Rolling Bearing Fault Diagnosis Of CNC Machine Tool Based On Variational Mode Decomposition And Gradient Boosting Decision Tree

Rolling bearing is one of the important parts of CNC machine tool mechanical transmission system.Research on the performance and diagnosis of the rolling bearings is an important part of evaluating the health of CNC machine tools,and has important practical value and economic significance.Bearing fault diagnosis can effectively prevent sudden accidents and is an important guarantee for the safe operation of mechanical systems.Therefore,this thesis takes rolling bearings as the research object,starts with the processing of bearing vibration signals,and conducts related research on the problems of rolling bearing fault diagnosis and identification.This thesis discusses the failure mechanism and analysis method of rolling bearings.On this basis,the types and causes of bearing failure,the characteristic frequency of failure and the analysis method is explained in detail.Combine with actual signals,the characteristics of vibration signals under various fault conditions are given.The method of Variational Modal Decomposition(VMD)is studied and used for the early diagnosis of rolling bearing faults.At the same time,Empirical Mode Decomposition(EMD)is introduced.Comparing the results of the two methods of processing analog signals,it can be seen the superiority of variational mode decomposition method.Finally,it is verified with full bearing life data.Singular value decomposition(SVD)and gradient boosting tree methods are studied,and a gradient boosting tree model for rolling bearing state classification is established.A fault diagnosis method of rolling bearing based on VMD-SVD and gradient boosting tree is proposed.The vibration signal is subjected to variational modal decomposition to obtain the modal component containing the fault feature,and then the modal component is used as the initial vector for singular value decomposition.The decomposed eigenvectors constitute the fault feature matrix.The gradient boosting tree model is used to test and train the feature matrix,so as to realize the identification and classification of bearings in different states.This method is applied to the fault diagnosis of rolling bearings,and the good diagnosiseffect proves the feasibility of the method proposed in this thesis.Finally,for the time-varying non-stationary characteristics of the variable-speed bearing vibration signal,an improved variational mode decomposition method is proposed,which optimizes the spectral concentration factor to estimate the instantaneous frequency of the signal,and then demodulates the nonlinear frequency-modulated signal.The demodulated fixed signals are extracted by a band pass filter.The actual analysis results show the effectiveness of this improved method in extracting fault signal features.