Study On Rolling Bearing Fault Diagnosis Based On Multivariate Adaptive Signal Decomposition

Rolling bearing is an important part of rotating machinery.Once it fails,it mayt only cause the paralysis of the mechanical system,and even threaten the safety of workers.Therefore,it is significant to study the fault diagnosis approaches of rotating machinery.Extracting fault features from non-linear vibration signals is the key to rolling bearing fault diagnosis.Because of the complex mechanical working environment,the single-channel signal based analysis methods can not completely represent the fault feature information,which shows some limitations in fault diagnosis.Traditional multivariate analysis methods,such as full vector spectrum,multivariate empirical mode decomposition and multivariate variational mode decomposition,are widely used in the field of fault diagnosis,but they all have certain inherent defects.Based on the intensive study of various multivariate signal processing algorithms,this paper proposes solutions to the shortcomings and defects of existing algorithms.On the basis of perfecting the theoretical content of multi-component signal processing,the corresponding improved algorithm is put forward.The superiority of the proposed method is verified by comparing the simulated and measured signals.The main research contents and innovations of this paper are as follows:(1)A fault diagnosis method based on full vector Autogram is proposed to solve the problem that full vector spectrum can not represent the fault characteristic clearly in strong noise environment.In this method,firstly,the optimal frequency band signal of each channel is selected and reconstructed by Autogram,and then the full vector envelope algorithm is used to fuse the reconstructed signal to obtain the signal envelope spectrum and extract fault characteristics.The proposed method is applied to simulated and measured signals,and the final results verify the superiority of the proposed method.(2)A multivariate uniform phase empirical mode decomposition(MUPEMD)is proposed to overcome the mode mixing problem of MEMD.In this method,a narrow-wave signal with homogeneous phase is introduced into the multivariate decomposition process,which can guarantee the decomposition effect and suppress the mode mixing.The experimental results show that MUPEMD has more advantages in suppressing mode mixing and improving decomposition accuracy.(3)An improved multivariate variational mode decomposition(IMVMD)is proposed to overcome the problem that MVMD can not select the initial iteration frequency adaptively.In this method,each mode is divided into several subbands.The initial iteration matrix is constructed with the center frequency of the subband with the highest kurtosis value,and then substituted into the multivariate variational decomposition model to obtain the decomposition results.The proposed method is applied to rolling bearing fault diagnosis,by comparing with the existing methods,the final results prove the superiority of IMVMD in multivariate fault diagnosis.(4)A rolling bearing fault diagnosis method based on IMVMD and SVM is proposed to solve the problems of low recognition rate of rolling bearings intelligent fault diagnosis based on single channel signals.In this method,firstly,the multi-component vibration signal of rolling bearing is decomposed by IMVMD,then the characteristic value of the multi-component signal is calculated to construct the multi-component characteristic matrix,and finally the multi-component characteristic matrix is input into SVM to realize the classification and identification of rolling bearing vibration data.The experimental results show that the proposed method can effectively extract bearing vibration signal characteristics and has a high fault recognition rate.In summary,the paper focuses on the existing problems in multivariate signal analysis methods such as MEMD and MVMD,and puts forward corresponding improvement methods.By applying the proposed method to the fault signal analysis of rolling bearings,several new methods for fault diagnosis of rolling bearings based on adaptive decomposition of multivariate signals are presented,which provide new technical means for fault diagnosis of rotating machinery and its key components.