Study On Fault Diagnosis Of Rolling Bearing Based On GVMD And Manifold Learning

Rolling bearing is a component used commonly to deliver load in mechanical equipment.It is of great risk to cause local damage by the reasons of scratch,crumbling,overload and others when the equipment was running.If rolling bearing got failure,then it would damage the equipment or lead accidents.Therefore,it is of significance to strengthen monitoring of running state and analyzing fault data for rolling bearing.In the actual complex production environment,the obtained vibration signal contains high frequency noise components.Getting the key information from mixed signal is a vital step for feature extraction and fault diagnosis.Analyzing vibration signal can obtain abundant information of equipment state,and extracting effective identification features can not only improve the accuracy of fault diagnosis,but also improve the reliability of early prediction.Therefore,the acceleration signal of rolling bearing was taken as the research object,combining genetic algorithm,variational mode decomposition theory,envelope spectrum analysis and pattern recognition methods to study these three issues of denoising,feature extraction and fault diagnosis.The main work of this paper could be summarized as follows:For the denoising problem,this paper adopted the method of combining variational mode decomposition(VMD)and wavelet soft threshold.Genetic algorithm was utilized to optimize parameters of VMD,then decomposed the mixed signal adaptively by GVMD method,and processed the modes of decomposition respectively by wavelet soft threshold method.Finally,the components signal were restructured after wavelet threshold processing.Experimental results verified that the presented method offered better denoising effect than traditional methods.For the reason that vibration response caused by local damage of rolling bearings might be overwhelmed by the greater vibration signal,so making a focus on the shock response caused by local damage would be the key to identify the fault type of rolling bearing.This paper proposed a GVMD envelope spectrum analysis method,based on the characteristics of adaptive decomposition and suitable for processing non-stationary and nonlinear signal for VMD method,GVMD method was employed to decompose vibration signal of rolling bearings and an envelope spectrum analysis was realized to the component which had the largest coefficient with original signal.After that according to the sharp change of spectrum peak,the feature frequencies were got and then faults types were distinguished.Aiming at the problem of traditional time domain and frequency domain analysis methods are difficult to take consideration of the time-varying characteristics for non-stationary signals.This paper proposed a feature extraction method combing statistical features of time domain,GVMD sample entropy and energy ratio features.It would reveal the fault characteristic information of rolling bearing vibration signal comprehensively and accurately.Manifold learning method was employed to reduce dimension,and support vector machine was combined to realize the fault diagnosis of rolling bearings.The experimental results showed that the average recognition rate of the proposed method is 98.67%.Compared with other feature extraction methods,the proposed method had better fault recognition performance.