Fault Diagnosis Of Rolling Bearing Based On Intrinsic Time Scale Decomposition

Rolling bearings are extremely important components in rotating machinery,and their operating states directly determines whether the mechanical equipment can work safely and normally.Therefore,it is significant to carry out health monitoring and fault diagnosis of rolling bearings.The actual bearings fault signals are often nonlinear and non-stationary,and contain a lot of noise,making it difficult to extract fault feature information.Therefore,this thesis takes rolling bearings as the research object,and conducts research on rolling bearings fault feature extraction,pattern recognition,and composite fault diagnosis.The main research contents are as follows:(1)The principle of intrinsic time scale decomposition(ITD)and multi-point optimal minimum entropy deconvolution adjusted(MOMEDA)are briefly described.The problems of ITD and MOMEDA are analyzed.(2)Aiming at the problem that the ITD method is difficult to extract bearings fault feature under the influence of strong background noise,a fault feature extraction method for rolling bearings combining ITD and parameter optimized MOMEDA is proposed.First,the proper rotation(PR)component containing rich fault information is selected according to the principle of maximum crest factor of envelope spectrum.Then,the MOMEDA noise reduction process is performed on the decomposed component.The two parameters — fault period T and filter length L that affect the filtering effect of MOMEDA are optimized with multi-point kurtosis(MKurt)and gini index of square envelope spectrum(GISES)respectively.Finally,envelope spectrum analysis is performed to identify the fault type.The proposed method is verified by using the inner ring fault simulation signal,the XJTU-SY rolling element bearing accelerated life test datas and the gearbox test datas.The results show that this method can extract the characteristic frequency of bearing faults under strong background noise,and has a high frequency domain resolution.(3)Aiming at the problem of fault information loss caused by single feature extraction,a pattern recognition method is proposed,which combines multiscale fuzzy entropy(MFE)and Hjorth parameters to form mixed fault feature and uses sparrow search algorithm(SSA)to optimize support vector machine(SVM).First,the original signal is decomposed by ITD to obtain several PR components,and the correlation coefficient is used to select the PR components with rich feature information.Then,MFE and Hjorth parameters are combined to form mixed feature vectors,and the mixed feature vectors of multiple PR components are calculated.Finally,the feature vectors are input into SSA-SVM to realize pattern recognition.The proposed method is verified by using the bearing data of Case Western Reserve University,XJTU-SY rolling element bearing accelerated life test datas and gearbox test datas.The results show that this method can realize the classification of different bearings states,and has a high classification accuracy.(4)Aiming at the problem that it is difficult to extract compound fault features of bearings,a compound fault diagnosis method of rolling bearings is proposed,which used ITD combined with parametric optimized resonance-based sparse signal decomposition(RSSD)to enhance fault features and parameter optimized MOMEDA to reduce noise and highlight fault features.First,based on the maximum Ec,the PR components with rich fault feature information are selected,and signal reconstruction is performed for the selected PR components.Then take1/Ec as the fitness function,the SSA is used to optimize the parameters Q and λ which affect the RSSD decomposition effect.The above reconstructed signal is decomposed by parameter optimized RSSD,and the low resonance component is obtained.Finally,parameter optimized MOMEDA is used to denoise the low resonance component,and composite fault type is extracted by envelope spectrum analysis.The proposed method is verified by using the simulation signal and the XJTU-SY rolling element bearing accelerated life test datas.The results show that the method can extract the characteristic frequency of compound fault,and the extraction effect is obvious.