Research On Fault Deteriorating State Recognition Of Roll Bearing Inside Track

With the development of modernization in industrial production, equipment maintenance mode is gradually shifting predictive maintenance, equipment fault diagnosis technology is the guarantee of the realization of predictive maintenance. Rolling bearing is often used as the mechanical parts in industrial and mining enterprises, so it is essential in the production equipment. For realizing the predictive maintenance in rolling bearing, ensuring rolling bearing's run in good condition, need to find the fault type and much more find the extent of the fault deterioration in time. Therefore, research on fault deterioration state recognition of roll bearing inside track has important theoretical and applied value.Basing on collecting extensive related information, this paper has put forward to combine Hilbert-Huang Transform and Sphere-structured Support Vector Machines to apply in the fault deteriorating recognition field of roll bearing inside track, and to make the recognized results visual, fully verifying the effects of this method.This paper has mainly researched on the several following aspects: Analyze the characteristic frequency of roll bearing and vibration signal characteristics of good roll bearing and of the roll bearing inside track's fault. Complete the roll bearing fault diagnose model, and respectively collect the gradual deteriorating vibration signal of rolling bearings from the good state, and the inside track's slight fault, to the moderate fault, and the severe fault.Elaborate principle of Hilbert-Huang which is the current excellent time-frequency signal processing method. Use The HHT method and wavelet method to analyze the same simulation signal and compare the results. Use the Hilbert marginal spectrum and the power spectrum to analyze deteriorating signal of the inside track's fault, showing that the Hilbert spectrum has the better identify, but the spectrum analysis can not well recognize the deteriorating extent , need to extract feature vector to map to the high-dimensional space, and conduct pattern recognition.EMD is a self-adapt time-frequency analyzed method, and could grasp firmly signal's characteristic information. Respectively use IMFS energy entropy and SVD method to extract the feature of the roll bearing inside track fault deteriorating state, the results show SVD method could much more effectively extract feature, with the deterioration of rolling bearing inside track fault, the SVD entropy is getting smaller.Basing on these, study theory and method of using Sphere-structured Support Vector Machines to recognize the roll bearing inside track's fault deterioration extent. At the same time, discuss problem of kernel functions to show that RBF kernel function is suitable to Sphere-structured Support Vector Machines. Try to find cluster core and cluster radius in different fault degree and study the changing law of them.In addition, this paper use the high-dimensional data visualization technology of principal component analysis and the chromaticity diagram to study display of Sphere-structured Support Vector Machine's recognizing results (high-dimensional) in two-dimensional space, make Sphere-structured Support vector Machine's recognizing results directly visual.Results of this paper show the method of Hilbert-Huang Transform and Sphere-structured Support Vector Machines can accurately recognize deteriorating state of the roll bearing inside track' fault. Provide an effective method for roll bearing fault diagnosis.