significance for the fault diagnosis of rolling bearings.Variational modal decomposition(VMD)is a non-recursive signal decomposition method with a solid theoretical foundation,which can effectively avoid problems such as endpoint effects and modal aliasing,and has good noise robustness.Aiming at the non-linear and non-stationary characteristics of rolling bearing vibration signals,the VMD method is applied to the field of rolling bearing fault diagnosis.The selection of key parameters in the decomposition process and the elimination of false modal components are deeply studied,and related solutions are proposed.Method,by combining with the maximum correlation kurtosis deconvolution(MCKD),autoregressive(AR)model,correlation dimension and least square support vector machine(LS-SVM)and other technologies,a related fault diagnosis method is proposed and carried out Related verification.The main research contents of this paper are as follows:(1)In-depth study of the VMD method,pointed out the shortcomings of the current common solutions,through the comparative analysis of simulation signals,pointed out the advantages of the VMD algorithm,and at the same time summarized the problems existing in the VMD,laying a foundation for the subsequent research Theoretical basis.(2)In view of the lack of basis for the selection of key parameters in the VMD decomposition process and the difficulty of effectively removing false modal components,a method for selecting the number of modal decomposition K based on the correlation coefficient and the penalty factor α based on approximate entropy is proposed,and According to the characteristics of rolling bearing signals,a false modal function selection method based on kurtosis and mutual information entropy is proposed.The analysis of measured signals proves the effectiveness of the proposed method.(3)The early fault characteristics of rolling bearings are weak,and the impact components that characterize the fault characteristics are often submerged in noise.The necessity of combining VMD with MCKD,AR models and correlation dimensions is analyzed,and AR based on the correlation dimensions of MCKD and VMD is proposed.The fault feature extraction method of the model.The method first uses MCKD to reduce the noise of early fault signals of rolling bearings to highlight the signal-to-noise ratio of the signals.Then,an AR model is established for the reconstructed signal of the sensitive IMF component obtained from the VMD decomposition.Finally,the correlation dimension of the autoregressive parameter sequence of the AR model is calculated to realize the extraction of the early fault characteristics of the rolling bearing.Through the analysis of laboratory signals and measured signals,the effectiveness of the proposed method is proved.(4)Aiming at the problem that a single feature cannot fully reflect the fault feature information of rolling bearings,a fault diagnosis method based on VMD multi-feature fusion and LS-SVM is proposed.The energy entropy,singular value entropy,power spectrum entropy and sample entropy of each sensitive IMF component obtained by VMD decomposition are calculated to reveal the characteristics of the rolling bearing vibration signal at different scales in the time-frequency domain,and the above parameters are composed of high-dimensional multi-feature state feature vectors.Enter the LS-SVM classifier to realize the fault diagnosis of rolling bearings.The effectiveness of the proposed method is verified by analyzing the signals of different fault types and the same degree of damage.The comparison of the diagnosis results of the ensemble empirical mode decomposition(EEMD)method shows that the proposed method has better fault feature extraction capabilities.