Research On Fault Diagnosis Of Rolling Bearing Based On Optimized VMD-MCKD And Feature Dimension Reduction

Rotating machinery is widely used in industrial production,rolling bearing is one of the most frequently used and easy to malfunction parts in rotating machinery,and its operating status directly affects the safety of industrial production.In practical application,due to the composition of the rolling bearing and the working condition,the generated signal always contain noise and shock,showing nonlinearity and non-stationarity,making the fault characteristics difficult to extract.Therefore,the study of rolling bearing fault diagnosis method is of great significance.This thesis takes the rolling bearing as the research object,with signal processing,feature enhancement,and fault identification as the starting point,a series of studies are carried out on the fault diagnosis technology of rolling bearing.The primary research contents of the thesis are as follows:(1)Regarding to the problem that it is difficult to extract fault features from rolling bearing signals under noise background,this thesis introduces the optimized variational modal decomposition(VMD)method.In this thesis,the empirical mode decomposition(EMD)method is used as a comparison to verify the advantages of the VMD method in reducing mode mixing and processing the noisy signals.Because the VMD method needs to set the number of decomposed modes in advance,this thesis proposes a method to decompose the signal by calculating the center frequency relationship of adjacent modes to determine the number of K values.The mode with the largest efficient weighted kurtosis(EWK)index is selected as the effective component,and the envelope analysis is carried out to realize fault diagnosis.The validity of the propounded approach is proved by simulation and experimental signal analysis.(2)In view of the fact that the early fault characteristics of rolling bearings are weak and the periodic impact is not obvious,and it is difficult to fully extract the fault features,this thesis introduces the maximum correlated kurtosis deconvolution(MCKD)method.Although the MCKD method can deconvolve periodic pulses from fault signals,its parameters need to be set reasonably,therefore,this thesis studies the weak fault feature extraction method based on optimized VMD and MCKD.Firstly,using the optimized VMD method as the prefilter to denoise the signal,the effective component is selected according to EWK index,and then the minimum envelope entropy of the effective component is obtained as the objective function,and the particle swarm optimization(PSO)algorithm is used to optimize the two parameters of MCKD:the optimal filter length L and the displacement number M.The effective component is enhanced by the optimized MCKD,and finally the enhanced signal is enveloped and demodulated to extract the fault characteristic frequency to realize the fault diagnosis.Combined with simulation and experimental signal analysis,and compared with the combination of VMD and minimum entropy deconvolution(MED)method,the availability of the presented approach is proved.(3)Aiming at the problem of low accuracy caused by feature redundancy due to the correlation between features in rolling bearing fault recognition,a fault diagnosis approach on the basis of Xgboost and support vector machine(SVM)is rendered.Firstly,extract time domain features from rolling bearing signals to form feature vectors,use Xgboost to rank the time domain features,then some features after sorting are selected to construct a feature set,and use SVM to identify fault types.Based on rolling bearing data,the simulation analysis of rolling bearing fault diagnosis under the same load and different loads are carried out and compared with the SVM algorithm and the method based on the combination of principal component analysis(PCA)and SVM,the experimental results show that the proposed method has higher accuracy.