Bearing Fault Diagnosis Based On Multivariate Empirical Mode Decomposition Cross Approximate Entropy And Gath-geva Clustering Algorithm

With the development of large-scale production and the progress of science and technology,the level of automation of modern equipment is increasing day by day,the scale of the system is also increasing,and the function and complexity are constantly expanding.Therefore,the risk of the system is getting higher and higher.Research on mechanical fault diagnosis technology is important for equipment to improve the reliability and safety of production.In this paper,a kind of multivariate empirical Mode decomposition(Multi-EMD)Cross Approximate Entropy,(cApEn)and GG Clustering Rolling Fault Bearing Diagnostic Method approach is proposed to solve the key problem of how to effectively extract the characteristics of vibration signals by using the signals collected from the rotating bearing parts.Secondly,the algorithm of EMD(EMD)and its application in signal processing are discussed in detail.Because of the problem of modal aliasing in empirical mode decomposition,which seriously affects the resolution and effect of decomposition,we introduce the Multivariate Empirical Mode Decomposition(Multi-EMD)method,which not only retains the good characteristics of EMD method,The local feature information of different scales of the signal is obtained,and the frequency aliasing between the components left by the EMD can be solved well.Then,Multi-EMD solves the problem of sampling imbalance and multi-channel signal decomposition.With the bearing vibration data of Case Western Reserve University as the signal source,the collected signal has complex nonlinear and periodic nonuniform operation characteristics.Because entropy theory can effectively capture the characteristics of the signal characteristics,the use of entropy to characterize the complexity and uncertainty of the signal.Cross approximation entropy(cApEn)is introduced on the dynamic characteristics that can characterize the signal more fully.Mutual entropy is an improvement of approximate entropy.Can effectively reflect the interaction of signal information.Cross approximation entropy can be well clustered as a clustering feature vector.After comparison of FCM clustering,GK clustering and GG clustering,GG clustering is used to classify.Finally,the fault diagnosis experiment of the vibration bearing is carried out for the fault data of the Case Western Reserve University and the actual data collected by Shanghai Baogang.In order to improve the accuracy of the calculation,the signal signal decomposition of the fault signal is obtained after a series of IMF components.Then we obtain the mutual entropy of the IMF components of different signals,reconstruct the mutual entropy matrix,and then cluster the different signals according to the mutual entropy matrix.According to the analysis of the close degree method,the method can effectively judge the different fault signals.